Engineering of rhesus monkey pluripotent stem cells for noninvasive survey and remote control after brain allotransplantation

Highlights

• •Long-term noninvasive survey of rhesus monkey brain allografts harboring build-in stimulatory DREADD with MR imaging, [¹⁸F]FDG-PET and EEG.
• •Achieving remote control over neuronal activity of brain allografts derived from engineered rhesus monkey pluripotent stem cells.
• •Improving the safety of transplantation in rhesus monkeys via a preset built-in suicide system.
• •Integration of genetic engineering and cell therapy represents a powerful strategy for the development of controllable cell therapeutics.

Abstract

Advanced technologies of targeted differentiation and genetic engineering in pluripotent stem cells (PSCs) offer production of purposefully designed cells for transplantation. Here, we engineer rhesus monkey PSCs (rhPSCs) to produce safe and functional neural progenitors and neurons, which could be noninvasively surveyed and controlled after brain transplantation. RhPSCs expressing hM3Dq-mCherry exhibited normal karyotype and had the ability to equally self-renew and differentiate into functional neurons as that of the mCherry expressing control. After allotransplantation into monkey prefrontal cortex, the hM3Dq-mCherry expressing rhPSC-derived cortical progenitors survived and matured gradually, which could be long-termly surveyed by [¹⁸F]-fluorodeoxyglucose positron emission tomography ([¹⁸F]FDG-PET) and electroencephalogram (EEG) following clozapine-N-oxide (CNO) administration. Remote activation of transplanted neurons caused increased [¹⁸F]-fluorodeoxyglucose uptake started at 3 months and reached a plateau 9–12 months post-transplantation. EEG analysis revealed a decrease in the high delta spectrum power while an increase in the beta power after CNO delivery. Notably, no seizure-like spikes were observed even after repeated CNO exposure. Magnetic resonance (MR) imaging found no overgrowth in all allografts. Expression of herpes simplex virus thymidine kinase (HSVtk), a Ganciclovir (GCV)-induced suicide gene in rhPSCs further assured the safety without sacrificing the functional outcome of brain grafts. Our study offers a feasible strategy for long-term noninvasive survey and remote control of brain-grafted neurons.

Graphical abstract

1. Introduction

Human pluripotent stem cells (PSCs), including human embryonic stem cells (ESCs) and human induced pluripotent stem cells (iPSCs), possess capabilities for unlimited self-renewal and differentiation into various types of neural cells [1], [2], [3], [4]. It is thus well-acknowledged that human PSCs have great promise for the treatment of diverse neurological disorders [5,6]. Pioneering studies reveal that regional neural progenitors derived from human PSCs could properly survive, differentiate into desired neurons and integrate into host neuronal circuits, yielding encouraging functional outcomes [7], [8], [9]. Currently, more than a dozen registered human PSC-based clinical trials for the therapies of neurological disorders are now underway worldwide [10,11].

With the advancement of genetic engineering technologies, human PSCs could be engineered for lineage labeling and tracing [12], correction of mutated gene loci [13,14], and control of cell function and safety after transplantation [15], [16], [17]. Ectopic expression of designer receptors exclusively activated by designer drug (DREADD) in human PSCs has been utilized to activate or inhibit neurons transplanted in Parkinson's disease models of mice and rats [18,19]. Suicide gene herpes simplex virus thymidine kinase (HSVtk) has been introduced into human PSCs to selectively eliminate unwanted dividing cells in order to improve the safety of brain transplantation in rodents [20,21]. Thus, integration of genetic engineering and cell-based therapy represents a powerful strategy for future controllable cell therapeutics that addresses concerns related to efficacy and safety in clinical trials for human PSC-based cell therapies.

Conducting stringent preclinical studies in nonhuman primates (NHPs) would greatly benefit newly developed therapeutics before applying them in real-world researches [22,23]. Meanwhile, multidimensional and noninvasive assessment methods, such as magnetic resonance (MR) imaging, positron emission tomography (PET) and electroencephalogram (EEG), are clinically employed for diagnosing neurological disorders. Integration of these noninvasive strategies in NHP studies will largely recapitulate future clinical studies on newly developed therapeutics and therefore facilitate their application [24].

In this study, we aim to establish a system for long-term survey as well as safety and efficacy control of neuronal cells after transplantation into NHP brain. We report that rhesus monkey PSCs (rhPSCs) with build-in DREADD and HSVtk cassettes are safe and functional for noninvasive survey and remote control of transplanted neuronal cells with standard clinical procedures and multidimensional assessment methodologies. Our study therefore offers a practical strategy for engineering human PSCs to ensure a way of visible, safe and controllable allografts for the treatment of neurological disorders.

2. Materials and methods

2.1. Maintenance and glutamatergic neuron differentiation of rhPSCs

RhESCs [25] (R278.5, passages 25–43) and rhiPSCs [26] (passages 28–43, reprogrammed from skin fibroblasts infected with retroviral particles expressing OSKM Yamanaka factors) were maintained on X-ray inactivated CF-1 mouse embryonic fibroblast feeder layer and passaged with Trypsin (Gibco, 27250018) at a split ratio of 1:8–10 every 4–5 days in the PSC growth medium [27] (DMEM/F12 (50%, Gibco, 11330057), Neurobasal (50%, Gibco, 21103049), N2 (1 ×, Gibco, 17502048), B27 (1 ×, Gibco, 12587010), Glutamax (1 ×, Gibco, 35050079), NEAA (0.1 mM, Gibco, 11140076), knockout serum replacer (2%, Gibco, A3181502), β-mercaptoethanol (0.1 mM, Amresco, 0482), bFGF (20 ng/mL, PeproTech, 100–18B) and IWR-1 (2.5 µM, Selleck, s7086)).

For cortical progenitor and glutamatergic neuron differentiation, rhPSC colonies were digested by Dispase (Gibco, 17105041) and lifted in suspension culture in the embryoid body (EB) differentiation medium (DMEM/F12 (50%), Neurobasal (50%), N2 (1 ×) and B27 (1 ×), supplemented with bFGF (10 ng/mL), SB431542 (5 µM, Selleck, s1067), CHIR99021 (3 µΜ, Selleck, s1263), Compound E (0.2 µM, MCE, HY-14176) and LDN193189 (0.1 µM, Selleck, s2618)). On day 6, EBs were collected and transferred onto laminin 521-coated (1 µg/mL, BioLamina, LN521–05) 6-well plate for attachment culture with culture medium switched to the neural induction medium [28] (Neurobasal, N2 (1 ×), B27 (1 ×), Glutamax (1 ×) and NEAA (0.1 mM), supplemented with bFGF (10 ng/mL), SB431542 (5 µM), CHIR99021 (3 µM) and human LIF (1000 U/mL, Millipore, LIF1050)). The medium was changed every other day. On day 17, neural rosettes were gently blown off by 1 mL pipette and the aggregates were expanded to form neurospheres in suspension culture using the same medium. Mechanical digestion was used to break neurospheres into small aggregates containing 15–30 cells for passage if desired.

For neural differentiation, neruospheres on day 25 were dissociated into single cells using Accutase (Gibco, A1110501) and transferred onto laminin 521-coated (1 µg/mL) coverslips at a density of 10⁴ cells per coverslip with the culture medium switched to the neural differentiation medium [29] (Neurobasal, Glutamax (0.1 ×), N2 (1 ×) and B27 (1 ×), supplemented with ascorbic acid (200 µM, Sigma, A4403), cyclic AMP (1 µM, Sigma, D0627), BDNF (10 ng/mL, PeproTech, 450–02), GDNF (10 ng/mL, PeproTech, 450–10), IGF-1 (10 ng/mL, PeproTech, 100–11) and Compound E (1 µM, only supplemented for the first three days)).

