Ex vivo enrichment of PRAME antigen‐specific T cells for adoptive immunotherapy using CD137 activation marker selection

Abstract

Objective

Adoptive immunotherapy with ex vivo expanded tumor‐specific T cells has potential as anticancer therapy. Preferentially expressed antigen in melanoma (PRAME) is an attractive target overexpressed in several cancers including melanoma and acute myeloid leukaemia (AML), with low expression in normal tissue outside the gonads. We developed a GMP‐compliant manufacturing method for PRAME‐specific T cells from healthy donors for adoptive immunotherapy.

Methods

Mononuclear cells were pulsed with PRAME 15‐mer overlapping peptide mix. After 16 h, activated cells expressing CD137 were isolated with immunomagnetic beads and cocultured with irradiated CD137^neg fraction in medium supplemented with interleukin (IL)‐2, IL‐7 and IL‐15. Cultured T cells were restimulated with antigen‐pulsed autologous cells after 10 days. Cellular phenotype and cytokine response following antigen re‐exposure were assessed with flow cytometry, enzyme‐linked immunospot (ELISPOT) and supernatant cytokine detection. Detailed phenotypic and functional analysis with mass cytometry and T‐cell receptor (TCR) beta clonality studies were performed on selected cultures.

Results

PRAME‐stimulated cultures (n = 10) had mean expansion of 2500‐fold at day 18. Mean CD3⁺ percentage was 96% with CD4:CD8 ratio of 4:1. Re‐exposure to PRAME peptide mixture showed enrichment of CD4 cells expressing interferon (IFN)‐γ (mean: 12.2%) and TNF‐α (mean: 19.7%). Central and effector memory cells were 23% and 72%, respectively, with 24% T cells expressing PD1. Mass cytometry showed predominance of Th1 phenotype (CXCR3⁺/CCR4^neg/CCR6^neg/Tbet⁺, mean: 73%) and cytokine production including IL‐2, IL‐4, IL‐8, IL‐13 and GM‐CSF (2%, 6%, 8%, 4% and 11%, respectively).

Conclusion

PRAME‐specific T cells for adoptive immunotherapy were enriched from healthy donor mononuclear cells. The products were oligoclonal, exhibited Th1 phenotype and produced multiple cytokines.

Adoptive immunotherapy with ex vivo expanded tumor‐specific T cells has potential as anticancer therapy. Preferentially expressed antigen in melanoma (PRAME) is an attractive target overexpressed in several cancers including melanoma and acute myeloid leukaemia (AML), with low expression in normal tissue outside the gonads. We developed a GMP‐compliant manufacturing method for PRAME‐specific T cells from healthy donors for adoptive immunotherapy.

Introduction

Adoptive T‐cell therapies are under investigation for the treatment of a variety of malignancies. These include ex vivo expanded tumor‐infiltrating lymphocytes (TILs), ¹ , ² ex vivo expanded circulating tumor antigen‐specific T lymphocytes ³ , ⁴ and genetically modified products such as chimeric antigen receptor (CAR) ⁵ , ⁶ and transgenic T‐cell receptor (TCR)‐modified T cells. ⁷ , ⁸ Naturally occurring T cells that recognise intracellular or extracellular tumor‐associated antigens or neoantigens formed by malignant genetic alterations can be ex vivo expanded and used therapeutically. ⁹ There are a number of small trials of this approach. ¹⁰ , ¹¹ , ¹² , ¹³ T cells reactive against BCR‐ABL ¹⁴ , PML‐RARa ¹⁵ , proteinase 3 ¹⁶ and WT1 ⁹ can be isolated and demonstrate specific cytolytic activity in vitro. ⁴ , ¹⁶ , ¹⁷ Identification of potential antigen targets is underway in many malignancies. ¹⁸ , ¹⁹

Preferential antigen in melanoma (PRAME) is overexpressed in a wide variety of tumor types including AML, ²⁰ melanoma, ²¹ renal cell carcinoma, ²² acute lymphoblastic leukaemia, ²³ chronic myeloid leukaemia, ²⁴ Hodgkin lymphoma ²⁵ and medulloblastoma. ²⁶ Moreover, its low expression in normal healthy tissue apart from gonadal tissue ²⁰ makes it an attractive target for cellular immunotherapy.

