Ethics approval

All animal experiments were performed in accordance with relevant guidelines and regulations, either in an approved animal facility at Sangamo Therapeutics France (accredited by the French Ministry of Research, Arrêté n° 4566) according to the APAFIS (“Autorisation de Projet utilisant des Animaux à des Fins Scientifiques”)-approved protocol APAFIS#12909-2017121211476703 v4 and in compliance with the Guide for the Care and Use of Laboratory Animals, or at the University of British Columbia (UBC), approved by the UBC Animal Care Committee (A16-0300).

Experiments using human samples were performed in accordance with the Declaration of Helsinki and approved by appropriate ethics committees. Leukopaks for Treg isolation were sourced from HemaCare (Northridge, USA), an FDA-registered collection center, from healthy human volunteers who consented under an Institutional Review Board-approved protocol compliant with Code of Federal Regulations Title 21 Part 1271. Allogeneic T cells and peripheral blood mononuclear cells (PBMCs) were obtained from healthy donors following written consent according to protocols approved by the Etablissement Français du sang. Human skin discarded from plastic surgery was obtained from the Harvard Skin Resource Centre, Skin Works, or the Cambie Surgery Clinic according to protocols approved by the UBC Clinical Research Ethics Board (H16-02930).

Generation and propagation of Tregs with HLA-A*02-specific CAR

The humanized HLA-A*02 ScFv was fused to a transmembrane domain and a signaling domain composed of the intracellular domain of human CD28 and CD3ζ. The control CAR TX235 was composed of the humanized HLA-A*02 ScFv and transmembrane domain but lacked intracellular signaling domains CD28 and CD3ζ. The resulting cDNA was cloned into a lentiviral vector. The TX200 lentivirus was produced by Lentigen/Miltenyi (Gaithersburg, USA). The TX235 lentivirus was generated in house by transfecting adherent HEK293T cells (Lenti-X) with a lentiviral 4-plasmid system including human immunodeficiency virus type 1 (HIV-1) gagpol, HIV-1 Rev, envVSV-G, and the TX235 CAR transfer plasmid. A schematic of TX200 and TX235 is presented in the supplementary data (Fig. S1). Surface expression was determined by flow cytometry using fluorescent HLA-A*02 dextramer (#WB2666, Immudex, Denmark).

White blood cells were harvested from healthy HLA-A*02-negative donors using leukapheresis and were cryopreserved (HemaCare). The leukapheresate was thawed and cells were isolated after staining for CD4 (PE-Vio770, #130-093-142), CD45RA (FITC, #170-076-502), CD25 (PE, #170-076-505) and CD127 (APC, #170-076-501) with antibodies that were of Good Manufacturing Practice-grade, except CD4 PE-Vio770, (all Miltenyi Biotech) to isolate naïve Tregs defined as CD4+/CD45RA+/CD25high/CD127low using an SH800 cell sorter (Sony, Japan). The gating strategy is shown in the supplementary data (Fig. S2). Cells were then resuspended in X-VIVO-15 medium (#BEBP02-054Q, Lonza) with 1000 U/ml interleukin (IL)-2 (#2238131-A, Novartis Pharma) and activated with anti-CD3/CD28-coated dynabeads (#40203D, Life Technologies). After 72 h, cells were transduced with the lentiviral vector and further expanded in X-Vivo-15 medium with 1000 U/ml IL-2 and anti-CD3/CD28-coated dynabeads. Medium was refreshed with IL-2 every 2 days. Cells were expanded for less than 2 weeks to keep the cells in an exponential phase of growth and minimize loss of the naïve Treg phenotype. At harvesting, magnetic beads were removed with the CTS™ DynaMag system (#12102, ThermoFisher, France) and cells were washed and cryopreserved in AT2 vials (AT-Closed Vial®, Aseptic Technologies, Belgium) in CryoStor CS10 (#210102, BioLife Solutions, USA). Vials were stored at −150 °C in a dry freezer until further use. After lentiviral transduction and expansion, cells were 100% CD4+/CD45RA+, 99.8 ± 0.10% CD25+, 97.8 ± 0.59% CD127low, and 93.6 ± 1.26% FOXP3+ (mean ± SEM with n = 16), with FOXP3 TSDR hypomethylation of 89.1 ± 3.4% after thawing. Cells were viable at 86.9 ± 1.3% and transduced with the HLA-A*02 CAR at 54.7 ± 1.8% (mean ± SEM, n = 20). Of note, an integration site analysis indicated polyclonal insertion of the lentiviral vector and no unexpected integration or site selectivity were found (data not shown).

