Engineering Biomaterials to Model Immune‐Tumor Interactions In Vitro

Engineered biomaterial scaffolds are becoming more prominent in research laboratories to study drug efficacy for oncological applications in vitro, but do they have a place in pharmaceutical drug screening pipelines? The low efficacy of cancer drugs in phase II/III clinical trials suggests that there are critical mechanisms not properly accounted for in the pre‐clinical evaluation of drug candidates. Immune cells associated with the tumor may account for some of these failures given recent successes with cancer immunotherapies; however, there are few representative platforms to study immune cells in the context of cancer as traditional 2D culture is typically monocultures and humanized animal models have a weakened immune composition. Biomaterials that replicate tumor microenvironmental cues may provide a more relevant model with greater in vitro complexity. In this review, the authors explore the pertinent microenvironmental cues that drive tumor progression in the context of the immune system, discuss how these cues can be incorporated into hydrogel design to culture immune cells, and describe progress toward precision oncological drug screening with engineered tissues.


Introduction
Each year, an estimated 2 million individuals are diagnosed with cancer in the US alone, and over half a million succumb to this immunodeficient mice and have resulted in their growing prevalence in recent years. [9,10]These models retain the biological heterogeneity and complexity of the original tumor; however, immunodeficient mice fail to fully capture immune-cancer interactions limiting their utility in emulating immune evasion, and thus reducing their clinical translatability.To overcome this challenge, mice with humanized immune systems have been developed. [11]Yet, the low engraftment rates (< 5-50% based on tumor compatibility/aggression with the selected model) make large-scale studies difficult to manage and impractical for drug discovery. [12]Moreover, inter-species cell interactions confound the reliability of the system through poor compatibility and cellular component deficiencies in host models. [13]iomaterials have the potential to build on the benefits of each of 2D tissue culture polystyrene (TCP)-based drug discovery (i.e., high throughput and reproducibility) and PDX models (i.e., greater complexity).The 3D ECM-mimetic biomaterials increase 2D system complexity while reducing time, cost, and the interspecies confounding factors of in vivo PDX models (Figure 1).While 3D biomaterials were first deemed attractive due to their ability to provide circumambient scaffolding in which cells could reside, they have since been developed to incorporate complex microenvironmental cues that impact cell behavior and phenotype.For example, cytotoxic drugs evaluated in 2D TCP co-culture were shown to reduce glioblastoma cell proliferation and migration in the presence of microglia; yet, when the same cells were cultured in 3D, hyaluronan-based hydrogels, the cytotoxic drugs had the opposite effect and increased glioblastoma cell proliferation. [14]Yet, the predictive power of 3D culture systems is still nascent, and they have yet to be incorporated into the drug discovery pipeline.The myriad of biomaterial compositions, chemistries, and experimental conditions studied have complicated their robust implementation in the preclinical space to date; however, there is a drive for change, influenced in part by the need for more predictive in vitro drug screens and the goal of reduced animal testing.In this perspective, we delve into how biomaterials have been used to transform our understanding of tumor immune evasion and we strive to elucidate their potential as tools for either patient-specific drug discovery or mechanistic studies.

Immune Cell Phenotype is Influenced by the TME
Multiple immune cell types that are part of the native TME have been extensively studied in vitro (Figure 2).In this section, we briefly describe the common immune cell types studied in vitro in the context of cancer.Myeloid-derived Suppressor cells (MDSCs) are immature cells that rapidly expand during infection and are driven by tumor-derived cytokines, such as interleukin-6 (IL-6), granulocyte colony-stimulating factor (G-CSF), and granulocyte-macrophage colony-stimulating factor (GM-CSF).These cytokines are secreted in the immunosuppressive TME and stimulate a positive feedback loop to promote tumor growth and MDSC accumulation, and further push the tumor into an immunosuppressive state.G-CSF and GM-CSF support the survival and proliferation of tumor cells and increase MDSC accumulation, further suppressing the immune response.Additionally, IL-6 dampens immune cell activity, allowing cancer cells to evade the immune system.Upon stimulation and activation, MDSCs become immunosuppressive.Additional cytokines secreted from tumor cells, such as transforming growth factor- (TGF-) and reactive oxygen species (ROS), also contribute to this immunosuppressive phenotype by inhibiting immune cell function.TGF- is upregulated in stiff ECM matrices and resultingly increases the secretion of matrix metalloproteinases (MMPs) to facilitate tumor cell invasion and metastasis. [15,16]Dendritic cells (DCs), a group of antigenpresenting cells (APCs), are crucial for initiating immune recognition and regulating immune responses.The TME can hinder DC maturation through inhibitory signals and elevated secretion of immunosuppressive factors, which decrease antigen presentation and reduce T cell activation.Tumor-associated macrophages (TAMs) are recruited to the TME as monocytes and differentiated and polarized typically towards a pro-inflammatory phenotype, which supports ECM remodeling and boosts genetic instability in cancer cells.TAMs can secrete immunosuppressive cytokines, such as IL-10, IL-1, and transforming growth factor-beta (TGF-), which suppress effector T cell (Teff) function, promote tissue remodeling and angiogenesis, and inhibit cell death signaling pathways in tumor cells.Histological analyses of numerous tumors correlate with increased TAM presence, worse clinical outcomes, and greater resistance to therapy. [17,18]AMs and their polarization state contribute to the regulation of the function and activation of tumor-infiltrating lymphocytes (TILs).TILs are primarily comprised of regulatory T cells (Tregs), helper T cells (CD4+ T cells), cytotoxic T cells (CD8+ T cells), and natural killer (NK) cells.Tregs play a crucial role in maintaining immune tolerance and preventing autoimmune diseases.In particular, the accumulation of Tregs in the TME can suppress the anti-tumor immune response with the secretion of immunosuppressive cytokines like IL-10 and TGF-.These cytokines negatively impact the function of Teff cells, which are responsible for recognizing and eliminating tumor cells.Conversely, certain cytokines, such as interferon-gamma (IFN-), can promote a cytotoxic T cell phenotype.Meanwhile, NK cells are also sensitive to these immunosuppressive cytokines, which can inhibit their cytotoxicity, degranulation, and IFN- production.In addition to regulation caused by cytokine secretion and direct contact with TAM surface factors, TILs are also controlled by both immune checkpoint molecules and the ratio of T cell subtypes (i.e., CD8/CD4) in the tumor. [19]Importantly, the phenotype of each immune cell is also regulated by mechanosensitive receptors and ion channels. [20,21]These are influenced by the cytoskeletal structure of the cell where mechanosensory inputs are converted into biochemical signals that interact with the surrounding TME.

