Genetic Profile of Endotoxemia Reveals an Association With Thromboembolism and Stroke.

Background Translocation of lipopolysaccharide from gram-negative bacteria into the systemic circulation results in endotoxemia. In addition to acute infections, endotoxemia is detected in cardiometabolic disorders, such as cardiovascular diseases and obesity. Methods and Results We performed a genome-wide association study of serum lipopolysaccharide activity in 11 296 individuals from 6 different Finnish study cohorts. Endotoxemia was measured by limulus amebocyte lysate assay in the whole population and by 2 other techniques (Endolisa and high-performance liquid chromatography/tandem mass spectrometry) in subpopulations. The associations of the composed genetic risk score of endotoxemia and thrombosis-related clinical end points for 195 170 participants were analyzed in FinnGen. Lipopolysaccharide activity had a genome-wide significant association with 741 single-nucleotide polymorphisms in 5 independent loci, which were mainly located at genes affecting the contact activation of the coagulation cascade and lipoprotein metabolism and explained 1.5% to 9.2% of the variability in lipopolysaccharide activity levels. The closest genes included KNG1, KLKB1, F12, SLC34A1, YPEL4, CLP1, ZDHHC5, SERPING1, CBX5, and LIPC. The genetic risk score of endotoxemia was associated with deep vein thrombosis, pulmonary embolism, pulmonary heart disease, and venous thromboembolism. Conclusions The biological activity of lipopolysaccharide in the circulation (ie, endotoxemia) has a small but highly significant genetic component. Endotoxemia is associated with genetic variation in the contact activation pathway, vasoactivity, and lipoprotein metabolism, which play important roles in host defense, lipopolysaccharide neutralization, and thrombosis, and thereby thromboembolism and stroke.

limulus amebocyte lysate

genetic risk score of endotoxemia

Mendelian randomization

Clinical Perspective

What Is New?

The biological activity of lipopolysaccharide in the circulation (ie, endotoxemia) has a small but highly significant genetic component.

The 5 genetic loci, which associate with endotoxemia, are mainly located at genes affecting the contact activation of the coagulation cascade and lipoprotein metabolism.

The genetic risk score of endotoxemia is associated with deep venous thrombosis, pulmonary embolism, venous thromboembolism, and ischemic stroke.

What Are the Clinical Implications?

The analyses suggest that endotoxemia may be one of the causal factors in thromboembolism and stroke.

The results indicate that the microbiome/host interactions play a role in thromboembolism and stroke risk.

Lipopolysaccharide, also known as endotoxin, is an important virulence factor for gram‐negative bacteria. The structural differences of the lipopolysaccharide molecules between bacterial species can have a major effect on their functional properties and biological activity. Lipopolysaccharide can act as an immunostimulator or immunomodulator, thereby contributing to the virulence of various bacterial species.

Translocation of lipopolysaccharide in the circulation, endotoxemia, can occur in the interface of host mucosal microbiota and the bloodstream (eg, in the gut). Endotoxemia is associated with an increased risk of cardiometabolic disorders, including incident cardiovascular disease events, obesity, metabolic syndrome, and diabetes. , , In addition, serum lipopolysaccharide activity is associated with many noncommunicable disease risk factors: it is inversely associated with high‐density lipoprotein (HDL) cholesterol concentrations and directly with triglyceride, cholesterol, CRP (C‐reactive protein), fasting glucose, insulin, glycated hemoglobin concentrations, and body mass index. Overall, endotoxemia is associated with a highly adverse metabolic profile of inflammatory character.

Previous candidate gene studies have demonstrated the importance of innate immune system pathways in host responsiveness to administered lipopolysaccharide. In addition, a genome‐wide association study (GWAS) for fever after evoked endotoxemia identified a genetic locus that modulates clinical responses in trauma and sepsis. However, the genetic determinants of human endotoxemia have not been investigated previously. The aim of this work was to assess the genetic profile of serum lipopolysaccharide activity using a GWAS approach, and to determine whether this profile has an association with cardiovascular risk.

