Safety and feasibility of anti-CD19 CAR T cells with fully human binding domains in patients with B-cell lymphoma

Abstract

Anti-CD19 chimeric antigen receptor (CAR)-expressing T cells are an effective treatment for B-cell lymphoma, but often cause neurologic toxicity. We treated 20 patients with B-cell lymphoma on a phase I, first-in-human clinical trial of T cells expressing the new anti-CD19 CAR Hu19-CD828Z (NCT02659943). The primary objective was to assess safety and feasibility of Hu19-CD828Z T-cell therapy. Secondary objectives included assessments of blood levels of CAR T cells, anti-lymphoma activity, second infusions and immunogenicity. All objectives were met. Fifty-five percent of patients who received Hu19-CD828Z T cells obtained complete remission. Hu19-CD828Z T cells had clinical anti-lymphoma activity similar to that of T cells expressing FMC63-28Z, an anti-CD19 CAR tested previously by our group, which contains murine binding domains and is used in axicabtagene ciloleucel. However, severe neurologic toxicity occurred in only 5% of patients who received Hu19-CD828Z T cells, whereas 50% of patients who received FMC63-28Z T cells experienced this degree of toxicity (P = 0.0017). T cells expressing Hu19-CD828Z released lower levels of cytokines than T cells expressing FMC63-28Z. Lower levels of cytokines were detected in blood from patients who received Hu19-CD828Z T cells than in blood from those who received FMC63-28Z T cells, which could explain the lower level of neurologic toxicity associated with Hu19-CD828Z. Levels of cytokines released by CAR-expressing T cells particularly depended on the hinge and transmembrane domains included in the CAR design.

Conflict of interest statement

Competing Interests

This work was supported by intramural funding of the Center for Cancer Research, National Cancer Institute, NIH. In addition, the NCI has Cooperative Research and Development Agreements with Kite Pharma, a Gilead Company that support development of anti-CD19 CAR T-cell therapies, and both James N. Kochenderfer and Steven A. Rosenberg are NCI Principle Investigators of these Research Agreements. James Kochenderfer has a patent application for the Hu19-CD828Z CAR and has received royalty payments from Kite, a Gilead Company. Adrian Bot, John Rossi, Allen Xue, and Nathalie Scholler are all employees of Kite, a Gilead Company.

Figures

Extended Data Fig. 1. CONSORT

Consort diagram of the Hu19-CD828Z clinical trial.

Extended Data Fig. 2. Hu19-CD828Z neurologic toxicities

Neurologic toxicity with Hu19-CD828Z. All grade 4, 3, and 2 neurologic adverse events within the first month after CAR T-cell infusion are listed. Grading by National Cancer Institute Common Terminology Criteria for Adverse Events Version 3; all adverse events listed under “Neurologic” are included except syncope. Syncope was not included because it was associated with cytokine-release syndrome and hypotension. The highest grade of each adverse event experienced by each patient is listed. For example, if a patient had both Grade 2 and Grade 3 tremor at different times, tremor is only listed under Grade 3.

Extended Data Fig. 3. FMC63–28Z neurologic toxicities

Neurologic toxicity with FMC63–28Z. All grade 4, 3, and 2 neurologic adverse events within the first month after CAR T-cell infusion are listed. Grading by National Cancer Institute Common Terminology Criteria for Adverse Events Version 3; all adverse events listed under “Neurologic” are included except syncope. Syncope was not included because it was associated with hypotension from cytokine-release syndrome. The highest grade of each adverse event experienced by each patient is listed. For example, if a patient had both Grade 2 and Grade 3 confusion at different times, confusion is only listed under Grade 3.

Extended Data Fig. 4. Peak serum protein levels

Peak immunologic protein levels. For all proteins, all 22 patients on the trial of FMC63–28Z T cells and all 20 patients on the trial of Hu19-CD828Z T cells were compared. Proteins were measured in serum samples by Luminex® assay between day 2 and 14 after CAR T-cell infusion. Statistics were by 2-tailed Mann-Whitney test.

Extended Data Fig. 5. Patient 3 immunologic protein levels

Patient 3 serum immunologic protein levels. Patient 3 was the only patient with Grade 3 or 4 neurologic toxicity on the Hu19-CD828Z trial. Peak serum levels of 9 immunological proteins are shown for Patient 3. Peak levels were determined between day 2 and day 14 after CAR T-cell infusion. These 9 proteins are shown because they were found to be prominently different between the Hu19-CD828Z and FMC63–28Z clinical trials (Figure 3). Proteins were measured by Luminex® assay. MCP-1, monocyte chemotactic protein-1; IL, interleukin; TNF-alpha, tumor necrosis factor-alpha; MIP-1-alpha, macrophage inflammatory protein-1-alpha; IFN-gamma, interferon-gamma. The red bars indicate the median protein levels for all 20 patients that received Hu19-CD828Z CAR T cells.

