ABSTRACT
Background
Epigenetic mechanisms may underlie associations between maternal caffeine consumption and adverse childhood metabolic outcomes. However, limited studies have examined neonate DNA methylation (DNAm) patterns in the context of preconception or prenatal exposure to caffeine metabolites.
Objectives
We examined preconception and pregnancy caffeine exposure with DNAm alterations in neonate cord blood (n = 378).
Methods
In a secondary analysis of the Effects of Aspirin in Gestation and Reproduction Trial (EAGeR), we measured maternal caffeine, paraxanthine, and theobromine concentrations from stored serum collected preconception (on average 2 months before pregnancy) and at 8 weeks of gestation. In parallel, self-reported caffeinated beverage intake was captured via administration of questionnaires and daily diaries. We profiled DNAm from the cord blood buffy coat of singletons using the MethylationEPIC BeadChip. We assessed associations of maternal caffeine exposure and methylation β values using multivariable robust linear regression. A false discovery rate (FDR) correction was applied using the Benjamini-Hochberg method.
Results
In preconception, the majority of women reported consuming 1 or fewer servings/day of caffeine on average, and caffeine and paraxanthine metabolite levels were 88 and 36 µmol/L, respectively. Preconception serum caffeine metabolites were not associated with individual cytosine-guanine (CpG) sites (FDR >5%), though pregnancy theobromine was associated with DNAm at cg09460369 near RAB2A (β = 0.028; SE = 0.005; FDR P = 0.012). Preconception self-reported caffeinated beverage intake compared to no intake was associated with DNAm at cg09002832 near GLIS3 (β = –0.013; SE = 0.002; FDR P = 0.036). No associations with self-reported intake during pregnancy were found.
Conclusions
Few effects of maternal caffeine exposure on neonate methylation differences in leukocytes were identified in this population with relatively low caffeine consumption.
Keywords: DNA methylation, caffeine intake, mother-child dyads, maternal exposures, periconception
Introduction
Caffeine is widely consumed and naturally found in beverages and foods, including coffee, tea, and cocoa products (1). Current guidelines from the American College of Obstetricians and Gynecologists recommend pregnant women limit consumption to less than 200 milligrams per day (2). During pregnancy, the fetus is directly exposed to maternal caffeine intake, as caffeine and its metabolites are readily able to cross the placenta and enter fetal circulation (3). Caffeine is metabolized by the cytochrome P450 1A2 enzyme (CYP1A2) to produce paraxanthine and theobromine, which account for approximately 80% and 12% of caffeine metabolites, respectively (4).
Maternal caffeine exposure has been previously associated with long-term outcomes, including childhood overweight and obesity and liver fat deposits by age 10 years (5–9). These effects of maternal caffeine exposure on childhood outcomes may be mediated through epigenetic mechanisms (10–15). For example, caffeine intragastrically administered from gestational days 9 through 20 (at 30, 60, or 120 mg/kg per day) in pregnant rats is associated with histone acetylation and reduced expression of genes responsible for cholesterol synthesis in male offspring liver samples, including cAMP, sirtuin1, and protein kinase A (10). More specifically, differences in DNA methylation (DNAm) patterns, a primary epigenetic mechanism, have been associated with exposure to caffeine (ranging from gestational day 8.5 up until day 20 at 20+ mg/kg per day, or about 2–4 cups of coffee for humans) and its metabolites in animal and in vitro studies (11–15).
Few human studies have examined associations between caffeine exposure and DNAm status, and these studies were limited by only examining self-reported coffee or tea intake in nonpregnant adults (16–18). Chuang et al. (16) identified 11 cytosine-guanine (CpG) probes in blood associated with daily coffee consumption that were linked to lipid metabolism and immune response. Yet, another epigenome-wide association study (EWAS) from Ek et al. (17) found no significant associations between coffee consumption (ranging from 28.8 to 107 cups per month on average) and DNAm among men and women but observed a significant association between tea consumption in women and DNAm at probes cg18192808 (DNAJC16) and cg14055589 (TTC17). Factors such as the retrospective and self-reported measures of coffee and tea intake may explain the discrepancy in findings. A meta-analysis of 15 cohorts of 15,789 nonpregnant adults in the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium identified coffee-associated differences in DNAm at 11 CpGs (18). Self-reported coffee intake ranged from 0.6 cups per day to 3.5 cups per day. Furthermore, no epidemiologic studies examined these associations in the context of preconception exposure to caffeine or its metabolites. These associations may be important to consider, because DNAm is beginning to be established in early development (19).
Therefore, the aim of this study was to examine the association of maternal caffeine status during preconception and pregnancy with DNAm patterns in cord blood. To do this, maternal caffeine exposure was assessed by both maternal serum biomarkers of caffeine metabolites and self-reported caffeinated beverage intake. Epigenetic markers may provide new mechanistic insights into the impacts of maternal caffeine status.
Methods
Study design
We conducted a secondary analysis of the Effects of Aspirin in Gestation and Reproduction (EAGeR) trial (2007–2011; NCT00467363). EAGeR was a multicenter, double-blind clinical trial that randomized White women with a history of pregnancy loss to low-dose aspirin + folic acid compared with folic acid prior to conception to investigate effects on live birth and pregnancy loss (n = 1228) (20, 21). Among them, 48% (n = 595) delivered a live birth. The current analysis is nested among newborns from the Salt Lake City, Utah, study site, which recruited >80% of study participants. Cord blood was collected at Utah beginning in 2009 (2 years after enrollment began) and was successful for over 90% of deliveries thereafter (n = 428). We removed samples that had insufficient DNA or failed quality control checks (n = 37). We also excluded 12 participants of other races/ethnicities [i.e., Black (n = 4), American Indian (n = 1), Asian (n = 6), and unidentified (n = 1)] as there have been documented racial/ethnic differences in caffeine consumption and their sources (e.g., more tea consumed in Asian cultures) and genetic ancestry plays a role in establishment of DNAm (22, 23). Supplemental Figure 1 is a participant flow diagram for the final analytic sample (n = 378). The study was approved by the institutional review board at the University of Utah (Salt Lake City, Utah IRB #1,002,521), and all participants provided written informed consent prior to enrolling. We previously observed that randomization to low-dose aspirin had no impact on DNAm in cord blood (24, 25).
Caffeine metabolite assessment
Caffeine and 2 primary metabolites, paraxanthine and theobromine, were measured via LC-MS from serum samples collected from women at the baseline enrollment visit prior to randomization (preconception; average 2.60 ± 1.57 months prior to pregnancy) and again at 8 weeks of gestation (26). Coefficients of variation in the lowest range of detection were 9%–17%. Of the 363 preconception samples available for analysis, 65 (17.9%), 113 (31.1%), and 28 (7.7%) were below the limit of detection (LOD) of 0.04 µg/mL for caffeine, paraxanthine, and theobromine, respectively. Of the 345 eight-week gestation samples available for analysis, 136 (39.4%), 268 (77.7%), and 123 (35.6%) were below the LOD of 0.04 µg/mL for caffeine, paraxanthine, and theobromine, respectively. Values below the LOD were imputed as LOD/√2 (27) and were included in the analysis. For this analysis, we categorized serum caffeine metabolites into tertiles due to the skewness of each metabolite toward low concentrations. Preconception categories were defined as follows: caffeine, ≤0.08, 0.08–0.56, or >0.56 µg/mL; paraxanthine, ≤0.04, 0.04–0.18, or >0.18 µg/mL; and theobromine, ≤0.34, 0.34–1.05, or >1.05 µg/mL. Early pregnancy categories were defined as follows: caffeine, ≤0.03, 0.03–0.21, or >0.21 µg/mL; and theobromine, ≤0.03, 0.03–0.22, or >0.22 µg/mL. Early pregnancy paraxanthine was dichotomized according to the LOD (≥0.04 compared with <0.04 µg/mL) because the 33.3 and 66.7 percentiles were both below the LOD.
