C-reactive protein polymorphisms are associated with the cortisol awakening response in basal conditions in human subjects

Abstract

Cortisol affects the acute-phase response, but it is unknown whether C-reactive protein (CRP), an acute-phase reactant, also affects hypothalamus–pituitary–adrenal axis activity. In the present study, associations were explored between CRP haplotypes with plasma CRP concentrations and basal salivary cortisol level. We included 266 physically healthy Caucasian subjects (103 females and 163 males) aged between 18 and 65 years of whom 94 had a psychiatric disorder in a genetic association study. Six tag single-nucleotide polymorphisms capturing the common genetic variation of the CRP gene were genotyped (i.e. rs2808628, rs2808630, rs1205, rs1800947, rs1417938, and rs3091244) to yield common CRP haplotypes. Plasma CRP concentrations, the salivary cortisol awakening response (CAR) (0, 30, 45, and 60 min after awakening), and the diurnal cortisol decline (11:00, 15:00, 19:00, and 23:00 h) were assessed for 2 days. rs2808628, rs1205, rs1417938, and rs3091244 showed expected associations not only with CRP concentrations, but also with salivary cortisol levels during the CAR. Five well-characterized CRP haplotypes were arranged in ascending order according to increasing CRP levels. There was an inverse linear association between CRP haplotypes and cortisol levels during the CAR, but no association with the diurnal cortisol decline. Hence, genetic variants in the CRP gene that are associated with lifetime plasma CRP levels were also associated with salivary cortisol levels after awakening, in basal, non-inflammatory conditions.

Keywords::

• Acute-phase response
• C-reactive protein
• hypothalamic–pituitary–adrenal axis
• Mendelian randomization
• polymorphisms
• salivary cortisol

Introduction

The immune system and the hypothalamus–pituitary–adrenal (HPA) axis play important roles in maintaining homeostasis by generating adaptive responses to noxious stressors (Webster et al. 2002; Elenkov and Chrousos 2006). C-reactive protein (CRP), an acute-phase reactant, and cortisol, the main stress hormone of the HPA axis, are both involved. CRP is predominantly produced in the liver and its release is regulated by an inflammatory cascade of reactions, which involve, among others, proinflammatory cytokines (Volanakis 2001; Black et al. 2004). Cortisol acts synergistically with the proinflammatory cytokine interleukin (IL)-6 to enhance this effect (Szalai et al. 2000). The main biological function of CRP is its ability to recognize pathogens and damaged cells of the host and to mediate their elimination by recruiting the complement system, which subsequently activates and attracts phagocytic cells via the activation of Fc receptors (Volanakis 2001; Du Clos and Mold 2004). CRP induces the release of proinflammatory cytokines IL-1, IL-6, and TNF-α by these cells (Ballou and Lozanski 1992). By contrast, cortisol is a potent endogenous anti-inflammatory agent with immunosuppressive effects. It has a strong capacity to suppress immune cell functions, such as inhibiting the release of proinflammatory cytokines. However, cortisol has many other functions besides immunosuppressive effects, such as effects on lipid metabolism, bone metabolism, memory, catabolism, digestion, sleep-waking cycle, memory, all aiding the restoration of homeostasis after stress. Little is known yet about a direct pathway from CRP to cortisol release. However, a bidirectional relationship between CRP and cortisol is assumed to play an important role in maintaining physiological homeostasis (Webster et al. 2002).

However, subjects with high circulating CRP levels have increased risk for cardiovascular disease and depression (Pearson et al. 2003; Almeida et al. 2007; Buckley et al. 2009). Environmental variables and lifestyle behaviors such as dietary intake, smoking, acute and chronic infections, gender, lipid levels, obesity, and blood pressure can contribute to variations in both CRP and cortisol levels (Wust et al. 2000b; Kunzel et al. 2003; Kushner et al. 2006). These variables might also influence the potential cross-sectional relationship between CRP and cortisol. CRP production has an important genetic component, as 35–40% of interindividual variation in blood CRP levels is heritable (Hage and Szalai 2007). Several well-characterized CRP gene polymorphisms are associated with plasma CRP levels, e.g. rs2808628, rs2808630, rs1205, rs1800947, rs1417938, and rs3091244 (Crawford et al. 2006; Hage and Szalai 2007; Bos et al. 2008; Zacho et al. 2008). Mendelian randomization refers to the random assortment of genes from parents to offspring that occurs during gamete formation and conception (Smith and Ebrahim 2004; Schunkert and Samani 2008), and this provides a method to assess, in an essentially unconfounded way, whether CRP is related to cortisol secretion, as measured in saliva.

