Viral RNA load in plasma is associated with critical illness and a dysregulated host response in COVID-19

A total of 250 adult patients with a positive nasopharyngeal swab polymerase chain reaction (PCR) test for SARS-CoV-2 performed at participating hospitals were recruited during the first pandemic wave in Spain from March 16th to the 15th of April 2020. The patients recruited were of three different categories. The first corresponded to patients examined at an emergency room and discharged within the first 24 h (outpatients group, n = 50). The second group were patients hospitalized to pneumology, infectious diseases or internal medicine wards (wards group, n = 100). Patients who required critical care or died during hospitalization were excluded from this group, in order to have a group of clear moderate severity. The third group corresponded to patients admitted to the ICU (n = 100). Patient`s recruited by participating hospital are detailed in the Additional file 1. Twenty healthy blood donors were included as controls. These controls were recruited during the pandemics, in parallel to the SARS-CoV-2 infected patients, and were negative for SARS-CoV-2 IgG. This study was registered at Clinicaltrials.gov with the identification NCT04457505.

Plasma from blood collected in EDTA tubes samples was obtained from the three groups of patients in the first 24 h following admission to the emergency room, to the ward, or to the ICU, at a median collection day since disease onset of 7, 8 and 10, respectively, and also from 20 blood donors (10 men and 10 women).

A panel of biomarkers was profiled by using the Ella-SimplePlex™ immunoassay (San Jose, California, USA), informing of the following biological functions potentially altered in severe COVID-19, based in the available evidence on COVID-19 physiopathology [13] [17] and also in our previous experience on emerging infections and sepsis [18‐21]: neutrophil degranulation: Lipocalin-2/NGAL, myeloperoxidase; endothelial dysfunction: ICAM-1, VCAM-1/CD106, angiopoietin-2; T cell survival and function: IL-7, Granzyme B; immunosuppression: IL-1ra, B7-H1/PD-L1, IL-10; chemotaxis: CXCL10/IP10, CCL2; Th1 response: interleukin 1 beta, IFN-γ, IL-12p70, IL-15, TNF-α, IL-2; Th2 response: IL-4, IL-10; Th17 response: IL-6, IL-17A; granulocyte mobilization / activation: G-CSF, GM-CSF; coagulation activation: D-Dimer; acute phase reactants: ferritin (C-reactive protein and LDH were profiled in each participant hospital by their central laboratories).

RNA was extracted from 100 µl of plasma using an automated system, eMAG® from bioMérieux® (Marcy l'Etoile, France). Detection and quantification of SARS-CoV-2 RNA was performed in five µl of the eluted solution using the Bio-Rad SARS-CoV-2 ddPCR kit according to manufacturer’s specifications on a QX-200 droplet digital PCR platform from the same provider. This PCR targets the N1 and N2 regions of the viral nucleoprotein gene and also the human ribonuclease (RNase) P gene using the primers and probes sets detailed in the CDC 2019 Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel [22]. Samples were considered positive for SARS-CoV-2 when N1 and/or N2 presented values ≥ 0.1 copies/µL in a given reaction. RNase P gene was considered positive when it presented values ≥ 0.2 copies/µL, following manufacturer`s indications. The test was only considered valid when RNase P gene was positive. Final results were given in copies of cDNA / mL of plasma. IgG specific for the nucleocapsid protein of SARS–CoV-2 was detected in 150 µl of plasma using the Abbott Architect SARS-CoV-2 IgG Assay (Illinois, USA). Viral RNA and SARS-CoV-2 IgG were profiled in the same plasma sample.

For the demographic and clinical characteristics of the patients, the differences between groups were assessed using the Chi-square test / Fisher's exact test where appropriated for categorical variables. Differences for continuous variables were assessed by using the Kruskal–Wallis test with post hoc tests adjusting for multiple comparisons. Multivariate logistic regression analysis was employed to evaluate the association between viral RNAemia and viral RNA load in plasma with severity, in the comparisons [outpatients vs ward patients] and [ward patients vs ICU patients]. Variables showing significant differences between groups in each comparison in the Kruskal–Wallis test were further introduced in the multivariate analysis as adjusting variables. The list of variables considered as potential adjusting variables were [Age (years)], [Sex (male)], [Alcoholism], [Smoker], [Drug abuse], [Cardiac disease], [Chronic vascular disease], [COPD], [Asthma], [Obesity], [Hypertension], [Dyslipidemia], [Chronic renal disease], [Chronic hepatic disease], [Neurological disease], [HIV], [Autoimmune disease], [Chronic inflammatory bowel disease], [Type 1 diabetes], [Type 2 diabetes], [Cancer], [Invasive mechanical ventilation], [Non-invasive mechanical ventilation], [SARS-CoV-2 IgG], [Temperature (ºC)], [Systolic pressure (mmHg)], [Oxygen saturation (%)], [Bilateral pulmonary infiltrate], [Glucose (mg/dl)], [Creatinine (mg/dl)], [Na (mEq/L)], [K (mEq/L)], [Platelets (cell × 10³ / µl)], [INR], [D Dimer (pg/ml)], [LDH (UI/L)], [GPT (UI/L)], [Ferritin (pg/ml)], [CRP (mg/dl)], [Haematocrit (%)], [Lymphocytes (cells/mm³)], [Neutrophils (cells/mm³)], [Monocytes (cells/mm³)]. Multivariate logistic regression analysis was performed using the “Enter” method, but also the backward stepwise selection method (Likelihood Ratio) was employed in each case to confirm the association between viral RNAemia and viral RNA load in plasma with disease severity (pin < 0.05, pout < 0.10), not forcing entry of these variables in the model. Correlation analysis was performed using the Spearman test applying the Bonferroni correction of the p value. Variables evaluated for correlation with viral RNA load were: [Temperature (ºC)], [Systolic pressure (mmHg)], [Oxygen saturation (%)], [Lymphocytes (cells/mm³)], [Neutrophils (cells/mm³)], [Monocytes (cells/mm³)], [Creatinine (mg/dl)], [LDH (UI/L)], [GPT (UI/L)], [Platelets (cell × 10³ / µl)], [INR], [CRP (mg/dl)], and all the biomarkers analysed by Ella-SimplePlex. Statistical analysis was performed with IBM SPSS® version 20 (IBM, Armonk, New York, USA).

