Paraffin Embedding Contributes to RNA Aggregation, Reduced RNA Yield, and Low RNA Quality

Abstract

The RNA isolated from FFPE tissues is of poor quality and quantity. Other studies have indicated that formaldehyde fixation or the duration of storage of tissue blocks accounted for RNA damage. Herein we report a third source of harm to RNA: embedding in warm paraffin. RNA bound to oligo(dT)-conjugated magnetic beads (an mRNA model) and total cellular RNA pellets were passed through formalin, graded ethanols, xylene, paraffin, and a formaldehyde demodification step. The mRNA model yielded at least 1550 bp amplicons at RT-PCR at each step of processing except paraffin, which yielded no more than 750 bp amplicons regardless of paraffin formulation or transition solvent. Quantitative RT-PCR on paraffinized RNA suggested a 1400-fold or more decrease in amplifiable RNA when compared with control. Compared with earlier processing steps, formalin-fixed paraffinized total cellular RNA produced only high-molecular-weight RNA and insoluble aggregates. These species were reproduced by heating RNA in hydrocarbon solvent at 60°C for 1 hour. Quantitative RT-PCR on paraffinized RNA suggested an at least 10- to 160-fold decrease in amplifiable RNA compared to controls. The data implicate paraffin embedding as primarily responsible for the high-molecular-weight RNA aggregates, reduced yields of RNA, and poor quality of RNA isolated from these chemical models of FFPE tissues.

Tissue specimens available for molecular analysis are often formalin-fixed and paraffin-embedded (FFPE).¹ Microarray and quantitative RT-PCR (RT-qPCR) assays detect mRNA expression levels that can be correlated to progression of disease.² The primary obstacle to such analyses on archival samples is the poor quality of FFPE-extracted RNA.^3,4 Although specimen quality at the time of processing dominates access to all molecular analyses,¹ once embedded in paraffin, the method of RNA extraction is the only factor that can be controlled to optimize RNA quality and quantity. Although many published and commercial reports describe reagents and methods to control pH and RNases, reverse formalin adducts, and increase nucleic acid solubility, RNA yields are poor, the bases may be chemically modified, and the isolated RNA seems to be fragmented to an average length of 100 to 200 bases.^3,5,6 RT-PCR on FFPE-derived RNA is generally inefficient for amplicons longer than 120 to 300 bases⁷ and cannot detect low-level transcripts, although DNA sequences can be detected at concentrations several orders of magnitude lower. Fixatives and preservatives have been recently developed for tissue specimens to yield higher quality extracted RNA.¹ However, FFPE remains a practical and effective technique that provides most archival materials.

In nearly 20 years of empirical efforts to optimize RNA extraction from FFPE samples, few improvements have been identified in the conditions (eg, temperature) and components (ie, salt, buffer, chelator, detergent, protease, and RNase inhibitor) described in the first report of RNA extraction from FFPE tissues,^8,9 although a final heating step such as preamplification¹⁰ improves yield. Studying the structural changes to RNA that occur during FFPE processing provides an alternative to purely empirical exploration of methods for improving RNA analysis. Understanding the chemical changes in RNA that occur during fixation, processing, and embedding may lead to better strategies for preventing or reversing degradation and, thus, to extraction of higher quality RNA. Previous investigations have demonstrated that the most common formaldehyde-RNA adducts were methylol (hydroxymethyl) groups, ethyl methyl ethers, and methylene bridge cross-links on the exocyclic amine of the adenine base.^11,12 These were reversible in model systems using mononucleotides and homopolymeric octamer ribonucleotides.^13,14 The isolation of high-quality RNA was described from FFPE tissues stored for up to a month after paraffin embedding.^14,15 Although formalin adducts were not fully reversed on heating of isolated total RNA in Tris-EDTA or proprietary proteinase K buffers,^3,10,14,15 these exceptional results encouraged us to further explore the possibility that little or no permanent damage occurred to RNA during FFPE processing. The objective of the present study was to investigate the contribution of each step of chemical processing in a model of fixed-tissue processing (ie, formalin fixation, graded alcohols, transition solvent, paraffin embedding, and their reversal) to RNA damage. The focus was on the chemical effects of formaldehyde, solvents, and heat on RNA, without tissue components. The reasoning was that this minimalist chemical approach could lead to improvements in the RNA obtained from FFPE tissues.