Caffeinated beverage intake assessment
Self-reported habitual caffeinated beverage intake was captured multiple times during the study. At baseline, prior to pregnancy, study participants completed lifestyle questionnaires capturing intake of coffee (cups/d), tea (cups/d), and caffeinated soda (servings/d) over the past 12 months. Servings were converted to standard servings/d, which corresponds to a 6 oz (177 mL) cup of coffee or tea and a 12 oz (355 mL) can of soda. For the analysis, we categorized each beverage intake by any cups/d or servings/d reported, compared with none. During the first 2 menstrual cycles after study enrollment, 330 (87.3%) women recorded the number of caffeinated beverages consumed on daily diaries completed at home. Similarly, a daily diary capturing total caffeinated beverage intake was also completed between 4 to 8 weeks of gestation. For this analysis, we averaged caffeinated beverage intake from the diaries to get a preconception average (i.e., diaries collected during the first 2 menstrual cycles) and an 8-weeks-gestation average. Intake was categorized as any amount compared with none. Lastly, questionnaires assessing the usual daily intake of caffeinated beverages over the past month were completed every 4 weeks beginning at 12 weeks gestation until 36 weeks. For this analysis, we categorized self-reported caffeinated beverage average intake over this period as none (i.e., no servings) or any amount.
DNAm measurement and processing
DNA measurement from cord blood samples has been previously described for this study (24). Briefly, prepared cord blood buffy coat was shipped and processed for DNA extraction and analysis at the University of Minnesota. DNA underwent bisulfate conversion (EZ DNA Methylation TM kit, Zymo Research) to differentiate unmethylated and methylated cytosines. DNAm was profiled using the Infinium MethylationEPIC BeadChip microarray (28, 29). Samples were randomly ordered to control for batch effects, and the sample plate and positions were tracked. The minfi package in R was used to process DNAm microarray data, including quantile normalization and background signal and dye-bias adjustment of probes (30). Probes with a detection P value greater than 0.01 were filtered and probes on the sex chromosomes were removed, leaving 815,112 probes for analysis. DNAm levels for each CpG probe were reported as β values ranging from 0 (unmethylated) to 1 (methylated). These β values were determined by calculating the ratio of methylated probe fluorescence intensity to the sum of the methylated and unmethylated probe intensities.
Statistical analysis
Linear regression models were used to assess a linear trend between self-reported caffeinated beverage intake and natural log-transformed serum caffeine metabolite concentrations. Spearman correlation coefficients (rs) were also used to compare within- and between-serum concentrations at preconception and 8 weeks of gestation. To assess the association of methylation β values (dependent variable) and each of the caffeine metabolites, we used multivariable robust linear regression. Separate models were created for each metabolite at each time point, comparing the second or third tertile with the first (lowest) tertile. In a secondary analysis, adjusted robust linear models were used to examine the associations of any self-reported intake of caffeinated beverages (compared with none) with methylation β values. Separate models were created for total caffeinated beverage intake, as well as coffee, tea, and soda beverage intakes. To account for multiple comparisons, we applied a false discovery rate (FDR) correction using the method by Benjamini-Hochberg (31).
Preconception models were adjusted for maternal age in years (continuous), household income of at least $40,000, ever-smoker status, and alcohol consumption obtained from the preconception diary. Pregnancy models were adjusted for maternal age, household income, and ever-smoker status. Alcohol consumption was not included due to few women reporting consuming alcohol (n <5) during pregnancy. Similarly, all women were provided with folic acid supplements as part of the trial, and over 92% of participants reported taking a multivitamin at baseline. In all models, we also adjusted for batch effects, cell count estimation, infant sex, and the infant's epigenetically derived ancestry. Ancestry was inferred using GLINT to generate 4 principle components of ancestry using information from select CpG sites (32). To adjust for batch effects from the DNAm measurement, plate number was included as a covariate in the analysis. The relative proportions of B cells, monocytes, CD4T, CD8T, granulocytes, NK cells, and nucleated red blood cells were estimated using a cord blood–specific reference (33, 34) to account for cellular heterogeneity. The treatment arm (low-dose aspirin or placebo) was not associated with either DNAm (24) or caffeine status (26) in this cohort, so it was not included as a covariate in this analysis. The statistical analysis was conducted in SAS 9.4 (SAS Institute) and R Studio 1.3.
Gene annotations were identified using the Illumina database and Ingenuity Knowledge Database and verified in the University of California Santa Cruz genome browser (GRCh37/hg19). Biologic networks and functional pathways were generated through the use of Ingenuity Pathway Analysis (IPA; QIAGEN Inc.) (35). CpG probes were imported and mapped to algorithmically generated networks and canonical pathways available in the proprietary IPA Knowledge Base. Resulting networks receive scores within IPA for ranking purposes, in which a greater number of molecules in the network results in a higher score. Canonical pathways were determined by IPA based on: 1) the ratio of the number of molecules from the data set that map to a canonical pathway, divided by the total number of molecules that map to that same pathway; and 2) a Fisher's exact test P value that determines whether the molecules in the data set and pathway overlap by chance alone.
Results
Women were, on average, 28.3 years of age (SD, 4.5 years) with household incomes of at least $40,000 (68.8%). The majority were never smokers (92.1%) and did not report consuming alcohol in their preconception diary (73.1%; Table 1). An older age, a higher household income, ever-smoking status, and higher preconception alcohol consumption were significantly associated with increasing caffeine metabolite tertiles. Maternal caffeine metabolite concentrations are reported in Supplemental Table 1. Except for preconception theobromine, the first tertile for each metabolite represents serum concentrations below the LOD. Low caffeine exposure was also indicated through maternal self-report. On the baseline questionnaire, 65%, 21%, and 7% reported any soda, coffee, or tea intake, respectively, with the majority consuming 1 or fewer servings/day on average. During the first 2 menstrual cycles of active follow-up, 75.8% reported consuming any amount of any type of caffeinated beverage. The number of caffeinated drinks consumed decreased over early pregnancy; by 8 weeks gestation, only about 23% reported consuming any amount. During the second and third trimesters, about half of all women reported any intake, and caffeinated soda was the primary source.
TABLE 1.
Baseline study population characteristics among women with newborn DNA methylation data
| Self-report caffeinated beverage, any intake | Preconception caffeine tertile | ||||
|---|---|---|---|---|---|
| Characteristic1 | Analytic sample | 1 | 2 | 3 | |
| ≤0.08 µg/mL | 0.08–0.56 µg/mL | >0.56 µg/mL | |||
| n | 378 | 250 | 125 | 119 | 119 |
| Age, years, mean (SD) | 28.3 (4.5) | 28.5 (4.6) | 27.5 (3.9) | 27.9 (4.3) | 29.8 (4.8) |
| Prepregnancy BMI, kg/m2, mean (SD) | 25.2 (5.5) | 25.5 (5.8) | 24.6 (5.1) | 25.2 (5.5) | 25.4 (5.2) |
| Education, >high school | 3387 (89.4) | 221 (88.4) | 115 (92.0) | 106 (89.1) | 105 (88.2) |
| Household income, ≥$40k | 260 (68.8) | 178 (71.2) | 74 (59.2) | 90 (75.6) | 88 (73.9) |
| Smoking status, never | 348 (92.1) | 227 (90.8) | 124 (99.2) | 109 (91.6) | 102 (85.0) |
| Preconception alcohol, any | 90 (27.3) | 84 (33.6) | 7 (6.4) | 28 (25.2) | 50 (52.1) |
All values are n (%) unless otherwise stated.
