This is a prospective observational cohort study conducted to assess patients who tested positive for COVID-19 through an RT-PCR test. Patient recruitment took place at the ICU facility within Kasr Alainy Cairo University Hospital - Internal Medicine Quarantine Hospital. Standard supportive therapy, including supplemental oxygen and symptomatic management, was administered as clinically indicated. Patients exhibiting moderate to severe hypoxia, defined as requiring a fraction of inspired oxygen (FiO₂) of ≥40%, were transferred to the intensive care unit (ICU) for escalated care, including invasive mechanical ventilation when deemed necessary. Within the ICU patient cohort, individuals were categorized into two groups based on their subsequent mortality outcomes: those who survived (ICU-S) and those who died within 20 days of sample collection (ICU-NS). Samples were collected during the period of October 2020 to December 2021 from ICU patients. Individuals with mild to moderate clinical symptoms of COVID-19 or with respiratory symptoms not related to SARS-CoV-2 (negative for SARS-CoV-2) were excluded from the study. All available samples and analyses were conducted by blinded operators and included in the final dataset.

A lung CT damage score for each patient was calculated according to a semi-quantitative CT severity scoring protocol proposed by Pan et al. (2020). This scoring system is based on the extent of parenchymal involvement in each of the five lobes (Elmokadem et al., 2022) as follows (0) no involvement; (1) <5% involvement; (2) 5%–25% involvement; (3) 26%–50% involvement; (4) 51%–75% involvement; or (5) >75% involvement. The resultant total CT score is the sum of the obtained lobar scores and ranges from 0 to 25.

A volume of 10 mL of fresh blood was collected from all subjects in Acid Citrate Dextrose (ACD) tubes (Greiner Bio-One GmbH, Kremsmünster, Austria). For flow cytometry measurements, 2 mL of collected citrated whole blood was incubated for 15 min with RBCs lysis buffer composed of NH₄Cl (ammonium chloride), NaHCO₃ (sodium bicarbonate), and EDTA (disodium) (Don-Doncow et al., 2019). The remaining 8 mL of blood was used for blood cell isolation as previously described (Abdel-Rahman et al., 2021). Briefly, whole blood was centrifuged at 300 × g for 15 min at 25°C with no breaks. (1) The upper clear layer (platelet-rich plasma) was transferred to a new tube and centrifuged to get platelets pellet and platelets poor plasma as described previously (Yasseen et al., 2022). (2) The lower layer was gently layered over equal volumes of lymphocyte separation medium (1.119 and 1.077). Tubes were then centrifuged at 500 × g for 35 min at 25°C with no brakes and lowered acceleration. After centrifugation, the second layer containing peripheral blood mononuclear cells (PBMCs) was collected, washed in PBS, and the pellet was suspended in 100 µL PBS. The layer comprising neutrophils was collected and washed in Hank’s Balanced Salt Solution (HBSS), then centrifuged again at 350 × g for 10 min at 25°C. The supernatant was discarded, and the pellet was suspended in RBCs lysis buffer and incubated at room temperature for 15 min. Following centrifugation for 10 min at 350 × g, the pellet was suspended in 100 µL PBS. Neutrophils and PBMCs counts were performed manually by a hemocytometer. Cellular viability was monitored for each sample using trypan blue, and only cells that exhibited more than 90% viability were used in the analysis.

The l-lactate level was assessed in plasma using colorimetric assays according to the manufacturer’s protocol with slight modifications (Lactate Assay Kit, Sigma-Aldrich, Missouri, United States, # MAK064-1KT). A volume of 25 μL of lactate standards or plasma samples was mixed with an equal volume of Master Reaction Mix (comprising 92% lactate assay buffer; 4% lactate probe, and 4% lactate enzyme mix). After a 30-minute incubation, absorbance was measured at 570 nm (A570), using the Cytation 5 Cell Imaging Reader (Agilent BioTek, CA, United States). Lactate levels present in the samples were estimated from the standard curve.

