Results

All Patients with Severe Respiratory Failure Caused by SARS-CoV-2 Have Immune Dysregulation or MAS
We assessed the differences of immune activation and dysregulation between SARS-CoV-2 and other known severe infections in three patient cohorts: 104 patients with sepsis caused by bacterial CAP; 21 historical patients with 2009 H1N1 influenza; and 54 patients with CAP caused by SARS-CoV-2. Patients with bacterial CAP were screened for participation in a large-scale randomized clinical trial with the acronym PROVIDE (ClinicalTrials.gov NCT03332225). Patients with 2009 H1N1 influenza have been described in previous publications of our group (Giamarellos-Bourboulis et al., 2009, Raftogiannis et al., 2010). The clinical characteristics of patients with bacterial CAP and CAP caused by COVID-19 are described in Table 1 . Each cohort (bacterial sepsis and COVID-19) is split into patients who developed SRF and required MV and those who did not. Three main features need to be outlined: (1) patients with COVID-19 and SRF are less severe than those with severe bacterial CAP, on the basis of the traditional severity scores of sequential organ failure assessment (SOFA) and acute physiology and chronic health evaluation (APACHE) II; (2) this leads to the conclusion that COVID-19 patients undergo an acute immune dysregulation with deterioration into SRF before the overall state of severity is advanced; and (3) although the burden of co-morbidities of patients with COVID-19, as expressed by the Charlson’s co-morbidity index, is higher among patients with SRF than among patients without SRF, it remains remarkably lower that traditional bacterial CAP and sepsis. It was also notable that the admission values of Glasgow Coma Scale scores of patients with bacterial CAP were 8.80 ± 4.76, and that of patients with COVID-19 was 14.71 ± 0.20 (p < 0.0001). This finding is fully compatible with clinical descriptions of severe COVID-19: patients are admitted at a relatively good clinical state and suddenly deteriorate.
Table 1 Baseline Clinical and Laboratory Characteristics of the Cohorts of Bacterial CAP and of Pneumonia Caused by SARS-CoV-2
No respiratory failure Severe respiratory failure
Bacterial SARS-CoV-2 p value Bacterial SARS-CoV-2 p value
Number of patients 48 26 56 28
Age (years, mean ± SD) 74.8 ± 16.8 59.2 ± 10.3 <0.0001 74.0 ± 12.6∗ 67.8 ± 10.8# <0.0001
Male gender (n, %) 25 (52.1) 15 (57.7) 0.807 27 (48.2)∗ 25 (89.3)∗∗ 0.0003
APACHE II score (mean ± SD) 18.50 ± 8.19 5.88 ± 3.40 <0.0001 26.63 ± 8.52∗∗ 10.17 ± 3.64## <0.0001
SOFA score (mean ± SD) 7.87 ± 3.81 1.50 ± 0.82 <0.0001 11.46 ± 3.15∗∗ 5.71 ± 2.19## <0.0001
CCI (mean ± SD) 5.53 ± 2.13 2.16 ± 1.46 <0.0001 5.57 ± 2.20∗ 3.39 ± 2.16## <0.0001
PSI (mean ± SD) 146.5 ± 43.2 80.0 ± 30.7 <0.0001 177.4 ± 40.4∗∗∗ 121.2 ± 28.3## <0.0001
Laboratory values (mean ± SD)
 Total white blood cell count (/mm3) 13,852.7 ± 7279.3 6379.6 ± 1993.9 <0.0001 17,666.9 ± 12,799.9∗ 9447.8 ± 3308.6## <0.0001
 Absolute platelet count (x103 /mm3) 201.3 ± 124.8 243.8 ± 109.1 0.141 224.3 ± 111.0∗ 213.9 ± 71.8∗ 0.654
 INR 1.28 ± 0.64 1.11 ± 0.15 0.187 1.33 ± 0.45∗ 1.17 ± 0.20∗ 0.077
 aPTT (secs) 37.19 ± 12.95 33.40 ± 6.22 0.165 38.42 ± 23.00∗ 37.52 ± 9.88∗ 0.844
 Fibrinogen (mg/dl) 475.6 ± 196.3 528.9 ± 152.5 0.234 495.3 ± 290.5∗ 693.5 ± 188.6# 0.002
 D-dimers (g/dl) 7.66 ± 13.9 2.76 ± 2.02 0.079 1.46 ± 1.62∗∗ 5.43 ± 6.41∗ <0.0001
 Creatinine (mg/dl) 1.55 ± 1.00 0.85 ± 0.19 0.001 1.71 ± 0.90∗ 1.11 ± 0.43# 0.001
