Research ArticleCancer

Immunotherapy of non-Hodgkin’s lymphoma with a defined ratio of CD8+ and CD4+ CD19-specific chimeric antigen receptor–modified T cells

See allHide authors and affiliations

Science Translational Medicine  07 Sep 2016:
Vol. 8, Issue 355, pp. 355ra116
DOI: 10.1126/scitranslmed.aaf8621

Standardizing the CAR assembly line

Chimeric antigen receptor (CAR)–modified T cells are engineered to recognize specific tumor antigens. They have shown promising results in clinical trials, primarily in leukemia so far, but it has been difficult to predict therapeutic efficacy and toxicity for individual patients. To address this issue, Turtle et al. treated non-Hodgkin’s lymphoma patients with CAR-T cells prepared from strictly defined subsets. By carefully controlling the ratio of CD4 to CD8 T cells, the authors were able to identify some of the treatment characteristics that correlate with therapeutic response and toxicity, including the role of the drug regimen used for lymphodepletion before CAR-T cell treatment.

Abstract

CD19-specific chimeric antigen receptor (CAR)–modified T cells have antitumor activity in B cell malignancies, but factors that affect toxicity and efficacy have been difficult to define because of differences in lymphodepletion and heterogeneity of CAR-T cells administered to individual patients. We conducted a clinical trial in which CD19 CAR-T cells were manufactured from defined T cell subsets and administered in a 1:1 CD4+/CD8+ ratio of CAR-T cells to 32 adults with relapsed and/or refractory B cell non-Hodgkin’s lymphoma after cyclophosphamide (Cy)–based lymphodepletion chemotherapy with or without fludarabine (Flu). Patients who received Cy/Flu lymphodepletion had increased CAR-T cell expansion and persistence, and higher response rates [50% complete remission (CR), 72% overall response rate (ORR)] than patients who received Cy-based lymphodepletion without Flu (8% CR, 50% ORR). The CR rate in patients treated with Cy/Flu at the maximally tolerated dose was 64% (82% ORR; n = 11). Cy/Flu minimized the effects of an immune response to the murine single-chain variable fragment component of the CAR, which limited CAR-T cell expansion and clinical efficacy in patients who received Cy-based lymphodepletion without Flu. Severe cytokine release syndrome (sCRS) and grade ≥3 neurotoxicity were observed in 13 and 28% of all patients, respectively. Serum biomarkers, one day after CAR-T cell infusion, correlated with subsequent sCRS and neurotoxicity. Immunotherapy with CD19 CAR-T cells in a defined CD4+/CD8+ ratio allowed identification of correlative factors for CAR-T cell expansion, persistence, and toxicity, and facilitated optimization of lymphodepletion that improved disease response and overall and progression-free survival.

INTRODUCTION

Lymphodepletion chemotherapy, followed by adoptive transfer of unselected autologous T cells that are genetically modified to express a chimeric antigen receptor (CAR) specific for CD19 (CD19 CAR-T cells), has produced a high rate of complete responses (CRs) in refractory B cell acute lymphoblastic leukemia (B-ALL) (15); however, results of therapy in refractory non-Hodgkin’s lymphoma (NHL) have been less impressive (68). Human CD4+ and CD8+ T cells are composed of distinct subsets that differ in their capacities to proliferate, persist in vivo, and mediate antitumor effects after in vitro expansion and adoptive transfer (913). In preclinical studies, we demonstrated that human CD19 CAR-T cells that were manufactured from purified CD4+ or CD8+ central memory (TCM) or naïve (TN) T cells were more potent in the elimination of CD19+ tumors from immunodeficient mice compared to CD19 CAR-T cells that were manufactured from effector memory (TEM) cells (13). We also observed synergistic enhancement in antitumor activity by administering a defined ratio of CD19 CAR-T cells derived from CD8+ and CD4+ T cell subsets, as compared to infusion of CAR-T cells derived from either the subset alone or the unselected T cells without regard for subset composition (13). Differences in the T cell subset composition of CAR-T cells prepared from unselected T cells and administered to patients with NHL in previous studies could, in part, have contributed to differences in efficacy in these studies (68, 14). Moreover, heterogeneity in the subset composition of infused CAR-T cells has made it challenging in these earlier trials to discern factors that correlate with expansion and persistence of CD19 CAR-T cells, quality and durability of antitumor responses, and the toxicities of CAR-T cell therapy. We hypothesized that selecting defined subsets of T cells for genetic modification and their formulation in a defined CD4+/CD8+ ratio would provide a more uniform CAR-T cell product for clinical applications, result in reproducible in vivo activity, and facilitate identification of factors that correlate with efficacy or toxicity.

Lymphopenia and the impaired proliferative capacity of T cells from patients with B cell malignancies present challenges to CAR-T cell manufacturing. In some clinical trials, the proliferation of autologous T cells in response to a test in vitro stimulation with anti-CD3/CD28 beads has been used to predict the success of manufacturing CD19 CAR-T cells and determine patient eligibility for enrollment (1517). Using this strategy, 24% of B-ALL patients were excluded from participation in a pediatric clinical trial, and the fraction of NHL patients who would be excluded was even higher, ranging between 75 and 100%, depending on the status of the patient’s disease at the time the T cells were collected (15). We developed a strategy in which CD4+ and CD8+ T cell subsets are separately selected from apheresis products, transduced with the CD19 CAR, and then stimulated in vitro with an irradiated CD19+ Epstein-Barr virus (EBV) lymphoblastoid cell line (LCL) to selectively expand transduced T cells capable of proliferating. The CAR-T cell product was then formulated in a defined CD4+/CD8+ CAR-T cell ratio and administered in a dose escalation/de-escalation format after lymphodepletion chemotherapy. We demonstrate that this approach for manufacturing a CAR-T cell product of defined composition was feasible in heavily pretreated patients with relapsed and refractory NHL, without excluding any patient on the basis of lymphopenia or the results of in vitro proliferation assays. The defined composition approach also allowed identification of correlative factors for CAR-T cell expansion and persistence in vivo, toxicity, and clinical response, and enabled optimization of a lymphodepletion regimen that improved disease response and overall and progression-free survival.