Extended Data Fig. 6. Serum proteins areas under the curves

Immunologic proteins areas under the curves. For all proteins, all 22 patients on the trial of FMC63–28Z T cells and all 20 patients on the trial of Hu19-CD828Z T cells were compared. Proteins were measured in serum samples by Luminex® assay from days 2 to 14 after CAR T-cell infusion. Area under the curve (AUC) was calculated by trapezoidal method. Statistics were by 2-tailed Mann-Whitney test.

Extended Data Fig. 7. Structural Models of CARs

Structural models of CARs. Top row: schematic representations of Hu19-CD828Z (left) and FMC63–28Z (right) CAR models are shown; scFv in blue; hinge in green; transmembrane domain in yellow; intracellular domain in red. The membrane position during molecular dynamics simulations is shown in grey. Bottom row: conformational flexibility for each corresponding CAR depicted as superimposed carbon-alpha traces for a set of 50 representative conformations observed during a 50 nanosecond molecular dynamics trajectory. The differences in flexibility originate in the very different structure and dynamic behavior of the corresponding hinge regions during the dynamics simulations. Transmembrane and scFv domains are affected by the hinge properties and display very different behaviors as well. A quantitative analysis of the molecular dynamics trajectories reveals that these behaviors affect the scFv mobility (assessed as molecular diffusibility) and the proper formation of a transmembrane dimer evaluated by the helix-helix occluded surface. All models assume a dimeric structure anchored by disulfide bonds. In short, Hu19-CD828Z exhibited less conformational flexibility than FMC63–28Z.

Extended Data Fig. 8. Anti-CAR immune responses measured by ELISPOT

Anti-CAR immune responses measured by ELISPOT. *Positive anti-CAR response was defined as 3x or greater increase in spot number from pretreatment to post-CAR T-cell infusion, and post-treatment spot number must have been 3x or more than the spot number of the media control. #Bin A contained peptides from the signal sequences, scFv linker, and hinge regions. Bin B contained peptides from the scFv light chain. Bin C contained peptides from the scFv heavy chain. Bin D contained peptides from transmembrane and intracellular domains. ^The increase in spots was the number of spots/40,000 total input PBMC at the positive time-point minus the number of spots/40,000 total input PBMC before CAR T-cell treatment.

Extended Data Fig. 9. Blood CAR T-cell levels and anti-CAR immune responses.

Blood CAR T-cell levels and anti-CAR immune responses. Anti-CAR T-cell responses were assessed by ELISPOT analysis of PBMC before CAR T-cell treatment and at time-points within 6 weeks after CAR T-cell infusion as summarized in Extended Data 8. CAR+ cell levels in the blood were assessed by quantitative PCR. The top row shows peak blood CAR+ cell levels with results divided into patients with or without anti-CAR responses by ELISPOT. (a) Hu19-CD828Z (b) FMC63–28Z. The bottom row shows blood CAR+ cell levels 1-month after CAR T-cell infusion with results divided into patients with or without anti-CAR responses detected by ELISPOT: (c) Hu19-CD828Z, (d) FMC63–28Z. No statistically significant differences in blood CAR+ cell levels were found between patients with or without anti-CAR responses. All patients with adequate cell samples for both ELISPOT and qPCR are included. P values by Mann-Whitney test are shown on the plots; significance was defined as P<0.05. Of the 4 comparisons, the FMC63–28Z 1-month comparison was closest to statistical significance with P=0.061. Each symbol represents an individual patient. The number of unique patients analyzed were as follows: Hu19-CD828Z Peak, n=18; FMC63–28Z Peak, n=18; Hu19-CD828Z 1 month, n=18; FMC63–28Z 1 month n=13.

Figure 1.. Hu19-CD828Z CAR T cells have substantial anti-lymphoma activity.

