key: cord-0852614-9lsyjp6q authors: Thwaites, R.; Sanchez Sevilla Uruchurtu, A.; Siggins, M.; Liew, F.; Russell, C. D.; Moore, S.; Carter, E.; Abrams, S.; Short, C.-E.; Thaventhiran, T.; Bergstrom, E.; Gardener, Z.; Ascough, S.; Chiu, C.; Docherty, A. B.; Hunt, D.; Crow, Y.; Solomon, T.; Taylor, G.; Turtle, L.; Harrison, E. M.; Semple, M. G.; Baillie, J. K.; Openshaw, P. J. title: Elevated antiviral, myeloid and endothelial inflammatory markers in severe COVID-19 date: 2020-10-13 journal: nan DOI: 10.1101/2020.10.08.20209411 sha: 9127d5aaf765b79abf4cac90091ba520203d543b doc_id: 852614 cord_uid: 9lsyjp6q The mechanisms that underpin COVID-19 disease severity, and determine the outcome of infection, are only beginning to be unraveled. The host inflammatory response contributes to lung injury, but circulating mediators levels fall below those in classical cytokine storms. We analyzed serial plasma samples from 619 patients hospitalized with COVID-19 recruited through the prospective multicenter ISARIC clinical characterization protocol U.K. study and 39 milder community cases not requiring hospitalization. Elevated levels of numerous mediators including angiopoietin-2, CXCL10, and GM-CSF were seen at recruitment in patients who later died. Markers of endothelial injury (angiopoietin-2 and von-Willebrand factor A2) were detected early in some patients, while inflammatory cytokines and markers of lung injury persisted for several weeks in fatal COVID-19 despite decreasing antiviral cytokine levels. Overall, markers of myeloid or endothelial cell activation were associated with severe, progressive, and fatal disease indicating a central role for innate immune activation and vascular inflammation in COVID-19. The mechanisms that underpin COVID-19 disease severity, and determine the outcome of infection, 26 are only beginning to be unraveled. The host inflammatory response contributes to lung injury, but 27 circulating mediators levels fall below those in classical 'cytokine storms'. We analyzed serial plasma 28 samples from 619 patients hospitalized with COVID-19 recruited through the prospective 29 multicenter ISARIC clinical characterization protocol U.K. study and 39 milder community cases not 30 requiring hospitalization. Elevated levels of numerous mediators including angiopoietin-2, CXCL10, 31 and GM-CSF were seen at recruitment in patients who later died. Markers of endothelial injury 32 (angiopoietin-2 and von-Willebrand factor A2) were detected early in some patients, while 33 inflammatory cytokines and markers of lung injury persisted for several weeks in fatal COVID-19 34 despite decreasing antiviral cytokine levels. Overall, markers of myeloid or endothelial cell activation 35 were associated with severe, progressive, and fatal disease indicating a central role for innate 36 immune activation and vascular inflammation in Main text 38 Fatal COVID-19 is associated with acute respiratory distress syndrome and raised systemic 39 inflammatory markers including IL-6 and C-reactive protein, often accompanied by neutrophilia and 40 lymphopenia 1 . The beneficial effect of corticosteroid treatment in severe disease highlights the role 41 of steroid-responsive inflammation in pathogenesis 2, 3 , and post-mortem studies report pulmonary 42 vessel vasculitis (most commonly myeloid cells) and microthrombosis in fatal 5, 6, 7 . The 43 virus-induced inflammatory state has laboratory features that resemble secondary haemophagocytic 44 lymphohistiocytosis (sHLH) 8, 9, 10 but the exact pattern and severity of inflammatory responses has 45 been only partially characterized. Levels of some inflammatory mediators, including IL-6, are 46 elevated in COVID-19, but are typically ten times lower than those reported in acute respiratory 47 5 study enrollment was similar in all groups: Severity 3, 7 days; Severity 4, 9 days; Severity 5, 11 days; 75 Severity 6/7, 11 days; and Severity 8, 8 days. Some differences in routinely performed clinical 76 8 CSF was similarly elevated in all hospitalized groups, relative to controls and was most pronounced 151 in the groups 6/7 and 8 (Fig. 2h) . Numerous other inflammatory cytokines and chemokines showed 152 similar results including TNF-α, IL-2, GDF-15, G-CSF, and VEGF-D ( Supplementary Fig. 2) . RAGE/S100A12 has previously been characterized as a marker of respiratory damage in ARDS 30 and 154 indeed was elevated in groups 6/7 and 8 relative to most others (Fig. 2i) . The neutrophil chemokine 155 IL-8 (CXCL8) was similarly elevated in severe disease, as was the neutrophil gelatinase associated 156 lipocalin (LCN-2/NGAL) ( Supplementary Fig. 2 ), in line with the reported association between blood 157 neutrophilia and severity 16 also seen in this cohort ( Supplementary Fig. 1b) . 