Plasma Proteomic Signatures of Coagulopathy Following Traumatic Injury: Links to Biomechanics and Patient Outcomes

Abstract

Background:

Traumatic injury is a leading cause of death, in large part driven by coagulopathy-associated hemorrhage and thrombotic complications. Identification of biological pathways and therapeutic targets has been limited. We performed a proteomic analysis using plasma isolated from trauma patients and assessed the proteome relation to clot formation and stability.

Methods:

Platelet-poor plasma was isolated from trauma patients upon emergency department arrival. Coagulation was characterized using rheology, turbidity, and confocal microscopy. Proteomics were measured using liquid chromatography-mass spectrometry and compared between: 1) healthy donors vs. trauma patients, 2) nonsevere vs. severe injury, 3) survived vs. deceased patients, 4) penetrating vs. blunt injuries, 5) traumatic brain injury (TBI) vs. no-TBI, and 6) presence vs. absence of physiological shock. Spearman correlations were calculated between coagulation tests, clinical vitals, and proteomics data. Gene ontology and Search Tool for the Retrieval of Interacting Genes/Proteins analysis identified enrichment of biological processes and interconnectedness of mortality-related proteins.

Results:

Hemostatic processes were enriched in healthy donors, severe injury patients, surviving patients, no-TBI patients, and patients not exhibiting shock. Plasminogen activator inhibitor 3 was in higher abundance in those who survived their injuries compared with those who died. Plasminogen activator inhibitor 3 was correlated with faster clotting time, thrombin generation, and decreased D-dimer. Proteins higher in deceased patients compared with those who survived, such as prostaglandin-H2 isomerase, related to immune activity, correlated with weaker clots and increased D-dimer.

Conclusions:

This exploratory analysis of plasma proteomics in trauma patients identified potential markers related to coagulation and immune activity, which may contribute to coagulopathy-associated mortality after injury and serve as therapeutic targets.

INTRODUCTION

Severe traumatic injury remains one of the leading causes of death and disability worldwide, predominantly in young adults between the ages of 15 and 45, impacting the most productive years of life (1,2). Globally, trauma contributes to over five million deaths annually, with the primary causes of death being traumatic brain injury (TBI) and hemorrhage, the latter being the most preventable cause of death (1,3–5). Hemorrhage-associated death is exacerbated by complex pathophysiological changes in coagulation, inflammation, and endothelial function (6,7).

Following traumatic injury, the body undergoes a cascade of molecular events, including the systemic release of cytokines, enzymes, and damage-associated molecular patterns into the bloodstream (8). This cascade of molecular derangements, combined with blood loss, inflammatory factors, and coagulation dysregulation, all contribute to the progression of impaired hemostasis. This is characterized by altered blood clot formation, increased bleeding, and thrombotic complications. The clinical consequences may be life-threatening because patients may simultaneously have an increased bleeding risk but also a predisposition for pathological clotting, which can contribute to multiorgan dysfunction and transfusion requirements (6,7,9,10). Though traditional clinical markers, such as platelet count, injury severity score (ISS), partial thromboplastin time (PTT), and viscoelastic assays, offer some insights to guide clinical decisions and therapy, they fail to capture the complexity and progression of the molecular changes that underlie the pathophysiology of coagulation after traumatic injury, thereby limiting the effectiveness of current therapies (11–13). To better understand the molecular changes that occur after trauma, high-throughput mass spectrometry-based proteomic technologies can provide a wealth of information to characterize the systemic response to trauma at a molecular level (11–13). Simultaneous measurement of the relative abundance of thousands of proteins in the blood can shed light on previously unrecognized actors and novel pathway intersections that may inform new intervention strategies to combat the trauma-induced coagulopathy cascade (11–13).

Recent studies have demonstrated that trauma patients exhibit proteomic signatures associated with ISS and clinical outcomes (11,13–15). Metabolic analysis has revealed distinct profiles that are associated with response to clinical interventions in patients suffering blunt traumatic injuries (16). For example, distinct inflammatory signatures have been associated with clinically relevant outcomes such as the development of early multi-system organ dysfunction (17). In addition, platelet dysfunction has also been linked to altered plasma proteomic profiles, which implicates specific proteins in impaired platelet-mediated clot formation after trauma (14,15). However, despite these advances, there are still substantial gaps in our understanding of which proteins are affected following traumatic injury, and whether identified proteins can be used as prognostic markers to guide treatment. Most existing studies have focused on either isolated clinical endpoints or very specific protein subtypes, limiting our ability to draw broader conclusions about systemic responses to trauma (15,18). Proteomic profiling has the potential to differentiate trauma phenotypes with divergent inflammatory and coagulation profiles, which can shed light on heterogeneous patient responses that may benefit from precise intervention (11,13).

In this study, we aim to characterize circulating proteomic profiles in samples obtained from trauma patients to elucidate the association of proteomic profiles with clotting kinetics, mechanics, and clinical data. We hypothesize that proteomic profiles in trauma patients will correlate with clinical and biochemical properties of coagulation, thus advancing our understanding of the complex physiologic changes in the early minutes of trauma and paving the way to improved diagnostic tools and individualized therapeutic interventions.

MATERIALS AND METHODS

Collection and preparation of plasma samples

Following institutional review board approval (IRB) (Pro2021001296), trauma patients at the Robert Wood Johnson University Hospital Level 1 trauma center who were designated as the highest level of trauma response and who met strict inclusion and exclusion criteria (Supplementary Table S1, https://links.lww.com/SHK/C705), had whole blood collected (n = 61) upon presentation between January 2021 and June 2023. Whole blood was drawn (5 mL) at a single time point into 3.2% sodium citrate (9:1, vol/vol) tubes immediately after arrival to the resuscitation area. In accordance with exclusion criteria, blood was drawn only from patients who had not received a transfusion or procedural intervention before hospital admission to reduce the influence of these confounding factors on sample proteomics. Blood samples were stored at 4°C for no more than 12 hours following blood draw before plasma isolation. Platelet-poor plasma (PPP) was isolated by centrifugation at 2,400 rpm for 10 minutes and then immediately stored at −80°C until transferred from the hospital to the laboratory on dry ice and placed back in a −80°C freezer until use for testing as described (19). After blood was collected and processed for storage, informed consent from the patient or surrogate was obtained. Samples were disposed of if any exclusion criteria were met, or approval from the patient or surrogate was not obtained in accordance with Rutgers IRB approval (Pro2021001296). All administration of blood products after a blood draw within the first 24 hours was recorded. Healthy donor samples were collected following informed consent and IRB approval (Pro2020001694) (n = 7). Whole blood was drawn into 3.2% sodium citrate anticoagulant tubes (Grenier Bio, Kremsmunster, Austria). Within 2 hours of blood draw, samples had PPP isolated by centrifugation at 2,400 rpm for 10 minutes. PPP was then frozen at −80°C until thawed for testing. All patient and healthy donor samples were thawed only once and used for testing within the day after thawing for all of the following methods. No samples were exposed to multiple freeze-thaw cycles. Patient sample collection and donor inclusion were conducted in accordance with the principles of the Declaration of Helsinki.