Biomaterial Scaffolds Modulate Immune Cell Phenotype
Biomaterials can incorporate ECM-driven cues to influence immune cells.In 3D, cells can spread and interact more extensively with both each other and the surrounding ECM.Importantly, the biomaterial itself may alter immune cell phenotype and thus, without proper consideration, tissue engineers risk inducing artificial immune signatures, which could create tumor-immune interactions not observed in vivo.In this section, we discuss common strategies, summarized in Figure 3, to create biomaterialbased scaffolds in which immune cells reside and their subsequent impact on phenotype.

Materials Native to the TME
To incorporate ECM signaling found in the TME, numerous research groups utilize naturally derived materials from tissue.Ma-trigel is the current standard in 2D screens as a coating material to improve cell spreading and remains actively used in research as an ECM material for 3D culture and xenotransplantation.Derived from the ECM of the Engelbreth-Holm-Swarm (EHS) mouse sarcoma tumor, Matrigel contains various glycoproteins, such as collagen IV and laminin, as well as tumorderived growth factors and small molecules. [22]While this suggests that hydrogels based on Matrigel would be good candidates to model the complex ECM of the TME, Matrigel is ill-defined chemically and, consequently, has limited utility (despite its popularity).Large variation both within and between batches results in variable protein and small molecule composition, often resulting in irreproducible results due to the presence of over 16 000 unique peptides, proteins, and growth factors. [22,23]The change in growth factor concentrations between Matrigel batches can influence macrophage and NK cell phenotypes.As a result, growth factor reduced Matrigel may be more desirable; however, this does not diminish the difference in glycoprotein concentrations, which can result in dramatic changes in stiffness (up to 2-fold) between different lots. [24]These variations undermine the utility of Matrigel in standardized drug screens because reproducibility is essential.
Variations can be reduced by using collagen I scaffolds, which are derived from rat tail and bovine sources.In vivo, increased collagen density and fiber alignment are strong negative prognostic factors in multiple cancers, including breast, colorectal, and hepatocellular carcinoma, [25] resulting in reduced patient survival.This has led to collagen's popularity when studying cancer-immune interactions.Immune cells can interact directly with collagen, binding to collagen integrins such as 11 and 21.This interaction influences macrophage phenotype.For example, blocking 21 binding in THP-1 monocytes induced a pro-inflammatory phenotype with increased expression of interferon regulatory factor 5 (IRF5) and IL-6, and decreased expression of signal transducer and activator of transcription 6 (STAT6) and IL-10. [26]In vitro, increasing collagen matrix density increased the immunosuppressive phenotype in macrophages, and thereby reduced T cell signaling and proliferation. [27]Notably, many T cells lack integrin alpha 1 (ITGA1) expression, a known receptor of collagen VI.As a result, T cell spreading, viability, and motility are compromised on collagen VI matrices and thus other matrices (i.e., collagen I) are required to culture T cells. [28]Mechanical strength and matrix density can be controlled by either increasing collagen concentration and hence physical crosslinking, [29] or introducing chemical crosslinks through modification of the backbone. [30]The chemical modification allows bioactive growth factors and peptides to be conjugated to the backbone and can increase the modulus and density of the fibrous network; however, both stress relaxation and gel permeability are often reduced, [30,31] which, in turn, impact immune cell behavior and phenotype.Hydrogel modification with bioactive peptides will be further discussed in the next section.
Gelatin, or denatured collagen, is frequently used as a substitute for collagen due to its commercial availability and susceptibility to proteolytic degradation via MMP-2/9s, which in turn improve screens that require matrix remodeling/invasion as a readout.In consideration of immune cell cultures, it has been found to be more immunogenic than collagen, increasing both pro-inflammatory gene expression and anti-inflammatory cytokine secretion, such as IL-10 and Arg1 in macrophages. [32]elatin forms into a hydrogel after chemical modification, most frequently through the introduction of methacrylate groups (GelMA) or thiols (Gel-SH).Notably, GelMA is a commonly used bioink for 3D bioprinting, [33] which may significantly improve the reproducibility of scaffold structures.
Hyaluronan (HA) is a naturally occurring polymer present throughout the body.HA is a popular material due to its biocompatibility and ease of chemical modification, enabling facilitated crosslinking and/or conjugation of growth factors.[36] While the role of HA on immune activation is a topic of active debate, the molecular weight of HA seems to impact its immunomodulatory properties. [34]Low molecular weight (LMW) HA (< 500 kDa) enhances a pro-inflammatory and immunostimulatory macrophage phenotype (i.e., a classically activated pro-inflammatory signature) with increased cytokine secretions, such as TNF-, IL-1, and nitric oxide.In contrast, high molecular weight (HMW) HA (> 1000 kDa) stimulates an immunosuppressive response [34,37,38] and may also induce pro-angiogenic behavior in macrophages in selective tumor types, causing increased secretion of VEGF, IL-8, FGF-2, and MMP-2 in breast cancer, but not colorectal cancer models. [39]While HMW HA is naturally present in homeostasis (1-10 MDa), secretion of hyaluronidases and free radicals during tumor growth and development will degrade the long polymer chain, increasing the presence of LMW HA and thus further promoting the pro-inflammatory microenvironment.Independent of molecular weight, many immune cells -including T cells, NK cells, and macrophages -express CD44, a major receptor for HA.The overexpression of hyaluronic acid synthase 2 (HAS2) in tumor-associated stromal cells has been correlated with malignancy in colorectal and breast cancers. [40,41]nockdown of HAS2 (the enzyme responsible for the majority of HA production) in fibroblasts, and the resulting reduction of HA in the TME, has been shown to reduce TAM recruitment and hinder neovascularization in breast cancer. [42]