L-lactate levels in the blood were measured using blood lactate meter according to the manufacturer’s protocol (Lactate Pro 2, ARKRAY Inc., Kyoto, Japan).

Suspended lysed blood cells were centrifuged at 500 × g for 5 min and washed twice with phosphate-buffered saline (PBS). Cells were then suspended in PBS, and neutrophils were detected using a 13-color flow cytometer CytoFLEX system (Beckman Coulter Life Sciences CytoFLEX benchtop flow cytometer) as follows. Suspended cells were incubated in the dark at room temperature for 30 min with CD66b-APC-Alexa Fluor 750 (Beckman Coulter Life Sciences, B08756) to detect neutrophils. After the incubation period, cells were washed with PBS, suspended in 300 μL PBS, and examined by flow cytometry to gate the neutrophil-specific CD66b-APC-Alexa Fluor 750 positive population. A total of 20,000 events were acquired and then analyzed using CytExpert software.

Extracellular acidification rates (ECAR) in neutrophils, PBMCs, and platelets were assessed using XF analysis (XF96, Seahorse analyzer, Agilent) as described previously (Yasseen et al., 2022). Cells were plated on 96 well format XF plates at a density of 2 × 10⁵ neutrophils/well, 2.5 × 10⁵ PBMCs/well and 20 × 10⁶ Platelets/well in unbuffered DMEM deprived of glucose and pyruvate (DMEM; with 4 mML-glutamine, pH 7.4 at 37°C) for the glycolytic stress tests. Plates were centrifuged at 800 × g for 5 min to form a monolayer of cells in the wells. Plates were then incubated for 30–40 min at 37°C in a non-CO₂ incubator. A baseline measurement for ECAR was acquired for 30 min. ECAR values were examined following consecutive injections of glucose (5.5 mM), oligomycin (2.5 µM), and 2-deoxy-D-glucose (2-DG) (50 mM). Glycolysis and glycolytic capacity were calculated by subtracting the average rates after the addition of 2-DG from the average rates after the addition of glucose and ATP synthase inhibitor oligomycin. All measurements were normalized to the seeding density before being multiplied by the respective neutrophils count per µL of blood determined by flow cytometry.

Statistical analysis and data graphing were performed using OriginPro 2022b (OriginLab Corporation, Northampton, United States). Graphical representations of data utilized box-and-whisker plots showing mean ± SD along with the actual scatter of the data points combined with a distribution curve to indicate normality. Density Color Mapping, which assigns color based on the density of points in a two-dimensional scatter plot, was used to infer associative variance and covariance directionality between the two-plotted parameters. Exact p-values are given for each comparison on most graphs and in the text. Following tests for normality (Shapiro–Wilk test), variables that passed the normality test were analyzed using the ANOVA test followed by Tukey post hoc tests to compare the differences between the three groups, or the independent samples t-test, while non-normally distributed data were analyzed using the Mann-Whitney test to compare the differences between two groups. Categorical variables are reported as counts and percentages while continuous variables are expressed as mean ± SD or a median (range). Differences between percentages were assessed by Pearson’s χ² tests or Fisher exact tests when the number of observations per group was less than 5. The χ² tests provided results that tested the hypothesis that mortality and a given variable (e.g., sex or comorbidity) are independent. Linear regression analysis and Pearson correlation coefficients were obtained without asymptotic assumptions. When the p-value was less than the significant level of 0.05, significant evidence of an association between mortality and the variable was observed.

The demographic and clinical characteristics of the patients studied are given in Table 1. Participants were divided into two groups: ICU-survivors (ICU-S, n = 36) and ICU-non-survivors (ICU-NS, n = 66). The two groups did not exhibit statistically significant differences between most of the clinical or demographic characteristics except for the blood saturation level (p = 0.045). Additionally, the number of patients suffering from diabetes is significantly higher in the ICU-NS group (p = 0.008). Meanwhile, the number of patients treated with steroids, remdesivir, hydroxychloroquine, and/or carbapenem antibiotics did not differ significantly between the two groups. Table 2 shows the laboratory results of the participants. When comparing all parameters in the two groups, we observed changes following the same reported trends in our previous studies (Yasseen et al., 2022). Non-significant changes were reported in all the parameters except for a significant decrease in albumin level in the ICU-NS group (p = 0.03) and in platelets count (p = 0.017). The results also show a significant increase in the white blood cells count (p = 0.02), C-reactive protein (p = 0.008), and D-Dimer level (p = 0.008) in non-survivors.

