Customizable biomaterials as tools for advanced anti-angiogenic drug discovery

Abstract

The inhibition of angiogenesis is a critical element of cancer therapy, as cancer vasculature contributes to tumor expansion. While numerous drugs have proven to be effective at disrupting cancer vasculature, patient survival has not significantly improved as a result of anti-angiogenic drug treatment. Emerging evidence suggests that this is due to a combination of unintended side effects resulting from the application of anti-angiogenic compounds, including angiogenic rebound after treatment and the activation of metastasis in the tumor. There is currently a need to better understand the far-reaching effects of anti-angiogenic drug treatments in the context of cancer. Numerous innovations and discoveries in biomaterials design and tissue engineering techniques are providing investigators with tools to develop physiologically relevant vascular models and gain insights into the holistic impact of drug treatments on tumors. This review examines recent advances in the design of pro-angiogenic biomaterials, specifically in controlling integrin-mediated cell adhesion, growth factor signaling, mechanical properties and oxygen tension, as well as the implementation of pro-angiogenic materials into sophisticated co-culture models of cancer vasculature.

INTRODUCTION

The inhibition of angiogenesis, the growth of new blood vessels from existing vascular networks [1] has been a critical target for cancer therapeutics since Folkman et al. demonstrated that tumor growth was dependent on the presence of vasculature [2, 3]. These studies motivated the discovery of numerous drugs that can disrupt angiogenesis and may consequently deny tumors of growth factors, oxygen, and nutrients required for growth. The first drugs that passed clinical trials disrupted Vascular Endothelial Growth Factor (VEGF) signaling pathways that contribute to angiogenesis, either as antibodies to VEGF such as Bevacizumab [4] or as inhibitors to VEGF Receptor Tyrosine Kinases such as Sunitinib [5]. Although several drug candidates, such as Bevacizumab, Sunitinib, and Sorafenib [6] have received FDA approval for the treatment of specific types of cancers, others failed to improve patient survival [7]. A summary of existing anti-angiogenic therapies, along with limitations, are found on Table 1 (Reviewed in detail elsewhere [8–14]). Ultimately the disruption of cancer vasculature may be important for limiting tumor growth in patients, but it does not fully eliminate tumors in many cases [15, 16]. Important limitations to therapeutics can render drugs ineffective, including the inability of drugs to access the entirety of cancer vasculature due to disturbed flow, and resistance of cancer endothelial cells to drug treatment [10, 11, 13, 17]. Additionally, due to the complexity of the tumor microenvironment, the use of vascular disruption can generate unintended consequences that reverse the progress of treatment or worsen the patient prognoses [18–20].

Table 1.

Summary of existing anti-angiogenic therapeutic strategies and their limitations.

Therapeutic Strategy	Target	Known Therapeutics	Limitations	References
Growth Factor SignalingAMG386	VEGF Family	Bevacizumab, Sorefenib, Sunitinib, Herceptin	• Imbalanced growth factor signaling in tumors • Requirement for targeting multiple growth factors • Narrow therapeutic window between effectiveness and cytotoxicity in off-target sites • Limited access to inhibitor targets due to compromised blood flow, organization of tumor vasculature	4–9 11,17
	TGFβ Family	Lerdelimumab, Metelimumab
	FGF Family	FP-1039, E-3810, TKI258
	Angiopoietin	AMG 386
	EGF	Cetuximab, Panitumumab, Erlotinib
	IGF	NVP-AEW541, NVP-ADW742
Calcium Signaling	NFAT Activation	FK506, Anti-SFRP2 antibody	• Limited access to inhibitor targets due to compromised blood flow	8, 11, 17
Tumor Endothelial Markers	TEM8,SFRP2	Anti-SFRP2 antibody	• Limited access to inhibitor targets due to compromised blood flow • Impaired genome/epigenetic changes to endothelial cell behavior • Abnormal expression of target molecules inhibits drug penetration to interior of tumor – binding site barrier	8, 10, 14, 30
Endogenous angiogenesis inhibitors	Angiostatin	Direct application of the inhibitors, or delivery via original ECM molecules including Collagen IV, Collagen XVIII (Endostatin)	• Narrow therapeutic window between effectiveness and cytotoxicity in off-target sites • Limited access to inhibitor targets due to compromised blood flow • Impaired genome/epigenetic changes to endothelial cell behavior	8,9, 13, 14
	Endostatin
	Thrombospondin-1
	Tumstatin
	Canstatin
	Arrestin
Extracellular Matrix	Matrix Metalloproteinase	MMP9 to release Angiostatin, Endotstatin, Tumstatin	• Narrow therapeutic window between effectiveness and cytotoxicity in off-target sites • Limited access to inhibitor targets due to compromised blood flow, organization of tumor vasculature	9, 14
	integrins αvβ3, αvβ5	Cilengitide, Vitaxin, S247
	Focal adhesion kinase	TAE-226, Pfizer-PF-573,228
Tumor Activated Fibroblasts	Fibroblast Activation Protein	Sibrotuzumab, val-prolineboronic acid	• Small molecule inhibitors ineffective in certain tumor types	8

An example of these consequences is angiogenic rebound, where cancer vasculature that was previously disassembled by anti-angiogenic treatment rapidly regrows after the cessation of treatment [18, 21–23]. This is the result of multiple contributing mechanisms, including the generation of a local hypoxic environment, increased expression of VEGF, PlGF, FGF2, SDF-1 and PDGF within the tumor [18, 21–23], and putative transformation of tumor cells into endothelial cells [21–25], all of which can result from initial anti-angiogenic treatment. Another example of an unintended effect is the activation of a metastatic switch in tumors, where cells within the tumor change phenotypes to encourage metastasis [20, 26, 27]. While anti-angiogenic treatment focuses on the deactivation of pro-angiogenic signaling pathways, emerging evidence suggests that those same pro-angiogenic pathways can inhibit the onset of metastasis through inhibition of tumor cell migration and invasion, growth factor production by cancer-associated fibroblasts, and the endothelial-mesenchymal transformation [20, 26, 27].

