Paracrine signaling mediators of vascular endothelial barrier dysfunction in sepsis: implications for therapeutic targeting

ABSTRACT

Vascular endothelial barrier disruption is a critical determinant of morbidity and mortality in sepsis. Whole blood represents a key source of paracrine signaling molecules inducing vascular endothelial barrier disruption in sepsis. This study analyzes whole-genome transcriptome data from sepsis patients’ whole blood available in the NCBI GEO database to identify paracrine mediators of vascular endothelial barrier dysfunction, uncovering novel insights that may guide drug repositioning strategies. This study identifies the regulated expression of paracrine signaling molecules TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 which either disrupt or protect vascular endothelial barrier function in sepsis and could serve as potential targets for repositioning existing drugs. Specifically, TFPI (barrier protective), MMP9 (barrier destructive), PROS1 (barrier protective), and JAG1 (barrier destructive) are upregulated, while S1PR1 (barrier protective) and S1PR5 (barrier protective) are downregulated. Our observations highlight the importance of considering both protective and disruptive mediators in the development of therapeutic strategies to restore endothelial barrier integrity in septic patients. Identifying TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 as druggable paracrine regulators of vascular endothelial barrier function in sepsis could pave the way for precision medicine approaches, enabling personalized treatments that target specific mediators of endothelial barrier disruption to improve patient outcomes in sepsis.

Introduction

Vascular endothelial barrier dysfunction, characterized by the disruption of inter-endothelial junctional complexes, plays a central role in the pathogenesis of life-threatening conditions such as systemic inflammatory response syndrome (SIRS) and sepsis.¹ Under normal physiological conditions, the vascular endothelial monolayer of blood vessels acts as a selective barrier, restricting the passage of fluid and solutes from the blood to the tissues. However, in SIRS and sepsis, the barrier function of the vascular endothelium is compromised. This dysfunction leads to increased endothelial permeability, microvascular leakage of fluid and solutes, and the formation of interstitial edema in various organs, including the myocardium, kidneys, brain, and lungs.^2,3

The myocardium is particularly sensitive to interstitial fluid accumulation. Even a modest increase in the volume of interstitial fluid in the heart can impair both cardiac contractility and relaxation. This contributes to the development of sepsis-induced cardiomyopathy, a condition that heightens the risk of death by up to 70%.⁴ Similarly, disruption of the pulmonary endothelial barrier results in edema and immune cell infiltration into the lungs, leading to acute lung injury, the latter is a major contributor to the increased mortality observed in sepsis patients.^5,6 In the kidneys, impairment of the glomerular endothelial barrier function reduces the glomerular filtration rate and leads to albuminuria, which is a hallmark of acute kidney injury. The increased incidence of acute kidney injury in sepsis is strongly correlated with higher mortality rates.⁷ Blood–brain barrier disruption similarly facilitates the influx of peripheral immune cells into the brain, leading to glial cell activation and exacerbating neuroinflammation.⁸ Thus, the fluid accumulation and edema resulting from vascular endothelial barrier dysfunction in multiple organs including the brain, heart, kidneys and lungs contribute to the progression of multiple organ dysfunction and death in sepsis. Currently, there are no approved drugs specifically targeting sepsis. The standard therapeutic approach remains fluid resuscitation and antibiotics to control infection. Therefore, therapeutic strategies aimed at preventing vascular endothelial barrier disruption have been proposed to reduce the incidence of multiple organ dysfunction and mortality in sepsis.³

In sepsis and other inflammatory conditions, recognition of pathogen- and damage-associated molecular patterns by pattern recognition receptors of blood cells induces the release of immune effector molecules, including proinflammatory cytokines (IL-1β, TNF-α), proteolytic enzymes, lipid mediators (e.g. prostaglandins, leukotrienes), histamine, bradykinin, reactive oxygen species and platelet activation products.⁹ These molecules act paracrinically on the semipermeable vascular endothelial wall, inducing the inflammatory activation of vascular endothelial cells and disruption of the vascular endothelial barrier. The mechanisms of disruption include cytoskeletal remodeling, which destabilizes intercellular junctional protein complexes (e.g., tight and gap junction proteins), biochemical modification of these junctional protein complexes, and enzymatic degradation of the endothelial glycocalyx. However, targeting key immune effectors, like IL-1β and TNF-α, has not been successful in lowering mortality in clinical trials.^10–12 This warrants further research to identify novel therapeutic targets to prevent endothelial barrier dysfunction and vascular leakage in sepsis.

