Markers of NETosis and DAMPs are altered in critically ill COVID-19 patients

Background Coronavirus disease 19 (COVID-19) is known to present with disease severities of varying degree. In its most severe form, infection may lead to respiratory failure and multi-organ dysfunction. Here we study the levels of extracellular histone H3 (H3), neutrophil elastase (NE) and cfDNA in relation to other plasma parameters, including the immune modulators GAS6 and AXL, ICU scoring systems and mortality in patients with severe COVID-19. Methods We measured plasma H3, NE, cfDNA, GAS6 and AXL concentration in plasma of 83 COVID-19-positive and 11 COVID-19-negative patients at admission to the Intensive Care Unit (ICU) at the Uppsala University hospital, a tertiary hospital in Sweden and a total of 333 samples obtained from these patients during the ICU-stay. We determined their correlation with disease severity, organ failure, mortality and other blood parameters. Results H3, NE, cfDNA, GAS6 and AXL were increased in plasma of COVID-19 patients compared to controls. cfDNA and GAS6 decreased in time in in patients surviving to 30 days post ICU admission. Plasma H3 was a common feature of COVID-19 patients, detected in 40% of the patients at ICU admission. Although these measures were not predictive of the nal outcome of the disease, they correlated well with parameters of tissue damage (H3 and cfDNA) and neutrophil counts (NE). A subset of samples displayed H3 processing, possibly due to proteolysis. Conclusions Elevated H3 and cfDNA levels in COVID-19 patients illustrate the severity of the cellular damage observed in critically ill COVID-19 patients. The increase in NE indicates the important role of neutrophil response and the process of NETosis in the disease. GAS6 appears as part of an early activated mechanism of response in Covid-19.

mild reduction in platelet count and high values of D-dimer seen in COVID-19, pointing at a somewhat different pathological mechanism. The involvement of immune regulatory and hemostatic pathways appears evident.
Given the systemic infection and activation of the innate immune system in severe COVID-19 patients and the organ injury, in particular in the lungs, cellular components released upon cellular disruption or neutrophil activation may contribute to COVID-19 disease. This is in line with the observation that in the acute respiratory distress syndrome, the involvement of the innate immune systems' neutrophil extracellular traps, NETs was previously shown to occur and to contribute to disease progress [18].
Extracellular histones are cytotoxic proteins that originate from the activation of neutrophils during NETosis [19][20][21] or from damaged tissues [22], while cell free DNA (cfDNA) and the protease neutrophil elastase (NE) are released concomitantly [23]. Histones and cfDNA, are known to activate Toll-like receptors (TLRs) and promote proin ammatory cytokine release via receptor-dependent and independent pathways [24,25]. Extracellular histones and NE have antimicrobial activities and aid in the killing of pathogens. However, while serving a protective function, NETs are potentially harmful to the host. NET formation in lung tissue is able to disturb microcirculation and NETs can easily expand in pulmonary alveoli, lling the lungs [26], while it was seen that histones can rapidly and strongly accumulate in the lung capillary network [27].
Histones have been shown to activate and recruit leukocytes [28], damage alveolar macrophages [29], activate erythrocytes [30], epithelial and endothelial cells, in particular pulmonary endothelial cells [31][32][33]. Collectively, extracellular histones cause an in ammatory response that leads to microvascular leakage and endothelial dysfunction, inducing an immunothrombotic procoagulant condition and eventually organ failure. If not cleared from circulation, histones as well as cfDNA facilitate severe systemic in ammation and worsen the clinical condition [34,35]. In a previous study [36] we reported a correlation with mortality of the plasma concentration of extracellular histone H3 in sepsis patients admitted to the ICU. Presence of NE in plasma is associated with exacerbations, lung function decline and disease severity in patients with COPD, bronchiectasis and cystic brosis [37][38][39] and decrease of NE levels in bronchiectasis patients improved lung function and airway in ammation [40].
