Expression profiling with RNA from formalin-fixed, paraffin-embedded material

Background Molecular characterization of breast and other cancers by gene expression profiling has corroborated existing classifications and revealed novel subtypes. Most profiling studies are based on fresh frozen (FF) tumor material which is available only for a limited number of samples while thousands of tumor samples exist as formalin-fixed, paraffin-embedded (FFPE) blocks. Unfortunately, RNA derived of FFPE material is fragmented and chemically modified impairing expression measurements by standard procedures. Robust protocols for isolation of RNA from FFPE material suitable for stable and reproducible measurement of gene expression (e.g. by quantitative reverse transcriptase PCR, QPCR) remain a major challenge. Results We present a simple procedure for RNA isolation from FFPE material of diagnostic samples. The RNA is suitable for expression measurement by QPCR when used in combination with an optimized cDNA synthesis protocol and TaqMan assays specific for short amplicons. The FFPE derived RNA was compared to intact RNA isolated from the same tumors. Preliminary scores were computed from genes related to the ER response, HER2 signaling and proliferation. Correlation coefficients between intact and partially fragmented RNA from FFPE material were 0.83 to 0.97. Conclusion We developed a simple and robust method for isolating RNA from FFPE material. The RNA can be used for gene expression profiling. Expression measurements from several genes can be combined to robust scores representing the hormonal or the proliferation status of the tumor.


Background
Breast cancer has been widely studied in the past and molecular characterization has increased the understanding of biological pathways that are altered during neoplastic transformation of cells [1][2][3][4]. However, the findings based on molecular profiling have not yet altered diagnosis, and decisions about treatment still rely mostly on histopathological and immunohistochemical techniques which are at best semi-quantitative [5,6]. Currently, many patients with primary, non-metastatic breast cancer with positive estrogen receptor (ER) status undergo several cycles of chemotherapy, although a substantial proportion of them does not benefit from it. Presently, no conventional parameters exist for many patients which allow to identify individuals who will benefit from chemotherapy. Personalized diagnosis on the basis of highly specific molecular analyses has the potential to improve the situation of many patients by optimizing medication, and at the same time, sparing others from unnecessary treatment regimens.
DNA chip studies are based on measuring gene expression for many genes in parallel [1,4,7,8]. Most protocols for gene expression analysis on the basis of DNA chips are sensitive to RNA degradation and RNA must be isolated from freshly prepared or FF tumor material. As a consequence, material is fairly limited and often originates from convenience samples of heterogeneous patients. Many of these studies including meta-analyses have revealed genes and biological functions of their products which are relevant for classification and prognosis [9,10]. However, many samples were derived from patients who did not participate in clinical studies and their treatment regimens were not standardized. Therefore, follow up data must still be interpreted with caution.
Obviously, procedures based on formalin-fixed, paraffinembedded (FFPE) material would greatly facilitate and speed up research in this area as large amounts of highly valuable material and clinical data have already been collected. In many cases, FFPE blocks are still available and they could be used for a molecular analysis. Especially material from clinical trials would allow investigating distinct clinical questions with existing material rather than material from newly designed studies.
Many efforts are currently made to individualize diagnosis of breast cancer by including molecular parameters into diagnosis. Fresh frozen material would obviously be ideal for a molecular analysis by gene expression measurements but it may be difficult to implement novel procedures which complicate current workflows of daily routine. Procedures based on FFPE material would be more feasible as they do not interfere with current protocols and they do not affect routine diagnosis as material for molecular analysis could be collected after standard diagnosis has been terminated. Only relatively few molecular approaches have been described which are based on FFPE material. For example, Paik and co-workers have established a recurrence score (RS, Oncotype DX), it allows to quantify the likelihood of distant recurrence and to predict the magnitude of chemotherapy benefit [11,12].
It is generally accepted that molecular profiles which reflect primarily biological characteristics of tumor cells, may complement clinical and histopathological diagnosis, resulting in a more detailed characterization of individual tumors, a pre-requisite for better treatment decisions. In this study we present the development of a novel procedure for RNA isolation from FFPE material and an optimized workflow for expression measurements by QPCR.

