Laser capture microdissection-reduced representation bisulfite sequencing (LCM-RRBS) maps changes in DNA methylation associated with gonadectomy-induced adrenocortical neoplasia in the mouse

Abstract

DNA methylation is a mechanism for long-term transcriptional regulation and is required for normal cellular differentiation. Failure to properly establish or maintain DNA methylation patterns leads to cell dysfunction and diseases such as cancer. Identifying DNA methylation signatures in complex tissues can be challenging owing to inaccurate cell enrichment methods and low DNA yields. We have developed a technique called laser capture microdissection-reduced representation bisulfite sequencing (LCM-RRBS) for the multiplexed interrogation of the DNA methylation status of cytosine-guanine dinucleotide islands and promoters. LCM-RRBS accurately and reproducibly profiles genome-wide methylation of DNA extracted from microdissected fresh frozen or formalin-fixed paraffin-embedded tissue samples. To demonstrate the utility of LCM-RRBS, we characterized changes in DNA methylation associated with gonadectomy-induced adrenocortical neoplasia in the mouse. Compared with adjacent normal tissue, the adrenocortical tumors showed reproducible gains and losses of DNA methylation at genes involved in cell differentiation and organ development. LCM-RRBS is a rapid, cost-effective, and sensitive technique for analyzing DNA methylation in heterogeneous tissues and will facilitate the investigation of DNA methylation in cancer and organ development.

Figure 1.

LCM-RRBS workflow. A complex tissue is dissected using LCM. Extracted DNA is digested by the methylation-insensitive enzyme MspI, end repaired and ligated with methylated Illumina adapters. After bisulfite conversion, each sample is ‘barcoded’ by introducing a sample-specific index (shown as green, blue or violet boxes) through low-cycle PCR. Samples are pooled and loaded onto a high-percentage gel for fragment separation and size selection. Using universal primers, the final library is amplified and sequenced on the Illumina platform.

Figure 2.

LCM-RRBS is reproducible and robust across 1 ng extracted from bulk fresh frozen and FFPE samples. CGI methylation (top panels) and the methylation at 2-kb regions flanking the TSS (bottom panels) were compared between (A) 1 ng (LCM-RRBS) and 400 ng of purified leukocyte genomic DNA (RRBS), and (B) 1 ng of FFPE DNA and 1 ng of fresh frozen genomic DNA extracted from the same endometrial tumor (LCM-RRBS).

Figure 3.

LCM-RRBS is robust across microdissected samples collected from fresh frozen and FFPE tissues. Fresh frozen and FFPE mouse liver was collected for DNA methylation profiling. LCM was used to collect tissue from areas ranging in size from 20 to 2 mm². CGI methylation (top panels) and methylation at 2-kb regions flanking the TSS (bottom panels) were compared between (A) fresh frozen samples (LCM-RRBS) and 400 ng of purified mouse liver genomic DNA (RRBS), and (B) FFPE samples (LCM-RRBS) and 400 ng of purified mouse liver genomic DNA (RRBS).

Figure 4.

DNA methylation profiling of GDX-induced adrenocortical neoplasms and adjacent normal tissue using LCM-RRBS. The adrenal glands of three ovariectomized DBA/2J mice were fresh frozen in Tissue-tek O.C.T. compound, cryosectioned and stained. Shown are representative cryosections pre- and post-LCM. Normal cells in the zona fasciculata contain large lipid droplets that are easily recognized. In contrast, neoplastic cells distort the normal adrenal zonal architecture and contain relatively few lipid droplets. The microdissected normal tissue included zona glomerulosa and zona fasciculata cells; care was taken to avoid dissection of X-zone (X), medulla (M) or capsule cells, as these cell types have distinct developmental origins (48), and therefore may have different epigenetic fingerprints. An average of 5.5 mm² of neoplastic (red) and normal (green) tissue per adrenal pair was collected and analyzed using LCM-RRBS.

Figure 5.

Validation of differentially methylated promoters. The DNA methylation of one hypermethylated and three hypomethylated promoters was interrogated by BSP and sequencing across enriched neoplastic and normal samples. All genes show a statistically significant difference (Fisher’s exact test, P < 10⁻¹⁵) in DNA methylation using BSP. Each colored box represents an individual CpG dinucleotide within a 2-kb region centered around the TSS. High (yellow), moderate (black), low (blue) and undetermined methylation levels are shown for each CpG. The mean methylation of each region interrogated is shown to the right of each heatmap. The red box indicates the region of the promoter that was interrogated by LCM-RRBS and BSP.