Article Open Access Identifying Early Infections in the Setting of CRS With Routine and Exploratory Serum Proteomics and the HT10 Score Following CD19 CAR-T for Relapsed/Refractory B-NHL Kai Rejeski1,2,3,4, Viktoria Blumenberg1,2,3,4, Gloria Iacoboni5,6, Lucia Lopez-Corral7,8, Soraya Kharboutli4,9, Rafael Hernani10, Agnese Petrera11, Niklas Müller1, Friederike Hildebrand1, Lisa Frölich1,3, Philipp Karschnia12, Christian Schmidt1, David M. Cordas dos Santos1,3, José Luis Piñana10, Fabian Müller4,9, Ana Africa Martin7,8, Martin Dreyling1, Michael von Bergwelt-Baildon1,3,4, Pere Barba5,6, Marion Subklewe1,2,3,4, Veit L. Bücklein1,2,3,4 GRAPHICAL ABSTRACT Article Open Access Identifying Early Infections in the Setting of CRS With Routine and Exploratory Serum Proteomics and the HT10 Score Following CD19 CAR-T for Relapsed/Refractory B-NHL Kai Rejeski1,2,3,4, Viktoria Blumenberg1,2,3,4, Gloria Iacoboni5,6, Lucia Lopez-Corral7,8, Soraya Kharboutli4,9, Rafael Hernani10, Agnese Petrera11, Niklas Müller1, Friederike Hildebrand1, Lisa Frölich1,3, Philipp Karschnia12, Christian Schmidt1, David M. Cordas dos Santos1,3, José Luis Piñana10, Fabian Müller4,9, Ana Africa Martin7,8, Martin Dreyling1, Michael von Bergwelt-Baildon1,3,4, Pere Barba5,6, Marion Subklewe1,2,3,4, Veit L. Bücklein1,2,3,4 Correspondence: Kai Rejeski (kai.rejeski@med.uni-muenchen.de). ABSTRACT Early fever after chimeric antigen receptor T-cell (CAR-T) therapy can reflect both an infection or cytokine release syndrome (CRS). Identifying early infections in the setting of CRS and neutropenia represents an unresolved clinical challenge. In this retrospective obser- vational analysis, early fever events (day 0–30) were characterized as infection versus CRS in 62 patients treated with standard-of-care CD19.CAR-T for relapsed/refractory B-cell non-Hodgkin lymphoma. Routine serum inflammatory markers (C-reactive protein [CRP], interleukin-6 [IL-6], procalcitonin [PCT]) were recorded daily. Exploratory plasma proteomics were performed longitudinally in 52 patients using a multiplex proximity extension assay (Olink proteomics). Compared with the CRSonly cohort, we noted increased event-day IL-6 (median 2243 versus 64 pg/mL, P = 0.03) and particularly high PCT levels (median 1.6 versus 0.3 µg/L, P < 0.0001) in the patients that developed severe infections. For PCT, an optimal discriminatory threshold of 1.5 µg/L was established (area under the receiver operat- ing characteristic curve [AUCROC] = 0.78). Next, we incorporated day-of-fever PCT levels with the patient-individual CAR-HEMATOTOX score. In a multicenter validation cohort (n = 125), we confirmed the discriminatory capacity of this so-called HT10 score for early infec- tions at first fever (AUCROC = 0.87, P < 0.0001, sens. 86%, spec. 86%). Additionally, Olink proteomics revealed pronounced immune dysregulation and endothelial dysfunction in patients with severe infections as evidenced by an increased ANGPT2/1 ratio and an altered CD40/CD40L-axis. In conclusion, the high discriminatory capacity of the HT10 score for infections highlights the advantage of dynamic risk assessment and supports the incorporation of PCT into routine inflammatory panels. Candidate markers from Olink proteomics may further refine risk-stratification. If validated prospectively, the score will enable risk-adapted decisions on antibiotic use. INTRODUCTION cell associated neurotoxicity syndrome (ICANS).9–11 Real-world evidence has further emphasized the importance of hematolog- Although chimeric antigen receptor T-cell (CAR-T) ther- ical toxicity as the most common high-grade toxicity of CAR-T apy has revolutionized the treatment landscape of relapsed/ therapy, which can present both as a syndrome of profound refractory (R/R) B-cell non-Hodgkin lymphoma (B-NHL),1–8 bone marrow aplasia,12 or as prolonged cytopenia.13–18 An addi- it is accompanied by a unique toxicity profile that classically tional expected on-target/off-tumor side effect of CD19-directed includes cytokine release syndrome (CRS) and immune effector 1 Department of Medicine III – Hematology/Oncology, University Hospital, LMU 10 Hematology Department, Hospital Clínico Universitario, INCLIVA Research Munich, Germany Institute, Valencia, Spain 2 Laboratory for Translational Cancer Immunology, LMU Gene Center, Munich, Germany 11 Metabolomics and Proteomics Core Facility, Helmholtz Zentrum Munich – 3 German Cancer Consortium (DKTK), Munich Site, and German Cancer Research German Research Center for Environmental Health, Munich, Germany Center, Heidelberg, Germany 12 Department of Neurosurgery, University Hospital, LMU Munich, Germany 4 Bavarian Cancer Research Center (BZKF), Partner Sites Munich and Erlangen, Supplemental digital content is available for this article. Germany Copyright © 2023 the Author(s). Published by Wolters Kluwer Health, Inc. 5 Department of Hematology, Vall d’Hebron Institute of Oncology (VHIO), University on behalf of the European Hematology Association. This is an open access Hospital Vall d’Hebron, Barcelona, Spain article distributed under the terms of the Creative Commons Attribution- 6 Department of Medicine, Universitat Autònoma of Barcelona (UAB), Bellaterra, Spain NonCommercial-ShareAlike 4.0 License, which allows others to remix, tweak, and 7 Hematology Department, Hospital Clínico Universitario de Salamanca, IBSAL, build upon the work non-commercially, as long as the author is credited and the CIBERONC, Salamanca, Spain new creations are licensed under the identical terms. 8 Centro de Investigación del Cáncer-IBMCC, Salamanca, Spain HemaSphere (2023) 7:4(e858). 9 Department of Internal Medicine 5, Hematology and Oncology, Friedrich-Alexander- http://dx.doi.org/10.1097/HS9.0000000000000858. Universität Erlangen-Nürnberg and University Hospital Erlangen, Germany Received: August 23, 2022 / Accepted: February 6, 2023 2 (2023) 7:4www.hemaspherejournal.com CAR-T is B-cell aplasia and hypogammaglobulinemia.19 The key exclusion criteria. Lymphodepleting chemotherapy with combination of cellular and humoral immunodeficiency trans- fludarabine and cyclophosphamide was applied according lates into a predisposition for infectious complications, which to the manufacturers’ instructions.1,2 Clinical metadata and substantially contribute to the toxicity burden and drive non- peripheral serum samples were collected with institutional relapse mortality.6,20–24 Infections, and particularly bacterial review board approval (Project No. 19-817). The study was infections, usually occur during the first 30 days after CAR-T performed in accordance with the 1964 Helsinki declaration during a vulnerable phase in which coincident CRS can be and its later amendments or comparable ethical standards. observed.21,23–25 Informed written consent was obtained from all patients. A more vexing and practical challenge encountered after Institutional guidelines for CRS, ICANS, and infection man- CAR-T therapy lies in determining if an episode of early fever agement are outlined in Suppl. Table S1. relates to infection versus CRS. The overwhelming majority of CAR-T patients develop CRS of any grade with pyrexia rep- Infection categorization and classification of severity resenting the cardinal symptom.26 Due to the fact that most patients develop transient lymphodepletion-associated neu- Infections were defined as bacterial, viral, or fungal on the tropenia,13,27 cytokine storm typically presents as a syndrome basis of microbiologic/histopathologic data or as a clinical syn- of “febrile neutropenia.” This usually triggers broad-spectrum drome of infection (eg, pneumonia, cellulitis, cystitis) based on antibiotic use—even in the absence of other clinical signs of retrospective chart review. The study timeframe was 30 days infection or microbiologic evidence. For example, we recently from CAR-T infusion due to the expected maximum time win- demonstrated that a striking 86% of patients receive intrave- dow of coincident CRS.41 All infections before infusion were nous (IV) broad-spectrum antibiotics in the first 10 days after excluded. Infection onset was specified as the day on which the CAR-T infusion.23 However, this comes at the price of antibi- diagnostic test was performed and/or the onset of symptoms. otic-specific side effects, the potential emergence of resistant The clinical source of infection was determined from the com- strains, and selection for Clostridium difficile and entero- bination of symptomology, microbiologic isolates, and radio- cocci.28–32 Furthermore, antibiotics can negatively impact the graphic findings. Fever alone, in the absence of clinical signs of intestinal micromilieu, which should be avoided considering the infection or microbiologic data, was not counted as an infec- recently uncovered immunomodulatory