Chimeric PD-1:28 Receptor Upgrades Low-Avidity T cells and Restores Effector Function of Tumor-Infiltrating Lymphocytes for Adoptive Cell Therapy

Magnitude, quality, and functional polarization of an immune response are crucial for the control of many diseases, including cancer. Examples of successful implementation of these principles are the adoptive transfer of selected highly tumor-reactive T cells (1) and checkpoint inhibition (2), which achieve remarkable tumor regressions even in patients who have failed previous treatments. Despite these successes, immunotherapeutic concepts need to be improved as no or low response is still a frequent outcome. Although there is consensus that durable antitumor responses require highly active T cells, it is not clear how this is best achieved. At least two major impediments have been identified that impinge upon achieving this prerequisite: inherent intermediate- to low-affinity T-cell receptors (TCR) of antitumor T cells that may not mount a successful tumor-specific response (3, 4), and functional inactivation of T cells in the tumor milieu (5–8).

Experimentally generated high-affinity TCRs might be advantageous by boosting effector function and proliferation, thereby persisting longer and in larger numbers as compared with T cells with low-affinity TCRs, but they may also bear an increased risk of unwanted toxicities (3, 4, 9–12). Low-avidity T cells would gain value if their effector functions could be enhanced, as they are reported to be less susceptible to tolerization and may represent a lower toxicity risk (3, 4, 13, 14).

Loss of function following tumor infiltration occurs independently of T-cell avidity (5–8). Thus, preventing or delaying functional loss and endowing reactive T cells with the capacity to maintain antitumor effector functions longer in the tumor environment are imperatives for achieving more effective T-cell response. Functional inactivation may occur due to blockade of TCR signaling, loss of cytotoxic proteins, lack of persistence, and functional polarization from an antitumor Th1 to a less favorable Th2 phenotype (15–21).

To overcome these activation restrictions, we focused on costimulation and coinhibition signals controlling strength and duration of the T-cell response (22). In challenging environments, costimulation has been shown to improve effector function as it enables refill of lytic granules and T-cell expansion (23, 24). Thus, we generated chimeric receptors composed of the PD-1 extracellular domain and the CD28 signaling domain (PD-1:28) following published designs (25, 26). PD-1:28–engineered T cells should receive costimulatory signals to support survival, proliferation, and TCR signaling when interacting with PD-1 ligand (PD-L1) on tumor cells. Simultaneously, coinhibitory interaction with expressed native PD-1 should be opposed, ameliorating another major hurdle of efficient antitumor function in PD-L1⁺ tumor milieus. All this will occur without changing the TCR specificity.

Here, we demonstrate that PD-1:28 chimeric receptors upgrade low-avidity T cells and restore effector function of tumor-suppressed human tumor-infiltrating lymphocytes (TIL). Through PD-1:28 signaling, T cells receive support in processes essential for tumor growth control, that is, stronger proliferation in solid tumor milieus and prevention of Th2 polarization. Moreover, PD-1:28–facilitated T-cell support seems to work synergistic with anti-PD-L1 checkpoint blockade, suggesting that PD-1:28 engineering of T cells can be beneficially combined with checkpoint blockade and potentially other immunotherapies.

Cell lines were grown in RPMI1640/1% l-glutamine/1% nonessential amino acids/1% sodium pyruvate/1% penicillin/streptomycin (RPMI basic), and 12% FCS at 37°C/6.5% CO₂. Cell lines were kept in continuous culture for no longer than 2 months. Mycoplasma testing was performed monthly using VenorGeM Classic (Minerva Biolabs). All cell lines were authenticated by flow cytometry to express the relevant molecules, HLA-A2, tyrosinase or TCR53 antigen, without or with PD-L1, as required (see Results). The following cell lines were used: HEK293 cells (positive for HLA-A2, negative for endogenous PD-L1 by flow cytometry, obtained from H. Engelmann, Institute of Immunology, Munich, Germany, in 1997) were transduced in our laboratory to express tyrosinase (HEK/Tyr) or tyrosinase and PD-L1 (HEK/Tyr/PD-L1). Single-cell clones were selected for comparable HLA-A2 and tyrosinase expression (see Results); human melanoma lines: SK-Mel23 (ATCC HTB71, obtained from M.C. Panelli in 2001, NIH, Bethesda, MD), FM3, FM55, FM86 (obtained from P. thor Straten in 2014, Danish Cancer Society, Copenhagen, Denmark), WM115 (ESTDAB-066), WM266.4A (ESTDAB-076; both purchased in 2003); human nonmelanoma lines: renal cell cancer RCC-53 and RCC-26 (generated in our laboratory in 1993 and 1991, respectively), squamous cell carcinoma UT-SCC-15 and glioblastoma U-373 (obtained from co-author P.J. Nelson in 2002; ref. 27). Murine mastocytoma line P815 (ATCC TIB-64, purchased in 1995) was authenticated by morphology and expression of FcγR, which is required for antibody coating.

