Adoptive therapy with chimeric antigen receptor (CAR) T cells shows great promise clinically. However, there are important aspects of CAR-T-cell biology that have not been explored, particularly with respect to the kinetics of activation, immune synapse formation, and tumor cell killing. Moreover, the effects of signaling via the endogenous T-cell receptor (TCR) or CAR on killing kinetics are unclear. To address these issues, we developed a novel transgenic mouse (designated CAR.OT-I), in which CD8+ T cells coexpressed the clonogenic OT-I TCR, recognizing the H-2Kb–presented ovalbumin peptide SIINFEKL, and an scFv specific for human HER2. Primed CAR.OT-I T cells were mixed with SIINFEKL-pulsed or HER2-expressing tumor cells and visualized in real-time using time-lapse microscopy. We found that engagement via CAR or TCR did not affect cell death kinetics, except that the time from degranulation to CAR-T-cell detachment was faster when CAR was engaged. We showed, for the first time, that individual CAR.OT-I cells can kill multiple tumor cells (“serial killing”), irrespective of the mode of recognition. At low effector:target ratios, the tumor cell killing rate was similar via TCR or CAR ligation over the first 20 hours of coincubation. However, from 20 to 50 hours, tumor cell death mediated through CAR became attenuated due to CAR downregulation throughout the time course. Our study provides important insights into CAR-T–tumor cell interactions, with implications for single- or dual receptor–focused T-cell therapy. Cancer Immunol Res; 3(5); 483–94. ©2015 AACR.
See related commentary by June, p. 470
Chimeric antigen receptor (CAR) T cells are at the forefront of cell-based therapies for the successful treatment of cancer. CAR-T cells redirected to CD19 (CAR19) have shown spectacular results in clinical trials for patients with chronic lymphocytic leukemia (CLL) and acute lymphocytic leukemia (ALL; refs. 1–6). Although the findings for CAR-T-cell therapy in solid tumors have been less convincing, objective clinical responses have been observed (7, 8). Despite these exciting advances, there are still important aspects of CAR-T-cell biology that remain unexplored.
Currently, it is known that CAR-T-cell binding to target cells following antigen recognition results in target cell death and release of effector cytokines IFNγ, IL2, and TNF (9). Furthermore, preclinical studies have demonstrated that perforin and Granzyme B were required for CAR-T-cell efficacy in vivo (10, 11). Nevertheless, relatively little is known about the kinetics of (CAR-T tumor cell) immune synapse formation, release of the cytotoxic effector molecules, and tumor cell apoptosis. It is unclear whether these fundamental steps of tumor killing by T cells are different following stimulation by the T-cell receptor (TCR) or CAR. Furthermore, although it is known that individual cytotoxic lymphocytes (NK and CD8+ T cells) can sequentially kill multiple targets (“serial killing”; refs. 12–15), this has not yet been described for CAR-T cells. Understanding the cell interactions and kinetics of CAR-T-cell killing of tumor cells will provide important base information against which new, improved CARs can be compared and lead to increasing the effectiveness of this approach.
In this study, we developed a model system to explore whether a difference exists in recognition and killing of tumor cells following stimulation through the TCR or CAR. We used time-lapse live microscopy to record the kinetics of CAR-T-cell interactions with tumor cells and tumor cell killing, and compared these parameters in cocultures in which CAR-T cells were activated via their TCR versus CAR. Using this strategy, we revealed important information regarding the kinetics of CAR-T-cell interactions with tumor cells and investigated whether CAR-T cells have the capacity to mediate serial killing of target cells.
Materials and Methods
MC57 cells (mouse fibrosarcoma cell line) or MC57 cells retrovirally transduced with the HER2 antigen (MC57-HER2) were cultured in complete media, RPMI-1640 comprising 10% FCS, 1 mmol/L sodium pyruvate, 2 mmol/L glutamine, 0.1 mmol/L nonessential amino acids, 100 U/mL penicillin–streptomycin (Invitrogen; Life Technologies), and 5 μmol/L 2ME. Tumor lines were incubated at 37°C with 5% CO2 and grown to 80% confluence in T-175 tissue culture flasks before passage. These cell lines were originally obtained from the ATCC, actively passaged for less than 6 months and were authenticated using short tandem repeat profiling. United Kingdom Coordinating Committee on Cancer Research (UKCCCR) guidelines for the use of cell lines in cancer research were followed. Tumor lines were also verified to be Mycoplasma negative by the Victorian Infectious Diseases References Laboratory (Melbourne, Victoria) by PCR analysis.