Materials Foreign to the TME
Materials not found in the TME, both naturally occurring and synthetically derived, are an attractive option to overcome the cost, reproducibility, and immunogenic challenges associated with naturally derived materials.Alginate is a naturally occurring biopolymer derived from seaweed frequently used in cell culture for immune-cancer co-cultures due to its relatively low cost.The ionic gelation and incorporation of proteins into alginate occur in aqueous conditions that are favorable to encapsulated cells.Rigidity, porosity, and stability can be tuned by calcium ion concentration.[45] Alginate may contain impurities, such as lipopolysaccharides (LPS) or endotoxins, which stimulate macrophages towards a pro-inflammatory state at concentrations as low as 0.05 EU mL −1 . [46]Even "low-endotoxin alginate" typically has 100 EU g −1 , which can activate certain immune cells.This underscores the importance of characterizing materials for endotoxin levels, which impact immune cell response both in vitro and in vivo.
Synthetic materials, which can be industrially synthesized using well-established polymerization techniques, improve batchto-batch reproducibility and limit xenogeneic interactions which can influence cell behavior.Most popular in this group are poly(ethylene glycol) (PEG) and poly(N-isopropylacrylamide) (PNiPAM), which are considered immunologically inert when cultured with macrophages due to their low immune response when implanted in vivo. [47]Hydroxy-terminated PEG, while appealing due to its low-cost, has been shown to activate immune cells in vivo. [48]While PNiPAM is a popular building block for hydrogel design due to its thermoresponsiveness, which results in a reversible sol-gel transformation at 32-33°C, PNiPAM-immune cell interactions are poorly characterized.As a result, in vitro coculture studies incorporating this polymer may result in nonbiologically relevant immune phenotypes.
Additionally, these materials all lack the ECM cues present in naturally derived materials which results in poor cell adherence to the material.When used without additional cell binding motifs, immune cells cultured in synthetic scaffolds have de-creased protein tyrosine kinase 2 (PTK2) gene expression, as well as lower expressions of F-actin and vinculin compared to their native counterparts. [26]This may be overcome with the bioconjugation of peptides, proteins, and/or other factors to "design" the hydrogel to support immune cell viability and growth.

Designer Hydrogels
Bioactive peptides conjugated to a polymer backbone enhance cellular interactions and help to chemically define the TME.For example, monofunctionalized binding peptides enhanced cell viability and proliferation of leukocytes [49] and macrophages, [50] and increased macrophage invasion in synthetic scaffolds. [51]he most commonly used peptide is arginine-glycine-aspartic acid (RGD), which is the principal integrin binding domain of a plethora of glycoproteins present in the TME, including collagen, fibronectin, laminin, and vitronectin.Incorporation of RGD into hydrogels increases immune cell viability and spreading. [50,52]his impacts cell cytoskeletal structure and improves T-cell signaling, activation, and immune synapse formation. [53]Additional peptides have been incorporated into hydrogels to improve cell spreading such as PHSRN (fibronectin), [50,51,54] DYIGSR (laminin), [55] and GFOGER (collagen). [55]Bioconjugation of peptides has also been used to tune macrophage polarity.Multiple groups have demonstrated that the collagen I-derived peptide, DGEA, which supports 21 binding and increased CD206 expression, resulted in a large population of macrophages polarized towards a pro-reparative state. [26,56]i-or tetra-functionalized peptides are often used as chemical crosslinkers in HA hydrogels and synthetic scaffolds.While PEG-based crosslinkers are very common, peptide crosslinkers can be designed to allow dynamic remodeling of the matrix and proteolytic-based cell invasion.The most common are peptide crosslinkers which can be cleaved by proteases, such as MMPs. [49]MMP recognition sequences typically cleave peptide bonds between a neutral or small amino acid (i.e., Gly) and a hydrophobic amino acid (i.e., Leu, Val, or Ile).For example, GPQG↓IWGQ is an MMP degradable sequence where the arrow indicates the site of cleavage by MMP-2 or -9.MMP degradable peptides can be designed in silico with tools such as CleavePredict. [57]When cultured in 3D, macrophages showed significant upregulation of MMP-9 and MMP-14.Interestingly, the incorporation of MMP-sensitive sites into these scaffolds may also increase secretions of pro-inflammatory cytokines such as TNF-. [58]

