PNJ Scaffolds Promote Microenvironmental Regulation of Glioblastoma Stem-Like Cell Enrichment and Radioresistance

Abstract

Glioblastoma (GBM) brain tumors contain a subpopulation of self-renewing multipotent Glioblastoma stem-like cells (GSCs) that are believed to drive the near inevitable recurrence of GBM. We previously engineered temperature responsive scaffolds based on the polymer poly(N-isopropylacrylamide-co-Jeffamine M-1000 acrylamide) (PNJ) for the purpose of enriching GSCs in vitro from patient-derived samples. Here, we used PNJ scaffolds to study microenvironmental regulation of self-renewal and radiation response in patient-derived GSCs representing classical and proneural subtypes. GSC self-renewal was regulated by the composition of PNJ scaffolds and varied with cell type. PNJ scaffolds protected against radiation-induced cell death, particularly in conditions that also promoted GSC self-renewal. Additionally, cells cultured in PNJ scaffolds exhibited increased expression of the transcription factor HIF2α, which was not observed in neurosphere culture, providing a potential mechanistic basis for differences in radio-resistance. Differences in PNJ regulation of HIF2α in irradiated and untreated conditions also offered evidence of stem plasticity. These data show PNJ scaffolds provide a unique biomaterial for evaluating dynamic microenvironmental regulation of GSC self-renewal, radioresistance, and stem plasticity.

Graphical Abstract

1. Introduction

The brain provides essential support for the malignant progression and lethality of glioblastoma (GBM) tumors. Some of this support arises from specialized microenvironments, or niches, that harbor glioblastoma stem-like cells (GSCs). GSCs exhibit features of neural stem cell (NSC) phenotypes, including self-renewal, a capacity for multipotent differentiation, and expression of NSC marker proteins. GSCs are believed to serve at the hierarchical apex of GBM progression, enabling regrowth of the tumor following conventional therapy.^[1–3] GSC niche microenvironments exist in at least three distinct anatomical locations: in hypoxic regions of the tumor core, at the invasive edge of the tumor, and near blood vessels in nutrient-rich perivascular spaces.^[1,4–6] These microenvironments provide a diverse set of regulatory functions that support GSC phenotypes and have been shown to contribute to treatment resistance, as well as tumor recurrence.^[7–13]

Identifying and disrupting critical regulatory mechanisms of the GSC niche may provide an avenue for sensitizing this persistent cell population to treatment. Radiation is the most effective post-surgical treatment for GBM, although it is well-understood that not all cells are susceptible.^[14] GSCs are characteristically radioresistant, and radiation exposure is reported to enrich GSC-specific markers in vitro.^[14–17] In addition, GSCs display increased radioresistance in vivo compared to in vitro culture, which suggests that microenvironmental support is critical to this phenotype.^[9,10] In situations where GSCs are depleted, the niche may promote dedifferentiation of remaining GBM cells to maintain the stem fraction.^[18–21] This bidirectional differentiation (interconversion) is a result of stem cell plasticity that is induced by the microenvironment, epigenetic factors, or in response to treatment. Thus, GSCs are able to maintain the hierarchical structure of GBM through the treatment resistance and stem-plasticity promoted by the microenvironment.

Designing assays to reliably measure microenvironmental regulation of stem phenotypes and radioresistance is challenging in standard in vitro cultures that drive cellular responses through non-physiological microenvironmental cues.^[22] This is a particular obstacle when comparing stem and non-stem tumor cell populations, which require divergent culture conditions that produce major, non-specific impacts on cellular growth (i.e., serum vs. serum free; growth factor addition; adherent vs. suspension culture).^[7] Conventional GSC culture conditions (e.g., large neurospheres or organoids) are simplified, often introducing hypoxia and gradients of nutrient access that effectively model cellular responses of the tumor core only. Further, traditional approaches do not typically permit selective manipulation of the microenvironment itself, and the vast majority of existing models focus on the hypoxic core of solid tumors, which neglects aspects of the microenvironment supporting radiation resistance along the invasive edge or perivascular niche. Thus, current tools for studying GSCs in vitro are not well-suited for investigating microenvironmental regulation of GSCs, and the mechanisms that facilitate GSC contributions to tumor recurrence beyond the necrotic tumor core remain unclear.^[23–26]

To address this gap in knowledge, we focused on modeling nutrient-rich niches that support GSCs, which we predicted this would facilitate evaluation of microenvironmental and cellular contributions to radioresistance. Our goal was to model the nutrient rich niches in which GSCs develop radiation resistance. We utilized poly(N-isopropylacrylamide-co-Jeffamine M-1000 acrylamide) (PNJ) scaffolds, which were previously developed to enrich GSCs from patient derived samples. PNJ is a temperature responsive polymer that forms a thermally reversible scaffold for use as a 3D cell culture platform. The thermal reversibility enables cell recovery under mild conditions by cooling cultures to room temperature.^[25,27] Our prior work established that patient-derived GBM cells cultured within soft (~150–325 Pa) PNJ scaffolds demonstrated enriched self-renewal, stem marker expression, and decreased radiation sensitivity compared to neurosphere cultures.^[25] In the present work, we studied self-renewal and radioresistance of GSC enriched cultures as a function of microenvironmental composition. We hypothesized that GSCs would respond dynamically to physical and chemical changes in microenvironmental composition.

Molecular characterization of GBM historically yielded four subtype classifications (proneural, neural, classical, and mesenchymal), although more recent investigation suggests that the neural subtype may result from contamination with nontumor cells.^[28] In these studies, we focused our analyses on two patient-derived models of GBM representing proneural (GBM3^[25,29,30]) and classical (GBM7^[25,29]) tumors. These models carry distinct features in terms of molecular profile and clinical features. The proneural subtype is characterized by mutated p53, IDH1, and PDGFR, while the classical subtype is driven by amplified EGFR and loss of PTEN. Phenotypically, proneural tumors carry the best prognosis in terms of median survival, while classical tumors are more quickly lethal. Here, we demonstrate that self-renewal is regulated as a function of both GBM model subtype as well as the microenvironment, and we also show that all GSC enriched cultures are more radioresistant when maintained in PNJ scaffolds compared to neurosphere cultures. Molecular responses to radiation included maintenance of a population of HIF2α expressing cells, which was not observed in neurosphere cultures. Moreover, expression of NESTIN was either maintained or increased following radiation of scaffold cultured cells even in conditions with a small self-renewing population, which provides evidence for stem plasticity. The de novo expression of HIF2α for cells cultured within PNJ provides a novel mechanistic connection between radiation resistance and GSC plasticity in these microenvironments. Notably, these studies highlight potential subtype-specific differences in GSC response to microenvironmental cues and advance this biomaterial platform to study cell behaviors within nutrient-rich microenvironments akin to the perivascular niche.

