OUP user menu

C-reactive protein genotypes associated with circulating C-reactive protein but not with angiographic coronary artery disease: the LURIC study

Tanja B. Grammer, Winfried März, Wilfried Renner, Bernhard O. Böhm, Michael M. Hoffmann
DOI: http://dx.doi.org/10.1093/eurheartj/ehn191 170-182 First published online: 22 May 2008

Abstract

Aims Circulating C-reactive protein is associated with future cardiovascular events. The causal role of C-reactive protein in the development of atherosclerosis remains controversial.

Methods and results We analysed the association between three genetic polymorphisms (PM) (−717C>T, rs2794521; +1059G>C, rs1800947; +1444C>T, rs1130864) at the C-reactive protein locus and related haplotypes with both circulating C-reactive protein and angiographic coronary artery disease (CAD). The concentration of C-reactive protein was similar in patients with stable CAD and in controls, but increased in patients presenting with acute coronary syndromes. In models adjusting for the main confounding variables, the minor alleles of the +1059G>C (rs1800947) and the +1444C>T PM (rs1130864) were associated with decreased and increased concentrations of C-reactive protein, respectively. Haplotypes 1 and 4 decreased, and haplotype 2 increased C-reactive protein, whereas haplotype 3 had no appreciable effect. None of the genetic variants affecting circulating C-reactive protein was consistently associated with the prevalence of angiographic CAD.

Conclusion A causal role of C-reactive protein in the development of CAD would require that genetic PM resulting in long-term modulation of the concentration of C-reactive protein be themselves associated with CAD. We were not able to detect such a relationship, which can be attributed to either a very small genetic effect size or the relationship between C-reactive protein and cardiovascular events may reflect confounding and reverse causation.

Keywords
  • C-reactive protein
  • Genetic PM
  • Haplotypes
  • Inflammation
  • Coronary artery disease

Atherosclerosis shares many characteristics of chronic inflammatory diseases. The acute phase reactant C-reactive protein predicts future cardiovascular events.13 C-reactive protein is produced in the liver in response to pro-inflammatory cytokines such as interleukin 6 which is released from activated cells at the site of inflammation.4 It is currently not clear whether an increased concentration of C-reactive protein merely reflects the inflammatory process smouldering within atherosclerotic lesions or whether C-reactive protein assumes an active role in the development of atherosclerosis.

Potentially, polymorphisms (PM) in a gene affecting its expression may provide a clue to the importance of the particular gene product to the development of a disease phenotype. Thus, a causal role of C-reactive protein in atherosclerosis would be supported if PM at the C-reactive protein gene would consistently be related to both circulating C-reactive protein levels and atherosclerosis. Basically, this is also the principle consideration underlying the strategy of Mendelian randomization.5 It implies that—unlike the relationship between an ‘exogenously’ modifiable, putatively causal factor and a disease—the association between functional genetic variants and a disease is insensitive to confounding and reverse causation.

Family and twin studies suggest that genetic factors may account for as much as 40% of the variance in plasma C-reactive protein levels.68 Indeed, single-nucleotide polymorphisms (SNPs) and haplotypes at the C-reactive protein gene have been identified and some of them have successfully been linked to the concentration of C-reactive protein.926

Experimental evidence has emerged that C-reactive protein may act as a pro-coagulant2730 and vasoconstrictor.3133 Consequentially, attempts have been made to link genetic PM associated with high C-reactive protein concentrations to self-reported coronary artery disease (CAD),21 the incidence rate of cardiovascular endpoints,22,24,26 myocardial infarction,16 arterial thrombosis,10 non-fatal myocardial infarction,34 carotid intima media thickness,22 or carotid artery compliance.25 Most of these studies did not establish a relationship between PM at the C-reactive protein gene and atherosclerosis-related phenotypes,10,16,24,26,34 whereas a few did, although in ethnic subgroups only21,22 or using a surrogate maker of atherosclerosis such as the carotid artery compliance.25

Other series of experimental studies have implicated C-reactive protein directly in the development and progression of atherosclerosis. C-reactive protein may be produced by smooth muscle cells and macrophages,35,36 contribute to the development of endothelial function,3739 induce the expression of adhesion molecules40,41 and angiotensin type I receptors,42 and enhance the endocytosis of LDL, either native43 or after oxidative modification.44

It is surprising in that respect that the relationship between PM at the C-reactive protein gene and the atherosclerotic burden assessed by coronary angiography has not been examined. For this reason, we set out to study the association between three genetic PM at the C-reactive protein locus and inferred haplotypes on the one hand and circulating levels of C-reactive protein and the prevalence of CAD on the other hand in a large cohort of individuals who underwent coronary angiography. We included the following PM in this study: −717C>T (rs2794521) which is located in the promoter,11,14,16,20,45 +1059G>C (rs1800947), a synonymous PM within the coding region,10,1316,18,20 and +1444C>T (rs1130864) which is in the 3′ untranslated region.11,16,20,46

Methods

Study design and participants

We studied participants of the LUdwigshafen RIsk and Cardiovascular Health (LURIC) study recruited between June 1997 and May 2001.47 Inclusion criteria were: German ancestry, clinical stability except for acute coronary syndromes, and the availability of a coronary angiogram. The indications for angiography in clinically stable individuals were chest pain and/or non-invasive test results consistent with myocardial ischaemia. Individuals suffering from acute illness other than acute coronary syndromes, chronic non-cardiac diseases, or malignancy within the past 5 years and those unable to understand the purpose of the study were excluded. The study was approved by the ethics committee at the ‘Ärztekammer Rheinland-Pfalz’. Informed written consent was obtained from all participants.

CAD was assessed by angiography with maximum luminal narrowing estimated by visual analysis. Clinically relevant CAD was defined as the occurrence of at least one stenosis of 20% or more in one or more of 15 coronary segments. Individuals with stenoses <20% were considered as not having CAD. We also provisionally defined CAD as the presence of at least one stenosis of 50% or more, again classifying individuals with stenoses <20% as controls. Further, we stratified the study participants according to the severity of CAD. This was accomplished by coding normal coronary vessels, 20–50% lesions in one to three vessels, at least one lesion >50%, or lesions >50% in two or three vessels as 0,1,2, or 3, respectively.