Trauma patient and healthy volunteer rheometer mechanical testing

Healthy volunteer and trauma patient samples were tested on a Malvern Kinexus Ultra rheometer (Netzsch, Selb, Germany) to analyze the rheological properties of PPP. Each sample was activated with 20 mM CaCl₂ and 0.6 units/mL thrombin (final concentration). Samples were tested on a 20 mm parallel plate oscillation strain test as previously described (19).

Trauma patient and healthy volunteer optical turbidity testing

Healthy volunteer and trauma patient samples were tested on a Molecular Devices SpectraMax Plus plate reader (Molecular Devices, San Jose, California) to measure the opacity of PPP as the plasma clot formed and lysed. Samples were activated similarly to Kinexus rheometer samples and tested as previously described (20).

Trauma patient and healthy volunteer confocal microscopy

PPP from healthy donors and trauma patients was imaged using a Zeiss LSM 710 confocal microscope (Zeiss, Oberkochen, Germany). Samples were prepared in an Eppendorf tube, with 1% of plasma volume replaced with 1.5 mg/mL Alexa Fluor 488-labeled fibrinogen (Thermo Fisher Scientific, Waltham, Massachusetts). Samples were activated similarly to Kinexus activation, and 100 μL of the samples were added to a clear-bottom borosilicate 96-well glass plate. Samples were imaged with a 40× water immersion objective, using a 1024 × 1024 resolution, a 1.2 numerical aperture, with a final image area of 213 × 213 μm. Z-stack images were obtained from three areas for each sample at an interval of 1 mm across a 10 mm depth, with maximum intensity projections produced from these images. Maximum intensity projection images from the Zeiss Zen software were analyzed as.czi files, using Java ImageJ software version 1.8.0_322. Fibrin fiber density was calculated by binarizing images, and total fiber area was measured as the percent of image area. Fibrin fiber length and pore diameter were measured manually using ImageJ and calculated from the mean of 36 fibers and 16 pores for each sample, respectively.

Plasma D-dimer concentration quantification

Plasma D-dimer concentration was measured using an Abcam human D-dimer Enzyme Linked Immunosorbent Assay kit (ab260076, Abcam, Cambridge, UK), following the manufacturer’s instructions on a Molecular Devices Spectramax Plus (Molecular Devices) plate reader. Samples were diluted using the provided diluent buffer, with the range of 1 part plasma and 10 parts diluent, to 1 part plasma and 200 parts diluent, to ensure results fell within the provided standard curve. Each sample was run in duplicate and measured at 450 nm absorbance.

Plasma thrombin generation assessment

Plasma thrombin generation was measured in patient and healthy donor plasma using a Technothrombin Thrombin Generation Assay (TGA) (Diapharma, Ohio) using the manufacturer’s instructions. A calibration curve was quantified prior to plasma samples. For patient and donor samples, 40 μL of sample was mixed with 10 μL TGA reagent concentrate low trigger and 50 μL TGA substrate. Fluorescence was measured at 360 nm excitation and 460 nm emission wavelengths using a Molecular Devices SpectraMax M2 (Molecular Devices) plate reader at 1-minute intervals for 90 minutes.

Plasma fibrinogen concentration quantification

Trauma patient and healthy donor plasma fibrinogen concentration was measured using the Dade Siemens Healthineers Fibrinogen Determination Kit (Siemens, Erlangen, Germany) on a STAGO Start 4 coagulation analyzer (Diagnostica STAGO, Parsippany, New Jersey). Frozen plasma samples were thawed to 37°C, then diluted one part plasma to nine parts Owrens veronal buffer, incubated at 37°C for 2 minutes, and mixed with 0.1 mL Dade Thrombin Reagent while measuring clotting time in duplicate. Fibrinogen concentration in samples was interpolated from the standard curve provided.

Plasma proteomics quantification

All plasma samples were directly digested with trypsin using the SP3 method (21). Patient and healthy donor plasma samples were analyzed by liquid chromatography-mass spectrometry (LC-MS/MS) using a nano LC-MS/MS (Dionex Ultimate 3000 RLSCnano system, Thermofisher, Waltham, Massachusetts) interfaced with an Eclipse (Thermofisher). Plasma samples were loaded onto a fused silica trap column Acclaim PepMap 100 (Thermofisher). Following a 5-minute wash at 5 μL/min with 0.1% trifluoroacetic acid, the trap column was brought in-line with an analytical column (Nanoease MZ peptide BEH C18, 130A, 17 μm, Waters) for LC-MS. Peptides were fractionated at a flow rate of 300 nL/min using a segmented linear gradient: 4% - 15% B over 30 minutes (where A: 0.2% formic acid and B: 0.16% formic acid in 80% acetonitrile), 15% - 25% B over 40 minutes, 25% - 50% B over 44 minutes, and 50% - 90% B over 11 minutes. Solution B was then returned to 4% for 5 minutes before the next run. Eluted peptides were analyzed using a data-independent acquisition workflow. The MS scan range was set to 400–1200 m/z, with a resolution of 12,000, an automatic gain control target of 3E6, and an automatic ion injection time. Also, 8 m/z windows were used to sequentially isolate ions for 400–800 m/z, and 20 m/z windows were used to isolate ions from 800 to 1200 m/z (automatic gain control target 4E5, automatic ion injection time). The ions were fragmented in the C-trap with a relative collision energy of 30. MS/MS spectra were recorded at a resolution of 30,000 in the Orbitrap.

Raw data were analyzed using a predicted library from the UniProt human proteome FASTA file for a library-free search with data-independent acquisition-NN 1.8.1 (https://github.com/vdemichev/DiaNN) under recommended settings. Protease was set as trypsin (C-terminus of K or R unless followed by P). Miss cut of 1 was allowed. Peptide length range from 7 to 30, precursor charge range of 1–4, precursor m/z range 300–1800, and fragment ion m/z range 200–18—were used to generate the predicted library. Deep learning-based spectra and retention times, and ion mobility prediction were used for spectra matching evaluation. The discovery spectra generation method was default, as described in Demichev et al, 2020 (22). The results were filtered with a posterior error probability filter of <0.01 and a protein group Q value (PG.Q) filter of <0.01. Protein group MaxLFQ values were used for quantification (22).