Flow cytometry

Cells from in vitro experiments were washed with phosphate-buffered saline (PBS)/4% bovine serum albumin and stained for cell surface markers including HLA-A*02 dextramer (#WB2666-APC, Immudex, Denmark), anti-human CD4, and anti-human CD69 (VioBlue, #130-113-219 and APC-Vio770, #130-112-616, Miltenyi Biotech, France). Prior to in vivo experiments, the number of HLA-A*02 CAR-expressing cells in the Treg batch used was determined using HLA-A*02 dextramer (#WB2666-PE, Immudex, Denmark).

Following in vivo experiments, spleen and brain samples were passed through a 70-µm cell strainer to obtain a single cell suspension. Brain cells were then incubated in RPMI/5% fetal calf serum (FCS) with 2.5 mg/ml collagenase D (#11088858001, Sigma–Aldrich, France) at 37 °C under agitation, washed in RPMI/5% FCS, and centrifuged. The cell pellet was resuspended in 70% (v/v) Percoll solution (#GE17-0891-01, Sigma–Aldrich, France), centrifuged at 500 × g over 37% (v/v) Percoll, and cells at the interface of the Percoll 70/37 gradient were recovered. Other tissues were roughly chopped with a scalpel blade and digested in RPMI/5% FCS containing 150 µg/ml collagenase D and 25 µg/ml DNase I (#10104159001, Sigma–Aldrich, France) and passed through a cell strainer. Red blood cells in blood samples and tissues were lysed with Red Blood Cell Lysing Buffer (#R7757, Sigma–Aldrich, France) or ammonium chloride. Blood and tissue cells were washed with PBS/2% FCS and incubated with mouse Fc block (#553142, BD Biosciences, France). Cells were then washed with PBS/4% bovine serum albumin and stained. Fixable viability dyes eFluor780 or eFluor506 (#65-0865-14 and #65-0866-14, ThermoFisher, France) were used. Cells were stained first with anti-human HLA-A*02 (BV421, #740082 or PE, #558570, BD Biosciences, France) and other cell surface markers, including anti-human CD4 (PE-Cy7, #557852 or FITC, #555346, BD Biosciences, France; or Percp-Cy5.5, #45-0049-42, ThermoFisher, France) and anti-human CD45 (VioGreen, #130-110-638, Miltenyi Biotech, France), and then fixed and permeabilized with the forkhead box P3 (FOXP3) staining buffer set (#00-5523-00, ThermoFisher, France), and stained for intracellular markers, including anti-human FOXP3 (PE, #12-4777-42, ThermoFisher, France or Alexa Fluor 647, #560045, BD Biosciences, France). Further antibody details are provided in the supplementary methods.

An Attune™ NxT flow cytometer and Attune™ NxT software were used for in vitro experiments, the biodistribution study, and tacrolimus study and analyses were performed with FlowJo V10. A MACSQuant Analyzer 10 with MACSQuantify software 2.8 was used for the GvHD model and hypomethylation study, and Cytoflex (Beckman Coulter) with FlowJo Software V9.9.4 and 10.3 was used for the skin transplant experiment. Naïve Tregs were isolated using an SH800 cell sorter and the gating strategy was modified from [17]. Exemplary gating strategies are presented in the supplementary methods (Fig. S2).

FOXP3 hypomethylation

To assess FOXP3 Treg-specific demethylated region (TSDR) hypomethylation, genomic DNA was extracted from cell pellets using the DNeasy Blood and tissue kit (#69506, Qiagen, USA), DNA bisulfite conversion was done using the Epitech Fast Bisulfite Conversion Kit (#59826, Qiagen, USA), and FOXP3 TSDR hypomethylation was then assessed by qPCR high-resolution melting curve analysis using Precision Melt Supermix (#1725112, Biorad, France) and FOXP3 TSDR-specific primers (forward: TTGGGTTAAGTTTGTTGTAGGATAG, reverse: ATCTAAACCCTATTATCACAACCCC, Sigma Aldrich, France). The fragment analyzed included 11 CpG sites and total TSDR methylation was calculated as the mean of methylation percentage of each individual CpG. Methylation levels were determined using a standard curve (#59695, Qiagen). For CAR-Treg products manufactured from leukapheresates of female donors, raw values were corrected to consider that one of the two TSDR alleles is fully methylated as a result of X-inactivation.

Plasma cytokine assessment

Human cytokines in plasma (interferon-γ [IFNγ], IL-1β, IL-2, IL-6, IL-10, IL-12p70, IL-13, IL-17A/F) were measured using multiplex cytokine immunoassay plates (V-plex custom human biomarker proinflammatory panel, #30098491, MesoScaleDiscovery, USA) according to manufacturer’s instructions.