2. Results

2.1. PNJ Scaffold Characterization

PNJ polymers were synthesized were synthesized with 10 wt/wt% (PNJ10) or 20 wt/wt% (PNJ20) Jeffamine M-1000 acryalmide (JAAm) and characterized by NMR as previously reported.^[25] PNJ scaffolds were formed with low (5 wt/v%) or high (10 wt/v%) polymer concentration in GSC media to generate 4 distinct scaffold compositions: PNJ10_Low, PNJ10_High, PNJ20_Low, PNJ20_High. To assess the rate of transport of the model protein epidermal growth factor (EGF), diffusion cells were filled with either water or PNJ scaffolds (Figure 1A). In water, EGF was characterized by a measured diffusion coefficient of 3.0E-6 cm² s⁻¹. In scaffolds, the measured diffusion coefficients were 1.28E-6 cm² s⁻¹ in PNJ10_Low, 2.70E-6 cm² s⁻¹ in PNJ20_Low, 6.43E-7 cm² s⁻¹ in PNJ10_High, 1.67E-6 cm² s⁻¹in PNJ20_High. The diffusion of EGF through all tested scaffold conditions was lower than the diffusion of EGF through water, and both JAAm content and total polymer concentration affected the diffusion coefficient. Increasing the total polymer concentration yielded a lower measured diffusion coefficient, while increasing JAAm content was observed to yield a higher measured diffusion coefficient. These differences are expected and can be attributed to changes in scaffold equilibrium water content, which will be higher in low polymer concentration scaffolds and in scaffolds possessing a higher relative JAAm content.^[31] The error observed in diffusivity measurements was likely due to the small size of the diaphragm channels, which produces inconsistencies in the edges of the scaffolds. Overall, these measured values are within range of the diffusion of EGF through brain tissue, which has been reported to be 5.18 ± 0.16E-7 cm² s⁻¹.^[32] These results confirm that protein diffusion is restricted in scaffolds as a function of scaffold composition within a general range expected for EGF diffusion through brain tissue. We previously reported the mechanical properties of PNJ scaffolds, which are characterized by moduli similar range to brain tissue (PNJ10_Low G’ = 325 Pa, G* = 367 Pa; PNJ10_High G’ = 972 Pa, G* = 1046 Pa; PNJ20_Low G’ = 153 Pa, G* = 164 Pa; PNJ20_High G’ = 617 Pa, G* = 652 Pa).^[25] This range of stiffness is comparable to healthy brain, which has been demonstrated to possess microregional variability of 100–1,000 Pa.^[33] Using these data, we determined the relationship of the diffusion coefficient, storage modulus, and scaffold composition (Figure 1B). Increasing JAAm decreased stiffness and increased diffusion; increasing the total polymer content increased scaffold stiffness and decreased diffusion.

Figure 1.

PNJ scaffold characterization. (A). EGF diffusion measured across a diaphragm diffusion cell with PNJ scaffolds or water (PBS + 1% BSA) separating the source and sink chambers.^[34] The diffusion of EGF in brain was reported by Thorne et al.^[32] (B) PNJ scaffolds effectively represent the shear modulus range reported for brain tissue,^[35–38] while also slowing EGF diffusion toward more physiological levels compared to fully liquid culture. Increasing JAAm leads to increased rates of diffusion (y-axis); increasing the total polymer content effects an increase in scaffold stiffness (x-axis). The PNJ shear storage modulus shown here has been reported previously by us.^[25]

2.2. Stem Cell Frequency

Self-renewal capacity of GSCs cultured in PNJ scaffolds or neurosphere conditions was measured by an in vitro limiting dilution assay. The limiting dilution assay is a gold-standard technique for in vitro assessment of stem cell frequency; by bringing cell concentrations to 1 cell per well, it is possible to distinguish between self-renewal and proliferative capacity, thus enabling calculation of stem cell frequency within a mixed population of cells.^[39] We previously reported that both GB3 and GB7 cells exhibited a significant increase in stem cell frequency when cultured in PNJ10_Low and PNJ20_Low scaffolds.^[25] Here, we expanded the library of materials tested and observed that support for stem cells in higher stiffness scaffolds (PNJ10_High and PNJ20_High) depended on cell type (Figure 2). Self-renewal capacity was reduced for GB3 (proneural subtype) cultures in both high stiffness scaffolds (p < 0.01, Figure 2A, B). GB7 (classical subtype) cells produced an opposing response to changing scaffold conditions. In this cell line, high stiffness scaffolds produced a significant increase in self-renewal compared to neurosphere conditions (p < 0.01, Figure 2C, D), which was similar to the increase observed in low stiffness scaffolds. Furthermore, PNJ20_High scaffolds elicited a significant increase in self-renewal compared to all other scaffold conditions (p < 0.01). Therefore, the combination of high PNJ and JAAm content were significant factors in promoting GB7 self-renewal. These data suggest that GSC self-renewal is affected by both PNJ concentration and JAAm content through mechanisms that may be differentially regulated according to cell type.

Figure 2.

GSC self-renewal measured in an in vitro limiting dilution assay following culture in neurosphere or PNJ scaffold culture conditions. (A, B) GB3 cells exhibited a significant decrease in self-renewing cells in high stiffness PNJ scaffolds compared to neurosphere and an increase in self-renewing cells in low stiffness PNJ scaffolds (** p < 0.01 compared to neurosphere; ## p < 0.01 compared to low stiffness scaffold conditions). (C, D) Self-renewing GB7 cells were observed to be significantly increased in all low and high stiffness scaffold conditions compared to neurosphere culture; PNJ20_High scaffolds maintained significantly higher self-renewal capacity than all other conditions (** p < 0.01 compared to neurosphere; ## p < 0.01 compared to other scaffold conditions). Percentages of self-renewing cells were calculated and tested for statistical differences using the Extreme Limiting Dilution Analysis software ^[39]. Note that the raw data express a log fraction of responsive wells vs number of cells, which is the standard representation of self-renewal data in this assay. Self-renewal data for PNJ10_Low and PNJ20_Low scaffold cultures has been previously reported and is shown here for comparison.^[25]

2.3. Functional Radiation Response

The response of cells to radiation exposure in different scaffold conditions was first evaluated by TUNEL assay, which labels cells in late-stage apoptosis (Figure 3 and 4). GB3 apoptosis was significantly reduced (p < 0.05) in PNJ10_Low and PNJ20_Low scaffolds compared to neurosphere conditions 48 hours after irradiation (Figure 3B). However, GB3 apoptosis was not significantly affected in PNJ10_High and PNJ20_High scaffolds compared to controls following irradiation. PNJ scaffolds also altered the number of apoptotic GB7 cells following radiation. We observed a significant decrease (p < 0.01) in apoptotic cells in PNJ10_High, PNJ20_Low, and PNJ20_High scaffolds compared to neurosphere conditions (Figure 4B).