(a) Hu19-CD828Z contained the Hu19 human scFv, CD8α hinge and transmembrane domains, the CD28 cytoplasmic domain, and a CD3ζ domain. FMC63–28Z was used in prior clinical trials. FMC63–28Z had a scFv from a murine antibody, hinge, transmembrane, and cytoplasmic domains from CD28, and a CD3ζ domain. (b) The clinical protocol included a CAR T-cell infusion preceded by conditioning chemotherapy of cyclophosphamide 300mg/m2/day and fludarabine 30mg/m2/day both daily for 3 days. (c) Hu19-CD828Z expression on infusion CAR T cells of Patient 12 was assessed by staining with the anti-CAR antibody and anti-CD3. Staining of pre-culture PBMC from Patient 12 is shown as a control. Similar results were obtained with all 20 patients that received Hu19-CD828Z T cells. (d) Durations of response are shown. Duration of response was from the day of 1st response until one of the following: progressive lymphoma, the patient started a different lymphoma therapy, or latest documented ongoing response; (*) indicates ongoing response at last follow-up. (e) Event-free survival of all 20 patients is shown. (f) Patient 6 obtained an ongoing CR of DLBCL. Yellow areas indicated by the arrows on the PET-CT scan are lymphoma. Lymphoma resolved after Hu19-CD828Z CAR T-cell infusion. The yellow areas in the heart and kidneys are from accumulation of radiotracer, not lymphoma. (g) PET-CT scan of Patient 13 shows areas of follicular lymphoma as red and yellow before CAR T-cell treatment and (h) disappearance of lymphoma after CAR T-cell treatment. Brain, kidneys, and bladder are normally red or yellow because of high metabolism and accumulation of radiotracer.

Figure 2.. CAR T-cell characteristics and persistence.

(a) Percentages of patients receiving either Hu19-CD828Z (Hu19) T cells or FMC63–28Z (FMC63) T cells experiencing different grades of neurologic toxicity; P values by 2-sided Fisher’s exact test. All 20 patients that received Hu19 T cells and all 22 patients that received FMC63 T cells are included. (b-f) Flow cytometry with anti-CAR antibody, anti-CD3, anti-CD4, anti-CD8 and other markers was performed on infusion cells from patients receiving Hu19 or FMC63 T cells. Results for c-f are from CD3+CAR+ cells. (b) %CAR+CD3+. (c) CD4:CD8 ratio. (d) %Central memory (CM, CCR7+CD45RA-negative) plus %naïve (CCR7+CD45RA+). (e) %Effector memory (EM, CCR7-negative, CD45RA-negative) plus %T-effector memory RA (TEMRA, CCR7-negative, CD45RA+). (f) %CD28+; b-f, colored bars represent means; error bars show +/−SEM; comparisons by 2-tailed Mann-Whitney test. For b-f n=20 unique patients for Hu19 and n=21 unique patients for FMC63. (g) CAR+ PBMC were quantified by qPCR at multiple time-points post-infusion. Peak CAR+ cell levels are shown for all 20 Hu19 patients and all 22 FMC63 patients. (h) CAR+ cells were quantified by qPCR 1-month (26–35 days for all of Figure 2) after CAR T-cell infusion. (i) For each patient, CAR+ cell number determined by qPCR 1-month post-infusion divided by the peak CAR+ cell number was the fraction of peak CAR+ cells persisting 1-month post-infusion. For h and i, all 20 Hu19 patients were included, and all 14 FMC63 patients with available 1-month post-infusion PBMC were included. (j) Blood CAR+ T cells were assessed by anti-CAR antibody flow cytometry. Plots are gated on live CD3+ lymphocytes. (k) Flow cytometry as in j was performed for patients with available samples 1-month post-infusion. For j and k (Hu19 n=16 and FMC63 n=14 unique patients) In g-i and k, horizontal bars represent medians, and comparisons were by 2-tailed Mann-Whitney test.

Figure 3.. Cytokine production by infusion CAR T cells and blood cytokine levels.

(a and b) T cells expressing Hu19-CD828Z released lower levels of immunological proteins compared with T cells expressing FMC63–28Z. Samples of the infusion CAR T cells from patients treated with either Hu19-CD828Z (Hu19) T cells or FMC63–28Z (FMC63) T cells were cultured overnight with CD19-K562 cells. Supernatant was collected and tested in Luminex® assays. Background release of each protein after overnight culture of CAR T cells with CD19-negative NGFR-K562 cells was subtracted from the protein release when CAR T cells were cultured with CD19-K562 cells. Cytokine values are normalized for CAR expression level by dividing the cytokine value by the %CAR+ T cells for each patient. Bars show mean+SEM. P<0.0001 is indicated by ***; P<0.001 is indicated by **; P≤0.01 is indicated by *; comparisons with no asterisk above the bars had P>0.05. N=18 unique patients for Hu19 and n=21 unique patients for FMC63. (c-k) Serum was collected at multiple time-points between day 2 and day 14 after CAR T-cell infusion and analyzed for immunologically-important proteins by Luminex® assay; all 20 Hu19 patients and all 22 FMC63 patients were compared. The peak protein levels for each patient are shown with horizontal bars representing the median. (c) Granzyme A, (d) MCP-1, (e) IL-2, (f) TNFα, (g) MIP-1α, (h) IL-6, (i) IL-7, (j) IFNγ, (k) IL-8. For all of Figure 3, statistical comparisons were by 2-tailed Mann-Whitney tests.