At the same time that it provides a rst line of defense against infections, the innate immune system initiates self-control responses to prevent damage to the host. One system involved in early immunomodulation is the GAS6/TAM ligand/receptor system [41]. GAS6 is a vitamin K-dependent protein that activates the TAM family (Tyro3, AXL and MerTK) of tyrosine kinase membrane receptors [42]. The GAS6/TAM system regulates the immune response by modulating cytokine production, inducing a regulatory cellular response pro le and by mediating efferocytosis, removing irreversibly damaged cells.
The system also provides a mechanism of regulating endothelial and platelet activation and interaction [43].
Extracellular histones and cfDNA are implicated in regulation of in ammatory and hemostatic pathways in the context of severe viral infections and ARDS, all of which are implicated in COVID-19. Therefore, the objective of this study was to determine circulating levels of extracellular histone H3, cfDNA, NE, and components of the GAS6/TAM system in a group of 83 consecutive patients admitted to an intensive care unit (ICU) and compare them with pre-COVID-19 UCI patients (n = 11) and healthy controls (n = 15).

Subjects and sampling
The study was approved by the Swedish National Ethical Review Agency (EPM; No. 2020 − 01623). Informed consent was obtained from the patient, or next of kin if the patient was unable give consent.
The Declaration of Helsinki and its subsequent revisions were followed. The protocol of the study was registered (ClinicalTrials ID: NCT04316884).
All adult patients with con rmed or suspected COVID-19 admitted to the ICU between March 21, and April 13, 2020 were screened for eligibility. The diagnosis was established by PCR detection of SARS-CoV-2 E and N-genes in nasopharyngealswabs according to previously described protocols [44]. Plasma samples were collected from 83 consecutive patients. We further used 11 previously included ICU-patients without COVID-19 as ICU-controls and included 15 healthy control subjects who were healthy university employed volunteers.
For a subset of patients, longitudinal plasma samples were available (n = 22), taken between days 1 and 12, which allowed analysis of time-dependent development of plasma values for the markers measured. For other patients, longitudinal sampling could not be performed due to mortality and/or limitations in logistic capacity at the peak of ICU occupancy.
The rst blood sample (in citrate buffer) collected after a patient was admitted to the ICU was used as baseline measurement. Platelet poor plasma (PPP) was prepared by centrifugation of the blood for 10 min at 3000 g at 4 ºC, after which the supernatant was carefully pipetted, whilst taking care not to disturb the cell-containing lower layer by keeping a generous margin from the buffy coat. PPP was aliquoted and snap-frozen until use. The healthy control PPP was prepared by the same method as the patient plasma.
Quantitation of cell-free DNA Cell-free DNA (cfDNA) was quantitated from plasma essentially as described earlier [45] using a real-time PCR-based assay. In short, plasma samples were diluted 8-fold in water to result in a nal assay dilution of 40 times. Reactions were performed in 96-well plates (Roche) employing a LightCycler 480 qPCR machine (Roche). Reaction volumes contained 5 µL of TATAA Probe GrandMaster Mix/no ROX (TATAA Biocenter), 0.5 µL TATAA Alu-60 assay probes (TATAA Biocenter), 2.5 µL H2O and 2 µL of sample.
Ampli cation consisted of a pre-denaturation step 2 minutes at 95 °C, to activate the DNA Polymerase in the master mix. Followed by 40 cycles of denaturation at 95 °C for 5 s, annealing at 60 °C for 10 s and extension at 60 °C for 30 s. A calibration range (from 1 to 300 ng/µL) using puri ed and quantitated DNA standard prepared as described [46] was included in each analysis, facilitating the direct correlation of Ct values to DNA concentration.
Quantitation of NE, sAXL, and GAS6 in Plasma NE, soluble AXL (sAXL), and GAS6 levels in plasma were determined by the ELISA technique using commercial kits from R&D systems (DuoSet ELISA, Bio-techne, Minneapolis, USA) according to the manufacturer's instructions. For determination of sAXL, the plasma form of the AXL cellular receptor, and GAS6, plasma samples were diluted 1:40 and 1:200 for NE. All samples were determined in duplicates.

