Severe COVID-19 infection is associated with aberrant cytokine production by infected lung epithelial cells rather than by systemic immune dysfunction

Abstract The mechanisms explaining progression to severe COVID-19 remain poorly understood. It has been proposed that immune system dysregulation/over-stimulation may be implicated, but it is not clear how such processes would lead to respiratory failure. We performed comprehensive multiparameter immune monitoring in a tightly controlled cohort of 128 COVID-19 patients, and used the ratio of oxygen saturation to fraction of inspired oxygen (SpO2 / FiO2) as a physiologic measure of disease severity. Machine learning algorithms integrating 139 parameters identified IL-6 and CCL2 as two factors predictive of severe disease, consistent with the therapeutic benefit observed with anti-IL6-R antibody treatment. However, transcripts encoding these cytokines were not detected among circulating immune cells. Rather, in situ analysis of lung specimens using RNAscope and immunofluorescent staining revealed that elevated IL-6 and CCL2 were dominantly produced by infected lung type II pneumocytes. Severe disease was not associated with higher viral load, deficient antibody responses, or dysfunctional T cell responses. These results refine our understanding of severe COVID-19 pathophysiology, indicating that aberrant cytokine production by infected lung epithelial cells is a major driver of immunopathology. We propose that these factors cause local immune regulation towards the benefit of the virus.


42
The clinical manifestations of COVID-19 range in severity from asymptomatic infection to critical 43 illness and death, yet the mechanisms by which SARS-CoV-2 cause morbidity and mortality have yet to be 44 fully elucidated. It has been proposed that an excessive immune response may cause immunopathology in 45 affected target organs, particularly the lower respiratory tract. Several large studies of hospitalized patients 46 demonstrated that disease severity and mortality are correlated with elevated levels of inflammatory 47 cytokines, suggesting a potentially dysregulated immune response to infection 1-4 . Consistent with this 48 notion, the steroid dexamethasone improved outcomes in severe and critically ill patients 5,6 . IL-6 specifically 49 has been proposed as a functionally important cytokine, 7 and the anti-IL-6R antibody (Ab) tocilizumab 50 provided a survival benefit in critically ill COVID-19 patients 8,9 .

121
CD38 and HLA-DR are markers of activated T cells during viral infections 25 , and this population 122 was increased among both CD4 + and CD8 + T cells in COVID-19 patients of all disease severities (Fig 2b).

129
There were no significant differences in the percentages of CD4 + or CD8 + T cells expressing PD-1, although 130 modest upregulation of PD-1 was seen on CD4 + EM and CD8 + EM and TEMRA cells (Fig 2c, 131 Supplementary Fig 4a,b). There was an increased percentage of CD8 + T cells expressing TIM-3 (Fig 2c).

132
The proportion of regulatory T cells (Tregs) also increased in COVID-19 patients (Fig 2a), suggesting a 133 counter-regulatory mechanism in response to increasing levels of T cell activation.

134
Several studies have shown that inhibitory receptors including PD-1 are upregulated on SARS-

135
CoV-2 specific T cells, and have suggested that PD-1 high cells in COVID-19 infection are exhausted 17,[27][28][29] 136 or have decreased polyfunctionality 28,30 . However, PD-1 can also be upregulated in acutely activated T 137 cells 3,31 . To determine whether there were differences in IFN-γ production by SARS-CoV-2-specific T cells, 138 we used an ELISPOT to measure IFN-γ production after stimulation with overlapping HLA class I & II 15-139 mer peptides from the S, M, and N proteins of SARS-CoV-2. IFN-γ production was seen as early as day 8 140 after symptom onset, and the degree of IFN-γ production was similar between patients with different 141 disease severities (Supplementary Fig 4f).

142
To determine if PD-1 on these cells represents a marker of activation or exhaustion, we used 143 intracellular cytokine staining to measure polyfunctionality after S/M/N peptide stimulation. Compared to 144 mild patients, severe patients had higher percentages of polyfunctional CD4 + T cells producing IFN-γ, TNF-145 α, and/or IL-2 in response to S/M/N peptide stimulation (Fig 2d-f). Furthermore, cytokine production was 146 concentrated in the PD-1 + CD4 + T cells, indicating that PD-1 represents an activation marker rather than a marker of dysfunction in this context. There was a similar trend with PD-1 + CD8 + T cells in severe patients, 148 but this was not significant due to increased patient-to-patient variability in the CD8 + T cell response 149 ( Supplementary Fig 4g). Cytokine-producing T cells were enriched amongst the CD38 + HLA-DR + population 150 (Fig 2d,2f), consistent with this population containing virus-activated T cells. We conclude that the adaptive 151 immune response is robust in severe COVID-19 patients and that lack of virus-specific immunity is not 152 contributory to the progression to disease severity.

