For decades, human tissue samples obtained from surgical procedures have been routinely fixed in formalin and embedded in paraffin for long-term storage. Consequently, most institutions have large paraffin-block archives that allow long-term follow up for all types of neoplasms, including rare tumors. These archives represent a valuable source for molecular studies. A major limitation for the use of such samples to study gene expression is three-fold: RNA can degrade prior to formalin fixation; formalin fixation produces significant chemical modification of RNA; and RNA continues to fragment and degrade over time, even after dehydration and paraffin-embedding.

Immediate freezing of fresh tissue samples preserves good quality RNA for gene expression studies; however, this procedure is not routinely performed in most hospitals and there are few institutions worldwide that have large frozen-tissue banks associated with long term clinical followup. Moreover, tumor samples accrued into a frozen tissue bank are inherently biased in their collection: such tumors must be sufficiently large and palpable in order for tissue to be excised and frozen for the bank 1.

Formalin-fixed paraffin-embedded (FFPE) samples are universally available and, with Institutional Review Board/Ethics Committee approval, cancer samples can be linked to clinical and/or follow up information, often available through an institution's tumor registry. Extracting RNA from such samples has been a promising but problematic challenge. In addition to RNA degradation concerns, formalin fixation creates cross-linking between nucleic acids and proteins and adds mono-methylol to amino groups for all four RNA bases (N-CH₂OH); methylene bridging then occurs between neighboring bases (N-CH₂-N), formed by condensation of amino bases and N-methylols 2.

Rupp and Locker described the first successful RNA extraction from FFPE samples in 1988 3. Since then, many protocols have been described 24567891011. There are currently multiple protocols commercially available. In most of these protocols the RNA is extracted by spin column purification according to similar basic principles: deparaffinization, followed by cell disruption with heated proteinase K, which is capable of efficiently degrading proteins that were covalently cross-linked with each other and RNA, thereby allowing more efficient RNA extraction than achieved by use of chaotropic agents such as guanidinium chloride 7. Proteinase K incubation at high temperature (60 to 70°C) also removes part of the methylol additions induced by formalin fixation 7. After proteinase K incubation, RNA is isolated by alcohol precipitation and though use of salt guanidine thiocyanate in a spin column purification step 212.

Magnetic beads can also be utilized to isolate RNA. In these protocols, nucleic acids are bonded to paramagnetic beads and isolated by a magnetic field. RNA is isolated using an optimized buffer solution and later a DNAse step. Magnetic micro-beads can be prepared in a number of ways, but usually magnetically susceptible particles (e.g. iron oxide) are coated with synthetic or biological polymers with high affinity for nucleic acid. A magnet is applied to the side of the tube containing the magnetically-labeled sample mixture and the magnetically bound nucleic acid aggregates near the wall of the tube, allowing the unlabeled materials to be pipetted off and discarded. In addition to quickly separating magnetically-bound nucleic acid, this method does not create the shear forces generated by spin centrifugation that may lead to nucleic acid degradation 13.

Breast cancer tissue RNA archived for one year is less degraded and has larger average molecular weight than RNA archived for six and 17 years, suggesting that fragmentation of FFPE tissue continues even after specimens are dehydrated and embedded in paraffin 9. Many molecular techniques depend upon the quantity and quality of extracted RNA in samples; translational studies that link molecular findings to clinical outcome require long term follow up. Here we compare RNA extraction protocols on a set of 14 ten year old FFPE breast cancer samples, testing spin column purification and magnetic bead-based technologies. Using the same protocols, we compare our results with a second set of 14 recently archived breast cancer samples fixed and embedded in paraffin three to ten months prior to this study, that we designate "months old" samples. We also test RNA fragment size in both sample groups using RT-PCR for the housekeeping gene G6PD.

Table 1 exhibits 168 RNA extraction experiments performed on 14 ten year old archived FFPE breast cancer samples, showing RNA quantity (in pg/μl) and quality (by RNA integrity number, RIN) for RNA that was extracted in triplicate using four different extraction protocols. Similarly, Table 2 shows 168 RNA extraction experiments on 14 FFPE breast cancer samples that are months old, archived three to ten months prior to analysis, again with RNA extracted in triplicate using each of the four protocols. Tables 3 and 4 show the minimum and maximum values, means, and medians for each of the protocols concerning the quantity and the quality of the extracted RNA, in the ten year old and months old set of samples, respectively. Tables 5 and 6 show the same parameters, but only for samples that had biologically useful RNA extracted, defined as RNA with a minimum RIN of 1.4.