2.2. Genetic engineering of rhPSCs

ROCK inhibitor Y27632 (1 mM, Selleck, S1049) pretreated rhESCs and rhiPSCs were digested by Trypsin and 10⁶ rhPSCs were resuspended in 200 µL electroporation buffer (5 mM KCl, 5 mM MgCl₂, 15 mM HEPES, 102.94 mM Na₂HPO₄ and 47.06 mM NaH₂PO₄, pH 7.2) mixed with 5 µg pCas9-GFP plasmid (Addgene #44719), 5 µg rhAAVS1 sgRNA (ACCGTGGGGCCACTAGGGA) plasmid and 30 µg donor plasmid (5′arm-SA-puro-CAG-hM3Dq-mCherry-3′arm, 5′arm-SA-puro-CAG-mCherry-3′arm or 5′arm-SA-puro-CAG-HSVtk-3′arm). Electroporation was performed using a Gene pulser Xcell (Bio-rad) system at 250 V, 500 µF with a 0.4 cm cuvette (Bio-rad). After electroporation, rhPSCs were replated on feeder layer with Y27632 (1 mM) supplied for the first 24 h. Puromycin (500 ng/mL, Sigma, p8833) was supplied on day 2, and after 5–7 days of drug selection, individual colonies were selected, expanded and subjected to genomic DNA PCR, Sanger sequencing and Southern blot. Locations of genomic DNA PCR primer sets and Southern blot probe are labelled in Fig. 1a and Fig. S3a. Primer sequences are listed in Table S1.

Fig. 1.

Engineering andin vitrocharacterization of DREADD-rhPSCs for noninvasive survey and remote control. (a) Schematic representation of donor vector construction and strategy for knock-in mCherry or hM3Dq-mCherry expressing cassette into rhAAVS1 locus to generate mCherry- and DREADD-rhPSCs. SA, splicing acceptor; puro, puromycin resistance gene; CAG, cytomegalovirus immediate-early enhancer/chicken β-actin promoter; scissor mark, rhAAVS1 sgRNA targeting site; F1/R1 and F2/R2, genomic DNA PCR primer sets for genotyping; SB probe, probe used for Southern blot validation. (b) Representative immunofluorescence images of NANOG, OCT4 and mCherry staining in mCherry- and DREADD-rhPSCs. The nuclei are stained with Hoechst 33342 (HOE). Scale bars, 20 µm. (c) Representative immunofluorescence images of Ki67, SOX2, PAX6, N-Cadherin (N-Cad), NESTIN and mCherry staining in mCherry- and DREADD-rhPSC-derived cortical progenitors. Neuroepithelial organized into rosette-like structures are observed in both groups. Scale bars, 20 µm. (d) Quantification of positively labeled cells in (c). For mCherry group, Ki67⁺, 84.98% ± 2.08%; SOX2⁺, 96.76% ± 0.43%; PAX6⁺, 95.90% ± 1.39%. For DREADD group, Ki67⁺, 84.14% ± 2.39%; SOX2⁺, 98.04% ± 0.69%; PAX6⁺, 97.76% ± 0.43%. Data are presented as mean ± SEM, n = 5 for each group. (e) Representative immunofluorescence images of vGluT1, FOXG1, Tuj1 and mCherry staining in mCherry- and DREADD-rhPSC-derived cortical neurons. Scale bars, 20 µm. (f) Quantification of positively labeled cells in (e). For mCherry group, vGluT1⁺, 93.10% ± 1.34%; FOXG1⁺, 91.08% ± 0.94%; Tuj1⁺, 98.16% ± 0.82%. For DREADD group, vGluT1⁺, 92.88% ± 1.01%; FOXG1⁺, 90.34% ± 1.42%; Tuj1⁺, 97.54% ± 1.03%. Data are presented as mean ± SEM, n = 5 for each group. (g) Quantification of rest membrane potential (RMP) in 8-week-old glutamatergic neurons. For mCherry group, −37.33 ± 2.98 mV. For DREADD group, −42.83 ± 2.45 mV. Data are presented as mean ± SEM, n = 10 for each group, unpaired t-test, n.s., no significant difference. (h) Whole-cell patch current-clamp recordings in 10-week-old glutamatergic neurons derived from mCherry- and DREADD-rhPSCs showing representative changes in membrane potential (MP) upon CNO (10 µM) perfusion and washout. (i) Quantification of MP change in (h). For mCherry group, before vs. after CNO perfusion, −45.94 ± 2.64 mV vs. −45.18 ± 2.09 mV. For DREADD group, before vs. after CNO perfusion, −48.56 ± 1.29 mV vs. −41.96 ± 1.47 mV. Data are presented as mean ± SEM, n = 5 for each group, paired t-test, **, p < 0.01. (j) Whole-cell patch voltage-clamp recordings in 10-week-old glutamatergic neurons derived from mCherry- and DREADD-rhPSCs showing representative changes in sEPSC frequency and amplitude upon CNO perfusion and washout. Voltage was held at −70 mV. (k) Quantification of sEPSC frequency change in (j). For mCherry group, saline, 8.2 ± 0.92; CNO, 6.9 ± 1.30; washout, 9.2 ± 0.98. For DREADD group, saline, 5.4 ± 0.92; CNO, 14.3 ± 1.77; washout, 7.8 ± 1.18. Data are presented as mean ± SEM, n = 10 for each group, ordinary one-way ANOVA with Turkey test, **, p < 0.01, ***, p < 0.001. (l) Quantification of sEPSC amplitude change in (j). For mCherry group, saline, 53.39 ± 1.55 pA; CNO, 52.48 ± 2.23 pA; washout, 49.81 ± 1.23 pA. For DREADD group, saline, 52.16 ± 2.85 pA; CNO, 60.16 ± 1.44 pA; washout, 57.65 ± 2.11 pA. Data are presented as mean ± SEM, n = 10 for each group, ordinary one-way ANOVA with Turkey test, *, p < 0.05.

For constructing rhPSCs that express both HSVtk and hM3Dq-mCherry, HSVtk expressing rhPSCs were infected with EF1α-hM3Dq-mCherry-IRES-BSD expressing lentiviruses. Blasticidin (2 µg/mL, InvivoGen, ant-bl-05) was supplied for positive selection from day 2 to day 7.

2.3. Karyotyping

Chromosome preparations were obtained 60 h after passage. Cells were harvested after a 4 h N-desacetyl-N-methylocolchicine pulse (1:50, Gibco, 15212012). The GTG-banding was analyzed according to standard rhesus monkey karyotype [30].

2.4. Staining

For immunofluorescence staining, cells cultured on coverslips were fixed with 4% paraformaldehyde for 15–20 min, then permeabilized and blocked with 10% donkey serum and 0.1% Triton X-100 in PBS for 1 h. Primary antibodies were diluted in 5% donkey serum and 0.05% Triton X-100 in PBS. The coverslips with fixed cells were incubated with the primary antibodies at 4 °C overnight. After adequate washing with PBS, cells were then incubated with fluorescent-conjugated secondary antibodies (Jackson ImmunoResearch, 1:1000 in PBS) for 1 h at ambient temperature. Nuclei were counterstained with Hoechst 33342 (1:1000, Sigma, 382065). Coverslips were mounted with Fluoromount-G (Southernbiotech 0100-01). Primary antibodies used are, NANOG (1:2000, R&D, AF1997), OCT4 (1:1000, Santa Cruz, sc-5279), mCherry (1:2000, Invitrogen, M11217), vGluT1 (1:400, Chemicon, MAB5502), FOXG1 (1:500, Abcam, ab18259), TUJ1 (1:1000, Sigma, T8660), SOX2 (1:500, R&D, AF2018), PAX6 (1:500, Biolegend, 901301), NESTIN (1:500, Chemicon, MAB5326), N-Cadherin (1:1000, Santa Cruz, sc-8424,), Ki67 (1:400, Chemicon, MAB4190), GABA (1:500, Sigma, A2052), ChAT (1:1000, Chemicon, AB144P), HOXB4 (1:50, DSHB, I12 anti-Hoxb4) and FOXA2 (1:1000, Abcam, ab60721). Positive cells were counted manually in a double-blinded manner using Photoshop CC 2018 (Adobe). Fluorescence intensity was calculated using Fiji [31].

For alkaline phosphatase assay, rhPSCs were fixed in 4% paraformaldehyde for 1 min. After rinsing with TBST, cells were stained with alkaline phosphatase kit (SiDanSai, 1101-050) for 15–20 min in dark according to manufacturer's instructions.

