A variety of biological processes, including steroid hormone secretion, have circadian rhythms, which are influenced by nine known circadian genes. Previously, we reported that certain variants in circadian genes were associated with risk for prostate cancer. To provide some biological insight into these findings, we examined the relationship of five variants of circadian genes, including NPAS2 (rs2305160:G>A), PER1 (rs2585405:G>C), CSNK1E (rs1005473:A>C), PER3 (54-bp repeat length variant), and CRY2 (rs1401417:G>C), with serum levels of sex steroids and insulin-like growth factor (IGF)-I and IGF-binding protein 3 (IGFBP3) in 241 healthy elderly Chinese men (mean age of 71.5). Age-adjusted and waist-to-hip ratio–adjusted ANOVA followed by likelihood ratio tests (LRT) showed that the NPAS2 variant A allele was associated with lower free and bioavailable testosterone (PLRT = 0.02 and 0.01, respectively) compared with the GG genotype. In addition, the PER1 variant was associated with higher serum levels of sex hormone-binding globulin levels (Ptrend = 0.03), decreasing 5α-androstane-3α, 17β-diol glucuronide levels (Ptrend = 0.02), and decreasing IGFBP3 levels (Ptrend = 0.05). Furthermore, the CSNK1E variant C allele was associated with higher testosterone to dihydrotestosterone ratios (PLRT = 0.01) compared with the AA genotype, whereas the longer PER3 repeat was associated with higher serum levels of IGF-I (PLRT = 0.03) and IGF-I to IGFBP3 ratios (PLRT = 0.04). The CRY2 polymorphism was not associated with any biomarkers analyzed. Our findings, although in need of confirmation, suggest that variations in circadian genes are associated with serum hormone levels, providing biological support for the role of circadian genes in hormone-related cancers. (Cancer Epidemiol Biomarkers Prev 2008;17(11):3268–73)

Circadian rhythms are the daily oscillations of multiple biological processes driven by endogenous clocks with or without external cues (1). These rhythms help maintain human sleep patterns and influence certain biological processes, such as sex hormone secretion (2, 3). Nine identified genes control the endogenous circadian rhythms via a transcription-translation feedback loop and include neuronal PAS domain protein 2 (NPAS2), casein kinase 1, ε (CSNK1E), cryptochrome 1 (CRY1), CRY2, period 1 (PER1), PER2, PER3, clock homologue (mouse) (CLOCK), and aryl hydrocarbon receptor nuclear translocator-like (ARNTL; ref. 4).

In an earlier report, we showed that certain circadian gene variants were associated with increased risk of prostate cancer (5). Although the exact mechanism is unclear, it is possible that the link between circadian genes and prostate cancer is mediated through hormones because steroid hormones play an important role in prostate cancer.

There is ample evidence that certain circulating hormone levels, such as testosterone, oscillate in a diurnal rhythm (2, 3). The central circadian clock located in the brain influences sex hormone secretion via the hypothalamic-pituitary-gonadal axis. Sleep fragmentation, in normal men and those with obstructive sleep apnea, resulted in elevated testosterone levels sustained throughout the night (6, 7). In mouse models, CLOCK mutants lacked the appropriate circadian signals required to coordinate hypothalamic hormone secretion, which resulted in reproduction disruptions in female mice (8). Because testosterone influences the growth hormone/insulin-like growth factor (IGF)-I pathway (9), IGF-I and its binding protein, IGFBP3, may also be influenced by circadian genes. Taken together with findings that variants in circadian genes are associated with delayed sleep phase syndrome and morning/evening preference (10-12), it is plausible that variants in these genes are associated with varying serum hormone levels.

To provide biological insight into the relationship between circadian genes and cancer, we examined the association between five circadian gene variants, included in our earlier investigation (5), and serum sex steroid, IGF-I, and IGFBP3 levels in 241 healthy Chinese men. Three of the five variants examined were originally chosen based on putative function by amino acid conservation analysis [PER1 rs2585405:G>C (Ala962Pro), NPAS2 rs2305160:G>A (Ala394Thr), and PER3 54-bp repeat length variant], whereas two variants were chosen because they had minor allele frequencies >5% (CRY2 rs1401417:G>C and CSNK1E rs1005473:A>C).

### Study Population

Details of this population-based study conducted in Shanghai, China have been reported previously (13). Briefly, healthy men were randomly selected from the general population in Shanghai between 1993 and 1995 and recruited as controls for a prostate cancer case-control study. Four hundred ninety-five eligible subjects were identified and 472 (95%) were interviewed, using a structured questionnaire, to collect information on selected demographic characteristics, including smoking history and alcohol use. Anthropometric measurements were also obtained as part of the interview, permitting the calculation of usual adult body mass index [BMI; weight (kg) / height (m)2] and waist-to-hip ratio (WHR). Of the 472 men with interview data, 330 men (70%) provided 20 mL of overnight fasting blood (collected between 7:00 a.m. and 10:00 a.m.) that was processed within 4 h of collection and stored at −70°C. Serum hormone and genotyping data were available for 241 (73%) of the 330 participants for this study. The study was approved by the Institutional Review Boards at the National Cancer Institute and the Shanghai Cancer Institute.