 Total bilirubin (mg/dl) 1.43 ± 1.92 0.67 ± 0.50 0.052 1.17 ± 1.88∗ 0.97 ± 0.68∗ 0.588
 AST (U/l) 155.1 ± 308.1 39.9 ± 28.5 0.062 311.7 ± 748.2∗ 76.6 ± 59.2# 0.102
 ALT (U/l) 234.6 ± 764.3 40.2 ± 24.9 0.200 175.8 ± 378.0∗ 64.3 ± 62.2∗ 0.126
Main comorbidities (n, %)
 Type 2 diabetes mellitus 13 (27.1) 4 (15.4) 0.386 21 (37.5)∗ 6 (21.4)∗ 0.214
 Chronic heart failure 8 (16.7) 3 (11.5) 0.737 18 (32.1)∗ 4 (14.3)∗ 0.114
 Coronary heart disease 7 (14.6) 2 (7.7) 0.479 10 (17.9) 5 (17.9)∗ 1.0
Comparisons with the respective groups without respiratory failure by the Student’s t test: ∗p-non-significant; ∗∗p < 0.05; #p < 0.001; ##p < 0.0001. Abbreviations are as follows: ALT, alanine aminotransferase; aPTT, activated partial thromboplastin time; AST, aspartate aminotransferase; APACHE, acute physiology and chronic health evaluation; CCI, Charlson’s comorbidity index; INR, international normalized ratio; PSI, pneumonia severity index; SD, standard deviation; SOFA, sequential organ failure assessment
Immune classification of patients with SARS-CoV-2 was performed by using the tools suggested for bacterial sepsis, i.e., ferritin more than 4,420 ng/mL for MAS (Kyriazopoulou et al., 2017), and HLA-DR molecules on CD14 monocytes lower than 5,000, in the absence of elevated ferritin, for the immune dysregulation phenotype (Lukaszewicz et al., 2009). It was found that contrary to the patients with bacterial CAP and SRF, all patients with SRF and SARS-CoV-2 had either immune dysregulation or MAS (Table 2 ).
Table 2 Association between Severe Respiratory Failure and Immune Classification among Patients with COVID-19 and Patients with Sepsis Caused by Bacterial CAP
Number of patients with SRF/total patients [%, (−95% CI, +95% CI)] p value∗
Bacterial CAP COVID-19
Intermediate 21/40 [52.5 (27.5-67.1)] 0/26 [0 (0-12.9)] <0.0001
Immunoparalysis (for bacterial CAP) and immune dysregulation (for COVID-19) or MAS 35/64 [54.7 (42.6-66.3)] 28/28 [100 (87.9-100)] <0.0001
∗comparisons by the Fisher exact test. Abbreviation is as follows: CI, confidence interval.

Major Decrease of HLA-DR on CD14 Monocytes Is Associated with SRF
Immunoparalysis of sepsis is characterized by significant decrease of the number of HLA-DR molecules on CD14 monocytes. This also happens in immune dysregulation caused by SARS-CoV-2 (Figures 1A and 1B). Although patients with bacterial-CAP-associated MAS also have decreased HLA-DR molecules on CD14 monocytes, their circulating ferritin is significantly higher than normal. This is a feature found only in a few patients with SARS-CoV-2 and MAS (Figure 1C). The absence of traits of MAS among cases of SARS-CoV-2 with immune dysregulation is further proven by the low scores of hemophagocytosis (HS). HScore is proposed as a classification tool for secondary MAS, and values more than 169 are highly diagnostic (Fardet et al., 2014). Seven patients with SARS-CoV-2 had HS above this cut-off, and all were properly classified by using ferritin (Figure 1D). Among patients with bacterial CAP at an intermediate immune state, the number of molecules of HLA-DR on their CD14 monocytes was lower than in healthy patients. However, patients with pneumonia caused by SARS-CoV-2 at an intermediate immune state maintained their number of molecules of HLA-DR on CD14 monocytes much closer to the healthy condition. When this number suddenly dropped, SRF supervened (Figures 1A and 1B). Moreover, the absolute counts of neutrophils and monocytes were higher among patients with immune dysregulation than patients with MAS (Figures 1E and 1F).