158 Other immunological mediators were not significantly different between 159 groups, indicating that only limited aspects of the immune repertoire were active in Interestingly, IL-4 levels were lower in the non-severe disease outcome groups (3, 4, and 5) relative 161 to both the control groups and the severe disease groups 6/7 and 8 ( Supplementary Fig. 2) , 162 indicating that suppression of the normal levels of type-2 cytokines may be associated with milder 163 COVID-19 disease, and that this mechanism is lost in severe disease. Similarly, IL-12p70, commonly 164 released by antigen presenting cells (APCs) 31 , was decreased in all hospitalized cases relative to the 165 HCs and group 1/2 ( Figure S2 ), possibly owing to the trafficking of APCs to the site(s) of viral 166 infection. 167 To determine the strength of the relationships between these individual plasma mediators we 168 performed a hierarchical correlation matrix analysis of mediators from plasma samples collected at 169 the time of study enrolment. This identified a strongly correlated cluster of inflammatory mediators 170 including GM-CSF, CXCL10, vWF-A2, and IL-6 ( Fig. 3a) ; increases in which were commonly associated 171 with the most severe COVID-19 outcome groups. Given the strong association between age and 172 COVID-19 severity 22 , and reports of increased inflammatory responses in males relative to females 173 with COVID-19 32 we investigated the influence of these demographic factors on plasma mediators 174 levels in hospitalized patients. As the major effect in our cluster analysis was severity (Fig. 1) , we 175 . CC-BY 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . further stratified each of these severity groups by age (≥ or < 70 years of age) and sex, to better 176 account for the influence of disease severity on plasma mediator levels. Following adjustment for 177 multiple testing, no mediator was found to be statistically different between males and females 178 within each outcome group ( Supplementary Fig. 3 ). By contrast, several differences were evident 179 between those aged ≥70 and <70 years, including elevated levels of D-dimer, CXCL10, and GM-CSF in 180 those aged ≥70 years; IFN-γ levels were, by contrast, greater in younger patients within severity 181 group 4 ( Fig. 3b and Supplementary Fig. 3) . 182 We next sought to determine the changes in levels of some key plasma mediators from the time of 183 enrolment over the course of disease, by relating data to the patient reported duration of symptoms 184 at the time of each sample collection, including consecutive samples collected from individual 185 patients. This analysis indicated that many mediators were stable over the time-course of 186 hospitalization, supporting the validity of using samples from the time of enrolment to study the 187 immunologic basis of COVID-19. However, some mediators did change over time; for example, there 188 was a gradual decrease in IFN-γ and CXCL10 over time in most groups ( Supplementary Fig. 4) , 189 including group 8 ( Fig. 4a and 4b , respectively). By contrast some other mediators remained 190 elevated or appeared to increase over the duration of symptoms in group 8, including angiopoietin-2 191 and D-dimer ( Fig. 4c and 4d , respectively). Similarly, the inflammatory mediators RAGE remained elevated or increased in group 8 in the latter stages of disease (Fig. 4e and 4f, 193 respectively). Together, these results indicated that the most severe outcomes of were associated with persistent coagulation and inflammation, even as IFN levels declined. 195 Finally, we hypothesized that differences in plasma mediator levels between patients with Severe 196 (groups 6/7 and 8) and Non-severe (groups 3, 4, and 5) COVID-19 would be apparent within the first 197 few days of symptoms. Indeed, within the first 4 days of symptoms several mediators were 198 significantly elevated in the Severe group, relative to Non-severe, including IL-2, respectively) , indicating a pronounced inflammatory response early in Severe 200 . CC-BY 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020 . . https://doi.org/10.1101 disease. Similarly, many markers of coagulation and endothelial injury were elevated in Severe, 201 relative to Non-severe, including D-dimer and vWF-A2 (P<0.0001, Fig. 4j and 4k, respectively), in 202 addition to angiopoietin-2 and IL-1α (which can be activated by thrombin 33 ) ( Supplementary Fig. 5) . 203 By comparison the lung damage-associated marker EN-RAGE 30 was not significantly different 204 between the Severe and Non-severe groups in the first 4 days of symptoms (P=0.098, 205 Supplementary Fig. 5 ). Together, these data indicated that severe COVID-19 is associated with 206 elevated levels of plasma mediators indicative of coagulation, endothelial activation and a broad 207 inflammatory response including CXCL10, GM-CSF, and IL-6. These differences were apparent within 208 the first days of symptoms, while markers of lung damage may only become elevated later in 209 disease, potentially indicating a pathological role for these processes and a window of opportunity 210 for early immunomodulation to prevent significant lung damage. 