Utilizing induced neural stem cell‐based delivery of a cytokine cocktail to enhance chimeric antigen receptor‐modified T‐cell therapy for brain cancer

Abstract Chimeric antigen receptor (CAR)‐modified T‐cell therapy has shown enormous clinical promise against blood cancers, yet efficacy against solid tumors remains a challenge. Here, we investigated the potential of a new combination cell therapy, where tumor‐homing induced neural stem cells (iNSCs) are used to enhance CAR‐T‐cell therapy and achieve efficacious suppression of brain tumors. Using in vitro and in vivo migration assays, we found iNSC‐secreted RANTES/IL‐15 increased CAR‐T‐cell migration sixfold and expansion threefold, resulting in greater antitumor activity in a glioblastoma (GBM) tumor model. Furthermore, multimodal imaging showed iNSC delivery of RANTES/IL‐15 in combination with intravenous administration of CAR‐T cells reduced established orthotopic GBM xenografts 2538‐fold within the first week, followed by durable tumor remission through 60 days post‐treatment. By contrast, CAR‐T‐cell therapy alone only partially controlled tumor growth, with a median survival of only 19 days. Together, these studies demonstrate the potential of combined cell therapy platforms to improve the efficacy of CAR‐T‐cell therapy for brain tumors.


| INTRODUCTION
Cell therapies are reshaping the landscape of cancer treatment strategies.Chimeric antigen receptor (CAR)-modified T cells have demonstrated robust efficacy and durable responses in patients with hematological malignancies.Strategies that translate these therapeutic benefits to patients with solid tumors is under intense investigation.A well-studied approach to enhance CAR-T therapy involves co-administration of supporting cytokines.However, immunomodulators are known to elicit toxicities when administered systemically, 1,2 emphasizing the need for targeted delivery.Direct engineering of T cells has been explored, but variability in production levels remain, and the short half-life of directly conjugated agents may limit the therapeutic window.
4][5][6][7] NSCs have a unique tumor-homing property which enables them to seek out solid and invasive tumor foci. 7When engineered to express specific gene products, NSCs provide long-term drug delivery to malignant tissue resulting in marked modulation of tumor volumes or the tumor microenvironment across a variety of preclinical models. 8,9NSCs have been engineered to release different cytokines, including interleukin (IL)-4 (IL-4), 10 IL-12, 11 and IL-23. 12These studies have shown that efficient NSC-based cytokine delivery induced regression of the aggressive brain cancer glioblastoma (GBM).However, tumors eventually escaped and mice succumbed to recurring tumors.As such, durability of antitumor response remains a key issue.
Here, we developed a novel strategy in which we leveraged tumoritropic properties of engineered induced neural stem cells (iNSC) to enhance tumor trafficking and persistence of CAR-T cells.
To this end, we used novel iNSCs transdifferentiated from human skin, a clinically relevant murine model of GBM, and clinically applicable CAR-T cells.Building on findings from previous virotherapies, 13 we engineered iNSCs to express the chemoattractant RANTES (Regulated upon Activation, Normal T cell Expressed and Secreted; also known as CCL5) and IL-15, which promotes T-cell proliferation and persistence. 14,15We found that RANTES/IL-15-secreting iNSCs enhanced CAR-T-cell migration in co-culture systems and their combination eliminated tumors in GBM xenograft models.