### Serum Hormone and Genetic Analysis

Total testosterone, dihydrotestosterone (DHT), estradiol, sex hormone-binding globulin (SHBG), and 5α-androstane-3α, 17β-diol glucuronide (3α-diol G) were assayed at the University of Southern California by RIAs, as previously reported (13). Testosterone, DHT, and estradiol were quantified following organic solvent extraction and Celite column partition chromatography using ethylene glycol as the stationary phase (13). SHBG and 3α-diol G were measured directly in serum using commercial kits (Diagnostic Systems Laboratories); the intraassay and interassay coefficients of variation ranged from 4% to 8% and 10% to 13%, respectively. Total testosterone, DHT, and SHBG measurements were used to calculate free and bioavailable testosterone (14), as well as molar ratios of testosterone to DHT (T:DHT), using the molecular weights of testosterone (288.4 g/mol) and DHT (290.4 g/mol) as conversion factors.

Plasma IGF-I and IGFBP3 were assayed by Diagnostic Systems Laboratories using ELISA methods as previously described (15). The lower limits of detection of IGF-I and IGFBP3 were 0.03 and 0.04 ng/mL, respectively. For both analytes, each sample was assayed twice, and the mean of the two determinations was used for data analysis. Samples for which the relative difference between the two determinations exceeded 10% were repeated. Split samples (n = 45) from a single individual were included among the study samples and used to assess laboratory reproducibility. The coefficients of variation for these split samples were 11.2% for IGF-I and 17.3% for IGFBP3. Molar ratios of IGF-I to IGFBP3 were calculated by the following formula (16):

$\mathrm{IGF\ {-}\ I:IGFBP3\ =\ [IGF\ {-}\ I(ng/mL)\ {\times}\ 0.130]/[IGFBP3(ng/mL)\ {\times}\ 0.036]}$

Genomic DNA extracted from buffy coat was used for genotyping circadian gene variants at Yale University as previously reported (5). Approximately 5% of the samples were duplicated for quality control and two reviewers independently scored the genotypes to confirm all results. The PER3 repeat length variation has either four or five copies of a 54-bp repetitive sequence in exon 18 (Genbank accession no. AB047686). Details of the PCR-based sequence length variant analysis are described elsewhere (17). The other four variants were single nucleotide polymorphisms of CRY2 (rs1401417:G>C), CSNK1E (rs1005473:A>C), NPAS2 (rs2305160:G>A), and PER1 (rs2585405:G>C). Taqman Assays-on-Demand primers and probes (Applied Biosystems) were used according to the manufacturer's instructions and have been previously described (18). Alleles for all single nucleotide polymorphisms are identified according to the SNPper online database (19).

### Statistical Analyses

All statistical analyses were carried out using STATA statistical software (StataCorp LP). Genotype frequencies for all gene variants were tested for Hardy-Weinberg equilibrium. ANOVA was used to compare differences in serum biomarkers and biomarker ratios among subjects with different genotypes for each single nucleotide polymorphism, and likelihood ratio tests (LRT) comparing models with and without genotyping data were done to determine if genetic variations contribute significantly to the model. Homozygous minor allele genotype categories that had less than five individuals were consolidated with corresponding heterozygous genotype categories with the understanding that for these markers, only a dominant effect could be detected; LRT comparing these models has 1 degree of freedom (df). If three genotype groups were available for analyses, LRT has 2 df and linear regression analysis was used to determine significance of monotonic dose-response relationships with the genotypes coded as a continuous variable (0, 1, and 2). Outcome variables were transformed (by log10, inverse, or square root) when necessary for normalization. All models were adjusted for categorical variables of age with age categories of ≤65, 66 to 75, and >75 and WHR or BMI using median values (0.89 and 21.5 kg/m2, respectively) as cutoff due to limited sample size. WHR and BMI were also assessed (as continuous variables) to determine if associations existed between variants in circadian genes and markers of adiposity. The associated P values were calculated for the two-sided tests with α = 0.05.

Selected demographic and serologic characteristics of the participants are shown in Table 1. The mean age of the study participants was 71.5. The study population was lean, with a mean BMI of 22.0 kg/m2 and WHR of 0.89. Minor allele frequencies for each variant analyzed ranged from 6.9% for the CSNK1E rs1005473 C allele to 47.2% for the PER1 rs2585405 C allele. Among participants, genotype frequency distributions for the five circadian gene variants were in Hardy-Weinberg equilibrium (P > 0.05; data not shown). No statistically significant associations were seen between markers of adiposity (WHR and BMI) and the five variants of circadian genes (data not shown).

Table 1.

Selected demographic, serologic, and genetic characteristics of study participants

CharacteristicParticipants (n = 241)
MeanSD
Age 71.5 6.5
BMI (kg/m222.0 3.4
WHR 0.89 0.06
Testosterone (ng/dL) 645.3 228.1
Free testosterone (ng/dL) 12.8 4.2
Bioavailable testosterone (ng/dL) 302.7 103.7
DHT (ng/dL) 67.4 27.6
T:DHT (molar ratio) 10.3 3.3
3α-diol G (ng/dL) 573.5 380.1
SHBG (nmol/L) 40.8 20.1
IGF-I (ng/mL) 124.0 42.5
IGFBP3 (ng/mL) 2,764.9 840.4
IGF-I:IGFBP3 (molar ratio) 0.165 0.046

