Cytokine Release Syndrome Following CD19 Directed Chimeric Antigen Receptor T-Cell Therapy

Components of a Chimeric Antigen Receptor

As shown in Fig. 35.2 , there are several key components of the CAR construct. These include an extracellular region composed of the single-chain variable fragment (scFv), which is a fusion protein of the light and heavy chains of immunoglobulin (Ig) that recognizes a specific tumor antigen, and a hinge region that links the scFv to the rest of the CAR structure. This is followed by a transmembrane domain that connects to a series of intracellular domains that provide costimulatory signals to activate the CAR T-cell. Most CAR therapies use either CD28 or 4-1BB-based costimulatory domains, which is followed by CD3ζ. This complex of recognition and signaling domains allow the CAR T-cell to recognize a tumor cell in an antigen-dependent fashion and then quickly become activated. The CAR construct is packaged in either a retrovirus or lentivirus for transduction into the T-cell.

Influence of Intracellular Costimulatory Domains on Toxicity

The type of costimulatory domain in the construct appears to be one of the components that most affects the overall risk and onset of CRS. While the activation of T-cells is the common underlying trigger of CRS in these therapies, differences in costimulatory domains between each therapeutic agent affects the incidence, time course, and severity of CRS. For example, the CD28 costimulatory domain induces a brisk but self-limited CAR T-cell expansion, whereas the 4-1BB costimulatory domain promotes slower expansion with longer persistence. The clinical picture may also be affected by the costimulatory domain used, as patients treated with Yescarta, which bears a CD28 costimulatory domain, have fewer grade 3–4 CRS events, but more grade 3–4 neurologic events, when compared to patients receiving 4-1BB containing constructs. However, in two trials of CAR T-cells in patients with LBCL, the incidence of all grades of CRS was 93% with a CD28-containing CAR, compared to 57% with a 4-1BB–containing CAR.

Influence of Extracellular and Transmembrane Domains on Toxicity

Commercially available CD19 CAR-T-cell products owe their success to the antigen specificity of the scFv derived from the murine anti-CD19 antibody clone FMC63. It is important to note, however, that these products differ in their costimulatory domains as well as their hinge and transmembrane domains. Yescarta and Tecartus use a CD28 hinge and transmembrane domain in addition to a CD28 costimulatory domain. Kymriah contains a CD8 hinge and transmembrane domain in addition to a 4-1BB costimulatory domain. Breyanzi uses an IgG4 hinge and transmembrane domain in addition to a 4-1BB costimulatory domain. It is possible that differences in the extracellular and transmembrane domains of the CAR influence toxicities, given that a murine scFv could be immunogenic and that various hinge and transmembrane domains could facilitate different rates and intensities of CAR T-cell activation. Further studies are needed to fully validate this hypothesis. However, in preclinical models, Alabanza et al. investigated the effect of utilizing a human antibody construct as compared to the clinically tested FMC63 murine construct and also determined the effect of utilizing different hinge and transmembrane domains on the expansion of CAR T-cells. They tested a series of four CARs, of which the first two incorporated the human scFv Hu-19 with either a CD28-based hinge and transmembrane domain or a CD8α hinge and transmembrane domain. The second two CARs incorporated the murine scFv FMC63 instead of the human Hu-19. They found that the CARs incorporating the human scFv Hu-19 had a similar specificity in targeting CD19-expressing cells and, interestingly, the amount of interferon (IFN)γ produced was higher than in the CARs that incorporated the murine FMC63 counterpart. Furthermore, CARs with Hu-19 and a CD28 hinge and transmembrane domain induced a similar CD4 T-cell degranulation and moderately higher CD8 T-cell degranulation in response to stimulation with CD19+ NALM6 cells compared to those with a CD8α hinge and transmembrane domain. Following stimulation, CAR T-cells with a CD8α hinge and transmembrane domain also produced less IFNγ and tumor necrosis factor (TNF)α compared to those with a CD28 hinge and transmembrane domain. These results were replicated in the constructs that incorporated the murine FMC63 scFv. This is particularly important, since these cytokines are implicated in the development of CRS and may indicate CARs with a CD8α hinge could have reduced toxicity compared to those with a CD28 hinge. Of note, there was no difference in proliferation between the two hinge and transmembrane domain constructs and no difference in the ability to eliminate tumors in mouse models. Overall, this study indicates it is feasible that future generations of CARs can be developed that both potently kill malignant cells while also having reduced cytokine secretion and therefore reduced risk of inducing CRS.

Clinical Trials With Humanized Chimeric Antigen Receptor Constructs

Based on the previous data and the hypothesis that a human scFv might be less immunogenic than a mouse scFv, a first in human phase I trial utilizing the aforementioned Hu-19 scFv with a CD8α hinge and transmembrane domain (Hu19-CD828Z) was completed in 20 patients with B-cell lymphoma. With the limitations of a small number of patients treated on this trial and cross-trial comparison, it appears that the Hu19-CD828Z CAR has lower rates of severe ICANS (5% vs. 50%) compared to its FMC63-based counterpart. While severe CRS was rare in both groups, this trial highlights the importance of a better understanding of the CAR construct in trying to minimize toxicity.