OT-I transgenic mice (16) were bred at the Peter MacCallum Cancer Centre (PMCC). A unique C57BL/6 transgenic mouse model, expressing a second generation CAR (anti-human HER-2 scFv and signaling domains for CD28 and CD3ζ) under the Vav promoter was previously generated at the PMCC (Yong C; unpublished data). The anti-HER2 scFv CAR construct used in this study was previously described in detail (17), and uses an anti-HER2 scFv, which was a kind gift from Winfred Wels (18). These CAR mice were then bred with OT-I mice to create CAR.OT-I mice, which expressed both the OT-I TCR and the CAR on CD8+ T cells.
Activation and differentiation of T cells
Naïve splenocytes were harvested from OT-I or CAR.OT-I mice and activated in the presence of 10 nmol/L SIINFEKL and 100 U/mL IL2 in complete media. Splenocytes were incubated for three days at 37°C, and viable T cells recovered following density gradient centrifugation. T cells were washed in and resuspended in complete media at a density of 1 × 106/mL with 100 U/mL recombinant human IL2 for an additional 4 days. These activated effector CD8+ T cells (CAR.OT-I or OT-I) were used for subsequent tumor killing studies and in microscopy analysis and are referred to as CTL throughout this study.
Effector CTLs were labeled with antibodies in FACS buffer (Magnesium and Calcium free PBS + 2% FCS) with anti-CD44 BV785, anti-CD62L BV510, and anti-CD8 BV711 (BioLegend), anti-CD107a PE (BD Biosciences), anti–c-myc TAG AF488 (Cell Signaling Technology) and the viability dye Fixable blue (Invitrogen) for 30 minutes at 4°C. Granzyme B was detected using the BD Intracellular Staining Kit according to the manufacturer's instructions by labeling with anti–Granzyme B-APC (Clone GB11; BD Biosciences) antibody. This antibody is anti-human but shown to be cross-reactive with mouse Granzyme B (19). Cells were subsequently analyzed using the LSR II (BD Biosciences) and final analysis was performed using the FlowJo software (Tree Star).
T-cell proliferation assay
Proliferation of splenocytes from naïve to day 3 (OT-I or CAR.OT-I) was detected by performing a cell trace violet (CTV; Invitrogen) dilution assay. Splenocytes were stained with 2.5 μmol/L CTV for 20 minutes at 37°C before washing and coculturing with MC57 cells pulsed with 10 nmol/L SIINFEKL. Cells were harvested 3 days later, and labeled for CD8 (BV711-BioLegend) and viability with Fixable blue dye (Invitrogen). Stained cells were analyzed on a BD LSRII (BD Biosciences) and CTV dilution assessed on CD8+ cells. Control wells contained splenocytes alone.
Quantification of receptor and tumor target antigen molecule density
A FACS-based assay was used to quantify the density of CAR.OT-I cell antigen receptors (OT-I and CAR), and tumor antigen (OVA or HER2) in molecules per cell. Effector OT-I or CAR.OT-I T cells were stained for CAR expression with an anti–c-myc tag AF488 antibody (Cell Signaling Technology) or an anti–Vα2-PE antibody (BD Biosciences), recognizing the TCR subunit specific to OT-I T cells. Target cells were stained for their expression of H-2Kb (BioLegend), H-2Kb SIINFEKL (25D1.16 PE; BioLegend) or HER2 (Mouse anti-human HER2 primary and anti-mouse secondary PE; BioLegend). A proprietary bead standard kit (Simply Cellular Beads, BANGSLABS laboratories) was stained using the same antibodies to create a standard curve with a known number of binding sites per bead. The mean fluorescence intensity (MFI) for each receptor and bead combination was recorded and compared with a standard curve, and a data reduction performed in MS EXCEL. The final receptor density per cell was recorded using GraphPad Prism software.