Analysis of extracellular histone H3
Extracellular histone H3 levels were determined using a semi-quantitative method previously described [36,47]. Brie y, plasma dilutions were subjected to SDS-PAGE gel electrophoresis and transferred to PVDF membranes (Bio-Rad Laboratories, Hemel Hempstead, UK) using semi-dry blotting. Membranes were blocked and incubated with primary anti-H3 antibody, o/n at 4 °C, (sc-8654-R, Santa Cruz Biotechnology, Heidelberg, Germany), followed by a secondary biotin-conjugated IgG for 30 minutes at RT (ab97083, Abcam, Cambridge, UK) and a streptavidin-biotin/alkaline phosphatase complex (Vectastain ABC-Alkaline Phosphatase for 30 min at RT, Vector Laboratories, Burlingame, USA). Histone H3 bands were detected by luminescent ECL substrate (Advansta, San Jose, USA). Resulting band densities were quanti ed by ImageQuant TL software (GE Healtcare, Little Chalfont, UK), as compared to known concentrations of puri ed calf thymus H3 (Roche, Basel, Switzerland). This analysis is independent of frequently observed cross reactivity of histone antibodies with non-histone plasma proteins and allows the inspection of potential in vivo histone proteolytic processing. The detection of citrullinated histone H3 was performed on a subset of 50 randomly selected COVID-19 patients that had veri ed presence of H3 and used an anti-citrulline antibody to inspect the presence of citrullinated protein at an overlaying position at 15 kDa employing the anti-citrulline detection kit (Part number 17-347B-1) and the monoclonal antibody (MABS487, Merck) in combination with a polyclonal anti-human IgGHRP-labelled antibody (DAKO, P0214). Imaging was performed as described for the histone H3 analysis described above.

Statistical analysis
Graphpad Prism version 8 (Graphpad Software Inc., La Jolla, CA, USA) and SPSS Statistics version 26 (SPSS Inc., Chicago, IL, USA) were used for statistical analysis. We employed a Kolmogorov-Smirnov test to inspect normality of data. Parametric data are presented as mean (SD) or geometric mean (95% CI) for log-transformed data unless stated. Non-parametric data are presented as median and inter quartile range (IQR). We used paired and unpaired t-test and one-way analysis of variance (ANOVA) to compare variables. Mann-Whitney tests were used for comparison of groups. *** indicates p < 0.001.
Of all ICU patients included, 83 patients had a con rmed COVID-19 diagnosis and 11 patients did not have COVID-19 and were used as ICU control patients. We obtained a total of 315 plasma-samples from the 83 patients with con rmed COVID-19, while 18 samples from 11 patients were obtained from the ICU control group. The healthy control group consisted of 15 individuals with a median age of 32 (IQR 24-37) of which 7 were male (47%).
The baseline (day 1) characteristics of the 83 ICU patients are shown in Table 1. Between the COVID-19 and the non-COVID-19 patients, males were overrepresented in the infected group, with 71.1% (p = 0.030), whereas BMI was not different between both ICU groups. Non-COVID-19 patients tended to be older than the COVID-19 patients. To allow comparison with a healthy population, a healthy control group was further included in our analyses. Physiological characteristics are also shown in Table 1. Between the COVID-19 and the non-COVID-19 patients, signi cant differences were observed in their clinical parameters, with respiratory rate, mean arterial pressure (MAP) and body temperature being higher in the COVID-19 group.