RESULTS

Patient characteristics

Thirty-seven patients with relapsed or refractory CD19+ NHL with a median age of 57 years (range, 22 to 70 years) were enrolled in the study and underwent leukapheresis for CD19 CAR-T cell manufacturing. Three of the 37 patients were in CR at enrollment and were ineligible to receive lymphodepletion chemotherapy and CAR-T cells, and 2 patients had rapidly progressive disease and chose to withdraw from the study to receive hospice care before starting treatment. The 32 patients who proceeded to lymphodepletion chemotherapy and CD19 CAR-T cell infusion (Table 1) had previously received a median of 5 treatment regimens (range, 2 to 11) for de novo large B cell lymphoma (LBCL; n = 11), LBCL that had transformed from indolent disease (TFLBCL; n = 11), mantle cell lymphoma (MCL; n = 4), or follicular lymphoma (FL; n = 6). Sixteen patients had relapsed after autologous (n = 14) or allogeneic (n = 4) hematopoietic stem cell transplantation (HCT). Before lymphodepletion and CAR-T cell therapy, all patients had measurable disease (≥2 cm) in lymph nodes or other extramedullary sites. The median tumor bulk estimated by the sum of the cross-sectional areas of six index lesions was 2933 mm2 (range, 124 to 17,907 mm2; n = 30; lesion size could not be calculated for the other two patients). Nine patients had bone marrow disease, and one patient had lymphoma detected by flow cytometry in the cerebrospinal fluid that was refractory to intrathecal chemotherapy.

Table 1. Patient characteristics, toxicity, and response.
View this table:

Heterogeneity of T cell subsets in NHL patients

An objective of our study was to determine the feasibility of manufacturing a consistent CD19 CAR-T cell product from NHL patients by selecting defined CD8+ and CD4+ T cell subsets from leukapheresis products, transducing each subset separately, and formulating a CAR-T cell product composed of a 1:1 ratio of CD8+ and CD4+ T cells (fig. S1). Before leukapheresis, we evaluated the proportions of CD4+ and CD8+ naïve and memory T cell subsets in the blood of each patient (Fig. 1, A to D). NHL patients had lower absolute numbers of CD4+ and CD8+ T cells compared to a cohort of healthy individuals (Fig. 1B) and a highly variable ratio of CD4+/CD8+ T cells (median, 1.47; range, 0.1 to 13.7) (table S1). At screening before leukapheresis, the absolute CD4+ and CD8+ T cell counts in the blood were <250/μl in 13 of 32 patients (41%) and in 17 of 32 patients (53%), respectively. The fractions of naïve and memory subsets in the CD4+ and CD8+ compartments differed between NHL patients and healthy donors (Fig. 1, C and D), and there was wide variation in the percentages of each subset in individual patients. These data suggest that, particularly in NHL patients who have higher fractions of terminally differentiated TEM/EMRA and/or a very high or low ratio of CD4+/CD8+ T cells in the blood, the ability to manufacture a CD19 CAR-T cell product of consistent potency may be enhanced by selection of defined T cell subsets from the leukapheresis product for CAR modification and formulation in a defined CD4+/CD8+ CAR-T cell ratio.

Fig. 1. Heterogeneity in distribution of TN, TCM, and TEM/EMRA cells within CD4+ and CD8+ T cell subsets in normal donors and patients with NHL.

(A) Representative fluorescence-activated cell sorting plots showing the proportion of TN (CD45RA+/CD62L+), TCM (CD45RA/CD62L+), and TEM/EMRA (CD62L) subsets in the CD3+/CD4+ and CD3+/CD8+ T cell populations in the blood of an NHL patient. (B) Absolute CD4+ and CD8+ T cell counts in blood from healthy individuals (n = 10) and NHL patients (n = 30). (C) Percentages of TN, TCM, and TEM/EMRA cells in the CD3+/CD4+ T cell population. (D) Percentages of TN, TCM and TEM/EMRA cells in the CD3+/CD8+ T cell population. Comparisons of continuous variables between two categories were made using the Wilcoxon rank-sum test.

CD19 CAR-T cell manufacturing

Each patient underwent a leukapheresis procedure without any serious adverse events. CD4+ and CD8+ T cells were enriched by CliniMACS selection from separate aliquots of each patient’s leukapheresis product for subsequent CAR-T cell manufacturing. Using immunomagnetic cell selection, highly enriched CD4+ T cells were isolated from leukapheresis products for CAR manufacturing from all patients. We used a two-step immunomagnetic selection procedure to enrich CD8+ TCM cells for CAR-T cell manufacturing from patients with a CD8+ TCM cell count of ≥20/μl in the blood on a screening assay (n = 20; CD8+ TCM cells were isolated from one additional patient with 13 CD8+ TCM cells/μl). From the remaining patients with CD8+ TCM counts of <20/μl, severe lymphopenia or high circulating lymphoma burden, we positively selected bulk CD8+ T cells (n = 10) or omitted the CD62L selection after initial depletion of CD4+, CD14+, and CD45RA+ cells (n = 1) to obtain an enriched CD8+ T cell population for transduction with the CAR.