METHODS

Data Availability

The authors declare that the data supporting the findings of the study are available within the article, its Supplementary Material, and on request from the corresponding author. Genome‐wide summary‐level statistics are available to download from GWAS Catalog with study accession GCST90032674 (ftp://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90032001‐GCST90033000/GCST90032674/). The FINRISK data are available from the THL Biobank (https://thl.fi/en/web/thl‐biobank/for‐researchers) based on a written application and following the relevant Finnish legislation. For the FinnDiane study and the twin samples, individual‐level data cannot be shared because of restrictions in patient and participant consent. FinnGen Data Freeze 5‐summary level data are publicly available at the FinnGen website (https://www.finngen.fi/en/access_results). UK biobank summary‐level data are available at http://www.nealelab.is/uk‐biobank/ and Megastroke at https://www.megastroke.org/.

Study Population

Participants from 3 studies, FinnDiane, FINRISK, and Finnish Twin Cohort, were used in the GWAS analyses. The FinnDiane cohort consists of participants with type 1 diabetes (FD‐T1D) and individuals with unclassified diabetes; the FINRISK studies consist of population‐based surveys in Finland; and, finally, the Finnish Twin Cohort study consists of Finnish adult twins. From each study, 2 independent cohorts were analyzed separately: from the FinnDiane, FD‐T1D (n=3940) and Finn Diane cohort of individuals with unclassified diabetes (n=302); from the FINRISK, FINRISK92 (n=656) and FINRISK97 (n=5667); and from the Finnish Twin Cohort, FinnTwin16 (n=451) and VpEpi (n=280). Altogether, the GWAS analyses included 11 296 unique samples. Independent from the GWAS population, FinnGen Study (https://www.finngen.fi/) Data Freeze 5, including 195 170 participants, was used to analyze the associations of the designed genetic risk score of endotoxemia (LPS‐GRS) with disease end points. Every participant provided written informed consent, and the study was approved by the Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa. Additional lead single‐nucleotide polymorphism (SNP) association lookups and Mendelian randomization (MR) analysis used UK Biobank and Megastroke populations. Data S1 contains more detailed descriptions of study populations.

Lipopolysaccharide Activity Measures

Endotoxin activities were determined with a limulus amebocyte lysate (LAL) assay on 1:5 diluted serum samples (HyCult Biotechnology b.v., Uden, the Netherlands), and the results were log transformed (natural logarithm) because of skewed distributions. Data S1 contains more details.

Genetic Analysis

Cohorts were genotyped with various genotyping platforms and went through rigorous quality control. Imputation was performed using 1000 Genomes Project phase 3 reference genotypes. Single‐marker association analysis was performed with linear mixed model to correct for the effect of cryptic relatedness and close relatives. The models were adjusted for empirical kinship matrix, sex, age, body mass index, total cholesterol, HDL cholesterol, triglycerides (log transformed), and study‐specific covariates (in FinnDiane, sample freeze time and genotyping batch; in FINRISK, genotyping batch and recruitment region; and in Finnish Twin Cohort, sample freeze time, genotyping batch, and additionally pregnancy status in FinnTwin16). The results were meta‐analyzed using inverse variance weighted fixed effect meta‐analysis. Genome‐wide significance level was set to P<5×10−8. More detailed specification of genetic analyses, expression quantitative trait loci analysis, gene set enrichment analysis, genetic risk score, MR, post hoc GWAS, and conditioned analyses can be found in Data S1.

Endotoxemia Measured by Using Other Techniques

A subpopulation of FinnDiane was used to determine endotoxemia by mass spectrometry–based method (n=363), as previously described, and a commercially available Endolisa assay (n=326) (609033; Hyglos GmbH, Bernried, Germany). Selection of participants for mass spectrometry was based on the F12 SNP rs1801020 genotype, obtaining 121 TT‐homozygous and 242 CC‐homozygous participants. For Endolisa assay, population genotype for rs1801020 was distributed as follows: TT, 21; TC, 124; and CC, 181. Endolisa measures were log transformed, and the correlations with LAL assay results were computed using Pearson product‐moment correlation. The SNP association with endotoxemia was analyzed with linear regression, applying an additive genetic model.