To date, methods of ex vivo expansion of TAA‐specific T cells rely on the use of antigen‐presenting cells (APCs) such as dendritic cells or engineered artificial antigen‐presenting cells. ⁹ , ²⁷ , ²⁸ , ²⁹ This approach is labour‐intensive, time‐consuming, associated with higher cost and may be difficult to translate to GMP‐compliant processes for clinical use. To address this, we developed a protocol for immunomagnetic bead selection of T cells expressing the activation marker CD137 (4‐1BB) after exposure to overlapping PRAME peptides as a rapid method of ex vivo expansion for clinical use (Figure 1). CD137 is a costimulatory molecule and a member of the tumor necrosis factor receptor (TNFR) family. Transient increased expression is seen on cells that have been activated by TCR engagement and signalling. ³⁰ We utilised this feature to select and expand PRAME‐specific T cells and performed phenotypic and functional analysis of the final cell product. The method we describe is robust using mononuclear cells from healthy donors and readily applicable to clinical use. A clinical trial utilising this protocol to prevent postallogeneic haemopoietic stem cell transplant relapse has been initiated (ANZ CTR NCT02895412).

Figure 1.

Method for ex vivo expansion of CD137‐expressing activated T cells.

Results

Determining maximal CD137 expression

In order to determine the optimum time for CD137⁺ cell selection, a time course was performed to determine the expression at 0, 16, 24 and 41 h after antigen challenge. Maximal cell surface CD137 expression by flow cytometry was achieved 16–24 h after previously expanded antigen‐specific T cells were rechallenged with antigen‐derived peptide mixture and anti‐CD28 antibody (Figure 2).

Figure 2.

CD137 expression is maximal at 18–24 h following exposure to PRAME peptide mixture. CD137 expression by cultured T cells following re‐exposure to PRAME peptide mixture measured by flow cytometry at 16, 24 and 41 h. (n = 1).

CD137⁺‐activated T‐cell isolation and expansion

The mean percentage of CD137⁺‐enriched cells isolated by immunomagnetic bead separation from PBMC was 0.6% of total starting cells (range 0.04–1.72%). There was no significant difference between the steady state or G‐MNC starting material (mean 0.9% and 0.36%, respectively, P = 0.093; Figure 3a). In the CD137‐enriched fraction, the proportion of CD3⁺ cells which expressed CD137⁺ by fluorescence flow cytometry was 67% (range 13–89.1%) with no significant difference between the different starting material (means of 62.7% and 70.6% for PBMC and G‐MNC, respectively, P = 0.60; Figure 3b). Using PBMC as starting material, the cell expansion by the end of culture was 148‐fold (range 96–210‐fold) at 11 days and 2539‐fold (range 990–4687‐fold) at 18 days. With G‐MNC as starting material, expansion after 11 days was 194‐fold (range 110–322‐fold) and 2138‐fold (range 660–5418‐fold) after 18 days (Figure 3c). There was no significant difference in expansion kinetics regardless of starting material of PBMC or G‐MNC (P = 0.917).

Figure 3.

Isolation and expansion of CD137‐expressing cells from PBMC. (a) Immunomagnetic column CD137 selection yielded a positive fraction ranging from 0.04 to 1.72% of starting cell number with no significant difference between starting material PBMC or G‐MNC (n = 5). (b) Flow cytometry assessment of CD137 immunomagnetic column selected fraction showing proportions of CD3‐positive cell expressing CD137 (n = 5). The majority of these cells are CD4⁺ T cells. (c) Ex vivo expansion of cells in culture. Mean expansion of 170‐fold and 2500‐fold on days 11 and 18, respectively. G‐MNC, G‐CSF‐primed apheresis‐derived mononuclear cells; PBMC, peripheral blood mononuclear cells; G‐MNC G‐CSF‐stimulated mononuclear cells.

Phenotype by fluorescence cytometry

Immunophenotype of resting cells at the end of culture was measured by fluorescence flow cytometry. The majority of cells were CD3⁺ T cells (mean 96%, range 92.4–99.6%). CD4 T cells dominated (80% of CD3⁺, range 55.1–99%; Figure 4). The majority of T cells in culture were CD45RA⁻ CD62L⁻ effector memory cells (72%, range 36.4–92.3%). A large population of CD45RA⁻ CD62L⁺ central memory cells (23%, range 3.5–59.5%) was also present. The expression of the co‐inhibitory marker PD1 was 24.4% (range 3.3–83.5%) with variable expression of Tim3 (range 51.3% (range 13.9–77.2%) and LAG3 (83.5%, range 62.8–95%). More detailed phenotype of the antigen‐specific T cells was explored with mass cytometry, described below.

