Epigenetic signature: implications for mitochondrial quality control in human aging

CpG methylation profiling of mitochondrial quality control candidate genes in differently aged humans

We analysed DNA methylation profiles using Sequenom MassARRAY EpiTYPER, a bisulphite-based technology that relies on base-specific cleavage and mass spectrometry and measures the level of methylation in amplicons containing one or more CpGs. To assess potential epigenetic changes during aging, the profiles were investigated in bisulfite-treated DNA samples collected from 381 subjects (48- to 107-year-old) of the discovery dataset. Specifically, we investigated a total of 1437 CpG sites located within CpG islands falling in candidate genes involved in mitochondrial biogenesis, mitophagy, fusion, and fission processes and annotated in the UCSC genome browser. Following stringent quality control criteria, a total of 500 CpG sites entered further analysis. The sequences of the CpG islands containing these sites and their chromosomal localization are shown in Figure S1. The results of the univariate linear regression analysis are presented in Table 1 and in Figure 1 in which we observed that 54 out of the 500 potential predictors are statistically significant after the Holm method to account for multiple testing. Changes of methylation levels with age at the 54 CpG sites are shown in Figure S2.

Starting from the set of 54 CpG units selected by the univariate analysis, we performed a stringent bootstrap-based stepwise selection procedure to remove potential spurious statistical associations. We found that among the most informative CpG markers (markers included in at least 80% of simulated models) there were BNIP3L_Amplicon1_CpG_10 (99.39% times), COX18_CpG_2 (85.16% times), COX18_CpG_15 (81.42% times), GABARAP_Amplicon2_CpG_7.8 (81.42% times), MARCH5_Amplicon1_CpG_2.3.4 (90.76% times), RAB32_Amplicon1_CpG_24 (86.6% times), RHOT2_Amplicon1_CpG_16 (91.31% times) and TFB1M_CpG_12.13 (80.83% times). The model including all these variables explained about 43.06% of age variance of the study sample.

In order to validate our findings, the CpG units selected by the permutation approach in the discovery dataset were tested in the replication dataset. Table 2 reports the results of the association obtained from the test. As it can be observed, we successfully replicated 2 out 8 significant associations and, in particular, those involving RAB32_Amplicon1_CpG_24 and RHOT2_Amplicon1_CpG_16 units.

Correlation of methylation profiles with geriatric components and survival at old age

We verified whether the variability of the two validated CpG markers (RAB32_Amplicon1_CpG_24 and RHOT2_Amplicon1_CpG_16) might affect the quality of life in the 60-89 years old group of the discovery dataset. In particular, this age group was analysed for examining the association between the methylation level variability at the analysed units and physical and cognitive abilities measured by geriatric assessments, including Hand Grip (HG), Activity Daily Living (ADL), Mini Mental State Examination (MMSE) and Geriatric Depression Scale (GDS). As shown in Table 3 we found that RAB32_Amplicon1_CpG_24 epigenetic variability was correlated with the risk of disability. In particular, subjects with methylation levels higher that 10% were more than twice as likely to develop disability than people with low methylation levels at that site. A borderline association was also detected between higher methylation level (>10%) at the RHOT2_Amplicon1_CpG_16 and MMSE performances.

Expression of RAB32 and RHOT2 genes

To explore the functional relevance of the methylation of the CpG islands located within RAB32 and RHOT2 genes, quantitative real-time PCR assays were carried out in order to evaluate the expression levels of mRNA in the samples of different ages. Consistently with the methylation patterns above described, we observed that mRNA levels of RAB32 and RHOT2 genes decrease and increase, respectively, during age further confirming their potential role as biomarkers of both chronological and biological age (Figure 2A and 2B).

Population samples

The discovery dataset included 381 unrelated individuals (185 men end 196 women) aged 48 to 107 years. The subjects older 60 years underwent a thorough geriatric assessment and a structured interview including the administration of a questionnaire validated at European level. The questionnaire collected socio-demographic information, anthropometric measures and a set of the most common tests to assess cognitive functioning, functional activity, physical performance, and depression. In particular, cognitive status was rated by Mini Mental State Examination (MMSE) [49]. Hand grip strength was measured by using a handheld dynamometer (SMEDLEY’s dynamometer TTM) while the subject was sitting with the arm close to his/her body. The test was repeated three times with the stronger hand and the maximum of these values was used in the analyses. The management of activities of daily living (bathing, dressing, eating, independence in and out of bed) was assessed by using the Katz’ Index of activities of daily living [50]. Depressive symptoms were assessed using the 15-item Geriatric Depression Scale [51].