Figure 3.

Functional response of GB3 GSCs to radiation treatment. Cultures were treated (2 Gy) and given 48 hrs to recover prior to analysis. (A) Apoptotic cells were identified via TUNEL staining (red). Unrepaired DNA damage was assessed via γH2AX staining (green). Proliferation was assessed via Ki67 staining. (B) Low stiffness scaffolds significantly reduced radiation induced cellular apoptosis compared to neurosphere culture (* p < 0.05). (C) DNA damage was observed at low levels and not different between conditions. (D) Cell proliferation was low and not different between groups. Quantification (Cell Fraction) is presented as number of staining events normalized to the number of cells estimated from nuclear counterstaining (DAPI). Analyses were performed on a minimum of two scaffold replicates, with details of cell counts provided in Supplementary Information. Significant differences were evaluated using one-way ANOVA with Bonferroni post-tests; scale bars = 100 μm.

Figure 4.

GB7 functional response to radiation treatment. Cultures were treated (2 Gy) and given 48 hrs to recover prior to analysis. (A) Apoptotic cells were identified via TUNEL staining (red). Unrepaired DNA damage was assessed via γH2AX staining (green). Proliferation was assessed via Ki67 staining. (B) Radiation induced cellular apoptosis was reduced in PNJ scaffolds, significantly so in PNJ10_High, PNJ20_Low, and PNJ20_High conditions compared to neurosphere culture (** p < 0.01). (C) DNA damage was observed at low levels and not different between conditions. (D) Cell proliferation was low and not different between groups. Quantification (Cell Fraction) is presented as number of staining events normalized to the number of cells estimated from nuclear counterstaining (DAPI). Analyses were performed on a minimum of two scaffold replicates, with details of cell counts provided in Supplementary Information. Significant differences were evaluated using one-way ANOVA with Bonferroni post-tests. Scale bars = 100 μm.

We next measured functional markers of DNA damage (γH2AX) and proliferation (Ki67) following irradiation. Both GB3 (Figure 3A, C) and GB7 (Figure 4A, C) showed relatively little activation of DNA damage repair, which was not significantly different than responses observed in neurosphere culture. GSC proliferation, as identified by the intranuclear marker Ki67, also did not significantly change following radiation in either GB3 (Figure 3A, D) or GB7 (Figure 4A, D) scaffold cultured cells compared to neurosphere culture. GB7 cells tended toward lower overall Ki67 activity following radiation compared to GB3. The results of these functional assays following irradiation indicate that scaffold conditions provide protection from radiation induced apoptosis in vitro but do not impact recovery, as measured by DNA damage repair or proliferation at 48 hrs after treatment. Moreover, conditions that enriched the self-renewing population also limited radiation-induced cell death, which is consistent for the known role of GSCs in vivo.

2.4. GSC Marker Expression

We next investigated the expression of proteins that regulate GSC phenotypes related to tumor malignancy, including NESTIN, HIF1α, and HIF2α. In untreated GB3 cultures, cells in PNJ10_Low, PNJ20_Low and neurosphere conditions all showed high expression of NESTIN (Figure 5A). However, PNJ10_High and PNJ20_High culture conditions produced a significant decrease in NESTIN (Figure 5C). After radiation treatment, PNJ10_Low and PNJ20_Low cultures maintained NESTIN expression, but expression decreased in neurosphere cultures. Notably, NESTIN expression increased (p < 0.01) in PNJ20_High conditions compared to untreated PNJ20_High conditions (Figure 5C). In untreated GB7 cultures, NESTIN expression was observed in all tested conditions (Figure 6A). Radiation treatment yielded a significant increase in NESTIN expression in both PNJ10_Low (p < 0.05) and PNJ10_High (p < 0.01) compared to radiated neurosphere cultures; PNJ10_High NESTIN expression was also significantly increased (p < 0.05) compared to untreated PNJ10_High culture (Figure 6C). Similar to GB3, NESTIN expression following radiation was lowest in neurosphere conditions.

Figure 5.

GB3 molecular response to radiation in PNJ scaffolds and neurosphere conditions. (A) Expression of NESTIN (red) and (B) HIF2α (red) was measured in untreated (0 Gy) and radiated (2 Gy) conditions after a 48 hr recovery period following treatment. (C) Untreated (0 Gy) high stiffness scaffolds produced significantly decreased NESTIN expression compared to neurosphere conditions (* p < 0.05, ** p < 0.01). Irradiated (2 Gy) cells in PNJ20_High scaffolds exhibited increased NESTIN compared to untreated conditions (## p < 0.01). (D) Untreated (0 Gy) cells in PNJ10_High scaffolds exhibited significantly higher HIF2α expression compared to neurosphere cultures (** p < 0.01). While all irradiated (2 Gy) PNJ scaffold conditions maintained a subset of HIF2α expressing cells, expression in PNJ10_High cultures decreased significantly compared to untreated conditions (* p < 0.05). Quantification of expression level is presented as area of staining (NESTIN or HIF2α) normalized to area of nuclear counterstain (DAPI). Analyses were performed on a minimum of two scaffold replicates, with details of cell counts provided in Supplementary Information. Significant differences were evaluated using one-way ANOVA with Bonferroni post-tests. Scale bars = 100 μm.

Figure 6.

