Developing Implantable Scaffolds to Enhance Neural Stem Cell Therapy for Post-Operative Glioblastoma

Abstract

Pre-clinical and clinical studies have shown that engineered tumoricidal neural stem cells (tNSCs) are a promising treatment strategy for the aggressive brain cancer glioblastoma (GBM). Yet, stabilizing human tNSCs within the surgical cavity following GBM resection is a significant challenge. As a critical step toward advancing engineered human NSC therapy for GBM, we used a preclinical variant of the clinically utilized NSC line HB1.F3.CD and mouse models of human GBM resection/recurrence to identify a polymeric scaffold capable of maximizing the transplant, persistence, and tumor kill of NSC therapy for post-surgical GBM. Using kinetic bioluminescence imaging, we found that tNSCs delivered into the mouse surgical cavity wall by direct injection persisted only 3 days. We found that delivery of tNSCs into the cavity on nanofibrous electrospun poly-l-lactic acid scaffolds extended tNSC persistence to 8 days. Modifications to fiber surface coating, diameter, and morphology of the scaffold failed to significantly extend tNSC persistence in the cavity. In contrast, tNSCs delivered into the post-operative cavity on gelatin matrices (GEMs) persisted 8-fold longer as compared to direct injection. GEMs remained permissive to tumor-tropic homing, as tNSCs migrated off the scaffolds and into invasive tumor foci both in vitro and in vivo. To mirror envisioned human brain tumor trials, we engineered tNSCs to express the prodrug/enzyme thymidine kinase (tNSCs^tk) and transplanted the therapeutic cells in the post-operative cavity of mice bearing resected orthotopic patient-derived GBM xenografts. Following administration of the prodrug ganciclovir, residual tumor volumes in mice receiving GEM/tNSCs were reduced by 10-fold at day 35, and median survival was extended from 31 to 46 days. Taken together, these data begin to define design parameters for effective scaffold/tNSC composites and suggest a new approach to maximizing the efficacy of tNSC therapy in human patient trials.

Graphical Abstract

Pre-clinical and clinical studies have shown that engineered tumoricidal neural stem cells (tNSCs) are a promising treatment strategy for the aggressive brain cancer glioblastoma (GBM). However, stabilizing human tNSCs within the surgical cavity following GBM resection is a significant challenge. In this study, Sheets et al. used a preclinical variant of the clinically utilized NSC line HB1.F3.CD and mouse models of human GBM resection/recurrence to identify a polymeric scaffold capable of maximizing the transplant, persistence, and tumor kill of NSC therapy for post-surgical GBM.

Introduction

Glioblastoma (GBM) is the most common primary brain tumor yet continues to have a poor 5-year survival rate of 5.1%,¹ making it one of the most devastating cancers in years of potential life lost.² In 2005, a landmark phase III clinical trial demonstrated maximal surgical resection, radiation, and concomitant/adjuvant temozolomide chemotherapy extended survival from 12.1 to 14.6 months, leading to the adaptation of today’s standard of care.^3,4 However, infiltrative cancer cells evade surgical resection and survive radio-chemotherapy regimens, ultimately leading to recurrence and death.^5,6 Therapies capable of selectively eradicating these residual GBM cells are urgently needed to prevent tumor recurrence and extend patient survival.

Recently, genetically engineered neural stem cells (NSCs) (for a summary of abbreviated terms for stems cells and scaffolds, see Table 1) have emerged as a promising therapy for GBM. NSCs possess a unique capacity to seek out tumors. By detecting soluble factors emitted by tumor cells, including CXCL12 (SDF-1α), HIF-1α, PDGF, and VEGF among others,7, 8, 9, 10, 11, 12 NSCs are able to migrate selectively into primary tumors and track down invasive GBM foci embedded throughout the brain. NSCs can also be genetically engineered to express therapeutic proteins, creating tumoricidal NSCs (tNSCs). Numerous preclinical studies have shown that tNSCs deliver potent anti-cancer gene products to invasive cancer cells in the brain, significantly reducing human GBM xenograft tumor burden and markedly prolonging animal survival.12, 13, 14 tNSC therapy recently entered first-in-human recurrent glioma patient trials based on the strength of these pre-clinical data.^15,16 In these clinical gene therapy trials, the GBM mass is surgically resected and allogenic, clonal HB1.F3.CD21 v-myc immortalized NSCs expressing prodrug-activating enzymes were injected directly into the surrounding resection cavity wall, followed by prodrug administration to initiate tumor kill. The initial clinical study of a single treatment round demonstrated safety, proof of concept for tumor-localized conversion of prodrug to active chemotherapeutic agent, migration to distant tumor sites, and non-immunogenicity.¹⁵

Table 1.

