Gene therapy using genome‐edited iPS cells for targeting malignant glioma

Abstract Glioblastoma is characterized by diffuse infiltration into the normal brain. Invasive glioma stem cells (GSCs) are an underlying cause of treatment failure. Despite the use of multimodal therapies, the prognosis remains dismal. New therapeutic approach targeting invasive GSCs is required. Here, we show that neural stem cells (NSCs) derived from CRISRP/Cas9‐edited human‐induced pluripotent stem cell (hiPSC) expressing a suicide gene had higher tumor‐trophic migratory capacity compared with mesenchymal stem cells (MSCs), leading to marked in vivo antitumor effects. High migratory capacity in iPSC‐NSCs was related to self‐repulsive action and pathotropism involved in EphB‐ephrinB and CXCL12‐CXCR4 signaling. The gene insertion to ACTB provided higher and stable transgene expression than other common insertion sites, such as GAPDH or AAVS1. Ferroptosis was associated with enhanced antitumor immune responses. The thymidylate synthase and dihydroprimidine dehydrogenase expressions predicted the treatment efficacy of therapeutic hiPSC‐NSCs. Our results indicate the potential benefit of genome‐edited iPS cells based gene therapy for invasive GSCs. Furthermore, the present research concept may become a platform to promote clinical studies using hiPSC.

0][11][12] Whether NSCs or MSCs are useful GDVs for diffuse infiltrative glioma cell treatment is unclear.Suicide gene therapy using herpes simplex virus thymidine kinase (HSVtk) or cytosine deaminase (CD) might be appropriate for glioblastoma. 13However, transduced HSVtk was silenced or cytotoxic during neuronal differentiation of hiPSCs. 14Stable constitutive expression of therapeutic genes in hiPSCs is difficult using viral vectors, 15,16 which integrate randomly into host genomes, which raises concerns of insertional mutagenesis and oncogene activation. 17Therefore, stable and safe transgene expression is required by designed insertion into the appropriate loci using genome-editing technology. 18 established a novel therapeutic approach using genomeedited iPSC-derived NSCs targeting invasive glioblastomas.
In silico ligand-receptor pairing analysis based on RNA-seq datasets was used to identify signals underlying NSC-specific tumor tropism.We screened curated ligand-receptor pairs and determined potential interactions for each cell type related to the presence of a complementary ligand or receptor in every other cell type (Figure 1c).
We evaluated autocrine interactions to detect self-repulsion signaling pathways because cell behavior differed between NSCs and MSCs, even without tumor cells (Figure 1a).GSCs displayed a diffusive invasion pattern, similar to the migration of NSCs.Ninety-two ligandreceptor pairs were autocrine active in iPSC-NSCs and GSCs, but not MSCs, and were associated with Eph/ephrin signaling (Figure 1d,e).
The paracrine interactions between GSCs and NSCs/MSCs were evaluated to identify the potential signaling for chemoattraction toward GSCs.Analysis of 27 ligand-receptor pairs activated in GSCs and NSCs (Figure 1g) demonstrated no enrichment of specific signaling pathways.This result led us to hypothesize that the gene expression profile of in vitro-cultured GSCs did not mirror the precise physiology of intracerebral glioblastoma, hindering the identification of chemoattractive signaling from glioblastoma.The addition of glioblastoma and parental intact brain tissue RNA-seq data to our in silico ligand-receptor pairing analysis has indicated that eight ligand-receptor pairs were selectively active between resected glioblastoma and NSCs (Figure 1g).The analysis suggested that NSCs, but not MSCs, migrated toward glioblastomas by EphB-ephrinB and C-X-C motif chemokine ligand (CXCL)12-C-X-C motif chemokine receptor (CXCR)4 signaling pathways.We focused on CXCL12-CXCR4 because it regulates tumor tropism. 21The interactions involving chemoattractant chemokine ligands and receptors are activated in the microenvironment of brain tumors. 22CR4 was highly expressed in NSCs but not in MSCs; CXCL12 ligand expression was upregulated in resected glioblastoma, compared with in vitro-cultured tumor-derived cells (Figure 1h).Reanalysis of published RNAseq data of resected glioblastoma demonstrated that CXCL12 was expressed in most samples [23][24][25] (Figure S4C).Furthermore, we have performed the migration assay using a CXCR4 antagonist (AMD3100) and a specific EphB4 inhibitor (NVP-BHG712).Although no significant change of net distance was observed in NSC (p = 0.32), antagonist for CXCR4 have blocked CXCL12-mediated iPSC-NSC pathotropism toward glioma cells (Figure S4D).On the other hand, the inhibition of EphB4 had little effect on pathotropism, but net distance was significantly decreased (p = 0.026).Inhibition of EphB/ephrinB pathway has blocked self-repulsive action (Figure S4E).This result has suggested the importance of the CXCL12/ CXCR4 and EphB/ephrinB pathway in regulating homing of engrafted NSCs to malignant glioma sites.
Whole-genome sequencing is a method to rapidly identify genetic variations.Whole-genome sequencing showed no off-target mutations induction in CD-iPSCs and CD-NSCs (Figure S6E,F).