Patients requiring hospitalization (either general ward or ICU) were older than those patients discharged to their home from the ER. Critically ill patients were more frequently male than those in the other groups. Comorbidities of obesity, hypertension, dyslipidemia and type 2 diabetes were more commonly found in patients requiring hospitalization, with no significant differences found in the comorbidities profile between critically ill and non-critically ill hospitalized patients. Fourteen per cent of the patients in clinical wards required non-invasive mechanical ventilation, while 96% of the patients admitted to the ICU required invasive mechanical ventilation. Critically ill patients had increased glucose levels, along with higher concentration of neutrophils in blood, increased levels of ferritin and C-reactive protein (denoting activation of the systemic inflammatory response). Increased levels of INR and D-dimers (reflecting activation of the coagulation system), as well as LDH and GPT, which levels raise as consequence of tissue and liver damage, were also observed in critically ill patients. Patients admitted to the ICU also showed a lower haematocrit, pronounced lymphopenia and lower monocyte counts at admission. ICU patients stayed longer in the hospital than ward patients, with 49% having a fatal outcome.

As depicted in Table 1, the frequency of the detection of SARS-CoV-2 viral RNA (RNAemia) was significantly higher in the critically ill patients (78%) compared to ward patients (27%) and outpatients (2%) (p < 0.001). Similarly, the group of critically patients showed higher viral RNA loads in plasma than either ward or outpatients (p < 0.001) (Table 1 and Fig. 1). Non-survivors showed the highest concentrations of viral RNA in plasma: viral RNA load (N1 region) in ICU non-survivors: 1587 copies / ml [10248]; viral RNA load (N1 region) in ICU survivors 574 copies / mL [1872] (results expressed as median [interquartile rank]); viral RNA load (N2 region) in ICU non-survivors: 2798 copies / ml [12012]; viral RNA load (N2 region) in ICU survivors 523 copies / mL [1478]. Patients admitted to the wards showed a significant higher frequency of viral RNAemia than outpatients, but viral RNA loads were not significantly different with the latter group (Table 1 and Fig. 1). Critically ill patients also had a higher frequency of specific SARS-CoV-2 IgG responses than the other groups (70% in ICU compared to 52% and 49% in the outpatients and ward groups, p < 0.05, Table 1). No significant differences were found between the group of outpatients and those admitted to the ward. The prevalence of viral RNAemia did not differ between those patients testing positive and those testing negative for SARS-CoV-2 IgG (43.8% and 40.4%, respectively, p = 0.586), who in addition showed no differences in viral RNA load (data not shown). Patients with viral RNAemia showed no differences in the days since onset of symptoms compared to those with no viral RNAemia (8.0 days [6.0]; 8.0 days [7.2], p = 0.965). In contrast, samples from patients with SARS-CoV-2 IgG were collected later since disease onset that those without SARS-CoV-2 IgG (10.0 days [7]; 7.0 days [6.0], p = 0.003).

While the proportion of patients with viral RNAemia was higher in the wards group compared to the outpatients’ group (Table 1), the multivariate analysis did not show a significant association between the presence of viral RNAemia and being hospitalized at the ward, with none of the both methods employed (Additional files 2 and 3). In contrast, when the ward group was compared with critically ill patients, a significant direct association was found between viral RNAemia and viral RNA load with critical illness, using the “Enter” method (Table 2) but also the backward stepwise selection method (Additional file 4).

viral RNA load in plasma (targeting either the N1 and the N2 regions) showed the strongest direct correlations with plasma levels of CXCL10, LDH, IL-10, IL-6, IL-15, myeloperoxidase and CCL-2 (MCP-1) and inverse correlations with lymphocytes, monocytes and O2 saturation (Fig. 2). These were the parameters whose levels varied the most in critically patients compared with ward and outpatient groups (Figs. 3, 4 and Additional file 5). CXCL10 was the most accurate identifier of viral RNAemia in plasma (area under the curve (AUC), [CI95%], p) = 0.85 [0.80 – 0.89), < 0.001), and IL-15 was the cytokine which most accurately differentiated clinical ward patients from ICU patients (AUC: 0.82 [0.76 – 0.88], < 0.001). Plasma viral RNA load also showed significant direct correlations with levels of VCAM-1, PDL-1, GM-CSF, G-CSF, neutrophil counts, IL-1ra, CRP, INR, D-dimer, TFNα, angiopoietin-2, GPT, ICAM-1, IL-7 and ferritin (Fig. 2), with most of these mediators showing the highest variations in the critically ill patients (Figs. 3, 4 and Additional file 5).