A final example of an unintended consequence of anti-angiogenesis treatment is vascular normalization, where prior to complete network disassembly the cancer vasculature restructures itself from a tortuous, leaking, dysfunctional vasculature [11, 12, 17, 28–30] into a well-ordered, patent vasculature [21–23, 29, 31] (Fig. 1). Unlike other side effects that hinder cancer treatment, normalization can potentially be exploited to improve the transport of anti-cancer drugs into the interior of a tumor, and thereby increase the effectiveness of cancer therapy [32, 33]. Currently, the clinical effectiveness of combined anti-angiogenesis and anti-cancer treatments is unclear [34] as the timing window of vascular normalization is limited and unknown, and there are not yet effective biomarkers to detect a normalization event in situ [34]. However, recent studies have demonstrated that treating ovarian and colon carcinomas with a combination of bevacizumab and paclitaxel in mouse models enabled a uniform intratumoral distribution of paclitaxel [35], and magnetic resonance images of tumor blood vessels suggest that normalization by bevacizumab may peak at 24 hours after treatment in human metastatic brain tumors [36].

Figure 1.

Normal vasculature and cancer vasculature. Whereas normal vasculature exhibits predictable branching patterns and well-defined arteries, arterioles, capillaries, venules and veins [17], cancer vasculature exhibits chaotic formation of a wide variety of blood vessels that are leaky, tortuous and poorly perfused [11–13, 17, 28–30]. Examples of cancer-specific blood vessels include: Mother Vessels – large, tortuous, leaky vessels; Vascular Malformations – Poorly perfused, abnormally large vessels coated with smooth muscle cells; Glomeruloid Microvascular Prolierations – disorganized, hyperproliferative and hyperperfused vessels; Transluminal Bridges – capillary vessels that penetrate and travel through larger blood vessels; Feeder arteries and Draining veins – tortuous, abnormally large vessels larger than vascular malformations [17].

Importantly, the occurrence of unintended side effects would be difficult to predict via existing angiogenesis assays used for drug discovery in vitro. Most functional assays of anti-angiogenic drug activity only demonstrate specific endpoints of vascular network disruption, such as decreased cell viability, proliferation, migration and/or vascular network formation [37–39]. While these in-vitro assays can be well-suited for discovering compounds that modulate angiogenesis [40], far-reaching effects beyond initial inhibition were not observed in vitro, and largely became known only after they were first observed in animal models [18, 41, 42] and human clinical trials [18, 20, 29, 43].

Advances in the design of biomaterials and cell culture platforms are necessary to enable rapid, detailed characterization of the complex effects of anti-angiogenesis treatments beyond the scope of initial vascular disruption. For example, models of angiogenic rebound will likely require culture environments that support long-term maintenance of vascular networks as well as re-assembly of disrupted vascular networks after the cessation of drug exposure. Models of vascular normalization will require environments that initially promote the formation of tortuous, pathological vasculature so that signs of normalization may be clearly detected. Cell culture systems should additionally accommodate heterotypic co-cultures including tumor-associated cell types to more closely mimic tumor environments. This review will examine recently developed biomaterials and cell culture tools that enable environmental control in vascular morphogenesis models, and may more effectively model anti-angiogenesis treatments.

BIOMATERIALS AS TOOLS TO MODEL VASCULAR MORPHOGENESIS

Angiogenesis consists of a complex series of cell actions, including soluble growth factor signaling, proliferation, migration, and assembly of multi-cellular tubular structures [1], all of which are modulated by the extracellular matrix (ECM) (Fig. 2). In physiological scenarios, microvasculature is in contact with a continuous ECM in the form of a basement membrane composed of Fibronectin, Collagen I, Collagen IV, multiple forms of Laminin, Perlecan and other proteoglycans [44, 45]. In dysfunctional vasculature the ECM contains μm-scale gaps in the basement membrane [44, 45] and elevated levels of ECM proteins, notably Collagen I, leading to increased mechanical stiffness [46]. Additionally, the typical tumor vasculature exists in hypoxic environments that impact behavior of not only endothelial cells but also tumor cells that can modulate angiogenesis via paracrine signaling [18]. Biomaterials are being developed by several groups to recapitulate important cellular and molecular components of pro-angiogenic environments. The following subsections will review biomaterials that present instructive environmental cues that are critical to modulating angiogenesis, including presentation of integrin-binding cell adhesion ligands, mechanical signaling, growth factor signaling, hypoxia and proximity to cancer cell types.

Figure 2.

Individual and synergistic effects of ECM properties on endothelial cell behaviors related to angiogenesis. Integrin-mediated cell adhesion influences endothelial cell viability, proliferation [47–49], migration [47, 49, 50], spatial polarization [51] and lumen formation [51, 52]. ECM modulus influences lumen formation [53], capillary sprouting rate and directionality [53–55], capillary aspect ratio [53, 55], proliferation, and migration [49]. Growth factor sequestration influences growth factor stability [56–58] and facilitates binding between growth factors and receptors [56, 57, 59]. Synergistically, integrin binding and growth factor sequestration influence VEGFR2 activity and focal adhesion assembly [60, 61]. Matrix modulus and growth factor sequestration influence VEGFR2 activity, cytoskeletal stress fiber formation and cell migration rate [62–64]. Finally, integrin binding and matrix modulus influence cell traction forces exerted and detected by endothelial cells [65].

Material Selection

Biologically-derived extracellular matrix components are commonly adapted as biomaterials to induce angiogenesis and endothelial network formation in endothelial cell culture. Collagen, Fibrin and Matrigel can present a wealth of signals that drive pro-angiogenic behaviors of endothelial cells. The often-cited limitations of naturally-derived materials are poor mechanical integrity, high batch-to-batch variation and complex compositions that mask specific mechanisms driving angiogenesis [66, 67]. Although there is difficulty in attributing changes in cell behavior to specific changes in a natural material due to its inherent complexity, studies using natural materials may be coupled to parallel studies on synthetic materials. Comparisons between natural and synthetic materials can reveal new mechanisms that affect angiogenesis, such as the existence of cryptic signaling moieties and growth factor binding sites that are critical to facilitating angiogenesis.

