Maternal PRKAA1 and EDNRA genotypes are associated with birth weight, and PRKAA1 with uterine artery diameter and metabolic homeostasis at high altitude

Study participants.

Recruited for longitudinal studies during pregnancy were 178 healthy Europeans and Andeans residing at two locations in Bolivia, the high-altitude city of La Paz or the neighboring city of El Alto, Bolivia (3,600 and 4,100 m, respectively) and the low-altitude city of Santa Cruz, Bolivia (400 m). Augmenting this longitudinal sample was a cross-sectional cohort of 120 women living in La Paz or El Alto, all but one of whom was Andean. Women were classified as Andean or European by self-identified ancestry as verified from a panel of ancestry-informative gene markers (AIMs) and parental and grandparental surnames as previously described (25). Participant exclusion criteria were maternal smoking, chronic hypertenslon, gestational or other forms of diabetes, absence of birth weight data, or failure for their DNA samples to pass quality control (QC) as described below, leaving a total of 245 participants whose data are reported here (Fig. 1). All high-altitude studies were performed at the Instituto Boliviano de Biología de Altura (Bolivian High Biology Institute) or the Clinica del Sur (Southern Clinic) in La Paz, and all low-altitude studies were conducted at the Clinica Sirianí (Siriani Clinic) in Santa Cruz. All study participants provided informed written consent to study procedures that had been approved by the Colorado Multiple Institutional, the Colegio Médico in Bolivia, the Pennsylvania State University, and the University of Michigan Institutional Review Boards.

Phenotype collection.

Each participant provided biographical information (age, altitude of birth, self-identified ancestry, parental and grandparental surnames and ethnicity, body weight before pregnancy, number of previous pregnancies and live births) on a questionnaire that was administered in her native language. Phenotypic studies were conducted in the longitudinal cohorts at both altitudes at pregnancy weeks 20 and 36 and 3–4 mo postpartum for a measurement in the nonpregnant state (25). The cross-sectional cohort was studied once between 17 and 43 wk of pregnancy with an average of 31.1 wk (± 5.7 SE) (10). For both cohorts, height was measured by stadiometer, body weight by balance scale in lightly clothed participants, subscapular and tricep skinfolds by Lange caliper (Beta Technology, Santa Cruz, CA) and summed from the right and left sides for an index of body fat during pregnancy, and blood pressure by arm cuff sphygmomanometer.

Infant sex, birth weight, length, and gestational age data were collected by hospital personnel at the time of delivery or by study personnel for infants born at home. Gestational age was based on the date of the last menstrual period and confirmed by fetal biometry at week 20 or by clinical assessment at delivery.

In the longitudinal cohorts, UA diameter was measured with an ATL3000 ultrasound unit with color imaging and Doppler at pregnancy weeks 20 and 36. Measurements were obtained bilaterally in the longitudinal view via a 4 Mhz curved linear array probe at the point where the UA appears to cross over the external iliac artery. Mean vessel diameter was expressed as (2 × end-diastolic + peak systolic)/3 and averaged from the right and left sides as previously described (25).

Selection of candidate genes and SNPs.

In previous work, we employed population genetic techniques to identify gene regions exhibiting strong signatures of natural selection among Andeans and Tibetans from dense genome scan data (7, 8). In brief, first we scanned across each chromosome to identify gene regions with previously unknown function with respect to altitude adaptation showing evidence of positive selection. We identified extended regions of statistical significance for three test statistics commonly used in population genetic research (locus-specific branch length, lnRH, and Tajima's D), using the hypergeometric distribution calculated for one megabase nonoverlapping windows along each chromosome. The results of this analysis revealed 14 candidate regions for high-altitude adaptation in Tibetans and 37 regions in Andeans. Second, we identified signatures of positive selection in genes with known physiological roles in oxygen homeostasis, including genes in the hypoxia inducible transcription factor (HIF) pathway and genes in or related to the globin family. We considered genes where two or more test statistics were statistically significant in Andeans or Tibetans (see Supplemental Tables S1 and S2 for additional detail). From these two analyses, we selected the 16 candidate genes listed in Table 1 to be assayed in our cohort based on rank, function, and assay design success. The chosen regions included: 1) 14 genes or gene regions with known physiological roles in oxygen homeostasis, including genes in the HIF pathway and the β-globin gene cluster located on chromosome 11 and 2) two genomic regions with unknown function with regard to high-altitude phenotypes that included portions of chromosome 12 [108,830 kilobase (kb) pairs–113,266 kb] and chromosome 15 (40,327–42,697 kb). Hereafter, we refer to these regions by their respective chromosome number (CHR12 or CHR15).

