Title: Cytokine release syndrome is not usually caused by secondary hemophagocytic lymphohistiocytosis in a cohort of 19 critically ill COVID-19 patients

Severe COVID-19 associated respiratory failure, poses the one challenge of our days. Assessment and treatment of COVID-19 associated hyperinflammation may be key to improve outcomes. It was speculated that in subgroups of patients secondary hemophagocytic lymphohistiocytosis (sHLH) or cytokine release syndrome (CRS) with features of macrophage activation syndrome might drive severe disease trajectories. If confirmed, profound immunosuppressive therapy would be a rationale treatment approach. Over a median observation period of 11 (IQR: 8; 16) days, 19 consecutive confirmed severe COVID-19-patients admitted to our intensive-care-unit were tested for presence of sHLH by two independent experts. HScores and 2004-HLH diagnostic criteria were assessed. Patients were grouped according to short-term clinical courses: discharge from ICU versus ongoing ARDS or death at time of analysis. The median HScore at admission was 157 (IQR: 98;180), without the key clinical triad of HLH, i.e. progressive cytopenia, persistent fever and organomegaly. Independent expert chart review revealed the absence of sHLH in all cases. No patient reached more than 3/6 of modified HLH 2004 criteria. Nevertheless, patients presented hyperinflammation with peripheral neutrophilic signatures (neutrophil/lymphocyte-ratio > 3.5). The latter best paralleled their short-term clinical courses, with declining relative neutrophil numbers prior to extubation (4.4, [IQR: 2.5;6.3]; n = 8) versus those with unfavourable courses (7.6, [IQR: 5.2;31], n = 9). Our study rules out virus induced sHLH as the leading cause of most severe-COVID-19 trajectories. Instead, an associated innate neutrophilic hyperinflammatory response or virus-associated-CRS appears dominant in patients with an unfavourable clinical course. Therapeutic implications are discussed.


Scientific Reports
| (2020) 10:18277 | https://doi.org/10.1038/s41598-020-75260-w www.nature.com/scientificreports/ Yet, in cases of serious COVID-19 with pneumonia and acute respiratory distress syndrome (ARDS), tocilizumab, a humanized monoclonal Interleukin-6-receptor antagonist, is administered in a phase II open label clinical trial (NCT04317092) to supress hyperinflammation that is assumed to cause fatal lung and multiorgan injury 6,7 . This is supported by cohort data from China, which report increased ferritin, and elevated TNF-α, IL-6 and IFN-γ levels in severe versus mild COVID-19-cases. Based on these preliminary data, cytokine release syndrome (CRS) or virus induced adult secondary hemophagocytic lymphohistocytosis (sHLH) have been proposed as underlying aetiology of severe COVID-19 7,8 . The latter is a clinical syndrome, characterized by massive systemic inflammation, persistent fever flares, hepatosplenomegaly, and severe cytopenia, which is rooted in hemophagocytic activity in bone marrow. It is a rare and often fatal clinical syndrome which can be observed in the context of malignancy, systemic autoimmunity, and viral infections such as Epstein-Barr virus 9 . Full blown sHLH in the latter context usually requires cytotoxic agents, i.e. etoposide, which is not to be used carelessly 9 .
The term CRS originally defines drug toxicity, which can be observed subsequently to monoclonal antibody or adoptive T-cell therapies. The syndrome is rooted in massive aberrant B-, T-cell and monocyte activation and a systemic "cytokine storm". The clinical course is pleiotropic, with fever, skin rash, cytopenia, neurologic symptoms, coagulopathy, hypotension and eventually organ failure i.e. ARDS. Application of tocilizumab usually resolves symptoms 10,11 . In the sense of an exuberant anti-viral innate immune response, elevated cytokine levels, fever and CRS-like symptoms, including organ injury have been reported in the context of viral infections, such as H5N1, SARS-CoV-1 and SARS-CoV-2 14,1213 . To avoid confusion, we will refer to this with the term "virus-associated CRS".
Whether "virus-associated" CRS, sHLH or both are present in severe COVID-19 profoundly impacts our understanding of the disease and impacts on therapeutic strategies. We therefore screened all consecutive 19 severe COVID-19-patients, who were admitted to our intensive care unit (ICU) for (a) CRS and (b) presence of sHLH, using the HScore, and 2004-HLH-diagnostic-criteria. Charts were reviewed by two independent HLHexperienced clinicians to rule in or out actual sHLH.