Synthetic, chemically defined biomaterials have been adapted as alternatives to naturally-derived ECM components to achieve greater control of cell behavior in angiogenesis models. Synthetic cell culture materials often consist of background materials that present chemically-defined biological cues to control cell behavior with consistency and predictability, while minimizing the presence of unwanted cues [68–70]. A commonly used material is poly(ethylene glycol) (PEG), a hydrophilic polymer that resists non-specific protein adsorption and presents specific biological signals to cells engineered into the material by chemical modification [71]. Similar materials include poly(vinyl) alcohol (PVA) [72–74], polyacrylamide (pAm) [74], alginate [75], and dextran [76, 77]. A number of studies have modified these materials with biomolecules that control cell behavior through mechanisms including adhesion, growth factor binding and matrix degradation, and these cues may synergistically induce desired angiogenic phenotypes from endothelial cells.

Adhesion Molecules

Among the most critical biomolecules included into a biomaterial are ligands to promote cell adhesion. A commonly used mediator of cell-material adhesion is the integrin-binding Arg-Gly-Asp amino acid sequence (RGD) derived from numerous cell-adhesive proteins present in the ECM [78]. Even though its cell adhesive capabilities are well known and characterized in cell culture systems [79], RGD presentation has been continuously improved in the form of optimized concentrations and spatial patterns to understand and control angiogenesis.

Recent studies have cultured endothelial cells in environments presenting ranges of RGD concentrations to find optimal pro-angiogenic conditions. On the surfaces of PEG hydrogels, optimal RGD concentrations are necessary to encourage endothelial network formation by human umbilical vein endothelial cells (HUVECs). Insufficient RGD density results in cell detachment and excessive RGD results in confluent cell sheet formation [80]. The role of RGD density in modulating pro-angiogenic activity may be defined by the average distance between RGD molecules on a cell culture material. Previous data suggest that this spacing modulates VEGF signaling and cell migration by influencing focal adhesion assembly and focal adhesion kinase (FAK) signaling, and a distance of approximately 44 nm between RGD molecules maximized FAK signaling in bovine aortic endothelial cells [81]. Three dimensional (3D) cultures of HUVECs and aortic ring explants have demonstrated that insufficient concentrations of RGD decrease cell viability, while excessive RGD concentrations result in either arrest of cell migration or excessive matrix degradation through integrin-mediated expression of matrix metalloproteinases (MMPs) [49, 82]. However, a separate study demonstrated that high concentrations of RGD permitted vacuole formation and subsequent lumen formation by endothelial colony forming cells (ECFCs), suggesting the assembly of maturing, patent blood vessels [83]. Taken together, these results reveal that RGD concentrations may encourage angiogenesis or blood vessel maturation, and that appropriate RGD concentrations should be optimized to encourage desired angiogenic outcomes.

Spatial patterning of RGD via photolithography also modulates endothelial tubule formation. One application of spatial patterning is to define where the formation of microvasculature is permitted in a material, but patterning can also impact morphologies and stability of capillary vessels. In one study, the aspect ratios of RGD patterns presented on PEG hydrogels determined whether endothelial cells formed cell sheets or tubule-like structures [84], and 100–200 μm-wide RGD patterns on PEG hydrogels forced HUVECs to form elongated cell aggregates that resembled tubules [85].

In addition to RGD, other studies have explored the ability of the laminin-mimicking peptides Tyr-Ile-Gly-Ser-Arg (YIGSR) and Ile-Lys-Val-Ala-Val (IKVAV) to modulate pro-angiogenic endothelial cell behaviors. YIGSR or IKVAV presented along with RGD in PEG hydrogels have induced greater tubule formation by HUVECs than with RGD alone [86]. An innovative study from Collier and coworkers explored pro-angiogenic effects in a self-assembly matrix comprised of the acetylated Gln-Gln-Lys-Phe-Gln-Phe-Gln-Phe-Glu-Gln-Gln (QQKFQFQFEQQ) peptide known as “Q11”, which forms β-sheet fibers. These peptides were terminated with cell-adhesive amino acid sequences including Arg-Gly-Asp-Ser (RGDS), Arg-Gly-Asp-Val (REDV), IKVAV and YIGSR [87–89]. The modified Q11 peptides modulated cell adhesion using multiple ligands that were orthogonally added to hydrogel formulations (Fig. 3), and endothelial cell proliferation, viability and adhesion were maximized through optimizing concentrations of multiple species of adhesion ligands [89]. This study demonstrated how systematic control over material properties, coupled with design of experiments methodology can identify important regulators of EC behavior.

Figure 3.

Multifactorial optimization and effects analysis of endothelial cell growth using Q11 modular peptides and cell adhesion ligands [89]. Copyright 2011, with permission from the Royal Society of Chemistry.

While peptides are effective for modulating cell-material adhesion and mimicking some of the functions of ECM proteins, studies have also begun to incorporate full-length ECM proteins into synthetic materials. In one study, human aortic endothelial cells (HAECs) were cultured in the presence of either full-length Collagen or a Collagen-mimicking peptide SCL2–2 presented on PEG hydrogels. SCL2–2 is a Collagen mimicking protein presenting mainly Gly-Phe-Pro-Gly-Glu-Arg (GFPGER) as the integrin-binding domain [90]. HAECs adhered to full-length collagen had greater levels of NOS3 protein expression, greater gene expression levels of NOS3, Thrombomodulin and Selectin-E, and proliferation when compared to HAECs adhered to SCL2–2. While full-length ECM proteins present some advantages, they can also present extraneous background signals into otherwise defined cell culture environments and may require greater control of regiospecific chemistries for incorporation into biomaterials without interfering with essential biological epitopes.

Growth Factor Signaling

Growth factor signaling has been integrated into biomaterials as either short growth factor-mimicking peptides or full-length growth factors. Vascular endothelial growth factor (VEGF) signaling provides an illustrative example. A short peptide termed “QK”, that mimicks the receptor-binding regions of VEGF [91], has been used to enhance angiogenesis. QK was coupled along with RGD to PEG hydrogels and increased the length of endothelial tubule structures relative to those reached on hydrogels functionalized with RGD alone [92]. QK may also be attached to other peptide sequences that facilitate non-covalent incorporation into biomaterials. This enables patterning of QK not just by photopatterning, but also by the use of fluid flow and partial dipping of biomaterials into solutions containing QK, as demonstrated when QK was coupled to hydroxyapatite scaffolds via an attached hydroxyapatite binding sequence [93]. When integrated into Collagen hydrogels via a helical Collagen mimicking domain (CMP) the addition of a CMP-QK peptide induced the formation of elongated endothelial cell structures in two-dimensional (2D) cell culture and enhanced capillary sprouting in three dimensional (3D) culture [94, 95]. Triple-helical hybridization has also coupled QK to PEG hydrogels and produced similar pro-angiogenic effects [85].

