Introduction

Epigenetic ageing and clocks have been studied and used in a wide range of biomedical and ecological research applications [1,2]. This field of research has assisted in the study of age related diseases and the role of environmental factors [3,4], and wildlife population management [5,6]. In human and most mammalian epigenetic ageing research, an Illumina methylation array or a custom array (HorvathMammalMethylChip40) is typically used [7–12]. This custom methylation array is highly effective in mammalian ageing research as it has probes for highly conserved and age associated cytosine-phosphate-guanine (CpG) sites. However, since it is mammalian specific, it cannot be used for non-mammalian species. Furthermore, in classes of vertebrates, such as bony fish, the most speciose group of vertebrates [13], the prospect of developing a universal fish array is low, due to a large evolutionary divergence between species [14,15]. Subsets of species that are similar on the genomic level could have arrays developed, however evolutionary divergent species may not be captured. Although other methylation arrays can be made similar to the one commonly used in humans [7]. This can be expensive and is unlikely to occur for other species. Therefore, it is important to be able to develop epigenetic ageing assays that can be applied to any species of interest.

In many wild species, alternatives to array-based technology have been used to develop epigenetic age assays [5,6,16–19]. One of these methods is a multiplex PCR followed by second-generation sequencing. The two important parts of this process from tissue to epigenetic age is the bisulfite treatment and the PCR. Bisulfite treatment involves the conversion of unmethylated cytosines to uracils. These uracils are then converted to thymines during PCR thereby making it possible to identify unmethylated sites during DNA sequencing. Bisulfite treatment can potentially degrade DNA, and may result in low quality DNA for sequencing and bias in conversion [20]. However, a new enzymatic-based method, which maintains DNA integrity, has recently been developed [21]. This method works in two parts. First, methylated cytosines are converted into products that cannot be deaminated. Second, unmethylated cytosines are converted to uracils. Both steps are carried out with enzymes and allow single base resolution of DNA methylation similar to bisulfite treatment. As DNA integrity is maintained, smaller quantities of starting DNA are required for this process [22–24]. This makes it desirable in settings where large quantities of DNA from individuals are difficult to collect or DNA is needed for multiple analyses.

In our previous studies, the PCR step for amplifying the age-associated loci has involved the handling of multiple PCR additives [18,19,25]. By handling each individual PCR additive, it is possible to make minor adjustments to the reaction to ensure proper amplification of the desired PCR product. However, manual handling of each reagent makes this assay susceptible to errors which may cause batch effects and are difficult to automate on liquid handling robots. Therefore, a commercial PCR master mix, whereby the only components to handle are the DNA, primers, and the master mix itself would be advantageous to reduce complexity. In this study, we aimed to improve an existing epigenetic ageing assay for the Australian Lungfish (Neoceratodus forsteri) [18] as a proof-of-concept for an improved workflow. This existing assay used a bisulfite treatment protocol and a multiplex PCR that contained handling of individual PCR additives. We compared a commercial bisulfite treatment approach with an enzymatic approach for DNA methylation detection and used a commercial PCR master mix to streamline the workflow. By using a known age data set of lungfish, we could detect if any modifications in the protocol resulted in differences in age prediction.

Methods

Results

Sequencing quality

Comparison between the two libraries found the enzymatic treated library to have a higher sequencing quality average across the read compared to the bisulfite treated library (Kolmogorov-Smirnov, p-value < 0.01, Fig 1A). Total reads per sample was also found to be significantly higher in the bisulfite (106,207 reads) treated library compared to the enzymatic (89,544 reads) treated library (paired t-test, p-value < 0.01, Fig 1B). As reflected in Fig 1B, the standard deviation of total reads was found to be greater in the bisulfite treated library. This too was found across all amplicons (Fig 2). A total of six samples were excluded in the bisulfite library as they did not have any coverage across multiple amplicons. This made age prediction impossible in these samples. In comparison, the enzymatic method had sufficient coverage across all amplicons (minimum coverage: 500 reads).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Sequencing quality between the bisulfite and enzymatic treated DNA over the A. length of the read (* Kolmogorov-Smirnov, p-value < 0.01) and B. total reads produced per sample in each library (* paired t-test, p-value < 0.01).

https://doi.org/10.1371/journal.pone.0297006.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Reads per amplicon between each the bisulfite and enzymatic treated DNA.

Each dot represents an individual sample and are coloured whether it was bisulfite or enzymatic treated. The dark horizontal line in each box whisker plot represents the median.

https://doi.org/10.1371/journal.pone.0297006.g002

Epigenetic age prediction

Epigenetic age prediction in both the bisulfite (Pearson correlation = 0.99, p-value < 0.01) and enzymatic (Pearson correlation = 0.99, p-value < 0.01) treated DNA libraries were found to have a high correlation between the known and predicted age (Fig 3A and 3B). The absolute error rates between each library were found to be not significantly different (paired t-test, p-value = 0.69, Fig 3C). Compared to the original assay which used an in-house bisulfite protocol [18], no statistical difference was observed in both the bisulfite protocol with the commercial kit in this study and enzymatic methods in absolute error (unpaired t-tests, p-values > 0.05). Both the bisulfite and enzymatic age predictions have a high correlation (Pearson correlation = 0.98, p-value < 0.01, S2 Fig in S1 File) suggesting both methods produce reliable age predictions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Epigenetic age prediction of the Australian Lungfish with either A. bisulfite or B. enzymatic treated DNA with a commercial PCR master mix used for both. C. Absolute error rates between each library. No statistical difference was found between each data set (paired t-test, p-value = 0.69).