Experimental design and statistical rationale

We conducted data preprocessing and statistical analysis using Python (via Google Colab) and Microsoft Excel. For correlation analysis, all missing values in the dataset were imputed using Multiple Imputation by Chained Equations (MICE), a robust and reliable method of handling missing data in high-dimensional omics datasets, widely adopted in contemporary omics studies (23–25). MICE preserves the statistical relationships among variables by iteratively modeling each feature based on the others, thereby maintaining variance and covariate structure that might otherwise be lost through listwise deletion. The regression model used within the MICE framework was random forest, as this was considered the most suitable model to use with proteomics data, yielding the most accurate detection of relevant pathways and signature genes (24). Due to the high dimensionality of the dataset, an established batch process was implemented in which imputation occurred in batches of 500 proteins until all the proteins were covered (25). This approach allowed us to retain the full dataset for downstream analyses while reducing potential bias associated with missing data.

Proteins detected in fewer than six individuals were excluded to improve the reliability of statistical inference. For each comparison, protein-wise log₂ fold change (FC) was calculated by taking the logarithm of the ratio between the group means. A threshold of log₂(FC) of 0.26 (equivalent to a 1.2-FC) was applied to identify differentially abundant proteins. Published data balancing sensitivity and specificity in detecting biologically meaningful yet subtle changes informed threshold selection (26–28). Statistical significance was assessed using a two-sample Welch’s t test, due to the assumption of unequal variances, for each protein. To evaluate the relationships between protein expression levels across the patient samples, we performed a Spearman rank-order correlation analysis in Python using Google Colab. The dataset included the 61 trauma patient samples. Preprocessing included handling missing data, which was imputed using MICE. The Spearman correlation coefficient, as well as the associated two-tailed P values were calculated using the SciPy library in Python, and the resulting matrix of correlations and P values was formatted into an Excel spreadsheet for interpretation. The software created for this manuscript is held in a Zenodo repository with DOI 10.5281/zenodo.16741038.

Gene ontology analysis

Proteins that were assessed to be significantly different between compared patient groups, were assessed with gene ontology (GO), with these categories including: Healthy donors, Trauma patients, nonsevere ISS (ISS ≤15), severe ISS (ISS >15), survived, deceased, blunt force injury, penetrating injury, no-TBI, TBI, nonshock (Shock index <0.9), and shock (Shock Index >0.9) at admission. GO is utilized to analyze proteomics data and show which biological processes are overrepresented from a list of proteins (29). The proteins from the categories listed above were loaded into GO for biological process analysis. For all GO analyses, a Fisher’s exact test was used along with calculating the false discovery rate. The negative logarithm of the raw P value was taken for each biological process detected, and the base 2 logarithm was taken for the fold enrichment value for each biological process. The percentage of proteins that were associated with a particular biological process (list coverage percentage) was calculated by dividing the total number of proteins detected with a respective biological process from the data by the total number of proteins that were identified in Gene Ontology’s reference human genome database. The percentage of proteins that were present in the whole reference human genome database (process coverage percentage) was calculated by dividing the total number of proteins detected with a particular biological process from the uploaded data by the total number of proteins that are related to that specific biological process in the reference human genome database.

RESULTS

Patient demographics

Plasma proteomics were analyzed in a total of 61 injured trauma patients. The included patients were predominantly male, had a median ISS of 17, median age of 38, and 74% suffered trauma from a blunt mechanism. Along with this, 48% of patients sustained TBI, and the in-hospital mortality rate of the cohort was 29.5%. Full characteristics of the cohort are shown in Table 1. Comparison of group characteristics, age, ISS, TBI prevalence, and Glasgow Coma Score (GCS) shows limited differences among confounding factors between compared groups (Supplementary Figure S1, https://links.lww.com/SHK/C706). Healthy donors (n = 7) had a median age of 25 [interquartile range (IQR) 24–26.5], 42% were female, and 58% were male. Of these donors, 71% were White, 29% were Asian, 29% were Hispanic/Latino, and 71% were non-Hispanic/Latino.

Table 1.

Patient demographics and injury characteristics

Patient cohort	Patient data (n = 61)
Time from injury to blood draw (Median minutes) (25–75% IQR)	47 (34–60.5)
Age (median years) (25–75% IQR)	38 (27–58)
Sex
Male (%)	46 (75%)
Female (%)	15 (25%)
Race
White	41 (67%)
Black	15 (25%)
Asian	5 (8%)
Ethnicity
Hispanic or Latino	22 (36%)
Non-Hispanic of Latino	39 (64%)
Injury mechanism
Blunt	45 (74%)
Motor vehicle collision	24 (39%)
Fall	11 (18%)
Pedestrian struck	10 (16%)
Penetrating	26 (26%)
Gunshot wound	7 (11%)
Stab	9 (15%)
Injury pattern
Spinal cord injury	5 (8%)
Traumatic brain injury	29 (48%)
Shock (Shock Index > 0.9)	18 (29.5%)
Mortality (%)	18 (29.5%)
Injury Severity Score
Median ISS (25–75% IQR)	17 (10–27)
</= 15 ISS	23 (38%)
> 15 ISS	38 (62%)
Glasgow Coma Score (Median GCS) (25–75% IQR)	10 (3–15)
Time to Mortality (Median Hours) (25–75% IQR)	28 (8.5–56.7)

Bold values indicate the number of patients (N) in each category.

IQR, interquartile range; ISS, injury severity score; GCS, Glasgow Coma Score.

Impact of traumatic injury severity and trauma phenotype on plasma proteomics

Compared with healthy donors, trauma patients (n = 61) were older (P < 0.05), with no difference in biological sex distribution (Supplementary Figure S1, https://links.lww.com/SHK/C706). Trauma patients exhibited a significantly higher abundance of proteins related to endothelial injury [Von Willebrand factor (VWF)], extracellular matrix remodeling [tissue inhibitor of metalloproteinase 1 (TIMP-1)], and macrophage mannose receptor 1 (MRC1) compared with healthy donors (n = 7) (Fig. 1A) (Table 2 and Supplementary Table S2, https://links.lww.com/SHK/C707). Trauma patients had a significantly lower abundance of proteins related to coagulation and the fibrin fiber network, relative to healthy donors (n = 7). These included fibrinogen alpha (FGA), beta (FGB), and gamma (FGG) chains, coagulation factor XIII alpha chain (F13A), and fibrinolytic regulators, including alpha-2-antiplasmin and protein MENT.