Abbreviations Used for Stem Cells and Scaffolds in Various Conditions

Phrase/Condition	Abbreviation
Neural stem cells	NSCs
NSCs genetically engineered to express generic tumoricidal agent	tNSCs
tNSCs expressing thymidine kinase but not exposed to prodrug ganciclovir (non-tumoricidal)	NSCstk
tNSCstk exposed to prodrug ganciclovir (tumoricidal)	tNSCstk
NSCs expressing diagnostic mCherry and firefly luciferase (non-tumoricidal)	NSCsmChFl
Mesenchymal stem cells	MSCs
MSCs genetically engineered to express generic tumoricidal agent	tMSCs
Polylactic acid biocompatible electrospun nanofiber scaffold	bENS
NSCs seeded onto bENS	bENS/NSCs
Gelatin matrix	GEM
NSCCs seeded onto GEM	GEM/NSCs
NSCstk seeded onto GEM but not exposed to prodrug ganciclovir (non-tumoricidal)	GEM/NSCstk
tNSCstk seeded onto GEM exposed to prodrug ganciclovir (tumoricidal)	GEM/tNSCstk

Despite the central role of surgery in clinical GBM and tNSC treatment, the mainstay of pre-clinical studies has been solid xenograft models. Our team and others have recently developed mouse models of GBM that include surgical resection, enabling the study of stem cell therapies in the context of the post-surgical brain microenvironment. Interestingly, studies showed that surgical resection has a significant negative impact on the persistence and efficacy of intra-cavity stem cell therapies for GBM. Kauer et al.¹⁷ showed that murine NSCs (mNSCs) were well tolerated in the mouse brain without surgical resection, with 50% of the initial cell dose persisting 4 weeks after injection. However, in mice whose tumors were resected at the time of mNSC infusion into the cavity walls, more than 90% of the stem cell dose was lost within the first week. The rapid loss of the tumoricidal stem cells markedly impaired treatment efficacy, wherein overall survival was statistically unchanged relative to control.¹⁷ Additional studies have shown that the rapid loss of therapeutic cells from the surgical cavity extends to human mesenchymal stem cell (hMSC) therapies as well.^18,19 Whether surgical resection results in the same rapid clearance for clinically relevant human NSCs directly infused in the post-surgical cavity remains unclear, and strategies to improve the transplant and persistence of HB1.F3 cells in the post-operative GBM cavity have not been explored.

Biocompatible polymeric scaffolds have been implanted into the GBM resection cavity of human patients for decades.^20,21 In addition to serving as controlled drug-release platforms, emerging evidence suggests that scaffolds can serve as biophysical matrices to enhance the transplant of cell therapies into the post-surgical tumor cavity. Recently, we discovered that applying biocompatible electrospun nano-scaffolds (bENSs) laden with hMSCs increased both the retention and persistence of cells in the GBM resection cavity 4-fold over standard direct-injection methods. The scaffolds were permissive to tumor-tropic migration, allowing hMSCs to exit the matrix and localize with residual human GBM cells. When scaffolds were used to deliver hMSCs releasing the pro-apoptotic agent TRAIL (tumor necrosis factor [TNF]-related apoptosis-inducing ligand) into the post-surgical cavity, we found that the composite scaffold/hMSC therapy markedly suppressed tumor re-growth and more than doubled median survival. Despite these promising results with hMSC therapy, it remains unclear whether biocompatible scaffolds can enhance clinically relevant hNSC therapies, what matrix composition is required to maximize persistence, and whether scaffold/hNSC composite therapy can effectively slow re-growth of post-surgical GBM.

In these studies, we sought to define the persistence of HB1.F3 NSCs in the post-surgical cavity and test a panel of polymeric scaffolds to identify matrices that improve the persistence of HB1.F3 NSCs within the cavity yet remain permissive to migration and tumor kill. Using our mouse model of GBM surgical resection, we found that the majority of HB1.F3 NSCs were lost shortly after injection into the cavity walls. Surprisingly, scaffolds that effectively supported transplant of hMSCs failed to markedly enhance HB1.F3 NSC persistence in the GBM cavity, and alterations in scaffold architecture and composition failed to significantly improve persistence. However, delivery of HB1.F3 NSCs on gelatin matrices (GEMs) was highly effective at enhancing initial cell numbers in the post-surgical cavity and prolonging persistence. HB1.F3 NSCs were able to migrate from the GEM and significantly inhibited progression of post-surgical GBM. Taken together, these data begin to define the scaffold design parameters required to efficiently transplant tNSCs into the post-operative GBM cavity and suggests this novel GEM/tNSC^tk (tNSC expressing the prodrug/enzyme thymidine kinase) treatment system can potentially significantly improve clinical outcome in patients receiving tNSC-mediated therapies for brain cancer.

Results

Persistence of HB1.F3 NSCs in the Post-Surgical Cavity

In the clinic, HB1.F3 NSCs are infused into the walls of the post-surgical cavity following GBM resection in human patients. To maximize tumor kill, the therapeutic cells must persist at levels high enough to induce killing of residual GBM cells and long enough to seek out local and distant tumor foci. To mirror clinical testing, we first investigated the persistence of HB1.F3 in the surgical cavity. Pre-clinical versions of the HB1.F3 NSC line with lentiviral (LV) constructs encoding for mCherry and firefly luciferase (NSCs^mChFl; Figure S1) enabled observation of their behavior on scaffolds in vitro as well as their persistence in vivo. We created cranial windows in nude mice, and NSCs expressing green fluorescent protein and firefly luciferase (NSCs^GFPFl) were directly infused into the walls of the post-surgical cavity (Figure 1A). Serial bioluminescence imaging (BLI) showed that the levels of NSCs^GFPFl declined rapidly, with 50% of the cells lost by day 1 and 95% lost by day 3 post-injection (Figure 1B).