Materials and Methods

Reagents

Unless otherwise stated, all buffers were prepared in diethyl-pyrocarbonate (DEPC)–treated water purchased from Fisher Scientific (Pittsburgh, PA). Paramat was purchased from Electron Microscopy Sciences (Hatfield, PA). Paraplast Extra and Paraplast Plus were purchased from Tyco Healthcare/Kendall (Mansfield, MA). Hemo-De was purchased from Fisher Scientific, and reagent grade octane from Sigma-Aldrich Corp. (St. Louis, MO). HeLa cells were grown to confluence in 100-mm tissue culture dishes, and their total cellular RNA was harvested using 0.75 mL per dish TRIzol LS (Invitrogen Corp., Carlsbad, CA) according to the manufacturer's instructions.

Formaldehyde Fixation and Paraffin Embedding

Seventy-five micrograms total HeLa RNA was adsorbed to 50 μL Dynabeads oligo(dT)25 (Invitrogen Corp.) according to the manufacturer's instructions in autoclaved 1.7-mL polypropylene microfuge vials. Samples were washed twice with 0.3 mL wash buffer [150 mmol/L LiCl in 10 mmol/L Tris (pH 7.4)] by exposing tubes to magnets and removing the supernatant. As expected, greater than 95% of the total RNA was removed in these washes, as measured by absorbance at 260 nm. The beads were washed using 0.3 mL 5× nuclease-free PBS and resuspended in the same buffer. An equal volume of 10% (v/v) aqueous methanol-free formaldehyde (Polysciences, Inc., Warrington, PA) was added, and the suspensions were manually mixed and allowed to react for 1 hour on a rocking platform at room temperature. Excess formaldehyde was removed by washing the beads once with 0.4 mL PBS. Samples were incubated at room temperature in 0.3 mL 70% (v/v) ACS/USP grade ethanol (Pharmco-AAPER; Brookfield, CT) for 5 minutes, 0.3 mL 85% ethanol for 5 minutes, and 0.3 mL 100% ethanol for 10 minutes. Beads were resuspended in 0.3 mL xylene (Arcturus Paradise; Molecular Devices Corp., Sunnyvale, CA) and incubated at room temperature for 20 minutes. Xylene was removed and replaced with approximately 0.3 mL liquid paraffin (Paraplast), and the samples were incubated at 60°C for 30 minutes. Samples were cooled to room temperature while being manually shaken to evenly coat the inside of the tubes with material. After overnight incubation at room temperature, the beads were extracted twice for 15 minutes using 0.75 mL 60°C xylene, and twice using room temperature xylene. Samples were incubated at room temperature in 0.3 mL 100% ethanol for 10 minutes, 85% ethanol for 5 minutes, and 70% ethanol for 5 minutes, and were resuspended in 0.3 mL wash buffer. RNA was released from beads by heating in 40 mmol/L Tris-acetate and 1 mmol/L EDTA (1× TAE) (pH 9) at 70°C for 30 minutes, exposing the tubes to a magnetic field, and quickly removing the supernatant.

In all RNA pellet experiments, essentially the same procedures were used. Aliquots of 25 to 100 μg RNA were formalin-fixed (or were untreated) in a volume of 0.3 mL and precipitated by adding 0.9 mL ethanol, hand mixed, and centrifuged at 12,000 × g for 10 minutes. Pellets were washed twice with 75% ethanol, and were recovered by centrifugation at 8000 × g for 2 minutes at each subsequent step. The volumes of graded alcohols, xylenes, and paraffin were changed to 1 mL. After reverse processing, samples were resuspended by heating at 60°C for 1 minute in 75 μL diethyl-pyrocarbonate–treated water. The formaldehyde demodification step consisted of heating in 1× TAE (pH 9) at 70°C for 30 minutes.

PCR Amplifications

RT-PCR experiments were performed using a two-step method. Equal amounts of RNA from each sample (1 ng, as measured by absorbance at 260 nm, assuming that 1 AU = 40 μg/mL) were reverse transcribed using a cloned AMV first-strand cDNA synthesis kit (Invitrogen Corp.) primed using random hexamers (unless otherwise stated) according to the manufacturer's instructions.