Fig. 1.

Influence of trauma type and patient outcome on plasma proteomics. Volcano plot X-axis shows the log₂(FC) of the average of trauma patient protein level compared with the average healthy donor protein level, and the Y-axis shows the −log₁₀(nonadjusted P value) of the significant difference between the two groups. Horizontal dotted lines indicate the significance cutoff (P < 0.05: −log₁₀(0.05) = 1.31), while vertical lines indicate a 1.2-fold change, converted to 0.26 with FC conversion, in both directions between healthy donors and trauma patients. Volcano plots show 588 proteins between healthy donors and trauma patients (A), identified with LC-MS/MS, 933 proteins between nonsevere and severe ISS injury patients (B), 1046 proteins between survived and deceased patients (C), 854 proteins between blunt and penetrating trauma patients (D), 1090 proteins between patients without TBI (no-TBI) and with TBI (E), and 1059 proteins between patients with shock and without shock (nonshock) (F). Proteins analyzed in A–F were detected in at least six individuals per group. Protein abbreviations are included in Supplementary Table S3, https://links.lww.com/SHK/C708. FC, fold change; ISS, injury severity score; LC-MS/MS, liquid chromatography-tandem mass spectrometry; TBI, traumatic brain injury.

Table 2.

Relative protein abundance in patient plasma

Comparison (Group 1 vs Group 2)	Protein	Log2 (fold change)	P value
Healthy vs. trauma	MRC1	0.383	0.010
	PTGDS	0.406	0.007
	SERPINA5/PAI-3	0.304	0.003
Low ISS vs. high ISS	MRC1	0.253	0.006
	PTGDS	0.703	0.0003
	SERPINA5/PAI-3	−0.194	0.029
Survived vs. deceased	MRC1	0.310	0.008
	PTGDS	0.804	0.001
	SERPINA5/PAI-3	−0.222	0.034
Penetrating vs. blunt	MRC1	0.146	0.158
	PTGDS	0.280	0.280
	SERPINA5/PAI-3	−0.095	0.331
No-TBI vs. TBI	MRC1	0.126	0.194
	PTGDS	0.558	0.011
	SERPINA5/PAI-3	−0.158	0.091
Nonshock vs. shock	MRC1	−0.022	0.840
	PTGDS	0.256	0.328
	SERPINA5/PAI-3	0.045	0.664

Fold Change is the mean of Group 2 divided by the mean of Group 2 mentioned in the comparison column (e.g., Group 2/Group 1 = trauma/healthy). A positive fold change indicates a higher average relative abundance in Group 2 than in Group 1 and vice versa. Bold values indicate statistically significant Log₂ (fold change) values (P < 0.05).

ISS, injury severity score; MRC1, macrophage mannose receptor 1; PTGDS, prostaglandin-H2 D-isomerase; SERPINA5/PAI-3, plasma serine protease inhibitor (plasminogen activator inhibitor 3).

Severely injured patients (ISS >15) (n = 38) had similar mean age, shock index, and ratio of male to female patients. Severe ISS patients had a lower mean GCS and increased prevalence of TBI compared with nonsevere ISS patients (Supplementary Figure S1, https://links.lww.com/SHK/C706). Severe ISS patients exhibited a higher abundance of platelet-related proteins, such as platelet basic protein, platelet factor 4 (PF4), and platelet glycoprotein V (GP5) compared with lower severity trauma patients (ISS ≤15) (n = 23), while no differences in coagulation factors were detected (Fig. 1B) (Supplementary Table S2, https://links.lww.com/SHK/C707). Patients who died from their injuries (n = 18) had a similar mean age, shock index, and biological sex distribution, with decreased mean GCS, increased ISS and TBI prevalence compared with surviving patients (n = 43) (q < 0.001, P < 0.001) (Supplementary Figure S1, https://links.lww.com/SHK/C706). Deceased patients exhibited a lower abundance of coagulation factor XII, both the regular chain (F12) and the short isoform (F12-Short), vitamin K-dependent protein C (PROC), thrombin-activatable fibrinolysis inhibitor, and FGA than surviving patients. Deceased individuals also exhibited higher abundance of inflammatory and immune markers, such as MRC1, metalloproteinase inhibitor 1 (TIMP-1), and prostaglandin-H2 isomerase (PTGDS) (Fig. 1C) (Supplementary Table S2, https://links.lww.com/SHK/C707). Penetrating injury patients (n = 16) had a lower mean age, ISS, and prevalence of TBI than blunt force injury patients (n = 45) (q < 0.01, P < 0.01). Penetrating injury patients had an increased mean GCS (q < 0.01), similar biological sex proportion and shock index when compared with blunt injury patients (Supplementary Figure S1, https://links.lww.com/SHK/C706). Blunt injury patients had a higher abundance of TFPI, MENT, and SERPINA11 compared with patients with penetrating trauma (n = 16) (Fig. 1D) (Supplementary Table S2, https://links.lww.com/SHK/C707). No-TBI patients (n = 32) and TBI (n = 29) patients had similar mean age, shock index, and biological sex ratio, while No-TBI patients had a higher mean GCS (q < 0.0001), and lower ISS and TBI prevalence (P < 0.001) (Supplementary Figure S1, https://links.lww.com/SHK/C706). TBI patients exhibited a higher abundance of PTGDS, complement C5, complement component C7, and serum amyloid A1 (SAA1) in comparison to no-TBI patients. Conversely, no-TBI patients exhibited higher abundance of VWF, coagulation factor VIII (F8), and platelet glycoprotein 1b alpha and Beta chains compared with TBI patients (Fig. 1E) (Supplementary Table S2, https://links.lww.com/SHK/C707). Lastly, when comparing patients presenting with Shock (n = 18) to nonshock patients (n = 43), defined by heart rate/SBP >0.9, these groups had similar mean age, GSC, ISS, sex ratio, and TBI prevalence (Supplementary Figure S1, https://links.lww.com/SHK/C706). Nonshock patients exhibited higher abundance of factor XII short chain, complement 7, and factor XII. Compared with nonshock patients, patients in Shock had higher measured abundances of serpin B6, platelet factor 4, and platelet glycoprotein V (Fig. 1F) (Supplementary Table S2, https://links.lww.com/SHK/C707).

