Engineering Tools for Regulating Hypoxia in Tumour Models

Abstract Major advances in the field of genomic technologies have led to an improvement in cancer diagnosis, classification and prognostication. However, many cancers remain incurable due to the development of drug resistance, minimal residual disease (MRD) and disease relapse, highlighting an incomplete understanding of the mechanisms underlying these processes. In recent years, the impact of non‐genetic factors on neoplastic transformations has increasingly been acknowledged, and growing evidence suggests that low oxygen (O2) levels (ie hypoxia) in the tumour microenvironment play a critical role in the development and treatment of cancer. As a result, there is a growing need to develop research tools capable of reproducing physiologically relevant O2 conditions encountered by cancer cells in their natural environments in order to gain in‐depth insight into tumour cell metabolism and function. In this review, the authors highlight the importance of hypoxia in the pathogenesis of malignant diseases and provide an overview of novel engineering tools that have the potential to further drive this evolving, yet technically challenging, field of cancer research.


| INTRODUC TI ON
Oxygen (O 2 ) tension in the body varies greatly, depending on the location and the physiological condition of the specific tissue. 1 The level of tissue oxygenation plays a critical role in both healthy and diseased physiological processes, such as ischaemia, tumours and inflammation. 2 In healthy tissues, O 2 concentration drops from 20% in the lungs to ~13% in the alveoli, and ~5% in the circulation. 3 The O 2 content in multicellular structures can further decrease to below 5%. Other tissue-specific O 2 levels include 5% in the venous blood, 1%-7% in the bone marrow, 0.5%-7% in the brain and 1% in the cartilage ( Figure 1A). 4,5 Increasing lines of evidence suggest that hypoxia is an innate facet of cancer, as the proliferation of malignant cells quickly exceeds the diffusion limit of O 2 (100-200 µm) resulting in inadequate oxygenation. 6 The vascular, metabolic and oncogenic adaptations that ensue are known to be critical to the biology of various cancers. Representing a therapeutic liability, tumour hypoxia is increasingly being explored for the development of personalized treatment approaches to influence tumour growth, metastatic potential and drug resistance. However, hypoxia-targeted strategies have only yielded limited success to date. Part of this unsatisfactory outcome may be attributed to the lack of appropriate experimental methods, which often involve the manipulation and study of cells exposed to non-physiologic O 2 concentrations and gradients that poorly reflect the physiologic conditions encountered by tumour cells in their natural environments. Hence, generating hypoxic conditions and hypoxic gradients in the in vitro setting has received increasing attention because hypoxia is capable of inducing, via hypoxia-inducible factor α (HIF-1α), a host of cell survival responses (eg autophagy). 7 In this review, we highlight several fundamental concepts of hypoxia, its metabolic adaptation and impact on tumour biology. We also discuss the need and recent progress of novel engineering tools and methodologies required to generate hypoxia and O 2 gradients, which are needed to further drive progress in this emerging field of research.

| HYP OXIA-INDUCIB LE FAC TOR S
The transcription factor hypoxia-inducible factor (HIF-1) is a key mediator for transmitting changes in O 2 tension into changes in genetic transcription allowing for cellular adaptation. [8][9][10] The level of HIF-1 ultimately regulates the expression of a wide range of adaptive processes, including the conversion from oxidative to glycolytic metabolism and angiogenesis. 11 Structurally, HIF-1 is a heterodimeric complex comprised of a stable beta subunit and O 2 -sensitive alpha subunits. Under normoxic conditions, prolyl hydroxylases (PHD) hydroxylate the alpha subunits of HIF, leading to ubiquitylation by the von Hippel Lindau (VHL) complex and subsequent proteasomal degradation ( Figure 1B). [12][13][14] Factor inhibiting HIF (FIH) also hydroxylates an asparagine residue of HIF-1α when O 2 is available, blocking its interaction with the transcriptional coactivator protein p300 and preventing transactivation of certain HIF target genes. 15,16 Hypoxia inactivates PHD and FIH, resulting in the accumulation of HIF-1 and its translocation to the nucleus where it interacts with HIF-1β and binds to hypoxia-response elements. 17,18 Notably, these regulatory mechanisms are also affected by the severity and duration of hypoxia. 19,20 In addition, it has been shown that HIF-1 is stabilized by an acidic intracellular pH, which often develops as a result of hypoxic metabolic changes. 21 Beyond being functionally important for the adaptation of normal and malignant cells to hypoxic conditions, HIF has been implicated in promoting genetic instability, immune evasion, migration and metastasis and stem cell maintenance. [22][23][24][25] Accordingly, elevated levels of HIF-1 have been demonstrated in some studies to be an independent negative prognostic indicator portending increased risk of metastasis, mortality and other adverse features in a variety of cancers including breast, lung and pancreas. 26,27 There is also evidence that HIF interacts with key tumour suppressor and proto-oncoproteins such as p53 and MYC. 28,29 However, for a minority of cancers, such as cervical cancer for example, it does not appear to have any prognostic significance. 30,31