Prior to ICU admittance, malignancy, representing the main indication for ICU admission for HIPEC surgery of the non-COVID-19 group and anticoagulant treatment were signi cantly higher in the control ICU group (p < 0.001 and p 0.050 respectively). All patients in both patient groups required supplemental inspired oxygen (Supplemental Table 1), but in the ICU control patients no high-ow nasal oxygen (HFNO) or oxygen masks were required, yet the latter group received more vasoactive treatment. No signi cant differences were present in antibiotic use. Supplementary Tables 2 and 3 presents routine parameters measured for most COVID-19 patients at study inclusion. The PaO2/FiO2 ratio, a clinical indicator of respiratory dysfunction, was on average 149 mmHg (IQR 123-186 mmHg) in the COVID-19 group. Leukocytes were increased in the ICU control patients whereas erythrocytes and thrombocytes were within reference range for both patient groups. A signi cant rise in CRP, IL-6 (n = 40), procalcitonin, and lactate dehydrogenase (LDH) was seen in the COVID-19 patients, indicative of the ongoing in ammatory response and tissue damage. All these measures had the lower interquartile range above the reference range. D-dimer, indicative of brinolytic activation, was also above the reference range, although in the 19 patients where aPTT was measured, the median value was prolonged but within reference range.
Detection of Histone H3, cfDNA and NE in plasma.
Plasma samples from the 94 ICU patients, were analyzed together with those of 15 healthy volunteers (Fig. 1). In the majority of patients (67%), no extracellular H3 could be detected at ICU admission. As compared with plasma samples of the ICU control patients however, the levels of extracellular histone H3 in the COVID-19 patients on day 1 were signi cantly higher (p = 0.047), as well as compared to the healthy controls (p = 0.025; Fig. 1A), while there was no signi cant difference between the ICU control and healthy control group ( Table 2). The cfDNA values from both ICU groups differ greatly (p < 0.001), with the COVID-19 group presenting 25-times higher levels than of the control group (Fig. 1B). The levels of cfDNA did not differ between the ICU control and healthy control patients (Table 2). There was a highly signi cant difference between both groups when NE was determined (p < 0.001; Fig. 1C) with NE being virtually absent from the ICU control group and the healthy controls (Table 2).  Detection of GAS6 and sAXL in plasma.
In COVID-19 and non-COVID-19 ICU patients, the GAS6 concentration at study inclusion was 24.8 (18.1-33.3) ng/mL and 12.1 (9.5-15.9) ng/mL respectively, indicating a signi cant difference between both groups (p < 0.001; Fig. 2A, Table 2). The concentration of the healthy control group 14.4 (11.0-19.7) ng/mL is signi cantly lower (p < 0.001) compared to the COVID-19 ICU group, while no difference is found with the non-COVID-19 ICU group ( Correlations between different plasma markers and laboratory parameters. The concentration of H3 correlated with cfDNA and NE concentrations at ICU admission in COVID-19 patients (Table 3). At this point, there was no signi cant correlation between cfDNA and NE measurements. In contrast, LDH correlated very well with cfDNA (0.506; p < 0.001), which could indicate a common origin in cellular damage. Remarkably, sAXL correlated signi cantly with the myocardial injury biomarker troponin I in this cohort.
These correlations persisted when all subsequent sampling was included (supplementary Table 4). In addition, other correlations with biochemical parameters indicating organ damage were observed, including H3 and cfDNA with AST, LDH and procalcitonin (weaker for H3 in the latter case; p = 0.029). Interestingly, cfDNA correlated also with D-dimer and CRP when all points were considered, indicators of ongoing coagulation and brinolysis. Although NE correlated strongly with H3 as was the case with the initial values, NE did not correlate signi cantly with other tissue damage parameters, nor with cfDNA, while it correlated signi cantly with leukocyte and neutrophil cell count, as well as procalcitonin and GAS6 (Supplementary Table 4).
Prognosis and time-dependent parameter development in COVID-19.
At ICU admission, the value of SAPS-3 score (but not SOFA, supplementary Fig. 1) was signi cantly higher in the group of COVID-19 patients that did not survive (n = 21) compared to those that survived (n = 62; supplementary Table 5). Non-survivors were older, hypertensive (86%), and had vessel or ischemic heart failure more frequently than survivors. Accordingly, they were more frequently treated with anticoagulants and ACEi/ARB treatment. They also suffered more thromboembolisms during ICU stay. From the biochemical measurements performed, non-survivors had higher markers of organ damage, including AST, troponin I and procalcitonin (Supplementary Table 6). In those patients where IL-6 was determined, mean IL-6 concentration in non-survivors (n = 10) was almost 3 times that of survivors (n = 30). Although cfDNA, H3, NE, sAXL and GAS6 were all elevated in COVID-19 positive patients at admission, none of these parameters was a good predictor of nal outcome. Indeed, at ICU admission, there was no correlation of any of these parameters with SAPS-3 or SOFA scores.