Mechanical Properties of the Biomaterial Impact Immune Cell Phenotype
[61][62] This can be a particularly interesting property to study due to the increase in rigidity of many local solid tumor sites.TCP is used in 2D screens; however, with Young's modulus on the order of 3 GPa, TCP stiffness is significantly different from the native TME in Decreased cell cluster size, increase in all measured markers in stiffer gels [58]   Human cord bloodderived macrophages GelMA 2/29 kPa Morphology, secretome TNF-alpha, IL-6, iNOS (M1) TGF-beta, Arg-1 (M2) Increase in cell spreading, increased iNOS, TNF-alpha, IL-6 decreased TGF-beta, Arg-1 [68]   vivo, resulting in a pro-inflammatory secretome (as evidenced by enhanced expression of TNF- and IL-2), and generating IC 50 value curves that are significantly different than those observed in 3D hydrogels and in vivo animal studies. [58,63,64]Macrophage polarization is sensitive to mechanical stiffness; however, reports on the impact of matrix rigidity are often contradictory.Several groups have reported macrophage polarization towards an anti-inflammatory phenotype when cultured on stiffer substrates through mediation of the Piezo1 ion channel and/or activation of the ROS-initiated NF-B pathway. [21,65]This phenomenon has been demonstrated with several scaffold materials including GelMA, [66] polyacrylamide, [65,67] PEG, [68] fibrin, [69] and agarose. [70]Similarly, T-cells show increased activation in stiffer gels, having amplified inflammatory cytokine expression. [71,72]In contrast, the opposite shift in macrophage secretome and activation states has been reported with increased stiffness on other materials -chitosan, [73] gellan gum, [50] and gelatin hydrogels. [74]oreover, Kim et.al. observed an increase in both pro-(iNOS, COX2, and TNF-) and anti-(CD206, Arg-1, and TGF-) inflammatory macrophage mRNA expression on stiffer PEG hydrogels compared to stiff hydrogels and 2D TCP culture. [58]These different findings likely result from different cell sources, as well as the wide variety of material sources, which remains a confounding factor for in vitro biological assays.Differences may also be due to variations in the definition of a "stiff" scaffold and differences in Young's modulus measurements; some define stiff to have moduli between 100-9000 Pa gels whereas others between 10 000-100 000 Pa.While many of the articles above use stringent definitions of pro-and anti-inflammatory states ("M1" and "M2"), it is becoming widely appreciated that macrophage activation exists on a spectrum.It may be that these reported discrepancies are due to this oversimplification of macrophage phenotype.Table 1 summarizes the role of material stiffness on macrophage phenotype.Interestingly, through each study, stiff substrates seem to promote an increase in cell area and cell spreading, which is often associated with an anti-inflammatory phenotype in 2D culture.
T cells can also sense the surrounding mechanical environment.T cells cultured in stiffer (≈40 kPa) alginate gels showed increased proliferation and CD25 expression compared to those cultured in softer (≈4 kPa) gels, as well as an increase in IL-2, IFN-, and TNF- cytokine expression observed on stiffer gels. [71]Beyond material stiffness, viscoelasticity can also impact immune cell phenotype.Viscoelasticity reflects both viscous and elastic characteristics of a material under stress and is measured through stress relaxation of the material over time.While it proves challenging to decouple matrix stiffness from viscoelasticity completely, Adu-Berchie et.al. maintained a consistent Young's modulus and porosity between collagen hydrogels by either relying on physical crosslinks for gelation (high-stress relaxation) or introducing chemical crosslinks (low-stress relaxation) with norbornene-tetrazine click chemistry.Interestingly, T cells cultured in slow-relaxing (vs fast relaxing) collagen matrices (60% relaxation in > 1000 s) were more highly activated, Reprogrammed through altered expression of metabolic enzyme and transcription factors such as lactate dehydrogenase-A (LDHA) [92,93]   expressed more inhibitory markers, and had several upregulated genes related to the activating protein-1 (AP-1) pathway, a critical regulator of T cell activation, differentiation, and maturation.In contrast, T cells cultured in fast-relaxing collagen (60% relaxation in < 10 s) expressed significantly more memory markers such as CD62L, Kruppel-like transcription factor 2 (KLF2), CXC motif chemokine receptor 3 (CXCR3), and sphingosine-1-phosphate receptor 4 (S1PR4). [75]atrix stiffness impacts immune cell activation and phenotype and thus hydrogels should be designed to emulate the stiffness of the host tissue.TCP culture exhibits stiffness values over 4 orders of magnitude greater than most solid tumor tissues and thus likely promotes immune cell phenotypes that do not typically reside in the native TME.Interestingly, Sridharan et.al. demonstrated macrophage polarization may be dependent on both stiffness and the crosslinking chemistry used to formulate the hydrogels. [67]They showed large differences in macrophage polarization based on whether the gel was physically or chemically crosslinked; however, Cuenot et al. showed that crosslinking chemistry influences viscoelasticity, which was not explored by Sridharan et al. [76] When hydrogel viscoelasticity was changed, T cell phenotype was impacted. [75]

Hydrogel Porosity Influences Immune Cell Phenotype
Hydrogel porosity has been shown to influence cell aggregation, spreading, and migration speed. [77]The liquid-filled pore structure mimics native tissue and promotes the diffusion of growth factors, cytokines, and other essential nutrients.Micropore structures (< 500 μm) are typically formed by cryogelation, [43,51] salt or sugar leeching, [78] or dissolution of colloidal/microgel suspensions [79,80] whereas macropore structures (> 500 μm) have been formed with 3D printing [43,81] In-depth methods for generating pore structures in hydrogel scaffolds have previously been reviewed. [81]igh porosity hydrogels, with pores between 100-200 μm, promote cell adhesion and migration.NK cells aggregate more easily in porous hydrogels, enhancing cell viability, proliferation, degranulation activity, and cytokine release. [77,82]Increasing either pore size or porosity can support the pro-inflammatory polarization of macrophages through the regulation of a STATrelated mRNA transcriptional pathway. [43,83]Along with an increase in Arginase 1 (Arg1) and a decrease in inducible nitric oxide synthase (iNOS), larger pore sizes also resulted in the secretion of more pro-angiogenic factors: VEGF, TGF-1, and bFGF. [83]Almeida et.al. [84] showed that larger pore sizes (250 vs 75 μm) resulted in increased TNF- production and elongation of embedded macrophages, which matches functional studies that correlate elongated cell shape to a pro-inflammatory phenotype. [85]

