Cytokine release syndrome-like serum responses after COVID-19 vaccination are frequent and clinically inapparent under cancer immunotherapy

Patients with cancer frequently receive immune-checkpoint inhibitors (ICIs), which may modulate immune responses to COVID-19 vaccines. Recently, cytokine release syndrome (CRS) was observed in a patient with cancer who received BTN162b2 vaccination under ICI treatment. Here, we analyzed adverse events and serum cytokines in patients with 23 different tumors undergoing (n = 64) or not undergoing (n = 26) COVID-19 vaccination under ICI therapy in a prospectively planned German single-center cohort study (n = 220). We did not observe clinically relevant CRS (≥grade 2) after vaccination (95% CI 0–5.6%; Common Terminology of Adverse Events v.5.0) in this small cohort. Within 4 weeks after vaccination, serious adverse events occurred in eight patients (12.5% 95% CI 5.6–23%): six patients were hospitalized due to events common under cancer therapy including immune related adverse events and two patients died due to conditions present before vaccination. Despite absence of CRS symptoms, a set of pairwise-correlated CRS-associated cytokines, including CXCL8 and interleukin-6 was >1.5-fold upregulated in 40% (95% CI 23.9–57.9%) of patients after vaccination. Hence, elevated cytokine levels are common and not sufficient to establish CRS diagnosis.


Supplementary Note 40 41
The supplementary note contains a detailed history of all patients who were vaccinated and had 42 post-vaccine AE that led to hospitalization and/or death. Each patient's history is depicted in a 43 timeline, and we discuss the beginning of immunotherapy, the various AE that we observed during 44 therapy, the vaccination dates, the post-vaccine AE and the reasons for hospitalization and/or 45 death in detail. All information is also available in the Supplementary Tables listing basic patient  46 characteristics (Table S1), the AE observed (Table S2-3) and the reasons for hospitalizations 47 (Table S4) and deaths (Table S5) was admitted due to grade 3 diarrhea clostridium difficile infection was diagnosed and after 201 antibiotic treatment the patient was discharged on 11.06.21 ( Figure T3B). On 18.9.21 the patient 202 was again admitted for diarrhea due to enteropathogenic E. coli (EPEC) colitis grade 3 ( Figure  203 T3B). He was discharged on 22.9.21 but diarrhea quickly reoccurred upon axitinib therapy. 204 Consequently, axitinib therapy was reduced to 3mg bi-daily under which diarrheas ceased 205 completely. after 50 days the first vaccination dose with sampling date indicated on the left. We observed 217 high sIL-2RA levels before ( 68-year old female patient who started ipilimumab and nivolumab as a 1 st line therapy for 236 metastatic melanoma on 19.03.21 ( Figure T4A). She received her first dose of BNT162b2 on 237 12.05.21. On 18.05.21 she was admitted to an external hospital because of grade 4 diarrhea 238 ( Figure T4B). After no other cause could be established, autoimmune colitis was suspected and 239 the patient received 2mg/kg body weight i.v. methylprednisolone under which diarrheas ceased 240 and she was discharged on 28.5.21 under oral glucocorticoid tapering. ICI was paused and not 241 reinitiated until the end of follow-up 1.10.21. Notably, this patient showed high sIL-2RA levels 242 before the vaccination and clinical onset of the colitis event ( Figure T4C) suggesting that some 243 subclinical inflammatory condition, such as colitis, may have existed before onset of clinical 244 symptoms ( Figure T4C). However, it is also possible that vaccination-induced cytokine release 245 contributed to the colitis event.  Figure T5A). Therapy was initially tolerated very well, and he received his first COVID-278 19 vaccination on 14.4.21. On 03.05.21, the patient was admitted to our inpatient ward because of 279 increasing fatigue and signs of cholestasis with darkening of urin and icteric sclerae ( Figure T5B). 280 Laboratory studies showed °II elevated bilirubin, °III alkaline phosphatase and °II aspartate 281 transaminase elevation whereas alanine transaminase levels stayed within normal reference 282 range indicating sever liver damage. Liver sonography showed no signs of obstruction. CT 283 confirmed progressive disease with new and enlarged liver metastases. Given the rapidly 284 deterioration of the patient he wished to not continue tumor-specific oncological therapy and was 285 instead placed on best supportive care. We adjusted the patient's analgesia and nausea 286 medication. The patient had no fever at any point during in-patient care. However, the patient 287 showed increased levels of several serum cytokines after the first COVID-19 vaccination such as 288 CXCL8, IL-6, CCL2, IL-18 and sIL-2RA ( Figure T5C) 76 year-old male patient who started nivolumab as a 2 nd line therapy for concurrent metastatic lung 321 adenocarcinoma and locally advanced squamous cell carcinoma of the esophagus ( Figure T6A).

