|Home | About | Journals | Submit | Contact Us | Français|
Inflammation causes vascular dysfunction and perpetuates proatherosclerotic processes. We hypothesized that a broad panel of inflammatory biomarkers and single nucleotide polymorphisms (SNPs) in inflammatory genes are associated with vascular stiffness.
We assessed 12 circulating inflammatory biomarkers [C-reactive protein (CRP), fibrinogen, interleukin-6, intercellular adhesion molecule-1, lipoprotein-associated phospholipase-A2 (Lp-PLA2 mass and activity), monocyte chemoattractant protein-1, myeloperoxidase, CD40 ligand, osteoprotegerin, P-selectin, tumor necrosis factor receptor II (TNFRII)] in relation to tonometry variables [central pulse pressure (CPP), mean arterial pressure (MAP), forward pressure wave, reflected pressure wave (RPW), carotid-femoral pulse wave velocity (CFPWV), augmentation index] measured in 2409 Framingham Heart Study participants (mean age 60 years, 55% women, 13% ethnic/racial minorities). SNPs (n=2195) in 240 inflammatory candidate genes were related to tonometry measures in 1036 white individuals.
In multivariable analyses, biomarkers explained less than 1% of any tonometry measure's variance. Applying backwards elimination, markers related to tonometry (P<0.01) were: TNFRII (inversely) with MAP; CRP (positively) and Lp-PLA2 (inversely) with RPW; and interleukin-6 and osteoprotegerin (positively) with CFPWV. In genetic association analyses, lowest p-values (false discovery rate <0.50) were observed for rs10509561 (FAS), p=6.6×10−5 for CPP and rs11559271 (ITGB2), p=1.1×10−4 for MAP.
These data demonstrate that in a community-based sample, circulating inflammatory markers TNFRII (MAP), CRP, Lp-PLA2 activity (RPW), interleukin-6 and osteoprotegerin (CFPWV) were significantly but modestly associated with measures of arterial stiffness and wave reflection. Additional studies are needed to determine if variation in inflammatory marker genes are associated with tonometry measures.
Arterial stiffness has been associated with increased cardiovascular morbidity and mortality.1 The clinical correlates of increasing vascular stiffness are advancing age, sex, body mass index,2 hypertension,3 diabetes mellitus, and smoking.4 The clinical applications and importance of arterial stiffness and wave reflection in cardiovascular disease assessment recently has been reviewed.5 Experimental and human evidence suggests that inflammation may contribute to vascular stiffness. For instance, autoimmune diseases are accompanied by increased arterial stiffness and premature cardiovascular disease.6 Current research suggests causal relations between acute inflammatory states induced by vaccination and impaired vascular function detected by reversible increases in pulse wave velocity (PWV).7 In addition, cross-sectional human studies have reported that C-reactive protein (CRP), interleukin-6 and tumor necrosis factor (TNF)-alpha concentrations are related to PWV,8–10 and arterial elasticity.11 Even in apparently healthy individuals large artery stiffness and central pulse pressure (CPP) are correlated with the extent of systemic inflammation.12,13
Recent studies suggest a potential role of genetic variation in tonometry response. Heritability for tonometry variables has been demonstrated, and linkage investigations identified several genomic regions for further research.14 Single nucleotide polymorphisms (SNPs) in genes coding for proteins such as the renin-angiotensin-aldosterone system,15 endothelial nitric oxide synthase,16 and inflammatory mediators (e.g. CRP) have been reported to be related to arterial stiffness measures.17
We sought to examine associations between systemic inflammatory markers and SNPs with vascular stiffness measures in a community-based cohort. We hypothesized that one or more of a broad range of inflammatory biomarkers are related to non-invasive measures of arterial stiffness and wave reflection. Furthermore, we tested the hypothesis that SNPs in inflammatory candidate genes are associated with tonometry measures, in part due to the effect of SNPs on biomarker concentrations.
The Offspring cohort was recruited in the 1970s and has been examined routinely every 4 to 8 years.18 Multi-ethnic Omni participants, were recruited in the 1990s.19 Participants attending examination 7 (n=3537) or Omni examination 2 (n=405) were eligible for analyses. Vascular tonometry data and inflammatory markers were available in 2095 Offspring participants and 314 Omni participants (for detailed exclusions see supplement). Genotype data were available on 1036 Offspring participants. The Boston University Medical Center Institutional Review Board approved study protocols.
