1 INTRODUCTION

Coronavirus disease 2019 (COVID-19) is an infectious disease characterized by fever, dry cough, anosmia, and dyspnea secondary to typical lung involvement. Although most hospitalized patients progress favorably, 5%–15% of them develop a severe cytokine storm resulting in a respiratory distress syndrome and multi-organ failure.1 Endothelial injury, hypercoagulability, and decreased blood flow (i.e., the Virchow triad) are suspected to play a key role in thrombosis, multiorgan failure, and death of patients with COVID-19.2 While increased inflammation, endothelial dysfunction, platelet activation, hypofibrinolysis, and hypercoagulation have been previously reported in patients with COVID-19,3-7 studies investigating blood flow and its determinants are rather scarce. A preliminary study from our group8 performed on only seven patients with COVID-19, showed no difference in blood viscosity compared to healthy individuals. In contrast, Joob et al.9 reported higher blood viscosity in individuals with COVID-19 compared to healthy controls. However, blood viscosity was not measured but, instead, was calculated from the hematocrit and plasma total protein content using the calculation of Nwose et al.10 Although practical, this formula does not take into account the influence of red blood cell (RBC) aggregation and deformability on blood viscosity.11, 12 By using the same formula, Mungmunpuntipantip et al.13 observed higher estimated blood viscosity in fatal COVID-19 patients compared to COVID-19 survivors. But, again, blood viscosity was not directly measured in this study. Thus, the link between increased blood viscosity and COVID-19-related complications and severity is not formerly established and can only be suspected.

Blood is a non-newtonian and shear-thinning fluid with its viscosity being dependent on hematocrit, plasma viscosity, RBC aggregation (at low to moderate shear rates), and deformability (mainly at high shear rates).12, 14 Both RBC aggregation and deformability may impact on RBC flow properties, independently of their effects on blood viscosity.14 For instance, increased RBC aggregation may modulate vascular function through its effects on the distribution of flowing RBCs and the resulting local wall shear stress,15 and may promote stasis and impede microcirculation.14, 16 Decreased RBC deformability may result in increased resistance in the pulmonary microcirculation17 and RBC trapping in various organs such as spleen, heart, and lungs.18 Recent lipidomic analyses showed alterations in RBC membrane lipids composition in COVID-19 patients.19 Moreover, despite no change in the total levels of key structural proteins, proteomic analyses showed minor increase in band 3 protein in RBC from COVID-19 patients probably due to protein fragmentation.19 In our preliminary report,8 we observed increased RBC aggregation and slightly decreased RBC deformability in patients with COVID-19 compared to controls but whether these RBC abnormalities could contribute to COVID-19 severity is currently unknown. In addition, whether impaired RBC rheology could contribute to blood hyper-coagulation in COVID-19 has not been investigated.

The aim of this study was to characterize blood rheological parameters in hospitalized patients with COVID-19 and to test the associations with several indicators of clinical severity. We also tested the associations between blood rheology and blood coagulation in a subset of COVID-19 patients.

2 MATERIAL AND METHODS 2.1 Population

Between January and May 2021, a total of 172 patients with COVID-19, hospitalized in the COVID-unit of the Internal Medicine Department of the Edouard Herriot Hospital (Hospices Civils de Lyon, Lyon, France) participated in this study, after giving informed consent. Clinical reports were reviewed by a physician to collect anthropometric data, heart rate, arterial oxygen saturation (SpO2), systolic and diastolic arterial pressures, medical history (obesity, smoking, high blood pressure [HBP], diabetes, cardiac, hepatic, renal or respiratory diseases, stroke or cancer), extent of pulmonary injuries, presence of pulmonary or deep vein thrombosis, supplemental oxygen flow rates, main treatments, time of hospitalization, and clinical outcome (recovery, worsening, and transfer to intensive care unit or death). A group of 38 healthy individuals with no apparent disease (age 56.0 ± 14.0 years; mean ± SD) was composed for comparisons of the blood rheological and RBC parameters. Blood was collected into EDTA and citrated tubes for different laboratory analyses on the first day of hospitalization. All patients were informed about the purpose and procedures of the study, which was approved by the local ethics committee (CPP Est IV 20–41 and 20–108).