role of the gut microbi- tion. Infection severity was classified as mild, moderate, severe, ome in the context of CAR-T specifically.33–35 For these reasons, life-threatening, or fatal according to previously established simple clinical tools that adequately discriminate between fever criteria.21,42 due to infection versus CRS are urgently needed. The clinical need for precise risk-stratification tools is especially pertinent as Index event categorization outpatient administration of CAR-T is being actively explored.36 CRS and ICANS were assessed prospectively according to We recently developed the CAR-HEMATOTOX (HT) score ASTCT consensus criteria.10 Cumulative incidence curves were for CAR-T–related hematotoxicity.13 The score is determined calculated as time-to-event-analysis from Kaplan-Meier esti- before lymphodepleting chemotherapy (day −5), incorporates mates for CRS 1°, CRS ≥2°, and severe infections; censoring the factors of hematopoietic reserve (ANC, hemoglobin, plate- observation time on the date of progression, last follow-up, or let count) and inflammation (C-reactive protein [CRP], ferri- death. We retrospectively defined fever events as infection versus tin), and risk-stratifies patients into a high versus low risk for CRS, with index events being categorized as the day on which prolonged neutropenia. Importantly, the model also identifies the respective toxicity event occurred. For patients that devel- patients at high risk for infectious complications and poor clin- oped CRS ≥2°, the first day with grade ≥2 CRS was defined as ical outcomes.23 Still, specificity was low and additional param- the event day. eters may refine the predictive capacity of the score. Standard inflammatory marker panels that incorporate CRP, procal- citonin (PCT), and interleukin (IL)-6 represent an attractive Serum cytokine analyses option for dynamic risk assessment, as they are readily avail- Serum inflammatory markers (CRP, PCT, IL-6) were measured able and already implemented at most large CAR-T centers. daily from CAR-T infusion until discharge and on subsequent While specific inflammatory signatures have been associated outpatient visits (weekly during the first month). Laboratory mea- with severe CRS37,38 and life-threatening infections39,40 following surements were assessed at the Institute of Laboratory Medicine CD19 CAR-T, these models often incorporate serum solutes that (University Hospital, LMU Munich). The temporal analyses of are not a part of the clinical routine or require long turn-around inflammatory markers over time were performed by computing times (eg, IL-8, IL-1 beta, interferon-γ), limiting their broad use. the aggregate median value for each day between day 0 and 21. Here, we therefore studied dynamic changes of routine serum Differences between groups were explored using a mixed effects inflammatory markers in the context of CD19 CAR-T, integrat- analysis considering both time and effect size using the restricted ing information from baseline risk predictors. Furthermore, we maximum likelihood method (GraphPad Prism v9.0). explored longitudinal proteomic signatures emerging over time in patients that developed severe infections compared to CRSonly Development of the HT10 risk classification system controls. The baseline HT and EASIX-FC scores were assessed before lymphodepletion (day −5) for all patients as previ- ously described.13,43 Inflammatory markers were tested using MATERIALS AND METHODS binary logistic regression for severe infections versus CRS Patient characteristics and clinical data collection index events. Discriminatory thresholds were determined In this retrospective observational study, early infection from receiver operating characteristic (ROC) curves by opti- events (day 0–30) and day-by-day CRS dynamics were char- mizing the respective Youden J statistic. The “HT10” score acterized in 62 adult patients receiving standard-of-care axi- was modeled for optimal discrimination of severe infections cabtagene ciloleucel (axi-cel, n = 23), tisagenlecleucel (tisa-cel, versus CRS control index events as determined by the highest n = 30), or brexucabtagene autoleucel (brexu-cel, n = 9) for area under the ROC curve (AUROC). Using the established R/R B-NHL. Patients were treated between January 2019 and PCT threshold, different values were added to the baseline March 2022 (data cutoff) at the University Hospital of the HT score and the sum score was tested for discrimination. LMU Munich, Germany. Participation in a clinical trial, or Metrics of score performance (eg, sensitivity, specificity) were CAR-T treatment for a non–B-NHL disease entity represented assessed. 3 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T External validation of the HT10 score a random effect per patient was included to account for poten- Four European CAR-T centers participated in the confirma- tially differing baselines. Regression models were fitted using tory cohort of 125 r/r B-NHL patients receiving standard-of- the R-packages lme4 and lmerTest, while plots were generated care CD19 CAR-T. The 30-day cumulative incidence of viral using the OlinkAnalyze and ggplot2 packages. Protein levels and nonviral infections was assessed. The HT10 score was were expressed in Normalized Protein eXpression (NPX) units, determined at time of first fever after CAR T-cell infusion. All which were derived from Ct values. Because NPX is expressed patients received a standard diagnostic evaluation of underlying in a log2 scale, a 1 NPX difference translates into a doubling of etiologies of fever as per institutional guidelines. Test character- protein concentration. istics were determined using ROC analysis studying if patients developed a severe infection during the first 30 days versus CRS Statistical considerations only. Statistical significance was explored by nonparametric Mann-Whitney test for continuous variables and Fisher exact Longitudinal Olink plasma proteomic assays test for comparison of group variables. Correction for multiple The serum proteome was characterized across 4 sequential testing was performed using the false discovery rate approach. time points (days 0, 4, 14, 28) in 52 patients using a 92-protein Statistical analysis and data visualization was performed using multiplex proximity extension assay from the Olink platform GraphPad Prism (v9.0), SPSS (v26.0), or R Statistical Software (“Immuno Oncology Panel,” Olink Bioscience). Experiments (v4.1.2). were performed as previously described.44,45 Briefly, olignonu- celotide-labeled monoclonal or polyclonal antibodies (PEA probes) were used to bind target proteins in a pairwise manner, RESULTS thereby preventing all cross-reactive events. Upon binding, the oligonucleotides come in close proximity and hybridize, fol- CRS and infection rates in a real-world cohort of CAR-T–treated lowed by extension, generating a unique sequence used for the B-NHL patients digital identification of the specific protein assay. Across all patients, the median age was 64 years (range 19–83), A linear mixed model (LMM) was applied per protein that median ECOG was 1 (IQR 1–2), and the median international tested differences of serum proteins by patient group (infection prognostic index (IPI) in evaluable LBCL patients was 2 (IQR versus CRSonly control) and time point. To account for potential 2–3) (Table 1). The study cohort focused on the patients that devel- confounding, age and sex were included as fixed effects, while oped any-grade CRS or an infection (57 patients). Five patients Table 1 Baseline Patient Demographic and Clinical Characteristics CRS 0°–1° (n = 35)a CRS ≥2° (n = 27) P Other (n = 47) Severe Infection (n = 15) P Age, years (95% CI) 61 (59-66) 65 (55-69) 0.65 64 (60-66) 60 (49-69) 0.65 Sex (female) 12 (34%) 12 (44%) 0.44 18 (38%) 6 (40%) 0.99 Prior autologous SCT 14 (40%) 6 (22%) 0.18 14 (30%) 6 (40%) 0.53 Median lines of prior therapy (excl. bridging. IQR) b 4 (3–5) 3 (2–4) 0.04 4 (2–4) 4 (3–4) 0.51 Median ECOG at lymphodepletion (IQR)b 1 (0–1) 1 (0–1) 0.70 0 (0–1) 1 (0.5–1) 0.09 IPI (IQR)b 2 (1–3) 2 (2–3) 0.99 2 (1.25–2.75) 3 (2–3.5) 0.04 Disease entity  DLBCL, PMBCL 22 (63%) 13 (48%) 0.31 24 (51%) 11 (74%) 0.15  Transformed lymphoma 8 (23%) 10 (37%) 0.27 16 (34%) 2 (13%) 0.19  Mantle cell lymphoma 5 (14%) 4 (15%) 0.99 7 (15%) 2 (13%) 0.99 CAR product  4-1BB (Tisa-cel) 17 (49%) 13 (48%) 0.99 24 (51%) 7 (47%) 0.99  CD28z (Axi-cel, KTX-19) 18 (51%) 14 (52%) 23 (49%) 8 (53%) Tumor burden  LDH (U/L), 95% CI 239 (174-319) 229 (201-381) 0.46 230 (199-290) 237 (143-390) 0.91  STLV (mm3, IQR)b 60 (7–218) 130 (33–488) 0.08 61 (15–260) 205 (45–442) 0.12 Risk Classification Systems  CAR-HEMATOTOX Score (Rejeski et al13), IQR 3 (1–4) 1 (0–3) 0.10 2 (1–3) 5 (2–6) 0.002  EASIX-F Score (Greenbaum et al43), IQR 1 (0–1) 1 (0–2) 0.90 1 (0–1) 1 (1–2) 0.04 Laboratory characteristics  C-reactive protein (mg/dL), 95% CI 1.4 (0.8-3.1) 0.9 (0.2-3.6) 0.43 0.8 (0.3-1.4) 3.1 (1.3-4.5) 0.01  Ferritin (ng/mL), 95% CI 830 (348-1450) 496 (180-1272) 0.48 456 (284-830) 1978 (1062-2878) 0.001  Abs. lymphocyte count (G/L), 95% CI 460 (378-736) 550 (421-752) 0.59 550 (430-740) 411 (212-720) 0.20  Abs. neutrophil count (G/L), 95% CI 1690 (950-2260) 2380 (1830-3510) 0.03 1970 (1770-2590) 1350 (580-2330) 0.05  Platelet count (G/L), 95% CI 139 (104-175) 126 (76-230) 0.87 146 (116-185) 70 (20-178) 0.05  Hemoglobin (g/dL), 95% CI 9.0 (8.5-10.1) 9.8 (9.4-10.5) 0.13 10.0 (9.6-10.3) 8.6 (7.6-9.0) 0.003 a Five patients did not develop CRS of any grade, these patients did not develop an infectious complication. b LBCL only. All P values <0.05 are highlighted in bold. CAR = chimeric antigen receptor; CI = confidence interval; CRS = cytokine release syndrome; DLBCL = diffuse large B-cell lymphoma; ECOG = Eastern Cooperative Oncology Group; IPI = International Prognostic Index; IQR = interquartile range; LBCL = large B-cell lymphoma; LDH = Lactate Dehydrogenase; PMBCL = primary mediastinal B-cell lymphoma; SCT = stem cell transplantation; STLV = sum of target lesion volume. 