miR-153 suppresses IDO1 expression and enhances CAR T cell immunotherapy

The human noncancerous colon cell line CCD-841 and colon cancer cell lines HCT-116, SW620, HT-29, and DLD-1 were purchased from the American Type Culture Collection (ATCC) (Manassas, VA). Immortalized lung cell line BEAS-2B and lung cancer cell lines A549 and H-1299, 293 T, HeLa, and blood cancer cell lines (Jurkat, K562, Daudi, and Raji) were also from ATCC. All cell lines were authenticated by short tandem repeat profiling and were tested for and found to be mycoplasma-free. All cells were cultured in media recommended by ATCC. The efficiency for transient transfection of miR-153 miRNA mimic or anti-miR-153 in 3 colon cancer cell lines ranged from ~ 40-70%.

For the CAR lentiviral vector, we replaced the copGFP gene in pSIH-copGFP with a CAR-expressing cassette against a variant of EGFR. The variant III mutation of the epidermal growth factor receptor (EGFRvIII) leads an in-frame deletion of a portion of the extracellular domain, creating a neoepitope [13]. The scFv we used is from cetuximab, a humanized monoclonal antibody against EGFRvIII. The K_D for this humanized scFv is 101 nM for EGFRvIII and 872 nM for WT EGFR [13]; therefore, it can target colon cancer cells expressing WT EGFR, albeit at a low efficiency. The scFv is fused to a CD8α hinge and transmembrane domain, and the intracellular domains of human CD28, 4-1BB, and CD3ζ [13]. Isolated T cells were derived from leukapheresis products obtained from de-identified healthy donors at Gulf Coast Regional Blood Center (Houston, TX). T cells were stimulated with Dynabeads Human T-Activator CD3/CD28 (ThermoFisher) at a bead-to-cell ratio of 3:1. T cells were then transduced by lentivirus carrying the CAR or the parental vector. The recombinant T cells were named CAR T or wild-type (WT) T and were expanded as described [13]. To detect CAR expression, we labeled recombinant T cells with Alexa Fluor® 647 AffiniPure Fab Fragment Donkey Anti-Mouse IgG (Jackson ImmunoResearch Laboratories, #715-607-003) and performed flow cytometry.

DLD-1-luc and HCT-116-luc tumor cells were generated and employed in a modified version of a luciferase-based cytotoxicity assay [14]. Briefly, lentivirus carrying the firefly luciferase (luc) gene was transduced into DLD-1 or HCT-116 cell lines, followed by puromycin selection. The resulting recombinant DLD-1-luc or HCT-116-luc cells were resuspended at 2 × 10⁴ cells per well in 96-well plates and co-cultured with naive T cells or CAR T cells at varying ratios. After overnight incubation, luciferase substrates were added, and the relative luminescence units (RLU) were determined. Percent killing was calculated as 100 − 100 × [(RLU from wells with T cells and target cells)/(RLU from wells with target cells only)]. To analyze cellular apoptosis under T cell killing, we used the IncuCyte Caspase-3/7 Green Apoptosis Assay from Essen Bioscience (Ann Arbor, MI). DLD-1 or HCT-116 cells were co-cultured with varying amounts of T cells in the presence of NucView 488-labeled caspase 3/7 substrate (Essen Bioscience). After 24 h, green fluorescent (free NucView 488) signals were measured and used as a marker for dead cells. For the tumor cell apoptosis assay, cells were resuspended in a buffer containing annexin V (APC) and propidium iodide (PI) (BioLegend, San Diego, CA) and incubated for 10–15 min at room temperature before flow cytometry. For cell cycle analyses, cells were fixed with 70% ethanol for 30 min, then incubated with 50 μg/ml PI and 100 μg/ml RNase A at room temperature for 15 min. Stained cells were measured using a FACS-Calibur instrument (BD Biosciences, San Jose, CA) and analyzed with Cell-Quest 3.3 software. To analyze the expression of PD-L1 and EGFR, cells were stained with fluorochrome-labeled antibodies for 15 min before being subjected to flow cytometry. All flow cytometry data were analyzed using FlowJo 10 (FLOWJO, LLC; Ashland, OR).