In 22 patients where several samples could be obtained during their ICU stay, we studied the timedependent development of the markers. We arbitrarily divided the results in an early sample (day 1-5) and a late sample (day 6-12), using the mean value of the different samples obtained during each period (supplementary Table 7). Platelet counts, in the normal range in the baseline measurements, increased signi cantly in the late samples, as well as leukocytes (supplementary Table 7). Markers of tissue damage also increased, including ALT and AST. There were no signi cant differences between early and late values of histone H3, cfDNA or NE, although more samples were positive for H3 in the latter group.
No differences were observed for sAXL or GAS6. We furthermore inspected if the correlations observed in the initial samples persisted over the early (day 1-5) and late phase of ICU stay for the 22 COVID-19 patients for which data were available over this complete period of time (Supplementary Table 8). We observed that H3 correlated with cfDNA over both time periods and generally showed persistent correlation with neutrophil counts. Strong correlations were seen between NE levels and neutrophils (0,810 and 0,708 respectively for early and late phase).
Next, we divided the 22 patients according to their nal outcome at 30 days (Fig. 3). Those individuals that survived (n = 15) showed a 40% decrease in cfDNA concentration in the late plasma samples.
Further, in these patients GAS6 concentration decreased more than 30% in late samples. In contrast, neither cfDNA nor GAS6 decreased in the group of non-survivors (n = 7), with their late values being very similar to the early ones. The same trends were observed in H3 and NE measurements, although in these cases, it did not reach signi cance. In addition, when regarding the NE concentrations, we found both in the early phase, and in the late phase a signi cant difference between survivors and non-survivors.  Histone positive samples contained signi cantly higher NE and cfDNA than those in which no H3 was detected (p < 0,001 and p < 0,001 respectively, Table 4), in line with the strong correlation between these parameters (Table 3 and supplementary table 4). The presence of histone in COVID-19 patients appeared not randomly divided over time (Fig. 3). Although no statistical signi cance was reached for comparison between early and late survivors/non-survivors, the average histone levels in early survivors were higher than in late survivors, whereas average histone H3 levels in early non-survivors were lower than in late non-survivors.
In several COVID-19 patients, histone H3 showed an additional 13 kDa band on Western blot (Figs. 4A and 4B). This was seen in 37% of histone H3 positive patients at day 1 and overall in 40,7% of all samples tested. The origin of this band is most likely a proteolytic cleavage of the single 15 kDa band. In one patient, only the lower weight band was seen (patient 47, day 7, Fig. 4A), suggesting full H3 proteolysis. In 5 patients, the proteolysis disappeared in samples collected after 6 days or later of ICU stay.
In addition to presence of histones per se, we found that in a subset of 50 randomly selected COVID-19 patients with histone H3 in their plasma, the presence of citrullinated histone and found that 36 of these (72%) were positive for citrullination (Fig. 4C).

Discussion
Our study broadly assessed a series of biomarkers re ecting the process of cellular damage and NETosis as well as a mechanism of early response to damage, the GAS6/AXL pathway, in a group of severely ill patients, consecutively admitted at an ICU during the present COVID-19 pandemic. A non-COVID-19 group was included consisting of ICU patients suffering from malignancies requiring surgery (Supplementary  Table 1). PaO2/FiO2 ratios were clearly lower in the COVID-19 group, resulting in higher percentage of invasive respiration applied. Increased tissue damage, and particularly lung tissue damage, with more pronounced in ammation, in combination with pulmonary thromboembolic disease could explain this difference. This is underscored by increases in CRP, IL-6, D-dimer, LDH, procalcitonin, ASAT and increased neutrophil counts at admission (Supplementary Table 2).
