Quantitative Systems Pharmacology (QSP) Modeling for Immunotherapy-induced Cytokine Release Syndrome

Cytokine Release Syndrome

Cytokine release syndrome (CRS) is an inflammatory immune reaction that clinically manifests as fever, headache, joint and muscle pain in mild cases to hypotension, vascular leakage and organ failure in severe cases (Shimabukuro-Vornhagen et al., 2018). CRS is associated with a tremendous surge of cytokines, chemokines and soluble mediators which are primarily proinflammatory (e.g., IL-6, IL-1, MCP-1, IFNɣ, TNFɑ, and others) and cause hyperactivation of immune cells (Teachey et al., 2016; Hay et al., 2017). The time course of CRS is broadly divided into initiation, peak and resolution stages, with each stage connected to different cytokine signatures. Of note, cytokines such as IL-10 which are typically considered anti-inflammatory also increase in the early stages (Panoskaltsis, 2021). However, it is thought that either the anti-inflammatory cytokines do not increase as much as the proinflammatory cytokines or the combined effect of many proinflammatory cytokines overwhelms the homeostatic mechanism and causes the adverse effects of CRS.

Clinical Challenges in Addressing CRS

There are many known causes of CRS, including graft-versus-host disease, infections, and therapeutic agents such chimeric antigen receptor-T (CAR-T) cells, T cell engagers (TCE), and immune checkpoint inhibitor antibodies (Shimabukuro-Vornhagen et al., 2018). Cases of high-grade clinical CRS have been documented frequently in patients treated with cancer immunotherapeutics and have become a major concern since it limits the efficacy of potent life-saving treatments. However, cytokine levels and clinical symptoms vary from patient to patient, and not all patients develop CRS even if they have elevated cytokines compared to untreated patients. The incidence and severity of CRS is decided based on clinical symptoms as there are no specific diagnostic tests thus far; there are also no conclusive links between the clinical CRS grades/symptoms and underlying biomarker signatures. However, tumor burden, peak concentration of IL-6 and C-reactive protein, lymphodepletion, and CAR-T construct and dosing have been identified as potential risk factors (Yan et al., 2021).  

Limitations of Experimental Models in Predicting CRS

In vitro and preclinical models have not always successfully predicted CRS risks for patients. As non-human primate models do not have a tumor, they are somewhat less impactful in capturing the intensity of CRS. Humanized mouse models bearing tumors can be useful for assessing immune infiltration, activation, and toxicity for immunotherapies, but the inherent differences in mouse and human immune systems compel us to be cautious of any findings from these models. One notable failure of preclinical testing in the past was the lack of CRS prediction for the superagonist anti-CD28 monoclonal antibody TGN1412, which led to the hospitalization and extensive treatment of six healthy young volunteers (Suntharalingam et al., 2006). In vitro co-culture models with PBMC or whole blood can provide mechanistic insights into drug binding and cytokine induction; but the dynamics and the complex feedback loops between the cytokines themselves and the cells that they activate, which are critical elements of CRS propagation, can be overlooked with these models (Shah et al., 2023).

How Quantitative Systems Pharmacology (QSP) Modeling Improves Prediction and Management of CRS

One approach to answer some of the essential questions associated with a particular drug’s mechanism of action and potential for inducing CRS is quantitative systems pharmacology (QSP) modeling. A QSP model can be developed based on known immune cell frequency, trafficking and proliferation patterns in specific tissues to predict drug-dependent cytokine release, as well as the secondary cytokine-dependent cellular dynamics that drive CRS. QSP models can integrate in vitro cytokine and in vivo tox data to predict the dose-response of cytokine release on a population level with a reasonable degree of certainty. Predicting CRS based on cytokine release is more complicated due to the lack of clearly established clinical guidelines, but seminal studies have shown that higher-grade clinical CRS is associated with higher peak levels of cytokines (Teachey et al,. 2016; Hay et al., 2017; Diorio et al., 2022). Consequently, CRS severity can be linked to proinflammatory cytokines’ magnitude or fold change from baseline in the model. Other potential metrics to distinguish CRS from non-CRS subjects (or serious vs mild CRS) in the model could be the ratio of proinflammatory to anti-inflammatory cytokines, or the induction of specific cytokines such as TNFɑ or IFNɣ by the drug at early times. These signatures have been observed for some TCE drugs (e.g., anti-CD19 therapies), but it remains to be seen if these are broadly applicable for immunotherapy-induced CRS. Stepping away from population-level predictions, a customized QSP model may be able to factor in a patient's characteristics that normally put them at higher risk for CRS (e.g., tumor burden, type of disease, lymphodepletion, etc) and predict individual responses more accurately.