For ELISA, target tumor cells were co-cultured with T cells overnight at 37 °C, and supernatants were harvested and subjected to ELISA for cytokine production according to the manufacturer’s instructions (eBioscience). For the soft-agar assay [15], 5000 viable cancer cells were seeded in 1.5 mL tissue culture medium with 1% glutamine on top of 0.4% soft agar layered onto 0.8% solidified agar with tissue culture medium in 6-well plates. After incubation for 15 days, the colony foci were counted under a microscope. For the wound-healing assay [16], cancer cells were transfected as indicated and cultured for 24 h and then starved in culture medium containing 0.2% FBS for 12 h; the monolayer of confluent cells was scratched using a 10-μL pipette tip and cells were then photographed at different time points under an Olympus IX81 microscope. The relative wound areas were measured using ImageJ software (NIH, Bethesda, MD). When DLD-1 tumor cells were co-cultured with EGFRvIII-CAR T cells, the medium was first preconditioned by the tumor cells for 5 days.

NSG mice were purchased from the Jackson Laboratory (Bar Harbor, ME) and were housed under standard housing conditions at the animal facilities in the Cleveland Clinic Lerner Research Institute. The precursor to miR-153 (Pre-miR-153) sequence was inserted into a lentiviral vector pSIH-copGFP from System Biosciences (Palo Alto, CA) to generate pSIH-miR-153 [17]. DLD-1-luc cells were infected by lentivirus carrying miR-153, and 5 × 10⁶ recombinant cells (DLD-1 + miR-153) were subcutaneously injected into 6- to 8-week NSG mice (n = 5). As a control for DLD-1-miR-153 (DLD + NC), DLD-1-luc cells were infected with virus carrying the parental vector. After 7 days, 2 × 10⁶ EGFRvIII-CAR T cells or control T cells were injected through the tail vein. For bioluminescence imaging, mice were given luciferin 5 min before anesthesia (3% isoflurane), and imaging was performed and analyzed using the IVIS Spectrum In Vivo Imaging System (Perkin Elmer, Waltham, MA). All animal experimental procedures were approved by the Institutional Animal Care and Use Committee of the Cleveland Clinic.

The data were analyzed with GraphPad (GraphPad Prism, version 5.02, La Jolla, CA). An unpaired two-tailed Student’s t test was performed for two-group comparisons. One-way analysis of variance was performed for multiple group comparisons with one independent variable. For patient survival, Kaplan-Meier analysis was followed by a log-rank Mantel-Cox test. Significance was set at *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. The Cancer Genome Atlas (TCGA) data were downloaded from TCGA portal (https://​portal.​gdc.​cancer.​gov/​) or OncoLnc (http://​www.​oncolnc.​org/​).

We determined the expression of IDO1, IDO2, and TDO in cancer cells treated with IFN-γ. We first used various concentrations (1 to 300 ng/ml) of IFN-γ to treat DLD-1 and A549 cells for 3 to 48 h and found in both cell lines that at 30 ng/ml IFN-γ, IDO1 expression rose to its highest level and plateaued at 24 h treatment as determined by flow cytometry (Additional file 1: Figure S1). We next treated the remaining cell lines with 30 ng/ml IFN-γ for 24 h and determined the mRNA levels of IDO1, IDO2, and TDO using quantitative real-time PCR (qPCR). As shown in Fig. 1a, IFN-γ upregulated IDO1 in all four colon cancer cell lines and two lung cancer cell lines, as well as in noncancerous colon epithelial CCD-841 cells, but not in immortalized lung epithelial BEAS-2B cells. IDO2 was significantly upregulated by IFN-γ in all six cancer cell lines (HT-29, HCT-116, SW620, DLD-1, H-1299, and A549), whereas TDO was upregulated in only A549 cells. Flow cytometry analyses validated IDO1 upregulation in these cancer cell lines (Fig. 1b). As shown on western blotting in Fig. 1c, IFN-γ treatment significantly elevated the IDO1 protein level in four colon cancer cell lines (HT-29, HCT-116, SW620, DLD-1) and two lung cancer cell lines (H-1299 and A549) and moderately increased IDO1 in CCD-841 and BEAS-2B cells. In contrast, IDO1 expression was undetectable in blood cancer cell lines (K562, Jurkat, Raji, and Daudi) with or without IFN-γ treatment.

We analyzed IDO expression in tumors from colorectal cancer patient data in the TCGA. Based on RNA read numbers, IDO1 expression was much higher than IDO2 and TDO expression (Fig. 1d, n = 434 colorectal patients). From 382 patients with survival data, we found that the average overall survival was reduced by 5 months in IDO1^high patients compared with IDO1^low patients (n = 191 in each group, Fig. 1e, left panel); however, this was not statistically significant (P = 0.4). For patients with earlier stage cancer (stage 2; too fewer stage 1 patients to be included), lower IDO1 expression was associated with better survival (Fig. 1e, right panel, P < 0.001); IDO1 expression was much higher in those patients who had already died (n = 18, mean overall survival = 7 months) than in patients who were still alive (n = 32, mean overall survival = 10 months, Fig. 1f). However, no significant correlation between IDO2 or TDO gene expression levels and patient survival was found (Additional file 2: Figure S2).