GB7 molecular response to radiation in PNJ scaffolds and neurosphere conditions. (A) Expression of NESTIN (red) and (B) HIF2α (red) was measured in untreated and radiated (2 Gy) conditions with a 48 hr recovery period following treatment. (C) Untreated (0 Gy) scaffold and neurosphere conditions produced similar expression of NESTIN. Irradiated (2 Gy) PNJ10 scaffolds exhibited increased NESTIN that was statistically significant compared to neurosphere cultures (* p < 0.05, ** p < 0.01), and for PNJ10_High, compared untreated conditions (# p < 0.05). (D) Untreated (0 Gy) PNJ20_High scaffolds produced significantly higher expression of HIF2α compared to neurosphere cultures, which did not express HIF2α (** p < 0.01). Irradiated (2 Gy) PNJ scaffold cultures all maintained a subset of HIF2α expressing cells, while expression in neurosphere cultures was rare. Quantification of expression level is presented as area of staining (NESTIN or HIF2α) normalized to area of nuclear counterstain (DAPI). Analyses were performed on a minimum of two scaffold replicates, with details of cell counts provided in Supplementary Information. Significant differences were evaluated using one-way ANOVA with Bonferroni post-tests. Scale bars = 100 μm.

The transcription factor HIF1α was consistently expressed at low levels by GB3 cells, but did not show spatial patterning from the exterior to the core of cell clusters (which would have been suggestive of hypoxia). There were no significant differences or trends in HIF1α expression across the neurosphere and PNJ culture groups (Supplemental Figure 1A, C). HIF1α was expressed consistently in both neurosphere and PNJ cultured GB7 with no significant differences between groups and no evidence for spatial patterning (Supplemental Figure 1B, D). These results suggest normoxic conditions for PNJ cultured cells, which is consistent with our prior observation that PNJ microenvironments are likely nutrient-rich and not hypoxic.^[25]

HIF2α protein expression detected by IF staining was restricted to the cytoplasm. This spatial localization is consistent with past reports suggesting cytoplasmic signal is indicative of normoxic stabilization of the transcription factor.^[6,40] In GB3 cells, HIF2α expression was observed with strong IF staining in a subset of cells in all untreated scaffold conditions (Figure 5B) but was generally absent in neurosphere conditions. Compared to neurosphere culture, expression of HIF2α was significantly increased in PNJ10_High (p < 0.01) (Figure 5D). Similar to GB3, expression of HIF2α in GB7 cells was consistently observed with strong staining in PNJ scaffold conditions, while neurosphere cultures again produced only sporadic HIF2α stabilization (Figure 6B). PNJ20_High scaffolds elicited significantly increased HIF2α expression compared to neurosphere culture, while other scaffold conditions produced comparable expression. In response to radiation, HIF2α expression by GB3 cells significantly decreased (p < 0.05) in PNJ10_High but was still expressed at some level in all PNJ cultures. Radiation treatment did not significantly alter HIF2α expression in GB7 scaffold cultures. Neurosphere cultures showed a slight increase in HIF2α in response to radiation, but this was not statistically significant. Thus, taken as a whole, HIF2α expression and HIF2α expressing cells were maintained following irradiation of cells in PNJ scaffolds

Quantitative comparisons of GB3 and GB7 self-renewal capacity, expression of NESTIN and HIF2α, along with their response to radiation were graphically summarized (Figure 7).

Figure 7.

Summary of GB3 and GB7 behaviors compared across PNJ scaffolds and neurosphere cultures. Colors are scaled such that the maximum measurement represents 100% in red, 50% of the maximum in yellow, and 0% in green. Color scales are set for individual columns except for NESTIN and HIF2α, which include both the radiated and untreated columns. Significant differences for responses measured in PNJ cultures compared to neurosphere cultures are represented by * (p < 0.05) and ** (p < 0.01) using one-way ANOVA with Bonferroni post-tests.

3. Discussion

The GSC population consists of highly tumorigenic, self-renewing cells^[3] residing in specialized microenvironments^[1] that are resistant to conventional treatments, including chemotherapy^[41] and radiation.^[14] GSC niches provide dynamic regulatory support for both self-renewal and radioresistance.^{[1,7–11,18]} Niche microenvironments are believed to enable stem-plasticity by directing the dedifferentiation of neoplastic cells to acquire the GSC phenotype.^[18] Following conventional treatment, recurrent GBM tumors have been shown to be enriched with GSCs compared to matched primary tumor samples.^[41] GSCs are thus believed to play a prominent role in driving near universal rates of tumor recurrence and a correspondingly low patient survival, and microenvironmental support plays an important role in mediating this behavior.

Isolating specific GSC-niche interactions in vivo is generally not feasible, given the complex landscape of the intact and complete tumor microenvironment. In contrast to the complexity of the in vivo circumstance, modeling GSC niches in vitro provides opportunities for mechanistic analysis of how distinct microenvironmental features support or hinder GSC maintenance.^[24] We previously engineered thermally reversible PNJ scaffolds for culturing stem cells.^[25] PNJ is an optically clear, viscous solution at room temperature, mechanically tunable, and capable of being biologically inert or presenting cell adhesion ligands.^[27] Together, these features enable facile recollection of cells under mild temperature stimulus and manipulation of microenvironmental parameters to study cell responses in situ. Notably, our scaffolds are not intended to model a solid tumor, in which the core is frequently necrotic and/or hypoxic. It was not possible for us to estimate a mesh size for these networks due to experimental challenges in measuring equilibrium swelling for such dilute gels, however, we do not expect that the size of the network poses hindrance to cellular movement or growth. We previously demonstrated that these scaffolds actively enrich GSCs compared to conventional neurosphere conditions, and we proposed that this enrichment could present a model of stem maintenance shared by the nutrient-rich perivascular niche.^[25] In this study, we utilized PNJ scaffolds to measure GSC self-renewal and radioresistance in response to changes in scaffold composition.

GSC niches may contribute to radioresistance through mechanisms that include interactions with the ECM,^[42,43] soluble signaling factors,^{[7,11,14,44–46]} and hypoxia.^{[6,18,47–52]} Of these components, hypoxia has received been focused on as providing the major degree of protection against radiation.^[53] Hypoxia most certainly plays a central role in cellular behavior within the tumor core. Mechanistically, radiation generates free radicals such as reactive oxygen species (ROS) that subsequently induce double strand break DNA damage. In oxygen restricted conditions, the capacity for ROS generation is reduced, thereby protecting cells from DNA damage. In addition, GSCs respond to decreases in oxygen through activation of a number of signaling pathways, including the hypoxia inducible transcription factors 1α (HIF1α) and 2α (HIF2α). HIF1α is stabilized under conditions of chronic hypoxia, while HIF2α provides an early response to hypoxia and – importantly – can also be stabilized in normoxic conditions.^[6,51,54] Clinically, expression of HIF2α, but not HIF1α, is correlated with poor prognosis in GBM.^[55] Experimentally, HIF2α has been shown to promote stem plasticity.^[18,19,54] Furthermore, activation of these transcription factors may decrease GBM radiation sensitivity through downstream activation of HIF target genes that include both stem and survival pathways.^[6,18,52,56] HIF2α can thus be a mechanism of radiation resistance that functions independently from HIF1α, under conditions of normoxia that may be present beyond the tumor core. Our results support HIF2α as an important aspect of mechanistic responses to the microenvironment and to radiation treatment.