The replication dataset included 468 subjects (206 men and 262 women) aged 18 to 108 years which were recruited at the INRCA Hospital, which is a reference point for the care of the aging people in the Calabria. Also these groups of subjects underwent through a geriatric assessment. Fully informed consent was obtained in writing from all the participants, and all the studies were approved by the Local Ethics Committee.

DNA samples

Six millilitres of venous blood were drawn from each human subject. Plasma/sera were used for routine laboratory analyses, while DNA was extracted from buffy coats following standard procedures.

Genomic DNA was obtained by phenol/chloroform purification. The DNA concentration and purity were determined spectrophotometrically.

Primer design for EpiTYPER assay

PCR primers for the genes of interest were designed using Sequenom’s EpiDesigner software (Table S1). They do not contain CpGs, amplify both methylated and unmethylated sequences equally, and delimit amplicon of size below 300 bp to increase the amplification success rate, covering as many CpGs as possible.

A T7-promoter tag (cagtaatacgactcactatagggagaaggct) was added to the reverse primers for the in vitro T7 transcription and a 10-mer tag sequence (aggaagagag) was added to the forward primers to balance the PCR primer length.

Bisulfite treatment and PCR conditions

Bisulfite conversion of each DNA sample was performed using the EZ-96 DNA Methylation-Gold kit (Zymo Research, Euroclone, Milan, Italy), according to the manufacturer’s protocol. Briefly, 1 μg of genomic DNA was added to 130 μl of CT conversion reagent in a final volume of 150 μl. The mix was incubated at 98 °C for 10 minutes and, successively, at 64 °C for 2.5 hours. After adding 400 μl of M-binding buffer to the wells of the silicon-A binding plate, each sample was loaded into the wells and centrifuged at 3000 g for 5 minutes. After adding of 400 μl of M-wash buffer to the wells and centrifugation at 3000 g for 5 minutes, 200 μl of M-desulfonation buffer was added to each well and incubated at room temperature for 20 minutes. Then, the solution was removed by centrifugation at 3000 g for 5 minutes and the wells were washed twice with 400 μl of M-wash buffer. Deaminated DNA was eluted in 30 μl of M-elution buffer. The PCRs were carried out in a total volume of 5μl using 1 μl of bisulfite-treated DNA, EpiTaq PCR buffer 1X, 0.4 μM of each primer, 0.3 mM dNTP mixture, 2.5 mM of MgCl₂, and 0.005 U TaKaRa EpiTaq HS (TaKaRa, Diatech Lab Line, Milan, Italy). The thermal profile used for the reaction included a 4 minute heat activation of the enzyme at 95 °C, followed by 45 cycles of denaturation at 94 °C for 20 seconds, annealing at optimal temperature for each primer pair (Table S1) for 30 seconds, extension at 72 °C for 1 minute, then one cycle at 72 °C for minutes. 0.5μl of each PCR product was electrophoresed on 1.5% agarose gel to confirm successful PCR and amplification specificity.

Dephosphorylation of unincorporated deoxynucleosidetriphosphates, and in vitro transcription and RNaseA cleavage

Unincorporated dNTPs in the amplification products were dephosphorylated by adding 1.7 μl DNase free water and 0.3 μl (0.5 U) shrimp alkaline phosphatase (SAP) (Sequenom, Inc., San Diego, CA, USA). Each reaction was incubated at 37 °C for 40 minutes, and SAP was then heat inactivated at 85 °C for 5 minutes. Subsequently, samples were incubated at 37 °C for 3 hours with 5 μl of T-cleavage reaction mix (Sequenom), containing 3.21 μl RNAse-free water, 0.89 μl 5X T7 polymerase buffer, 0.22 μl T-cleavage mix, 0.22 μl 100 mM DTT, 0.40 μl T7 RNA polymerase and 0.06 μl RNase A, for concurrent in vitro transcription and base-specific cleavage. The samples of cleaved fragments were then diluted with 20 μl water. Conditioning of the cleavage reaction was carried out by adding 6 mg of clean resin.

Mass spectrometry

10 nl of the resultant cleavage reactions were spotted onto silicon matrix preloaded chips (Spectro-CHIP; Sequenom) using the MassARRAY nanodispenser (Sequenom) and analysed using the MassARRAY Compact System matrix-assisted laser desorption/ionization-time-of-flight mass spectrometer (MALDI-TOF) (Sequenom). The spectra’s methylation ratios were calculated using EPITYPER software v1.0 (Sequenom). The method yields quantitative results for each of the sequence-defined analytic units referred as CpG units, which may contain either one individual CpG site or an aggregate of CpG sites. Triplicate independent analyses from sodium bisulfite-treated DNA samples were undertaken. The effectiveness of the entire experimental procedure was assessed by analyzing as control CpGenome Universal Unmethylated DNA (Chemicon) and CpGenome Universal Methylated DNA (Chemicon, Millipore, Germany) in serial mixtures of methylated and unmethylated products, with 10% methylation increments. Data quality control and filtering were carried out by the removal of the CpG dinucleotides whose the measurement success rate was <90%. Poor-quality and non-valuable data for the quantitative methylation of each CpG unit measured by MALDI-TOF-MS were excluded.