2.2 Biochemical and hematological parameters

Plasma fibrinogen and C-reactive protein levels were determined in 98 and 151 COVID-19 patients, respectively, by standard coagulation and biochemical methods. Hemoglobin, mean corpuscular volume (MCV), mean corpuscular hemoglobin concentration (MCHC), RBC, platelets (PLT), and white blood cell (WBC) counts were determined using a hematology analyzer (XN-9000; Sysmex Corporation, Kobe, Japan) in COVID patients. Hematocrit (Hct) was determined by microcentrifugation in both the COVID-19 and control groups.

2.3 Blood viscosity, RBC deformability, and RBC morphology

Blood rheological parameters were measured for the two groups after full blood oxygenation and within 4 h after sampling, as recommended.11 Blood viscosity was measured at native Hct using a cone/plate viscometer (Brookfield DVII+ with CPE40 spindle; Brookfield Engineering Labs, Natick, MA) at 11.5, 22.5, 45, 90, and 225 s−1.

RBC deformability was determined at 37°C and at shear stresses ranging from 0.95 to 30 Pa by ektacytometry (LORRCA MaxSis; RR Mechatronics, Hoorn, The Netherlands). The system has been described elsewhere in detail.11 Briefly, 25 μL of prepared blood suspension was mixed with 5 mL polyvinylpyrrolidone (PVP; viscosity = 30 cP) and sheared into a Couette system made of glass. The diffraction pattern was analyzed by the computer to calculate an elongation index reflecting RBC deformability. The elongation index-shear stress curves were subsequently parameterized with a Lineweaver-Burk fitting to determine the maximal theoretical elongation that could be achieved at an infinite shear stress (EImax) and the magnitude of shear stress required to induce half of EImax (SS1/2). The SS1/2/EImax ratio provides an index of RBC deformability that facilitates effective and simple comparisons between RBC samples with increased SS1/2/EImax ratio reflecting decreased RBC deformability.20 Blood smears were performed for 132 COVID-19 patients and RBC morphologies were screened by optical microscopy, focusing on the presence of spherocytes, stomatocytes, echinocytes, and elliptocytes.

Osmolality gradient ektacytometry (osmoscan; LORRCA MaxSis; RR Mechatronics) was also performed to get insights into RBC deformability modulators such as internal viscosity, surface, and volume of RBCs. RBC deformability was measured at a shear stress of 30 Pa with increasing osmolality from 90 to 600 mOsm/kg and at 37°C, as recommended.21 Several parameters were determined: Omin (i.e., the osmolality at which RBC deformability value reaches a minimum in the hypotonic region of the curve), EImax (i.e., the highest RBC deformability), and Ohyper (also called O′), which corresponds to the osmolality at half of the EImax on the hypertonic region of the curve.21 Omin reflects the osmotic fragility and the surface-to-volume ratio, EImax depends on the membrane deformability and RBC surface area, and Ohyper reflects MCHC and MCV and is, therefore, highly dependent on the hydration status of the cells.22

2.4 RBC aggregation

RBC aggregation was determined at 37°C via syllectometry, (i.e., laser backscatter versus time, using the LORCA) after adjustment of the Hct to 40%. In addition, the RBC disaggregation threshold (i.e., RBC aggregates strength) was determined using a re-iteration procedure23: seven separate pre-defined shear rates between 7.5 and 800 s−1 were applied on the RBC suspension, with or without alternating disaggregation shear rate, to locate the minimal shear rate needed to prevent RBC aggregation.

To determine the role of plasma and cellular factors on RBC aggregation in COVID-19 patients, further experiments were done with the Myrenne aggregometer (GmbH, Roetgen). The Myrenne aggregometer uses light transmission principle and allows the measurement of two RBC aggregation indices: RBC aggregation in stasis (M index) and at low shear rate (M1).11 RBC aggregation was determined either in autologous plasma or in PBS-Dextran suspension without plasma (Sigma-Aldrich, 70 kDa, 3%) after adjustment of Hct at 40%. The role of fibrinogen on RBC aggregation was also tested with the Myrenne aggregometer after incubation of RBC from COVID-19 patients with Ancrod (1.2 U/mL, Sigma Aldrich) for 2 h at room temperature (RT) (22–25°C). Ancrod is a fibrinogenolytic compound derived from the venom of the Malayan pit viper Calloselasma rhodostoma. Ancrod fibrin monomers form defective polymers, more susceptible to plasmin digestion than are thrombin fibrin clots.24

2.5 RBC senescence markers

The following RBC senescence markers were measured: phosphatidylserine (PS) and CD47 at the surface of RBCs, internal Ca2+, and reactive oxygen species (ROS).25 Blood was centrifuged (800 g, 10 min at 20°C) and plasma and buffy coat were discarded. RBCs were washed in PBS 1X, and RBC pellets resuspended at 0.4% Hct in PBS buffer containing 2.5 mM Ca2+ (for PS, ROS, CD47, and Ca2+ analysis) or 5 mM EDTA (for the PS negative control).

