Phase I Escalating-Dose Trial of CAR-T Therapy Targeting CEA⁺ Metastatic Colorectal Cancers

Characteristics of CAR-T Cells

The manufacture period for CAR-T products was approximately 2 weeks, and the cell products were comprehensively measured by flow cytometry (FCM) on the day before infusion (Figure 1A). CD3⁺ (T cells; median, 94.6%; range, 92.2%–98.4%), CD16⁺CD56⁺ (NK cells; median, 3.4%; range, 0.5%–5.6%), CD3⁺CD16⁺CD56⁺ (NKT cells; median, 15.1%; range, 8.2%–24.5%), and CAR⁺ % (median, 33.7%; range, 14.7%–43.2%) subsets were present (Figure 1B). The proportion of CD4/8 in CAR⁺ products (median, 0.4; range, 0.3–0.7) showed no difference compared with that in the CD3⁺ product (median, 0.2; range, 0.1–0.5) (Figure 1C). Memory T cells contained several subtypes, and we selected CD3, CD45RA, CD45RO, and CD62L as markers for classification.¹¹ Proportion of T naive (Tn; CD3⁺/CD45RA⁺/RO⁻/CD62L⁺; median, 6.1%; range, 1.6%–18.3%), T central memory (Tcm; CD3⁺/CD45RA⁻/RO⁺/CD62L⁺; median, 9.9%; range, 5.6%–36.7%), and T effector memory (Tem; CD3⁺/CD45RA⁻/RO⁺/CD62L⁻; median, 24.4%; range, 16.1%–43.4%) cells were shown (Figure 1D). Checkpoint inhibitors were also measured in CAR-T products. Programmed death 1 (PD-1) expression was not detected on the surface of CD4⁺ and CD8⁺ cells. LAG-3 was expressed only in CD8⁺ cells (median, 14.7%; range, 7.1%–31.3%). TIM-3 was expressed in both CD4⁺ (median, 14.5%; range, 5.8%–16.5%) and CD8⁺ (median, 75.8%; range, 49.2%–78.6%) (Figure 1E). In the cytotoxicity assay, CEA CAR-T cells specifically targeting CEA high-expression tumor cells (LoVo and LS174T), CEA medium-expression tumor cells (HT-29), and CEA low-expression tumor cells (MCF7) were confirmed (Figure S1).

Adverse Events after CAR-T Therapy

We recorded adverse events and did not observe any CAR-T-related severe events (Table S1). For patients receiving DL5, patient (P) 9 tolerated the treatment well, but P7 and P10 had a fever of 39°C to 40°C within a few hours after cell infusion. P7 had a body temperature of 40°C on day 3 that lasted for 2 days discontinuously. P10 had a fever after each infusion from days 1 to 3. Nonsteroidal anti-inflammatory drug (NSAID) medication was helpful to relieve the fever, but the fever reappeared several times within the first 4 days. We considered the fever as the most significant event in response to cell infusion. Most patients with a grade 3/4 lymphocyte count decrease were expected because of lymphodepletion. After lymphodepletion, the lymphocyte levels dropped to approximately 0.1 × 10⁹/mL on day 0, and most patients recovered to approximately 0.5 × 10⁹/mL within the first week (Figure S2A). The neutrophil count was also influenced by lymphodepletion in some patients. We also assumed that the decrease of platelet (PLT) in P7 and both the alopecia and the gastritis in P10 were related to lymphodepletion. P8 had a duodenum perforation on day 20 after cell infusion. CD3⁺ cells were not found in the specimen of the perforating site (data not shown). This adverse event was not related to CAR-T therapy, and we suspected it to be due to dyspepsia. The hepatic pain of P1 and the abdominal pain of P6 were due to tumor lesions. Abnormal hepatic functions were observed only in patients with a high tumor burden (diameter >8 cm) of liver metastasis (P1, P5, and P9). Their gamma-glutamyl transpeptidase (GGT) and alkaline phosphatase (ALP) were in gradual deterioration (Figure S2B). Sudden hepatic function degradation was not detected in any patient. Although the gastrointestinal tract tissue showed CEA expression and was most likely to have the side effect of colitis, there was no colitis even at the high-dose infusion. We also monitored renal function (blood urea nitrogen [BUN], serum creatine, electrolyte, and uric acid levels) and coagulation function (activated partial thromboplastin time [APTT], thrombin time [TT], and international normalized ratio [INR]). Abnormalities were not observed. Moreover, abnormalities of the cardiac system (hypertension/hypotension and arrhythmia) and respiratory system (cough and respiratory distress) were not seen in any patient.