For each of the 16 candidate genes, we selected haplotype-tagging (tag) SNPs for Andean (i.e., Quechua and Aymara) haplotypes. To identify tagSNPs in this population, genotype data spanning 500 kb upstream and downstream of the start and end coordinates of each gene were phased with fastPHASE (v1.1) (46), and the resulting haplotypes were inferred with Haploview v4.1 (4). In this way, we included SNPs located within each candidate gene as well as SNPs upstream and downstream as the latter have been shown to regulate adjacent genes (57). All SNPs upstream or downstream of the transcription initiation start site or the termination codon are referred to by the HIF-pathway candidate gene for which they were selected. See Supplemental Table S2 for a listing of each SNP selected.

SNP genotyping.

Peripheral blood was drawn from an antecubital vein into a collection tube containing EDTA and transported on dry ice to our University Park lab. DNA was isolated by the Puregene DNA purification system (Qiagen, Valencia, CA) according to the manufacturer's instructions. Genotyping of 78 AIMs that distinguish West African, European, and Native American parental populations was performed at Prevention Genetics (Marshfield, WI) as previously described (9, 11, 48) and used for calculating individual ancestry estimates using a dihybrid model implemented in the program Maximum Likelihood (17).

Genotyping of candidate SNPs was performed by using the ABI SNPlex genotyping technology (Applied Biosystems, Foster City, CA) and Genemapper v3.5 (Applied Biosystems) software for genotype calling. In total, 82 SNPs produced reliable genotype calls and were subject to QC. SNPs were filtered from our dataset if they exhibited ≥20% missing data in our combined longitudinal and cross-sectional cohort (n = 15 SNPs) or if they violated Hardy-Weinberg equilibrium (P < 0.001, n = 4 SNPs), as such violations were likely to indicate genotyping error. Therefore quantitative association analyses were performed on 63 SNPs distributed across the 16 candidate gene regions (Table 1). We also excluded 53 individuals [30 Andeans (8 low, 22 high altitude) and 23 Europeans (5 low, 18 high altitude)] from our analysis as they exhibited ≥20% missing SNP data or lacked birth weight data. The mean birth weight of infants born to excluded vs. included women did not differ within any altitude and ancestry group.

Association analysis.

We combined European and Andean groups and the longitudinal and cross-sectional cohorts to increase our power to detect associations between selection-nominated candidate genes and birth weight in a cross-population manner. SNP analysis and genetic-model assessment were conducted with PLINK version 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/). We performed standard linear regression on birth weight to calculate regression coefficients and 95% confidence intervals using an additive model of inheritance. SNPs showing significant uncorrected (P < 0.05) birth-weight associations in an additive model next were run with dominant and recessive models of inheritance. Models were adjusted for Native American ancestry proportion (NAAP) as estimated by the AIMs to control for population substructure, altitude of birth (either low or high), gestational age, infant sex, and maternal height as these variables were shown to influence birth weight in our full cohort (63). We applied a Bonferroni correction for multiple testing with a P value cutoff of P < 0.0008 (α = 0.05, 63 tests).

Gene expression analysis.

Forty-six (46) high-altitude Andean women who delivered at term from the cross-sectional cohort were included for gene expression analyses, 27 of whom delivered non-IUGR infants, and 19 delivered IUGR infants. Infants were classified as IUGR if their birth weight was <10th percentile of sea-level values for their gestational age and sex, and non-IUGR if the birth weight was between the 10th and 90th percentiles (62). No babies weighed more than the 90th percentile or were postterm (>42 wk).