322
He initially tolerated therapy well and received his 1 st AZD1222 vaccination on 28.04.21. Most 323 analyzed cytokine-release syndrome-associated serum cytokines decreased after vaccination 324 ( Figure T5C). On 18.05.21 he was hospitalized for lack of bowel movements for 6 days at an 325 external hospital ( Figure T5B) Here, we seek to explore changes in the peripheral blood immune cell composition as histology agnostic on-treatment biomarker candidates across mono-and combination immunotherapies (IT) in a single-arm cohort trial with a retrospective training and prospective validation cohort. Cancer patients undergoing immunotherapy receive pre-and on-treatment peripheral blood draws (1-7 weeks) within the framework of the NCT Liquid Biobank. We will analyze changes in the immune contexture in an unbiased manner by single cell RNA sequencing, flow cytometry and quantitative serum proteomics. Based on these markers, we will develop a biomarker-guided multiparametric prediction model for radiological response to therapy, which will be the primary outcome of this study. Secondary outcomes include the immune cell and serum proteome/metabolome composition under IT and at disease progression, grade ≥3 adverse events, as well as progression-free and overall survival. This study will define tumor type independent on treatmentbiomarkers, which can be validated in future prospective multicenter trials to guide cancer immunotherapy.

Introduction
Immune-checkpoint inhibitors have prolonged overall survival in many types of tumor either in monotherapy or in combination with other immunotherapies, targeted therapies and/or cytotoxic chemoor radiotherapy (1-6). However, only a small fraction of patients responds which generally cannot be identified reliably by available biomarkers such as immunohistochemistry assessment of PD-L1 and mismatch repair protein expression. Therefore, patients currently must undergo radiological imaging before and after 2-6 months of therapy to determine whether a relevant response has been achieved. In case of treatment failure, this results in a loss of time, which could be used for other promising therapies in these patients. Moreover, patients may suffer from unnecessary adverse events while undergoing an ineffective therapy for months.
Small retrospective studies suggest similar changes in peripheral blood mononuclear cells in responders to different immunotherapies in several types of tumor (7)(8)(9)(10). A possible explanation for this observation is that different immunotherapies induce similar anti-tumor immune responses in multiple tumor types.
These responses are not present before the start of therapy but can be exclusively detected after initiation of treatment. Therefore, changes in leukocyte composition during therapy represent a promising histology-agnostic on-treatment biomarker candidate across different immunotherapies. Because leukocyte phenotypes are shaped by the surrounding metabolomic and cytokine environment (11, 12), these factors should be accounted for when analyzing changes in immune cell composition.
Here, we will explore whether changes in peripheral blood immune cell composition and the serum proteome and metabolome can be used as early-on treatment biomarkers which predict response to immunotherapies across tumor types. These biomarkers could be used to identify non-responders to immunotherapies early and switch treatments in case of low response probability thereby saving patients valuable time and decreasing overall healthcare costs for expensive but in some cases ineffective therapies.
Another problem in most immunotherapies is the development of potentially life-threatening immune related adverse events (irAE) in which the immune response gets directed against healthy tissues.
Currently, there are no biomarkers available predicting this event. As a secondary outcome we will therefore assess the peripheral blood immune cell composition and the surrounding serum proteome/metabolome as a possible biomarker for irAEs.
At some point of therapy patients will develop secondary resistance, meaning the progression of disease after an initial response. This may be associated with the development of immune escape mechanisms such as the recruitment of T cell suppressive cell populations or soluble mediators (13). We will characterize these changes at secondary resistance and thereby help identify new immunotherapy targets aiming to improve current immunotherapy approaches.