| Therapeutic efficacy of CD-mNSCs in immunocompetent mouse model of GSC
To evaluate T cell-mediated antitumor immune response, NSCs differentiated from 38C2 mouse iPSC line 27,28 were transduced with the lentiviral vector CSII-EF-yCD-UPRT-IRES-hKO1.yCD-UPRT-transduced mouse iPSC-derived NSCs (CD-mNSCs) were highly sensitive to 5-FC.Stable constitutive yCD-UPRT expression was achieved in mNSCs (Figure S8A).Representative BLI, radiance intensities, tumor volume, and Kaplan-Meier plots of treated mice indicated notable antitumor effects (Figure 4a-d; Figure S8B).Tumor cells were gradually killed by the bystander killing effect of CD-mNSCs (Figure 4b).Histological analysis showed the complete disappearance of TSG cells (ffLuc) in 4/6 mice (Figure 4b).CD8 + cell numbers were significantly higher in brains of treated mice than controls (TSG cells [ffLuc] only) after 5-FC administration (Figure S8C).Furthermore, only CD-mNSCs without tumor cells were implanted into the mouse brain, resulting in the CD8 + cell infiltration.This result suggested a direct immune response to the iPSC-derived NSCs (Figure S8C).Treated mice had higher proportions of intratumoral CD8 + cells in CD45 + CD3 + cells than controls (Figure S8D).Fewer CD163 + immunosuppressive cells were observed in treated mice than controls (Figure S8C).Survival was significantly prolonged in treated mice compared with controls (Figure 4d), which suggests that CD-mNSCs with 5-FC administration enhanced antitumor immune responses.