Full-length VEGF protein has been coupled to cell culture materials to enhance angiogenic cell behavior. One study covalently immobilized VEGF to PEG molecules through a succinimidyl carbonate group and spatially patterned PEG-VEGF molecules into PEG hydrogels via a scanning confocal laser. The resulting VEGF patterns directed endothelial cell network formation in specific areas of the hydrogels [84]. Full-length VEGF has also increased endothelial cell proliferation when attached to electrospun micro-and nano-fibers of poly(ε-caprolactone) via carbodiimide-mediated carboxylic acid coupling reactions [96]. VEGF was also patterned into Collagen sponges in spatially defined areas to permit formation of microvascular networks. This was achieved by extruding Collagen mixed with an aqueous VEGF solution onto a copper plate, freezing, then inserting the frozen Collagen-VEGF stripes into a larger Collagen sponge [97]. These are only a few examples selected from a larger published literature combining biomaterials with VEGF [98–101], reviewed elsewhere.

Other growth factors besides VEGF have also been shown to modulate angiogenic cell behaviors when attached to biomaterials. PDGF-BB and bFGF were incorporated into PEG hydrogels via succinimidyl ester coupling to observe effects on HUVEC and pericyte co-cultures. bFGF and PDGF-BB worked in concert to increase tubule network length, albeit by different mechanisms. While PDGF-BB increased pericyte proliferation, bFGF increased HUVEC migration speeds in 2D and 3D cultures. The use of PDGF-BB in endothelial culture systems notably provides the ability to tailor environments to enhance angiogenic activity of perivascular cell types in addition to endothelial cells, and this highlights the need to integrate multiple growth factor signals into pro-angiogenic materials. When PDGF-BB alone was included in hydrogels, endothelial tubule formation occurred whether PDGF-BB was soluble or attached to the hydrogel. When HUVECs were cultured in the presence of releasable PDGF-BB and insoluble Ephrin A1, a matrix-bound receptor tyrosine kinase ligand, the PDGF-BB did not affect endothelial tubule widths in a 2D or 3D environment [102]. Similar hydrogels were implanted in mouse cornea models, but these were modified to contain immobilized PDGF-BB. When exposed to the variety of other cells present in vivo, the presence of covalently immobilized PDGF-BB enhanced the extent of tubule formation (Fig. 4) [103].

Figure 4.

Hydrogels containing covalently immobilized PDGF-BB enhances vascular network formation in the mouse cornea model compared to hydrogels encapsulating soluble PDGF-BB only [103]. Copyright 2011, with permission from Elsevier.

While the direct attachment of growth factors to scaffolds has enhanced angiogenic response by endothelial cells, one potential limitation of covalently attaching growth factor receptor-binding molecules to biomaterials is that they may prevent receptor internalization or dimerization [104, 105]. In one extreme example, VEGF-mediated endothelial network formation in Matrigel was inhibited by an immobilized QK peptide that could not be released from the surrounding environment, and the putative mechanism involved “receptor pinning” by immobilized QK [106]. Numerous other strategies have been used to modulate soluble growth factor signaling dynamics via strategies that mimic in-vivo mechanisms. The in-vivo ECM is capable of passive and cell-mediated release of soluble growth factors including VEGF [56], bFGF [57], and other pro-angiogenic growth factors. The ECM is also capable of enhancing growth factor stabilization and concentration in the matrix via growth factor-binding glycosaminoglycans and proteoglycans (e.g. heparin) [56–59]. Strategies to mimic relevant ECM-growth factor interactions have been extensively reviewed elsewhere, and include temporal control over growth factor release [100], spatial control over growth factor gradients [107], and inclusion of growth-factor binding and sequestering molecules to the matrix [101, 108].

Mechanical Properties

The stiffness of the cellular microenvironment is a critical mediator of cell phenotype, and is a distinguishing feature when comparing normal and diseased (e.g. cancerous) tissues [46, 109]. Optimized stiffness ranges can enable endothelial cell network formation in 2D and 3D environments. For example compliant (elastic modulus 140 Pa) polyacrylamide hydrogels functionalized with 0.1 mM RGD promoted formation of endothelial cell networks while stiffer hydrogels (elastic modulus 2500 Pa) promoted formation of confluent endothelial cell sheets [110]. On collagen-coated polyacrylamide hydrogels, stiffness dictated the expression of pro-angiogenic genes as well as pro-osteogenic genes in HUVECs. Specifically, VEGFR2 gene expression was upregulated on 3 kPa elastic modulus hydrogels, while angiogenic and osteogenic genes were upregulated on 30 kPa elastic modulus hydrogels [111]. In 3D environments a balance between matrix degradability and stability is required to foster HUVEC network formation. One study putatively demonstrated a need for degradable matrices that permit remodeling and cell migration, but retain enough stability to prevent the collapse of a forming vascular network [49]. Interestingly, HUVECs in 3D environments have variable responses to drug treatment depending on the surrounding stiffness and the presence of tumor-derived growth factors. Specifically, HUVECs in one study were more sensitive to the angiogenesis inhibitor Vandetenib when seeded on softer materials than stiffer materials, and treatment with tumor-derived growth factors removed stiffness effects on HUVEC network formation and decreased drug sensitivity [112]. Finally, the density of a hydrogel network also affects endothelial cell responses to VEGF gradients. Specifically, enhanced collagen density increased human dermal microvascular endothelial sprout polarization toward increasing concentrations of VEGF and increased sprout stability (Fig. 5) [53].

Figure 5.

Human dermal microvascular endothelial cell sprouting behavior in VEGF gradients. Sprouting varies depending on collagen density and stiffness [53]. Copyright 2010, with permission from the Royal Society of Chemistry.

