Lipidomic Signatures Align with Inflammatory Patterns and Outcomes in Critical Illness

Alterations in lipid metabolism have the potential to be markers as well as drivers of the pathobiology of acute critical illness. Here, we took advantage of the temporal precision offered by trauma as a common cause of critical illness to identify the dynamic patterns in the circulating lipidome in critically ill humans. The major findings include an early loss of all classes of circulating lipids followed by a delayed and selective lipogenesis in patients destined to remain critically ill. Early in the clinical course, Fresh Frozen Plasma administration led to improved survival in association with preserved lipid levels that related to favorable changes in coagulation and inflammation biomarkers. Late over-representation of phosphatidylethanolamines with critical illness led to the validation of a Lipid Reprogramming Score that was prognostic not only in trauma but also severe COVID-19 patients. Our lipidomic findings provide a new paradigm for the lipid response underlying critical illness.


Introduction 55
Acute critical illness is a major healthcare burden and commonly leads to short and long-term morbidity and 56 mortality 1,2 . Common causes of acute critical illness, including severe injury and infections, are among the 57 leading causes of death worldwide 3 . Most recently, the COVID-19 pandemic has emerged as a major to acute critical illness. 72 Lipids comprise 30% of the body's non-water mass and are not only a main component of cell 3 membranes but also important energy substrates and signaling molecules 11 . Previous studies in critically ill 74 humans provide evidence that lipolysis and lipogenesis are altered dramatically in acute critical illness. For 75 example, circulating levels of glycerolipids, sphingolipids, phospholipids, and lyso-phospholipids vary from 76 baseline in patients with acute critical illness 12-18 . However, a comprehensive assessment of the changes in 77 circulating lipids that correlate with outcomes and markers of disease in acute critical illness is lacking. 78 To define the changes in the circulating lipidome associated with acute critical illness, we utilized a a Lipid Reprogramming Score that was found to correlate highly with later patient outcomes. These findings 88 were validated in a second trauma database and two publicly available databases that include critically ill 89 COVID-19 patients, suggesting that some of the unique lipidomic patterns identified in this study may be 90 generalizable to critical illness resulting from multiple etiologies. To determine the dynamics changes in circulating lipids after severe injury in humans, we carried out a 95 quantitative analysis of plasma lipid levels in samples obtained during the PAMPer trial 19 . This prospective, 96 multi-institutional, pragmatic trial enrolled seriously injured humans suffering polytrauma at risk for 97 hemorrhagic shock. Only patients that were transported by helicopter to a Level 1 trauma center were 98 included and randomization took place in the pre-hospital setting. Patients in the treatment arm received two 99 units of FFP initiated during helicopter transport, while the control group was assigned randomly to 100 standard-of-care, which did not include FFP in the pre-hospital setting. The use of pre-hospital FFP was 101 associated with a 9.8% reduction in 30-day mortality (p=0.03) 19 . A total of 193 of the original 523 patients 102 were selected for lipidome analysis (Fig S1). This cohort included both non-survivors (n=72) and survivors 103 (n=121) selected to represent the overall cohort. Samples were obtained at admission to the trauma center 104 (0h) and at 24 and 72h after admission. Only the time 0h sample was obtained in the early non-survivors 105 (n=51). A group of 17 non-fasting healthy subjects was used as controls for baseline values. The detailed 106 demographic information of these patients is shown in Table 1. Since underlying medical conditions and 107 medication history can influence circulating lipid profiles, we also provide this information (Table S4). 108 Chronic health conditions and medications were rare in the trauma patient population and evenly distributed 109 across the outcome groups (Table S1).

110
The overall data analysis workflow is shown in Fig 1A. Liquid chromatography mass spectrometry 111 (LC-MS) was used to carry out targeted lipidomic analysis on the plasma samples. In total, 996 lipids were 112 quantified using internal standards. In the quality control analysis, the median relative standard deviation 113 (RSD) for the lipid panel was 4%. Lipids are named according to sub-class and acyl chains detected. For  We first explored the dynamic changes in the global pattern of the circulating lipidome in trauma 121 patients. Uniform Manifold Approximation and Projection (UMAP) is a non-linear method for dimension 122 reduction that can identify the global structure of multi-dimensional data. In Fig 1C, each dot represents a 123 single subject and the distance between dots in the UMAP plot reflects the global similarity/ differences in 124 overall lipid profiles between samples 20 . We observed that trauma patients at 0h were quite dispersed and 125 partially overlapping with healthy subjects, suggesting an early and rapidly evolving response pattern 126 immediately post-injury. There was excellent separation across the three time points on UMAP, underscoring 127 the role of time in the major changes in lipid patterns after trauma.