Non-invasive measures of arterial stiffness and wave reflection were assessed in supine participants as previously described, blinded to clinical and biomarker data (see Supplement for further details).20
Twelve circulating markers, representing different inflammatory pathways, were selected a priori: including CD40 ligand, CRP, fibrinogen, intercellular adhesion molecule-1, interleukin-6, lipoprotein-associated phospholipase-A2 (Lp-PLA2) activity and mass, monocyte chemoattractant protein-1, myeloperoxidase, osteoprotegerin, P-selectin, tumor necrosis factor receptor-II (TNFRII). Details of specimen type, measurement kit and reproducibility are provided in Table S1 (http://hyper.ahajournals.org).
Genotyping of common SNPs was conducted by Perlegen Sciences, Inc. (Mountain View, Ca; 232 genes, 2942 SNPs) and the Broad Institute of Harvard and Massachusetts Institute of Technology (9 genes, 125 SNPs) using methods described in the Electronic Supplement. Inflammatory candidate genes were selected by Framingham investigators, guided by criteria outlined in Table S2. A total of 2195 SNPs in 240 genes passed quality control.
We used natural log transformed biomarkers. We conducted multivariable linear regression to relate tonometry measures (dependent variables) to circulating biomarkers, adjusting for cohort, sex, age, age2, MAP (except if MAP was the dependent measure), heart rate, height, weight, total/high density lipoprotein cholesterol, glucose, diabetes, smoking, prevalent cardiovascular disease, hormone replacement therapy, hypertension treatment, aspirin (≥3 days/week), and lipid-lowering medication. Tonometry measures associated with the inflammatory marker panel with a global p<0.01, were studied further. Forcing in clinical covariates, we selected biomarkers associated with the respective tonometry measure with backward elimination models (p<0.01 for inclusion).
Potential variation due to major confounders of tonometry variables was accounted for by the creation of multivariable-adjusted residuals (age, age2, sex, height, and weight). We examined tonometry residuals in association with inflammatory SNPs using analysis of variance with a general genetic model (2 degrees of freedom); if the lowest frequency genotype category had <10 individuals, nonparametric Kruskal-Wallis tests were performed. Mean residual tonometry measures and beta coefficients were computed for each genotype. In addition, we calculated within phenotype q-values. We used q-value<0.50 as a threshold indicative of potentially important associations.21 The q-value represents the expected proportion of false positive associations among tests exhibiting the specified level of statistical significance, and is less conservative than Bonferroni adjustments.
Because of numerous SNP, tonometry and biomarker measures we were concerned about multiple testing, but were aware that different research groups have assessed various tonometry and biomarker measures. To maximize results disclosure, yet modestly reduce multiple testing penalties, we a priori specified secondary phenotypes. Table S3 displays biomarker and tonometry characteristics of primary (6 tonometry and 12 biomarker) and secondary (2 tonometry and 4 biomarker [available on a subset]) measures. We provide secondary analyses of Pearson partial correlation coefficients adjusted for age, age2, sex, and cohort for individual biomarkers (12 primary and 4 secondary) and tonometry variables (dependent measures; 6 primary, in Table S4, and 2 secondary tonometry measures Table S5). We tested final models for interactions with hypertension (3 levels: hypertension treated, hypertension untreated and no hypertension), lipid lowering treatment, and hormone replacement therapy.
To explore Mendelian randomization,22 we examined biomarkers that were retained in backward elimination models relating biomarker concentrations to tonometry measures. The association of SNPs significantly associated with each such biomarker (i.e. SNPs both cis [within] and trans [outside] the gene coding the marker) were assessed in relation to the same tonometry measures with which the biomarker was significantly related.
SAS version 8.1 (Cary, North Carolina, http://www.sas.com/presscenter/guidelines.html) was used for regression analyses and for creating phenotype residuals for genetic analyses. R was used for genetic analyses (www.r-project.org).
Clinical characteristics by study sample are presented in Table S6. Briefly, the phenotype sample included Omni ethnic minority participants and was slightly younger compared with the genotype sample, which consisted only of white Offspring cohort participants. Tonometry measures and inflammatory marker concentrations are provided in Table S3.