Self-reported caffeinated beverage intake had a positive correlation with serum caffeine and paraxanthine concentrations at both the preconception and 8-week-gestation assessments (P value for linear trend < 0.001) and are presented in Supplemental Table 2. Caffeinated beverage intake was not associated with preconception serum theobromine (P value for linear trend = 0.52), though a positive association was identified at 8 weeks of gestation (P value for linear trend = 0.003). Among preconception serum concentrations, caffeine was highly correlated with paraxanthine (rs = 0.86; P value < 0.001), and theobromine demonstrated low correlations with caffeine (rs = 0.34; P value < 0.001) and paraxanthine (rs = 0.30; <0.001). Further, serum concentrations at 8 weeks gestation showed low correlations with each other (rs, 0.35–0.47; P values < 0.001). In comparing preconception and 8-week-gestation serum concentrations, the rs ranged from 0.11 to 0.29, indicating a low correlation between the preconception and gestational-week-8 measures of serum metabolites (P values < 0.05).
Serum caffeine metabolites
In the array-wide analysis, preconception serum markers of caffeine, paraxanthine, and theobromine were not associated with individual CpG probes after FDR adjustment. Early pregnancy theobromine concentrations (tertile 2 compared with tertile 1) were associated with differential DNAm at probe cg09460369 (FDR P = 0.012) near the RAB2A gene on chromosome 8 (Tables 2 and 3). A similar trend was observed in comparing tertile 3 to tertile 1 for the same probe (β = 0.017; SE = 0.004; P value = 0.005), though statistical significance was not reached after the FDR correction. Caffeine and paraxanthine concentrations at 8 weeks gestation were not associated with individual CpG probes in neonates after FDR adjustment.
TABLE 2.
Top-rank CpG sets from the preconception exposure and cord blood DNAm analysis, n = 378
| CpG Site | β | SE | P Value | FDR P Value | Chromosome | Position | Gene | Relation to CpG Island |
|---|---|---|---|---|---|---|---|---|
| Preconception Serum | ||||||||
| Maternal serum caffeine (comparing tertile 2 vs. tertile 1) | ||||||||
| cg15058799 | −0.02601 | 0.00497 | 1.67E-07 | 0.138984 | 17 | 62,257,238 | TEX2 | |
| cg16968650 | −0.01889 | 0.003896 | 1.25E-06 | 0.520046 | 1 | 6,599,434 | NOL9 | |
| cg24229198 | −0.01445 | 0.003046 | 2.10E-06 | 0.520962 | 15 | 52,785,986 | MYO5A | |
| cg02189597 | 0.0138 | 0.00296 | 3.13E-06 | 0.520962 | 15 | 60,881,071 | RORA | North Shelf |
| cg05364412 | −0.00914 | 0.00196 | 3.10E-06 | 0.520962 | 19 | 2,136,934 | AP3D1 | Island |
| Maternal serum caffeine (comparing tertile 3 vs. tertile 1) | ||||||||
| cg14415456 | −0.01267 | 0.002554 | 6.97E-07 | 0.290335 | 12 | 5,371,496 | ||
| cg01591591 | 0.0187 | 0.003699 | 4.25E-07 | 0.290335 | 16 | 3,921,282 | CREBBP | |
| cg09611679 | −0.02215 | 0.004629 | 1.72E-06 | 0.476605 | 8 | 11,653,098 | FDFT1 | |
| cg15048078 | 0.02297 | 0.005539 | 3.37E-05 | 0.683558 | 4 | 38,665,599 | KLF3; FLJ13197 | Island |
| cg24636629 | −0.01773 | 0.004428 | 6.26E-05 | 0.683558 | 4 | 40,058,917 | LOC344967; N4BP2 | Island |
| Maternal serum paraxanthine (comparing tertile 2 vs. tertile 1) | ||||||||
| cg08346494 | −0.00532 | 0.001056 | 4.65E-07 | 0.19357 | 11 | 70,049,962 | FADD | South Shore |
| cg22362389 | 0.03115 | 0.00614 | 3.90E-07 | 0.19357 | 17 | 79,619,190 | PDE6G | South Shelf |
| cg06726154 | −0.01839 | 0.003751 | 9.43E-07 | 0.261841 | 8 | 62,669,129 | ||
| cg01554873 | −0.02448 | 0.005282 | 3.58E-06 | 0.597194 | 1 | 33,478,978 | AK2 | |
| cg21140380 | 0.03379 | 0.007379 | 4.66E-06 | 0.597194 | 11 | 43,938,616 | ALKBH3 | |
| Maternal serum paraxanthine (comparing tertile 3 vs. tertile 1) | ||||||||
| cg05029100 | 0.01184 | 0.00256 | 3.77E-06 | 0.522128 | 5 | 126,054,138 | ||
| cg27054066 | −0.01384 | 0.003033 | 5.02E-06 | 0.522128 | 8 | 70,905,247 | ||
| cg06307802 | 0.01237 | 0.002709 | 4.99E-06 | 0.522128 | 10 | 76,784,752 | KAT6B | |
| cg14415456 | −0.01152 | 0.002439 | 2.31E-06 | 0.522128 | 12 | 5,371,496 | ||
| cg01060862 | −0.02348 | 0.005008 | 2.75E-06 | 0.522128 | 12 | 99,235,137 | ANKS1B | |
| Maternal serum theobromine (comparing tertile 2 vs. tertile 1) | ||||||||
| cg25544461 | 0.009909 | 0.002102 | 2.44E-06 | 0.338786 | 7 | 73,894,573 | GTF2IRD1 | North Shore |
| cg27482594 | −0.0153 | 0.003218 | 1.98E-06 | 0.338786 | 10 | 98,861,170 | SLIT1-AS1; SLIT1 | |
| cg08943696 | −0.01026 | 0.002175 | 2.36E-06 | 0.338786 | 12 | 131,437,802 | GPR133 | |
| cg16361417 | −0.00968 | 0.002045 | 2.22E-06 | 0.338786 | 14 | 77,553,212 | ||
| cg26022581 | −0.01114 | 0.002283 | 1.06E-06 | 0.338786 | 14 | 105,832,816 | PACS2 | South Shore |
| Maternal serum theobromine (comparing tertile 3 vs. tertile 1) | ||||||||
| cg07320628 | −0.01531 | 0.003128 | 9.91E-07 | 0.825329 | 7 | 28,395,006 | CREB5 | |
| cg01193403 | −0.00237 | 0.004219 | 0.5736 | 0.999954 | 4 | 7,736,302 | SORCS2 | |
| cg09627796 | 0.003735 | 0.004211 | 0.3751 | 0.999954 | 4 | 7,736,472 | SORCS2 | |
| cg04457631 | −0.00213 | 0.005227 | 0.6833 | 0.999954 | 4 | 7,736,753 | SORCS2 | |
| cg14181141 | 0.001724 | 0.002884 | 0.55 | 0.999954 | 4 | 7,738,508 | SORCS2 | |
| Preconception self-report of caffeinated beverage intake | ||||||||
| Self-reported intake of any cups/d of caffeinated coffee | ||||||||
| cg24571508 | −0.01533 | 0.002902 | 1.27E-07 | 0.083732 | 4 | 81,094,387 | ||
| cg04537283 | −0.01474 | 0.002835 | 2.01E-07 | 0.083732 | 1 | 41,105,444 | RIMS3 | |
| cg17377891 | −0.01663 | 0.003407 | 1.06E-06 | 0.273555 | 6 | 11,851,894 | ||
| cg21093052 | 0.02912 | 0.006099 | 1.80E-06 | 0.273555 | 6 | 38,048,627 | ZFAND3 | |
