MC-38 cell line is CEA negative murine colon cancer cell derived from C57BL/6 mice, and MC-38-CEA-2 (MC-32) cells are CEA positive MC-38 cells retrvirally transduced with human CEA gene (23). Both cell lines were obtained from Kerafast Inc. (Boston, MA). C57BL/6 (B6) and Thy1.1 congenic mice (B6.PL-Thy1a/CyJ) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA).

β-actin (#4967), Bcl-xL (#2762), Myc (#14819 and #12855) and peroxidase-conjugated anti-rabbit (#7054) or anti-mouse Ig (#7056) for Western blot, were obtained from Cell Signaling Technology (Beverly, MA). Anti-mouse CD3 (145-2C11) and CD28 (37.51) antibodies, mouse IL-2 and IFN-γ ELISA Kits, all FITC-, PE-, PE/Cy5-, PE/Cy7, and APC-conjugated antibodies, PE Annexin V Apoptosis Detection Kit (640934) and Cyto-Fast™ Fix/Perm Buffer Set (#426803) were purchased from Biolegend (San Diego, CA)

Tumor cells were thawed in 37°C and washed with fresh RPMI-1640 media twice before incubation, and cultured with 5% CO₂ at 37°C in complete RPMI-1640 media supplemented with 10% heat-inactivated FBS, 1% penicillin/streptomycin and 1% L-glutamine. Retrovirus-packaging Plat-E cells expressing retroviral particles pol, gag and env to package the retroviral vector DNA coding sequences to produce retroviruses without the probability of generating replication competent viruses (24) were used for retroviral transduction. The expansion of Plat-E cells began on day 4 before the retrovirus packaging operation and were passaged at the density of 3×10⁶ cells in 10 ml of DMEM per 10cm plate at least 16 h before transfection. Naive CD8⁺ T cells were purified from the spleen and lymph nodes of C57BL/6 mice by using the murine naive CD8a⁺ T cell isolation kit (#130-096-543, Miltenyi Biotech, CA) and cultured in complete media additionally supplemented with 0.06% β-mercaptoethanol based on the complete RPMI-1640 media.

Anti-human CEA (hMN14) (CD28-CD3ζ) CAR and cDNA for Bcl-xL was linked with 2A sequence and subcloned into the retroviral vector MFG (20, 25), which was derived from murine leukemia virus (MLV)-based retroviral vector and had been demonstrated ready to be modified to improve safety and gene expression (26). Three constructs used in the present study are Vector 1: MFG-control vector (empty plasmid without interest gene); Vector 2: MFG-anti-CEA CAR (MFG-based vector subcloned an anti-CEA CAR sequence); and Vector 3: MFG-anti-CEA CAR+Bcl-xL (MFG-based vector with the same CAR with Bcl-xL gene). The CAR structure in the MFG constructs consists of a synthetic single chain variable fragment (scFv, derived from mouse monoclonal antibody, clone MN-14) specific to hCEA epitope expressed on MC38-CEA-2 cells in which a Myc epitope-tagged framework was engineered to enable the tracking of scFV expression, a spacer and a transmembrane domain (both derived from CD28), a single costimulation domain (derived from human CD28), and a T cell receptor (TCR) signaling element (derived from human CD3ζ chain). CAR in Vector 3 contains a Bcl-xL gene with CD3 ζ through 2A element. The parental MFG-based retroviral plasmids were expanded in MAX Efficiency™ DH5α Competent Cells (ThermoFisher, #18258012) and purified with Invitrogen™ PureLink™ Expi Endotoxin-Free Maxi Plasmid Purification Kit (ThermoFisher, #A31231) according to the manufacturer’s instruction. All constructs were verified by sequencing.