| Investigating the impact of cell-based RANTES/IL-15 delivery on CAR-T-cell therapy in GBM
To begin investigating cell-based delivery of RANTES/IL-15 to enhance CAR-T-cell migration and proliferation, we first transduced human U87 GBM cells with a replication-incompetent retroviral vector (RV) encoding human RANTES and IL-15 (U87 RANTES-IL-15 ).Enzyme-linked immunosorbent assay (ELISA) analysis of conditioned media confirmed cytokine expression, with U87 RANTES-IL-15 cells producing 7721 pg/mL of RANTES and 150 pg/mL of IL-15 over 72 h (Figure 1a).We then performed cell migration and proliferation assays using CAR-T cells targeting chondroitin sulfate proteoglycan 4 antigen (CSPG4-CAR-T) that we previously identified as a clinically relevant target in GBM. 16T-cell expression of CSPG4-CAR was confirmed via flow cytometry (Figure S1).Using a transwell assay system, U87 cells were seeded and cultured overnight in growth medium.The next day, T cells were seeded in transwell inserts, allowed to migrate for 5 h, and enumerated via flow cytometry.We found that U87 RANTES-IL-15 increased the migration of control T cells 1.9% at 24 h or 6.6% at 48 h, compared to migration toward control U87 cells (Figure 1b,c).This effect was more pronounced in CSPG4-CAR-T cells, with migration increasing 8.8% at 24 h or 9.6% at 48 h.
To investigate the impact of RANTES/IL-15 on T-cell migration in vivo, we implanted U87 RANTES-IL-15 or control U87 cells into brains of immunodeficient NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice, which display severe defects in innate and adaptive immunity.Five days later, CSPG4-CAR-T cells were infused intravenously (Figure 1d).One, 4, or 7 days after CSPG4-CAR-T-cell infusion, brains were harvested, bisected along the interhemispheric fissure, and separated into "tumor" and "contralateral" (nontumor) hemispheres.We then utilized flow cytometry to quantify number of T cells present within each hemisphere.Using this approach, we found that RANTES/IL-15 secretion significantly increased the amount of T cells present within tumors, with T-cell accumulation threefold higher in U87 RANTES-IL-15 compared to control U87 cells on Day 4 (314 cells/μL, U87 RANTES-IL- 15 ; 113 cells/μL, U87) and over fourfold higher on Day 7 (133 cells/ μL, U87 RANTES-IL-15 ; 39 cells/μL, U87; Figure 1e).In contrast, we detected only slight differences in T-cell accumulation between the contralateral hemispheres on Day 7 (16 cells/μL, U87 RANTES-IL-15 ; 5 cells/μL, U87).Taken together, these data demonstrate a proof of concept that RANTES/IL-15 enhance T-cell migration in vitro and in vivo.In addition, T-cell migration into the brain is selective toward the tumor hemisphere with few cells observed in the contralateral hemisphere.

| Generation and characterization of iNSCs expressing RANTES/IL-15
Engineered iNSCs possess great potential for cell therapy applications due to their innate tumor-homing and ability to provide sustained delivery of biologic agents.To investigate whether these unique properties of iNSCs can enhance anti-tumor properties of CAR-T cells, we first constructed a lentiviral vector (LV) encoding human RANTES and IL-15 (Figure 2a).In this construct, RANTES and IL-15 were cloned under control of the EF-1 alpha promoter and separated by a 2A-peptide sequence.An IRES element driving green fluorescent protein (GFP) expression was also included to quantify transduction efficiency.Using our well-established methodology for converting human fibroblasts into second-generation iNSCs, human-induced neurospheres (hiNeuroS), 17 we engineered RANTES/IL-15-secreting hiNeuroS (hiNeuroS RANTES-IL-15 ) (Figure 2b).Robust expression of the GFP reporter gene was detected by fluorescence microscopy (Figure 2c).ELISA analysis of conditioned media confirmed RANTES/ IL-15 secretion, with hiNeuroS RANTES-IL-15 producing 12,777 pg/mL of RANTES and 1377 pg/mL of IL-15 over 72 h (Figure 2d).To explore the ability of hiNeuroS RANTES-IL-15 to enhance the anti-tumor cytotoxicity of CAR-T therapy, FLuc + U87 cells were cocultured with increasing amounts of CSPG4-CAR-T cells in the presence of NS-CM or control media for 72 h.Luciferase assays showed viability of U87 cells in control media was reduced by 96%, 91%, and 77% at CAR-T: tumor ratios of 1:1, 1:2, and 1:5, respectively (Figure 2g).We detected a further increase in U87 tumor-killing in the presence of NS-CM, with viability reduced by 91% and 85% at CAR-T:tumor ratios of 1:2 and 1:5, respectively.Taken together, these results suggest that hiNeuroS RANTES-IL-15 can enhance migration, expansion, and tumor-killing of CAR-T cells in vitro.NSCs are an emerging new platform for drug delivery in cancer therapy.These cells possess the natural ability to home toward malignant tumors.Furthermore, NSCs can be engineered to express a variety of cytotoxic agents.Although NSCs have been successfully utilized in a wide-range of preclinical and clinical cancer therapy applications, 8,9 their potential for co-opting the immune system has yet to be explored.In this study, we shift away from traditional NSC delivery strategies that rely on secretion of toxic antitumor agents with deleterious off-target effects in favor of naturally occurring soluble factors that enhance the functionality of immune cells.