In the present study, we assessed these CRP polymorphisms using a Mendelian randomization design in order to elucidate the relationship between plasma CRP and basal salivary cortisol concentrations over the day under noninflammatory conditions. We tested the hypothesis that haplotypes of the CRP gene that affect plasma CRP levels are also associated with basal salivary cortisol levels, especially with cortisol during the awakening response, which shows substantial heritability (40–48%; Wust et al. 2000a). In recent investigations, it has been demonstrated that cortisol levels in the first hour after awakening can serve as a useful index of adrenocortical activity. When measured with strict reference to the time of awakening, the assessment of this endocrine response can uncover subtle changes in the HPA axis activity. The cortisol awakening response (CAR) is rather consistent and about 75% of subjects show a CAR. Furthermore, the intraindividual stability over time is remarkably high with correlations up to r = 0.63 (for the area under the response curve) (Wust et al. 2000b).

As indices of activity of the HPA axis, we used sequential assessments of salivary cortisol over the day to derive measures of the CAR and the diurnal decline in cortisol secretion.

Materials and methods

Subjects

We included 266 physically healthy subjects aged between 18 and 65 years, of whom 94 subjects had a depressive and/or anxiety disorder according to the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV. Data were collected as part of studies on psychiatric and neuroendocrine correlates of the HPA axis (Klaassens et al. 2008; Veen et al. 2009). Nonpsychiatric subjects were recruited by advertisement in local newspapers, asking for physically and mentally healthy persons willing to participate in a study on the biological stress system in relation to depression and anxiety. Psychiatric subjects were included from the outpatient department of the Rivierduinen mental health center in Leiden, The Netherlands. We allocated nonpsychiatric and psychiatric subjects to one experimental group to increase the sample size and statistical power. Exclusion criteria were a history of neurological or endocrine diseases or other serious or unstable medical conditions. Furthermore, subjects with substance or alcohol abuse according to the DSM-IV, as well as pregnant or breast-feeding women and premenopausal women with ovariectomy were excluded. Subjects were excluded if they had used psychotropic medication within the previous 14 days; subjects taking corticosteroids, antidiabetics, estrogens, thyroid hormone, or herbal medication (e.g. Valerian, St. Johns Wort) were also excluded. All subjects had a routine physical examination and laboratory blood tests, to exclude acute infections and chronic inflammatory diseases. Prior to participation, written informed consent was obtained from all subjects. The study was approved by the ethics committee of the Leiden University Medical Center.

Saliva cortisol sampling

Instructions concerning saliva sampling prohibited eating, smoking, drinking tea or coffee, or brushing teeth within 15 min before sampling. Furthermore, no dental work 24 h prior to sampling was allowed. Saliva samples were obtained using Salivettes (Starstedt, Germany) at eight time points covering the CAR and the diurnal decline in cortisol secretion. The CAR includes four sampling points; at awakening (T1) and 30 min (T2), 45 min (T3) and 60 min (T4) after awakening. Four additional samples were taken to assess the diurnal decline in cortisol at 11:00 h (T5), 15:00 h (T6), 19:00 h (T7), and 23:00 h (T8). Subjects registered each time a sample was collected in order that a limited check could be made on whether subjects were sampled at the required times. The cortisol secretion curve on a single day is determined by situational factors and, to a smaller extent, by trait factors (Hellhammer et al. 2007). Therefore, to reduce measurement error and the effects of day-to-day variation, subjects were asked to provide saliva samples on two consecutive nonworking days. Samples were initially stored at 4°C for a week at most and delivered by the subject to our clinic. After receipt in our clinic, salivettes were centrifuged at 2000g for 10 min, aliquoted and stored at − 20°C. Cortisol analysis was performed by competitive electrochemiluminescence immunoassay (Roche, Basel, Switzerland), as described in van Aken et al. (2003). The lower detection limit was 0.50 nmol/l. The intra-assay coefficients of variance were lower than 7%, the inter-assay coefficients of variance were lower than 8%, except for the very low range. Per sampling point, physiologically unlikely high values (i.e. >50 nmol/l) were excluded from further analyses (1.3% of the data). Saliva cortisol levels are reported as nmol/l and showed a normal distribution. The two cortisol values obtained at the time points on the 2 days were significantly correlated indicating moderate to good intra-individual stability over time (Pearson's r between 0.45 and 0.65, p ≤ 0.001). Therefore, mean cortisol values for each time point were computed for each subject and used in the analyses. If a sample was missing, then the value of the other day was used (2.9% of the data).