The metabolic syndrome was defined as recommended by the National Cholesterol Education programme.48 Diabetes mellitus was diagnosed if plasma glucose was >1.25 g/L in the fasting state or >2.00 g/L 2 h after an oral glucose load,49 or if individuals were receiving anti-diabetic treatment. Hypertension was diagnosed if the systolic and/or diastolic blood pressure exceeded 140 and/or 90 mm Hg or if there was a history of hypertension, evident through the use of antihypertensive drugs.

Measurements of C-reactive protein and of three PM at the C-reactive protein locus were complete in 3252 of 3279 individuals with coronary angiograms; out of the 3252 individuals with complete measurements, 626 presented with unstable angina, 114 with non-ST-elevation myocardial infarction (Troponin T > 0.1 µg/L), and 289 with ST-elevation myocardial infarction (Troponin T > 0.1 µg/L).

Clinical chemistry

Fasting blood samples were obtained by venipuncture in the early morning. ‘Sensitive’ C-reactive protein was measured by immunonephelometry on a Behring Nephelometer II (N High Sensitivity CRP, Dade Behring, Marburg, Germany) after completion of the patient recruitment in 2001 in samples stored at −80°C. In the C-reactive protein assay used, the limit of detection for C-reactive protein is 0.17 mg/L; it is linear up to 500 mg/L. The lowest and the highest C-reactive protein concentrations encountered in this study were 0.17 and 269 mg/L, respectively. Blood glucose was determined enzymatically using the hexokinase/glucose-6-phosphate dehydrogenase method (Roche Diagnostics, Mannheim, Germany). Lipoproteins were separated by a combined ultracentrifugation-precipitation method (β-quantification).47,50 Cholesterol and triglycerides were measured with enzymatic reagents from WAKO (Neuss, Germany) on a WAKO R30 or Olympus AU640 analyser.47 Glomerular filtration rate (GFR) was calculated as GFR (mL/min/1.73 m2) = 186×creatinine−1.154×age−0.203 and GFR (mL/min/1.73 m2) = 138 ×creatinine−1.154×age−0.203 in males and in females, respectively.51

C-reactive protein-polymorphisms

Genomic DNA was prepared from EDTA anticoagulated blood by using a salting-out procedure. The SNPs were genotyped by polymerase chain reaction and restriction fragment-length analysis, using the following primer pairs and enzymes: CRP P1 -717 (5′-gtt ccc ctt cct gtg tcc aag ta-3′), CRP P2−717 (5′-act gga ctt tta ctg tca ggg c-3′), and DraIII for rs2794521; CRP P1+1059 (5′-ttt tac agt ggg tgg gtc tg-3′), CRP P2+1059 (5′-aac act tcg cct tgc act tc −3′), and BsiHKI for rs1800947; CRP P1+1444 (5′-gtg tct ggt ctg gga gct cgt ta −3′), CRP P2+1444 (5′-ctt ctc agc tct tgc ctt atg agt −3′), and HpyCH4III for rs1130864. The following fragments were obtained: rs2794521 T allele 172 bp, C allele 109 bp+63 bp; rs1800947 G allele 276 bp+99 bp, C allele 375 bp; rs1130864 C allele 156 bp+39 bp, T allele 195 bp. To validate the method three samples of each genotype were sequenced; as internal control 184 DNA samples, randomly distributed within the samples, were genotyped twice. To confirm genotype assignment, the restriction fragments were analysed independently by two scientists on two separate occasions. The results were scored blinded as to case–control status.

Statistical analysis

Continuous variables were assessed for normality by the inspection of histograms. C-reactive protein and triglycerides were not normally distributed and therefore transformed logarithmically before being used in parametric statistical procedures. Clinical and anthropometric characteristics were compared between CAD patients and controls by analysis of variance (ANOVA) or logistic regression using co-variables as indicated (Table 1). We studied the effect of the CAD status, sex, age, and risk factors on C-reactive protein using ANOVA models in which we included those factors not under examination as co-variables (Table 2). Haplotypes (Table 3) were inferred using the PHASE 2.0 software.52 To examine the effect of PM and haplotypes on C-reactive protein, we used ANOVA with co-variables as indicated (Table 4). In models assuming a co-dominant (additive) effect of the alleles, genotypes were coded as 0, 1, and 2, respectively, and genotypes were either treated as interval-scaled or categorical variables, the most frequent genotype being considered as the reference category in the latter case. When assuming a dominant effect, the most frequent genotype was coded as 0, and the combined remaining ones were coded as 1. When assuming a recessive effect, the least frequent genotype was coded as 1, the combined other ones were coded as 0. When looking at the effect of haplotypes, the numbers of a given haplotype per individual was used as the independent variable, also allowing for co-dominant, dominant, and recessive effects. The assumption for using ANOVA that the residuals are normally distributed was examined using plots of observed vs. predicted values. In none of the analyses presented in Tables 2 and 4 did we obtain any indication that this assumption was violated. Further, the estimated marginal means of the dependent variables along with their 95% confident intervals (CI) are reported in the ANOVA procedures and the least significant difference t-test was used for post hoc comparisons. Estimated marginal means are not observed means; rather they represent predicted means estimated at the co-variables held at their respective actual means. Finally, we analysed the association between C-reactive protein genotypes or haplotypes and angiographic CAD in an analogous fashion by logistic regression (Table 5). Multivariable adjustment was carried out in two steps, first for sex and age, and then, in addition, for cardiovascular risk factors (body mass index, diabetes mellitus, hypertension, smoking, LDL-C, HDL-C, logarithmically transformed triglycerides, GFR). All statistical tests were two-sided. P < 0.05 was considered statistically significant. The SPSS 14.0 statistical package (SPSS Inc., Chicago, IL, USA) was used.