We sought to study the regulation of IDO1 gene expression mediated by miRNAs. According to the computational program TargetScan version 7.1, which predicts biological targets of miRNAs by searching for the presence of conserved sites that match the seed region of each miRNA [18], the human IDO1 gene is putatively targeted by miR-153 and other 9 miRNAs (Additional file 3: Figure S3A). We cloned the full-length IDO1 3′ untranslated region (3′UTR) downstream of the firefly luciferase 2 (luc2) gene in the pmirGLO vector. The pmirGLO vector constitutively expresses the Renilla luciferase gene as a reference reporter. We introduced these 10 miRNAs and pmirGLO-IDO1-3’UTR into 293 T or Hela cells. We found only miR-153 downregulated luc2 expression in both 293 T and Hela cell lines (Additional file 3: Figure S3B and Fig. 2a). When the miR-153 binding site in the IDO1 3’UTR was mutated, the reporter downregulation by miR-153 was abolished (Fig. 2a).

We transfected miR-153 mimics or a negative control into colon cancer cell lines DLD-1, HCT-116, and HT-29 and treated the cells with IFN-γ. Compared with mock treatment (transfection without agents) or the negative control, miR-153 substantially decreased IFN-γ-induced IDO1 expression in all cell lines as determined by flow cytometry (Fig. 2b) and about 60–90% by western blotting (Fig. 2c). The difference between flow cytometry and western blotting was likely a result of methodological variances (western blotting measures total soluble proteins from a pool of cells; flow cytometry measures the fluorescent density of individual cells). We noted that PD-L1, another IFN-γ-induced protein, was not suppressed by miR-153 (Fig. 2b). To corroborate this finding, we next transfected a miR-153 expression vector containing both miR-153 and GFP into DLD-1 cells and stained cells using antibodies against IDO1. We found that the IDO1 staining signal was substantially reduced in the transfected GFP+ cells compared with the untransfected (GFP−) cells (Fig. 2d). miR-153 also decreased IDO1 expression in monocyte-derived dendritic cells (Fig. 2e). Next, we used LPS or Cp-G DNA instead of IFN-γ to treat DLD-1 cells and found that both LPS and Cp-G DNA induced IDO1 expression. Introduction of miR-153 resulted in lower IDO1 expression in DLD-1 cells treated with LPS or Cp-G DNA (Additional file 4: Figure S4A and B). IDO-1 downregulation by miR-153 was also observed in HCT-116 cells treated with LPS (Additional file 4: Figure S4C). miR-153 decreased IDO1 mRNA expression in DLD-1 cells, but not in HCT-116 cells (Additional file 4: Figure S4D). We noted that mammalian miRNAs downregulate gene expression through both mRNA degradation and/or translational repression [19]. Nonetheless, these data suggest that IDO1 is a bona fide target gene of miR-153. We analyzed miR-153 expression in relationship to IDO1 mRNA levels in patients with colorectal cancer and found a significant inverse correlation between miR-153 and IDO1 mRNA levels (P = 0.04, R² = 0.23; Fig. 2f). However, patients’ survival was not associated with miR-153 expression levels.

We determined whether miR-153 affects colon cancer cell phenotypes through a series of in vitro cell-based assays. As shown in Fig. 3, cell proliferation, apoptosis, and cell cycle arrest were not changed by miR-153 overexpression in DLD-1 and HCT-116 cells with or without IFN-γ treatment (Fig. 3a, b). IFN-γ treatment increased cancer cell migration (Fig. 3c), but slightly decreased cell proliferation after 1 to 6 days (Fig. 3d), and decreased colony formation (Fig. 3e). However, miR-153 introduction into either HCT-116 or DLD-1 cells barely altered cell proliferation, cell migration, and colony formation. Previous reports demonstrate that miR-153 overexpression reduces cell proliferation in lung cancer cells [20], melanoma cells [21], and glioblastoma stem cells [22], but does not affect cell proliferation in colon cancer cells [23]. Our data demonstrate that overexpression of miR-153 and subsequent IDO1 inhibition do not alter cell proliferation and other cellular processes in colon cancer cells. These data suggest that miR-153-mediated cell proliferation regulation is cancer cell-type specific. It is notable that overexpression of miR-153 downregulated IDO1 expression about 50% in HCT-116 and DLD-1 cells (Fig. 2c), which did not affect cell proliferation, cell migration, and colony formation of these cells (Fig. 3). However, when IDO1 expression was completely depleted in HCT-116 and HT-29 cells, cell proliferation was significantly reduced [24]. These results imply that moderate downregulation of IDO1 by miR-153, but not elimination of IDO1, has little effect on colon cancer cells.