Plasma CRP measurement

Venous blood was sampled within 2 weeks after the saliva cortisol sampling with standard venipuncture techniques. Blood samples were collected at 9.00 or 13.30 h, and anti-coagulated with ethylenediaminetetraacetic acid, and the plasma was separated by centrifugation. Plasma CRP concentrations were measured by automated enzymatic colorimetric methods using a Modular P analyzer (Roche) with a lower limit of detection of 3 mg/l.

Single-nucleotide polymorphism (SNP) genotyping

Genomic DNA was isolated from the blood samples according to standard procedures (QIAamp blood maxi kit, Qiagen, Venlo, The Netherlands) (Loffler et al. 1997). Six well-characterized CRP single-nucleotide polymorphisms (SNPs) that are related to plasma CRP levels were determined (Figure 1). I-plex assays were assigned using the Assay designer software (Sequenom, San Diego, CA, USA). Genotyping was performed using the MassArray platform according to the manufacturer's protocols (Sequenom). After PCR on 2.5 ng of DNA, a primer extension reaction was performed to introduce mass differences between alleles. After removing salts by adding a resin, ∼15 nl of the product was spotted onto a target chip with 384 patches containing matrix. Mass differences were detected using the Bruker Autoflex MALDI-TOF mass spectrometer, and genotypes were assigned real-time user Typer 3.1 software (Sequenom). As a quality control, 10% of samples were genotyped in duplicate, and no inconsistencies were observed.

Figure 1.  Schematic overview of the structure of the CRP gene consisting of two exons seperated by a single intron. The orientation of the gene is marked by arrows, with the gene transcribed from left to right. All single nucleotide polymorphisms (SNPs) were in Hardy–Weinberg equilibrium (all p ≥ 0.15). Arrows mark the approximate location on chromosome 1 of the six tag-SNPs that were in close linkage disequilbrium (D′ for all SNP pairs ≥ 0.96, except for two pairs with ≥ 0.90). The five most frequent haplotypes are presented, which are similar to those found in populations from Northern and Western European ancestry. rs3091244 is a triallelic SNP located in the promotor region of the CRP gene.

Statistical analyses

The haplotype frequencies were reconstructed using SNPHAP (Clayton 2002; version 1.3; available online at http://www-gene.cimr.cam.ac.uk/clayton/software/) with an estimation–maximization algorithm. Haplotypes were placed in the order that previously have been associated with increasing plasma CRP levels (Carlson et al. 2005; Bos et al. 2008). Each of the CRP polymorphisms was assessed to determine if the observed genotype frequencies were consistent with the expected Hardy–Weinberg's proportions using the χ²-test. The linkage disequilibrium (LD) was assessed using Lewontin's D′ statistics (Lewontin 1964).

As markers of overall cortisol concentrations, the total area under the curve of the CAR (AUC_CAR) and diurnal decline (AUC_diurnal) were calculated using the trapezoid formula (Pruessner et al. 2003). For the AUC_CAR, the following formula was used: AUC_CAR = (T1+T2)/2 × 0.5+(T2+T3)/2 × 0.25+(T3+T4)/2 × 0.25. For the AUC_diurnal, the following formula was used: AUC_diurnal = ((T5+T6)/2 × 4+(T6+T7)/2 × 4+(T7+T8)/2 × 4)/12.

Furthermore, subjects were dichotomized according to the plasma CRP-concentration (>3 mg/l = high risk and ≤ 3 mg/l = low/average risk). The cut-off points of low/average risk ( < 3.0 mg/l) and high risk (>3.0 mg/l) correspond to approximate tertiles of CRP in the adult population (Pearson et al. 2003). This cut-off reflects the finding that subjects with plasma CRP concentrations >3.0 mg/l are at an increased risk for cardiovascular disease [estimated odds ratio (OR) between 1.6 and 2.0] and depression (estimated OR 1.6 in men) (Pearson et al. 2003; Almeida et al. 2007; Buckley et al. 2009).

