RNA quantity was assessed spectrophotometrically using NanoDrop^® ND-1000 Spectrophotometer (Wilmington, USA), which showed that the yields from snap frozen extracts were greater than those from FFPE when RNA was extracted from identical numbers of cells using all the protocols examined (Stratagene Absolutely RNA^® FFPE Kit, Ambion RecoverAll™ Total Nucleic Acid Isolation Kit, Gentra Purescript^® RNA Purification Kit and Invitrogen Trizol^® Reagent). In comparison of FFPE extracts (Table 1), Ambion gave the highest yields, and column based Stratagene and Ambion protocols produced clean RNA with OD 260/280 ratio greater than 1.8. RNA quality was assessed using Agilent 2100 Bioanalyzer (Agilent technologies, Waldbronn, Germany) and TaqMan^® RT-PCR, which showed variations in RNA quality dependant on the protocol used. Stratagene and Ambion RecoverAll™ kits gave superior FFPE RNA results with regard to quality than the others examined.

Modification to the Stratagene and Ambion protocols generated approximately 25–40% greater yields and larger fragments of RNA (Figure 2) than the standard procedure. Adjustment to Stratagene and Ambion protocols produced decreased C_Ts (e.g. with a mean of 2.95 cycles in GEMM experiment and a mean of 3.14 cycles in PreAmp), indicating the improved quality of RNA extracted from FFPE (Figure 3).

A separate experiment using Ambion RecoverAll™ kit enabled extraction of large RNA molecules including cross-linked RNAs (Figure 4). Incubation of eluted RNA at 70°C for 20 minutes was found to be the best condition to disrupt cross-links while not compromising RNA integrity in comparison with other modifications (70°C for 10 min, 95°C for 10 min and 95°C for 20 min).

C_T values were also dependent on the type of TaqMan^® PCR Master Mix that was used (Figure 3 and Figure 5). TaqMan^® with UPMM generated higher C_Ts for long amplicons than GEMM. This was observed by a comparison of two GAPDH assays – one 122 bp and the other 67 bp. There was a mean threshold detection difference of 17 cycles using UPMM and a mean difference of 4 cycles using GEMM with RNA template extracted using Ambion RecoverAll™ kit (Figure 3). In addition, HLA_A is the largest amplicon (164 bp) which produced no product with UPMM, however, GEMM and PreAmp both allowed the detection of this amplicon size (Figure 3). GEMM improved the C_T values for long amplicons (over 100 bp) compared with UPMM (Figure 5-a). When the ΔC_Ts were compared to Theoretical ΔC_Ts, PreAmp results generally correlated with GEMM for all the 8 assays analyzed (Figure 5-c).

TaqMan^® PreAmp analysis consistently achieved decreased C_T values in both snap frozen and FFPE aliquots compared with TaqMan^® without pre-amplification step (Figure 5-b, c). A good correlation between PreAmp and UPMM was observed using CDKN1B to analyze RNA extracted with different protocols (Figure 6), suggesting that PreAmp does not introduce any bias into the reaction.

TaqMan^® analysis over all assays showed C_Ts to be 2 to 11 cycles higher when using FFPE extracts compared to snap frozen counterparts when the amounts of input RNA were identical (Figure 3). Focussing on an analysis of GAPDH using two assays with different amplicon sizes revealed the smaller amplicon assay (67 bp) produced C_Ts from snap frozen and FFPE samples that were closer together (Figure 7).

Generally, RNA extracted from FFPE materials is fragmented and chemically modified. Fragmentation occurs possibly since tissues are surgically removed, and it continuously occurs during fixation and preservation. Experimental design was such that equal numbers of cells were processed for FFPE cell block construction and snap freezing for comparative purposes. By using cells fixed under controlled conditions possible variables due to the effects of storage on RNA degradation were negated, thus revealing the true impact of formalin fixation on RNA quality. We extracted intact RNA fragments in a separate experiment from pellets with a large number of formalin fixed paraffin embedded normal thyroid cells (Figure 4), which was consistent with a finding by Scicchitana et al 17. In that study, intact RNA molecules were extracted from newly formalin fixed paraffin embedded human bone marrow stromal cells suggesting that RNA degradation could be a minor problem for recently fixed FFPE in some types of cells. Our study showed that the differences in RNA quality between the small and large fixed pellets (Figure 2 and Figure 4) might be due to the fact that the majority of cells in the centre of the large pellet were remained alive during construction of the large solid cell pellet. This could limit the access of RNases to intracellular mRNA because degradation will not fully occur until the fixation process begins resulting to the intact mRNA being maintained. Unfortunately, intact mRNA is not generally obtained in majority of fixed tissues or cell lines, and RNA extracted from FFPE is normally degraded to fewer than 300 bp.