Human breast cancer samples
Human breast cancer specimens were divided into two aliquots, one of which was processed for histological diagnosis by fixation with formalin and embedding in paraffin. FFPE material was obtained from the Institute of Pathology (University of Bern) and the Pathology Länggasse, Bern. Tissue (3-5 mm thick slices of tumor) was fixed over night in buffered formalin and processed for paraffin embedding in a Tissue Processing Center TPC 15 (Medite Medizintechnik, Germany). The second aliquot was frozen on dry ice and stored at -80°C. Fresh frozen material was obtained from the Tumorbank Bern. Both, FF and FFPE samples were checked by hematoxylin and eosin staining and only samples with more than 50% tumor cells were used for this study. An informed consent to use the material for research was obtained from all the patients.

RNA Extraction
Intact RNA was isolated from four 25 μm thick kryo-sections of approximately 0.5 cm 2 . The tissue was homogenized in 420 μl lysis buffer (4 M guanidinium thiocyanate, 30 mM Tris pH 8.0, 1% Triton-X-100), 8.0,1 using a TissueLyser (Mixer Mill, Retsch GmbH, Haan, Germany) at 15 Hz for 3 min. Total RNA was bound to silicabased columns (Epoch Biolabs, Huston Texas), treated with DNase I (30 Kunitz units for 20 min. at room temperature; Qiagen, Hilden, Germany), washed once with lysis buffer (containing 30% ethanol) and once with 20 mM NaCl (containing 20% ethanol) and eluted in 50 μl 10 mM Tris pH 7.4, 0.1 mM EDTA and stored at -20°C. RNA quantity was measured on an ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and quality assessed by capillary electrophoresis with an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA) using Agilent RNA 6000 Series Nano kits. RNA was isolated from ten 10 μm thick FFPE sections according to the RNeasy FFPE protocol of Qiagen (Fig. 1, lanes B), the ncLysis protocol of Applied Biosystems (lanes C) and the protocol developed in our laboratory (lanes D). Paraffin sections were de-paraffinized with xylene, washed with ethanol and dried in a speed vac. For our own protocol, 200 μl lysis buffer (4 M guanidinium thiocyanate, 30 mM Tris, pH 8.0, 1% Triton-X-100) was added to the dried sections and immediately homogenized in a Mixer Mill at 20 Hz for 4 min. Proteinase K (Roche Diagnostics, Mannheim, Germany) was added (1 mg/ml final concentration) and tissue was digested for 1 hour at 55°C. One milliliter dilution buffer (30 mM Tris, pH 8.0, 1% Triton-X-100) was added to each lysate and digestion continued for 1 hr after adding fresh proteinase K (final concentration 1 mg/ml). RNA was de-modified by adding 318 μl of de-modification solution (5 M NH 4 Cl) and incubating at 94°C for 20 min or as described in the text. RNA was bound to silica-based columns and digested with DNase I as described for fresh-frozen tissue samples. The reproducibility of our own procedure was tested by isolating several independent RNAs from consecutive sections of the same tissue block. About 10 μg of total RNA could be isolated from 5 to 10 FFPE sections (0.5-1 cm 2 / section). RNA was isolated from closely matched sections using the RNeasy FFPE kit (Qiagen) or the ncLysis system (Applied Biosystems) according to the protocols included with the kits. In both cases, the RNA was purified on silicabased columns. 22 samples were available. In 14 cases sufficient RNA was obtained from all 4 parallel isolations. In 2 cases of FF material (samples 4 and 11) and in 6 cases of FFPE material (samples 1, 5, 7, 9, 12 and 21) less than 1.5 μg RNA could be isolated with the ncLysis protocol. These samples were excluded from further analysis.