211 While markers of fibrinolysis have previously been associated with disease severity 16 and 212 thrombosis is common in severe and fatal COVID-19 4, 5, 20 the causes of this manifestation of severe 213 disease are not known. We demonstrate that increasing disease severity is associated with broad 214 elevations in inflammatory mediator levels, alongside a signature of endothelial injury. This signal 215 was most pronounced in fatal COVID-19 and was apparent even in the early stages of disease. 216 The elevation of angiopoietin-2, thrombomodulin, and vWF-A2 in fatal COVID-19 cases provides 217 evidence for the involvement of endothelial injury in COVID-19 severity. Endothelial injury following 218 inflammatory damage, including the increasingly recognized pulmonary artery vasculitis 4, 20 in 219 COVID-19, may result in the initiation of a pro-coagulant role for these cells 34 . Alternatively, this 220 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . fibrin rich thrombi 28 . Neutrophilic inflammation could have an etiological role in endothelial injury 226 though neutrophilia is predominantly a feature of the later phases of COVID-19 1 , while endothelial 227 injury was evident in the first days of symptoms. However, the continued thrombosis in late stage 228 fatal COVID-19 may result from neutrophil mediated coagulation, observed in other settings 35, 36, 37 229 and recently demonstrated in COVID-19 38 . Combined, these results indicate a multiplicity of possible 230 pro-coagulant triggers that may contribute to pathology at different stages of disease. 231 We found that the antiviral immune mediator CXCL10 and the myeloid cell growth factor GM-CSF, 232 were strikingly elevated in fatal cases of COVID-19. This is confirmed by a recent report describing 233 the potential utility of CXCL10 as an early prognostic marker of COVID-19 severity 39 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . While many cytokines and other inflammatory mediators were most significantly elevated in fatal 251 and critical COVID-19, these data do not necessarily support the concept of a "cytokine storm" in 252 13 . While some elements, such as elevated IL-6 and ferritin levels (reported in other 253 studies, but not seen here) 8, 9, 10 , are reminiscent of sHLH, the relatively gradual clinical progression 254 and persistent elevation of some cytokines, even during the early stages of symptomatic disease, are 255 uncommon amongst conditions associated with cytokine storms such as toxic-shock syndrome and 256 bacterial sepsis. 257 To our knowledge, this is to date the largest study of inflammatory responses in COVID-19. The 258 multicenter nature of ISARIC4C adds to the ability to interpret and apply these results to other 259 settings. However, further studies are needed to determine the prognostic value of these key plasma 260 biomarkers, including multivariable analyses of biological data alongside clinical and demographic 261 data. This detailed level of analysis may also enable the phenotyping of patients most likely to 262 respond to individual therapies. Future analyses should focus on the biological features of patients 263 that respond to therapeutic interventions, such as dexamethasone 2, 3 , to enable mechanistic insight 264 and targeting of treatment. The clear distinction between patients that would progress to severe 265 COVID-19 and those that would not, even in the earliest stages of disease, indicates that early 266 therapeutic intervention may be crucial to limit mortality. Overall, these data indicate an early 267 inflammatory response in COVID-19, most prominent in those who will later suffer severe or fatal 268 disease. These responses may enable the development of prognostic biomarkers, inform our 269 understanding of immunopathogenesis in COVID-19 and enable novel approaches for therapeutic 270 intervention. 271 . CC-BY 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . 14 To calculate partial HScores 24 , ferritin, triglyceride and AST measurements from this study were 294 combined with recorded results from case report forms for temperature and routine hemoglobin, 295 white cell counts, and platelet counts. 296 Immunoassays 297 IFN-γ, TNF-α, IL-1β, IL-2, IL-4, IL-6, CXCL8/IL-8, IL-10, IL-12p70 and IL-13 were quantified using MSD 298 All values at or below the lower limit of detection (LLOD) were replaced with the geometric mean of 306 the lower limits of detection across plates for each assay. 