In the set of ten year old samples (Tables 1 and 5), protocols 1, 2, 3, and 4 extracted RNA with a minimum RIN of 1.4 in, respectively, 26, 26, 35, and 30 of the 42 triplicate RNA extraction procedures performed for each protocol. Thus, the success rate for extracting biologically useful RNA with protocols 1, 2, 3, and 4 was 62%, 62%, 83%, and 71%, respectively. When analyzed on a per sample basis, protocols 1–4 produced biologically useful RNA in at least one of three extractions for, respectively, 12, 14, 13, and 13 out of 14 cases in the set for a case success rate of, respectively, 86%, 100%, 93%, and 93%.

For the set of recently archived months old samples (Tables 2 and 6), protocols 1, 2, 3, and 4 extracted RNA with a minimum RIN of 1.4 in, respectively, 34, 34, 39, and 35 of the 42 RNA extraction procedures performed for each protocol, giving a success rate for extracting biologically useful RNA of 81%, 81%, 93%, and 83%, respectively. The case success rate was 100% for all four protocols; in other words, each protocol was able to extract RNA with a RIN ≥ 1.4 in at least one of the three attempts for every case in this set.

RT-PCR for the housekeeping gene glucose-6-phosphate dehydrogenase (6GPD) was performed using primers that targeted different fragment sizes of extracted RNA. All samples from both one and ten year old sets successfully amplified 67 bp and 151 bp fragments of G6PD mRNA. However, when primers targeting 242 bp fragments were tested, only six of 14 (43%) months old FFPE samples and none of the ten year old samples demonstrated G6PD gene expression (Figure 1).

Biopsies and surgical specimens are routinely fixed in formalin and embedded in paraffin for histological analysis. Although the morphological integrity of the tissues is preserved, this procedure causes degradation of nucleic acids, predominantly RNA. The development of protocols for RNA extraction from these samples makes it possible to molecularly study long-term archived tissues, opening new perspectives for research and diagnostic investigation. Soguero et al. were able to extract RNA in more than 20 year-old archival liver tissue 14. In breast cancer, long-term follow-up is important because the time to developing distant metastases and death can vary greatly, with both recurrence time and death extending beyond 15 years after surgery 151617. Even patients with small tumors (stage T1N0M0) may have 10-year relapse-free survival rates of less than 75% in the absence of systemic therapy 18.

In addition to the impact of long-term storage, RNA can also be degraded during the relative delay associated with putting the specimen in fixative following surgical excision and by prolonged fixation prior to paraffin embedding 19. However, the greater challenge is that RNA is chemically modified during formalin-fixation. Fortunately, some of the alterations can be reversed by heating the RNA to 50–55°C 2. Many attempts have been made to avoid these chemical modifications by proposing use of alternative fixatives such as Bouin, Carnoy, acetone, and alcohol as substitutes for formalin 51020. These fixatives, however, introduce tissue artifacts that can make microscopic histopathological analysis difficult or even impossible. Moreover, these fixatives are not suitable for immunohistochemistry reactions, much more common in pathologic practice than gene expression analyses. Finally, samples in most archives are already fixed in formalin. So, for retrospective studies, it is imperative to develop and/or refine protocols for RNA extraction from FFPE.

The first step in RNA extraction from FFPE samples is deparaffinization of the sections obtained from the paraffin blocks. When deparaffinization is incomplete, the extracted RNA quality is worse 21. In our work, in the set of 10-year-old samples, both the quantity and quality of RNA extracted with the deparaffinization solution d-limonene (protocols 2 and 3) were higher than that obtained using xylene – also known as xylol, dimethylbenzene, or methyltoluene – (protocol 1) (Tables 3 and 5). Due to other protocol differences, we cannot assess the impact of the deparaffinizing agent alone. It is worthy to note, however, that with the set of recently archived months old samples, the differences between protocols is smaller (Tables 4 and 6).

Historically, RNA integrity has been evaluated using agarose gel electrophoresis stained with ethidium bromide. Typically these gels show two peaks that correspond to ribosomal RNA species 28S and 18S. According to this technique, a given RNA sample is considered of good quality when the relation 28S:18S is equal to or higher than 2 22. The 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) is a microfluidic platform that uses electrophoretic separation of RNA by molecular weight and provides laser-induced fluorescence measurements. Using electropherogram curves, 1208 RNA samples from different sources and in different degrees of degradation were analyzed and an algorithm was created to determine RNA integrity number (RIN). The RIN for a given sample ranges from 1 to 10, from totally degraded RNA to co