In vivo, the tumor microenvironment ECM sequesters soluble signaling factors, which may serve as a depot for GSC signaling, while also providing mechanical input signals.^[7,57] Mitogenic growth factors such as EGF and FGF are potent regulators of GSC phenotypes and requisite for stem maintenance in vitro.^[58] Additionally, soluble factors promote GBM radioresistance via growth factor (EGF/bFGF,^[7,14] EGFR,^[59,60] TGFβ/TGFβR1,^[11] IGF-1/IGF1R,^[44]) and cytokine signaling (SDF-1/CXCR4^[45,46]). We previously observed that PNJ scaffolds increased expression of EGFR in both GB3 and GB7 cells.^[25] In separate studies, we also observed that PNJ sequestered ovalbumin protein in sink diffusion conditions.^[31] Thus, we predicted that the movement of biomolecules through PNJ scaffolds would be hindered, which would effectively sequester signaling molecules to stimulate autocrine/paracrine signaling.^[31] We tested this hypothesis by measuring the diffusion of EGF (6.6 kDa) as a model growth factor. Our measurement of EGF diffusion through water was comparable to reported values, nearly an order of magnitude greater than the reported diffusion coefficient through brain tissue.^[32] As expected, each of the PNJ scaffold formulations limited growth factor diffusion compared to diffusion through water (Figure 1). PNJ concentration and JAAm content independently produce a measureable effect on both EGF diffusion and scaffold stiffness (measured by the storage modulus G’). Increasing JAAm content produced an increase in diffusion and a decrease in G’, while increasing PNJ concentration decreased diffusion and increased G’. The 4 PNJ scaffold formulations cover the reported stiffness of brain tissue (100–1,000 Pa^[33,35–37]), and slow growth factor diffusion toward more physiological levels compared to standard liquid media culture conditions.^[32] Solid tumors are often more stiff than surrounding tissues,^[38] and the majority of available biomaterial models focus on relatively high stiffnesses, in the range of 10s of kPa or even in a MPa range,^[61] however, such high stiffnesses are not likely to occur outside the tumor core, in regions where small clusters of cells have invaded healthy brain or perivascular spaces. Modeling the larger neurosphere clusters that would be characteristic of a solid tumor and time of recurrence remains a subject of future work. Here, we focused on developing softer materials that retain features of the brain microenvironment present in the invasive edge or perivascular niche, where GSCs are known to reside and escape conventional treatments. The sequestration of growth factor further distinguishes PNJ from traditional neurosphere and organoid cultures by modeling a nutrient rich environment in which smaller clusters of cells can engage in close-contact paracrine signaling.

Our data support both microenvironmental stiffness and nutrient availability (sequestration) as being important factors mediating GSC enrichment and their phenotypic response to radiation. There is active discussion regarding the role of stiffness in regulating GBM cell behavior,^[62] and less is understood about the specific responses that would be expected for GSCs. Self-renewing NSCs were reported by Saha et al. to be preferentially enriched on substrates with stiffness ≥100 Pa.^[35] Microenvironmental stiffness and chemistry have been shown to regulate the migration and invasive capacity of GSCs in vitro,^[62–66] and growth factor accessibility is known to be critical for maintaining the stem fraction, particularly in neurosphere cultures.^[67,68] These different microenvironmental characteristics would be expected to independently regulate GSC behaviors in vitro. To our knowledge, GSC self-renewal capacity has not been studied extensively as a function of the 3D biophysical microenvironment. Our prior work indicated that both GB3 and GB7 exhibited significantly increased populations of self-renewing cells in low stiffness (~150–325 Pa) PNJ10_Low and PNJ20_Low scaffolds.^[25] Here, we observed that GB3 cells cultured in higher stiffness scaffolds (~617–971 Pa) showed a significant decrease in the self-renewing population, whereas GB7 cells showed the opposite response. These data demonstrate that regulation of self-renewal by the physical microenvironmental is dependent on cell type. Grundy et al. described subtype-specific regulation of GSC migration, whereby neural subtype GSCs migrated efficiently on soft substrates and mesenchymal GSCs were poorly motile.^[62] In the present study, GB3 cells belong to the proneural classification, which phenotypically best represent an oligodendrocyte lineage.^[69] Jagielska et al. reported that oligodendrocyte precursors are better maintained on soft (100 Pa) substrates but show increased differentiation on substrates with greater stiffness.^[70] This may provide a possible mechanism for the decline in GB3 self-renewal in high concentration scaffolds. GB7 cells, on the other hand, are characterized as classical subtype. For these cells, the self-renewing population showed affinity for each of the tested scaffold conditions. The classical subtype most directly corresponds to an astrocytic lineage,^[69] which reportedly show preference for stiff (>1 kPa) substrates.^[35,71] Both soft and stiff microenvironments enriched GB7 self-renewal, which stands in contrast to GB3. Thus, our data suggest that the observed behaviors may relate to subtype characteristics of their originating tumor. Although additional work is necessary to expand and validate these studies across a greater number of models, these data provide significant early evidence that microenvironmental responses in nutrient rich niches may be GBM subtype dependent.