Chromium release assay
Tumor cell targets were labeled with 100 μCi 51Cr ± 1 μmol/L SIINFEKL, at 37°C in T-cell media, before washing and seeding at 1 × 105/mL. CTLs were added at various effector:target (E:T) ratios and incubated at 37°C for 4 hours. Chromium release was read using the Wallac Wizard 1470 (PerkinElmer), and the percentage of lysis calculated using the following formula: [(experimental counts/minute − spontaneous counts/minute)/(total counts/minute − spontaneous counts/minute)] × 100. Results are shown as the percentage of cytotoxicity (mean ± SEM), using pooled data from five experiments.
xCELLigence killing assays
MC57 or MC57-HER2 cells (2 × 104/well) were left to adhere for 16 hours to xCELLigence E-plates (Cat 67710; ACEA Bioscience). On the day of assay, peptide was added (±1 μmol/L SIINFEKL (MC57-OVA257) before washing and adding CTLs (d7 OT-I or CAR.OT-I) at an E:T ratio of 10:1 and 5:1 for 8 hours (Fig. 4), or an E:T ratio of 1:1 and 1:4 for 50 hours (Fig. 5). An electrical current was applied to each coculture (20, 21) in the xCELLigence assay. Adherent cells (MC57) are resistant; however, as the effector cells kill the MC57 cells and they detach from the plate, the electrical resistance in the coculture decreases. Electrical resistance as arbitrary units (AU) over time (minutes) was monitored and the rate of killing detected by the slope of the resistance curve. E-plates were read every 2 minutes and data were normalized to starting resistance, using the formula: x = resistance reading +10 and y = (xbasereading/xbasereading)/(xn time point/xbase reading), where y is the final AU plotted as fold change against time in GraphPad Prism software to create kinetic curves.
Perforin protein detection by Western blot analysis
Total cell extracts were prepared from OT-I or CAR.OT-I CTLs using lysis buffer (25 mmol/L Hepes, 0.25 mol/L NaCl, 2.5 mmol/L EDTA, 0.1% Triton X-100), in combination with complete Protease Inhibitor Cocktail (Roche diagnostics). SDSPAGE was performed on cell extracts under reducing conditions, using Bis-Tris 4% to 12% Bis-Tris gel (Invitrogen). Following transfer onto polyvinylidene difluoride (PVDF) membrane (Amersham GE Healthcare), a Western blot analysis was performed using primary antibodies for rat anti-human perforin (clone P1-8; ref. 22) and mouse anti-human β-actin (Sigma-Aldrich), overnight at 4°C. PVDF membranes were probed using horseradish peroxidase–conjugated anti-rat IgG or anti-mouse IgG secondary antibodies (Dako) followed by the chemiluminescence developer, ECL (GE Healthcare). Protein bands were imaged using Image Lab software (Bio-Rad).
CD107a degranulation assay
Tumor cells (MC57, MC57-OVA257, and MC57-HER2) were cocultured at 1 × 105/mL in a 96-well plate, with 1 × 106/mL of CTLs, in the presence of 0.5 μg of anti-CD107a PE antibody (BD Biosciences), at 37°C, 5% CO2. Cells were harvested at 30, 60, and 120 minutes, labeled with anti-CD8α antibody and stained with viability dye as described above. FACS lismode files were analyzed using FlowJo software (Tree star).
Generation of tumor targets expressing human HER2
MC57 fibrosarcoma cells were transduced with a retrovirus encoding human HER2 and GFP (MC57-HER2), as described previously (23). GFP- and HER2-positive cells were then purified by FACS using the FACS ARIA (BD Biosciences). HER2 was detected by indirect staining with a mouse anti-human HER2 antibody (9G6.10; NeoMarker) and a secondary anti-mouse antibody conjugated to PE (donkey anti-mouse).
