The telomere-mitochondrial axis of aging in newborns

Study population

Mothers with a singleton full-term birth were selected from the ongoing population-based prospective ENVIRONAGE (ENVIRonmental influence ON early AGEing) birth cohort study, which is located in Limburg, Belgium. Detailed study procedures have been described previously [7]. Between 2010 and 2017, 1530 mother newborn pairs were recruited. In 691 samples, cord plasma protein levels (p53 and PGC-1α) were measured. After removing 4 missing data, 18 outlying data (3 SDs from the mean) for protein levels and 10 outlying data for TL and mtDNAc, a total of 613 participants were used for statistical analysis. After delivery, mothers completed study questionnaires to provide detailed information on maternal age, paternal age, maternal education, smoking status, parity, and newborn’s ethnicity. Maternal education was classified as “low” when mothers did not obtain any diploma, “middle” when they obtained a high school diploma, and “high” when they obtained a college or university diploma. Mothers were categorized as “never smoker”, “former smoker” when they had quit smoking before pregnancy, and “smoker” if they had smoked at any time point during pregnancy. Parity was categorized in mothers having their first newborn, having their second newborn, or having their third or more newborn. Newborns were classified as either “European” when two or more grandparents were European, or as “non-European” when at least three grandparents were of non-European origin. Other maternal and perinatal parameters such as maternal pre-pregnancy Body Mass Index (BMI), newborns’ sex, birth weight and birth date were collected from medical records after birth. Maternal pre-pregnancy BMI (kg/m²) was calculated based on data obtained during the first antenatal consultation. The date of conception was estimated by combining data on the first day of the mother’s last menstrual period and the first ultrasonographic examination. The ENVIRONAGE study protocol has been conducted according to the Helsinki Declaration and was approved by the Ethical Committees of Hasselt University in Diepenbeek, Belgium (reference no. B371201216090 and B371201524537) and East-Limburg Hospital in Genk, Belgium. Written informed consent was provided by all participating mothers.

Sample collection and preparation

Procedures for umbilical cord blood and placental tissue collection for TL and mtDNAc assessment have been described in detail previously [7]. BD Vacutainer^® Plus plastic whole-blood tubes with spray-coated K2EDTA (BD, Franklin Lakes, NJ, USA) were used to collect umbilical cord blood samples immediately after delivery. To obtain buffy coat and retrieve blood plasma, samples were centrifuged at 3,200 rpm for 15 min and stored separately. Within 10 minutes after delivery, placentas were collected and stored at –20°C. Four different placental biopsies (approximately 1–2 cm³) were taken at the fetal side at 4 cm from the umbilical cord and directly underneath the chorioamniotic membrane. Contamination by the chorioamniotic membrane was avoided by dissection followed by visual examination. All samples were stored at –80°C until DNA extraction. DNA was extracted using the QIAamp DNA Mini Kit (Qiagen Inc., Venlo, the Netherlands), according to the manufacturer’s instructions. The quantity and purity of the sample was measured with the Nanodrop spectrophotometer (ND-1000; Isogen Life Science, the Netherlands). DNA samples were normalized to ensure a uniform DNA input of 5 ng for each qPCR reaction, and this was checked using the Quant-iT™ PicoGreen^® dsDNA Assay Kit (Life Technologies, Europe). Extracted DNA was stored at –80°C until further use.

Average relative TL and mtDNAc measurement

TL and mtDNAc were measured in cord blood buffy coat and placental tissue using a previously described modified quantitative real-time polymerase chain reaction (qPCR) protocol [8–10]. Details are provided in supplement (Supplementary Material). On each run, a 6-point serial dilution of pooled buffy coat or placental DNA was used to assess PCR efficiency as well as eight inter-run calibrators (IRCs) to account for inter-run variability. Non-template controls were also used in each run. All samples were measured in triplicate on a 7900HT Fast Real-Time PCR System (Applied Biosystems) in a 384-well format. qPCR curves for each sample were visually inspected and when technical problems were detected or triplicates showed too high variability, samples were removed for further analysis. Using qBase plus software (Biogazelle, Zwijnaarde, Belgium), all measurements were processed and normalized to a reference gene, taking into account run-to-run differences. Average relative TL was calculated by determining the ratio of one telomere gene copy number (T) to one reference gene (36B4) (S). Average mtDNAc was calculated by determining the average of the ratio of two mitochondrial gene copy numbers (MTF3212/R3319 and MT-ND1) (M) to two nuclear reference genes (36B4 andβ-Actin) (S). Telomere assay-reliability was assessed using an intra class correlation coefficient (ICC). The inter-assay ICC was 0.936 (95% CI: 0.808 to 0.969) and the intra-assay ICC was 0.952 (95% CI: 0.947 to 0.956).