PS exposure on the outer membrane leaflet of the RBCs was evaluated by using Annexin V-FITC. RBC suspensions were protected from light and incubated for 30 min at RT with Annexin V-FITC (1:200 dilution, Beckman Coulter, CA, USA). Immediately after incubation, samples were diluted and analyzed by flow cytometry (FACS, BD Accuri C6, Franklin Lakes, USA). PS exposure was measured in the FITC channel (with an excitation wavelength of 488 nm and an emission wavelength of 530 nm) according to manufacturer's instructions. Negative controls were obtained by replacing Ca2+ by EDTA to prevent Annexin V from binding to PS. For each sample, 50 000 events, gated for the appropriate Forward Scatter (FSC), were counted and the percentage of PS positive RBCs was determined.

RBC CD47 exposure was quantified by incubating RBC suspensions at 0.4% Hct with anti-CD47 antibody coupled with phycoerythrin (1:33 dilution, Miltenyi, Bergisch Gladbach, Germany), for 30 min at RT in the dark. The samples were then analyzed using FACS, according to manufacturer's instructions. The Median Fluorescence Intensity (MFI) of 50 000 gated events were recorded to quantify CD47 exposure.

Internal ROS content was determined using 2′,7′–dichlorofluorescin diacetate (DCFDA, Sigma-Aldrich, Saint-Quentin-Fallavier, France). RBC suspensions at 0.4% Hct were incubated for 30 min at RT in the dark with 10 μM of DCFDA (Sigma-Aldrich, Saint-Quentin-Fallavier, France). The samples were then analyzed using FACS, according to manufacturer's instructions. The MFI of 50 000 gated events was recorded to quantify ROS levels.

RBC Ca2+ content was measured with a Fluo3/AM (Biotium, Fremont, USA) probe. RBC suspensions were incubated for 30 min at RT with 5 μM of Fluo3/AM, and analyzed using FACS, according to the manufacturer's instructions. MFI of the 50 000 gated events was recorded to quantify Ca2+ levels.

2.6 Rotational thromboelastometry

Coagulation activation and clot polymerization parameters, as well as clot lysis parameters were determined in a subset of COVID-19 patients (n = 76) with rotational thromboelastometry (ROTEM® delta, Werfen, TEM International, Germany) in FIBTEM mode at 37°C for 45 min.

The addition of a potent platelet inhibitor (cytochalasin D) blocks platelet activation, shape change, expression, and activation of glycoprotein IIb/IIIa (fibrinogenreceptor).26 Thereby, platelet contribution to clot formation and clot strength is eliminated in this assay.27 Samples are recalcified with 0.2 M CaCl2-solution and tissue thromboplastin is added.

Several parameters were derived: (1) coagulation time (CT), which is defined by the time from test start until a clot firmness amplitude of 2 mm is reached, (2) clot formation time (CFT), which corresponds to the time between 2 and 20 mm clot firmness amplitude is achieved, (3) alpha angle (α), which reflects the kinetics of clot formation, too, and corresponds to the angle between the baseline and a tangent to the clotting curve through the 2 mm point, (4) maximum clot firmness (MCF) in millimeters, which corresponds to the maximum amplitude of clot firmness reached during the test, (5) the amplitude of clot firmness 5, 10, and 20 min after CT (A5, A10, and A20, respectively), (6) maximum lysis (ML, expressed in % of MCF), which corresponds to the difference between MCF and the lowest amplitude after MCF, (7) lysis indices 30 and 60 (LI30 and LI60), which indicate the percentage of MCF still present 30 and 60 min after CT, respectively.

2.7 Statistical analysis

Data are expressed as mean ± SD. Comparisons were achieved by using paired or unpaired student T-tests, or Analysis of Variance followed by Tukey' post hoc tests, when appropriate. Pearson correlations were performed to test the associations between different parameters.