https://doi.org/10.1371/journal.pone.0297006.g003

Discussion

Epigenetic age prediction is valuable in biomedical diagnostics and research and for wildlife management, but its widespread adoption is dependent upon both the ease of the assay, and the accuracy of the model to predict the age of the animal. The main limitation with developing an epigenetic clock is the calibration data set. Epigenetic clocks need to be calibrated with known age animals. Unfortunately, the species that need a molecular based method for age prediction, most likely do not have any known age information. Therefore, most epigenetic clocks most likely do not have an adequate sample size for calibration. However, there are still other aspects of epigenetic clocks such as the assay itself that need further improvement to be widely adopted. In this study, we have improved the sequencing quality for an established epigenetic ageing clock by using an enzymatic method and reduced the number of reagents required by using a commercial PCR master mix. These alternative and simplified approaches improve the overall assay, thus making it more feasible for large scale applications.

The advantage of the enzymatic method over bisulfite sequencing is that it produces less DNA degradation [32], and this is clearly demonstrated by our results (Fig 1). DNA degradation can be a significant issue in the collection of genetic samples from wildlife [33]. Therefore, an already degraded DNA sample may be unable to be sequenced after bisulfite treatment as the cumulation of DNA degradation from both the original sample and the bisulfite treatment may be too high. This makes the enzymatic method more practical when working with wildlife genetic samples that are at risk of DNA degradation because they are inevitably collected in the field, and often by non-experts, and sometimes under sub-optimal and challenging conditions. The enzymatic method also can work with less starting DNA, down to quantities as low as picograms of DNA [21]. This is another advantage for wildlife research, which often operates under tight animal welfare conditions, and in remote settings, where obtaining a high quantity of DNA can be difficult. By requiring less DNA and without degradation during the process, the enzymatic method is well-suited to characterising DNA methylation in genetic wildlife samples. In contrast to the bisulfite treatment, where in our experiment six samples were excluded from the analysis due to no coverage of amplicons, there was no drop out of samples with the enzymatic method with accurate age prediction. The drop out we observed could have resulted from several factors, but DNA degradation during bisulfite treatment is a strong candidate. Drop out of coverage could be overcome by splitting the amplicons into two separate reactions. However, this would require more DNA and would be counter to the desire for simpler laboratory workflows. Other possibilities would most likely be method specific with the bisulfite protocol. PCR showed amplified products on a gel, but because of a multiplex PCR, it is impossible to determine if all amplicons amplified. Potentially there could be further optimisation required to amplify all amplicons, such as modifying the ratios of primers in the PCR master mix. Our results indicate that an enzymatic method offers significant advantages over traditional bisulfite sequencing where certainty of obtaining results is important or where sample quality and quantity may be variable, as may be the case for wildlife samples.

As epigenetic ageing assays become more readily used in wildlife management there is a greater need to increase throughput, especially within fisheries where in the order of 1000s of samples could be processed. It is ideal to reduce the complexity of these assays to reduce batch effects and human error. In this study, we show a commercial PCR master mix performed as well as an existing customized master mix on both bisulfite and enzymatic treated DNA. The disadvantage to using a commercial PCR kit is that the make-up of the contents may be unknown. This can make it difficult in identifying the ideal candidate master mix for the PCR. Therefore, multiple commercial master mixes may need to be tested, thereby making the initial optimisation time-consuming. However, the benefit is that the ongoing assay is more simplified than a customised one.

Epigenetic age prediction is being used more in a variety of research applications. However, for it to be adopted more widely, especially for non-model organisms, the methods need to be more robust and simplified. In this study, we have improved a previous epigenetic age prediction assay by increasing the sequencing quality using an enzymatic method for DNA methylation and simplified the PCR by using a commercial master mix. Cost per sample remains a challenge outside of the biomedical field and currently the enzymatic is approximately double the cost. However, this may become comparable to bisulfite treatment with time. Depending on a researcher’s experiment and budget will determine if a bisulfite or enzymatic approach will be used. The addition of alternatives epigenetic research such as the enzymatic based method for DNA methylation may enable novel research. However, future research should focus on increasing assay efficiency. For example, making DNA sequencing higher throughput to process a larger number of samples or designing additional of i5/i7 indexes. With more improvement in the assays, we anticipate that epigenetic age prediction will become a routine methodology for a wide variety of applications.

Supporting information

Acknowledgments

The authors would like to thank the Australian Research Council project (LP130100118) team including Mark Kennard, Stewart Fallon, Sharon Marshall, Andrew McDougall, Peter Kind, Jane Hughes, Dan Schmidt, and Nick Bond for approving provision of samples. AGRF is supported by the Australian Government National Collaborative Research Infrastructure Initiative through Bioplatforms Australia. We would also like to thank Yi Jin Liew for internal review of our manuscript.