WHAT IS ALREADY KNOWN ON THIS TOPICWHAT THIS STUDY ADDS

Here, we determined the therapeutic efficacy of TCRs in a near-clinical setting of stringent T-cell therapy to study the predictive power of preclinical analyses.

We evaluated neoantigen-specific TCRs from tumor-bearing hosts and found that TCRs from expanded T-cell clones were not all therapeutically effective.

Distinguishing failing from effective TCRs was successful by assessing the long-term persistence of T-cell products in vitro.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYBackground

The T-cell receptor (TCR) that is used to engineer T cells for adoptive therapy (TCR-therapy) largely determines treatment responses, as the TCR defines the therapeutic target as well as the function of TCR-modified T cells (TCR-Ts). This also applies to TCRs recognizing antigens resulting from somatic mutations in tumor DNA that are presented by major histocompatibility complex (MHC) molecules on the cell surface.1 These mutations are mostly caused by cancer-specific non-synonymous single nucleotide variations that result in single amino acid substitutions, also known as point mutations.2 3 Because these aberrant proteins are not expressed during T-cell development in the thymus, they are recognized by the adult immune system as new antigens and are therefore often referred to as neoantigens (NeoAg). Importantly, most NeoAgs are specific for the individual murine or human cancer in which they are found. Extensive studies in human cancers have shown that about 99% of these neoantigenic determinants are not shared between patients and are therefore appropriately referred to as “unique”.4

Provided that NeoAgs are highly and homogeneously expressed in the tumor, a TCR derived from NeoAg-specific CD8 T cells seems to be sufficient to eliminate even large and established solid tumors when used for TCR therapy.5 6 However, unmanipulated tumors often show lower NeoAg expression and are mostly characterized by heterogeneity, since they diversify genetically early in their evolution.5 7 We have shown that targeting multiple NeoAgs with TCR-Ts can reduce the risk of therapy-induced tumor escape.8 However, even TCR therapy using three NeoAg-specific TCRs derived from CD8 T cells was rarely effective when targeting autochthonous NeoAgs because the enormous heterogeneity of established tumors allowed antigen-negative cancer cells to escape.8 In contrast, we further showed that two TCRs, one isolated from CD4 T cells (reactive against a cancer-derived NeoAg presented on stromal cells) and one from CD8 T cells (targeting a NeoAg on cancer cells), are essential and sufficient to eliminate large heterogeneous cancers in mice.8

For targeting recurrent NeoAgs, the same TCRs could be used to treat specific patient subgroups.9 These off-the-shelf TCRs would be readily available; but, because they were not selected in the autologous host, they may carry the risk of unintended reactivity. Furthermore, the unique nature of most cancer mutations would require even more individualized treatment strategies. An approach in which autologous NeoAg-specific TCRs are isolated from the cancer patient’s repertoire and used therapeutically would result in a truly personalized TCR therapy. First clinical results showed no objective or only transient therapeutic effects,10 11 which may be attributed to omitting therapeutically effective CD4 T cells. The importance of the latter for tumor destruction has become more evident in preclinical studies.8 12 13 In addition, predicting which patient-derived TCRs will show therapeutic efficacy prior to clinical application remains difficult. We previously showed that convergent TCR recombination can be used as an indicator for the therapeutic efficacy of tumor-derived TCRs isolated from CD4 T cells.13

In the current study, we pursued a functional approach to predict the therapeutic efficacy of tumor-derived TCRs experimentally. We developed a model system with inducible expression of the H-2Kb-presented NeoAg p68S551F (mp68)3 that mimics obstacles faced by T cells in naturally growing tumors, such as suboptimal priming conditions due to lack of acute inflammation. Under those suboptimal priming conditions, antigenic tumors may grow to large sizes without inducing a tumor-destructive immune response and thus sneak through immune surveillance.14 15 Furthermore, the resulting chronic antigen exposure may render any newly generated NeoAg-specific CD8 T-cell dysfunctional.16 Thus, our model allowed us to delay antigen induction until the transplantation-induced inflammation had subsided. As in patients, the endogenous NeoAg-specific T-cell response was unable to prevent tumor progression. The mp68-specific tumor-infiltrating T cells comprised a diverse repertoire of TCR clonotypes and we tested whether expanded T-cell clones contain therapeutically effective TCRs. Only about one-half of the NeoAg-specific TCRs we isolated induced tumor destruction when expressed in peripheral T cells and adoptively transferred. Similar results were obtained when analyzing TCRs isolated from the spleens of mice immunized with mp68-expressing cancer cells. Predicting the therapeutic efficacy of TCRs in vivo required in vitro assays that assessed TCR-T persistence rather than immediate T-cell responses and antigen sensitivity. These results were confirmed in experiments comparing the therapeutic efficacy of human TCRs.