4 (2023) 7:4www.hemaspherejournal.com A B C D E Figure 1. CRS and infection rates in a real-world cohort of CAR-T–treated B-NHL patients. (A) Schematic overview of the study cohort with key inclusion and exclusion criteria. (B) Cumulative incidence rate of CRS grade 1 (gray) and CRS grade ≥2 (magenta) in the first 30 d after CAR-T infusion. (C) Cumulative incidence rate of severe infection (green) in the first 30 d after CAR-T infusion. Severe infections were defined as requiring i.v. anti-infective therapy and/or hospitalization and microbiologic evidence of infection. In the absence of microbiologic evidence overwhelming evidence of clinical infection had to present (eg, clinical symptoms and concordant imaging findings. (D) Aggregated median ANC over time for 62 patients between day 0 (CAR infusion) and day 30. Light shading depicts the 95% CIs of the median for each time point. (E) Relative distribution of CRS grades according to ASTCT grading in patients with severe infection (n = 15) and without severe infection events (n = 47). B-NHL = B-cell non-Hodgkin lymphoma; CAR-T = chimeric antigen receptor T-cell therapy; CI = confidence interval; CRS = cytokine release syndrome. developed neither CRS nor an infection and were excluded from inflammation (median ferritin 1978 versus 456 ng/mL, P = 0.002) subsequent analyses (Figure 1A). Demographic and laboratory and impaired hematopoietic reserve (median hemoglobin 8.6 ver- characteristics were balanced between the patients with mild CRS sus 10.0 g/dL, P = 0.003), which was further reflected by higher (ASTCT 1°) versus moderate-to-severe CRS (ASTCT ≥2°). No HT scores (median 5 versus 2, P = 0.002). significant difference in baseline EASIX-FC scores was observed. The cumulative 30-day incidence of grade 1° and grade ≥2° CRS On the other hand, patients that developed a severe infection had was 47% and 43%, respectively (Figure 1B). Patients with grade a higher median ECOG (2 versus 1, P = 0.01) and IPI (3 versus 2, ≥2° CRS displayed earlier CRS onset (median 1 versus 3 days) and P = 0.07) at baseline compared with the CRSonly control. Notably, longer CRS duration (median 8 versus 4 days) compared with their at baseline, the infection group exhibited high levels of systemic mild CRS counterparts (Suppl. Table S2). Most patients with grade 5 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T ≥2° CRS exhibited previous grade 1° CRS (19/27, 70%), lasting severe neutropenia (ANC <500/µL) (Figure 1D). Indeed, neu- a median of 2 days (IQR 1–2). The anti–IL-6 receptor antagonist tropenia was observed in 37 of 57 (65%) CRS events and 9 of tocilizumab was applied in the overwhelming majority of CRS 15 (60%) infection events. Grade 2 or higher CRS was noted in patients, irrespective of CRS severity (1°: 90%, ≥2°: 100%), reflect- 33% of the patients with a severe infection compared with 47% ing the institutional practice of tocilizumab application in case of the patients without (Figure 1E). We did not find a significant of fever >24 hours. Still, grade ≥2° CRS patients more frequently correlation between CRS grade and severe infections (G2 = 0.13, received high-dose glucocorticosteroids and developed more severe P = n.s., Suppl. Figure S2). Notably, severe infections were asso- ICANS (Suppl. Table S2), consistent with prior reports.23 ciated with a prolonged duration of neutropenia (median 13 A total of 15 patients developed a severe infectious compli- versus 8.5 days, P = 0.01) and a higher frequency of an aplas- cation for a cumulative 30-day incidence of 24.6% (Figure 1C). tic phenotype of neutrophil recovery13 (47% versus 9.5%, P = This included 9 bacterial infections, 2 fungal infections, and 4 0.004) (Suppl. Table S2 and Suppl. Figure S1). Furthermore, patients with a clinical syndrome of infection (2 central line patients that developed severe infections displayed a prolonged infections, 2 pneumonias). Median infection onset was on day duration of high-dose glucocorticosteroid use, and more fre- 8 (IQR 3–23), commonly occurring in the setting of coincident quently required an intensive care unit admission. A Procalcitonin Dynamics B Procalcitonin Dynamics 0.8 0.8 CRS Severe Infection (15) Control (47) Procalcitonin [ g/l] p = 0.84 p = 0.02 0.6 0.6 Procalcitonin [ g/l] 0.4 0.4 0.2 0.2 0.0 0.0 0 3 6 9 12 15 18 21 0 3 6 9 12 15 18 21 Days relative to CAR-T cell infusion Days relative to CAR-T cell infusion C CRP Dynamics D CRP Dynamics 8 8 CRS Severe Infection (15) Control (47) p = 0.96 p < 0.0001 6 6 CRP [mg/dl] CRP [mg/dl] 4 4 2 2 0 0 0 3 6 9 12 15 18 21 0 3 6 9 12 15 18 21 Days relative to CAR-T cell infusion Days relative to CAR-T cell infusion E IL-6 Dynamics F IL-6 Dynamics 2500 2500 CRS Severe Infection (15) Control (47) 2000 p = 0.002 2000 p = 0.99 IL-6 [pg/ml] IL-6 [pg/ml] 1500 1500 1000 1000 500 500 0 0 0 3 6 9 12 15 18 21 0 3 6 9 12 15 18 21 Days relative to CAR-T cell infusion Days relative to CAR-T cell infusion Figure 2. High-grade CRS is characterized by high interleukin-6 levels, while severe infections are marked by baseline inflammation high pro- calcitonin levels. (A–B) Aggregated median procalcitonin values over time during the first 21 d after CAR-T infusion by CRS grade (A) and presence of severe infection (B). (C–D) Aggregated median CRP values by CRS grade (C) and presence of severe infection (D). (E–F) Aggregated median interleukin-6 values by CRS grade (E) and the presence of severe infection (F). Serum samples were prospectively collected in a laboratory panel that was performed at least daily during the first 2 wk and then as indicated. Significance values were determined with a mixed effects analysis considering both time and effect size. CAR-T = chimeric antigen receptor T-cell therapy; CRS = cytokine release syndrome. 6 (2023) 7:4www.hemaspherejournal.com A Pt #1 Observation Time Index Event Pt #2 Severe Infection CRS ≥2° Pt #3 CRS 1° CAR-T t B 2.5 Severe Infection (15) Procalcitonin [ g/l] 2.0 CRS 1.5 1.0 0.5 0.0 -3 -2 -1 0 1 2 3 4 5 6 7 Days relative to Index Event C 6 Severe Infection (15) CRS CRP [mg/dl] 4 2 0 -3 -2 -1 0 1 2 3 4 5 6 7 Days relative to Index Event D 5000 Severe Infection (15) 4000 CRS IL-6 [pg/ml] 3000 2000 1000 0 -3 -2 -1 0 1 2 3 4 5 6 7 Days relative to Index Event Figure 3. Serum inflammatory dynamics by index event. (A) Graphical representation of index event analysis: hypothetical patient #1 displays early grade 1 CRS and no infections; patient #2 displays grade early 2 CRS followed by a severe infection; patient #3 displays an early severe infection followed by grade 1 CRS. Serum procalcitonin (B), CRP (C), and IL-6 (D) levels were retrospectively assessed relative to the respective index event (severe infection: green; grade 1 CRS: gray; grade ≥2 CRS: magenta). The aggregated median values are depicted. CRP = C-reactive protein; CRS = cytokine release syndrome; IL-6 = interleukin-6. 