Whole-genome expression profiling of blood samples from sepsis patients provides critical insights into the regulation of genes encoding paracrine signaling molecules, which are secreted into the bloodstream to act on the adjacent vascular endothelial wall. These data are valuable for identifying pathogenic mediators and potential therapeutic targets.¹³ It has been reported that different leukocyte populations exhibit distinct gene expression patterns and functions in sepsis, indicating that whole-blood expression profiles may represent a composite of these patterns relevant to the sepsis phenotype.^14–17 Moreover, whole-blood gene expression analysis offers a significant advantage by minimizing experimental artifacts, as the sample collection is straightforward and requires less preprocessing compared to the isolation and ex vivo stimulation of specific leukocyte subsets.^18,19

Here, we analyzed whole-genome transcriptome data from the blood of sepsis patients across multiple regions to identify paracrine mediators and signaling pathways associated with vascular endothelial barrier dysfunction, as well as potential therapeutic targets. To the best of our knowledge, this is the first study to apply gene ontology (GO)-based pathway analysis to uncover pathways linked to endothelial barrier dysfunction and druggable targets using genome-wide expression data from whole blood samples of sepsis patients. Our findings reveal the regulated expression of several paracrine signaling molecules, TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 which play key roles in either disrupting or protecting the vascular endothelial barrier in sepsis. These molecules could serve as potential targets for repurposing existing drugs. By modulating these factors with existing drugs, it may be possible to reduce vascular endothelial barrier disruption and edema, thereby mitigating the severity and mortality of sepsis. Additionally, the identification of TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 as druggable regulators of endothelial barrier function in sepsis offers a foundation for precision medicinal approaches. These strategies could enable personalized treatments that target specific mediators of endothelial dysfunction, ultimately improving patient outcomes. This study demonstrates how leveraging existing genomic data can yield novel therapeutic insights and open new avenues for clinical interventions.

Materials and methods

Identification of statistically significant differentially expressed genes by GEO2R analysis

The whole genome transcriptome data of human sepsis patients associated with the search parameters systemic inflammatory response syndrome and expression profiling by array were retrieved from the NCBI GEO database²⁰ until 24/07/2024. The data sets of Homo sapiens, each consisting whole-blood samples of patients and concurrent healthy individuals (controls), and reporting quality and integrity assessment of RNA were included. Data obtained from (i) samples of subjects who were recovering from sepsis, (ii) blood cells incubated with inflammatory agonists including lipopolysaccharide in vitro post-isolation prior to gene expression profiling, (iii) those without concurrent healthy controls, and (iv) lacking information on the quality and integrity assessment of RNA used for gene expression profiling were excluded. The GEO accession number along with information on the sample size associated with data sets retrieved from the GEO database and used for this study are provided in Table 1. GEO2R analysis²¹ based on limma (linear models for microarray analysis)²² statistical test was then performed to identify differentially expressed genes in the pre-processed, log₂-transformed, six datasets consisting of healthy control (N = 128) and patient samples (N = 325). The data sets GSE95233, GSE28750, and GSE57065 are based on the Affymetrix platform, while the remaining three data sets GSE72326, GSE100159, and GSE69528 are based on the Illumina platform. Gene expression profiles based on Affymetrix platform have been pre-processed by the standard Robust Multi-array Average preprocessing algorithm that employs quantile normalization of probe level data, while those based on Illumina platform were normalized using quantile normalization with IlluminaGUI in R. No clear documentation was available either from publications or GEO data files associated with each accession regarding the standard procedures employed for addressing batch effects during gene expression profiling. GEO2R analysis of the afore-said gene expression data employed the Benjamini & Hochberg false discovery rate method to identify genes showing statistically significant (p < 0.05) change in expression (fold change cutoff 1.5) and to correct for the occurrence of false-positive results, as described.²¹ The statistically significant changes in differential gene expression in the sepsis data sets have been depicted using the Volcano plot (Supplementary Figure S1).

Table 1.

Whole-genome transcriptome data of human sepsis patients and control subjects retrieved from NCBI GEO database.

GEO accession number	Sample size		Age (years)		Gender (%)
	Sepsis	Control	Sepsis	Control	Sepsis		Control
					M	F	M	F
GSE72326	15	21	22–89	Not available	67	33	Not available	Not available
GSE100159	33	12	Not available	Not available	Not available	Not available	Not available	Not available
GSE95233	102	22	25–85	51–76	65	35	50	50
GSE28750	10	20	38–82	21–60	45	55	50	50
GSE57065	82	25	29–84	19–59	68	32	20	80
GSE69528	83	28	18–81	35–71	54	46	39	61
Total	325	128

Gene ontology and drug interaction/target analysis

The statistically significant differentially expressed genes satisfying a fold change cutoff of 1.5, identified by GEO2R analysis, were subjected to GO for identifying pathways associated with vascular endothelial barrier dysfunction and drug interaction/target analysis described as follows. The web-based interface ‘Pathway Commons’ employing the BioPAX model^23,24 was used to identify the pathways associated with vascular endothelial barrier in the peer-reviewed ‘Reactome’ database of human pathways and processes.^25,26 The statistically significant differentially expressed genes satisfying a fold change cutoff of 1.5 identified by GEO2R analysis and represented in the ‘Reactome’-identified pathways associated with vascular endothelial barrier were then determined by Reactome pathway analysis tool.^25,27,28 The drug interaction/target analysis for identifying genes that encode targets for the action of existing drugs was performed by the drug gene interaction database (DGIdb)²⁹ as described previously.^30,31