Figure 1 Characteristics of Immune Dysregulation of COVID-19
(A) Absolute numbers of the molecules of the human leukocyte antigen (mHLA-DR) on CD14 monocytes. Patients with bacterial CAP and CAP caused by SARS-CoV-2 are classified into three states of immune activation: intermediate, immunoparalysis for bacterial CAP and dysregulation for COVID-19, and MAS.
(B) Mean fluorescence intensity (MFI) of HLA-DR on CD14 monocytes of healthy volunteers and of patients with CAP caused by SARS-CoV-2 classified according to their state of immune activation.
(C) Ferritin concentrations in the serum of patients with bacterial CAP and sepsis and CAP caused by SAR-CoV-2 according to their state of immune activation.
(D) Hemophagocytosis score among patients with CAP caused by SARS-CoV-2 classified according to their state of immune activation.
(E) Absolute neutrophil counts among patients with CAP caused by SARS-CoV-2 classified according to their state of immune activation.
(F) Absolute monocyte counts among patients with CAP caused by SARS-CoV-2 classified according to their state of immune activation.
Bars in each graphic represent mean values and standard errors. Statistical comparisons are indicated by the arrows; ns: non-significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Comparisons were done by the Mann-Whitney U test followed by correction for multiple comparisons.

CD4 Cell and NK Cell Cytopenias Are Characteristics of Infection by SARS-CoV-2
The absolute counts of CD3+CD4+CD45+ lymphocytes, CD3+CD8+CD45+ lymphocytes, CD3−CD16+CD56+CD45+ cells, and CD19+CD45+ lymphocytes were lower among patients with COVID-19 compared with those in 10 healthy subjects adjusted for age and gender. Compared with patients with CAP caused by 2009H1N1, patients with COVID-19 had lower CD3+CD4+CD45+ lymphocytes but higher CD3+CD16+CD45+ cells and CD19+CD45+ lymphocytes (Figures 2A–2E). Those patients with immune dysregulation caused by COVID-19 had lower counts of CD3+ CD4+ CD45+ lymphocytes, CD3+CD8+CD45+ lymphocytes, and CD3−CD16+CD56+CD45+ cells than those at an intermediate immune state. When comparisons were limited to patients with SRF and infection by one of the two viruses, it was found that infection by SARS-CoV-2 was accompanied by lower CD3+CD4+CD45+ lymphocytes but with higher CD3−CD16+CD56+CD45+ cells and CD19+CD45+ lymphocytes than 2009H1N1 (Figures 2F–2J). The Th17 function as assessed by IL-17 production capacity was downregulated among patients with immune dysregulation (Figure 2K).
Figure 2 CD4 Cell and NK Cell Cytopenias Are Characteristics of Infection by SARS-CoV-2
(A–E) Absolute counts of CD3+CD4+CD45+ lymphocytes (A), CD3+CD8+CD45+ lymphocytes (B), CD3+CD16+CD56+CD45+ lymphocytes (C), CD3−CD16+CD56+CD45+ cells (D), and CD19+CD45+ lymphocytes (E) among healthy volunteers, patients with CAP caused by the 2009H1N1 influenza virus, and patients with CAP caused by COVID-19. Patients with COVID-19 are classified into three states of immune activation: intermediate, dysregulation, and MAS.
(F–J) absolute counts of CD3+CD4+CD45+ lymphocytes (F), CD3+CD8+CD45+ lymphocytes (G), CD3+CD16+CD56+CD45+ lymphocytes (H), CD3−CD16+CD56+CD45+ cells (I), and CD19+CD45+ lymphocytes (J) among patients with severe respiratory failure developing in the field of CAP caused by the 2009H1N1 influenza virus and COVID-19.
(K) IL-17 production by PBMCs after stimulation with heat-killed Candida albicans.