n

%

NPAS2 (rs2305160:G>A)
GG 140 58.1%
GA 90 37.3%
AA 11 4.6%
PER1 (rs2585405:G>C)
GG 62 28.4%
GC 106 48.6%
CC 50 22.9%
CSNK1E (rs1005473:A>C)
AA 202 86.7%
AC 30 12.9%
CC 0.4%
PER3 [54-bp repeat polymorphism (Genbank AB047686)]
4-/4-repeat 180 76.3%
4-/5-repeat 53 22.5%
5-/5-repeat 1.3%
CRY2 (rs1401417:G>C)
GG 188 79.0%
GC 46 19.3%
CC 1.7%
CharacteristicParticipants (n = 241)
MeanSD
Age 71.5 6.5
BMI (kg/m222.0 3.4
WHR 0.89 0.06
Testosterone (ng/dL) 645.3 228.1
Free testosterone (ng/dL) 12.8 4.2
Bioavailable testosterone (ng/dL) 302.7 103.7
DHT (ng/dL) 67.4 27.6
T:DHT (molar ratio) 10.3 3.3
3α-diol G (ng/dL) 573.5 380.1
SHBG (nmol/L) 40.8 20.1
IGF-I (ng/mL) 124.0 42.5
IGFBP3 (ng/mL) 2,764.9 840.4
IGF-I:IGFBP3 (molar ratio) 0.165 0.046

n

%

NPAS2 (rs2305160:G>A)
GG 140 58.1%
GA 90 37.3%
AA 11 4.6%
PER1 (rs2585405:G>C)
GG 62 28.4%
GC 106 48.6%
CC 50 22.9%
CSNK1E (rs1005473:A>C)
AA 202 86.7%
AC 30 12.9%
CC 0.4%
PER3 [54-bp repeat polymorphism (Genbank AB047686)]
4-/4-repeat 180 76.3%
4-/5-repeat 53 22.5%
5-/5-repeat 1.3%
CRY2 (rs1401417:G>C)
GG 188 79.0%
GC 46 19.3%
CC 1.7%

Table 2 shows the age-adjusted means and SDs of serum androgens in relation to genotypes of circadian genes. LRT comparing models with and without the genotyping data shows that the NPAS2 rs2305160 variant A allele (or genotypes GA + AA) was associated with decreased levels of free and bioavailable testosterone (PLRT = 0.02 and 0.01, respectively) but not with other markers when compared with the GG genotype; comparisons between different genotype groups of NPAS2 variation also show significant differences in free and bioavailable testosterone levels (PLRT = 0.04 and 0.03, respectively). The PER1 rs2585405 C allele was associated with lower 3α-diol G (PLRT = 0.02) and higher SHBG (PLRT = 0.04) levels; the relationship between the analytes and genotype categories seems to be linear (Ptrend = 0.04 and 0.03, respectively). Furthermore, although no associations were seen between the five genetic variants analyzed and levels of total testosterone or DHT, the C allele of the CSNK1E variant was associated with a higher value of T:DHT (11.7) compared with men with the AA genotype (10.0; PLRT = 0.01); this variant was also borderline associated with 3α-diol G (PLRT = 0.06). No associations were seen for the CRY2 variant with any of the markers analyzed. These results did not change materially after adjustment for BMI instead of WHR (data not shown).

Table 2.

Serum levels of sex steroid hormones and SHBG among study participants in relation to variants of circadian genes