128
To depict the differences between the healthy controls and patients across time, we projected relative 129 levels of all lipids assayed on a heatmap ( Fig 1D). Compared to healthy controls, most lipid species were 130 persistently lower after trauma. This dramatic shift between healthy controls and injured humans was also 131 observed when total lipid concentrations were compared ( Fig 1E). Association between lipidome pattern and outcome of trauma patients 134 We next investigated the association between the circulating lipidome and patient outcomes. The three 135 5 outcomes used for this analysis included (1) early non-survivors (death within 3 days of admission), (2) 136 non-resolving patients (survivors with duration of intensive care unit [ICU] stay 7 days or patients that died 137 after day 3 following admission), and (3) resolving patients (survivors with duration of ICU stay <7 days).

138
UMAP plots of the global lipidomic patterns indicated enrichment of early non-survivors in the region 139 encircled in red at 0h and an enrichment of the non-resolving patients in the region encircled by the blue line 140 at 72h after admission ( Fig 2A&B). Furthermore, we observed a dramatic drop in the levels of nearly all 141 major lipid species at 0h for early non-survivors compared to the other patient groups or healthy controls 142 ( Fig 2C). Patients in both the resolving and non-resolving groups at 0h also exhibited a drop in most lipid 143 species compared to healthy controls, but not to the degree seen in the non-survivors. Patients in the 144 resolving group exhibited a persistent suppression in most lipids at 24 and 72h ( Fig 2D&E). Remarkably, show how these pathways relate to the changes in lipid levels in the non-resolving group. 168 We next examined the impact of injury severity reflected by injury severity scores (ISS) on lipid levels 169 and profiles. Patients were separated into minimal (ISS<10), moderate (ISS 10-25), or severe (ISS  25) 170 injury ( Fig S2A). Exploration of the lipid profiles by either UMAP or heatmap demonstrated no major 171 impact of ISS on the post-injury lipid patterns (Fig S2B). We also observed poor correlation between ISS 172 and total lipids concentrations of either saturated or unsaturated fatty acids (Fig S2C&D, 0h timepoint   173 shown). Thus, while injury induces major changes in the circulating lipidome, in this cohort of patients with 174 shock on presentation, ISS alone does not associate with lipid patterns. The key observation of the PAMPer trial was the demonstration that initiating FFP administration in the 178 pre-hospital setting reduced early mortality when compared to standard care 19 . To assess for an impact of 179 FFP, we compared lipid profiles in patients in the treatment arm to those in the standard-of-care arm. UMAP 180 plots demonstrated a skewing in the lipid profiles towards the healthy controls in the FFP treatment group at 181 0h ( Fig 4A&B). However, the impact of pre-hospital FFP on lipid profiles was seen to dissipate at 24 and 182 72h, with no difference in lipid levels or patterns between the FFP and standard-of-care groups at these time  Fig 4C, Fig S4A). We then assessed the relationship between the predicted mortality, calculated from 186 the Trauma and Injury Severity Score (TRISS), and lipid levels in the two cohorts (Fig. 4D). Average lipid 187 levels were higher in the FFP group across all TRISS values. All unexpected deaths (low TRISS Score: 188 predicted mortality rate less than 50%) were in the standard-of-care patients and 11/14 had lipid levels 189 below the mean for the overall cohort. Deaths seen in the FFP group were limited to those with a high 190 expectation for death for all except one patient (high TRISS Score: predicted mortality rate of greater than 191 75%). A Forest plot of log-odds ratios from multi-variable logistical regression is shown in Fig. 4E. This 192 analysis revealed that lower lipid levels at 0h significantly favored mortality within the first 72h while FFP 193 administration favored survival. Only TRISS had a higher association with early mortality than FFP or lipid 194 levels even when traumatic brain injury (TBI) and sex were added to the model. 195 We next carried out correlation analysis to identify the factors that associate with circulating lipid levels 196 in the early response to severe injury. Included in the analysis were 21 inflammatory and immune mediators, 7 6 markers of endotheliopathy/ tissue injury, and 2 measures of coagulation abnormalities, all measured at 198 time 0h. Also included in the analysis were typical measures of injury severity and interventions associated 199 with adverse outcomes. Interestingly, the mediators segregated into three subsets, each with strong internal 200 correlation ( Fig 4F). These included a subset represented by pro-inflammatory cytokines and chemokines 201 that mostly positively correlated with early death, injury severity, endotheliopathy, and abnormal coagulation 202 (Subset 1: IL-6, IL-8, IL-10, MCP-1/CCL2, IP-10/CXCL10, and MIG/CXCL9) and two subsets that and not with any of the mediator subsets ( Fig 4F). 210 We next used probabilistic graphical models for mixed data types 21,22 to infer potential direct 211 (cause-effect) relationships within the multi-modal observational data included in Figure 4F. These features 212 were loaded into the algorithm and nodes and edges projected onto a graph with early mortality as the  To further generalize our findings of outcome-associated changes in circulating lipids to other trauma 229 datasets and causes of acute critical illness, we analyzed a separate trauma dataset 23 (Trauma dataset-2:TD-2, 230 n=86) and two public datasets derived from COVID-19 patients 16,17 . To assist with the comparison between 231 these trauma and COVID-19 datasets, we set the 0 timepoint in the COVID-19 datasets as the day of 232 symptom onset for non-severe patients or day of progression for severe patients. A total of 29 lipids were 233 identified in common among the 4 datasets ( Fig 5A-D , Table S2). Eight lipids from the PE class  We conducted an in-depth comparison between the two trauma datasets to ensure the reproducibility of 239 our findings. A total 75 lipids from 9 sub-classes were found to be in common between PAMPer and TD-2   Table S3). The eight PE species ranked at ranking at 3, 41, 63, 109, 110, 142, 206, and 294 255 respectively (Volcano plot shown in Fig S6A). In addition, we found that 27 lipids belonging to TAG class 256 of lipids and 7 additional PE lipids were significantly higher in non-resolving patients at 72h (adjusted 257 p<0.01, log foldchange>0.4). This differential analysis also yielded three LPC that were significantly lower. 258 Next, we constructed a matrix that correlated the initial eight PE in the starting pool with these 37 259 differentially expressed lipids (Fig S6B). The starting PE were correlated positively with several other PE 260 and 27 TAG, and negatively correlated with the three lower LPC species. This indicates that the eight PE 261 common to all four datasets may also be representative of an overall reprogramming that includes 262 upregulation of TAG release and a suppression of LPC release into the circulation. We generated a LRS 263 represented as a mean z-score for each patient across all three timepoints and plotted them in a UMAP plot 264 ( Fig S6C) in order to further reveal their relationships with global lipidome patterns. We found that the 265 gradient in the LRS increased from left-to-right along the x-axis in the UMAP plot, which was consistent 266 with the outcome-based pattern at 72h. We then transformed the score into a categorical variable with 267 three thresholds based on tertiles (Low, Medium, High) for all PAMPer patients surviving at 72h (Fig S6D).