Figure 4.

Fluorescence flow cytometry on PRAME‐specific T‐cell cultures (n = 10). At the end of culture, the majority of cells were CD3⁺ T cells with preponderance for CD4⁺ helper T cells. CD14⁺ monocytes, CD19⁺ B cells and CD3^negCD56⁺ NK cells were virtually absent. There was predominance of CD45RA^negCD62L⁺ central memory T cells and CD45RA^negCD62L^neg effector memory T cells. There is variable expression of co‐inhibitory markers PD1, Tim3 and LAG3 above isotype control.

Specificity

Specificity of the cultured T cells was demonstrated by measuring the cytokine production when re‐exposed to PRAME‐derived peptide mixture using fluorescent flow cytometry, ELISPOT, cytokine detection in culture supernatant and high‐dimensional mass cytometry.

Intracellular flow cytometry showed that upon re‐exposure to PRAME peptide mixture, compared with media‐only control, CD4⁺ T cells in all cultures secreted significantly more IFN‐γ and TNF‐α and there was a significant increase in the surrogate degranulation marker CD107 (IFN‐γ: mean 12.2% (range 0.2–36.3%) vs 0.5% (0.1–2.3%), P = 0.02; TNF‐α: mean 19.7% (1.7–54.6%) vs 0.7% (0.3–1.7%), P = 0.007; CD107: mean 16.6% (0.7–48.8%) vs 1.4% (0.3%–2.4%), P = 0.01). There was also increased expression of CD107 by CD8⁺ T cells (mean 12.9% (0.1–43.7%) vs 2.2% (0.2–8.0%), P = 0.04). CD8⁺ PRAME antigen‐specific IFN‐γ and TNF‐α responses were seen in 5 of 10 cultures (the proportion of responsive CD8⁺ T cells in these 5 cultures: IFN‐γ mean 13.6%, range 3.6–33.3%; TNF‐α mean 12.7%, range 2.3–33%). However, in all cultures, there was not a statistically significant difference in PRAME‐specific CD8⁺ IFN‐γ⁺ and TNF‐α⁺ compared with unstimulated cells at the end of culture (IFN‐γ mean 7.1% (0.1–33.3%) vs 0.9% (0.1–3.0%), P = 0.09; TNF‐α mean 6.7% (0.3–33%) vs 0.4% (0.0–1.1%), P = 0.10; Figure 5a).

Figure 5.

Antigen specificity of the cultured T cells by cytokine release in response to re‐exposure to PRAME. (a) Expression of IFN‐γ and TNF‐α and CD107 degranulation in the CD4⁺ and CD8⁺ T cells following 18 days of culture in 10 cultures following restimulation to PRAME peptide mixture (PRAME) or no restimulation (Nil). Pooled data represented as mean ± SD. *P < 0.05; **P < 0.01. (n = 10) (b) IFN‐γ ELISPOT was performed on 3 PRAME‐specific T‐cell cultures (PRAME peptide mixture vs negative controls media only and irrelevant peptide mixture). (c) Photograph of representative wells from Elispot assay with cultured T cells incubated with (i) PRAME peptide mixture; (ii) media only; (iii) irrelevant peptide mixture. (d) Supernatant harvested from 2 cultures analysed with cytokine antibody membrane array showed increases in Th1 cytokines (IFN‐γ, TNF‐α, IL‐2) as well as Th2 cytokines (IL‐4, IL‐5, IL‐13).

A response to PRAME was demonstrated by ELISPOT with a mean of 981 SFU/10⁵ cells (range 37–2678 SFU/10⁵ cells) compared with 25 SFU/10⁵ cells (range 16–32 SFU/10⁵ cells) in the negative control and 27 SFU/10⁵ cells (13–42 SFU/10⁵ cells) when exposed to irrelevant peptide mixture (WT1) (Figure 5b and c).

In order to assess the secretion of an extended selection of cytokines, the supernatant of 2 cultures was harvested with or without re‐exposure to PRAME‐derived peptide mixture and incubated with cytokine array. This showed increased levels of Th1 (IFN‐γ, TNF‐α, IL‐2) as well as Th2 cytokines (IL‐4, IL‐5, IL‐13) (Figure 5d and Supplementary figure 1).