EGFR was expressed at a high level on the surface of DLD-1 and HCT-116 (Fig. 4a). We constructed a CAR targeting the variant III mutation of EGFR (EGFRvIII) in a lentiviral vector. The scFv against EGFRvIII was from cetuximab, a humanized monoclonal antibody against EGFRvIII, which has a K_D of 101 nM for EGFRvIII and 872 nM for WT EGFR [13]. Thus, this EGFRvIII-CAR can target colon cancer cells expressing WT EGFR, albeit at a low efficiency. The scFv was fused to a CD8α-derived hinge and transmembrane domain, both CD28 and 4-1BB co-stimulation domains linked to the CD3ζ signaling domain, as previously reported [13]. The GM-CSF leader sequence was used for efficient expression and localization of CAR molecules (Fig. 4b). We generated CAR T cells expressing EGFRvIII-CAR (CAR T) or control cells (WT T) that were transduced by virus carrying the parental vector as described [13]. We labeled CAR T or WT T cells with antibodies recognizing the CAR, and the EGFRvIII-CAR transgene expression was detected in CAR T cells but not WT T cells (Fig. 4c). qPCR analyses further detected the expression of the EGFRvIII-CAR mRNA in CAR T cells, but not in the WT T cells (Fig. 4d). Compared with WT T cells, CAR T cells released significantly more of the cytotoxic cytokines IFN-γ, TNF-α, and IL-2 (Fig. 4e) and had greater proliferation when they were co-cultured with HCT-116 cells (Fig. 4f). In addition, CAR T cells showed robust in vitro tumor cell killing activities when co-cultured with both HCT-116 and DLD-1 cells, as the co-cultured tumor cells released more caspase 3/7 proteins (Fig. 4g) and had greater cell lysis (Fig. 4h). Taken together, these results indicate that CAR T cells specifically recognize EGFR-positive tumor cells and enhance tumor cell apoptosis.

We next investigated potential anti-tumor benefit of CAR T cells when they encounter tumor cells overexpressing miR-153. A stable DLD-1 cell line overexpressing luciferase (DLD-1-luc) was transduced by a lentivirus carrying an expression cassette for miR-153, resulting in the DLD-1 + miR-153 line. A control cell line, DLD-1-NC, was obtained by infecting DLD-1-luc cells using a virus carrying the parental vector. We found that CAR T cells co-cultured with DLD-1 + miR-153 produced more IL-2, TNF-α, and granzyme B and had higher proliferation and potentially higher T cell activation than those co-cultured with DLD-1-NC (Fig. 5a, b). The expression of T cell activation markers, checkpoint markers, and co-stimulation markers was similar to each other (Additional file 5: Figure S5). These data indicate that IDO1 downregulation by miR-153 in colon cancer cells enhances the cytotoxicity of co-cultured CAR T cells. Our results are in agreement with a previous study in which higher expression of IL-2, TNF-α, and granzyme B in CAR T cells was also observed, when similarly made CAR T cells were co-cultured with target tumor cells with EGFR or EGFRvIII [13]. The underlying mechanism could be reduction of immunosuppression by IDO1.

We next injected DLD-1 + miR-153 and DLD-1 + NC cells into the flank of immunodeficient NSG mice (n = 5). Animal body weight was measured (Fig. 5c), and tumor growth was followed using in vivo bioluminescence monitoring (Fig. 5d). Mouse body weight within each group changed little during tumor cell engraftment and subsequent treatment with EGFRvIII-CAR T cells or control T cells (Fig. 5c). Tumor growth was similar between the two groups at day 5 after tumor cell inoculation, when we infused mice with CAR T cells or WT T cells via the tail vein. At day 10, tumors were smaller in mice injected with either DLD-1 + NC or DLD-1 + miR-153 treated with CAR T cells than those in mice with DLD-1 + miR-153 and WT T cells; mice with DLD-1 + miR-153 and CAR T injection had the smallest tumors. At day 15, tumors were eradicated in 5/5 mice with DLD-1 + miR-153 and EGFRvIII-CAR T and in three of five mice with mice with DLD-1 + NC and EGFRvIII-CAR T. Four of five mice with DLD-1 + miR-153 and WT T cells developed larger tumors, some of which had metastasized (Fig. 5d). These data support that autonomous overexpression of miR-153 within tumor cells enhanced anti-tumor activities of CAR T cells.