Since treatment resistance is considered a characteristic of self-renewing GSCs and is actively supported by features of the in vivo tumor microenvironment,^{[1,7,9–11,18]} we investigated PNJ scaffolds for their ability to model radio-protection in vitro. Notably, our prior studies demonstrated that GSCs irradiated within PNJ microenvironments showed greater viability than neurosphere cultured cells, but these assays were simple viability tests that did not probe post-radiation cell behavior during recovery in a 3D environment. Here, we studied changes in cellular apoptosis and expression of stem markers during a recovery period within the microenvironment. One limitation of our analysis is that we did not include a non-irradiated control for apoptosis (TUNEL) staining, however, our experimental designs enabled direct comparison of irradiation in different microenvironments. We direct the reader to published examples of TUNEL staining in GBM [80, 81]; baseline TUNEL levels are low and expected to be confined to the tumor core (for example, no expression of TUNEL was observed in the perivascular niche [80]). In our work, decreased radiation induced apoptosis was observed in GB3 and GB7 cells cultured in scaffolds that enriched the self-renewal capacity (Figure 7). However, PNJ cultured GB3 cells showed a level of radio-protection that extended beyond the pool of self-renewing cells. For GB3, neurosphere culture maintain approximately twice as many self-renewing cells compared to the high stiffness scaffolds, but there was no significant differences in apoptosis between those groups. Thus, one significant observation from this work is that radiation resistance exhibited by cells was not exclusively a function of the self-renewing fraction; these data suggest that other microenvironmental variables likely play a role in how cells respond to radiation induced damage.

In considering mechanisms mediating cellular responses to radiation, we examined expression of γH2AX, which marks DNA double strand breaks and leads to activation of DNA repair machinery.^[72] Phosphorylation of γH2AX occurs within minutes of DNA damage, and it is dephosphorylated following cellular repair. Prolonged activation (>24 hrs) of γH2AX is a precursor to cell death.^[10,73] Therefore, the relative lack of γH2AX expression observed in culture was not necessarily surprising, given that cultures were provided with a 48 hr recovery period following treatment. We also examined intranuclear Ki67 as a marker for proliferation. We did not observe significant Ki67 activity in any of the culture conditions, suggesting that cells had shifted to a less proliferative phenotype to avoid apoptosis.

Expression levels of NESTIN, HIF1α, and HIF2α were measured using immunofluorescence to identify molecular changes in stem regulation that may also contribute to radio-response. Immunofluorescence was chosen for the ability to detect differences in intra-sphere protein spatial distribution (a concern in large neurospheres^[67,74]), and to better identify expression patterns in small cell fractions that would be lost if the cells were homogenized. However, we recognize that these results only allow for semi-quantitative comparisons. Our IF results therefore lays the groundwork for future studies that would dive more deeply into complementary techniques that can more sensitively detect quantitative differences in gene or protein expression.

NESTIN, HIF1α, and HIF2α were evaluated for their critical importance to the GSC phenotype. NESTIN marks the self-renewing population of NSCs,^[75] while in GBM, expression identifies GSCs, is strongly correlated to poor clinical prognosis, and is observed in invasive cells in vivo.^[2,3,76] HIF1α is important for promoting tumor angiogenesis as well as GSC expansion in hypoxic growth conditions.^[48] HIF2α has been reported as a GSC selective biomarker that plays an essential roles in tumorigenicity and angiogenesis via downstream production of VEGF.^[6] Although stabilization of this transcription factor is associated with hypoxia, expression has also been observed in GSCs residing in perivascular niches, acidic microenvironments, and in normoxic in vitro culture conditions.^[6,19,54] Furthermore, Li et al. reported that GSCs with genetic knockdown of HIF2α generated orthotopic GBM tumors exhibiting expression of the transcription factor, which indicates that the in vivo microenvironment selected GSCs that were not efficiently targeted by HIF2α knockdown.^[6] Outside of known function in tumorigenicity and angiogenesis, HIF2α also has a documented role in promoting GBM stem-plasticity.^[18,19] Downstream target genes of HIF2α include Oct4, c-Myc, and Nanog all of which are critical to the development of induced pluripotent stem cells (iPSCs).^[77]

GB3 cells exhibited NESTIN expression that closely mirrored the population of self-renewing cells (Figure 7). Similar patterns of near ubiquitous NESTIN expression were observed in low stiffness scaffolds and neurosphere culture, while expression was absent in high stiffness scaffolds (Figure 5A). Following irradiation, NESTIN significantly increased in PNJ20_High cultured GSCs (Figure 5C). Interestingly, this high stiffness was shown to inhibit self-renewal. Therefore, increased NESTIN in these conditions provides evidence that the PNJ microenvironment may promote stem plasticity in response to radiation, although further analysis of stem features would be necessary to confirm. HIF1α was observed in all PNJ and neurosphere cultures without any obvious spatial patterning, providing evidence that hypoxia was not a significant factor in the differential cellular responses observed in different culture conditions (Supplemental Figure 1A, C). Expression of HIF2α was not reflective of GB3 self-renewal, as high stiffness PNJ10 scaffolds exhibited the highest expression and low self-renewal (Figure 7), although expression decreased following radiation. The transcription factor was observed in all scaffold conditions in both irradiated and non-treated scaffold conditions but not in neurosphere conditions, suggesting that it plays a role in both maintenance of cells as well as their response to irradiation.

In contrast to GB3, GB7 cultures showed consistent expression of NESTIN in all culture conditions, which was not directly reflective of self-renewal status (Figure 7). This may indicate a subtype specific baseline expression or some phenotypic drift as a result of in vitro culture. In response to radiation, GSCs cultured in PNJ10_Low, PNJ10_High were enriched in NESTIN expressing cells. HIF1α expression was again observed in all of the culture conditions with no obvious differences between the groups or evidence for spatial patterning (Supplemental Figure 1B, D), further suggesting that hypoxia is unlikely to be driving cellular responses in these culture models. HIF2α expression in untreated GB7 was reflective of self-renewal, with high stiffness PNJ20 cultures showing the strongest expression. Similar to GB3, we observed a consistent population of HIF2α expressing cells in both untreated and irradiated PNJ cultures that were not found in neurosphere conditions.

GSC plasticity is a critically important mechanism for maintenance of the stem cell pool and development of tumor heterogeneity in vivo.^[20] Stem plasticity or enrichment of the stem fraction has been described both as a response to radiation^[14–16] and separately as a function of HIF2α signaling induced by the tumor microenvironment.^[19,54,78] Here, we observed that NESTIN expression reflected the self-renewing populations for GB3 but not GB7, emphasizing that the role of baseline NESTIN is cell or subtype dependent. Yet following irradiation, all scaffold conditions maintained or increased NESTIN expressing cells. Moreover, PNJ microenvironments consistently maintained a population of HIF2α expressing cells even in conditions where the fraction of self-renewing cells was depleted. Together, these results suggest dynamic regulation of stem phenotypes in PNJ cultures, which promote stem plasticity and survival. To our knowledge, the role of GSC HIF2α expression has not been described in any 3D GSC microenvironment models investigating radioresistance and stem plasticity. One limitation of this study is that we have not identified a singular mechanism that explains scaffold influence on GSC behaviors. It should be noted that our primary goal was to establish the feasibility of PNJ scaffolds as a novel model and to use this novel model to better understand how GSCs present beyond the solid tumor core respond to radiation. The data shown here provide compelling evidence that the composition of perivascular- or brain-mimetic PNJ scaffolds influence GSC behaviors as well as their response to radiation. These data form a foundation upon which future investigation can dive more deeply into mechanistic aspects of this response to better understand GBM recurrence.