Figure 1.

bENS-Based Delivery of tNSCs into the Post-Surgical Cavity

(A) Schematic representation showing the delivery of tNSCs in the walls of the post-surgical cavity by direct injection. (B) Summary graph of serial BLI showing the persistence of tNSCs delivered in vivo by direct injection. (C and D) White light (C) and SEM images (D) of bENS. (E) Representative images and summary data showing the proliferation of tNSCs on bENSs or cultured without scaffolds (n = 3). (F and G) Fluorescent images showing nestin⁺ hNSCs at the time of seeding on a bENS (F) and 8 days after seeding (G). (H) Summary graph and summary table of BLI showing the persistence of tNSCs and tMSCs delivered into the post-surgical cavity by direct injection (n = 6) or on a bENS (n = 7). Inset is a summary table showing the time to 50% and 95% clearance of tNSCs delivered by DI or bENSs. Data are mean ± SEM. *p < 0.05 versus control by Student’s t test.

Persistence of bENS/NSCs Delivered into the Post-Surgical Cavity

Building on previous experiments wherein bENSs significantly improved the delivery and persistence of cells in the post-operative brain,¹⁹ we hypothesized that delivering NSCs into the resection cavity on bENSs would improve NSC persistence. Nanometer-diameter bENSs were fabricated by an electrospinning process as previously reported,²² then cut into 2 × 2-mm scaffolds (Figures 1C and 1D). Cells were seeded dropwise onto disinfected scaffolds. In vitro growth rate assays showed that NSCs^mChFl proliferated on bENSs at a similar rate to those on tissue culture dishes (Figure 1E). Immunohistochemical (IHC) staining showed that cells on bENSs continued to express the NSC marker nestin after 1 week, suggesting that the scaffolds and culture conditions did not induce differentiation (Figures 1F and 1G).

We next investigated the impact of bENSs on the persistence of NSCs in the post-surgical cavity. NSCs^mChFl were seeded on bENSs and implanted into a surgical resection cavity in nude mice. Serial BLI showed that bENSs only partially supported NSC persistence with 50% of NSCs lost by day 6 and 95% lost by day 8. In contrast, previous studies reported that 50% loss of other stem cell types on bENSs was not observed until day 20.¹⁹ bENSs therefore provided a modest improvement in NSC persistence compared to direct injection, but was several-fold less efficient than other configurations. These results suggested that the scaffolds could be modified to better suit NSC transplant.

Modifications to bENSs Have Minimal Impact on tNSC Persistence

We next sought to parametrically modify individual scaffold properties to determine their impact on NSC persistence. To determine whether surface adhesion modification could significantly prolong persistence in vivo, collagen (3.6 mg/mL) was adsorbed onto bENS (cbENS) surfaces by incubation at 37°C for 2 h prior to cell seeding (Figure 2A). cbENS/NSC^mChFl composites (Figure 2B) were implanted into the surgical cavity of nude mice and tracked with serial BLI as before. Minimal differences were detected in the persistence of NSCs delivered on bENSs or the surface-coated cbENSs (Figure 2C). Next, meltblown scaffolds (MBSs) were explored to determine whether larger fiber diameter and more porous structures would affect NSC persistence (Figure 2D). NSCs^mChFl were seeded on MBSs containing fibers with a 20.7 ± 1.9-μm diameter and porosity of 96.1% with average pore size of 59.9 ± 50.8 μm (Figure 2E) and transplanted into the post-surgical cavity. Serial BLI again revealed minimal differences between the persistence of NSCs delivered on bENSs or MBSs (Figure 2F). Lastly, we sought to determine the impact of three-dimensional (3D) scaffold morphology on NSC delivery (Figure 2G). NSCs^mChFl were seeded on bENSs (Figure 2H), and three individual bENSs/NSCs^mChFl were transplanted into the resection cavity to create a 3D stack. Although persistence improved at early time points, the cells were eventually lost and overall persistence was not statistically different from NSCs delivered on a single bENS (Figure 2I). It was determined based on this initial improvement that a different scaffold material with a 3D microstructure should be pursued.

Figure 2.

Parametric bENS Optimization for NSC Persistence

bENSs were engineered with different parameters designed to enhance NSC persistence, including surface coating, fiber diameter, and morphology (A, D, and G). NSCs were seeded on the scaffolds (B, E, and H), delivered into the surgical cavity, and NSC persistence was monitored by BLI (C, F, and I). NSCs delivered by DI and on parental bENSs were included as controls. (A–C) Summary images and graphs showing that collagen-coated cbENSs (n = 6) failed to markedly extend NSC persistence compared to DI or bENSs. (D–F) SEM images and summary graphs showing delivery of NSCs on bENSs composed of micro-scale fiber diameters (n = 3) did not increase persistence. (G–I) SEM images and summary graphs showing that changing dimensionality through 3D stacked NSC/bENS constructs lead to minimal extensions in NSC persistence.