In multivariable-adjusted analyses, the inflammatory markers as a group were significantly related to three tonometry measures: MAP, RPW and CFPWV (global P<0.01). Therefore, we conducted backward elimination of the 12 inflammatory biomarkers in relation to the three tonometry measures, adjusting for 17 potential confounders (Table 1). TNFRII was inversely associated with MAP; CRP (positive) and lipoprotein-associated phospholipase-2 (Lp-PLA2) activity (inverse) were related to RPW; interleukin-6 and osteoprotegerin were positively associated with CFPWV. After accounting for 17 potential clinical confounders, the inflammatory markers explained less than 1% of the tonometry measures' variability (partial R2 ranged from 0.69% for RPW to 0.94% for MAP).
We examined the relations of 2195 SNPs in 240 inflammatory candidate genes to tonometry variables. The top five SNPs associated with each tonometry variable are displayed (Table 2; full disclosure available at FHS inflammation website [www.inflammation-framinghamheartstudy.org]). We observed that SNP rs10509561 in FAS (previously TNFRSF6), p=6.6×10−5 was associated with CPP. For SNP rs10509561 heterosis was seen with the lowest geometric mean for the heterozygote (standardized residual mean −0.13 (±0.09) compared to means of 0.06 (±0.04) in the major allele and 0.27 (±0.11) in the minor allele homozygote). Rs10509561 in FAS was also among the top SNPs for forward pressure wave (p=1.2×10−3) and RPW (8.9×10−3; See Figure S1). In addition, SNP rs11559271 in ITGB2 gene, p=1.1×10−4 was associated with MAP. Only seven SNP biomarker associations showed q-values of less than 0.5.
We observed modest correlations for pairwise comparisons of tonometry variables with inflammatory markers (partial correlation coefficients ranging from −0.11 to 0.17; Table S4). An unanticipated observation was that two of the markers selected with backwards elimination in Table 1 in the 17 covariate-adjusted models were not significantly associated with the tonometry measure if examined individually (secondary analyses). As displayed in Table S4, with the model adjusting for four covariates (age, age2, sex and cohort), TNFRII concentrations were not associated with MAP (r=0.00, p=0.82), and CRP concentrations were not associated with RPW (r=0.02, p=0.43). We conducted post-hoc exploratory analyses to understand which covariate(s) affected the TNFRII-MAP association; hypertension treatment and hormone replacement therapy were the factors most responsible for rendering the TNFRII-MAP relation significant in the 17 covariate-adjusted multivariable model.
Using a p<0.01, we observed potential interactions between several inflammatory markers and hypertension category. Specifically the association between Lp-PLA2 and RPW, and between IL-6 and CFPW was stronger in those with compared with those without hypertension. Individuals taking lipid lowering medication (vs. those without) had a more positive slope between IL6 and CFPWV (Table S7).
For biomarkers significantly associated with tonometry measures (Table 1), we subsequently examined whether SNPs associated with specific biomarker concentrations, were also associated with the respective tonometry measure (Table S8) at p≤0.05, to test the “Mendelian randomization” concept. Top associations were observed for TNFSF15 SNP rs10817678 for osteoprotegerin concentrations (p=0.003) and CFPWV (p=0.02), and for rs6586166 in the FAS gene in relation to interleukin-6 (p=0.009) and CFPWV (p=0.02). The SNP rs1051931 in the PLA2G7 gene was associated with Lp-PLA2 activity (p=0.0001), but the association with RPW was p=0.054.
In a community-based cohort 12 circulating inflammatory biomarkers representing different pathophysiological pathways and 2195 SNPs in 240 inflammation-related candidate genes were related to tonometric measures of arterial stiffness and wave reflection. The circulating inflammatory biomarker panel revealed significant associations between interleukin-6 and osteoprotegerin with CFPWV; CRP and Lp-PLA2 activity (inverse) with RPW; and TNFRII with MAP (inverse). After accounting for potential clinical confounders, the inflammatory markers explained minor additional variability in the tonometry measures (<1%). Genetic findings potentially interesting for follow-up were observed in both pulsatile (CPP) and steady-flow (MAP) measures of vascular function, which were associated with SNPs in the inflammatory genes FAS and ITGB2.