Mass cytometry

Production of cytokines by individual cells was assessed by mass cytometry on three cultures. Choice of antibody targets was based on the results obtained from the cytokine array. tSNE was used to visualise the distribution of cytokine production in individual cells (Figure 6a). Production of TNF‐α, IFN‐γ, IL‐2, IL‐4, IL‐8, IL‐13 and GM‐CSF in response to re‐exposure to PRAME‐derived peptide mixture was primarily by CD4⁺ cells. There was evidence of degranulation with increased expression of CD107. The responding cells corresponded numerically with those identified as antigen responsive in fluorescence flow cytometry. CD4⁺ T cells were further characterised by sequential gating ³¹ (Figure 6b), with the majority of CD4⁺ T cells showing Th1 or Th17 phenotype (CXCR3⁺) (mean 73%, range 59.6–87.5%) with Th2 cells (CXCR3^negCCR4⁺CCR6^neg) comprising 8% (2.9–13.1%) of CD4⁺ T cells (Figure 6b).

Figure 6.

Extended phenotype and cytokine production assessed by mass cytometry. (a) Mass cytometry was performed on three cultures after stimulation with PRAME‐derived peptide mix and analysed with tSNE dimensionality reduction with algorithm‐based clustering of cells deemed immunophenotypically similar on the 38 selected parameters as detailed in the Methods section. Dots on each tSNE plot in the same position represent the same cell. The expression of various markers is coded by the colour gradient with warmer colours correlating to higher expression. (b) Sequential gating on mass cytometry demonstrated predominance of Th1 helper phenotype with 51.9% (range 46.3–61.6%) of CD4⁺ T cells displaying a Th1 phenotype as defined by CXCR3⁺CCR6^neg and a further 20.7% (12.8–25.5%) with Th1/Th17 phenotype CXCR3⁺CCR6⁺. 7.5% (2.7–12.3%) of CD4⁺ T cells had CXCR3^negCCR4⁺CCR6^neg Th2 phenotype. (c) Cytokine‐releasing cells (16.1%, range 8.8–35.7%) were gated for further analysis on tSNE plots. (d) Threshold for positivity of individual parameters was determined based on signal separation in contour plots with the exception of cytokines (IFN‐γ, TNF‐α, GM‐SCF, IL‐2, IL‐4, IL‐6, IL‐8, IL‐10, IL‐13, IL‐21) and degranulation markers (granzyme and CD107) where the thresholds were set with unstimulated specimens (representative example of TNF‐α shown). (e) Extended phenotypic expression of each parameter of the cytokine secreting subset of cells on mass cytometry as a proportion of viable cells.

The CD4⁺ population secreting multiple cytokines in response to PRAME was gated as shown in Figure 6c for further analysis. The threshold for positivity of the cytokines and degranulation markers is determined based on the unstimulated sample (Figure 6d). The cytokine producing populations from the three cultures had similar extended phenotype as shown in Figure 6e with CD3, 4, 8, memory phenotypes and co‐inhibitory markers PD1, Tim3 and LAG3 correlating numerically with fluorescence flow cytometry. The activation marker CD154 was increased more than CD137 at 4 hours after re‐exposure to the peptide mixture. The cells were transcriptionally active with high expression of Tbet expression supporting Th1 phenotype. GATA3 (transcription regulator for Th2) showed relatively modest increase in expression. Eomes, the CD8 transcription regulator, was unexpectedly elevated in these CD4⁺ cells. FoxP3 was highly expressed with negligible IL‐10 expression.

TCR sequencing

In order to assess the clonality of the final product, reactive cells were sorted by the expression of CD137 after restimulation at day 18 of ex vivo expansion in 4 cases. After RNA extraction, the CDR3 of the beta TCR of the CD137‐positive fraction was sequenced. All ex vivo expanded PRAME‐specific cell products were oligoclonal (Figure 7). In 3 of 4 cases, the majority of cells in the product were represented by the top 10 clones (mean 59% of clonal repertoire, range 27–73%). The CD4⁺ and CD8⁺ T‐cell fractions were separately sequenced for case 4 and both fractions showed oligoclonality of TCR expressed.

Figure 7.

TCR sequencing. TCR clonal proportions of each culture depicting the relative repertoire spaces occupied by clones ordered by prevalence as indicated in the legend.