Results

Analysis of differentially regulated genes in sepsis for the identification of druggable paracrine mediators of vascular endothelial barrier function

GEO2R analysis of the pre-processed, log₂-transformed whole-genome transcriptome data of whole blood samples of human sepsis patients (N = 325, six data sets, Table 1) employing multiple testing corrections (refer method) and a fold change cutoff of 1.5 revealed that 712 genes were consistently differentially regulated in the whole blood samples of human sepsis patients, of which 351 were consistently significantly upregulated and 361 were consistently significantly downregulated (Figure 1, Supplementary Table S1). The Pathway Commons employing the BioPAX model showed that 76 pathways are associated with vascular endothelial barrier function in the peer-reviewed Reactome database (Supplementary Table S2).

Figure 1.

Genes consistently differentially regulated in the whole blood samples of human sepsis patients. Genes which are consistently (a) Upregulated and (b) Downregulated following GEO2R analysis of the pre-processed, log₂-transformed whole genome transcriptome data of whole blood samples of human sepsis patients (N = 325, six data sets, table 1) involving multiple testing corrections (refer method) and a fold change cutoff of 1.5. Refer supplementary table S1 for complete list of differentially regulated genes.

The drug–gene interaction/target analysis identified several genes significantly differentially regulated in human sepsis patients, many of which are associated with vascular endothelial barrier function pathways. The genes identified include both upregulated and downregulated genes, each of which interacts with/can be targeted by existing drugs. Among these, MAPK14 (upregulated), specifically associated with both the ‘VEGFA-VEGFR2 Pathway’ and ‘Platelet Sensitization by LDL Pathway’ interacts with the highest number of existing drugs. In addition, F5 (upregulated), specifically linked to the ‘Common Pathway of Fibrin Clot Formation’, interacts with the second highest number of drugs. TFPI (upregulated), associated with the ‘Extrinsic Pathway of Fibrin Clot Formation’, ranks third in terms of druggability (Figure 2, Table 2). Notably, TFPI inhibits coagulation factor Xa, the activity of which mitigates the barrier-disruptive effects of thrombin on the vascular endothelium during inflammatory conditions.³² TFPI is consistently upregulated in sepsis (Supplementary Table S1) and can be activated by existing drugs (Table 2), positioning it as a promising candidate for therapeutic intervention in sepsis.

Figure 2.

Genes consistently differentially regulated in the whole blood samples of human sepsis patients and associated with human vascular endothelial barrier function pathways. Reactome vascular endothelial barrier function pathways listed by the pathway commons employing the BioPAX model, containing the genes, consistently and significantly (p < 0.05) differentially regulated following multiple testing corrections and a fold change cutoff of 1.5 (refer method) in the whole blood samples of human sepsis patients. Druggability and/or the potential of differentially regulated genes/molecular mediators to interact with existing drugs is indicated by the percentage of normalized weighted drug target score of these pathways obtained following drug target/gene interaction analysis.

Table 2.

Genes, consistently and significantly (p < 0.05) upregulated following multiple testing corrections (refer method) and a fold change cutoff of 1.5 in the whole blood samples of human sepsis patients and associated with human vascular endothelial barrier function pathways which can be activated or inhibited by existing drugs.