| ME TABOLI C ADAP TATI ON S TO HYP OXIA
When O 2 availability decreases, cellular metabolism shifts from oxidative phosphorylation to the less efficient glycolysis. To maintain this process, pyruvate oxidizes NADH and is reduced to lactate via lactate dehydrogenase (LDH). As lactate accumulates, the cytoplasm becomes increasingly acidic which inhibits glycolysis, so lactate is excreted from cells by monocarboxylate transporters. 32 HIF-1 up-regulates the production of many glycolytic genes including isozymes of LDH that favour pyruvate reduction, lactate transporters and multiple other enzymes including hexokinase 1 and 3, aldolase A and C and pyruvate dehydrogenase kinase 1. 33-36 HIF-1 also up-regulates COX4-2, a subunit of complex IV of the electron transport chain which appears to be more efficient under hypoxic conditions and potentially generates less reactive oxygen species  37,38 When the demand for intracellular glucose increases, cancer cells can utilize glycogen to remain viable and proliferate. 39 This too limits the production of ROS, avoiding senescence. 40 Hypoxic cells can also utilize glutamine via both oxidative metabolism and reductive carboxylation. [41][42][43] Finally, in addition to glucose, glycogen and glutamine, there is evidence that hypoxic cancer cells may use other carbon sources such as exogenous acetate to produce acetyl-CoA, and perhaps nutrients released from organelles as a consequence of autophagy. 44,45 Furthermore, it has been shown that there is metabolic interplay between hypoxic and normoxic tumour regions. For example, tumour vascular endothelial cells have been noted to be highly glycolytic, thus allowing more O 2 to reach further into the tumour. 44 Another study has shown that a symbiotic relationship can exist between normoxic tumour regions that oxidize lactate to spare glucose and hypoxic tumour regions which metabolize glucose into lactate, thus providing a metabolic substrate for the normoxic regions. 45