View this table:
Table 1

Clinical and biochemical characteristics of coronary artery disease patients and controls

ControlsCADP-valuea
MenWomenMenWomen
(n = 364)(n= 333)(n= 1909)(n= 646)
Age (years) means ± SD55 ± 1262 ± 1063 ± 1066 ± 10<0.001b
Body mass index (kg/m2) means ± SD27 ± 427 ± 528 ± 427 ± 50.392
Diabetes mellitus, (%)63 (17)64 (19)657 (34)253 (39)<0.001
Systemic hypertension, (%)212 (58)292 (69)1046 (74)519 (80)0.002
Past smoking, (%)158 (43)53(16)1083 (57)152 (24)<0.001
Current smoking, (%)82 (23)46 (14)411 (22)102 (16)<0.001
Systolic blood pressure (mm Hg) means ± SD135 ± 21138 ± 23142 ± 23143 ± 250.007c
Diastolic blood pressure (mm Hg) means ± SD81 ± 1279 ± 1182 ± 1180 ± 120.434c
Fasting blood glucose (g/L) means ± SD1.06 ± 0.301.04 ± 0.241.15 ± 0.341.19 ± 0.44<0.001
LDL cholesterol (g/L) means ± SD1.17 ± 0.311.24 ± 0.311.13 ± 0.331.23 ± 0.390.002d
HDL cholesterol (g/L) means ± SD0.40 ± 0.110.46 ± 0.120.36 ± 0.100.42 ± 0.11<0.001
Triglycerides (g/L) median (25th and 75th percentile)1.45 (1.03–2.12)1.28 (0.92–1.77)1.50 (1.12–2.01)1.53 (1.15–2.06)<0.001e
Glomerular filtration rate (mL/min)88 ± 1778 ± 1784 ± 1974 ± 180.837
  • aANOVA or logistic regression, respectively, adjusted for sex and age.

  • bANOVA adjusted for sex only.

  • cadditionally adjusted for use of beta blockers, ACE inhibitors, AT1 receptor antagonists, calcium channel blockers and diuretics.

  • dafter adjusting for the use of lipid lowering drugs (>97% statins) in addition to age and sex, estimated marginal means for LDL cholesterol are 113 and 118, respectively, in individuals without and with CAD (P = 0.002).

  • eANOVA of logarithmically transformed values.

View this table:
Table 2

Effect of confounding factors and clinical presentation on circulating CRP

nCRP (mg/L)aΔ (%)bP-valuec
SexEta2 = 0.005
 Men22733.2 (3.0–3.3)Ref
 Women9794.6 (4.2–5.0)+46.2<0.001
Age (years)Eta2 = 0.0090.009d
 <6012013.1 (2.8–3.3)Ref
 60–7011773.6 (3.4–3.9)+17.5<0.001
 >708744.2 (4.0–4.6)+37.0<0.001
r0.121<0.001
Coronary artery diseaseEta2 = 0.102<0.001d
 None6972.8 (2.5–3.0)Ref
 Stable CAD15262.9 (2.7–3.1)+5.40.363
 Unstable CAD (Troponin T−)6263.9 (3.6–4.3)+41.5<0.001
 NSTEMI (Troponin T+)11410.1 (8.1–12.5)+266.2<0.001
 STEMI (Troponin T+)28910.2 (8.9–11.7)+270.6<0.001
Body mass index (kg/m2)Eta2 = 0.005
 <26 or 27e15233.2 (3.0–3.4)Ref
 >26 or 27e17293.9 (3.7–4.1)+20.0<0.001
r0.138<0.001
Metabolic syndrome/DiabetesEta2 = 0.0090.009d
 None11353.3 (3.0–3.5)Ref
 Metabolic syndrome10803.5 (3.3–3.8)+7.30.239
 Diabetes mellitus10373.9 (3.6–4.2)+19.5<0.001
Hypertension
 No8863.7 (3.4–4.0)Ref
 Yes23663.5 (3.3–3.7)−4.60.338
SmokingEta2 = 0.024<0.001d
 Never11652.9 (2.7–3.1)Ref
 Former14463.6 (3.3–3.8)+23.1<0.001
 Current6415.1 (4.6–5.6)+76.8<0.001
LDL cholesterol (g/L)0.705d
 1st quartile (<1.00)10473.6 (3.4–3.9)Ref
 2nd quartile (1.00–1.19)7783.6 (3.3–3.9)−0.90.881
 3rd quartile (1.20–1.40)6973.6 (3.3–3.9)−1.60.784
 4th quartile (≥1.41)7303.4 (3.1–3.7)−6.30.267
r−0.0520.003
HDL cholesterol (g/L)Eta2 = 0.038<0.001d
 1st quartile (<0.34)12444.8 (4.5–5.1)Ref
 2nd quartile (0.34–0.41)8873.5 (3.2–3.8)−27.2<0.001
 3rd quartile (0.42–0.49)6442.7 (2.5–3.0)−44.2<0.001
 4th quartile (≥0.50)4772.4 (2.1–2.7)−50.5<0.001
r−0.282<0.001
Triglycerides (g/L)Eta2 = 0.007<0.001d
 1st quartile (<0.97)5724.2 (3.8–4.7)Ref
 2nd quartile (0.97–1.32)7653.9 (3.5–4.2)−8.90.155
 3rd quartile(1.33–1.94)10223.3 (3.1–3.6)−21.4<0.001
 4th quartile (≥1.95)8933.2 (2.9–3.5)−24.6<0.001
r0.0560.002
Glomerular filtration rate (mL/min)Eta2 = 0.010<0.001d
 ≥9010483.3 (3.0–3.5)Ref
 60–8918223.5 (3.3–3.7)+6.10.228
 <603825.0 (4.4–5.6)+52.7<0.001
r−0.141<0.001
  • aEstimated marginal means and 95% confidence intervals obtained in a general linear model (ANOVA) adjusted for each of the other factors and for the use of lipid-lowering drugs (>97% statins), whereby age, body mass index, LDL cholesterol, HDL cholesterol, triglycerides (log transformed), and glomerular filtration rate, were included as continuous rather than categorial co-variables. The estimated marginal means and confidence intervals of logarithmically transformed CRP have been back-transformed on the original scale. Correlation coefficients (r) are provided for continuous variables.

  • bDifference compared to the first (reference) category of each variable; Eta2 values represent the ratio of the variance explained by the predictor to the total variance. Note that Eta2 values are displayed only if the overall P-value is <0.05.

  • cP-values compared to the first category of each variable.

  • dOverall P-values, provided for predictors with more than one degree of freedom.

  • eThresholds of 26 and 27 kg/m2 apply to males and females, respectively.