We pre-specified a conservative strategy of analyzing the biomarkers as a group, and conducting multivariable models adjusting for 17 potential confounders. An unanticipated finding was that two selected biomarkers, TNFRII and CRP, were not significantly associated with the tonometry measure in individual biomarker models adjusting for only four potential confounders (age, age2, sex and cohort). We acknowledge several alternative interpretations of the apparent inconsistency. One possibility is that TNFRII-CPP, and CRP-RPW associations represent false positive findings due to multiple testing. Another explanation is that the 17 covariate model `over-adjusted', by including potential confounders that pathophysiologically serve as intermediate mechanisms. For instance, based on prior longitudinal studies, one speculation is that inflammation contributes to the development of hypertension,23 and diabetes.24 Hence, adjusting for diabetes or hypertension may be inappropriate if inflammation leads to diabetes and hypertension, which etiologically contribute to the development of arterial stiffness.
Recent investigations have demonstrated that arterial stiffness3 and inflammation are both crucial factors in cardiovascular pathology and aging. Prior studies have related CRP and interleukin-6 to conduit vessel distensibility and arterial stiffness.7,8,12,13,25 It has been shown that osteoprotegerin regulates vascular morphology and function in interaction with the immune system and may be responsible for vascular calcification.26 In animal and in vitro models osteoprotegerin has a favorable effect on arteries, yet higher osteoprotegerin was associated with high CFPWV in our data. In accordance with these findings, in human studies, osteoprotegerin concentrations have been associated positively with systolic blood pressure and brachial-ankle PWV27 and are indicators of cardiovascular disease risk and mortality,28 which might be due to compensatory over-expression of osteoprotegerin in an imbalanced system. TNFRII and TNF-alpha were associated with aortic and brachial-ankle PWV.6,29 We found a modest relation to RPW (CRP), and CFPW (osteoprotegerin and interleukin-6) but not to CPP and MAP. After accounting for multiple testing, none of the markers was significantly related to augmentation index. A lack of association of augmentation index with CRP concentrations has been reported previously.12
Among the top associations of SNPs in inflammatory genes was FAS rs10509561 with CPP, which also was related to forward pressure wave and RPW. FAS codes for widely expressed membrane receptors, which belong to the TNF super-family. FAS protein, and its ligand, induce apoptotic cell death and are integral to immune processes.30 In addition, Fas signaling has been reported to regulate blood pressure and endothelial function through the modulation of eNOS expression in mice,31 and therefore might have a pathophysiological relation to arterial stiffness2 We acknowledge that biological plausibility is only one criteria for establishing a causal relation; clearly our findings will need to be replicated in other studies and with other study designs.
Unreplicated reports have been published relating inflammatory candidate genes to tonometry measures. We were not able to replicate genetic findings reported for initial associations of CRP gene SNP rs1800947 with brachial-ankle PWV.32 In the Rotterdam study no relevant relations of CFPWV and pulse pressure to common polymorphisms of TGFB1 gene could be demonstrated,33 similar to our data.
We hypothesized that inflammatory markers would be related to tonometry measures and sought to test whether SNPs significantly associated with the related inflammatory markers would also explain variability in tonometry measures. According to the theory of Mendelian randomization, an association of genetic variation with an inflammatory marker and a vascular phenotype would provide suggestive evidence of a causal biomarker-vascular phenotype relation. Genotype-vascular phenotype associations would be less likely to be secondary to reverse causation and residual confounding.23,34 Most of the biomarker-SNP associations were in genes not coding for the respective biomarker and have, to our knowledge, not been investigated in association with biomarker concentrations. Of note, the two top SNPs for Lp-PLA2 activity and RPW are in the PLA2G7 gene and were significantly associated with Lp-PLA2 activity in the Framingham Offspring cohort and in prior studies.35 In our exploratory analyses we were not able to show strong evidence for Mendelian randomization for the inflammatory biomarkers examined (including Lp-PLA2). We acknowledge that many of the conditions required to infer causality under Mendelian randomization were not met in the present project.36,37 The lack of strong confirmation of an association between inflammatory marker SNPs, biomarkers and tonometry phenotypes in our data may be explained by the fact that SNPs in the corresponding genes neither explained substantial variability in the biomarker nor in tonometry variables.