Vascular endothelial barrier function pathway target	Drug	Mechanism
MAPK14	Aldesleukin	Activation
	Argipressin	Activation
	Dasatinib anhydrous	Inhibition
	Doxorubicin hydrochloride	Activation
	Doxycycline anhydrous	Inhibition
	Epoetin alfa	Inhibition
	Gefitinib	Activation
	Haloperidol decanoate	Activation
	Lithium	Inhibition
	Methoxsalen	Activation
	Paregoric	Activation
	Pirfenidone	Inhibition
	Resveratrol	Activation
	Sorafenib	Inhibition
	Vandetanib	Inhibition
F5	Alpelisib	Inhibition
	Avatrombopag	Pharmacogenomic biomarker
	Copanlisib	Experimental validation required
	Drotrecogin alfa	Inhibition
	Duvelisib	Experimental validation required
	Eltrombopag	Pharmacogenomic biomarker
	Idelalisib	Experimental validation required
	Leniolisib	Inhibition
	Lusutrombopag	Pharmacogenomic biomarker
	Rivaroxaban	Inhibition
	Sapanisertib	Experimental validation required
	Tamoxifen	Pharmacogenomic biomarker
	Thrombin	Activation
TFPI	Andexanet alfa	Inhibition
	Atorvastatin calcium trihydrate	Inhibition
	Azithromycin	Experimental validation required
	Chondroitin sulfates	Activation
	Dactinomycin	Inhibition
	Defibrotide	Activation
	Fenofibrate micronized	Inhibition
	Fulvestrant	Inhibition
	Heparin	Activation
	Lovastatin	Inhibition
	Tinzaparin sodium	Activation
ITGAM	Atorvastatin calcium trihydrate	Inhibition
	Clarithromycin	Inhibition
	Dimethyl sulfoxide	Experimental validation required
	Fentanyl citrate	Experimental validation required
	Hydrocortisone butyrate	Inhibition
	Paregoric	Inhibition
	Phenylephrine	Experimental validation required
	Theophylline	Inhibition
ITGA2B	Abciximab	Inhibition
	Alteplase	Experimental validation required
	Aspirin	Experimental validation required
	Enoxaparin sodium	Experimental validation required
	Eptifibatide	Inhibition
	Thrombin	Activation
	Tirofiban	Inhibition
MMP9	Bevacizumab-awwb	Experimental validation required
	Celecoxib	Inhibition
	Curcumin	Inhibition
	Hydralazine	Experimental validation required
	Methyldopa anhydrous	Experimental validation required
	Nifedipine	Inhibition
PROS1	Avatrombopag	Pharmacogenomic biomarker
	Estradiol valerate	Pharmacogenomic biomarker
	Lusutrombopag	Pharmacogenomic biomarker
	Progesterone	Pharmacogenomic biomarker
	Vitamin K3	Activation
	Warfarin	Pharmacogenomic biomarker
NCF4	Cyclophosphamide anhydrous	Experimental validation required
	Idarubicin	Experimental validation required
	Prednisone	Experimental validation required
	Rituximab	Experimental validation required
	Vincristine	Experimental validation required
S100A12	Atogepant	Experimental validation required
	Eptinezumab	Experimental validation required
	Methotrexate	Inhibition
	Rimegepant	Experimental validation required
	Ubrogepant	Experimental validation required
FGR	Acalabrutinib	Experimental validation required
	Bosutinib	Inhibition
	Dasatinib anhydrous	Inhibition
	Ibrutinib	Inhibition
ITGB3	Abciximab	Inhibition
	Anti-thymocyte globulin	Experimental validation required
	Eptifibatide	Inhibition
	Tirofiban	Inhibition
JAG1	Hydrocortisone butyrate	Experimental validation required
	Maralixibat	Pharmacogenomic biomarker
	Odevixibat	Pharmacogenomic biomarker
CEACAM1	Arcitumomab	Inhibition
	Tretinoin	Activation
FCER1G	Aspirin	Experimental validation required
	Benzylpenicilloyl polylysine	Activation
TGFA	Cetuximab	Inhibition
	Panitumumab	Inhibition
APH1B	Nirogacestat	Inhibition
EPAS1	Belzutifan	Inhibition
PRCP	Benazepril	Experimental validation required
SIRPA	Capecitabine	Experimental validation required
STOM	Amlexanox	Experimental validation required

Several genes associated with the ‘Integrin Cell Surface Interactions’ pathway are also significantly upregulated in sepsis. ITGAM and ITGA2B, both upregulated, each interact with 8 and 7 existing drugs, respectively. Additionally, ITGB3 (upregulated), involved in the ‘Integrin Cell Surface Interactions’, ‘VEGFA-VEGFR2 Pathway’, and ‘PECAM1 Interactions’ pathways interacts with 4 existing drugs. Other upregulated genes linked to specific pathways include MMP9 (associated with ‘Extra-nuclear Estrogen Signaling’, interacts with six drugs), PROS1 (associated with the ‘Cell Surface Interactions at the Vascular Wall’, ‘Common Pathway of Fibrin Clot Formation’, and ‘Intrinsic Pathway of Fibrin Clot Formation’, interacts with six drugs), NCF4 (associated with the ‘VEGFA-VEGFR2 Pathway’, interacts with five drugs), and S100A12 (linked to ‘Advanced Glycosylation End-product Receptor Signaling’, interacts with five drugs). FGR, associated with the ‘Platelet Sensitization by LDL Pathway’, interacts with four drugs (Figure 2, Table 2). Importantly, MMP9 and PROS1, which were previously shown to disrupt and protect the vascular endothelial barrier, respectively, through their paracrine actions^33–35 and are consistently upregulated in sepsis (Supplementary Table S1), can be targeted by existing drugs (Table 2), making them promising candidates for therapeutic intervention in sepsis.

A few genes are associated with the ‘NOTCH4 Activation and Transmission of Signal to the Nucleus’ pathway. JAG1 (upregulated) is involved in interaction with three existing drugs, while APH1B, EPAS1, PRCP, SIRPA, and STOM (all upregulated) each interact with a single drug. These genes are also associated with pathways such as ‘Regulation of Gene Expression by Hypoxia-Inducible Factor’, ‘Intrinsic Pathway of Fibrin Clot Formation’, ‘Cell Surface Interactions at the Vascular Wall’, and ‘RHOJ GTPase Cycle’. Additionally, CEACAM1 (upregulated), associated with ‘Cell Surface Interactions at the Vascular Wall’, interacts with two existing drugs. Similarly, FCER1G (upregulated), involved in both ‘Cell Surface Interactions at the Vascular Wall’ and ‘Platelet Adhesion to Exposed Collagen’, interacts with two drugs. TGFA (upregulated), associated with ‘Extra-nuclear Estrogen Signaling’, interacts with two existing drugs (Figure 2, Table 2). Of note, JAG1 which was previously shown to disrupt the vascular endothelial barrier to increase edema,^36,37 is consistently upregulated in sepsis (Supplementary Table S1). Although the druggability analysis indicates that JAG1 interacts with three drugs, the effect of hydrocortisone butyrate on its activation or inhibition is poorly understood and requires experimental validation. JAG1 has been identified as a pharmacogenomic biomarker for the other two drugs (maralixibat and odevixibat), and there currently exists no known drug that specifically inhibits JAG1 (Table 2), highlighting the critical need for novel inhibitors targeting JAG1.

