Evidence showing lipotoxicity worsens outcomes in covid-19 patients and insights about the underlying mechanisms

Summary We compared three hospitalized patient cohorts and conducted mechanistic studies to determine if lipotoxicity worsens COVID-19. Cohort-1 (n = 30) compared COVID-19 patients dismissed home to those requiring intensive-care unit (ICU) transfer. Cohort-2 (n = 116) compared critically ill ICU patients with and without COVID-19. Cohort-3 (n = 3969) studied hypoalbuminemia and hypocalcemia’s impact on COVID-19 mortality. Patients requiring ICU transfer had higher serum albumin unbound linoleic acid (LA). Unbound fatty acids and LA were elevated in ICU transfers, COVID-19 ICU patients and ICU non-survivors. COVID-19 ICU patients (cohort-2) had greater serum lipase, damage-associated molecular patterns (DAMPs), cytokines, hypocalcemia, hypoalbuminemia, organ failure and thrombotic events. Hypocalcemia and hypoalbuminemia independently associated with COVID-19 mortality in cohort-3. Experimentally, LA reacted with albumin, calcium and induced hypocalcemia, hypoalbuminemia in mice. Endothelial cells took up unbound LA, which depolarized their mitochondria. In mice, unbound LA increased DAMPs, cytokines, causing endothelial injury, organ failure and thrombosis. Therefore, excessive unbound LA in the circulation may worsen COVID-19 outcomes.


INTRODUCTION
The coronavirus disease  pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) represents the greatest global public health crisis of our time. COVID-19 outcomes range from asymptomatic disease to death, with mortality being almost 50% for those requiring invasive mechanical ventilation (Domecq et al., 2021). COVID-19 outcomes therefore may depend on disease modifiers and not just infection.
Here, after initially noting UFA, total and unbound LA elevation in hospitalized patients progressing to severe COVID-19, we prospectively compared NEFAs in an ICU cohort of severe COVID-19 patients to those without COVID-19. Based on these, and the mechanistic link of UFAs triggering synchronous hypocalcemia and hypoalbuminemia Khatua et al., 2020), we did a multivariate analysis to study the impact of calcium and albumin levels on mortality in a large retrospective hospitalized COVID-19 cohort. Finally, in animal experiments (mice), we administered an established LA dose Khatua et al., 2021) lower than palmitic acid (PA; the most abundant saturated FA) and compared the resulting phenotype in mice to that of severe COVID-19 patients while avoiding the confounding effects of coexisting diseases (e.g., pneumonia) or therapies like fluid resuscitation. Interestingly, pre-binding LA to albumin normalized the unbound FA elevation and reversed the severe COVID-19 like phenotype induced by LA.

RESULTS
Total and unbound serum oleic (OA) and linoleic acid (LA) increase in severe COVID-19 patients requiring ICU admission (cohort 1) The COVID-19 patients dismissed home (n = 22) or transferred to the ICU (n = 8) had similar demographics (Table 1), and the interval between the first (admission) and second (dismissal or ICU transfer) blood sample (6.3 G 2.8 versus 7.4 G 5.9 days respectively, p = 0.51). Although NEFAs were similar at admission (Adm., Figure 1), severe COVID-19 patients (red boxes Figure 1) had higher oleic acid (C18:1) and LA (C18:2) (Figures 1E-1F) at the time of ICU transfer. Saturated fatty acids like myristic acid (C14:0), PA (C16:0), stearic acid (C18:0), and unsaturated ones including palmitoleic (C16:1), linolenic (C18:3) and arachidonic acid (C20:4) were not significantly different between the two groups at either timepoint 1G,and 1H). The other C:20 or C:22 fatty acids were also similar between the groups (data not shown). Overall, the proportion of UFAs (i.e., % UFA) at the time of ICU transfer ( Figure 1I) increased in comparison to the proportion in those dismissed home. This change (D in Figure 1I) was measured by subtracting the % serum UFAs (of total NEFA) at the time of dismissal or ICU transfer from the respective patient's %UFA at the time of admission. ICU transfers also had significantly higher mortality (4/8 versus 0/22, p = 0.003; bottom row of Table 1).