Figure 4.. Comparison of CAR designs.

(a) Five CAR plasmids are listed. LSIN, lentiviral vector; MSGV, gamma-retroviral vector. All 5 CARs contained a CD28 costimulatory domain and a CD3ζ domain. LSIN-Hu19-CD828Z had the Hu19 human scFv plus CD8α hinge and transmembrane (TM) domains. LSIN-FMC63-CD828Z had the FMC63 scFv plus CD8α hinge and transmembrane domains. LSIN-Hu19–28Z had the Hu19 scFv plus CD28 hinge and transmembrane domains. MSGV-Hu19-CD828Z had the Hu19 scFv plus CD8α hinge and transmembrane domains; MSGV-Hu19-CD828Z was identical to LSIN-Hu19-CD828Z except for the gene-therapy vectors. MSGV-FMC63–28Z had the murine FMC63 scFv plus CD28 hinge and transmembrane domains. (b) T cells from the same patient were transduced with each of 5 CARs or left untransduced as indicated. Plots are gated on live, CD3+ lymphocytes. These results are representative of results from 6 unique donors. (c) The %CAR+ T cells for each of the 5 CARs is shown. Flow cytometry gating was as in b. LSIN-Hu19-CD828Z was compared to the other 4 CARs. The only consistent difference was between LSIN-Hu19-CD828Z and LSIN-Hu19–28Z (P=0.031). LSIN-Hu19-CD828Z was compared to the other 3 CARs; P values for the comparisons were: LSIN-FMC63-CD828Z, 0.6875; MSGV-Hu19-CD828Z, >0.999; MSGV-FMC63–28Z, 0.5625. (d) The median fluorescence intensities of only the CAR+ T cells are shown. When LSIN-Hu19-CD828Z was compared to the other 4 CARs, the only consistent difference was between LSIN-Hu19-CD828Z and MSGV-FMC63–28Z (P=0.031). LSIN-Hu19-CD828Z was compared to the other 3 CARs; P values for the comparisons were: LSIN-FMC63-CD828Z, 0.063; LSIN-Hu19–28Z, 0.4375; MSGV-Hu19-CD828Z, >0.999. (e) A CD4+ versus CD8+ plot gated on CD3+CAR+ events was used to determine the CD4 to CD8 ratio of CAR+ T cells. LSIN-Hu19-CD828Z was compared to the other 4 CARs; P values for the comparisons were: LSIN-FMC63-CD828Z, 0.438; LSIN-Hu19–28Z, 0.438; MSGV-Hu19-CD828Z, 0.563; MSGV-FMC63–28Z, 0.094. For c-e, comparisons were by 2-tailed Wilcoxon matched-pairs signed-rank test. Graphs c-e show mean +/−SEM. For c-e, n=6 independent experiments with cells from unique donors.

Figure 5.. Functional comparison of CARs.

(a) T cells were transduced with either LSIN-Hu19-CD828Z, LSIN-FMC63-CD828Z, LSIN-Hu19–28Z, MSGV-Hu19-CD828Z, or MSGV-FMC63–28Z and cultured overnight with CD19+ CD19-K562 cells or CD19-negative NGFR-K562 cells. IFNγ was measured in the culture supernatant. (b) The same T cells from a were tested for IFNγ release when cultured with CD19+ NALM6 cells. For a and b, lines connect paired results with T cells from the same donor, and P values for each comparison are shown. (c-g) T cells from the same donors as in a were transduced with either LSIN-Hu19-CD828Z, Hu19–28Z, or MSGV-FMC63–28Z and cultured overnight with CD19-K562 or NGFR-K562. Cytokines were measured in culture supernatants (c) IFNγ, (d) IL-2, (e) TNFα, (f) IL-17A, (g) IL-4. For a-g, CD19-specific cytokine release is cytokine release by CAR T cells cultured with CD19-K562 or NALM6 minus cytokine release by CAR T cells cultured with NGFR-K562. (h) Degranulation was assessed by flow cytometry for CD107a. CD19-specific degranulation was %CD3+CD107+ cells with NALM6 stimulation minus %CD3+CD107+ cells with NGFR-K562 stimulation. For c-i, P values are shown above the brackets connecting results from different CARs. (i) There was no consistent difference in CD4+ to CD8+ ratios of CD3+CD107+ cells in h. For all of Figure 5, the percentage of T cells expressing different CARs was equalized prior to experiments by adding untransduced T cells as needed. For all of Figure 5, comparisons were made by 2-tailed Wilcoxon matched-paired signed-rank tests. Bar graphs show mean +/−SEM; n=6 different donors in all comparisons.