4. Conclusions

These studies describe the application of tunable PNJ scaffolds for studying niche microenvironmental regulation of self-renewal and responses to radiation in patient-derived GSCs. PNJ-mediated regulation of self-renewal was dependent on the physical properties of the scaffold as well as on cell type. Our data supports that enhancement of paracrine/autocrine signaling due to sequestration of proteins within the hydrogel scaffold may facilitate GSC enrichment and contribute to radioresistance. Stiffness-dependent mechanosensation is an alternative possibility to account for scaffold-dependent cellular phenotypes that remains the subject of future work. Scaffold cultures were radio-protective, as evidenced by a decreased radiation induced cell death in two GSC models. In response to radiation, all scaffold conditions maintained or increased expression of NESTIN and maintained a population of HIF2α expressing cells, which was not observed in neurosphere culture. Taken together, these data suggest that GSC maintenance is dynamically regulated in PNJ scaffolds to promote radioresistance and stem-plasticity.

5. Experimental Section

5.1. Materials

Chemicals were reagent grade and purchased from Sigma Aldrich (St. Louis, MO, USA) unless otherwise stated. Cell culture reagents were purchased from Thermo Fisher Scientific (Waltham, MA, USA) unless otherwise stated. N-isopropylacrylamide (NIPAAm) purchased from Tokyo Chemical Company (Portland, OR, USA) was purified by recrystallization from hexane, while 2,2’-Azobisisobutyronitrile (AIBN) was purified by recrystallization from methanol. Jeffamine® M-1000 was generously donated by the Huntsman Corporation (Salt Lake City, UT, USA). Jeffamine M-1000 acrylamide (JAAm) was synthesized as previously described ^[31,79].

5.2. Polymer Synthesis

Poly(N-isopropylacrylamide-co-Jeffamine M-1000 acrylamide) or PNJ polymers were synthesized as previously described ^[25,27,31] with NIPAAm:JAAm monomer feed ratios of 90:10 (PNJ10) or 80:20 (PNJ20) by mass. Following synthesis, PNJ polymers were purified by dialysis (3,500 MWCO) against diH₂O at 4°C, lyophilized, and the dry polymer was sterilized with ethylene oxide. Syntheses were confirmed by ¹H NMR in D₂O (400 MHz Varian Inova, Agilent Technologies, Santa Clara, CA, USA), and JAAm incorporation was consistently measured at 68–76% of the feed ratio.

5.3. Patient-Derived GSC Cultures

Patient tissue samples were acquired from the Biobank Core Facility at St. Joseph’s Hospital and Medical Center and Barrow Neurological Institute (BNI) (Phoenix, AZ, USA). All samples were collected and transmitted according to the Biobank Institutional Review Board’s approved protocol. Two low-passage patient-derived GSC cell lines, GB3 and GB7, were established from primary GBM tumors surgically resected at BNI. Both lines were characterized as human GSC models with GB3, classified as a proneural GBM subtype, and GB7 was classified as a classical GBM subtype (detailed characterization of subtype and both cell lines is provided in our prior publications; GB3: ^[25,29,30]; GB7: ^[25,29]). GSCs were propagated in standard serum-free neurosphere suspension culture for less than 20 passages. Briefly, cells were dissociated into a single cell suspension with Accutase, counted (Cellometer Mini, Nexcelom, Lawrence, MA, USA), resuspended in neurosphere media (1:1 mixture of DMEM/F12 supplemented with B27, N2 and penicillin/streptomycin) at 100k cells/mL for GB7 or 300k cells/mL for GB3, and cultured in petri dishes coated with poly(hydroxyethylmethacrylate) (polyHEMA) to prevent cell attachment. Cultures were supplemented with 20 ng/mL of epidermal growth factor (EGF) and 20 ng/mL basic fibroblast growth factor (bFGF) (Merk Millipore, Billerica, MA, USA) every 2–3 days. Neurospheres were passaged at confluence.

5.4. PNJ Scaffold Cultures

PNJ was dissolved at low (5 wt/v%) or high (10 wt/v%) concentration in GSC media overnight at 4°C to generate 4 different PNJ solutions: PNJ10_Low (5 wt/v%), PNJ10_High (10 wt/v%), PNJ20_Low (5 wt/v%), PNJ20_High (10 wt/v%). GSC neurospheres were dissociated with Accutase to a single cell suspension, counted (Cellometer Mini, Nexcelom, Lawrence, MA, USA), and diluted in PNJ-media solution at room temperature (GB3: 500k cells/mL; GB7: 250k cells/mL). PNJ-GSC suspensions were incubated at 37°C to form temperature responsive PNJ scaffolds and encapsulate cells in 3D culture. After 48 hours, an equal volume of warm neurosphere media (no PNJ) was added above the scaffold to maintain nutrient balance. Every 2–3 days, scaffolds were supplemented with EGF and bFGF (20 ng/mL) after being solubilized at room temperature to allow for nutrient distribution. At confluence (7 – 14 days), PNJ scaffolds were solubilized and diluted in cold PBS. The resulting liquid cell suspension was centrifuged to recover live cells for further analysis of cell behaviors. Neurosphere cultures were utilized as control cultures for all evaluations of scaffold growth characteristics.