Retroviral transduction was performed as described before (27). Briefly, Plat-E packaging cells using the ecotropic envelope were transfected with the CAR constructs and supernatants from the transfected packaging cells were collected 48 h after transfection, and used for transduction of T cells. Murine mature T cells from lymph nodes and spleen were used in all experiments. 5 ×10⁵ CD8⁺ T cells were stimulated with anti-CD3 plus anti-CD28 antibodies. After 2 days, the supernatant was replaced with 1 ml viral supernatant containing 5 µg/ml Polybrene (Sigma), and the cells were spun for 1 h at 32°C and incubated at 32°C for 8 h. This was repeated the following day. Viral supernatant was removed and replaced with fresh medium, and T cells were re-cultured. Expression of GFP was determined by flow cytometry gating on CD8⁺ T cells. GFP-expressing T cells were purified from cell sorting using a FACS Vantage SE I high-speed cell sorter (BD Immunocytometry Systems, San Jose, CA).

5 ×10⁵ CD8⁺ T cells were stimulated with anti-CD3 plus anti-CD28 antibodies in 48-well plates. Cytokines were measured by enzyme-linked immunosorbent assay (ELISA; Biolegend); T cell survival in vitro was determined by trypan blue (Sigma) exclusion assay; and proliferation was measured in triplicate cultures by incorporation of ³H-thymidine (1 µCi/well; ICN Pharmaceuticals, Laval, QC, Canada) during the last 12 h of culture (20).

Live CD8⁺ cells were recovered by Ficoll treatment and positive selection with anti-CD8 microbeads (Miltenyi Biotec Inc). Cells lysates were extracted and used for Western blotting as described (28).

Target MC-38 (hCEA^-) or MC-32 (hCEA⁺) tumor cells were co-cultured with effector CEA CAR-T cells at the target:effector (E:T) ratio in 1:2, 1:5, or 1:10 in triplicate. For test of background, wells contained target cells only. The plates were incubated at 37°C for 12 h, and the CAR-T cell-mediated cytotoxicity was measured using the Cayman’s 7-AAD/CFSE Cell-Mediated Cytotoxicity Assay Kit according to the manufacture’s instruction by flow cytometry. In brief, both prepared MC-38 and MC-32 cells were stained with CFSE dye prior to co-culture with CAR-T cells according to the protocols previously described (29). The percentage of specific lysis was calculated as follows: cytotoxicity (%): [100% × dead targets/(dead targets + live targets)] (experiment) − [100% × dead targets/(dead targets + live targets)] (background). In addition, an Amnis ISXII (MilliporeSigma) equipped with 405, 488, and 642 nm lasers with a single camera (six channels) was used to acquire experimental samples using the INSPIRE software.

T cells were cultured with anti-CD3 plus anti-CD28 antibodies and transduced on day 2 and 3 with retroviral vectors (27). Cells were re-cultured for 2 more days. GFP⁺ CD8⁺ T cells were sorted and 3 × 10⁶ sorted cells were injected i.v. into C57BL/6 mice, which were challenged s.c. with 5 × 10⁶ MC-38 or MC-32 tumor cells in PBS, or PBS without tumor cells as a control one week prior to the adoptive T cell transfer. The volume of the tumor (mm³) was measured using a caliper by a blinded investigator and calculated as follows: V = length × width² × 0.52. Mice were sacrificed when the tumor size reached 20mm in any direction.

Tumor tissues were collected from mice at the indicated time points. Samples were collected in embedding cassette and blocked with 10% neutral buffered formalin. Samples were infiltrated with the wax and embedded the infiltrated tissues into wax blocks. Both vertical and horizontal sectioning were prepared for immunostaining.

H&E staining: Routine Hematoxylin & Eosin (H&E) staining was performed at an interval of every five serial sections. Immunological staining: Tissue sections were fixed with acetone (Sigma), and incubated with 3% bovine serum albumin (BSA; Sigma) to block non-specific protein binding (30). Sections were stained with fluorescein isothiocyanate (FITC) anti-Myc (1:1000; Thermo Fisher #13-2511).