Mimicking Immune Metabolic Programming in Biomaterials
As the TME has many different cell populations, these diverse cell types in the TME are metabolically heterogeneous (Table 2).Metabolic stress plays a significant role in all cell types and is especially important for the growth and progression of cancer cells within the TME.This stress is influenced by the cell phenotypes present in different regions of the tumor, where the cell populations compete for limited oxygen and nutrients.Hypoxia is a common feature of solid tumors due to poor vascularization at the core, which refers to low oxygen levels.In these poorly vascularized and metabolically active areas, the availability of essential resources necessary for cell survival is constrained and metabolic lymphatic drainage is severely impaired. [86,87]hen metabolic waste accumulates, an unfavorable pH and harsh environment for immune cell functions and survival results.Hypoxic conditions shift tumor cells towards anaerobic cellular respiration, which generates an excess of lactate, leading to an acidic TME that further promotes immune suppression and immune evasion.The resultant secretion of hypoxiainducible factors (HIFs) has been observed to regulate tumor immunity by influencing tumor cells, immunosuppressive cells, Teff cells, and APCs.[79][80] Furthermore, immunosuppressive macrophages are recruited to the tumor, thereby enhancing immunity.Therefore, modeling the metabolic stress characteristic of tumors and immune cells in the TME may provide insight into how tumors will respond to different immunotherapy candidates.
In vitro, hypoxia can be generated through culture in a low O 2 (1% O 2 ) incubator or chemically with cobalt (II) chloride hexahydrate (CoCl 2 • 6H 2 O) [94] ; however, 3D culture allows for the generation of hypoxic gradients which provides a more physiologically relevant milieu to study tumor and immune metabolic profiles.Spheroids and organoids create oxygen and nutrient gradients in a size-dependent manner, yet are difficult to probe dynamically, and thus limit the interpretability of the system. [95,96]o overcome this challenge, biomaterials can be integrated into devices to induce hypoxic and metabolic gradients.For example, the Tissue Roll for Analysis of Cellular and Environment Response (TRACER) device comprises a cellulose scaffold rolled around an oxygen-impermeable mandrel to create a layered stack of cells that can be analyzed in a layer-by-layer manner.By co-culturing peripheral blood mononuclear cell (PBMC)derived macrophages with pancreatic ductal adenocarcinoma in TRACER, Co et.al. [97] displayed gradients of known hypoxia response genes involved in glycolysis, such as glucose transporter protein type 1 (GLUT1), and the expression of immune checkpoint inhibitors, such as PD-L1.Interestingly, macrophages in this study cultured in hypoxic conditions were more chemoprotective against gemcitabine and T cells exposed to conditioned media from hypoxic macrophages had fewer inflammatory T cells.
Microfluidics is another common method to establish physiologically relevant oxygen gradients.One such method by Pedron et.al. [98] involved the creation of a cell metabolic gradient by using a microfluidic device to create either cell or material gradients.After 7 days in culture, they detected a significant increase in HIF-1 gene expression in areas of either high cell or GelMA crosslinking density and observed a lateral gradient of secreted HIF-1 and VEGF protein via immunofluorescence.Microfluidics can also be used to establish vascular-mimetic milieus: porous membranes lined with microvascular endothelium on one side and a thin epithelium layer on the other create a model of the vascular wall bordering the perfusable channel with the hydrogel compartment.Grant et.al. [99] controlled the oxygen permeability of their device through the incorporation of gas-impermeable films that led to cells inducing their own hypoxic environment.Notably, these devices can be embedded with oxygen sensors to provide continuous feedback throughout the experiment.Beyond studying hypoxic gradients, microfluidic devices also provide a relevant platform to study immune cell infiltration into the tumor through the biomimetic vascular endothelium.

TME Guides Immune Cell Infiltration and Invasion
Beyond angiogenesis, which provides a pathway for immune cells to reach the tumor site, histological staining has revealed that ECM remodeling is driven by the excessive deposition of collagen and HA in the tumor stroma, resulting in both a stiffer and denser matrix (Figure 4).Genetic studies have corroborated this finding with the upregulation of the HA synthesis class of genes, such as HAS2, and collagen in tumor cells.Similarly, the increased density of either collagen or HA hydrogels results in significant reductions in immune cell motility, including in macrophages, T cells, DCs, and NK cells. [51,100]This impediment to immune cell migration compromises the effectiveness of immune surveillance within the tumor.
The density, spatial distribution, and composition of immune cells in solid tumors are closely related to the clinical outcome of many cancers and their response to treatment, including breast cancer, [101] pancreatic cancer, [102] non-small-cell lung carcinoma, [103] and gliomas. [104]As many immune cells infiltrate solid tumors through blood vessels, those in the TME allow the infiltration of pro-tumorigenic immune cells, which can be targeted in drug screens.Tumor vasculature differs greatly from normal vasculature as it is more permeable and has heterogeneous flow patterns.Cancer and immune cell extravasation have been extensively investigated in co-culture microfluidic devices. [105]Endothelial barrier function has been evaluated on a thin, porous membrane to mimic the blood vessel wall against the tumor.Cancer and/or immune cells have been either seeded directly within the microfluidic channel to mimic intravasation processes or perfused through a channel against an endothelial cell barrier to model extravasation. [106]he fluidic flow of the "circulating" immune cells must be appropriately designed.Interstitial fluid flow, as low as 3 μm s −1 within the TME, induces an immunosuppressive state in macrophages through integrin/sarcoma (SRC)-mediated mechanotransduction pathways. [107]Precise unidirectional control of flow can be reproducibly achieved with microfluidic devices.In vitro, macrophages exposed to fluid flow ranging from 0.5 to 6.5 μm s −1 adopted a more immunosuppressive phenotype, similar to that observed in vivo, and when in triculture conditions with HUVECs and breast cancer cells, angiogenic sprouting into a collagen gel increased compared to cultures without macrophages (Figure 5a). [108]nce flow conditions are established, biomaterials incorporated into microfluidic platforms allow immune cell infiltration past a thin endothelial cell wall barrier to be studied.The ability of macrophages to invade the biomaterial is dependent on matrix stiffness, with fewer macrophages invading into stiffer gels (174 vs 1666 Pa). [109]Ayuso et.al. developed a biomimetic blood vessel that enables NK cell infiltration into a collagen gel.This formed a hypoxic and nutrient gradient, resulting in reprogramming the NK cell metabolic state towards a lower redox ratio. [110]

