To facilitate the clinical application of TILs, the standard rapid expansion protocol was automated using the TRT Process ( Figure 1A ) (34). All handling steps and components of the REP, including addition of young TILs (yTILs) and irradiated feeder cells, cytokines, media, activation reagents, media exchanges, harvest and final product formulation were incorporated into the process. The specifications of the REP culture, such as frequency and number of feeding steps can be flexibly programmed by the user and are performed automatically by the CliniMACS Prodigy. Visual inspection of the cell culture to follow the growth of T cell clusters that form during the primary activation phase is possible using the integrated microscope camera of the device ( Supplementary Figure S1A ). The first 5 days of expansion are performed under static conditions to avoid disruption of cell-cell contacts that are critical in the early phase of the REP expansion. Shaking of the culture chamber has been implemented to provide adequate gas-exchange needed for higher cell densities during later phases of expansion. Timepoints and shaking parameters can be flexibly adjusted by the operator. IPC samples can be collected sterilely to monitor cell phenotype, cell composition, cell counts, viability, glucose and lactate levels throughout the process ( Supplementary Figures S1B, C ).

To demonstrate the utility of the TRT Process, between 2x10⁵ and 2x10⁶ yTILs derived from melanoma patients were expanded for 12-15 days in the automated REP. Using the TRT Process expansion rates of 2,453 ± 1,819-fold could be reached yielding final cell counts of 2.4x10⁹ ± 1.8x10⁹ TILs. An upper limit of ~5x10⁹ cells was found similar to other manufacturing processes of T cells on the CliniMACS Prodigy platform and is not TIL specific but rather reflects culture-density limitations and the chamber size ( Figure 1B ) (35). The final cell product was highly viable and contained predominantly CD3⁺ T cells. The ratio of CD4⁺ and CD8⁺ subsets was largely patient specific and was not affected by the expansion culture ( Supplementary Figure S1D ). yTILs from other solid tumors such as pancreatic cancer, renal cell carcinoma and urothelial carcinoma were also expanded with the TRT Process. Like yTILs from melanoma, yTILs from other solid tumors were robustly expanded to clinically relevant numbers reaching final cell counts of 0.7 – 5.3x10⁹ cells, which corresponded to expansion rates between 905- and 12,674-fold. Expanded TIL cell products from other solid tumors were viable, contained almost exclusively CD3⁺ T cells and showed patient dependent populations of CD4⁺ and CD8⁺ T cells ( Figure 1C ).

At the end of each expansion process, the differentiation and exhaustion status of the resulting TILs was defined using flow cytometric analysis. The final product was mainly composed of effector memory T cells (42.0 ± 38.2%, n=8) and stem cell-like memory T cells (25.5 ± 38.0%, n=8) and contained only a minor population of terminally differentiated effector T cells. The exhaustion state of the expanded TILs was defined based on expression of the co-inhibitory markers TIM3, LAG3 and PD1 (36). A considerable proportion of TILs expressed TIM-3 (51.2 ± 34.5%, n=8), followed by LAG-3 (38.4 ± 37.5%, n=7) and only very few TILs expressed PD-1 (11.8 ± 12.5%, n=8). Only low frequencies of triple-positive TIM3⁺/LAG3⁺/PD1⁺ TILs were found in the final TIL product (5.4 ± 5.5%, n=5) indicating that TILs expanded using the TRT Process are not terminally exhausted ( Figure 2A ). To prove this hypothesis the specificity and functionality of expanded cells were assessed in vitro by co-culturing the TILs with autologous tumor cell lines. All tested TIL products showed considerably increased IFN-γ (p=0.0038, n=5) and TNF-α (p=0.0273, n=5) production as well as upregulated expression of CD107a (p=0.0058, n=5) and CD137 (p=0.0468, n=3) upon co-culture with autologous cell lines ( Figure 2B , Supplementary Figure S3 ). In the absence of autologous tumor cells or upon co-culture with an allogeneic melanoma cell line (526mel), little or no background activity was observed, suggesting that the TIL recognition was HLA and tumor-line specific ( Figure 2C ).

To automate the enrichment of antigen-specific T cells using the TRT Process, CD137 was selected as marker for magnetic separation based on previous findings (24, 25). As the development of the process was hampered by the scarcity of appropriate tumor samples, leukapheresis (LP) was used as a more accessible starting material. The process was mimicked using a mixture of viral peptides (PepTivator^® Peptide Pools) to stimulate peripheral blood cells overnight for 16 to 20 hours followed by CD137 enrichment of activated virus-specific T cells (VSTs).