| THE ROLE OF HYP OXIA IN SOLID MALIG NAN CIE s
In solid cancers, hypoxic tumour cells respond by producing angiogenic factors, but this pathologically induced process yields new vessels that are structurally and functionally suboptimal compared with vessels produced by well-coordinated physiologic angiogenesis. 6 Chaotic non-laminar blood flow, leakiness, and vascular remodelling lead to dynamic changes in O 2 delivery, with hypoxia lasting from seconds to days or of a cyclical nature. 46 The result of the high metabolic demands of malignant cells combined with limited O 2 delivery due to abnormal vasculature is that even highly vascularized cancers or tumour regions can contain areas of severe hypoxia. 11 Similarly, regardless of the degree of hypoxia, the pO 2 level of a tumour is always lower than corresponding normal tissue, resulting in hypoxia relative to physioxia.
Hypoxic conditions lead to elevated genomic instability, the selection of cells that have diminished DNA repair (down-regulated MLH1, MSH2, RAD51) and apoptotic potential (TP53 mutations), and a dampening of the antitumour immune response. [47][48][49][50][51] It also leads to the development of protective stem cell niches and enhanced expression of multidrug resistance proteins. [52][53][54] Furthermore, the lower rate of proliferation of hypoxic cancer cells decreases the effectiveness of cytotoxic chemotherapeutics that work best in actively dividing cells. Finally, in addition to resistance to chemotherapy and radiotherapy, hypoxia has been shown to contribute to resistance to immunotherapy via a variety of mechanisms including down-regulation of MHC-I and up-regulation of immune checkpoints. 55,56 The net result of which is that low O 2 levels in solid cancers can generate a more mutagenic and treatment-resistant phenotype. As a result, tumour hypoxia has been linked to unfavourable cancer outcomes. In prostate cancer, for example, hypoxia has been associated with biochemical relapse independent of factors such as Gleason score, prostate-specific antigen (PSA) levels or Tcategory. 57 Another study found an association between HIF-1 and vascular endothelial growth factor (VEGF) expression on diagnostic tumour biopsies and biochemical relapse following radiotherapy or radical prostatectomy, although it has been acknowledged that factors unrelated to hypoxia may up-regulate HIF-1. 58 Patients with head and neck cancer treated with radiation alone were found to have an association between tumour hypoxia, as measured with electrodes, and inferior overall survival (OS) and higher rates of local recurrence. 59,60 Studies have also shown that hypoxia seems to increase the propensity for metastatic disease across multiple cancer types, such as cervical and breast cancer. [61][62][63] A limitation of the available clinical data is that a variety of different techniques were used to measure hypoxia, each with their attendant advantages and disadvantages, which have been reviewed elsewhere. 2 There have been various attempts to therapeutically exploit hypoxia as a differentiating metabolic characteristic of malignant cells.
These have included radiosensitizers, antiangiogenics and hypoxiaactivated pro-drugs amongst others. For example, one approach involving a combination of accelerated radiotherapy, the inhalation of carbogen (98% O 2 and 2% CO 2 ) and the vasoactive compound nicotinamide (ARCON) was compared with accelerated radiotherapy alone in a phase III trial of patients with laryngeal cancer. 64 The combined approach resulted in a statistically significant improvement in regional control, albeit without improvement in local control.
Another phase II trial of radiation, carbogen and nicotinamide compared to radiation alone in patients with locally advanced bladder cancer resulted in a significant improvement in OS and local relapse rates with the hypoxia-directed treatment. 65 An example of a radiosensitizer that has been studied is nimorazole, which was tested in combination with radiation in a phase III trial of patients with supraglottic laryngeal and pharyngeal cancers versus placebo and radiation, and demonstrated improved locoregional control. 66 However, a number of hypoxia-activated pro-drugs demonstrating promising early activity in phase I and II trials ultimately led to negative phase III trials. These include tirapazamine in head and neck cancer and evofosfamide (TH-302) in advanced pancreatic cancer and soft tissue sarcomas. 67,68

| THE MOLECUL AR HALLMARK S OF HYP OXIA
One major challenge in directly measuring the extent of hypoxia in malignant tissues is the significant amount of both intratumoral heterogeneity and intertumoral heterogeneity in O 2 status for each cancer type, which can change over time. There has therefore been a growing effort to understand the molecular hallmarks of hypoxia, ultimately using diagnostic tumour biopsies as both an indirect reflection of the broader O 2 microenvironment over time and to deduce a given tumour's dependence on hypoxia for its proliferation. To this end, single-nucleotide variants (SNVs) and copy number aberrations (CNAs) of TP53, MYC and PTEN have consistently been associated with hypoxia in multiple cancer types. 69 There was, however, a notable degree of variation in SNV hypoxia signatures between tumour types, which emphasizes the need for further, in-depth studies in each malignancy. Another study identified a genomic signature of the metabolic shift associated with tumour hypoxia across multiple cancer types. 70 In addition to protein and mRNA, microRNA expression has been associated with hypoxia. 71,72 For example, miR-210 abundance was associated with hypoxia across 18 tumour types in one study, although more studies are needed to determine their precise regulatory role. 69 Some limitations of the clinical utility of hypoxia biomarkers include a degree of dependence on adequate sampling of the tumour to account for special heterogeneity and that many markers are regulated by both hypoxia-dependent and hypoxia-independent mechanisms.
It is becoming increasingly clear that microenvironmental pressures such as hypoxia may be shaping the mutational architecture of cancer, selecting for subclones with aggressive features. 69 A major challenge is to identify those tumours with a "hypoxic driver" molecular or genetic phenotype in which hypoxia is a primary driver of the cancer's behaviour, as this subgroup will be enriched in predictive value for response to hypoxia-targeted treatments. 73 For example, a retrospective analysis of the aforementioned nimorazole trial examined a 15-gene hypoxia panel in pre-treatment biopsies and found that only patients with hypoxic tumours as determined by the panel had improved local control and survival. 74 It has also been hypothesized that hypoxic niches in tissues such as the bone marrow may provide shelter to cancer stem cells and are at least partially responsible for treatment resistance in leukaemia and other diseases. 75 Therefore, a greater understanding of the key molecular pathways underpinning hypoxic cancer cells' resistance to treatments may promote the development of novel targets and therapies.
However, a major reason why a significant knowledge gap persists in our understanding of the role of hypoxia is the difficulty of studying cancers ex vivo, which often involves the use of un-physiologic cell culture techniques carried out in ambient air or in chambers that maintain a constant level of hypoxia. Therefore, progress in our understanding of the molecular characteristics of hypoxia, as well as its therapeutic exploitation, will likely require a tandem progress in experimental models.