2.5. Electrophysiological recording

Whole-cell patch-clamp recordings were performed on in vitro differentiated neurons 12–14 weeks after differentiation (8–10 weeks after being plated). Cells cultured on coverslips were perfused with 95% O₂/5% CO₂ bubbled artificial cerebrospinal fluid (ACSF, 127 mM NaCl, 1.9 mM KCl, 26.2 mM NaHCO₃, 1.2 mM KH₂PO₄, 2.2 mM CaCl₂, 1.4 mM MgSO₄ and 10 mM glucose, pH 7.25) by a peristaltic pump (Longer, BT100-2J) at ambient temperature. Intracellular solution (140 mM K⁺-gluconate, 0.1 mM CaCl₂, 2 mM MgCl₂, 1 mM EGTA, 4 mM Mg²⁺-ATP, 0.1 mM GTP Na₃ and 10 mM Na⁺-HEPES, pH 7.25) filled electrodes were prepared by vertical puller (Narishige, PC-100) with 5–8 MΩ resistance. Voltage and current recordings were amplified by MultiClamp amplifier (Axon instruments, 700B), converted by Digidata analog-digital converter (Axon instruments, 1550B) and analyzed by pClamp 9.0 software (Axon Instruments). RMPs and APs were recorded under current clamp mode. sEPSCs were recorded at a holding potential of −70 mV under voltage clamp mode.

For DREADD activation assays, 10 µL saline and 10 µL saline dissolved CNO (12 mM, MCE, HY-17366A) were sequentially supplied in perfused ACSF, with each solution having a total volume of 12 mL.

2.6. Animal ethics

Adult male rhesus monkeys (macaca mulatta, aged 6–7 yr, weighed 4.5–10.0 kg) were supplied by Hubei Topgene Biotechnology Co., Ltd. Monkeys were housed in 12 h light-dark cycle environment at ambient temperature of 22 ± 1 °C and relative humidity of 55% ± 5%, sheltered individually in group 4 enclosures (floor 0.56 m², height 81.3 cm, recommended by Guide for the Care and Use of Laboratory Animals, 8th edition). The monkeys were fed twice daily with laboratory monkey diet (Beijing Keao Xieli Feed Co., Ltd.), offered ad libitum access to water and were given seasonal fruits, vegetables, nuts and snacks. Enrichment toys (Eetoys) and treats containers (Kong) were provided to enhance their well-being. All experimental procedures were approved by the Institutional Animal Care and Use Committee of School of Medicine, Tongji University (ethics number 2020YANYUSHEN093) and Tongji hospital, Tongji Medical College, Huazhong University of Science and Technology (ethics number TJH-202012019). All efforts were made to minimize the number of animals used and ameliorate any distress.

2.7. Cell transplantation

For cortical progenitor transplantation, neurospheres at 6–7 weeks were dissociated into single cells with Accutase and resuspended in Neurobasal supplemented with Y27632 (1 mM), B27 (1 ×) and BDNF (10 ng/mL) to reunite into small aggregates of 15–30 cells for 24 h at a density of 4.8 × 10⁶ viable cells/5 mL before transplantation.

Cortical progenitor aggregates were collected and resuspended in 30 µL Neurobasal supplemented with Y27632 (1 mM), BDNF (10 ng/mL) and B27 (1 ×). Stereotaxic injection was performed using a 27 G, 10 µL Hamilton syringe, with the coordinates summarized in Table S2. Rhesus monkeys were positioned in a stereotaxic apparatus (ZS-FD/D) according to standard MR imaging and histological atlas of the brain [32]. Coordinates of BA 9/46 in each monkey were referenced to the infraorbital ridge (ear bar zero, EBZ) and confirmed by CT scan [33]. BA 9/46 on each side was transplanted with corresponding cortical progenitors in 3 tracks and with 4 deposit spots in each track (Fig. S2a-b). 2 µL of aggregate suspension per deposit spot was injected at a rate of 0.5 µL/min.

For deltoid rhPSCs transplantation, 2 × 10⁷ undifferentiated HSVtk-rhPSCs were resuspended in 100 µL growth medium 1:1 mixed with monkey self-derived plasma, and administered intramuscularly to deltoid as one injection spot. Each side of the deltoid was injected at 5 spots.

2.8. HSVtk/GCV system characterization

Wild-type rhPSCs and HSVtk-rhPSCs were challenged with GCV (Pusheng, H20045077) at different concentrations for 48 h. HSVtk-rhPSC-derived neurons, after being plated, were exposed to 10 µM GCV for 5 or 10 days. Colony size, area and fluorescence intensity were calculated by Fiji.

GCV was dissolved in saline at a concentration of 25 mg/mL. A dosage of 10 mg/kg GCV was intravenously dripped at a rate of 1 mL/min daily for 2 months in monkeys transplanted with HSVtk-rhPSC-derived cortical progenitors or HSVtk-rhPSCs.

2.9. Surgical procedures

Monkeys had been fasted for 12 h before preanesthetic with atropine (0.03 mg/kg, i.m., Jilin Huamu Animal Health Products Co., Ltd.), then were induced anesthesia with ketamine (10 mg/kg, i.m., Jiangsu Hengrui Pharmaceuticals Co., Ltd.), and were maintained with isoflurane inhalation. (2% ± 1% in oxygen, RWD Life Science Co., Ltd.) Electrocardiograph (COMEN) was used to monitor vital parameters. Penicillin was given for 5 days at a dosage of 25,000 U/kg after transplantation. Immunosuppressant FK506 (MCE, HY-13756A) was supplied 1 week prior to transplantation and maintained throughout the experiment with blood concentration at 5–10 ng/mL.

2.10. Magnetic resonance imaging

Atropine (0.03 mg/kg), ketamine (4.4–9.0 mg/kg), midazolam (0.1–0.2 mg/kg, Jiangsu Nhwa Pharmaceutical Co., Ltd.) and xylazine (1–2 mg/kg, Jilin Huamu Animal Health Products Co., Ltd.) were i.m. administered into rhesus monkeys fasted for 12 h. A 3.0T MR scanner (uMR790, United Imaging) was used for imaging.

For deltoid graft detection, sedated monkeys were positioned supine with external rotation in order to optimize the scanning planes. T1-weighted sequences were acquired in axial, coronal, and sagittal planes (axial, TR/TE 576/11.2 mm, FA 120°, FOV 160 mm², matrix 420 × 600, slice 40 × 2 mm; coronal, TR/TE 435/10.3 mm, FA 150°, FOV 160 mm², matrix 432 × 576, slice 28 × 2 mm; sagittal, TR/TE 373/10.3 mm, FA 150°, FOV 160 mm², matrix 432 × 576, slice 24 × 2 mm). Volumetric analysis of deltoid graft was performed in 3D Slicer [34]. Manual delineation of low-signal regions on T1-weighted images was done slice by slice, and the segment editor toolbox was used for 3D reconstruction and volumetric estimation. For brain MR imaging, monkeys were fitted in an MR coil, T1-weighted (TR/TE 12.57/5.5 mm, FA 10°, FOV 96 mm², matrix 384 × 384, slice 60 × 1 mm), T2-weighted (TR/TE 3400/377.2 mm, FA 60°, FOV 96 mm², matrix 288 × 288, slice 150 × 0.5 mm) and SWI (TR/TE 32.6/20 mm, FA 15°, FOV 120 mm², matrix 240 × 240, slice 50 × 1 mm) sequences were acquired. Contrast agent gadobenate dimeglumine (0.2 mL/kg, Bracco) was i.v. administered when needed.

Fusion of PET, CT and MR images, and extraction of radiomic features were performed by PMOD 3.7 (PMOD Technologies).

2.11. Positron emission tomography/computed tomography

Monkeys were i.m. sedated with ketamine (5 mg/kg) and maintained with isoflurane (1% in oxygen) inhalation during PET/CT (80 kV, 160 mAs, FOV 300 mm², matrix 600 × 600, slice 342 × 1.4 mm, from calvarium to the hypogastrium) with an animal PET/CT scanner (miniEXPLORER, United Imaging). In each paired session, saline or CNO (10 mg/kg) was i.v. injected 10 min before a 75 min dynamic scan using [¹⁸F]FDG (0.30±0.02 mCi/kg) as radioactive contrast agent. Monkeys were allowed to cage in a separate room for 1 day between saline and CNO session.

PET data were analyzed by PMOD 3.7. Volumes of interests (VOIs) were manually delineated in a double-blinded manner by placing a 3 mm diameter circle into the transplanted brain area based on fused T1-weighted MR and PET/CT images. Reconstructed PET images acquired between 45 and 60 min after saline or CNO injection were used for the calculation of the radioligand SUV.