In vitro assays

For in vitro experiments, each Treg donor constituted an individual experiment and n represents the number of Treg donors. Per experiment, 3 to 4 different Treg batches were used (see figure legends for number of biological and technical replicates). Mycoplasma testing was performed for all isolated cells as part of the quality assessments.

For activation, suppression, and cross-reactivity assays, TX200-TR101 Tregs were first cultured at 37 °C in X-VIVO-15 medium (#BEBP02-054Q, Lonza) with 300 IU/ml IL-2 (#2238131-A, Novartis Pharma) for 48 h.

In the activation assay, TX200-TR101 Tregs were stimulated for 24 h with anti-CD3/CD28-coated dynabeads (#40203D, Life Technologies, 1 bead:1 cell ratio) as a positive control for maximum TCR activation, with HLA-A*02-positive PBMCs (1:1 ratio of PBMCs to Tregs), with HLA-A*02 dextramer (#WB2666-PE, Immudex, Denmark), or with HLA-A*02-negative PBMCs (1:1 ratio of PBMCs to Tregs) or blank dextramer (#NI3233, Immudex, Denmark) as negative controls. Cells were then harvested and analyzed by flow cytometry.

In the suppression assay, Tconv were stained with a fluorescent proliferation dye (eFluor450, #65-0842-85, ThermoFisher, France). TX200-TR101 Tregs were cultured without IL-2 for 24 h and preactivated with HLA-A*02 or blank dextramer for 24 h while allogeneic CD4+/CD25– Tconv were preactivated for 24 h with anti-CD3/CD28-coated dynabeads (1:1 ratio). Preactivated Tconv and allogeneic TX200-TR101 Tregs were then co-cultured at Tconv:Treg ratios of 1:1, 2:1, 4:1, 8:1, 16:1 or 1:0 for 3 days at 37 °C in X-VIVO-15 medium. Proliferation suppression of Tconv was assessed by flow cytometry. The percentage of suppression was calculated for each ratio Tconv:Treg as the percentage of inhibition of the Tconv proliferation when cultured with Tregs compared with Tconv cultured alone.

In the cross-reactivity assay, TX200-TR101 Tregs were co-cultured with characterized cryopreserved human PBMCs (#CTL-UP1, Cellular Technology Limited) in a 1:1 ratio for 24 h. The HLA genotypes of the PBMCs that were available in the biobank for testing are specified in the supplementary. Cells were then harvested and analyzed by flow cytometry.

To assess the risk of cellular transformation and acquisition of a tumorigenic potential, TX200-TR101 Tregs cultured in X-VIVO-15 medium with 300 IU/ml IL-2 for 24 h were left unstimulated or chronically stimulated with anti-CD3/CD28-coated dynabeads (1:1 ratio) in the presence or absence of high-dose IL-2 (1000 IU/ml) for up to 75 days. Expansion with chronical stimulation was performed every 7 days and sub-passage every 2 days until cell death. From the third round of activation with beads, and/or depending on the cell growth, beads were replaced every 7 days. Three days after activation, cells were counted and diluted with fresh medium. This medium refresh was performed every 2 days, based on cell counting. In culture conditions where viability reached <3–5%, cells were kept in culture for an additional 15 days, with a medium refresh every 2–3 days, to ensure no clone of immortalized cells would grow. For cells cultured with high-dose IL-2, several starvation cycles were performed to ensure that no IL-2-independent cell growth occurred. Weekly, 2 samples of 1–0.5 × 106 cells were frozen in 10% dimethyl sulfoxide and sent to Life Length® (Madrid, Spain) for assessment of telomere length using high throughput quantitative fluorescent in situ hybridization technology [18] and of telomerase activity using a telomerase repeat activity protocol modified for real-time qPCR analysis [19]. Each assay analyzed 5 technical replicates.

Mouse models

NOD scid gamma (NSG) mice (NOD.Cg-PrkdcscidIl2rgtmWjl /SzJ, JAX stock no 005557) [20] and HLA-A*02 NSG mice with humanized HLA-A*02 (NOD.Cg-PrkdcscidIl2rgtm1Wjl Tg[HLA-A/H2-D/B2M]1Dvs/SzJ, JAX stock no 014570) [21] were purchased from Jackson Laboratory (Bar Harbor, USA), distributed by Charles River (Lyon, France) or bred in house at the UBC. Unless otherwise specified, mice were housed in a specific opportunistic pathogen-free facility, in individually, positively ventilated polysulfone cages with HEPA-filtered air, controlled 12 h light/dark cycle, temperature of 20–26 °C, and relative humidity of 30–70%. Filtered tap water and standard rodent chow were provided ad libitum. For experiments, mice were age-matched but distributed randomly to treatment groups. Except for the skin graft experiment that was conducted with two Treg batches, each in vivo experiment was conducted with one Treg batch/donor and n represents the number of mice per group.