View this table:
Table 3

Prevalence of CRP genotypes and haplotypes in coronary artery disease patients and controls

CRP genotypeControlsCADP-value

nPer centiPer cent
−717C>T (rs2794521)
 TT35150.4139554.6
 CT29442.298138.4
 CC527.51797.00.136a
 T allele99671.4377173.8
 C allele39828.6133926.20.079b
+1059G>C (rs1800947)
 GG62089.0223887.6
 GC7210.930511.9
 CC10.1120.50.355a
 G allele131694.4478193.6
 C allele785.63296.40.249b
+1444C>T (rs1130864)
 CC30543.8115045.0
 CT31645.3112444.0
 TT7610.928111.00.811a
 C allele92666.4342467.0
 T allele46833.6168633.00.684b
Haplotype
 1 (TGC)45332.5176834.6
 2 (TGT)46633.4167732.8
 3 (CGC)39728.5132926.0
 4 (TCC)775.53246.3
 5 (CGT)10.1100.20.253c
  • aχ2 test, 2 by 3 contingency table.

  • bχ2 test, 2 by 2 contingency table.

  • cχ2 test, 2 by 5 contingency table.

View this table:
Table 4

Effect of CRP genotypes and haplotypes on the concentration of CRP

CRP genotypeAll individualsControls or stable CAD only
nCRP (mg/L) mean (95% CI)aΔ (%)bP-valuecnCRP (mg/L) mean (95% CI)aΔ (%)bP-valuec
−717C>T (rs2794521)0.928d0.892d
 TT17463.6 (3.4–3.8)Ref11912.7 (2.6–2.9)Ref
 CT12753.5 (3.3–3.7)−1.50.7428682.7 (2.5–2.9)−1.90.718
 CC2313.6 (3.1–4.2)+1.00.9031642.8 (2.4–3.3)+2.10.824
+1059G>C (rs1800947)Eta2 = 0.008<0.001dEta2 = 0.0060.002d
 GG28583.7 (3.5–3.8)Ref19412.8 (2.7–3.0)Ref
 GC3812.8 (2.5–3.2)−22.4<0.0012742.3 (2.0–2.6)−18.70.005
 CC131.4 (0.7–2.6)−62.90.00281.1 (0.5–2.5)−59.80.024
+1444C>T (rs1130864)Eta2 = 0.008<0.001dEta2 = 0.008<0.001d
 CC14553.2 (3.0–3.4)Ref9882.4 (2.3–2.6)Ref
 CT14403.8 (3.6–4.0)+18.5<0.0019892.9 (2.7–3.1)+20.2<0.001
 TT3574.3 (3.8–4.8)+34.3<0.0012463.2 (2.8–3.7)+31.30.001
Haplotype 1 (TGC)Eta2 = 0.0020.031dEta2 = 0.0080.004d
 014133.7 (3.5–4.0)Ref9912.8 (2.6–3.0)Ref
 114583.5 (3.3–3.7)−6.60.1159742.8 (2.6–3.0)−1.70.740
 23813.2 (2.8–3.5)−15.50.0122582.2 (1.9–2.6)−21.30.003
Haplotype 2 (TGT)Eta2 = 0.008<0.001dEta2 = 0.009<0.001d
 014583.2 (3.0–3.4)Ref9902.4 (2.3–2.6)Ref
 114453.8 (3.6–4.0)+18.5<0.0019942.9 (2.7–3.1)+20.0<0.001
 23494.4 (3.9–4.9)+37.2<0.0012393.2 (2.8–3.7)+33.0<0.001
Haplotype 3 (CGC)0.958d0.900d
 017563.6 (3.4–3.8)Ref11992.7 (2.6–2.9)Ref
 112663.5 (3.3–3.8)−0.90.8348612.7 (2.5–2.9)−1.70.736
 22303.6 (3.1–4.2)+1.30.8781632.8 (2.4–3.3)2.20.819
Haplotype 4 (TCC)Eta2 = 0.007<0.001dEta2 = 0.0050.003d
 028623.7 (3.5–3.8)Ref19442.8 (2.7–3.0)Ref
 13792.8 (2.5–3.2)−22.2<0.0012722.3 (2.0–2.6)−18.3<0.006
 2111.3 (0.7–2.7)−64.30.00371.1 (0.5–2.6)−60.30.036
  • aEstimated marginal means and 95% confidence intervals obtained in a general linear model (ANOVA) adjusted for sex, age, clinical status at presentation (no CAD, stable CAD, unstable CAD, NSTEMI, STEMI), use of lipid-lowering drugs (>97% statins) body mass index, metabolic syndrome/diabetes mellitus, hypertension, smoking, LDL cholesterol, HDL cholesterol, triglycerides (log transformed), glomerular filtration rate. The estimated marginal means and confidence intervals of logarithmically transformed CRP have been back-transformed on the original scale.

  • bDifference compared to the first (reference) category of each variable; Eta2 values represent the ratio of the variance explained by the predictor to the total variance. Note that Eta2 values are displayed only if the overall P-value is <0.05.

  • cPost hoc between group P-values compared to the first category of each variable.

  • dOverall P-values.

View this table:
Table 5

Odds ratios (OR) for angiographic coronary artery disease according to CRP genotypes and haplotypes