Discussion

CD137 is a costimulatory receptor belonging to the TNF receptor superfamily and is almost uniformly expressed by activated CD4⁺ and CD8⁺ T cells as well as some APCs. ³² The inducible expression of CD137 in T cells upon activation has recently been used as a positive selection marker for immunomagnetic column selection for ex vivo expansion of antigen‐specific T cells. ³³ , ³⁴ The time to maximal CD137 expression on T cells following exposure to immunogen has been reported to vary from 12 h to 5 days depending on the nature of stimulation. ³² , ³⁵ , ³⁶ We found that the maximal CD137 expression using PRAME‐derived peptide mixture and CD28 antibody costimulation was at 16–24 h (Figure 2). A small number of CD137‐positive T cells were isolated following overnight incubation with PRAME peptide mixture and CD28 antibody. This is not unexpected because of the low frequency of TAA‐specific T cells as is consistent with published findings. ³⁴ , ³⁷ Enrichment and numeric expansion of these cells can be achieved in culture with T‐cell medium, supported by cytokines IL‐2, IL‐7 and IL‐15 and feeder cells in the form of irradiated autologous PBMCs or G‐MNCs.

There was predominance of CD4⁺ T cells in the majority of cultures, which is consistent with previous reports. ³⁸ , ³⁹ Other reports of in vitro expanded CD8‐dominant PRAME‐specific T‐cell cultures have utilised substantially different methodology. ⁹ , ⁴⁰ CD4⁺ T cells have been shown to be an important component of the antitumor T‐cell response ⁴¹ , ⁴² in their role of both activating cytolytic T cells ⁴³ and inducing cytotoxicity independently of CD8⁺ T cells. ⁴⁴ Mass cytometry and cytokine array data show that the majority of the CD4⁺ T cells show characteristics of Th1 function which is associated with in vivo efficacy following adoptive transfer. ⁴² The majority of the cultured cells have a central or effector memory phenotype, again consistent with previous reports. ³⁴ We and others have previously highlighted the importance of T cells from these memory compartments in long‐term persistence of adoptive antigen‐specific T cells in vivo, ⁴⁵ , ⁴⁶ , ⁴⁷ and which may be associated with superior clinical outcome. ⁴⁸ , ⁴⁹ There is modest expression of co‐inhibitory receptors PD‐1, Tim3 and LAG3, which has been reported to be expressed in activated CD4 T cells with Th1 function. ⁵⁰ The active proliferation and cytokine response to re‐exposure to PRAME pepmix suggest these markers represent a response to antigen stimulation rather than a marker of exhaustion.

The specificity of cultured T cells has been demonstrated using multiple modalities. Given the low precursor frequency of tumor antigen‐specific T cells, the proportion of T cells with demonstrable specific response to PRAME peptide mixture represent significant enrichment of specific T cells. Flow cytometry and mass cytometry confirmed a subset of cultured T cells being able to secrete multiple pro‐inflammatory cytokines in response to re‐exposure to PRAME‐derived peptide mixture. Multifunctional T cells are associated with superior efficacy in vivo. ⁵¹ , ⁵²

TCR sequencing shows the T‐cell products to be oligoclonal without significant overlap of the clones between donors. This suggests that the original number of clones in the starting material is low and is consistent with other reports of tumor antigen‐specific T‐cell precursor frequency. ¹¹ , ³⁷ Further work will include defining the epitope targets and HLA restriction of these products but this is beyond the scope of this manuscript.

This method has the major advantage of requiring no specialised antigen‐presenting cells such as dendritic cells. ⁴⁷ , ⁵³ , ⁵⁴ , ⁵⁵ Instead, we relied on the antigen‐presenting capacity of monocytes and other APCs in isolated PBMCs or G‐MNC to present the oligopeptides with anti‐CD28 antibody providing costimulation. This results in reduced personnel time required for manipulation of cells and shorten overall manufacturing timeframe. The use of scanning oligopeptide peptide mixture allows ex vivo expansion of T cells without the need to customise oligopeptide sequences based on the major histocompatibility complex (MHC) type of the donor. The source of starting material for T‐cell culture did not have any significant effect on the isolation or expansion of PRAME‐specific T cells. We have incorporated manufacture of ex vivo expanded antigen‐specific T cells for clinical use into routine transplant donor assessment processes to improve clinical translation feasibility. This methodology is currently in use in an ongoing phase I clinical trial (NCT02895412) of ex vivo expanded donor‐derived pathogen‐specific and tumor antigen‐specific T cells following allogeneic stem cell transplant for patients with acute myeloid leukaemia.

In conclusion, we have developed a method of ex vivo expansion of PRAME‐specific T cells for clinical use and demonstrated the in vitro antigen specificity of the cultured T cells, which display a Th1 phenotype associated with favorable in vivo function and persistence after adoptive transfer.