2.12. Electroencephalogram

An animal electrophysiological amplified system (Bio-Signal Technologies, Medusa) was used for EEG recording. Monkeys were i.m. sedated with ketamine (3 mg/kg) and be seated in a primate chair (Crist Instrument, ConformaMax) for 1 h before EEG recording. The monkeys underwent 3–5 residential training sessions to ensure tolerance during the recording procedures.

After shaving and cleaning, 10 electrodes were placed on the monkey's scalp with medical EEG conductive paste (Greentek, GT20) according to 10–20 system. Reference electrodes were placed at Cz, A1 and A2 while the ground electrode was placed at FPz. Recording electrodes were placed at FP1, FP2, C3, C4, O1 and O2. The sampling rate was 1000 Hz. Resting EEG recordings consisted of 2 sessions: a 45 min saline session followed by a 180 min CNO session (10 mg/kg CNO i.v. injected after the completion of saline session recording). Fruit and juice were given as rewards during the session breaks, and shampoo cleaning was given after recording.

EEG data were preprocessed and analyzed using MATLAB codes, EEGLAB GUI and its extensions [35,36]. Channel location, high-pass filtration, intervals rejection and low-pass filtration were applied sequentially. Artifacts in EEG traces caused by animal locomotion (mainly EMG and EOG) were manually removed by inspecting video records and high-amplitude waveforms. ICA (independent component analysis) algorithm was then used to further prune the data by removing artifacts. A basic FIR (finite impulse response) filter was applied to filter EEG data (1–30 Hz filtered for spectral analysis and clinical diagnosis, 80–250 Hz with 100 Hz notch filtered for epileptic spike and HFOs (high-frequency oscillation)) analysis. 1–30 Hz EEG power spectrum density were computed by 1-second FFT (fast Fourier transformation) window length with 50% overlap and 0.2 Hz frequency resolution. Frequency bands were summed in delta (1–4 Hz), high delta (2–4 Hz), theta (4–8 Hz), alpha (8–13 Hz) and beta (13–30 Hz) to obtain normalized relative power statistics. Firstly, the preprocessed 1–30 Hz EEG data were examined by an experienced clinician for epileptic spike recognition. Then, the preprocessed 80–250 Hz EEG data were analyzed by EPINETLAB through searching epileptic spikes under the criteria of ± 200 µV absolute amplitude shift from the baseline and a duration of > 5 s for tracing potential seizure onset zones (SOZs). In addition, 80–250 Hz spectrograms were computed by 1-second cmor (complex Morlet wave transformation) window length with no overlap in all channels. HFOs were defined as time duration > 6 ms with deflections > 5 SD above the baseline amplitude of root mean square and computed using Staba's method with 3 ms window length in all channels [37].

2.13. Statistical analysis

Results were presented as mean ± SEM (standard error of the mean). All statistical analyses and data plotting were performed by GraphPad Prism (version 9.4.1 for Windows, GraphPad Software, SD, CA). Differences were considered significant when p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Statistical tests employed for each analysis were indicated in figure legends.

3. Results

3.1. Genetic engineering of rhPSCs for noninvasive survey and remote control

The chemogenetic technology DREADD has emerged as a powerful approach for transient manipulation of neuronal activities in rodents. After systemic delivery, the “designer drug” clozapine N-oxide (CNO) is in vivo converted into clozapine, which could pass through the blood-brain barrier and act on DREADD to manipulate neuronal activities [38]. While the bioactive signals elicited by CNO could be captured by neuroimaging or EEG, we hypothesized that the ectopic expression of stimulatory DREADD in transplants would serve as a way to on purpose light on transplanted neurons and facilitate full-cycle transplantation survey.

To this end, we engineered rhPSCs through CRISPR/Cas9-mediated homologous recombination and integrated the mCherry or hM3Dq-mCherry expressing cassette into the rhesus monkey AAVS1 (rhAAVS1) safe harbor [39]. Both rhesus monkey ESCs (rhESCs) and rhesus monkey iPSCs (rhiPSCs) were electroporated with plasmids harboring CAG-Cas9-2A-EGFP and rhAAVS1 sgRNA expressing cassette together with the respective donor plasmid mCherry (5′arm-SA-puro-CAG-mCherry-3′arm) or DREADD (5′arm-SA-puro-CAG-hM3Dq-mCherry-3′arm) (Fig. 1a). After drug selection for 5–7 days, individual colonies were expanded and validated by genomic DNA PCR and Southern blot (Figs. S1a-b). Heterozygotes mCherry-rhPSCs and DREADD-rhPSCs showed comparable growth behavior with uniform expression of NANOG and OCT4, indicating their normal self-renewal capability (Fig. 1b). In addition, mCherry was distributed in both nucleus and cytosol in mCherry-rhPSCs, while hM3Dq-mCherry fusion protein was largely localized at the cell membrane in DREADD-rhPSCs (Fig. 1b). Both mCherry- and DREADD-rhPSCs had normal karyotype (Fig. S1c). Together, stimulatory DREADD-rhESCs and -rhiPSCs are successfully engineered through homologous recombination. These engineered cells could self-renew normally and could therefore be used to study their value in tracing and controlling allografted neurons in rhesus monkeys.

3.2. Stimulatory DREADD-rhPSCs effectively differentiate into cortical progenitors and glutamatergic neurons in vitro

To study their differentiation potential, mCherry- and DREADD-rhPSCs were subjected to cortical differentiation using standard protocols [40]. By day 17, neural epithelia organized into rosette-like neural tubes were observed in both mCherry- and DREADD-rhPSC cultures (Fig. 1c). Immunofluorescence staining revealed the uniform expression of SOX2, PAX6, NESTIN and N-Cadherin in the neural epithelia of both groups, indicating their cortical progenitor identity and synchronous differentiation (Fig. 1c-d). Quantification revealed that over 80% of cells were labelled with proliferative marker Ki67, indicating the proliferative state of the cortical progenitor in both groups at this differentiation stage (Fig. 1d). By day 30, the majority of cells were positive for neuronal marker Tuj1, indicating their neuronal identity (Fig. 1e-f). Meanwhile, over 90% of cells expressed the forebrain marker FOXG1 and the glutamatergic neuron marker vGluT1, highlighting their fate as cortical neurons (Fig. 1e-f). No GABAergic (GABA⁺), cholinergic (ChAT⁺), midbrain (FOXA2⁺) or spinal cord (HOXB4⁺) neurons were identified in the culture, which further supported their regional cortical identity (Fig. S1d). Throughout all differentiation stages, mCherry signals were detected with proper localization in mCherry- and DREADD-rhPSC derivated cortical progenitors and neurons, suggesting stable expression of the engineered gene within the rhAAVS1 safe harbor. These results suggest that engineered rhPSCs with build-in DREADD expression element within the rhAAVS1 locus preserve their typical neuronal differentiation potencies.

3.3. DREADD-glutamatergic neurons respond to CNO in vitro

The electrophysiological features of cortical neurons derived from mCherry- and DREADD-rhPSCs were characterized in 8–10 week-old neurons (12–14 weeks after differentiation). There were no obvious differences in the resting membrane potentials (RMPs) between the groups (Fig. 1g). Both groups were able to fire repeated mature APs in response to depolarizing current injections under standard current-clamp recording protocols [41]. The membrane potentials (MP) of cortical neurons derived from both mCherry- and DREADD-rhPSCs reached above 0 mV following 30 pA current injection with rapid depolarization and rapid repolarization (Figs. S1e-i). In addition, they displayed similar firing frequencies in response to current steps from 0 to 80 pA (Fig. S1j). These results indicate that glutamatergic neurons derived from both mCherry- and DREADD-rhPSCs mature equally in their electrophysiological properties.