Bars in each graphic represent mean values and standard errors. Statistical comparisons are indicated by the arrows; ns: non-significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Comparisons were done by the Mann-Whitney U test followed by correction for multiple comparisons.
The next question was how this change of CD19+CD45+ lymphocytes is translated to serum immunoglobulins (Igs). Concentration of IgGs and their subclasses in the plasma of COVID-19 patients was low, as it was in bacterial CAP (Figures 3A–3D). However, IgM and IgA were higher than in bacterial CAP (Figures 3E and 3F). Overall, patients at MAS had lower IgG2, IgM, and IgA than those at an intermediate immune state, and patients at immune dysregulation had lower IgM than those at an intermediate immune state.
Figure 3 Moderate Derangement of Circulating Immunoglobulins in Pneumonia Caused by SARS-CoV-2
Serum amounts of IgG subclasses (A–D), IgM (E), and IgA (F) of patients with CAP caused by SARS-CoV-2 are shown. Patients are classified into three states of immune activation: intermediate, dysregulation, and MAS. Findings are compared with those in patients with bacterial CAP, who are classified into three states of immune activation: intermediate, immunoparalysis, and MAS.
Bars in each graphic represent mean values and standard errors. Only statistically significant comparisons are indicated by the arrows; ∗p < 0.05; ∗∗p < 0.0001; ∗∗∗p < 0.0001. Comparisons were done by the Mann-Whitney U test followed by correction for multiple comparisons.

The IL-6 Blocker Tocilizumab Partially Rescues the Immune Dysregulation Driven by SARS-CoV-2
Sepsis-induced immunoparalysis is characterized by profound deficiency of monocytes for cytokine production upon ex vivo stimulation (Giamarellos-Bourboulis et al., 2011). Indeed, production of tumor necrosis factor-α (TNF-α) by LPS-stimulated peripheral blood mononuclear cells (PBMCs) of patients with bacterial CAP classified for immunoparalysis was significantly lower than in patients at an intermediate state (Figure 4 A). That was not the case for patients with pneumonia caused by SARS-CoV-2, in whom PBMCs showed sustained TNF-α production after stimulation with LPS (Figure 4B). The function of PBMCs in patients with SRF caused by 2009H1N1 was also impaired, and there was lower TNF-α production, a pattern different from COVID-19 patients (Figure 4C). Surprisingly, stimulation of IL-1β was lower among patients with immune dysregulation than among patients with an intermediate immune state (Figures 4D–4F). IL-6, however, followed the stimulation pattern of TNF-α (Figures 4G–4I). This generated the hypothesis that in the case of SRF-aggravated pneumonia caused by SARS-CoV-2, there is a unique combination of defective antigen presentation and lymphopenia that leads to defective function of lymphoid cells, whereas monocytes remain potent for the production of TNF-α and IL-6.
Figure 4 Main Features of Immune Dysregulation of Pneumonia Caused by SARS-CoV-2
(A) Production of TNF-α by PBMCs of patients with sepsis caused by bacterial CAP classified into intermediate state of immune activation and immunoparalysis.
(B) Production of TNF-α by PBMCs of patients with CAP caused by SARS-CoV-2 classified into three states of immune activation: intermediate, dysregulation, and MAS.
(C) Production of TNF-α by PBMCs of patients with SRF developing after infection caused by the 2009H1N1 virus and by SARS-CoV-2.
(D) Production of IL-1β by PBMCs of patients with sepsis caused by bacterial CAP classified into intermediate state of immune activation and immunoparalysis.
(E) Production of IL-1β by PBMCs of patients with CAP caused by SARS-CoV-2 classified into three states of immune activation: intermediate, dysregulation, and MAS.
(F) Production of IL-1β by PBMCs of patients with SRF developing after infection caused by the 2009H1N1 virus and by SARS-CoV-2.
(G) Production of IL-6 by PBMCs of patients with sepsis caused by bacterial CAP classified into intermediate state of immune activation and immunoparalysis.
(H) Production of IL-6 by PBMCs of patients with CAP caused by SARS-CoV-2 classified into states of immune activation: intermediate, dysregulation and MAS.
(I) Production of IL-6 by PBMCs of patients with SRF developing after infection caused by the 2009H1N1 virus and by SARS-CoV-2.