GraphPad Prism 9 (La Jolla, CA, USA) and SPSS 23.0 (IBM, Armonk, NY, USA) software were used for statistical analyses. A p-value < .05 was considered significant.

3 RESULTS

General parameters, medical history/comorbidities, clinical, and hematological parameters in patients with COVID-19 are displayed in Tables S1 and 1. Overweight/obesity, diabetes, and hypertension were highly prevalent in our cohort. Only 7% of the cohort had no sign of pulmonary lesions while 24% had more than 50% of pulmonary lesions. The main treatments were oxygen therapy, anticoagulants, and corticoids. Thirty-two percent of the hospitalized cohort were transferred to intensive care unit and 3% died. Hematological and biochemical analyses showed increased CRP and fibrinogen levels, but no sign of RBC dehydration (normal MCV and MCHC) was observed. Mean percentage of reticulocyte was into the normal range (0.2%–2%) but 23% of patients had greater values than 2%.

TABLE 1. Clinical and hematological parameters in patients with COVID-19 Physiological parameters Normal range/values SpO2 (%) 97 ± 2 95–100 HR (bpm) 93 ± 17 60–90 SBP (mmHg) 134 ± 23 <140 DBP (mmHg) 75 ± 15 <90 Extent of pulmonary lesions (%) None 7 <25% 33 25%–50% 36 50%–75% 19 >75% 5 Treatments Oxygen supplementation (%) 90 Oxygen flow rate (L/min) 2.0 ± 2.2 Hydration (%) 23 Antibiotherapy (%) 45 Anticoagulant (%) 73 Corticoïds (%) 84 Clinical outcomes Pulmonary/peripheral vein thrombosis (%) 5.2 Intensive care unit (%) 32 Death (%) 3 Hospitalization length (days) 14.1 ± 12.5 Hematological parameters Normal range/values White blood cells (G/L) 8.1 ± 3.3 3.78–9.92 Red blood cells (T/L) 4.4 ± 0.6 3.99–5.57 Hemoglobin (g/dL) 12.8 ± 1.7 11.5–15.0 MCV (fL) 87.6 ± 6.2 79.9–97.0 MCHC (g/L) 331 ± 130 319–363 Reticulocytes (%) 1.6 ± 1.3 0.8–2.6 Platelets (G/L) 325 ± 144 161–420 CRP (mg/L) 46.0 ± 45.8a <10 Fibrinogen (g/L) 5.8 ± 1.5a 2.0–4.0 Abbreviations: CRP, C-reactive protein; DBP, diastolic blood pressure; HR, heart rate; MCHC = mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; SBP, systolic blood pressure; SpO2, arterial oxygen saturation. 3.1 Blood viscosity is higher in COVID-19 patients compared to controls despite lower hematocrit and unaltered RBC deformability

Hematocrit was significantly lower in COVID-19 patients compared to the control group (Figure 1A). Hematocrit and percentage of reticulocytes were negatively correlated in COVID-19 patients (r = −0.26; p < .01). Blood viscosities were greater in the COVID-19 group compared to the control group at all shear rates (Figure 1B). Neither SS1/2/EImax (Figure 1C), nor the osmoscan parameters (Figure 1E–F) differed between the two groups. RBC senescence markers (Figure 1G–J) were not different between the two groups although percentage of RBC expressing PS tended to be greater in patients with COVID-19 compared to controls (Figure 1G). The analysis of RBC morphology in the blood of COVID-19 patients showed that 26% of the 132 for which blood smears have been done had at least 1% of their RBC with abnormal shape (7% with spherocytes, 9% with echinocytes, 8% with stomatocytes, and 2% with elliptocytes). The mean proportion of abnormal RBC in this subgroup was 3.0 ± 1.7%. Slightly higher SS1/2/EImax (i.e., lower RBC deformability) was observed in COVID-19 patients with abnormal RBC morphology compared to those without (2.33 ± 1.15 a.u. vs 2.05 ± 0.61 a.u., respectively; p < .05). However, no difference was found with the control group (2.20 ± 0.64 a.u.).