Unless otherwise specified, mice were conditioned with intraperitoneal injection of 30 mg/kg busulfan (#B2635, Sigma–Aldrich) to favor human cell engraftment approximately 24 h prior to intravenous human cell injection via the tail vein. The number of cells for injection was determined for each batch of Tregs by flow cytometric analysis of HLA-A*02 CAR-expressing cells using dextramer. The percentage of dextramer-positive cells ranged from 45–72% and the number of Tregs injected ranged from 2.5–5 × 106 cells per mouse in the individual experiments. FOXP3 hypomethylation levels were also assessed prior to cell injection. Throughout experiments, body weights and GvHD scores were monitored 3 times weekly by investigators that were blinded to treatment. GvHD was scored based on weight, fur texture, posture, activity level, and skin integrity. Unless otherwise specified, categories were rated from 0–3 as described in the supplementary (Table S2), with higher scores indicating more symptoms. During experiments, blood samples were collected by retro-orbital sampling under local anesthesia (unless specified otherwise below). At the end of experiments or when ethical endpoints were reached, mice were euthanized by cervical dislocation, necropsy was performed, and tissue and blood samples were collected.

In the GvHD experiment, a total of 30 female NSG mice aged 7–8 weeks and with a median (range) weight of 22 g (20–25 g) were randomly assigned to one of the 3 following groups: HLA-A*02-positive human PBMCs alone, together with TX200-TR101 Tregs, or with TX235-TR101 Tregs (n = 5 per group/2 pooled experiments). Mice were conditioned with busulfan, injected with HLA-A*02-positive human PBMCs and/or Tregs (1:1 ratio, 5 × 106 cells each), and monitored as described above for 4 weeks. Blood samples for flow cytometry were collected weekly and blood and spleen were collected at sacrifice. Representative fluorescence-activated cell sorting (FACS) plots obtained in blood are shown in the supplementary data (Fig. S3).

In the hypomethylation experiment, a total of 11 male and female HLA-A*02 NSG mice aged 8–11 weeks and with a median (range) weight of 29 g (23–36 g) were randomly assigned to one of the 3 following groups according to the degree of FOXP3 TSDR hypomethylation of the injected TX200-TR101 Tregs: 80.8% hypomethylation (n = 5, 3 males and 2 females), 92% (n = 3, 1 male and 2 females), and 89.1% (n = 3, 2 males and 1 female). Mice were conditioned with busulfan, injected with TX200-TR101 Tregs (2.5 × 106 cells per mouse), and monitored as described above for 27 days. Blood samples for flow cytometry were collected weekly and blood, spleen, liver, and lung were collected at sacrifice. The results of these experiments are presented together with results of previous experiments with a similar set-up, where Tregs isolated with a previous gating strategy (TR100) were included. In these experiments, male and female HLA-A*02 NSG mice aged 8–11 weeks received TX200-TR100 Tregs with 69% hypomethylation (n = 20, 9 males and 11 females), or TX200-TR101 Tregs with 100% hypomethylation (n = 5, 2 males and 3 females) or 98% hypomethylation (n = 2, 1 male and 1 female), and were monitored for up to 27 days.

In the biodistribution experiment, a total of 24 male and female HLA-A*02 NSG mice aged 8-10 weeks and with a median (range) weight of 24 g (20–31 g) were conditioned with busulfan, injected with TX200-TR101 Tregs (3 × 106 cells per mouse), and monitored as described above. Mice were randomly assigned to one of 3 sacrifice timepoints at 1, 2, and 3 months (n = 8, 3 males and 5 females, planned per timepoint). Blood samples for flow cytometry were collected weekly and blood (for flow cytometry and plasma cytokine assessment), spleen, liver, lung, kidney, heart, testis/ovaries, brain, and intestine were collected at sacrifice. Gross pathology and histopathology were performed. For animals euthanized 2 and 3 months post-injection, parts of the spleen, lung, kidney, liver, testis/ovary, brain, and the whole intestine were collected for histopathology. Any gross lesion/abnormal tissue growth during the study was collected with half the lesion analyzed by flow cytometry and half fixed in 10% buffered formalin for histological analysis. The rest of the spleen, liver, lung, kidney, heart, brain, and testis/ovaries were processed for flow cytometric analysis. The gating strategy is shown in the supplemental data (Fig. S4).