Hydrogel stiffness can be tuned by changing polymer concentration [113], crosslinking density [49, 109, 114, 115], and temporal control of stiffness may be controlled using engineered degradation mechanisms [116–118] in most materials. Other strategies have recently changed mechanical properties while also decoupling polymer network density and mechanical stiffness. Collagen modification with tyramine groups and treatment with peroxidases have enabled covalent crosslinking of Collagen hydrogels while still enabling cell-ECM interactions [119], and the crosslinking density modulated network formation by ECFCs. Additionally, the impact of changing stiffness via crosslinking density independently of Collagen density was explored by crosslinking Collagen after non-enzymatic glycation [55, 109]. Increased matrix stiffness increased the extent of capillary sprouting from bovine arterial endothelial cell and HUVEC spheroids, increased sprouting in mouse and chick embryo models, and increased blood vessel permeability. These techniques allowed investigators to explore the effects of changing mechanics without simultaneously affecting the density of structural molecules and insoluble signaling molecules presented to cells.

An additional strategy for controlling material stiffness is the use of biocomposites, in which reinforcing materials such as ceramics and glasses may be added to otherwise soft materials. One study incorporated hydroxyapatite nanocrystals into Fibrin matrices to increase angiogenic sprout number, length and invasion speed by HUVECs [120]. Another study enhanced angiogenic sprouting in aortic ring assays by adding magnesium-doped bioglass to poly(butylene succinate) hydrogels [121]. In both cases, the use of reinforcing materials in the biocomposites was implicated not only in increased stiffness of a pro-angiogenic material, but enhanced growth factor binding and sequestration as well. As such, the mechanical and biochemical properties of biocomposites have the potential to recapitulate angiogenic environments found in bone-related diseases such as bone cancer and ectopic tissue mineralization [122, 123].

Previous studies have demonstrated that degradability is critical to enabling endothelial network formation in 2D [124] and 3D [113] endothelial cell cultures. Many PEG-based hydrogels are crosslinked into a polymer network through the use of cell-degradable crosslinking molecules. Hydrogels are often fully or partially crosslinked using pegylated mimics of a MMP-labile site on Collagen, and the sequence may be modified to enhance or inhibit degradability. For example, the native MMP-labile amino acid sequence of Gly-Ile-Ala-Gly (GIAG) can be exchanged for Gly-Ile-Trp-Gly (GIWG) for enhanced degradability or Gly-Pro-Ala-Gly (GPAG) for inhibited degradability. These changes have been shown to affect the rate of endothelial cell sprouting in an aortic arch assay [125]. Another strategy to modulate degradability is the use of crosslinking peptides with a single MMP-labile site versus peptides with multiple MMP-labile sites, and this strategy has been used to dictate HUVEC migration [126].

To further define the shape of endothelial cell networks and directionality of angiogenic sprouting in biomaterials, modulus and degradability can be spatially defined using innovative techniques. For example, perfusion-based frontal polymerization – the uneven photopolymerization of hydrogels as a result of a concentration gradient of Eosin Y photoinitiator – generates hydrogels with graded density that can dictate the direction of growth by sprouting endothelial cells [127]. Spatially patterned degradability has also been achieved via multiple other methods. Degradability in hyaluronic acid hydrogels was defined by adding MMP-labile crosslinking groups. In addition, a secondary means of regulating degradability was achieved via UV photopolymerization of acrylate-functionalized hyaluronic acid to make certain areas of the hydrogel impassible, thereby dictating locations of ECFC vessel formation [128].

Maintenance of Hypoxia

Numerous vascular disorders lead to decreased levels of dissolved oxygen present in surrounding tissue. This is known to significantly affect endothelial cell phenotype [129], including via Hypoxia Inhibitory Factor (HIF) signaling. Though tests requiring the establishment of a hypoxic environment can be achieved using specialized incubators and bioreactors with controlled atmospheric conditions, recent work has alternatively used biomaterials to expose cells to hypoxic conditions for prolonged periods of time and generate defined oxygen gradients in endothelial cell culture.

One study achieved control of dissolved oxygen concentrations in gelatin-based materials by including a ferulic acid crosslinking molecule and initiating a crosslinking reaction with Laccase (Fig. 6). The crosslinking reaction consumes oxygen and maintains low oxygen conditions for up to 1 hour after polymerization. This caused increased expression of HIF1α and HIF2α in ECFCs, both of which promote vascular network formation [130]. This Laccase-based crosslinking mechanism was adapted in Dextran-based hydrogels in which oxygen consumption was achieved during crosslinking of tyramine-functionalized dextran. Here, hypoxic conditions in the hydrogels were maintained for up to 12 hours in the hydrogels, and the duration of hypoxia was tunable through polymer concentration and stiffness [131]. Interestingly, materials that regulate dissolved oxygen levels in endothelial cell culture represent an adaptable approach to regulating the cellular microenvironment without the need for bioreactors or similar instrumentation.

Figure 6.

Hydrogel crosslinking reactions can consume oxygen and reduce dissolved oxygen concentrations present in endothelial cell culture environments [130]. Copyright 2014. Used with permission from Nature.

Vascular morphogenesis in co-culture systems

Endothelial cells in tumors interact with cancer cells and cancer stem cells, which respond to environmental cues and subsequently affect the vascular network through physical contact and paracrine signaling [132, 133]. Studies utilizing biomaterials to define populations of endothelial-cancer cell co-cultures and affect phenotypes of heterotypic cell populations are beginning to be applied toward modeling cancer vasculature. For example, discrete spatial patterning of heterotypic cell populations has been achieved through micro-contact printing and cell culture in microfluidic channels. Segregation of breast cancer cells in hyaluronic acid hydrogels from ECFCs in fibrin hydrogels was performed using sequential micro-contact printing (Fig. 7) [134]. Discrete patterning of HUVECs and HeLa cancer cell populations was recently achieved in 2D cell culture by combining micro-contact printing, activation of un-patterned PVA with dopamine, and sequential cell seeding [135]. The separation of cell populations is a promising tool for studying juxtacrine and paracrine control of angiogenesis by cancer cells, as demonstrated in microfluidic platforms. In one example, salivary gland adenoid cystic carcinoma cells and oral squamous cell carcinoma cells induced – to varying degrees based on cell type – sprouting capillary growth by HUVECs through a basement membrane extract matrix [136]. Taken together, these technologies create environments that incorporate relevant ECM cues, as well as endothelial and tumor cells to accurately model the in-vivo tumor environment.