Materials and methodsPlasmids

Genes encoding mp68-specific TCRs (table 1) were designed as described (TCR-β-P2A-TCR-α),5 synthesized (GeneArt, Thermo Fisher), and cloned into pMP71.17 CDK4R24L-specific TCRs (table 2) are described (14/35, P1, H1-1, H2-1, H2-2)6 18 or were generated in this work (M1, see below). TCR genes comprising native leader sequences, and constant regions of mouse TCR-αβ counterparts19 were cloned into pMP71. For studying TCR function in vivo, CDK4R24L-specific TCRs were cloned into pMP71-IRES-GFP. For doxycycline-inducible expression, three sequences encoding for the mp68 epitope were separated by AAY, fused to Thy1.1 ((SNFVFAGI-AAY)3-Thy1.1) and cloned in the vector pMOV20 (pMOV-mp68-Thy1.1). For constitutive expression, (SNFVFAGI-AAY)3-Thy1.1 was cloned into pMP71 (pMP71-mp68-Thy1.1). The transgene cassettes of both epitope-encoding fusion constructs started with a methionine to initiate transcription.

Table 1

Sequences of TCR clonotypes expanded in mp68-tetramer-binding TILs of mice bearing 4E9ON tumors or in T cells of immune spleens after immunization with 4E9ON cancer cells

Table 2

Sequences of TCRs targeting the HLA-A2-presented neoantigen CDK4R24L

Cells

Cells were cultured in Roswell Park Memorial Institute (RPMI) medium with 5% heat-inactivated fetal calf serum (FCS, PAN Biotech) and 100 U/mL penicillin/streptomycin (cell medium, CM). 8101 cancer cells are derived from a tumor that arose in an UV-irradiated C57BL/6 mouse3: 8101 (regressor phenotype, mp68+), 8101PRO (progressor phenotype, mp68−), and 8101–12 (clone 12, regressor phenotype, mp68+). Presence or absence of mp68 was verified by sequencing as described.7 Transduction of 8101PRO using MP71-mp68-Thy1.1 (8101PRO-mp68) and 8101–12 using MP71-GFP (8101–12-GFP) was done as described6 followed by enrichment using flow cytometry. Relapse variants of 8101PRO-mp68 that occurred after TCR-therapy were seeded as single cell suspensions and kept in culture for at least three passages to remove stromal content. 8101PRO-4E9 cancer cells (4E9) were generated by transduction of 8101PRO using MOV-mp68-Thy1.1. 4E9 cancer cells were cultured for 48 hours in CM containing 500 ng/mL doxycycline to induce mp68 expression in vitro (4E9ON). MC703 cancer cells6 are derived from a methylcholanthrene-induced tumor of an HHD21 mouse. MC703-ALD6 is its derivative with constitutive expression of the ALD epitope. WM-902B+A26 are melanoma cells natively expressing CDK4R24L and with ectopic expression of HLA-A2 and GFP. Plat-E22 and 293-RD11423 packaging cells were used for generating ecotropic or amphotropic virus supernatant, respectively. The murine T-cell line 5824 lacks endogenous expression of TCR-αβ genes.

T-cell culture and transduction

T cells were cultured in RPMI supplemented with 10% FCS, 1 mM sodium pyruvate, 100 µM non-essential amino acids and 50 µM 2-mercaptoethanol (T-cell medium, TCM). Transfection of packaging cells was performed with Lipofectamine (Thermo Fisher) and 3 µg of pMP71 plasmids as detailed above (see Plasmids). Virus supernatant harvested 48 hours later. Primary human T cells: 1.5×106 human peripheral blood mononuclear cells were isolated as described6 seeded in TCM in 24-wells coated with anti-CD3 (OKT3, 5 µg/mL) and anti-CD28 (CD28.2, 1 µg/mL) antibodies 48 hours before transduction. Transduction was performed on two consecutive days by spinoculation for 90 min at 800 g and 32°C with 1 mL amphotropic virus. T cells were kept in the continued presence of human interleukin (IL)-2 (13 days at 400 U/mL, 2 days at 40 U/mL, Novartis) before being frozen and used for experiments. Primary mouse T cells: single cell suspensions of spleens of C57BL/6 or HHD mice were generated as described6 and cultured at 2×106 cells/mL with anti-CD3 (clone: 145–2 C11, 1 µg/mL), anti-CD28 (37.51, 0.1 µg/mL), and human IL-2 (40 U/mL) to activate T cells. Transduction was performed 24 and 48 hours later. For this, 1.5×106 cells were seeded in virus-coated6 wells together with protamine sulfate (4 µg/mL, Sigma), 4×105 mouse T-Activator beads (Thermo Fisher), and IL-2 (40 U/mL) together with 1 mL ecotropic virus supernatant and spinoculated for 30 min at 32°C and 800 g. Fresh virus supernatant was added on the following day for a second spinoculation. T cells were further cultured in IL-15 (50 ng/mL, Miltenyi Biotec) and expanded for 9 days before being frozen and used for in vitro assays. For adoptive transfer, T cells were used 3 days after completing transduction and after removing T-Activator beads using a magnet. 58 cells: 1×105 cells were seeded in 24 wells, before adding ecotropic virus supernatant 24 hours and 48 hours later, each time followed by spinoculation (30 min, 32°C, 800 g). Cells were further expanded for functional assays.