Among the genes significantly downregulated in sepsis, SMAD3, associated with the ‘NOTCH4 Intracellular Domain Regulates Transcription pathway’, interacts with the highest number of existing drugs (28 drugs). LCK and MTR (downregulated), associated with ‘PECAM1 Interactions’ and ‘Sulfur Amino Acid Metabolism’ pathways, respectively, each interact with seven drugs. FYN (downregulated), involved in the ‘Cell Surface Interactions at the Vascular Wall’, ‘VEGFA-VEGFR2 Pathway’, ‘PECAM1 Interactions’, and ‘Platelet Adhesion to Exposed Collagen’ pathways, interacts with five drugs. S1PR1 (downregulated), associated with the ‘Lysosphingolipid and LPA Receptors Pathway’, interacts with five drugs. Similarly, CD2 (downregulated), associated with ‘Cell Surface Interactions at the Vascular Wall’, interacts with four drugs. Additionally, ITGA4 (downregulated), involved in both ‘Cell Surface Interactions at the Vascular Wall’ and ‘Integrin Cell Surface Interactions’, interacts with two drugs, as can ITGB7 (downregulated), which is specifically associated with the ‘Integrin Cell Surface Interactions pathway’. ITPR1 (downregulated), associated with both ‘cGMP Effects’ and ‘VEGFR2-Mediated Cell Proliferation’, is involved in interaction with two drugs. Lastly, S1PR5 (downregulated), involved in the ‘Lysosphingolipid and LPA Receptors Pathway’, also interacts with an existing drug (Figure 2, Table 3).

Table 3.

Genes, consistently and significantly (p < 0.05) downregulated following multiple testing corrections (refer method) and a fold change cutoff of 1.5 in the whole blood samples of human sepsis patients and associated with human vascular endothelial barrier function pathways which can be activated or inhibited by existing drugs.

Vascular endothelial barrier function pathway target	Drug	Mechanism
SMAD3	4-aminophenol	Experimental validation required
	Azacitidine	Experimental validation required
	Cetylpyridinium	Experimental validation required
	Chlorquinaldol	Experimental validation required
	Clioquinol	Experimental validation required
	Curcumin	Experimental validation required
	Dasatinib anhydrous	Activation (by phosphorylation)
	Dexamethasone	Inhibition
	Digitoxin	Experimental validation required
	Doxorubicin hydrochloride	Experimental validation required
	Epirubicin hydrochloride	Experimental validation required
	Fludarabine	Experimental validation required
	Fluorescein	Experimental validation required
	Fluspirilene	Experimental validation required
	Genistein	Activation
	Hexachlorophene	Experimental validation required
	Ibrutinib	Experimental validation required
	Imatinib	Experimental validation required
	Lanatoside c	Inhibition
	Leuprolide acetate	Activation
	Lopinavir	Experimental validation required
	Mitoxantrone	Inhibition
	Oxyphenbutazone	Experimental validation required
	Oxyquinoline	Experimental validation required
	Pyrithione	Experimental validation required
	Thimerosal	Experimental validation required
	Triclosan	Experimental validation required
	Vincristine	Experimental validation required
LCK	Dasatinib anhydrous	Inhibition (Multitarget)
	Entrectinib	Experimental validation required
	Gefitinib	Inhibition
	Imatinib	Inhibition
	Pazopanib	Inhibition
	Tirbanibulin	Experimental validation required
	Vandetanib	Experimental validation required
MTR	Cyclophosphamide anhydrous	Experimental validation required
	Dactinomycin	Experimental validation required
	Doxorubicin hydrochloride	Experimental validation required
	Folic acid	Experimental validation required
	Hydroxychloroquine	Experimental validation required
	Sulfasalazine	Experimental validation required
	Vincristine	Experimental validation required
FYN	Aldesleukin	Activation
	Dasatinib anhydrous	Inhibition
	Entrectinib	Experimental validation required
	Pazopanib	Experimental validation required
	Vandetanib	Experimental validation required
S1PR1	Etrasimod	Agonist
	Fingolimod hydrochloride	Agonist
	Ozanimod	Agonist
	Ponesimod	Agonist
	Siponimod	Agonist
CD2	Alefacept	Inhibition
	Aldesleukin	Experimental validation required
	Octreotide acetate	Experimental validation required
	Zinc chloride	Experimental validation required
ITGA4	Natalizumab	Inhibition (Antibody)
	Vedolizumab	Inhibition (Antibody)
ITGB7	Natalizumab	Inhibition (Antibody)
	Vedolizumab	Inhibition (Antibody)
ITPR1	Caffeine	Inhibition
	Adenosine triphosphate	Activation
S1PR5	Fingolimod hydrochloride	Agonist