Figure 7.

Spatially segregated co-culture of endothelial cells and cancer cells is achieved by sequential micro-contact printing steps. Materials may also be patterned to induce multicellular structure formation with desired morphologies and geometries [134]. Copyright 2012. Used with permission from the Royal Society of Chemistry.

To recapitulate increasingly complex cancer environments for drug discovery, studies have leveraged materials-based techniques to generate vascularized microtumors [30, 137]. Many vascularized microfluidic tumor models consisted of cancer cells encapsulated in fibrin that were grown directly in co-culture with endothelial cells [138, 139], or spatially partitioned from a neighboring population of endothelial cells in microfluidic channels [140]. These systems demonstrated perfusable vascular networks growing within and around the tumor spheroids [139], as well as cancer cells migrating to and infiltrating vascular networks [140]. In many cases, cancer cells were partitioned inside alginate-based microcapsules to grow in their own isolated environments [141–144]. Alginate capsules are resistant to cell attachment but are also permeable to growth factors such as MCF-7 breast cancer cell-secreted VEGF and HIF1α [141]. The vascularized microtumor model was further developed through the incorporation of alginate capsules containing MCF-7 cells into vascularized Collagen-I hydrogels. Briefly, MCF-7 cells were encapsulated in Collagen-I cores surrounded by alginate shells. The fully vascularized tumor model was generated by encapsulating multiple MCF-7 capsules along with HUVECs and human adipose-derived stem cells into a larger Collagen-I hydrogel [145]. Here, further steps were taken to control biomaterial properties to modulate angiogenesis. The stiffness of the collagen core of the MCF-7 capsules modulated proliferation and expression of vimentin and CXCR7 by the MCF-7 cells. Additionally, direct contact between endothelial cells and MCF-7 cells was controlled through the addition of sodium citrate, which could dissolve alginate shells while leaving the overall tumor model intact. This enabled observation of paracrine signaling or direct cell-cell contact within the model. This modular approach shows promise in allowing for heterogenous biomaterial constructs that can modulate cancer vasculature formation in drug screening models.

PERSPECTIVE: MATERIAL APPLICATIONS TO MODEL CANCER VASCULATURE

A critical limitation of in-vitro models of human vasculature, particularly those used to identify anti-angiogenic drugs for cancer treatment, is that they only model the effects of initial vascular disruption and are unable to model unintended side effects and long-term effects of treatment. The studies reviewed in the current manuscript represent a toolset that may be applied toward modeling long-term effects of anti-angiogenic drug treatments, including angiogenic rebound, vascular normalization, and the metastatic switch (Fig. 8, Table 2). In this section we speculate on how customizable biomaterials, along with endothelial cells of pathological origin, can be applied toward the creation of advanced in-vitro models of anti-angiogenesis treatment.

Figure 8.

Schematics of Angiogenic Rebound, Vascular Normalization, and Metastatic Switch events after anti-angiogenesis treatment. Angiogenic Rebound is described as a regrowth in cancer vasculature due to restoration of angiogenic growth factor signaling after the cessation of angiogenesis inhibition. Vascular Normalization is described as the brief reorganization of dysfunctional cancer vasculature into well-organized, patent vasculature over the course of angiogenesis inhibition. The Metastatic Switch is described as the phenotypic transformation of numerous cell types in a cancer tumor, including endothelial cells, into highly invasive, migratory cells as a result of angiogenesis inhibition.

Table 2.

Summary of biomaterials engineering tools and their applications in modeling pathological vasculature.

					Biological Context
Biomaterial Design Parameter	Technique	References	Angiogenic Rebound	References	Vascular Normalization	References	Endothelial-Mesenchymal Transition	References
Integrin-binding Laminin and Fibronectin analogs	- Covalent binding of adhesion peptides	49, 79, 80, 82, 83, 86, 87–89	Environment promotes cell attachment, spreading, tubulogenesis and migration before and after anti-angiogenic treatment	47–52, 60, 61, 65	Integrin binding molecules promote lumen formation and stability of patent vascularture during anti-angiogenic treatment	51, 52, 60, 61, 65	Cell adhesion to ECM modulates propensity of EndMT	160, 161
	- Covalent binding of full-length ECM proteins	90
Spatial patterning of integrin-binding ligands and growth factors	- Photopatterning and liquid immersion patterning	84, 85, 93, 97, 127, 128	Vascular networks are tortuous and disorganized prior to anti-angiogenic treatment	11, 17, 28, 29	Tortuous and disorganized vascular networks will be converted to organized, patent vasculature during anti-angiogenesis treatment.	11, 17, 28, 29, 31, 49	Localization of endothelial cells to defined areas and promotion of cell-cell contacts modulates susceptibility to EndMT	162, 163
Juxtacrine and Paracrine signaling between cancer cells and endothelial cells, off-target drug effects	- Microcontact printing, segregation of co-cultured cancer cells and endothelial cells	134–136	Cancer cells and stromal cells increase production of pro-angiogenic growth factors following anti-angiogenic treatment	18, 21–23, 132, 133	Cancer cells and stromal cells secrete pro-angiogenic growth factors during anti-angiogenic treatment, modifying drug dose ranges and schedules necessary for normalization	23, 132, 133	Cancer cells and stromal cells secrete cytokines that mediate EndMT	163, 164
Endothelial cells predisposed to mimick tumor angiogenesis	- Patient-specific endothelial cells	149	Endothelial cells are more proliferative, migratory, and less likely to form stable, patent vasculature	17, 149, 150	Endothelial cells are more proliferative, migratory, and less likely to form stable, patent vasculature. Anti-angiogenic treatment would enable normalization	17, 149, 150	Endothelial cells with prolonged exposure to cancer-derived growth factors and inflammatory growth factors have increased susceptibility to EndMT	20, 26, 27
	- Endothelial cells exposed to cancer-derived conditioned media/inflammatory growth factors	150
Elevated growth factor presence, enhanced growth factor stability	- Covalent incorporation of full-length growth factors	84, 96, 97, 102, 103	Elevated presence and stability of growth factors causes resistance to drug treatments, contributes to rebound following anti-angiogenic treatment	18, 21–23	Increased persistence and stability of growth factors in the microenvironment impacts drug dose rates ane schedules that normalize vasculature	18, 21–23	Balance of VEGF, TGFβ, bFGF and other growth factors critical to determining Endothelial Cell susceptibility to EndMT	159
	- Covalent incorporation of peptide growth factor mimicks	85, 92–95
	- Incorporation of non-covalent growth factor binding ligands	59, 100, 101, 107, 108
Sustained hypoxia	- Use of crosslinking reactions that consume oxygen	130, 131	Hypoxic environment present in tumors due to dysfunctional vasculature, hypoxic environment sustained following anti-angiogenic treatment	11, 17, 21–23	Paracrine signaling from cancer cells are modulated by hypoxia in the tumor environment. This will in turn impact drug dose rates and schedules that normalize vasculature	11, 17, 21–23	Hypoxia and HIF1α are implicated as initiators of EndMT	11, 17, 21–23, 165–169
Elevated stiffness, enhanced environmental stability	- Polymer concentration	49, 53, 110–113	Cell culture material persists after anti-angiogenic treatment, may be repopulated by new vasculature	49, 53–55, 60–65	Environment promotes lumen formation and structurally supports patent vasculature during normalization	49, 53–55, 60–65	Endothelial cells on stiffer matrices more susceptible to EndMT via TGFB signaling	171
	- Crosslinking concentration	49, 109, 114, 115
	- Degradation	113, 116–118, 124–126
	- Decoupled stiffness	55, 109, 119,
	- Biocomposites	120, 121