Tumor-Guided Physical Remodeling of the Extracellular Matrix Hinders Immune Cell Invasion and Supports Tumor Immune Evasion
Macrophages and T cells, through collagen fiber phagocytosis, can dynamically remodel collagen-based matrices to increase the velocity of immune cell infiltration.Sajedi et.al. [111] characterized T cell invasion and found, upon first entry into the matrix, cells undergo a "random walk", moving more slowly in areas of increased density and more quickly in areas of lower density.These first cells forge a path for others to follow, resulting in increased velocity over time.Interestingly, when stiffness is increased in alginate-RGD hydrogels, the opposite effect was observed where the velocity of T cells increased by ≈20% when cultured on stiffer gels. [71]This resulted in increased contact with APCs and subsequent activation of T cells.Similarly, Hsieh et.al. observed increases in macrophage spreading, M integrin expression, and motility on stiffer and denser fibrin scaffolds. [69]This indicates that the choice of matrix is a strong determinant in the observed phenotype and must be carefully validated to the cell and disease of interest.
To carefully control pore size in their collagen scaffolds, Stachowiak and Irvine used an inverse opal hydrogel synthesis method. [112]Briefly, PEG was gelled around monodisperse poly(methyl methacrylate) microspheres, which were then dissolved to create a porous structure.T cells suspended in collagen were then embedded into this matrix.Notably, poor adhesion to the PEG-only (no collagen) scaffold was reported, even with the conjugation of bioactive fibronectin.Larger pore sizes resulted in higher velocity and greater displacement of T cells over time.Bahlmann et.al. observed a similar trend in primary human PBMC-derived macrophages cultured in an HA-cryogel in which higher porosity and larger pore structures resulted in greater invasion distances. [51]A follow-up study showed that the invasion of macrophages was not only influenced by pore structure but also by exposure to conditioned media from Hodgkin and Reed-Sternberg (HRS) cells -the malignant cell type in Hodgkin's lymphoma.Conditioned media increased pro-inflammatory cytokine production, such as IL-6 and IL-13, and significantly increased invasion distances. [54]Interestingly, macrophages have been shown to induce an epithelial-to-mesenchymal (EMT) transition in adenocarcinoma cells: increased matrix stiffness of a Matrigel-alginate composite gel induced macrophages towards an immunosuppressive phenotype and increased the invasiveness of tumor cells. [113]In addition to controlled pore size, obtaining meaningful results based on immune and tumor cell motility also requires the incorporation of TME-native material.
The need to incorporate materials native to the TME is further emphasized by the ability of tumor cells to extensively re-model collagen fibers that then influence their migration.Histological analysis and quantitative phase imaging have identified collagen fiber alignment as a negative prognostic factor in breast carcinoma, pancreatic ductal adenocarcinoma, colon carcinoma, and prostate cancer, [114][115][116][117][118] suggesting its utility for in vitro studies.These fibers are often aligned perpendicular to the tumor boundary, and impact migration and infiltration of multiple stromal cells.However, collagen is commercially available in an isotropic form, which does not recapitulate the structure of the bundled and aligned fibers of the native TME.[121] Through this method, monocultured macrophages and T cells move faster and more persistently in an anisotropic manner parallel to the direction of aligned collagen fibers. [120]This may be due to greater access to 11 integrin binding domains along the fibers and/or the ∼35-fold increase in directionally-dependent stiffness perpendicular to the aligned fibers. [122]As collagen is often observed parallel to the tumor core, immune cells may be hindered from invading the tumor core and instead isolated to the periphery, thereby elucidating another method for immune cell evasion (Figure 5b).
Other cell types, such as fibroblasts which play a key role in cancer progression through ECM deposition and remodeling, can dynamically align collagen fibers by stretching the chains along their lamellipodia.Notably, fibroblasts were only able to align collagen at low density (1.1 mg mL −1 ), but not at high density (10 mg mL −1 ).Macrophages showed consistent displacement along fibers in matrices that were remodeled by fibroblasts. [123]Tumor cells possess a similar ability to dynamically remodel collagen fibers to facilitate macrophage migration.For example, breast cancer cells contract their cytoskeleton to generate a local perturbation to the matrix and guide macrophages toward the tumor cell in a mechanically guided manner). [124]

