Platelet lipidome alterations in septic shock: a matched case-control study

Abstract

Background

Platelets play a central role in hemostatic and inflammatory responses during septic shock, with lipids being essential for their function. However, the specific lipidomic alterations occurring in platelets during septic shock remain poorly understood.

Objectives

This study aimed to characterize platelet lipidomic changes in septic shock and investigate their associations with disease severity.

Methods

In this matched case-control study, platelets were isolated from 49 septic shock patients and 47 nonseptic controls (matched for age, gender, and comorbidities). Lipidomic profiling was performed using untargeted lipidomics to identify significant alterations in the platelet lipidome. Associations among lipid changes, clinical data, and plasma biomarkers of coagulopathy and inflammation were explored.

Results

More than 60% of the annotated platelet lipids were significantly altered in septic shock. Cholesteryl esters, sphingomyelins, lysophosphatidylcholines, and ether-lipids were significantly reduced, while ceramide levels increased. Fatty acyl chain remodeling displayed distinct patterns, with polyunsaturated fatty acids increasing in triacylglycerols and decreasing in phospholipids. Lipid alterations were strongly associated with thrombocytopenia, and lysophosphatidylcholine levels inversely correlated with disease severity, as indicated by the Sequential Organ Failure Assessment score.

Conclusions

Septic shock induces significant disruptions in the platelet lipidome, with the extent of these alterations correlating with sepsis-associated thrombocytopenia severity. The observed changes affect multiple lipid classes, surpassing those reported under physiological conditions or in other diseases. These findings highlight the impact of sepsis-driven dysregulated inflammation and coagulopathy on platelet lipid composition, providing new insights into sepsis pathophysiology.

Essentials

• •The impact of septic shock on the human platelet lipidome is unknown.
• •We performed an untargeted LC-MS lipidomic analysis in a matched case-control study of 49 septic shock patients and 47 controls.
• •Septic shock caused widespread platelet lipid alterations, including classwide reductions in LPC, sphingomyelins, cholesterol esters, and ether lipids.
• •Lipid changes correlated with thrombocytopenia and sepsis severity, suggesting potential lipid biomarkers and mechanistic relevance in thromboinflammation.

1. Introduction

Septic shock is a severe form of sepsis, an unconstrained response to infection that leads to circulatory failure and life-threatening organ dysfunction [1]. Platelets play a crucial role in sepsis pathophysiology, as this condition is characterized by widespread coagulation abnormalities ranging from mild biological changes to severe coagulopathy. These disturbances, commonly referred to as sepsis-induced coagulopathy (SIC), can progress to disseminated intravascular coagulation (DIC) [2]. Most septic patients experience a reduction in platelet count within the first 4 days following sepsis onset. Thrombin-mediated platelet activation and subsequent consumption are considered the primary drivers of sepsis-associated thrombocytopenia [[3], [4], [5]]. Moreover, microvascular thrombosis is a hallmark of SIC, leading to ischemia and organ dysfunction. Additionally, the procoagulant state in sepsis increases the risk of macrovascular thrombosis, such as deep vein thrombosis. In later stages, excessive platelet consumption and depletion of coagulation factors heighten the risk of bleeding [6].

Beyond their role in thrombosis, platelets are key regulators of inflammation and immune responses [5]. They orchestrate both innate and adaptive immunity by releasing cytokines and lipid mediators into their environment [7,8]. These mediators contribute to cellular communication via autocrine and paracrine mechanisms [9]. The intricate interplay between coagulation and inflammation in sepsis, known as thromboinflammation, highlights the multifaceted functions of platelets in this condition [7,10].

Lipids are fundamental determinants of platelet integrity and function [11,12]. They provide structural support to the platelet plasma membrane, modulating platelet interactions with their environment and enabling shape changes during activation. Additionally, lipids serve as energy substrates and precursors for bioactive mediators. For instance, diacylglycerol (DG) is crucial for intracellular signaling, while eicosanoids mediate autocrine and paracrine communication [9,13]. The platelet lipidome is highly dynamic, shaped by lipid exchange with plasma, de novo lipogenesis, and enzymatic conversion of membrane lipids into bioactive molecules [14]. Numerous cellular enzymes regulate platelet lipid synthesis, trafficking, and metabolism [9,13]. Recent studies have reported alterations in the platelet lipidome in various pathologic conditions, including liver disease, cancer, acute coronary syndrome, and COVID-19 [[15], [16], [17], [18], [19], [20]]. These changes not only provide potential diagnostic and prognostic biomarkers but are also linked to variations in platelet reactivity. For example, in mouse models, elevated levels of arachidonic acid–containing phosphatidylethanolamine plasmalogen (PEP) lipids have been associated with enhanced thromboxane A₂ generation, promoting thrombosis and thrombus growth [21], whereas reduced levels of these PEP species correlate with decreased platelet activity [22]. In COVID-19 patients, elevated levels of lysophosphatidylglycerol and bis(monoacylglycero) phosphate have been correlated with heightened platelet reactivity to thrombin [18]. Sepsis is known to significantly alter lipid content in plasma, serum, and erythrocytes [[23], [24], [25], [26], [27], [28], [29]]. During sepsis, inflammatory lipid mediators exhibit dynamic changes, either promoting or counter-regulating inflammation. For instance, plasma levels of eicosanoids and resolvins increase [23,[30], [31], [32]], whereas lysophosphatidylcholine (LPC), a proinflammatory lipid that activates endothelial and immune cells, has been shown to decline, with lower plasma and serum levels correlating with increased mortality [27,29,[33], [34], [35]].

