(696h) A Model-Based Investigation of Cytokine Storm for T-Cell Therapy

In the last decade, scientists developed a process to re-engineer T-cells from a patient which currently serve as a last resort against blood cancers, such as acute lymphoblastic leukemia. The genetically engineered T-cells are called Chimeric-Antigen Receptor (CAR-T) cells, where the T-cells are reproducing with the specific antigen against cancer cells (1) (2). Those treated have seen wildly successful rates of cancer remission (3). Introduction of CAR-T cells into a patient’s blood system often leads to a moderate to severe cytokine release syndrome (CRS), and in its most lethal stage what is called a cytokine storm (3). Normally, CRS affects limbs, organs and other areas of the body, but can also affect the brain. This is referred to as cerebral CRS (4). Cytokines are regulated and produced in a number of different ways; from T-cells, within feedback loops of other cytokines, and with other signalizing and indicator molecules (5). Cytokine storm and cytokine release syndrome are characterized by an unregulated immune response to the CAR-T cell therapy or other foreign molecule, inducing a large quantity of cytokines to be released which can be potentially life-threatening to the patient receiving the therapy. The most infamous case of cytokine storm was the millions of healthy individuals who perished in 1918 thanks to the Spanish Pandemic flu. Thankfully today, there are several drugs on the market which serve to counteract this upregulation of cytokines, such as Tocilizumab, an antagonist of the IL6 receptor, as well as an IL2 receptor antagonist antibody (6) (3). In several cases, these two drugs have saved patients experiencing cytokine release and cytokine storm from near-death experiences (7).

In 2006, a Phase 1 clinical trial of a CD28 monoclonal antibody superagonist was introduced to 6 healthy patients, all male subjects (7). The drug had previously been tested in other animal studies and had been successful (8). In less than an hour, all the patients exposed to the superagonist had spiked cytokine levels, where some of the markers of cell death and inflammation, such as cytokines Interferon-γ and TNF-α were at rising at a dangerous rate (7). Doctors stepped in and delivered analgesic drugs and an IL2 receptor antagonist antibody to blunt the reaction. Although all 6 patients survived, all of them had permanent effects from only 8mg of the CD28 superagonist. Doctors captured frequent blood samples and measured cytokine levels using cytobeads, as well as levels of T-cells, macrophages and other indicators of inflammation (7). Half a decade later, Stengel’s group at Princeton University developed the first human multi-cytokine computational model in an article entitled, Dynamics of a Cytokine Storm (6). It is the only model to date which models human cytokine levels on this scale. The researchers developed a second order linear system of equations, both uncoupled and coupled for all nine cytokines monitored. This model illustrated the positive and negative relationship between cytokines; it was an 18 variable, 90 parameter model.

In this work, we further simulated Stengel’s CRS model to: 1) study the variation ranges (i.e., uncertainties) of TNF-α, IFN-γ, IL1, IL2, IL4, IL6, IL8, IL10 and IL12 from 5,000 in silico patients treated by the T-Cell therapy; and 2) test the synergistic effect of simultaneous inhibitions of multiple cytokines on reducing CRS. Specifically, a local sensitivity analysis was conducted to determine the most sensitive parameters out of the 90 parameters in the model. Each parameter was changed -50% and 50% and the change of the peak values of all 9 cytokines was recorded. Parameters were then ranked based on their sensitivity values. It is found that 7 of 10 of the most sensitive parameters in the model were related to IL12 and IL12 coupling with other cytokines. This gave insight to look into IL12 interactions. From the learnings of the sensitivity analysis, a Monte Carlo simulation was conducted with variation in these 10 parameters to simulate 5,000 in-silico patients and obtain the corresponding peak cytokine levels. The results enabled the team to conduct a principal component analysis (PCA) as well as hierarchical clustering, which is different from the original model developed in the Yiu article (6). The original article looked at the wave function profile of just the patients from the clinical trial, while this PCA took a look at the variation in the in-silico patients generated in MATLAB. In addition, the PCA enabled the team to utilize current drugs out on the market to simulate in-silico ­individual cytokine knockouts using the model from the original Yiu paper. The results demonstrate the effect that IL2 and IL6 have on the relationship with other cytokines, which confirms the positive and negative relationships from other models and literature reviews (5).

References

1. Kinetics and biomarkers of severe cytokine release syndrome after CD19 chimeric antigen receptor-modified T-cell therapy. Hay, K., et al. 2017, Blood, pp. 2295-2306.

2. Engineered T Cells for cancer therapy. June, C, et al. s.l. : Cancer Immunology Immunotherapy, 2014.

3. Current Concepts in the diagnosis and management of cytokine release syndrome. Lee, D. W., et al. 2014, Blood , pp. 188-195.

4. Predominant Cerebral Cytokine Release syndrome in Cd19-directed Chimeric Antigen receptor-modified T-Cell Therapy. Hu, Y, et al. s.l. : Journal of Hematology and Oncology, 2016.

5. Different Subsets of T Cells, Memory, Effector Functions, and CAR-T Immunotherapy. Golubovskaya, V. and Wu, L. 2016, Cancers, pp. 1-12.

6. Dynamics of a Cytokine Storm. Yiu, H. H., Graham, A. L. and Stengel, R. F. 2012, PLOS One, pp. 1-15.

7. Cytokine Storm in a Phase 1 Trial of the Anti-CD28 Monoclonal Antibody TGN1412. Suntharalingam, G., et al. s.l. : The New England Journal of Medicine, 2006, Vol. 355.

8. Monoclonal antibody TGN1412 trial failure explained by species differences in CD28 expression on CD4+ Effector memory T-cells. Eastwood, D, et al. Hertfordshire, UK : British Journal of Pharmacology, 2010.