| cg15057554 | 0.02294 | 0.004774 | 1.55E-06 | 0.273555 | 7 | 157,787,328 | PTPRN2 | |
| Self-reported intake of any cups/d of caffeinated tea | ||||||||
| cg106103091 | −0.1566 | 0.01065 | 5.84E-49 | 4.86E-43 | 13 | 100,083,372 | ||
| cg145327171 | −0.2554 | 0.02669 | 1.07E-21 | 4.45E-16 | 13 | 95,897,739 | ABCC4 | |
| cg05954339 | −0.08301 | 0.01525 | 5.21E-08 | 0.01447 | 7 | 128,292,014 | LINC01000 | |
| cg00073010 | −0.01684 | 0.003364 | 5.52E-07 | 0.091885 | 17 | 17,627,194 | RAI1 | Island |
| cg06301296 | −0.00723 | 0.001438 | 5.01E-07 | 0.091885 | 4 | 6,988,743 | TBC1D14 | Island |
| Self-reported intake of any servings/d of caffeinated soda | ||||||||
| cg24789467 | −0.01957 | 0.004037 | 1.24E-06 | 0.454676 | 5 | 132,159,003 | SHROOM1 | Island |
| cg25658464 | −0.01012 | 0.002142 | 2.28E-06 | 0.454676 | 5 | 150,861,640 | SLC36A1 | |
| cg05741395 | −0.01121 | 0.002377 | 2.41E-06 | 0.454676 | 15 | 42,173,869 | SPTBN5 | North Shore |
| cg25966829 | −0.01481 | 0.003085 | 1.59E-06 | 0.454676 | 1 | 192,613,150 | RGS13 | |
| cg19924120 | −0.01012 | 0.002158 | 2.73E-06 | 0.45476 | 18 | 76,160,108 | North Shore | |
| Self-reported preconception intake of any type of caffeinated beverage | ||||||||
| cg09002832 | −0.013 | 0.002 | 4.39E-08 | 0.036574 | 9 | 4,300,181 | GLIS3 | Island |
| cg14236539 | 0.013 | 0.003 | 6.26E-07 | 0.212765 | 6 | 47,026,946 | ||
| cg04289820 | 0.023 | 0.004 | 7.67E-07 | 0.212765 | 14 | 107,221,194 | Island | |
| cg23877720 | −0.006 | 0.001 | 2.44E-06 | 0.405877 | 7 | 49,815,117 | VWC2 | Island |
| cg22438063 | 0.010 | 0.002 | 2.38E-06 | 0.405877 | 16 | 75,121,671 | ZNRF1 | |
Abbreviations: CpG, cytosine-guanine; DNAm, DNA methylation; FDR, false discovery rate.
Potentially polymorphic.
TABLE 3.
Top-rank CpG sets from the pregnancy exposure and cord blood DNAm analysis, n = 378
| CpG Site | β | SE | P Value | FDR P Value | Chromosome | Position | Gene | Relation to CpG Island |
|---|---|---|---|---|---|---|---|---|
| Maternal serum at 8 weeks gestation | ||||||||
| Maternal serum caffeine (comparing tertile 2 vs. tertile 1) | ||||||||
| cg14229247 | −0.01007 | 0.002071 | 1.16E-06 | 0.608595 | 9 | 100,745,139 | ANP32B | Island |
| cg19140548 | 0.009416 | 0.002032 | 3.58E-06 | 0.608595 | 10 | 105,552,406 | SH3PXD2A | |
| cg03609636 | −0.00767 | 0.001625 | 2.37E-06 | 0.608595 | 11 | 102,323,712 | TMEM123 | Island |
| cg13000180 | 0.01182 | 0.002473 | 1.75E-06 | 0.608595 | 16 | 3,355,450 | ZNF75A | Island |
| cg13780046 | −0.02103 | 0.004542 | 3.65E-06 | 0.608595 | 21 | 43,756,472 | ||
| Maternal serum caffeine (comparing tertile 3 vs. tertile 1) | ||||||||
| cg08651856 | 0.02689 | 0.005955 | 6.30E-06 | 0.397483 | 4 | 103,780,165 | UBE2D3 | |
| cg24441866 | 0.03135 | 0.007311 | 1.80E-05 | 0.397483 | 4 | 175,075,603 | LOC101928509 | |
| cg17684034 | −0.01438 | 0.003296 | 1.28E-05 | 0.397483 | 5 | 74,862,658 | POLK | |
| cg05164570 | 0.01445 | 0.002936 | 8.60E-07 | 0.397483 | 5 | 140,071,033 | HARS | Island |
| cg01252586 | −0.00429 | 0.000952 | 6.58E-06 | 0.397483 | 5 | 175,815,479 | NOP16; HIGD2A | Island |
| Maternal serum paraxanthine (≥LOD vs. <LOD) | ||||||||
| cg03133256 | −0.01691 | 0.003512 | 1.48E-06 | 0.61718 | 22 | 24,802,889 | SPECC1L | |
| cg05112617 | 0.009442 | 0.001908 | 7.46E-07 | 0.61718 | 3 | 93,698,932 | ARL13B | Island |
| cg07677273 | 0.01606 | 0.003938 | 4.56E-05 | 0.621864 | 4 | 37,004,924 | ||
| cg08679638 | −0.01315 | 0.003147 | 2.93E-05 | 0.621864 | 4 | 88,883,651 | ||
| cg17596409 | −0.01198 | 0.002925 | 4.23E-05 | 0.621864 | 4 | 153,186,305 | ||
| Maternal serum theobromine at 8 weeks gestation (comparing tertile 2 vs. tertile 1) | ||||||||
| cg09460369 | 0.028 | 0.005 | 1.43E-08 | 0.011927 | 8 | 61,429,249 | RAB2A | Island |
| cg14651896 | −0.012 | 0.002 | 2.67E-07 | 0.111267 | 9 | 124,094,730 | GSN | |
| cg22906451 | −0.014 | 0.003 | 4.22E-07 | 0.117129 | 19 | 1,994,824 | BTBD2 | South Shore |
| cg08151407 | 0.017 | 0.003 | 6.78E-07 | 0.14123 | 3 | 183,171,057 | LINC00888 | |
| cg15326613 | −0.018 | 0.004 | 2.56E-06 | 0.222618 | 6 | 114,132,612 | ||
| Maternal serum theobromine at 8 weeks gestation (comparing tertile 3 vs. tertile 1) | ||||||||
| cg00282267 | 0.03608 | 0.007114 | 3.95E-07 | 0.217159 | 7 | 36,803,916 | ||
| cg20400017 | 0.01518 | 0.003026 | 5.21E-07 | 0.217159 | 9 | 120,535,563 | ||
| cg03818021 | −0.0116 | 0.002421 | 1.66E-06 | 0.27671 | 5 | 1,798,716 | MRPL36 | North Shore |
| cg27468830 | 0.01561 | 0.003248 | 1.53E-06 | 0.27671 | 7 | 157,227,072 | ||
| cg18209808 | −0.02923 | 0.006053 | 1.38E-06 | 0.27671 | 12 | 27,993,787 | ||
| Pregnancy self-report | ||||||||
| Any type of caffeinated beverage (8 weeks gestation) | ||||||||
| cg26341453 | −0.0101 | 0.002095 | 1.44E-06 | 0.599156 | 1 | 28,593,732 | ||
| cg03424508 | −0.05267 | 0.01078 | 1.02E-06 | 0.599156 | 17 | 86,925 | South Shore | |
| cg05164540 | −0.01735 | 0.003965 | 1.21E-05 | 0.621971 | 5 | 107,896,573 | ||
| cg15754587 | 0.01422 | 0.003302 | 1.66E-05 | 0.621971 | 6 | 25,170,368 | ||
| cg22149610 | −0.00779 | 0.001812 | 1.74E-05 | 0.621971 | 6 | 31,857,034 | Island | |
| Any type of caffeinated beverage (12–36 weeks gestation) | ||||||||
| cg19954884 | −0.00869 | 0.001737 | 5.69E-07 | 0.198177 | 4 | 129,775,763 | JADE1 | |
| cg07791897 | −0.01494 | 0.003027 | 7.99E-07 | 0.198177 | 5 | 55,897,584 | ||
| cg22141918 | 0.01674 | 0.003414 | 9.51E-07 | 0.198177 | 14 | 24,539,803 | CPNE6 | |
| cg24179591 | −0.01066 | 0.002149 | 6.96E-07 | 0.198177 | 16 | 66,967,636 | FAM96B | North Shore |
| cg20646219 | 0.01083 | 0.002318 | 2.97E-06 | 0.412843 | 17 | 38,490,228 | RARA | |
Abbreviations: CpG, cytosine-guanine; DNAm, DNA methylation; FDR, false discovery rate; LOD, limit of detection.