Because unbound FAs can worsen acute disease outcomes , we compared their levels in the sera of patients at the time of dismissal to home versus ICU transfer ( Figure 1J-1L) using two different methods. The fluorometric ADIFAB method, showed total unbound FAs to be statistically higher in the sera of patients transferred to the ICU ( Figure 1J.). On measuring individual FAs in the dealbuminated sera by gas-chromatography and mass-spectrometry (GC-MS), we only noted serum oleic acid (OA; C18:1) and LA (C18:2) to be significantly higher in the sera of severe COVID-19 patients needing ICU care ( Figure 1K and 1L). Other unbound serum saturated fatty acids (e.g., myristic acid) and unsaturated ones (e.g., palmitoleic acid) were similar between the two groups of patients (results not shown). Totals of unbound FAs measured by both methods correlated well (r = 0.73, p < 0.001), which further improved to 0.83 (CI-0.66-0.92, p < 0.001) after excluding the one definite GC-MS outlier 15  LA and unbound FAs are elevated in ICU non-survivors, and a prospective ICU cohort of severe COVID-19 patients with hypocalcemia and hypoalbuminemia (cohort 2) Among ICU patients, 39 had severe COVID-19 pneumonia. The non-COVID ICU patients (n = 77) were postcardiac surgery (34, 44%), had septic shock (28, 36%) or stroke (7, 9%). Patients with COVID-19 were younger, had higher BMI, and had lesser coronary artery disease. Additional characteristics are in Table 2. COVID-19 patients received less intravenous fluids and had a more even fluid balance in the first 24 h (Table 2). Total calcium, ionized calcium, and albumin levels were significantly lower and lipase levels were significantly higher in the COVID group irrespective of septic shock in the non-COVID group (Last 4 rows Table 2, S1). Although our COVID-19 group had a higher proportion of non-Caucasian and Native-American patients, this did not affect biometrics, co-morbidities, outcomes, or measured parameters (Tables S2 and S3).  Table S4). Serum LA, UFA, and unbound FAs were also higher in mice given LA (right side Table 3). In contrast, mice given PA (the most abundant saturated fatty NEFA) even at a higher dose than LA did not have elevated PA or unbound fatty acid levels. As shown later (Figure 7), the increase in serum LA has little significance when bound to albumin. Septic-shock patients also had elevated palmitoleic acid (21.6 mM IQR 12.2-40.4mM, p = 0.001) versus non-COVID controls (15.4 mM IQR 9.0-16.8mM). The shorter chain of palmitoleic acid with 16 carbons and 1 double bond could contribute to elevated unbound FAs . Thus, septic shock, which is associated with hypocalcemia and hypoalbuminemia (Cumming, 1994), and COVID-19 both result in elevated unbound FA, albeit because of different NEFA. ICU non-survivors (n = 13) had similar demographics as survivors (Table S5). However non-survivors had higher serum LA and unbound FAs [6.2 mM IQR 4.5-11.8 mM] versus survivors [3.8 mM IQR 2.8-5.9 mM, p = 0.02 (Table S6), with a higher prevalence of MOF, ECMO requirements (Table S7) and IL-1Ra, IL-6 elevation (Table S8). Thus, elevated LA and unbound fatty acid lipotoxicity seemed to associate with worse inflammation, organ failure, and reduced survival. A total of 3969 patients were included in the retrospective multicenter study (3480 from Beaumont Health and 489 from University of Texas Health at San Antonio). Patient characteristics between hospitalized patients that survived (n = 3398) versus non-survivors (n = 571) are described in Table 4. Non-survivors presented with more comorbidities (Table 4). Calcium and albumin levels were significantly higher in survivors (as compared to nonsurvivors (Figures 2A and 2B) at admission and daily for the first 4 days of hospitalization (Table S9). This trend was paralleled in patients requiring mechanical ventilation (n = 781, Figures 2C and 2D). Univariate analysis showed that patients' age, male gender, HTN, DM, CAD, CHF, CKD, history of malignancy, creatinine levels, and BUN levels had higher mortality (Table S10). On multivariate analysis, increasing patient age, BMI and male gender, unlike race or comorbidities were the only consistent variables associated with increased mortality whereas decreasing levels of calcium and albumin showed a consistent independent association with increased mortality throughout the first 4 days of hospitalization (Table 5). Individual multivariate analyses models are described in the supplemental material (Tables S11-S20).