Generating Pathological Vasculature

To model side effects of anti-angiogenic drug treatment, it will be necessary to generate diseased vasculature, perhaps by using endothelial cells from known pathological origins. These cell types, which may be derived from induced pluripotent stem cells or directly harvested from diseased tissue, can display diseased phenotypes in vascular models prior to exposure to the model environment. For example, endothelial network formation was recently modeled when endothelial cells were derived from induced pluripotent stem cells (IPSCs) of type 1 diabetic patients [146], wherein diabetic patient-derived endothelial cells demonstrated a more severe response to hypoxic conditions [130] compared to non-diabetic endothelial cells. IPSC differentiation protocols can generate endothelial cells from patients with other conditions (e.g. cancer) using a similar approach, which makes iPSC-ECs a particularly valuable tool. Recent studies show that iPSC-ECs can form capillary-like networks on 2D hydrogel surfaces [80] and while encapsulated in 3D hydrogels [146–148].

Primary endothelial cells have been derived from hepatocellular carcinomas and exhibited properties such as enhanced migration, resistance to apoptosis and the ability to form more dense networks of tubules and sprouting capillaries compared to normal endothelial cells [149]. In addition, endothelial cells with prolonged exposure to pathological environments have demonstrated an ability to exacerbate tissue pathology. As an example, human umbilical vein endothelial cells (HUVECs) treated with conditioned media generated from lung carcinoma cell lines and inflammatory growth factors TNFα, VEGF, and bFGF were shown to activate pro-metastatic signaling pathways in lung carcinoma cells [150]. Taken together, these studies demonstrate the potential value of implementing cell types of pathological origin when modeling the effects of anti-angiogenesis treatment.

Several challenges bar the generation of vascular features that are unique to cancer vasculature: abnormally large, tortuous and leaky blood vessels such as mother vessels and vascular malformations; disorganized, hyperpermable and hyperproliferative vessels such as glomeruloid microvascular proliferations; transluminal bridges that travel through larger blood vessels; blood vessels that have variable, rather than constant diameter and wall thickness (Figure 1) [11, 12, 17, 28–30]. Numerous biomaterial properties may be leveraged in tumor models to control the generation of cancer vasculature. For example, blood vessel tortuosity is predicted to be the product of unusually high integrin binding to a heterogeneous ECM [11, 12, 17]. Integrin binding molecules may be patterned heterogeneously into materials to provide anchor points that restrict natural blood vessel lengthening with growth, and therefore generate vessel tortuosity. In a separate example, imbalanced growth factor signaling from the surrounding environment may lead to outcomes such as endothelial hyperproliferation, transluminal bridge formation, loosening of endothelial cell-cell interconnections and abnormal lumen polarity [11, 17]. Imbalanced growth factor signaling may be recapitulated by co-culture with growth factor-secreting fibroblasts and cancer cells, as well as designing materials that lack the ability to bind to growth factors in ways that generate stable growth factor gradients [17]. Growth factor presence in the environment should be at higher sustained concentrations than normal, and should not directionally guide blood vessel growth in a systematic way. To mediate the formation of abnormally large blood vessels, chosen culture materials should be extensively degradable in order to permit blood vessel widening [11, 17]. Heterogeneity throughout the material may also be used to resemble the heterogeneity of tumor microenvironments and encourage the formation of blood vessels with changing diameter at different points of a vessel [12].

Angiogenic Rebound

Angiogenic rebound is the excessive reassembly of vascular networks that were initially disrupted by anti-angiogenic drug treatment [18, 21–23]. Modeling this phenomenon requires the ability to sustain cell adhesion, viability and migration after anti-angiogenic drug treatment. Biomaterial customization approaches reviewed here may be leveraged to add adhesion ligands such as YIGSR, IKVAV and RGD to chemically-defined hydrogels to improve cell survival after initial network disruption. Materials may also be tailored for culture of iPSC-ECs and cancer-derived endothelial cells to address a hypothesis that pathological origins may improve endothelial cell drug resistance and angiogenic rebound (Table 1).

One cause of angiogenic rebound is the elevated quantity of angiogenic growth factors secreted by fibroblasts and cancer cells into the extracellular environment after anti-angiogenic treatment [18, 21–23]. These effects can potentially be recapitulated by using photopolymerizable hydrogels and sequential microprinting techniques to culture segregated populations of endothelial cells and growth factor-secreting cells. The addition of growth factor-binding ligands to a biomaterial may also maintain an elevated presence of growth factors for long periods of time. Additionally, the use of hydrogels with hypoxia-induced crosslinking would result in temporary maintenance of hypoxia and consequent upregulation of endogenous growth factor secretion in the model tumor environment. Successful implementation of these features potentially leads to models in which angiogenic rebound is observable in time frames resembling those observed in vivo, and mechanisms may be targeted to prevent network reformation in tumors.