C-Reactive Protein and Cytokines

C-reactive protein (CRP) and IL-6 have been demonstrated to be sensitive markers for the diagnosis of cytokine release syndrome (CRS), and these two factors had a favorable correlation.12, 13 Thus, CRP and IL-6 were immediately measured in the laboratory department of Southwest Hospital each time that the trial procedure was recommended. Most patients in low DL and high DL showed a certain degree of single-peak CRP levels within the first week (Figures 2A and 2B). IL-6 levels behaved similarly (Figure S2C). The CRP and IL-6 levels of P2, P6, and P8 were constantly negative. Most factors among CRP and the 15 cytokines had the same parabolic trend of a single peak level, especially in high DL infusion (Figure 2C; Figure S2D). We performed Spearman’s correlation analysis to determine the correlations between CRP and each of the 15 cytokines (n = 93; Table S2). Among the CRP and cytokine levels we measured, interferon (IFN)-γ was correlated with all the factors except IP-10 and MIG (Figure 2D). CRP was correlated with another eight factors. Fractalkine, IL-1RA, IL-2, and MIP-1α were correlated with another 11 factors. However, levels of IL-1RA and IL-2 were mostly undetectable. IP-10 and MIG were only correlated with two and four other factors, respectively. Because of the fever of Ps 7 and 10, we suspected that IFN-γ is an even more sensitive indicator for cytokine release syndrome than CRP (Figures 2E and 2F).

Clinical Evaluation of CAR-T Therapy

We conducted the clinical evaluation of CEA CAR-T therapy through CEA level monitoring and lesion analysis. The CEA levels were analyzed and illustrated (Figure 3A). The serum CEA level of P6 was negative, but CEA expression was confirmed by immunohistochemistry (IHC) (see Table 1). Most patients had an apparent decrease in the CEA levels as response to CAR-T therapy, especially in long-term observation. Serum CEA dropped to a satisfactory level, and the decline convinced us of the efficacy of CEA CAR-T therapy. The short-term outcomes were determined by comparing the diameters of imagological lesions between enrollment and 4 weeks after cell infusion according to irRECIST (Immune-related Response Evaluation Criteria In Solid Tumors) (see Table 1). Long-term observation of tumor lesions was also conducted according to standard procedures. Seven out of 10 patients who experienced progressive disease (PD) in previous treatments had stable disease with slight variation in the tumor diameters after CAR-T therapy. Ps 3 and 5 had remained with stable disease for more than 30 weeks. The positron emission tomography (PET)/computed tomography (CT) of P9-2 showed apparent attenuation of tumor activity (Figure 3B), with standard uptake value (SUV) max attenuation from 7.38 (enrollment) to 5.80 (fourth week). Moreover, we observed the shrinkage of one liver lesion in P10 receiving DL4 by MRI (Figure 3C).