Animal and in vitro studies: LA reacts favorably with calcium and albumin, inducing hypocalcemia, hypoalbuminemia To understand hypocalcemia and hypoalbuminemia from LA elevation during COVID-19 infection, we studied their interactions using isothermal titration calorimetry. LA interacted with a favorable enthalpy (DH) with both albumin (DH = À154 G 54 kJ/mol) and calcium (DH = À17.1 G 1.5 kJ/mol) (Figures 3A  3D) over a few days as in patients with COVID-19 associated mortality ( Figure 2). We thus went on to study if the increased unbound FAs noted in COVID-19 patients and mice with elevated LA cause MOF, thrombosis, and cytokine elevation noted in severe COVID-19.
Severe COVID-19 and unbound LA cause endothelial damage We first noted that unbound LA at clinically relevant concentration range of 2.5-10 mM dose dependently depolarized mitochondria (510/590 Emission ratio in Figure 4A) in JC-1 loaded endothelial cells (HUV-EC).
The phenomenon seemed to be triggered by unbound LA since LA (30 mM) and its fluorescent tracer LAcoumarin were reversibly taken into HUV-ECs (middle row 60-150S images, Figure 4B). Although LA uptake depolarized HUV-EC mitochondria as evidenced by reduced fluorescence of the mitochondrial membrane potential sensitive dye MitoTracker Red CMXRos (MT-Red, top row), the addition of albumin and a iScience Article fluorophore (Albumin-647, at 180s, bottom row) reversed the uptake and depolarization. Thus, unbound LA may depolarize endothelial mitochondria and injure them. This hypothesis is supported by the higher endothelial injury marker; soluble E-selectin in ICU COVID-19 patients (versus non-septic COVID-19 controls), and LA treated mice ( Figure 4C and 4D). In later sections ( Figure 5J, black arrows) we also noted morphological evidence of LA induced endothelial injury in vivo. Similar to COVID-19 patients, LA treated mice had higher soluble ICAM-1, and DAMPs (i.e., ds-DNA, HMGB-1, and histone-DNA complex levels) ( Figures 4E-4J) which as we shall discuss later can be pro-thrombotic and pro-inflammatory.
Severe COVID-19 induced cytokine elevation, organ failure and thrombosis in ICU patients are replicated by unbound LA in vivo COVID-19 ICU patients had significantly higher levels of the proinflammatory cytokines CXCL1, IL-1b, IL-6, MCP-1, and TNF-a compared to the non-COVID group (Table 6). LA treated mice also had a similar pattern  Tables 6, S21). However, in PA administered mice, although IL-6 elevation was noted, it was lower than in mice given LA (right side Table 6).
COVID-19 patients were in a prothrombotic state evidenced by a higher rate of deep venous thrombosis (DVT) and pulmonary embolism (PE), irrespective of septic shock (Bottom rows of Tables 7, S22), whereas MOF was higher after excluding septic shock (Top rows of Tables 7, S22). ECMO utilization was also seen more frequently in the COVID-19 group (Tables 7, S22). Similar to COVID-19 patients, LA induced MOF in mice ( Figure 5) necessitating euthanasia by 72 h. Shock was evidenced by a drop in carotid artery pulse distention (Pulse dist., Figure 5A). LA also caused hypothermia ( Figure 5B) suggesting a severe systemic inflammatory response. Apart from shock, LA also induced other parts of MOF. LA induced renal failure was noted as a large increase in serum BUN and creatinine ( Figures 5C and 5D). LA also increased TUNEL (Terminal deoxynucleotidyl transferase dUTP nick end labeling) positivity in the renal tubules ( Figures 5E and 5H yellow arrows). LA induced lung injury was noted as TUNEL positivity in the lung alveoli (Figures 5F and 5J red arrows) and vessels (J, black arrows). Interestingly, this was associated with pulmonary thrombi (Figures 5J and 6J). LA induced lung injury was also noted as increased lactate dehydrogenase (LDH) and higher annexin V positive cells in the bronchoalveolar lavage (BAL; Figures 5K-5L). Specifically, type-II pneumocyte injury was noted as increased dual Thyroid transcription factor-1 (TTF-1), Annexin V positive cells, or CD208 + Annexin V+ cells (Figures 5MÀ5O) in the BAL of LA treated mice. Therefore, LA induced shock, and renal and lung injury in mice.