Injury phenotype dictates enrichment of diverse biological processes

Large variations were observed between healthy donors, trauma patients, and within patient trauma phenotypes. Protein counts differ due to the group-specific inclusion threshold, where proteins detected in at least six individuals per subgroup were included. Of these, the abundance of 30 proteins was significantly lower in trauma patients than in healthy donors, while 106 were higher in trauma patients compared with healthy donors (Supplementary Table S4, https://links.lww.com/SHK/C709). In addition, 16 protein levels were higher in nonsevere ISS compared with severe ISS, while 75 protein levels were higher in the severe ISS compared with the nonsevere ISS groups. Comparing further patient groups identified higher protein levels between survived and deceased patients, 21 vs. 26 proteins were higher in each group compared with the other group, respectively. Differences were observed between blunt and penetrating, 16 vs. 69 proteins, respectively, between no-TBI and TBI, 22 vs. 23 proteins, and between nonshock and shock, 39 vs. 31 proteins, respectively (Supplementary Table S2, https://links.lww.com/SHK/C707 and Supplementary Table S4, https://links.lww.com/SHK/C709). These relative differences between groups are indicative of either an increase in proteins in one group, either through increased production and release into the blood plasma, or by a reduction in proteins in the compared group, potentially through consumption through use, and degradation leading to a lower concentration in the compared group.

Following the determination of which protein levels were significantly higher in each binary comparison of subject groups, they were categorized to assess how these changes affected biological processes with GO. Proteins were grouped into the following: higher abundance in healthy, in trauma, in nonsevere ISS, in severe ISS, in survived, in deceased, in blunt, in penetrating, in no-TBI, in TBI, in low shock Index, and in high shock Index (Supplementary Table S4, https://links.lww.com/SHK/C709). Significant differences were observed based on presence of traumatic injury (Fig. 2A) (Supplementary Table S5, https://links.lww.com/SHK/C710 and Supplementary Table S6, https://links.lww.com/SHK/C711), severity of injury (Fig. 2B) (Supplementary Table S7, https://links.lww.com/SHK/C712), patient mortality (Fig. 2C) (Supplementary Table S8, https://links.lww.com/SHK/C713 and Supplementary Table S9, https://links.lww.com/SHK/C714), injury mechanism (Fig. 2D) (Supplementary Table S10, https://links.lww.com/SHK/C715 and Supplementary Table S11, https://links.lww.com/SHK/C716), presence of TBI (Fig. 2E) (Supplementary Table S12, https://links.lww.com/SHK/C717), and shock index (Fig. 2F) (Supplementary Table S13, https://links.lww.com/SHK/C718).

Fig. 2.

Biological process enrichment between subject injury phenotype. GO enrichment plots display the Log₂(fold enrichment) of biological processes on the x axis, identified by the proteins present in a list of significantly higher proteins in each subject comparison pair. The size of the bubble indicates the number of proteins significantly higher in each category of patient and donor type, which are related to each GO process. The y axis indicates the −Log₁₀(P value), indicating the significance of the fold enrichment. Processes that are enriched in healthy donors compared with trauma patients (A), nonsevere ISS compared with severe ISS (B), survived patients compared with deceased patients (C), blunt injury patients compared with penetrating injury patients (D), TBI patients compared with no-TBI patients (E), and lastly patients with high shock index in comparison to patients with low shock index (F). Biological processes detected were categorized into broader biological themes and defined as their occurrence in each patient comparison pair (G) and their normalized occurrence in the total altered biological processes (H). Vertical dashed lines indicate a 10%, 25%, 50%, and 100% fold enrichment in the analyzed protein data set, corresponding to 3.32, 4.64, 5.64, and 6.64 Log₂(fold enrichment). Indicating an increasing overrepresentation of certain biological processes in the protein data set. GO, gene ontology; ISS, injury severity score; TBI, traumatic brain injury.

Biological processes enrichment was plotted with the -Log₁₀ (P value) on the y axis, and the Log₂ (fold enrichment) on the x axis, respectively. The size of the bubble is representative of the list coverage percentage of the proteins in the associated GO biological process. Of the 299 biological processes enriched in either healthy donors or trauma patients in comparison to each other, hemostasis, coagulation, and fibrinolytic processes were primarily enriched in healthy donors [blood coagulation-GO:0007596, plasminogen (PLG) activation-GO:0031639] (Supplementary Table S5, https://links.lww.com/SHK/C710), while metabolic processes immune processes and oxygen free radical response were enriched in trauma patients (Fig. 2, A,G,H) (Supplementary Table S6, https://links.lww.com/SHK/C711). GO did not identify any biological processes being enriched in nonsevere ISS patients (Supplementary Table S7, https://links.lww.com/SHK/C712). While in the severe ISS group, 203 processes were enriched, with only five of these (2.4%) related to hemostasis and coagulation. The majority of enriched processes in those with severe ISS were involved with apoptotic processes, cell motility, and metabolism (Fig. 2, B,G,H) (Supplementary Table S7, https://links.lww.com/SHK/C712). A total of 44 processes were enriched between either survived or deceased patients, with hemostasis and fibrinolytic processes enriched in survived patients (factor XII activation-GO:0002542) (Fig. 2, C,G,H) (Supplementary Table S8, https://links.lww.com/SHK/C713), while immune response was enriched in Deceased patients (Immune effector process-GO:0002252) (Fig. 2, C,G,H) (Supplementary Table S9, https://links.lww.com/SHK/C714).

Comparison of injury type identified unique processes enriched in either blunt or penetrating trauma, with 89 processes enriched between both groups. Cell motility and metabolism were primarily enriched in blunt injuries (Fig. 2, D,G,H) (Supplementary Table S10, https://links.lww.com/SHK/C715), while immune processes predominated in penetrating injury patients (Fig. 2, D,G,H) (Supplementary Table S11, https://links.lww.com/SHK/C716). Between no-TBI and TBI patients, 23 processes were enriched (Supplementary Table S12, https://links.lww.com/SHK/C717), with no statistically significant enriched processes observed in the TBI patients (Supplementary Table S12, https://links.lww.com/SHK/C717). Hemostasis (blood coagulation intrinsic pathway-GO:0007597), telomere maintenance cajal body interactions, and intrinsic coagulation processes (blood coagulation intrinsic pathway-GO:0007597) in particular were enriched in no-TBI patients compared with patients with TBI (Fig. 2, E,G,H). In patients with a low shock Index (SI  < 0.9) (nonshock), processes related to coagulation (regulation of blood coagulation-GO:0030193), and wound healing were enriched compared with patients with a high shock index (SI >0.9) (shock). Despite 31 proteins being at higher concentrations in the shock subgroup, only one biological process was significantly upregulated (Fig. 2F) (Supplementary Table S13, https://links.lww.com/SHK/C718).