Time-lapse live video microscopy
Interactions between CTLs and tumor cells were assessed by time-lapse live microscopy, using a protocol previously published by Lopez and colleagues (15) and M.R. Jenkins (submitted for publication) (see Supplementary Methods for details). Image analysis was performed using Leica LAS AF Lite software or MetaMorph Imaging Series 7 software (Universal Imaging).
Statistical analyses were performed using GraphPad Prism 6 software. Statistical tests applied include Student t test and ANOVA. Asterisks within figures refer to statistical difference between test and control groups, P values and the number of replicate experiments performed to derive the data are indicated in the legends for Figs. 2, 4, 5, and 7.
Expression of CAR in OT-I cells does not affect stimulation through the TCR
OT-I T cells become activated and differentiate into effector and memory cells following engagement of antigen-presenting cells displaying SIINFEKL (OVA257) in the context of H-2Kb (16, 24). We used well-established protocols to induce activation and differentiation of naïve CAR.OT-I T cells to effector CTLs. CAR.OT-I cells expressed both the OT-I TCR and a second-generation CAR (scFv-CD28-ζ) recognizing the HER2 antigen (11). We first compared the effect of stimulation through either the TCR or CAR on target cell recognition and effector function, by examining whether the presence of the CAR in OT-I T cells affected signaling and subsequent activation through the TCR. Naïve splenocytes isolated from CAR.OT-I or OT-I transgenic mice and activated with OVA257 over 7 days displayed equivalent levels of activation markers CD44 and CD62L (Fig. 1A and B), and equivalent proliferative capacity, as measured by CTV dilution (Fig. 1C). There was no statistically significant difference in either the proliferation or division index for OT-I CTLs, compared with CAR.OT-I CTLs, in response to SIINFEKL-pulsed syngeneic splenocytes. These data indicate that the expression of CAR did not disrupt TCR-mediated activation of CAR.OT-I CTLs.
Activated OT-I and CAR.OT-I cells display similar levels of cytotoxic granule proteins
Although the presence of the CAR had no adverse effect on the activation and differentiation of CAR.OT-I effector cells (comparison of cytotoxicity will be described below), it was important to examine expression of the key cytotoxic granule proteins, perforin and Granzyme B. Perforin (Supplementary Fig. S1A) and granzyme B (Supplementary Fig. S1B) were equivalently expressed by OT-I and CAR.OT-I CTL. We next compared the functional ability of CAR.OT-I CTL with exocytose cytotoxic granules in response to either TCR or CAR target cell antigens. There was no statistically significant difference in the level of CD107a exposure on the surface of CAR.OT-I cells following activation with either MC57-OVA257 or MC57-HER2 tumor targets. CAR.OT-I CTLs (Supplementary Fig. S1C) or OT-I CTLs (Supplementary Fig. S1D) degranulated over a 2-hour period in response to TCR (MC57-OVA257) or CAR stimulation (MC57-HER2), but not in response to control MC57 cells. Finally, we compared the effect of OVA257 peptide dose in the initial activation culture, on the functional capacity of the generated effector T cells. We showed that T-cell CD69 and Granzyme B expression levels followed the same trend from days 4 to 7, despite being stimulated by accessory cells cultured in limiting concentrations of OVA257 (Supplementary Fig. S2A–S2D). In addition, OVA257 dose did not affect the maturation status of the activated effector cells, as measured by the percentage of CD44+ and the percentage of CD62L+ CD8+ T cells in the culture from days 4 to 7 (Supplementary Fig. S2E–S2H).