Binary logistic and linear regression analyses were used to investigate associations between polymorphisms and haplotypes with changes in plasma CRP and salivary cortisol concentrations. Age, gender, smoking status (yes/no), body mass index (BMI), and group (presence/absence of depression and/or anxiety disorder) were entered as additional independent variables in the regression analyses to adjust for their potential effects on the variable of interest. Nonpsychiatric and psychiatric subjects were combined because of the relatively low number of subjects per group to increase the statistical power. Analyses were performed using the Statistical Package for the Social Science version 16.0 (SPSS 16.0; SPSS, Inc., Chicago, IL, USA). p-Value < 0.05 was considered statistically significant.

Results

Study population

The mean age of the study population was 40.1 years [standard deviation 13.1 years]. The study population consisted of 103 females (38.7%) and 163 males (61.3%), and 38.3% of the subjects were smokers. The mean BMI of the study population was 25.3 ± 4.5 kg/m². The genotypes frequencies are listed for each polymorphism in Table I. Rare genotypes ( < 1%) for rs3091244 and rs1800947 were subsequently collapsed. The genotype frequencies of all polymorphisms were in Hardy–Weinberg equilibrium (HWE; all p>0.15). All SNPs were in HWE (all p ≥ 0.15). The six polymorphisms were in close LD (D′ for all SNP pairs ≥ 0.96, except for two pairs with ≥ 0.90).

Table I.  Plasma CRP levels, AUC_CAR values, and AUC_diurnal values according to genotype frequencies for CRP polymorphisms in 266 subjects.

SNP genotype	n	High CRP (%)	AUCCAR (nmol/l/h)	AUCdiurnal (nmol/l/h)
rs2808628
GG	111	34 (30.6%)	15.2 (14.3–16.1)	6.4 (6.0–6.9)
GA	113	20 (17.7%)	16.0 (15.0–16.9)	6.7 (6.2–7.2)
AA	40	7 (17.5%)	16.9 (15.4–18.3)	6.6 (5.9–7.4)
Missing values	2
Test statistics		Wald = 7.00, df = 1	β = 0.11, t = 1.89	β = 0.03, t = 0.57
p-Value for trend		p = 0.01	p = 0.06	p = 0.57
rs2808630
TT	140	31 (22.1%)	15.6 (14.8–16.5)	6.6 (6.2–7.1)
TC	108	28 (25.9%)	16.1 (15.2–17.0)	6.4 (5.9–6.9)
CC	17	3 (17.7%)	15.3 (12.8–17.9)	6.9 (5.6–8.2)
Missing values	1
Test statistics		Wald = 0.11, df = 1	β = 0.02, t = 0.31	β = − 0.02, t = − 0.27
p-Value for trend		p = 0.75	p = 0.76	p = 0.79
rs1205
CC	110	35 (31.8%)	15.1 (14.1–16.0)	6.4 (6.0–6.9)
CT	115	21 (18.3%)	16.1 (15.2–17.0)	6.6 (6.2–7.1)
TT	41	7 (17.0%)	16.8 (15.3–18.3)	6.8 (6.1–7.6)
Missing values	0
Test statistics		Wald = 7.45, df = 1	β = 0.13, t = 2.16	β = 0.05, t = 0.91
p-Value for trend		p = 0.01	p = 0.03	p = 0.37
rs1800947
GG	231	59 (25.5%)	15.7 (15.0–16.3)	6.5 (6.2–6.9)
GC/CC	35	4 (11.4%)	16.6 (15.0–18.2)	6.9 (6.1–7.7)
Missing values	0
Test statistics		Wald = 2.11, df = 1	β = 0.07, t = 1.07	β = 0.05, t = 0.80
p-Value for trend		p = 0.15	p = 0.29	p = 0.42
rs1417938
TT	135	25 (18.5%)	16.3 (15.4–17.1)	6.7 (6.2–7.1)
TA	108	31 (28.7%)	15.5 (14.6–16.4)	6.6 (6.1–7.1)
AA	21	5 (23.8%)	14.2 (12.1–16.2)	5.9 (4.8–7.0)
Missing values	2
Test statistics		Wald = 4.33, df = 1	β = − 0.12, t = − 2.03	β = − 0.06, t = − 1.04
p-Value for trend		p = 0.04	p = 0.04	p = 0.30
rs3091244
CC	107	16 (15.0%)	16.5 (15.5–17.4)	6.6 (6.1–7.1)
CT/CA	97	31 (32.0%)	16.0 (15.1–17.0)	7.0 (6.5–7.5)
TT/TA/AA	33	8 (24.2%)	14.2 (12.6–15.9)	5.9 (5.0–6.8)
Missing values	29
Test statistics		Wald = 5.19, df = 1	β = − 0.13, t = − 2.08	β = − 0.04, t = − 0.57
p-Value for trend		p = 0.02	p = 0.04	p = 0.57