Aim of the study
Here we will analyze the immune cell, serum protein and metabolite composition in peripheral blood during immunotherapy (IT) and combinations of IT with other cancer therapies such as chemotherapy, radiotherapy, and targeted therapy across different types of tumor. We will derive a multiparametric predictive biomarker model for radiological response to IT based on changes in the abundances of different immune cell subsets, proteome and metabolite patterns during therapy. Moreover, we will develop predictive models for secondary therapy resistance and immune-related adverse events based on these biomarkers. To understand the molecular mechanisms leading to these events we will test different functions of peripheral blood immune cells of our patients using preclinical in vitro and xenograft mouse models.

Primary outcome parameter
The study is exploratory in nature. The primary endpoint is radiological response to therapy for which a biomarker-based prediction model will be developed based on the immune cell composition and serum proteome and metabolome of these patients.

Secondary outcome parameters
Secondary endpoints include the immune cell and serum proteome and metabolome composition under IT and at disease progression, grade ≥3 adverse events including infections, as well as progression-free and overall survival. Select leukocyte subsets will be tested for cytotoxicity, migration, proliferation, metabolic function and cytokine production and differentiation properties in vitro and in xenograft mouse models. The effects of patient serum factors on these properties will also be investigated. 6

Study Design
Interventions: This is a non-interventional cohort trial.

Patient recruitment and informed consent
Eligible are patients with malignant neoplasms (ICD-10 2016, C00-C97) undergoing immunotherapy at Heidelberg University Hospital, Heidelberg, Germany who receive blood draws before and 1-7 weeks after initiation of immunotherapy as well as every 8-12 weeks under immunotherapy within the framework of the "NCT Liquid Biobank" program. Patients recruited retrospectively provide informed consent upon inclusion into the "NCT Liquid Biobank" program and explicitly agree to cellular and genomic analyses (please see appendix 1). Patients were informed about the risks of identification especially regarding genomic analyses (appendix 1). It is therefore ethically justified to not ask these patients for study-specific consent. We will recruit 220 patients for the training cohort (including 10% drop-out), which may be recruited from the NCT liquid biobank retrospectively. We will recruit another 220 patients prospectively for the validation cohort (including 10% drop-out). These patients will provide written informed studyspecific consent to "NCT ANTICIPATE", Patient numbers are defined by statistical sample size estimation as outlined below (Appendix 2).

Figure 1 Sampling strategy
Patients of multiple tumor types undergoing immunotherapy will receive repeated blood draws at day 0, at 1-7 weeks and every 8-12 weeks after initiation of therapy until disease progression. Peripheral blood will be analyzed by single cell RNA sequencing (scRNAseq), flow cytometry and serum proteomics, as well as by in vitro and in vivo cytotoxicity, migration, metabolic function, proliferation and cytokine production and differentiation assays. These outcomes will be used to model radiological response to therapy as assessed by CT scan and development of immune related adverse events (irAEs).

Study outline
Peripheral blood mononuclear cells (PBMC) will be isolated by density gradient centrifugation at NCT Heidelberg laboratories. Samples will be stored in liquid N2 . PBMC will be sorted for viable CD45 + cells using fluorescence-assisted cell sorting and analyzed by single cell RNA sequencing and flow cytometry.
Serum will be analyzed by quantitative mass spectrometry, multiplex cytokine and peptide arrays. Select protein concentrations changing under IT will be validated by ELISA and immunoblots. Radiological response to therapy will be determined using the RECIST 1.1 criteria (14). Changes in leukocyte subset frequencies, serum proteome and metabolome concentrations at 1-7 weeks after therapy initiation as 8 compared to baseline will be used to derive a predictive biomarker model for radiological response to therapy and a receiver-operator curve (ROC) will be calculated. This first model will be constructed using a training cohort of 200 patients which may be recruited retrospectively from the NCT liquid biobank. This model will be validated in a validation cohort of another 200 patients, which will be recruited prospectively.
To explore the biological relevance of specific leukocyte subsets PBMC will be challenged with different functional assays in vitro and after xenotransplantation in mice including measurements of cytotoxicity and its surrogates, proliferation, migration, metabolic function, cytokine production and differentiation. The effect of serum factors on these functions will also be investigated.