5.5. EGF Diffusion across PNJ Scaffolds

EGF diffusion was measured across PNJ scaffolds in a diaphragm diffusion cell and compared to its diffusion through water in the same system. PNJ10_Low, PNJ10_High, PNJ20_Low, PNJ20_High, and PBS + 1% BSA (control) solutions were injected in 10 μL aliquots into the central sample chamber of a 3D gel chemotaxis slide (μ-Slide Chemotaxis ^3D, IBIDI, Martinsried, Germany) and heated on a heating block to 37°C to form scaffolds. Diffusion media (PBS + 1% BSA) was added to empty channels on either side of the newly formed scaffold and the slides were equilibrated at 37°C overnight to fully hydrate the scaffolds. The diffusion media on one side of the gel was replaced with diffusion media supplemented with 4 ng/mL of EGF (source channel) to establish up a growth factor gradient such that at t = 0, the concentration in the source channel was 4 ng/mL and the concentration in the sink channel was 0 ng/mL. Diffusion media from the source and sink channels was collected at 48, 96, or 120 hrs for n = 10 scaffolds each. EGF concentration was measured using a sandwich ELISA (DuoSet, R&D Systems, Minneapolis, MN, USA) in comparison to EGF controls. The diffusion coefficient was calculated using the following relationship for diffusion across a diaphragm ^[34]:

$$D=\frac{1}{\beta t}\ln (\frac{C_{source}(0)-C_{sink}(0)}{C_{source}(t)-C_{sink}(t)})$$

Where C_source and C_sink are the EGF concentrations in the source and sink chambers at a given time t; A_s is the scaffold cross-sectional area; W_s is the scaffold width; V_source and V_sink are the volume of the source and sink chambers; t is the time at measurement; and

$$\beta =\frac{A_S}{W_S}(\frac{1}{V_{source}}+\frac{1}{V_{sink}})$$

5.6. Stem Cell Frequency

Stem cell frequency was determined using a limiting dilution assay, as previously described ^[25]. In prior work, we analyzed stem cell frequency in low concentration scaffolds (PNJ20_Low, PNJ10_Low). Here, GSCs cultured in high concentration PNJ scaffolds or neurosphere conditions were recovered, dissociated with Accutase to a single cell suspension, counted (Cellometer Mini, Nexcelom, Lawrence, MA, USA), and resuspended in NSC media. Cells were cultured at initial densities of 1, 2, 5, 10, 20, 50, or 100 cells/well (n = 24 wells for each density) in polyHEMA coated 96-well plates to promote formation of secondary neurospheres and enable analysis of stem cell frequency. Sphere formation potential at each initial density was analyzed with brightfield microscopy (Zeiss Axio Observer A1), and wells that did not produce a sphere (% nonresponsive) were counted. This experiment was replicated 3 times. Data was analyzed using the Extreme Limiting Dilution Analysis (ELDA) software to quantify differences in stem cell frequency produced between neurosphere and scaffold culture conditions using a chi-squared test for pairwise differences ^[39]. Statistical significance is reported for p < 0.05.

5.7. Radiosensitivity

GSCs were cultured in PNJ scaffolds or neurosphere conditions as described in Sections 2.3 and 2.4 for 7–8 days. Scaffold and neurosphere cultures were then treated with 2 Gy ionizing radiation (RS 2000, RAD Source, Suwanee, GA, USA). Following radiation, cultures were returned to the cell culture incubator. After 48 hrs, cells were collected from PNJ and neurosphere cultures for immunostaining and TUNEL assay.

5.8. Immunofluorescence Staining

GSCs were collected from neurosphere or PNJ scaffold conditions as described in Sections 2.3 and 2.4 and processed for immunofluorescence staining using an established protocol ^[74]. Briefly, cells were centrifuged and culture media was removed. Cells were then resuspended in 4% paraformaldehyde (PFA) and incubated for 3 hrs at 4°C for fixation. Fixed cell suspensions were centrifuged to remove PFA, and dehydrated in 30% sucrose for 30 min at 25°C. Dehydrated cell suspensions were centrifuged to remove sucrose, and resuspended in a small volume (~500 μL) of optimal cutting temperature compound (OCT). OCT-cell suspensions were then pipetted on top of a block of frozen OCT and frozen at −80°C. The block of OCT served as a cutting stage for the cells to be cryosectioned. Samples were sectioned at 5 μm thickness, collected on gelatin coated slides, and stored at −80°C until staining. Slides were defrosted at room temperature (5–10 min) and samples were fixed to the gelatin coating with PFA for 5 min. Antigen retrieval was performed by heating samples in 10 mM Sodium Citrate buffer (pH 6.0) for 30 minutes at 80°C. Samples were blocked for 30 min (10% Normal Goat Serum, 0.1 M glycine, 0.3% Triton-X 100 in PBS). Primary antibodies were incubated overnight at 4°C (Supplemental Table 1; buffer: 10% Normal Goat Serum, 0.3% Triton-X 100 in PBS), and secondary antibodies were incubated for 45 minutes at 25°C (Supplemental Table 1; buffer: 10% Normal Goat Serum, 0.3% Triton-X 100 in PBS). Cells were counterstained with DAPI. Samples were imaged with an inverted fluorescence confocal microscope (Zeiss LSM 710 Axio Observer Z1) with imaging settings (laser power, gain) applied uniformly across conditions to enable direct comparison of staining results.

5.9. TUNEL Assay

Samples for TUNEL staining were prepared as described in Section 2.8. However, instead immunostaining, fragmented DNA, indicative of apoptosis, was labeled via TUNEL staining (DeadEnd Fluormetric TUNEL System, Promega, Fitchburg, WI, USA) according to the manufacturer’s protocol. TUNEL staining was performed 48 hrs after radiation as described in Section 2.7. TUNEL stained cells were counterstained with DAPI.

5.10. Image Analysis

Image analyses were performed in ImageJ (National Institutes of Health, version 1.47). For each image, a native threshold function (RenyiEntropy) was applied and the area of positive staining was measured and normalized to DAPI area. This method enabled semi-quantitative comparison of normalized immunostaining results across different PNJ scaffold and neurosphere culture conditions. Linear corrections to image brightness and contrast were performed on images represented in the presented figures to improve visual clarity only. These corrections were not applied to the raw images during semi-quantitative analysis. Because the expression of some markers remained close to zero in neurosphere conditions, data are not represented as a fold-change, and we instead focus on normalization to the DAPI signal. Each experiments represents a minimum of two scaffold replicates and measurements were conducted on no fewer than 4 randomly selected images. The number of cells counted for all semi-quantitative analysis of immunofluorescence is provided Supplemental Table 2. Statistical tests for determining differences in immunostaining and TUNEL staining were performed in Prism 5 (GraphPad) using a one-way ANOVA with Bonferroni multiple comparisons for post-hoc analyses. Statistical significance is reported for p < 0.05.