Of note, S1PR5 has been shown to be crucial for the mobilization of patrolling Ly6C⁻ monocytes from bone marrow to the periphery, a process that mediates the resolution of inflammation, tissue repair, and stability of the vascular endothelium.^38,39 S1PR5 is consistently downregulated in sepsis (Supplementary Table S1) and can be activated/targeted by existing drug (Table 3), making it a promising candidate for therapeutic intervention in sepsis. Moreover, S1PR1 acts as a receptor for the ligand S1P that is secreted by circulating cells. S1P stimulates its own secretion by autocrine action and promotes the barrier function of vascular endothelial cells via paracrine signaling.^40–42 Notably, S1PR1 is consistently downregulated in sepsis (Supplementary Table S1), but it can be activated by existing drugs (Table 3), positioning it as a promising target for therapeutic modalities in sepsis.

Other druggable targets in the whole blood genome of human sepsis patients

The druggability analysis identified several genes that are significantly differentially regulated in human sepsis patients, in addition to those associated with vascular endothelial barrier dysfunction, which interact with existing drugs. These druggable, significantly upregulated genes include those involved in the heightened activation of immune pathways, leading to enhanced immune cell recruitment, cytokine production, and inflammatory responses, such as JAK2 (interacts with 26 drugs), TLR2 (interacts with 3 drugs), TLR4 (interacts with 8 drugs), IL1RN (interacts with 4 drugs), NFKBIA (interacts with 4 drugs), and FPR1 (interacts with 4 drugs). Additional targetable genes include those involved in DNA replication and cell cycle regulation, such as TYMS (interacts with 32 drugs), TOP2A (interacts with 25 drugs), and HDAC4 (interacts with 9 drugs), suggesting increased cell proliferation and survival during persistent inflammation (Supplementary Table S3).

Genes upregulated in sepsis patients also include those involved in metabolism and transport, such as LDLR (a key mediator of lipid metabolism, interacts with 20 drugs), GAPDH (a mediator of carbohydrate metabolism, interacts with 18 drugs), MGAM (involved in carbohydrate metabolism, interacts with 5 drugs), and AQP9 (a mediator of water transport, interacts with 9 drugs). These genes are implicated in the metabolic reprogramming that occurs during inflammation. Other druggable upregulated genes/gene products include ALCAM (a cell adhesion molecule, interacts with 15 drugs) and migration-related factors like ELANE (interacts with 6 drugs) and HPSE (interacts with 6 drugs), which may mediate increased tissue infiltration by immune cells. Additionally, genes involved in oxidative stress management and detoxification are upregulated, including NQO2 (interacts with 7 drugs), POR (interacts with 10 drugs), and CA4 (interacts with 9 drugs). Finally, other upregulated genes include those involved in tissue remodeling, extracellular matrix degradation, and metabolic processes such as TSPO (interacts with 26 drugs), UGCG (interacts with 5 drugs), CDA (interacts with 5 drugs), PYGL (interacts with 3 drugs), and GM2A (interacts with 3 drugs) (Supplementary Table S3). Taken together, these genes, which encode effector molecules mediating overactivated immune responses and altered cell proliferation, metabolism, and extracellular matrix remodeling processes, represent potential targets for addressing the immune dysregulation and other pathological features of sepsis with existing drugs.

Similarly, the druggable, significantly downregulated molecular targets include those mainly involved in immune regulation and the resolution of inflammatory response. This includes NT5E (CD73, interacts with 12 existing drugs) involved in the production of anti-inflammatory adenosine,⁴³ PTGER4 (interacts with 6 existing drugs) suppressing proinflammatory gene expression during challenge with bacterial endotoxin⁴⁴ and CD86 (interacts with 7 existing drugs) supporting Treg to suppress inflammation.⁴⁵ Other significantly downregulated genes with suppressive effects on the immune system include KLRD1 (interacts with 5 existing drugs) inhibiting natural killer cell function,⁴⁶ and IL2RB (interacts with 4 existing drugs) required for IL-2-triggered regulatory T cells controlling the inflammatory response.⁴⁷ Genes associated with other key cellular processes such as genomic stability and DNA repair, including APEX1 (linked to base excision repair, interacts with 41 existing drugs) and ATM (involved in sensing DNA damage, interacts with 36 existing drugs), and the antiapoptotic protein BCL2 (interacts with 32 existing drugs) were also significantly downregulated (Supplementary Table S4). Impaired sensing of DNA damage and repair might exaggerate the inflammatory response. Enhanced apoptosis of circulating cells might be associated with immunosuppression occurring in later stages of sepsis and organ failure.