The selected CD4+ and CD8+ T cells were transduced with a lentiviral vector that encoded the CD19 CAR and a truncated cell surface human epidermal growth factor receptor (EGFRt), which enabled identification of transduced cells by flow cytometry using the anti-EGFR monoclonal antibody, cetuximab. Transduced EGFRt+/CD4+ and EGFRt+/CD8+ T cells were stimulated once with an irradiated allogeneic CD19+ LCL line during culture (n = 31), resulting in enrichment of the percentage of EGFRt+ cells within CD3+/CD4+ T cells from 34.3 ± 2.6% (mean ± SE) to 78.8 ± 2.1% and within CD3+/CD8+ T cells from 36.3 ± 3.1% to 81.3 ± 2.2% at the time of product release. LCL stimulation also markedly increased the number of CAR-T cells at the time of harvest for product formulation and release. The median fold expansions of CD3+/CD4+ and CD3+/CD8+ T cells between anti-CD3/anti-CD28 bead stimulation and LCL stimulation were 10.4 (range, 2.5 to 35.9; n = 27) and 7.5 (range, 1.1 to 59.5; n = 28). Between LCL stimulation and CAR-T cell product release, the median fold expansions of CD3+/CD4+/EGFRt+ and CD3+/CD8+/EGFRt+ CAR-T cells were 230.5 (range, 107.7 to 698.6; n = 28) and 198.9 (range, 29.3 to 458.6; n = 29). This cell manufacturing strategy allowed a CAR-T cell product to be prepared for all patients; however, 2 of the 32 patients did not receive their target CAR-T cell dose. These two patients were assigned to the top dose level of 2 × 107 CAR-T cells/kg and received CAR-T cell doses of 8.8 × 106/kg and 7.0 × 106/kg. With the exception of one additional patient who received 38.8 and 100% of the assigned CD4+ and CD8+ CAR-T cell target doses, respectively, all other patients received a cell product with a CD4/CD8 CAR-T cell ratio of 1:1 in the dose specified for the cohort to which they were assigned. The infused CAR-T cells were predominantly CD45RA/CD62L+ and CD45RA/CD62L (fig. S2).

Decreased antitransgene product immune responses and enhanced CAR-T cell expansion in patients who received cyclophosphamide/fludarabine lymphodepletion

The study evaluated three CAR-T cell dose levels administered 36 to 96 hours after lymphodepleting chemotherapy. Twelve patients received chemotherapy with cyclophosphamide (Cy) or Cy and etoposide (Cy/E), followed by infusion of CAR-T cells with an overall response rate (ORR) of 50% [one CR and five partial response (PR)] (Table 1). In these patients, the peak numbers of CD4+ and CD8+ EGFRt+ CAR-T cells in the blood after adoptive transfer correlated with the fraction of CD19+ bone marrow cells before lymphodepleting chemotherapy [CD4+/EGFRt+, Spearman correlation (ρ) = 0.52 and P = 0.082; CD8+/EGFRt+, ρ = 0.69 and P = 0.014] (Fig. 2, A and B). B cell aplasia, defined as <0.01% CD19+ B cells of total white cells in the blood, developed after CAR-T cell infusion in 11 of these 12 patients; however, we observed a loss of CAR-T cells and recovery of B cells in the blood within 100 days in 9 of 10 patients who consented to long-term monitoring. Eight of these patients developed progressive disease, suggesting that longer CAR-T cell persistence may be critical for durable antitumor efficacy. Five of the patients in this group with persistent or progressive disease received a second CAR-T cell infusion at the same dose (n = 1) or a 10-fold higher CAR-T cell dose (n = 4). In all five patients, there was no observable CAR-T cell expansion or persistence immediately after infusion (Fig. 2C), or measurable antitumor activity. In five of five patients, cytotoxic T cell responses that were specific for autologous T cells expressing the CAR transgene were detected in the blood after CAR-T cell therapy, and in four patients, the peptide epitopes that were recognized were mapped to sequences in the murine single-chain variable fragment (scFv) (fig. S3 and table S2).

Fig. 2. Increased CD19 CAR-T cell expansion and persistence after Cy/Flu lymphodepletion.

(A and B) Peak CD4+/EGFRt+ (A) and CD8+/EGFRt+ (B) CAR-T cell numbers after the first CAR-T cell infusion in relation to the percentage of CD19+ cells (normal and malignant CD19+ B cells) in the bone marrow before lymphodepletion chemotherapy for each patient. (C) CAR-T cell persistence in the blood as integrated transgene copies per microgram of DNA in two patients who received a cycle of Cy or Cy/E lymphodepletion chemotherapy and CAR-T cell infusion followed by a second cycle of Cy lymphodepletion chemotherapy and CAR-T cell infusion. Integrated transgene copies were determined by quantitative real-time fluorescence polymerase chain reaction (QPCR) for distinct sequences located in the Woodchuck hepatitis virus posttranscriptional regulatory element (WPRE) or Flap/EF1α regions of the lentivirus. (D and E) Mean ± SEM of CD4+/EGFRt+ (D) and CD8+/EGFRt+ (E) CAR-T cell counts in the blood on the indicated days after CAR-T cell infusion for patients treated with 2 × 107 EGFRt+ cells/kg after either Cy or Cy/E lymphodepletion (No Flu) or Cy/Flu lymphodepletion (Cy/Flu). (F) CAR-T cell persistence in blood of patients who received Cy or Cy/E lymphodepletion (No Flu, black; n = 9) compared to Cy/Flu lymphodepletion (red; n = 18) is shown as FlapEF1α integrated transgene copies per microgram of DNA. CAR-T cell persistence data are truncated at the time of HCT for patients who underwent autologous or allogeneic HCT after CAR-T cell infusion.

We then administered Cy and fludarabine (Flu) for lymphodepletion before CAR-T cell infusion in 20 subsequent patients to determine whether intensified immunosuppression would enhance CAR-T cell expansion and persistence and prevent or delay immune responses to the CAR. Compared to the patients without Flu in the lymphodepletion regimen, patients who received Cy/Flu had higher serum concentrations of interleukin-7 (IL-7) (P = 0.014) and IL-15 (P < 0.001) on the day of CAR-T cell infusion, markedly greater expansion of CD4+ and CD8+ CAR-T cells in the blood in the first 10 days after infusion, and higher numbers of CAR-T cells in the blood 1 and 3 months later (Fig. 2, A and B, D to F). With the exception of patients who underwent autologous or allogeneic HCT after CAR-T cell infusion, in whom CAR-T cells were not detectable after HCT, 16 of the 18 patients who received Cy/Flu lymphodepletion and survived more than 28 days after CAR-T cell infusion had detectable CAR-T cells in the blood by QPCR (>10 copies/μg of DNA) at the last follow-up (range, 34 to 349 days) (Fig. 2F). We identified a relationship between CAR-T cell dose and in vivo expansion in Cy/Flu-lymphodepleted patients (Fig. 3, A and B) that was not apparent in those who received Cy or Cy/E. Furthermore, unlike patients who initially received Cy or Cy/E lymphodepletion, three of four patients who received Cy/Flu before their first CAR-T cell infusion and had a second infusion to treat persistent disease had evidence of further tumor regression after the second infusion (fig. S4); three patients exhibited CAR-T cell expansion after the second infusion without developing a detectable T cell immune response to the transgene product (Fig. 3C). Together, these data demonstrate that Cy/Flu lymphodepletion enhances early proliferation and persistence of both CD4+ and CD8+ CAR-T cells with kinetics that are affected by the dose of CAR-T cells infused and enables repetitive dosing of CAR-T cells.

