Differential long-term stability of microRNAs and RNU6B snRNA in 12–20 year old archived formalin-fixed paraffin-embedded specimens

Total RNA from FFPE radical prostatectomy specimens was previously obtained from a cohort of 92 patients who underwent surgery at Johns Hopkins between 1993 and 2001, and were followed for recurrence until 2004 [ 18]. Briefly, malignant regions were identified by a pathologist and two 0.6 mm tumor cores were obtained from each sample. Samples were deparaffinized with xylene, extracted with ethanol, and protease digested. Total RNA was extracted with the RecoverAll Total Nucleic Acid Isolation Kit (Ambion, Grand Island, NY) and stored at −80 °C. RNA quantity was assessed by Nanodrop spectroscopy (Thermo Scientific, Wilmington, DE) and RNA integrity score were obtained with Agilent RNA Pico kit and the 2100 Bioanalyzer Instrument (Agilent Technologies, Wilmington, DE) as per manufacturer instructions.

Reverse transcription and quantitative PCR were previously performed using the Taqman MicroRNA Reverse Transcription Kit and Applied Biosystems qRT-PCR probe and primer sets (Applied Biosystems, Grand Island, NY) [ 18]. Briefly, 2 ng of RNA from FFPE samples was applied to RT and QPCR was performed with the Bio-Rad MyiQ single color RT-PCR detection system (Bio-Rad Laboratories, Hercules, CA). Standard curves (2 ng/μl to 0.016 ng/μl) for RNU6B and for each miRNA transcript were generated from cell lines known to express each miRNA (PC3 for miR-221; LNCaP for miR-141; and MCF-7 for miR-21). The resulting standard curves and results were used to determine assay linearity, qRT-PCR efficiency, and SQmean.

Over the last several decades the Johns Hopkins retropubic radical prostatectomy program has treated thousands of men with localized prostate cancer [ 19– 21]. Archival specimens from these patients provide a unique resource to study the effects of long term FFPE tissue block storage on miRNA stability. As a single institution study of a single type of cancer, these samples underwent standardized processing, fixation, and storage procedures. Therefore, they present a unique and well-controlled resource for miRNA analyses. Tissue blocks from a previously defined nested-case control cohort of 92 radical prostatectomy patients, treated at Johns Hopkins Hospital between 1993 and 2001, were identified and total RNA was extracted from pathologist-defined tumor tissue [ 18]. The characteristics of patient and tumor samples are provided in Table  1. Transcript levels of miR-21, miR-141, miR-221, and the commonly applied reference gene, RNU6B snRNA, were quantified by Taqman quantitative RT-PCR assay. The results demonstrate a clear and statistically significant inverse association between transcript quantity and FFPE block age for all four small non-coding RNA transcripts by Spearman rank correlation analysis (Fig.  1, Additional file 1: Table S1). These results indicate a continuous loss of miRNA and RNU6B snRNA transcript stability over storage time.

The quality of RNA from older FFPE samples is generally considered to be poor. To investigate whether RNA quality is indicative of miRNA and snRNA transcript levels, we calculated the RIN for each sample. RIN scores ranged from 1 to 3.9 (Average 2.1 ± 0.5). No association was found between sample RIN and FFPE block age (Additional file 1: Figure S1). Moreover, no statistically significant correlation was found between RIN and RNU6B snRNA, miR-21, or miR-221 transcript levels by Spearman rank correlation (Fig.  2a, b and d). However, a significant correlation was observed between RIN and miR-141 transcript levels (Fig.  2c), indicating that RNA quality can be associated with the stability and detectability of some miRNAs. Collectively, these results identify FFPE block age as the most consistent feature associated with miRNA and snRNA transcript detectability and stability. The results of these statistical analyses are summarized in Additional file 1: Table S2.

Mature miRNAs are approximately 22 nucleotides in length while common endogenous reference controls, such as RNU6B snRNA, are slightly longer (Table  2). Given the differences in length and function between miRNAs and reference snRNAs, we investigated whether there may be corresponding differences in signal stability over time by general linear models. RNU6B and miR-21 were found to have comparable stabilities over time (Fig.  3a). However, the rate of RNU6B snRNA degradation was significantly more rapid than that of miR-141 and miR-221 (Fig.  3b– c). Accordingly, miR-21 was also found to have a more rapid degradation rate when compared to miR-141 and miR-221 (Fig.  3d). These data support that commonly studied microRNAs and reference controls can have significantly different stabilities in FFPE tissues stored over long periods of time. These results may reflect differences in small non-coding RNA sequence, function, or association with proteins. They further suggested that quantitative miRNA studies from long-term archived FFPE tissues may benefit from epidemiologic study design or statistical analysis methods that take into account differential storage-dependent transcript degradation.