Results
Baseline characteristics, treatment and clinical course of severe COVID-19-cases. 19 consecutive laboratory confirmed COVID-19-patients with typical chest X-ray (n = 1) or CT-scan (n = 18) were admitted to our ICU 15,16 . Overall, they were mostly male (89%), with a median age of 70 (IQR: 58; 78) years. All had at least one pre-existing comorbidity or history of organ failure, with arterial hypertension (58%) being the most frequent comorbidity.
Patients were treated in accordance with the "surviving sepsis campaign guidelines for severe COVID-19" 17 . All patients received intravenous broad-spectrum antibiotics and regular surveillance for superinfection. Two patients additionally received remdesivir within clinical trials. Most patients (89%) showed rapid deterioration of oxygenation and were intubated after a median of 0 (IQR: 0;0) days.
Our approach resulted in clinical improvement of 47% of patients. After a median of 8.5 (IQR 8;12.5) days, eight patients were extubated, one patient was discharged without need for mechanical ventilation. The remaining cases were judged as unfavourable, due to prolonged weaning, progressive ARDS requiring organ support (n = 7), or death (n = 3) (see Fig. 1 mmHg, p = 0.04) was recorded in the "favourable group" during the observation period. Kidney injury (70% versus 22%, p = 0.04), liver failure (60% versus 0%, p = 0.01) and need for organ support were more likely in those who took an unfavourable clinical course (60% versus 11%, p = 0.03). Of note, requirement of organ support did not per se preclude a favourable course ( Fig. 1 -ID-6). For more details see Table 1 and Fig. 1 Fig. 2d). Immunophenotypic data revealed rather low numbers of circulating CD8 + T-cell subsets with a trend towards lower values in the unfavourable group (supplementary Table 1). Table 1. Cohort characteristics at time of ICU-admission and clinical course. (Sub)groups: "Overall", "favourable" versus "unfavourable" clinical group. Statistical comparisons were done using ANOVA-, Mann-Whitney-and Kruskal-Wallis-test as appropriate comparing "favourable" versus "unfavourable" clinical groups. *Comorbidities were defined as follows: Pulmonary disease: Asthma, chronic obstructive pulmonary disease or fibrotic lung disease; Cardiovascular disease (CVD): Pulmonary artery embolism, peripheral arterial occlusive disease, history of lung oedema; Atrial fibrillation (AF); Liver disease: history of gastrointestinal bleeding, liver-cirrhosis or pancreatitis;** Ulcerative colitis; ***prior immunosuppression included recent history of stem-cell transplantation, intake of cyclosporine and low dose steroids. Abbreviation: Advanced organ support (ADVOS), acute kidney injury -transient / requiring dialysis (AKI), Aspartateaminotransferase (AST), chronic kidney disease (CKD), extracorporeal membrane oxygenation (ECMO), not applicable (n.a.), sustained low efficiency dialysis (SLED), Sepsis-related organ failure assessment score (SOFA).

Parameter
Overall (n = 19) Favourable (n = 9) Unfavourable course (n = 10)  Table 2). Of note, these intergroup differences could not be replicated in absolute cell numbers of the respective cell populations, likely due to the pronounced interindividual variation in leukocytes (supplementary Fig. 1). After the first week of intubation, patients in the favourable group separated and showed significantly decreased relative neutrophilia, lymphopenia and NLR the day prior to successful extubation, whereas NLR remained increased in the unfavourable group (favourable: 10 Fig. 2b-d). The decline of CRP over this period was less pronounced and only separated the groups prior to extubation (Fig. 2f). Interestingly, despite a declining trend over the first week, serum IL-6-levels did not separate favourable and unfavourable clinical courses (median at tube removal: 32 [IQR: 13;49], versus 65 [IQR: 52;150] pg/ml, p = 0.15; Fig. 2g). D-Dimers tended to increase in both groups (repeated-measure-ANOVA: p = 0.13 or 0.2, Fig. 2h). Thus, an acute neutrophilic response rather than serum mediator kinetics paralleled the clinical course in our patients. The lack of organomegaly and lymphopenia suggest that hyperinflammation is linked to innate immunity.

Assessment of virus induced HLH.