Animals

The ARRIVE reporting guidelines were followed to provide information on the use of laboratory animals.25 Mice (Rag−/− (B6.129S7-Rag1tm1Mom), TNA2 (B6.Cg-Tg(HLA-A/H2-D/B2M)1Bpe H2-D1tm1Bpe B2mtm1Unc Rag1tm1Mom/Luck),6 C57BL/6, HHD,21 and ABab-A226 were bred in a specific pathogen-free environment in the animal facility of the Max-Delbrück-Center for Molecular Medicine. Tumor induction: 3–5×106 8101PRO-mp68, MC703-ALD,6 or 4E9OFF cancer cells were injected in 100 µL phosphate-buffered saline (PBS) subcutaneously into the right flank of either Rag1−/−, TNA2, or C57BL/6 mice (12–20 weeks old, female or male), respectively. Presence or absence of antigen expression was verified by flow cytometry assessing Thy1.1 or GFP expression prior to cancer cell injection. Tumor size was determined using a caliper three times per week according to π/6 × (abc). Mice were sacrificed when either tumors reached the maximum permitted size, the maximum observation time was reached (150 days) or if due to tumor burden the overall health condition was poor. To compare TCR quality, the T-cell therapy experiments were designed with a group size sufficient to distinguish TCRs based on the extent of tumor regression. Examiners were not blinded with respect to treatment groups. Other potential confounders were not accounted for. Animals were excluded from analysis, if they died due to reasons unrelated to tumor burden. Immunization: C57BL/6 mice were inoculated with 3–5×106 lethally irradiated 4E9ON cancer cells (20 Gy). Antigen expression was confirmed by staining for Thy1.1 prior to irradiation. Doxycycline treatment of mice (see below) started 48 hours prior to tumor cell inoculation. Two consecutive boosts using live 4E9ON cancer cells were performed 6 and 10 weeks later. 10 days after the second boost, mice were sacrificed to isolate their spleen. TCR-M1 was generated by immunizing an ABab-A2 mouse as described.27 The mouse received three times 1–2 µg of plasmid encoding the CDK4R24L epitope ALD by gene gun and three times with peptide (100 µg) combined with Cytosine-phosphate-Guanine (CpG) oligonucleotides and incomplete Freund’s adjuvant. The boosts were performed 1, 5, 7, 9 and 10 months after the initial immunization. 8 days after the last immunization, the mouse was sacrificed and splenocytes were cultured for 10 days in the presence of 1×10–8 M ALD peptide and then sorted by flow cytometry using interferon (IFN)-γ capture assay as described. Adoptive T-cell transfer: mice received indicated TCR-Ts intravenously 2–4 weeks after tumor induction. Treatment groups were allocated with a similar average tumor size. TCR-Ts were used on day 5 of ex vivo culture (3 days after completing TCR transduction). The total number of transferred cells was adjusted according to the transduction rate to transfer 1×106 CD8+ TCR-Ts per mouse. TCR-Ts were injected intravenously in 100 µL PBS. Control mice received unmodified T cells or no T cells. Doxycycline treatment: C57BL/6 mice bearing 3 weeks old 4E9OFF tumors or to receive 4E9ON cancer cells received 200 µg doxycycline per milliliter drinking water supplemented with 5% sucrose changed two times a week. 4E9OFF tumors were at least palpable at this point.