Fig. 3. Improved clinical responses to CD19 CAR-T cell immunotherapy after Cy/Flu lymphodepletion.

(A and B) Mean ± SEM of CD4+/EGFRt+ (A) and CD8+/EGFRt+ (B) CAR-T cell counts in the blood on the indicated days after CAR-T cell infusion for patients treated with Cy/Flu lymphodepletion and either 2 × 105 or 2 × 106 EGFRt+ cells/kg. (C) CAR-T cell persistence in the blood as integrated transgene copies per microgram of DNA in two patients who received a cycle of Cy/Flu lymphodepletion chemotherapy and CAR-T cell infusion followed by a second cycle of Cy/Flu lymphodepletion chemotherapy and CAR-T cell infusion. (D) Probability curves showing the likelihood of response (CR/PR) according to the peak CD4+/EGFRt+ and CD8+/EGFRt+ cell counts in blood, the AUC0–28, and Cmax. P values were reported from the Wilcoxon two-sample test. (E and F) OS and PFS for patients who received Cy/Flu compared to Cy or Cy/E (No Flu) lymphodepletion followed by infusion of CD19 CAR-T cells at ≤2 × 106 EGFRt+ cells/kg. The median OS follow-up times for No Flu and Cy/Flu are 25 and 6.3 months, respectively. The median PFS follow-up for Cy/Flu is 5.8 months. The median PFS for No Flu is 1.5 months. The numbers of patients at risk at each time point are indicated. Log-rank tests were used to compare between-group differences in survival curves. HR, hazard ratio; CI, confidence interval.

Improved therapeutic efficacy of CAR-T cells in patients who received Cy/Flu lymphodepletion

In the entire cohort of 32 patients, higher peak expansion and longer duration of CAR-T cell persistence in the blood were associated with a greater probability of tumor regression (Fig. 3D). Consistent with the increase in CAR-T cell expansion and persistence observed with Cy/Flu lymphodepletion, we found that addition of Flu in the lymphodepletion regimen was associated with improvement in the depth of response (ordinal logistic regression, CR versus PR versus NR; P = 0.02) (Table 1). The importance of CAR-T cell persistence for ongoing tumor eradication was highlighted by the observation that in six patients, the reduction in tumor burden at the initial restaging performed 1 month after CAR-T cell infusion was not the maximal response, and further tumor regression was observed on subsequent restaging studies (fig. S5). Cy/Flu lymphodepletion followed by infusion of 2 × 106 CAR-T cells/kg was selected as the preferred regimen for NHL patients because of the lower toxicity observed after the 2 × 106/kg cell dose compared to the 2× 107/kg cell dose, and antitumor activity in multiple histologies (Table 1 and table S3). Although follow-up is short, patients who received CAR-T cells at ≤2 × 106/kg after Cy/Flu lymphodepletion had better overall survival (OS; P = 0.17) and progression-free survival (PFS; P = 0.008) compared to those who received Cy or Cy/E lymphodepletion (Fig. 3, E and F). Only one of the nine patients who achieved CR after Cy/Flu lymphodepletion and infusion of CAR-T cells has relapsed (follow-up, 2.3 to 11.2 months).

Toxicity of CD19 CAR-T cells

Patients were evaluated for immediate and delayed toxicity of lymphodepleting chemotherapy and CAR-T cell infusion. Serious acute toxicity in the first 2 hours after CAR-T cell infusion was not observed at any CAR-T cell dose. A majority of patients developed the toxicities expected with cytotoxic chemotherapy, including bone marrow suppression, alopecia, mild mucositis, and neutropenic fever. Depletion of normal CD19+ B cells, consistent with the in vivo presence of functional CD19 CAR-T cells, was noted in 29 of 30 patients who survived until day 28.

Cytokine release syndrome (CRS) and neurotoxicity are serious toxicities that have been observed in a subset of patients with B cell malignancies who are treated with CD19 CAR-T cells, and the factors that predict the patients at greatest risk for these toxicities have been difficult to define (2, 48). We evaluated whether administering a defined-composition CAR-T cell product might facilitate identifying patients who were more likely to develop toxicity. Overall, 20 of 32 patients developed increased concentrations of serum cytokines, fever, and/or hypotension consistent with CRS. Severe CRS (sCRS), which requires management in the intensive care unit (ICU) and treatment with tocilizumab (n = 3) and/or corticosteroids (n = 4), developed in only four (12.5%) patients, all of whom had received Cy/Flu conditioning (Table 1 and table S3). Severe neurotoxicity [National Cancer Institute Common Toxicity Criteria for Adverse Events (NCI CTCAE) version 4.03 grade ≥3] was observed in 9 of 32 patients (28%) and was also more frequently observed in patients who received Cy/Flu lymphodepletion. Neurotoxicity presented as reversible encephalopathy alone (n = 5) or with tremor (n = 1) or speech disturbance (n = 1). Choreoathetosis and fatal intracranial hemorrhage were observed in one patient each. T cell dose was related to the development of sCRS and neurotoxicity, with three of six patients (50%) treated at 2 × 107 CAR-T cells/kg after Cy/Flu lymphodepletion developing sCRS and four of six (67%) patients developing grade ≥3 neurotoxicity (Table 1). Two patients who were treated with 2 × 107 CAR-T cells/kg died: one patient developed a pontine hemorrhage, and one patient had a fatal gastrointestinal (GI) hemorrhage associated with a known GI invasive lymphomatous mass. Thus, the CAR-T cell dose of 2 × 107/kg after Cy/Flu lymphodepletion was considered excessively toxic for NHL patients, and subsequent patients were treated at a lower dose level.