Testosterone (ng/dL)
Calculated free testosterone (ng/dL)
Calculated bioavailable testosterone (ng/dL)
T:DHT (molar ratio)
DHT (ng/dL)
3α-diol G (ng/dL)
SHBG (nmol/L)
NPAS2 (rs2305160:G>A)
GG 140 656.3 226.5 13.3 3.9 312.6 92.4 64.8 25.0 10.1 3.2 485.6 298.1 35.2 14.8 5.0 1.4
GA 90 621.5 225.0 12.0 3.9 279.9 92.9 64.2 25.2 10.1 3.2 462.7 281.7 38.6 16.0 4.8 1.4
AA 11 635.0 226.0 13.1 3.9 306.9 92.2 71.9 25.4 8.8 2.9 548.6 294.7 36.2 13.8 4.6 1.3
PLRT = 0.50  PLRT = 0.04  PLRT = 0.03  PLRT = 0.68  PLRT = 0.42  PLRT = 0.50  PLRT = 0.27  PLRT = 0.39
Ptrend = 0.30  Ptrend = 0.06  Ptrend = 0.05  Ptrend = 0.74  Ptrend = 0.32  Ptrend = 0.89  Ptrend = 0.25  Ptrend = 0.18
GA + AA 101 623.0 224.7 12.1 3.9 282.9 92.9 65.0 25.3 10.0 3.2 471.5 287.8 38.3 15.9 4.8 1.4
PLRT§ = 0.24  PLRT§ = 0.02  PLRT§ = 0.01  PLRT§ = 0.97  PLRT§ = 0.50  PLRT§ = 0.78  PLRT§ = 0.14  PLRT§ = 0.18
PER1 (rs2585405:G>C)
GG 62 603.4 230.5 13.0 4.0 304.7 93.5 61.7 24.7 9.8 3.2 542.0 322.0 33.0 13.6 5.0 1.4
GC 106 649.7 227.0 12.7 4.0 297.0 94.2 65.7 25.8 10.2 3.3 448.7 275.3 37.1 15.6 4.9 1.4
CC 50 653.2 226.7 12.7 4.0 296.8 92.7 67.2 25.5 9.8 3.2 421.4 250.2 38.9 15.9 5.0 1.4
PLRT = 0.36  PLRT = 0.88  PLRT = 0.85  PLRT = 0.45  PLRT = 0.80  PLRT = 0.06  PLRT = 0.08  PLRT = 0.81
Ptrend = 0.21  Ptrend = 0.64  Ptrend = 0.64  Ptrend = 0.23  Ptrend = 0.99  Ptrend = 0.02  Ptrend = 0.03  Ptrend = 0.80
GC + CC 156 650.9 226.3 12.7 4.0 296.9 93.5 66.2 25.8 10.1 3.3 439.7 271.7 37.7 15.9 4.9 1.4
PLRT§ = 0.16  PLRT§ = 0.61  PLRT§ = 0.58  PLRT§ = 0.23  PLRT§ = 0.71  PLRT§ = 0.02  PLRT§ = 0.04  PLRT§ = 0.59
CSNK1E (rs1005473:A>C)
AA 202 643.0 225.2 12.9 4.0 302.8 93.3 65.5 25.3 10.0 3.3 461.4 285.5 36.4 15.5 4.9 1.3
AC + CC 30 662.5 226.1 12.7 4.0 298.5 93.1 58.2 23.4 11.8 3.4 588.0 341.4 35.9 15.2 5.2 1.4
PLRT§ = 0.64  PLRT§ = 0.78  PLRT§ = 0.81  PLRT§ = 0.13  PLRT§ = 0.01  PLRT§ = 0.06  PLRT§ = 0.90  PLRT§ = 0.21
PER3 [54-bp repeat polymorphism (Genbank AB047686)]
4-/4-repeat 202 642.1 227.0 12.9 4.0 301.4 92.5 66.3 25.6 9.9 3.2 484.3 297.2 36.6 15.3 5.0 1.4
4-/5- and 5-/5-repeat 31 650.1 226.5 12.9 4.0 300.2 94.0 61.7 24.5 10.5 3.3 486.4 292.4 35.4 14.5 4.8 1.4
PLRT§ = 0.86  PLRT§ = 0.95  PLRT§ = 0.94  PLRT§ = 0.21  PLRT§ = 0.22  PLRT§ = 0.72  PLRT§ = 0.51  PLRT§ = 0.31
CRY2 (rs1401417:G>C)
GG 188 641.7 227.1 12.7 4.0 298.1 94.2 65.0 25.1 10.0 3.3 482.7 299.7 36.5 15.5 5.0 1.4
GC + CC 50 648.8 226.0 13.2 4.0 308.5 93.3 65.7 25.0 10.2 3.4 459.2 274.0 36.6 15.0 4.7 1.4
PLRT§ = 0.84  PLRT§ = 0.50  PLRT§ = 0.48  PLRT§ = 0.86  PLRT§ = 0.78  PLRT§ = 0.65  PLRT§ = 0.96  PLRT§ = 0.15