307 Statistical analyses used GraphPad Prism v8.3.0 (GraphPad, La Jolla, California, USA) R version 3.6.1 309 and Python 3.7.3 with Pandas 1.0.3 and Seaborn 0.10.0. Non-parametric mediator data (as 310 determined by D'Agostino and Pearson normality test) were analyzed by ANOVA using Kruskal-Wallis 311 tests with Dunn's test for multiple comparisons of patient groups within in time group. Non-312 parametric two-way analyses were performed using Mann-Whitney U tests. Correlation matrix 313 analysis was performed using the R packages ggplot2 and ggcorrplot and Spearman's test for 314 correlation of non-parametric data, after P-value adjustment for multiple testing. The false discovery 315 rate, or expected proportion of discoveries which are falsely rejected, was controlled using the 316 methods of Benjamini and Hochberg. Heatmaps of scaled plasma mediator data were generated 317 using the ComplexHeatmap package in R with rows and columns split by K-means clustering and 318 . CC-BY 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. clustering. Each patients' column is additionally annotated with data on disease outcome 327 ("Severity") as one of the following outcome groups: not requiring oxygen support ('3', n=128), 328 requiring oxygen via a face mask ('4', n=103), requiring non-invasive ventilation or high-flow nasal 329 canulae ('5', n=78), requiring invasive mechanical ventilation ('6/7', n=87) or fatal disease ('8', n=69) . 330 Columns are additionally annotated with patient age, sex and duration of illness at the time of 331 sample collection ("Onset"). 332 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . https://doi.org/10.1101/2020.10.08.20209411 doi: medRxiv preprint RAGE/S100A12. Data were analyzed for statistical significance using Kruskal-Wallis tests with Dunn's 343 tests for multiple comparisons between all groups. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. 344 Correlogram of the association between plasma mediator levels at the time of enrolment in all 346 patients hospitalized with COVID-19 (n=465). b) Inflammatory mediator levels within an outcome 347 group, stratified as those ≥ or < than 70 years of age. Data in panel a were analyzed using 348 Spearman's rank correlations with correction for multiple testing; significant correlations are 349 denoted by a circle, the color of which denotes the Spearman's R value. Data in panel b were 350 analyzed using Mann-Whitney U tests with P-value adjustment for false discovery rate. 351 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . https://doi.org/10.1101/2020.10.08.20209411 doi: medRxiv preprint COVID-19 that would: not require oxygen support ('3', n=9-93); require an oxygen face mask ('4', 368 n=22-71); require non-invasive ventilation or high-flow nasal cannulae ('5', n=15-63); require 369 invasive mechanical ventilation ('6/7', n=19-91); or progress to fatal disease ('8', n=15-63) each plasma mediator between males and females ("Sex") and those aged ≥70 years and <70 years 382 ("Age") within each disease outcome group ('8'=Red, '6/7'=Orange, '5'=Purple, '4'=Dark blue, 383 '3'=Cyan). Data were analyzed using Mann-Whitney U tests with P-value adjustment for false 384 discovery rate. 385 Supplementary Fig. 4 -Longitudinal analysis of selected plasma mediators within each disease 386 outcome group. All data within each severity group was related to the duration of symptoms at the 387 time of sample collection ("Onset to sample", measured in days) for each plasma mediator. 388 Generalized additive modelling was used to fit a restricted cubic spline which is plotted together 389 with the standard error (grey). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . https://doi.org/10.1101/2020.10.08.20209411 doi: medRxiv preprint patients in the groups 6/7 or 8 ("Severe", n=22) and groups 3, 4, or 5 ("Non-Severe", n=54 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020 . . https://doi.org/10.1101 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020 . . https://doi.org/10.1101 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 13, 2020. . https://doi.org/10.1101/2020.10.08.20209411 doi: medRxiv preprint Clinical features of patients infected with 2019 novel coronavirus in Wuhan, 489 China Dexamethasone in Hospitalized Patients with Covid-19 -Preliminary 492 Report Association Between Administration of Systemic 495 Corticosteroids and Mortality Among Critically Ill Patients With COVID-19: A Meta-analysis Tissue-specific tolerance in fatal Covid-19. medRxiv Pulmonary Vascular Endothelialitis, Thrombosis, and Angiogenesis in 501 Covid-19 Association of COVID-19 inflammation with activation of the C5a-C5aR1 axis Histopathological findings and viral tropism in UK patients with severe fatal 507 COVID-19: a post-mortem study Pulmonary retention of primed neutrophils: a novel protective host 616 response, which is impaired in the acute respiratory distress syndrome Tocilizumab in Hospitalized Patients With COVID-19 Pneumonia GM-CSF blockade with mavrilimumab in severe COVID-19 pneumonia and 623 systemic hyperinflammation: a single-centre, prospective cohort study GM-CSF Neutralization With Lenzilumab in Severe COVID-19 Pneumonia: 627 A Case-Control Study Therapeutic blockade of granulocyte macrophage colony-stimulating factor 630 in COVID-19-associated hyperinflammation: challenges and opportunities Pathogenic T-cells and inflammatory monocytes incite inflammatory storms in 634 severe COVID-19 patients Open source clinical science for emerging infections ISARIC4C. ISARIC Coronavirus Clinical Characterisation Consortium. 2020 mice: Multivariate Imputation by Chained 643 Equations in R