RT-PCR was performed for six amplicons from the 2077-bp Ku80 mRNA sequence determined for HeLa cells (GenBank Accession No. M32865). Primers were purchased from Integrated DNA Technologies (Coralville, IA). Each 25-μL reaction contained 0.05 ng cDNA, 1× TaqPCR Master Mix (Promega Corp., Madison, WI), and 2 μmol/L of each primer. The primers used were 1206F, 5′-TGCAGCATTGTGCAGATACACACC-3′, and 1435R, 5′-AGCGAAGCTTCTCAACGATAGCCT-3′ (230 bp); 1057F, 5′-TTGATGCTCATGGGTTTCAAGCCG-3′, and 1435R (379 bp); 1057F and 1563R, 5′-TGCTTCAACCTTGGGCAATGTCAG-3′ (507 bp); 464F, 5′-TGTGGGTCTGTGCCAACCTCTTTA-3′, and 1209R, 5′-TGCAACCTCCTTCTCCAGACACTT-3′ (746 bp); 464F and 1435R (972 bp); and 464F and 2014R, 5′-ACCATCACCATGGCAACAGAAAGC-3′ (1551 bp). Samples were incubated at 95°C for 2 minutes; 35 cycles of 95°C for 30 seconds, 56°C for 45 seconds, and 72°C for 3½ minutes; followed by 72°C for 5 minutes. Products were separated on 1.5% agarose gels stained using ethidium bromide, and were photographed before digital scanning.

RT-qPCR was performed for an 84-base β₂-microglobulin amplicon as previously described.^16,17 Each 25-μL reaction contained 0.1 ng cDNA, 1× Taqman Universal PCR Master Mix (Applied Biosystems, Inc., Foster City, CA), forward primer 5′-TGACTTTGTCACAGCCCAAGATA-3′ (0.3 μmol/L), reverse primer 5′-AATCCAAATGCGGCATCTTC-3′ (0.3 μmol/L), and probe 5′- (VIC)TGATGCTGCTTACATGTCTCGATCCCA(TAMRA)-3′ (0.2 μmol/L). Samples were incubated at 50°C for 2 minutes, then 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute, monitoring VIC fluorescence on a genetic analyzer (model 7700; Applied Biosystems, Inc.). Thresholds were fixed at 0.04 ΔRn for each experiment.

Denaturing Agarose Gel Electrophoresis

For examination of RNA, 1% w/v agarose gels were prepared in 1× running buffer [20 mmol/L MOPS (3-(N-morpholine)propanesulfonic acid), 1 mmol/L EDTA, and 5 mmol/L NaOAc (pH 7)] with 18% v/v of a 37% solution of formaldehyde. Before loading, samples were heated at 55°C for 15 minutes with an equal volume of a sample buffer consisting of 1× running buffer, 50% v/v formamide, 18% v/v of 37% solution of formaldehyde, and 10 μg/mL ethidium bromide.¹⁸

Statistical Analysis

P values were calculated using the two-tailed paired Student's t-test function of Excel software (Microsoft Corp., Redmond, WA).

Results

Modeling Effects of Tissue Processing Using RNA Bound to Paramagnetic Beads

RNA was bound by the interaction between poly(A) mRNA tails with oligo(dT)25-mers on paramagnetic beads, and a substantial portion remained attached throughout formalin fixation, graded alcohols, xylene, paraffin embedding, and reversal of each of these steps (Table 1). The RNA, presumed to be mRNA, was released from the beads by heating in 1× TAE buffer (pH 9) at 70°C for 30 minutes. This demodification step frequently removes nearly all formaldehyde-RNA adducts.¹⁹ Formalin fixation resulted in loss of approximately two-thirds of bound RNA, possibly because substantial modification of A bases decreased their binding to T. The amount of bound RNA remained substantially the same throughout graded ethanols and xylenes. About one-third to one-half of the remaining bound RNA was lost during paraffin embedding. RNA that was bound to beads, washed, and eluted by heating in TAE (native/untreated) demonstrated a ratio of absorbances at 260 to 280 nm of 2.15. Formalin fixation, incubation in organic solvents, and paraffin embedding decreased the mean 260/280 ratios, and these differences were statistically significant at P < 0.05.

Table 1.