RESULTS

Genetic Association Analyses

The genetic factors associated with endotoxemia, measured using the LAL assay, were analyzed in 6 Finnish cohorts using the GWAS setting. The genetic analyses contained 8.1 to 9.9 million genotyped or imputed genetic markers passing quality control per cohort. The clinical characteristics of participants are presented in Table 1.

Table 1

Clinical Characteristics of the Subjects in the 6 Cohorts

Characteristic	Cohort
	FD‐T1D	FD‐rest	FINRISK92	FINRISK97	FT16	VpEpi
Total No.	3940	302	656	5667	451	280
Log(lipopolysaccharide)	−0.65 (0.48)	−0.60 (0.49)	0.23 (0.55)	−0.61 (0.52)	−0.47 (0.30)	−0.61 (0.21)
Women, n (%)	1901 (48.2)	133 (44.0)	250 (38.1)	2826 (49.9)	233 (51.7)	172 (61.4)
Age, y	40.58 (12.7)	53.78 (11.9)	53.66 (8.7)	53.07 (10.8)	24.40 (0.8)	61.72 (4.2)
BMI, kg/m2	25.37 (3.8)	26.28 (4.1)	27.59 (4.5)	27.15 (4.4)	22.72 (2.9)	27.77 (4.9)
Total cholesterol, mmol/L	4.81 (0.9)	4.87 (1.0)	5.99 (1.1)	5.68 (1.0)	4.94 (0.9)	4.68 (0.9)
HDL cholesterol, mmol/L	1.42 (0.4)	1.38 (0.4)	1.32 (0.4)	1.39 (0.4)	1.78 (0.4)	1.53 (0.4)
Log(triglycerides), mmol/L	0.08 (0.5)	0.13 (0.5)	0.45 (0.6)	0.29 (0.5)	0.14 (0.4)	0.14 (0.4)

Values are given as mean (SD) or number (percentage). Log refers to logarithmic transformation of the values. All clinical variables differed between the cohorts (P<0.001, 1‐way ANOVA or χ2 test, where appropriate). BMI indicates body mass index; FD‐T1D, FinnDiane cohort of participants with type 1 diabetes; FD‐rest, FinnDiane cohort of individuals with unclassified diabetes; FINRISK92 and FINRISK97, FINRISK‐cohorts enrolled in 1992 and 1997, respectively; FT16 and VpEpi, younger and older subpopulations of the Finnish Twin Cohort; and HDL high‐density lipoprotein.

When the cohort‐wise results were combined in fixed‐effects meta‐analysis composed of 11 296 participants, 741 markers clustered at 5 independent loci (in chromosomes 3, 4, 5, 11, and 15) were genome‐wide significantly (P<5×10−8) associated with endotoxemia (Figure 1 and Table 2). In addition, a single SNP, rs77601517 on chromosome 12, reached genome‐wide significance in the FINRISK‐97 cohort. However, the SNP was additionally available only in the FINRISK‐92 cohort, demonstrating no association, and was not studied further. All the significant SNPs (Table S1) and the Manhattan and QQ plots (Figure S1) are available in the Supplementary Material.

Figure 1

Manhattan and QQ plots of genome‐wide association study (GWAS) results combined in fixed‐effects meta‐analysis.

We performed a GWAS of endotoxemia, measured by limulus amebocyte lysate assay in 11 296 individuals with Finnish ancestry. The horizontal red line represents genome‐wide significance (P<5×10−8). Single‐nucleotide polymorphisms in 5 independent loci available in all cohorts passed the genome‐wide significance threshold. Inflation of the P values is presented in the QQ plot (λqc=1.049).