Time frame
This trial is planned for 7 years until 31.05.2027. We assume a recruitment of approximately 100 patients per year.

Data protection
Blood samples have been collected within the "NCT Liquid Biobank" in a pseudonymized format. The following information will be retrieved from Heidelberg University's electronic patient record files: Whole blood counts and leukocyte differential counts as well as c-reactive protein at time of blood sample collection, age at day of first blood draw, gender (m/f/d), tumor type, stage and histology, mutational status of the tumor, sites of metastasis, ethnic background, history of tobacco use (current, previous, never), preexisting health conditions, concurrent medication and grade 3/4 adverse events during immunotherapies according to Common Terminology Criteria for Adverse Events (CTCAE) v.5.0 and the European Society for Medical Oncology guidelines on irAEs (15). Data retrieval will be performed by a physician or a Heidelberg University medical student who is in regular contact with oncology patients and is bound to "ärztliche Schweigepflicht" and the Landesdatenschutzgesetz (LDSG BW, BDSG). The actual identity will only be known to the NCT biobank personal, the person retrieving the data and the study lead. All laboratory analyses will be performed under the patient pseudonym. Data collected within the trial include scRNAseq, FACS, protein and metabolite concentrations and leukocyte subset function, data from blood samples and imaging results. The timeframe for this study is set from 1.6.2020-31.05.2027.
Data will be stored for 10 years after completion of the trial and will be anonymized on 01.06.2037.
Patients have received contact numbers and e-mail addresses should they chose to resign from participation in the NCT liquid biobank program. In this case all information of the patient will be destroyed unless specified otherwise.

Risks for patients
This study is non-interventional. Risks are described in further detail in the "NCT Liquid Biobank" protocol (S-207/2005).

Benefits for patients
Participants do not directly benefit from participation in the trial. However, future patient generations may benefit from the results obtained in this study. Identification of a powerful on-treatment biomarker would enable faster identification of beneficial treatment strategies for individual patients and reduce the time of patients undergoing ineffective/suboptimal therapies with their inherent adverse events.

Study type
Single center non-interventional single-arm cohort study. -complete sampling status (blood samples obtained before starting IT (baseline) and 1-7 weeks after therapy initiation) will be included in the final analysis. Patients may not have discontinued immunotherapy between baseline sample and 1-7 weeks sampling.
-Complete response evaluation: imaging studies by computed tomography or magnetic resonance imaging performed <6 weeks before start of the immune combination therapy (baseline) and 6-14 weeks after initiation. For inclusion in response analysis, patients may not have discontinued immunotherapy between the imaging studies used for response evaluation. Death from any cause will be considered progressive disease and may be used for response evaluation.

Randomization
This is a single-arm cohort trial. No randomization is required.

Exit criteria patients
-Withdrawal of consent for participation in the "NCT Liquid Biobank" or "NCT ANTICIPATE" o When patients decide to exit the study, they can decide whether their data and biomaterial collected up to this point can be used or whether all their personal data and biomaterials shall be destroyed.

Exit criteria study
-Publication of a tumor type and combination therapy independent biomarker predicting response to IT across histologies and above mentioned immunotherapies with an AUC of ≥0.90 within the 95% confidence interval in a prospective, independent, multicenter validation cohort.
-Failure to retrieve the required patient number for the training cohort by 1.6.2025

Compensation for health damage because of the study interventions
Following the "NCT Liquid Biobank" protocol no compensation is provided as no relevant health damage is inflicted.

Secure laboratory
All analyses will be performed and sample will be stored in rooms 02.116 and 99.101 at National Center for Tumor Diseases (NCT) Im Neuenheimer Feld 460 and rooms 01.116 and 01.113 at Otto-Meyerhof-Zentrum Im Neuenheimer Feld 350, 69120 Heidelberg.