A hallmark of COVID-19 infection is hypoxia in critical patients, and our cohort showed decreased partial pressure of oxygen in ICU-NS patients compared to ICU-S (Figure 1A, ICU-S: 72.9 ± 29.5; n = 23, ICU-NS: 58.4 ± 23.7; n = 58. ICU-S vs ICU-NS: p = 0.0235). In order to compare our measured parameters with those reported for control subjects, meta-analysis paper was used to extract relevant data (Forrer et al., 2023). When comparing PaO₂ = 91.5 ± 9.75 for controls with ICU-S and ICU-NS patients, the following statistics were computed: Control vs. ICU-S p < 0.0001; Control vs. ICU-NS p < 0.0001. Moreover, the PaO₂/FiO₂ ratio is frequently used to determine the severity of lung injury in mechanically ventilated patients. Here we show that the PaO₂/FiO₂ ratio is remarkably lower in the case of ICU-NS, indicating marked hypoxemia (Figure 1B, ICU-S: 318.1 ± 149.9; n = 14, ICU-NS: 115.1 ± 107.1; n = 25. ICU-S vs ICU-NS: p < 0.0001). Oxygen-hemoglobin dissociation curves (ODC) were constructed using ICU-survivors (ICU-S, n = 20) and ICU-non-survivors (ICU-NS, n = 52) COVID-19 patient data and compared to the standard human ODC at T = 37°C, pH = 7.4, created by the Severinghaus model (Figure 1C). First, the theoretical standard human blood ODC was estimated based on the original computations implemented by Severinghaus (1958). A left shift of the Hemoglobin oxygen dissociation curve of the ICU-S group is observed, while the Hemoglobin oxygen dissociation curve of the ICU-NS group is shifted towards higher oxygen pressure to reach the same degree of hemoglobin saturation. These results suggest an exacerbated respiratory dysfunction and relatively impaired gas exchange in ICU-NS COVID-19 patients. Next, we hypothesized that the observed shift in the ODC may result from lowered blood pH during the critical COVID-19 illness, particularly in non-survivors. Analysis of the chemistry laboratory results retrieved during hospitalization revealed that the average blood pH value was indeed lower in ICU-NS (pH = 7.37 ± 0.12, n = 57) relative to ICU-S group (7.43 ± 0.05, n = 23, p = 0.01) (Figure 1D). Since the ICU-NS group contained significantly more diabetic comorbodity, the control cohort data extracted yielded mean pH values of 7.41 ± 0.02 with statistical analyses of Control vs. ICU-S p < 0.0001; Control vs. ICU-NS p < 0.0001. Declines in blood pH can be a result of either respiratory or metabolic acidosis.