In the clinic, patients are given IL-6-blocking therapies like the anti-IL6 receptor Tocilizumab to mitigate CRS (Si et al., 2020). While Tocilizumab has proved efficacious and has not yet shown evidence of compromising the efficacy of anti-cancer immunotherapies, there are instances of patients unresponsive to it and the causes remain unknown (Si et al., 2020). Moreover, the appropriate timing of Tocilizumab administration also varies and must be modified based on patients’ clinical status. QSP models can help with planning mitigation strategies, including predicting the efficacy of cytokine blockers, determination of priming doses (an initial low dose followed by high maintenance dose), and dose fractionation or split dosing. Priming dose selection has been tested empirically so far and clinical optimization of doses can be time-consuming and expensive because of many dose levels and timings. Additionally, some of these strategies could potentially affect the efficacy of the immunotherapies and thus require evaluation on a case-by-case basis.

Applying QSP modeling in combination with large-scale clinical data gathering and analysis of available immune-activating drugs, along with the development of robust in vitro and in vivo experimental models of CRS, can improve our understanding of the concurrence of therapeutic and patient characteristics that increase the risk of CRS.

Of note, there are still major knowledge gaps in the field of CRS that require careful consideration. Cytokine measurements are typically made in blood, not in tissues, and even in the case of blood cancers, bone marrow analysis is limited due to the invasive nature of the tests. So the soluble biomarker and cellular levels in these locations are not well-characterized. Additionally, cytokine levels are extremely dynamic and change in the course of hours to days (Yan et al., 2021), so sampling times are critical. Many studies sample at periodic intervals which do not adequately capture the complexity of events underlying CRS (Panoskaltsis, 2021). Likewise, many studies focus on cytokines alone, but the importance of tracking activation, proliferation and possibly exhaustion of cell types that are key to cytokine release and eventual resolution of inflammation cannot be overstated. Linking clinical features of CRS categories to cellular and cytokine signatures and dynamic changes will go a long way toward developing a predictive approach to reduce CRS. QSP modeling can quantitatively link drug dose to cell activation and expansion, which in turn can be related to cytokine accumulation and finally to symptoms such as fever, pain, etc, as long as there are distinguishable trends in the different CRS groups. Applying QSP modeling in combination with large-scale clinical data gathering and analysis of available immune-activating drugs, along with the development of robust in vitro and in vivo experimental models of CRS, can improve our understanding of the concurrence of therapeutic and patient characteristics that increase the risk of CRS. Ultimately, it will lead to the development of a predictive framework to safely screen immune-oncology drugs and select effective treatment strategies. 

References

• Diorio et al. Comprehensive Serum Proteome Profiling of Cytokine Release Syndrome and Immune Effector Cell-Associated Neurotoxicity Syndrome Patients with B-Cell ALL Receiving CAR T19. Clin Cancer Res. 2022 Sep 1;28(17):3804-3813. PMID: 35705524

• Hay et al. Kinetics and biomarkers of severe cytokine release syndrome after CD19 chimeric antigen receptor-modified T-cell therapy. Blood. 2017 Nov 23;130(21):2295-2306. PMID: 28924019

• Panoskaltsis. Are all cytokine storms the same? Cancer Immunol Immunother. 2021; 70(4): 887–892. PMID: 33416946

• Shah et al. Cytokine release syndrome and cancer immunotherapies - historical challenges and promising futures. Front Immunol. 2023 May 25:14:1190379. PMID: 37304291

• Shimabukuro-Vornhagen et al. Cytokine release syndrome. J Immunother Cancer. 2018; 6: 56. PMID: 29907163

• Si et al. Spotlight on Tocilizumab in the Treatment of CAR-T-Cell-Induced Cytokine Release Syndrome: Clinical Evidence to Date. Ther Clin Risk Manag. 2020 Aug 4:16:705-714. PMID: 32801727

• Suntharalingam et al. Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. N Engl J Med. 2006 Sep 7;355(10):1018-28. PMID: 16908486

• Teachey et al. Identification of Predictive Biomarkers for Cytokine Release Syndrome after Chimeric Antigen Receptor T-cell Therapy for Acute Lymphoblastic Leukemia. Cancer Discov. 2016 Jun;6(6):664-79. PMID: 27076371 

• Yan et al. Characteristics and Risk Factors of Cytokine Release Syndrome in Chimeric Antigen Receptor T Cell Treatment. Front Immunol. 2021 Feb 23:12:611366. PMID: 33708205