Enrichment of diverse biological processes in each patient group may indicate both the alteration of the processes in a particular group, or potentially the inhibition or shutdown of the same process in the comparison group, leading to an observed enrichment.

Proteomic profiles exhibit dynamic relationships with traumatic injury and coagulation function

We next identified the proteins that were significantly different between healthy donors and trauma patients, and characterized their relationship with the top predictors of mortality established in a previous study (19). The proteins that were higher in the healthy donor group and trauma group were correlated to activated PTT (aPTT) (Fig. 3A), prothrombin time (PT) (Fig. 3B), ISS (Fig. 3C), rheological clot stiffness (Pa) (Fig. 3D), rheological clotting rate (Pa/s) (Fig. 3E), fibrinogen concentration (Fig. 3F), D-dimer concentration (Fig. 3G), and rate of thrombin generation (velocity index) (Fig. 3H). Proteins were categorized into four groups based on whether they belonged to the healthy group or trauma group, and proteins in each group related to coagulation were highlighted. A large number of proteins that were significantly higher in trauma compared with healthy, had significantly positive correlations to parameters that are part of the admission laboratory panel, such as PTT, PT, and D-dimer, as well as to ISS. These findings suggest a relationship among the presence of traumatic injury, increased trauma severity leading to an increase in the levels of these proteins, delayed coagulation, and increased prehospital fibrinolysis (Fig. 3, A–C and G). In addition to these clinical parameters, kinetic and mechanical properties of clots, such as clot stiffness, clotting rate, fibrinogen concentration, and thrombin generation velocity index, were negatively correlated to a large number of trauma-related proteins, demonstrating a relationship among traumatic injury, reduced clot stability and formation, and reduction in coagulation proteins (Fig. 3, D–F and H).

Fig. 3.

Correlation of proteins with clotting kinetics and mechanics stratified by trauma presence. Bubble U-plots showing Spearman Rank Correlation Coefficients for the 137 proteins that are significantly different between healthy donors and trauma patients and the correlation of these proteins to clinical markers: aPTT (A), PT (B), and ISS (C), mechanical coagulation parameters: maximum clot stiffness (D) and rheological clotting rate (E), and coagulation-related factors: fibrinogen (F), D-dimer (G) and thrombin generation velocity index (H). The x axis indicates the Spearman correlation r value, and the y axis indicates the -Log10(P value) of the correlation. The size of the bubble indicates the Log2(absolute value FC) of the difference in concentration of protein between healthy donors and trauma patients. Protein abbreviations are included in Supplementary Table S3, https://links.lww.com/SHK/C708. aPTT, activated partial thromboplastin time; FC, fold change; ISS, injury severity score; PT, prothrombin time.

Trauma mortality is associated with unique proteomic profiles corresponding to coagulation function

To elucidate the potential mechanistic role of proteomics markers in dysregulated coagulation kinetics and patient mortality, proteins that were differentially expressed in survived vs. deceased patients were assessed for their correlation to various clinical and biomechanical features previously established as predictive of trauma mortality (Fig. 4) (19). Of the 47 proteins significantly altered between survived and deceased patients, 26 proteins were significantly higher in deceased patients, and 21 were significantly higher in surviving patients. While no proteins higher in the deceased group related directly to coagulation, eight of the 21 proteins higher in the survived group directly related to coagulation processes, including FGA chain, PROC, plasminogen (PLG), plasma kallikrein (PK), plasma serine protease inhibitor (PAI-3), alpha-2-antiplasmin, and coagulation factor XII both the regular (F12) and short isoforms (F12-Short) (Fig. 4). The majority of proteins significantly higher in deceased patients were correlated with coagulation parameters previously associated with higher mortality (19). These include elongated aPTT and PT, elevated ISS (Fig. 4, A–C), lower clot stiffness and slower clotting rate, reduced fibrinogen concentration (Fig. 4, D–F), higher plasma D-dimer and reduced thrombin generation rate (Fig. 4, G and H). Among those that appeared frequently as highly correlated to each parameter associated with mortality, included immune and inflammation markers (SEMA4B, COL18A1, PTGDS), organ damage markers (AMY2A), and matrix proteins indicating tissue damage (NID1) (Fig. 4). These indicate a diverse network of protein pathways, influenced by trauma severity, may contribute to poor clot formation, prehospital fibrinogen consumption, reduced clot stability, and ultimately to mortality from trauma.

Fig. 4.

Correlation of proteins with clotting kinetics and mechanics stratified by mortality. Bubble U-plots showing Spearman rank correlation coefficients for the 47 proteins that are significantly different between survived and deceased patients and the correlation of these proteins to clinical markers: aPTT (A), PT (B) and ISS (C), mechanical coagulation parameters: maximum clot stiffness (D) and rheological clotting rate (E). As well as coagulation related proteins: fibrinogen (F), D-dimer (G) and thrombin generation velocity index (H). The x-axis indicates the Spearman correlation r value, and the y-axis indicates the -Log10(P value) of the correlation. Size of the bubble indicates the Log2(absolute value fold change) of the difference in concentration of protein between survived and deceased patients. Protein abbreviations are included in Supplementary Table S3, https://links.lww.com/SHK/C708. aPTT, activated partial thromboplastin time; ISS, injury severity score; PT, prothrombin time.

In patients who survived, proteins expressed at higher levels correlated to shorter aPTT, shorter PT, and lower ISS (Fig. 4, A–C), reduced plasma D-dimer, and faster thrombin generation rate (Fig. 4, G and H), all previously associated with survival (19). Several proteins, including antifibrinolytic factors (PAI-3 and thrombin-activatable fibrinolysis inhibitor), unactivated fibrinolytic proteins PLG (PLG and PROC), tissue repair proteins (RARRES2), and procoagulant factors (F12, F12-Short, and FGA), frequently appeared as highly correlated. However, when observing their correlation to clot stiffness, clotting rate, and fibrinogen concentration, only two proteins showed strong correlations with a trend toward survival: RARRES2 and FGA (Fig. 4, D–F). This indicates a lack of reliance or contribution of the other factors on these parameters.

