High-dimensional analysis of injured patients reveals distinct circulating proteomic profiles in plasma vs. whole blood resuscitation

Abstract

Early blood product resuscitation is often essential for optimal trauma care. However, the effects of different products on the underlying trauma-induced coagulopathy and immune dysfunction are not well described. Here, we use high-dimensional analysis and causal modeling in a longitudinal study to explore the circulating proteomic response to plasma as a distinct component versus low-titer O whole blood (LTOWB), which contains plasma. We highlight the differential impacts of plasma and LTOWB on immune mediator levels and the distinct capacity of plasma to modulate coagulation by elevating fibrinogen and factor XIII and reducing platelet factor 4. A higher proportion of plasma in prehospital resuscitation is associated with improved admission time coagulation parameters in patients with severe shock and elevated brain injury markers and reduced post-admission transfusion volumes in those suffering from traumatic brain injury (TBI) and blunt injury. While LTOWB offers broad hemostatic benefits, our findings demonstrate specific advantages of plasma and support individualized transfusion strategies.

Keywords: blood coagulation; blood transfusion; fibrinogen; innate immunity; plasma; platelet activation; platelet factor 4; precision medicine; proteomics; traumatic brain injury.

Graphical abstract

Figure 1

A summary of study flow and cohort characteristics 402 blood samples from 134 SWAT participants drawn at admission, 4 h, and 24 h, and 20 donor plasma samples were analyzed for circulating proteomics (A). Using network analysis, we identified homogeneous immune and coagulation-associated protein modules whose levels were compared between patients receiving different blood products (B). In parallel, we identified all SWAT patients who received either LTOWB or plasma component treatment by admission but not both (C). Propensity-matched subsets of these two treatment groups were compared for their clinical coagulation parameters. In addition, we assessed how receiving plasma could influence the total 24-h transfusion received in different patient subsets (D). (E) and (F) summarize the injury characteristics of patients in the omics sub-cohort. Patients could have received plasma from different sources (plasma component or LTOWB). (G) Summarizes the distribution of products as received before each blood draw.

Figure 2

Samples from donors and patients with trauma show clear and time-dependent differences in the overall circulating proteomic profiles Volcano plots (A)–(F) demonstrate time-dependent differences between patient and donor samples, highlighting the most significantly different (size >2.2 SDs, false discovery rate (FDR)-adjusted Mann-Whitney U p value log <−8) proteins at 0 (A), 4 (B), and 24 h (C). Integrin αIIbβ3 protein and a collection of mitochondrial matrix-associated proteins (green box) were higher in donors at all time points. Heart-specific fatty acid-binding protein at 0 h, extracellular matrix-associated proteins MMP9 and HSP47 at 4 h, skeletal muscle-specific troponin I, and beta enolase at 24 h had a relatively higher level in patients. (D) Identifies higher levels (size >1 SD, FDR-adjusted Mann-Whitney U p log <−15) of tissue-specific (TNNI, TIMP1) and immune-associated (SAA2, IL1RL1, and GCSF) proteins in 4 h samples compared to 0 h samples. As seen in (E) and (F), both 0 and 4 h samples have higher levels of leukocyte cell-derived chemotaxin-2, cathepsin F, and CNS-associated protocadherin-8. 24 h samples have higher levels of acute inflammatory proteins (SAA1 and 2, CRP, and LBP), chemokines (CCL 7 and 23), and immunoglobulin-like CD226 protein compared to the early time points. (G–L) Show principal component analysis using 198 immune and hemostasis-associated proteins. The red asterisks represent significant component-level differences at FDR-adj. p < 0.05 as estimated by the Mann-Whitney U test. Patient and donor samples are easily distinguishable in the first two dimensions (G). Plasma recipients show a distinct proteomic profile (H), but no clear separation is visible between LTOWB and no-LTOWB groups (I). A minor separation can be seen when patients are dichotomized using severity (J), penetrating trauma (K), or brain injury (L).