TCR sequencing

Single T-cell sequencing: 4E9ON tumors were excised, homogenized (gentleMACS dissociator (Miltenyi Biotec)) and enzymatically digested in collagenase D (2 mg/mL, Roche), DNAse I (20 U/mL, Sigma) for 90 min, and additional 30 min in 0.025% trypsin at 37°C before analyzing single cell suspensions by flow cytometry. Single cell suspensions of immune spleens were viably frozen until used for analysis by flow cytometry. Cells were stained using H-2Kb:mp68 tetramers (PE, MBL, dilution: 1:200) and antibodies against CD8 (BV421, clone: 53–6.7, 1:100), CD3 (APC, 145–2 C11, 1:50), and 7-AAD (BioLegend). Live, single CD3+CD8+ mp68-tetramer-binding cells were sorted (BD Aria Fusion) into 96-well PCR plates containing 12 µL of 1× One-Step RT-PCR buffer (Qiagen) and snap frozen on dry ice as described.28 TCR sequences of single cells were determined according to previously published protocols.28–30 Briefly, TCR-encoding messenger RNAs were reverse-transcribed and amplified in a PCR reaction using the One-Step RT-PCR kit (Qiagen) and TCR-specific primers.30 Using nested primers,30 a second PCR was performed and barcodes28 were added in a third PCR reaction using HotStarTaq DNA polymerase (Qiagen). Amplified and barcoded TCR sequences were then pooled, gel-purified and sequenced using MiSeq Reagent Kit v2, 500 cycles (Illumina) and the Illumina MiSeq instrument for paired-end sequencing. The sequencing data were processed as described.28 Bulk T-cell sequencing: the TCR repertoire of T cells sorted with mp68-tetramers from the tumor of mouse T1 was determined as described.7 The identification of TCR genes from the ABab-A2 mouse immunized against CDK4R24L was done as described.27 Identity of TCR genes was determined using the International ImMunoGeneTics database.31

Flow cytometry

Single cell suspensions of tumors (see above) or spleens were stained with anti-CD39 (APC, Duha59, 1:100), anti-PD-1 (BV711, 29F.1A12, 1:50), anti-CD8 (APC-Cy7, 53–6.7, 1:100), anti-CD3 (BV421, 145–2 C11, 1:50) and 7-AAD. Appropriate isotype controls for anti-Programmed Death (PD)-1 and anti-CD39 were included for all samples. TCR expression in primary mouse TCR-Ts was assessed by staining with anti-CD8 (BV421, 53–6.7, 1:100) and antibodies detecting the variable regions of TCR-β: TCRvβ2 (B20.6, 1:10), 5 (MR9-4, 1:50), 6 (RR4-7, 1:50), 8.1 (MR5-2, 1:50), 11 (KT11, 1:50), 13 (MR12-4, 1:50) all in PE from BioLegend. TCR expression in TCR-Ts derived from primary human T cells was analyzed using anti-CD8 (BV421, 53–6.7, 1:100) and antibodies detecting the mouse constant region of TCR-β (APC, H57-597, 1:50) or using GFP when used as a marker for TCR expression (integrated via an IRES to the expression vector).6 TCR expression in 58 cells was determined by detecting CD3 (APC, 145–2 C11, 1:50) and intracellular/extracellular presence of mouse constant TCR-β (APC, H57-597, 1:50). Intracellular staining was performed using the Cytofix/Cytoperm Fixation/Permeabilization Kit (BD). Reisolated cancer cells of 4E9ON and 8101PRO-mp68 tumors were stained with anti-Thy1.1 (APC, OX-7, 1:500) to confirm Thy1.1 expression. SYTOX Blue (BD) or 7-AAD was used to discriminate live and dead/apoptotic cells in each measurement. If not stated otherwise, antibodies were purchased from BioLegend.

T-cell function

Short-term: 24 hours co-cultures of 5×104 TCR-Ts with target cells (Effector-to-Target ratio: 1:1) were performed in 96-well plates. Target cells were irradiated (63 Gy) splenocytes of C57BL/6 mice loaded with the indicated mp68 peptide (SNFVFAGI (purity: >95%, HPLC-purified), Biosyntan), 8101 and 8101PRO-mp68. Serial dilutions of the mp68 peptide were performed in the range of 1×10–6 to 1×10–12 M. Non-engineered T cells, and TCR-Ts cultured without target cells or with 1 µM ionomycin (Calbiochem) and 5 ng/mL phorbol-12-myristate-13-acetate (Promega) for TCR-independent cytokine release were used as controls. Supernatants of co-cultures were analyzed for IFN-γ content by ELISA (BD). Long-term: 2×103 tumor cells (WM-902B+A2 or 8101–12-GFP) were seeded in 96-well plates before adding 1×104 CD8+ TCR-Ts 24 hours later. The total number of T cells per well was kept constant by adding non-engineered T cells. Mouse TCR-Ts were used without cryopreservation. Every 3 days, 2×103 fresh tumor cells were seeded into the wells (re-challenge). Co-cultures were analyzed by continued imaging (every 2 hours with a 10× objective) using an Incucyte SX5 (Sartorius). Outgrowth of GFP-expressing tumor cells was determined as the time when confluence reached 10% (8101–12-GFP) or 15% (WM-902B+A2).