Testosterone (ng/dL)
Calculated free testosterone (ng/dL)
Calculated bioavailable testosterone (ng/dL)
T:DHT (molar ratio)
DHT (ng/dL)
3α-diol G (ng/dL)
SHBG (nmol/L)
NPAS2 (rs2305160:G>A)
GG 140 656.3 226.5 13.3 3.9 312.6 92.4 64.8 25.0 10.1 3.2 485.6 298.1 35.2 14.8 5.0 1.4
GA 90 621.5 225.0 12.0 3.9 279.9 92.9 64.2 25.2 10.1 3.2 462.7 281.7 38.6 16.0 4.8 1.4
AA 11 635.0 226.0 13.1 3.9 306.9 92.2 71.9 25.4 8.8 2.9 548.6 294.7 36.2 13.8 4.6 1.3
PLRT = 0.50  PLRT = 0.04  PLRT = 0.03  PLRT = 0.68  PLRT = 0.42  PLRT = 0.50  PLRT = 0.27  PLRT = 0.39
Ptrend = 0.30  Ptrend = 0.06  Ptrend = 0.05  Ptrend = 0.74  Ptrend = 0.32  Ptrend = 0.89  Ptrend = 0.25  Ptrend = 0.18
GA + AA 101 623.0 224.7 12.1 3.9 282.9 92.9 65.0 25.3 10.0 3.2 471.5 287.8 38.3 15.9 4.8 1.4
PLRT§ = 0.24  PLRT§ = 0.02  PLRT§ = 0.01  PLRT§ = 0.97  PLRT§ = 0.50  PLRT§ = 0.78  PLRT§ = 0.14  PLRT§ = 0.18
PER1 (rs2585405:G>C)
GG 62 603.4 230.5 13.0 4.0 304.7 93.5 61.7 24.7 9.8 3.2 542.0 322.0 33.0 13.6 5.0 1.4
GC 106 649.7 227.0 12.7 4.0 297.0 94.2 65.7 25.8 10.2 3.3 448.7 275.3 37.1 15.6 4.9 1.4
CC 50 653.2 226.7 12.7 4.0 296.8 92.7 67.2 25.5 9.8 3.2 421.4 250.2 38.9 15.9 5.0 1.4
PLRT = 0.36  PLRT = 0.88  PLRT = 0.85  PLRT = 0.45  PLRT = 0.80  PLRT = 0.06  PLRT = 0.08  PLRT = 0.81
Ptrend = 0.21  Ptrend = 0.64  Ptrend = 0.64  Ptrend = 0.23  Ptrend = 0.99  Ptrend = 0.02  Ptrend = 0.03  Ptrend = 0.80
GC + CC 156 650.9 226.3 12.7 4.0 296.9 93.5 66.2 25.8 10.1 3.3 439.7 271.7 37.7 15.9 4.9 1.4
PLRT§ = 0.16  PLRT§ = 0.61  PLRT§ = 0.58  PLRT§ = 0.23  PLRT§ = 0.71  PLRT§ = 0.02  PLRT§ = 0.04  PLRT§ = 0.59
CSNK1E (rs1005473:A>C)
AA 202 643.0 225.2 12.9 4.0 302.8 93.3 65.5 25.3 10.0 3.3 461.4 285.5 36.4 15.5 4.9 1.3
AC + CC 30 662.5 226.1 12.7 4.0 298.5 93.1 58.2 23.4 11.8 3.4 588.0 341.4 35.9 15.2 5.2 1.4
PLRT§ = 0.64  PLRT§ = 0.78  PLRT§ = 0.81  PLRT§ = 0.13  PLRT§ = 0.01  PLRT§ = 0.06  PLRT§ = 0.90  PLRT§ = 0.21
PER3 [54-bp repeat polymorphism (Genbank AB047686)]
4-/4-repeat 202 642.1 227.0 12.9 4.0 301.4 92.5 66.3 25.6 9.9 3.2 484.3 297.2 36.6 15.3 5.0 1.4
4-/5- and 5-/5-repeat 31 650.1 226.5 12.9 4.0 300.2 94.0 61.7 24.5 10.5 3.3 486.4 292.4 35.4 14.5 4.8 1.4
PLRT§ = 0.86  PLRT§ = 0.95  PLRT§ = 0.94  PLRT§ = 0.21  PLRT§ = 0.22  PLRT§ = 0.72  PLRT§ = 0.51  PLRT§ = 0.31
CRY2 (rs1401417:G>C)
GG 188 641.7 227.1 12.7 4.0 298.1 94.2 65.0 25.1 10.0 3.3 482.7 299.7 36.5 15.5 5.0 1.4
GC + CC 50 648.8 226.0 13.2 4.0 308.5 93.3 65.7 25.0 10.2 3.4 459.2 274.0 36.6 15.0 4.7 1.4
PLRT§ = 0.84  PLRT§ = 0.50  PLRT§ = 0.48  PLRT§ = 0.86  PLRT§ = 0.78  PLRT§ = 0.65  PLRT§ = 0.96  PLRT§ = 0.15
*