153
Patients with SARS-CoV-2 also had a serum cytokine and chemokine profile consistent with 154 increased T cell activation. Levels of CCL19 and CCL20, which recruit T cells to lymph nodes for activation, 155 increased with disease severity (Fig 2g). Severe patients had higher levels of sCD25/IL-2Ra, which is 156 cleaved and released upon T cell activation. CCL5 and CXCL10 recruit T cells to sites of inflammation, 157 and were elevated in the serum of COVID-19 patients. CXCL10 also increased with disease severity.

158
Levels of CCL19, CCL20, and CD25/IL-2Ra remained elevated over time in severe patients, while CCL5 159 and CXCL10 levels declined over time in both mild and severe patients (Fig 2h). Patients with severe  Fig 5a). However, the level of CD86 increased in plasmacytoid DCs, 169 indicating a more activated status. CD1c + DCs also had higher levels of Tim-3 at late time points after 170 SARS-CoV-2 infection. The proportions of neutrophils, non-classical monocytes, and intermediate 171 monocytes were increased in patients with COVID-19 compared to healthy controls (Fig 3a). While the 172 percentage of classical monocytes was unchanged, the mean fluorescence intensity (MFI) of CD86 and 173 HLA-DR was decreased in infected patients (Fig 3a), suggesting the emergence of less-mature monocytes 174 from the bone marrow. This is further supported by increased levels of the myeloid growth factor GM-CSF in the serum of patients with SARS-CoV-2 infection (Fig 3b), and a negative correlation between GM-CSF 176 and HLA-DR levels on intermediate monocytes (R 2 = 0.15, p = 0.004) (Supplementary Fig 5c). These 177 parameters are consistent with tissue repair-type macrophages being favored during SARS-CoV-2 178 infection.

179
Patients with SARS-CoV-2 had increased levels of cytokines responsible for recruiting neutrophils, 180 monocytes and macrophages to sites of inflammation, including the neutrophil chemoattractants IL-8, 181 CXCL1, CXCL2, and the monocyte chemoattractants CCL2, CCL4, and CX3CL1 (Fig 3b). IL-8, CCL2, and 182 CX3CL1 also increased with disease severity. Distinct groups of cytokines clustered together in correlation 183 plots at late time points (Fig 3c), particularly in severe patients. CCL2 levels remained high over time in 184 severe patients, and higher levels of CCL2 also correlated with a longer duration of moderate or severe 185 illness (R 2 = 0.17, p = 0.00737) (Fig 3d-e).

186
IL-6 has been identified as a pathologic mediator of cytokine release syndrome after CAR-T cell 187 treatment, and it has been hypothesized that a similar phenomenon may be driving severe pathology in 188 some COVID-19 patients 7 . IL-6 signals through the IL-6R and gp130 complex. Gp130 is ubiquitously 189 expressed, while IL-6R expression is normally limited to immune cells and hepatocytes. IL-6 can also form 190 a complex with soluble IL-6R (sIL-6R!) and signal in trans through gp130 in cells that do not express the 191 IL-6R. We found that IL-6 levels increased with disease severity, while sIL-6R! and gp130 levels were 192 similar between severity groups (Fig 3b). While there was a correlation between CRP and IL-6 levels, there 193 were many patients who had a high CRP but only a modest increase in IL-6 ( Supplementary Fig 5d). Levels 194 of IL-6 remained high at late timepoints in severe patients when compared to mild (Fig 3d), and levels of 195 soluble gp130 were lower in severe patients at late timepoints ( Supplementary Fig 5e). sIL-6R! levels 196 remained high over time in both mild and severe patients. Interestingly, the duration of moderate or severe 197 disease positively correlated with IL-6 levels and negatively correlated with soluble levels of gp130, which 198 is an endogenous inhibitor of IL-6 trans-signaling 34-37 (Fig 3e, Supplementary Fig 5f).