Sample size estimation
Based on a meta-analysis of current clinically applied predictors to PD-1/PD-L1 immune checkpoint blockade (16), we defined an AUC of >0.78 as a clinically relevant threshold above the current biomarker predictive power. To assess the potential predictive power of using the immune cell composition as a biomarker we used published data from renal cell carcinoma patients undergoing combined nivolumab and ipilimumab immune checkpoint therapy to calculate a receiver operator curve (Appendix 2, Figure   2). Based on this AUC confidence intervals for different numbers of patients were calculated assuming a response rate of 10% to immune checkpoint therapy. This response rate is the average response rate in poorly responding tumor types, which likely represent the majority of our patients at NCT (17). In our sample size estimation, a patient number of n=200 resulted in a lower 95% confidence interval of AUC=0.782, which was above our predefined clinically relevant threshold. We estimate a drop-out of 10% due to loss to follow-up of the patients resulting in a total of 220 patients for the training and 220 patients for the validation cohorts to be recruited.

Statistical analysis
For the primary endpoint treatment response, a prediction model will be developed based on a logistic regression model with variable selection and compared to a deep neural network using a training cohort of 200 patients. Validity of the models will be assessed via cross-validation and further assessed in a prospective validation cohort of another 200 patients. The AUC for the prediction model will be calculated together with 95% confidence intervals. We will perform subset analyses for biomarker sensitivity and specificity in each tumor type and each combination therapy as well as after stratification for age, gender, tumor type, stage, histology, mutational status, sites of metastasis, ethnic background, preexistent health conditions, concurrent medication and ECOG. Secondary endpoints overall and progression-free survival will be evaluated using the Kaplan-Meier method. Differences in the distribution of the above-named variables between biomarker positive and negative patients between will be analyzed by Mann-Whitney-U tests (continuous, interval, ordinal variables) and chi-square tests (categorical variables).

Ethical aspects
The trial will be conducted in accordance with the declaration of Helsinki in its current edition. The study protocol will be evaluated by the ethics committee of the Medical Faculty of Heidelberg University. The names of the patients and other confidential or privileged information are in accordance with "ärztliche Schweigepflicht", the Datenschutz-Grundverordnung (DSGVO) and the Landesdatenschutzgesetz (LDSG BW, BDSG). Confidential or privileged health related information will only be passed-on in a pseudonymized format. Third party members do not get insight into the original documentation unless required by law. Participation in the "NCT Liquid Biobank" is voluntary. Patients may withdraw consent by themselves or by their proxy by law at any time without reason and without disadvantages for their future treatment and healthcare. Patients were informed about the benefits and risks of participation in the "NCT Liquid Biobank" Participants were informed about the nature of the performed analysis on cellular compositions including genomic analyses. Patient consent has been documented and signed in the informed consent form in written form. This trial is supported financially by NCT Heidelberg. In case a patient withdraws consent of participation in the trial, all information obtained for this patient will be destroyed unless specified otherwise by the patient.

Conflicts of interest
All investigators have no conflicts of interest to disclose.

Funding
Cost will be covered by Clinical Cooperation Unit Virotherapy at National Center for Tumor Diseases (NCT) and German Cancer Research Center Heidelberg, Germany. Funding originates from public and private third parties.

Report Definitions
Confidence Level is the proportion of confidence intervals (constructed with this same confidence level, sample size, etc.) that would contain the true coefficient alpha. N is the total number of subjects sampled. R is N2 / N1, so that N2 = R x N1. N1 is the number of subjects sampled from the 'positive' group. N2 is the number of subjects sampled from the 'negative' group. Sample AUC is the anticipated value of the sample area under the ROC curve. C.I. Width (UCL-LCL) is the width of the confidence interval. It is the distance from the lower limit to the upper limit. Lower and Upper Confidence Limits are the actual limits that would result from a dataset with these statistics.
They may not be exactly equal to the specified values because of the discrete nature of the N1 and N2.