P value from LRTs comparing models with and without genotype data (df, 2).

P trend from genotype from linear regression analysis, adjusted for age and WHR.

§

P value from LRTs comparing models with and without genotype data (df, 1).

Similar to the patterns observed for sex steroids, associations were seen between variants of circadian genes and age-adjusted mean serum levels of IGF-I and IGFBP3 (Table 3). A borderline significant decreasing trend in IGFBP3 levels was seen across genotypes of the PER1 variant (PLRT = 0.05), although no significant associations were seen when comparing GC or CC with the GG genotype. In addition, men who had the longer PER3 repeat length variant had significantly higher serum IGF-I levels (mean, 133.7 ng/mL) and IGF-I:IGFBP3 ratios (mean, 0.173) compared with those with the homozygous four-repeat genotype [mean, 120.9 ng/mL (PLRT = 0.03) and mean, 0.158 ng/mL (PLRT = 0.04), respectively].

Table 3.

Serum levels of IGF-I and IGFBP3 among study participants in relation to variants of circadian genes

IGFBP3 (ng/mL)
IGF-I:IGFBP3 (molar ratio)
nMean*SDMean*SDMean*SD
NPAS2 (rs2305160:G>A)
GG 140 126.1 41.3 2,673.9 768.1 0.165 0.046
GA 90 120.0 41.3 2,662.6 761.8 0.156 0.045
AA 11 130.2 41.5 2,843.1 766.0 0.166 0.045
PLRT = 0.44  PLRT = 0.72  PLRT = 0.31
Ptrend = 0.59  Ptrend = 0.71  Ptrend = 0.30
GA + AA 101 121.1 41.3 2,681.9 764.9 0.157 0.045
PLRT§ = 0.36  PLRT§ = 0.92  PLRT§ = 0.17
PER1 (rs2585405:G>C)
GG 62 130.8 40.8 2,836.6 799.5 0.161 0.045
GC 106 124.2 40.8 2,633.3 753.2 0.165 0.046
CC 50 118.6 40.8 2,560.7 737.2 0.163 0.045
PLRT = 0.25  PLRT = 0.12  PLRT = 0.89
Ptrend = 0.10  Ptrend = 0.05  Ptrend = 0.84
GC + CC 156 122.4 40.8 2,609.9 749.9 0.164 0.046
PLRT§ = 0.16  PLRT§ = 0.05  PLRT§ = 0.67
CSNK1E (rs1005473:A>C)
AA 202 126.1 41.4 2,703.2 769.0 0.163 0.045
AC + CC 30 113.5 41.8 2,565.2 738.9 0.155 0.044
PLRT§ = 0.11  PLRT§ = 0.33  PLRT§ = 0.34
PER3 [54-bp repeat variant (Genbank AB047686)]
4-/4-repeat 202 120.9 40.8 2,676.9 768.4 0.158 0.045
4-/5- + 5-/5-repeat 31 133.7 41.0 2,682.2 759.5 0.173 0.047
PLRT§ = 0.03  PLRT§ = 0.92  PLRT§ = 0.04
CRY2 (rs1401417:G>C)
GG 188 125.4 40.9 2,708.1 765.3 0.162 0.046
GC + CC 50 120.9 40.9 2,584.9 744.0 0.162 0.045
PLRT§ = 0.49  PLRT§ = 0.31  PLRT§ = 0.96
IGFBP3 (ng/mL)
IGF-I:IGFBP3 (molar ratio)
nMean*SDMean*SDMean*SD
NPAS2 (rs2305160:G>A)
GG 140 126.1 41.3 2,673.9 768.1 0.165 0.046
GA 90 120.0 41.3 2,662.6 761.8 0.156 0.045
AA 11 130.2 41.5 2,843.1 766.0 0.166 0.045
PLRT = 0.44  PLRT = 0.72  PLRT = 0.31
Ptrend = 0.59  Ptrend = 0.71  Ptrend = 0.30
GA + AA 101 121.1 41.3 2,681.9 764.9 0.157 0.045
PLRT§ = 0.36  PLRT§ = 0.92  PLRT§ = 0.17
PER1 (rs2585405:G>C)
GG 62 130.8 40.8 2,836.6 799.5 0.161 0.045
GC 106 124.2 40.8 2,633.3 753.2 0.165 0.046
CC 50 118.6 40.8 2,560.7 737.2 0.163 0.045
PLRT = 0.25  PLRT = 0.12  PLRT = 0.89
Ptrend = 0.10  Ptrend = 0.05  Ptrend = 0.84
GC + CC 156 122.4 40.8 2,609.9 749.9 0.164 0.046
PLRT§ = 0.16  PLRT§ = 0.05  PLRT§ = 0.67
CSNK1E (rs1005473:A>C)
AA 202 126.1 41.4 2,703.2 769.0 0.163 0.045
AC + CC 30 113.5 41.8 2,565.2 738.9 0.155 0.044
PLRT§ = 0.11  PLRT§ = 0.33  PLRT§ = 0.34
PER3 [54-bp repeat variant (Genbank AB047686)]
4-/4-repeat 202 120.9 40.8 2,676.9 768.4 0.158 0.045
4-/5- + 5-/5-repeat 31 133.7 41.0 2,682.2 759.5 0.173 0.047
PLRT§ = 0.03  PLRT§ = 0.92  PLRT§ = 0.04
CRY2 (rs1401417:G>C)
GG 188 125.4 40.9 2,708.1 765.3 0.162 0.046
GC + CC 50 120.9 40.9 2,584.9 744.0 0.162 0.045
PLRT§ = 0.49  PLRT§ = 0.31  PLRT§ = 0.96
*

P value from LRTs comparing models with and without genotype data (df, 2).

P trend from genotype from linear regression analysis, adjusted for age and WHR.

§

P value from LRTs comparing models with and without genotype data (df, 1).

Results from this population-based study suggest that variations in circadian genes are associated with serum levels of androgens and IGF markers. In particular, NPAS2 rs2305160:G>A (Ala394Thr), PER1 rs2585405:G>C (Ala962Pro), and CSNK1E rs1005473:A>C are associated with increased levels of serum androgens, whereas there is a suggestive association between PER variants and IGF markers. Although in need of confirmation, these findings are consistent with observations that some circulating hormones oscillate in a circadian rhythm.

These findings support our previous report that selected variants of circadian genes are associated with altered prostate cancer risk (5). For example, the NPAS2 rs2305160 A allele, which was associated with lower free and bioavailable testosterone levels in this study, conferred a 20% reduced risk of prostate cancer (95% confidence interval, 0.5-1.1; ref. 5). In vivo studies have shown that androgens are important in the growth of normal and cancerous prostate cells (20, 21). However, epidemiologic studies of serum androgens have been inconclusive. Several studies report suggestive associations of androgens with increased prostate cancer risk, whereas others report null or even inverse associations (22, 23). One reason for these mixed results may be that serum androgen levels are indirect indicators of intraprostatic androgenicity and may not truly reflect androgen metabolism and action within the prostate (22). It is also unclear if androgen measurements taken at time of the study reflect causative androgen exposure because serum androgen levels are known to decrease with age (3) and are influenced by time of day and sleep patterns (6, 7).

It is noteworthy that the PER1 and SHBG genes reside ∼500 kb apart on chromosome 17p12-p13, given our finding that there was an association between PER1 rs2585405 and serum SHBG. Because of the proximity of these genes, it is plausible that there may be some degree of linkage between these two genes and/or that they have gene expression regulatory elements in common. However, SHBG does not seem to be expressed in a circadian pattern (2, 3), suggesting that the relationship between PER1 and SHBG cannot be explained by proximity of the genes alone and requires further investigation.

Because serum testosterone influences the growth hormone/IGF-I pathway (9), the suggestive association between variants in circadian genes and IGF-I, IGFBP3, and IGF-I:IGFBP3 (an indicator of free IGF-I) is not entirely unexpected. IGF-I is secreted primarily by the liver and is mostly bound by IGFBP3. Some but not all previous studies have shown that serum levels of IGF-I and IGFBP3 vary throughout the day in a circadian rhythm (24-29). In men with hypopituitarism, growth hormone treatment not only increased gene expression of IGF-I but also increased CLOCK expression and decreased PER1 expression in muscle cells (30). Our observation of a suggestive relationship between circadian genes (specifically that of PER genes) and the IGF pathway is consistent with these previous findings. Because altered risks for cancers of the prostate and breast have been linked to variants in circadian genes (5, 17, 31) and IGF-I (32), the potential interaction between circadian genes and IGF-I on cancer risk warrants further examination.