| Safety of CD-NSCs (ACTB)
To evaluate the toxic effects of 5-FC, 5-FC was added into the medium containing iPSC-NSCs without yCD-UPRT gene or glioma cells.No killing power was observed for both iPSC-NSCs and glioma cells (Figure S9A).
To analyze safety, CD-NSCs (ACTB) transduced ffLuc 19 by lentiviral vector were implanted into normal brains (Figure S9B).BLI signal intensity and Venus fluorescence disappeared in mice receiving 5-FC from Days 7 to 21 (Figure S9C-E).No differences in GFAP, NeuN, and CD31 expressions were observed at implanted sites between 5-FC (+) and (À) brain sections (Figure S9F).Ki-67 + cells were not present in brains after 5-FC (+) administration and cleaved caspase-3 + cells were present in implanted striatum (Figure S9F).There were no differences in Ki-67 + and nestin + endogenous mouse neural progenitor cell number in periventricular areas between 5-FC (+) and (À) brain sections (Figure S9F).Liquid chromatography-mass spectrometry showed 15 μM 5-FU in 1 Â 10 5 CD-NSCs (ACTB) supernatant after adding 5-FC (Figure S9G  junctions (bystander killing effect). 13The UPRT gene can directly convert 5-FU to FUMP. 13 hG008, GL261, and TSG showed higher F I G U R E 2 Gene loci achieving high and stable yCD-UPRT expression.(a) Schematic depiction of the CRISPR/Cas9-mediated strategy for inserting the yCD-UPRT gene into the indicated locus.Single guide RNA (sgRNA) target sequence and HR donor constructs are shown.(b) Each CD-NSC was cultured in the presence of 0, 7, 14, 35, or 50 μM 5-FC for 7 days.The sensitivity to 5-FC was evaluated by CCK-8 assay.Data represent the mean ± SD (n = 6).(c) qRT-PCR analysis of each CD-NSC for CD or GAPDH expression.Gene expression was normalized relative to β-actin expression.Data represent the mean ± SD (n = 3).(d) Western blot analysis of each CD-NSC for CD or GAPDH expression.β-Actin was used as an internal control.(e) Immunocytochemical analysis of neurospheres stained with anti-GAPDH antibody or anti-CD antibody and DAPI.Densitometry of signals was evaluated.(f) Immunocytochemical analysis of EdU + cells (red) in neurospheres.Nuclei were stained with Hoechst 33258 (blue).The frequency of neurospheres containing immunopositive cells is shown.Data represent the mean ± SD (n = 10).(g) Representative images of CD-NSC (ACTB; at passage 20).CD-NSCs (ACTB) were cultured in the presence of 0, 7, 14, 35 or 50 μM 5-FC for 7 days.The sensitivity to 5-FC was evaluated by CCK-8 assay.Data represent the mean ± SD (n = 3).Scale bar, 100 μm.WT, wild-type iPSCderived NSCs sensitivity to 5-FU than other glioma cells (Figure S10B).Significantly lower thymidylate synthase (TS) and dihydropyrimidine dehydrogenase (DPD) expressions in hG008, GL261 and TSG than in other glioma cells were confirmed by quantitative reverse transcription PCR (Figure S10C).The dotted lines indicated that glioma cells (hG008, GL261 and TSG) with the half gene expressions of the TS and DPD compared with U87 are excellent responders to 5-FU (Figure S10D).The present in vivo antitumor effect may be associated with not only high migratory capacity of CD-NSCs but also the TS and DPD expressions in glioma cells.