Murine hepatocellular carcinoma (HCC) line 434 originated in LoxP-TAg mice (28) and was grown in RPMI-basic/12% FCS/0.05% β-mercaptoethanol (mouse medium). Authentication was done by testing SV40LT, H2-D^b, and PD-L1 expression.

Human peripheral blood mononuclear cells (PBMC) were from healthy donors (HD). Transduced T cells were grown in RPMI-basic/10% human serum (T-cell medium) and 50 U/mL IL2, or as indicated. TILs were tissue suspensions prepared from postsurgery clear cell renal cell carcinoma (ccRCC) as described previously (16). Detailed pathologic features are described in Supplemental Material. Blood and tissues were collected with donors' and patients' informed consent in accordance with the Declaration of Helsinki and after approval by the Institutional Review Board of Ludwig-Maximilian University (Munich, Germany).

pMP71 plasmids encoding TCR-T58, TCR-D115 (both tyrosinase-epitope(AA_366-377)-specific/HLA-A2-restricted), and TCR53 have been described previously (27). Various costimulatory chimeric receptors were designed on the basis of the extracellular domain (ECD) of PD-1 and the intracellular domain (ICD) of CD28 containing the signaling motifs YMNM, PRRP, and PYAP (Fig. 1A). The chimeric sequence PD-1:8^tm:28 was composed of PD-1 ECD (AA1-170), CD8α (AA128-210) as linker and transmembrane domain (TMD), followed by CD28 ICD (AA180-220). PD-1:cys28^tm contains PD-1 ECD (AA1-155) followed by CD28 TMD-ICD (AA141-220), which includes the membrane-proximal cysteine residue at position AA141 (adapted from Prosser and colleagues; ref. 26). PD-1:28^tm contains PD-1 ECD (AA1-165) followed by CD28 TMD-ICD (AA153-220) omitting the extracellular portion with the cysteine (adapted from Ankri and colleagues; ref. 25). PD-1trunc is a truncated PD-1 sequence (AA1-201; ref. 25). Murine mPD-1:28^tm contains murine PD-1 ECD (AA1-169) and murine CD28 TMD-ICD (AA151-218). Chimeric sequences were ordered from GeneArt.

Chimeric receptors were expressed in human T cells either by electroporation with in vitro–transcribed RNA cloned in pGEM vector (provided by S. Milosevic, Medigene GmbH) or by retroviral transduction using the pMP71 vector as described previously (29). Murine splenocytes were retrovirally transduced as described previously (29). For experimental details, see Supplementary Material.

The animal experiments were performed according to and with approval of the Landesamt für Gesundheit und Soziales, Berlin, Germany (HCC model) or they were approved by the Regierung von Oberbayern, Germany and performed in accordance with the Animal Care and Use Committee at the Helmholtz Zentrum München (NSG model).

For the xenograft model, NSG mice (Charles River Laboratories; 7–11 weeks old) were injected subcutaneously with 5 × 10⁶ SK-Mel23 cells. Grown to approximately 800 mm³ (15–20 days), tumors were injected intratumorally with 6 × 10⁶ carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled T cells. Tumors were harvested and weighed 4 hours, 1, 2, 4, 6, and 10 days after adoptive T-cell therapy (ATT). Single-cell suspensions were prepared for flow cytometry as described previously (16). Tumors were measured using a caliper prior to T-cell injection and before mice were sacrificed. Tumor volumes were calculated using the formula; volume = (length × width²) × 0.52. For Matrigel plug assays, SK-Mel23 (HLA-A2⁺Tyr⁺) and HEK293 (HLA-A2⁺Tyr⁻) were labeled with CellTracker Deep Red fluorescent dye (1.5 μmol/L, Thermo Fisher Scientific) or Vybrant CFDA-SE Cell Tracer Kit (1.5 μmol/L, Invitrogen), respectively, and both additionally labeled with 2 μmol/L PKH26 (Sigma-Aldrich). SK-Mel23, HEK293 (each 2 × 10⁶), and T cells (4 × 10⁶) in 150 μL PBS were mixed with 200 μL Matrigel (Corning) and injected subcutaneously into NSG mice. IL15 (300 U) was given intraperitoneally. Plugs were recovered after 2 days and single-cell suspension prepared for flow cytometry. CFSE-(HEK293) and Deep Red–positive cell (SK-Mel23) percentages were determined among gated PKH26⁺/CD45⁻ cells. Specific killing was calculated using the formula: |$\% \ {\rm{Specific}}\ {\rm{killing}}\ = [ {1 - {{( {{\rm{ctrl}}{\frac{{\% {\rm{SK}} - {\rm{Mel}}23}}{{\% {\rm{HEK}}\,293}}} )} \mathord{/ {\vphantom {{( {{\rm{ctrl}}\frac{{\% {\rm{SK}} - {\rm{Mel}}23}}{{\% {\rm{HEK}}\,293}}} )} {( {{\rm{sample}}\frac{{\% {\rm{SK}} - {\rm{Mel}}23}}{{\% {\rm{HEK}}\,293}}} )}}} \kern-\nulldelimiterspace} {( {{\rm{sample}}{\frac{{\% {\rm{SK}} - {\rm{Mel}}23}}{{\% {\rm{HEK}}\,293}}} )}}} ]$||$\times 100$|⁠, whereby ctrl is cell mixture without T cells and sample is cell mixture with T cells.