CRP genotypenModel 1 OR (95% CI)P-valueModel 2 OR (95% CI)P-valueModel 3 OR (95% CI)P-value
−717C>T (rs2794521)
 TT17461.0reference1.0reference1.0reference
 CT12750.84 (0.71–1.00)0.0500.85 (0.71–1.02)0.0830.86 (0.70–1.05)0.137
 CC2310.87 (0.62–1.21)0.3940.85 (0.60–1.21)0.3790.92 (0.62–1.36)0.684
 TT vs. CT vs. CCb0.88 (0.78–1.01)0.0790.89 (0.77–1.02)0.1010.91 (0.78–1.06)0.238
 TT vs. CT or CCa1746/15060.84 (0.71–1.00)0.0470.85 (0.71–1.02)0.0720.87 (0.71–1.05)0.150
 TT or CT vs. CCa3021/2310.93 (0.68–1.29)0.6700.92 (0.65–1.29)0.6190.99 (0.68–1.44)0.943
+1059G>C (rs1800947)
 GG28581.0reference1.0reference1.0reference
 GCa3811.11 (0.85–1.45)0.4361.11 (0.84–1.47)0.4771.12 (0.83–1.53)0.461
 CCa133.32 (0.43–25.61)0.2492.69 (0.34–21.64)0.3514.04 (0.49–33.49)0.196
 GG vs. GC vs. CCb1.16 (0.90–1.50)0.2501.14 (0.88–1.50)0.3251.19 (0.89–1.60)0.233
 GG vs. GC or CCa2858/3941.14 (0.88–1.49)0.3301.13 (0.86–1.49)0.3921.16 (0.86–1.58)0.329
 GG or GC vs. CCa3239/133.28 (0.43–25.30)0.2542.66 (0.33–21.38)0.3573.98 (0.48–32.98)0.201
+1444C>T (rs1130864)
 CC14551.0reference1.0reference1.0reference
 CTa14400.94 (0.79–1.13)0.5200.91 (0.76–1.10)0.3510.93 (0.75–1.14)0.462
 TTa3570.98 (0.74–1.30)0.8921.05 (0.78–1.41)0.7621.01 (0.72–1.41)0.964
 CC vs. CT vs. TTb0.97 (0.86–1.11)0.6840.99 (0.86–1.13)0.8350.98 (0.84–1.13)0.747
 CC vs. CT or TTa1455/17970.95 (0.80–1.13)0.5560.94 (0.79–1.12)0.4950.94 (0.77–1.15)0.543
 CC or CT vs. TTa2895/3571.01 (0.77–1.32)0.9441.10 (0.82–1.46)0.5301.05 (0.76–1.44)0.774
Haplotype 1
 014131.0reference1.0reference1.0reference
 114581.05 (0.88–1.26)0.5621.02 (0.85–1.23)0.8220.98 (0.79–1.20)0.820
 23811.28 (0.96–1.70)0.0991.27 (0.94–1.72)0.1271.24 (0.88–1.73)0.219
 0 vs. 1 vs. 2b1.10 (0.97–1.25)0.1301.09 (0.95–1.24)0.2101.06 (0.92–1.23)0.419
 0 vs. 1 or 2a1413/18391.09 (0.93–1.30)0.2951.07 (0.89–1.27)0.4801.02 (0.84–1.24)0.831
 0 or 1 vs. 2a2871/3811.24 (0.94–1.63)0.1221.25 (0.94–1.67)0.1251.25 (0.91–1.72)0.171
Haplotype 2
 014581.0reference1.0reference1.0reference
 114450.94 (0.79–1.12)0.4750.91 (0.75–1.10)0.3170.92 (0.74–1.13)0.402
 23490.98 (0.74–1.31)0.9071.06 (0.78–1.44)0.7031.03 (0.73–1.44)0.887
 0 vs. 1 vs. 2b0.97 (0.86–1.10)0.6660.99 (0.86–1.13)0.8480.98 (0.84–1.13)0.757
 0 vs. 1 or 2a1458/17940.95 (0.80–1.12)0.5200.94 (0.78–1.12)0.4720.94 (0.77–1.14)0.504
 0 or 1 vs. 2a2903/3491.02 (0.77–1.33)0.9121.11 (0.84–1.48)0.4651.07 (0.78–1.48)0.675
Haploytpe 3
 017561.0reference1.0reference1.0reference
 112660.83 (0.70–0.99)0.0400.84 (0.70–1.02)0.0740.86 (0.70–1.05)0.139
 22300.85 (0.62–1.19)0.3640.85 (0.60–1.21)0.3580.91 (0.62–1.35)0.642
 0 vs. 1 vs. 2b0.88 (0.77–1.01)0.0640.89 (0.77–1.02)0.0880.91 (0.77–1.06)0.224
 0 vs. 1 or 2a1756/0.84 (0.71–0.99)0.0370.85 (0.71–1.01)0.0630.87 (0.71–1.05)0.146
 0 or 1 vs. 2a14960.93 (0.67–1.28)0.6520.91 (0.65–1.28)0.5990.97 (0.66–1.43)0.894
Haploytpe 4
 028621.0reference1.0reference1.0reference
 13791.12 (0.86–1.47)0.3951.12 (0.84–1.48)0.4361.14 (0.84–1.56)0.399
 2112.77 (0.35–21.69)0.3322.38 (0.29–19.56)0.4193.63 (0.43–30.76)0.237
 0 vs. 1 vs. 2b1.16 (0.90–1.50)0.2601.15 (0.88–1.50)0.3201.20 (0.90–1.61)0.221
 0 vs. 1 or 2a2862/3901.15 (0.88–1.49)0.3181.14 (0.86–1.50)0.3711.18 (0.87–1.60)0.296
 0 or 1 vs. 2a3241/112.74 (0.35–21.40)0.3382.35 (0.29–19.30)0.4263.57 (0.42–30.21)0.244
  • aGenotypes treated as categorical variables and compared with the reference category.

  • bGenotypes treated as interval-scaled variables.

  • Model 1: unadjusted.

  • Model 2: adjusted for age, and sex.

  • Model 3: in addition adjusted for clinical status at presentation (no CAD, stable CAD, unstable CAD, NSTEMI, STEMI), body mass index, metabolic syndrome/diabetes mellitus, hypertension, smoking, LDL cholesterol, HDL cholesterol, triglycerides (log transformed), glomerular filtration rate.

Results

Characteristics of coronary artery disease patients and controls

Patients with CAD were significantly older than controls (Table 1). Current or past smoking, diabetes mellitus, and hypertension were more prevalent in CAD compared with controls. After adjusting for sex and age, systolic blood pressure, fasting glucose, and triglycerides were higher in CAD patients than in controls; HDL cholesterol was lower. Body mass index, diastolic blood pressure, and GFR were not significantly different. LDL-C concentrations did also not differ between the two groups, due to the fact that 57% of CAD patients received lipid-lowering drugs (>97% statins) compared with 18% of controls. When we adjusted for the use of lipid-lowering drugs in addition to age and sex, estimated marginal means for LDL-C were 1.13 and 1.18 g/L, respectively, in controls and CAD patients (P = 0.002).

Association of cardiovascular risk factors and coronary artery disease status with C-reactive protein

We examined the effect of sex, age, risk factors, and of the clinical status on C-reactive protein in a general linear model in which we included those factors not under examination as co-variables (Table 2). C-reactive protein was higher in women than in men, and increased in parallel to age and body mass index. C-reactive protein was elevated in smokers, more so in current than in former ones. There was a non-significant tendency towards higher C-reactive protein in patients with the metabolic syndrome, and it was significantly increased in patients with diabetes mellitus. C-reactive protein was similar in hypertension as in normotension. C-reactive protein was high at high triglycerides and at low HDL-C, but was not related to LDL-C. Patients with calculated GFRs <60 mL/min had increased C-reactive protein compared with those with GFRs ≥ 90 mL/min. Interestingly, compared with individuals without CAD, C-reactive protein was not significantly increased in patients with stable CAD. Only in patients with unstable angina, NSTEMI or STEMI was C-reactive protein significantly higher than in individuals without CAD. On the whole, these clinical characteristics accounted for 24% of the total variance of C-reactive protein with the clinical condition at presentation (10%) being the most important predictor.