Persistence of CAR-T Cells in Peripheral Blood and Tumor Tissues

Existence and persistence of CAR-T cells were measured in peripheral blood by FCM using Protein L that is a novel reagent for CAR-T detection¹⁴ or recombinant CEA protein with His tag, and by qPCR for detection of DNA copy numbers. CAR-T cells were undetectable in peripheral blood by FCM using either Protein L or CEA. qPCR data showed that most of the patients had increasing DNA copies of CAR-T cells in the first few days in accordance with CAR-T infusion. For patients receiving low CAR-T doses (DL1 and DL2), changes of DNA copies were not significant and high number of DNA copies was not observed. However, patients receiving high doses (DL3–DL5) showed apparent DNA changes as signs to CAR-T cell infusion (Figure 4). Most of the patients quickly manifested a peak level of DNA copies after CAR-T infusion, and some patients even had a second peak level in a few weeks after CAR-T infusion. Moreover, for the three patients receiving a second CAR-T therapy 6–7 weeks after the first CAR-T therapy, DNA copies sustained high levels for a couple of days in patients receiving DL5 (P7-2 and P9-2; Figures 4B, 4D, and 4E). Nonetheless, DNA copies of all patients declined to undetectable levels 4–6 weeks post CAR-T infusion. In addition, tumor biopsies from three patients (P5: right adrenal gland; P6: greater omentum; P7-2: lung) on the fourth week after cell infusion were obtained. These specimens were formalin fixed and paraffin embedded (FFPE). We found CD3⁺ T cells in these tumor specimens by immunohistochemistry (IHC) (Figure 5A). DNA copy number in tumor tissues of Ps 6 and 7-2 was 1,029.22 and 983.29/μg genomic DNA, respectively (Figure 5B). The specimen material of P5 was not sufficient for qPCR detection.

Phenotypes of Peripheral Blood Mononuclear Cells in the Peripheral Blood

To monitor the immune environment and peripheral blood mononuclear cells (PBMCs) phenotypes, we conducted weekly follow-up of the FCM detection of the PBMCs (Figures 6A and 6B; Figure S3). We used the PBMC data at the time of enrollment as the baseline. Immune-suppressive factors, including the myeloid-derivced suppressor cells (MDSC) (CD14⁺CD45⁺HLA-DR⁻) and regulatory T cell (Treg) (CD4⁺CD25⁺CD127⁻) proportions in PBMCs and checkpoint inhibitors (PD-1) in T cells, tended to increase within the first 2 weeks (Figure 6A). Within the first 2 weeks, several patients seemed to have an apparent rise in the MDSC and PD-1 levels among PBMCs and T cells, respectively (Figure 6B). The proportions of TIM-3 and LAG-3 on T cells of most patients were undetectable (data not shown). The lymphocyte subtypes and CD4/8 ratio were determined (Figure S3A). There was no difference in the CD4/8 ratio between CAR⁺ and CD3⁺ (p < 0.01). Moreover, the Tn proportion in T cells tended to drop constantly, whereas Tcm and Tem increased to a certain extent (Figure S3B). High proportion cells of CD45RA⁺CD45RO⁺ was detected. This kind of cells was suspected a medium status during the differentiation from T naïve cells to T memory cells (Figure S3C).15, 16, 17

Carbohydrate Antigen

Carbohydrate antigens are widely expressed in CRC and are biomarkers for CRC, but they lack sensitivity and specificity. Carbohydrate antigens, including CA125, CA242, CA19-9, CA724, and CA50, were measured accompanying CEA. Only CA242 (n = 61; correlation coefficient, 0.539) correlated with CEA among the five carbohydrate antigens (Figure S4A). Even so, none of the carbohydrate antigens exhibited apparent variation or deterioration, and they were well controlled (Figure S4B).

Study Description

This study was approved by the Ethics Committee of the Southwest Hospital of Third Military Medical University (Chongqing, China) and Second Affiliated Hospital of Chongqing Medical University (Chongqing, China). This study was done according to the principles of the Declaration of Helsinki. All participants were informed of the trial procedure, possible risks, and adverse effects of the therapy. They voluntarily provided written informed consent for enrollment. The registration number on Clinicaltrials.gov is NCT02349724. No commercial sponsor was involved in the study. Eligible patients had measurable lesions and failed more than one line of systemic therapy. All patients had completed their antitumor therapies (e.g., chemotherapy) more than 2 weeks before enrollment. The interval of radiological examinations between enrollment and lymphodepletion was less than 1 month. CEA intensity was confirmed by immunohistochemistry (IHC). CEA staining was graded into three levels (+, ++, and +++), and the proportion of tumor cells at each level was judged by a blinded pathologist (see Table 1). Normal left ventricular ejection fraction (LVEF) had been confirmed for every patient. Moreover, patients were excluded if they had a hematological abnormity with a hemoglobin level less than 9 g/dL, a white blood cell (WBC) count less than 2,000/μL, an absolute neutrophil count less than 1,000/μL, or a platelet count less than 50,000/μL. Regarding hepatic and renal function, the patients were required to have a serum bilirubin level less than 3× the upper limit of normal, alanine aminotransferase (ALT) or aspartate aminotransferase (AST) less than 3× (without liver metastasis)/5× (with liver metastasis) the upper limit of normal, and BUN or creatine levels less than 3× the upper limit of normal.