Gelatin-Based Scaffolds Enhance NSC Survival and Persistence Post-Implantation

To allow rapid implementation into human trials, we examined NSC performance on a clinically approved resorbable hemostatic GEM material (Figure 3A). Pre-clinical studies have shown GEM to be a viable scaffold material for stem cells in regenerative medicine applications,23, 24, 25 but its utility for tNSC cancer therapy had remained unknown. For diagnostic testing, NSCs^mChFl were seeded onto the scaffolds. Fluorescent imaging (Figure 3B) and high-resolution scanning electron microscopy (SEM) imaging confirmed efficient seeding, with cells extensively covering the matrix and spreading across pore walls (Figure 3C). To explore the growth rates of NSCs on GEMs compared to bENSs, a range of initial seeding densities of NSCs were seeded onto the scaffolds and incubated at 37°C. Scaffolds were removed from incubation on days 1, 3, and 7 and the number of cells on the scaffolds was assessed by BLI intensity. Results showed that independent of the initial seeding densities tested, NSCs proliferated to a statistically similar number of viable cells on day 7 (Figure 3D). Contact inhibition likely dictated overall proliferation on the scaffolds. Cells on GEM, which has a high surface area due to large interconnected 3D pores, grew to an approximately 5-fold greater population than did those on comparably low surface area, thin film-like bENSs. NSCs on GEM maintained their stemness during culture, with IHC staining showing high levels of nestin expression by NSCs on day 2 that were maintained through day 10 of culture on GEM (Figure 3E).

Figure 3.

In Vitro Characterization of NSCs on GEM

(A) Photograph of GEM scaffold. (B) Macroview fluorescent image of NSCs^mChFl growing on a GEM scaffold. (C) SEM images showing GEM porosity and cell attachment on the inner walls of the pores. (D) BLI data showing NSC proliferation over time at different initial seeding densities (n = 3). (E) Confocal images of hNSCs growing on GEM, maintaining stemness as evidenced by nestin⁺ staining both (i) initially and (ii) at 10 days in culture. (F) In vivo persistence was significantly improved on a GEM (n = 8) compared to both direct injection and a bENS.

Having observed adhesion and growth in vitro, GEM/NSC^mChFl composites were next implanted into the resection cavity of nude mice to determine the impact of GEM on NSC persistence in vivo (Figure 3F). Serial BLI revealed that GEM scaffolds markedly improved NSC persistence in contrast to the limited persistence of NSCs delivered by direct injection or on the bENSs (p = 0.0094 by log-rank test when analyzed as BLI signal survival). GEMs increased post-transplant levels of NSCs 8-fold compared to direct injection and extended time to 95% clearance from day 3 to day 19 day. Compared to bENSs, GEMs increased the levels of NSCs in the cavity 7.6-fold at day 5 post-implant and extended time to 95% clearance from day 8 to day 19. Taken together, these results suggest that GEMs can enhance NSC delivery and persistence in the clinical post-surgical cavity as compared to direct injection or polylactic acid (PLA)-based nanofiber scaffolds.

GEMs Are Permissive to Tumor-Homing Migration of tNSCs In Vitro and In Vivo

NSCs must be able to migrate off of the scaffold on which they are implanted in order to deliver therapeutics to, and remain therapeutically effective against, cancer cells that have invaded surrounding brain parenchyma. To test whether NSCs would migrate off GEM scaffolds and retain their tumor-tropic properties, an in vitro assay was developed wherein human GBM tumors expressing green fluorescent protein (GFP) were established in 3D agar culture systems (Figure 4Ai). Agar immediately adjacent to the established GBM tumor was aspirated from the culture dish to simulate a post-surgical resection cavity, and GEM/NSC^mChFl composites were implanted into the cavity. As shown in Figures 4Aii and 4Aiii, we found that NSCs^mChFl migrated off the GEMs, homing to and extensively populating surrounding GBM foci within 48 h. NSC^mChFl exit from the GEMs was confirmed by fluorescent imaging of thin sections collected from the 3D culture system, which showed the NSCs^mChFl populating the tumors at a cellular resolution (Figures 4Aiv and 4Av; Figure S2).

Figure 4.

NSCs Migrate off GEM and toward GBM In Vitro and In Vivo

(Ai) 3D agar culture systems were created to mimic the in vivo resection cavity. U87^GFP GBM cells were implanted into 3D agar matrices. A small resection cavity mimic was created by removing agar adjacent to the U87^GFP implant. GEM/NSCs^mChFl were seeded into the agar cavity. Fluorescent imaging was used to monitor the migration of NSCs^mChFl to the GBM foci. Representative images captured at the time of implant (Aii) and 2 days post-implant (Aiii) showed that the NSCs^mChFl migrated through 3D agar to populate the GBM cells. Fluorescent imaging of thin slices of the 3D agar system demonstrated the change in NSC distribution between day 0 (Aiv) and 2 (Av) at a cellular resolution. (Bvi) This migratory behavior is also seen in an in vivo resection model. Tumors were resected and GEM/NSCs^mChFl were implanted in the cavity. Mice were sacrificed and tissue sections imaged with confocal microscopy on days 7 (Bvii and Bix) and 11 (Bviii and B-x) to determine the extent of cell migration. NSCs are observed to home to both local and distant recurrent foci.