Despite growing evidence of multiple lipid alterations in sepsis, the lipidome of human platelets in this condition remains unexplored. Given the critical role of platelets in sepsis pathophysiology, a better understanding of their lipidomic profile could provide new insights into disease mechanisms and potential therapeutic targets. This study aims to characterize changes in the platelet lipidome of septic shock patients using a high-resolution liquid chromatography–mass spectrometry (LC-MS)-based untargeted lipidomic analysis. To ensure robust comparisons, septic patients were matched with controls based on age, gender, and major comorbidities.

2. Methods

2.1. Design and setting of the study

This study was a monocentric, prospective, translational and observational case–control study, conducted at the Cliniques universitaires Saint-Luc (Brussels, Belgium). Patients with septic shock admitted to the intensive care unit (ICU) were systematically included within 48 hours of admission, between February 1, 2019, and June 1, 2020. Clinical and demographic data were collected at the time of inclusion. To establish baseline values, anonymized clinical data from septic shock patients admitted to the ICU were first analyzed to identify key matching variables (age, sex, and major comorbidities). Based on these parameters, control subjects were prospectively enrolled from the central clinical laboratory. These individuals were age and sex matched to the septic cohort and presented similar comorbidity profiles. They were not healthy volunteers but outpatients undergoing routine clinical follow-up.

2.2. Population

Septic shock was defined according to the Sepsis-3 definition as sepsis requiring vasopressor therapy to maintain a mean arterial pressure >65 mm Hg and lactate levels >2 mmol/L despite adequate fluid resuscitation (30 mL/kg of intravenous crystalloid within 6 hours) [1]. Exclusion criteria were therapeutic anticoagulation (oral or parenteral), including heparins, fondaparinux, vitamin K antagonists, and direct oral anticoagulants; recent chemotherapy (<1 month); active inflammatory disease; hemophilia or other coagulopathies; a history of thrombocytopenia (<100,000 platelets/mm³); cirrhosis (Child–Pugh>A); recent major surgery (<48 hours) unrelated to septic shock or infection source control; cardiac arrest during ICU stay; and any decision regarding limitations of care. All septic patients received thromboprophylaxis with low-molecular-weight heparin (nadroparin, 3800 IU/d subcutaneously). Blood sampling was performed at least 6 hours after low-molecular-weight heparin injection.

2.3. Clinical outcomes

The prognosis of septic shock patients was assessed using the acute physiology and chronic health evaluation (APACHE) II [36] and the sequential organ failure assessment (SOFA) [37] scores. DIC and SIC were diagnosed at inclusion based on the International Society of Thrombosis and Haemostasis (ISTH) scoring system [38,39]. Relevant data were collected from central medical records, including routine biological parameters measured in patients admitted to ICU, such as platelet count, C-reactive protein (CRP) levels, coagulation markers, renal function, and liver enzyme levels. Mortality was assessed at 30 days and 1 year after inclusion.

2.4. Sampling

Blood samples were collected from ICU patients via central venous catheters and from control participants via venipuncture. Vacutainer tubes containing citrate phosphate dextrose adenine were used for blood collection. The platelet isolation protocol was adapted from Burzynski et al. [40]. Immediately after collection, blood samples were centrifuged at 330g for 20 minutes at room temperature (RT) to obtain platelet-rich plasma. To prevent platelet activation, apyrase (1:1000) and eptifibatide (1:500) were added to the platelet-rich plasma prior to a second centrifugation at 400g for 10 minutes at RT. The resulting supernatant was collected and centrifuged at 13,500g for 15 seconds to prepare platelet-poor plasma (PPP). PPP samples were snap-frozen in liquid nitrogen and stored at −80 °C for subsequent biomarker quantification. In parallel, the platelet pellet was immediately snap-frozen and stored at −80 °C until lipidomic analysis.

2.5. Measurement of biomarkers

Soluble plasma biomarkers of inflammation and NETosis were quantified using enzyme-linked immunosorbent assay (ELISA) or suspension array sandwich immunoassays, following regulatory standards for commercially available research-use-only ELISA kits. Frozen PPP was thawed at RT on the day of the experiment. Each assay was performed in duplicate, adhering to the validated lower and upper limits of quantification. Cytokines and chemokines were measured using the Bio-Plex Pro Human Cytokine 27-Plex Panel (27-Plex) according to the manufacturer’s protocol [41]. Details of the analyzed biomarkers are provided in Supplementary Table 1.

2.6. Platelet aggregometry

Platelet function was evaluated using multiple electrode platelet aggregometry (Multiplate), which quantifies platelet aggregation by measuring the time-dependent increase in impedance between 2 electrodes. Whole blood samples were stimulated with 3 agonists—ADP (6.5 μmol/L), arachidonic acid (AA, 0.5 mmol/L), and thrombin receptor–activating peptide (TRAP, 32 μmol/L)—to assess distinct platelet activation and aggregation pathways. Patients with a platelet count <100 × 10⁹/L were excluded, as results below this threshold are considered unreliable [42]. Aggregation values (AU × min) were normalized to platelet count (AU × min per 10⁹ platelets/L) to account for interindividual variability in platelet numbers [43,44].