Maternal report of caffeinated beverages
Based on maternal self-report of caffeinated beverage intake, 1 significant inverse association with preconception consumption of any type of caffeinated beverage was found involving probe cg09002832 (FDR P = 0.036) near the GLIS3 gene on chromosome 9 (Tables 2 and 3). No FDR-corrected significant associations were found with specific types of preconception caffeinated beverages (i.e., coffee, tea or soda) or with self-reported intake during pregnancy (up to 8 weeks and 12–36 weeks). In an ad hoc analysis, we defined regular or consistent caffeine drinkers using the preconception and gestation week 8 diaries and pregnancy questionnaires (n = 54/250), but did not find FDR-significant CpG probes.
Functional enrichment analysis
Next, we imported the top-ranked 100 CpG probes, based on the FDR P value identified in the early pregnancy theobromine and preconception intake array-wide analyses, into IPA (Supplemental Tables 3 and 4). The resulting networks, along with the top related disease or functions, are provided in the Supplemental Materials (Supplemental Tables 5 and 6;Supplemental Figure 2). The top network showed that the maternal theobromine at 8 weeks of gestation was related to functions of “cell death and survival, lipid metabolism, small molecule biochemistry” (score = 44). The top IPA network of “cancer, gastrointestinal disease, organismal injury and abnormalities” was associated with the probes from preconception caffeinated beverage intake (score = 59). IPA also returned the overlapping canonical pathways of the top CpG probes identified in the preconception intake and pregnancy theobromine analyses based on the following categories: “intracellular and second messenger signaling”; “cellular growth, proliferation, and development”; “cellular immune response”; and “cellular stress and injury” (Table 4).
TABLE 4.
Overlapping canonical pathways for CpG sites identified in preconception (intake of any caffeinated beverage) and early pregnancy (serum theobromine), by category
| β-values | SE | P Value | CpG sites | Genes | |
|---|---|---|---|---|---|
| Category 1: intracellular and second messenger signaling | |||||
| Protein kinase A signaling | |||||
| Theobromine | −0.019 | 0.005 | 6.46E-05 | cg12534645 | CAMK2A |
| −0.012 | 0.003 | 9.26E-05 | cg14129040 | CREB1 | |
| −0.018 | 0.004 | 7.97E-06 | cg24319143 | DUSP22 | |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| Caffeinated beverage intake | 0.006 | 0.003 | 1.79E-05 | cg18871670 | ANAPC13 |
| −0.006 | 0.002 | 0.000158 | cg04459013 | EPM2A | |
| 0.006 | 0.002 | 0.000217 | cg10290200 | FLNC | |
| 0.011 | 0.003 | 0.000277 | cg11411922 | MYLK2 | |
| −0.006 | 0.001 | 6.27E-06 | cg13552999 | PTPN21 | |
| cAMP-mediated signaling | |||||
| Theobromine | −0.019 | 0.005 | 6.46E-05 | cg12534645 | CAMK2A |
| 0.008 | 0.002 | 2.21E-05 | cg22426570 | CHRM2 | |
| −0.012 | 0.003 | 9.26E-05 | cg14129040 | CREB1 | |
| Caffeinated beverage intake | 0.016 | 0.004 | 9.79E-05 | cg09028487 | MC1R |
| −0.009 | 0.002 | 9.92E-05 | cg19653161 | RGS10 | |
| G-protein coupled receptor signaling | |||||
| Theobromine | −0.019 | 0.005 | 6.46E-05 | cg12534645 | CAMK2A |
| 0.008 | 0.002 | 2.21E-05 | cg22426570 | CHRM2 | |
| −0.012 | 0.003 | 9.26E-05 | cg14129040 | CREB1 | |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| Caffeinated beverage intake | 0.016 | 0.004 | 9.79E-05 | cg09028487 | MC1R |
| −0.009 | 0.002 | 9.92E-05 | cg19653161 | RGS10 | |
| AMP-activated protein kinase signaling1 | |||||
| Theobromine | 0.008 | 0.002 | 2.21E-05 | cg22426570 | CHRM2 |
| −0.012 | 0.003 | 9.26E-05 | cg14129040 | CREB1 | |
| −0.004 | 0.001 | 9.26E-05 | cg17287814 | PPP2R2A | |
| 0.028 | 0.005 | 1.43E-08 | cg09460369 | RAB2A | |
| Caffeinated beverage intake | 0.004 | 0.001 | 0.000236 | cg06875318 | ORAI1 |
| 0.021 | 0.005 | 2.68E-05 | cg13345558 | PRKAA1 | |
| Category 2: cellular growth, proliferation, and development | |||||
| Kinetochore metaphase signaling | |||||
| Theobromine | 0.004 | 0.001 | 0.000033 | cg27329848 | BUB1B |
| Caffeinated beverage intake | 0.006 | 0.001 | 1.79E-05 | cg18871670 | ANAPC13 |
| 0.01 | 0.002 | 3.81E-05 | cg08133816 | CENPT | |
| −0.012 | 0.003 | 0.000294 | cg21751147 | MAD1L1 | |
| ILK signaling | |||||
| Theobromine | −0.012 | 0.003 | 9.26E-05 | cg14129040 | CREB1 |
| −0.007 | 0.002 | 0.00007 | cg18778433 | PARVA | |
| −0.004 | 0.001 | 9.26E-05 | cg17287814 | PPP2R2A | |
| Caffeinated beverage intake | −0.024 | 0.006 | 0.000136 | cg11689813 | ACTA1 |
| 0.006 | 0.002 | 0.000217 | cg10290200 | FLNC | |
| Category 3: Cellular immune response | |||||
| PI3K signaling in B lymphocytes | |||||
| Theobromine | −0.019 | 0.005 | 6.46E-05 | cg12534645 | CAMK2A |
| −0.012 | 0.005 | 9.26E-05 | cg14129040 | CREB1 | |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| Caffeinated beverage intake | −0.025 | 0.007 | 0.000271 | cg02211519 | CD81 |
| 0.011 | 0.003 | 0.000205 | cg00150231 | INPP5D | |
| iCOS-iCOSL signaling in T helper cells | |||||
| Theobromine | −0.019 | 0.005 | 6.46E-05 | cg12534645 | CAMK2A |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| Caffeinated beverage intake | 0.011 | 0.003 | 2.05E-04 | cg00150231 | INPP5D |
| Category 4: cellular stress and injury | |||||
| ATM signaling2 | |||||
| Theobromine | −0.012 | 0.005 | 9.26E-05 | cg14129040 | CREB1 |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| −0.004 | 0.001 | 9.26E-05 | cg17287814 | PPP2R2A | |
| Caffeinated beverage intake | −0.007 | 0.002 | 2.03E-04 | cg24532898 | TLK1 |
| Hypoxia signaling in the cardiovascular system3 | |||||
| Theobromine | −0.012 | 0.005 | 9.26E-05 | cg14129040 | CREB1 |
| −0.013 | 0.003 | 0.000102 | cg08315283 | NFKBIA | |
| Caffeinated beverage intake | 0.022 | 0.006 | 3.27E-04 | cg17177602 | UBE2J2 |
AMP-activated protein kinase signaling overlaps with the “cellular growth, proliferation, and development” category.