The HCC mouse model has been described previously (ref. 28, and Supplementary Material). The ATT experiment was performed as outlined in the results and detailed in Supplementary Material.

Flow cytometry was performed with LSRII (BD Biosciences) and FlowJo 8.8.7 software. PD-1:28 chimeric receptor and transgenic TCRs (TCR-T58, TCR-D115) expression was analyzed using anti-PD-1-APC (eBioscience), anti-mouse TCRβ-constant region (mTCR)-PB (BioLegend), anti-CD3-PE-Cy7 (BioLegend), anti-CD4-APC-A780 (eBioscience), anti-CD8-V500 (BD), anti-Ki-67-Alexa-Fluor-700 (BioLegend), anti-PD-L1-BV421 (BD Biosciences), and 7-AAD (Sigma-Aldrich). Absolute cell numbers of intratumoral T cells were determined using Flow-Count Fluorospheres (Beckman Coulter). HEK293 and tumor lines were analyzed using anti-HLA-A2 (ATCC HB54) plus anti-mouse IgG1-A488 (Invitrogen), anti-PD-L1-FITC (BD Biosciences), anti-tyrosinase (Upstate Biotechnology) plus anti-mouse IgG2a-A647 (Invitrogen), and 7-AAD. T cells were stimulated with HEK/Tyr/PD-L1 cells at 1:2 (30 minutes/37°C), then stained with anti-CD45-PE-Cy7 (BioLegend), anti-CD8-V500, anti-pERK (pT202/pY204; Cell Signaling Technology) plus anti-rabbit IgG-A647 (Invitrogen) or anti-pRPS6 (pS235/pS236)-A647 (Cell Signaling Technology), and 7-AAD. Cells were fixed using Phosflow Cytofix Buffer (BD Biosciences) and permeabilized with Phosflow Perm Buffer III (BD Biosciences).

T cells from NSG xenograft tumors were analyzed using anti-CD45-PE-Cy7, anti-CD8-PB (BD Biosciences), anti-PD-1-APC, and 7-AAD. TILs and splenocytes from mice HCC tumors were stained using anti-CD3-APC-Cy7 (BioLegend), anti-CD19-V450 (eBioscience), anti-CD45-PE-Cy7 (eBioscience), anti-PD-1-PerCP-Cy5.5 (BioLegend), anti-Thy1.1-A647 (BioLegend), and propidium iodide (Sigma-Aldrich).

Tumor suspensions from ccRCC were stained with anti-CD45-PE-Cy7, anti-CD3-PB (both BioLegend), anti-CD11c-APC, and near-IR fluorescence-reactive viability dye (Invitrogen) and sorted as CD45⁺/live/single/CD3⁺/CD11c⁻/small lymphocytes (FACS Aria IIIu, BD Biosciences).

T cells were stimulated with HEK/Tyr or HEK/Tyr/PD-L1 at 1:2 with or without antibodies blocking PD-1/PD-L1 (20 μg/mL, BioLegend). Human TILs and PBMCs were stimulated at 1:1 with OKT3/PD-L1-Fc (R&D Systems)–loaded P815 cells (each 5 μg/1 × 10⁶ cells). Murine TILs and splenocytes were stimulated with HCC434 or PMA/I (50/500 ng/mL; Sigma-Aldrich). Supernatants were harvested after 16 hours and analyzed by ELISA (BD Biosciences) or Bio-Plex (Bio-Rad). Cytokine levels were normalized to the percentage of mTCR⁺CD8⁺ T cells.

Spheroids were generated by seeding 800 SK-Mel23 cells into hanging drops. On day 3, single spheroids were confronted with 1.5 × 10⁴ T cells labeled with CMFDA (Thermo Fisher Scientific) for 30 minutes. Nonattached T cells were removed and T cells/spheroids incubated in hanging drops with CellEvent Caspase-3/7 Red Detection Reagent (Thermo Fisher Scientific) for 24 hours followed by fixation. Samples were imaged up to a depth of 70 μm using a spinning disk confocal Nikon TiE microscope. Invaded CMFDA⁺ T cells and CellEvent Caspase-3/7–positive apoptotic cells were counted using Fiji software (see Supplementary Material for details).

Chromium release assay was performed as described previously (16).

Statistical tests, indicated in figure legends, were performed using GraphPad Prism 6 software.