7 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T High-grade CRS is characterized by high interleukin-6 levels, while severe ROC analyses were performed to study how the inflammatory infections are marked by baseline inflammation and high PCT levels markers discriminate for the respective endpoints of grade ≥2° Next, we studied dynamic changes in serum inflammatory versus grade 1° CRS (= CRS index event), as well as severe infec- marker profiles over time in patients with CRS grade 1° versus ≥2°, tion versus any-CRS control (= infection index event) (Table 3). and in CRS patients with and without a severe infection. Following For the CRS index event, the highest AUC was observed for CD19 CAR-T infusion, patients with grade ≥2° CRS displayed IL-6 (AUCROC = 0.95, P < 0.001) with an optimal discrimina- significantly elevated day 0–21 levels of IL-6 compared with their tory threshold of 77 pg/mL. The sensitivity and specificity were grade 1° CRS counterparts (P = 0.002, Figure 2E). On the other 89% and 93% at this threshold, respectively. Conversely, the hand, CRP and PCT levels were not significantly altered in a mixed highest discrimination between severe infectious complications effects analysis accounting for both time and marker elevation and CRS was observed for serum PCT (AUCROC = 0.78, P = (Figure 2A, C). In patients with infectious complications, PCT 0.0007, Figure 4A). An optimal discriminatory threshold was levels were significantly increased over time (P = 0.02, Figure 2B). established at a PCT level of 1.5 µg/L (sensitivity = 60%, spec- Furthermore, CRP levels were significantly increased (P < 0.0001, ificity = 91%). Figure 2D), though this was particularly evident close to CAR-T infusion (day 0–3), further underlying the higher risk of infection Combining the baseline CAR-HEMATOTOX with index-event serum in patients with baseline inflammation (Table 1). The mixed effects PCT increases the discriminatory capacity for severe infections in model did not detect significant changes in IL-6 levels over time the setting of CRS between patients with and without infection (Figure 2F). Although the baseline HT score was associated with severe To elucidate distinct inflammatory signatures in relation to infections (AUCROC = 0.75, P = 0.003, Figure 4B), the observed a specific toxicity event (CRS 1° versus CRS ≥2° versus severe specificity was low at 47%. On the basis of the above find- infection), we performed “index event analyses” that take a ings, we tested if serum PCT values improve upon the baseline retrospective snapshot of laboratory constellations on a day of prognostication of infection risk with the HT score (Table 1). interest (Figure 3A, Table 2). While all inflammatory markers Differently weighted combinations of the score and the estab- were significantly increased on the index-event day in patients lished PCT threshold (1.5 µg/L) were analyzed in regards to their with grade ≥2° versus 1° CRS, this was especially noted for capacity to identify severe infections (Suppl. Table S3). Optimal CRP (median 5.4 versus 1.5 mg/dL, P = 0.001, Figure 3C) and test characteristics were observed with the addition of 10 points IL-6 (median 1196 versus 30.7 pg/mL, P < 0.001, Figure 3D), to the patient-specific HT score in case of serum PCT levels ≥1.5 though this was likely influenced by prior tocilizumab exposure µg/L on the event day. This so-called “HT10” score displayed for previous grade 1° CRS. On the other hand, the severe infec- superior discrimination for severe infections on ROC analysis tion index events were characterized by particularly high PCT (AUCROC 0.92, P < 0.0001, sensitivity 80%, specificity 91%, levels (median 1.6 versus 0.3 µg/L, P = 0.001, Figure 3B) com- Figure 4C) when compared with the HT score or PCT alone pared with the respective CRS controls. Of interest, we did not (Figure 4A-B, Suppl. Figure S3). Furthermore, the “HT10” score find significant differences in neutrophil counts by index event was strongly associated with severe infections on binary logis- group, suggesting that both CRS and infection typically occur tic regression analysis (G2 = 34.13, P < 0.0001, Figure 4D). We during a phase of coincident neutropenia (Figure 1D). found that the risk of an infection versus only CRS was highest in Table 2 Serum Inflammatory Markers on the Index Event Day CRS Index Event Infection Index Event Lab Parameter CRS 1° (n = 30) CRS ≥2° (n = 27) P CRS Control (n = 57) Severe Infection (n = 15) P Procalcitonin (µg/L, 95% CI) 0.2 (0.2-0.3) 0.4 (0.2-0.7) 0.002 0.3 (0.2-0.4) 1.6 (0.3-5.8) 0.001 Median CRP (mg/dL, 95% CI) 1.5 (1.1-3.7) 5.4 (3.3-7.5) 0.001 3.45 (1.8-5.4) 1.8 (1.3-5.5) 0.37 Median IL-6 (pg/mL, 95% CI) 30.7 (18.4-42.7) 1196 (200-3525) <0.001 63.8 (36.4-137) 2423 (44-13481) 0.04 ANC (G/L) 510 (0-1290) 340 (0-1270) 0.21 340 (95-1115) 680 (0-1860) 0.56 Laboratory parameter on the day of the index event. P values were determined by Mann-Whitney test. Correction for multiple testing was performed using the false discovery rate approach (2-stage step-up method of Benjamini, Krieger and Yekutieli). ANC = absolute neutrophil count; CI = confidence interval; CRP = C-reactive protein; CRS = cytokine release syndrome; IL-6 = interleukin-6. Table 3 Receiver Operating Characteristic Analyses Laboratory Parameter Discriminatory Threshold AUC P Value AUC Sens. Spec. CRS index event  Procalcitonin 0.4 µg/L 0.70 0.007 66% 80.0%  CRP 1.83 mg/dL 0.73 0.003 89% 60%  IL-6 77 pg/mL 0.95 <0.001 89% 93% Infection index event  CRP N/A 0.58 0.37 N/A N/A  IL-6 1553 pg/mL 0.68 0.03 60% 81%  Procalcitonin 1.5 µg/L 0.78 <0.001 60% 91% ROC analyses were computed for the predicted probability of severe infection (infection index event) vs any-CRS control and CRS ≥2° vs CRS 1° (CRS index event). Serum inflammatory markers are sorted from top to bottom in the order of optimal discrimination determined by AUC for the respective index event. Discriminatory thresholds were determined based on ROC analyses computed for the predicted probability of grade ≥2 CRS vs grade 1 CRS (CRS index event), as well as the probability of severe infection vs CRSonly control (infection index event). CRP = C-reactive protein; CRS = cytokine release syndrome; IL-6 = interleukin-6; ROC = receiver operating characteristic. 8 (2023) 7:4www.hemaspherejournal.com Figure 4. Combining the baseline CAR-HEMATOTOX with index-event serum procalcitonin increases the discriminatory capacity for severe infection events. Results of the receiver operating characteristic analyses comparing procalcitonin alone (A) or CAR-HEMATOTOX score alone (B) against the binary outcome of severe vs nonsevere infection. (C) ROC analysis studying the influence of a combined CAR-HEMATOTOX and procalcitonin (“HT + PCT10”) score on the binary outcome of severe vs nonsevere infection. In the case of a PCT level ≥1.5 μg/L, an additional 10 points were added to the baseline CAR- HEMATOTOX score. The respective area under the curve and P value of the ROC curve are depicted. (D) Binary logistic regression analysis comparing the association between the “HT + PCT10” score and the binary outcome of severe vs nonsevere infection. The P value is shown for the likelihood ratio test (G2); light shading indicates the 95% asymptotic confidence bands. (E) Simplified HT10 Score: The composite number of patients fulfilling the criteria for a high score (either baseline CAR-HEMATOTOX score ≥5 or day-of-event PCT ≥1.5 μg/L) and the relation to the endpoint of severe infection is depicted. CAR = chimeric antigen receptor; PCT = procalcitonin; ROC = receiver operating characteristic. case of a baseline HT score ≥5 and/or an event-day PCT greater The HT10 score discriminates for early infections at first fever in an than 1.5 µg/L (simplified HT10 score, Figure 4E). These data external validation cohort highlight that the combination of both patient-level baseline risk To test the external validity of the “HT10” score, we studied (HT score), and dynamic changes in select serum inflammatory early infections and CRS in a confirmation cohort of 125 patients markers such as PCT, can help to identify severe infections in the treated with CD19 CAR-T for r/r B-NHL across 4 CAR-T sites setting of CRS. (Vall d’Hebron Barcelona, Erlangen, Valencia, Salamanca). A 9 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T total of 144 patients were screened: 8 non–B-NHL patients and infections (Figure 5E). The first nonviral infection was noted 11 patients without available PCT measurements were excluded on day +19 in the HT10low group, with a 30-day incidence of from the analysis (Suppl. Figure S4). Relevant baseline patient only 3.3%. Overall, we found that the majority of infections features and toxicity results are provided in Suppl. Table S4. occurred within the first 20 days in the confirmation cohort. Patients in the confirmation cohort exhibited fewer lines of prior therapy, had lower hemoglobin and ECOG performance status, but also displayed higher median LDH and IPI at lymphode- Longitudinal proteomic analysis reveals that patients with severe pletion. The rate of severe infections during the first 30 days infections display progressive endothelial dysfunction and severe was 21%; the rate of severe CRS was 10%, both comparable to immune dysregulation the training cohort (Suppl. Table S5). Toxicity management was Inflammatory marker patterns were further characterized in 13 similar between both cohorts with