Due to fever, hyperinflammation and high mortality, we asked whether patients with severe COVID-19-disease and organ dysfunction suffer from virus induced secondary HLH.
Nevertheless, the HScore is not intended for and has never been validated for use in an ICU-setting 18 . Likewise, higher cut-off values for ferritin have been suggested 19,20 .
We therefore applied "modified"-"2004 HLH-diagnostic-criteria", which are considered positive in the presence of 4/6 of the following criteria: persistent fever, splenomegaly, persistent or progressive cytopenia, ferritin > 10.000 ng/ml, AST > 50 IU/l. As can be retrieved from Table 3 only one patient from the unfavourable group www.nature.com/scientificreports/ reached the modified cut-off for serum ferritin. The sustained fever-and cytopenia-criteria were reached by 3 (16%) and 6 (32%) patients, respectively. 13 (68%) patients showed hypertriglyceridemia (Table 3). Fever and cytopenia were judged in a strictly longitudinal manner. The remaining domains were identical to the HScoredomains (Table 3), this resulted in 0/19 patients fulfilling four of the "modified"-"2004-HLH-diagnostic-criteria". In line with this, sHLH usually implies massive activation of T-cellular immunity, which can be assayed via elevated sIL-2R 21 . Although 5 (26%) of our patients showed values above 2000 U/ml, none exceeded the limit of 10,000 IU/ml, which has high specificity for sHLH.
Most importantly, sHLH is a clinical syndrome, which is ultimately defined by the assessment of the experienced clinician. We therefore had all case-charts (supplementary file 1) reviewed by experts. Both reached agreement, that the diagnostic criteria for classical sHLH were not fulfilled in any of the presented cases. (Table 3). However, CRS-like hyperinflammation associated with COVID-19 was assessed positive in 11 (58%) cases. 8 (42%) patients were defined "uncertain" regarding the latter due to concomitant superinfection (Table 3).
In summary, none of the patients showed evidence for secondary HLH within a median of 11 (8;16) days of observation after ICU-admission, despite the presence of ARDS, systemic hyperinflammation and eventually organ failure. Thus, virus-associated CRS but not sHLH is frequently associated with severe COVID-19.

Discussion
Based on SARS-CoV-1 and preliminary results of recent anti-cytokine-targeted interventional studies, an exuberant immune response has already been extrapolated as a major cause of lung injury, organ failure and mortality in COVID-19 22 . In fact, hyperinflammation and hyperferritinemia are hallmarks of severe over moderate COVID-19 23,24 . Moreover, hypercoagulability causing thrombosis, a classical feature of secondary HLH, has been observed amongst around 30% of severe COVID-19 patients 12 . And more recent evidence suggests, that this hyperinflammation can replicate different facets of both sHLH and viral CRS in different patient groups and different timepoints of disease courses 12,23 . These observations gave rise to the idea, that "virus associated" CRS or even sHLH might drive severe COVID-19 7,25 .
Here we report on the clinical course and immunologic findings of 19 consecutive severe COVID-19-patients. We demonstrate that up to a median of 11 (8; 16) days post ICU-admission none of our patients showed classical evidence of sHLH, which practically excludes sHLH as an initial driver of severe COVID-19 in the majority of Table 2. Short-term trends of immunologic parameters after ICU-admission. Column 2-4: We report relevant clinical immunologic parameters at time of ICU-admission (ADM) and after *1 week (5-8), while patients that required intubation were still on mechanical ventilation. Paired t-tests, or paired Wilcoxon-rank-test, were used to test for intra-individual differences, as applicable (full cohort). Columns 5 and 6: Group differences (favourable versus unfavourable clinical courses) were assessed using ANOVA or Kruskal-Wallis-test. **Serum ferritin and sIL-2R receptor were assayed after a median of 2 days (missing data n = 1) post admission (ADM) and 4 days later -we report peak values. We report median and interquartile range (IQR) or frequencies as counts and percent of total. ANOVA was used for group comparisons: favourable versus unfavourable clinical course. Abbreviations: Interleukin (IL), C-reactive-protein (CRP), neutrophil/lymphocyte-ratio (NLR), soluble Interleukin 2 receptor (sIL-2R).