To facilitate costimulation of T cells in tumor environments, we designed costimulatory chimeric receptors encompassing the ECD of PD-1 and the ICD of CD28. Different receptor designs (Fig. 1A) were compared regarding surface expression and effect on T-cell function: (i) PD-1:8^tm:28; (ii) PD-1:cys28^tm; (iii) PD-1:28^tm; and (iv) truncated PD-1 (PD-1trunc) without signaling capacity.

Human activated T cells were electroporated with in vitro–transcribed RNA encoding the various receptors. Four, 24, and 48 hours later, surface expression of the receptors on CD3⁺CD8⁺ and CD3⁺CD4⁺ T cells was analyzed by flow cytometry. Four hours after electroporation, the PD-1trunc receptor was best expressed in terms of percentage of PD-1⁺ cells and median fluorescence intensity (median FI; Fig. 1B and C). Percentages of PD-1 surface–positive cells were not significantly different between PD-1trunc and either PD-1:cys28^tm or PD-1:28^tm at 4 and 24 hours after electroporation. However, median FI was significantly higher for T cells expressing PD-1trunc compared with PD-1:cys28^tm or PD-1:28^tm. PD-1:8^tm:28 showed significantly lower surface expression compared with PD-1trunc, PD-1:cys28^tm, and PD-1:28^tm regarding percentage of positive cells and levels of surface expression.

Effects of chimeric receptors on T-cell function were analyzed by assessing TCR-induced cytokine secretion. Human T cells were transduced to stably express one of two different HLA-A2–restricted tyrosinase-specific TCR, T58 or D115, followed by transient expression of chimeric costimulatory receptor (Fig. 2A). These T cells were stimulated with HLA-A2⁺ HEK293 cells expressing either the target antigen tyrosinase (HEK/Tyr) alone or together with PD-L1, the ligand for PD-1 (HEK/Tyr/PD-L1; Fig. 2B). Despite artificial expression of tyrosinase, HEK/Tyr cells were low activators of TCR-T58- or TCR-D115-T cells stimulating less IFNγ secretion and lower cytotoxicity than the natural tumor target SK-Mel23 (Fig. 2C–E). The HEK-cell system was further used as it allowed assessing the contribution of PD-L1 interacting with PD-1:28 chimeric receptors to T-cell function.

TCR-T58- and TCR-D115–expressing T cells with PD-1:28^tm or PD-1:cys28^tm secreted significantly more IL2 and IFNγ when stimulated with HEK/Tyr/PD-L1 compared with HEK/Tyr cells (Fig. 3A). The PD-1:8^tm:28 chimeric receptor had no effect on IL2 and IFNγ production of T cells expressing either transgenic TCR. Expression of PD-1trunc did not alter cytokine secretion of TCR-T58-T cells, but increased cytokine secretion of TCR-D115-T cells when stimulated with HEK/Tyr/PD-L1 cells, indicating that the truncated PD-1 is not an inert construct. The effect is likely due to competitive attenuation of negative signals through T cell–upregulated native PD-1 interacting with PD-L1 on target cells. Therefore, T cells electroporated with water (control) must be the reference for experiments assessing PD-1:28 effects.

To demonstrate that improved function was due to specific ligand receptor binding, blocking antibodies to PD-1 and PD-L1 were added to cocultures. No effect was seen in cocultures of untransfected T cells with HEK/Tyr or HEK/Tyr/PD-L1 cells. Similarly, no effect was seen when PD-1:28^tm T cells were cocultured with HEK/Tyr cells. However, when blocking antibodies were added to cocultures of PD-1:28^tm T cells with HEK/Tyr/PD-L1 cells, cytokine levels were reduced to those in HEK/Tyr cocultures, thus representing the levels induced by the TCR alone (Fig. 3B).

The molecular basis for the significant increase in cytokine secretion upon stimulation through PD-1:28^tm was investigated by assessing phosphorylation of MAP kinase ERK and ribosomal protein S6 (RPS6) using phospho-flow cytometry. After stimulation with HEK/Tyr/PD-L1, phosphorylation of ERK was enhanced 4.4-fold in PD-1:28^tm–engineered T cells compared with water control. Regarding pRPS6, 43% of PD-1:28^tm T cells were positive for the phosphorylated form compared with 33% in control (Fig. 3C and D). Although these differences were not significant, there was a clear and reproducible trend toward increased phosphorylation of both ERK and RPS6 when T cells received dual stimulation through PD-1:28^tm and their TCRs.

Next, we assessed whether PD-1:28 engineering could improve functional responses of T cells with low-avidity TCRs recognizing tumor-expressed self-antigens.