Table 2

Lead SNPs in Each Locus Associated With a Genome‐Wide Significant Level With Endotoxemia

Marker	rs5030082	rs71640036	rs1801020	rs2081361	rs10152355
Closest gene	KNG1 (intron)	KLKB1 (intron)	F12 (5’UTR)	YPEL4	LIPC
Chromosomal position	3:186458949	4:187161120	5:176836532	11:57411742	15:58671178
A1/A2	A/G	T/G	A/G	T/C	A/C
A2 frequency	0.38	0.42	0.75	0.7	0.54
β (SE)	0.12 (0.01)	−0.25 (0.01)	0.26 (0.02)	0.09 (0.01)	−0.08 (0.01)
P value	2.95×10−19	5.41×10−78	6.62×10−65	4.37×10−9	2.51×10−9
Direction	+−−+++	−−−−−−	++−+++	++++−+	−−−−−−
I2	70.9	92.9	90	1.2	0
Heterogeneity P value	0.00415	8.85×10−14	1.25×10−09	0.409	0.919
eQTL associations	KNG1	CYP4V2, F11, F11‐AS1, FAM149A, RPSAP70, KLKB1, FLJ38576, TLR3	F12, FGFR4, LMAN2, MXD3, RAB24, RGS14, PRELID1, SLC34A1	AP000662.4, MED19, TIMM10, ZDHHC5, SERPING1	LIPC, ADAM10

Fixed‐effects meta‐analysis, full list of 741 SNPs is presented in Table S1. The associations between all genome‐wide association study significant SNPs and eQTL were considered and grouped under the lead SNP of the locus. Full list of eQTL associations is available in Table S7; a false discovery rate of <0.01 was required (full list in Table S2). 5’UTR indicates 5′ untranslated region (regulatory region); A1, reference allele; A2 frequency, allele frequency of the effect allele in the combined sample; A2, alternative allele; β, estimated effect size for each copy of effect allele (increase in SD of normalized residual unexplained by other covariates); direction, sign of estimates (? for not available) for each cohort in order FinnDiane cohort of participants with type 1 diabetes, Finn Diane cohort of individuals with unclassified diabetes, FINRISK92, FINRISK97, FinnTwin16, and VpEpi; eQTL, expression quantitative trait loci; I2, I2 estimate for heterogeneity; and SNP, single‐nucleotide polymorphism.

The cohort‐wise forest plots for the lead variants are presented in Figure 2. Stepwise conditional regression with GCTA–conditional and joint analysis did not indicate independent secondary signals in any of the 5 loci after accounting for the lead SNPs. QQ plot of the GWAS meta‐analysis with the significant SNPs omitted (Figure S2) demonstrated good adherence to the diagonal for the nonsignal SNPs. Figure S3 presents the Manhattan and QQ plots of GWAS conditioned on the lead SNPs. In addition, the leave‐one‐out analysis showed that single cohort is not excessively driving the result of the meta‐analysis (Table S2 and Figure S4).

Figure 2

Forest plots of the lead single‐nucleotide polymorphisms of loci reaching genome‐wide significance in meta‐analysis of the genome‐wide association study for endotoxemia.

The meta‐analysis included 11 296 samples. Presented point size is proportional to the inverse of SE of the estimate. DinnDiane: rest, FinnDiane cohort of individuals with unclassified diabetes; FinnDiane: T1D, FinnDiane cohort of participants with type 1 diabetes; FR indicates FINRISK; and FT16 and VpEpi, subpopulations of Finnish Twin Cohort.

Linking SNPs to Genes

The regional plots of the 5 loci identified (Figure S5) and their position in relation to nearby genes (Table S1) are presented in the supplemental data. To identify biological pathways and processes behind endotoxemia, we performed a pathway enrichment analysis. The complete list of top pathways enriched with P<0.01 is presented in Table S3. The strongest gene set enrichment was observed in the biological processes “intrinsic pathway” (23 genes; P=5×10−4), followed by “chromatin packaging and remodeling” (237 genes; P=5×10−4). However, none of the associations in the enrichment analysis reached statistically significant false discovery rate (<0.05).

The expression quantitative trait locus analysis identified 24 genes affected by our GWAS significant markers (Table 2; full list in Tables S4 and S5 for blood and genotype‐tissue expression portal, respectively). Loci in chromosomes 3, 4, 5, and 11 had expression quantitative trait loci associations with several genes' expression (KNG1, F11/KLKB1, F12, and SERPING1, respectively) playing a role in the contact activation of the intrinsic pathway of coagulation. Figure 3 illustrates the relation of these genes and the intrinsic pathway of coagulation.