| PATHOPHYS IOLOG IC AL EFFEC TS OF OX YG EN G R AD IENTS
Increasing lines of evidence suggest that O 2 gradients might play an important role in the process of drug resistance and cancer cell survival, potentially by providing "escape routes" along which neoplastic cells migrate when a cell death signal is activated by cytotoxic therapy. In fact, cellular migration along gradients, including chemokine, 76 cytokine 77 or growth factor 78 gradients, has long been recognized as a fundamental process in cellular adaption.

| HYP OXIA-INDUCING CHEMIC AL S
Intracellular hypoxia-like responses can be created or mimicked by using chemical reagents, such as sodium sulphite (Na 2 SO 3 ),  can be rapidly and conveniently established (Figure 2A,B) through adjusting the flow rate of medium pre-equilibrated with lower oxygen tension. 95 Shih et al showed that the spatially confined chemical reaction could generate stable O 2 gradients within the microfluidic device (21% O 2 nomaxia and 1% O 2 hypoxia). 101 The O 2 scavenging chemical reaction between pyrogallol (benzene-1,2,3-triol) and

| Microfluidic devices
NaOH occurred in the chemical reaction chamber (Figure 2C,D).
When pyrogallol is added in alkaline solution, it absorbs O 2 rapidly and creates a "sink" that induces a unidirectional diffusion of O 2 to generate an O 2 gradient ( Figure 2E). It is possible to alter the range and steepness of the gradient O 2 in the same device by changing the composition of the gas mixture fed into the culture areas with different sizes and shapes. 103 The disadvantages of microfluidic systems include complicated manufacturing processes, the need of flow control instruments and device set-up. In addition, it is not suitable for long-term or large-scale cell studies. 104

| Enzymatic reactions
Recently, O 2 -consuming enzymes have been exploited as an alternative strategy to create hypoxic culture environments. 105 The most widely used O 2 -consuming enzyme is glucose oxidase (GOX), which converts glucose, oxygen and water into gluconic acid and hydrogen peroxide (H 2 O 2 ). 106 An endogenous enzyme, GOX, has been used in cancer diagnosis and treatment. For example, the consumption of glucose and oxygen may be exploited for cancer-starvation and hypoxia-activated therapy, respectively. 107 On the other hand, the reaction product gluconic acid may be employed for pH-responsive drug release. Finally, H 2 O 2 generated in the reaction can be converted into toxic hydroxyl radicals for cancer cell killing. 107 While the reaction of GOX is fast and effective, one significant drawback for its application in cell studies is the production of cytotoxic H 2 O 2 , the accumulation of which can lead to undesired cellular toxicity, but can also inactivate GOX. 108,109 To minimize the cytotoxic by-product of GOX reactions, catalase (CAT) can be used to reduce H 2 O 2 into water.
However, this reaction partially offsets hypoxia by producing half an oxygen ( Figure 3A). Dawes et al designed GOX immobilized polyethylene glycol diacrylate (PEGDA) hydrogel for extended hypoxic cell cultures ( Figure 3B). 105 Immobilization of O 2 -consuming GOX within covalently cross-linked hydrogels provides an easy method to control solution O 2 tension without using external devices (2.5%-9%).  Figure 4A). 115,116 The duration of hypoxia can be extended by increasing the thickness of the HI hydrogels.
As the thickness of the hydrogel increases, the diffusion of O 2 in the media or atmosphere decreases and the hypoxic duration in the matrix increases (Figure 4B,C). Laccase-mediated reactions were shown to be cytocompatible, and the HI hydrogels were supportive of vas-

ACK N OWLED G EM ENTS
This work was supported in part by the National Cancer Institute (R01CA227737, to CL) and IU Simon Comprehensive Cancer Center/

Walther Cancer Foundationvia an Oncology Physical Sciences &
Engineering Research Embedding Program Award (to CL, HK).

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analyzed in this study.