SARS-CoV-2 infection usually leads to complications such as pneumonia, and in severe cases, ARDS and sepsis, which are now known to leave lasting damage to the lungs including interstitial lung disease (Stewart et al., 2022). We analyzed the extent of lung damage in a subset of the studied cohort by analyzing the retrieved chest radiographs for 23 ICU-S and 45 ICU-NS as described in Methods; representative images are given in Figure 2A. ICU-NS group showed significantly greater lung damage with a mean score ±SD = 15.93 ± 5.87 vs 11.61 ± 4.68 for ICU-S group (p = 0.003), Figure 2B. We then proceeded to explore the prognostic power of the CT score as a mortality predictor. We started by generating a receiver-operating characteristic (ROC) curve for the prediction of mortality in terms of the CT score (Figure 2C). This analysis established a CT damage score cut-off value of 13.5 (AUC = 0.721, p = 0.003) as a stratifying parameter for mortality prediction in the subsequent Kaplan-Meier survival analysis (Figure 2D). CT lung damage score ≥13.5 was subsequently used as a predictor of mortality within 20 days of image acquisition (31 out of 38, 81.5%; Log Rank χ² = 7.26, p = 0.007). Respiratory acidosis results from hypoventilation and subsequent elevation of PCO₂. Laboratory results obtained for participants at the site of sample collections during hospitalization revealed non-significant differences in PCO₂ (Figure 2E) (ICU-S, n = 23; ICU-NS, n = 57), or in [HCO₃ ⁻] levels (Figure 2F) (ICU-S, n = 23; ICU-NS, n = 58) between ICU-survivors and ICU-non-survivors. This confirms that the decline in pH is not due to respiratory acidosis or metabolic acidosis from enhanced loss of HCO₃ ⁻. To answer the remaining arm of metabolic acidosis, we investigated lactate levels in the plasma and blood of ICU patients. Plasma levels of lactate, determined colorimetrically, were significantly higher in ICU-NS patients than in controls (Control: 5.62 ± 1.5; n = 14, ICU-S: 6.34 ± 2.32; n = 13, ICU-NS: 7.47 ± 2.2; n = 21. ICU-S vs Control: p = 0.64; ICU-NS vs Control: p = 0.036; ICU-S vs ICU-NS: p = 0.27), indicating a marked lactic acidosis in the non-survivors group (Figure 2G). To further confirm these findings, we also measured blood lactate levels in stored blood samples from a subset of controls, ICU-S and ICU-NS subjects using a blood lactate meter (Control: 3.55 ± 1.4; n = 11, ICU-S: 5.51 ± 2.01; n = 15, ICU-NS: 7.37 ± 3.19; n = 15. ICU-S vs Control: p = 0.11; ICU-NS vs Control: p = 7.7 × 10⁻⁴; ICU-S vs ICU-NS: p = 0.1). Our results showed that blood lactate levels were significantly elevated in the ICU-NS group when compared with controls (Figure 2H). Moreover, we reanalyzed our data reporting lactate levels in both blood and plasma while taking into account the discrepancy between the number of diabetic patients in each group. When we compared their levels of lactate in either plasma or whole blood, we found no statistical difference between those groups. Specifically, for plasma lactate levels: ICU-S = 6.2 ± 2.5 (n = 11); ICU-S DM = 7.3 ± 0.79 (n = 2); p-value = 0.88; ICU-NS = 8.4 ± 1.94 (n = 13); ICU-NS DM = 5.9 ± 1.75 (n = 8); p-value = 0.057. As for blood lactate levels: ICU-S = 5.49 ± 2.17 (n = 13); ICU-S DM = 5.65 ± 0.07 (n = 2); p-value = 0.99; ICU-NS = 6.75 ± 2.78 (n = 8); ICU-NS DM = 8.07 ± 3.70 (n = 7); p-value = 0.78. In conclusion, diabetes mellitus was not a significant confounding factor regarding blood and plasma lactate content.

We asked if and how leukocytes contribute to the observed increase in blood/plasma lactate levels and lowered pH, especially in non-survivors (Figure 3). To identify potential contributors to lactic acidosis in circulating blood, we assessed and compared glycolytic activities in isolated neutrophils, platelets, and peripheral blood mononuclear cells (PBMCs) from the current cohort of critically ill patients relative to the control group. First, group-dependent changes in cell numbers in these three cell types were assessed using flow cytometry to show that only neutrophil counts exhibit a remarkable increase with severe COVID-19 infection (Figure 3A, Control: 499.32 ± 338.9; n = 9, ICU-S: 1,156.42 ± 938.9; n = 18, ICU-NS: 1,507.42 ± 1,191.19; n = 24. ICU-S vs Control: p = 0.26; ICU-NS vs Control: p = 0.04; ICU-S vs ICU-NS: p = 0.50). In fact, PBMCs showed a significant decrease in non-survivors, Figure 3A (Control: 300.13 ± 122.46; n = 8, ICU-S: 195.71 ± 145.47; n = 7, ICU-NS: 117.34 ± 122.75; n = 10. ICU-S vs Control: p = 0.28 ICU-NS vs Control: p = 0.02; ICU-S vs ICU-NS: p = 0.45).