Figure 3

Hypothesized connection between intrinsic pathway of coagulation and genetic variants associating with endotoxemia.

Genetic associations connect endotoxemia to increased coagulation: via gene expression (expression quantitative trait loci [eQTL]), protein expression, and protein activity. Two possible mechanisms are hypothesized to explain the association. Bradykinin is cleaved from kininogen and can affect intestinal permeability by allowing increased microbial leakage from the gut. Immunothrombosis, a mechanism proposed to be involved in normal immunology, is hypothesized to alter both immunological defense and formation of thrombosis when dysregulated. Genetic risk score and Mendelian randomization analysis connect endotoxemia to venous thromboembolism, deep vein thrombosis, and stroke. GWAS indicates genome‐wide association study.

Post Hoc Regression Analyses

To evaluate the proportion of the endotoxemia variance explained by the genetic markers, we performed additional regression analyses in the 2 largest cohorts (FD‐T1D and FINRISK97). The clinical covariates, age, sex, body mass index, and total cholesterol, HDL cholesterol, and triglyceride concentration, used in the GWAS analyses explained 50.8% of the lipopolysaccharide variability in the FinnDiane cohort and 27.8% of the lipopolysaccharide variability in the FINRISK cohort (Tables S6 and S7). Adding the 5 lead SNPs representing each loci to the models increased the proportions by 1.5 percentage points in FinnDiane and by 9.2 percentage points in FINRISK, resulting in 52.3% and 37.1% of the variance explained, respectively.

Alternative Methods to Measure Lipopolysaccharide

Endotoxemia was also measured in 2 separate subsamples of the FinnDiane cohort by determining the lipopolysaccharide mass and by using the Endolisa assay. The original LAL results, which measure the biological activity of lipopolysaccharide, had a modest but significant correlation with both the lipopolysaccharide mass (correlation=0.23; P=7.5×10−6) and the Endolisa (correlation=0.19; P=7.1×10−4) results. In both subsamples, the lead SNP of the FinnDiane cohort (rs1801020) was strongly associated with endotoxemia, when measured using the LAL assay (β=0.083 [P=7.48×10−5] and β=0.072 [P=0.016], respectively). However, the lead SNP did not associate significantly with either lipopolysaccharide mass (P=0.30) or the Endolisa results (P=0.7). The measures and differences by genotype are presented in Figure S6.

Genetic Risk Score and MR

In the Megastroke population, single lead SNPs were associated with the risk of “ischemic stroke,” “any stroke,” “TOAST small artery occlusion,” “TOAST cardioaortic embolism,” or “intracranial aneurysm” (Figure 4A). We composed a genetic risk score for endotoxemia (LPS‐GRS) from GWAS‐significant SNPs and analyzed its association with designed end points in an independent population, the FinnGen. LPS‐GRS was significantly associated with deep vein thrombosis, pulmonary embolism, and venous thromboembolism (Figure 4B and Table S8). Next, MR using the lead SNPs was conducted on the UK Biobank data, which displayed associations with deep vein thrombosis, pulmonary embolism, and venous thromboembolism, and Megastroke populations, which revealed an association with ischemic stroke (Figure 4C and Figure S7). More detailed information is available in Tables [Link], [Link].

Figure 4

Associations of the cardiovascular disease end points with 5 endotoxemia‐associated single‐nucleotide polymorphisms (SNPs) and the composed genetic risk score (GRS).

A, Significant associations of the lead SNPs with stroke events in the Megastroke population. P value for logistic regression analysis is presented on the right side. The asterisk marking represents P<0.01. B, GRS of endotoxemia (LPS‐GRS) is calculated from the 5 lead SNP genotypes, based on stepwise conditional regression (GCTA–conditional and joint analysis [COJO]) of lipopolysaccharide genome‐wide associated study results weighted with effect sizes. False discovery rate (accounting for total of 46 end points tested) is presented on the right side. Population consists of 195 170 individuals from the FinnGen study. Full list of analyzed phenotypes is in Table S8. C, Association between endotoxemia and prothrombotic end points is analyzed by Mendelian randomization in Megastroke (for ischemic stroke) and UK Biobank populations using MR‐base platform. The 5 lead SNPs, based on GCTA‐COJO, were used in the analysis. DVT indicates deep vein thrombosis; OR, odds ratio; and TOAST, classification of 5 subtypes of ischemic stroke.