tocilizumab being applied severe infection patients (all with coincident CRS) compared with in 78% of cases. Of interest, 68% of the patients in the confir- 39 CRSonly controls. A total of 201 samples were studied across mation cohort received an antibiotic prophylaxis with a fluoro- 4 time points resulting in 21.068 unique data points. Principal quinolone. On ROC analysis, the HT10 score reliably identified component analysis of the NPX distribution did not identify any patients at time of first fever after CAR T-cell infusion that went potential outliers, and only a single sample was flagged with a on to develop a severe infection as opposed to only CRS (AUC quality control warning (Suppl. Figure S5). To identify proteins 0.87, P < 0.001, sensitivity 86%, specificity 86%, Figure 5A). differential between infection versus CRSonly conditions, LMM In HT10high patients (eg, score ≥5), we observed an increased were fitted to each patient accounting for both infection status rate of all-grade infections (58% versus 21%, P < 0.001) and time effects and adjusting for age, sex, and patient baseline. and especially severe infections (58% versus 8%, P < 0.001) Prior to CAR infusion (day 0), we found that the patients that (Figure 5B). The difference in infection rates were more pro- subsequently developed infections already displayed significant nounced when specifically analyzing nonviral infections (severe protein-level changes compared with their CRSonly counterparts infections: 55% vs 3%, P < 0.0001; Figure 5C). Accordingly, the (top left, Figure 6A). For several proteins, such as TNFRSF12A cumulative 30-day incidence of severe infections was increased and CXC3L1, these protein-level differences were conserved in the HT10high patients (Figure 5D), particularly for nonviral across multiple time points (Figure 6A). A total of 16 candidate A HT10 Score: External Validation (n=125) B Maximal Infection Grade by C Maximal Infection Grade by HT10 Score - All Infections HT10 Score - Non-Viral Infections 100 100 100 p < 0.0001 p < 0.0001 3% 90 8% 90 p < 0.0001 11% 80 Relative Distribution (%) 80 p < 0.0001 Relative Distribution (%) 80 21% 70 70 Sensitivity% 60 60 60 5 5 50 50 40 4 4 40 40 55% / 55% HT10 Score 58% / 58% AUC = 0.87 3 3 p < 0.0001 30 30 2 2 20 Simpl. HT10 AUC = 0.86 20 1 20 1 p < 0.0001 10 0 10 0 Both: Sensitivity 86% / Specificity 86% 0 0 20 40 60 80 100 0 0 100% - Specificity% HT10high HT10low HT10high HT10low D Cumulative Infection Rate - Severe Infections (All Types) E Cumulative Infection Rate - Severe Infections (Non-Viral) 100 100 All Patients (125) All Patients (125) 90 90 Simplified HT10 Score High (33) Simplified HT10 Score High (33) 80 p < 0.0001 80 p < 0.0001 Probability of Infection Probability of Infection Simplified HT10 Score Low (92) Simplified HT10 Score Low (92) 70 70 60 58% 60 55% 50 50 40 40 30 30 20 21% 20 17% 10 7.6% 10 3.3% 0 0 0 10 20 30 0 10 20 30 Days after CAR-T infusion Days after CAR-T infusion Figure 5. External validation of the discriminatory capacity of the HT10 score at time of first fever. (A) ROC analysis studying the influence of the HT10 score (magenta) and simplified HT10 score (blue), determined at time of first fever on the binary outcome of severe infection vs CRS only during the first 30 d after CAR T-cell infusion. The respective area under the curve, P value of the ROC curve, and sensitivity/specificity are depicted. (B–C) Relative distribution of infection grades for all infection subtypes (B) and nonviral infections only (C) during the first 30 d comparing HT10 high vs low patients. Infection grades (1°–5°) are color-coded in shades of green with the connecting green and gray lines and percentage numbers comparing all-grade and grade ≥3 infections, respectively, in HT10 high vs low patients. Significance values were determined by Fisher exact test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (D) Cumulative 30-d incidence on all infection types (viral, bacterial, fungal, clinical syndromes of infection) by HT10 score. (E) Cumulative 30-d incidence of only nonviral infections HT10 score. CAR = chimeric antigen receptor; CRS = cytokine release syndrome; ROC = receiver operating characteristic. 10 (2023) 7:4www.hemaspherejournal.com proteins significantly differed by patient group. These top can- (adjusted P = 4.8E-05, Figure 7A). In terms of proteins related didate proteins were heavily involved in biological processes to inflammatory response, we detected dysregulation of the related to inflammation, immune response, angiogenesis, and CD40/CD40L axis with upregulation of CD40 (adjusted P = endothelial function (Suppl. Table S6–7). Hierarchical clus- 0.049, Figure 7B) and downregulation of CD40L (adjusted P tering of patients using the differentially expressed candidate = 0.029, Figure 7B) in the severe infection group. Furthermore, proteins demonstrated a clear separation of the majority of the innate immune regulators CXC3L1, CCL20, and CXCL13 severe infection from CRSonly patients (Figure 6B). Upregulated were elevated. Taken together, these longitudinal proteomic proteins in the severe infection group included TNFRSF12A, studies highlight progressive immune dysregulation and endo- CXC3L1, CCL20, CXCL13, and IL-15 (Figure 6C). On the thelial dysfunction in CAR T-cell patients developing infec- other hand, soluble CD40-L, ANGPT1, and EGF were down- tious complications as opposed to only CRS. regulated. Of interest, we observed upregulation of ANGPT2 over time (adjusted P = 0.019, Figure 7A) while ANGPT1 was DISCUSSION downregulated (adjusted P = 0.027, Figure 7A)—indicating an increased angiopoietin-2 to angiopoietin-1 ratio in the patients In this retrospective observational study of 62 patients treated that developed severe infections. Endothelial dysfunction was with CD19 CAR-T for B-NHL, we found that severe infections further reflected by downregulation of FGF2 (adjusted P = typically present with increased index-day PCT levels. Combining 0.026, Figure 7A) and marked upregulation of TNFRSF12A the baseline HT score with day-of-event PCT levels enhanced Figure 6. Severe infections are associated with distinct inflammatory signatures compared with CRS only controls. (A) Volcano plots showing the estimated mean difference between infection and CRSonly patients (x-axis), and corresponding −log10 (P values) (y-axis). Colors indicate proteins that passed the multiple testing adjusted P value threshold of P < 0.05. Each panel corresponds to one timepoint (days 0, 4, 14, 28). (B) Heatmap depicting normalized and centered NPX values (color scale) of the 16 proteins that differed between infection and CRSonly patients (significant main and/or interaction effect). Each row corresponds to a sample, and each column to a protein. Both rows and columns are sorted by hierarchical clustering, with dendrograms showing the clustering result. Time and severe infection status are indicated by the leftmost columns. (C) Differential expression per timepoint (from left to right) of the 16 proteins that differed between infection and CRSonly patients (significant main and/or interaction effect). The x-axis depicts the estimated mean difference between infection and CRSonly patients, and colors indicate the corresponding P values (not corrected for multiple testing). Each panel corresponds to one timepoint, with pro- teins sorted by NPX level changes (right: downregulated; left: upregulated in the Infection group). CRS = cytokine release syndrome; NPX = Normalized Protein eXpression. 