Perfusion of 10 µM CNO elicited the depolarization of RMPs in DREADD-neurons, from an average of −48.56 ± 1.29 mV to −41.96 ± 1.47 mV, accompanied by increased subthreshold membrane potential oscillation (SMPO) (Fig. 1h-i). After CNO perfusion, no depolarization of RMPs was observed in mCherry-control group (Fig. 1h-i), which was in consistent with that of cortical neurons derived from human ESCs at the similar stage [42]. Under voltage-clamp mode with a holding potential of −70 mV, the frequency of EPSCs increased from 5.4 ± 0.92 to 14.3 ± 1.77 per 10 s and its amplitude increased from 52.16 ± 2.85 pA to 60.16 ± 1.44 pA in DREADD neurons after CNO perfusion (Fig. 1j-l). Meanwhile, the frequency of EPSCs timely recovered upon CNO washout (Fig. 1j-k). Again, no changes in EPSC frequency and amplitude were observed in mCherry neurons after CNO perfusion (Fig. 1j-i). These results suggest that glutamatergic neurons generated from engineered DREADD-rhPSCs functionally respond to the designer drug CNO at both neuronal and synaptic levels.

3.4. Capturing of CNO elicited responses in DREADD-rhPSC-derived brain allografts via [¹⁸F]FDG-PET and EEG

Chen et al. [18] successfully transplanted human midbrain dopaminergic neurons differentiated from human PSCs engineered with hM3Dq-expressing cassette into the rodent brain, and found that systemic delivery of CNO enhanced the function of grafted neurons. To date, it remains unknown whether the CNO-hM3Dq stimulatory DREADD system can be applied in NHPs for on purpose control of neuronal activities. If applicable, the administration of CNO could also be used to finely tune the neuronal activities of brain-transplanted DREADD-expressing neurons in NHPs. Meanwhile, clinically accessible methods such as PET and EEG could be applied to capture the neuronal activities elicited by DREADD agonist CNO and therefore offer a noninvasive solution for full-cycle graft survey.

Four male rhesus monkeys aged 6–7 (Fig. S2a) were rendered to MR imaging in a 3.0 Tesla MR scanner and brain anatomical structures were referenced [43]. Subsequently, the monkeys received [¹⁸F]-fluorodeoxyglucose ([¹⁸F]FDG) administration and PET was performed to study the baseline [¹⁸F]FDG uptake after saline or 10 mg/kg CNO intravenous (i.v.) injection [44], [45], [46], [47], [48]. The fusion of T1-weighted MR images with the corresponding reconstructed PET data showed relatively low [¹⁸F]FDG baseline uptake in Brodmann area (BA) 9/46 of the bilateral dorsolateral prefrontal cortices in rhesus monkeys (Fig. 2a). CNO i.v. injection exhibited no effect on baseline [¹⁸F]FDG uptake as compared with the saline injection control in this area (Fig. 2a-b). We therefore selected BA 9/46 on both sides for transplantation.

Fig. 2.

Capturing the maturation and CNO-elicited responses in DREADD-rhPSC-derived neurons via [¹⁸F]FDG-PET and EEG after rhesus monkey allotransplantation. (a) Representative coronal (COR) T1-weighted MR and [¹⁸F]FDG-PET images of bilateral BA 9/46 in the rhesus monkey before transplantation (monkey #T1467001). PET was performed 10 min after i.v. injection of either saline or CNO (10 mg/kg). Dotted circles represent regions used for cell transplantation on both sides. Color scale, standard uptake value (SUV). R, right; L, left; H, head; F, foot. Scale bars, 10 mm. (b) Quantification of SUV ratio (SUVR, relative to cortex BA 10 control region) in (a). For saline group, 96.96% ± 0.89%. For CNO group, 96.64% ± 0.98%. Data are presented as mean ± SEM, n = 8 scans (obtained from 8 regions in BA 9/46 of monkeys #T1563105, #T1408071, #T1467001 and #T1403009 before transplantation), paired t-test. (c) Representative coronal and axial (AXI) [¹⁸F]FDG-PET images in saline or CNO i.v. injection condition in a rhesus monkey transplanted with mCherry-rhPSC (right side)- and DREADD-rhPSC (left side)-derived cortical progenitors (monkey #T1408071, 6 months after transplantation). A, anterior; P, posterior. Scale bar, 10 mm. (d) Quantification of SUVR in (c) For saline group vs. CNO group in mCherry-neuron transplantations, 98.10 ± 0.76% vs. 98.78 ± 0.74%, n = 14 scan pairs (obtained from 2 regions in BA 9/46 of monkeys #T1563105, #T1408071). For saline group vs. CNO group in DREADD-neuron transplantations, 95.54% ± 0.48% vs. 101.53% ± 0.61%, n = 25 scan pairs (obtained from 4 regions in BA 9/46 of monkeys #T1563105, #T1408071, #T1467001 and #T1403009). Data are presented as mean ± SEM. Paired t-test was used for comparisons between saline-CNO treatment paired sessions, ****, p < 0.0001. Ordinary one-way ANOVA with Turkey test was used for comparisons between mCherry and DREADD groups, ^##, p < 0.01. (e) Quantification of normalized SUVR ($\frac{{\text{SUV}}_{\text{CNO}}-{\text{SUV}}_{\text{saline}}}{{\text{SUV}}_{\text{saline}}}$) within saline-CNO treatment paired sessions. For nonoperative group, −0.33 ± 0.49%, n = 8 scan pairs (obtained from 4 regions in BA 9/46 of monkeys #T1563105, #T1408071, #T1467001 and #T1403009). For mCherry group, 0.71 ± 0.40%, n = 14 scan pairs (obtained from 2 regions in BA 9/46 of monkeys #T1563105, #T1408071). For DREADD group, 6.28±0.50%, n = 25 scan pairs (obtained from monkeys #T1563105, #T1408071, #T1467001 and #T1403009). Data are presented as mean ± SEM, ordinary one-way ANOVA with Turkey test, ****, p < 0.0001. (f) Representative axial T1-weighted MR image and time course of [¹⁸F]FDG-PET images of bilateral BA 9/46 after CNO i.v. injection in a rhesus monkey transplanted with mCherry-rhPSC (right side)- and DREADD-rhPSC (left side)-derived cortical progenitors (monkey #T1563105). Scale bar, 10 mm. (g) Time course of normalized SUVR change in (f). For mCherry group, n = 2 regions (obtained from monkeys #T1563105 and #T1408071). For DREADD group, n = 4 regions (obtained from monkeys #T1563105, #T1408071, #T1467001 and #T1403009). Data plotted at the top panel are presented as mean (line) ± SEM (shade). Ordinary two-way ANOVA, ****, p < 0.0001. Data plotted at the bottom panel are individual SUVR change of both groups. (h) Schematic representation of the scalp electrode positions marked on the rhesus monkey head for EEG recording. 10 electrodes were placed on monkey scalp with medical EEG conductive paste according to 10–20 system. Reference electrodes were placed at Cz, A1 and A2 while the ground electrode was placed at FPz. Recording electrodes were placed at FP1, FP2, C3, C4, O1 and O2. (i) Normalized EEG power spectrum density (PSD) recorded by scalp electrodes close to the mCherry-graft (obtained from monkeys #T1563105 and #T1408071) and DREADD-graft (obtained from monkeys #T1563105, #T1408071, #T1467001 and #T1403009) in saline or CNO i.v. injection condition. (j) Quantification of the power in (i) summed in high delta (2–4 Hz), theta (4–8 Hz), alpha (8–13 Hz) and beta (13–30 Hz) frequency bands. For mCherry group, before vs. after CNO perfusion, high delta, 16.15% ± 1.41% vs. 15.64% ± 1.25%; theta, 17.87% ± 2.21% vs. 16.06% ± 1.67%; alpha, 15.48% ± 2.34% vs. 14.94% ± 2.18%; beta, 30.16% ± 2.46% vs. 31.10% ± 2.14%; n = 5 records (obtained from monkeys #T1563105, #T1408071). For DREADD group, before vs. after CNO perfusion, high delta, 16.92% ± 0.48% vs. 14.83% ± 0.54%; theta, 17.27% ± 1.51% vs. 17.30% ± 1.25%; alpha, 14.50% ± 1.52% vs. 16.29% ± 1.37%; beta, 27.48% ± 0.86% vs. 34.51% ± 1.30%; n = 9 records (obtained from monkeys #T1563105, #T1408071, #T1467001 and #T1403009). Data are presented as mean ± SEM, paired t-test, *, p < 0.05, **, p < 0.01. (k) Representative EEG power spectra of FP1, FP2, C3, C4, O1, O2 channels in 1–30 Hz frequency band in saline or CNO i.v. injection condition in a rhesus monkey transplanted with mCherry-rhPSC (right side, close to FP2 electrode)- and DREADD-rhPSC(left side, close to FP1 electrode)-derived cortical progenitors (monkey #T1408071, 9 months after transplantation).