Figure 3

Prehospital plasma, but not LTOWB, is associated with immune pathway alterations in severely injured patients Volcano plot (A) highlights the most different (size >1 SD, FDR-adjusted Mann-Whitney U p value <0.001) circulating proteins between 59 plasma recipients and 75 non-recipients (with or without LTOWB). Patients treated with plasma had higher levels of protein C, GRFA1, SMOC1, FlII, and MMP8. As (B) shows, the differences mainly persisted in severely injured patients who had a <90% chance of resolution (n = 41 plasma recipients vs. 42 non-receivers). Severely injured plasma recipients had lower levels of PF4, cellular proteins DPYSL4 and PLSCR3, as well as immune mediators INFA5 and CCL5. As seen in (C), receiving LTOWB (with or without plasma) did not lead to significant differences, whether in all or only in severely injured patients. (D) Shows relative enrichments in Gene Ontology and Reactome databases between severely injured plasma (n = 17) and LTOWB recipients (n = 22) at 0 h (none had received both treatments). We found higher levels of IL-2 (Family-wise error rate [FWER] p value <0.1, absolute normalized enrichment score >2) in plasma recipients.

Figure 4

Plasma recipients show a distinct pattern in coagulation and platelet-associated proteins Using a causal forest model to account for confounders and heterogeneity of treatment effects, CLOT (A) and ALPHA (C) modules were found to be significantly associated with plasma in admission and 24 h samples, but not with LTOWB (B, D). In (A)–(D), the error bars show mean ± SEM (standard error of the mean) at each time point. Gold and purple asterisks indicate p < 0.05 for, respectively, nominal and adjusted p values from the Mann-Whitney U test of between-subgroup effect differences. (E) Compares the relative level of admission time ALPHA and CLOT constituents in plasma recipients (with or without LTOWB) to that of plasma non-recipients in ISS, gender, and age-matched TBI vs. NBI groups of patients. The green asterisks present FDR-adjusted p values as estimated by the Mann-Whitney U test between treatment recipients and non-recipients. In the TBI group, several individual proteins remained significantly different after multiple testing adjustments.

Figure 5

Severe shock and head trauma moderate the association of plasma with coagulation and platelet-associated protein modules The error bars show mean ± sub-sample SD (standard deviation) at each time point, and purple asterisks indicate p < 0.01 after multiple testing adjustments on the Mann-Whitney U test and between-subgroup effect size difference >0.5. Based on this conditional effect analysis, the positive association of plasma component treatment with CLOT was significantly stronger in patients with head injury, old age, and blunt mechanism at 0 h (A) and severe shock at 24 h (B). The plasma association with lower ALPHA levels was moderated by age and severe shock at both time points (C and D).

Figure 6

Clinical impacts of prehospital treatment with plasma We divided patients into high- and low-brain injury groups based on a Brain Injury Score created with PCA on admission levels of six brain injury-specific proteins (see Table S7). (A) Shows partial correlation analysis between treatment share (product units to total units received) and PT at admission. Error bars indicate mean+SEM for the correlation measure in each subgroup. A significant plasma association with lower PT was only observed in the high Brain Injury Score plasma recipients. (B–D) Compare coagulation parameters in propensity-matched plasma (n = 36) and LTOWB recipients (n = 35) selected from patients who had not received both treatments. PT and INR are significantly lower in the plasma component group, while platelet count is similar. For (A)–(D), p values were estimated by the Mann-Whitney U test. (E–G) Show the standardized effect measures regarding the association of pre-admission plasma with the sum of total units of blood products received on the first admission day as estimated by the multi-variable mixed-effects regression model. Plasma is significantly associated with lower transfusion volume only in the TBI subgroup (E). A trend toward a higher response in patients with blunt mechanisms of injury can be observed (F). The shock index does not moderate this association (G) (see Tables S5 and S6; Figure S20).