C-reactive protein genotypes

The three PM of the C-reactive protein gene were in Hardy-Weinberg equilibrium. The allele frequencies (Table 3) were similar to those reported in other Caucasian populations. We found five haplotypes at the C-reactive protein locus. The fifth of these haplotypes was very rare and therefore disregarded in all further analyses. χ2 testing did not reveal any significant association between allele frequencies and angiographic CAD.

Effect of C-reactive protein genotypes and circulating C-reactive protein

We used ANOVA to examine the effects of genotypes on circulating C-reactive protein with statistical adjustments made for age, sex, use of lipid-lowering drugs, cardiovascular risk factors (body mass index, diabetes mellitus, hypertension, smoking, LDL-C, HDL-C, triglycerides, and GFR), and for CAD status (no CAD, stable CAD, unstable angina, NSTEMI, STEMI, coded 0 through 4, Table 4). The −717C>T (rs2794521) PM was not related to C-reactive protein, but both the +1059G>C PM (rs1800947) and the +1444C>T PM (rs1130864) were. One C allele at position +1059 (rs1800947) increased C-reactive protein by 20–30%, one T allele at position +1444 (rs1130864) increased C-reactive protein by ∼18%. We observed four haplotypes at the C-reactive protein locus. Haplotype 1 moderately, but significantly lowered C-reactive protein. One copy of haplotype 2 increased C-reactive protein by ∼18%. Haplotype 3 had no appreciable effect on C-reactive protein. One copy of haplotype 4 decreased C-reactive protein by 20–30% (Table 4). Thus, three out of the four haplotypes substantially affected circulating C-reactive protein. The proportion of the total variance of the C-reactive protein concentration explained by each individual genetic PM or haplotype was <1% throughout.

To be certain that confounding by any acute phase response was ruled out, we conducted an additional set of analyses which we restricted to controls or patients with stable CAD. Consequentially, in this analyses the co-variable CAD status could assume two values only (0 and 1, for no CAD and stable CAD, respectively, Table 4). This analysis was highly consistent with the results obtained in the entire cohort. Finally, the same associations were seen when we considered study participants not receiving lipid-lowering drugs separately (data not shown).

Association of C-reactive protein gene polymorphisms and haplotypes with angiographic coronary artery disease

There was a tendency towards a lower prevalence of angiographic CAD in carriers of the C allele of the −717C>T (rs2794521) PM which was, however, not robust against adjustment for confounding variables (Table 5). Heterozygous carriers of the +1059G>C PM (rs1800947), who had significantly reduced C-reactive protein concentrations, revealed a spuriously, but insignificantly higher prevalence of angiographic CAD than carriers of the wild-type. Homozygous carriers of the C allele at position +1059 (rs1800947) even showed a more than three-fold higher prevalence of CAD, but this association was also not statistically significant. There was virtually no association of the +1444C>T PM (rs1130864) and CAD.

In carriers of two copies of haplotype 1 which lowers C-reactive protein the prevalence of CAD was slightly increased, but this was not significant. Haplotype 2 which was associated with high C-reactive protein was not at all associated with CAD. Carriers of one or two copies of haplotype 3 (dominant model) which was neutral with regard to the C-reactive protein concentration had a significantly lower prevalence of CAD compared with those lacking this haplotype. Finally, there was a non-significant increase in the prevalence of CAD in carriers of haplotype 4, which decreased C-reactive protein.

Neither C-reactive protein genotypes nor C-reactive protein haplotypes were associated with a history of myocardial infarction (data not shown). To examine the impact of definitions of CAD different from the current one (at least one stenosis of 20% or more in more than one coronary segment), we also used the presence of one or more stenoses of 50% or more as a criterion. This also did not materially change the results (data not shown). Further, we compared the severity of CAD (coded 0 through 3) between C-reactive protein genotypes and haplotypes by ANOVA and covariance. In none of these models (data not shown) were the C-reactive protein genotypes and haplotypes associated with severity of CAD with the exception that carriers of the +1444CT (rs1130864) genotype (who have increased C-reactive protein) had a slightly lower mean severity of CAD (P = 0.018).

Discussion

We have completed the largest study simultaneously investigating the relationship between C-reactive protein genotypes, circulating C-reactive protein, and angiographic CAD. The study has two key results. We confirm that genetic diversity at the C-reactive protein locus impacts on the concentration of C-reactive protein in the circulation. Further, none of the genotypes raising C-reactive protein was consistently associated with an increased prevalence of CAD and that none of the genotypes resulting in low C-reactive protein protected from CAD.

Circulating levels of C-reactive protein show substantial intra- and inter-individual variation.53 Many conventional cardiovascular risk factors themselves produce high C-reactive protein concentrations. We therefore sought to identify variables potentially confounding associations between genotypes and phenotypes. As expected, C-reactive protein was higher in women than in men, increased in parallel to age and body mass index, was elevated in patients with type 2 diabetes mellitus, in previous or current smokers, at low HDL-C and at impaired renal function. C-reactive protein increased in the order of no CAD, stable CAD, unstable CAD, NSTEMI, and STEMI. Of interest, however, as reported previously from the LURIC cohort54 the difference in C-reactive protein between patients without angiographic CAD and those with clinically stable CAD was small and statistically not significant, although our ability to detect even a small difference was considerable: At a power of 0.80 and an alpha of 0.05 we would have been able to detect a roughly 25% difference in C-reactive protein between the controls (n = 697) and the stable CAD patients (n = 1526). This may indicate that C-reactive protein is not causal to the development of atherosclerosis. Rather, C-reactive protein may be more closely related to the inflammatory activity of atherosclerotic lesions. On total, the anthropometric and clinical measures accounted for 24% of the total variance of C-reactive protein with the clinical status at the time of presentation contributing as much as 10%. This is approximately in line with Kathiresan et al.23 reporting that 12 clinical factors were significantly related to serum C-reactive protein level and combined explained ∼26% of C-reactive protein level variation.