On focusing on prothrombotic mechanisms, we noted COVID-19 patients in the ICU had higher coagulation factor III (CFIII; Figure 6A) and tissue factor pathway inhibitor (TFPI; Figure 6B) supporting a prothrombotic state compared to controls. Similarly, 8 of 12 mice given LA had portal venous thrombi ( Figures 6C and 6D) unlike control mice. A prothrombotic state induced by LA was noted as increased fibrinogen, plasminogen activation inhibitor-1 (PAI-1), and TFPI levels. Elevated Soluble Platelet Endothelial Cell Adhesion Molecule 1 (PECAM-1) or CD31 ( Figures 6E-H) levels suggested LA induced endothelial injury in vivo. On immunohistochemistry of control mouse lungs, PECAM-1/CD31 normally localized to pulmonary vessel endothelium (yellow Figure 6I), and immune cells as described previously (Lertkiatmongkol et al., 2016). LA, however, dramatically increased PECAM-1/CD31 expression in the alveolar capillaries and pulmonary vascular thrombi (red squares, oval, Figure 6J), with the mice  iScience Article developing thrombocytopenia and an elevated activated clotting time ( Figures 6K-6L). Electron microscopy of pulmonary vessels showed LA causes loss of or lifting of endothelial cells (* in Figure 6O) from the basement membrane, wherein fibrin strands attached (green arrows Figure 6O) and extended inwards towards large platelet aggregates ( Figure 6P).
Administration of LA with calcium and albumin (fatty acid free) inhibited the increase in unbound FA without affecting the increase in serum LA or total NEFA ( Figures 7A-7C). This reduction in unbound FA normalized serum calcium, albumin, and ionized calcium ( Figures 7D-7F Finally, it has been reported that patients with diabetes mellitus could be linked to underlying elevated FAs as compared to non-diabetics (Bergman and Ader, 2000). Furthermore, LA plasma levels have been reported to be significantly higher in women (Lohner et al., 2013). In order to evaluate if the presence of diabetes or the female gender could confound the results, we analyzed the ICU cohort (Cohort 2) firstly by removing all diabetic patients (Table S23) and secondly by removing women from the cohort (Table S24). No major differences were noted as compared to the initial results. Therefore, the prothrombotic state, MOF and reduced survival noted in severe COVID-19 patients is likely because of elevated unbound UFAs like LA. Here we present evidence in humans with mechanisms in animals and in vitro studies explaining how unbound LA may worsen COVID-19 outcomes. Prospectively, in hospitalized patients we note that in comparison to mild COVID-19 patients, those progressing to severe COVID-19 requiring ICU admission have higher UFAs, including LA and OA, which are also increased in the unbound form. Severe COVID-19 ICU patients had higher lipase, LA levels, UFAs, inflammatory cytokines, thrombotic events, hypoalbuminemia, and hypocalcemia compared with non-COVID patients. Unbound FAs were elevated in both COVID-19 and septic shock patients who also had a similar rate of MOF. Cumulatively, COVID-19 patients required more organ support therapies including ECMO and mechanical ventilation. The retrospective hospitalized COVID-19 cohort had hypocalcemia and hypoalbuminemia independently associated with hospital mortality and ventilator requirements after adjusting for age, gender, BMI, race, and medical comorbidities.