| DISCUSSION
Few studies have evaluated differences in the tumor-trophic properties of NSCs and MSCs in the brain. 29We demonstrated that iPSC-NSCs had higher tumor-trophic migratory capacity than MSCs in the brain.RNA-seq-based ligand-receptor pairing analysis suggested that self-repulsive action and pathotropism were important for iPSC-NSC migration related to ephrin ligand/receptor signaling-mediated repulsion in iPSC-NSCs and CXCL12-CXCR4 interactions between GSCs and iPSC-NSCs.EphB-ephrinB signaling enhanced neural crest and hippocampal stem/progenitor cell migration. 30,31CXCR4/CXCL12 signaling promoted tropism of NSCs toward glioma cells. 32These signaling pathways were not detected in MSCs, which suggests their role in migration differences between NSCs and MSCs.Indeed, CD-NSCs induced strong antitumor effects even in GSC mice with diffuse invasiveness. 4,33Cs promoted glioma cell proliferation.Inflammatory cytokines secreted by MSCs were associated with tumor growth [34][35][36] and MSCs differentiated into CAFs expressing αSMA. 37Furthermore, FAP + fibroblasts concentrated around implanted MSCs in the brain.Therefore, MSCs might not be suitable for treating malignant tumors.
Housekeeping gene loci (GAPDH or ACTB) or safe harbor sites (AAVS1) were selected as efficient gene knock-in loci in iPSCs and embryonic stem cells, [38][39][40] but comparative gene expression at these target sites was not reported.ACTB had the highest stability under any condition. 41The present study suggests that ACTB is an appropriate locus for the stable insertion of therapeutic genes in hiPSCs.
Ferroptosis, characterized by excessive iron accumulation and lipid peroxidation, 42,43 enhanced cell immunogenicity and recruited immune cells to tumor sites. 445-FU induced apoptosis in colorectal cancer. 45Here, converted 5-FU released from CD-NSCs induced ferroptosis (greater than exogenous 5-FU administration) and apoptosis in GSCs, which led to antitumor immune responses.Ferroptosis requires continuous iron-dependent reactive oxygen species formation over an extended period to trigger death. 43Cell-based suicide gene therapy killed tumor over a longer period than exogenous 5-FU administration because 5-FU was gradually released after prodrug conversion in CD-NSCs.
The mechanisms of resistance to 5-FU have been previously evaluated in other malignant tumors including colorectal, breast, gastric, pancreatic, and lung cancers. 46TS, which is an essential enzyme for DNA de novo synthesis, and DPD, which catabolizes 5-FU to the inactive metabolite, were measured in those studies. 46However, few studies focused on TS or DPD expression in glioblastomas. 47,48The present study first demonstrated that the TS and DPD expressions predict the treatment efficacy of CD-NSC in glioblastomas.Biomarkers to predict treatment efficacy can be utilized in the personalized medicine.
Nonlytic, amphotropic retroviral replicating vectors and immortalized human NSCs derived from human fetal brains to delivery CD [49][50][51] did not affect overall survival.Virus coverage might not have encompassed the large invading glioma cell area.hiPSC-derived NSCs might provide better therapeutic effects because of their high tumortrophic migratory capacity.Preclinical hiPSC-derived NSC studies for glioblastomas are required.
Limitation of the present study was the paucity of the number of cell lines in iPSCs.The ultimate goal of the cell bank is to supply these therapeutic iPSCs of good quality at large scale.iPSCs may show different biochemical characteristics among cell lines from different donors.However, the migratory capacity of iPSCs can be quantitatively screened via our established assays including organotypic brain slice culture before implantation.
In this study, migration capacity between iPSC-NSC was compared with adult BMSC or AMSC.However, that is inappropriate because they were derived from different tissues and different donors.In the future study, we will use iPSC-derived NSC and the iPSC-derived MSC, 52,53 allowing to compare migration capacity head-to-head with more accuracies.
Higher cell numbers and repeat cell injection may provide additional treatment effects in this strategy.Future studies with higher injected cell concentrations and repeated cell therapy are warranted to further evaluate this treatment.NSCs express various receptors for chemoattractant signals because of brain pathology.These chemoattractants are chemokines such as CXCL12 and monocyte chemoattractant protein 1, or other chemotactic proteins, such as vascular endothelial growth factor (VEGF). 7 Therefore, further analyses are warranted to elucidate the mechanisms of migration.

| CONCLUSIONS
NSCs derived from CRISRP/Cas9-edited hiPSC have high tumortrophic migratory capacity and stable constitutive therapeutic transgene expression, which leads to strong antitumor effects against GSCs.The present research concept may become a platform to promote clinical studies using hiPSC.

| Human iPSCs
1210B2-hiPSCs 54 were derived from human peripheral blood mononuclear cells of a healthy 29-year-old African/American female (Cellular Technology Limited).1210B2-hiPSCs (kindly provided by Shinya Yamanaka, Kyoto University, Kyoto, Japan) were cultured with a feeder-free protocol. 12,54Embryoid body (EB) formation and neural stem/progenitor cell generation were performed as previously described. 12,55

| Mouse iPSCs
The mouse iPS clone, 38C2, 27 established from mouse embryonic fibroblasts, was differentiated into neurospheres via EBs in the presence of 10 M retinoic acid (Sigma-Aldrich, Kanagawa, Japan) as described previously. 12,55

| U87 cells
A U87 human glioma cell line was obtained from the American Type Culture Collection (HTB-14; VA, USA).Single-cell clones stably expressing ffLuc gene (a Venus fluorescent protein and firefly luciferase fusion gene) 19 were established as previously described. 4he U87 model is not infiltrative has an entirely abnormal and leaky vasculature and is not of glial origin. 56In the present study, pathotropism was evaluated using a Rose diagram map according to a previous study. 57The Rose diagram map can show the quantified spatial distribution of CD-NSCs around the tumor, including both the number of CD-NSCs in various directions and their distance from the tumor center.In this system, bulk tumor mass is used.hG008 cells diffusely infiltrated into the brain parenchyma.Therefore, U87 cells were mainly used to make a Rose diagram map in this study.