p53, PGC-1α and SIRT-1 protein measurement

After thawing cord blood plasma, 100 μl of plasma was quantified for p53 protein levels (U/mL) using a Human p53 ELISA Kit according to the manufacturer’s instructions (ref. ab46067, Abcam, Cambridge, United Kingdom). For PGC-1α (ng/mL), 100 μl of 1:250 diluted plasma was quantified according to the manufacturer’s instructions (ref. E-EL-H1359, Elabscience, Texas, USA). Two commercially available Human SIRT1 ELISA Kits were tested for the detection of cord plasma SIRT-1 protein levels, but for none of them the limit of detection was reached (ref. ab171573, Abcam, Cambridge, United Kingdom and ref. E-EL-H1546, Elabscience, Texas, USA). All protein levels were measured in duplicate using a FLUOstar Omega microplate reader (BMG Labtech, Ortenberg, Germany). To minimize variability, all samples were randomized across 96 well plates and all ELISAs for one protein were measured in one day, while using the same serial dilution. Five IRCs per plate were taken into account to control for potential variability between plates. The intra-assay coefficient of variation was 5% for p53, and 10% for PGC-1α while the inter-assay coefficient of variation reached 14.5% for p53 and 25% for PGC-1α.

DNA methylation measurement

Cord blood DNA was used to determine the epigenome-wide DNA methylation levels. DNA samples were bisulfite-converted, amplified and then hybridized to the Illumina Infinium Human MethylationEPIC 850K BeadChip array (Illumina, San Diego, CA, USA) at the GenomeScan lab (Leiden, The Netherlands). The array measurements were scanned using an Illumina iScan and the data quality was assessed using the R script MethylAid. DNA methylation data preprocessing, quality control, outlier detection, batch effect removal and probe filtering were described previously [11]. In total, 57 CpG loci with their UCSC reference gene name referring to TP53, PGC1A or SIRT1 were selected for the present study. Methylation levels of TP53, PGC1A and SIRT1 were available for 205 participants and used for further analysis. Details are provided in Supplementary Table 1.

Statistical analysis

All statistical analyses were performed using R studio version 3.6.2 (R Core Team, Vienna, Austria). Shapiro-Wilk’s test was used to check the normality of the distributions. Average relative TL, mtDNAc and cord plasma protein levels were log₁₀-transformed to better approximate a normal distribution. For the descriptive statistics, continuous variables were presented as means ± standard deviation (SD) and categorical variables as numbers (frequency in percentage). Pearson correlation was used to systematically evaluate following correlations: (1) TL in cord blood and placenta and mtDNAc in the respective tissue, (2) TL in cord blood and placenta and cord plasma protein levels, (3) mtDNAc in cord blood and placenta and cord plasma protein levels and (4) cord plasma protein levels. Multiple linear regression was applied to further confirm the associations independent of potential confounding effects. We adjusted for a priori selected covariates based on known associations of these factors with TL, mtDNAc, and protein levels as shown previously in multiple studies [8, 12]: Technical covariates (sample storage and batch effects), newborn’s sex, gestational age, maternal BMI, maternal and paternal age, ethnicity, parity, smoke status, maternal education and month of delivery. All model estimates were presented as percentage difference with 95% CI and expressed for a 10% increment in explanatory variable.

In a secondary explorative analysis, we evaluated in a subpopulation whether cord blood methylation levels of genes corresponding to our measured protein levels (p53, PGC-1α and SIRT1) were related to the studied age-related markers and protein levels, to further explore the interrelationship of the telomere-mitochondrial aging axis. We restricted these DNA methylation association studies to age-related markers measured in cord blood, as it is known that DNA methylation levels are highly tissue- and cell type- specific [13–15]. Since multiple CpGs were measured for each gene, we first made a correlation matrix showing the Pearson correlations between the CpGs for each gene. Second, we performed Principal Component Analysis (PCA) to reduce the dimensionality of our dataset [16]. To determine the number of Principal components (PCs) to retain in our analysis, we created scree plots for each gene (Supplementary Figure 1). PCs on the left side of the “elbow” of the graph were retained in the analysis. For all genes, this resulted in 5 PCs to retain. Next, we correlated the PCs with the CpGs, to determine their loadings (Supplementary Table 2). CpG loci were selected as relevant to a factor if the absolute value of their factor loadings were larger than 0.45 [16]. First, these first five PCs were used to assess the Pearson correlations between methylation data and both cord blood age-related markers and cord plasma protein levels. Second, multiple linear regression was applied to confirm these associations while adjusting for the aforementioned covariates.