199
In order to better understand the pathophysiology that differentiates severe patients from mild or for an active innate immune response (elevated G-CSF, IL-8, and the percentage of neutrophils) as well as 205 an activated T cell response (elevated CXCL10, with the anti-IL-6R antibody tocilizumab 9 . Based on prior work studying cytokine-release syndrome in CAR-

217
T cell therapy 38,39 and IL-6 production in infectious models 40-42 , it has been assumed that IL-6 in COVID-19 218 patients is being produced by macrophages 43 . However, in our cohort examining representative patients 219 having "high" versus "low" serum IL-6 levels at the protein level ( Supplementary Fig 6a), no difference in 220 mRNA for either IL-6 or CCL2 was observed among peripheral level blood mononuclear cells 221 ( Supplementary Fig 6c-d). This result is consistent with the flow cytometric analysis of circulating 222 monocytes, which indicated an immature and possibly tissue repair phenotype rather than an inflammatory 223 one (Fig 3a). Together, these results suggested that the source of these cytokines might not be immune 224 cells, but rather raised the possibility that virus-infected cells in the lung might be the major source. We 225 therefore examined expression of IL-6 and CCL2 mRNA in lung tissue from a cohort of 10 fatal COVID-19 226 cases listed in Supplementary Table 3. We performed a multispectral immunofluorescence assay 227 combining RNA in situ hybridization (RNA-ISH) for SARS-CoV-2 RNA and IL-6 or CCL2 mRNA, along with 228 protein immunofluorescence (IF) staining to identify the cells of origin. Thyroid transcription factor 1 (TTF1) 229 was used to identify type 2 pneumocytes, and CD45 was utilized to identify leukocytes (Fig 4a, 230 Supplementary Fig 7a). SARS-CoV-2 RNA was detected in all of the autopsy lung specimens.
Unexpectedly, the vast majority of IL-6 transcripts were detected in cells that did not co-stain for the 232 macrophage markers CD68 or the M2 macrophage marker CD163 ( Supplementary Fig 6e-f). Interestingly, 233 large numbers of TTF1 + type 2 pneumocytes expressed IL-6 mRNA, with a high percentage of these cells 234 also positive for SARS-CoV-2 RNA (Fig 4a-c). Quantitative analysis showed TTF1 + type 2 pneumocytes 235 were the predominant IL-6-expressing cell type, greatly outnumbering CD45 + immune cells (Fig 4b,c).

236
Among the IL-6 positive populations, type 2 pneumocytes relative to CD45 + cells showed greater IL-6 237 expression on a per cell basis, as indicated by a greater number of TTF1 + cells with higher mean staining 238 intensity for IL-6 (Fig 4d). Similarly, CCL2 expression was particularly abundant on TTF2 + type 2 239 pneumocytes ( Supplementary Fig 7a-d). Together these data show that virus-infected lung epithelial cells 240 are the major source of IL-6 and CCL2 in SARS-CoV-2 infected lungs.

243
Here we show that IL-6 and CCL2 are major factors that discriminate severe infection from mild or 244 moderate disease. IL-6 is known to be produced by innate immune cells such as macrophages or dendritic 245 cells, and by non-immune cells such as epithelial cells or fibroblasts. In allergic asthma 44,45 , SARS-CoV-246 1 40 , influenza 41 , and pneumovirus infection models 42 , IL-6 has been shown to be produced by macrophages 247 and other myeloid cells, whereas IL-6 can be produced by cultured nasal epithelial cells infected with 248 RSV 46,47 . In mouse models of CAR-T cell cytokine release syndrome, macrophages and monocytes are 249 the predominant source of IL-6 38,39 , while vascular endothelial cells have also been shown to produce IL-6 250 in CRS autopsy specimens 48 . Our results from human autopsy specimens unexpectedly show that the 251 predominant source of IL-6 and CCL2 in vivo during SARS-CoV-2 infection is from infected epithelial cells.

252
Our data are consistent with scRNA-seq studies of PBMCs from COVID-19 patients that showed a 253 discrepancy between serum cytokine measurements and the cytokine transcripts of CCL2 and IL-6 among 254 PBMCs 49-53 . Large numbers of epithelial pneumocytes co-stained with IL-6 or CCL2 and SARS-CoV-2 RNA 255 probes, implicating direct cytokine induction by the virus. When considering potential mechanisms of 256 cytokine production, it has been demonstrated that SARS-CoV-2 induces Nuclear Factor kappa B (NF-kB) 257 upregulation and IL-6 production in cultured lung alveolar and epithelial cells 54,55 . CCL2 and other 258 inflammatory mediators are also induced via the NF-kB pathway 56 .
In mouse models of coronavirus infections, sustained CCL2 expression enhanced the lethality of 260 disease, and promoted immunopathology with a destructive monocyte/macrophage response and 261 ineffective virus clearance 57 . The effect of excess CCL2 in human SARS-CoV-2 has not yet been 262 elucidated. CCL2 may recruit wound-healing M2 macrophages, which can facilitate lung tissue repair by 263 stimulating type 2 pneumocyte expansion 58 , thereby triggering a process capable of enhancing virus 264 propagation 59 . The anti-IL-6R antibody tocilizumab improves survival in critically ill patients 9 , implying that 265 excessive IL-6 is detrimental to the host. Elevated levels of IL-6 in cancer models have been mechanistically 266 linked to decreased DC survival and activation, and consequently impaired CD8 + T cell priming 60 . As such,