2.7. Lipidomic analysis

Lipid extraction, LC-MS analysis, and data processing were performed as previously described [45]. Briefly, lipids were extracted from frozen samples, not previously thawed, containing subject platelets and spiked with 6 internal standards—LPC 13:0; PC 19:0/19:0; PC 14:0/14:0; PS 12:0/12:0; PG 15:0/15:0; and PE 17:0/17:0 (Avanti Polar Lipids Inc). All sample volumes were adjusted during the final extraction step to normalize the total platelet count to 1 × 10⁸ platelets per sample. Signal intensities obtained from LC-MS analyses were further normalized using the cyclic loess algorithm (normalizeBetweenArrays function of R software; R Core Team), ensuring comparability across samples. For each sample, 1 μL corresponding to 3.33 × 10⁵ platelets was injected into a 1290 Infinity high-performance liquid chromatography system coupled with a 6530 Accurate Mass Q-TOF (Agilent Technologies Inc) via a dual electrospray ionization source. Lipid elution was performed on a Zorbax Eclipse plus column (C18, 2.1 × 100 mm, 1.8 μm; Agilent Technologies Inc), maintained at 40 °C. Chromatographic separation was achieved using an 83-minute gradient with solvent A (0.2% formic acid and 10 mM ammonium formate in water) and solvent B (0.2% formic acid and 5 mM ammonium formate in methanol/acetonitrile/methyl tert-butyl ether, 55:35:10 [v/v/v]). Raw MS data were processed as previously described in detail using Mass Hunter Qualitative Analysis (version B.06 or B.07; Agilent Technologies) for peak picking and in-house bioinformatic scripts (available upon request on github.com), which were optimized for the following steps using log2-transformed data: Retention time correction, filtering for presence in at least 80% of samples in at least 1 group, normalization of signal intensities using cyclic loess algorithm (with normalizeBetweenArrays function), missing data imputation on scaled data using K-nearest neighbors (with a setting of k = 5) and batch correction using Combat algorithm. A maximal value of 20% was set for missing values for a given lipid feature in any groups and of 80% for the coefficient for interindividual variation among subjects. The resulting final corrected datasets contain lipid features, defined by their m/z, retention time and corrected signal intensity. Lipid features were annotated to lipids with their respective acyl chains by MS/MS analysis and with our in-house database as previously described [45]. For lipids represented by multiple features, only the major one was retained, and reported as unique lipid. Priority for annotation by MS/MS analysis was given to lipid features found to be significantly associated with sepsis.

2.8. Statistical analysis

Statistical analysis was performed using R software (version 4.2.1; R Foundation for Statistical Computing) and the limma (v 3.54.2) Bioconductor package on the log2-transformed and corrected dataset. Multivariable regression models were built for each unique lipid species, incorporating disease group (control vs septic shock), gender, age, diabetes status, and statin treatment as predictors. Fold-change (FC) estimates and corresponding P values were derived for each lipid species and predictor. To account for multiple testing, P values were adjusted using Benjamini–Hochberg false discovery rate (FDR) method, with an FDR of <0.05 considered statistically significant.

2.9. Ethical approval

The study protocol was approved by the ethics committee, and all participants provided written informed consent (B403201938590 and NCT04107402). All authors had full access to the primary clinical data.

3. Results

3.1. Patients’ baseline characteristics and outcomes

A total of 47 and 49 patients were included in both the control and septic shock groups. Table summarizes their baseline characteristics and clinical outcomes. The mean age was 62 years in the control group and 65 years in the septic shock group. There were no significant differences in major comorbidities between the groups, with hypertension and overweight affecting approximately half of the patients.

Table.

Patients’ baseline characteristics and outcomes.

Characteristic	Control (n = 47)	Septic shock (n = 49)	P
Demographics
Men	26 (55)	24 (49)
Women	21 (45)	25 (51)	.5478
Age (y)	62.0 ± 14.7	65.6 ± 14.6	.2384
Medical history
Hypertension	20 (43)	26 (53)	.3161
BMI > 25	26 (58)	27 (55)	.8372
Diabetes	11 (23)	8 (16)	.4479
Hypercholesterolemia on statin	13 (28)	11 (22)	.6401
History of smoking	10 (21)	15 (31)	.3559
COPD	4 (8.5)	5 (10)	1
CKD	9 (19)	11 (22)	.8032
Cancer history	14 (30)	9 (18)	.2349
Pre-existing treatment
P2Y12 inhibitor	2 (4.3)	3 (6.1)	1
Aspirin	9 (20)	16 (33)	.1652
Statins	13 (28)	10 (20)	.4769
Routine laboratory testing
CRP (mg/dL)	5.9 ± 9.7	223.9 ± 110.5	<.0001
Highest CRP (mg/dL)	NA	313 ± 122	NA
Creatinine (mg/dL)	1.2 ± 1.3	2.2 ± 1.9	.0045
Hemoglobin (g/dL)	13.3 ± 1.5	10.3 ± 2.0	<.0001
WBCs (103/µL)	7.3 ± 2.0	17.5 ± 10.8	<.0001
Neutrophils (103/µL)	4.4 ± 1.4	15.5 ± 10.1	<.0001
Lowest lymphocytes (103/μL)	NA	451.5 ± 356.0	NA
Platelets (103/µL)	255.7 ± 84.3	201.0 ± 137.3	.0206
Thrombocytopenia (<150 × 103/µL)	3 (6.4)	21 (43)	<.0001
Total cholesterol (mg/dL)	183.4 ± 41.1	89.0 ± 38.5	<.0001
LDL cholesterol (mg/dL)	101.2 ± 40.2	34.4 ± 34.6	<.0001
HDL cholesterol (mg/dL)	54.0 ± 14.4	21.2 ± 13.5	<.0001
Triacylglycerols (mg/dL)	148.8 ± 85.5	162.3 ± 118.4	.5761
Plasma cytokines
IL-1B (pg/mL)	2.6 ± 2.3	4.6 ± 3.6	.0019
IL-6 (pg/mL)	2000 ± 3400	2356.9 ± 4068.7	.0003
TNFα (pg/mL)	68.2 ± 13.8	152.1 ± 193.7	.0053
sTREM-1 (pg/mL)	127.4 ± 111.2	541.0 ± 350.0	<.0001
MCP-1 (pg/mL)	28.0 ± 10.0	302.1 ± 508.1	<.0001
MPO (ng/mL)	28.9 ±2 0.7	763.6 ± 1,313.4	.0004
NE (ng/mL)	7.6 ± 3.0	172.0 ± 153.3	<.0001
Platelet functional testing
ADP (AU × min/platelets × 103/µL)	2.8 ± 1.0	4.2 ± 5.1	.0796
ASPI (AU × min/platelets × 103/µL)	3.2 ± 1.5	7.6 ± 14.2	.0474
TRAP (AU × min/platelets × 103/µL)	4.3 ± 1.8	6.5 ± 9.5	.1381
Coagulation assessment
INR	1.0 ± 0.1	1.6 ± 0.5	<.0001
TATc (ng/mL)	8.3 ± 18.2	22.9 ± 29.7	.0059
D-dimer (ng/mL)	701 ± 769.5	9185.8 ± 10182	<.0001
Organ failure and severity scores	NA		NA
Pao2/Fio2		223.2 ± 118.8
Ventilation duration (d)		2.86 ± 6.1
NorE highest doses (Μg/kg/min)		0.3 ± 0.4
NorE duration (d)		4.8 ± 6.1
Renal replacement therapy		15 (31)
APACHE II score		20.4 ± 7.0
SOFA score		8.9 ± 3.2
SIC score		12 (24)
DIC score		7 (15)
Outcome
30-d mortality	0 (0)	23 (47)	<.0001
1-y mortality	0 (0)	28 (57)	<.0001
ICU length of stay (d)	NA	8.2 ± 9.0	NA