The data were presented as the mean ± standard error of mean (SEM). Differences in means were analyzed by Student’s t-test. One-way ANOVA with Tukey post hoc tests were used for differences between three or more groups in a single condition or time point. Survival curves were constructed by the Kaplan−Meier method and analyzed with the log-rank test. All tests were 2-sided, with p < 0.05 considered to indicate statistical significance. All statistics were calculated using GraphPad Prism 8 (San Diego, CA, USA).

MFG-based retroviral vector had been demonstrated a reliable gene-edition system with high transduction efficiency and enable engineered cells highly expressing the interest protein (31). The 2A peptide regions from Picornavirus has been widely used to create an individual fragment encoding multiple proteins (32). To generate reliable and versatile constructs to transduce primary CD8⁺ T cells that permit the expression of multiple genes, we used a T2A sequence to generate multicistronic retroviral vectors with efficient translation of two cistrons (e.g., the CEA scFv-TM28-CD28-CD3ζ CAR with Myc tag and Bcl-xL) (Figure 1A). The CEA scFv was derived from mouse monoclonal antibody clone MN14.

Thus, two fragments of the CEA CAR and Bcl-xL were linked with the 2A sequence and were subcloned into the MFG vector. The new construct MFG-CEA CAR-2A-Bcl-xL was confirmed by DNA sequencing (Supplementary Data 1) and GFP expression in the packaging Plat-E and primary T cells (Figure 1B). Furthermore, naive CD8⁺ T cells were infected with the retrovirus-mediated transduction, which led to the surface expression of CEA CAR (Myc⁺) (Figure 1C) and exogenous expression of Bcl-xL (32 kDa; Figure 1D).

To determine whether enforced expression of Bcl-xL could contribute to the survival of CEA CAR-T cells, we compared the proliferation, recovery and apoptosis rate of transduced CD8⁺ T cells with the MFG retroviral constructs containing Bcl-xL or not. The primary CD8⁺ T cells were activated with anti-CD3 plus CD28 antibodies, and after retroviral transduction on days 2-3, T cells were passively re-cultured in the absence of further stimulation and their proliferation was assessed by thymidine incorporation after 1 (day 4) and 3 days (day 6). Bcl-xL-forced expression in CD8⁺ T cells resulted in enhanced passive proliferation at late times at day 6, as measured by thymidine incorporation, compared to only CEA CAR expression (Figure 2A). In line with this, enumerating the recovery of live T cells through monitoring GFP expression showed that expression of Bcl-xL allowed CD8⁺ T cells to expand from day 4 through day 6 over that engendered by transducing only CEA CAR in isolation. Culture over 8 days showed that, Bcl-xL significantly enhanced the ability of CD8⁺ T cells to survive (Figure 2B). In addition, the percentage of apoptosis in CD8⁺GFP⁺ T cells was reduced in anti-CEA CAR+Bcl-xL transduction group compared to transfection of only anti-CEA CAR (Figure 2C).

To investigate whether overexpression of Bcl-xL promote greater recall responses of CEA CAR-T cells, effector CD8⁺ T cells from C57BL/6 mice expressing CEA CAR or CEA CAR with Bcl-xL from primary naive cultures were sorted based on GFP expression and re-stimulated. CD8⁺ T cells transduced with CEA CAR with Bcl-xL displayed enhanced recall proliferation (Figure 3A), and greater numbers were recovered over time (Figure 3B) compared to the introduction of CEA CAR alone in isolation. In addition, effector function was not affected in which production of IL-2 or IFN-γ was unaltered regardless of the forced expression of Bcl-xL (Figure 3C). However, the percentage of apoptosis in CD8⁺GFP⁺ T cells transduced with CEA CAR with Bcl-xL was decreased compared to the introduction of CEA CAR alone in isolation. The percentage of early apoptotic cells (Annexin V⁺7-AAD^-) reduced less (4.18% with CEA CAR with Bcl-xL vs. 5.78% in MFG vector control and 5.16% with CEA CAR alone), but the percentage of late apoptotic cells (Annexin V⁺ 7-AAD⁺) markedly lessened (9.72% with CEA CAR with Bcl-xL vs. 30.2% in vector control and 31.5% with CEA CAR alone) (Figure 3D).