We next explored NSC^mChFl migration from GEMs in vivo. Human GBM tumors were established in nude mice and resected 1 week later. GEM/NSC^mChFl composites were implanted in the resection cavity. At selected time points, mice were sacrificed and post-mortem tissue sections examined for cell migration (Figure 4Bvi). NSCs^mChFl distributed throughout tumor foci in the peri-tumoral space by day 7 and populated residual GBM foci 1.5 mm from the surgical cavity by day 11 post-implant (Figures 4Bvii–4Bx). Taken together, these data suggest that GEM is permissive to tumor-tropic migration of NSCs.

GEM/tNSC Therapy Inhibits Growth of Post-Operative GBM

We next explored the efficacy of GEM/tNSC treatment for post-surgical residual GBM. In the clinical setting, the cell-seeded scaffolds would be implanted at the time of tumor resection to target residual GBM cells that evade surgery. To model this clinical approach, 8 × 10⁵ highly invasive PDX GBM8 tumor cells engineered to express GFP-Fluc (GBM8^GFPFl cells) were implanted into the cerebrum of nude mice (outlined in Figure 5A).²⁶ Seven days later, established tumors were surgically resected and cell-seeded scaffolds were implanted into the resection cavity (Figure 5B). Therapy was initiated in the treatment group (GEM/tNSCs^tk) 3 days later via daily intraperitoneal injections of 100 mg/kg ganciclovir (GCV) and continued for 14 days, while the control group (GEM/NSCs^tk) received saline injections. Despite the diffuse nature of the GBM, serial BLI showed that the GEM/tNSC^tk treatment group had a 5.0-fold reduction in absolute tumor signal at day 7 post-resection and a statistically significant 9.7-fold reduction at day 35 post-resection (p < 0.001 by two-way ANOVA) compared to control (Figures 5C and 5D). The reduction in tumor volume significantly extended survival, with GEM/tNSC^tk-treated mice surviving an average of 46 days after therapy compared to only 31 days in control animals (p = 0.0163 by log-rank test) (Figure 5E). Post-mortem IHC staining showed no detectable toxicity to brain tissue at day 7 or 35 post-GEM/tNSC^tk transplant (Figures 5F and 5G).

Figure 5.

tNSC.TK Therapy Is Effective When Delivered on GEM

(A) Overview of therapeutic efficacy study design. NSCs^tk were seeded on a GEM and implanted into the GBM resection cavity. Mice (n = 5) were injected with 100 mg/kg prodrug ganciclovir (GCV) daily for 2 weeks, with control mice receiving saline. (B) Intraoperative series showing (i) pre-operative anesthetized mouse, (ii) opened skin revealing the green fluorescent GBM8 tumor, (iii) resected tumor and GEM/tNSCs^tk implanted into the resection cavity, and (iv) post-operative mouse with sealed skin wound. (C) Representative serial BLI images of tumor growth in one −GCV mouse and one +GCV mouse over time. (D) Quantitative BLI results of tumor growth over time. (E) Kaplan-Meyer survival curves showing a 15-day (48%) improvement to median survival times when mice are treated with GEM/tNSCs^tk. (F and G) H&E-stained coronal brain sections for early (F, day 7) and late (G, day 35) time points.

GEM scaffolds were examined for changes in the cellular content throughout the course of tumor treatment. At day 10 and 20 after tumor resection and GEM/tNSC^tk implantation, the scaffold was surgically removed and fixed in 10% formalin. One half of the sample was prepared for SEM imaging and the other half for histological examination. SEM imaging revealed an increase in overall cell coverage from day 10 to day 20 (Figures 6A and 6C). Magnified cross-sectional views indicated a larger percentage of cellular content located at the periphery of the scaffold and a comparably unpopulated interior at both time points (Figures 6B and 6D). At the 10-day time point, H&E images show a predominantly intact scaffolding structure with tissue appearing slightly denser at the edges than in the center (Figure 6E). While the edges of the scaffold contain some pockets of tumor cells, there was no evidence of cancerous cells located in the interior (Figure 6F). In contrast, H&E images of the 20-day time point revealed signs of significant scaffold degradation with tumor cells scattered throughout (Figures 6G and 6H). Fluorescence imaging of the 10-day scaffold also showed tNSCs clustered in the center of the GEM with GBM8 tumor cells mostly at the edges (Figures 6I and 6J). The 20-day scaffold was overrun with GBM8 cells while few tNSCs were detected, coinciding with the time point at which BLI imaging shows progression of the tumor in vivo (Figures 6K and 6L). Taken together, these data suggest a transition in cellular content from higher to lower tNSC presence that is correlated with the progression from lower to higher tumor burden.

Figure 6.

GEM Scaffolds Surgically Removed from Treatment Group Mice at Early (10-Day) and Late (20-Day) Time Points after Initial Resection and Scaffold Implantation Reveal Temporal Changes in Cellular Content

(A–D) Representative SEM images showing the overall cell presence on the GEM (n = 1) and degree of cross-sectional infiltration. (E–H) H&E images highlighting the onset of GEM degradation, indicating a shift from sparse to significant tumor cell presence on the matrix. (I–L) Fluorescent confocal images that distinguish NSC and GBM8 cells from native host tissue via the constitutive expression of engineered mCherry and GFP constructs, further suggest the overall transition from NSC to GBM8 cell presence on scaffolds.