Values are expressed as numbers (percentages), and continuous data are presented as mean ± SD. For platelet functional testing, a ratio of the impedance aggregometry reading and the platelet count was calculated. A P value of <0.05 was considered statistically significant.

APACHE, acute physiology and chronic health evaluation; ASPI, arachidonic acid; BMI, body mass index; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; DIC, disseminated intravascular coagulopathy; HDL, high-density lipoprotein; ICU, intensive care unit; IL, interleukin; INR, international normalized ratio; LDL, low-density lipoprotein; MCP, monocyte chemoattractant protein; MPO, myeloperoxidase; NE, neutrophil elastase; NorE, norepinephrine; Pao₂/Fio₂, arterial oxygen partial pressure/fractional inspired oxygen; SIC, sepsis-induced coagulopathy; SOFA, sequential organ failure assessment; TATc, thrombin-antithrombin complex; sTREM, soluble triggering receptor expressed on myeloid cell; TNF, tumor necrosis factor; TRAP, thrombin receptor activating peptide; WBC, white blood cell.

Routine laboratory tests of septic shock patients, recorded at ICU admission, revealed significant differences compared with control patients. Septic shock patients exhibited elevated CRP levels, increased white blood cell count, acute kidney injury (assessed by elevated creatinine), and anemia (assessed by decreased hemoglobin levels). Their platelet count was significantly lower than that of the control group (202 ± 136 vs 253 ± 85 × 10³/μL), although still within the normal range on average. Notably, 43% of patients with septic shock had thrombocytopenia (platelet count <150 × 10³/μL), compared with only 6.4% of control patients. Regarding lipid profiles, low-density lipoprotein (LDL) cholesterol and high-density lipoprotein (HDL) cholesterol levels were significantly reduced in septic shock patients (LDL: 45 ± 73 vs 101 ± 34 mg/dL; HDL: 21 ± 13 vs 54 ± 14 mg/dL), while plasma triacylglycerol (TG) levels remained unchanged. Inflammatory biomarker analysis showed that plasma levels of interleukin (IL)-1β, IL-6, tumor necrosis factor α, soluble triggering receptor expressed on myeloid cells 1, monocyte chemoattractant protein (MCP)-1, myeloperoxidase, and neutrophil elastase were all significantly increased in septic shock patients. When platelet aggregation was normalized to platelet count, values tended to be higher in septic patients than controls, reaching statistical significance for ASPI (P = .047), whereas differences for ADP and TRAP did not reach significance. International normalized ratio (INR) values and D-dimer levels were significantly higher in septic shock patients than those in controls. Septic shock severity was assessed using SOFA and APACHE II scores, with mean values of 9 ± 3 and 20 ± 7, respectively. SIC was diagnosed in 24% of patients, while DIC was observed in 16%. Septic shock patients had a mean ICU length of stay of 8 days. Their 30-day mortality rate was 46%, and their 1-year mortality rate was 58%. In contrast, none of the patients in the control group died during the 1-year follow-up.