CAR and TCR on CAR.OT-I cells are differentially expressed
Before comparing the kinetics of target cell engagement and cytotoxicity following activation via CAR or TCR, it was important to confirm that CAR.OT-I cells coexpressed CAR and TCRVα2, and to compare the relative expression of both antigen receptors, as differences have a significant impact on their responses to cognate antigen. We demonstrated that CAR.OT-I CTLs expressed both the CAR and OT-I TCR, although CAR expression levels varied between cells (Fig. 2A). CAR.OT-I mouse T cells expressed significantly lower levels of CAR than TCRVα2 (P = 0.003, Fig. 2B). The level of TCRVα2 expressed was similar between OT-I and CAR.OT-I CTL (Fig. 2B). We next compared the level of CAR on CAR.OT-I CTLs with that on transduced mature T cells. Wild-type C57BL/6 splenocytes were retrovirally transduced with the same second-generation CAR as that expressed by CAR.OT-I cells. In these transduced C57BL/6 cells, CAR density (8,797 ± 4,203 molecules/cell) was moderately increased compared with that of CAR.OT-I cells (3,040 ± 660 molecules/cell), although this difference was not statistically significant (Fig. 2B).
The TCR and CAR target antigen expression levels were also assessed on MC57-HER2 tumor cells. We found that MC57-HER2 cells uniformly expressed both human HER2 and H-2Kb (Fig. 2C and D). Furthermore, when pulsed with OVA257, the OT-I TCR antigen OVA257/H-2Kb were similarly displayed on both MC57-HER2 and MC57 cells (Fig. 2D and E). Finally, the number of tumor target antigen molecules was not statistically different for HER2 and OVA257/H-2Kb (Fig. 2F). Taken together, our data indicated that CAR.OT-I cell TCR expression levels were significantly higher than that of the CAR molecules. In subsequent experiments, we compared effector cell killing of tumor cells, and explored whether the difference in CAR.OT-I cell antigen receptor levels influenced the functional outcomes of T-cell activation and the kinetics of tumor cell killing.
Cytotoxic T cells can effectively kill tumor targets via CAR or TCR ligation
Having defined the parameters of our model, we next compared the kinetics of attachment, recognition, and cytotoxicity of tumor target cells by CAR.OT-I cells through engagement of either TCR or CAR, by live-cell microscopy. We used a well-characterized method (M.R. Jenkins; submitted for publication; refs. 15, 25) to precisely pinpoint the time taken for a CAR-T cell to deliver a lethal hit to a target cell, after binding of the CAR-HER2 receptor. In this experiment, either OT-I or CAR.OT-I cells were labeled with the Ca2+ indicator fluo-4 AM and cocultured with MC57, MC57-OVA257, or MC57-HER2 target cells, in the presence of 100 μmol/L propidium iodide (PI). The formation of perforin pores on the target cell was inferred by the strong PI binding and fluorescence within the cytosol of the target cell emanating from the region of the immune synapse (termed the “PI blush,” Fig. 3A–C). Montages of single-cell conjugates are shown for control OT-I cells cocultured with MC57-OVA257 (Fig. 3A; Supplementary Movie 1), and CAR.OT-I cells cocultured with either MC57-OVA257 (Fig. 3B, Movie 2), or MC57-HER2 cells (Fig. 3C; Supplementary Movie 3). In each corresponding image, we analyzed the fold change in fluorescence intensity of fluo-4-AM within the killer cell, and the fold change in PI fluorescence in the target cell, in real time (right for Fig. 3A–C). Thus, upon antigen engagement, the Ca2+ flux (green fluorescence) preceded tumor cell uptake of PI (red fluorescence). In contrast, when there was no effector cell recognition (control MC57 cells), both green and red fluorescence remained at basal levels (data not shown). When the same effector cells were cocultured with tumor cells that did not express cognate antigen, degranulation and cell killing were not observed (data not shown).
We analyzed the kinetics of many single-cell conjugates by live-cell microscopy, measuring the time taken for each stage of killing: (i) effector cell “binding” (scanning by the effector cell); (ii) “recognition” (antigen recognition indicated by Ca2+ flux in effector cell); (iii) delivery of the “lethal hit” (visualized when PI binds cytosolic RNA after entering through perforin pores in the target cell membrane; ref. 15); and (iv) “target cell rounding,” signifying loss of cell adhesion and commitment to cell death. Using the same coculture combinations as in Fig. 3, we subsequently characterized the kinetics of individual effector cell–tumor cell killing interactions (Fig. 4). There was no statistically significant difference in the time interval for “Ca2+ flux to lethal hit delivery” (Fig. 4A), “lethal hit delivery to tumor cell rounding” (Fig. 4B), and “lethal hit delivery to tumor cell detachment” (Fig. 4C) following recognition of the target cell via the TCR or the CAR. However, we found that when CAR.OT-I cells recognized tumor antigen via the CAR (vs. the TCR), the time interval for “Ca2+ flux to tumor cell detachment” was significantly reduced (Fig. 4D). Interestingly, effector CAR.OT-I cells responding via CAR stimulation displayed a more uniform response for this parameter, indicating a more consistent duration of synapse (Fig. 4D).