Discussion

Advances in pre-clinical GBM models have shown that tumor resection reduces viability, persistence, and efficacy of tNSCs that are directly injected into the resection cavity wall. As long as surgery remains part of the standard of care for GBM treatment, alternative methods for the delivery of stem cells into the brain must be considered. Biocompatible polymeric scaffolds ameliorate the negative impact of the hostile resection microenvironment by providing a physical, protective structure to which cells adhere during transplantation. Scaffolds support the cells as the effects of surgical brain injury wane, at which point cells migrate off the scaffold and into the parenchyma. In the context of resection, scaffolds consistently resulted in larger numbers of stem cells persisting in the brain for longer periods of time. This is in line with other works that have shown that extracellular matrix (ECM), hyaluronic acid, fibrin, and PLA scaffolds improve stem cell therapy post-resection.17, 18, 19^,27

In this work, we examined for the first time the persistence and efficacy of tNSCs in a mouse model of GBM resection and recurrence. We quickly discovered that bENSs would not adequately support tNSC viability and persistence post-implantation. Adjustments to surface coating and fiber diameter did not affect persistence behavior in vivo. The only modification that provided temporary improvement was arranging multiple bENSs into a 3D stacked morphology prior to implant, but ultimately this approach also failed to match improvements observed previously with MSCs.

To ease clinical translation, we turned toward a US Food and Drug Administration (FDA)-approved GEM to study tNSC persistence. Highly porous GEM provides a large surface area-to-volume ratio that maximizes the density of stem cells that can be delivered into the resection cavity. In vitro growth assays showed that for GEM of a fixed size, a common maximum cell number was eventually achieved independent of the number of cells that were initially seeded, and that this maximum cell number was 5-fold larger than for equivalently cut sizes of monolayer bENSs. SEM imaging confirmed that cells attached to pore walls. In addition to allowing fluid and nutrient exchange, these large interconnected pores also allow for cell transit, as evidenced by in vitro and in vivo migration studies as well as analysis of scaffold explants.

Optimal cell delivery requires tNSCs to persist in the post-surgical microenvironment and retain the capacity for tumor-tropic migration. tNSCs delivered to the cavity on GEM remain viable on the scaffold immediately after implantation, which allows for a precise number of cells to be delivered. Once in the cavity, GEM increased the persistence of the tNSCs 8-fold over direct injection. Cells maintained stemness, as indicated by nestin expression, and retained their tumor-homing ability. As a result of the prolonged persistence, tNSC^tk therapy was more effective and median survival of mice was significantly increased from 31 to 46 days.

The cellular content of the scaffolds changes throughout the course of tNSC^tk therapy. Treatment is initially effective, as the number of transplanted tNSCs^tk on scaffolds is quite high. Over time, the tNSC^tk population on the scaffolds decreases, likely due to a combination of migration off the scaffold, immune clearance, and/or suicide from their conversion of prodrug GCV to active GCV triphosphate (GCV-TP). The model presented introduces human tNSCs into immunodeficient mice. The murine immune system constantly clears human cells despite the aid of the GEM. An allogenic transfer of tNSCs into a human will reduce immune-based clearance of these cells and potentially result in more robust GBM cell killing than seen in this model. As they convert prodrug to active drug, a subset of susceptible tNSCs^tk likely takes up GCV-TP and becomes damaged to the point of suicide. Prodrug-triggered suicide ensures eradication of the transplanted cells and eases concerns of secondary tumor formation. Clinical data from autopsy tissue in the initial clinical trial with this NSC line showed no evidence of secondary tumorigenicity. Regardless, when tNSCs are no longer present, treatment benefits cease and evasive GBM cells begin to accumulate on the scaffold on day 20. Despite this, we show that a single dose of GEM/tNSCs^tk extended overall survival in the highly aggressive orthotopic PDX GBM8 model by more than 2 weeks compared to control animals. It is evident that GEM maintains tNSC viability and persistence, significantly increasing their therapeutic efficacy as compared to direct injection into post-operative brain parenchyma. These findings support adaptation of our GEM/tNSC approach in the clinical setting to improve clinical outcomes.

Conclusions

The promise of tNSC therapy to treat GBM relies on the ability of transplanted cells to remain viable, persist over time, and migrate to invasive tumor foci in the post-surgical cavity. Unfortunately, the harsh post-surgical microenvironment does not optimally support direct injection of tNSCs. This study demonstrates that seeding tNSCs on GEM prior to transplantation significantly supports cell viability, persistence, and efficacy, as demonstrated by improved survival outcomes in mice. GEM scaffold delivery allows the therapeutic cells to evade physical washout from implantation, persist in the immunologically active post-surgical environment, and maintain stemness and tumor-homing capacity. In addition to cellular support, the porous, 3D GEM structure allows for fluid exchange to minimize mass-effect side effects and provides a large surface area to maximize the number of cells delivered. The delivery conditions presented in this study provide the foundational design parameters necessary to effectively deliver tNSCs to the post-surgical GBM resection cavity and may improve the method of tNSC-mediated therapies in current and future human patient trials.

Materials and Methods

HB1.F3.CD Cells

The v-myc immortalized human NSC line HB1.F3 originated from primary fetal telencephalon cultures at 15 weeks of gestation.^28,29 Upon isolation and expansion, the HB1.F3 subclone was transduced with the retroviral vector pMSCV-puro/CD, and clones were then isolated and expanded. HB1.F3.CD clone 21 was then given to the University of North Carolina (UNC) at Chapel Hill under a material transfer agreement.