3.2. Platelet lipidome was strikingly altered during septic shock

Using a LC-MS–based untargeted lipidomic approach, a total of 847 lipid features were successfully detected, of which 224 were annotated to unique lipid species. According to the lipid maps consortium classification [46], the platelet lipid profile established in this study comprised lipids from 19 different classes/subclasses, covering 5 major lipid categories: fatty acyls (FAs; 4 species/1.8%), glycerolipids (56 species/25%), phospholipids (PL; 130 species/58%), sphingolipids (30 species/13.4%), and sterols lipids (4 species/1.8%) (Figure 1A). Principal component analysis of the platelet lipidomic profiles revealed a clear separation between septic shock patients and control subjects (Figure 1B). Importantly, no distinct clustering was observed according to the site of infection or type of pathogen (Figure 1C, D), nor according to age (< or > 60 years), sex, diabetes status, or treatment with aspirin, P2Y12 inhibitors, insulin, or statins (Supplementary Figure 1A–G). Overall, 539 of the 847 lipid features (63%) exhibited significant variation (FDR < 0.05) between septic shock patients and matched controls. Among the 224 annotated unique lipids, 144 species (64%) showed significant alterations, with 54 upregulated and 90 downregulated. These changes were observed across all 5 lipid categories (Figure 1A, E; Supplementary Figure 2). Further details on specific lipid changes are provided in Supplementary Table 2.

Figure 1.

Platelet lipidome undergoes critical alterations in septic shock patients. (A) Profile of identified lipids: In gray, numbers of detected, identified, and significantly altered lipid species are shown. From left to right, untargeted lipidomic analysis enabled the detection of lipid features from 5 major categories (each represented by a specific color), with a hierarchical subclassification. Fatty-acyls are shown in salmon, glycerolipids in khaki, glycerophospholipids in green, sphingolipids in blue, and sterols in pink. On the right, the number of annotated lipid species for each class/subclass is indicated, along with the count of significantly altered species in septic shock patients (n = 49) compared with controls matched for sex, gender, and main comorbidities (n = 47). A Benjamini-Hochberg correction was applied, with a false discovery rate (FDR) of <0.05 considered statistically significant. (B) Principal component analysis (PCA) plot of platelet lipid profiles in control and septic shock patients. (C) PCA plot of platelet lipid profiles in control and septic shock patients grouped by the site of infection (abdominal, lung, urinary, and other). (D) PCA plot of platelet lipid profiles in control and septic shock patients, categorized based on the involved pathogen (Gram-positive cocci, Gram-negative bacilli, other pathogens, co-infection, or unknown). (E) Volcano plot comparing septic shock patients and controls, depicting the 224 annotated platelet unique lipids. The X-axis represents the base 2 logarithm (Log₂) of the fold-change (FC) in mass spectrometry signal intensity between the 2 groups (sepsis/controls) for each lipid species. The Y-axis shows the -Log₁₀-adjusted P values. Each dot represents a lipid species, colored according to the categories defined in (A). (F) Boxplots of lipid classes/subclasses, comparing septic shock patients and controls. The Y-axis displays the Log₂FC of signal intensity values for each lipid species (sepsis vs controls). The X-axis shows lipid classes/subclasses, organized and color-coded according to their respective categories. Within each class/subclass, each dot represents a lipid species. Black dots indicate significantly different lipids between the groups, with an FDR of <0.05 considered statistically significant.

3.3. Septic shock induced classwide homogeneous shifts in specific lipid species

Each lipid species within a given lipid class or subclass was visualized using boxplots (Figure 1F), enabling a classwide analysis of lipid changes in the platelets of septic shock patients. A significant reduction in specific cholesteryl esters (CEs), namely CE 18:1, CE 18:2, and CE 20:4, was observed in septic shock platelets, while free cholesterol levels remained unaltered. Free cholesterol is the predominant platelet lipid, and among the CE, these 3 identified species are known to be the most abundant [11]. Cholesterol is enriched in specialized signaling microdomains within the platelet membrane, known as lipid rafts, which play a crucial role in regulating membrane protein localization [47].

Similar to cholesterol, sphingomyelins (SMs) are major components of lipids rafts, contributing to their structure and function [47]. Septic shock patients exhibited extensive alterations in the platelet sphingolipidome, characterized by a reduction in the majority of the detected SM species (SM 18:1;O2/18:0, SM 18:1;O2/20:0, SM, 18:1;O2/22:0, SM 18:1;O2/24:0, SM 18:2;O2/14:0, SM 18:2;O2/16:0, SM 19:1;O2/22:0, SM 19:1;O2/24:0, SM 19:1;O2/24:1, SM 42:0;O2, SM 44:2;O2). SMs can be hydrolyzed by sphingomyelinase, which removes a phosphocholine group to generate Ceramide (Cer), a process essential for platelet membrane integrity, as platelets lack de novo ceramide synthesis [11,48]. In line with this SM depletion, a simultaneous increase in most detected ceramide species (Cer 18:1;O2/18:0, Cer 18:1;O2/23:1, Cer 18:1;O2/24:1, Cer 18:1;O/24:1, and Cer 18:0;O/22:0) was observed in septic shock platelets.

PLs serve as the primary building blocks of the plasma membrane and function as precursors for bioactive molecules [47]. Their hydrolysis by phospholipase A₂ (PLA₂) produces a free fatty acid and a lysophospholipid, such as LPC from PC hydrolysis [23]. LPC levels were markedly reduced in septic shock platelets, with LPC 18:0 showing the most significant decrease.

Ether lipids (PCO, PCP, LPCO, PEO, and PEP) are peroxisome-derived PLs in which the sn-1 hydrocarbon chain is linked via an ether bond rather than the more common ester bond found in diacyl PL. They exist in 2 forms: plasmanyl lipids (with an alkyl residue) and plasmalogens (also called plasmenyl lipids, with an alkenyl residue). Ether bonds influence membrane dynamics, promoting membrane fusion and stabilizing lipid raft microdomains [49]. Interestingly, septic shock platelets showed a marked decrease in PCO, LPCO, PEO, and PEP species, whereas ester-linked PL (diacyl PL) did not exhibit a classwide reduction.