cDNA synthesis and QPCR
Aliquots of 100 to 500 ng of total RNA were reverse transcribed using MultiScribe™ MuLV reverse transcriptase (High-Capacity cDNA Archive Kit; Applied Biosystems, Foster City, CA, USA) and random or gene-specific primers. Reverse primers were kindly provided by Applied Biosystems, they were used at 1 μM each, cDNAs were made in the presence of 3, 10 or 22 reverse primers as 3-plex, 10plex or 22-plex, respectively. Regular Assays on Demand (Applied Biosystems) were used for QPCR (Table 1). Manually designed assays coding for short, medium-size and long amplicons of the insulin growth factor-binding protein 5 (IGBP5) were selected with Primer Express (Version 3, Applied Biosystems). Forward primer and probe were kept constant for all assays while reverse primers were selected such that amplicons of different sizes were generated [13]. QPCR reactions were carried out in triplicates in a final volume of 10 μl in 1× FAST Master mix (Applied Biosystems) and cDNA corresponding to 4 ng total RNA. QPCR was performed on an ABI 7500 FAST instrument (2 min at 95°C, followed by 45 cycles of 95°C for 3 sec and 60°C for 30 sec). The quality of the assays and the absence of contaminating DNA were assessed with water and RNA instead of cDNA, respectively (data not shown). Three positive controls containing cDNA derived of ZR-7-51 cells were included on each 96-well plate. Cycle threshold values (Ct) were determined using the SDS software of the 7500 FAST System (Version 1.3.1). Constant threshold values were set for each gene throughout the study. Figure 1 RNA isolation and characterization. Total RNA was isolated from kryo-sections (lanes A) and from paraffin sections according to the RNeasy FFPE protocol of Qiagen (lanes B), the ncLysis protocol of Applied Biosystems (lanes C) or according to our own protocol (lanes D). Aliquots of each RNA were separated by capillary electrophoresis (Agilent Bioanalyzer) on Nano chips along with RNA ladder (L; Ambion). Shown are RNAs from two representative tumors (Tu#10 and #18).

Data processing and determination of breast cancer classification scores
All the measured cycle threshold (Ct) values represent log 2 expression levels. These values need to be normalized such that they are comparable across samples and suitable for generating scores. For a gene, a large Ct value corresponds to a low expression level, so the first processing step needed was to reverse the sense of this relationship by letting Ct' = max(cut_off -Ct, 0) be the new value for each measured gene. The cut off value was set empirically to 35.0 as any higher raw Ct value was deemed unreliable. This cut off was fixed a priori and kept constant throughout all the experiments reported here. Then, the final value of each target gene was taken to be where R represents the reference value and was taken as the mean of Ct' values of 5 selected reference genes (GAPDH, GUSB, RPLP0, TFRC, UBB, see Results section for details). The approach guarantees that all ΔCt values are positive and upper bounded by max_val (set to 33 for all the results reported here).
We used the scores associated with three of the gene groups listed in Table 1: the ER, HER2 and Proliferation group. While for the HER2 and Proliferation groups the scores were taken as the average ΔCt value of the genes in the group, for the ER group more weight was given to the ESR1 gene: where the gene symbols stand for the corresponding ΔCt values.
Finally, for each tumor a Total score was computed as Total_score = (Proliferation_score + HER2_score -ER_score + max_val)/3 The Total score, together with the group scores as computed above, are used in all subsequent discussions.

Isolation of RNA from FFPE material
Total RNA was isolated from FF human breast cancer specimen which resulted in intact RNA in all samples (Fig. 1, lanes A, shown are RNAs from two representative tumors from a series of 14 tumors). RNA from FF tissue was used as reference for partially fragmented RNA isolated from FFPE material of the same tumors. RNA was assessed by capillary electrophoresis. The size distribution of RNA isolated according to our own protocol was in the range of 200 to 1000 nucleotides (Fig. 1, panel D) while the majority of RNA fragments was in the range of 100 nucleotides when RNA was isolated according to RNeasy FFPE (panel B) or the ncLysis system (panel C). Gene expression was measured by QPCR using 25 commercially available and three own TaqMan assays [13] (Tab. 1). The cycle threshold values (Ct values) were determined from RNAs isolated according to one of the three protocols for FFPE material and compared to Cts obtained with intact RNA of the same tumors. Fig. 2 shows correlation coefficients between intact RNA (A) and FFPE-derived RNAs isolated according to the RNeasy FFPE protocol, (A vs B); the ncLysis system (A vs C); or our own protocol (A vs D) for all 14 tumors using the expression levels of 5 genes (GAPDH, GUSB, RPLP0, TFRC, UBB; see below). The cDNAs were made in the presence of random (white boxes) or genespecific primers (gray boxes). Clearly, correlation coefficients between intact and partially fragmented RNA were higher with gene-specific primers than random primers and RNA isolated according to our own protocol resulted in cDNA which performed better in QPCR than cDNA made from RNA isolated according to RNeasy FFPE and ncLysis protocols.