Experimentally in mice, unbound LA induced the widely reported hypoalbuminemia, hypocalcemia, DAMP release, cytokine storm, thrombosis, and MOF phenotype, which we note and others have reported in severe COVID-19. These are consistent with well-known models of acute lung injury induced by intravenous UFAs (Kamuf et al., 2018;Moriuchi et al., 1998). PA, the most abundant saturated NEFA, despite being administered at higher doses than LA, did not enter the circulation, perhaps because of its extreme hydrophobicity . Administering intraperitoneal PA with a solvent (dimethyl sulfoxide) or directly through an incision were equally harmless (data not shown). iScience Article Although albumin binds LA more strongly than calcium (DH = À154 G 54 kJ/mol versus À17.1 G 1.5 kJ/mol; Figures 3A and 3B), albumin's molar amounts in normal serum (600-800 mM) are lower than calcium (2-2.5mM). However, albumin's stronger binding to LA put it upstream of calcium in preventing lipotoxicity ( Figure 8). Therefore, despite the increase in total LA concentrations induced by giving prebound LA ( Figure 7B); the prebinding kept unbound FAs low at control mouse levels ( Figure 7C). This is mechanistically consistent with the energetically favorable pre-binding of LA by albumin preventing the increase in unbound FAs despite an increase in total LA. Although we cannot comment on the exact proportion of unbound LA neutralized by calcium, the therapeutic role of calcium is supported by it preventing LA induced hypocalcemia (Figures 7D and 7F) and its energetically favorable binding to LA ( Figure 3B). This is clinically relevant because we note hypocalcemia with severe COVID-19 (Figure 2), and calcific fat necrosis is noted in autopsies of COVID-19 patients (Lax et al., 2020). The protective role of calcium against lipotoxicity is also supported by previous studies showing extracellular calcium deposits in fat necrosis and extracellular calcium supplementation to reduce lipotoxic cell injury and delay organ failure . Mechanistically, unbound LA relevant to COVID-19 concentrations was reversibly taken into cells and depolarized mitochondria ( Figures 4A and 4B). LA which reduces transendothelial resistance, causing macromolecular capillary leakage Khatua et al., 2021), reacted with albumin and calcium, explaining the rapid hypoalbuminemia and hypocalcemia we note in humans and mice (Figures 2 and 3). Such hypoalbuminemia cannot be explained by reduced albumin synthesis, because albumin has a 25-day half-life (Levitt and Levitt, 2016). These findings and concepts are summarized in Figure 8.   (Domecq et al., 2021;Richardson et al., 2020) developed respiratory failure (Domecq et al., 2021;Li et al., 2021), MOF, and VTE events. Similarly, endothelial injury (Ackermann et al., 2020), lung injury, and vascular occlusions were also induced by unbound LA in mice (Figures 5 and 6). LA induced injury in the BAL without coexisting pneumonia or pancreatitis suggest unbound LA may worsen COVID-19. Interestingly, the elevation of unbound FAs, UFAs, and LA in ICU non-survivors (Table S6) irrespective of etiology suggests a broader relevance of such elevations. Please note that because our COVID-19 patients had an even fluid balance, we gave our mice water ad libitum with subcutaneous saline supplementation at 10% bodyweight/ day.
The proportion of LA in human adipose increased from 5 to 10% in the 1950s to R20% by 2000 (Guyenet and Carlson, 2015), closely following the pattern of increased dietary intake of LA (Blasbalg et al., 2011). Although LA (50 mM) was proposed to synergize with the anti-viral drug Remdesivir and reduce SAR-S-CoV-2 replication in human cells (Toelzer et al., 2020); the reduced replication can be explained by LA's toxicity on the cells in which the virus was cultured. This diet related increase in visceral LA and association of COVID-19 mortality with UFA intake noted on multivariate analysis  along with the inability of remdesivir to reduce COVID-19 severity (Al-Abdouh et al., 2021;Okoli et al., 2021;Piscoya et al., 2020) suggest a deleterious role of excess LA in COVID-19 infection.