Cell Culture

Adherent HB1.F3.CD tNSCs were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and 1% glutamine as described previously.^30,31 U87 glioma cells were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in DMEM media supplemented with 10% FBS and 1% penicillin/streptomycin. Clinically derived GBM8 human GBM cells were kindly gifted from Dr. Hiroaki Wakimoto (Massachusetts General Hospital, Boston, MA, USA) and cultured as non-adherent neurospheres in complete EF media as previously described.²⁶

LV Vectors

Four previously developed LV constructs were used in this study to label cancer and stem cell populations with fluorescent and luminescent reporters for in vitro and in vivo tracking. The constructs encode for either GFP or mCherry (mCh) along with either firefly luciferase (FL) or Renilla luciferase (RLuc). LV-GFP-FL, LV-GFP-RLuc, LV-mCh-FL,³² and LV-mC-RLuc-TK (encodes for thymidine kinase) were each packaged as LV vectors in 293T/17 cells using a helper virus-free packaging system as previously described.³³ Unlabeled target cells were transduced at 80% confluency in a six-well plate with 0.8 μg/mL Polybrene (Sigma) and varying multiplicities of infection (MOIs) overnight. Infection efficiency was determined optically, and low-expressing populations were puromycin selected.

In Vitro Tumor Killing Assay

Tumor cells were seeded at a density of 5,000 cells per well in a 96-well plate. tNSCs^tk were seeded on scaffolds that were then transferred into the tumor-bearing wells in a range of ratios (0:1 to 4:1 tNSCs to tumor cells, 0–20,000 cells per well). After a 6-h equilibration period, media in all wells were exchanged for fresh media, with treatment groups receiving 20 μg/mL GCV. After 72 h of co-culture, media in all wells were aspirated and replaced with 0.75 mg/mL d-luciferin (PerkinElmer) in PBS, which reacted only with viable tumor cells. The extent of killing was determined by light emitted from these cells via quantitative BLI using an in vivo imaging system (IVIS) Kinetic.

In Vitro Migration

A 0.6% agar solution was created by dissolving 240 mg of agar in 10 mL of PBS in an Erlenmeyer flask using a microwave. The dissolved agar solution was then poured into 30 mL of warm DMEM containing 2% FBS. Agar gels were then created by pouring 2 mL of the solution into each well of a six-well plate. PLA microfibers were interspersed into the agar as it cooled to provide white matter-like migratory paths throughout the gels. Tumors were established by injecting 1 × 10⁶ U87^mChFl cells in 3 μL of DMEM via a 10-μL capacity Hamilton syringe. An additional 1 mL of DMEM medium supplemented with 2% FBS and 1% penicillin/streptomycin was added on top of each gel to prevent dehydration. After overnight incubation at 37°C, a resection cavity was created by aspirating a portion of the gel bordering the newly established tumor. GEM/NSCs were tacked in the cavity and incubated at 37°C for 2, 4, 6, and 8 days. Samples were fixed in 10% formalin, embedded in OCT blocks, frozen, and cut into 60-μm-thick sections using a cryostat. Fluorescence images of the NSC migration toward the tumor were captured using an Olympus MVX10 fluorescence dissecting stereomicroscope.

Antibody Staining

In vitro samples were fixed in 10% formalin at 4°C for 1 h followed by a PBS rinse. Samples were then placed in 200-μL centrifuge tubes and blocked in blocking buffer (0.5% BSA and 0.25% Triton X-100 in PBS) for 30 min. Primary rabbit anti-human nestin antibodies (MilliporeSigma, ABD69) were diluted 1:400 in blocking buffer and incubated for 1 h at room temperature on a shaker. After three PBS rinses, goat anti-rabbit Alexa Fluor 488 secondary antibodies (Thermo Fisher Scientific, A-11008) were diluted 1:800 in blocking buffer and incubated for 1 h at room temperature on a shaker and protected from light. After three PBS rinses, Hoechst was added to the samples and incubated for 5 min on a shaker protected from light. Samples were then washed in PBS and imaged.

Scaffold Preparation and Cell Seeding

Two days before implantation, bENSs and GEM scaffolds were cut into resection cavity-sized (approximately 2 × 2 mm) pieces and disinfected by immersion in 70% ethanol for 15 min. Scaffolds were then washed in PBS and placed in DMEM containing 10% FBS and 1% penicillin-streptomycin while preparing cells for seeding. NSCs were lifted at approximately 70% confluency using 0.05% trypsin-EDTA incubation at 37°C for 5 min. After resuspension in DMEM, cells were counted using a hemocytometer and pelleted via centrifugation at 100 × g for 5 min. Supernatant was aspirated and NSCs were re-suspended in DMEM to a final concentration of 2 × 10⁵ cells/μL. Scaffolds were pat-dried by repeated placement on a sterile lid of a 10-cm plate and then placed in a six-well plate once dried. Scaffolds were seeded by pipetting 2.5 μL of freshly mixed cell suspension directly onto scaffolds and incubating at 37°C for 30 min. Scaffolds were then flipped over in the well plate, and another 2.5 μL of freshly mixed NSC suspension was added to the other side (this resulted in a total of 1 × 10⁶ cells seeded per scaffold). After another 30-min incubation at 37°C for 30 min, scaffolds were fully covered in DMEM and gently lifted to allow media to flow underneath them. Seeded NSCs were incubated at 37°C in this state for 2 days prior to implantation surgery.