Third trimester peripheral blood samples (8 ml) were collected into a BD Vacutainer CPT Cell Preparation tube (Franklin Lakes, NJ) containing sodium citrate and Ficoll Hypaque density fluid. The resultant peripheral blood mononucleated cells (PBMCs) were then resuspended in RNAlater (Ambion, Austin, TX) and stored at −80°C until analysis. Total RNA was prepared with an RNeasy kit (Qiagen, Hilden, Germany) and used for cDNA synthesis with an oligo (dT) primer (Affymetrix, Santa Clara, CA). The gene expression profile for each participant was assessed with the Affymetrix GeneChip Human Gene 1.1 ST microarray, which measures 764,885 probes covering 28,869 genes throughout the genome. Raw chip files were background corrected, log₂ transformed, and normalized by robust multiarray average in the Affymetrix Expression Console (Affymetrix) (20). A linear model was then fit to the normalized data to generate a matrix, including an expression value, for each probe.

RNA transcriptional profiles were contrasted by maternal genotypes with Limma and Surrogate Variable Analysis (31, 50) packages in R (44) with gene-chip processing date, gestational age at the time of study, and the presence or absence of hypertensive complications of pregnancy (preeclampsia, gestational hypertension) as covariates. Genes that were differentially expressed between our genotypes of interest at a Benjamini-Hochberg (BH) corrected P value <0.05 to control for multiple comparisons (5) were then analyzed by Ingenuity Pathway Analysis (IPA, Ingenuity Systems, http://www.ingenuity.com) software. IPA subdivides input data, in this case a list of differentially expressed genes, on the basis of molecular connectivity (networks), the assumption being that interrelationship likely represents significant biological function. A numerical score is given to indicate the degree of fit between the network and a list of differentially expressed genes; the score is based on the hypergeometric distribution and is calculated with the right-tailed Fisher's exact test. Finally, given the overrepresentation of mechanistic target of rapamycin (mTOR) pathway genes in the list of Andean selection-nominated candidate genes (see below), we generated a single P value for mTOR pathway genes using the Stouffer-Liptak method (32, 53) to evaluate the relationship of SNP genotypes with transcriptional activity within that pathway specifically.

Maternal and infant characteristics.

The Andean women were younger, shorter, weighed less when nonpregnant or during the 3rd trimester, and were of higher parity than the Europeans at high altitude (Table 2). The Andeans were also shorter than the Europeans at low altitude, but nonpregnant body mass index, pregnancy weight gain, and pregnancy skinfolds were similar in all groups. As expected the average NAAP was higher for Andeans than Europeans at high altitude. The cross-sectionally studied Andeans were also predominantly of Native American ancestry and did not differ from the longitudinal cohort of Andean women in any maternal or infant characteristic. At low altitude, NAAP reported for Europeans was considerably higher than at high altitude, but values in the low-altitude group principally reflected Native American ancestry of low-altitude origin.

The Andean and European infants at high or low altitude were similar with respect to length, gestational age, sex, and birth weight prior to adjustments for covariates. According to multiple linear regression analysis, maternal height (P < 0.05) and gestational age (P < 0.001) were positively associated with birth weight in our longitudinal cohort (Table 3). After adjusting for these covariates by setting them to the mean values for the two ancestry groups combined, we found birth weight to be greater in Andean compared with European infants at high but not low altitude (P < 0.001) (Table 2).

Candidate gene associations with birth weight.

We performed standard linear regression for each SNP to isolate the effect of maternal SNP genotype on infant birth weight after controlling for maternal height, gestational age, altitude of residence, infant sex, and population structure (using NAAP). Two of the 63 SNPs from the 16 natural selection-nominated candidate genes, PRKAA1 SNP rs1345778 and EDNRA SNP rs11731134, were related to infant birth weight in the 245 European or Andean women for the two altitudes combined with additive or dominant models (Table 4). Figure 2 shows the best model for each SNP, and Supplemental Tables S3–S5 provide the full results for additive, dominant, and recessive models.