Vascular Normalization

Vascular normalization is the transformation of dysfunctional vasculature into more stable, patent vasculature following short-term anti-angiogenic drug treatment [11, 17, 21–23, 28, 29, 31]. Modeling this phenomenon in vitro may require vascular network features such as network area and stability to be quantifiable as potential markers of normalization events. Increased biomimicry of vascular models which recapitulate aspects of the in-vivo environment such as pharmacokinetics [151] and juxtacrine/paracrine signaling between various cancer, stromal, and endothelial cell types [23, 132, 133] will determine meaningful drug concentration ranges and dosing timelines that predictably trigger normalization. Initial pathological network formation may be achieved through the use of cancer-derived endothelial cells in order to generate disorganized and disrupted vasculature prior to exposure to anti-angiogenic drugs [17, 149]. Studies of vascular normalization may need to be done in materials wherein anti-angiogenic drugs can either improve or disrupt vascular network stability depending on drug concentration drug doses. For example, we previously developed 3D models of endothelial cell network formation in which a hydrogel with low modulus and high RGD concentration caused the VEGF inhibitor SU5416 to stabilize vascular networks rather than disrupt vascular network formation [49]. Finally, customizable materials may need to be adapted for systems that permit fluid perfusion through model vasculature in order to observe the impact of drug treatments on vessel leakiness and stability. Successful generation of normalization models may reveal appropriate dose ranges, time tables and biomarkers to recognize and control normalization in vivo.

Metastatic Switch and Mesenchymal Transition

An important side effect associated with anti-angiogenic drug treatment, specifically the inhibition of VEGF signaling, is the onset of pathological phenotypes in endothelial cells. One example of this phenotypic switch is the Endothelial-Mesenchymal Transition (EndMT), where endothelial cells change phenotypes to resemble mesenchymal cells [122, 152–156] that contribute to cancer growth and metastasis [26, 27, 157]. EndMT has been shown to be dependent on multiple environmental factors, notably TGF-β signaling. Evidence has emerged to demonstrate that inhibiting VEGF signaling by anti-angiogenic treatment increases migratory phenotypes in endothelial cells and cancer cells in mouse models [20], as VEGF is known to attenuate TGF-β activity [158].

The identification of enhanced cancer metastasis as a side effect of angiogenic inhibition highlights an urgent need to characterize the effects of VEGF inhibition in tumor-like environments that influence endothelial responses to TGF-β. These environmental variables include TGF-β concentration [159], integrin interactions with the ECM [160, 161] and β-catenin [162], cell-cell contacts [163], and co-cultures with supporting cell types [164]. Recent evidence also implicates hypoxia as an inducer of EndMT through SNAIL and Smad2/3 activation [165–169]. The effects of specific interactions between endothelial cells and the ECM on EndMT are only beginning to be answered [170]. In one study, cell-adhesive poly(L-Lysine)/hyalouronic acid films increased the propensity of HUVECs to undergo TGF-β1-mediated EndMT on materials with increasing stiffness [171]. Results such as these begin to further implicate pathological extracellular environments as agonists of EndMT. Heparin binding molecules [59, 108] may be another important component to add to EndMT models, as they facilitate binding, concentration and stabilization of soluble TGF-β locally near endothelial cell populations. Other ligands, including PHSRN peptide, act synergistically with RGD in full-length fibronectin to potentially mediate EndMT as well [172]. A greater understanding of ECM-modulation of EndMT may include discovery of appropriate drug dosing ranges for achieving vascular disruption while avoiding EndMT, as well as elucidation of therapeutic targets to prevent the transition.

Anti-angiogenic drug discovery

The generation of physiologically-relevant tumor models will be critical toward designing and improving therapeutics that overcome the limitations of current technologies (Table 1). Future drug discovery models will be required to mimick vascular features that act as known limitations that diminish the efficacy of cancer therapeutics [30]. Lowered drug efficacy is largely due to limited access of drugs to critical blood vessels in the tumor [11, 17]. Barriers to access include poor blood perfusion by cancer vasculature and the binding site barrier (accumulation of drugs at the exterior areas of a tumor due to binding to target molecules) [30]. Lowered drug efficacy can also result from drug resistance of the endothelial cells themselves, either due to aberrant gene expression [10, 13], and an unbalanced growth factor signaling environment maintained by cancer and perivascular cell types [11].

Consequently, two critical elements of simulating the above limitations are perfusion of model blood vessels, and comprehensive monitoring of marker expression changes by endothelial cells during and following anti-angiogenic drug treatment. Materials techniques described here would contribute to the generation of cancer-like, perusable vasculature that would allow scientists to track the penetration of drugs into a tumor, and whether treatment normalizes vasculature. Because tumors differ biologically from one another [30], the use of cancer cells and endothelial cells derived from tumors would be ideal for vetting potential therapeutics. Most cancer models continue to use HUVECs as endothelial components of tumor models, and future models should begin to use cancer-derived or patient-derived endothelial cells for the benefit of targeting markers and signaling pathways that are specific to separate tumor types. The use of combined cancer cells and endothelial cells in microtumor models has been used to test both anti-cancer and anti-angiogenic drugs on both cell types simultaneously [139, 145], and these would be ideal platforms for testing the effects of combined drug delivery for cancer treatment. The application of these models following failed clinical trials can rapidly improve the efficacy of drugs that would have otherwise been successful. Similarly, cancer models developed as screening tools can determine therapeutics that overcome known barriers early, then see validation in animal models and clinical trials [30]. Finally, successful modeling of therapeutic limitations as well as growing knowledge on drug interactions with materials may bring insight into materials-mediated drug delivery techniques. Existing biomaterials-mediated drug delivery techniques, such as transport of micro-and-nanoparticles through tumors, and MMP-based drug release from ECM-derived materials are reviewed in greater detail elsewhere [30, 173, 174].

CONCLUSION

Recently developed technologies in biomaterials development are leading to unprecedented control over in-vitro endothelial cell culture environments as well as the ability to model mechanisms of normal and pathological angiogenesis. With the systematic utilization of the techniques reviewed here, models of cancer vasculature will likely improve understanding of far-reaching side effects of anti-angiogenic treatment, which may translate into improvements in cancer therapy.