CAR.OT-I T cells can similarly kill targets through CAR and TCR stimulation in short-term but not long-term assays
Our data suggested that despite equivalent activation, CAR.OT-I CTLs engaging targets via CARs remained in synapse for shorter times. However, a cytotoxicity assay performed using CAR.OT-I T cells cocultured for 4 hours with MC57-OVA257 or MC57-HER2 showed no statistical difference in killing by CAR.OT-I cells targeting the TCR (MC57-OVA257) or CAR (MC57-HER2; Fig. 5A). Predictably, there was minimal cytotoxicity in the absence of antigen (Fig. 5A, diamonds). Although 51Cr-release assays are suitable for determining overall cytotoxic function, they do not provide direct information on the rate of target cell death. To address this, we used an xCELLigence assay, which measures killing by recording the reduction in resistance to an electrical current passed through the adherent target cells over time (20, 21). At each time point, arbitrary resistance values are normalized to the starting time point and expressed as fold change and the rate of killing is reflected by the gradient of the curve. We observed that at two different E:T ratios (10:1 and 5:1), the rate of killing of MC57-OVA257 or MC57-HER2 cells by CAR.OT-I effector cells was equivalent for up to 8 hours (Fig. 5B and C).
CAR.OT-I T cells stimulated through the CARs are capable of serial killing
The capacity of CTLs to serially kill target cells is important for eradicating large tumor burdens. However, the capability of CAR receptors to sequentially kill multiple tumor targets (“serial killing”) has not been previously investigated. Hence, we examined this issue by using CAR.OT-I cocultured with either MC57-HER2 or MC57-OVA257 targets and live time-lapse microscopy. Serial killing of tumor targets by CAR.OT-I effector cells was observed following engagement through the TCR (Fig. 6A; Supplementary Movie 4) or CAR (Fig. 6B; Supplementary Movie 5). Individual effector CAR.OT-I cells sequentially delivered a lethal hit to two (Fig. 6A) or three (Fig. 6B) tumor targets. Interestingly, CAR.OT-I cells mediated efficient tumor cell killing of adjacent tumor cells almost immediately after the first hit (Fig. 6B, right; Supplementary Movie 5). As previously described in mouse natural killer cells (25), we did not see repetitive repriming of Ca2+ flux within a single CTL between these rapid killing events. We further examined the videos of effector–target cell interactions, and assessed the frequency of serial killing events (Fig. 6C). CAR.OT-I CTL engaged in serial killing of MC57-OVA257 (22.52%) and MC57-HER2 (21.74%), and this was equivalent to that observed when OT-I cells were cocultured with MC57-OVA257 (23.53%).