An additional analysis was performed using fold change cutoff of 1.49 and p < 0.05 to identify any genes near the threshold that might have been excluded in the afore-said analysis with fold change cutoff of 1.50. It showed the upregulation of OLR1 and GNA15, associated with ‘Cell surface interactions at the vascular wall’ and ‘Thromboxane signaling through TP receptor’ vascular endothelial barrier-related pathways, respectively (Supplementary Table S5).

Discussion

Signaling molecules secreted by immune-activated circulating cells represent a key source of paracrine mediators that drive vascular endothelial barrier disruption in sepsis. This study aimed to identify these paracrine mediators, which could serve as potential targets for the repositioning of existing drugs. To this end, we analyzed whole-genome transcriptome data from the whole blood of sepsis patients across multiple regions, available in the NCBI GEO database. Using statistical analyses and GO-based pathway and drug interaction/target analysis, we identified pathways associated with vascular endothelial barrier dysfunction and potential drug targets in the genome-wide expression data of sepsis patients’ whole blood samples. Importantly, our study adds a new layer of insight by re-analyzing existing data to uncover previously unrecognized associations. Our analysis reveals the upregulation of TFPI, MMP9, PROS1, and JAG1, along with the downregulation of S1PR5 and S1PR1, which encode paracrine mediators regulating vascular endothelial barrier integrity and function in sepsis. These factors may be targeted by repositioning existing drugs to preserve endothelial barrier function in sepsis patients.

Previous studies have demonstrated that TFPI can mitigate the barrier-disruptive effects of thrombin on the vascular endothelium under inflammatory conditions. Thrombin concentration is increased during sepsis,⁴⁸ and TFPI has been shown to prevent vascular endothelial barrier disruption induced by high concentration of thrombin by proteolytic activation of the protease-activated receptors – 1 as well as 2.³² The present study shows that the expression/release of TFPI can be activated by four drugs, suggesting the potential for repurposing these agents to enhance TFPI-mediated protection of endothelial barrier integrity in sepsis. Conversely, the observation that six drugs inhibit TFPI implies that these drugs may negate TFPI’s protective effects on the endothelial barrier, potentially exacerbating vascular endothelial dysfunction and edema in sepsis and other systemic inflammatory diseases. While a direct association of some of these drugs such as defibrotide, heparin and tinzaparin with TFPI, which might mediate beneficial effects in sepsis, has not been demonstrated, these three drugs have been suggested to exert protective effects against vascular injury and thrombosis in sepsis, demanding further investigations into the underlying molecular mechanisms.^49–52

In addition to the afore-said coagulation pathway, proteolytic enzymes such as MMP9 also play a critical role in modulating endothelial barrier function during sepsis. The zinc-dependent matrix metalloprotease MMP9 has been shown to disrupt vascular endothelial barrier integrity and increase vascular permeability by degrading the tight junction proteins claudin-5, occludin and ZO-1, as well as the proteins associated with the basal membrane such as collagen, fibronectin and laminin.^53–55 Supporting its role in sepsis, circulating concentration of MMP9 protein is increased in sepsis patients,⁵⁶ consistent with the upregulated gene expression observed in this study. Furthermore, we show that MMP9 expression/activity can be inhibited by three drugs, highlighting their potential for repurposing in sepsis therapy. Conversely, drugs activating MMP9 may further compromise vascular integrity, contributing to edema and disease severity. Even though a direct association of some of these drugs such as celecoxib, curcumin and nifedipin with MMP9, which might mediate the beneficial effects in sepsis, has not been demonstrated, these drugs have been shown to exert protective effects in sepsis through anti-inflammatory and anti-oxidant mechanisms.^57–60 Therefore, whether these beneficial effects are mediated through MMP9 inhibition remains a question worthy of further studies.

Another key axis in endothelial barrier regulation involves the PROS1–Tyro3–S1PR signaling pathway. Previously, it was shown that PROS1 binds to the Tyro3 receptor, activating the sphingosine 1-phosphate receptor S1PR1 to protect the integrity of the vascular endothelial barrier.⁶¹ Interestingly, while PROS1 is upregulated, our study also demonstrates a concurrent downregulation of S1PR1 and S1PR5 in sepsis, both of which protect endothelial barrier function. Our study shows that PROS1, S1PR1 and S1PR5 could be activated by 1, 5, and 1 drugs, respectively, suggesting the potential for repurposing these agents to enhance PROS1/S1PR1/S1PR5-mediated endothelial barrier protection in sepsis. In contrast, existing drugs that inhibit PROS1, S1PR1, and S1PR5 might abolish PROS1/S1PR1/S1PR5‘s protective effects on vascular stability, exacerbating vascular endothelial dysfunction and edema in sepsis. Although a direct association of some of these small molecules such as vitamin K3 with PROS1 which might mediate the beneficial effects in sepsis has not been demonstrated, vitamin K3 is commonly used in critically ill patients, requiring further investigations into the safety and mechanism of its potential therapeutic effects.^62,63