For PRKAA1 SNP rs1345778, the GG and GT genotypes were associated with lower birth weight, and the TT genotype was associated with higher birth weight (Fig. 2A). Furthermore, the higher birth weight allele, T, was observed at a significantly higher frequency in the Andean (T = 0.88) compared with the European women (T = 0.73), and there were no high-altitude Andean women with the GG genotype (Table 5). For EDNRA SNP rs11731134, infants born to women with the CC or GC genotypes had higher birth weights compared with those born to women of the GG genotype (Fig. 2B). There were no differences in EDNRA SNP rs11731134 allele frequencies between ancestry groups (Table 5).

PRKAA1 and EDNRA SNP associations with UA diameter.

Previously we showed that greater UA blood flow protected Andeans against altitude-associated reductions in fetal growth (25). Since one means by which SNPs in or near EDNRA and PRKAA1 may influence birth weight is via alterations in UA diameter, we explored its associations with ENDRA and PRKAA1 SNP genotypes in our longitudinally studied participants at weeks 20 and 36 of pregnancy at high altitude. Maternal genotypes for PRKAA1 SNPs rs929785 and rs3805490 were associated with UA diameter measured at high altitude for pregnancy week 20 after a Bonferroni correction for multiple tests with a P value cutoff of P < 0.0056 (α = 0.05, 9 tests) was applied using additive and dominant models of inheritance (Supplemental Table S6). Furthermore, maternal genotypes for PRKAA1 SNPs rs929785, rs1345778, and rs3805490 were significantly associated with UA diameter measured at high altitude at 36 wk after the application of a Bonferroni correction for multiple tests with a P value cutoff of P < 0.0056 (α = 0.05, 9 tests) regardless of whether an additive or dominant model of inheritance was employed (Supplemental Table S7). For simplicity and because there were no women with GG genotypes at high altitude (Table 5), the associations between UA diameter measured at 36 wk and PRKAA1 SNP rs1345778 genotype using the dominant model are shown in Fig. 3A. EDNRA SNP rs765774 was associated with UA diameter measured at high altitude for pregnancy week 20 (Supplemental Table S6). There were no associations between UA diameter measured at high altitude for pregnancy week 36 with any of the three EDNRA SNPs considered (rs12508797, rs11731134, rs765774), including the rs11731134 SNP associated with birth weight (Fig. 3B, Supplemental Tables S6 and S7).

PRKAA1 SNP rs1345778 associations with gene transcriptional activity.

To explore the metabolic pathways by which maternal PRKAA1 rs1345778 SNP genotypes might influence fetal growth, we contrasted the transcriptional profiles for the GG/GT vs. TT women. We identified 828 transcripts that differed, 410 of which were upregulated and 418 downregulated, using a BH P < 0.05 to control for multiple comparisons. Using IPA we identified networks implicated by the 780 mapped genes (Supplemental Table S8). Table 6 shows the top networks identified (cellular assembly and organization, cellular function and maintenance, and amino metabolism, lipid metabolism and molecular transport) and the transcripts differing between GG/GT vs. TT genotypes for each network.

We then asked whether PRKAA1 SNP rs1345778 genotype influenced the expression of genes belonging to the mTOR pathway as defined by the Kyoto Encyclopedia of Genes and Genomes (KEGG). We focused on the mTOR pathway because 1) PRKAA1 is the catalytic subunit of AMPK, which is a regulator of the mTOR pathway, and 2) mTOR genes were overrepresented in the list of Andean selection-nominated candidate genes. Specifically, the probability that six genes, PRKAA1, PRKAA2, MTOR, PIK3CA, TNF, VEGFA, within our list of 40 selection-nominated candidate genes (see Supplemental Table S1) would be part of the 61 gene KEGG mTOR pathway is P = 6.25e-12. The transcriptional activity of 12 mTOR pathway genes (RHEB, IGF1R, RPS6, PIK3CB, PDPK1, VEGFA, PRKCG, PIK3R3, PRKAA2, EIF4EBP1, RRAGA, and RRAGB) differed between women with the GG/GT vs. TT genotype with a nominal P value of P < 0.05; the hypergeometric probability that 12 genes within the mTOR pathway would be differentially expressed due to random chance is P = 0.002 (Fig. 4). Furthermore, after observed correlation among genes were accounted for, transcriptional differences of mTOR pathway genes varied between women with the GG/GT vs. TT genotype at a P = 7.51e-09.