ATM signaling is also part of the “cell cycle regulation” category.
Hypoxia signaling in the cardiovascular system is also part of the “cancer,” “cardiovascular signaling,” and “ingenuity toxicity list pathways” categories.
Abbreviations: CpG, cytosine-guanine; ILK, integrin-linked kinase; PI3K, phosphoinositide 3-kinase; iCOS, inducible costimulator; iCOSL, inducible costimulator ligand; ATM, ataxia-telangiectasia mutated
Discussion
We investigated array-wide methylation profiles in neonatal cord blood in association with maternal caffeine exposure during preconception and early pregnancy. Exposure was examined using 2 approaches: serum markers of caffeine metabolites and maternal report of caffeinated beverage intake. Overall, we found few differences in methylation at individual CpG sites with periconception caffeine exposure. Differential methylation at CpG probe cg09460369 (RAB2A) was associated with serum theobromine at 8 weeks of gestation. RAB2A encodes a protein required for transport from the endoplasmic reticulum to the Golgi complex and has been implicated in conditions like rheumatoid arthritis (36) and osteoarthritis (37). In addition, cg09002832 (GLIS3) was associated with preconception caffeinated beverage consumption. The GLIS3 gene encodes a protein important in transcription and is involved in the early development of tissues, including pancreatic beta cells and the thyroid, brain, liver, and kidney (38). Both cg09460369 and cg09002832 are located in CpG islands, suggesting a role in the regulation of gene expression, though this needs to be confirmed with gene transcription data (39). Though replication is needed, our study provides novel but limited evidence of associations between maternal caffeine exposure and DNAm alterations.
Protein kinase A, AMP-activated protein kinase, and cAMP-mediated signaling were elucidated as potential functions impacted by the methylation patterns in our study. Protein kinase A is essential to regulating metabolic homeostasis and gene expression through its role in phosphorylation of enzymes and transcription factors and is regulated by AMP. For example, it is the primary regulator of acetyl-CoA carboxylase in fatty acid metabolism (40) and its activation by drugs such as metformin decreases blood glucose concentrations (41). Our finding is consistent with an EWAS of habitual coffee consumption among older adults in which genes linked to significant CpG probes (P values < 10–6) were related to protein kinase activity in an enrichment analysis (16). Animal models have also provided evidence linking in utero caffeine exposure to gene expression of protein kinase A and AMP (10). This is potentially relevant, as a recent meta-analysis reported a 39% increased risk of childhood overweight and obesity among offspring of mothers with the highest caffeine intake compared to those with the lowest (7). This association was further supported by a significant dose-response relationship in which each 100 mg per day increase of maternal caffeine consumption was associated with a 31% increased risk of childhood overweight and obesity (7). Therefore, DNAm may be one mechanism linking preconception or prenatal caffeine exposure to later disease risks, though the results of this study alone should not be interpreted as a causal mechanism of the developmental effects of maternal caffeine exposure.
Pathways linked to cell cycle function, growth, and development were also prominent in our study. An in vitro experiment demonstrated that 24-hour treatment with theobromine (100–200 µg/mL) stopped preadipocytes in the nondividing, G0/G1 phase (42). Notably, these concentrations of theobromine are higher than the levels observed in this cohort (<1.05 µg/mL) or those that are attainable through typical ingestion of theobromine-rich foods, such as dark chocolate. We were unable to assess this further, as dietary recalls to measure self-reported intake of theobromine-rich foods were not collected in this study. Further research is needed to expand on the potential impacts of caffeine metabolites, particularly theobromine, on cellular function preconceptionally and prenatally. Notably, the pathways identified in IPA are only potential biologic processes that may be impacted by the methylation of CpG probes in that pathway. We cannot rule out that the identification of these potential pathways in our study may be due to the differing number of CpG probes per gene that are present on the MethylationEPIC 850k.
Strengths of this study include the examination of neonate DNAm in the context of preconception and prenatal exposure to caffeine in a well-established, prospective cohort. We uniquely utilized 2 sources of maternal exposure: serum metabolites and self-reported intake. Though exposure misclassification is possible, the prospective design limits potential recall bias of self-reported intake, and findings were similar between self-reported and serum measures. However, this study had some limitations. First, a single preconception marker or a single early gestation marker of circulating caffeine metabolites may not reflect usual intake during these critical periods [caffeine metabolites have half-lives of around 3 hours (43)]. Repeated measures of these metabolites and adjustment of time since caffeinated beverage consumption should be considered in future studies. Second, participants in our study population had relatively low reported consumption of coffee and tea intake, which was also reflected in serum caffeine concentrations, and this may limit generalizability. The median concentration in the first tertile of preconception caffeine intake is consistent with a clinical trial of adults (a serum caffeine value of 0.03 µg/mL corresponded to no cups of coffee) (44), and our median value of preconception caffeine was less than what was observed in another low-consumption cohort of premenopausal women (45). Further, our highest preconception tertiles (but not 8-week-gestation tertiles) of caffeine and paraxanthine were similar to the highest quartile of plasma caffeine (>0.66 µg/mL) and paraxanthine (>0.23 µg/mL) in a cohort of pregnant women at gestational weeks 8–13 (46). Additionally, the prevalence of preconception coffee consumption was about 20% in this study. In contrast, the prevalence of coffee consumption among women over 20 years of age was 60.3% based on an examination of NHANES 2011–2016 data (22). Tea consumption was also low in our study population (7%), limiting our ability to examine its intake directly with DNAm. In an EWAS combining 4 European cohorts of older adults, Ek et al. (17) found that tea consumption in women, but not men, was associated with DNAm and mapped to genes related to estradiol. We also did not systematically capture nonbeverage sources of caffeine by self-report in questionnaires/diaries; however, the serum caffeine metabolites would reflect recent exposure regardless of source (e.g., foods, medications). Further, it is also possible that this cohort of women attempting pregnancy may have exhibited different behaviors (i.e., intentionally reducing caffeine consumption) compared to women with unplanned pregnancies (47, 48). Therefore, research among more diverse women of reproductive age with higher levels of caffeine exposure from coffee and tea sources is warranted. Third, the generalizability of results may be further limited, as our study population consisted of White women with a history of 1–2 pregnancy losses who had a live birth delivery during follow-up. Lastly, gene expression data were not available in this study, so we were unable to determine whether differences in methylation correlate with cellular activity. Similarly, genotype data were not available to account for the role of underlying genetics on DNAm, particularly genotypes that might be relevant to caffeine metabolism (i.e., CYP1A2). As this discovery study is limited in sample size, the findings should be validated in additional independent studies. Further research in this area would benefit from the integration of genomic and transcriptomic data.