Methods

Donors

Healthy donors provided informed consent in accordance with the Declaration of Helsinki following ethics approval from the Sydney West Local Health District Human Research Ethics Committee. T cells for the starting material were derived from peripheral blood mononuclear cells (PBMCs) collected at steady state or from G‐CSF‐stimulated apheresis products (G‐MNC) from healthy allogeneic stem cells donors procured as part of routine transplantation practice. PBMC isolation was performed as previously described by gradient density centrifugation in Ficoll‐Hypaque (GE Healthcare, NSW, Australia). ⁵⁶ For T‐cell products generated from apheresis, a small percentage of harvest (approximately 0.5–1%) was taken as starting material if the amount of the harvest was > 2.5 × 10⁶ CD34⁺ cells/kg of recipient bodyweight, as previously described. ⁵³ HLA typing of 4 healthy donors on which TCR sequencing was performed was done commercially by PathWest Laboratory Medicine, Western Australia, using next‐generation sequencing, and is shown in Supplementary table 1.

Peptides

PRAME peptide mixture [lyophilised pentadecapeptide mixtures with 11 amino acid overlaps covering the sequence of PRAME (UniProtKB Acc. no. P78395)] was reconstituted as per manufacturer’s instructions (Miltenyi Biotec, Bergisch Gladbach, Germany) and frozen in 1 µg/peptide aliquots. WT1 (Swiss‐Prot Acc. no. P19544) peptide mixture (Miltenyi Biotec) used as the irrelevant peptide mixture for negative controls in the ELISPOT assay was similarly reconstituted and aliquoted.

Antigen stimulation and activation marker selection

50 × 10⁶ viable PBMC or G‐MNC was resuspended in 1 mL of AIM‐V (Gibco‐BRL, CA, USA) supplemented with 10% human AB serum (Australian Red Cross Lifeblood, Sydney, Australia). PRAME peptide mixture was added to PBMC to a concentration of 1 µg mL^–1 per peptide, and anti‐CD28 antibody (Miltenyi Biotec) was added to a concentration of 1 µg mL^–1. The cells were incubated at 37°C at 5% CO₂ for 1 h. The cultures were then incubated in a 6‐well plate for a further 16–24 h at concentration of 1 × 10⁷ mL^–1 by adding medium with CD28 antibody to maintain concentration of 1 µg mL^–1. CD137⁺ cells were labelled with CD137‐PE and anti‐PE MicroBeads (Miltenyi Biotec) as per manufacturer’s instructions. CD137⁺ cells were isolated by positive selection with MS MACS columns using miniMACS magnetic separators.

Ex vivo expansion

The ex vivo expansion method is shown in Figure 1. The CD137^neg cells from the flow through of immunomagnetic column selection were irradiated (30 Gy) and cocultured with CD137⁺ cells at a CD137^neg:CD137⁺ ratio of 100:1 at a total cell concentration of 2 × 10⁶ mL^–1 in AIM‐V medium supplemented with 10% human AB serum in G‐REX10 bioreactors (Wilson Wolf, MN, USA). Culture medium was supplemented with IL‐2, IL‐7 and IL‐15 (all cytokines from Miltenyi Biotec) every 2–3 days to a final concentration of 20 U mL^–1, 200 U mL^–1 and 200 U mL^–1, respectively. Restimulation with PRAME primed autologous irradiated PBMCs was performed after 10 days in culture. Autologous cells were primed for 1 h with PRAME peptide mixture at final concentration of 1 μg mL^–1 and irradiated (30 Gy) before being added at a ratio of 2:1. The coculture was resuspended in AIM‐V at 2 × 10⁶ mL^–1 with anti‐CD28 antibody.

Fluorescence cytometry

Immunophenotyping of cultured cells was performed at the end of culture with antibodies against CD3, CD4, CD8, CD14, CD19, CD56, CD62L, CD45RA, PD1, LAG3, Tim3, CTLA4 (all BD Biosciences, NJ, USA) and the viability marker Fluorogold (Molecular Probes, OR, USA). Antibody clone and fluorochrome conjugation are listed in Supplementary table 2. 1 × 10⁶ cells from each culture were stained in a 100 μL volume with antibody cocktail for 30 min and washed in flow cytometry buffer (phosphate‐buffered saline with 0.5% bovine serum albumin and 0.1% sodium azide) before being acquired with FACS Fortessa (BD biosciences) flow cytometer.