Assessment of RNA Released from Oligo(dT) Paramagnetic Beads (mRNA model) Through Each Step of Fixed-Tissue Processing and Reversal

Formalin	−	+	+	+	+
Ethanols	−	−	+	+	+
Xylene	−	−	−	+	+
Paraffin	−	−	−	−	+
Xylene	–	−	−	−	+
Ethanols	−	−	+	+	+
TAE 70°C	+	+	+	+	+
Absorbance ratio 260/280	2.15 ± 0.06	1.92 ± 0.08	1.99 ± 0.20	1.61 ± 0.18	1.83 ± 0.14
RNA amount recovered (μg)	1.11 ± 0.19	0.38 ± 0.09	0.30 ± 0.08	0.34 ± 0.04	0.20 ± 0.06
Cq⁎	19.2 ± 0.2	20.7 ± 0.6	21.1 ± 0.2	22.8 ± 0.6	27.3 ± 0.8

^⁎Controls for water template and unfixed RNA without RT gave C_q values greater than 35; input RNA reached a C_q of 25.3 ± 0.3. The same amount of RNA, as measured by absorbance at 260 nm, was used to prepare cDNA for each sample, priming with random hexamers. Values are given as mean ± SD of at least three experiments. RT-qPCR amplifications were of an 84-mer β₂-microglobulin amplicon.^16,17

Concentrations of RNA released from oligo(dT) magnetic beads at each step were adjusted so that the same amount of RNA was reverse transcribed into cDNA in each sample. Two PCR methods were used to assess the quality of available mRNA. First, RT-qPCR for a β₂-microglobulin transcript revealed a quantification cycle (C_q) of approximately 19 (Table 1) for the original RNA bound to beads. Mean C_q values increased at each processing step; however, the only difference that was statistically significant at P < 0.01 was paraffin embedding. Paraffin embedding increased the C_q by approximately 8. Assuming (1 + E)ⁿ amplification with E of 1, where E is efficiency and n is the cycle number,²⁰ the accessible mRNA decreased at least 256–fold. Correcting for RNA loss, accessible mRNA decreased at least 1400-fold. Second, RT-PCR targeting six amplicons of the Ku80 gene transcript was performed. At each step of processing except paraffin embedding, RT-PCR results were positive for segments ranging from 230 to 1551 bp (Figure 1). In contrast, paraffin-embedded samples produced only positive RT-PCR results for amplicons up to and including 746 bp.

Figure 1.

Effect of histologic processing steps on the size of recoverable PCR amplicons. RNA bound to paramagnetic beads was formalin-fixed, passed through graded alcohols and xylene, and embedded in paraffin. After reversing these steps, samples were heated at 70°C for 30 minutes in 1× TAE buffer (pH 9). Equal amounts of RNA were reverse-transcribed into cDNA that was used as a template for the PCR amplifications of sequences corresponding to the Ku80 mRNA from HeLa cells. The left lanes are ϕX174 DNA/HaeIII ladders. Images are inverted from agarose gels stained using ethidium bromide.

Effects of Paraffin Formulation and Transition Solvent on RNA Degradation

Embedding samples in paraffin wax and/or subsequent wax removal was the step that produced the greatest RNA degradation in the paramagnetic beads model of tissue processing. To address the possibility that this was an artifact of one formulation or lot of paraffin, the same experiments were performed using different formulations of paraffin. Paraplast Plus includes dimethyl sulfoxide to increase permeability, and Paraplast Extra includes additional polymers for block consistency. Under the same conditions, embedding samples in these paraffin types and a formulation (Paramat) from another manufacturer produced the same results as with the original embedding medium (Table 2). Similarly, two additional transition solvents, Hemo-De and octane, were examined. Under the same conditions, these transition solvents reproducibly yielded substantially the same results as those obtained using xylene (Table 2).

Table 2.

Effects of Embedding Medium and Transition Solvent on Amplifiable RNA Attached to oligo(dT) Magnetic Beads (mRNA Model) through Paraffin and Back to Aqueous Buffer

Variable	PCR-positive amplicon (bp)
	230	379	507	746	972	1551
Transition solvents, embedded in Paraplast Regular medium
Octane	+	+	+	+	−	−
Hemo-De	+	+	+	+	−	−
Xylene transition solvent, embedded in medium
Paraplast Extra	+	+	+	+	−	−
Paraplast Plus	+	+	+	+	−	−
Paramat	+	+	+	+	−	−

Data are given as reproducibly amplified (+) or not amplified (−) in at least three independent experiments. Samples were heated at 70°C for 30 minutes in TAE (pH 9) before cDNA preparation. In all experiments, equal amounts of cDNA were used as templates for PCR.