DISCUSSION

We identified 5 genetic loci that displayed significant association with serum endotoxin activity levels in multiple cohorts. According to the expression quantitative trait loci, several of these SNPs are associated with expression of nearby genes that affect the contact activation of coagulation and lipoprotein metabolism. These both play important roles in host defense against infectious organisms, including induction of inflammatory responses and lipopolysaccharide neutralization. Furthermore, the composed genetic profile and the MR results indicated associations of endotoxemia with thromboembolism and stroke. The results further link microbiomes with cardiovascular diseases.

Endotoxemia is associated with increased risk of cardiovascular events in the largest population of the present study, the FINRISK, and other studies. , , , It may also contribute to stroke, , , and metabolic endotoxemia has been suggested as a novel therapeutic target to improve stroke outcome. Dysbiosis may maintain an inflammatory environment, which has a significant impact on cerebrovascular risk and stroke severity through the microbiota‐gut‐brain axis. Dysbiosis, including increased abundance of gram‐negative Enterobacteriaceae‐family members, has been reported in patients with large‐artery atherosclerotic ischemic stroke and transient ischemic attack compared with asymptomatic people. Therefore, in addition to a genotype predisposing to endotoxemia, patients with stroke may have an ample source of lipopolysaccharide. A more recent study showed that interaction between lipopolysaccharide and SARS‐CoV‐2 S protein resulted in a hyperinflammatory effect, which was hypothesized to contribute to the activation of the coagulation and complement system observed in severe COVID‐19 disease. Future research will show whether endotoxemia plays a role in COVID‐19.

A large portion of significant polymorphisms associated with endotoxemia in the present study were located at genes affecting the contact activation of the coagulation cascade. The dynamics of this pathway are complex but relatively well described, and the currently detected genetic associations fit well with an overall phenotype of increased coagulation. The lead SNP in the chromosome 5 locus (rs1801020) is a known functional variant; the minor allele results in an additional upstream open reading frame for gene F12 at the sequence level, leading to increased expression and activity. , The KNG1, KLKB1/F11, and SERPING1 loci also contained SNPs that affect the gene/protein expression levels or the protein activity. , The lead SNP on chromosome 11, rs2081361, has been previously associated with high SERPING1 expression, which may affect the levels of the central complement system regulator, C1 inhibitor, and thereby FXII activity. Furthermore, in chromosome 3 locus, rs5030049 in the intronic region of KNG1 is associated with higher protein levels of high‐molecular‐weight kininogen, which is the nonenzymatic cofactor in the contact system. In the same locus, rs710446 and rs2304456 have also been associated with FXI plasma levels. In addition, endotoxemia was associated with rs4253238 and rs4253417 in chromosome 4, linking it with high activities of plasma kallikrein and FXI, which are important activators of coagulation (ie, FX and FIX). In addition to the complement system, all these proteins play a major role in the intrinsic pathway. “Intrinsic cascade” was also recognized as the top pathway enriched in the present study.

In post hoc analysis, lead SNPs explained a larger percentage of endotoxemia variance in the population‐based cohort (FINRISK97) than in the cohort consisting of patients with diabetes (FD‐T1D). Subjects with diabetes have higher serum lipopolysaccharide levels compared with subjects without diabetes because of hyperglycemia, hyperinsulinemia, hypertriglyceridemia, and low HDL cholesterol concentrations. All these metabolic features are associated with impaired clearance of lipopolysaccharide, and the role of genetics in determining the endotoxemia levels may be smaller. Indeed, the covariates explained a larger portion of the endotoxemia variance in the cohort of diabetic subjects than in the population‐based cohort. However, the variance explained may be notably overestimated because of a phenomenon called “winner's curse,” which is characteristic for large‐scale quantitative trait studies.