279
Robust adaptive immune responses were seen in patients with mild and moderate disease and 280 were even higher in patients with severe disease, arguing that the lack of a protective immune response 281 did not cause severe disease. CD38 + HLA-DR + CD4 + and CD8 + T cells were polyfunctional, and patients 282 with a higher disease severity had more cytokine producing CD4 + T cells. These cytokines were being 283 produced by PD-1 + cells, indicating that PD-1 in this context is a marker of activation, not exhaustion. This 284 is consistent with other recent work showing that tetramer + PD-1 + , SARS-CoV-2-specific CD8 + T cells 285 produce cytokines 62 . In patients with severe disease, markers of T cell activation such as sCD25/IL-2R 286 remain high at late time points, suggesting an ongoing immune response against the virus. In some severe 287 patients who ultimately die from SARS-CoV-2, persistent viral RNA has been demonstrated in longitudinal 288 saliva samples from the Iwasaki group 63 , as well as in our autopsy lung samples. Increased antigen load 289 and duration of antigenic exposure leads to increased T and B cell expansion and differentiation in other 290 experimental models 64 . While we cannot rule out that the increased adaptive immune response causes 291 immunopathology, the increase in regulatory modulators such as IL-10 and Tregs suggests that the immune 292 system is appropriately executing negative feedback pathways. Our data suggests that the immune 293 response to SARS-CoV-2 is a functional and proportional response to infection, and infected pneumocytes 294 are the major source of IL-6 and CCL2, thus revising the paradigm of how we understand the pathogenesis 295 of severe SARS-CoV-2 infection.

520
For serum, fresh blood was collected into a preservative-free vacutainer tube and allowed to clot 521 for at least 30 minutes at room temperature. Leftover plasma samples collected in heparinized tubes were 522 also obtained from the clinical chemistry lab. Tubes were centrifuged for 20 minutes at 1300 x g at room 523 temperature, and the yellow serum/plasma layers were collected and stored at -80 degC until analysis.

524
For PBMCs, fresh blood was collected into heparinized vacutainer tubes and separated using

567
The reaction was stopped after 15 minutes with 2M sulfuric acid. The optical density (OD) was read at 450 568 nm using a Synergy H4 plate reader (BioTek). The OD values for each sample were background subtracted.