Cortical progenitors derived from mCherry- and DREADD-rhPSCs 6–7 weeks after differentiation were transplanted into bilateral BA 9/46 of 4 rhesus monkeys by stereotaxic injection (Fig. S2a). Each side of transplantation involved 3 injection tracks in a 5 mm³ area region with 4 deposit spots in each track, and a total of 320,000 cells were injected per deposit spot. One month after transplantation, MR imaging was conducted to view the injection tracks, and the transplantations in BA 9/46 in rhesus monkeys were confirmed (Fig. S2b). PET was performed every 2–3 months to assess [¹⁸F]FDG uptake in both saline and CNO conditions. No obvious differences in [¹⁸F]FDG uptake were observed between mCherry- and DREADD-grafts in saline condition. However, an increase elicited by CNO in [¹⁸F]FDG uptake could be observed in DREADD- but not mCherry-grafts 6 months after transplantation (Fig. 2c-e). In addition, CNO elicited increase in [¹⁸F]FDG uptake in DREADD-grafts usually became visible 3 months and reached a plateau 9–12 months post-transplantation (Figs. 2f-g). These results suggest that the CNO-hM3Dq stimulatory DREADD system can also be applied in NHPs and allowing remote control of neuronal activity through systemic delivery of the designed drug CNO. Furthermore, our results also suggest that [¹⁸F]FDG-PET is practical for visualization of CNO elicited DREADD activation in brain-grafted cortical neurons, thus serving as an noninvasive way for graft survey.

Another noninvasive approach for probing brain functional states is EEG. Resting-state EEG recordings were performed in rhesus monkeys with cortical progenitor transplantation for 9 months. Analysis of the brain power spectrum density (PSD) in the frequency band of 1–30 Hz revealed an increase in beta power (13–30 Hz) and a decrease in high delta power (2–4 Hz) in the scalp electrode close to the DREADD grafts following CNO injection (Fig. 2h-j). These results suggest that delivery of CNO leads to local activation of grafted cortical neurons harboring stimulatory DREADD, and the response of DREADD-grafts to CNO might promote an alert state in awake rhesus monkeys. In addition, the spectrum power transition induced by CNO could only be clearly observed in the prefrontal electrode close to the side of DREADD-rhPSC-derived grafts, suggesting a tight correlation between the induced responses and the grafted hM3Dq expressing neurons (Fig. S2c and Fig. 2k).

Together, these results suggest that the CNO-hM3Dq stimulatory DREADD system is functional in NHPs. Transplanted cortical progenitors mature gradually into functional neurons and systemic delivery of CNO could timely activate the grafted hM3Dq expressing neurons, offering a way to purposely control the activity of transplanted neurons. Moreover, the widely applied clinical diagnosis methods [¹⁸F]FDG-PET and EEG could efficiently capture neuronal activation elicited by CNO, highlighting their robustness in full-cycle survey of brain grafts in noninvasive manners.

3.5. Rhesus monkeys bearing CNO-hM3Dq survey and control system do not have serious adverse events

All 4 rhesus monkeys showed no signs of fever or obvious abnormalities in routine blood tests throughout 26 months post-transplantation. None of the monkeys showed appetite or weight loss. MR imaging showed no overgrowth at the transplanted regions and no bleeding or edema occurred in the brain (Fig. 3a).

Fig. 3.

Rhesus monkeys show no serious adverse events after allotransplantation and repeated DREADD activation. (a) Axial, coronal and sagittal (right/left SAG) T1-weighted and T2-weighted whole brain MR images of bilateral BA 9/46 (obtained from monkeys #T1563105, #T1408071 and #T1467001, 18 months after transplantation). Monkey #T1403009 is not suitable for MR imaging given he has several suspected subcutaneous metallic implants. Scale bars, 10 mm. (b) Schematic overview of in vivo study design showing the executing time point of CNO delivery, PET and EEG. (c) Representative 60 s EEG traces (80–250 Hz) of FP1, FP2, C3, C4, O1, O2 channels and corresponding time-frequency analysis (TFA, complex Morlet wave transformation) in saline or CNO i.v. injection condition. Naïve plot is obtained in a rhesus monkey with no surgical operation or transplantation. 9 m and 16 m plots are obtained in a rhesus monkey transplanted with mCherry-rhPSC (right side, close to FP2 electrode) and DREADD-rhPSC (left side, close to FP1 electrode)-derived cortical progenitors (monkey #T1408071, 9 and 16 months after transplantation). (d) Quantification of HFO frequency based on 80–250 Hz resting EEG of FP1, FP2, C3, C4, O1, O2 channels. For saline group vs. CNO group in experimental naïve monkeys, 1.00 ± 0.60 vs. 1.67 ± 0.67, n = 9 (obtained from 3 rhesus monkeys with no surgical operation or transplantation). For saline group vs. CNO group in DREADD-neuron transplanted monkeys, 2.11 ± 0.63 vs. 0.89 ± 0.54 in 9 m group and 1.22 ± 0.64 vs. 1.44 ± 0.60 in 16 m group; n = 9 for both groups (obtained from monkeys #T1563105, #T1408071 and #T1467001). Data are presented as mean ± SEM, ordinary one-way ANOVA with Turkey test. (e) Duration quantification of every HFO counted in (d). For saline group vs. CNO group in experimental naïve monkeys, 42.33 ± 3.41 ms vs. 39.00 ± 3.50 ms. For saline group vs. CNO group in DREADD-neuron transplanted monkeys, 35.21 ± 3.12 ms vs. 38.63 ± 5.05 ms in 9 m group and 37.91 ± 2.30 ms vs. 33.46 ± 2.14 ms in 16 m group. Data are presented as mean ± SEM, ordinary one-way ANOVA with Turkey test. (f) HFO duration summed in (e) distributed in 10 ms bins.

Synchronized activation of glutamatergic neurons in a limited region is responsible for rhythm generation in epileptic seizures [49,50]. Given that CNO was repeatedly employed in this study, it was necessary to examine whether severe adverse events occurred following repeated stimulation of grafted neurons. To study whether the grafted neurons underwent pathological transformation to form a potential seizure onset zone (SOZ), epileptic spikes and high-frequency oscillations (HFOs) were scrutinized and extracted from the resting EEG data recorded from 6 scalp electrodes (Fig. 3b) [51]. Despite repeated CNO delivery, no epileptic spasms were observed in grafted monkeys by video recording or direct observation, and no epileptic discharges were detected across the 6 scalp electrodes (Fig. 3c). The frequently and rhythmically occurrence of 80–250 Hz HFOs (ripples) are considered to have an apparent link to epileptogenicity [52,53]. More than 12 counts of HFOs per minute are referred to be a reliable indicator of SOZ [54,55]. In all cases, HFOs occurred independently with 6 counts or less within a 15 min period (Fig. 3d), suggesting a nonSOZ state. The duration and distribution features of independent HFOs in the transplanted group of DREADD-rhPSC derivates following repeated CNO delivery were similar to those observed in naive group without transplantation (Figs. 3e-f). These results suggest that transplantation of cortical neural progenitors engineered with build-in DREADD elements and repeated CNO activation are safe and unlikely to induce tumor growth or epilepsy.

3.6. Constructing suicide gene-based safeguard system in rhPSCs to ensure safe brain transplantation

Suicide gene approaches can eliminate unwanted cells after transplantation [20,21]. We integrated HSVtk-expressing cassette into the rhAAVS1 locus of rhPSCs via CRISPR/Cas9 based homologous recombination (Figs. S1b, S3a-b). The HSVtk-expressing rhPSCs were then infected with lentiviruses to express hM3Dq-mCherry stably (Fig. S3d-h). RhPSCs with combined expressions of HSVtk and hM3Dq-mCherry (referred as HSVtk-rhPSCs hereafter) had normal karyotype and typically expressed the pluripotent markers OCT4 and NANOG (Fig. S3c-d). Meanwhile, HSVtk-rhPSCs could be differentiated equally into cortical progenitors and glutamatergic neurons with stable membrane hM3Dq-mCherry expression at a comparable level to that of DREADD-rhPSCs derived neurons (Fig. S3e-i).