3.4. Septic shock triggered a remodeling of FA chains in TG, PC, and PE

To assess changes in diacyl PL and TG at the species level, the FA chain length and degree of unsaturation (number of double bounds) were compared between the 2 groups (Figure 2). Among diacyl-PL classes, a significant remodeling of FA chains was observed in PC and PE. Specifically, PC and PE containing at least 1 long polyunsaturated fatty acid (PUFA) chain were reduced. In contrast, PC and PE with shorter and more saturated fatty acids (18-carbon chains with 0-3 double bounds) were increased in septic shock platelets. This remodeling pattern indicates a decrease in platelet PC and PE species enriched in omega-3 and omega-6 fatty acids, including eicosapentaenoic acid (20:5), docosapentaenoic acid (22:5), docosahexaenoic acid (22:6), and AA (20:4) (Figure 2A, B).

Figure 2.

Septic shock triggers a remodeling of fatty-acyls chains in triacylglycerols (TGs) and phospholipids. (A) Dot-plot of glycerophospholipid remodeling in septic shock patients compared with controls. The Y-axis displays 1 fatty acyl chain of phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), and phosphatidylinositols (PIs), while the X-axis shows the other fatty acyl chain. The intensity of the dot color is proportional to the Log₂FC in signal intensity between groups, with upregulated lipids in red and downregulated lipids in blue. Dot size reflects adjusted P values calculated using the Benjamini–Hochberg method, with larger dots indicating statistical significance (FDR < 0.05). (B) Forest plot of platelet PCs and PEs, comparing septic shock and control patients. Each line on the Y-axis represents a PC or PE species. The X-axis shows the Log₂FC of peak intensities for each species, comparing septic shock and control groups. Horizontal lines indicate the 95% CI. Only significantly altered species, with a |Log₂FC| ≥ 0.5, are represented, with statistical significance defined as an adjusted P value of <0.05. (C) Dot-plot of TG remodeling in septic shock patients compared with controls. Similar to (A), but here, the Y-axis displays the degree of unsaturation (number of double bounds), while the X-axis indicates the carbon chain length. (D) Forest plot of platelet TGs, comparing septic shock and control patients. Similar to (B) but showing the Log₂FC of significantly altered TG species.

Although glycerolipids, such as DG and TG, constitute only a small fraction of the platelet lipidome, they play essential roles. DG species act as bioactive signaling mediators, and TG species serve as the primary energy substrates for platelets [13,14,50]. During septic shock, a notable shift in TG composition was observed. Long-chain and polyunsaturated TG species were predominantly increased. In contrast, shorter and more saturated TG species (total acyl chain length <50 carbons, 0-3 double bonds) were significantly reduced in platelets (Figure 2C, D). Of note, statin treatment influenced platelet lipid composition similarly in both septic shock patients and controls. Specifically, platelets from statin-treated patients exhibited higher levels of several TG species and lower levels of some PL species (Supplementary Table 3; Supplementary Figure 3A), consistent with recent reports [51]. Comparing the FC effects of statins in control and septic patients revealed a strong positive correlation across lipid species (Supplementary Figure 3B), indicating similar statin effects in both groups and no significant interaction with disease status.

3.5. Platelet lipidome alterations were associated with sepsis-associated thrombocytopenia and sepsis severity

Correlations between significantly altered platelet lipid species and clinical as well as biological parameters in septic shock patients were analyzed and summarized as correlation matrices (Supplementary Figure 4A–G). Specifically, potential associations were evaluated with comorbidities (diabetes mellitus and hyperlipidemia), treatments (statin use, aspirin use, P2Y12 inhibitor use, and insulin therapy), plasma lipid profiles (LDL, HDL, and TGs), septic severity and prognostic indicators (APACHE score, SOFA score, and 30-day mortality rate), platelet count and activation/reactivity markers (platelet count, large platelet ratio, sCD62P, platelet aggregometry with ADP, ASPI, and TRAP), coagulation and fibrinolysis biomarkers (INR, thrombin-antithrombin complex, D-dimers, ATIII, tissue factor, ristocetin cofactor, tissue-type plasminogen activator, plasminogen activator inhibitor 1, and tissue factor pathway inhibitor), endothelial activation markers (vascular cell adhesion molecule and intercellular adhesion molecule), and inflammatory mediators (eg, CRP, soluble triggering receptor expressed on myeloid cells 1, tumor necrosis factor α, IL-1β, IL-6, MCP-1, myeloperoxidase, neutrophil elastase, and CitH3).

3.5.1. Platelet lipid changes and clinical factors

Platelet lipid alterations in septic shock were not associated with dyslipidemia, diabetes, aspirin or P2Y12 inhibitor use, insulin therapy, or statin treatment (Supplementary Figures 1, 3, and 4A–G). Correlations between platelet and plasma lipid classes were further investigated to assess whether platelet lipid remodeling in sepsis reflects systemic alterations of the plasma lipidome. In our cohort, platelet TG levels correlated significantly with plasma TG concentrations (Supplementary Figure 4B). Both LPC and CE species showed significant positive correlations with HDL and LDL levels, consistent with their role as structural components of these particles. In contrast, platelet TGs and PCs did not correlate with HDL or LDL, and although some SM species correlated with HDL, no correlation was observed with LDL (Supplementary Figures 5A–E and 6A–E).