Cell Secretome Modulates Immune Infiltration
Immune cell trafficking and recruitment are influenced by the secretion of a diverse array of proteins, growth factors, cytokines, and extracellular vesicles by tumor cells, stromal cells, and immune cells.Hydrogel scaffolds may be used to study both the secretome of tumor-associated cells in response to substrate properties and then how these factors influence downstream immune cell behavior.It is well established that substrate stiffness, porosity, and composition influence the secretome of a plethora of cell types; [125] however, studies have focused primarily on mesenchymal stromal cell (MSC) activation induced by tumor cells. [126]MSC-and fibroblast-derived chemokines and immunomodulatory factors contribute substantially to immune cell recruitment and proliferation.For example, Tregs are recruited through the secretion of various chemokines including CXCL9, CXCL10, CXCL11, CCL21, and CCL22 that are induced by IFN-.In one study, the secretome of human lung MSCs cultured on TCP, lung-derived hydrogels, or decellularized porcine lung scaffolds showed functional differences in the downstream phagocytic capabilities of macrophages.Interestingly, phagocytosis was greatest in macrophages conditioned with media from MSCs cultured in TCP conditions while the secretome of MSCs cultured in hydrogel and scaffold conditions had a marked increase in macrophage anti-inflammatory surface marker expression. [127]owever, the influence of biomaterial properties on the secretome of tumor-associated cells, be they MSCs, immune cells, or tumor cells, requires further study.While hydrogels are often incorporated for phenotypic migration studies, conditioned media and downstream analysis of the secretome are frequently acquired from cells cultured in 2D TCP. [54,126]o study the direct effect of chemokines on immune cell infiltration, these factors are loaded in hydrogel scaffolds.For example, Stachowiak and Irvine developed a porous (80 μm) collagen scaffold loaded with 4 μg mL −1 of CCL21, resulting in higher CD4+ T cell velocity and reduced turning angles, [112] which reflects T cell attraction and chemotactic migration associated with CCL21.Similarly, CCL21 has been anchored to heparinconjugated PEG scaffolds, resulting in increased CD4+ T cell proliferation and a greater proportion of Teff cells. [128]Future studies could investigate the hypo-responsiveness of Tregs as a result of CCL21-induced migration, which contributes to impaired immune responses. [129]hile research surrounding immunotherapeutic target identification is dominated by TCP and mouse models, biomaterials are emerging as promising tools to supplement these studies.For example, injecting cytokine (m-CSF and IL-2-like) loaded alginate hydrogels into mouse models considerably enhanced DC and Treg recruitment. [130]Loading certain chemokines or their agonists, such as CCL22 which binds to CCR4 on Tregs, attracts Tregs to the tumor site and leads to their accumulation in TME.This makes the CCL22 and CCR4 axis a target to restore Treg and DC crosstalk in lymph nodes. [131,132]Using this biomaterial-based strategy may provide additional insight into the role of CCL22-CCR4 communication in cancer metastasis. [133]iomaterials remain a promising avenue to validate these mechanisms and explore new interactions surrounding immune cell recruitment.Different engineering tools have been established to study chemotactic gradients or compartmentalize cells to understand their communication and resultant migratory phenotypes (Figure 5). [134,135]In tandem with microfluidic devices, chemotactic gradients have been created and studied by perfusing two parallel liquid streams sandwiching a biomaterial containing cells. [136]Incorporating tumor-associated immune cells into the device may further elucidate the impact of the secretome on immune infiltration.

Biomaterials and Immune Checkpoint Inhibitors
The immune checkpoint blockade is crucial for immune self-tolerance and limits the duration and intensity of immune responses.Some of the most common immune checkpoints are programmed death-ligand 1 (PD-L1) on tumor cells, macrophages, and MDSCs, B7-1/B7-2 on antigen-presenting cells (APCs), as well as CTLA-4 and PD-1 on T cells.Immunosuppressive cells such as TAMs, NK cells, and MDSCs upregulate the PD-L1/PD-L2 expression in tumor cells, which inhibits T cell activity and prompts T cell exhaustion.Expression of these immune checkpoint molecules is often upregulated in the TME through either direct secretion of certain cytokines, such as interferongamma (IFN-), or the hypoxic microenvironment.As such, IFN- is often administered or hypoxia is induced in vitro to simulate immune checkpoints to study the effectiveness of immunotherapy.
Biomaterials may provide an in vitro platform to test these new immunotherapeutic-based approaches on a personalized level, as patient-derived single-cell suspension and spheroids support short culture times (< 2 weeks) for a desired output, lending themselves to high-content drug screens.Marrella and Dondero et.al. [63] explored the expression and density of a wide array of surface ligands critical for NK and T-cell mediated immune responses against tumors after exposure to IFN-, comparing 2D culture to 3D spheroids in alginate gels.Most notably, IFN- exposure in 3D culture reduced expression of immune receptors involved in NK-mediated tumor recognition, a phenotype observed in metastatic, neuroblastoma patients.Furthermore, NK-protective ligands had higher expression in MYCNamplified neuroblastoma cells, which represent a more aggressive and treatment-resistant class of the disease.Thus, compared to 2D, 3D culture in alginate may provide a more representative picture of what occurs in more aggressive stages of neuroblastoma.In the case of in vitro immunotherapy studies, 3D biomaterials may be necessary as PD-L1 expression has been shown to be influenced directly by substrate stiffness, with soft (≈2 kPa) gels showing much lower expression by western blotting of three different lung cancer cell lines compared to stiff (≈25 kPa) gels and 2D TCP. [137]Thus, 2D studies may provide an inaccurate picture of what occurs in vivo, especially in the case of cancers of soft tissues such as the lung and brain.