In the tacrolimus experiment, a total of 20 female NSG mice aged 7 weeks and with a median (range) weight of 21 g (17–25 g) were randomly assigned to one of the 4 following treatment groups: PBMCs only (n = 7), PBMCs plus tacrolimus (n = 6), TX200-TR101 Tregs (n = 3), and TX200-TR101 Tregs plus tacrolimus (n = 4). Mice were conditioned with busulfan and injected with PBMCs and/or Tregs (5 × 106 cells each), as described above and tacrolimus groups received daily intraperitoneal injections of 0.5 mg/kg tacrolimus (Prograf, #29485063, Astellas Pharma, through Euromedex, France) for 3 weeks, starting 7 days after cell injection. The tacrolimus dose was chosen based on the results of pilot studies, where doses between 0.1 and 2 mg/kg had been tested. Mice were monitored as described above for a total duration of 3 weeks. Spleen was collected at sacrifice.

In the skin transplant experiment, a total of 32 female NSG mice aged 8–12 weeks and with a median (range) weight of 22 g (20–25 g) received dorsal skin transplants of approximately 1 cm2 from human HLA-A*02-positive donors as described previously [8]. Nineteen mice received a graft from Donor 1 (Cohort 1) and 13 mice received a graft from Donor 2 (Cohort 2). Nine weeks after skin transplantation, mice were randomly assigned to one of the 4 following groups: PBS (n = 3 each for Cohorts 1 and 2), HLA-A*02-negative human PBMCs (n = 6 for Cohort 1, n = 3 for Cohort 2), HLA-A*02-negative human PBMCs plus autologous TX200-TR101 Tregs (n = 7 for Cohort 1, n = 5 for Cohort 2), and TX200-TR101 Tregs alone (n = 3 for Cohort 1, n = 2 for Cohort 2) and injected with 10 × 106 PBMCs and/or 5 × 106 Tregs (2:1 ratio). Each cohort received a different batch of TX200-TR101 (Donors 1 and 2). Human PBMC engraftment and skin rejection were analyzed and blood samples for flow cytometry collected weekly. GvHD score categories in this experiment were rated from 0–2 as described in Cooke et al. 1996 [22] and blood samples were collected from the saphenous vein. The mice were monitored for 28 to 35 days, and skin graft, surrounding mouse skin, spleen, intestine, liver, and lung were harvested for histopathology at sacrifice. GvHD and skin rejection scores of both cohorts were pooled for analysis.

Histology and immunohistochemistry

Human skin grafts and surrounding mouse skin were fixed overnight at 4 °C in 10% formalin and stored in 70% ethanol before paraffin-embedding and preparation of hematoxylin/eosin-stained sections. For histopathological skin graft scoring, slides were evaluated by a blinded clinical pathologist using a scoring system defined by 6 factors (Lerner grade, spongiosis, necrotic keratinocytes, adnexal involvement, parakeratosis, lymphoid cuffs in dermis) as described previously [8]. Tissues collected in the biodistribution study were fixed in 10% buffered formalin and shipped to Vetopath (Antibes, France) for histological analysis. Tissues were dehydrated in Tissue Tek VIP2000 (Miles Scientific) and embedded in resin-mixed paraffin after xylene baths. Tissues collected at the UBC were shipped to Pacific Tox Path, LLC (USA) for evaluation, where tissues were trimmed and processed to paraffin blocks. All tissue blocks were cut at 4–5 µm and sections stained with hematoxylin/eosin and examined microscopically by a board-certified veterinary pathologist.

Immunohistochemistry for detection of human FOXP3 and human CD45 in mouse organs (intestine, liver, lung, and spleen) and human skin grafts and analysis of slides was performed as described previously [8]. Immunohistochemical localization of CD45 as leukocyte marker and FOXP3 as Treg marker was performed by light microscopy for individual mice and scored according to a grading scale: 0 (no findings), 1 (minimal), 2 (mild), 3 (moderate), 4 (marked) and 5 (severe).

Statistics

Sample size calculation for mouse experiments was performed with GPower software 3.1. Five mice per group were needed to get an effect size of 2 between control PBMC- and CAR-Treg-treated groups with a test power of 0.89. Statistical analyses were conducted using GraphPad Prism 8.4.3. Information on sample size, replicates, variables, and statistical significance testing is provided in the figure legends. One-way ordinary analysis of variance (ANOVA), 2-way ANOVA (with Tukey’s or Dunnett’s multiple comparison test), Kruskal–Wallis multiple comparison test with Dunn’s post-test, or Holm-Sidak method following multiple t-test, were performed as appropriate to assess statistical significance. P-values < 0.05 were considered statistically significant. P-values in figures are presented as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.