3.5.2. Association with thrombocytopenia and sepsis severity

Platelet lipid modification during septic shock correlated with sepsis-associated thrombocytopenia. Patients were stratified according to platelet count (≥150,000 vs <150,000 platelets/μL). Although the thrombocytopenic group was smaller (21 vs 28 patients), these patients exhibited more pronounced and statistically significant lipid alterations compared with nonthrombocytopenic patients (Figure 3A). However, sepsis-induced coagulopathy, assessed by the INR, and platelet function, assessed by impedance aggregometry, showed no correlation with lipid changes. Interestingly, the SOFA score was significantly correlated with a decrease in each identified LPC species (R = 0.31-0.57) (Figure 3B).

Figure 3.

Association between platelet lipidome alterations, sepsis-associated thrombocytopenia, and sepsis severity. (A) Volcano plots of the 224 annotated platelet-specific lipids comparing septic shock patients and controls. The left panel shows septic shock patients with thrombocytopenia (platelet count <150,000/μL), and the right panel shows patients without thrombocytopenia (platelet count ≥150,000/μL). The X-axis represents the Log₂ fold change (sepsis vs controls) in mass spectrometry signal intensity for each lipid species, and the Y-axis the –Log₁₀-adjusted P values. Each dot represents a lipid species, colored according to the categories defined in Figure 1A. (B) Scatter plots showing the correlation between lysophosphatidylcholine (LPC) levels and sepsis severity, assessed by the sequential organ failure assessment (SOFA) score. For each LPC species, the Y-axis represents the Log of semiquantitative lipid abundances (arbitrary unit [AU]). The X-axis displays the SOFA score. Each dot represents a patient, with gray indicating control patients and black representing septic shock patients. The strength of the correlation between LPC levels and SOFA scores in septic shock patients is indicated by Pearson correlation coefficient (R). Controls are displayed for visual comparison but were not included in the correlation analysis since their SOFA score was zero.

3.5.3. Correlations with inflammatory and coagulation markers

Given the relationship between platelet lipid alterations and sepsis-associated thrombocytopenia, we further explored correlations between modified lipids and biomarkers of inflammation and coagulopathy. Several PC species containing long-chain PUFAs (including PC 18:2_22:5, 20:4_20:5, 20:4_22:6, and 20:4_22:5) were negatively correlated with MCP-1, CitH3, and vascular cell adhesion molecule. Additionally, several LPC species (including LPC 16:0, 18:0, 18:1, and 20:0) showed negative correlations with MCP-1, IL-1β, IL-6, NE, and CitH3. Numerous TG species demonstrated positive correlations with CitH3 levels, whereas CE species correlated positively with ATIII. Collectively, these findings suggest a link between altered platelet lipid composition, inflammation, and endothelial/coagulation pathways during septic shock.

4. Discussion

This study provides the first comprehensive analysis of the impact of septic shock on the platelet lipidome. Using an untargeted lipidomic analysis, we analyzed platelets from 49 septic shock patients and 47 controls matched for age, gender, and comorbidities. The results reveal a profound alteration in platelet lipid composition, with 64% of annotated lipids significantly affected. The major findings can be summarized as follows: (1) classwide downregulations were observed in CE, SM, LPC, and ether lipids, while ceramide levels were significantly increased in septic platelets; (2) opposite remodeling of PL and TG was noted, with PL-PUFA species decreasing and TG-PUFA species increasing; (iii) lipid alterations correlated with thrombocytopenia severity, suggesting a link between platelet lipid metabolism and platelet count in septic shock; and (iv) LPC levels were inversely correlated with sepsis severity, reinforcing their potential role as biomarkers of disease progression.

Recent MS-based lipidomic studies have demonstrated platelet lipid alterations in inflammatory diseases such as COVID-19 and coronary artery diseases [15,18,19,52], as well as in obesity and bleeding disorders [12,53]. The platelet lipidome is primarily composed of cholesterol, PC, PE, PS, PI, and SM, with only 15 lipids accounting for 70% of the total lipid mass [11]. Interestingly, our study identified notable alterations in key lipid species within this subset, including PE 20:4_16:0 (Log₂FC −0.25); SM 18:1;O2/24:0 (Log₂FC −0.82); PC 20:4_16:0 (Log₂FC −0.18); and PE 20:4_18:1 (Log₂FC 0.49)). These findings suggest that sepsis-associated lipid change has a substantial impact on the overall lipid profile, surpassing the alterations previously reported in COVID-19 and acute coronary syndrome platelets [15,18,20].

The significant reduction of ether lipids in septic shock platelets is particularly striking. Ether lipids, a subclass of PL characterized by an ether bond, play essential roles in membrane stability and oxidative stress protection [49]. A similar decrease was previously observed in COVID-19 platelets [18]. Given the central role of oxidative stress in septic shock and COVID-19, a plausible explanation is that plasmalogens act as scavengers for reactive oxygen species [54,55]. Additionally, ether lipids, rich in PUFA at the sn-2 position, may serve as a reservoir for bioactive lipid mediators [21,56]. PUFAs such as AA, eicosapentaenoic acid, and docosahexaenoic acid, are enzymatically released from PL by PLA₂ during inflammation, generating precursors for proinflammatory and anti-inflammatory mediators [47,57]. This is consistent with another key observation from our study: a marked reduction in PUFA-containing PLs. Similar PUFA-PL depletion has been reported in the plasma and erythrocytes of septic shock patients [26,34,58], and in the platelets of patients with acute coronary syndrome, a condition recognized for its thromboinflammatory response [15].