Figure 4 displays the degree of RNA modification in FFPE cells caused by methylol groups during formalin fixation. Masuda and colleagues 18 suggested that formaldehyde reacts with RNA forming an N-methylol followed by an electrophilic attack to form a methylene bridge between amino groups. Interestingly, our results show a clear large RNA band, approximately 5 kb in length (Figure 4-lane 6), which is hypothetically the cross-linked RNA. The other two bands visualized were also larger than that extracted from snap frozen cells indicating the modification of RNA in FFPE. Furthermore, with incubation at high temperature, the large RNA band (~5 kb) was removed, and the sizes of the other two bands became closer to 18 s and 28 s of the snap frozen extracts. This is in agreement with the results of a mass spectrometric analysis that was carried out by Masuda group, suggesting methylol modification is reversible by heating 1819. However, a balance must be achieved between breaking cross-links while not contributing to degradation of labile RNA. Our results showed that incubation at 70°C for 20 minutes was the optimal condition, among those tested, to de-modify cross-linked RNA while maintaining RNA integrity (Figure 4).

A prerequisite for gene expression studies in formalin-fixed tissue samples was the establishment of a reliable and reproducible extraction method to provide detectable RNAs for subsequent analysis 1. The quality of extracted RNA can be variable among different extraction methods due to several factors, such as the contamination of RNases, proteins and genomic DNA 20. Our results showed that proteinase K digestion based protocols followed by on column DNase digestion and RNA elution produced good detectable RNA of FFPE.

Proteinase K digestion has also been demonstrated by many laboratories to be critical in RNA extraction protocol 132122 possibly because of two functions. On one hand, it can degrade proteins that are covalently cross-linked with each other and nucleic acid to release RNA from the matrix, thereby allowing efficient RNA extraction from FFPE materials 21. On the other, it can inactivate RNases that tend to be stable and do not require cofactors to function. Thus its activity avoids any potential reactivation of RNase during reversal of fixation in aqueous buffers 1.

Our data demonstrates that incubation in proteinase K buffer at 70°C for 20 minutes facilitated the disruption of cross-links, resulting in improved quantity and quality of RNA. The RNA extracted using the modified protocols (with the incubation) enhanced detection by a mean of 3 cycles (Figure 3), which is equivalent to 8 fold increased sensitivity. It is most likely that the additional incubation step removed the remaining cross-links between RNA and protein leading to the longer RNA molecules being extracted 23. In addition, this incubation denatured proteinase K, thus avoided its further damage to RNA in the following purification procedure, which suggests the necessity of a heating (at 70°C for 20 min) application in any proteinase K based extraction of FFPE samples.

TaqMan^® QRT-PCR technique is based on the 5' nuclease activity of Taq DNA polymerase and involves cleavage of a specific fluorogenic hybridization probe that is flanked by PCR primers 2425. Because of the small target size, many laboratories have demonstrated that it is possible to measure gene expression levels using FFPE tissues as a source of mRNA 3. Still, it seems to be problematic to perform large scale of analysis on FFPE because of the high C_Ts and limited concentration of extracts 1415.

We employed a novel TaqMan^® PreAmp technique which we found to be a practical solution to decrease C_T values, and in particular suitable in our hands to generate real-time PCR results from limited amounts of input RNA, such as extracted from laser captured microdissected material 26. The principle of TaqMan^® PreAmp technique is to amplify target cDNA prior to real-time TaqMan^® PCR analysis. Briefly, cDNA is synthesized from total RNA by use of random priming. The cDNA for the specific target assays is then amplified by pre-amplification reaction using pooled gene-specific primers to increase the number of targeted copies. The pre-amplification product is diluted and finally analyzed by real-time TaqMan^® PCR using single assay containing one pair of gene-specific primers and probe.

The TaqMan^® PreAmp technique addresses the challenge faced by researchers working with rare or precious samples, which only limited RNA could be extracted from, to perform gene expression analyses using real-time QRT-PCR. The simple process enables the user to perform uniform amplification from as little as one nanogram of cDNA (which was used in this study) or alternatively conduct up to 200 real-time PCR reactions per pre-amplification reaction without compromising the available sample material. Our results demonstrated that TaqMan^® PreAmp overcame the difficulties usually caused by low yields of RNA extraction from FFPE.

This study demonstrated that the sensitivity of TaqMan^® detection was influenced by choice of Master Mix and amplicon length in assay design. We evaluated UPMM and GEMM which showed similar sensitivity for short amplicons (less than 90 bp), while GEMM displayed better sensitivity for longer amplicons (over 100 bp) compared with UPMM in parallel snap frozen and FFPE extracts (Figure 5). This effect was dissipated when UPMM was used after pre-amplification, possibly because of the increased copy number of template available or because of the composition of the PreAmp MasterMix which is closer in components to GEMM than UPMM. Reagents can have a significant effect on assay reproducibility 27 due to some parameters such as different polymerases sensitivity 28, primer binding efficiency and the concentration of Mg²⁺ 29. Karrer et al. described the Monte Carlo effect using plant material suggesting that the PCR reproducibility could be limited when the number of available templates is low. Increased concentrations of Mg²⁺ reduced PCR variation possibly by allowing a higher proportion of annealed primers extended by the more active polymerase 29.