569
A positive control standard was prepared from plasma samples pooled from 6 COVID-19-infected patients,

645
Cells were incubated at 37°C with 5% CO2 with activating stimuli for 18 hours in CTL-Test Medium

646
(ImmunoSpot) with 1% L-glutamine (Gibco) prior to developing plates per manufacturer's recommended 647 procedure. Plates were scanned using ImmunoSpot analyzer and spots were counted using ImmunoSpot        FiO2 is depicted on the right. Each dot represents an individual S/F ratio calculated based on the simultaneous oxygen saturation and FiO2 recorded for hospitalized patients; all available timepoints per day per patient are shown. Oxygen delivery categories are labelled on the X axis. Colored shaded areas indicate S/F ratios which correspond to mild (green), moderate (orange) and severe (red) disease severity categories. (b) Serum SARS-CoV-2 total Ig antibody levels (RBD and Spike) in (n=68) PCR+ patients over time expressed as days post symptom onset (DPSO). Samples from individual patients are connected by lines, and colored by disease severity (mild in green, moderate in orange, severe in red). The dashed line indicates the max titer of a pool of negative controls, used as the threshold for positivity. Samples post receipt of convalescent plasma transfusion were excluded from this and all antibody analyses. (c) Maximum serum SARS-CoV-2 IgG (red) and IgM (turquoise) antibody levels (RBD and Spike) in PCR+ patients during acute phase of infection (day 10-19) (n=31), recovery (day 20-59) (n=9), and late recovery (day 60-120) (n=5), where day is DPSO. Data from the pre-humoral phase (<day 10) is excluded. (d) Mean populations of immune cell subsets involved in the humoral response, plotted as % of CD45+ cells, during the early phase (days 1-9 from symptom onset) and late phase (days 10-30 from symptom onset) of disease in COVID-19 infection compared to non-infected healthy donor controls (HD). Each dot represents the average of measurements during each phase of disease from an individual subject (early phase: mild, n=15; moderate, n=6; severe disease, n= 6 and late phase: mild, n=18; moderate, n=15; severe, n=17; non-infected healthy controls, n=9). The boxplots show the medians (middle line) and the first and third quartiles (upper and lower bounds of the boxes). (c,d) Significance was determined by two-sided Mann Whitney Wilcoxon test and p-values are indicated by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001, ****, p ≤ 0.0001). (e) Linear regression shown for disease severity expressed as S/F ratio and maximum anti-SARS-CoV-2 (RBD and Spike) total Ig antibody titers from days 10-30 in (n=53) PCR+ patients. Shaded areas represent 95% confidence interval. Ig titers from the pre-humoral phase (<day 10) were excluded. CD4 + Figure 2: SARS-CoV-2 infection elicits a robust expansion of activated polyfunctional T cells (a-c) Proportion of immune cell subsets related to adaptive immune responses are shown as a percentage of either live CD45+ cells, live CD4+ cells, or live CD8+ cells, as indicated. Where multiple timepoints within the early (D1-9) or late (D10-30) phase per patient were available, the mean was taken and each patient is represented by one dot per time phase. n = 9 for HD, n = 15, 6, 6 for mild, moderate, and severe respectively in the early phase, and n = 18, 15, 17 for mild, moderate, and severe in the late phase. (b, left) Representative flow cytometry plots showing CD38+ HLA-DR+ subsets. (d-f) PBMCs from D11-25 DPSO were stimulated with a combined pool of peptides from the S, M and N proteins for 9 hours and stained for intracellular cytokine production. Background activity in unstimulated wells was subtracted from stimulated wells; negative values after subtraction were set to 0. Representative flow cytometry plots are shown for TNF /IFN-γ (d, upper panels) and TNF /PD-1 (d, lower panels) staining. The percentage of cells producing various combinations of IFN-γ, TNF , and IL-2 were reported for CD4+ and CD8+ non-naive (e) and CD4+ and CD8+ CD38+HLADR+ cells (f). (e,f) Comparisons between groups were done by summing the total cytokine production in each column and performing a 2-way ANOVA with Sidak's multiple comparisons test. Error bars represent mean +/-SD for each cytokine subset. HD n = 9; mild n = 8; severe n = 7. Wells with <50 CD38+HLA-DR+ cells were excluded from that subset analysis, leaving n = 9 / 7 / 6 HD/mild/severe for CD4+CD38+HLA-DR+ and n = 9 / 8 / 6 HD/mild/severe for CD8+CD38+HLA-DR+. (g) Peak cytokine and chemokine levels related to T cell activation and survival are shown during the early phase (days 1-9 from symptom onset) and late phase (days 10-30 from symptom onset) of disease in SARS-CoV-2 infection compared to non-infected healthy controls. Each dot represents maximum value per individual subject during each phase of disease (early phase: mild, n=15; moderate, n=10; severe, n= 11 and late phase: mild, n=23; moderate, n=16; severe, n= 19 and non-infected healthy controls, n=18). Samples from patients post receipt of tocilizumab (which directly modulates cytokine levels) were excluded from this and subsequent cytokine analyses. (h) Kinetics of cytokine expression over time (days post symptom onset) from mild (green, n = 33), moderate (orange, n = 19), and severe (red, n = 23) patients. Multiple timepoints per patient plotted when available. Linear regression for cytokine values over time in severe (red) and mild (green) patients shown. Shaded areas represent 95% confidence interval. (a-c, g) Significance was determined by two-sided Mann Whitney Wilcoxon test and p-values are indicated by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001, ****, p ≤ 0.0001).  