Modeling Effects of Fixed Tissue Processing with Total RNA Pellets

Total RNA pellets were examined using denaturing agarose gel electrophoresis. Total RNA isolated from HeLa cells in DEPC-treated water was either untreated or fixed in formalin, and passed through graded alcohols and xylene, then paraffinized. These processes were reversed, and the formalin adduct demodification step was performed. Analysis of RNA from each step of this process, before paraffin embedding, showed bands consistent with migration of the 28S and 18S ribosomal subunits during denaturing agarose gel electrophoresis (Figure 2). In all cases, heating in TAE (pH 9) for 30 minutes at 70°C increased the smearing of bands and the intensity of staining. Unfixed RNA migrated similarly in every chemical step of fixed-tissue processing including paraffin embedding. Formalin-fixed RNA consistently demonstrated minor high-molecular-weight bands, which were reversed on demodification in every step of processing except paraffin embedding. During paraffin embedding, much of the fixed RNA was reproducibly stuck in the wells of the denaturing agarose gels (Figure 2). Heating in TAE reversed this and demonstrated slightly increased smearing relative to that of unprocessed RNA, as well as poorer resolution of the band corresponding to the 28S ribosomal subunit. The apparent absence of the 28S ribosomal subunit (Figure 3) suggested that degradation or altered migration was reproducible in most but not every experiment. Tubes containing the recovered paraffinized samples consistently exhibited a white precipitate, which was not solubilized on formaldehyde demodification or during heating in gel loading buffer. However, partial dissolution was demonstrated after heating for 2 hours at 70°C in TAE (pH 4) (data not shown). In gel loading buffer, these precipitates absorbed approximately 50% of the soluble ethidium bromide dye after 3 days at 4°C, which suggests that they were composed of RNA. Essentially the same amounts of RNA were recovered at each step except paraffin embedding (Table 3). The data reflect a 10-fold or greater decrease in RNA. In typical experiments, 6-fold or greater to 12-fold less RNA was recovered from paraffin (data not shown). Native/untreated RNA had a ratio of absorbances at 260 to 280 nm of 2.13 (Table 1). Each subsequent tissue-processing step produced lower (approximately 2.0) mean 260/280 ratios, and some were statistically significant at P < 0.01. Because equal amounts of RNA from RNA pellets processed through xylenes but not paraffin appeared the same at agarose gel electrophoresis (Figure 2), the same amount of RNA from each sample was analyzed using RT-qPCR for β₂-microglobulin. When primed with random hexamers, C_q values were essentially the same as those for native/untreated RNA processed up to and including xylenes (Table 3). The amplification of RNA processed through paraffin was decreased in a statistically significant manner (P < 0.01). The mean C_q for paraffin-embedded RNA was increased by approximately 4 compared with native/untreated RNA. Correcting for RNA loss, this indicated no less than a 160-fold decrease in accessible mRNA. When primed with a β₂-microglobulin transcript-specific primer, reverse primer 5′-AATCCAAATGCGGCATCTTC-3′, C_q values were essentially the same as native/untreated RNA for samples processed up to and including xylenes (Table 3). Whereas P was a modest 0.04, the mean C_q for paraffin-embedded RNA was increased by 1 to 2 compared with native/untreated RNA. Correcting for RNA loss, this indicated a 10-fold or greater to less than 40-fold decrease in accessible mRNA. Formalin-fixed RNA without demodification was detectably amplified when the cDNA was prepared using transcript-specific primers but not random hexamers (Table 3).

Figure 2.

Effect of histologic processing steps on the electrophoretic mobility of total cellular RNA. RNA isolated from HeLa cells was formalin-fixed or not (native/untreated RNA), passed through graded alcohols and xylene, and paraffin-embedded. After reversing these steps, samples were either untreated or heated at 70°C for 30 minutes in 1× TAE buffer (pH 9). Equal amounts of RNA were heated in loading buffer at 65°C for 10 minutes and loaded onto denaturing agarose gels. 28S and 18S indicate migration of ribosomal subunits. Images stained using ethidium bromide are inverted. Lanes labeled with an asterisk indicate samples that contained visibly insoluble material before and after all heating steps.

Figure 3.

Heating of formaldehyde-treated RNA in nonpolar solvent was sufficient to produce high-molecular-weight species. Samples were processed and reverse processed as indicated, and equal amounts of RNA were heated in loading buffer at 65°C for 10 minutes and loaded onto denaturing agarose gels. Images stained using ethidium bromide are inverted. Lanes labeled with an asterisk indicate samples that contained visible insoluble material before and after all heating steps. Left panel: Heating fixed RNA in nonpolar solvents produces aggregates. The first lane contained a 1-kb DNA ladder. Middle panel: Effect of time on conversion to high-molecular-weight RNA in xylene heated at 60°C. Right panel: Effect of temperature on conversion to high-molecular-weight RNA in xylene heated for 60 minutes.