Note: Percentages of subjects with plasma CRP concentrations >3 mg/l are presented. Rs3091244 is a triallelic SNP. Geometric estimated means adjusted for age, gender, smoking status (yes/no), body mass index (BMI), and group (presence/absence of affective disorder) with 95% confidence intervals are presented for AUC_CAR (nmol/l/h; area under curve for CAR) and AUC_diurnal (nmol/l/h; area under curve for diurnal decline in salivary cortisol concentration). p-Values of the binary logistic regression analysis (plasma CRP) and linear regression analyses (cortisol measures) were used to test for linear trend.

Associations with individual CRP genotypes (Table I)

Plasma CRP concentrations showed the expected differences among subjects with different genotypes of rs2808628 (Wald statistics = 7.00, df = 1, p = 0.01), rs1205 (Wald statistics = 7.45, df = 1, p = 0.01), rs1417938 (Wald statistics = 4.33, df = 1, p = 0.04), and rs3091244 (Wald statistics = 5.19, df = 1, p = 0.02). rs1800947 showed a trend toward significance (p = 0.08), but this trend was lost after adjustment for covariates (Wald statistics = 2.11, df = 1, p = 0.15). The genotypes of rs2808630 did not show differences in plasma CRP concentrations (p>0.05).

The CRP polymorphisms that were related to differences in plasma CRP concentrations also showed significant differences in cortisol AUC_CAR values, i.e. rs1205 (β = 0.13, t = 2.16, p = 0.03), rs1417938 (β = − 0.12, t = − 2.03, p = 0.04), rs3091244, and a trend for rs2808628 (β = 0.11, t = 1.89, p = 0.06). No significant associations were found between CRP polymorphisms and cortisol AUC_diurnal values.

The CRP concentrations, as dichotomous variable, were not significantly associated with cortisol AUC_CAR values and cortisol AUC_diurnal values (respectively, t = 1.13, p = 0.26, t = 0.46, p = 0.65).

Associations with CRP haplotypes (Table II and Figure 2A–C)

Five common haplotypes with ≥ 1% frequencies were observed (Figure 1). We numbered haplotypes in our dataset according to previous published increasing plasma CRP concentrations from haplotypes 1 to 5 (Carlson et al. 2005; Bos et al. 2008) and we confirmed the order in our dataset by showing increasing concentrations of CRP in the order from haplotypes 1 to 5 (Wald statistics = 8.08, df = 1, p = 0.004). A statistically significant linear trend, but in reverse direction, was also found for AUC_CAR values in the order from haplotypes 1 to 5 (β = − 0.09, t = − 2.21, p = 0.03). No significant associations were found between haplotypes and the cortisol AUC_diurnal values (β = − 0.04, t = − 0.87, p = 0.38) (Table II). For illustrative purposes, Figure 2 demonstrates the linear trend according to the five CRP haplotypes for plasma CRP concentrations and salivary cortisol concentrations across the day (Figure 2A–C).

Table II.  Associations between CRP haplotypes with plasma CRP levels and salivary cortisol measures in 266 subjects.

Haplotype	Genotype combination	n	High CRP (%)	AUCCAR (nmol/l/h)	AUCdiurnal (nmol/l/h)
1	ATTCTC	36 (6.8%)	4 (11.4%)	16.1 (14.5–17.7)	6.8 (6.1–7.6)
2	ATTGTC	153 (28.8%)	31 (21.0%)	16.2 (15.4–17.0)	6.3 (6.2–7.0)
3	GCCGTC	140 (26.3%)	34 (24.6%)	15.6 (14.8–16.4)	6.6 (6.2–7.0)
4	GTCGAT	151 (28.4%)	43 (29.1%)	15.0 (14.3–15.8)	6.4 (6.0–6.8)
5	GTCGTA	39 (7.3%)	13 (34.2%)	15.0 (13.5–16.5)	6.6 (5.9–7.4)
Other haplotypes		13 (2.4%)
p-Value for trend			Wald = 8.08, df = 1, p = 0.004	β = − 0.09, t = − 2.21, p = 0.03	β = − 0.04, t = − 0.87, p = 0.38