To explore the serial killing, we previously observed by video, a 51Cr-release assay using target:effector (T:E) ratios of 1:1 through 32:1 over 18 hours was used. This showed CAR.OT-I CTLs killed either MC57-OVA257 or MC57-HER2 at equivalent levels across a wide range of T:E ratios. Furthermore, the amount of CAR.OT-I CTLs killing of tumor cells (MC57-OVA257 or MC57-HER2) was significantly different compared with MC57 cells for T:E of 1:1 thru to 8:1 (Fig. 7A). To assess whether the difference in CAR and TCR antigen receptor levels might have long-term consequences for tumor cell killing (whether single or serial killing), we performed an xCELLigence assay over 50 hours. A lower E:T ratio was used than was used previously to reduce the rate of target cell death and so potentially reveal differences in effector cell–tumor killing kinetics. TCR-ligated CAR.OT-I CTLs killed tumor cells at a consistent rate over the extended time period at an E:T ratio of 1:1 (Fig. 7B), as indicated by the continual reduction in impedance. The rate of killing by CAR-ligated CAR.OT-I CTLs was similar for the first 20 hours of the assay, but slowed markedly beyond this time point at both E:T ratios (Fig. 7B). By measuring the slope of the two curves, we were able to compare the rates of target cell death from 20 to 50 hours. Pooled data from three independent experiments showed a significantly higher CAR.OT-I CTL killing of MC57-OVA257 compared with that of MC57-HER2 (Fig. 7C). To explore the mechanism(s) leading to this observed difference in killing rate, we recovered the CAR.OT-I CTLs from the coculture at 20 and 50 hours and examined their viability (Fig. 7D) and antigen receptor (TCR and CAR) expression levels (Fig. 7E and F). CAR.OT-I CTL viability remained equivalent irrespective of the antigen stimulus and the time point (Fig. 7D). Interestingly, when ligated by the CARs, CAR.OT-I CAR expression was downregulated by 20 hours, and this persisted through to 50 hours (Fig. 7E). In contrast, when ligated via the TCRs, CAR.OT-I endogenous TCR expression levels were unchanged at 20 hours, and although TCR levels decreased by 50 hours, they were clearly detectable above the level of the negative control. In summary, persistent endogenous TCR, but not CAR, expression provides a potential explanation for the difference in long-term CAR.OT-I CTL cytotoxic function following activation via H-2Kb-OVA257 versus HER2 antigen.
CAR-T cells have evoked exciting clinical responses in CLL and ALL (1, 2, 4–6). A hallmark of many tumors is the loss of T-cell recognition molecules (MHC I), and therefore the use of CAR-T targeting tumor antigens may be of considerable clinical benefit. However, responses to CAR-T therapy in patients with solid tumors have been modest to date (8), and this is likely due to several factors, including the effect of the suppressive tumor microenvironment and low frequency of CAR-T cells trafficking to the tumor site. Despite the enormous progress in CAR-T cell therapy, there are fundamental aspects of CAR-T-cell biology that remain unstudied. Pivotal to the CAR-T-cell antitumor effect is their ability to form an effective immune synapse with a tumor target. Similarly, the respective role of the CAR versus that of the endogenous TCR in formation of the synapse and the subsequent activation of T cells remains unknown. This study addressed these issues in a model system in which individual CAR-T cells (CAR.OT-I) coexpressed two antigen receptors (OT.I TCR and anti-HER2 CAR), and the tumor cells were capable of displaying the cognate antigen for both antigen receptors.
In our methodology, CAR.OT-I cells expressed TCR at higher levels than CAR, although cognate antigen levels on MC57 tumor cells were equivalent, this created an equivalent comparison. The kinetics of CAR.OT-I cell activation, delivery of the “lethal hit” and tumor cell killing were comparable when binding to targets occurred via the TCR or CAR. Calcium flux into the effector cell is a marker of antigen recognition, and surprisingly, the time interval from Ca2+ flux until CTL detachment from the tumor cell was reduced when CAR.OT-I cells were ligated via the CAR. Nevertheless, both were equally effective in killing tumor cells in the short term. In long-term assays, CAR.OT-I cells displayed equivalent rates of killing when recognition occurred via TCR or CAR, for the first 20 hours. However TCR-stimulated CAR.OT-I cells were more effective killers between 20 and 50 hours. We subsequently showed that CAR.OT-I cells downregulate CARs to a higher degree than TCRs in response to cognate antigen long-term. This provides the most likely explanation for loss of cytotoxic function in the long-term assays. Our results are also consistent with that observed previously, in which CAR expression was downregulated following long-term in vitro culture (26). Interestingly, this study demonstrated that CAR expression could be reinduced by further activation through TCR signaling, and it would be intriguing to determine whether this was the case in CAR.OT-I cells. Regarding the endogenous TCR, prior studies have shown that effector T cells require only 10 to 100 peptide–MHC complexes on target cells for their activation and induction of cytotoxicity (27–29). Although our results show that the CAR.OT-I CTLs display reduced endogenous TCR expression by 50 hours, this is above the threshold required for T-cell activation. It is also possible that with the high avidity, CAR-T CTLs become exhausted more rapidly, and therefore studies on the formation of long-term memory cells will be of critical importance. As CAR-T-cell persistence after adoptive transfer is critical for a clinical response (1, 2, 4–6, 30), the pathway to effective CAR-T-cell memory, and the effect of TCR versus CAR engagement on this, will be further explored both in vitro and in vivo using our model system. Adoptive therapy with CAR-T cells uses autologous polyclonal T cells expressing an endogenous TCRs and CARs, and recently, dual-specific T cells (31) have also been used with success. In this context, our study suggests that patient CAR-T cells may respond differently in vivo when activated via their CARs or TCRs.