Our data corroborated the observation that amplicon size is crucial in designing assays to analyze gene expression levels not only using FFPE extracts but also using RNAs with high integrity 1215. Many researchers have found that short amplicons generated lower C_Ts than longer amplicons on analysis of the same gene in FFPE 3031. Data generated in this study evaluating snap frozen samples using two sizes of GAPDH (Figure 3) demonstrated GAPDH-67 generated a reduction in C_T by 2 – 3 cycles over GAPDH-122 in both GEMM and PreAmp experiments, demonstrating that smaller amplicons give more consistent results 32.

In addition, we found C_Ts between FFPE and snap frozen were closer for small amplicons than that for large amplicons (Figure 7). For example, the analysis of GAPDH using a target amplicon of 67 bp displayed C_Ts 1 – 2.5 cycles closer between FFPE and snap frozen than the same experiment using the longer amplicon size of 122 bp. We further measured the amplification efficiency associated with these two sizes of GAPDH assays using a broad dilution range (5 Log₁₀s) which showed closed efficiencies with 99.98% for GAPDH-67 and 99.92% for GAPDH-122 (data not shown). This measurement eliminates non-specific amplification that could contribute to decreased amplification efficiency of the true target. Therefore, it seems reasonable to conclude that the shift of efficiency detected from frozen to fixed material (Figure 7) is due to degradation of RNA in FFPE, and logical to extrapolate that the longer an amplicon is, the more likely its template will be degraded in extracted RNA. As a general rule, house-keeping genes or normalising assays should have amplicon sizes that match the size of the target whose expression is to be measured 3 and amplicons less than 100 bp should be employed in gene expression studies using FFPE materials.

Nthy-ori 3-1 (ECACC, Wiltshire, UK) is a normal thyroid follicular epithelial cell line derived from adult thyroid tissue that has been transfected with a plasmid encoding for the SV40 large T gene 33.

This cell line was grown to confluence in a humidified atmosphere containing 5% CO₂ at 37°C in the following plating medium: RPMI 1640 with 2 mM L-glutamine, 10% Foetal calf serum (FCS), Penicillin (100 U/ml) and Streptomycin (100 μg/ml). Tripsinized cells were counted with a hemocytometer. Approximately 1 × 10⁵ suspended cells were aliquot and were pelleted (a) snap frozen and (b) formalin fixed and paraffin embedded into a cell block. When formalin fixation was required, a cohesive solid cell pellet was constructed using 20% agar 23. The cells were centrifuged in an eppendorf tube, and the supernatant was removed using a pipette. Approximately 30 μl of pre-warmed agar (60°C) was added to each tube. The solid cell pellet was formed within a few seconds. Cell blocks were placed in 10% buffered formalin at room temperature for 5 hours followed by tissue processing on a Tissue-Tek^® V.I.P.TM tissue processor for 8 hours comprising: 10% buffered formalin fixation (4 hours at 37°C), 60% ethanol (20 minutes at 37°C), 80% ethanol (20 minutes at 37°C), 100% ethanol (20 minutes at 37°C), xylene (40 minutes at 37°C) and paraffin (80 minutes at 60°C). The pellets were subsequently paraffin embedded.

RNA extraction was performed using 4 protocols (Table 2): Ambion RecoverAll™ Total Nucleic Acid Isolation Kit, Stratagene Absolutely RNA^® FFPE Kit, Gentra Purescript^® RNA Purification Kit, and Invitrogen Trizol^® Reagent, in addition to modified Stratagene and Ambion protocols. Apart from deparaffinazition, RNA was extracted from snap frozen and FFPE cells in parallel according to manufacturer's protocols including proteinase K digestion (not included in Trizol^® protocol), RNA isolation and elution or hydration procedures. The modification in Stratagene and Ambion protocols involved incubation at 70°C for 20 minutes after the recommended proteinase K digestion.

RNA quantity was assessed using the NanoDrop^® ND-1000 Spectrophotometer (Wilmington, USA), and the quality was measured using the RNA 6000 Pico LabChip^® Kit on an Agilent 2100 Bioanalyser (Agilent technologies, Waldbronn, Germany).

Seven TaqMan^® Gene Expression Assays (P/N: 4331182, Applied Biosystems, CA, USA) and one custom designed assay (GAPDH-67) were utilised in this study with a range of amplicon sizes from 62 to 164 (Table 3). The extracted RNA was reverse transcribed into cDNA and were then quantified using TaqMan^® real time PCR with or without PreAmp procedure.

Replicates were omitted if C_T standard deviation was greater than 1.5 in the triplicate. All the data were collected in Excel form. The formulas used to generate the figures are as below:

1. ΔC_T (A-B) = C_T X_mean(A) - C_T X_mean(B)

2. Theoretical ΔC_T = Log₂(Amount cDNA in A/Amount of cDNA in B)