Biomarkers and the risk of CRS and neurotoxicity

CRS is initiated by activation and proliferation of CAR-T cells after recognition of CD19+ target cells (4). The peak number of CD4+/EGFRt+ and CD8+/EGFRt+ CAR-T cells in the blood and the area under the curve of CAR-T cell numbers (transgene copies per microgram of DNA) between days 0 and 28 (AUC0–28) were associated with the occurrence of sCRS (Fig. 4A). Peak serum concentrations of IL-6, interferon-γ (IFN-γ), ferritin, and C-reactive protein (CRP) after CAR-T cell infusion correlated with the occurrence and severity of sCRS (Fig. 4B). The highest IL-6 and IFN-γ concentrations were observed in patients who received Cy/Flu lymphodepletion followed by CAR-T cells infused at 2 × 107/kg (Fig. 4C), consistent with the higher incidence of toxicity observed in these patients (Table 1). Patients who developed grade ≥3 neurotoxicity had higher AUC0–28, more peak CD4+/EGFRt+ and peak CD8+/EGFRt+ cells in the blood, higher peak serum IL-6, IFN-γ, IL-15, IL-2, IL-18, TIM-3, ferritin, and CRP concentrations, and lower serum transforming growth factor–β (TGF-β) compared to those without neurotoxicity (Fig. 4, D and E). In multivariate analyses, peak numbers of CD4+/EGFRt+ and CD8+/EGFRt+ cells, serum ferritin, and IL-6 had the strongest associations with severe neurotoxicity (table S4).

Fig. 4. Factors correlating with toxicity after CD19 CAR-T cell therapy.

(A) Peak counts (median and interquartile range) of CD4+/EGFRt+ and CD8+/EGFRt+ cells in blood after CAR-T cell infusion and the AUC0–28 in patients with sCRS (requiring ICU) and without sCRS. (B) Peak concentrations (median and interquartile range) of serum IL-6, IFN-γ, ferritin, and CRP after CAR-T cell infusion in patients with sCRS, with mild CRS (signs and symptoms of CRS, but not requiring ICU admission), and without CRS. Units shown on the y axis are as follows: IL-6, pg/ml; IFN-γ, pg/ml; ferritin, ng/ml; CRP, mg/liter. **P ≤ 0.01; *P < 0.05, Wilcoxon two-sample test. P values are shown in table S6. (C) Peak concentrations of serum IL-6 and IFN-γ after CAR-T cell infusion in patients treated at dose level 1 (DL1), DL2, or DL3. (D) Peak counts (median and interquartile range) of CD4+/EGFRt+ and CD8+/EGFRt+ cells in the blood after CAR-T cell infusion and the AUC0–28 in patients with grade ≥3 neurotoxicity or with grade 0 to 2 neurotoxicity. (E) Peak concentrations (median and interquartile range) of serum IL-6, IFN-γ, IL-15, IL-2, IL-18, TIM-3, ferritin, CRP, and TGF-β after CAR-T cell infusion in patients with grade ≥3 neurotoxicity and with grade 0 to 2 neurotoxicity. Units shown on the y axis are as follows: cytokines (IL-6, IFN-γ, IL-15, IL-2, IL-18, TIM-3, and TGF-β), pg/ml; ferritin, ng/ml; CRP, mg/liter. **P ≤ 0.01; *P < 0.05, Wilcoxon two sample test. P values are shown in table S6.

We evaluated serum biomarkers on the first day after CAR-T cell infusion to determine whether patients at highest risk for subsequent severe toxicity might be identifiable for early intervention. We found higher IL-6, IFN-γ, IL-15, IL-8, and IL-10 concentrations on day 1 after CAR-T cell infusion in patients who subsequently developed sCRS (Fig. 5A). Higher IL-6, IFN-γ, IL-15, IL-8, and IL-10 and lower TGF-β concentrations were found in patients who subsequently developed severe neurotoxicity, as compared to those who did not (Fig. 5B), with the strongest associations in multivariate analyses identified in patients with higher serum IL-6 and IL-15 concentrations (table S4). Furthermore, receiver operating characteristic (ROC) analyses identified serum concentrations of IL-6 >15.2 pg/ml, IL-15 >76.7 pg/ml, and TGF-β <25,532 pg/ml on day 1 as optimal discriminators of the risk of severe neurotoxicity (Fig. 5, C to E). Simultaneous or sequential use of more than one of these cytokine assays resulted in a net sensitivity of 97 to 100% and a net specificity of 95 to 99%, respectively (table S5) (18). These data provide an opportunity for studying the use of serum cytokine concentrations on the first day after CAR-T cell infusion to identify patients at the highest risk of subsequent toxicity who might benefit from early intervention.

Fig. 5. Prediction of subsequent toxicity using serum biomarkers collected within 24 hours of CAR-T cell infusion.

(A) Concentrations (median and interquartile range) of serum IL-8, IFN-γ, IL-15, IL-10, and IL-6 on day 1 after CAR-T cell infusion in patients who did or did not subsequently develop sCRS. **P ≤ 0.01; *P < 0.05, Wilcoxon two-sample test. P values are shown in table S6. (B) Concentrations (median and interquartile range) of serum IL-15, TGF-β, IL-6, IL-10, IL-8, and IFN-γ on day 1 after CAR-T cell infusion in patients who developed grade ≥3 neurotoxicity or grade 0 to 2 neurotoxicity. **P ≤ 0.01; *P < 0.05, Wilcoxon two-sample test. P values are shown in table S6. (C to E) ROC curves demonstrating the interaction between serum IL-6, IL-15, and TGF-β concentrations on day 1 after CAR-T cell infusion and the occurrence of grade ≥3 neurotoxicity. The red point indicates the cut points for each assay at which the following sensitivities and specificities were obtained: IL-6, 15.2 pg/ml (sensitivity, 78%; specificity, 81%); IL-15, 76.7 pg/ml (sensitivity, 88%; specificity, 78%); TGF-β, 25,532 pg/ml (sensitivity, 88%; specificity, 78%). Data are summarized in table S5.