Statistics and software

Statistical calculations were performed using Prism V.9 (GraphPad). Fluorescence-Activated Cell Sorting (FACS) data was analyzed using FlowJo V.10 (FlowJo). SnapGene (GSL Biotech LLC) was used to analyze DNA sequences.

ResultsEndogenous mp68-specific CD8 T cells infiltrate tumor tissue but fail to prevent cancer progression

The 8101 cancer cells harbor the autochthonous NeoAg mp68 (DDX5S551F) that induces a CD8 T-cell response when the cancer was used for immunization of C57BL/6 mice.3 The mp68 peptide binds with high affinity to H-2Kb32 and proved to be a rejection antigen for T-cell therapies targeting large established 8101 tumors with high and homogenous mp68 expression.5 To mimic induction and prolonged exposure of T cells to mp68 in immunocompetent mice, we used the progressor tumor variant 8101PRO that was generated by serial transplantation of 8101 cancer cells into C57BL/6 hosts.33 Due to in vivo selection, 8101PRO lost mp68 as verified by sequencing of genomic DNA (figure 1A). We reinstalled mp68 expression in 8101PRO under control of a doxycycline-inducible promoter20 along with Thy1.1 as a surface marker. Transcription of mp68 in the derivative clone 8101PRO-4E9 (hereafter 4E9ON/OFF) was tightly controlled, as shown by doxycycline-dependent expression of Thy1.1 (figure 1B) and recognition of 4E9ON but not 4E9OFF cancer cells by mp68-specific TCR-Ts (figure 1C). The mp68-specific TCR for generating TCR-Ts was obtained by immunizing C57BL/6 mice with 8101 cancer cells (designated 1D95). When 4E9OFF cancer cells were injected into immunocompetent mice, most animals developed measurable tumors about 3 weeks later (online supplemental figure 1A, figure 1D). Expression of mp68 was initiated 3 weeks after transplantation, when acute inflammation at the injection site had subsided34 and tumors were 142±132 mm3 in size (figure 1D). However, the expression of mp68 could not prevent tumor progression, even when induced in small, only palpable 4E9 tumors (figure 1D). A comparison of 4E9ON and 4E9OFF tumors 3 weeks after induction of mp68 expression showed no significant difference in size (online supplemental figure 2). All progressing 4E9ON tumors were infiltrated by CD8 T cells expressing high levels of PD-1 and CD39 (figure 1E), indicating their terminally dysfunctional phenotype35 and providing context for their failure to arrest tumor growth. A fraction of these CD8 tumor-infiltrating lymphocytes (TILs) bound a tetramer loaded with the mp68 peptide (figure 1F), thus mirroring NeoAg-specific T-cell responses observed in patients.36 TILs isolated from progressing 4E9OFF tumors were unable to bind the mp68 tetramer (figure 1F).

Figure 1

Induction of mp68 neoantigen expression in a progressively growing cancer causes infiltration of mp68-specific T cells that are unable to suppress tumor growth. (A) 8101 cancer cells harbor a serine (S) to phenylalanine (F) exchange in position 551 of the p68 protein that is caused by a point mutation (cytosine to thymidine transition). Electropherograms show the presence or absence of the point mutation (red) in 8101 (mp68+, heterozygous) and 8101PRO (mp68−). (B) 8101PRO-derived 4E9 cancer cells are mp68− (4E9OFF) and express the mp68-Thy1.1 fusion gene on incubation with doxycycline (4E9ON). Histograms show surface expression of Thy1.1 in indicated cells as determined by flow cytometry. (C) 1D9 TCR-Ts secrete IFN-γ when incubated with 4E9ON but not 4E9OFF cancer cells. Depicted values are technical replicates of an experiment that was repeated nine times showing comparable results. P/I, phorbol-12-myristate-13-acetate/ionomycin. (D) Induction of mp68 expression cannot prevent progression of 4E9ON tumors in C57BL/6 mice. Data are compiled from two independent experiments and plotted on separate x-axes according to growth kinetics. Curves are shown for tumors isolated 3 weeks (upper left, mouse T1 (red) and T2 (cyan) are highlighted), 5 weeks (upper right, T3 in orange), and 14.5 weeks (lower left, T4 in purple) after inducing mp68 expression 21 days after inoculation. Other mice in the treatment group are shown as open circles. Growth of 4E9OFF tumors is shown as control (lower right, gray squares). (E–F) 4E9ON-infiltrating CD8 T cells show surface markers of exhaustion (E) and contain a fraction that binds mp68-tetramers (F). CD3+CD8+ TILs from reisolated 4E9ON tumors were analyzed by flow cytometry to assess expression of CD39 and PD-1 (E) and binding of mp68-tetramers (F) using material from three to nine animals, respectively. Mice T1-T4 are color-coded. Additional mice are shown as open circles. Representative FACS plots show TILs of mouse T3 (E) and T1 (F) gated on CD3+CD8+ cells (left). A comparison with corresponding splenocytes (E) and TILs from 4E9OFF tumors (F) is shown as a scatter plot (right). The number (n) indicates the sample size for each graph. P values result from paired t-test (E) or unpaired t-test with Welch’s correction (F). FACS, Fluorescence-Activated Cell Sorting; IFN, interferon; P/I, phorbol-12-myristate-13-acetate/ionomycin; PD-1, Programmed Death-1; TIL, tumor-infiltrating lymphocyte.