There was a standard procedure for the management of each patient (Figure 1A). Basic radiological examinations and blood tests were administrated to judge whether the patient fit the enrollment criteria. CAR-T preparation lasted 14–16 days, and lymphodepletion lasted 5 consecutive days. The first four patients received CTX 300 mg/kg/day for 3 days, whereas the dose for the other patients was 900 mg/kg/day (see Table 1). All of the patients had fludarabine (FLU) 25 mg/kg/day for the last 2 days. The day after completing lymphodepletion, blood tests (routine blood analysis and measurements of C-reactive protein, CEA, and hepatic and renal function) were applied for the evaluation of the patient status for CAR-T infusion. An escalating dose was administrated based on safety considerations, and five DLs were set: 1 × 10⁵, 5 × 10⁵, 1 × 10⁶, 1 × 10⁷, and 1 × 10⁸/CAR⁺/kg cells. Each DL had three patients except for DL4 (1 × 10⁷) with two patients because a newly enrolled patient withdrew before lymphodepletion. Systemic IL-2 or other cytokine delivery was not administrated. To ensure safety, each patient received 10%, 30%, and 60% of the total cells on the first 3 days. Weekly follow-up lasted for 4 weeks, and radiological assessment was executed on the fourth week. The long-term follow-up schedule was not shown. All responses were evaluated according to the lesion diameter and immune-related response criteria (irRECIST).33, 34 Adverse events were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 4.0.

Patient Characteristics

In total, 10 CRC patients with metastasis fit the criteria and were recruited (Table 2). Most patients had liver metastasis (8/10), and half of the patients had lung metastasis (5/10). Eight of 10 patients had primary lesion resection (enterectomy). No patient has a Karnofsky score (KPS) below 80, and all patients were adenocarcinoma according to the pathology results. Eight patients were enrolled at Southwest Hospital, whereas the other two were enrolled at The Second Affiliated Hospital, Chongqing Medical University. All of the blood samples were tested at Southwest Hospital. The median number of lines of chemotherapy of patients was 2.5 (range, 1.25–3 lines). The median age was 58 years (range, 48.8–67 years). All of the radiological examinations (CT, MRI, or PET) were administrated within 4 weeks before lymphodepletion. Among the 10 patients, 4 (P1, P7, P9, and P10) received two different doses of infusions (Table 1). The interval between the two infusions was more than 6 weeks. The other six patients received only one infusion.

Generation and Quality Control of CEA CAR-T Cells

Lentivirus was produced and purified from the supernatant of 293T cells with transfection of the pCDH lentiviral plasmid, envelope plasmid PMD2.G, packaging plasmid (pMDLg/pRRE), and plasmid-producing Rev regulatory protein (pRSV-Rev), as we described previously.³⁵ The pCDH lentiviral plasmid consisted of BW431/26 scFv-IgG4 hinge-CD28 transmembrane domain-CD28-CD3ζ. Peripheral blood was collected from patients, and peripheral blood mononuclear cells (PBMCs) were isolated on day 1. PBMCs were activated by immobilized CD3 (MACS, 170-076-116) and CD28 (MACS, 170-076-117) antibodies. On day 2, T cells were infected with lentiviral vector in six-well plates (1 × 10⁶ cells/well, 5 MOI) with polybrene. After viral transduction, T cells were expanded by IL-2 (100 U/ml) for approximately 12–14 days to obtain enough cell dose. Quality control of cell products was performed by FCM. For the cell products, the CAR⁺ %, proportion of T cells (CD3⁺), NK cells (CD16⁺CD56⁺), NKT cells (CD3⁺CD16⁺CD56⁺), CD4/8 ratio, T memory cells, and checkpoint inhibitors (TIM-3, PD-1, LAG-3) on the T cell surface were measured (Figures 1B–1E) on the day before CAR-T infusion. To ensure transfection, we also applied CAR⁺ % detection 4–5 days after viral transfection.