Several SNPs in the endotoxemia‐associated loci discovered have been earlier associated with thrombosis in general: they include SNPs in genes F11, KNG1, and KLKB. The LPS‐GRS of the present study also showed associations with conditions linked to blood coagulation alterations, which further validates the association between endotoxemia and the contact activation pathway. Although the genetic background of stroke has been intensively studied and heritability has been evaluated to range between 16% and 40%, our results are novel: the MR results suggested a causal role of endotoxemia in thromboembolism and stroke. The MR approach uses the genetic risk score SNPs (effectively the lead SNPs of the lipopolysaccharide GWAS) as “instrumental variables,” which allows us to study the causality between endotoxemia and clinical end points in comparable manner as in a randomized controlled trial. Using genetic variants as instrumental variables, which do not have endogenous issues, such as reverse causation or missing confounders, provides us with consistent estimates from a regression. However, one of the MR assumptions is that the genetic variant does not have any other effect to the outcome other than through the exposure (endotoxemia). In the present study, this assumption was not totally fulfilled, because it can be assumed that the SNPs located at genes affecting the coagulation cascade could in addition have more direct associations with thromboembolism and stroke.

The observed signal of causality between thrombosis‐related end points and endotoxemia may be hypothesized to be a sign of imbalanced “immunothrombosis” (Figure 3), a term that describes a thrombosis in microvessels triggered by inflammation. In this concept, proposed by Engelmann and Massberg, immunothrombosis is a naturally occurring process that suppresses pathogen invasion, but potentially leads to pathological thrombosis if not carefully in balance. We hypothesize that LPS‐GRS may partly reflect defective immunothrombosis, which may affect both the formation of pathogenic thrombosis and the elevation of lipopolysaccharide activity in circulation.

In addition to immunothrombosis, it is plausible that the contact activation pathway has a significant impact on gut barrier function (Figure 3). Negatively charged molecules (eg, heparin, dextran sulfate, and endotoxins) activate the kallikrein‐kinin system, which eventually leads to production of the vasodilator, bradykinin. Earlier studies support the view that aberrant activation of kallikrein‐kinin pathway could be associated with decreased intestinal barrier function, which may eventually lead to higher circulating endotoxin levels and thereby increase the risk of organ damage.

The fifth locus associated with endotoxemia is next to LIPC, encoding hepatic lipase. Hepatic lipase also has plausible links to the kinetics of the contact activation cascade, because it is characterized by its ability to bind heparin and heparan sulfate in the endothelium. Heparin or heparan sulfate interact with antithrombin (III), potentiating its inhibitory effect on both FXII and FXI activity. Most important, however, genetic variation in LIPC affects lipoprotein particle distribution and composition, which might have a direct effect on endotoxemia. Hepatic lipase is recognized in HDL metabolism and reverse cholesterol transport by hydrolyzing phospholipids and triglycerides, resulting in particles that are more susceptible to clearance. Most of the lipopolysaccharide activity in the circulation is bound to lipoproteins, especially HDL, which contributes to the neutralization of lipopolysaccharide activity. However, lipoprotein distribution is different during inflammation, infection, or metabolic diseases. , Therefore, detoxification of lipopolysaccharide and the following net endotoxemia is dependent on the inflammatory and metabolic state, lipoprotein and apolipoprotein profile, and concentrations of lipopolysaccharide binding proteins, which may all be disturbed in stroke.

We also measured endotoxemia using different techniques, Endolisa and a mass spectrometry–based method. The former technique is based on the binding of lipopolysaccharide to a recombinant bacteriophage protein, and the latter quantifies the most abundant hydroxylated fatty acid of the lipid A moiety of most lipopolysaccharide molecules. The genetic associations seem to be restricted to the biological activity of lipopolysaccharide determined by the LAL assay, because with the 2 other methods we did not find significant associations with the lead SNP (rs1801020). Indeed, correlations between the results obtained with different methods were only modest. It is known that structural variations and enzymatic modifications of lipopolysaccharide molecules lead to differing immunologic responses. In relation to systemic inflammation, the level of endotoxin activity has been considered a more important determinant than the total endotoxin mass.