(J–L) Serum amounts of TNF-α, IL-6, and CRP of patients with CAP caused by SARS-CoV-2 classified into states of immune activation: intermediate, dysregulation and MAS.
Bars in each graphic represent mean values and standard errors. Statistical comparisons are indicated by the arrows; ns: non-significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Comparisons were done by the Mann-Whitney U test followed by correction for multiple comparisons.
As a next step, we measured circulating concentrations of TNF-α, interferon-γ (IFN-γ), IL-6, and C-reactive protein (CRP) in patients infected by SARS-CoV-2. IFN-γ was below the limit of detection in all patients (data not shown), indicating that Th1 responses do not contribute to over-inflammation. No differences of circulating TNF-α concentrations were found between COVID-19 patients at the three states of immune classification (Figure 4J). In contrast, IL-6 and CRP concentrations were significantly higher among patients with immune dysregulation than among patients at an intermediate state of immune activation (Figures 4K and 4L). Given that IL-6 was below the limit of detection in some patients with immune dysregulation, we divided them into two groups as follows: seven patients with IL-6 below the limit of detection and 14 patients with detectable IL-6. Their severity was similar given that SOFA score and pneumonia severity indexes were similar (p values of comparisons were 0.937 and 0.877, respectively; data not shown).
IL-6 is known to inhibit HLA-DR expression (Ohno et al., 2016), leading to the hypothesis that IL-6 over-production mediates the low HLA-DR expression on CD14 monocytes of COVID-19 patients. In agreement with this, negative correlation was found between serum amounts of IL-6 and the absolute number of HLA-DR molecules on CD14 monocytes of patients with COVID-19 but also between the absolute lymphocyte count and the absolute number of mHLA-DR on CD14 monocytes of patients with COVID-19 (Figures 5A and 5B). Furthermore, PBMCs from patients with immune dysregulation were cultured overnight in the presence of plasma of the COVID-19 patients, which was already shown to be rich in IL-6. The expression of HLA-DR on CD14 monocytes was strongly inhibited by COVID-19 plasma from patients with immune dysregulation but not by plasma from patients with an intermediate immune state of activation (Figures 5C–5F); the addition of the specific blocker of the IL-6 pathway Tocilizumab partially restored the expression of HLA-DR on monocytes of all patients with immune dysregulation (Figures 5E and 5F). Treatment with Tocilizumab in six patients was accompanied by increase of the absolute lymphocyte blood count within the first 24 h (Figure 5G). IL-6 was produced partly by CD14 monocytes and partly by CD4 cells (Figure 5H).
Figure 5 Immune Dysregulation Caused by SARS-CoV-2 Is Mediated by IL-6
(A) Negative correlation between serum amounts of IL-6 and the absolute numbers of the mHLA-DR on CD14 monocytes. The Spearman’s (rs) co-efficient of correlation and the respective p value are provided.
(B) Correlation between the absolute lymphocyte count and the absolute numbers of mHLA-DR on CD14 monocytes. The rs co-efficient of correlation and the respective p value are provided.
(C) Changes of the absolute numbers of mHLA-DR on CD14 monocytes of four patients infected by SARS-CoV-2 with intermediate state of immune activation after incubation with medium and their plasma.
(D) Changes of the MFI of HLA-DR on CD14 monocytes of four patients infected by SARS-CoV-2 with intermediate state of immune activation after incubation with medium and their plasma.
(E) Changes of the absolute numbers of mHLA-DR on CD14 monocytes of eight patients infected by SARS-CoV-2 with immune dysregulation after incubation with medium and their plasma; modulation by the addition of the specific IL-6 blocker Tocilizumab is also shown.
(F) Changes of the MFI of HLA-DR on CD14 monocytes of eight patients infected by SARS-CoV-2 with immune dysregulation after incubation with medium and their plasma; modulation by the addition of the specific IL-6 blocker tocilizumab is also shown.
(G) Changes of the absolute lymphocyte count of six patients before and after start of treatment with Tocilizumab.
(H) Intracellular staining for IL-6 in CD14 monocytes and in CD4 lymphocytes of three patients infected by SARS-CoV-2 with immune dysregulation.
Statistical comparisons are indicated by the arrows; ns: non-significant; ∗p < 0.05; ∗∗p < 0.01.