FCM Detection

BD FACSCalibur and BD FACSAria II were the instruments used for FCM detection, and BD FACSDiva software and FlowJo were used for analysis. We administrated recombinant CEA with a His tag (11077-H08H-50; Sino Biological) for CAR⁺ %-specific detection. Antibodies specific for CD3 (344818; BD Biosciences), CD4 (300506; BD Biosciences), CD8 (344740; BD Biosciences), CD14 (555397; BD Biosciences), CD25 (555433; BD Biosciences), CD45 (560367; BD Biosciences), CD45RA (304130; BD Biosciences), CD45RO (304222; BD Biosciences), CD62L (555544; BD Biosciences), CD127 (562436; BD Biosciences), HLA-DR (555813; BD Biosciences) Protein L (21189; Thermo Fisher Scientific) and Alexa Fluor® 647 conjugate Streptavidin (S21374; Thermo Fisher Scientific) were applied. TIM-3 (536422; BD Biosciences), PD-1 (560908; BD Biosciences), and LAG-3 (565616; BD Biosciences) were also detected in CD4⁺ and CD8⁺ cells. Lymphocyte subtypes were measured using the MultiTEST IMK Kit (340503; BD Biosciences). Isotype antibodies and untransduced T cells (UTD) were used as controls.

Quantitative DNA Copies

Blood samples were obtained according to the trial procedure (Figure 1A). Genomic DNA was extracted from the blood samples using a QIAamp DNA Blood Mini Kit (51106; QIAGEN) and following the protocol. We applied TaqMan probes for qPCR in an ABI ViiA 7. TaqMan primers were 5′-GAGGACATCGCCACCTACTACTG-3′ (forward) and 5′-GTCCCTTGGCCGAACGT-3′ (reverse). Probe was 5′-7VIC-ATCAGTGGAGTAGTTATC-3′. The TaqMan method was performed in accordance with the published protocol.³⁶ CDKN1A was used as the control and produced a correction factor (CF) for correction of DNA copy number. Healthy donor DNA sample was detected as negative control and definition of pre-infusion acceptance ranges of qPCR method. The lower limit of quantification (LLOQ) was determined at 50 copies/μg genomic DNA.

Cytokines

Plasma samples were stored at −80°C and were analyzed in batches using a 15-plex human cytokine magnetic bead panel (SPR372; Milliplex) with the MAGPIX instrument (Luminex). The levels of factors including IL-2, IL-6, IL-10, Fit-3L, Fractalkine, GM-CSF, IFN-γ, tumor necrosis factor alpha (TNF-α), IL-2Rα, IL-1RA, MIG (CXCL9), IP-10, MCP-1, MIP-1α, and MIP-1β were measured. Samples were undiluted, and each was assayed in duplicate. Data analysis was conducted according to the related protocol and algorithm.

Statistical Analysis

Descriptive statistics (median and ranges) were used to aggregate the data. Comparison of the CD4/8 ratio in CAR⁺ and CD3⁺ cells was performed using a paired t test. The levels of CEA, carbohydrate antigens, and cytokines were not normally distributed. The correlation between CEA and other carbohydrate antigens was determined by Spearman’s correlation. Spearman’s correlation was also applied to analyze the correlation between C-reactive protein (CRP) and cytokines. The correlation coefficient was also calculated. p < 0.05 was considered to be statistically significant.