DISCUSSION

CD19 CAR-T cell therapy represents an important therapeutic advance for patients with relapsed/refractory B cell NHL; however, the optimal T cell dose and lymphodepletion regimens to achieve durable efficacy and the prediction and management of toxicities remain to be elucidated. Systematically studying these issues has been challenging because of small numbers of patients and the differences in the phenotypic composition of CAR-T cell products administered to individual patients in previous trials (2, 47). Our data show that formulating a CAR-T cell product with a uniform CD4+/CD8+ CAR-T cell ratio for NHL patients is feasible by selecting and transducing individual T cell subsets, and if administered at the maximum tolerated dose after Cy/Flu conditioning, this approach results in marked tumor regression in a majority of patients with low incidence of serious toxicity. The strategy of selecting CD4+ and CD8+ T cell subsets before transduction was applied to all patients with an intent to treat basis, and no patient was excluded on the basis of the absolute lymphocyte count, tumor burden, or a pre-enrollment test expansion. This strategy was successful in obtaining a CAR-T cell product for 100% of enrolled patients, and 91% of the patients received the prescribed cell dose and formulation of CAR-T cells. This high success rate in preparing CAR-T cells in this heavily pretreated group of patients may reflect removal of inhibitory cell subsets from the leukapheresis product by selecting CD4+ and CD8+ T cells and/or the selective expansion of CAR-transduced T cells capable of proliferating by the brief in vitro expansion with CD19+ EBV-LCL before cell infusion. Additional improvements in the cell manufacturing process might be achieved using cytokines, such as IL-7, IL-21, and/or IL-15, that have been reported to improve the growth of human T cells in vitro and enhance the therapeutic efficacy of adoptively transferred murine T cells in preclinical models (15, 19, 20). We did not observe differences in CAR-T cell expansion and persistence or clinical outcomes between patients who received CAR-T cells manufactured from CD4+ T cells formulated with either CD8+ TCM-derived CAR-T cells or bulk CD8+ T cell–derived CAR-T cells. Additional studies will be required to define the optimal T cell subset(s) and formulation for CD19 CAR-T cell therapy.

An important predictor of the antitumor efficacy of CD19 CAR-T cells in NHL is their ability to expand and persist in vivo after adoptive transfer. Our data show that CAR-T cell expansion was lower in the subset of patients who received Cy or Cy/E lymphodepletion, and persistence was shorter in these patients due to the development of an anti-CAR T cell immune response. Because all CD19 CARs used in published clinical trials incorporate a murine scFv, our data provide one potential mechanism for the loss of CAR-T cells observed in other trials (27, 2123). Intensification of lymphodepletion by addition of Flu to Cy, as in our study, increased the peak of expansion, the AUC0–28, and long-term persistence of infused CAR-T cells and improved the CR rate, OS, and PFS. The specific mechanisms by which Flu increased CAR-T cell accumulation and persistence are likely multifactorial, and our data indicate that they could include increased homeostatic cytokine availability or reduction in the anti-CAR immune response. An alternative possibility is that the addition of Flu to Cy lymphodepletion might improve CAR-T cell proliferation by reducing expression of indoleamine deoxygenase in the tumor microenvironment (24). In a subset of patients, we observed an improvement in clinical response between 1 and 3 months after initial restaging. Although the timing of the peak of CAR-T cell expansion in the blood did not differ between patients whose best tumor response was at the first restaging or later, it is conceivable that Cy/Flu lymphodepletion may enable prolonged persistence of functional CAR-T cells in the tumor and ongoing antitumor activity. It is unknown whether intensification of lymphodepletion using other chemotherapy regimens or radiation would be similarly effective in augmenting CAR-T cell proliferation and persistence.

We observed that treatment with the same lymphodepletion and CD19 CAR-T cell immunotherapy regimen resulted in lower CR rates in NHL and chronic lymphocytic leukemia (CLL) patients compared to B-ALL patients (1, 25). The reasons for the differences are yet to be determined but may include differences in CAR-T cell access to tumor antigen and/or immune suppression in the tumor microenvironment in each disease. A recent report suggested that T cells isolated from patients with B-ALL, NHL, and CLL differ in their capacity to proliferate in vitro to anti-CD3/anti-CD28 stimulation; however, it is not known whether the reduced proliferative capacity of T cells isolated from NHL and CLL patients translates to a difference in the in vivo efficacy of infused CAR-T cell products from NHL and CLL patients, as compared to B-ALL patients (15). The manufacturing strategy that was used in our trial was highly successful in generating a cell product from NHL patients, even those with reduced CD4+ and CD8+ T cell numbers.