However, "absence of sHLH" is not contradictory to the hypothesis that virus-associated-CRS and hyperinflammation rather than viral replication determines the course of COVID-19 as some observations can hardly be explained by virus replication 24 . Cases from France indicate, that development of severe COVID-19-pneumonia can be accompanied by a decreasing viral load 33 . We and others find that deterioration of oxygenation occurs with some delay, about 3 days after hospital admission accompanied by the peak inflammatory reaction 6 . Thereby, elevated inflammatory mediators, a hypercoagulable state (D-Dimer) and elevated NLR were characteristics of the disease at onset of ARDS. Cumulatively, these data confirm the presence of a strong virus-associated CRS in SARS-CoV-2-infection, simultaneously to the onset of ARDS 6,24,34 . Our data adds to the current knowledge, that amongst severe COVID-19-patients, NLR and relative neutrophilia more than CRP paralleled the observed short-term clinical course. It is tempting to speculate that virus-associated CRS and the neutrophilic response are the main culprits in severe COVID-19-pneumonia, which occurs with a delay only when the "misguided anti-viral-immune response" hits 25 . One might argue, that ARDS per se implies alveolar endothelial injury, complement-, neutrophil-activation, and NETosis and thus entails a systemic hyperinflammatory response [35][36][37] . Still, others have reported, that an increased NLR is apparent already before the onset of ARDS, and was an independent predictor for severe COVID-19 in 61 patients 34,38 . In addition, Zheng M et al. reported functional exhaustion of cytotoxic T cell subsets upon SARS-CoV-2 infection 34 . This may not only hinder clearance of the virus but also override the cellular control mechanisms that normally inhibit the development of CRS 25,34 .
Considering these data, combined anti-viral and anti-inflammatory treatment strategies are seemingly plausible options to improve outcomes. Although we could not find an association of serum IL-6 with the clinical course of our patients and although experimental data indicate potentially harmful effects of IL-6R blockade in experimental lung injury there have been occasional reports of therapeutic success with tocilizumab in COVID-19 25,39 . In addition, both IL-6 and IL-1 were associated with outcomes in ARDS decades ago 40 . Still, we believe that a sense of proportion is required here.
In the absence of sHLH, cytoreductive drugs, i.e. etoposide, are not indicated. Further, we advocate caution when applying combined cytokine-directed treatment protocols in this vulnerable patient population. Our patients are elderly, carry comorbidities and are not immune to bacterial superinfections. The latter was also observed in a larger cohort 6 . After all, 9/19 patients improved and overcame viral-CRS by standard supportive care. Thus, treatment within clinical trials should be prioritized when using experimental immunomodulation. Moreover, short acting substances, e.g. IL-1R-blockade with anakinra for which positive data exist from other neutrophilic hyperinflammatory conditions such as Still's disease with ARDS and septic shock deserves special consideration, as it allows the therapy to be quickly de-escalated 41,42 . Also, Ruxolutinib a tyrosine kinase inhibitor interfering with Jak-STAT dependent cytokine signalling seems appealing (NCT 04338958). If neutrophils prove to be causally involved, colchicine could be another short-acting and cost-effective option for emerging countries, (NCT04322682) 43 .
Our short-term observational study has limitations. First, the overall sample size is relatively small with only a short follow up. In addition, 2 patients were transferred to our ward more than 48 h after external ICU admission, thus datasets for sIL-2R and ferritin were not complete. We can indeed exclude sHLH as a common underlying pathology in severe COVID-19 with ARDS. However, we cannot eliminate the possibility that some patients will develop secondary HLH later on, e.g. as a result of bacterial sepsis or prolonged viral-CRS 44 . Moreover, the judgment of the (un)favourable disease-courses based on clinical criteria reflects a rather weak clinical endpoint and holds risk for bias. Nevertheless, all patients classified as favourable were permanently discharged from ICU at time of manuscript preparation. Further, in clinical reality, therapeutic decisions will likely also depend on Figure 2. Immunologic parameters, neutrophilia and IL-6 in relation to clinical course at time of ICU admission, at day 4-7 (median 6 days)-when patients were still intubated-and prior to extubation if applicable (median 10 days). The median of 10 days differs from the median intubation time of 8.5 days (see main manuscript) since not all patients were intubated within 24 h post admission. Data from the unfavourable group (no extubation possible) were matched accordingly to achieve equivalent median days post ICU admission. Abbreviations: C-reactive-protein (CRP), Interleukin-(IL)-6, neutrophile/lymphocyte-ratio (NLR). This analysis includes 17 patients who had undergone intubation (1 patient each from the favourable and the unfavourable group were excluded, see methods section). Each datapoint represents one patient. Statistical significance for independent t-test between groups: n.s. = not significant, * = p < 0.05, ***p < 0.001. Statistical significance for repeated measure ANOVA † = not significant, § = p < 0.05, § § = p < 0.01, § § § = p < 0.001. Patients' characteristics were entered based on interviews with patients or families and the clinic's medial record system. BMI, Horovitz Index (HI) and sequential-organ-failure-assessment-(SOFA)-Score were calculated on admission as body weight [kg]/height 2 [m 2 ], paO2/FiO2 or using the online-calculator https ://mdcal c.com/seque ntial -organ -failu re-asses sment -sofa-score , respectively. Clinical data were independently reviewed by 2 researchers.