Hematocrit (A), blood viscosity (B), RBC deformability (C–F), RBC senescence markers (G–J), RBC aggregation (K–L), RBC aggregates strength (M) in patients with COVID-19 (COVID), and healthy individuals (Controls). Comparisons of RBC aggregation (N) and RBC aggregates strength (O) in patients with COVID-19 measured in autologous plasma or Dextran solution. Correlation between RBC aggregation and plasma fibrinogen concentration (P) and between CRP level and plasma fibrinogen concentration (Q) in COVID-19 patients. Effect of Ancrod molecule on RBC aggregation (R) and RBC aggregates strength (S) in blood samples from patients with COVID-19. Significant difference: *p < .05, **p < .01, ***p < .001

3.2 Increased fibrinogen causes RBC hyper-aggregation in COVID-19 patients

Both RBC aggregation (Figure 1L) and RBC aggregates strength (Figure 1M) were highly increased in patients with COVID-19 compared to controls. Figure 1K shows the presence of RBC rouleaux in the blood of one patient diluted with autologous plasma (hematocrit = 1%), which demonstrates that RBC hyper-aggregation is due to true RBC aggregation process and not RBC agglutination. Since RBC aggregation may be caused by plasma and cellular (i.e., RBC aggregability) factors, further experiments were performed on seven patients with COVID-19 to compare RBC aggregation in autologous plasma versus Dextran. The results show a decrease and a normalization of both RBC aggregation (Figure 1N) and RBC aggregates strength (Figure 1O) in Dextran compared to plasma. Fibrinogen is one of the key plasma factors involved in RBC aggregation process28, 29 and Figure 1P shows a positive correlation between RBC aggregation and plasma fibrinogen concentration in COVID-19 patients. Moreover, the use of Ancrod to deplete fibrinogen in the plasma of five patients with COVID-19 lead to a reduction of both RBC aggregation (Figure 1R) and RBC aggregates strength (Figure 1S). Finally, we observed a positive correlation between CRP and fibrinogen levels in COVID-19 patients (Figure 1Q).

3.3 RBC hyper-aggregation correlates with enhanced blood coagulation in COVID-19 patients

Rotational thromboelastometry analyses showed hypercoagulability in patients with COVID-19 with α angle, MCF, A5, A10, and A20 values being higher than the normal range values (Table 2). No correlation was observed between blood viscosity and FIBTEM parameters. In contrast, RBC aggregation was positively correlated with A5 (r = 0.45, p < .001), A10 (r = 0.47, p < .001), A20 (r = 0.49, p < .001), α (r = 0.19, p < .05), MCF (r = 0.56, p < .001), and negatively with CFT (r = −0.38, p < .001). Fibrinogen level was positively correlated with A5 (r = 0.47, p < .001), A10 (r = 0.43, p < .001), A20 (r = 0.47, p < .001), MCF (r = 0.34, p < .01), and negatively with CFT (r = −0.33, p < .01).

TABLE 2. Rotational thromboelastometry parameters (FIBTEM) in COVID-19 patients COVID-19 patients Normal rangeb CT (s) 67.1 ± 22.0 46–84 CFT (s) 100.5 ± 82.8 n.a. α (°) 78.1 ± 4.1a 30–70 MCF (mm) 33.8 ± 8.4a 9–25 A5 (mm) 28.7 ± 7.6a 6–16 A10 (mm) 31.3 ± 8.5a 6–16 A20 (mm) 33.0 ± 9.0a 7–18 ML (%) 2.2 ± 3.3 n.a. LI30 (%) 99.2 ± 4.0 94–100 LI60 (%) 98.0 ± 3.2 94–100 Abbreviations: A10, amplitude of clot firmness 10 min after CT; A20, amplitude of clot firmness 20 min after CT; A5, amplitude of clot firmness 5 min after CT; CFT, clot formation time; CT, coagulation time; LI30, lysis index 30 min after CT; LI60, lysis index 60 min after CT; MCF, maximum clot firmness; ML, maximum lysis; α, alpha angle. 3.4 RBC aggregation, blood viscosity, and blood coagulation are associated with some clinical markers of severity in COVID-19 patients