| hG008 cells
The human GSC line (hG008) was established from human glioblastoma specimens. 33MIF expression in hG008 cells is higher than in nonbrain tumor-initiating cells. 33In tumor-derived neurosphere culture in vitro, hG008 cells can be expanded longer than nonbrain tumor-initiating cells.
Single-cell clones stably expressing ffLuc were established. 4We have previously reported for the first time the spatiotemporal characterization of human GSC invasion in an orthotopic xenograft mouse model using timelapse imaging of organotypic brain slice cultures and 3D imaging of optically cleared whole brains. 4GSCs in the corpus callosum migrated more rapidly and unidirectionally toward the contralateral side with pseudopod extension.These characteristics of GSC invasion shared the histological features observed in glioblastoma patients.hG008 cells (ffLuc) were cultured in ultra-low attachment cell culture flasks (Corning, NY, USA) using the same culture conditions as that for neurospheres. 4

| TSG cells
A mouse GSC line (TSG) was kindly provided by the Division of Gene Regulation Keio University School of Medicine and cultured with the same procedures used for hG008 cells. 58TSG was established by overexpressing H-Ras (V12) in normal NSCs isolated from the subventricular zone of adult mice harboring a homozygous deletion of the Ink4a/Arf locus. 58Single-cell clones stably expressing ffLuc were established as previously described. 4,58

| Other human and mouse glioma cell lines
Other human and mouse glioma cell lines (U251, SF126 and GL261), and human GSC line (hG020) 33 were cultured using the same procedures used for U87 cells and hG008 cells, respectively.SF126 was obtained from the JSRB Cell Bank (IFO50286; Osaka, Japan).U251 cells were obtained from the RIKEN BRC (Ibaraki, Japan).
Recombinant lentiviral vector production and titer determination were performed as described previously. 12

F
I G U R E 1 RNA-seq data analysis.(a) Analysis of mice implanted with iPSC-NSC, AMSC2, or BMSC2 (hKO1 + ) in the absence of U87 cells (ffLuc).Scale bar, 100 μm.(b) 3D plot of principal component analysis.Each dot represents a sample.(c) Overview of the in silico ligand-receptor pairing analysis.All curated ligand-receptor pairs were screened and pair matching was extracted based on the presence of a complementary ligand or receptor.(d) Heatmap showing the matched or unmatched status of autocrine cell-cell interactions for each cell type for all curated ligand-receptor pairs.(e) Enrichment analysis of NSC/GSC-specific autocrine ligand-receptor pairs across C2 genesets in MSigDB.(f) Heatmap of EphB/ephrinB gene expression.(g) Heatmap showing the matched or unmatched status of paracrine cell-cell interactions for each cell type for ligand-receptor pairs.(Left) Status of paracrine cell-cell interactions between NSCs/MSCs and GSCs for all curated pairs.(Middle and Right) Status of paracrine cell-cell interactions between NSCs/MSCs and tumor/intact brain tissue for selected pairs.A list of eight ligand-receptor pairs specific to the tumor-NSC interaction is shown.(h) The expression level of CXCR4 and CXCL12 Organotypic brain slice culture was used to visualize antitumor effects on hG008 cells (ffLuc).5-FC reduced hG008 cell (ffLuc) growth in the slice cultures with CD-NSCs (hKO1 + ; Figure S7A,B; Movie S3).

2. 10 |
Biomarker to predict therapeutic response The metabolism of 5-FU is shown in Figure S10A.CD can convert 5-FC to 5-FU, which is converted to fluorouridine monophosphate (FUMP) or fluorodeoxyuridine monophosphate (FdUMP) through fluorouridine.Furthermore, FUMP is converted to fluorouridine diphosphate (FUDP) and fluorouridine triphosphate (FUTP), which inhibits RNA synthesis via the cell cycle-independent pathway.FdUMP is converted to FdDMP and FdTMP and inhibits DNA synthesis via the cell cycle-dependent pathway.5-FU has a dual mechanism of action.It directly kills the CD-transduced cells and neighboring untransduced cancer cells through cell membranes without cell-cell