Despite low starting frequencies, a high purity of CD137⁺ T cells (92.1 ± 4.0%) could be reached following the magnetic isolation. On average, the recovery of CD137⁺ T cells within the positive fraction was 33.4 ± 15.7% while the viability was 79.8 ± 12.1% with some donor dependent variability ( Figure 3A ). The enriched CD137⁺ fraction contained 36.0 ± 13.6% T cells, with a slightly higher proportion of CD4⁺ T cells. The remaining contaminating populations in the enriched fraction were CD137⁺ neutrophils, eosinophils and monocytes ( Supplementary Figure S3 ).

Low input numbers of CD137-enriched T cells (between 2x10⁵ – 1.5x10⁶ T cells) were expanded for 13 – 14 days to final cell counts of 1.4 – 5.4x10⁹ cells, corresponding to expansion rates of 1,105- to 16,864-fold. The viability of the cells at the end of the culture was consistently high ( Figure 3B ). Notably, despite the heterogenous starting cell population, the expanded cells contained > 90% CD3⁺ T cells. On average, the frequencies of CD4⁺ and CD8⁺ T cells were well balanced with slight bias towards the CD8⁺ T cells (35.5 ± 24.3% CD4⁺ and 57.8 ± 25.9% CD8⁺). Most T cells in the final product showed central memory (62.5 ± 19.5%) and effector memory (25.6 ± 14.4%) phenotype. A considerable proportion of expanded VSTs expressed TIM-3 (51.3 ± 21.8%) but only moderate populations stained positive for LAG-3 (27.8 ± 30.1%) and PD-1 (11.2 ± 6.1%). Like expanded yTILs, the expression of the exhaustion markers seemed not to be co-regulated and consequently only low frequencies of triple positive VSTs (2.7 ± 2.9%) could be detected. ( Figure 3C ).

To confirm the specificity and function of VSTs generated by the TRT Process the production of IFN-γ among CD3⁺ T cells was determined in the final product ( Figure 3D ). To this end, expanded VSTs were re-stimulated with the same virus-specific PepTivators which were initially used for stimulation and enrichment ( Figure 3D ). All VST preparations significantly upregulated IFN-γ in the presence of the viral PepTivators (VSTs p<0.0001, n=14) whereas no background production could be detected in non-stimulated samples indicating that the TRT Process with its CD137 enrichment step enables the isolation and expansion of antigen-specific T cells.

After confirming the performance of the TRT Process to enrich and expand VSTs from peripheral blood, the process was validated using melanoma tumor samples. Enzymatically digested tumor tissue containing both tumor and T cells was cultured overnight for 16 to 18 hours and subjected to the CD137 magnetic cell separation. After enrichment, the frequency of CD137⁺ cells among viable T cells was evaluated to determine the enrichment performance ( Supplementary Figure S4 , Figure 4A ). The CD137 isolation of TRTs yielded consistently high purity with CD137 frequencies among T cells increasing from 33.1 ± 11.3% prior to enrichment to 92.8 ± 6.4% following enrichment. The recovery of CD137⁺ T cells within the positive fraction was 24.6 ± 29% and the viability of the TRTs was 74.7 ± 22.0%. The enriched CD137⁺ cell fraction consisted of 93.5 ± 3.2% T cells on average, with higher proportion of CD8⁺ T cells ( Figure 4B ).

Due to the small size of the tumor samples, the overall number of enriched CD137⁺ TRTs was rather low and consequently only 1.8 – 5.7x10⁴ TRTs were subjected to the rapid expansion. Despite the exceptionally low starting cell numbers, cells could be expanded with consistently high viability to 6.3x10⁸ – 1.5x10⁹ T cells which corresponded to expansion rates of 18,158- to 79,325-fold. Expanded TRTs had high T cell purities of > 90% CD3⁺ T cells for two of the runs while the third run showed a slightly lower frequency of 88.3% ( Figure 4C ). On average, higher percentages of CD8⁺ T cells were seen at the end of the expansion (17.6 ± 9.1% CD4⁺ and 73.4 ± 24.8% CD8⁺). The predominant subpopulations among expanded TRTs displayed a central memory (33.6 ± 22.8%) and effector memory (34.2 ± 17.4%) phenotype. Like expanded yTILs, expanded TRTs were largely TIM-3⁺ (66.9 ± 48.5%) but expressed less LAG-3 (39.0 ± 35.0%) and PD-1 (18.0 ± 21.9%) and only a minority co-expressed all three markers (13.3 ± 21.5%) ( Figure 4D ).