Patients receiving oxygen supplementation had greater RBC aggregation (Figure 2A), RBC aggregates strength (Figure 2B), and blood viscosity (Figure 2C) than those who did not need oxygen therapy. Fibrinogen concentration was greater in patients with oxygen than those without (5.7 ± 1.7 vs 4.4 ± 1.0 g/L, respectively, p < .05). In addition, compared to patients without oxygen supplementation, those who needed oxygen supplementation had higher A5 (21.8 ± 10.7 vs 29.2 ± 7.2, respectively, p < .05), A10 (22.6 ± 9.9 vs 31.9 ± 8.1 mm, respectively, p < .05), A20 (23.2 ± 8.8 vs 33.7 ± 8.6 mm, respectively, p < .01), and MCF (23.8 ± 9.4 vs 34.5 ± 7.9 mm, respectively, p < .01). While blood viscosity (Figure 2F) was not different between patients classified according to the extent of pulmonary lesions, RBC aggregation (Figure 2D), but not RBC aggregates strength (Figure 2E), was higher in all patients with pulmonary lesions compared to those without. Fibrinogen in patients without pulmonary lesion (4.4 ± 1.2 g/L) was lower than those with less than 25%, 25%–50%, 50%–75%, or greater than 75% pulmonary lesions (5.9 ± 1.1, 5.7 ± 1.6, 6.1 ± 1.8 and 6.7 ± 1.1 g/L, respectively; all p < .05). A5 (data not shown), A10 (Figure 2G), A20 (Figure 2H), and MCF (Figure 2I) were greater in patients with pulmonary lesions compared to those without. RBC aggregation also correlated with the length of hospitalization (r = 0.19, p < .05).

Associations between blood rheology or Rotational thromboelastometry (ROTEM) parameters, and markers of clinical severity. Comparisons of RBC aggregation, RBC aggregates strength, and blood viscosity at 45 s−1 between patients with COVID-19 with O2 supplementation and those without O2 supplementation (A–C). Comparisons of RBC aggregation (D), RBC aggregates strength (E), blood viscosity at 45 s−1 (F), and ROTEM parameters (G–I) between patients with COVID-19 and pulmonary lesions (the percentages of pulmonary lesions are indicated below the X-axis) and those with no pulmonary lesions. Significant difference: *p < .05, **p < .01, ***p < .001

4 DISCUSSION

The present work is the first to characterize the blood rheological profile in a large cohort of hospitalized COVID-19 patients. We demonstrated that: (1) blood viscosity and RBC aggregation are increased in patients compared to healthy individuals; (2) RBC hyper-aggregation is caused by increased fibrinogen level; (3) RBC hyper-aggregation correlates with blood hypercoagulability and is associated with various markers of clinical severity (oxygen supplementation, extent of pulmonary lesions, length of hospitalization).

Our results clearly showed higher blood viscosity in patients with COVID-19 compared to controls, despite lower hematocrit. Although hematocrit and hemoglobin remained in the normal range, the values were closed to the lower limit of this normal range. The negative correlation found between hematocrit and reticulocytes percent, and the slightly elevated reticulocytes percent in almost 25% of the patients with COVID-19 suggest mild enhanced erythropoiesis. Thomas et al.19 reported structural protein damages and membrane lipid remodeling in RBCs from COVID-19 patients, which would be in agreement with the marginally decreased RBC deformability found in a preliminary study performed on seven COVID-19 patients.8 Another study performed on 17 patients and using real-time deformability cytometry to estimate physical properties of RBCs concluded that RBC deformation and size are heterogeneous in COVID-19 patients,30 which would be in agreement with the presence of more than 1% of schizocytes observed in 10 of 14 patients with COVID-19 in a recent study.31 The present study is the largest one focusing on RBC physical properties in COVID-19 patients and neither isotonic shear stress gradient, nor osmolality gradient ektacytometry (i.e., the gold standard methods) show a difference in RBC deformability between patients and controls. Nevertheless, the tendency of patients with COVID-19 for having higher percentage of RBCs exposing PS suggests that RBCs could be slightly damaged, even if the mean deformability of the whole RBC suspension remains in the normal range of values. The analysis of RBC morphology also showed that abnormal shapes can reach more than 1% in 26% of patients with COVID-19 and these patients had slightly lower RBC deformability than those with no abnormal RBC morphology, although the mean RBC deformability remained in the range of the control values. Since PS is a signal for erythrophagocytosis in the spleen,32 a greater removal of damaged RBC could be suspected leading to slightly decreased hematocrit and enhanced erythropoiesis in the most anemic patients.