TNFa  VEGF  IL-8  gp130  IL-15  CXCL2  CX3CL1  GM-CSF  CXCL1  CCL4  TRAIL  CXCL10  CCL5  IL-6  Granzyme B  IL-7  IL-10  IL-13  CCL2  IFNa  IL-6Ra IFNy VEGF-C IL-1Ra CD25/IL-2Ra G-CSF Angiopoietin-2 IL-1a CCL20 IL-33 CCL19 TGFa  Where multiple timepoints within the early (D1-9) or late (D10-30) phase per patient were available, the mean was taken and each patient is represented by one dot per time phase. n = 9 for HD, n = 15, 6, 6 for mild, moderate, and severe, respectively in the early phase, and n = 18, 15, 17 for mild, moderate and severe in the late phase. (b) Peak cytokine and chemokine levels are shown during the early phase (days 1-9 from symptom onset) and late phase (days 10-30 from symptom onset) of disease in SARS-CoV-2 infection compared to non-infected healthy controls. Each dot represents maximum value per individual subject during each phase of disease (early phase: mild, n=15; moderate, n=10; severe, n= 11 and late phase: mild, n=23; moderate, n=16; severe, n= 19 and non-infected healthy controls, n=18). (c) Correlations between cytokines from days 10-30 for COVID-19 patients were calculated and clustered hierarchically. S/F ratio is fixed as the first column for comparison. Samples were stratified by disease severity. Spearman correlation coefficients were quantified by the scale of color and size of colored squares; significance of the correlation is labeled with * (P < 0.05), ** (P < 0.01), and *** (P < 0.001      Comp-Pacific Orange-A :: CD20  Where multiple timepoints within the early (D1-9) or late (D10-30) phase per patient were available, the mean was taken and each patient is represented by one dot per time phase. n = 9 for HD (blue), n = 15, 6, 6 for mild (green), moderate (orange), and severe (red) respectively in the early phase, and n = 18, 15, 17 for mild, moderate and severe in the late phase. (c) UMAP projections of live, CD45+ CD3+ cells from samples between D11-17 DPSO (equal numbers of cells sampled from n=7, 7, 7, 7 HD, mild, moderate, and severe patients, respectively) with FlowSOM clusters overlayed. (d) Marker expression in each FlowSOM cluster shown. (e) The percentage of each cluster derived from HD (blue), mild (green), moderate (orange) or severe (red) patients. (f) PBMCs were stimulated with peptides from the S, M, or N proteins in separate wells for 18 hours and IFN-γ production measured by ELISPOT. Response to S, M, and N peptides was summed and normalized per 100,000 cells plated. IFN-γ ELISPOT response shown as a linear correlation with DPSO with disease severity indicated by color with shaded area representing 95% confidence interval (top panel) and as a boxplot by disease severity (bottom panel); n = 10, 17, 18 in mild (green), moderate (orange), and severe (red), respectively. (g) Cytokine production after stimulation with pooled peptides from the S, M and N proteins, ɑCD3/CD28/CD49d antibodies, or PMA and ionomycin is shown for each individual patient. Background activity in unstimulated wells was subtracted from stimulated wells. The percentage of cells producing various combinations of IFN-γ, TNFα, and IL-2 were reported. HD: n = 9 for SMN stimulation, n = 8 for ɑCD3/CD28/ CD49d and PMA+ionomycin stimulation; mild n = 8; severe n = 7. (h) Maximum cytokine and chemokine levels related to T cell homeostasis are shown during the early phase (days 1-9 from symptom onset) and late phase (days 10-30 from symptom onset) of disease in SARS-CoV-2 infection compared to non-infected healthy controls. Each dot represents maximum value per individual subject during each phase of disease (early phase: mild, n=15; moderate, n=10; severe, n= 11 and late phase: mild, n=23; moderate, n=16; severe, n= 19 and non-infected healthy controls, n=18). (a, b, f, h) Significance was determined by two-sided Mann Whitney Wilcoxon test and p-values are indicated by asterisks (*, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001, ****, p ≤ 0.0001). Where multiple timepoints within the early (D1-9) or late (D10-30) phase per patient were available, the mean was taken and each patient is represented by one dot per time phase. n = 9 for HD, n = 15, 6, 6 for mild, moderate, and severe respectively in the early phase, and n = 18, 15, 17 for mild, moderate and severe in the late phase. (c) The MFI of HLA-DR on classical, intermediate, and non-classical monocytes is shown as a linear regression with peak serum GM-CSF levels in n = 24, 13, 18 mild (green), moderate (orange), and severe (red) patients respectively. (d) Linear regression is shown for peak serum IL-6 and peak clinical CRP level. Each dot represents the maximum value per individual subject during the disease (mild, n=17; moderate, n=18; severe, n= 21). (e) Linear correlations between peak serum gp130 or Il-6R levels versus DPSO in severe (red) and mild (green) patients; disease severity is indicated by color in mild (green, n = 23), moderate (orange, n= 19), and severe (