Levels of cfDNA, H3 and NE differed signi cantly already at admission to ICU care between the COVID-19 and non-COVID-19 groups. This is likely indicative of increased tissue damage and neutrophil activation in this group. Given the equimolar presence of core histones H1, H2A, H2B, H3 and H4 in nucleosomes, exposure of H3 to the extracellular milieu, is equally indicative for presence of the other core histones.
However, histones H3 and H4 have been found to be the most toxic [21,48]. While cfDNA and NE levels appeared to decrease over time, we noted that when analyzing the full collection of H3 measured, the highest levels of histones were found for the days 4-8 and that this marker was on average less present during the rst days of ICU admission. Given the difference to NE and the at least partial NETosis origin of H3 (as indicated by their citrullination) we hypothesize that a rapid clearance of histone H3 in the early phase of ICU stay could explain the observed overall lower levels of circulating free histone H3. The clearance of cytotoxic histones could represent a protective response that is brought about by uptake by immune cells, binding to intact glycocalyx or binding to endothelium, platelets, or vesicles that originate from these cells. Even though more data are required to support our hypothesis, the latter could contribute to the observed reduction in platelet counts and, by virtue of their capacity to activate endothelium and platelets, to the observed increase in thromboembolic events. Remarkably, we noted that in a sub-set of plasma samples that contained H3, exclusively found in the COVID-19 group, H3 was proteolyzed. Several proteases are known to process extracellular histones including NE. Partial processing of H3 by NE was shown to occur during NET formation [49]. Given the observed correlation between H3 proteolysis and NE levels, we conclude that in COVID-19 patients, extracellular histones are most likely processed by NE. Of note, we realize the limitations of the semi-quantitative H3 assay used here, mostly being timeintensive, however, unless truly speci c antibodies are available, proteolysis would be undetected using commonly applied solid-phase based methods like ELISA.
We further found that in a subset of 50 histone H3 containing samples, histone H3 was citrullinated in 36 (72%), suggesting that histone exposure involves the activity of PAD4 as is the case during ROSindependent NETosis. It is likely that the non-citrullinated H3 originates from damaged tissue or from PAD4-independent NETosis pathways.
It has been suggested that cfDNA could serve as a surrogate marker for NETosis in critically ill mechanically ventilated patients in whom NETs contribute to local alveolar in ammation [50]. However, our data show that cfDNA correlated best with H3, while the correlation with NE was less strong. This could indicate that NET formation could be better re ected by the release of neutrophil-speci c markers such as NE, while H3 and cfDNA could relate to cellular damage in a broader sense. cfDNA is a DAMP able to activate TLRs. The sustained increase in cfDNA observed in COVID-19 patients would propagate in ammation through TLR activation. In this sense, it is important to stress that cfDNA decreased over time in those patients that survived the infection, while those that did not survive maintained constantly high plasma cfDNA concentrations. Likewise, increased levels of NE are able to reduce the lung permeability barrier function and induce release of pro-in ammatory cytokines, collectively inducing emphysematous lesions.
We observed H3 positive samples levels at ICU admission, like we found earlier in a critically ill ICU population of sepsis patients [36]. While this points at similar pathways being involved in disease onset and progression, other reports [51] pointed out that COVID-19 clinical features are similar but different from those seen in sepsis. The observed normal platelet counts found in most samples in this study, irrespective of the phase of the disease or its outcome, further underscore this point. A possible early onset of NETosis, with an associated rise in extracellular histones is supported by our observation that NE levels were evident already from day 1 till day 12, implying neutrophil activation and NETosis, accompanied by cfDNA also being present already at day 1.