Table 3.

Assessment of RNA Recovered From Total RNA Pellets Through Each Step of Fixed-Tissue Processing and Reversal

Formalin	−	+	+	+	+
Ethanols	−	−	+	+	+
Xylene	−	−	−	+	+
Paraffin	−	−	−	−	+
Xylene	−	−	−	−	+
Ethanols	−	−	+	+	+
TAE 70°C	+	+	+	+	+
Absorbance ratio 260/280	2.13 ± 0.02	1.98 ± 0.01	1.99 ± 0.02	2.00 ± 0.02	2.03 ± 0.13
RNA amount recovered (μg)	31.5 ± 3.2	30.4 ± 0.3	32.6 ± 2.3	27.8 ± 0.8	2.6 ± 0.2
Cq primed with random hexamers⁎	28.9 ± 1.2	30.4 ± 0.3†	31.0 ± 0.5	30.0 ± 0.5	33.4 ± 0.5
Cq primed for β2-microglobulin⁎	27.6 ± 0.4	28.0 ± 0.1‡	27.8 ± 0.5	27.6 ± 0.5	29.3 ± 0.9

^⁎Samples were heated at 70°C for 30 minutes in TAE (pH 9) before cDNA preparation, using the same amount of RNA for each sample. C_q values are given as the mean ± SD of at least three experiments. Controls for water template and native/untreated RNA without RT yielded C_q values greater than 35.

^†Without demodification by heating in TAE, C_q was not reached.

^‡Without demodification by heating in TAE, C_q was 31.9 ± 0.7.

Effect of Paraffinization Temperature and Conditions on RNA Pellets

To address the possibility that the transition from liquid to solid paraffin could have been responsible for RNA aggregation, samples were processed through paraffin. After the 30-minute heating step in paraffin, samples received warm xylenes (“liquid”) or were placed on ice for 3 minutes to solidify before immediate reverse processing (“solid”). The same amount of RNA from each sample was primed using random hexamers, and analyzed using RT-qPCR for β₂-microglobulin. C_q values of 30.8 ± 1.1 and 31.1 ± 0.8 were reached for samples processed through liquid paraffin and solid paraffin, respectively, which were heated in TAE (pH 9) for 30 minutes at 70°C (data not shown). This suggested that the transition from liquid to solid paraffin had little influence on RNA degradation.

To address the possibility that heating RNA pellets in hydrocarbon solvents, either transition solvent or paraffin, could have been responsible for aggregation, samples were heated at various steps of processing. Formalin-fixed RNA processed to ethanol and heated at 60°C in ethanol for 1 hour migrated essentially the same as RNA processed to xylene and paraffinized native/untreated RNA (Figure 3). However, heating in xylene at 60°C for 1 hour and paraffinization in which samples were reverse processed immediately migrated essentially the same as RNA processed through paraffin. The high-molecular-weight species observed under paraffinization were reproduced by heating in xylene in a time- and temperature-dependent manner (Figure 3). This implied that heating formalin-fixed RNA in either xylene or paraffin (hydrocarbon “solvents”) produced substantially reversible cross-linking and substantially irreversible adducts.

Discussion

RNA fragmentation correlates with poor RT-qPCR amplification.²¹ When RNA was extracted from samples shortly after FFPE tissue processing, substantial RNA fragmentation was reported.²² However, the electrophoretic migration (Figures 2 and 3) of our processed RNA was inconsistent with fragmentation to an average of 100 to 200 bases reported for typical FFPE isolated RNA.^3,5,6 Others^14,15 have reported more or less intact RNA after FFPE processing. RNA with clear 28S and 18S ribosomal bands was extracted from FFPE tissues stored for up to a month after paraffin embedding.^14,15 Extensive RNA fragmentation may require the storage of FFPE tissues at room temperature for at least a year.¹⁵

RNA yield was reduced in the mRNA and total RNA chemical models of fixed-tissue processing. In the total RNA (pellet) model, the 10-fold RNA loss was observed only at the paraffin step (Table 3). We suspect that the threefold loss of RNA on formalin fixation in the mRNA (magnetic beads) model may have been due to formalin modification of adenine bases disrupting the interaction of poly(A) tails with the oligo(dT) matrix.¹⁴ Less RNA (less than twofold) was lost on paraffin embedding, possibly because of tight electrostatic interactions in nonpolar solvents. The insoluble RNA and high-molecular-weight RNA observed on paraffin embedding (Figures 2 and 3) are consistent with multiple methylene cross-links. These species were observed only in processed RNA that had been fixed using formaldehyde (Figure 2).