268
When displayed on a UMAP plot, the separation of patients into low, medium, and high LRS tertiles  Risk assessment using LRS for patients with trauma or COVID-19 274 We next investigated whether the LRS was associated with outcomes in trauma or COVID-19 patients.

275
Time-series analysis suggested that non-resolving trauma patients experienced dramatic increases of LRS at 276 24 to 72h post-trauma compared to resolving patients ( Fig 6B). Recovery analysis revealed that LRS-high 277 and LRS-medium groups experienced a longer period prior to recovery than patients in the LRS-low group 278 ( Fig 6C). In addition, trauma patients with medium or high LRS were associated with higher injury severity, 279 lower admission blood pressure, mass transfusion, higher INR, and higher incidence of NI and MOF (Table   280 S4). High LRS was also associated with lower probability of recovery (HR:0.75, Cl:0.60-0.94) even when 281 adjusted for age, ISS, TBI, and treatment effect in a Cox regression model ( Fig 6D). To validate our finding 282 using a second trauma population, we adopted the same strategy to construct the LRS using TD-2, which 283 was dominated by resolving trauma patients. The time-series analysis, recovery curve, and Cox regression 284 model all showed similar correlations of LRS with outcomes in TD-2 as seen in PAMPer trial patients (sFig 285 6D, F and G). We then tested whether we could generalize the LRS for the two COVID-19 patient datasets 286 using the same approach. The Shui, et al. 17 COVID-19 dataset lacked detailed clinical data; therefore, we 287 only compared differences in LRS among the four outcome groups defined by the authors of the study. We 288 found that moderate and severe COVID-19 patients had a higher LRS compared to healthy subjects (Fig   289   S6E). Consistent with these findings, the LRS was also significantly higher in the severe group when 290 compared to the non-severe COVID-19 patients in the dataset of Guo, et al 16 (Fig 6E). We also observed an 291 upward trend in LRS during the time window preceding progression (< 48h after progression, Fig 6E).