Expression profile analysis of the RAB32 and RHOT2 genes

The total RNA was extracted from the blood of individuals of various age (30–90 years old) using ReliaPrep RNA Tissue Miniprep System (Promega Corp, Italy). Briefly, 2.5 ml of fresh whole blood was mixed to 7.5 ml of RNA red blood cell lysis solution and incubated at room temperature for 10 minutes. White cells were isolated by centrifugating samples at 3000 g for 10 minutes and lysed in 200 μl of LBA+TG buffer and 85% μl of isopropanol. The lysate was then transferred to a ReliaPrep minicolumn, and RNA was purified according to manufacturer's recommendation.

The RNA concentration was measured for each sample using a spectrophotometer and purity of the sample was evaluated using the 260/280 nm absorbance ratio. RNA samples were treated with DNA‐free DNase to remove any residual genomic DNA contamination.

Reverse transcriptase‐PCRs (RT–PCR) were carried out using the RevertAid RT Kit (Thermo Fisher Scientific, Milan, Italy). First, a RT mix including 500 ng of total RNA and 1 μl of Oligo(dT)18 primers was preheated at 65 °C for 5 minutes. Then, the reaction was carried out in a 20 μl final volume containing 1X reaction buffer, 20 U of RiboLock RNase inhibitor, 1 mm of dNTP mix, and 200 U of RevertAid M‐MuLV RT reverse transcriptase. The mix was incubated at at 42 °C for 60 minutes and, successively, at 70 °C for 5 minutes to inactivate the reverse transcriptase. The cDNAs obtained were then used as a template for real‐time PCRs carried out using the SYBR Green qPCR Master Mix (Promega) in a StepOne Plus machine (Applied Biosystems, Milan, Italy).

The final PCR mixture (10 μl) contained 1μl of cDNA, 1X GoTaq qPCR Master Mix, 0.2 μmoles of each primer, and 1X CXR reference dye. Forward and reverse primers were as follows: RAB32For 5’-CAGGTGGACCAATTCTGCAAA-3’; RAB32Rev 5’-GGCAGCTTCCTCTATGTTTATGT-3’;

RHOT2For 5’-TGGAGCTGACTGCGGACTAT-3’; RHOT2Rev 5’-TCTGCACAAACTGGTAGC

CAA-3’; GAPDHFor 5’-ATGGGGAAGGTGAAGGTCG-3’; GAPDHRev 5’-GGGGTCATTGAT

GGCAACAATA-3’. The thermal profile used for the reaction included a 2 minutes heat activation of the enzyme at 95 °C, followed by 35 cycles of denaturation at 95 °C for 15 seconds and annealing/extension at 60 °C for 60 seconds, followed by melt analysis ramping at 60–95 °C. All measurements were taken in the log phase of amplification. Negative controls (in which water instead of cDNA was added) were also run in each plate. StepOne Software V 2.0 was used to analyze data. Gene expression values were normalized to GAPDH gene expression, used as internal control. In addition, the normalized values measured in the 30‐year‐old human were used as reference values (relative quantification) for the other samples.

Statistical analyses

Linear regression analyses were carried out to evaluate the level of DNA methylation as a function of age. To evaluate the combined effect of the analysed CpG markers with respect to the age of sample under study, we used a stepwise variable selection procedure based on the Akaike Information Criterion (AIC) combined with a bootstrap re-sampling method. In brief, this algorithm firstly simulates a new dataset taking a sample with replacement from the original dataset; secondly, it runs a stepwise selection procedure based on AIC on this simulated dataset, and, finally, repeats the previous steps n times. At the end of this procedure, this algorithm records how many times each variable (i) was selected, (ii) the estimate of the regression coefficient was statistically significant, and (iii) changed signs. The final model was selected by retrieving the variables that were selected in at least 80% of the bootstrap samples using linear regression. The prediction accuracy of the developed prediction model was assessed using the adjusted R² parameter, which is a measure of the proportion of age variance explained by a particular set of CpG predictors and their combined effect. Statistical analyses were performed using the R statistical language program (http://www.R-project.org/). In particular, the CpGassoc package was used to perform a linear regression analysis using the Holm method to account for multiple testing; the bootStepAIC package was used to perform stepwise variable selection procedure combined with a bootstrap resampling approach and, finally, ggplot2 packages were used for graphics purpose [52].

Student’s t-test was adopted to compare methylation profiles with respect to the analysed geriatric parameters.