Because similar aggregation and degradation of RNA were observed for warm paraffin and xylenes (Figure 3), the large reduction in solvent dielectric constant from water and ethanol (approximately 80 and 25, respectively) to xylenes and paraffin hydrocarbons (approximately 2.0 and 1.7, respectively) may explain the similar effects of either hydrocarbon on RNA aggregation. Standard FFPE processing is performed in room temperature xylenes and warm paraffin because both solvents must be liquid to penetrate tissues. At room temperature, xylene is a liquid and paraffin is a solid. It should not be surprising that the same effects on RNA are observed with use of either hydrocarbon solvent at elevated temperatures.

As previously reported,¹⁴ our formalin-fixed and native/untreated total cellular RNA migrated similarly during denaturing agarose gel electrophoresis (Figure 2). In separate experiments, formalin-fixed total cellular RNA migrated slowly compared with native/untreated RNA at electrophoresis using a bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA) (data not shown). It is possible that the harsher conditions of denaturing agarose gel electrophoresis, in particular the 2.2-mol/L concentration of formaldehyde as a “denaturing agent,”¹⁸ obscured the differences between formalin-fixed and native/untreated total cellular RNA. We used denaturing agarose gels to examine RNA because slowly migrating species were not detected on the bioanalyzer despite being quantifiable by UV absorbance. In samples in which the bulk of the RNA remained in the sample wells of agarose gels, only background signal was present on the bioanalyzer, probably because the bioanalyzer reads nucleic acid absorbance only as it is eluted from the gel matrix.

Tissue fixation and storage methods such as HOPE (HEPES–glutamic acid buffer–mediated organic solvent protection effect)²³ may preserve RNA for molecular analyses that are superior to those extracted from FFPE tissues. This study suggests two possible contributions to the success of such procedures. The HOPE method fixes in alcohol, omitting formalin, and uses low-melting paraffin (melting temperature approximately 52°C).²³ Without formalin fixation, no apparent RNA cross-linking was observed (Figure 2). Higher quality RNA obtained from low-melting paraffin may correlate with the temperature-dependent formation of high-molecular-weight formalin-fixed RNA (Figure 3). The duration of paraffin-embedding procedures varies among laboratories and with respect to the size and shape of pathologic specimens. Embedding time can range from hours to days.^3,24 Most FFPE tissue blocks provide little or no information on paraffin-embedding conditions. Inasmuch as the time and temperature that samples spend in warm hydrocarbon solvents should be minimized during paraffinization and its reversal (Figure 2), such conditions may prove important. Embedding times can be reduced by using techniques such as application of vacuum.²⁵ Attaching FFPE sections to slides via overnight heating is likely to degrade RNA.

These data suggest additional practical strategies for improving the quality of the RNA extracted from FFPE tissues. Others have cautioned against use of oligo(dT) primers for generation of cDNA in formalin-fixed systems.^10,14,15 In the present study, use of transcript-specific primers to prepare cDNA seemed to provide superior detection of transcripts compared with use of random hexamers, especially for RNA that was substantially modified using formalin (Table 3). This may be due to increased targeting of reverse transcription to a specific RNA rather than all RNAs. Our results reproduce findings that a formaldehyde demodification step on isolated RNA, such as heating in dilute Tris buffer, improves results of RT-PCR and RT-qPCR.^10,14,15

Data from the present study demonstrate that paraffin embedding causes significant RNA damage in warm nonpolar solvents after formaldehyde fixation and dehydration. This damage includes low-quality RNA as measured using conventional and RT-qPCR amplification, low RNA yield (Tables 1 and 3), and RNA aggregation or cross-linking (Figure 3). Although it could be argued that isolated RNA might behave differently than that in FFPE tissue sections, it seems unlikely that the cellular environment, such as the presence of RNA-protein adducts, resulted in significant differences in the processes that caused degradation in heated paraffin. RNA is highly concentrated in a cellular environment, and there are no obvious functional differences between RNA-RNA and protein-RNA cross-links that would prevent isolation of intact RNA from fixed tissue samples.¹