Note: Five common haplotypes were constructed from rs2808628, rs2808630, rs1205, rs1800947, rs1417938, and rs3091244. Percentages of subjects with plasma CRP concentrations >3 mg/l are presented. Geometric estimated means adjusted for age, gender, smoking status (yes/no), BMI, and group (presence/absence of affective disorder) with 95% confidence intervals are presented for AUC_CAR (nmol/l/h; area under curve for CAR) and AUC_diurnal (nmol/l/h; area under curve for diurnal decline in salivary cortisol concentration). p-Values of the binary logistic regression analysis (plasma CRP) and linear regression analyses (cortisol measures) were used to test for the linear trend. As each subject represents two CRP haplotypes, statistical analysis was performed for 532 cases (i.e. 2 × 266 subjects).

Figure 2.  Graphical representation of associations between five common CRP haplotypes and plasma CRP and salivary cortisol concentrations across the day in 266 subjects. Percentages of subjects with elevated plasma CRP levels (>3 mg/l) are presented (A) Cortisol assessments in the first hour after awakening (B) and during the diurnal decline (C) are presented. (B) Demonstrates the linear trend in the CAR according to CRP haplotypes. Data are mean ± SEM, from Tables I and II. Statistical analysis results are in Tables I and II.

In order to disentangle which part of the CAR mostly contributes to the significant associations between haplotypes and cortisol concentrations after awaking, we repeated the regression analyses by entering other often used (composite) measures of salivary cortisol during the first hour of awakening: cortisol awakening peak (highest level at T2/T3), cortisol awakening rise (peak − T1), and the cortisol concentrations at awakening (T1). We found that the cortisol awakening peak and the cortisol concentrations at awakening showed a significant linear trend in the order from haplotypes 1 to 5 (β = − 0.09, t = − 2.06, p = 0.04, and β = − 0.11, t = − 2.65, p = 0.01; respectively). For the cortisol awakening rise, no association with haplotypes was found (β = 0.003, t = 0.08, p = 0.94).

The haplotype frequencies of haplotypes 1–5 in psychiatric subjects (n = 94) were, respectively, 3.2, 33.9, 25.3, 28.0, and 9.7%, and in the nonpsychiatric subjects (n = 172), respectively, 9.0, 27.0, 29.9, 29.7, and 6.3%. The frequency of the individual haplotypes did not differ significantly between psychiatric and nonpsychiatric subjects (χ²-test for linear-by-linear trend = 1.7, p = 0.33), and the effect of group in binary logistic and linear regression analyses with plasma CRP concentrations and AUC_CAR was also not significant (OR = 1.49, 95% CI = 0.87–2.54, p = 0.14, and β = − 0.41, t = − 0.85, p = 0.40, respectively).

Discussion

The main finding of the present study is that CRP haplotypes that are associated with differences in lifetime plasma CRP levels were also associated with differences in salivary cortisol concentrations during the awakening response. More specifically, the present data showed that variants of the CRP gene which were associated with higher CRP concentrations under basal, noninflammatory conditions, were also associated with lower cortisol concentrations after awakening. These findings support the hypothesis of a relationship between the CRP and the CAR. To the best of our knowledge, this is the first study that investigated associations between CRP gene polymorphisms and cortisol levels using a Mendelian randomization design.

Our finding of associations between CRP polymorphisms (i.e. rs3091244, rs1417938, rs2808628, and rs1205) and plasma CRP concentrations is consistent with the previously published data (Carlson et al. 2005; Hage and Szalai 2007; Bos et al. 2008). In addition, we showed significant associations between the same CRP polymorphisms and cortisol concentrations during the CAR. Extended analyses showed that the differences in CAR could be mostly ascribed to differences in cortisol awakening peak levels and cortisol levels at awakening and not to the cortisol awakening rise. This might indicate that CRP polymorphisms are not related to differences in the cortisol rise, but to differences in total cortisol release after awaking. No associations between CRP polymorphisms and cortisol release during the diurnal decline were found, suggesting that genetic variation in the CRP gene only affects cortisol levels after awakening under basal, noninflammatory conditions. The effects on the awakening response may thus be a manifestation of the multiple effects of the CRP gene. This is in accordance with findings of previous studies that evaluated the heritability of cortisol levels. In a twin study of 52 monozygotic (MZ) and 52 dizygotic (DZ) twin pairs, it was concluded that about 40–48% of individual differences in cortisol release in the first hour after awakening was attributable to genetic factors, whereas no genetic influence was found for the remaining diurnal decline (Wust et al. 2000a). This pattern was replicated in a larger study of 199 MZ and 272 DZ adult twin pairs, in which the cortisol levels at awakening and 30 min later showed around 30% heritability, with no significant effects for values recorded later during the day (Kupper et al. 2005).