11 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T Figure 7. Patients with severe infections display progressive endothelial dysfunction and immune dysregulation. Point range plots showing the LMM estimates of mean NPX level (y-axis) per timepoint (x-axis) for the 16 candidate proteins related to either endothelial and angiogenetic dysfunction (A), or immune dysregulation and inflammatory response (B). The infection group is shown in red, while the CRSonly group is shown in teal. The P value corrected for multiple testing (q value) is superimposed on each panel for all Olink candidate proteins. LMM = linear mixed model; NPX = Normalized Protein eXpression. the discrimination for severe infections compared with CRSonly (Suppl. Figure S2).14,21 Still, we did observe higher rates of grade ≥3 events. In a confirmatory cohort of 125 patients, the resulting ICANS in the infection cohort (33% versus 7%, P = 0.02, Suppl. HT10 score identified patients at high risk for severe infectious Table S2), which likely contributed to the more frequent use of complications at time of first fever. By studying longitudinal pro- high-dose glucocorticoids in these patients—an important risk fac- teomic patterns, we identified endothelial dysfunction and pro- tor for infectious complications.23 The majority of both CRS and gressive immune dysregulation in the severe infection group. infection events occurred in the setting of neutropenia, underlining The incidence, temporal distribution, and clinical sever- the ubiquitous nature of febrile neutropenia during the early post– ity of infections and CRS was comparable to prior published CAR-T phase. In terms of routine inflammatory marker profiles, reports.14,21,23,25 In contrast to previous studies, we did not find we did not observe the “double peak of IL-6 pattern” described an association between CRS severity and infection occurrence by Luo and colleagues.39 The authors further delineated a model 12 (2023) 7:4www.hemaspherejournal.com incorporating 3 cytokines (IL-8, IL1β, and interferon [IFN] γ) that 86% for the HT10 score in our confirmation cohort indicates an predicted life-threatening infections with high sensitivity. In a pedi- attractive negative predictive value, with the patients presenting atric cohort, Diorio et al40 established that a normal to mildly ele- with a low baseline risk profile and continuously low serum PCT vated IFNγ in combination with an elevated IL1β was associated levels unlikely to develop a severe infection. Indeed, only 3% of with sepsis as opposed to CRS. Although IL1β was not part of our HT10low patients developed a severe nonviral infection during the Olink panel, we could not confirm alterations of IL-8 (q = 0.30), study period, and none were observed until day+19 (Figure 5E). or IFNγ (q = 0.93) in the patients developing severe infections in Concomitantly, this low-risk group may be spared unnecessary our LMM. However, we would note important cohort-level differ- antibiotic exposure during the vulnerable time period of peak ences that likely result in significant heterogeneity of the studied CAR T-cell expansion (typically weeks 1–2).74,75 patient populations. These relate to the endpoint (ie, life-threat- This study has several relevant limitations. It was retrospec- ening infection, sepsis), patient population (ie, adult versus pedi- tive, uncontrolled, and limited to a moderately sized number atric), timing of sample collection (event-triggered versus fixed of patients, which restrict drawing firm conclusions. Even time points), institutional choice of antimicrobial prophylaxis, though the infection rate was very low in the HT10low group, as well as the respective CAR target antigen (ie, BCMA, CD22). it is unclear if that would extend to patients that do no receive Moreover, almost all patients received IL-6-receptor blockade in early anti-infective therapy in case of CRS, as is currently the our study, reflecting evolving management paradigms. standard-of-care.23 As a result, the HT10 score needs to be Interestingly, the exploratory serum proteomic analysis revealed validated in a prospective manner, ideally at time of first fever, that the disruption of endothelial markers was already present before it is implemented in routine clinical use. Furthermore, the prior to CAR-T infusion, which was especially prominent for the assays used to determine PCT and other inflammatory markers increase of the ratio of Angiopoietin-2 to Angiopoietin-1 (Ang-2/1 may differ from site to site, which may impact the determined ratio). Increased Ang-2/1 ratios have been extensively studied in marker thresholds. The use of high-dose glucocorticoids may diverse disease contexts including as a predictor of morbidity and have impacted cytokine marker measurements and the pro- mortality in sepsis, critical illness, hemolytic uremic syndrome, teomic analysis. Viral infections were not identified in the early and COVID-19.46–51 Ang-2 is stored and rapidly released upon post-CAR-T phase, and it is unlikely that the diagnostic utility stimulation from endothelial cell Weibel-Palade bodies, while of PCT extends to viral infections,76 though COVID-19 may Ang-1 promotes endothelial stability.52,53 In the context of CD19 represent an important exception.77–79 Most centers currently CAR-T, increased Ang-2/1 ratios have been linked to enhanced do not perform serial PCT measurements, making it hard to blood-brain-barrier permeability in patients developing severe neu- extrapolate our findings to larger patient numbers. Still, these rotoxicity,54 and represent a predictor of severe CRS.38 Notably, findings offer a rationale for longitudinal assessment of PCT, platelets can stabilize Ang-1,48 which may provide an explanation especially during the critical phases of CAR-T–related immu- as to why patients with severe baseline thrombocytopenia display a notoxicity (eg, day 1–10). If validated, we see several salient particularly high risk for infectious complications.23 Overall, these clinical applications. First, the score may guide the initiation of data support a model wherein CAR-T patients are potentially pre- IV broad-spectrum antibiotic treatment and avoid unnecessary disposed for infections due to heightened endothelial permeability, antibiotic use in patients with a very low risk of severe infec- downregulation of endothelial-associated antioxidant defenses, and tions (ie, low HT10 score, stable hemodynamics and respiratory glycocalyx shedding.55 This in turn may also result in malformation status).65,66 Reducing antibiotic exposure appears especially per- of neutrophil extracellular traps, an important mechanism of host tinent given recent evidence pointing towards the immunomod- defense against infection.56–59 Consistent with a response to micro- ulatory role of the gut microbiome during CAR-T therapy.33–35 bial exposure, we found that serum levels of the innate immune Second, the score may be incorporated into decision making regulators CX3CL1, CCL20, and CXCL13 were increased in the algorithms regarding in- versus outpatient management of severe infection group.60,61 This was particularly evident on day 14 CAR-T patients at time of first fever.36,80 Finally, future preven- (Figure 6A), by which timepoint the majority of infection events tive strategies may target the endothelial compartment or dys- had occurred. Furthermore, we found a significant alteration of the regulated immune response at an early time point as a means to CD40/CD40-L axis, which plays a vital role in host defense against prevent severe infectious complications in CAR T-cell patients. pathogens mainly via the regulation of the CD8+ cytotoxic T-cell In conclusion, severe infections and CRS present with distinct (CTL)–directed immune response.62 For example, the activation of inflammatory signatures following CD19 CAR-T. By incorporat- CD40-expressing CD8+ CTLs can be facilitated either directly by ing dynamic changes in serum inflammatory markers into base- CD40-L-expressing CD4+ T-cells or indirectly via dendritic cells.63,64 line prognostication models, early infections could be identified Although the identified soluble serum cytokines are promis- in the setting of coincident CRS, enabling more patient-specific ing, they currently are still experimental in nature, and need to toxicity management. be validated externally and prospectively, across multiple clinical settings. On the other hand, PCT represents an established sep- sis marker and is already available at most academic centers.65–68 ACKNOWLEDGMENTS PCT is a 116-aminoacid peptide liberated into the circulation First and foremost, the authors thank the patients and their families for mainly in response to a stimulus of bacterial infection, and carries their participation in this study. We would like to thank Simon Forsberg from the advantages of a wide biological range, short time of induc- Olink Proteomics for the help with the bioinformatic analysis. tion after bacterial stimulus, and long half-life.69 The identified discriminatory threshold of 1.5 µg/L is in line with the large meta-analysis of Wacker and colleagues67 outlining the utility of AUTHOR CONTRIBUTIONS PCT to distinguish sepsis from other inflammatory conditions, Conceptualization: KR, VLB; investigation: KR, GI, VB, LLC, SK, RH, AP, and is in accordance with consensus recommendations for antibi- NM, FH, LF, PK, CS, DMCdS, JLP, FM, AAM, MS, MvB-B, PB, MS, VLB; otic initiation.66 Notably, integrating the patient-individual base- formal analysis and visualization: KR, VLB; methodology: KR, VLB; writing line risk of infection enhanced the utility of PCT as a biomarker original draft: KR, VLB; writing review and editing: KR, GI, VB, MvB-B, to identify early infections in the setting of CRS. High CAR- AM, WB, PB, MS, VLB. All authors read and approved the final manuscript. HEMATOTOX scores confer an increased risk of prolonged severe neutropenia across disease entities (eg, LBCL, mantle cell lymphoma, multiple myeloma), and also reflect a pro-inflam- DATA AVAILABILITY matory state of the host, which in turn can exacerbate the risk For original data and material, please contact kai.rejeski@med.uni- of infectious complications.13,23,28,70,71–73 The high sensitivity of muenchen.de. 13 Rejeski et al Distinguishing Severe Infection From CRS After CD19 CAR-T DISCLOSURES 11. Dos Santos DMC, Rejeski K, Winkelmann M, et al. Increased visceral fat distribution and body composition impact cytokine release syndrome KR: Kite/Gilead - Research Funding and travel support; Novartis - Honoraria. onset and severity after CD19 CAR-T in advanced B-cell malignancies. VB: Novartis - Honoraria, Research Funding; Gilead - Consultancy, Research Haematologica. 