The components of the GAS6/TAM system have been shown to increase in a diverse spectrum of in ammatory conditions [52], including sepsis and septic shock; but also systemic in ammatory response syndrome (SIRS) without infection [53]. In several studies, GAS6 concentration in plasma at IC admission correlated with severity of organ damage, either in SOFA or with damage of speci c organs [53][54][55][56][57]. This also the case in viral infections. Using a murine model of respiratory syncytial virus (RSV) infection, Shibata et al [58] showed that GAS6 is expressed upon infection in lung alveoli, leading to conversion of alveolar macrophages into M2-like cells. These studies illustrate the modulatory role of the innate response provided by the GAS6/TAM system and suggest that the presence of these components in plasma could be an early event in the orchestration of the immune response to viral infections.
In our COVID-19 cohort, GAS6 concentrations at ICU admission more than doubled those of non-COVID-19 IC patients. Among plasma determinations that correlate with GAS6 are the interleukins IL-6 [53,54] and IL-8 [54]. GAS6, IL-6 and IL-8 concentrations are increased in septic patients who develop acute lung injury (ALI; [53). Plasma GAS6 concentration was able to signi cantly discriminate patients that would develop ALI in that cohort {Yeh, 2017 #63]. Also, non-survivors of sepsis in ICU tend to have initial higher concentration of GAS6 and GAS6 could predict mortality with an AUC of 0.7 [54]. However, in our cohort, no evident correlation with severity of the lung symptoms could be established, which could re ect a speci c characteristic of SARS-CoV-2 interaction with the system.
In our study group, GAS6 concentration was maximal at IC admission, and was maintained high in nonsurvivors. Ni et al have shown that recombinant GAS6 infusion improves the outcome of experimental sepsis in mice, controlling multi-organ dysfunction [59]. One of the target cells of GAS6 in bacterial infection is the vascular endothelium, that showed reduced LPS-induced permeability in the presence of GAS6. Interestingly, GAS6 is also necessary to maintain the response of vascular endothelium during in ammatory conditions, allowing endothelial cell interactions with platelets and leukocytes [43].
While the concentration of sAXL was found increased in sepsis cohorts [53,54], the increase was not so evident and did not signi cantly correlate with organ damage, similarly to our observation in COVID-19 patients. Soluble AXL is an ADAM-shed form of the receptor, found in plasma in a complex with GAS6 [60]. The speci c increase of GAS6 over sAXL in COVID-19 could re ect a need of free, active GAS6 in this condition. Interestingly, sAXL showed a correlation with troponin I. This re ects the observed correlation of sAXL with parameters and progression of heart failure also observed in previous studies [61], and could indicate a speci c implication of AXL in the cardiac damage observed in COVID-19. It is possible that in the context of COVID-19, GAS6 could be acting through another TAM receptor in this context, MerTK. MerTK inhibition increases the in ammatory cytokine storm in LPS-induced ALI [62], acting as a tolerogenic receptor under in ammatory conditions [63]. However, it has to be noted that LPS and other PAMPs drastically reduce MerTK expression [63].

Conclusion
We have shown the presence of extracellular histone H3 and cfDNA in the plasma of COVID-19 patients to be signi cantly different from plasma of ICU patients who did not have COVID-19. Cytotoxic extracellular histone H3 was found in 40% of COVID-19 patients, theoretically contributing to disease progress, but not predicting nal disease outcome. Histones were found to correlate with parameters for tissue damage and with neutrophil counts, implying their partial origin from NETosis. This was supported by the observation of cfDNA that showed a strong correlation with extracellular histone levels, suggesting their possible use as proxy markers for the assessment of plasma histone levels. In COVID-19, an early activation of the GAS6 immunomodulatory pathway takes place, re ected in a clear increase of the plasma concentration of this vitamin K dependent protein. The study was approved by the Swedish National Ethical Review Agency (EPM; No. 2020-01623).
Informed consent was obtained from the patient, or next of kin if the patient was unable give consent.
The Declaration of Helsinki and its subsequent revisions were followed. The protocol of the study was registered (ClinicalTrials ID: NCT04316884).

Consent for publication
Not applicable.

Availability of data and materials
The data used and/or analyzed in the present study are available from the corresponding author on reasonable request.