Blood viscosity depends on several parameters: hematocrit, plasma viscosity, RBC deformability, and aggregation.12 Since hematocrit was lower in patients with COVID-19 than in controls, it may be suspected that at least one of the other three blood rheological parameters would be highly increased to cause blood hyperviscosity. Plasma viscosity was not measured in this study because cone-plate viscosimetry is not the best method to assess it.11 However, the study of Maier et al.33 performed in 15 patients with COVID-19 supports that plasma viscosity is increased in this disease, probably as the consequence of the rise in plasma fibrinogen concentration,34, 35 as noted in our study. Although RBC deformability was not affected by COVID-19 in our study, we observed very high level of RBC aggregation with RBC aggregates being very sticky and difficult to dissociate. The levels of RBC aggregation and RBC aggregates strength found in COVID-19 patients were far greater than those reported in patients with sepsis (+1.2 and +2.6 fold for RBC aggregation and RBC aggregates strength, respectively,36) or in patients with sickle cell disease (+1.3 fold for both RBC aggregation and RBC aggregates strength37). RBC rouleaux may form in vascular areas where shear rate is low (<10 s−1) and usually dissociate when they flow in regions where shear rate is high (>100 s−1).34 However, our findings show that a shear rate greater than 500 s−1 is needed to disrupt the RBC aggregates in patients with COVID-19, which strongly suggests that RBC hyper-aggregation contributes to blood hyper-viscosity in this population. The increased fibrinogen level caused by inflammation38 is at the origin of this RBC hyper-aggregation in COVID-19 since replacing plasma by dextran or depleting plasma from fibrinogen both resulted in lowering RBC aggregation and RBC aggregates strength. Nevertheless, other factors could be involved since fibrinogen level is also enhanced in sepsis but RBC aggregation and RBC aggregates strength remain lower than those found in COVID-19 patients.

Several retrospective and observational studies showed an abnormal coagulation profile in patients with COVID-19 that would increase the risk for thromboembolic events.39, 40 A recent study reported positive correlations between mortality and markers of inflammation, coagulopathy, kidney and tissue damage, and hypoxia.41 Only 5% of the present cohort developed pulmonary or peripheral thrombosis, which may be explained by the fact that almost 75% of patients were treated with anticoagulants. However, despite anticoagulant therapy, rotational thromboelastometry measurements performed in FIBTEM mode support still hypercoagulability in COVID-19 patients with fibrinogen playing a major role, independently of the role of PLT. The correlation found between several FIBTEM parameters and fibrinogen confirm the key role of fibrinogen in the hypercoagulable state of COVID-19 patients.42 It is frequently assumed that RBCs play a passive and relatively unimportant role in thrombosis and hemostasis.43 However, it has become apparent that RBCs have a variety of important functions and have a substantial influence on blood clotting, hemostasis, and thrombosis that is clinically significant.43 Clot stability has been shown to be highly dependent on the presence of RBC and their rheological properties, which may impact on the efficacy of anticoagulants and thrombolysis therapeutics.44, 45 Rigidification of RBCs by chemical agents has been shown to interfere with blood clot retraction, with more RBCs being trapped in the fibrin network,46 a finding recently confirmed by another group.47, 48 Our results showed positive correlations between RBC aggregation and A5, A10, A20, and MCF, and a negative correlation with CFT. The shear rate applied to the blood suspension in the ROTEM® is 0.1 s−1,49 which corresponds to a very low shear rate where RBC aggregates can easily form during the measurements.11, 34 Indeed, one may suspect that RBC aggregation has participated in the formation of clot and to its firmness. In vivo, enhanced RBC aggregation increases blood viscosity and promotes flow stasis, increasing the risk of thrombosis.34, 50, 51

The present study also suggests a role of RBC aggregation in the modulation of the clinical severity of COVID-19. Patients who needed oxygen supplementation and with pulmonary lesions had higher RBC aggregation than those without. Since clot firmness was also increased in patients requiring oxygen supplementation and with pulmonary lesions compared to those without, one may suspect that RBC hyper-aggregation and hypercoagulability are linked phenomenon, with both ones depending on the concentration of fibrinogen. Increased RBC aggregation may affect blood circulation and tissue perfusion in several ways: (1) it may promote blood flow stasis and increases blood viscosity locally, (2) persistence of sticky RBC aggregates may affect blood flow into the microcirculation since RBCs are only able to pass through narrow capillaries as single cells rather than as aggregates, (3) in large vessels, aggregated RBCs concentrate along the flow axis (axial migration) and enhance platelet margination to the vascular wall, (4) increased RBC aggregation may favor plasma skimming at vascular bifurcation (Fahraeus effect), that may be accompanied by lower oxygen delivery to tissues and (5) increased RBC aggregation modulates vascular reactivity through its effects on nitric oxide production by endothelial cells.12,