The observed associations between lead SNPs and coagulation‐related end points should be interpreted with caution because of multiple testing of various end points and the 5 lead SNPs. Another obvious limitation of the work is that we were not able to target the mechanisms behind the observed associations. Traditionally, endotoxemia has been considered an acquired characteristic of an individual, but as the mechanisms of the endotoxemia‐disease associations are still nonconclusive, a more complex connection is probable. This would include, as proposed in this study, underlying genetics, which can act as a mutual risk factor for both disease incidence and mechanisms affecting lipopolysaccharide processing or access into the circulation.

The current data convincingly show that the genetic variation makes a highly significant contribution to endotoxemia. Assuming that genetic factors can modify the translocation or the neutralization of endotoxins, the results suggest a novel part in the puzzle of host/microbiome interactions. More important, the results further characterize the concept of tight interaction between immunity and coagulation.

Sources of Funding

The project was supported by grants from the Academy of Finland (No. 1266053 to Dr Pussinen, No. 299200 to Dr Sandholm, and Nos. 265240, 263278, 308248, 312073, 100499, 205585, 118555, and 141054 to Dr Kaprio), the Sigrid Juselius Foundation (to Dr Pussinen), the Paulo Foundation (to Dr Pussinen), the Päivikki and Sakari Sohlberg's Foundation (to Dr Pussinen), the Finnish Dental Society Apollonia (to Dr Pussinen), and Finnish Foundation for Cardiovascular Research (to Dr Salomaa). The FinnDiane study was supported by grants from Folkhälsan Research Foundation, the Wilhelm and Else Stockmann Foundation, the Liv och Hälsa Foundation, Helsinki University Central Hospital Research Funds, the Novo Nordisk Foundation (NNFOC0013659/PROTON), European Foundation for the Study of Diabetes Young Investigator Research Award funds, and the Academy of Finland (Nos. 275614 and 316664). Genotyping of the FinnDiane genome‐wide association study (GWAS) data was funded by the Juvenile Diabetes Research Foundation within the Diabetic Nephropathy Collaborative Research Initiative (grant 17‐2013‐7), with GWAS quality control and imputation performed at the University of Virginia. Phenotype and genotype data collection in the twin cohort has been supported by the Wellcome Trust Sanger Institute, the Broad Institute, ENGAGE—European Network for Genetic and Genomic Epidemiology, FP7‐HEALTH‐F4‐2007 (No. 201413), the National Institute of Alcohol Abuse and Alcoholism (Nos. AA‐12502, AA‐00145, AA15416, and K02AA018755), and the National Heart, Lung, and Blood Institute (HL104125). GTEx Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by National Cancer Institute, National Human Genome Research Institute, National Heart, Lung, and Blood Institute, National Institute on Drug Abuse, National Institute of Mental Health, and National Institute of Neurological Disorders and Stroke. The funding of the FinnGen project consists of 2 grants from Business Finland (HUS 4685/31/2016 and UH 4386/31/2016) and support from 11 industry partners: AbbVie, AstraZeneca, Biogen, Celgene, Genentech (part of Roche), GSK, Janssen, Maze Therapeutics, MSD, Pfizer, and Sanofi. The MEGASTROKE project received funding from sources specified at http://www.megastroke.org/acknowledgments.html.

Disclosures

Dr Groop has received investigator‐initiated research grants from Eli Lilly and Roche; is an advisory board member for AbbVie, AstraZeneca, Boehringer Ingelheim, Cebix, Eli Lilly, Janssen, Medscape, Merck Sharp & Dohme, Novartis, Novo Nordisk, and Sanofi; and has received lecture fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Elo Water, Genzyme, Merck Sharp & Dohme, Medscape, Novo Nordisk, and Sanofi. Dr Salomaa has received honoraria for consultations from Novo Nordisk and Sanofi. He also has an ongoing research collaboration with Bayer Ltd (all unrelated to the present study). The remaining authors have no disclosures to report.

Appendix S1

Data S1

Tables S1–S11

Figures S1–S7

References 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.