RANTES and IL-15 were previously shown to enhance CAR-T-cell
8 However, we did not explore the long-term persistence of CAR-T cells, which have been detected intracranially for up to 159 days in CNS lymphoma models. 19Since CAR-T-cell longevity is closely tied to the presence of co-stimulatory domains, we expect ours could persist for up to 3 months based on the presence of CD28 in our CSPG4-CAR construct. 20In the presence of IL-15, their longevity may be extended even further as previously evidenced. 21Although we did not directly measure in vivo cytokine levels in tumor tissue, we believe these would be comparable to those measured in vitro due to similar cell numbers and incubation times used in these experiments.
Next, we generated a therapeutic iNSC platform, hiNeuroS, for the delivery of RANTES/IL-15 (Figure 4).Reprogrammed iNSCs offer a distinct clinical advantage over endogenous "off-the-shelf" NSCs.They are manufactured from a patient's own cells, thereby reducing the risk for immunogenicity and increasing persistence. 8,9Conversion of somatic cells into iNSCs is accomplished via direct and indirect strategies.The former, used to generate hiNeuroS, introduces lineage-specific transcription factors, while the latter relies on classic Yamanaka factors. 22Indirect transdifferentiation has been associated with the formation of teratomas, a type of germ cell tumor, due to passage through a pluripotent intermediate stage. 23Direct transdifferentiation, used here, bypasses this intermediate phase, making it a safer cellular conversion strategy.
In addition to the aforementioned safety benefits, hiNeuroS also enjoy a variety of functional enhancements.Previously, we demonstrated that hiNeuroS possess unique transcriptional profiles compared to parent fibroblasts and first-generation iNSCs, including upregulation of genes related to tumor-homing migration, stemness, and proliferation. 17In vivo, hiNeuroS proliferated for up to 1 week and showed a twofold increase in transhemispheric migration to tumor foci and a threefold increase in persistence compared to traditional iNSCs following intracerebroventricular infusion.These properties make hiNeuroS an ideal vehicle for sustained and targeted cytokine delivery.Here, we used a non-invasive GBM cell line to assess efficacy of our combination cell therapy.However, it would be interesting to fully explore the therapeutic potential of hiNeuroS against an invasive GBM tumor model in future studies.
A key benefit of using hiNeuroS RANTES-IL-15 is illustrated in Figure 2g: RANTES/IL-15 increases tumor killing at low CAR-T:tumor ratios.Since 2017, the US Food and Drug Administration (FDA) has approved five CAR-T therapies for the treatment of cancers.The acquisition cost for a single dose of these therapies ranges from $373,000 to $475,000, 24 placing a substantial financial burden on patients and payers alike.Reducing the number of CAR-T cells required per dose via supplementation with immune-modulating iNSCs could significantly reduce these costs.Since there are no FDAapproved NSC therapies on the market, it is difficult to predict the acquisition cost of a clinical grade product.However, a recent study estimated the manufacturing costs for commercial scale expansion of stem cells would be up to 1000-fold less than CAR-T therapy. 25 the in vivo efficacy study, we co-implanted tumor and iNSCs into the brain prior to administering CAR-T cells intravenously.We chose this design to simplify the technical aspects of the study and demonstrate proof-of-concept that a stem cell-delivered cytokine cocktail enhances CAR-T penetration and efficacy.In this approach, we saw a clear and substantial benefit from iNSC-based delivery of RANTES/IL-15.However, this study failed to address how the combination cell therapy interacts with a pre-established brain tumor as observed in clinical application.Now that we have established preclinical proof-of-concept, we are exploring more complicated studies that mirror different aspects of clinical treatment, including resection/recurrence models, additional supportive cytokines, and different CAR-T variants.
Another limitation of this study is that we used an immune deficient mouse model to assess in vivo efficacy.Although this allowed us to use clinically relevant human GBM, iNSC, and CAR-T cells, it also limited our ability to address potential interactions of RANTES/IL-15 with innate immunity.Previously, IL-15 was shown to induce production of pro-inflammatory mediators, tumor necrosis factor alpha (TNF-α) and IL-6, in microglia, 26 while RANTES-induced astrocytes to release monocyte chemoattractant protein-1 and TNF-α. 27Since both are implicated in the amplification of inflammatory responses in the CNS, additional studies are necessary to assess the potential for iNSC-induced neurotoxicity in immune competent animals.
Finally, we cannot exclude the possibility that other factors released by iNSCs may contribute to the beneficial immunomodulatory effects observed in the present study.We previously mapped the transcriptomic profile of hiNeuroS to investigate differences in gene expression associated with stemness, tumor-homing, and proliferation. 17However, additional studies are necessary to explore genes which may be implicated in CAR-T therapy.