CD19 CAR-T cell therapy has been associated with toxicity in published studies, and understanding factors that predict which patients are at highest risk for toxicity is an area of intense interest (2, 46). Here, the use of CAR-T cells of defined composition and the abrogation of the anti-CAR immune response by Cy/Flu lymphodepletion revealed a correlation between cell dose and in vivo CAR-T cell expansion, which should facilitate more predictable cell dosing and better definition of the therapeutic index. Severe toxicity due to CAR-T cells in patients treated with Cy/Flu lymphodepletion was predominantly seen at the 2 × 107/kg CAR-T cell dose. A reduction of the cell dose to 2 × 106/kg reduced toxicity and achieved CR and ORR rates of 64 and 82%, respectively, which compares favorably to the 45% CR rate observed in a recent report (6). Additional studies will be required to determine whether the higher toxicity at a given CAR-T cell dose seen in patients who received Cy/Flu lymphodepletion is due to greater lymphodepletion and in vivo CAR-T cell expansion or whether Flu directly increases toxicity by mechanisms that are not dependent on CAR-T cell expansion. The peak CAR-T cell count in the blood between days 0 and 28 after CAR-T cell infusion (Cmax) and AUC0–28 that was associated with CR were also associated with neurotoxicity and sCRS, suggesting that further reducing the CAR-T cell dose to minimize toxicity might also result in loss of efficacy. Despite delivery of a consistent CAR-T cell product at a defined optimal dose(s), severe toxicity may still occur in a minority of patients due to patient-specific biological variation, such as may result from polymorphisms in cytokine and cytokine receptor genes. In the serum of NHL patients collected 1 day after CAR-T cell infusion, high IL-6, IL-8, IL-10, IL-15, and IFN-γ concentrations were associated with the subsequent occurrence of sCRS and neurotoxicity, and low TGF-β concentration was associated with neurotoxicity. Serum IL-6 and IFN-γ concentrations were also associated with subsequent toxicity in a cohort of B-ALL patients who received the same lymphodepletion regimens and CAR-T cell product in this study (1). Because IL-8, IL-10, IL-15, and TGF-β concentrations were not assessed in our B-ALL cohort, it remains unknown whether these cytokines predict toxicity in B-ALL as well as NHL patients. Our identification of biomarkers on day 1 after CAR-T cell infusion that were associated with the subsequent occurrence of severe toxicity provides an opportunity to test in a future clinical trial whether early intervention strategies for suppressing cytokine release or blocking cytokine receptors in high-risk patients may mitigate severe toxicity.

This study reports a large series of NHL patients treated at a single institution with CD19 CAR-T cells and demonstrates the feasibility of selecting T cell subsets for CAR engineering and formulation of defined therapeutic products from NHL patients for adoptive therapy. The results show that delivery of a consistent CAR-T cell product is feasible in an overwhelming majority of heavily pretreated patients, provides high OR and CR rates, and allows identification of factors that are important for optimizing CAR-T cell proliferation and persistence and for reducing toxicity.

MATERIALS AND METHODS

Study design

We performed a phase 1 open-label trial to evaluate the feasibility and safety of infusing a defined 1:1 ratio of CD4+/CD8+ CD19-specific CAR-T cells in patients with relapsed or refractory CD19+ B cell malignancies. This article reports the data from NHL patients in the study. Because of differences in disease characteristics and the response to CAR-T cells, CLL and B-ALL patients were treated in distinct cohorts and reported separately (1, 25). The study is available at https://clinicaltrials.gov/ct2/show/NCT01865617 and was conducted with approval from the Fred Hutchinson Cancer Research Center institutional review board. Informed consent was obtained from all patients after a discussion of the possible risks and adverse effects of the therapy. In contrast to other studies, we did not exclude any patient on the basis of the absolute lymphocyte count or a previous assessment of T cell function (15).

T cell subset selection and CD19 CAR-T cell manufacturing

Peripheral blood mononuclear cells (PBMCs) were collected from each patient by leukapheresis, and the product was divided into two aliquots for enrichment of CD4+ and CD8+ T cell subsets on a CliniMACS device for CAR-T cell manufacturing (fig. S1) (1, 26). Our preclinical studies suggested that infusion of CD8+ TCM CAR-T cells combined with CD4+ CAR-T cells might provide optimal antitumor efficacy (13); therefore, CD8+ TCM cells were isolated from patients with an absolute CD8+ TCM cell count of ≥20/μl using a two-step sequential depletion of CD4+/CD14+/CD45RA+ cells followed by selection of CD62L+ cells. For patients with an absolute CD8+ TCM cell count of <20/μl, either the CD62L+ selection was omitted or bulk CD8+ T cells were selected. Enriched CD4+ and CD8+ T cell subsets were separately stimulated with anti-CD3/anti-CD28 CTS paramagnetic beads (Dynabeads) and transduced with a lentivirus encoding a CAR comprising an FMC63-derived CD19-specific scFv fused to a modified immunoglobulin G4 hinge spacer, a CD28 transmembrane domain, a 4-1BB costimulatory domain, and a CD3ζ signaling domain (13). A cell surface human EGFRt was also encoded in the lentiviral vector separated from the CAR cassette by a ribosomal skip sequence to allow precise enumeration of transduced CD4+ and CD8+ CAR-T cells by flow cytometry (27, 28). After removal of CD3/CD28 beads, the CD4+ and CD8+ T cells were separately stimulated with a clinically qualified irradiated allogeneic CD19+ LCL line and cultured in medium (RPMI 1640, 10% pooled human serum, 3.5 mM l-glutamine, and 44 μM β-mercaptoethanol) supplemented with IL-2 (50 U/ml). CAR-T cell manufacturing was completed within about 15 days after initial bead stimulation. Quality assessments were performed separately on the cultured CD4+ and CD8+ CAR-T cells, and the two fractions were formulated in a 1:1 ratio of CD3+/CD4+/EGFRt+:CD3+/CD8+/EGFRt+ cells for infusion.

Lymphodepletion chemotherapy and CAR-T cell infusion

To deplete endogenous lymphocytes before adoptive transfer of CAR-T cells, patients received one of four chemotherapy regimens: Cy, 2 to 4 g/m2 iv (intravenously) on day 1; Cy, 2 to 4 g/m2 iv on day 1 and etoposide, 100 to 200 mg/m2 per day iv on days 1 to 3 (Cy/E); and Cy, 60 mg/kg iv on day 1 and fludarabine 25 mg/m2 per day iv on either days 2 to 4 or days 2 to 6 (Cy/Flu). Between 36 and 96 hours after completion of chemotherapy, CAR-T cells were infused (intravenously) at or as close as possible to one of three cell dose levels (2 × 105 EGFRt+ cells/kg, 2 × 106 EGFRt+ cells/kg, and 2 × 107 EGFRt+ cells/kg) (Table 1).