Double bonds in a FA, like LA, increase its lipolytic generation, and the aqueous stability of its monomers even without a carrier like albumin . Saturation in contrast makes long chain fatty acids too iScience Article hydrophobic to exist as unbound monomers. This likely explains the lack of PA elevation, cytokine response in PA administered mice (Tables 3 and 6) and the lack of MOF as previously reported . Double bonds explain the elevated UFAs and unbound FA levels in COVID-19 non-survivors (Table S6), those with septic-shock [who had increased palmitoleic acid (C16:1), the shorter chain of which increases aqueous stability] and mice given LA alone (Table 3). Previously, in the presence of albumin 300-600 mM LA concentrations were shown to depolarize mitochondria Patel et al., 2016). In contrast, in the absence of albumin we note 2.5-30 mM unbound LA is sufficient to depolarize endothelial (HUV-EC) cell mitochondria ( Figures 4A and 4B), which is reversed by albumin. This supports the deleterious role of unbound LA in causing endothelial injury (i.e., elevated E-selectin, ICAM-1; Figures 4C and 4E) in severe COVID-19 patients and in our mice. The endothelial injury was corroborated by electron microscopy ( Figure 6O), elevated circulating PECAM-1 levels and higher PECAM-1 expression on IHC ( Figures 6H and 6J). Such cell injury, along with previous studies showing LA to cause cytochrome c leakage, reduce ATP levels, inhibit mitochondrial complexes I and V (Khatua et al., 2019;Navina et al., 2011;Patel et al., 2016) may also explain the pulmonary and renal TUNEL positivity in LA treated mice Figure 8. Schematic describing the pathophysiology observed in severe COVID-19, wherein lipolytically generated LA results in the organ failure, thrombosis and death preceded by hypocalcemia and hypoalbuminemia The red L shaped structures denote unsaturated fatty acids like LA, which are present in excess although the blue straight lines denote saturated fatty acids. When the amount of unsaturated fatty acids like LA exceeds the ability of albumin to bind them, the unbound LA reacts with calcium causing hypocalcemia. Excess unbound LA is taken up by cells, causing mitochondrial depolarization, and consequent epithelial and endothelial injury with DAMP release, the latter among which causes the cytokine storm. The endothelial injury causes vascular leak, and hypoalbuminemia in addition to the release of procoagulant coagulation factor III (CFIII) from basement membranes, which with DAMPs promote thrombosis. These plus plasminogen activation inhibitor-1 (PAI-1) result in a prothrombotic state and worsening organ failure which can result in death.
LA itself is a precursor of arachidonic acid (Gao et al., 2010), which is elevated in our COVID-19 ICU patients (Table 3), those with elevated IL-6 ( Thomas et al., 2020), and mice given LA. Arachidonic acid can induce platelet aggregation, and increase the formation of prothrombotic thromboxane A2 (Lagarde et al., 2010), resulting in increased thrombosis.
There is a potential role for early albumin and calcium supplementation to prevent lipotoxicity in COVID-19. We have described experimentally what occurred when albumin was administered with LA and calcium: it inhibited the increase in unbound FA and prevented DAMP increase, coagulation abnormalities, and MOF development; resulting in improved survival in mice. The potential utility of albumin therapy in COVID-19 patients was recently described by an Italian group (Violi et al., 2021). In an observational prospective study performed in 29 SARS-CoV-2 patients treated with anticoagulant alone or anticoagulant plus albumin supplementation for 7 days, the investigators demonstrated a significant decrease of D-dimer only in the albumin-treated patients, who also had significantly reduced mortality (0/10 versus 8/19 without albumin supplementation; p = 0.02).
In summary, we note that during severe COVID-19 infection, the lipolytic release of UFAs like LA, perpetuated by the hypoalbuminemia and hypocalcemia induced by LA may result in cellular uptake of the unbound FAs like LA, resulting in mitochondrial injury. This injury to endothelial and other cells in vivo may result in shock, renal failure, DAMP release, and the consequent cytokine storm, MOF, and thrombosis that we note during severe COVID-19 infection.

Limitations of the study
We acknowledge several limitations. First, our prospective cohorts are small, and the cohort-2 non-COVID-19 ICU patients are not an ideal control group because these included septic patients with whom they could share some pathophysiologic mechanisms. However, on excluding septic patients, the most important lipotoxic, inflammatory, and thrombotic abnormalities persisted. In addition, although organ failure was increased in the COVID-19 ICU group, mortality did not achieve statistical significance compared to the non-COVID group. To address this, we compare the NEFA profile between ICU survivors and non-survivors and note that NEFA, UFAs and specifically LA and unbound fatty acids were higher in non-survivors (Table S6). We also do not have an obese animal model of COVID-19 to test the hypothesis that unbound fatty acids worsen COVID-19 in these models, nor do we provide proof of albumin having a therapeutic role in these. iScience Article improved survival than those who received anticoagulation only (Violi et al., 2021). Another limitation included the lack of lipase levels in the large retrospective cohort; however, large meta-analyses have shown lipase elevation to be associated with worse COVID-19 outcomes in the absence of clinical pancreatitis (Yang et al., 2022). Therefore, although this study provides preliminary evidence of and mechanisms supporting lipotoxic exacerbation of COVID-19, its findings need to be confirmed in larger studies.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests.