Using multivariate adjustment for confounding variables we confirmed previous results as to the association between PM at the C-reactive protein gene and systemic C-reactive protein concentrations. As seen by others,11,14,16,45 the −717C>T PM (rs2794521) was not related to C-reactive protein. The +1059G>C PM (rs1800947) was associated with lower C-reactive protein in some,10,13,16,18 but not all11,14,15 prior studies. In the current study, each minor (C) allele of the +1059G>C PM (rs1800947) lowered C-reactive protein by ∼1 mg/L. In line with previous work,11,16,34 the minor (T) allele of the +1444C>T PM (rs1130864) was positively and dose-dependently related to C-reactive protein. Most likely, this association is due to the fact that the +1444C>T PM (rs1130864) stands in linkage disequilibrium to a functional PM in the promoter of the C-reactive protein gene (−286T>A; rs3091244).16 Roughly, the presence of a single T allele increased the C-reactive protein concentration by ∼0.6 mg/L, an effect estimate which is slightly greater than that obtained in a previous analysis of pooled data from 4659 individuals.34

We inferred four major haplotypes at the C-reactive protein locus. The findings compare well with those by Miller et al.16 Our haplotypes 2, 3, and 4 concur with their haplotypes 2, 1, and 4; our haplotype 1 represents a combination of their haplotypes 3, 5, and 6. In line with the findings by Miller et al.,16 our haplotypes 2 and 4 increased and decreased C-reactive protein in a dose-dependent fashion, respectively. Miller et al.’s haplotype 316 is far more frequent than their haplotypes 5 and 6. Thus, the effect of our ‘pooled’ haplotype 1 on C-reactive protein is likely driven by Miller et al.’s haplotype 316 which decreases C-reactive protein. This was indeed true, as our haplotype 1 significantly lowered C-reactive protein. Our results are also entirely compatible with those of Kardys et al.24 who—despite using a different set of PM—observed the same four major haplotypes at the C-reactive protein gene locus at frequencies of 32.8, 31.7, 29.5, and 5.9%, respectively. However, our results do not completely compare with Carlson et al.20 The SNPs genotyped in the current work tag their haplotypes 1 (+1059G>C; rs1800947), 4 (-717c>T; rs2794521), and 5 to 8 (+1444C>T; rs1130864). Carlson et al.’s20 most frequent haploytpe 2 is not tagged, as is their haplotype 3 which occurred at a frequency of <5%.

In line with previous work,23 the proportion of the total variance of the C-reactive protein concentration explained by the genetic PM or haplotypes was small; it was in the order of 1% or less for those individual genetic predictors revealing a significant association with C-reactive protein. However, the proportions of variance explained by individual clinical factors well accepted to significantly impact on C-reactive protein such as sex, body mass index, metabolic syndrome/diabetes were <1% and thus similarly low. The three PMs chosen were the most frequent studied ones when we initiated this study. In the meantime several publications with additional SNPs and haplotypes have been published.55 However, as we were able to identify the most common haplotypes, we are convinced that additional PMs do not add information to the study or change the presented results.

The main purpose of this study was to examine whether or not variants at the C-reactive protein gene modulating the concentration of C-reactive protein would be associated with the atherosclerotic changes of the coronary arteries. Such a relationship would strongly imply C-reactive protein in the development of CAD. So far, a few studies addressed the value of C-reactive protein-PM in the prediction of acute cardiovascular events, but virtually no information has been available relating C-reactive protein-PM to the angiographic findings.

The −717C>T PM (rs2794521) had no appreciable effect on C-reactive protein, but there was a trend to a lower prevalence of CAD in carriers of the minor (C) allele which was mostly not significant. This is in line with a recent small cross-sectional study in Chinese45 and with the Physicians' Health Study (n= 610 cases with either myocardial infarction or thromboembolic stroke),16 both of which revealed the minor (C) allele of the −717C>T PM (rs2794521) associated with a decreased risk of CAD. These observations may have its analogy in a study looking at genetic PMs related to lipoprotein metabolism in which genotypes predicting an adverse lipid profile protected against cardiovascular events.56

The common denominator of the other results was that neither the +1059G>C PM (rs1800947), the +1444C>T PM (rs1130864), nor any of our haplotypes was associated with the prevalence of angiographic CAD. From the literature, a clear relationship between C-reactive protein PM and cardiovascular outcomes does not emerge at present. This may be related to the fact that different PMs were genotyped in the various studies and that true associations were missed because not all of the common genetic variation was represented. Our findings which regard to the +1059G>C PM (rs1800947) are in agreement with previous studies in Caucasians10,16 in which carriers of the C allele did not have a lower risk of cardiovascular events despite having lower C-reactive protein concentrations, but stand in contrast to Lange et al.22 who observed that white carriers of the C allele were at a significantly decreased risk of cardiovascular death. Three previous studies of the +1444C>T PM (rs1130864) in Caucasians also found no association between the +1444C>T PM (rs1130864) and cardiovascular events despite a similar effect of the minor allele variant on C-reactive protein.16,17,34 C-reactive protein haplotypes were also not associated with incident CAD in the Rotterdam heart study,24 and in patients beginning dialysis.26 In the study of Crawford et al.21 the AA genotype of the triallelic promoter PM -286C/T/A (rs3091244) (which was not tested in the current study) correlated positively with the C-reactive protein concentration in the non-Hispanic black and in the Mexican American groups, respectively, but was associated with prevalent CAD in the non-Hispanic white population. The odds ratio (OR) for CAD of the AA genotype was 30.11 (95% CI 3.26–278.08) compared with the CC referent genotype, whereas the association of the AT genotype with prevalent CAD was opposite (OR, 0.10; 95% CI 0.02–0.57). However, only <5 participants had the AA or AT genotype, respectively, while there were only 52 cases with the referent CC genotype. These associations should be interpreted with reservation for several reasons: they rested on very small numbers and self-reported CAD, were not accompanied by corresponding changes in C-reactive protein, and were limited to the Caucasians only. Ethnicity-related effects were also encountered by Lange et al.22 who examined non-coding SNPs different from those in the current study. Although carotid intima media thickness was not related to any of the PMs tested, some of them increased and some of them decreased the risk of either stroke or cardiovascular disease, either in the Caucasians or African–Americans only. Finally, Eklund et al.25 reported that the tri-allelic −286C/T/A promoter PM (rs3091244) was associated with carotid artery compliance, a surrogate marker of uncertain clinical significance in men only.