The unique strength of our study includes population-based samples with minimal selection bias. Because close to 70% of the study participants gave fasting blood for the study, it is unlikely that observed allele frequencies are related to response status. Quality control measures taken to minimize misclassification of genotyping included confirmation of genotyping results by two independent reviewers and assessment of duplicate samples, which showed high concordance. In addition, we had high-quality serum assays of sex hormones as well as IGF markers, with low intraassay and interassay variation (coefficients of variation, <18%).

Limitations of the study should be noted. First, we have limited gene coverage, with only one variant genotyped in each of the five circadian genes. Based on data from the International HapMap Consortium (33), ∼275 tag single nucleotide polymorphisms would be needed to cover the nine circadian genes adequately. There are also ∼45 putatively functional variants that need to be typed to capture any risks related to function of the circadian genes. Therefore, to fully capture the variations of these nine genes, ∼320 variants need to be genotyped. Second, we do not have data on sleep patterns and light exposure, which could also affect circadian rhythmicity of some hormones. Third, we have serum marker data from only one time point in the day (between 7 and 10 a.m.) and thus cannot determine how diurnal variations of the analytes will be affected by the genotypes studied; within the day, variations in serum steroid levels between early morning blood draws and other blood draw times have been observed (34, 35) but do not influence our results because we only collected blood in the early morning hours. Fourth, because the Shanghai population is relatively homogeneous, we have limited generalizability to other populations. In addition, prostate cancer screening in Shanghai is relatively uncommon during our study period (1990-1995), and thus, prostate cancer is detected at a later age, with a mean age of 71, which is slightly older than most prostate cancer case patients in the United States. This limits our ability to generalize our results to younger men, although relationships between serum hormones and circadian genes in younger men also exist in our study, but the number of men in this age group is small.

In conclusion, our population-based study of healthy Chinese men suggests that variants in circadian genes may be related to varying serum hormone levels. Future studies with larger sample sizes and more complete gene coverage are needed to confirm our findings. In addition, genes that are related to the metabolism and expression of these hormones should also be studied to determine if an interaction exists between circadian genes and hormone-related genes that would affect serum hormone levels.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the staff of the Shanghai Cancer Institute for specimen collection and processing, collaborating hospitals and urologists for data collection, local pathologists for pathology review, Shelley Niwa (Westat) and Gigi Yuan (Information Management Systems, Inc.) for data management and preparation, and Janis Koci (Scientific Applications International Corporation) for management of the biological samples.