2022;107:2096–2107. Funding; Celgene - Research Funding; Janssen - Honoraria, Research Funding; 12. Rejeski K, Kunz WG, Rudelius M, et al. Severe Candida glabrata pan- Roche - Research funding; Takeda - Research funding. GI: Consultancy and colitis and fatal Aspergillus fumigatus pulmonary infection in the setting Honoraria - Novartis, Roche, Kite/Gilead, Bristol-Myers Squibb, Abbvie, of bone marrow aplasia after CD19-directed CAR T-cell therapy - a case Janssen, Sandoz, Miltenyi, AstraZeneca. MD: Research Support - Abbvie, report. BMC Infect Dis. 2021;21:121. Bayer, BMS/Celgene, Kite/Gilead, Janssen, Roche; Honoraria - Amgen, Astra 13. Rejeski K, Perez Perez A, Sesques P, et al. CAR-HEMATOTOX: a model Zeneca, Kite/Gilead, Janssen, Lilly, Novartis, Roche; Consultancy - Astra for CAR T-cell related hematological toxicity in relapsed/refractory Zeneca, Beigene, BMS/Celgene, Kite/Gilead, Janssen, Novartis, Roche; Journal large B-cell lymphoma. Blood. 2021;138:2499–2513. Editor - HemaSphere. MvB-B: Consultancy, Research Funding and Honoraria 14. Logue JM, Zucchetti E, Bachmeier CA, et al. Immune reconstitution and - MSD Sharp & Dohme, Novartis, Roche, Kite/Gilead, Bristol-Myers Squibb, associated infections following axicabtagene ciloleucel in relapsed or Astellas, Mologen, and Miltenyi. PB declares having received honoraria refractory large B-cell lymphoma. Haematologica. 2020;106:978–986. from Allogene, Amgen, BMS/CELGENE, Jansen, Kite/Gilead, Incyte, Jazz 15. Jain T, Knezevic A, Pennisi M, et al. Hematopoietic recovery in patients Pharmaceuticals, Miltenyi Biomedicine, Novartis and Nektar. MS: Morphosys receiving chimeric antigen receptor T-cell therapy for hematologic - Research Funding; Novartis - Consultancy, Research Funding; Janssen - malignancies. Blood Adv. 2020;4:3776–3787. Consultancy; Seattle Genetics - Research Funding; AMGEN - Consultancy, 16. Rejeski K, Wu Z, Blumenberg V, et al. Oligoclonal T-cell expansion in a Honoraria, Research Funding; Celgene - Consultancy, Honoraria; Kite/ patient with bone marrow failure after CD19 CAR-T for richter trans- Gilead - Consultancy, Honoraria, Research Funding; Roche AG - Consultancy, formed DLBCL. Blood. 2022;140:2175–2179. Research Funding. VLB: AMGEN - Honoraria; Celgene - Research Funding; 17. Rejeski K, Burchert A, Iacoboni G, et al. Safety and feasibility of stem Pfizer - Honoraria; Kite/Gilead - Research Funding, Honoraria; Novaritis - cell boost as a salvage therapy for severe hematotoxicity after CD19 Honoraria, Consultancy/Advisory; BMS - Consultancy/Advisory; Takeda - CAR T-cell therapy. Blood Adv. 2022;6:4719–4725. Consultancy/Advisory. None of the mentioned conflicts of interest were related 18. Fried S, Avigdor A, Bielorai B, et al. Early and late hematologic tox- to financing of the content of this manuscript. All the other authors have no icity following CD19 CAR-T cells. Bone Marrow Transplant. conflicts of interest to disclose. 2019;54:1643–1650. 19. Park JH, Riviere I, Gonen M, et al. Long-term follow-up of CD19 CAR therapy in acute lymphoblastic leukemia. N Engl J Med. SOURCES OF FUNDING 2018;378:449–459. This work was supported by a grant within the Gilead Research 20. Hill JA, Seo SK. How I prevent infections in patients receiving CD19- Scholar Program (to KR, MS). Deutsche Forschungsgemeinschaft (DFG, targeted chimeric antigen receptor T cells for B-cell malignancies. Blood. German Research Foundation) research grant provided within the 2020;136:925–935. Sonderforschungbereich SFB-TRR 388/1 2021 – 452881907, and DFG 21. Hill JA, Li D, Hay KA, et al. Infectious complications of CD19-targeted research grant 451580403 (to MS). The work was further supported by the chimeric antigen receptor-modified T-cell immunotherapy. Blood. Bavarian Elite Graduate Training Network (to MS), the Wilhelm-Sander 2018;131:121–130. Stiftung (to MS, project no. 2018.087.1), the Else-Kröner-Fresenius Stiftung 22. Wudhikarn K, Pennisi M, Garcia-Recio M, et al. DLBCL patients treated (to MS), and the Bavarian Center for Cancer Research (BZKF). KR received with CD19 CAR T cells experience a high burden of organ toxicities but a fellowship from the School of Oncology of the German Cancer Consortium low nonrelapse mortality. Blood Adv. 2020;4:3024–3033. (DKTK). KR, VB, and VLB were funded by the Else Kröner Forschungskolleg 23. Rejeski K, Perez A, Iacoboni G, et al. The CAR-HEMATOTOX (EKFK) within the Munich Clinician Scientist Program (MCSP). risk-stratifies patients for severe infections and disease progression after CD19 CAR-T in R/R LBCL. J ImmunoTher Cancer. 2022;10:e004475. 24. Gudiol C, Lewis RE, Strati P, et al. Chimeric antigen receptor T-cell ther- apy for the treatment of lymphoid malignancies: is there an excess risk REFERENCES for infection? Lancet Haematol. 2021;8:e216–e228. 1. Locke FL, Ghobadi A, Jacobson CA, et al. Long-term safety and activ- 25. Wudhikarn K, Palomba ML, Pennisi M, et al. Infection during the first ity of axicabtagene ciloleucel in refractory large B-cell lymphoma year in patients treated with CD19 CAR T cells for diffuse large B cell (ZUMA-1): a single-arm, multicentre, phase 1-2 trial. Lancet Oncol. lymphoma. Blood Cancer J. 2020;10:79. 2019;20:31–42. 26. Fajgenbaum DC, June CH. Cytokine Storm. N Engl J Med. 2. Schuster SJ, Bishop MR, Tam CS, et al. Tisagenlecleucel in adult 2020;383:2255–2273. relapsed or refractory diffuse large B-Cell lymphoma. N Engl J Med. 27. Juluri KR, Wu V, Voutsinas JM, et al. Severe cytokine release syndrome 2019;380:45–56. is associated with hematologic toxicity following CD19 CAR T-cell ther- 3. Wang M, Munoz J, Goy A, et al. KTE-X19 CAR T-cell therapy apy. Blood Adv. 2021;6:2055–2068. in relapsed or refractory mantle-cell lymphoma. N Engl J Med. 28. Taplitz RA, Kennedy EB, Bow EJ, et al. Antimicrobial prophylaxis for 2020;382:1331–1342. adult patients with cancer-related immunosuppression: ASCO and IDSA 4. Crump M, Neelapu SS, Farooq U, et al. Outcomes in refractory dif- clinical practice guideline update. J Clin Oncol. 2018;36:3043–3054. fuse large B-cell lymphoma: results from the international SCHOLAR-1 29. Spellberg B, Doi Y. The rise of fluoroquinolone-resistant esche- study. Blood. 2017;130:1800–1808. richia coli in the community: scarier than we thought. J Infect Dis. 5. Bethge WA, Martus P, Schmitt M, et al. GLA/DRST real-world outcome 2015;212:1853–1855. analysis of CAR-T cell therapies for large B-cell lymphoma in Germany. 30. Lautenbach E, Metlay JP, Bilker WB, et al. Association between fluo- Blood. 2022;140:349–358. roquinolone resistance and mortality in Escherichia coli and Klebsiella 6. Nastoupil LJ, Jain MD, Feng L, et al. Standard-of-care axicabtagene cilo- pneumoniae infections: the role of inadequate empirical antimicrobial leucel for relapsed or refractory large B-cell lymphoma: results from the therapy. Clin Infect Dis. 2005;41:923–929. US lymphoma CAR T consortium. J Clin Oncol. 2020;38:3119–3128. 31. Trecarichi EM, Tumbarello M, Spanu T, et al. Incidence and clini- 7. Jacobson CA, Locke FL, Ma L, et al. Real-world evidence of axicabta- cal impact of extended-spectrum-beta-lactamase (ESBL) production gene ciloleucel for the treatment of large B cell lymphoma in the United and fluoroquinolone resistance in bloodstream infections caused by States. Transplant Cell Ther. 2022;28:581.e1–581.e8. Escherichia coli in patients with hematological malignancies. J Infect. 8. Karschnia P, Rejeski K, Winkelmann M, et al. Toxicities and response 2009;58:299–307. rates of secondary CNS lymphoma after adoptive immunotherapy 32. Bow EJ. Fluoroquinolones, antimicrobial resistance and neutropenic with CD19-directed chimeric antigen receptor T cells. Neurology. cancer patients. Curr Opin Infect Dis. 2011;24:545–553. 2022;98:884–889. 33. Schubert ML, Rohrbach R, Schmitt M, et al. The potential role of 9. Morris EC, Neelapu SS, Giavridis T, et al. Cytokine release syndrome the intestinal micromilieu and individual microbes in the immuno- and associated neurotoxicity in cancer immunotherapy. Nat Rev biology of chimeric antigen receptor T-cell therapy. Front Immunol. Immunol. 2021;22:85–96. 2021;12:670286. 