Parameters affecting the RNA quality and QPCR
Several parameters were systematically optimized to improve the protocol for RNA isolation from FFPEderived sections. For example, QPCR made in the presence of primers specific for large amplicons (Fig. 3, dashed line) is very sensitive to RNA fragmentation and modification resulting in higher Ct values than primers specific for medium-size amplicons (dotted line) or short amplicons (non-interrupted line). In addition, the effect of de-modification of FFPE-derived RNA is apparent: the Ct determined from de-modified RNA is 3 or more units lower than the Ct measured from the same RNA but without demodification. The effect was consistently observed with several tumors and also when expression was measured Comparison of RNAs isolated according to different protocols with TaqMan assays from Applied Biosystems (data not shown). The optimum time of demodification was 20 min, longer times led to higher Ct values (not shown).
The different protocols of RNA isolation from FFPE material were further compared by measuring expression levels of reference genes in the 14 tumors and by comparing the results to Cts generated from corresponding intact RNAs (Fig. 4)

Normalization
Results generated in the presence of partially fragmented RNA cannot be directly aligned with results produced from intact RNA and a suitable normalization is required to eliminate or reduce the effects of fragmentation and residual modification in RNA from FFPE material. Nine putative reference genes were selected from the literature [14] and from microarray results [15]. Expression was measured from intact and FFPE-derived RNA and raw Cts from all the 14 tumors are plotted for each putative reference gene (Fig. 6). Analyses based on intact RNAs revealed that 8 of the 9 tested genes performed similarly well (panel A). RPS23 which was hardly measurable (mean Ct in intact RNA > 37) was characterized by a large variation between the different tumors. A slightly higher variation was observed when expression levels were compared for FFPE-derived RNAs (panel B): GAPDH, GUSB, RPLP0, TFRC, RPL7A and UBB showed a similar performance and small variations between the 14 tumors as was seen with intact RNA. In contrast, the Ct values with RNA from FFPE material revealed larger variations for ACTB and RPS11 and therefore, the two genes were excluded as reference genes. The ACTB and RPS11 amplicons are larger than amplicons for the other reference genes and also for the test genes (Tab. 1, see also Fig. 3). Five genes were used as reference genes: GAPDH, GUSB, RPLP0, TFRC and UBB. For comparison, raw Ct values are shown for 4 genes related to the ER response (BCL2, CEPG1, ESR1, PGR) (Fig. 6, left). As expected, a high variation was observed for these genes between the 14 tumors. Protocols B and C did not yield enough usable data, precluding the data from further analysis. For example, protocol B did not have data for all the reference genes and for protocol C several test genes could not be measured reliably (e.g. BCL2, PGR of the ER group).

Comparison of RNA isolation methods
RNAs isolated from FFPE material according to our own protocol were also compared to RNA derived of kryo-preserved material of the same tumors in a different way. The arithmetic mean of the five reference genes (GAPDH, GUSB, RPLP0, TFRC and UBB) was used for normalizing expression values of all the genes in each RNA. Normalized expression values were compared between intact and FFPE-derived RNA for each gene and each tumor [see Additional File 1]. Good conservation of inter-tumor differences were observed between kryo-preserved and FFPE samples for most genes.