Future of Biomaterials as Tumor Models in Drug Discovery
It is becoming increasingly well validated (functionally and transcriptomically) that 3D co-culture of cancer cells with tumor-associated immune cells upregulates aggressive tumor phenotypes and increases their chemoresistance. [138]Both in vitro and in vivo studies of immune and tumor cell interactions relating to evasion (immune phenotype, infiltration, and invasion) have identified a plethora of interactions with the surrounding microenvironment which cannot be captured in 2D.However, both the biophysical and mechanical properties of the biomaterial drastically influence immune and tumor cell phenotype and downstream drug interactions, indicating that 3D scaffolds must be carefully designed with their native physiological milieu taken into strict consideration.While biomaterials show immense promise in their ability to provide more physiologically relevant tumor immune response profiles, the models must be validated, and standards established for scale-up.
It remains unclear how much complexity is required to improve the efficacy of drug screens to identify druggable targets and then drugs, which translate into clinical use.As more studies are published, the benefit of 3D biomaterials culture, over 2D, to provide more representative models of what is observed in vivo is becoming clearer.It is important to balance the complexity of the system and the time required to reach a decision with patient cells while maximizing clinical relevance.For example, while organoids provide impressive complexity, the months required to establish a culture are often too long to impact patient care as the tumor may have shifted into a new phenotype.This is especially true for highly plastic and aggressive cancers. [139]herefore, organoids may be useful for mechanistic studies; however, their low throughput reduces their appeal in drug screening.In contrast, 2D TCP is fast, but may not provide accurate readouts.Biomaterials will likely provide an additional metric to parse out the effects of the biochemical and mechanical properties of the ECM.The ultimate goal of these preclinical screens is to provide greater predictive power to downstream clinical trials.
However, not all 3D hydrogels are the same as we observed changes to cell growth rates, phenotype, and drug responses based on the scaffold used. [140]We must consider the relevance of the biochemical and mechanical properties of the 3D hydrogel in the context of the TME that is being emulated and the ultimate impact they have on immune cell phenotypes.As tissue engineers continue to tackle problems in cancer immunology, standardization in immune cell phenotype reporting must occur.A complete consensus of ECM stiffness on macrophage phenotype cannot be established due to the different characterization methods employed.Broadly characterizing a macrophage as M2/anti-inflammatory due to CD206+ expression without additional follow-up experiments is an oversimplification and removes any phenotypic nuance appreciated in immunology.The inclusion of immunology experts will increase the potential of the material as a relevant scaffold for immune cell activity in downstream studies and drug screens.
From a material perspective, mechanical stiffness and viscoelasticity are major contributors to cell polarization or activation state across all immune cell types.The mechanical cues imparted to the cell by the material not only influence phenotype but also the magnitude of immune checkpoint molecule expression on tumor cells, indicating that stiffness must be carefully modulated to match that in the tumor microenvironment.Good reproducibility between batches must be achieved as this is a consistent problem with xenogenic material sources.Synthetic materials often lack the biochemical cues that exist in the native TME and have shown contrasting invasion behaviors to their native counterparts.Native materials, such as hyaluronan and collagen, typically show better biomimetic outcomes; however, synthetic materials can be modified with relevant bioactive peptides to improve cell viability and spreading, as well as induce phenotypes similar to what is observed in vivo.With the massive output of new hydrogel formulations put forward to study tumors, as well as the rising trend of artificial intelligence, there is an opportunity to develop a comprehensive library to guide future research and development towards a standard suitable for the clinic.
Bioprinted materials have gained popularity as they allow consistent extrusion of "bioinks"; [141] yet, bioprinting too has some limitations -the biomaterial must be extrudable; however, extrusion may push immune cells towards a pro-inflammatory phenotype.Furthermore, the material must be designed with additional requirements in mind such as fast gelation or chemistry to allow photo-inducible crosslinking, as well as relatively low viscosities.Any material designed for drug discovery, screening, or testing (whether bioprinted or not) must be reproducible and validated against current gold-standard methods.
From a cell perspective, the number and types incorporated are open to debate.In some cases, conditioned media from the tumor cell results in the same outcome as direct contact co-culture, [54] however, in other situations (based on the disease and behavior of interest), direct contact with multiple tumor, immune, and stromal cells is required for success.In either situation, it remains imperative that the system is thoroughly characterized for immune phenotype through functional assays and the whole secretome to ensure the validity of the results.

Figure 1 .
Figure 1.Candidate models for drug screening with tumor-immune cell co-cultures, comparing the time, cost, and complexity of each model.Does increased complexity correlate with enhanced clinical translatability?

Figure 2 .
Figure 2. Immune cells infiltrating the tumor microenvironment undergo phenotypic changes to combat the hostile environment.

Figure 3 .
Figure 3. Material considerations when designing a hydrogel for immune cell culture.

Figure 4 .
Figure 4. Tumor initiation leads to dramatic alterations to the extracellular matrix.

Figure 5 .
Figure 5.The process of immune cell infiltration and extravasation is extensively modeled in biomaterial platforms.a) Infiltration through the blood vessels induced by chemotactic gradients in a dual-flow microfluidic chip.b) Matrices to study immune infiltration through physical remodeling of the scaffold have found significant impacts elicited by matrix density, porosity, and collagen fiber alignment.c) Microfluidic channel with porous walls bordering an ECM-mimetic hydrogel allows axonal sprouting, modeling immune-assisted metastasis of circulating tumor cells.

Table 1 .
Summary of studies analyzing the effect of hydrogel stiffness on macrophage phenotype.

Table 2 .
Diverse tumor-associated cell types result in heterogeneous metabolic programs that drive tumor growth.