T58 and D115 are high- and low-avidity TCRs, respectively, that recognize the same HLA-A0201–restricted tyrosinase epitope. Low-avidity TCR-D115–expressing T cells secreted significantly less IL2 and IFNγ than T cells expressing TCR-T58 when stimulated with HEK/Tyr/PD-L1 target cells (Fig. 4A). TCR-D115-transgenic T cells coexpressing PD-1:28^tm or PD-1:cys28^tm showed augmented cytokine secretion that reached levels comparable with high-avidity TCR-T58–transduced T cells stimulated with PD-L1⁺ target cells. Thus, activation of the costimulatory cascade through PD-1:28 chimeric receptor allowed cells with a low-avidity TCR to approximate the cytokine function of cells with a high-avidity TCR. Similar results were observed for cytotoxicity (Fig. 4B). Low-avidity TCR-D115/Mock T cells showed lower specific lysis of HEK/Tyr and HEK/Tyr/PD-L1 target cells than the high-avidity TCR-T58/Mock T cells. PD-1:28^tm expression in TCR-D115-T cells increased cytotoxicity against PD-L1⁺ HEK/Tyr cells reaching levels equal to TCR-T58-T cells, while the cytotoxicity against PD-L1⁻ HEK/Tyr was unchanged.

PD-1:28^tm engineering not only enhanced T-cell response against HEK/Tyr/PD-L1 cells but also enabled T cells to respond stronger to tumor lines with endogenous expression of antigens and PD-L1. The extent of functional improvement followed the level of PD-L1 surface expression on the tumor cells. Melanoma lines grown in standard culture exhibited low PD-L1 surface expression (FI range, 180–540) and PD-1:28^tm-engineered TCR-D115-T cells reached on average 1.4-fold (range, 1.3- to 1.6-fold change) higher IFNγ and IL2 secretion compared with TCR-D115-T cells without PD-1:28^tm (Fig. 4C, and data not shown). PD-L1 was strongly induced on SK-Mel23 grown as 3D spheroids, and significantly higher T-cell numbers and more apoptotic tumor cells were recovered from spheroids cocultured with PD-1:28^tm–engineered TCR-D115-T cells compared with TCR-D115-T cells without PD-1:28^tm (Fig. 4D). Higher PD-L1 levels were also seen on nonmelanoma lines (FI range, 850–1830) that express the natural tumor antigen recognized by TCR53 (27). Consistent with the PD-L1 expression on TCR53 target cells, TCR53-T cells derived strong benefit from PD-1:28^tm engineering with 1.4- to 2.5-fold improvement of IFNγ secretion (Fig. 4E). These results indicate that T cells will benefit more from PD-1:28^tm engineering in tumor milieus where PD-L1 is highly expressed.

We previously demonstrated that TILs from human RCC tissue show deficits in degranulation and IFNγ secretion, in part due to insufficient activation of signaling molecules AKT, ERK, and JNK (16). As CD28 costimulation fosters activation of intracellular signaling molecules, including AKT, ERK, and JNK, we sought to determine whether costimulation through PD-1:28^tm could restore functionality in unresponsive TILs. Therefore, sorted CD3⁺ TILs from RCC tissues were electroporated with PD-1:28^tm or water and stimulated with target cells (P815 loaded with anti-CD3/PD-L1-Fc). CD3⁺ TILs electroporated with water did not secrete IFNγ upon stimulation, while CD3⁺ TILs electroporated with PD-1:28^tm clearly produced IFNγ at levels comparable with PBMCs of HD (Fig. 5A). Similar improvement was observed for IL2 secretion (Fig. 5B). As TILs in the tumor microenvironment can be polarized toward Th2 or regulatory IL10 programs, we addressed the question whether facilitated costimulation through PD-1:28 might unleash undesirable Th2 or IL10 cytokines. However, neither mock-transfected nor PD-1:28^tm–transfected TILs (n = 3) secreted detectable levels of IL4, IL5, IL13, or IL10 (data not shown). These cytokines were detected at low levels in supernatants of one (IL4) and two (IL13) HD-PBMCs, yet levels were not augmented when costimulated through PD-1:28^tm. TNF levels in supernatants of TILs were very low and marginally increased after facilitated costimulation. HD-PBMCs secreted detectable amounts of TNFα; yet, levels did not change with PD-1:28^tm costimulation (data not shown). Thus, costimulation through PD-1:28^tm can reinstate functionality in TILs that had lost activity in the tumor microenvironment. Costimulation through PD-1:28^tm rescued Th1 cytokine secretion (IFNγ, IL2), while TNFα, Th2 cytokines, or IL10 were not affected.

Presumably, ATT will be combined with checkpoint blockade using anti-PD-1 or anti-PD-L1 antibodies. To explore whether PD-1:28^tm-engineered T cells would benefit in this situation, we added titrated concentrations of anti-PD-L1 antibody to cocultures of TCR-D115/Mock or TCR-D115/PD-1:28^tm T cells with HEK/Tyr or HEK/Tyr/PD-L1 and measured secreted IFNγ. Anti-PD-L1 antibody did not change IFN-γ levels if the target cell was negative for PD-L1 (Fig. 5C). If the target expressed PD-L1 and the T cell had no PD-1:28^tm, IFNγ levels increased with raising anti-PD-L1 antibody concentration, presumably due to interrogating negative signals mediated through endogenously upregulated PD-1. Of note, in cocultures of PD-L1⁺ target cells with PD-1:28^tm T cells, low concentrations of anti-PD-L1 antibody strongly augmented IFNγ while high antibody concentration prevented this induction. PD-1:28^tm-engineered T cells combined with checkpoint blockade secreted significantly more IFNγ compared with T cells without PD-1:28^tm, indicating that PD-1:28^tm T-cell engineering can work in synergy with checkpoint blockade therapy.