Finally, signaling through JAG1 and its cleavage by ADAM17 represents an additional mechanism that promotes vascular barrier disruption. JAG1, consistently upregulated in sepsis, disrupts endothelial barrier integrity by decreasing the expression of VE-cadherin and dissociating VE-cadherin from β-catenin, leading to the disassembly of adherens junctions.^36,37 Interestingly, our study reveals a concurrent upregulation of ADAM17, a zinc-dependent metalloprotease that catalyzes the cleavage of JAG1, releasing its immunologically active soluble ectodomain.⁶⁴ These observations support the hypothesis that JAG1 and ADAM17 may cooperate in promoting endothelial barrier disruption during sepsis. Of note, the drug target analysis in this study reveals that there are currently no drugs available that can inhibit JAG1 and ADAM17 despite their critical role in mediating vascular endothelial barrier dysfunction in sepsis, highlighting the urgent need for novel therapeutic strategies aimed at targeting these molecules. In this context, the development of drugs that can inhibit JAG1 and ADAM17 could be crucial in preventing the endothelial barrier disruption seen in sepsis. To this end, in silico investigations employing structure-based drug discovery processes to design new small molecule inhibitors and repurpose existing small molecule drugs to inhibit ADAM17 have been ongoing.⁶⁵ The cephalosporin antibiotic ceftolozane and the thrombopoietin receptor agonist eltrombopag, both with carboxylic acid group capable of interacting with the catalytic zinc of human ADAM17 represent promising candidates for further drug repurposing studies.^66,67 Likewise, an in silico screening approach could identify small molecules that specifically interrupt the interaction of JAG1 ligand with its Notch receptor, and inhibit Notch signaling in vitro.⁶⁸ The potential of these small-molecule inhibitors of ADAM17 and JAG1 to ameliorate vascular endothelial barrier dysfunction in sepsis remains worthy of future investigations.

GO analysis aimed at deciphering the involvement and potential interactions among TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 revealed that four of these factors (JAG1, S1PR1, S1PR5, and MMP9) are involved in the Signal transduction pathway, while three (PROS1, S1PR1, and MMP9) are associated with the Immune system pathway (Supplementary Table S6). Notably, all these key mediators of vascular endothelial barrier function in severe systemic inflammation do not converge within a single pathway. This observation underscores the complexity of the molecular mechanisms driving vascular endothelial barrier disruption, and point to the challenge of developing therapeutic strategies that target a single pathway to prevent such disruption in the setting of severe systemic inflammation.

A limitation of the present study is the lack of further experimental validation of therapeutically relevant gene expression changes observed in the whole blood of sepsis patients. The expression of paracrine effectors of vascular endothelial barrier function – TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 which serve as potential targets for repositioning existing drugs has been found regulated in whole-genome expression profiling. However, further studies are required to validate their regulated expression using qRT-PCR, and protein quantification methods such as ELISA. Subsequently, the potential of these secreted paracrine effectors to contribute to the vascular endothelial barrier disruption in sepsis, and whether barrier disruption could be mitigated by targeting them with existing drugs should be confirmed. An additional limitation is the study’s failure to account for patient heterogeneity, including variations in age, sex, sepsis severity, previous treatment history, and comorbid conditions. Moreover, the autocrine mediators of barrier dysfunction, secreted by inflammatory-activated vascular endothelial cells themselves or the endothelial gene regulation occurring downstream of paracrine action of mediators secreted by immune cells could not be identified through this analysis.

In conclusion, this study identifies the regulated expression of multiple paracrine signaling molecules – TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 which either disrupt or protect vascular endothelial barrier function in sepsis and could serve as potential targets for repositioning existing drugs. Our observations highlight the importance of considering both protective and disruptive mediators in the development of therapeutic strategies to restore endothelial barrier integrity in septic patients. By repurposing existing drugs to modulate these factors, vascular endothelial barrier disruption and edema could be decreased to reduce the severity and mortality associated with sepsis. Furthermore, identifying TFPI, MMP9, PROS1, JAG1, S1PR1, and S1PR5 as therapeutically relevant paracrine regulators of vascular endothelial barrier function in sepsis could pave the way for precision medicine approaches, enabling personalized treatments that target specific mediators of endothelial barrier disruption and improving patient outcomes.

Funding Statement

This research was funded by the Faculty Research Grant (FRG)-2022, No. FRG/2022/SoBT_03, awarded to Tom Skaria by the National Institute of Technology Calicut. Shahid Puthiyottil gratefully acknowledges Council of Scientific and Industrial Research (CSIR), Government of India, for providing the Research Fellowship (09/0874(13665)2022-EMR-I).

Authors’ contribution statement

Shahid Puthiyottil conducted the literature search, performed the formal investigation and data analysis, and contributed to writing and revision of manuscript. Tom Skaria conceived and designed the study, validated the methodology and data analysis, and wrote and revised the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/21688370.2025.2503523.