Traces of Seahorse metabolic analysis of freshly isolated neutrophils from subsets representing all groups demonstrate that the extracellular acidification rate by individual neutrophils is independent of mortality outcome (Figure 3B). Generally, for both neutrophils and PBMCs, basal ECAR, glycolytic metabolism, and glycolytic capacity normalized to seeded cell counts did not show any significant differences among all groups (Figure 3, panels C, E, G). Only platelets exhibited significantly reduced glycolytic flux in critically ill COVID-19 patients, which confirms our recently published data (Yasseen et al., 2022). However, a significant increase in population-specific basal ECAR per µL lysed blood (Figure 3D, Control: 1.44 × 10⁵ ± 0.54 × 10⁵; n = 8, ICU-S: 3.26 × 10⁵ ± 2.78 × 10⁵; n = 18, ICU-NS: 6.1 × 10⁵ ± 6.2 × 10⁵; n = 24. ICU-S vs Control: p = 0.63; ICU-NS vs Control: p = 0.047; ICU-S vs ICU-NS: p = 0.13), glucose-supplemented glycolytic metabolism (Figure 3F, Control: 2.76 × 10⁵ ± 2.57 × 10⁵; n = 8, ICU-S: 6.62 × 10⁵ ± 4.62 × 10⁵; n = 18, ICU-NS: 10.94 × 10⁵ ± 10.44 × 10⁵; n = 24. ICU-S vs Control: p = 0.49; ICU-NS vs Control: p = 0.04; ICU-S vs ICU-NS: p = 0.19), and a trend in glycolytic capacity when mitochondria are inhibited with oligomycin (Figure 3H, Control: 3.53 × 10⁵ ± 3.86 × 10⁵; n = 8, ICU-S: 6.79 × 10⁵ ± 4.65 × 10⁵; n = 18, ICU-NS: 11.43 × 10⁵ ± 11.3 × 10⁵; n = 24. ICU-S vs Control: p = 0.64; ICU-NS vs Control: p = 0.07; ICU-S vs ICU-NS: p = 0.2) have been detected in ICU-NS neutrophils relative to ICU-S and control groups. These results suggest that the observed increase in glycolysis in ICU-NS neutrophils is driven by a rise in neutrophil counts rather than a metabolic shift in individual neutrophils.

We then investigated the relationship between neutrophil counts and lactic acidosis in COVID-19 ICU-patients. A rise in lactate levels that coincides with increased neutrophil counts in the circulation and inflamed tissues has been demonstrated in multiple pathological conditions including shock, sepsis, and ischemia (Andersen et al., 2013) but not in COVID-19. Here, we compared neutrophil counts with blood pH in critically ill patients as shown in the scatter and distribution plots (Figure 4A for ICU-S; Figure 4B for ICU-NS), which shows an increased tendency toward lower pH in patients exhibiting greater neutrophil counts, especially in ICU-NS group. In Figure 4C we show that blood lactate levels positively correlate with neutrophil counts (Pearson’s r = 0.43, p = 0.009), a trend that was also conserved for plasma lactate levels albeit with weaker but significant correlation (Figure 4D, Pearson’s r = 0.36, p = 0.017). These results further suggest that neutrophilia are significantly contributing to lactic acidosis in critically ill patients. These results recalled an important question: Is there a link between neutrophilia and lung damage? To answer this question, we plotted the CT damage score as a function of neutrophil counts (Figure 4E for ICU-S; Figure 4F for ICU-NS). Although no statistically significant correlation has been observed to linearly connect these parameters, a qualitative association can be seen that indicates distribution at higher values of both neutrophil counts and CT damage scores in the ICU-NS group. This suggests a subtle association between neutrophilia and COVID-19-caused lung damage.