Effects of PD-1:28^tm expression in T cells in vivo was studied using a human melanoma xenograft mouse model based on NSG mice carrying established vascularized subcutaneous melanoma tumors formed by the SK-Mel23 melanoma line (Fig. 6A). CFSE-labeled TCR-D115/Mock or TCR-D115/PD-1:28^tm T cells were injected intratumorally into tumors (802 mm³ ± 83, mean ± SEM) and analyzed at 4 h, 1, 2, 4, 6 and 10 d after ATT by flow cytometry. It was observed that TCR-D115/Mock T cells stained positive for PD-1 on 1, 2, and 4 days after ATT. Thereafter, PD-1 staining decreased, but never completely disappeared (Fig. 6B and C). This temporary expression of endogenous PD-1 suggests that T-cell activation occurred within the tumor microenvironment, as PD-1 is expressed on recently activated T cells (30). Activation lasted until day 4 after ATT. For TCR-D115/PD-1:28^tm T cells, temporary expression of endogenous PD-1 could not be distinguished from the transgenic PD-1:28^tm expression. However, a temporary increase in median FI of PD-1 was observed with similar kinetics.

TCR-D115/PD-1:28^tm T cells lost CFSE intensity more rapidly and strongly than TCR-D115-T cells without PD-1:28^tm (Fig. 6D and E). The difference was significant at all time points (Fig. 6D), indicating that TCR-D115/PD-1:28^tm T cells responded faster and stronger to antigen-specific stimulation in the tumor microenvironment than TCR-D115 T cells without the chimeric costimulatory receptor. Survival could not be assessed in this model as T cells were injected into large tumors to ensure establishment of a challenging microenvironment. Animal regulatory restrictions did not allow extension of the observation time. However, compared over 7 days (Fig. 6F), tumors injected with TCR-D115/PD-1:28^tm T cells grew slower (median 1.5-fold increase) compared with tumors injected with TCR-D115 T cells without PD-1:28^tm (median 2.0-fold increase). While acknowledging the small sample size and absence of statistical significance, the data suggest that PD-1:28^tm costimulation can improve antitumor function of low-avidity T cells in the tumor milieu. Analysis of T cells recovered from 7-day tumors support this evidence: Higher T-cell numbers were found in tumors injected with TCR-D115/PD-1:28^tm T cells with higher percentages of Ki-67⁺ cells and lower frequencies of T cells expressing PD-L1, compared with tumors injected with TCR-D115/Mock T cells (Fig. 6G). Higher T-cell numbers and better tumor control after ATT with TCR-D115/PD-1:28^tm T cells are consistent with results from the spheroid model (Fig. 4D). Low T-cell numbers and high PD-L1 expression may explain the observed poorer tumor control through TCR-D115/Mock T cells, as PD-L1 on T cells has been described to inhibit T-cell expansion and function (31). Furthermore, in a Matrigel-killing assay in situ over 2 days (Fig. 6H), TCR-D115/PD-1:28^tm T cells achieved on average 76.2% of SK-Mel23 killing, while killing through TCR-D115/MOCK T cells was much lower (59.3% on average).

Limitations of NSG xenograft models led us to analyze the impact of PD-1:28^tm chimeric receptors in a fully murine system. In the autochthonous HCC mouse model, T cells become exhausted after an initial phase of antitumor activity, and survival could be prolonged with anti-PD-L1 (28). To investigate whether PD-1:28^tm could rescue exhausted T cells, splenocytes were isolated from mice 15 weeks after initiation of oncogenesis (it is assumed that T cells have lost control over tumor growth at this time point; ref. 28) and transduced with Thy1.1 to allow tracking after ATT, alone or together with a murine PD-1:28^tm (mPD-1:28^tm). ATT was given to LoxP-TAg mice that carried occult tumors (week 12 after oncogene activation; Fig. 7B). Although ATT did not prolong survival (not shown), notable differences were observed in the TIL populations after ATT with T cells with or without mPD-1:28^tm. Globally, there were no differences in percentages of CD3⁺ T cells or CD19⁺ B cells in tumors or spleens of mice with or without ATT (Fig. 7C and D). However, the cytokine profile of TILs in the PD-1:28^tm-negative ATT was shifted toward Th2⁺Th17 cytokines compared with splenocytes that were dominated by Th1 cytokines (Fig. 7E and F). PD-1 can inhibit Tc1/Th1 responses in the tumor microenvironment through SHP-2 (32) and, indeed, very high levels of PD-1 were observed among TILs (Fig. 7E, inset). Thus, endogenous PD-1 induction in the tumor microenvironment might contribute to the Th2 polarization of TILs. TILs from mice having received ATT with mPD-1:28^tm-positive T cells had a more favorable Th1/Th2 balance with less Th2⁺Th17 cytokines compared with TILs from mice given mPD-1:28^tm-negative ATT, yet with a notable increase in IL10.