In Vivo Persistence Surgery

All methods described were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of North Carolina at Chapel Hill (protocol no. 15-337.0). Stereotaxic resection and scaffold implantation were performed aseptically on 20–30 g of athymic nude mice (Jackson Laboratory) as previously described.^18,19 Persistence studies were conducted in non-tumor-bearing mice to prevent morbidity-related complications. Briefly, a linear 3- to 5-mm rostral-caudal incision was made in the midline of the scalp. A craniotomy was performed over the right hemisphere between the bregma and lambda points to provide access to the brain. The skin incision was sealed with Vetbond glue. Post-operative mice were placed in a heated chamber during recovery from inhaled anesthesia, returned to the cage once ambulatory, and observed for complications during recovery. Three days later, a second surgery was performed to create a sham resection cavity. Using a vacuum pump, the surface of the exposed brain was aspirated until a hemispherical cavity approximately 2 mm in diameter and depth was created. Bleeding was controlled with copious PBS irrigation and application of Surgicel when necessary. Cell-seeded scaffolds were then implanted in the sham resection cavity and secured in place with 1 μL of fibrinogen and thrombin extracted from TISSEEL (Baxter). The skin incision was re-sealed with Vetbond glue, and mice were observed during recovery as before. There was no surgery-induced morbidity observed.

In Vivo Tumor Killing Efficacy Surgery

Stereotaxic tumor establishment, resection, and scaffold implantation surgeries were performed aseptically on 20–30 g of athymic nude mice (Jackson Laboratory) as previously described.^18,19. Scalp incision and craniotomy were performed as described above. To establish orthotopic tumors, human U87^GFPFl cells (1 × 10⁵ cells in 3 μL of serum-free media) or patient-derived xenograft GBM8^GFPFl tumor cells (8 × 10⁵ cells in 3 μL of serum-free medium) were injected at stereotaxic coordinates 2.5, −0.5, −0.5 from bregma using a 10-μL Hamilton syringe. Cells were infused at 1 μL per min with a motorized integrated stereotaxic injector (Stoelting Company), and then, following a 5-min pause, the syringe was retracted at 0.5 mm per min. After 7 days of tumor growth, a second surgery was performed in which the primary tumor was resected via gentle aspiration using a vacuum pump. Tumors were resected under the guidance of a fluorescence dissecting stereomicroscope. With room lights dimmed, tumor cells were identified by their fluorescent signal and aspirated until no fluorescent signal was seen by eye. Bleeding was controlled with copious PBS irrigation and application of Surgicel when necessary. Cell-seeded scaffolds were then implanted in the resection cavity and secured in place with fibrinogen and thrombin. The skin incision was re-sealed with Vetbond glue. Post-operative mice were placed in a heated chamber during recovery from inhaled anesthesia, returned to the cage once ambulatory, and observed for complications during recovery. There was no surgery-induced morbidity observed.

In Vivo Imaging and Survival

Tumor progression was tracked weekly with IVIS Kinetic BLI. Mice were injected with 150 mg/kg luciferin intraperitoneally (i.p.) and anesthetized in the imaging chamber. Images were captured after allowing 10 min for the luciferin to react with luciferase-expressing tumor cells and reach peak emission. Exposure times were adjusted as needed to prevent detector saturation. Radiance was analyzed in Living Image software by drawing regions of interest (ROIs) over the signal emitting from the brain. Mice were routinely monitored during tumor progression and were sacrificed prior to the development of significant neurological symptoms in accordance with the IACUC protocol. Kaplan-Meier survival curves were generated based on the date of sacrifice.

Tissue Processing

Tissue samples were collected and stored in 10% formalin for 24 h and then cryoprotected in 30% sucrose. Samples were then frozen in OCT blocks and stored at −80°C. A Leica CM 1850 clinical cryostat was used to cut 40-μm-thick tissue sections, which were collected on glass slides. Progressive H&E staining was performed using Meyer’s hematoxylin and ethanol-based eosin with xylene-based mounting medium.

SEM Imaging

SEM images were acquired using the Hitachi S-4700. Samples were first prepared by formalin fixation followed by gradient ethanol dehydration (50%, 70%, 80%, 95%, and 100%, 5 min each). Samples were then critical point dried and sputter coated in 6-nm-thick gold-palladium alloy to prevent surface charging during imaging.

Statistical Analysis

Data are expressed as mean ± standard error, and differences are considered significant at p <0.05 by tests denoted in legends and marked graphically with asterisks. Error bars represent standard error.

Data Availability Statement

The raw/processed data required to reproduce these findings cannot be shared at this time due to legal or ethical reasons.

Conflicts of Interest

K.T.S., M.G.E., and S.D.H. have ownership interest and S.D.H. is CSO of Falcon Therapeutics. K.S.A. has ownership interest and is CSO of TheraBiologics, Inc. The remaining authors declare no competing interests.