Module scores
Normalized expression values were also used to compute scores representing ER-related genes (ESR1, PGR, BCL2, CEPG1), HER2-related genes (HER2 and GRB7), genes related to proliferation (STK15/AURKA, CCNB1, MYBL2, MKI67, BIRC5/SURV) and a Total score representing all the genes of the three scores (for details see Methods). The computation of biologically meaningful scores with multiple genes instead of relying on just one has the scope to reduce noise variation. Module scores and Total scores were computed separately from normalized expression values of intact RNAs (circles) and of RNAs isolated according to our own protocol (triangles) and Total scores are depicted separately for each tumor (Fig. 7). The figure demonstrates that similar values are obtained for each tumor irrespective of whether they are computed from intact RNA or from RNA derived of FFPE material. This suggests that scores can be computed with RNA from FF samples as well as with RNA from FFPE samples. ER and HER2 scores were visualized in scatter plots, where the ER and HER2 scores were represented on the x-and y-axis, respectively ( Fig. 8A and 8B). It was apparent that the three immunohistochemically ER-negative tumors have low ER scores (#15, #18, #20) and the only immunohistochemically HER2 positive tumor (#6) among the 14 tested tumors had a high HER2 score and an intermediate ER score (see also Table 2). The remaining tumors were all ER positive as assessed by immunohistochemistry (IHC) and they had relatively high ER scores. ER-negative and HER2-positive tumors all had high Proliferation scores (visualized by the red color of the dots). A larger spectrum of Proliferation scores (from blue to red) was found for ER positive tumors. Similar distributions were found when Reproducibility of RNA isolation from FFPE material Figure 5 Reproducibility of RNA isolation from FFPE material. The RNAs were isolated from paraffin blocks according to our own protocol. BM33 and BM36 (panel A) are two separate RNAs isolated from tissue block "BM", D33 and D36 are RNAs isolated from block "D" (panel B). For comparison, 45T and 56T originate from two distinct tumors isolated from one patient (panel C). Gene expression was measured by QPCR for 24 genes and raw Ct values are shown for each gene measured from the two matching RNAs.
Stability of reference gene expression in RNA isolated from FF and FFPE material
RNA was quantified by dot blot hybridization [23], semiquantitative PCR [19] and more recently, by QPCR [24,18,26,13,17,33,32] and other methods [28][29][30]. RNA derived of FFPE material is not only partially hydrolyzed but also chemically modified: formalin reacts with nucleotides leading to the formation of methylol groups in nucleobases. These groups tend to further react and form intra-and inter-molecular methylene bridges in RNA, DNA [34,35,31] and protein [36]. As a result, reverse transcription is impaired and threshold cycle values (Ct values) increase during subsequent QPCR.
The protocol for RNA isolation described here was complemented by adding a separate demodification step which involves incubation at elevated temperature in a buffer containing ammonium chloride which favors the reversion of methylol groups to amino groups in nucleobases. It does not only improve the efficiency of downstream applications (mainly reverse transcription), it also improves the recovery of RNA from FFPE sections. RNA yield and quality can be further improved by extensive digestion of FFPE material with protease in a buffer containing guanidinium thiocyanate. Reverse transcription in the presence of gene-specific primers prevents the initiation of cDNA synthesis inside amplicons and therefore, cDNA made in the presence of gene-specific primers is a better template for QPCR than cDNA made from random primers (Fig. 2). Several papers have demonstrated that QPCR with primers coding for short amplicons are more efficient than primers coding for long amplicons [17,20,24,13,32].
Finally, normalization of raw data is used to eliminate or at least reduce the effect of poorer quality of starting RNA. Various approaches of normalization were proposed in the literature [37,14,38,32]. They are based on calculating relative expression values: expression levels of genes of interest are expressed relative to the expression of one or a panel of several suitable reference genes. An ideal refer-  Module scores Figure 8 Module scores. ER, HER2 and proliferation scores were computed from expression values of 14 breast cancers and visualized in a scatter plot. The ER score was determined from four genes, the HER2 score from 2 and the proliferation score from 5 genes (see Methods). Tumors are positioned according to their ER score (x-axis) and HER2 score (y-axis). Proliferation scores are color coded. The histological ER status is indicated by a "-" or "+" sign next to the tumor numbers in the plot. The results were computed from intact RNA derived of FF material (A) and RNA isolated from FFPE material according to our own protocol (B). Individual scores for each tumor are given in Table 2.
ence gene has a stable expression level in all the samples under investigation. As such "ideal" reference gene normally does not exist, the mean or median expression level of several suitably chosen reference genes is used as a relatively stable reference Ct value. We used a formalized approach to characterize all candidate reference genes. Candidate reference genes were ranked according to their standard deviations of raw Ct values in RNA from FF and FFPE material. The final rank of each candidate reference gene was taken as the mean of the two ranks obtained with RNA from intact and FFPE material. Genes with higher ranks were excluded as reference genes.