A fully humanized CAR would also have an added advantage in avoiding an immunogenic response to the murine scFv, which would allow the study of these CARs in patients previously exposed to CAR-T-cell therapy. A pilot clinical trial of a humanized CD19 CAR-T-cell product in children and young adults with R/R ALL with (n = 33) or without (n = 41) prior exposure to CAR-T-cell therapy was completed by Myers et al. All grade CRS developed in 84% of the patients and grade 4 CRS developed in 6.8% of the patients. The safety profile is impressive in light of the complexity of the patient population, especially in patients pretreated with CAR-T-cell products as compared to the Eliana trial of tisagenlecleucel, which reported an all grade CRS incidence of 77% and a grade 4 CRS incidence of 25%. A fully humanized CD19 targeted CAR-T-cell construct was tested in a similar population by Cao et al. where 14 of the 18 patients had previously received a CD19 targeted CAR T-cell with a murine construct. Seventeen of the 18 patients developed CRS, of which three patients had grade 3 or higher CRS. Larger studies are needed to evaluate the safety of these humanized or fully human CARs.

Comparison of Cytokine Release Syndrome in Clinical Trials and Standard-of-Care Studies

Given that CAR-T-cell products were being developed at major research centers, a major concern following their commercialization was whether the safety data was only representative of a highly selected patient population. In addition, with the manufacturing process taking 3 to 4 weeks, there was a fear that only patients with relatively better controlled disease were selected for these trials. However, several reports of real-world data with both axicabtagene ciloleucel (Yescarta) and tisagenlecleucel (Kymriah) have been reported and are reassuring, as they appear to be comparable across multiple reports. Table 35.2 lists the frequency of CRS, as well as its time to onset and median duration in the standard of care reports compared to the registration trials. The rates of grades 3 and 4 CRS in the FDA-licensed product reports, including those of the U.S. Lymphoma CAR T-cell Consortium and of several other real-world multicenter studies, for axicabtagene ciloleucel (Yescarta) ranged from 4% to 13% and 2% to 4%, respectively. This is comparable to the ZUMA-1 registration trial where the rates of grades 3 and 4 CRS were 9% and 3%, respectively. All grade CRS was also similar, with real-world studies reporting a range of 83% to 93% compared to 93% in the ZUMA-1 trial. The rate of grade 3 or higher CRS in ALL patients treated with tisagenlecleucel (Kymriah) was lower in the post-FDA approval Center for International Blood and Marrow Transplant Research (CIBMTR) registry analysis compared to the registration trial, with a reported rate of 16% compared to 46% in the Eliana registration trial. Similarly, for patients with DLBCL, the rates of grade ≥ 3 CRS in several real-world multicenter reports of tisagenlecleucel (Kymriah) ranged from 1% to 5%, which is lower than the 22% reported in the JULIET registration trial and comparable to that reported in a CIBMTR registry analysis, which was 4.5%. The rates of all grade CRS ranged from 45% in a CIBMTR registry analysis to 70% in a multicenter retrospective report and was 58% in the JULIET registration trial. The median time to CRS onset across all registration trials and standard of care product reports, including from the CIBMTR registry, U.S. Lymphoma CAR T-cell Consortium, and several multicenter reports, was 2 to 3 days, with a range of 1 to 27 days. The median duration of CRS was 7 to 8 days, with a range of 1 to 121 days. The outcomes for both products were comparable between clinical trials and the real-world data.

Insights from Standard-of-Care Studies

An important caveat of retrospective studies with the FDA-licensed products is that the centers included in these reports are mainly the same ones that ran the registration trials. This can explain the similar rate of all grade CRS across the FDA-licensed product trials. Furthermore, the decreased grades 3–4 CRS in the FDA-approved product reports compared to registration trials with tisagenlecleucel (Kymriah) could be caused by experience with the patients receiving those products and thus a better understanding of the management of CRS gained during the execution of the clinical trial. For both axicabtagene ciloleucel (Yescarta) and tisagenlecleucel (Kymriah), the reports on the FDA-licensed products suggest that they can be offered with confidence to a broader patient population without a heightened risk of toxicity. This is particularly true for older patients, given that those treated with the FDA-licensed products had a higher median age, which reached an upper limit of 89 years old in both a multicenter retrospective study and a CIBMTR registry analysis. In addition, patients in the FDA-licensed product reports had a higher level of comorbidities and presented with more aggressive disease before receiving CAR-T-cell treatment. However, the lack of uniformity of CRS grading scales makes comparison across trials difficult. Larger registry trials using uniformly reported data are needed to reach a definitive conclusion as to the actual real-world CRS rates.

Predictive Factors for Cytokine Release Syndrome

Clinical trials have often looked at potential predictive factors for CRS as secondary outcomes. As commercial CAR-T-cell products are becoming more widely used, retrospective real-world data is also beginning to emerge regarding prediction of immune-mediated toxicity in CAR-T-cell patients. The ability to forecast the possible toxicities and their expected severity can influence clinical decisions, such as the setting in which the treatment is administered (outpatient vs. inpatient) and expected intensive care unit and long-term hospitalizations. This may help the informed selection of patients who are more suitable for treatment. The factors that may influence CRS incidence and severity may include disease-related factors such as remission, tumor burden, grade, and stage, as well as patient-related factors such as comorbidities, recent therapies, and performance status. For example, a higher Eastern Cooperative Oncology Group score, signifying a lower patient performance status, was found to be correlated with severe CRS in a retrospective study. The onset of severe CRS may potentially be predicted within 24 hours of CAR T-cell infusion, as patients with severe CRS have increased heart and respiratory rates, decreased blood pressure, increased weight, and decreased serum albumin within this time frame. Monitoring of these signs is therefore critical for predicting and managing CRS.