Severe COVID-19, shock and ARDS were defined according to the National Health Commission of China's and the World Health Organization's interims papers definition, respectively 45,46 . As observation periods were limited, favourable spontaneous clinical course of severe COVID-19-patients were defined by either "no need Table 3. Expert ratings, median HScores and modified HLH-2004 criteria fulfilled. The table reports median HScores for assumed negative and positive results for the bone marrow criterion. Groups: "overall", "favourable" versus "unfavourable" clinical group, statistical comparison was done comparing the latter two groups. Further frequencies of the HScore subdomains are reported. We do the same for the modified 2004 HLH criteria its subdomains. None of our experts detected evidence for HLH in any of the patients' charts reviewed. CRS like inflammation was rated, when there was evidence for inflammatory disease in the absence of bacterial coinfection. These cases were classified as undefined. Abbreviations: bone marrow (BM) biopsy; Triglycerides (TAG), Aspartate-amino-transferase (AST), hemophagocytic lymphohistiocytosis (HLH), soluble Interleukin-2 receptor (sIL-2R). *The results were identical, when adjusted laboratory cut offs for hypofibrinogenemia and abnormal AST were used.

Overall
Favourable (n = 9) Unfavourable (n = 10) www.nature.com/scientificreports/ for intubation" or "was extubated" at time of analysis versus those who were in need of mechanical ventilation, or in need of extracorporeal lung assist devices, prolonged weaning, or had died at time of analysis. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Scores and assessment for HLH. The HScore was calculated at time of ICU-admission and after 1 week as described by Fardet et al. (see Table 4 and supplementary Table 1, respectively 18 ). Yet, in an ICU-setting only a cut-off value for ferritin ≥ 10.000 ng/ml demonstrated sufficient sensitivity in these patients 19 . The bone marrow biopsy criterion was further shown to have poor specificity for HLH among seriously ill patients 47 . Thus, the pediatric 2004-HLH criteria, were modified according to centre specific cut-offs, and a cut-off for ferritin of 10.000 ng/ml was used ("modified 2004-HLH criteria"; Table 4). Chart reviews to determine actual presence of sHLH were independently performed by L.G. M.P. and L.R.P., the latter two having at least ten years of experience in diagnosis and treatment of secondary HLH.
Statistics. Was performed using IBM SPSS Version 23. We report median and interquartile range (IQR) or counts and percent of total (%) as applicable. Group comparisons (favourable versus unfavourable clinical group or overall cohort at ICU admission or after 1 week) were done using ANOVA, Mann-Whitney-U-, Kruskal-Wallis-, repeated measure paired ANOVA or paired-Wilcoxon-rank-test as applicable. Table 4. HScore criteria as reported by Fardet et al. 18 and "modified 2004-HLH-diagnostic-criteria". *We used centre specific cut-offs for definition of hypo-fibrinogenaemia, hypertriglyceridemia, leucopenia and neutropenia with respect to the "modified HLH2004 criteria". We used lower or upper laboratory specific reference ranges to define the adjusted cut offs. In addition the ferritin criterion was modified within the adjusted 2004 HLH-diagnostic guidelines to a cut off of ≥ 10.000 ng/ml, according to better specificity in ICUpatients 19 . The HScore is calculated as a sum of points -see Fardet et al. 18 . The sum-score can be transformed into a probability score of HLH -however, this has never been validated for assessment of HLH in an ICUsetting. Abbreviations: Aspartate-aminotransferase (AST), Haemoglobin (Hgb), triglycerides (TAG).