For intracellular cytokine staining, 0.5 × 10⁶ cells were incubated with an overlapping peptide mix of PRAME (Miltenyi Biotec) for 5 h in the presence of anti‐CD28 (L293) and CD49d (L25) antibodies (BD Biosciences). Monensin (2 μm; BD biosciences) and brefeldin A (1 μg mL^–1; BD biosciences) were added for the last 4 h, followed by labelling with fluorochrome‐conjugated antibodies against surface antigens CD3, CD4 and CD8 and degranulation marker CD107a + b (H4A3 and H4B4) and viability marker 7‐AAD (BD biosciences). Intracellular labelling for IFN‐γ (B27) and TNF‐α (Mab‐11) was performed using the Cytofix/Cytoperm Kit (BD Biosciences) according to the manufacturer's instructions. T cells stimulated with 5 ng mL⁻¹ phorbol 12‐myristate 13‐acetate (PMA) and 1 ng mL^–1 ionomycin (Sigma‐Aldrich, MO, USA) were used as positive controls, and unstimulated cultured T cells were used as negative controls. Samples were acquired with a FACS Canto II (BD Biosciences) flow cytometer and analysed with FlowJo version 10 (Tree Star, Inc., OR, USA).

PRAME‐specific T‐cell cultures were restimulated with PRAME peptide mixture, and activation marker expressing cells were isolated for T‐cell receptor (TCR) sequencing. Sixteen hours after stimulation, cells were resuspended in phosphate‐buffered saline containing 5% foetal bovine serum (PBS + 5%FBS) and 20% human FCR blocking reagent (Miltenyi Biotec), stained with CD3, CD4, CD8, CD137, CD154 and viability marker 7‐AAD (all BD Biosciences) for 20 min at 4°C and washed and resuspended in 100 µL PBS + 5% FBS. They were then sorted on a BD FACSAriaIII (BD Biosciences) using a 70‐µm nozzle into CD3⁺CD4⁺ or CD3⁺CD8⁺ CD137⁺ and/or CD154⁺ populations (gating strategy shown in Supplementary figure 2). In cases with CD8⁺ T cells < 10%, the CD4⁺ and CD8⁺ fractions were not separated.

Mass cytometry

T cells were prepared for mass cytometry by incubating 2 × 10⁶ cells with PRAME peptide mixture for 5 h in the presence of anti‐CD28 (L293) and CD49d (L25) antibodies. Monensin (2 μm) and brefeldin A (1 μg mL^–1) were added for the last 4 h. Cells were then stained with cisplatin as a DNA identifier followed by heavy metal‐conjugated monoclonal antibodies for mass cytometry analysis as previously described. ⁵⁷ Dimensionality reduction analysis was performed with the t‐stochastic neighbourhood embedding (tSNE) algorithm (implemented in FlowJo as a PlugIn). A fixed number of 7000 cells per condition was sampled at random without replacement from each file and combined for analysis. The markers used for clustering were CD3, CD4, CD8, CXCR3, CCR4, CCR6, CCR7, CD45RA, CD62L, CD25, CD27, CD39, CD122, CD127, CD134, CD154, PD1, Tim3, LAG3, HLA‐DR, IL‐2, IL‐4, IL‐6, IL‐8, IL‐10, IL‐13, IL‐21, GM‐CSF, IFN‐γ, TNF‐α, CD107, granzyme, RANTES, GATA3, Eomes, FoxP3, Tbet. Antibody clone and heavy metal conjugation are listed in Supplementary table 3.

ELISPOT

IFN‐γ enzyme‐linked immunospot assay (ELISPOT) was performed on cells at the end of culture using a previously described method modified for PRAME antigen. ³² Briefly, 0.5–2 × 10⁵ cells were suspended in 200 μL of AIM‐V/10% human AB serum and stimulated with PRAME peptide mixture for 18 h in 96‐well filter plates (MAIPS4510; Millipore) coated with capture antibody (51‐2555KZ; BD Biosciences). Following wash and incubation with detection antibody (51‐1890KZ; BD Biosciences), immunospots were developed using ExtrAvidin and SigmaFast BCIP/NBT alkaline phosphatase substrate (Sigma‐Aldrich, St Louis, MO) in accordance with manufacturer's directions. The plates were analysed with the AID ELISPOT Reader (Autoimmune Diagnostika, Strassberg, Germany). Tests were performed in triplicates and spots were counted manually.