Endogenous mp68-specific CD8 T cells express a diverse repertoire of unique TCR clonotypes

To characterize the mp68-specific CD8 T-cell response in 4E9 tumor-bearing mice, we sorted TILs of mouse T1 (figure 1D) that bound the mp68 tetramer (figure 1F) and captured their TCR-α and TCR-β repertoire. Surprisingly, TCR loci within this T-cell population showed combinations of many different variable and joining regions (TRA: 17, TRB: 22; figure 2A,B) comprising a large number of different CDR3s (TRA: 60, TRB: 100). However, only three of the V-J combinations found in the TCR-α and TCR-β loci accounted for more than 10% of the repertoire and encoded distinct CDR3 sequences, indicating expansion of certain T-cell clonotypes (figure 2A,B). To identify the clonal ancestry of these frequent TCR genes, we sorted~180 single mp68 tetramer-binding T cells from the CD8 population of TILs and sequenced their individual TCR-αβ pairs.29 30 As expected, the most abundant TCR-α and TCR-β genes from the repertoire analysis (figure 2A–B) were found in the majority of the single T cells (table 1), confirming the TCR identity of three expanded T-cell clonotypes (figure 2C). We continued the analysis for three additional mice by isolating single mp68-specific CD8 T cells from TILs of mouse T2 (analyzed 3 weeks after induction of mp68 expression), mouse T3 (5 weeks), and mouse T4 that was analyzed after 14.5 weeks in which T cells were exposed to mp68 in the tumor microenvironment (figure 1D, middle panels). Mouse T1, T2, T3, and T4 comprised a total of 10 expanded CD8 T-cell clonotypes, and at least two clonotypes were expanded in each mouse. The TCR sequences of the expanded T-cell clonotypes differed between and within the investigated mice, with differences characterized by usage of different variable and joining regions and largely different CDR3s (table 1). Furthermore, all non-expanded TCR clonotypes identified in the four mice (n=84) were unique and not found in TILs of another mouse. This tremendous TCR diversity was not limited to the repertoire of mp68 tetramer-binding TILs, as additional TCR clonotypes (n=33) were found in spleens of C57BL/6 mice (V1, V2) following immunization with 4E9ON cancer cells. The mice received lethally irradiated 4E9ON cancer cells and subsequently rejected live 4E9ON cancer cells in two consecutive boosts 6 and 10 weeks after initial immunization (online supplemental figure 1B). T cells in immune spleen cells of these mice contained three TCR clonotypes each that were expanded (figure 2D and table 1), but like all other sequences, these TCRs were again unique and not found in the other immunized mouse or in any of the mp68 tetramer-binding TILs.

Figure 2

T cells that bind mp68-tetramers isolated from TILs of 4E9ON tumors or immunized mice comprise a diverse repertoire of TCRs. (A–B) The frequency of TRAV-TRAJ (A) and TRBV-TRBJ (B) combinations found in TCR genes of TILs isolated from the 4E9ON tumor of mouse T1 is shown as a heat map. Combinations with a frequency above 10% in the repertoire are indicated in red and their CDR3 usage is shown. (C–D) TCR sequences of single CD8+ T cells binding the mp68-tetramer were determined and the relative abundance of T-cell clonotypes is indicated in pie charts. T cells were isolated from 4E9ON tumors of mice T1 (red), T2 (cyan), T3 (orange), and T4 (purple) (C) or from spleens of 4E9ON-immunized mice V1 (blue) and V2 (green) (D). Expanded T-cell clonotypes are color-coded and the fraction of non-expanded clonotypes, which are represented in ≤10 (C) or ≤5 wells (C-D), are indicated in gray (C) and white (C-D). The number of clonotypes contained in the non-expanded fraction is given. TCR, T-cell receptor; TIL, tumor-infiltrating lymphocyte; TRAJ, TCR-α chain joining region; TRAV, TCR-α chain variable region; TRBJ, TCR-β chain joining region; TRBV, TCR-β chain variable region.