10. Lee DW, Santomasso BD, Locke FL, et al. ASTCT consensus grading 34. Smith M, Dai A, Ghilardi G, et al. Gut microbiome correlates of for cytokine release syndrome and neurologic toxicity associated with response and toxicity following anti-CD19 CAR T cell therapy. Nat immune effector cells. Biol Blood Marrow Transplant. 2019;25:625–638. Med. 2022;28:713–723. 14 (2023) 7:4www.hemaspherejournal.com 35. Blumenberg V, Busch G, Baumann S, et al. High bacterial abundances granulomatous disease after gene therapy is calprotectin-dependent. J of dorea and pediococcus in the gut microbiome linked to expan- Allergy Clin Immunol. 2011;127:1243–52.e7. sion, immune checkpoint expression and efficacy of CD19-directed 60. Gasteiger G, D’Osualdo A, Schubert DA, et al. Cellular innate immunity: CAR T-cells in patients with r/r DLBCL. Blood. 2021;138(Suppl an old game with new players. J Innate Immun. 2017;9:111–125. 1):2792–2792. 61. Vilgelm AE, Richmond A. Chemokines modulate immune surveillance 36. Myers GD, Verneris MR, Goy A, et al. Perspectives on outpatient admin- in tumorigenesis, metastasis, and response to immunotherapy. Front istration of CAR-T cell therapy in aggressive B-cell lymphoma and acute Immunol. 2019;10:333. lymphoblastic leukemia. J ImmunoTher Cancer. 2021;9:e002056. 62. Elgueta R, Benson MJ, de Vries VC, et al. Molecular mechanism and 37. Teachey DT, Lacey SF, Shaw PA, et al. Identification of predictive bio- function of CD40/CD40L engagement in the immune system. Immunol markers for cytokine release syndrome after chimeric antigen recep- Rev. 2009;229:152–172. tor T-cell therapy for acute lymphoblastic leukemia. Cancer Discov. 63. Ahmed KA, Wang L, Munegowda MA, et al. Direct in vivo evi- 2016;6:664–679. dence of CD4+ T cell requirement for CTL response and mem- 38. Hay KA, Hanafi LA, Li D, et al. Kinetics and biomarkers of severe cyto- ory via pMHC-I targeting and CD40L signaling. J Leukoc Biol. kine release syndrome after CD19 chimeric antigen receptor-modified 2012;92:289–300. T-cell therapy. Blood. 2017;130:2295–2306. 64. Bourgeois C, Rocha B, Tanchot C. A role for CD40 expression 39. Luo H, Wang N, Huang L, et al. Inflammatory signatures for quick on CD8+ T cells in the generation of CD8+ T cell memory. Science. diagnosis of life-threatening infection during the CAR T-cell therapy. J 2002;297:2060–2063. ImmunoTher Cancer. 2019;7:271. 65. Schuetz P, Briel M, Christ-Crain M, et al. Procalcitonin to guide ini- 40. Diorio C, Shaw PA, Pequignot E, et al. Diagnostic biomarkers to differ- tiation and duration of antibiotic treatment in acute respiratory entiate sepsis from cytokine release syndrome in critically ill children. infections: an individual patient data meta-analysis. Clin Infect Dis. Blood Adv. 2020;4:5174–5183. 2012;55:651–662. 41. Shimabukuro-Vornhagen A, Godel P, Subklewe M, et al. Cytokine 66. Schuetz P, Wirz Y, Sager R, et al. Procalcitonin to initiate or discontinue release syndrome. J ImmunoTher Cancer. 2018;6:56. antibiotics in acute respiratory tract infections. Cochrane Database Syst 42. Young JH, Logan BR, Wu J, et al. Infections after transplantation of Rev. 2017;10:CD007498. bone marrow or peripheral blood stem cells from unrelated donors. Biol 67. Wacker C, Prkno A, Brunkhorst FM, et al. Procalcitonin as a diagnostic Blood Marrow Transplant. 2016;22:359–370. marker for sepsis: a systematic review and meta-analysis. Lancet Infect 43. Greenbaum U, Strati P, Saliba RM, et al. CRP and ferritin in addi- Dis. 2013;13:426–435. tion to the EASIX score predict CAR-T-related toxicity. Blood Adv. 68. Assicot M, Gendrel D, Carsin H, et al. High serum procalci- 2021;5:2799–2806. tonin concentrations in patients with sepsis and infection. Lancet. 44. Filbin MR, Mehta A, Schneider AM, et al. Longitudinal proteomic analy- 1993;341:515–518. sis of severe COVID-19 reveals survival-associated signatures, tissue-spe- 69. Dandona P, Nix D, Wilson MF, et al. Procalcitonin increase after cific cell death, and cell-cell interactions. Cell Rep Med. 2021;2:100287. endotoxin injection in normal subjects. J Clin Endocrinol Metab. 45. Nahi H, Chrobok M, Meinke S, et al. Autologous NK cells as consoli- 1994;79:1605–1608. dation therapy following stem cell transplantation in multiple myeloma. 70. Locke FL, Rossi JM, Neelapu SS, et al. Tumor burden, inflammation, Cell Rep Med. 2022;3:100508. and product attributes determine outcomes of axicabtagene ciloleucel in 46. Page AV, Tarr PI, Watkins SL, et al. Dysregulation of angiopoietin 1 and large B-cell lymphoma. Blood Adv. 2020;4:4898–4911. 2 in Escherichia coli O157:H7 infection and the hemolytic-uremic syn- 71. Jain MD, Zhao H, Wang X, et al. Tumor interferon signaling and sup- drome. J Infect Dis. 2013;208:929–933. pressive myeloid cells are associated with CAR T-cell failure in large 47. Mikacenic C, Hahn WO, Price BL, et al. Biomarkers of endothelial acti- B-cell lymphoma. Blood. 2021;137:2621–2633. vation are associated with poor outcome in critical illness. PLoS One. 72. Rejeski K, Hansen DK, Bansal R, et al. The CAR-Hematotox Score As 2015;10:e0141251. a Prognostic Model of Toxicity and Response in Patients Receiving 48. Page AV, Liles WC. Biomarkers of endothelial activation/dysfunction in BCMA-Directed CAR-T for Relapsed/Refractory Multiple Myeloma. infectious diseases. Virulence. 2013;4:507–516. Blood. 2022;140(Suppl 1):7506–7508. 49. Ricciuto DR, dos Santos CC, Hawkes M, et al. Angiopoietin-1 and 73. Rejeski K, Wang Y, Albanyan O, et al. The CAR-Hematotox Score angiopoietin-2 as clinically informative prognostic biomarkers of mor- Identifies Patients at High Risk for Hematological Toxicity, Infections bidity and mortality in severe sepsis. Crit Care Med. 2011;39:702–710. and Poor Clinical Outcomes Following Brexucabtagene Autoleucel in 50. Smadja DM, Guerin CL, Chocron R, et al. Angiopoietin-2 as a marker Relapsed/Refractory Mantle Cell Lymphoma. Blood. 2022;140(Suppl of endothelial activation is a good predictor factor for intensive care unit 1):651–653. admission of COVID-19 patients. Angiogenesis. 2020;23:611–620. 74. Peinelt A, Bremm M, Kreyenberg H, et al. Monitoring of circulating 51. Villa E, Critelli R, Lasagni S, et al. Dynamic angiopoietin-2 assess- CAR T cells: validation of a flow cytometric assay, cellular kinetics, ment predicts survival and chronic course in hospitalized patients with and phenotype analysis following tisagenlecleucel. Front Immunol. COVID-19. Blood Adv. 2021;5:662–673. 2022;13:830773. 52. Fiedler U, Scharpfenecker M, Koidl S, et al. The Tie-2 ligand angiopoie- 75. Ayuk FA, Berger C, Badbaran A, et al. Axicabtagene ciloleucel in vivo tin-2 is stored in and rapidly released upon stimulation from endothelial expansion and treatment outcome in aggressive B-cell lymphoma in a cell Weibel-Palade bodies. Blood. 2004;103:4150–4156. real-world setting. Blood Adv. 2021;5:2523–2527. 53. Brindle NP, Saharinen P, Alitalo K. Signaling and functions of angiopoi- 76. Kamat IS, Ramachandran V, Eswaran H, et al. Procalcitonin to dis- etin-1 in vascular protection. Circ Res. 2006;98:1014–1023. tinguish viral from bacterial pneumonia: a systematic review and 54. Gust J, Hay KA, Hanafi LA, et al. Endothelial activation and blood- meta-analysis. Clin Infect Dis. 2020;70:538–542. brain barrier disruption in neurotoxicity after adoptive immunotherapy 77. Hu R, Han C, Pei S, et al. Procalcitonin levels in COVID-19 patients. Int with CD19 CAR-T cells. Cancer Discov. 2017;7:1404–1419. J Antimicrob Agents. 2020;56:106051. 55. Ince C, Mayeux PR, Nguyen T, et al. The endothelium in sepsis. Shock. 78. Tong-Minh K, van der Does Y, Engelen S, et al. High procalcitonin levels 2016;45:259–270. associated with increased intensive care unit admission and mortality in 56. Camicia G, Pozner R, de Larranaga G. Neutrophil extracellular traps in patients with a COVID-19 infection in the emergency department. BMC sepsis. Shock. 2014;42:286–294. Infect Dis. 2022;22:165. 57. Fuchs TA, Abed U, Goosmann C, et al. Novel cell death program leads 79. Kaal A, Snel L, Dane M, et al. Diagnostic yield of bacteriologi- to neutrophil extracellular traps. J Cell Biol. 2007;176:231–241. cal tests and predictors of severe outcome in adult patients with 58. Bianchi M, Hakkim A, Brinkmann V, et al. Restoration of NET for- COVID-19 presenting to the emergency department. Emerg Med J. mation by gene therapy in CGD controls aspergillosis. Blood. 2021;38:685–691. 2009;114:2619–2622. 80. Borogovac A, Keruakous A, Bycko M, et al. Safety and feasibility of out- 59. Bianchi M, Niemiec MJ, Siler U, et al. Restoration of anti-Asper- patient chimeric antigen receptor (CAR) T-cell therapy: experience from gillus defense by neutrophil extracellular traps in human chronic a tertiary care center. Bone Marrow Transplant. 2022;57:1025–1027. 15