HSVtk-rhPSCs were significantly eliminated in a dose dependent manner after a 48 h challenge with the suicide gene inducer Ganciclovir (GCV) (Fig. 4a-c). Applying GCV in wild-type rhPSCs showed no effect, highlighting the specificity of the GCV/HSVtk suicide system (Fig. S4a-c). Meanwhile, applying GCV in HSVtk-rhPSC-derived neuronal cultures for 10 days at a dosage of 10 µΜ resulted in no changes in neuronal morphology or density (Fig. 4d). Immunofluorescence staining revealed a consistent intensity of the Tuj1 signals before and after GCV treatment (Fig. 4e), while the proportion of Ki67⁺ cells was almost completely eliminated by GCV (Fig. 4f). These results indicate that the GCV/HSVtk suicide system effectively eliminates the proliferating cells while largely preserving postmitotic neurons, representing a practical approach for safeguarding neuronal transplants.

Fig. 4.

Generation andin vitrocharacterization of HSVtk-rhPSCs for safety control. (a) Alkaline phosphatase live staining of HSVtk-rhPSC colonies upon 48 h treatment of GCV at different concentrations. Scale bar, 20 mm. (b) Quantification of colony area proportion in (a). 0 µM, 44.98% ± 2.13%; 4 µM, 35.18% ± 1.58%; 40 µM, 27.22% ± 0.81%; 400 µM, 22.50% ± 1.28%; 4000 µM, 16.16% ± 0.55%. Data are presented as mean ± SEM, n = 5 for each group, ordinary one-way ANOVA with Dunnett test, ***, p < 0.001, ****, p < 0.0001, compared with 0 µM group. (c) Quantification of colony size in (a). 0 µM, 2.97 ± 0.14 mm²; 4 µM, 2.12 ± 0.09 mm²; 40 µM, 1.68 ± 0.08 mm²; 400 µM, 1.50 ± 0.10 mm²; 4000 µM, 1.19 ± 0.08 mm². Data are presented as mean ± SEM, n = 20 for each group, ordinary one-way ANOVA with Dunnett test, ****, p < 0.0001, compared with 0 µM group. (d) Representative immunofluorescence images of Tuj1 and Ki67 staining in 40 days HSVtk-rhPSC-derived cortical neurons untreated or treated with 10 µM GCV for indicated days. Scale bars, 20 µm. (e) Quantification of relative Tuj1 fluorescence intensity in (d). For untreated control group, 1.00±0.03. For 5-day treated group, 0.98±0.10. For 10-day treated group, 0.96 ± 0.08. Data are presented as mean ± SEM, n = 10 for each group, ordinary one-way ANOVA with Turkey test. (f) Quantification of the percentage of Ki67⁺ cells in (d). For untreated control group, 22.24% ± 1.61%. For 5-day treated group, 13.90% ± 1.55%. For 10-day treated group, 0.70% ± 0.23%. Data are presented as mean ± SEM, n = 10 for each group, ordinary one-way ANOVA with Turkey test, ***, p < 0.001, ****, p < 0.0001. .

To evaluate the GCV/HSVtk suicide system in vivo, HSVtk-rhPSC-derived cortical progenitors were also transplanted into BA 9/46 and undifferentiated HSVtk-rhPSCs were transplanted intramuscularly into monkey deltoid for teratoma formation [56] (Fig. 5a). Similar to DREADD-rhPSC-derived grafts, HSVtk-rhPSC-derived brain grafts exhibited an increase in CNO elicited [¹⁸F]FDG uptake and the power of beta frequency band, as well as a decrease in the power of the high delta frequency band in PET and EEG analyses (Fig. 5b-h), suggesting gradual maturation of HSVtk-neurons after transplantation. Furthermore, neither epileptic discharge nor changes in durations and distribution features of HFOs were caused by repeated CNO exposure (Fig. S3j-m). These results further suggest the safety of the build-in stimulatory DREADD and HSVtk elements.

Fig. 5.

The suicide system preserves CNO-elicited response of DREADD neuronsin vivo. (a) Schematic overview of transplantation strategy for rhPSCs or rhPSC-derived cortical progenitors bearing suicide system, the time duration of daily GCV (10 mg/kg) i.v. drip, and the executing time point of PET, MR imaging and EEG. (b) Representative coronal and axial [¹⁸F]FDG-PET images in saline or CNO i.v. injection condition in a rhesus monkey transplanted with HSVtk-rhPSC (right side)- and DREADD-rhPSC (left side)-derived cortical progenitors (monkey #T1467001, 14 months after transplantation). Scale bar, 10 mm. (c) Time course of [¹⁸F]FDG-PET images of bilateral BA 9/46 in saline or CNO i.v. injection condition in a rhesus monkey (monkey #T1403009) transplanted with HSVtk-rhPSC (right side)- and DREADD-rhPSC (left side)-derived cortical progenitors before (16 months after transplantation) and 1 or 2 months after GCV treatment. Scale bar, 10 mm. (d) Quantification of SUVR in HSVtk-neuron transplantations. For saline group vs. CNO group before GCV treatment, 97.40% ± 0.98% vs. 102.52% ± 1.09%, n = 8 scan pairs (obtained from 2 regions in BA 9/46 of monkeys #T1467001 and #T1403009). For saline group vs. CNO group after GCV treatment, 95.28% ± 0.87% vs. 102.61% ± 1.42%, n = 4 scan pairs (obtained from 1 region in BA 9/46 of monkey #T1403009). Data are presented as mean ± SEM. Paired t-test was used for comparisons between saline-CNO treatment paired sessions, *, p < 0.05, ***, p < 0.001. Ordinary one-way ANOVA with Turkey test was used for comparisons between untreated and GCV treated groups. (e) Quantification of normalized SUVR within saline-CNO paired sessions. For the group before GCV treatment, 5.28% ± 0.83%, n = 8 scan pairs (obtained from 2 regions in BA 9/46 of monkeys #T1467001 and #T1403009). For the group after GCV treatment, 6.04% ± 1.23%, n = 4 scan pairs (obtained from 1 region in BA 9/46 of monkey #T1403009). Data are presented as mean ± SEM, unpaired t-test. (f) Normalized spectrum power recorded by scalp electrodes close to the HSVtk-graft before (obtained from monkeys #T1467001 and #T1403009) and after (obtained from monkey #T1403009) GCV treatment in saline or CNO i.v. injection condition. (g) Quantification of normalized spectrum power in (f) summed in high delta, theta, alpha and beta frequency bands. For the group before GCV treatment, before vs. after CNO perfusion, high delta, 17.66% ± 1.24% vs. 15.24% ± 0.67%; theta, 16.56% ± 1.20% vs. 17.95% ± 1.07%; alpha, 14.43% ± 1.17% vs. 18.03% ± 0.68%; beta, 26.93% ± 2.56% vs. 33.28% ± 1.61%; n = 8 records (obtained from monkeys #T1467001 and #T1403009). For the group after GCV treatment, before vs. after CNO perfusion, high delta, 19.74% ± 1.63% vs. 13.58% ± 1.30%; theta, 14.38% ± 0.33% vs. 15.49% ± 0.65%; alpha, 11.44% ± 0.98% vs. 15.60% ± 1.21%; beta, 27.51% ± 3.40% vs. 36.91% ± 3.82%; n = 8 records (obtained from monkey #T1403009). Data are presented as mean ± SEM, paired t-test, *, p < 0.05, **, p < 0.01. (h-i) Representative EEG power spectra of FP1, FP2, C3, C4, O1, O2 channels in 1–30 Hz frequency band in saline or CNO i.v. injection condition in a rhesus monkey (monkey #T1403009) transplanted with HSVtk-rhPSC (right side, close to FP2 electrode)- and DREADD-rhPSC (left side, close to FP1 electrode)-derived cortical progenitors before ((h), 16 months after transplantation) and 2 months after (i) GCV treatment.

16 months after transplantation, daily i.v. drip of 10 mg/kg GCV was given to rhesus monkeys transplanted with DREADD-rhPSC- and HSVtk-rhPSC-derived cortical progenitors for 2 months, which was used to activate the suicide system in NHPs according to the pharmacokinetics of GCV with no hematological or organ toxicity (Fig. 5a) [57,58]. Of note, an increase in [¹⁸F]FDG uptake and beta power, as well as a decrease in high delta power upon CNO treatment were maintained even after long-term GCV treatment (Fig. 5c-g and 5i), implying that GCV had no effect on the function of transplanted postmitotic neurons.