Effective and failing TCRs are found in expanded mp68-specific clonotypes of tumor-bearing or immunized mice

All TCR genes derived from CD8 T cells that expanded in response to mp68 in tumor-bearing and vaccinated mice were molecularly cloned. When introduced into TCR-negative 58 cells, all TCRs assembled with endogenous CD3 and thus supported functional TCR expression (online supplemental figure 3A,B). TCR-Ts were similarly generated from primary T cells of C57BL/6 mice (online supplemental figure 3C). 3 of the 16 TCRs (T1-2, T1-3, and T3-3) failed to bind mp68-loaded tetramers (online supplemental figure 3C,D) and showed no reactivity when incubated with target cells loaded with mp68 peptide (online supplemental figure 3E). A second TCR-α chain found in TILs T3-3 failed to confer mp68 reactivity (TCR T3-3/2, online supplemental figure 3D,E). In TILs encoding for TCRs T1-2 and T1-3, no second subdominant TCR-α chain was found. These three non-reactive TCRs were excluded from subsequent analyses. Next, we evaluated the therapeutic efficacy of mp68-reactive TCR-Ts by treating Rag−/− mice that had established 8101PRO tumors that constitutively expressed a fusion gene encoding mp68 and Thy1.1, which served as a surface marker (8101PRO-mp68, online supplemental figure 1C). A small fraction of 8101PRO tumor cells (mp68-Thy1.1-negative) remained in the 8101PRO-mp68 population (online supplemental figure 4), allowing the therapeutic efficacy of the TCR-Ts to be measured by their ability to select for these cancer cell variants. Although all mp68-reactive TCRs supported antigen-specific expansion in tumor-bearing (T1, T2, T3, T4; n=7) or immunized mice (V1, V2; n=6), less than half of the analyzed TCR-Ts (n=6) were therapeutically effective (figure 3). Therapeutic efficacy was indicated by tumor regression (figure 3A) and recurrence of mp68-Thy1.1-negative cancer cell variants (figure 3B). In three treatment groups (T2-2, T4-1, V2-1), the varying percentage of mp68-Thy1.1-negative cancer cell variants in the reisolates aligned with the therapeutic outcome observed in the individual mice (figure 3A,B). The therapeutically effective cluster of TCRs included those from TILs (n=4) and immunized mice (n=2), with one TCR from each group (T1-1, V2-1) achieving tumor eradication in a fraction of treated mice, suggesting that the therapeutic quality of NeoAg-specific TCRs derived from TILs was not inferior to that of TCRs isolated from tumor-free environments. Importantly, one but not all of the isolated TCRs from each of the four tumor-bearing mice were therapeutically effective, necessitating preliminary screening to avoid treatment failure.

Figure 3

Less than half of the TCRs isolated from mp68-tetramer-binding TILs or T cells of immunized mice are therapeutically effective. (A) The therapeutic quality of TCR-Ts was determined by TCR-therapy in mice bearing large established 8101PRO-mp68 tumors and is represented by their ability to destroy tumors. Shown are tumor growth curves of individual mice pooled from two to three independent experiments. The number of treated mice in each group is shown in the upper right of each graph. Injection of 8101PRO-mp68 cancer cells is indicated with an arrow. 2–3 weeks after cancer cell injection, TCR-Ts (color-coded for each originating mouse and TCR clonotype) were adoptively transferred (day 0, dashed arrows). Mice receiving unmodified T cells (UT) are shown as control. The average tumor size at treatment start was 182±103 mm3. The number (n) indicates the sample size for each graph. (B) Loss of Thy1.1 expression on reisolated cancer cells of relapsing 8101PRO-mp68 tumors is indicative of treatment success. Reisolated cancer cells were adapted to culture and the percentage of Thy1.1+ cells was determined by flow cytometry. Lack of Thy1.1+ cells indicates the selection of mp68− escape variants by the respective TCR-Ts. Each data point represents one tumor. Mean and SEM are shown. TCR-Ts, T-cell receptor-modified T cells; TIL, tumor-infiltrating lymphocyte.