Furthermore, we confirmed the clear linear trend in increasing plasma CRP levels from haplotypes 1 to 5 (Carlson et al. 2005; Bos et al. 2008), and we also found a linear trend, but in the inverse direction, between CRP haplotypes and cortisol levels after awakening. The relationship between CRP and cortisol levels has scarcely been investigated. In general, stressful and proinflammatory conditions, such as acute infections, may increase both CRP and cortisol levels, due to the activation of the immune system and HPA axis (Birjmohun et al. 2007; Steptoe et al. 2008). In contrast, several studies have shown that in chronic inflammatory conditions, such as rheumatoid arthritis and Sjögren's syndrome (Straub et al. 2002; Eijsbouts et al. 2005; Johnson et al. 2006), cortisol levels were not elevated, despite the presence of chronic high circulating inflammatory cytokine and CRP levels (Straub et al. 2002; Eijsbouts et al. 2005). It could be hypothesized that a chronic proinflammatory state with high CRP levels ultimately results in the adaptation of the HPA axis, through negative feedback mechanisms, leading to lower cortisol levels. As persistent high cortisol concentrations would predispose to infections (Straub and Cutolo 2006), such a cortisol decrease may be adaptive. Our findings are consistent with this hypothesis of a negative feedback of CRP on the HPA axis. Here, we showed that genetically induced higher CRP levels are associated with lower cortisol levels. Hence, CRP effects on cortisol may play a role in maintaining homeostasis during stress. However, in contrast with the direct effects of CRP polymorphisms on CRP levels, the mechanism of the effect on cortisol levels likely involves a much more complex network of endogenous interactions that include cytokines.

Some methodological limitations of the present study should be mentioned. Firstly, plasma CRP levels were not analyzed with a high sensitivity assay and the lower limit of detection was 3 mg/l. Nevertheless, the sensitivity was sufficient to confirm the well-known associations between CRP polymorphisms and CRP levels found in previous studies. Secondly, our study may have been underpowered with a relatively small sample size. Data were collected as part of studies on psychiatric and neuroendocrine correlates of the HPA axis. To increase the statistical power in our analyses, we allocated physically healthy subjects with and without a psychiatric disorder to one experimental group. Although psychiatric and nonpsychiatric subjects did not differ in frequency of individual haplotypes, replication in a larger, more uniform and healthy cohort is required. Thirdly, the HPA axis may be involved in the pathology of psychiatric disorders, that some of the subjects were suffering from. However, post hoc analyses showed no statistical differences between nonpsychiatric and psychiatric subjects in cortisol levels, neither during the awakening response nor for the diurnal decline. Moreover, there was no interaction with psychopathology for our main effects. Therefore, the associations between CRP polymorphisms and cortisol levels are unlikely due to confounding by psychopathology. Fourthly, although we adjusted for several potential confounders, other confounders that are likely to affect CRP and/or cortisol levels might have influenced our outcome, such as acute undetected (viral) infections. Finally, the design of the study needs to be considered. Our study made use of naturally occurring genetic variation resulting from independent gene assortment (i.e. Mendelian randomization), as the basis of our analyses of association. Variants of nearby genes in LD with the CRP polymorphisms may also explain our findings.

We conclude that genetic variants in the CRP gene are associated with differences in plasma CRP and salivary cortisol levels after awakening under basal, noninflammatory conditions, indicating that the CRP metabolism and cortisol release during the awakening response are closely linked.

Acknowledgements

The authors thank Prof. Dr P.E. Slagboom and Dr H.E.D. Suchiman of the Molecular Epidemiology Section for SNP genotyping and their valuable comments. Dr N. van Leeuwen of the Division of Medical Pharmacology/LACDR is thanked for DNA isolation and her assistance with genotyping.

Funding: The funding for this study was provided by the Leiden University Medical Center (LUMC) and Rivierduinen Mental Health Center. Part of this study was funded by a grant from the Netherlands Brain Foundation (Hersenstichting Nederland, Grant No. 15F07(2).24).

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.