Immunotherapy has emerged as a pillar of cancer therapy, although improvements are needed as response rates are still low (1, 2, 33). Loss of function, exhaustion, and poor persistence of transferred T cells within the tumor microenvironment hinder the efficacy of ATT. T-cell function depends on activation of signaling cascades downstream of the TCR supported by costimulation, which we are able to modify to overcome impediments of ATT. Costimulation increases the MAP kinase activation required for cytokine secretion and degranulation and supports TCR-independent programs downstream of AKT/mTOR, such as survival, cell-cycle progress, glucose consumption, and restoration of lytic proteins, which are particularly valuable in a restricting environment (23, 24). Human CD8⁺ T cells lose CD28 costimulatory receptor expression during effector–cell differentiation (34), and epithelial cancers generally do not express cognate ligands CD80 or CD86 (35, 36); thus, T cells are unable to receive proper costimulation in this situation. In the context of ATT, low-avidity T cells may not be sufficiently active for tumor elimination due to low functional responses. Yet, they represent the majority of antitumor-reactive T cells and are relatively resistant to the suppressive tumor environment (13, 37). Harnessing them for the antitumor response would enlarge and enrich the spectrum of T cells for ATT. As previous data have shown that PD-1:28 receptors can improve TCR- and chimeric antigen receptor (CAR)-engineered T cells (25, 26, 38, 39), we engineered low-avidity T cells with a PD-1:28 chimeric signaling receptor to explore whether their functional activity could be enhanced allowing their inclusion in a library of TCRs for ATT.

We show that PD-1:28 chimeric receptor can be stably expressed on the surface of human T cells. Surprisingly, the CD8^tm-linker destabilized surface expression, although similar sequences are effectively used in expression constructs of CAR receptors. We demonstrate that ligation of PD-1:28 overcomes important hurdles in ATT. PD-1:28 engineering invigorated TILs that were functionally disabled in human RCC environments. These TILs recovered Th1-cytokine secretion without unleashing Th2 or IL10 programs presumably based on restoration of ERK and AKT pathways as shown in our in vitro coculture system and previous work using similar receptors (25, 26). PD-1:28 ameliorated T-cell inhibition that involved native PD-1/PD-L1 interactions and induced the costimulatory signaling pathway. The combined effect was stronger on the low-avidity TCR-D115-T cells compared with the high-avidity TCR-T58-T cells. Indeed, PD-1:28 cosignaling upgraded low-avidity TCR-D115-T cells to approximate cytokine and cytotoxic response levels of high-avidity TCR-T58-T cells. Thus, PD-1:28 engineering should allow harnessing the advantages of low-avidity T cells for the antitumor response.

PD-1:28 engineering enhanced the T-cell effector response against melanoma, RCC, squamous cell carcinoma, and glioblastoma cell lines. Response enhancement depended on PD-L1 surface expression level on the tumor cells. Thus, PD-1:28 engineering will deliver benefit in conditions with high PD-L1 expression, that is, in tumor milieus. Further evidence that PD-1:28^tm engineering can support T cells in a “hostile” tumor microenvironment was provided by a human xenograft model using large SK-Mel23 tumors with established T-cell–inhibitory conditions. Here, as was observed in SK-Mel23 spheroids, PD-1:28^tm–engineered low-avidity TCR-D115-T cells proliferated significantly more and reached higher absolute cell numbers compared with TCR-D115-T cells without the chimeric receptor. TCR-D115/PD-1:28^tm T cells also showed higher tumor cell killing. Proliferation and killing are important prerequisites for CD8⁺ T cells to cause tumor shrinkage (19, 40, 41). Tumors injected with low-avidity D115/PD-1:28^tm T cells grew slower over 7 days than tumors injected with TCR-D115 T cells. In future the xenograft animal experiments will be extended to larger sample sizes with longer observation times based on regulatory authority approval.

To determine the potential effect of PD-1:28^tm on tumor progression, a murine model of autochthonous HCC was used where tumors develop after activation of the SV40-TAg (28). This tumor model resembles the human situation in its slow course of tumor progression, spanning up to 35 weeks from initiating the oncogenic process until death from tumor load. ATT using T cells engineered with a murine PD-1:28^tm showed positive effects on T-cell polarization. Most notable was the prevention of Th2 and Th17 cytokine secretion. The reduction of Th2 cytokines may be attributed to PD-1:28^tm deflecting native Th2-polarizing PD-1/PD-L1 interactions into Th1-favoring costimulatory pathways. Similarly, augmentation of Th1 cytokines was observed in patients treated with PD-1 blockade (41). ATT with PD-1:28^tm-positive T cells also resulted in more IL10, which may enhance cytotoxicity and IFNγ secretion by CD8⁺ T cells and tumor growth, but may also suppress antitumor immunity (42, 43).