Human studies
The age, sex of the subjects in the various cohorts in the study are mentioned in the respective Tables. The cohorts are detailed below.
Hospitalized pre-ICU COVID-19 cohort (cohort 1) BAL was performed via tracheal incision at the time of euthanasia. To collect mice lungs BAL fluid, an incision was made on the disinfected neck skin. The trachea was exposed by blunt dissection. After tracheal incision with a scalpel, a catheter was placed in the exposed trachea. which was connected to a syringe filled with 1 mL of sterile saline solution. A cotton ligature was placed around the trachea and catheter to avoid flowing back of the fluid to the upper airways. The lungs were flushed with saline gently, preventing the collapse of the lung airway. The aspirated fluid was collected and centrifuged (400g, 10 min, 4 C) to pellet the cells, which were immediately as below. The supernatant was analyzed for lactate dehydrogenase (LDH) leakage and total proteins. A colorimetric LDH leakage assay was performed with Roche Cytotoxicity Detection kit (04744942001) following manufacturer's protocol. The data obtained was used to calculate LDH activity (U/L). Total protein was estimated by using Pierce BCA Protein Estimation kit (23225) following manufacturer's protocol and reported as mg/mL. For flow cytometry the pelleted cells were fixed (20 min, RT) in BD FACS lysing solution (349202), followed by staining with antibodies Thyroid transcription factor 1 (AF594; SantaCruz 8G7G3/1), CD208 (CF647; Biorbyt orb665591), and Annexin V apoptosis marker (AF488; Invitrogen 13201) in FACS staining buffer (Invitrogenä 00422226) for 40 min in dark on ice. Protocols followed were as recommended by the respective manufacturers. The stained cells were washed (400g, 5min) 3 times and transferred in FACS tubes for analysis. 25,000 counts were read per sample with BD FACS Fortessa Instrument (BD biosciences) and the data acquired was analyzed using Flow Jo software (BD).  . Before use the cells were transferred to HEPES buffer pH 7.4 (20 mmol/L HEPES, 120 mmol/L NaCl, 5 mmol/L KCl, 1 mmol/L MgCl2, 10 mmol/L glucose, 10 mmol/L sodium pyruvate) and used as described in the section for imaging studies below.

Reagents
Linoleic acid R99% purity was procured from Sigma-Aldrich and stored at À80C. A fresh vial was used for each study. All reagents were of the highest purity and procured from the specified manufacturer. Routine chemicals were obtained from Sigma-Aldrich (Saint-Louis, MO).
Blood sample handling of prospective cohorts-1 and -2 Serums samples were collected within 12 hours of ICU admission. Serum was collected using tubes with serum separators. The samples were stored, transported at 4 C and analyzed or frozen at À80 C within the next 12 hours. One sample of the 8 severe COVID-19 patients belonging to cohort 1 was spilled and lost and not available for serum FA analysis. Serum FAs were analyzed by Gas chromatography at the Vanderbilt University medical center lipidomics core .

Unbound fatty acids using ADIFAB reagent
Total serum unbound FAs were measured using the fluorescent ADIFAB2 method (FFA sciences; San Diego, CA) using the manufacturer's instructions as reported in published studies . A calibration curve using oleic acid standards (100nM to 50 mM) in DMSO was used as a reference. Figure S1 shows a summary of 9 calibration curves as black dots, with the mean depicted as the dashed line. The red dots show the values of unbound oleic acid standard supplied by the manufacturer. Based on these calibration curves and standards, low and high serum controls were generated and used for quality control for each experiment. These fluorescent unbound fatty acid results are shown for all studies. In case of quantification of specific unbound fatty acids (for cohort 1) using gas-chromatography mass-spectrometry, the following protocol was used.
Quantification of unbound fatty acids using gas-chromatography mass spectrometry (GC-MS) for cohort-1 Dealbumination: Patient sera that were to be de-albuminated were thawed and vortexed. The sample to be de-albuminated was removed and warmed to 37 C. This temperature is important to keep unbound fatty ll OPEN ACCESS