Establishing a stable, repeatable platform for measuring changes in sperm DNA methylation

Abstract

Background: Several independent research groups have shown that alterations in human sperm methylation profiles correlate with decreased fecundity and an increased risk of poor embryo development. Moving these initial findings from the lab into a clinical setting where they can be used to measure male infertility though requires a platform that is stable and robust against batch effects that can occur between sample runs. Operating parameters must be established, performance characteristics determined, and guidelines set to ensure repeatability and accuracy. The standard for technical validation of a lab developed test (LDT) in the USA comes from the Clinical Laboratory Improvement Amendments (CLIA). However, CLIA was introduced in 1988, before the advent of genome-wide profiling and associated computational analysis. This, coupled with its intentionally general nature, makes its interpretation for epigenetic assays non-trivial.

Results: Here, we present an interpretation of the CLIA technical validation requirements for profiling DNA methylation and calling aberrant methylation using the Illumina Infinium platform (e.g., the 450HM and MethylationEPIC). We describe an experimental design to meet these requirements, the experimental results obtained, and the operating parameters established.

Conclusions: The CLIA guidelines, although not intended for high-throughput assays, can be interpreted in a way that is consistent with modern epigenetic assays. Based on such an interoperation, Illumina's Infinium platform is quite amenable to usage in a clinical setting for diagnostic work.

Keywords: CLIA; Clinical Laboratory Improvement Amendments; DNA methylation; Epigenetics; LDT; Laboratory-developed test; Male infertility.

Fig. 1

Linearity and reportable range. Heat map of methylation level for eight samples comparing methylation level reported from WGBS (y- xis) with methylation level reported from 450K Human Methylation array (x axis) at loci with at least 30X coverage in the WGBS

Fig. 2

Reference range. a Reference interval for the methylation levels of 6690 CpG sites which have significant difference in methylation levels between fertile and infertile groups. Red and blue dots show the low and high limit for the 95% confidence interval respectively. b Distribution of the interquartile range for the methylation values of all CpGs in the 450HM array data for the known-fertile samples. c Density of methylation levels for hypo-methylated (blue) and hyper-methylated (red) loci in the fertile reference range exhibit unimodal distributions

Fig. 3

Analytical sensitivity experimental design and results. a Sample layout on 450HM chips. This experiment is looking at analytical sensitivity to various DNA concentrations. There are six different, color-coded samples (A, B, C, D, E, and F). Each sample contains eight different concentrations centered around the estimated limit of detection (LoD). Assuming the estimated LoD from previously collected data is 10 ng/ μl, the concentrations will be blank (0 ng/ μl), 0.005 ng/ μl, 0.01 ng/ μl, 0.05 ng/ μl, 0.1 ng/ μl, 0.5 ng/ μl, 1 ng/ μl, and 10 ng/ μl (standard concentration). b The percentage of failed samples at varying DNA concentrations. c The percentage of failed probes per sample at varying DNA concentrations

Fig. 4

Precision experimental design. The stratification of the samples analyzed for precision on different chips and varying positions on chips

Fig. 5

Precision results. a Empirical probability that a new replicate will give an aberrant methylation call consistent with existing replicates at different thresholds (of difference from fertile 95% confidence interval) for aberrant methylation calls. Dashed line shows the threshold used, which corresponds to an average probability of 0.997. b Euclidean distance between technical replicates of outlier sample from panel A. c Heatmap of the number of loci with different aberrant methylation status between technical replicates for two samples. Upper triangular matrix shows one sample and lower triangular matrix shows the other. Each row/column represents one technical replicate, and the cell at the intersection of a row/column is the number of differences between the sets of aberrantly methylated loci in the two replicates. The diagonal is the replicate compared with itself. Marginal cells (colored white) give the number of aberrantly methylated loci called in the corresponding replicate

Fig. 6

Analytical specificity. a The stratification of samples with different contamination levels on the 450HM array. There are eight different, color-coded samples labeled A through H. Each initial sample is divided into six replicates, which are contaminated with six different levels of bacterial cells; for example, A1 = no bacterial contamination, A2 = 10% of cells in sample are bacterial in origin, A3 = 20% of cells, A4 = 30% of cells, A5 = 40% of cells, and A6 = 50% of cells. b The number of aberrantly methylated loci per sample for different concentrations of bacterial cells (E. coli DNA) in the sample. c Scatter plot of the measure of probe intensity vs. the number of aberrantly methylated loci (failed probes are removed before calling aberrant methylation) per sample, stratified by the amount of bacterial contamination in the sample

Fig. 7

Summary of criteria. This matrix summarizes the findings of this study in each of the experiment groups in terms of the platform’s suitability for the intended use, the selection of operational parameters and the identification of criteria for continued performance monitoring