In summary, few differences in cord blood DNAm at individual CpG sites were identified in association with maternal caffeine intake. Future work should examine these associations among women attempting pregnancy, who have a greater variability in caffeinated beverage consumption both within and outside of the recommended ranges. DNAm changes in neonatal cord blood from preconception or early pregnancy theobromine exposure may be linked to energy metabolism, gene expression, and cell cycle function. This suggests that epigenetic mechanisms may underlie the previous associations between maternal caffeine exposure and adverse metabolic outcomes in childhood, including obesity and liver fat deposits. However, additional studies are needed to explore potential underlying mechanisms among women attempting pregnancy with higher caffeine exposure.
Supplementary Material
Acknowledgments
This work utilized the computational resources of the NIH High Performance Computing Biowulf cluster (http://hpc.nih.gov).
The authors’ responsibilities were as follows—KJP, EFS, and EHY: designed and conducted the research; WG: provided essential materials; KJP, AP-S, and SLR: analyzed data or performed the statistical analysis; KJP: wrote the manuscript; KJP and EHY: have primary responsibility for the final content; SKZ, KCS, RMS, and SLM: provided essential feedback in writing the manuscript; and all authors: read and approved the final manuscript.
Author disclosures: The authors report no conflicts of interest.
Notes
Supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (contract numbers HHSN267200603423, HHSN267200603424, HHSN267200603426, and HHSN275201300023I-HHSN2750008).
Supplemental Figures 1–2 and Supplemental Tables 1–6 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.
Abbreviations used: CpG, cytosine-guanine; CYP1A2, cytochrome P450 1A2 enzyme; DNAm, DNA methylation; EAGeR, Effects of Aspirin in Gestation and Reproduction; EWAS, epigenome-wide association study; FDR, false discovery rate; IPA, Ingenuity Pathway Analysis; LOD, limit of detection.
Contributor Information
Kristen J Polinski, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Alexandra Purdue-Smithe, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Sonia L Robinson, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Sifang Kathy Zhao, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Karen C Schliep, Department of Family and Preventive Medicine, University of Utah, Salt Lake City, UT, USA.
Robert M Silver, Department of Family and Preventive Medicine, University of Utah, Salt Lake City, UT, USA.
Weihua Guan, University of Minnesota School of Public Health, Minneapolis, MN, USA.
Enrique F Schisterman, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Sunni L Mumford, Department of Family and Preventive Medicine, University of Utah, Salt Lake City, UT, USA.
Edwina H Yeung, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Data Availability
Data described in the manuscript, code book and analytic code will be made available upon request pending application and approval.
References
- 1. Frary CD, Johnson RK, Wang MQ. Food sources and intakes of caffeine in the diets of persons in the United States. J Am Diet Assoc. 2005;105(1):110–3. [DOI] [PubMed] [Google Scholar]
- 2. ACOG Committee opinion no. 462: moderate caffeine consumption during pregnancy. Obstet Gynecol. 2010;116(2 Pt 1):467–8. [DOI] [PubMed] [Google Scholar]
- 3. Syme MR, Paxton JW, Keelan JA. Drug transfer and metabolism by the human placenta. Clin Pharmacokinet. 2004;43(8):487–514. [DOI] [PubMed] [Google Scholar]
- 4. Lelo A, Miners JO, Robson RA, Birkett DJ. Quantitative assessment of caffeine partial clearances in man. Br J Clin Pharmacol. 1986;22(2):183–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Voerman E, Jaddoe VW, Hulst ME, Oei EH, Gaillard R. Associations of maternal caffeine intake during pregnancy with abdominal and liver fat deposition in childhood. Pediatr Obes. 2020;15(5):e12607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Greenwood DC, Thatcher NJ, Ye J, Garrard L, Keogh G, King LG, Cade JE. Caffeine intake during pregnancy and adverse birth outcomes: a systematic review and dose-response meta-analysis. Eur J Epidemiol. 2014;29(10):725–34. [DOI] [PubMed] [Google Scholar]
- 7. Jin F, Qiao C. Association of maternal caffeine intake during pregnancy with low birth weight, childhood overweight, and obesity: a meta-analysis of cohort studies. Int J Obes (Lond). 2021;45(2):279–87. [DOI] [PubMed] [Google Scholar]
- 8. Li DK, Ferber JR, Odouli R. Maternal caffeine intake during pregnancy and risk of obesity in offspring: a prospective cohort study. Int J Obes. 2015;39(4):658–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Papadopoulou E, Botton J, Brantsæter AL, Haugen M, Alexander J, Meltzer HM, Bacelis J, Elfvin A, Jacobsson B, Sengpiel V. Maternal caffeine intake during pregnancy and childhood growth and overweight: results from a large Norwegian prospective observational cohort study. BMJ Open. 2018;8(3):e018895. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Hu S, Liu K, Luo H, Xu D, Chen L, Zhang L, Wang H. Caffeine programs hepatic SIRT1-related cholesterol synthesis and hypercholesterolemia via A2AR/cAMP/PKA pathway in adult male offspring rats. Toxicology. 2019;418:11–21. [DOI] [PubMed] [Google Scholar]
- 11. Fang X, Robinson J, Wang-Hu J, Jiang L, Freeman DA, Rivkees SA, Wendler CC. cAMP induces hypertrophy and alters DNA methylation in HL-1 cardiomyocytes. Am J Physiol Cell Physiol. 2015;309(6):C425–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Xu D, Zhang B, Liang G, Ping J, Kou H, Li X, Xiong J, Hu D, Chen L, Magdalou J et al. Caffeine-induced activated glucocorticoid metabolism in the hippocampus causes hypothalamic-pituitary-adrenal axis inhibition in fetal rats. PLoS One. 2012;7(9):e44497–e. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Sharmin S, Guan H, Williams AS, Yang K. Caffeine reduces 11β-hydroxysteroid dehydrogenase type 2 expression in human trophoblast cells through the adenosine A(2B) receptor. PLoS One. 2012;7(6):e38082. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Buscariollo DL, Fang X, Greenwood V, Xue H, Rivkees SA, Wendler CC. Embryonic caffeine exposure acts via A1 adenosine receptors to alter adult cardiac function and DNA methylation in mice. PLoS One. 2014;9(1):e87547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Fang X, Mei W, Barbazuk WB, Rivkees SA, Wendler CC. Caffeine exposure alters cardiac gene expression in embryonic cardiomyocytes. Am J Physiol Regul Integr Comp Physiol. 2014;307(12):R1471–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Chuang YH, Quach A, Absher D, Assimes T, Horvath S, Ritz B. Coffee consumption is associated with DNA methylation levels of human blood. Eur J Hum Genet. 2017;25(5):608–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Ek WE, Tobi EW, Ahsan M, Lampa E, Ponzi E, Kyrtopoulos SA, Georgiadis P, Lumey LH, Heijmans BT, Botsivali M et al. Tea and coffee consumption in relation to DNA methylation in four European cohorts. Hum Mol Genet. 2017;26(16):3221–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Karabegovic I, Portilla-Fernandez E, Li Y, Ma JT, Maas SCE, Sun DK, Hu EA, Kuhnel B, Zhang Y, Ambatipudi S et al. Epigenome-wide association meta-analysis of DNA methylation with coffee and tea consumption. Nat Commun. 2021;12(1):2830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Smith ZD, Chan MM, Mikkelsen TS, Gu H, Gnirke A, Regev A, Meissner A. A unique regulatory phase of DNA methylation in the early mammalian embryo. Nature. 