Study population characteristics

Demographic, pregnancy-related and perinatal characteristics of the mother-newborn pairs included in this study (n = 613) were summarized in Table 1. Our study subset was representative for the original study population (Supplementary Table 3). Mothers were on average 29.3 (SD: 4.6) years old and had an average pre-pregnancy BMI of 24.6 (SD: 4.8) kg/m². Half of the participating women were highly educated (50%). The majority of the pregnant women never smoked cigarettes (63%), 25% stopped smoking before pregnancy, and 11% kept smoking on the average 3.4 cigarettes/day during pregnancy. Among the newborns, 52% were girls. Newborns had an average gestational age of 39.2 (SD: 1.67) weeks, an average birth weight of 3420 (SD: 496) grams, and most were of European origin (86%). Information about the age-related and protein markers is given in Table 2. Cord plasma SIRT-1 levels were excluded from the analysis since the measurements did not reach the limit of detection (LOD) of 0.31−0.63 ng/ml.

The telomere-mitochondrial axis of aging: links between TL, mtDNAc, p53 and PGC-1α

Unadjusted correlations

Cord blood and placental TL were positively correlated (r = 0.40, p < 0.0001, Supplementary Figure 2A), but no correlation between cord blood and placental mtDNAc was observed (Supplementary Figure 2B). The following correlations were systematically evaluated: (1) TL and mtDNAc, (2) TL and cord plasma protein levels, (3) mtDNAc and cord plasma protein levels and (4) between the cord plasma protein levels.

First, a positive correlation was found between TL and mtDNAc in both cord blood (r = 0.23, p < 0.0001, Supplementary Figure 3A) and placental tissue (r = 0.28, p < 0.0001, Supplementary Figure 3B).

Second, cord plasma p53 levels were negatively correlated with cord blood TL (r = −0.13, p = 0.0015, Supplementary Figure 4A) but not with placental TL (Supplementary Figure 4B). Cord plasma PGC-1α levels were negatively correlated with both cord blood TL (r = −0.10, p = 0.013, Supplementary Figure 4C) and placental TL (r = −0.09, p = 0.029, Supplementary Figure 4D).

Third, no correlation was observed between cord plasma p53 levels and cord blood mtDNAc (Supplementary Figure 5A), but tended to be negatively correlated with mtDNAc in placenta (r = −0.07, p = 0.094, Supplementary Figure 5B). No correlation was observed between cord plasma PGC-1α levels and cord blood mtDNAc (Supplementary Figure 5C), but a negative correlation was observed with placental mtDNAc (r = −0.11, p = 0.0082, Supplementary Figure 5D).

Fourth, cord plasma p53 and PGC-1α levels were not correlated (Supplementary Figure 6). In Figure 1, a visual summary of all correlations is presented, which is based on the experimentally derived telomere-mitochondrial axis of aging hypothesis.

Figure 1. Summary of the results found in this study integrated into the experimentally based telomere-mitochondrial axis of biological aging hypothesis. DNA damage and telomere shortening activate p53 leading to growth arrest, senescence or apoptosis. p53 might also impair mitochondrial function and mitochondrial DNA content indirectly through suppression of PGC-1α – one of the master regulators of the mitochondria – leading to mitochondrial comprise and increased ROS levels, which leads to more DNA damage including telomere shortening. p53 and PGC-1α could therefore be central players in the association between telomere length and mitochondrial DNA content and subsequently in the aging process. Solid lines represent significant associations between age-related or protein markers, while non-significant associations are represented by dotted lines. p53 and PGC-1α levels were only measured in cord blood, while TL and mtDNAc were measured in both cord blood and placental tissue. Abbreviations: p53: tumor suppressor protein 53; PGC-1α: peroxisome proliferator-activated receptor gamma co-activator 1 alpha protein. Figure based on the experimental work of Sahin et al. [5, 33].

Cord blood TP53, PGC1A and SIRT1 methylation levels and the cord blood telomere-mitochondrial axis of aging