| CONCLUSIONS
In summary, this study demonstrates the first ever use of NSCs as immune modulators.We showed that our next-generation iNSCs T cells were stained with an anti-human CD3 fluorochromeconjugated monoclonal antibody (Biolegend).Dead cells were excluded using a nucleic acid stain, Helix NP NIR (Biolegend).Fluorescence-activated cell sorting (FACS) data were collected using an LSRFortessa (BD Biosciences) flow cytometer and analyzed with FlowJo (BD Biosciences) software.

| Bioluminescent imaging
Serial bioluminescent imaging (BLI) was used to assess tumor growth in vivo as previously described. 29Briefly, mice were given D-luciferin (3 mg/mouse in 200 μL phosphate buffered saline (PBS)) by intraperitoneal injection.Fifteen minutes later, photon emission was measured using an IVIS Spectrum (PerkinElmer) imaging system.For in vivo studies, tumor luminescence was quantified using Living Image (PerkinElmer) software and normalized based on the earliest recorded tumor signal.AUC was calculated from normalized luminescence curves using last observation carried forward for deceased animals.
For in vitro studies, D-luciferin (1:100) was added directly to tissue culture plates and luminescence was measured 10 min later using a Gen5 (Agilent) microplate reader.

| Tissue processing
Following anesthetization with 5% isoflurane, mice were perfused via intracardiac puncture with 10 mL PBS and 10% formalin.Brains were harvested, incubated at room temperature in 10% formalin for 72 h, and transferred to 70% ethanol.Mouse brains were trimmed in a coronal orientation through the injection site, then processed and embedded as fixed formalin paraffin blocks.Sections were obtained at 5 μm and mounted onto positively charged slides.

| Immunohistochemistry
Chromogenic IHC was performed on paraffin-embedded brain samples.IHC was carried out using the Bond-III Autostainer (Leica Biosystems).Slides were dewaxed in Bond Dewax Solution (Leica Biosystems) and hydrated in Bond Wash Solution (Leica Biosystems).Heat-induced antigen retrieval was performed for 20 min at 100 C using Bond-Epitope Retrieval Solution 1 pH 6.0 (Leica Biosystems).This was followed by a 5-min peroxide blocking step using Bond Polymer Refine Detection (Leica Biosystems).After pretreatment, slides were incubated for 1 h with 1:500 anti-human CD3 antibody (ab5690, Abcam).Antibody detection with 3,3 0 -diaminobenzidine was accomplished using the Bond Intense R Detection System (Leica Biosystems) supplemented with Novolink Polymer (Leica Biosystems) secondaries.Stained slides were dehydrated and coverslipped with Cytoseal 60 (Thermo Fisher Scientific).Positive controls were included for each assay.IHC-stained slides were digitally imaged in the Aperio AT2 (Leica Biosystems) using a Â20 objective.

| Co-culture experiments
Tumor cells were seeded in tissue culture plates (1e4 cells/well in 96-well plates for cytotoxicity assay and 1e5 cells/well in 24-well plates for proliferation assay) and cultured overnight.CSPG4-CAR-T cells (1e4, 5e3, or 2e3 cells/well for cytotoxicity assay and 5e4 cells/ well for proliferation assay) were suspended in complete T-cell medium or hiNeuroS RANTES-IL-15 conditioned media (supernatant from 2.5e5 cells/mL after 72 h); cell suspensions were added and cultured for an additional 3 days.Residual FLuc + U87 cells were assessed via BLI, while residual T cells were counted based on CD3 expression.