To confirm the tumor-specific reactivity of the CD137-enriched TRTs generated using the TRT Process, autologous tumor cell lines were generated and used as target cells. For direct comparison, matched CD137-depleted TILs were expanded in parallel using the same automated REP process ( Supplementary Figure S5 ). CD137⁺ TRTs and CD137^- TILs were co-cultured with autologous tumor cell lines overnight for 18 hours and subsequently analyzed for cytokine production, expression of the degranulation marker CD107a and the activation marker CD137. As additional negative control, TRTs were cultured in the absence of target tumor cells. CD137⁺ TRTs from all three samples showed considerably increased production of IFN-γ (p=0.0858, n=3) and TNF-α (p=0.0661, n=3) as well as upregulation of CD107a (p=0.0692, n=3) and CD137 (p=0.0836, n=3) upon co-culture with autologous tumor cell lines ( Figure 5A ). In contrast, CD137^- TILs did not show any activation after co-culture with autologous tumor cell lines suggesting that tumor reactivity was almost exclusively attributed to the CD137⁺ fraction ( Figure 5B ).

To further confirm tumor specificity of the generated TRTs, one selected TRT preparation was tested against allogeneic patient derived tumor cell line as well as the HLA-A2⁺ melanoma cell line 526mel ( Figure 5C ). A considerable proportion of CD137-enriched TRTs secreted IFN-γ in response to autologous tumor cells, whereas no response could be seen against allogeneic tumor cells and the 526mel cell line. The corresponding expanded CD137^- fraction showed only negligible responses against both autologous and allogeneic tumor cells with no preference towards the autologous tumor cells. Similar trends with even higher frequencies were observed for the upregulation of CD137.

To prepare feeder cells from frozen LP or PBMCs, cells were thawed, washed with an excess of TexMACS™ GMP Medium supplemented with 3% (v/v) heat-inactivated AB serum (Cat. no. H4522-100ML, Sigma-Aldrich) and subsequently irradiated (40 Gy). To obtain tumor digest, tumor samples were cut into small pieces before processing them using the Tumor Dissociation Kit in combination with the gentleMACS™ Dissociator according to manufacturer instructions. Young tumor-infiltrating lymphocytes were generated by either culturing of a tumor fragment or tumor digest in TexMACS GMP Medium supplemented with 3-5% AB serum (Cat. no. H4522-100ML, Sigma-Aldrich), antibiotics and 6,000 IU/mL MACS^® GMP Recombinant Human IL-2. Fresh tumor digest and yTILs were cryopreserved in heat-inactivated AB serum supplemented with 10% DMSO. Tumor cells were cultured in RPMI 1640 (Cat. no. L0501-500, Biowest) supplemented with 10% FBS (Cat. no. BS-2022-500, Catus Biotech GmbH), 2 mM L-Glutamine (Cat. no. BE17-605E, Lonza) and antibiotics. If not specified otherwise, reagents and material were obtained from Miltenyi Biotec.

T cell stimulation, CD137 enrichment and expansion of cells were performed using the CliniMACS Prodigy^® Tubing Set 520 and the TRT Process. CliniMACS^® PBS/EDTA Buffer containing 0.5% (v/v) human serum albumin (HSA) and TexMACS GMP Medium supplemented with 3% (v/v) heat-inactivated AB serum (Cat. no. H4522-100ML, Sigma-Aldrich) were prepared and starting cells were transferred into a 150 mL bag. The tubing set was installed according to the instructions and buffer, media and up to 1x10⁹ viable WBCs (LP) or viable cells (tumor material) were attached to the tubing set by sterile welding. Leukapheresis from healthy donors was stimulated with multiple virus-specific PepTivator Peptide Pools (MACS^® GMP PepTivator^® AdV5 Hexon, MACS^® GMP PepTivator^® HCMV pp65, MACS^® GMP PepTivator^® EBV Select, MACS^® GMP PepTivator^® HPV16-E6 and MACS^® GMP PepTivator^® HPV16-E7) at the recommended concentration of 0.6 nmol/mL for each PepTivator Peptide Pool while tumor material was stimulated by co-culture of T cells and tumor cells (contained in the tumor digest). A sample was taken after the overnight stimulation culture (between 16 to 20 hours) to assess the CD137 frequency. CliniMACS^® CD137 GMP Biotin and CliniMACS^® Anti-Biotin GMP MicroBeads were used for magnetic labelling as recommended in the respective data sheets. Sampling was performed after labelling and after enrichment to evaluate frequencies of CD137⁺ cells. For expansion, TexMACS GMP medium supplemented with 3,000 IU/mL MACS GMP Recombinant Human IL-2 and 3% (v/v) heat-inactivated AB serum (Cat. no. H4522-100ML, Sigma-Aldrich) was prepared and attached to the tubing set to replace the medium after the stimulation culture. Between 0.01 – 2x10⁶ CD3⁺ responder cells (TRTs or TILs) were cultured with 4x10⁸ irradiated feeder cells and T cell activation reagent (MACS^® GMP CD3 pure). After the first medium bag was empty (usually before day 8), the concentration of AB serum was diluted by using TexMACS GMP Medium containing IL-2 but lacking AB serum for further media exchanges. Sterile sampling was performed in regular intervals starting on day five to assess the cell count, viability, glucose and lactate values. For the harvest, cells were automatically rebuffered and transferred into a target cell bag. If not specified otherwise, reagents and material were obtained from Miltenyi Biotec.