Collectively, our results demonstrated that expression of Bcl-xL substantially promotes the survival of CEA CAR-T cells when measuring short-term T cell expansion in vitro.

To determine this functional activity of CEA CAR-T cells, we co-cultured the CAR-T cells with murine tumor cells: MC-32 (expressing human CEA) and MC-38 (lacking human CEA) (Figure 4A). No apparent cytotoxicity was observed in CEA negative MC-38 cells at the target:effect or ratio of 1:5 for each transduced cell. The control T cells showed no cytotoxicity to MC-32 cells at any target:effector ratio and the CEA CAR-T cells showed significantly increased cytotoxicity to MC-32 cells in the co-culture system at various target:effector ratio of 1:2, 1:5, 1:10 (Figure 4B), indicating a CEA-specific antitumor reaction. Furthermore, the CAR-T cells over-expressing Bcl-xL or not displayed a dosage-dependent cytotoxicity to MC-32 cells, but no significant difference was observed between these two groups, indicating that the exogenous Bcl-xL protein did not affect on the overall antitumor function of the CAR-T cells in vitro.

Amnis image stream analysis showed colocalization of CAR-T cells and tumor cells, which suggest that the scFv of the CAR expressed on CEA CAR-T could bind with hCEA on tumor cell stably and activated the CAR-T sufficiently, because both CAR-T with or without Bcl-xL were far bigger than T cell transduced with empty MFG vector. The data also indicate that the cytotoxicity of CAR-T is cell-cell direct-contact dependent (Figure 4C).

To determine whether Bcl-xL was capable of increasing the expansion or persistence of CEA CAR-T cells in response to antigen presented in vivo, GFP-sorted CEA CAR-T cells (Thy1.2⁺), obtained from the in vitro cultures in Figure 3 and adoptively transferred into syngeneic recipients (Thy1.1⁺). These mice were subsequently challenged with anti-CD3 CD3ϵ F(ab’)₂ fragment. Activated T cells transduced with either the vector control or the CEA CAR expanded less over 3 days in the lymph nodes and spleen than those CEA CAR-T cells expressing Bcl-xL (Figures 5A, B), supporting the in vitro results (Figure 3). The effect of Bcl-xL is long lasting, with enhanced numbers of CEA CAR-T cells not only present a week after the anti-CD3 challenge through the peak of response, but also after two weeks while the secondary in vivo response was completed and contraction of T cell populations had followed in all recipients (Figures 5C, D). Overall, these data strongly support the conclusion that an action of Bcl-xL sustains the persistence of CEA CAR-T cells.

To demonstrate that the CEA CAR-T cells overexpressing Bcl-xL have the ability in inducing CEA-specific T cell persistence in a physiologically and clinically relevant setting, we used a murine model of colorectal cancer (Supplementary Data 2). We s.c. injected C57BL/6 mice in the flank region with MC-32 or control MC-38 tumor cells. After a week, we performed i.v. adoptive cell transfer of genetically engineered T cells. Three weeks after the adoptive cell transfer, we observed that more CEA-reactive CD8⁺ T cells accumulated in MC-32 tumor tissues in mice receiving CEA CAR-T cells overexpressing Bcl-xL than those receiving the CEA CAR-T cells without overexpressing Bcl-xL or the non-specific T cell control (Figure 6). Mice receiving CEA CAR-T cells overexpressing Bcl-xL had the smaller MC-32 tumor sizes as compared to those receiving the CEA CAR-T cells without overexpressing Bcl-xL or the non-specific T cell control (Figure 7A), correlating with their enhanced survival (n=6; Figure 7B). Taken together, these findings indicate that the adoptive transfer of CEA CAR-T cells overexpressing Bcl-xL can generate CEA-specific CTL persistence and result in enhanced tumor suppression.