Therapeutic efficacy of TCR-Ts is predicted by their persistent control of tumor growth in vitro

The activation profile of TCR-Ts after incubation with target cells either loaded with graded amounts of mp68 peptide (figure 4A) or endogenously expressing mp68 at high (8101PRO-mp68, figure 4B, left) or natural (8101, figure 4B, right) levels revealed a significant difference between the group of therapeutically effective and failing TCRs. These data were derived from four to eight individual experiments (online supplemental figure 4A–C), which demonstrated reproducibility but also inherent variations across repetitions. A cluster of three failing TCRs (V2-2, V1-3, T3-2) showed an activation profile that overlapped with the group of therapeutically effective TCRs (figure 4, online supplemental figure 5), making them difficult to discriminate. We therefore evaluated the long-term persistence of antitumor activity by repeatedly challenging TCR-Ts with tumor cells (online supplemental figure 6). Similar to the control of cancer progression in vivo, we monitored whether outgrowth of 8101 cancer cells (8101–12-GFP) could be prevented by TCR-Ts in vitro. Indeed, the time that TCR-Ts could inhibit tumor progression in cell culture was found to be an accurate predictor of therapeutic efficacy in vivo (figure 4C, online supplemental figure 5D). Importantly, the cluster of three failing TCRs that was incorrectly suggested to induce robust antitumor responses by short-term assays was also clearly identified as ineffective when TCR-T persistence was used as a parameter (figure 4C, online supplemental figure 5D).

Figure 4

Long-term, but not short-term in vitro assays are accurate predictors of in vivo efficacy of mp68-specific TCR-Ts. TCR-Ts expressing mp68-specific TCRs derived from TILs of 4E9ON tumors or immunized mice were incubated with splenocytes loaded with graded amounts of mp68 peptide (A), 8101 cancer cells with ectopic (B, left) or native expression of mp68 (B, right and C). TCRs were classified as effective or failing based on their in vivo efficacy (figure 3) and are represented as single dots showing an average of three to eight independent experiments (detailed in online supplemental figure 5). (A) EC50 is the concentration of mp68 peptide required to elicit half-maximal IFN-γ release by TCR-Ts when loaded on splenocytes in 24 hours co-cultures. (B) IFN-γ release by TCR-Ts was determined after 24 hours co-culture with 8101PRO-mp68 (left) or 8101 cancer cells (right). (C) TCR-Ts were incubated with 8101–12-GFP cancer cells and time to tumor outgrowth (≥10% confluence) was assessed by monitoring co-cultures using Incucyte SX5. Additional 8101–12-GFP cancer cells were added to the culture every 3 days. The range of the data set for therapeutically effective TCRs is indicated in gray. Failing TCRs that fall within this range are indicated. P values result from unpaired t-tests. IFN, interferon; TCR-Ts, T-cell receptor-modified T cells; TIL, tumor-infiltrating lymphocyte.

CDK4R24L-specific TCRs from patients are therapeutically effective

We extended the analysis to a set of human TCRs targeting an HLA-A2-presented NeoAg derived from mutant CDK4 (CDK4R24L, table 2). A TIL-derived CDK4R24L-specific TCR (hereafter referred to as P1)37 was compared with several TCRs obtained from tumor-free settings: (1) CDK4R24L-specific TCRs were generated by in vitro stimulation of T cells of healthy donors. For this, dendritic cells of two donors (H1, H2) were loaded with CDK4R24L-encoding RNA and used to prime and expand autologous T cells in cell culture.38 TCRs from derivative clones of these T-cell lines (hereafter referred to as H1-1, H2-1, and H2-2) have been described.18 (2) We further isolated a TCR (referred to as M1) from a transgenic mouse that expressed the entire human TCR-α and TCR-β loci together with HLA-A226 and was immunized with the CDK4R24L NeoAg. We further used a validated patient-derived TCR from previous studies that developed in response to CDK4R24C but showed cross-reactivity to CDK4R24L (TCR 14/356 39). The CDK4R24L-specific TCRs were all unique with no apparent sequence similarities (table 2). We used this set of TCRs to verify whether long-term in vitro experiments would predict the therapeutic value of the different human CDK4R24L-specific TCRs. TCR-Ts were generated by engineering human T cells obtained from peripheral blood of healthy donors (online supplemental figure 7A) and repeatedly challenged with cancer cells that endogenously expressed CDK4R24L (WM-902B+A2) (online supplemental figure 6). One TCR (H2-2) failed to inhibit cancer cell growth (figure 5A). All other TCRs (P1, 14/35, H1-1, H2-1, M1) similarly controlled tumor cel