We also applied GeNorm [14] to characterize candidate reference genes: ACTB and RPS11 had poorest stability measure M [14] for FFPE-derived RNA and RPL7A had a poor stability measure when RNA from FF material was tested (data not shown). For these reasons GAPDH, GUSB, RPLP0, TFRC and UBB were used as reference genes in this study.
Our own RNA isolation protocol was compared to RNA that was isolated from the same material but according to commercial protocols and products (Qiagen RNeasy FFPE and ncLysis system of Applied Biosystems). Additional products for FFPE material from commercial providers (e.g. Stratagene, Ambion) were tested and the results obtained with our own protocol were superior to all tested commercial products (data not shown).
We determined module scores for each of the 14 tumors in this study. The limited number of samples does not allow statements about the clinical significance of module scores but they can be used to compare scores computed from intact RNA from FF material and RNA isolated from FFPE according to our own protocol. Pearson correlations between these RNAs in the 14 tumors were 0.966, 0.856 and 0.833 for ER, HER2 and Proliferation scores, respectively. As kryo-preserved RNA and RNA from FFPE material always originated from different portions of the same tumor, a certain variation of gene expression cannot be excluded and, as a consequence, part of the observed variability between kryo and FFPE material may be attributed to biological heterogeneity in the tumors. The three module scores were combined to a Total score. The Total score is similar to the recurrence score described by Paik [11], with high expression of genes related to proliferation and HER2 and low expression of ER-related genes indicating higher risk.
The data generated from FF and FFPE material were also compared to ER and HER2 levels assessed by IHC results from the same tumors. Three tumors (#15, #18, #20) were ER-negative and one was strongly HER2-positive (#6) (Tab. 2). The same tumors had low ER scores when assessed by QPCR (Fig. 8). Tumor #6 had a high HER2 score and an intermediate ER score. These results are in good agreement with the expected distribution of the three scores [15,39]. By comparing QPCR based data with well known tumor subtypes allowed to validate the protocols developed here, even if no new biological findings are provided. The primary issue of this work was to document that stable and robust expression values can be determined from FFPE-derived RNA which are close to the values computed from intact RNA of the same tumors. The optimization and validation of the scoring procedure remains an important issue but obviously, the available number of samples is not sufficient to deal with this aspect and it will be addressed separately and on a larger collection of samples.
While IHC results are at most semi-quantitative, QPCRbased results reflect more accurately the expression level of genes in question. The module scores proposed here integrate quantitative gene expression data from several genes, this makes the resulting scores more robust than measurements based on single genes. QPCR is not only quantitative, it is also very sensitive over a large dynamic range. The number of genes which can be measured by QPCR is not limited and additional genes and module scores can be included in the analysis if this will be required.
Importantly, certain predictive parameters still cannot be determined with current technologies. For example, breast cancers are classified into histological grade 1, 2 or 3. This grading most likely reflects the proliferative state of tumor cells [40]. Grading may be especially important as high grade tumors seem to respond more favorably to chemotherapy than low grade tumors. Unfortunately, many tumors are histological grade 2 and for those tumors the benefit is not clear. Paik and co-workers documented that their recurrence score (RS) was also predictive for a response to chemotherapy [41]. The RS defined by Paik and coworkers is composed of 16 test genes mainly representing ER response genes, proliferationassociated genes, HER2-related genes and invasion genes and 5 genes for normalization [11,41].
The genes selected for this study (Tab. 1) were selected from published DNA chip studies with breast cancer samples [15]. They mostly coincide with the genes used by Paik ad co-workers.

Conclusion
The results presented in this study reveal that RNA isolated from FFPE material according to the protocol developed in our laboratory can be used for expression measurements by QPCR although the RNA is partially degraded. The optimized isolation and de-modification procedures combined with a normalization procedure results in stable and robust gene expression data. Robustness of results was further increased by computing scores from several genes representing the hormonal and the proliferation status of the tumor. Molecular profiling from FFPE material may be of interest for routine diagnostics in the near future as FFPE material is always available [42]. Similarly, molecular profiling from FFPE material may be of great interest in the context of existing and newly planed clinical trials for which only formalin-fixed samples exist.