292
C-reactive protein (CRP) and lymphocyte count are known to correlate with worse outcome in COVID-19 293 patients 24 . We compared LRS with these two variables to classify severe versus non-severe patients. The

294
AUC score for LRS, lymphocyte count, and CRP was 0.788, 0.817, and 0.822, respectively ( Fig 6F). Finally, 295 multi-variable logistical regression suggested that LRS is an independent risk factor for COVID-19 patients 296 (Log2 OR: 1.54, Fig 6G). Thus, a score based on the levels of a subset of circulating lipids associates with 297 features in trauma and Covid-19 patients that predict a complicated clinical course.

315
Using 42 subjects with both metabolomics and proteomics data, we identified 150 proteins that correlated 316 positively (spearman correlation coefficient r > 0.3) with the LRS (Fig S7C). Pathway enrichment analysis 317 revealed that the LRS was associated with neutrophil degranulation, platelet degranulation, and the 318 complement cascade (Fig S7C and Fig S7E). Negatively correlated (spearman correlation coefficient r < 319 -0.3) proteins (n=24) were enriched in regulation of insulin-like growth factor-1 (IGF-1) transport and 320 uptake, and post-translational protein phosphorylation (Fig S7D and Fig S7F). To further seek biological 321 significance, we selected 40 representative proteins from the positive and negative correlating groups to 322 construct a correlation matrix (Fig 7B). Components of the LRS were clustered in the module comprised 323 acute phase proteins, the complement cascade, and immunoglobins and were correlated negatively with 324 modules associated with IGF-1. Our findings using data from COVID-19 patients suggests that excessive 325 acute phase and immune responses and impaired metabolism associates with a pathologic circulating lipid 326 signature across several causes of acute critical illness. The main goal of this study was to correlate the temporal patterns in the circulating lipidome with  Among the most pronounced changes observed in our study was the early loss of all classes of lipids in 346 the circulation after injury. A study of 32 trauma patients showed that blood triglyceride levels were 347 significantly lower in 9 non-survivors within 28 minutes of injury, suggesting that injury-induced decreases 348 in circulating lipids may begin very early after a severe trauma 27 . Our healthy controls were non-fasting and 349 sampled throughout the day to align with the presentation of the typical trauma patient. Therefore, the 350 differences between controls and injured at time 0h are unlikely to be due to dietary effects. While the 351 degree of the decline in lipids associated with clinical outcomes, the incidence was not dependent on injury 352 severity. A stress hormone-induced hypermetabolic state with associated increased catabolism is seen after 353 trauma and other causes of critical illness 6,28 and may explain the persistent decline in circulating lipids. The 354 catabolism response generates energy substrates from carbohydrates, fats, and protein in an "all or none" 355 manner that, like our findings, is not influenced by injury severity 29 . It is reasonable to speculate that the 356 abrupt loss of lipids may be due, in part, to the uptake and catabolism of lipids to meet the energy demands.

357
The finding that patients that die within first 72h experience the greatest magnitude in lipid loss from the 358 circulation raises the interesting possibility that a circulating energy substrate crisis contributes to the early 359 mortality.  We observed that PC (16:0/18:1) and PC (18:0/18:1) were higher at 72h in the non-resolving trauma patients 379 or severe Covid-19 patients, raising the possibility for a lipid reprogramming process across organs during 380 persistent critical illness. 381 We derived a LRS that reflects the magnitude of lipid reprogramming associated with delayed adverse 382 outcomes. We found that higher LRS at 72h is an independent risk factor for recovery. Higher LRS was also 383 13 observed in the sickest COVID-19 patents and even preceded the onset of critical illness. This indicates that 384 lipid reprogramming involving higher levels of a subset of PE in the circulation is a feature common to

405
In conclusion, our findings provide a new paradigm for the lipid response to a severe and acute 406 systemic stress leading to critical illness (summarized in Figure 7C). Our causal modeling and correlation 407 analyses place lipolysis a central regulator of the evolution from acute disease onset to critical illness in 408 humans. The features of lipogenesis we identified appear to be common to critical illness due to multiple    Resolving (Survival with ICU stay < 7 days); Non-resolving (Survival with ICU stay >= 7 days or 445 non-survival with death day >3 days) and Early-nonsurvivors (Non-survival with death day <=3 days).

446
Blood samples were collected using vacuum isolation tubes with anticoagulant of Heparin sodium, which 447 were centrifuged at 4C and plasma fractions were stored at -80C for further analysis.

448
This study was approved by the IRB of University of Pittsburgh as previously described 19   linoleic acid (C18:2)). Lipid with over 80% missing quantitative values were discarded due to the concern of 472 low quality. Other missing values for each lipid species were imputed with the minimum concentration.