Clinical response assessment

Patients underwent whole-body imaging with diagnostic-quality computed tomography and concurrent positron emission tomography before and about 4 weeks after CAR-T cell therapy. Best responses to lymphodepletion chemotherapy and a single CAR-T cell infusion are reported according to the Lugano criteria (29). Marrow biopsies were obtained before lymphodepletion and about 4 weeks after CAR-T cell infusion from patients with marrow disease on initial staging. Toxicity was graded using the NCI CTCAE version 4.03, with the exception of CRS, which was defined as sCRS when ICU management was required.

CAR-T cell enumeration

Blood samples were obtained from patients before and at intervals after CAR-T cell infusion, and flow cytometry was performed to identify CD4+ and CD8+ CAR-T cells as viable CD45+/CD3+/CD4+/EGFRt+ or CD45+/CD3+/CD8+/EGFRt+ events, respectively. The absolute CAR-T cell count was determined by multiplying the percentage of CAR-T cells identified by flow cytometry in a lymphocyte forward scatter–side scatter gate by the absolute lymphocyte count established by a complete blood count performed on the same day. The area under the curve of CAR-T cell numbers (transgene copies per microgram of DNA) between days 0 and 28 (AUC0–28) was calculated using a trapezoidal rule computational algorithm (30). Loss of CAR-T cell persistence was defined as <10 copies/μg of DNA evaluated by QPCR to detect integrated transgene sequence.

Cytokine assay

Serum concentrations of IFN-γ, IL-6, IL-8, IL-10, IL-15, and TGF-β were evaluated by Luminex assay, according to the manufacturer’s instructions.

Anti-CAR immune response

We evaluated CD8+ T cell immune responses to the CAR transgene using a modification of an assay we previously described (31). Cryopreserved PBMCs collected from patients before lymphodepletion chemotherapy and about 4 weeks after CAR-T cell infusion were stimulated twice at 7-day intervals with autologous irradiated CAR-T cells and IL-2. The preinfusion and postinfusion PBMC cultures were assayed for lysis of autologous CAR-T cells and autologous nontransduced T cells in a chromium release assay. An immune response against the CAR transgene was defined as the presence of specific lysis of autologous CAR-T cells by postinfusion PBMC cultures but not of autologous nontransduced T cells or autologous CAR-T cells by preinfusion PBMC cultures. A T cell line that exhibited specific lysis of autologous CAR-T cells was stimulated with pools of overlapping peptides from the CAR construct, and peptide pools that induced IFN-γ secretion higher than that induced by T cells alone in an enzyme-linked immunospot assay were identified. Predictions of major histocompatibility complex class I binding by peptides shared between the candidate pools were made using the Immune Epitope Database and Analysis Resource Consensus tool (32), which combines predictions from NetMHC (3.4) (33, 34), SMM (35), and Comblib (36).

Statistical analysis

Comparisons of continuous variables between two categories were made using the Wilcoxon rank-sum test. Relationships between continuous variables were analyzed using Spearman correlation (ρ). Univariate (with Firth correction) and stepwise multivariate logistic regression were performed to assess predictors for the occurrence of severe neurotoxicity, where log10 values were used to transform data as appropriate (for cytokine concentrations and CD4+/EGFRt+ and CD8+/EGFRt+ cell counts), with 0.01 substituting for values of 0. All P values reported were two-sided, and no adjustments were made for multiple comparisons. ROC curves were constructed using the empirical method, and the optimal cut point was established on the basis of the Youden index (37). For time-to-event analyses, the Kaplan-Meier method was used to estimate survival distributions, and the reverse Kaplan-Meier method was used to estimate median follow-up time (38); log-rank tests were used to compare between-group differences in survival curves.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/8/355/355ra116/DC1

Fig. S1. Plan of CD19 CAR-T cell manufacturing.

Fig. S2. CAR-T cell product characterization.

Fig. S3. Anti-CAR transgene product immune response.

Fig. S4. Tumor regression after a second CAR-T cell infusion in patients who received Cy/Flu lymphodepletion.

Fig. S5. Delayed normalization of imaging after CD19 CAR-T cell infusion.

Table S1. Variable CD4+/CD8+ T cell ratio in blood of NHL patients.

Table S2. Summary of anti-CAR transgene product immune responses.

Table S3. CRS and neurotoxicity.

Table S4. Multivariate analyses.

Table S5. Simultaneous and sequential testing of cytokines to predict toxicity.

Table S6. P values for Figs. 4 (B and E) and 5 (A and B).

Reference (39)

REFERENCES AND NOTES

Acknowledgments: We acknowledge the Fred Hutchinson Cancer Research Center (FHCRC) Cell Processing Facility and the Seattle Cancer Care Alliance (SCCA) Cell Therapy Laboratory, and the staff of the Program in Immunology and SCCA Immunotherapy Clinic. C.J.T. is a Damon Runyon clinical investigator. Funding: Funding for this study was provided by NCI (National Cancer Institute) R01 CA136551; NIDDK (National Institute of Diabetes and Digestive and Kidney Diseases) P30 DK56465; NCI P30 CA15704; Life Science Discovery Fund; Bezos Family Foundation; and Juno Therapeutics. Author contributions: C.J.T. and D.G.M. designed the trial and experiments, analyzed the data, and wrote the paper. S.R.R. designed the trial and experiments and wrote the paper. L.-A.H., C.B., M.H., B.P., E.R., and R.H. performed experiments and analyzed the data. L.-A.H., S.C., L.S., X.C., B.W., and S.H. designed experiments and analyzed the data. C.C. collected research data. D.L. analyzed the statistical data. Competing interests: C.J.T. and C.B. received research funding from Juno Therapeutics and hold patents. D.G.M. receives research funding from Juno Therapeutics. S.R.R. receives research funding from and is a cofounder of Juno Therapeutics and holds patents. M.H. holds a patent. D.L. is an employee of and has equity in Juno Therapeutics. FHCRC receives research funding from Juno Therapeutics.
View Abstract

Navigate This Article