Intersection and interplay of protein alteration based on patient outcome and trauma phenotype

To further assess the interplay between proteins associated with patient outcome, the 47 proteins that were measured to be significantly different between survived and deceased patient groups were assessed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) protein–protein interaction database (30). A network plot was constructed identifying 43 nodes and 96 edges, with several proteins exhibiting substantial interconnectedness between proteins in the plot (Fig. 5A). Overall network statistics indicate that the interactions between the proteins were statistically significant, with a protein–protein interaction P value <1.0e-16 where only 11 edges were expected. Of the 47 proteins identified, eight proteins did not have a connection to the overall network developed but did appear adjacent to the network, and four proteins were not present in the STRING database, indicating lack of connection to overall processes or lack of substantial data on these proteins (Fig. 5A). Of the proteins highlighted in the STRING diagram, proteins were related to specific biological processes as defined by GO. The most interconnected processes observed in this protein set were predominantly involved in blood coagulation and fibrinolysis-related processes. When sorting GO biological processes by STRING strength, the nine coagulation-related processes were nine of the top 11 strongest, out of 50 total processes identified.

Fig. 5.

Interplay of plasma proteomics on biological process and trauma phenotype. STRING-based web plot indicating the 47 proteins significantly altered between survived and deceased patient groups, highlighting their interactivity and relevance to GO biological processes (A). Connections between nodes indicate predicted associations based on experimental and predicted information on interactions of proteins (30). In the web, nodes indicate proteins significantly altered between survived and deceased patients, and are color-coded based on their relevance to GO biological processes. Edges in the diagram indicate the predicted and known associations, with multiple types of evidence used to indicate associations, including known associations from curated databases (cyan line) and experimentally determined (purple line), along with predicted interactions such as gene neighborhoods (green line), gene fusions (red line), and gene co-occurrences (blue line), and lastly, other interactions such as text mining (lime line), co-expression (black line), and protein homology (light blue line). Upset plot comparing the intersection of altered plasma proteomics across healthy donor and patient injury phenotype, severity, and outcome (B). The left bar graph indicates the total number of proteins significantly altered between each donor and patient group, while the top bar graph indicates the total number of proteins that are consistently significantly altered across the nodes highlighted below each bar. Filled-in circles are representative of which sets are part of a specific intersection, while lines connecting circles indicate shared proteins between data groups. GO, gene ontology; ISS, injury severity score; TBI, traumatic brain injury.

An upset plot was generated defining the number of overlapping proteins depending on injury type (Fig. 5B). Of note, five proteins were consistently altered between healthy donors and trauma patients, nonsevere ISS and severe ISS groups, and survived and deceased patient groups. These include plasma serine protease inhibitor (PAI-3/SerpinA5), vesicular integral-membrane protein (VIP36/LMAN2), collagen alpha-1 XVIII chain (COL18A1), MRC1, and PTGDS. Of these five, only PAI-3 was associated with trauma, nonsevere ISS, and survived patient groups, while the other four were associated with trauma, severe ISS, and deceased patient groups.

DISCUSSION

This study characterized the impact of traumatic injury on the levels of plasma proteins and their interplay with coagulation kinetics, clot mechanics, and markers of hemostatic function. In this single-center study, we characterized patient proteomics and their relation to these factors, from single-time point blood draws of severely injured trauma patients. We identified several differentially expressed proteins, with enriched protein abundance across several patient subgroups (stratified by injury patterns, clinical presentation, and patient outcome), associated with hemostasis, metabolic, and immune system processes. Previous studies have analyzed proteomic profiles in relation to coagulation and traumatic injury using clinical samples, demonstrating the value of this integrative approach (11,31). Building on this foundation, our study further investigates the relationship between plasma proteomics, coagulation function, and patient outcomes within a trauma cohort.

Acute-phase reactants and coagulation-related proteins, found at the interface between the systemic immune and coagulation responses to traumatic injury, were detected at higher relative abundance in trauma patients compared with healthy controls (32). VWF plays a critical role in platelet aggregation and adhesion, and is related to endothelial activation after trauma, leading to a pro-inflammatory, procoagulant state (32). Post-trauma, circulating levels of VWF are elevated, reflecting endothelial injury, which may be triggered by hypoperfusion, acidosis, and hypothermia, leading to activation of the coagulation cascade (33). In response to injury, TIMP-1, an inhibitor of matrix metalloproteinases, and MRC1, an innate immune activation marker, which modulate the inflammatory response and tissue modeling, were also expressed in higher abundance (34,35). Alpha-2-antiplasmin, factor XIII alpha chain, and fibrinogen alpha, beta, and gamma chains were lower in trauma patients than in healthy donors, supporting reduced antifibrinolytic activity and weakened blood clot mechanical stability after traumatic injury (36). The consumption of fibrinogen and antifibrinolytic proteins, whereby these proteins are incorporated into clots or cleaved during fibrin remodeling, may underlie the overall patterns observed (37–39). Our results as a whole support the hypothesis that traumatic injury enhances PLG activation, coagulation and hemostatic pathways, immune and metabolic process activity, and oxidant detoxification, indicative of a multifactorial immune-coagulation response to injury. Increasing injury severity was associated with lower protein concentrations of clot-stabilizing factors, higher platelet activation and immune response, and apoptotic pathways, further defining the overall profile of dysregulated coagulation and immune responses to traumatic injury.

Of the nearly 2000 proteins analyzed in this study, four were detected at higher levels in trauma patients (compared with healthy donors), those with severe ISS (compared with those with nonsevere ISS), and those who died from their injuries (compared with those who survived). These four proteins were: Vesicular integral-membrane protein VIP36, collagen alpha-1(XVIII) chain, MRC1, and PTGDS. Notably, PTGDS was considered to be a part of leukocyte-mediated immunity and immune effector process in GO. PTGDS generates prostaglandin D2, which can inhibit platelet aggregation and can reduce PAI-1 generation (40–42). In addition to its known role in the coagulation and immune response, our analysis showed that PTGDS abundance was negatively correlated with markers of clotting mechanics and kinetics. We have previously shown clotting mechanics and kinetics to be independent predictors of mortality after traumatic injury, suggesting PTGDS as a potential novel biomarker of immune-coagulation in trauma patients (19).