2012;484(7394):339–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Schisterman EF, Silver RM, Lesher LL, Faraggi D, Wactawski-Wende J, Townsend JM, Lynch AM, Perkins NJ, Mumford SL, Galai N. Preconception low-dose aspirin and pregnancy outcomes: results from the EAGeR randomised trial. Lancet North Am Ed. 2014;384(9937):29–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Schisterman EF, Silver RM, Perkins NJ, Mumford SL, Whitcomb BW, Stanford JB, Lesher LL, Faraggi D, Wactawski-Wende J, Browne RW et al. A randomised trial to evaluate the effects of low-dose aspirin in gestation and reproduction: design and baseline characteristics. Paediatr Perinat Epidemiol. 2013;27(6):598–609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Rehm CD, Ratliff JC, Riedt CS, Drewnowski A. Coffee consumption among adults in the United States by demographic variables and purchase location: analyses of NHANES 2011–2016 data. Nutrients. 2020;12(8):2463. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Lieberman HR, Agarwal S, Fulgoni VL 3rd. Daily patterns of caffeine intake and the association of intake with multiple sociodemographic and lifestyle factors in US adults based on the NHANES 2007–2012 surveys. J Acad Nutr Diet. 2019;119(1):106–14. [DOI] [PubMed] [Google Scholar]
- 24. Yeung EH, Guan W, Mumford SL, Silver RM, Zhang C, Tsai MY, Schisterman EF. Measured maternal prepregnancy anthropometry and newborn DNA methylation. Epigenomics. 2019;11(2):187–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Yeung EH, Guan W, Zeng X, Salas LA, Mumford SL, de Prado Bert P, van Meel ER, Malmberg A, Sunyer J, Duijts L et al. Cord blood DNA methylation reflects cord blood C-reactive protein levels but not maternal levels: a longitudinal study and meta-analysis. Clin Epigenetics. 2020;12(1):60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Purdue-Smithe A, Kim K, Schisterman E, Schliep K, Perkins N, Sjaarda L, Mills J, Silver R, Andriessen V, Alkhalaf Z et al. Caffeinated beverage intake and serum caffeine metabolites and risk of pregnancy loss (OR17-04-19). Curr Dev Nutr. 2019;3(Suppl 1):nzz039.OR17-04-19. [Google Scholar]
- 27. Hornung RW, Reed LD. Estimation of average concentration in the presence of nondetectable values. Appl Occup Environ Hyg. 1990;5(1):46–51. [Google Scholar]
- 28. Michels KB, Binder AM, Dedeurwaerder S, Epstein CB, Greally JM, Gut I, Houseman EA, Izzi B, Kelsey KT, Meissner A et al. Recommendations for the design and analysis of epigenome-wide association studies. Nat Methods. 2013;10(10):949–55. [DOI] [PubMed] [Google Scholar]
- 29. Moran S, Arribas C, Esteller M. Validation of a DNA methylation microarray for 850,000 CpG sites of the human genome enriched in enhancer sequences. Epigenomics. 2016;8(3):389–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry RA. Minfi: a flexible and comprehensive bioconductor package for the analysis of infinium DNA methylation microarrays. Bioinformatics. 2014;30(10):1363–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol. 1995;57(1):289–300. [Google Scholar]
- 32. Rahmani E, Yedidim R, Shenhav L, Schweiger R, Weissbrod O, Zaitlen N, Halperin E. GLINT: a user-friendly toolset for the analysis of high-throughput DNA-methylation array data. Bioinformatics. 2017;33(12):1870–2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Gervin K, Salas LA, Bakulski KM, van Zelm MC, Koestler DC, Wiencke JK, Duijts L, Moll HA, Kelsey KT, Kobor MS et al. Systematic evaluation and validation of reference and library selection methods for deconvolution of cord blood DNA methylation data. Clin Epigenetics. 2019;11(1):125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Salas LA, Gervin K, Jones MC. FlowSorted.CordBloodCombined.450k: Illumina 450k/EPIC data on FACS and MACS umbilical blood cells. Version 1.5.0 ed. R package. New York, USA: Bioconductor; 2020. [Google Scholar]
- 35. Krämer A, Green J, Pollard J Jr., Tugendreich S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics. 2014;30(4):523–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Sabir JSM, El Omri A, Banaganapalli B, Al-Shaeri MA, Alkenani NA, Sabir MJ, Hajrah NH, Zrelli H, Ciesla L, Nasser KK et al. Dissecting the role of NF-κb protein family and its regulators in rheumatoid arthritis using weighted gene co-expression network. Front Genet. 2019;10:1163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Wang X, Yu Y, Huang Y, Zhu M, Chen R, Liao Z, Yang S. Identification of potential diagnostic gene biomarkers in patients with osteoarthritis. Sci Rep. 2020;10(1):13591. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Kim Y-S, Nakanishi G, Lewandoski M, Jetten AM. GLIS3, a novel member of the GLIS subfamily of Krüppel-like zinc finger proteins with repressor and activation functions. Nucleic Acids Res. 2003;31(19):5513–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Moore LD, Le T, Fan G. DNA methylation and its basic function. Neuropsychopharmacology. 2013;38(1):23–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Zhang BB, Zhou G, Li C. AMPK: An emerging drug target for diabetes and the metabolic syndrome. Cell Metab. 2009;9(5):407–16. [DOI] [PubMed] [Google Scholar]
- 41. Whitehead JP, Richards AA, Hickman IJ, Macdonald GA, Prins JB. Adiponectin–A key adipokine in the metabolic syndrome. Diabetes Obes Metab. 2006;8(3):264–80. [DOI] [PubMed] [Google Scholar]
- 42. Fuggetta MP, Zonfrillo M, Villivà C, Bonmassar E, Ravagnan G. Inflammatory microenvironment and adipogenic differentiation in obesity: the inhibitory effect of theobromine in a model of human obesity in vitro. Mediators Inflamm. 2019;2019:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Callahan MM, Robertson RS, Arnaud MJ, Branfman AR, McComish MF, Yesair DW. Human metabolism of [1-methyl-14C]- and [2-14C]caffeine after oral administration. Drug Metab Dispos. 1982;10(4):417–23. [PubMed] [Google Scholar]
- 44. Kempf K, Herder C, Erlund I, Kolb H, Martin S, Carstensen M, Koenig W, Sundvall J, Bidel S, Kuha S et al. Effects of coffee consumption on subclinical inflammation and other risk factors for type 2 diabetes: a clinical trial. Am J Clin Nutr. 2010;91(4):950–7. [DOI] [PubMed] [Google Scholar]
- 45. Schliep KC, Schisterman EF, Wactawski-Wende J, Perkins NJ, Radin RG, Zarek SM, Mitchell EM, Sjaarda LA, Mumford SL. Serum caffeine and paraxanthine concentrations and menstrual cycle function: correlations with beverage intakes and associations with race, reproductive hormones, and anovulation in the Biocycle Study. Am J Clin Nutr. 2016;104(1):155–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Gleason JL, Tekola-Ayele F, Sundaram R, Hinkle SN, Vafai Y, Buck Louis GM, Gerlanc N, Amyx M, Bever AM, Smarr MM et al. Association between maternal caffeine consumption and metabolism and neonatal anthropometry: A secondary analysis of the NICHD fetal growth studies-singletons. JAMA Netw Open. 2021;4(3):e213238. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Lum KJ, Sundaram R, Buck Louis GM. Women's lifestyle behaviors while trying to become pregnant: evidence supporting preconception guidance. Am J Obstet Gynecol. 2011;205(3):203.e1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Wise LA, Wesselink AK, Hatch EE, Weuve J, Murray EJ, Wang TR, Mikkelsen EM, Toft Sørensen H, Rothman KJ. Changes in behavior with increasing pregnancy attempt time: a prospective cohort study. Epidemiology. 2020;31(5):659–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data described in the manuscript, code book and analytic code will be made available upon request pending application and approval.