Effector-to-target ratios, avidity of interactions, and ability of CTLs to penetrate tumors may all affect the success of cell-mediated therapy. The ability of adoptively transferred CAR-T cells to undertake serial killing of multiple tumor targets is likely to be a key requirement for effective therapy, as CAR-T cells will initially be “outnumbered” by tumor cells. As has been shown with bispecific T-cell engagers (32), we also now show that individual T cells redirected to tumor antigens by CAR expression are capable of killing multiple tumor targets. This serial-killing capacity along with the ability of CAR-T cells to proliferate in vitro (9) and in vivo (CAR19 cells in CLL and ALL; refs. 4, 33) likely underpins the therapeutic success of comparatively small CAR-T-cell doses in the face of high tumor burden. Our work establishes the means by which to measure the potency of CAR-mediated T-cell activation against the internal control of endogenous TCR-mediated activation, and a means by which to measure enhanced synapse formation and/or serial killing by the addition of immune-potentiating therapies such as lenalidomide (34) or checkpoint blockade inhibitors (35). The combination therapies are likely to enhance CAR-T-cell proliferation and serial killing, resulting in maximal tumor cell death and improved clinical benefit.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: A.J. Davenport, M.R. Jenkins, H.M. Prince, D.S. Ritchie, J.A. Trapani, M.H. Kershaw, P.K. Darcy, P.J. Neeson
Development of methodology: A.J. Davenport, M.R. Jenkins, R.S. Cross, C.S. Yong, D.S. Ritchie, J.A. Trapani, P.K. Darcy, P.J. Neeson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.J. Davenport, M.R. Jenkins, R.S. Cross, P.K. Darcy, P.J. Neeson
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.J. Davenport, M.R. Jenkins, R.S. Cross, H.M. Prince, D.S. Ritchie, J.A. Trapani, M.H. Kershaw, P.K. Darcy, P.J. Neeson
Writing, review, and/or revision of the manuscript: A.J. Davenport, M.R. Jenkins, R.S. Cross, C.S. Yong, H.M. Prince, D.S. Ritchie, J.A. Trapani, M.H. Kershaw, P.K. Darcy, P.J. Neeson
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Davenport, D.S. Ritchie, P.K. Darcy
Study supervision: M.R. Jenkins, D.S. Ritchie, J.A. Trapani, M.H. Kershaw, P.K. Darcy, P.J. Neeson
The authors acknowledge the assistance of the PMCC Experimental Animal facility technicians for animal care, the PMCC FACS Core for cell sorting, and the Microscopy Core facility at PMCC for technical support. The authors thank Dr. Hideo Yagita (Juntendo University, Tokyo) for the Granzyme B antibody p1-8.
This work was funded by a program grant from the National Health and Medical Research Council (NHMRC). A.J. Davenport was supported by a scholarship from the Fight Cancer Foundation, and M.R. Jenkins is supported by a National Health and Medical Research Council of Australia (NHMRC)/RG Menzies postdoctoral training fellowship and an NHMRC New Investigator Project grant. P.K. Darcy and M.H. Kershaw were supported by NHMRC Senior Research Fellowships (#1041828 and 1058388, respectively).
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.