1
Panda S, Hogenesch JB. It's all in the timing: many clocks, many outputs.
J Biol Rhythms
2004
;
19
:
374
–87.
2
Lejeune-Lenain C, Van Cauter E, Desir D, Beyloos M, Franckson JR. Control of circadian and episodic variations of adrenal androgens secretion in man.
J Endocrinol Invest
1987
;
10
:
267
–76.
3
Plymate SR, Tenover JS, Bremner WJ. Circadian variation in testosterone, sex hormone-binding globulin, and calculated non-sex hormone-binding globulin bound testosterone in healthy young and elderly men.
J Androl
1989
;
10
:
366
–71.
4
Fu L, Lee CC. The circadian clock: pacemaker and tumour suppressor.
Nat Rev Cancer
2003
;
3
:
350
–61.
5
Chu LW, Zhu Y, Yu K, et al. Variants in circadian genes and prostate cancer risk: a population-based study in China. Prostate Cancer Prostatic Dis 2007 [epub ahead of print].
6
Luboshitzky R, Zabari Z, Shen-Orr Z, Herer P, Lavie P. Disruption of the nocturnal testosterone rhythm by sleep fragmentation in normal men.
J Clin Endocrinol Metab
2001
;
86
:
1134
–9.
7
Luboshitzky R, Aviv A, Hefetz A, et al. Decreased pituitary-gonadal secretion in men with obstructive sleep apnea.
J Clin Endocrinol Metab
2002
;
87
:
3394
–8.
8
Miller BH, Olson SL, Turek FW, et al. Circadian clock mutation disrupts estrous cyclicity and maintenance of pregnancy.
Curr Biol
2004
;
14
:
1367
–73.
9
Meinhardt UJ, Ho KKY. Modulation of growth hormone action by sex steroids.
Clin Endocrinol (Oxf)
2006
;
65
:
413
–22.
10
Archer SN, Robilliard DL, Skene DJ, et al. A length polymorphism in the circadian clock gene Per3 is linked to delayed sleep phase syndrome and extreme diurnal preference.
Sleep
2003
;
26
:
413
–5.
11
Carpen JD, Archer SN, Skene DJ, Smits M, von Schantz M. A single-nucleotide polymorphism in the 5′-untranslated region of the hPER2 gene is associated with diurnal preference.
J Sleep Res
2005
;
14
:
293
–7.
12
Carpen JD, von Schantz M, Smits M, Skene DJ, Archer SN. A silent polymorphism in the PER1 gene associates with extreme diurnal preference in humans.
J Hum Genet
2006
;
51
:
1122
–5.
13
Hsing AW, Stanczyk FZ, Belanger A, et al. Reproducibility of serum sex steroid assays in men by RIA and mass spectrometry.
Cancer Epidemiol Biomarkers Prev
2007
;
16
:
1004
–8.
14
Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of simple methods for the estimation of free testosterone in serum.
J Clin Endocrinol Metab
1999
;
84
:
3666
–72.
15
Chokkalingam AP, Pollak M, Fillmore CM, et al. Insulin-like growth factors and prostate cancer: a population-based case-control study in China.
Cancer Epidemiol Biomarkers Prev
2001
;
10
:
421
–7.
16
Morimoto LM, Newcomb PA, White E, Bigler J, Potter JD. Variation in plasma insulin-like growth factor-1 and insulin-like growth factor binding protein 3: genetic factors.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
1394
–401.
17
Zhu Y, Brown HN, Zhang Y, Stevens RG, Zheng T. Period3 structural variation: a circadian biomarker associated with breast cancer in young women.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
268
–70.
18
Zhu Y, Leaderer D, Guss C, et al. Ala394Thr polymorphism in the clock gene NPAS2: a circadian modifier for the risk of non-Hodgkin's lymphoma.
Int J Cancer
2007
;
120
:
432
–5.
19
Riva A, Kohane IS. SNPper: retrieval and analysis of human SNPs.
Bioinformatics
2002
;
18
:
1681
–5.
20
Huggins C, Hodges CV. Studies on prostatic cancer. I. The effect of castration, of estrogen and androgen injection on serum phosphatases in metastatic carcinoma of the prostate.
Cancer Res
1941
;
1
:
293
–7.
21
Niu Y, Xu Y, Zhang J, et al. Proliferation and differentiation of prostatic stromal cells.
BJU Int
2001
;
87
:
386
–93.
22
Hsing AW, Reichardt JK, Stanczyk FZ. Hormones and prostate cancer: current perspectives and future directions.
Prostate
2002
;
52
:
213
–35.
23
Chokkalingam AP, Stanczyk FZ, Reichardt JKV, Hsing AW. Molecular epidemiology of prostate cancer: hormone-related genetic loci.
Front Biosci
2007
;
12
:
3436
–60.
24
Stratakis CA, Mastorakos G, Magiakou MA, et al. 24-Hour secretion of growth hormone (GH), insulin-like growth factors-I and -II (IGF1, -II), prolactin (PRL) and thyrotropin (TSH) in young adults of normal and tall stature.
Endocr Res
1996
;
22
:
261
–76.
25
Goodman-Gruen DD, Barrett-Connor EE. Epidemiology of insulin-like growth factor 1 in elderly men and women. The Rancho Bernardo Study.
Am J Epidemiol
1997
;
145
:
970
–6.
26
Oscarsson JJ, Johannsson GG, Johansson JJO, et al. Diurnal variation in serum insulin-like growth factor (IGF)-I and IGF binding protein 3 concentrations during daily subcutaneous injections of recombinant human growth hormone in GH-deficient adults.
Clin Endocrinol (Oxf)
1997
;
46
:
63
–8.
27
Heuck CC, Skjaerbaek CC, Orskov HH, Wolthers OOD. Circadian variation in serum free ultrafiltrable insulin-like growth factor I concentrations in healthy children.
Pediatr Res
1999
;
45
:
733
–6.
28
Skjaerbaek CC, Frystyk JJ, Kaal AA, et al. Circadian variation in serum free and total insulin-like growth factor (IGF)-I and IGFII in untreated and treated acromegaly and growth hormone deficiency.
Clin Endocrinol (Oxf)
2000
;
52
:
25
–33.
29
Haus EE, Dumitriu LL, Nicolau GGY, Bologa SS, Sackett-Lundeen LL. Circadian rhythms of basic fibroblast growth factor (bFGF), epidermal growth factor (EGF), insulin-like growth factor-1 (IGF-1), insulin-like growth factor binding protein 3 (IGFBP3), cortisol, and melatonin in women with breast cancer.
Chronobiol Int
2001
;
18
:
709
–27.
30
Sjogren K, Leung K-C, Kaplan W, et al. Growth hormone regulation of metabolic gene expression in muscle: a microarray study in hypopituitary men.
Am J Physiol Endocrinol Metab
2007
;
293
:
E364
–71.
31
Zhu Y, Stevens R, Leaderer D, et al. Non-synonymous polymorphisms in the circadian gene NPAS2 and breast cancer risk.
Breast Cancer Res Treat
2008
;
107
:
421
–5.
32
Renehan AG, Zwahlen M, Minder C, et al. Insulin-like growth factor (IGF)-I, IGF binding protein 3, and cancer risk: systematic review and meta-regression analysis.
Lancet
2004
;
363
:
1346
–53.
33
The International HapMap Consortium. A haplotype map of the human genome.
Nature
2005
;
437
:
1299
–320.
34
Guay A, Miller MG, McWhirter CL. Does early morning versus late morning draw time influence apparent testosterone concentration in men aged ≥45 years? Data from the Hypogonadism In Males study.
Int J Impot Res
2007
;
20
:
162
–7.
35
Crawford ED, Barqawi AB, O'Donnell C, Morgentaler A. The association of time of day and serum testosterone concentration in a large screening population.
BJU Int
2007
;
100
:
509
–13.