Cytokine array

PRAME‐specific T cells were incubated with PRAME peptide mixture for 5 h in the presence of anti‐CD28 (L293) and CD49d (L25) antibodies. The supernatant was harvested for the detection of cytokine using dot blot cytokine antibody membrane array (Abcam ab133997, Cambridge, UK) as per manufacturer’s instructions; supernatant from end of culture T cells without restimulation was used as a control. The resultant array images were captured with Chemidoc Touch imaging system (Bio‐Rad, CA, USA) and analysed with Imagelab v5.2.1 (Bio‐Rad) and ImageJ (NIH).

TCR sequencing

TCR sequence acquisition was performed as previously described. ⁵⁸ RNA was isolated using Qiagen RNeasy Mini kits (Quantitect RT, QIAgen, Hilden, Germany). Using 5 μL of cDNA template per reaction, TCRβ transcripts were PCR‐amplified from the ex vivo expanded T‐cell product using high‐fidelity Q5 polymerase (New England Biolabs, MA, USA) and a mix of 19 Trbv‐specific forward primers and a single Trbc‐specific reverse primer. Forward and reverse primers had distinct 5′ overhang adapter sequences that enabled addition of sample‐specific indices and P5/P7 sequencing adapters in a second PCR using the Illumina Nextera XT DNA library preparation kit. Before the second PCR, magnetic beads (Agencourt AMPure XP, Beckman Coulter, CA, USA) were used to enrich amplicons > 100 bp. After determining amplicon concentrations using a QIAxcel capillary electrophoresis machine (Qiagen), equimolar amounts of amplicons from up to 270 samples were pooled into a single tube, concentrated using magnetic beads (Agencourt AMPure XP, Beckman Coulter) and then 300‐ to 500‐bp amplicons were gel‐purified before sequencing on an Illumina NextSeq machine, with a short read 1 of 6 bases followed by a read 2 of 145 bases.

Raw sequence fastq files were analysed with FastQC to quality check average base pair scores. Migec checkout function was used to demultiplex and collapse reads into their 10bp Unique Molecular Identifier groups (UMI) using the ‐cute function to compress output, alter FastQ header with UMI info, adapter trimming and remove trails of template‐switching. Mixcr was used to assemble and align reads and to extract Cdr3 sequences. Align function aligned the input demultiplexed reads to the V, D and J T‐cell receptor reference sequences with –report option for alignment statistics. Assemble function built the clonotypes using the alignments generated from the align function and according to specific gene features, that is CDR3. Finally, the export function was used to extract the clonotypes, but with the assemble function, with information such as clonotype nucleotide sequence and count to parsable text files for downstream analysis.

Statistical analysis

Statistical analysis was performed using GraphPad Prism version 7.02 (GraphPad, La Jolla, CA, USA). Results are expressed as mean (range minimum–maximum) unless otherwise specified. For comparison of cell expansion parameters and phenotype between the different starting materials, the two‐tailed Student’s t‐test was used for comparison between 2 groups. A paired t‐test was used for comparison of measured parameters between PRAME‐stimulated and media‐only conditions for experiments determining the specificities of the products (intracellular cytokine flow cytometry, cytokine array and ELISPOT) and for comparison of proportion of cells expressing measured parameters between cytokine secreting subset and noncytokine secreting subset as determined by mass cytometry.

Conflict of interest

EB reports advisory board membership Abbvie, Novartis, Astellas and MSD. DG reports advisory Board membership Abbvie, Gilead, Indee Labs and Novartis. KM reports advisory board membership Indee Labs. DG and KM report research funding from Haemalogix. EB, DG, KM and LC report patents in the field of adoptive cell therapy manufacture.

Author Contributions

Koon Hiang Lee: Data curation; Formal analysis; Investigation; Methodology; Writing‐original draft; Writing‐review & editing. Kavitha Gowrishankar: Investigation; Project administration; Resources; Supervision; Writing‐review & editing. Janine Street: Investigation; Methodology. Helen McGuire: Formal analysis; Investigation; Methodology; Resources. Fabio Luciani: Formal analysis; Resources; Software; Supervision. Brendan Hughes: Data curation; Formal analysis; Investigation; Software; Visualization. Mandeep Singh: Formal analysis; Investigation; Methodology. Leighton Clancy: Methodology; Resources; Validation; Writing‐review & editing. David Gottlieb: Resources; Supervision; Writing‐review & editing. Kenneth Micklethwaite: Resources; Supervision; Writing‐review & editing. Emily Blyth: Conceptualization; Funding acquisition; Methodology; Resources; Supervision; Visualization; Writing‐review & editing.

Supporting information