Samples for flow cytometric analysis were taken from unmanipulated tumor material, after stimulation, after CD137 enrichment and at different days during expansion culture. Samples were acquired by flow cytometry using the MACSQuant^® Analyzer 10 and different parameters were analyzed using the MACSQuantify™ Software. The cellular composition was assessed before and after CD137 separation and the separation performance was determined. During the initial static activation phase, the T cell activation can be monitored using the integrated camera of the CliniMACS Prodigy. During expansion, total cell count and viability were analyzed by flow cytometry while the glucose and lactate concentration of the medium were determined with the Accutrend^® Plus System (Cat. no. PZN 1696541, Roche) using the BM-Lactate (REF 03012654, measurement range 0.8−22 mmol/L) and Accutrend Glucose (REF 11447475, measurement range 20-200 mg/dL) test strips. For cell counts and cellular composition during stimulation and enrichment, the 8-Color Immunophenotyping Kit (CD19-PE-Vio^® 770 was exchanged for CD20-PE-Vio^® 770) was used, according to the manufacturer’s instructions ( Supplementary Figure S6A ). For the CD137 staining, cells were stained in CliniMACS PBS/EDTA Buffer + 0.5% HSA according to the manufacturer’s instructions using the following fluorochrome-conjugated monoclonal antibodies and reagents: CD3-FITC, CD4-APC, CD8-VioBlue^®, CD45-VioGreen™, CD137-PE or Labeling Check Reagent-PE and 7-AAD Staining Solution. CD137 was stained with a CD137 fluorochrome-conjugated antibody before and with Labeling Check Reagent after magnetic labelling. The purity was evaluated among viable CD4/CD8⁺ T cells and the recovery was calculated as follows:

For measurement of cell counts and viability during the expansion, Propidium Iodide was added to the cells directly before acquisition ( Supplementary Figure S6B ). For the phenotype staining, cells were stained in CliniMACS PBS/EDTA Buffer + 0.5% HSA according to the manufacturer’s instructions using the following fluorochrome-conjugated monoclonal antibodies: CD3-FITC, CD4-VioBlue^®, CD8-VioGreen™, CD27-VioBlue^®, CD28-APC, CD45RA-APC-Vio^®770, CD62L-PE, CD95-PE-Vio^®770, CD223 (LAG3)-PE, CD279 (PD1)-PE-Vio^®770 and CD366 (TIM3)-APC. Propidium iodide was added directly before acquisition to exclude dead cells. To characterize T cell differentiation, cells were classified into naïve (CD45RA⁺, CD62L⁺, CD95^-), stem cell-like memory (CD45RA⁺, CD62L⁺, CD95⁺), central memory (CD45RA^-, CD62L⁺), effector memory (CD45RA^-, CD62L^-) and effector (CD45RA⁺, CD62L^-) cells. If not specified otherwise, antibodies, reagents and material were obtained from Miltenyi Biotec.

After rapid expansion, the cell functionality was analyzed by flow cytometry based on the procedure described in previous publications (52, 53). Briefly, 1 – 2x10⁵ expanded TILs or TRTs cells were stimulated with 1 – 2x10⁵ cells from the corresponding autologous or allogeneic tumor cell line. CD107a-VioBlue^® was added at the beginning of the stimulation period and BD GolgiPlug™ (contains brefeldin A; Cat. no. 555029, BD Biosciences) and BD GolgiStop™ (contains monensin; Cat. no. 554724, BD Biosciences) were added after 1 hour at the recommended concentration. After 18 hours, cells were stained in CliniMACS PBS/EDTA Buffer + 0.5% HSA according to the manufacturer’s instructions using the following fluorochrome-conjugated monoclonal antibodies: CD4-VioGreen™ and CD8-FITC. For the staining of intracellular markers, the Inside Stain Kit was used according to the manufacturer’s instructions. Cells were fixed and permeabilized and the following fluorochrome-conjugated monoclonal antibodies were used: CD137-PE, CD154-APC, IFN-gamma-APC-Vio^®770 and TNF-alpha-PE-Vio^®770. The production of IFN-gamma and TNF-alpha as well as the expression of CD107a, CD137 and CD154 were assessed among CD4⁺ and/or CD8⁺ cells. If not specified otherwise, antibodies, reagents and material were obtained from Miltenyi Biotec.

Statistical analysis was performed using GraphPad Prism software. Differences between groups and the corresponding p values were determined by parametric, Student's paired t-tests (one-tailed or two-tailed, as appropriate). All p values <0.05 were considered statistically significant and are indicated as *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001.