In deltoid-transplanted rhesus monkeys, MR imaging represented the presence of rhPSC grafts as a relatively low signal in T1-weighted sequences with a clear boundary visible on axial, coronal, and sagittal planes (Fig. 6a-b). We identified 15 rhPSC transplants out of 30 injections, which showed continuous enlargement in size (Figs. 6c-d and S3n), suggesting their active proliferating states. MR images of HSVtk-rhPSCs deltoid grafts showed a gradual decrease in volume and became almost invisible 1.5 months after GCV exposure (Fig. 6c-e). These results suggest that the build-in GCV/HSVtk suicide system in rhPSCs can effectively eliminate proliferating cells in transplanted tissues. Importantly, this system will not eliminate postmitotic neurons, enabling the maintenance of reconstructed neuronal networks and functions even if the safeguard system is switched on to eliminate unwanted proliferating cells in brain grafts.

Fig. 6.

Activation of the HSVtk/GCV suicide system reduces the size of mitotic allografts. (a) Schematic overview of the MR imaging coordinates used in monkey deltoid scan. (b) Representative axial, coronal and sagittal T1-weighted MR images and reconstructed 3D view of a deltoid graft (obtained from monkey #T1563105 on the left deltoid, 12 months after transplantation). Dotted circle delineates the region of the graft and enlarged images corresponding to the graft area respectively were presented at the lower panel. Scale bars, 10 mm. (c) Time course of a deltoid graft estimated from sagittal T1-weighted MR images (obtained from monkey #T1408071 on the right deltoid). Dotted circle delineates the region of the graft and enlarged images corresponding to the graft area respectively were presented at the lower panel. Scale bars, 10 mm. (d) Growth kinetics of HSVtk-rhPSCs graft volume in allografted monkeys before GCV treatment. n = 10 (obtained from monkeys #T1563105 and #T1408071). (e) Growth kinetics of HSVtk-rhPSCs graft volume in allografted monkeys after saline or GCV (10 mg/kg) i.v. drip. For saline group (n = 6, obtained from monkey #T1563105) vs. GCV group (n = 4, obtained from monkey #T1408071), 676.48 ± 69.73 mm³vs. 663.64 ± 108.08 mm³ in 0 m group; 697.15 ± 69.86 mm³vs. 501.54 ± 98.05 mm³ in 0.5 m group; 705.55 ± 51.40 mm³vs. 326.51 ± 72.27 mm³ in 1 m group; 733.84 ± 58.34 mm³vs. 135.00 ± 24.29 mm³ in 1.5 m group. Data are presented as boxplot with individual points, unpaired t-test, *, p < 0.05, ***, p < 0.001. RM two-way ANOVA with sidak test, ^##, p < 0.01.

4. Discussion and conclusion

Integration of cell-based therapy and genetic engineering represents a powerful strategy for producing designed cell sources for safe, effective and controllable cell transplantation. In this study, we engineered rhPSCs with build-in stimulatory DREADD and HSVtk suicide gene system. These engineered cells maintained their self-renewal capacity and could be differentiated into functional cortical glutamatergic neurons. After being allotransplanted into the dorsolateral prefrontal cortices of rhesus monkeys, the transplanted DREADD-neurons were able to be activated by systemic delivery of CNO, which the bioactive signal can be captured timely by [¹⁸F]FDG-PET and EEG. Postmitotic neurons with HSVtk expression in the brain retained their functions in the presence of GCV, although the activation of the suicide system successfully depleted dividing cells in peripheral transplants. Our strategy in engineering pluripotent stem cells with build-in chemogenetic and suicide systems would therefore offer an approach to unlimitedly produce designed transplantable neural progenitors for controllable cell therapy.

Although pluripotent stem cells hold great promise in cell-based replacement therapy, safety concerns have been a major hurdle before pluripotent stem cells enter clinical reality. In our current study, no uncontrolled overgrowth was observed in all brain allografts in 26 months after transplantation. In vitro validation experiments revealed high differentiation efficiency and consistency of over 90% cortical glutamatergic neurons derived from both rhESCs and rhiPSCs batch by batch. Thus, an efficient and synchronized lineage differentiation would improve the safety of cell therapy [59,60]. De Luzy et al. [21] revealed that activation of the suicide system by systemic GCV administration in a rat Parkinsonian model transplanted with HSVtk expressing dopaminergic progenitors in the striatum successfully led to effective elimination of proliferation cells without affecting the total yield and function of dopaminergic neurons. Here, we elucidate that the GCV/HSVtk suicide system is also functional in NHPs, which would therefore serve as a preset safeguard to ensure a safe transplantation and pave the road to move cell therapies forward into clinical reality.

Longitudinal monitoring of cell grafts and the assessment of their growth and function are essential for cell therapy [61]. Noninvasive diagnostic methods such as MR imaging, PET and EEG play an essential role in this regard. By engineering the build-in stimulatory DREADD expression system in pluripotent stem cells, we reveal the efficacy of the hM3Dq/CNO system for long-term monitoring of the function of brain-transplanted neurons when combined with [¹⁸F]FDG-PET and EEG. The development of selective and high-affinity DREADD radioligands further enables noninvasive and longitudinal DREADD detection in the central nervous system in both rodents and NHPs [62,63].

The ability to control the function of transplanted neurons opens up the possibility of refining the therapeutic outcome. Chen et al. [18,64] revealed that the DREADD/CNO system could be effectively employed for precise control over the activity of grafted cells in mice. Our results reveal that the DREADD system is functional in NHPs, and systemic delivery of CNO is effective to tune the function of brain-grafted DREADD-neurons. Of note, cortical glutamatergic neurons derived from both mCherry- and DREADD-rhPSCs, even in the presence of CNO, did not show spontaneous action potentials (sAPs). This is in line with the observations from cortical neurons differentiated from human embryonic stem cells (hESCs) [42]. The lack of sAP in cortical glutamatergic neurons is therefore a potential cellular mechanism for non-epileptic firing in DREADD-neuron brain grafted monkeys even after repeated CNO treatment, and strengthens the safety and reliability of the DREADD system for future clinical applications.

While this study mainly focused on the validation of the DREADD and HSVtk systems in combination with pluripotent stem cells for safe, visible and controllable brain transplantation, in the future, it will be extremely valuable to apply these systems in NHP disease models to fully evaluate their potentials in treating neurological disorders. Although we did not provide histological evidence regarding the survival of transplanted glutamatergic neurons before and after GCV treatment, the conclusion is largely based on functional data obtained from [¹⁸F]FDG-PET and EEG studies. It is crucial to strike a balance between scientific and ethical rigor when conducting research involving nonhuman primates. Based on these considerations, it is still our primary concern to largely imitate the actual clinical scenarios and minimize the replaceable sacrifice of rhesus monkeys.

Author contributions

J.C., B.Z. and W.C. contributed equally to this work. X.Z., H.C. and L.L. conceived and supervised the study. X.Z., J.C. and B.Z. designed and performed experiments. J.C. and B.Z. analyzed data. W.C. and S.Z. helped with PET/CT, MR imaging and drug administration, S.Z. also assisted EEG recording and analysis. X.Z. and X.D. supervised PET/CT, provided and evaluated the quality of [¹⁸F]FDG. B.W. operated craniotomy and supervised transplantation. Y.L. helped with drug administration. W.Z. supervised MR imaging. X.Z., J.C., H.C. and L.L. interpreted the results and wrote the manuscript. All authors commented on the manuscript.

Declaration of competing interest

The authors declare that they have no conflicts of interest in this work.

Biographies

Jiayi Cheng is a Ph.D. candidate at Tongji University, Medical School. Her research focuses on neural lineage differentiation from nonhuman primate and human pluripotent stem cells and generation of transplantable neural cells for treating neurological disorders.

Ling Liu (BRID: 08216.00.37538) is a full professor of regenerative medicine at Tongji University, Medical School. She is also an adjunct professor of Shanghai East Hospital and Shanghai Fourth Hospital. Dr. Ling Liu and her team focus on human neural development, neural lineage differentiation from human pluripotent stem cells and generation of transplantable neural cells for treating neurological disorders.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.