| GBM xenograft model
To assess antitumor effects and infiltration of CSPG4-CAR-T, we used 23 immunocompromised female NSG mice (Jackson Laboratory).Stereotaxic injections were performed in the right hemisphere of each mouse brain to engraft 1e5 FLuc + U87 or 2.5e5 GBM8 tumor cells mixed with or without 5e5 hiNeuroS RANTES-IL-15 intracranially (AP = 0.6 mm, ML = 2.0 mm, DV = 3.0 mm).Mice were intravenously infused with 3e6 CSPG4-CAR-T or PBS 4 days later.Tumor growth was assessed 3 days per week via BLI.

| Statistical analysis
All animal experiment methodology followed the ARRIVE guidelines.
Minimization randomization strategy was applied to allocate animals to treatment groups based on bioluminescence values to ensure balance among groups. 34,35Data were analyzed using Prism (GraphPad) software.Student's t-test was used when comparing only two groups.
One-way analysis of variance (ANOVA) was used when comparing multiple groups and Kruskal-Wallis test was used for longitudinal studies.Statistically significant ANOVAs and Kruskal-Wallace tests were followed by Bonferroni's and Dunn's multiple comparisons test, respectively.Survival analysis was performed using the Mantel-Cox

F
We used a transwell assay to investigate the effect of hiNeuroS-RANTES-IL-15 on CAR-T-cell migration.U87 cells were seeded and cultured overnight in hiNeuroS RANTES-IL-15 conditioned media (NS-CM) or control media.The next day, CSPG4-CAR-T cells were seeded in transwell inserts and allowed to migrate for 5 h.Flow cytometry was used to enumerate T cells, leading to sixfold greater CAR-T-cell migration in NS-CM compared to control media (927 cells, U87 + NS-CM; 162 cells, U87).To investigate the effect of iNSC-delivered RANTES/ IL-15 on T-cell proliferation, CSPG4-CAR-T cells were cultured in NS-I G U R E 1 RANTES/IL-15-secreting GBM tumor cells enhance T-cell trafficking and expansion.(a) RANTES and IL-15 were detected in culture supernatant collected from transduced U87 tumor cells.(b) U87 or U87 RANTES-IL-15 were seeded and cultured for 24 or 48 h.Transwell inserts were added and CSPG4-CAR-T or control T cells were seeded therein.T cells were allowed to migrate through inserts for 5 h.Afterward, inserts were removed and media was collected for flow cytometry.(c) Migration of CSPG4-CAR-T or control T cells toward U87 or U87 RANTES-IL-15 cultured for 24 or 48 h was evaluated by flow cytometry (n = 3).(d) NSG mice were implanted i.c. with U87, U87 RANTES-IL-15 , or PBS in the right hemisphere of the brain.Five days later, mice were infused i.v. with CSPG4-CAR-T.Mice were sacrificed 1, 4, or 7 days following CSPG4-CAR-T infusion.Brains were harvested, bisected along the interhemispheric fissure, and processed for flow cytometry (n = 3).(e) CSPG4-CAR-T concentrations in single cell suspensions generated from bisected mouse brains (contralateral hemisphere vs tumor hemisphere) were analyzed by flow cytometry.Schematic(s) created with BioRender.com.*p < 0.05, **p < 0.01, ***p < 0.001.CAR-T, chimeric antigen receptor-modified T cell; CSPG4-CAR-T, CAR-T cells targeting chondroitin sulfate proteoglycan 4 antigen; GBM, glioblastoma; IL-15, interuekin 15; NSG, NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ.
F I G U R E 3 Combination hiNeuroS RANTES-IL-15 + CAR-T therapy increases CAR-T infiltration and eradicates non-invasive GBM.(a) Schematic representation of study design.NSG mice were implanted i.c. with FLuc + U87 ± hiNeuroS RANTES-IL-15 in the right hemisphere of the brain.Four days later, mice were infused i.v. with CSPG4-CAR-T or PBS (n = 5-7).(b) Representative bioluminescent images depict U87 tumor growth.(c) Luminescence was quantified by calculating total flux within a region of interest centered on the head, then normalizing to Day 0 values.Dotted line signifies CAR-T infusion.(d) Kaplan-Meier survival curves.(e) AUC was calculated for each normalized tumor growth curve.(f) CAR-T tumor infiltration (brown) was assessed by immunohistochemistry. Schematic(s) created with BioRender.com.***p < 0.001.CAR-T, chimeric antigen receptor-modified T cell; CSPG4-CAR-T, CAR-T cells targeting chondroitin sulfate proteoglycan 4 antigen; GBM, glioblastoma; NSG, NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ.CM or control media for 72 h.Flow cytometry analysis revealed threefold greater CAR-T expansion in NS-CM compared to control media (16,219 cells, U87 + NS-CM; 5411 cells, U87; Figure 2e,f).