The routine measurement of a broad inflammatory biomarker panel representing diverse pathways, systematically ascertained tonometry data performed in accordance with strict quality control protocols, well-characterized cardiovascular risk factors enabling multivariable models, carefully pre-specified analytical approaches with a priori designation of covariates and primary versus secondary models, and the large community-based cohort limiting referral bias constitute study strengths. It should be noted that compared with prior studies, we did not observe very strong associations between tonometry measures and inflammatory biomarker concentrations or SNPs. Most prior studies were smaller (n=78–39110,11,38–41), were referral-based (selected for vascular conditions42,43) and included only small numbers' of inflammatory biomarkers and vascular measures apart from blood pressure,9,12,44–47 and tested only limited numbers of SNPs.32,33
Limitations of our study must be noted. A middle-aged to elderly study cohort may not be optimal to dissect genetic associations which might be found to be more prominent in younger samples. Clinical factors like age, sex, MAP, height and heart rate account for up to 50% of the tonometry measures variability.2 It might be argued that the relative contribution of the biomarkers and SNPs toward explaining tonometric variability was low, because we accounted for variability related to potential confounders. In attempting to conservatively adjust for a broad range of potential confounders, as noted above, we acknowledge that we potentially over-adjusted, if some of the clinical covariates included in the multivariable models actually represent intermediate mechanisms. Non-random exclusions for nursing home and home visit participants could have introduced confounding because these subjects suffer from illnesses that may affect inflammatory biomarkers and tonometry measures. We recognize that experts may disagree on the choice of inflammatory biomarkers and candidate genes selected for such a study. However, we sought to cover different inflammatory pathways that have been reported to be of central importance in vascular pathophysiology.
Although testing a broad panel of inflammatory biomarkers and SNPs is an asset of our study, we concede that the breadth of markers and SNPs studied introduces substantive concerns about multiple testing. We cannot exclude the possibility that the reported associations may be due to chance. To reduce the probability of false positives we restricted our analyses to 6 key measures of arterial stiffness, and introduced more stringent thresholds for statistical significance than the typical p<0.05 as applied in other Framingham multi-inflammatory biomarker projects.48,49 However, we emphasize that our clinical and genetic findings will need to be replicated. Conversely, because we accounted for measuring multiple biomarkers, SNPs and tonometric measures we cannot exclude the possibility of false negatives; minor associations may have been missed. A strength of our study is that we did not selectively report only positive findings. We provide investigators with a web-based resource of all primary, secondary and exploratory results. We measured circulating inflammatory biomarkers as a surrogate for systemic vascular inflammation; we cannot exclude the possibility that local vascular inflammation may be causally related to local arterial stiffness without being manifest in systemic biomarker concentrations. Similarly, the clinical analyses were cross-sectional; we cannot infer the temporality of the observed associations. Although we modeled the association of inflammation to tonometry (dependent measure), it is equally plausible that arterial stiffness could lead to enhanced vascular shear and systemic inflammation.
Arterial stiffness and wave reflection are risk indicators for cardiovascular disease and mortality. Robust experimental and epidemiologic evidence has shown that inflammatory pathways are implicated in vascular remodeling and disease. However, whether inflammatory processes are causal or are surrogate markers for vascular remodeling is still under investigation. In our study, although modest, we demonstrated associations in inflammatory markers carefully selected to represent different pathophysiological pathways and SNPs in inflammatory genes with tonometry variables. Candidate genes were chosen in a thorough review of known evidence using current literature and genetic databases. Yet, our findings must be replicated. Successful replication of the tonometry-inflammatory SNPs association would provide additional motivation for further research into the mechanisms initiating and perpetuating the contribution of vascular inflammation to vascular remodeling and stiffness. Whereas, inflammatory biomarkers and SNPs contribute only minor amounts to predicting tonometric indices, we submit that the associations after accounting for 17 covariates suggests that they may play a pathophysiological role in vascular stiffness. Although no direct clinical implications presently should be derived from the results, if our data are replicated they might spur further investigations into the role of inflammatory markers and SNPs in risk stratification, and potential targeted therapeutic interventions.
Sources of funding Supported by NIH/NHLBI contract N01-HC-25195 and NIH grants HL60040, HL064753. HL076784, AG028321 (EJB), U01-HL66582 (EJB); HL080124, HL077447 and HL71039 (RSV). NIH Research career award HL04334 (RSV); American Diabetes Association Career Development Award and NCRR GCRC M01-RR-01066 (JBM); Reynolds Foundation (DL); NIH HL080025, Doris Duke Charitable Foundation Clinical Scientist Development Award and Burroughs Wellcome Fund Career Award for Medical Scientists (CNC).
Lp-PLA2 measurements were provided by GlaxoSmithKline at no cost.
Disclosures Dr Mitchell is owner of Cardiovascular Engineering, Inc., which designs and manufactures vascular stiffness measurement devices. Other authors report no conflicts.