Despite positive effects of PD-1:28^tm on T-cell performance in ATT, this was not sufficient to change tumor growth characteristics in this murine tumor model. Effects on tumor growth have been documented in a mouse model with subcutaneous tumors overexpressing an artificial antigen (ovalbumin) and PD-L1 (44). In our model with tumor development over an extended time period using native antigen and PD-L1 expression, the tumor immune escape may be more complex (45). Overcoming suppression likely requires additional supportive measurements beyond PD1:28 chimeric receptors, such as inhibition of IL10. This is in line with clinical results of ATT using TILs where benefit was achieved only after patient preconditioning with lymphodepletion and radiation in a multiple TIL injection regimen and coadministration of IL2 (19, 46). We have noted that much higher T-cell numbers can be recovered from xenograft tumors after 7 days when IL15 is administrated intraperitoneally at the day of ATT (data not shown).

In summary, we show that cytokine secretion of functionally disabled human TILs can be restored through a PD-1:28^tm receptor. PD-1:28^tm mediated a functional upgrading of low-avidity T cells to the level seen in high-avidity T cells. As a consequence, low-avidity T cells previously considered to be therapeutically inefficient can now be considered for ATT. As the PD-1:28 engineering does not modify the TCR sequence, adverse effects on specificity are not expected, thus making this a safer approach than TCR affinity enhancement (3, 4, 10–12, 47). PD-1:28 chimeric receptors have wider potential applications also for CAR-expressing T cells, where these engineered cells have shown better migration and tumor infiltration (38). Although some benefits on T-cell function and tumor control were observed, more advanced in vivo models are needed to better document an improved antitumor response. It may be that the potential impact of PD-1:28 receptors on tumor control cannot be reliably visualized in a murine system (48) as mouse T cells do not lose CD28 (34, 49) and, therefore, may not share the same necessity of facilitated costimulation with human T cells. Moreover, mouse models often fail to correctly predict clinical efficacy. New models, such as advanced tumor spheroid 3D cultures, may be helpful in this regard.

Checkpoint inhibition therapies using PD-1/PD-L1 antibodies may benefit from a combination with PD-1:28–engineered T-cell therapy as therapeutically applied anti-PD-L1/anti-PD-1 antibodies may not reach saturation of all target molecules (50). The supportive effects of PD-1:28 engineering imparted on T-cell function make it an attractive tool for ATT.

D.J. Schendel is the chief executive officer and chief scientific officer at and has ownership interest (including patents) in Medigene AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Schlenker, M. Leisegang, P.J. Nelson, G. Willimsky, E. Noessner

Development of methodology: R. Schlenker, L.F. Olguín-Contreras, M. Leisegang, S. Rühland, P.J. Nelson, H. Harz, W. Uckert, E. Noessner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Schlenker, L.F. Olguín-Contreras, J. Schnappinger, A. Disovic, S. Rühland, W. Uckert, G. Willimsky, E. Noessner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Schlenker, L.F. Olguín-Contreras, M. Leisegang, P.J. Nelson, G. Willimsky, E. Noessner

Writing, review, and/or revision of the manuscript: R. Schlenker, L.F. Olguín-Contreras, M. Leisegang, J. Schnappinger, P.J. Nelson, D.J. Schendel, W. Uckert, G. Willimsky, E. Noessner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Leonhardt, S. Wilde, D.J. Schendel

Study supervision: E. Noessner

We thank A. Brandl, K. Hummel, A. Slusarski, A. Frank, I. Jeremias, D. Brech, and K. Borgwald for excellent technical support, R. Mocikat for supervision of animal work, and M. Heikenwälder for technical advice. We highly appreciate the work of all animal caretakers and appreciate the contributions of blood donors and patients. BioImaging/Microscopy was performed at the Center for Advanced Light Microscopy (CALM) of the LMU Munich.

E. Noessner, W. Uckert, G. Willimsky, and D.J. Schendel received grants from the DFG (SFB-TR36). E. Noessner received additional grants from the Deutsche Krebshilfe and the Erich and Gertrud Roggenbuck-Stiftung. L.F. Olguin-Contreras received a fellowship from CONACyT-DAAD. M. Leisegang received grants from BIH-CRG1 and Einstein-Stiftung Berlin. H. Leonhardt received a grant from DFG SFB1243/A01. P.J. Nelson was supported by a grant from the Deutsche Forschungsgemeinschaft within the Priority Program SPP1629 (NE 648/5-2) and within a grant from the Wilhelm-Sander-Stiftung (2014.129.1).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.