2. 3 |
hiNeuroS RANTES-IL-15 enhances CAR-T-cell therapy for GBMWe next investigated the impact of hiNeuroS RANTES-IL-15 on CAR-Tcell therapy in vivo, using an established human GBM model.FLuc + U87 cells were implanted intracranially with or without hiNeuroS-RANTES-IL-15 in NSG mice (Figure3a).Serial BLI was then used to monitor tumor growth.Four days later, CSPG4-CAR-T cells were infused intravenously while a group of mice remained untreated.We found that CAR-T cells alone induced a slight inhibition in tumor growth at early time points, reducing GBM volumes 82-fold by Day 10 (Figure 3b,c).However, the tumors rapidly escaped CAR-T therapy, increasing in volume 10-fold by Day 20.The animals succumbed to the large tumor volumes, showing a median survival of just 19 days (Figure 3d) and threefold reduction in area under the curve (AUC) (Figure 3e).In contrast, CAR-T cells infused in hiNeuroS RANTES-IL-15treated mice markedly inhibited tumor growth, reducing tumor volumes 2538-fold by Day 10 with most tumors decreasing below the level of detection.The dramatic reduction in tumor volumes led to significant improvements in survival, with all tumor-bearing mice treated with the combination therapy surviving through 60 days posttreatment and 290-fold reduction in AUC.To confirm that antitumor effects of the combination of CAR-T cells and hiNeuroS RANTES-IL-15 were associated with enhanced CAR-T-cell migration and persistence, we measured the number of CAR-T cells within GBM.Following the same treatment regimen as above, a subset of mice were sacrificed 10 days after intravenous CAR-T-cell infusion.Brains were collected and immunohistochemistry (IHC) was performed to define the distribution and number of CAR-T cells within the tumor (Figure 3f).Tumor volumes were larger in mice treated with CAR-T cells alone, while mostly scar tissue remained following CAR-T-cell therapy in hiNeuroS RANTES-IL-15treated mice.Anti-human CD3 staining showed a high degree of CAR-T-cell infiltration into tumors in both treatment groups.Taken together, these findings demonstrate that CAR-T therapy alone is insufficient to kill GBM in vivo.However, when combined with hiNeuroS RANTES-IL-15 cells, CAR-T therapy can completely eradicate tumors as a result of enhanced T-cell migration and persistence.
U R E 4 hiNeuroS cytokine delivery platform enhances CAR-T therapy for GBM.CAR-T therapy alone results in low lymphocyte infiltration and apoptosis due to the immunosuppressive tumor microenvironment in GBM.When combined with tumor-homing hiNeuroS RANTES- IL-15 , CAR-T efficacy is enhanced as a result of increased lymphocyte infiltration, activation, proliferation, and survival.Schematic(s) created with BioRender.com.CAR-T, chimeric antigen receptor-modified T cell; GBM, glioblastoma.
However, CAR-T-cell entry into the CNS is obstructed by unique anatomical barriers that are further complicated by the immunosuppressive properties of GBM.Using U87 RANTES-IL-15 , we demonstrated increased CAR-T-cell migration into the brain following systemic infusion.CAR-T infiltration peaked on Day 4, then began to decline by Day 7.This timeline is consistent with previously published data on CAR-T-cell tissue biodistribution, which showed peak peripheral tumor infiltration 2 days postinjection and subsequent decline by Day 5.