Of note, the decline in PUFA-PL coincided with a notable rise in PUFA-TG in septic shock platelets. PUFA-PLs are highly susceptible to lipid peroxidation under oxidative stress, a process that can trigger ferroptosis, a regulated form of cell death [[59], [60], [61]]. To mitigate this, cells redistribute PUFAs from PLs to lipid droplets, storing them as TGs. This adaptative membrane remodeling reduces PUFA-PL vulnerability to oxidation, while increasing PUFA-TG levels [60,61].

Sphingolipids also serve as a reservoir for bioactive lipids [62,63]. In this study, we observed a significant reduction in SM species, associated with an increase in ceramide in the septic shock platelets. This pattern strongly suggests activation of acid sphingomyelinase (SMPD1), an enzyme that hydrolyzes SM into ceramides in response to platelet activation, inflammation, and oxidative stress [11,62,64]. Increased platelet ceramide levels have been reported in coronary artery diseases [52]. Notably, ceramide accumulation in platelets has been linked to a proinflammatory phenotype rather than direct hemostatic alterations. In a mouse model of abdominal aortic aneurysm, platelet ceramide accumulation promoted proinflammatory cytokine release and platelet-leukocyte aggregate formation [65], 2 phenomena frequently observed in septic shock [66]. Similar increases in ceramide levels, along with decreases in SM, have been documented in both plasma and erythrocytes of septic shock patients [26,27,29].

Another striking finding was the significant reduction in LPC levels in septic shock platelets. LPC, through G protein–coupled and Toll-like receptors, modulates vascular inflammation, promotes chemoattractant release, and enhances oxidative stress [67]. Interestingly, LPC is a key component of platelet extracellular vesicles [68], which mediate cell-to-cell communication. In sepsis, platelet extracellular vesicles are known to be highly proinflammatory and procoagulant [69,70]. Thus, the observed platelet LPC depletion could result from increased secretion of LPC-enriched vesicles. A similar decrease has been observed in COVID-19 platelets [18]. Additionally, plasma and erythrocytes LPC levels have been shown to decline in sepsis possibly due to conversion into lysophosphatidic acid, a bioactive lipid implicated in sepsis pathophysiology [26,29]. Importantly, platelet LPC levels were significantly and negatively correlated with sepsis severity (SOFA score), in agreement with plasma LPC depletion in sepsis patients [29,31,34,35].

The mechanisms driving platelet lipid alterations in sepsis could initially appear to be linked to thrombin-induced platelet activation. However, ex vivo thrombin stimulation induces only minimal changes to the platelet lipid landscape, primarily increasing oxidized PUFAs and re-esterified PL, while reducing TGs [11,21]. In contrast, sepsis induces profound lipidomic shifts, characterized by TG accumulation and major PL remodeling. This suggests that sepsis-associated platelet activation involves additional mechanisms, potentially influenced by (i) direct interactions with pathogens, (ii) crosstalk with endothelial cells and immune cells [[71], [72], [73], [74], [75], [76], [77], [78]], or (iii) removal of activated platelets contributing to microthrombi formation.

The platelet lipid alterations observed in our study may partly reflect the profound changes occurring in the plasma lipidome during sepsis. Platelets can exchange lipids with plasma components through the open canalicular system [79,80] and via apolipoprotein-mediated interactions with HDL and LDL particles [81]. This is supported by the significant correlations observed between platelet LPC and CE species and plasma HDL/LDL levels, consistent with their role as major lipoprotein constituents. However, the absence of correlations for other lipid classes such as PC, SM, and TG, together with findings from other clinical settings showing dissociation between plasma and platelet lipid composition [52,53], suggests that selective and context-dependent platelet remodeling mechanisms also contribute to these changes.

Although no direct correlation was observed between specific platelet lipid species and platelet aggregation, this does not preclude a functional impact of the observed lipidomic remodeling. Platelet lipids are increasingly recognized as key modulators of functions extending beyond hemostasis including inflammatory signaling, immune cell interactions, and intercellular communication through platelet-derived extracellular vesicles [13,22,47,82] . In the context of sepsis, these lipid-mediated pathways may contribute to the proinflammatory and procoagulant platelet phenotype characteristic of the disease. Further studies integrating lipidomics with functional platelet assays and immune profiling will be needed to delineate these mechanisms more precisely.

This study has certain limitations. First, the untargeted lipidomics approach used is semiquantitative, as it relies on class-representative internal standards rather than compound-specific calibration. Therefore, the reported data reflect relative rather than absolute lipid concentrations. Second, the lipidomic analysis was restricted to platelets, without concurrent plasma lipid profiling. Since platelets continuously exchange lipids with plasma, examining plasma lipid alterations in the same patients would provide valuable insights and allow for correlations between platelet and plasma lipid changes in septic shock. Finally, the lipidomic workflow used in this study did not encompass lower-abundance inflammatory mediators such as prostaglandins and isoprostanes. Complementary analyses using targeted LC-MS methods capable of quantifying these lipid species [77] could further advance our understanding of the interplay between platelet-derived lipid mediators and plasma inflammatory biomarkers in sepsis.

In conclusion, this study provides the first comprehensive lipidomic analysis of septic shock platelets, revealing extensive alteration affecting > 60% of lipid species. These alterations, spanning all major lipid categories, likely influence platelet function in hemostasis and inflammation. Notably, lipidomic changes correlated with thrombocytopenia severity and key clinical markers of sepsis severity and inflammation. These findings offer novel insights into sepsis pathophysiology and provide a foundation for future research into potential diagnostic, prognostic, and therapeutic applications.