String network analysis showed that proteins associated with survival and mortality were strongly correlated, with the most prominent links involving coagulation and fibrinolysis pathways implicated in postinjury biological responses. Our proteomic analysis demonstrated a potential mechanistic role of PLG activator inhibitor 3 (PAI-3), which was detected at a higher relative abundance in trauma patients compared with healthy subjects, in nonsevere ISS compared with severe ISS, and in those who survived their injuries compared with those who died. Our correlation analysis reflected PAI-3 association with decreased fibrinolysis and accelerated clot formation. The decreased fibrinolysis could be attributed to PAI-3’s ability to inhibit tissue PLG activator and urokinase PLG activator, which ultimately inhibits fibrinolysis (43,44). PAI-3 is an inhibitor of activated protein C (APC), which could reflect the accelerated clot formation associated with it (45). APC inhibits factor Va and VIIIa for an anticoagulant effect and has been implicated in acute traumatic coagulopathy (46,47). Interestingly, the unactivated form of APC, vitamin K-dependent protein C, was lower in deceased patients than in surviving patients. Since APC has been identified as a main driver of acute traumatic coagulopathy, and has pro-fibrinolytic and anticoagulant effects, the differential presence of PAI-3 and APC, when patients are stratified by mortality, suggests an early balance toward excessive anticoagulant effects of APC and excessive antifibrinolytic effects of PAI-3 as mechanisms of acute traumatic coagulopathy and mortality after injury (46,47).

Comparisons between blunt and penetrating injury show distinct molecular responses, especially in the pathways related to metabolism, inflammation, and coagulation. In blunt trauma, elevated levels of fatty acid binding protein 4 indicate that there is facilitation of lipolysis and the release of free fatty acids into circulation, which, in turn, is linked to systemic inflammation (48). Furthermore, fatty acid binding protein 4 had moderate positive correlations with D-dimer concentration, which may reflect the body’s response to endothelial injury (48). Blunt trauma is also accompanied by higher levels of heat shock protein 90 alpha, which is a stress-induced chaperone that aids tissue repair and inflammatory regulation (49). Heat shock protein 90 alpha had a negative correlation with clotting parameters and a positive correlation with white blood cell count.

In contrast, penetrating trauma was associated with enriched processes related to inflammatory activity. Vitamin D binding protein and gelsolin were present in penetrating trauma, which are components of the actin scavenging system that clears extracellular actin released after injury (50,51). Gelsolin had negative correlations with maximum stiffness and clotting rate. Vitamin D Binding Protein had negative correlations with tissue PLG activator concentration and D-dimer concentration. As actin can interfere with fibrin structure and fibrinolysis, the removal of actin through the actin scavenging system has an association with both the coagulation and fibrinolytic system after injury as the body clears intracellular debris from the extracellular space (50).

In terms of proteins that were lower in TBI, there were biological processes altered in relation to coagulation, fibrin clot formation, and protein folding, which suggests the dysregulation or consumption of these processes after injury. For example, the biological process, “blood coagulation” was identified, which involved five proteins that were lower in TBI: fibulin-1, von Willebrand factor, platelet glycoprotein Ib alpha chain, tubulin beta-1 chain, and coagulation factor VIII. Of these proteins, only VWF had correlations with ISS, maximum stiffness, clotting rate (Pa/s), and tissue PLG activator concentration. Of the proteins involved in protein folding, prostaglandin E synthase 3 had a negative correlation with clotting rate and a positive moderate correlation with D-dimer. This enzyme produces prostaglandin E₂, which can both inhibit and promote platelet aggregation in the presence of an agonist (52,53). There is also evidence that PLG with urokinase-type PLG activator can promote prostaglandin E₂ synthesis, which may be reflected in the D-dimer release associated with prostaglandin E synthase 3 (54).

Lastly, the presence or absence of shock is associated with alterations in processes related to coagulation, wound response, and metabolic processes. Lower concentrations of procoagulant and wound healing-related proteins in the shock group suggest systemic activation and consumption of these proteins in response to and potentially further contributing to organ hypoperfusion. These findings are consistent with prior studies, which identified the importance of coagulation factors and metabolic process alterations in the occurrence and severity of shock following trauma (55,56).

This study is subject to several limitations. First, in measuring plasma proteomics, only the relative abundance of proteins is measured, not protein expression or activity levels, limiting our ability to infer functional effects from the concentration of proteins detected. The size of our patient cohort is adequate for an exploratory characterization of the proteomic and kinetic coagulation changes following trauma. However, we acknowledge that the small sample size limits our ability to make physiological or clinical inferences. This study is designed to identify potential biomarkers associated with injury phenotype and inform further larger cohort studies and functional validation experiments. In addition, we did not control for variables such as time from injury to hospital arrival. Several studies have identified proteome changes associated with subject age and biological sex. Therefore, the lack of matched groups in the present study may limit the overall generalizability of our findings (57–59). Additional confounders, such as ISS, shock, and TBI, create complexities when interpreting the proteomic relationships. We identified significant differences in factors such as GCS, ISS, TBI, and age across subgroups, which indicate they are interrelated and may influence observed associations (Supplementary Figure S1, https://links.lww.com/SHK/C706). Although our main use of Spearman correlations shows a nonparametric overview of the proteomic relationships in these subgroups, we acknowledge that this does not fully account for these confounding clinical factors. Lastly, while numerous significant correlations were observed between study variables, indicating possible previously unrecognized biological process connections, causation cannot be inferred from these results, as this is a correlation study. Our analysis allows us to observe trends and associations across groups. Mechanistic and in vivo experiments are necessary to understand how identified proteins directly or indirectly affect clinical phenotypes.

CONCLUSIONS

This single-center, single-time point of blood draw analysis of plasma proteomics of trauma patients revealed discrete biomarkers and biologic processes that are differentially expressed following traumatic injury and when stratified by mortality and injury severity. Notably, proteins such as PTGDS and PAI-3 were among the most significantly altered, highlighting their potential in serving as prioritized targets for future research aimed at modulating coagulation and immune responses after injury. Furthermore, correlation analysis with novel markers of clot mechanics and kinetics revealed mechanistic insight into immune and structural mechanisms of trauma-induced coagulopathy. These insights underscore the potential role of proteomic profiles in predictive modeling of clinical outcomes as well as identifying novel biological processes that may be therapeutic targets to mitigate the adverse effects of trauma-induced coagulopathy. The top differentially expressed proteins (LMAN2, COL18A1, MRC1, PTGDS, and SERPINA5/PAI-3) represent actionable starting points for exploring novel therapeutic strategies. Future work studying their functional role in coagulation, inflammation, and immune regulation may elucidate modulation strategies with clinical impact.

ACKNOWLEDGEMENTS

We would like to thank Dr. Haiyan Zheng, the Rutgers Biological Mass Spectrometry Core Facility and the Rutgers Center for Advanced Biotechnology and Medicine for performing sample proteomic analysis.