Thus, genotypes that would expose to a long-term moderate elevation in C-reactive protein were not related to CAD and those leading to low C-reactive protein did not protect from CAD. The approach used in this study to search for causality resembles the strategy of Mendelian randomization, which assumes that if genetic variants specifically alter an ‘exogenously’ modifiable factor deemed causal to the development of a disease, then those genetic variants should themselves be associated with the disease.5 As an acute phase reactant, C-reactive protein is the prototype of an ‘exogenously’ modifiable factor and the relationship between C-reactive protein and CAD would, in addition, be subject to confounding by most of the well known cardiovascular risk factors (cf. Table 2). In contrast, there is no reason to assume that those ‘exogenous’ confounders are associated with the functional C-reactive protein genetic variants which exert stable long-term effects on the concentration of C-reactive protein. Further, the well-established association between C-reactive protein and cardiovascular events might be due to reverse causality. In the case of C-reactive protein, reverse causality is likely to be relevant, since rupture-prone atherosclerotic lesions bear significant inflammatory activity although an individual has not become symptomatic yet. Although the clinical phenotype of angiographic CAD has hardly been addressed, previous studies pursuing similar strategies have arrived at consistent conclusions. For instance, Timpson et al.57 found no evidence for a causal role of C-reactive protein in the metabolic syndrome, and studies looking at the relationship between genetic variants at the C-reactive protein locus and acute cardiovascular events were negative with only a few minor exceptions. This is well in line with the finding of no relationship between C-reactive protein genotypes or haplotypes and myocardial infarction in the current study (data not shown).

Can our conclusions be reconciled with the experimental evidence attributing to C-reactive protein an active role in atherothrombosis? Indeed, a plethora of publications have incriminated C-reactive protein to exert pro-coagulant or atherogenic effects.2733,3544 However, very recent studies cast serious doubt on the reliability of these findings because most of the pro-thrombosis, pro-inflammatory, or proatherogenic effects seen in vitro could be attributable to bacterial lipopolysaccharide or sodium azide contaminating commercial C-reactive protein preparations and because many of the effects are not seen any more once highly pure C-reactive protein is used.5866 Consistently, recent studies of atherosclerosis-prone mice show that overexpression of human C-reactive protein did not enhance the development of atherosclerosis.6769 Even the suggestion that lowering C-reactive protein by statins in humans prevents cardiovascular events70,71 would not be a compelling proof of causality since a strong correlation exists between the effect of statins on LDL cholesterol and on C-reactive protein.72

Our study may have some limitations. First, as only a single measurement of C-reactive protein has been obtained, the day-to-day within-person variability of C-reactive protein might have attenuated the association between C-reactive protein and genetic PM. For example, at a sample size of approximately n = 200, the within-person variability would attenuate the correlation coefficient and the regression coefficient between C-reactive protein and another variable by 15 and by 27%, respectively.73 However, we believe that the actual attenuation is much smaller in the current study due to its large sample size. In addition, Carlson et al.20 measured C-reactive protein on 7 and 15 year of their prospective study and the results of their genetic association study were very similar at these two time points. Second, although the current study is one of the largest coping with the relationship between C-reactive protein genetic PMs and CAD its statistical power may still have been too low to detect small effects of these PMs. For instance, the estimated marginal mean C-reactive protein in CC homozygotes at +1444 (rs1130864) was 3.2 mg/L compared with 4.3 mg/L in TT homozygotes. Assuming a log-linear relationship between the C-reactive protein concentration and CAD described by König et al.74 this change should be accompanied by an OR for CAD of 1.20. Power calculation for the present study shows that, with a power of 80% and an alpha of 0.05, in reference to CC (n= 1413, 44%) we would have been able to demonstrate a relative risk for coronary heart disease of ∼1.6 for TT (n= 381, 11%).

The third limitation resides in the fact that the proportion of the total variance of the C-reactive protein concentration contributed by both clinical characteristics and the genetic PMs or haplotypes was small so that most of the variation in C-reactive protein levels is yet to be explained, it is thus possible that PMs in long-range cis-acting regulatory regions or other genes such as the apolipoprotein E gene75,76 have significant impact on both C-reactive protein levels and cardiovascular risk. Also, there may be other factors that preclude the detection of an association between genetic markers and disease outcomes even if a causal relationship exists.77

We also wish to notice that adjustment for multiple testing has not been performed because we examined only three PMs and because these PMs have all been studied in relation to cardiovascular disease before. Further, our control individuals were recruited at a tertiary referral centre, underwent coronary angiography and may hence not be representative for a random population sample. This, however, may also be regarded as strength of the study. The prevalence of clinically asymptomatic coronary atherosclerosis has been reported to be very high at or above 50 years of age.78 Hence, angiography-based recruitment of controls rules out that individuals with significant, yet clinically unapparent, CAD are inadvertently allocated to the control group. Further, the major cardiovascular risk factors occur at a similar frequency in our controls compared with the general population. The prevalence of hypertension is close to that found in a random probability sample from Germany.79 Prima vista, diabetes mellitus appears two to three times more frequent in our study than in the general German population.80 This is, however, most likely due to the fact we did not rely on self-reports. Rather, we measured fasting glucose and performed an oral glucose challenge in individuals not previously known to have DM.

Without doubt is there a correlation between the concentration of C-reactive protein and the risk of future cardiovascular events.13 The current data along with many pertinent previous epidemiological10,16,34 and experimental findings, however, raise the possibility that C-reactive protein does not have a causal role in the development of atherosclerosis. Rather the reported association between C-reactive protein and coronary events in13 may be due to reverse causality bias or to confounding by other cardiovascular risk factors related to C-reactive protein.

Acknowledgements

The technical assistance of Sabine von Karger and Andrea Müller is gratefully acknowledged. The authors extend appreciation to the participants of the Ludwigshafen Risk and Cardiovascular Health Study; without their collaboration this article would not have been written. We thank the LURIC study team either temporarily or permanently involved in patient recruitment, sample and data handling, and the laboratory staff at the Ludwigshafen General Hospital and the Universities of Freiburg and Ulm, Germany.

Conflict of interest: none declared.

References

References

The above article uses old reference style.

View Abstract