Specimen Quality Evaluation in Canadian Biobanks Participating in the COEUR Repository

Abstract

Human biological specimens are important for translational research programs such as the Canadian Ovarian Experimental Unified Resource (COEUR) funded by the Terry Fox Research Institute. Sample quality is an important consideration, as it directly impacts the quality of ensuing research. The aim of the present study was to determine the quality of tissues collected from different sites contributing to the COEUR cohort. Samples from high-grade serous ovarian tumors (fresh frozen and corresponding paraffin-embedded tissues) were provided by nine participating Canadian biobanks. All samples were shipped to a central site using a Standard Operating Protocol (SOP). DNA and RNA extraction was conducted by the quality control division of the Canadian Tumor Repository Network (CTRNet). DNA quality was determined by ß-globin gene PCR amplification, and RNA quality by the RNA integrity number (RIN), as measured by the Agilent BioAnalyzer. DNA of acceptable quality had at least three bands of ß-globin amplified from DNA (n=115/135), and a RIN number ≥7 was considered very good for RNA (n=80/135). Sample preparation and storage time had little effect on RNA or DNA quality. Protein expression was assessed on tissue microarray by immunohistochemistry with antibodies against p53, WT1, E-cadherin, CK-7, and Ki67 from formalin fixed-paraffin embedded (FFPE) tissues. As seen with a nonhierarchical clustering statistical method, there was no significant difference in immunostaining of paraffin tissues among specimens from different biobanks. Interestingly, patients with worse outcome were highly positive for p53 and weak for WT1. In conclusion, while there was no common SOP for retrospectively collected material across Canadian biobanks, these results indicate that specimens collected at these multiple sites are of comparable quality, and can serve as an adequate resource to create a national cohort for the validation of molecular biomarkers in ovarian cancer.

Introduction

Ovarian cancer is the second most common gynecological cancer and the leading cause of death from gynecological malignancies. Early detection of ovarian cancer is rare and little is known about the natural history of disease progression. More recently, the notion that ovarian cancer is a single disease has given way to a more sophisticated concept of ovarian disease subsets that may be associated with very different molecular events. With the advent of new treatment modalities, including sequential administration of multiple rounds of active agents, ovarian cancer is now clinically considered to be a chronic disease.¹ A high proportion of patients respond well to the initial treatment and experience long progression-free survival, although the majority will develop recurrent ovarian cancer at varying intervals. Whereas platinum/taxane-based treatment is currently the gold standard in first-line therapy, failure of this treatment in a significant proportion of patients remains a serious problem. Identifying nonresponders and offering them alternative first-line treatments remains one of the most important challenges in the initial clinical management of the ovarian cancer patient, and the focus of future clinical trials. In fact, because ovarian cancer is a heterogeneous group of diseases, it is likely that different approaches to treatment are required.

The ovarian cancer community in Canada, which is vibrant and collegial, is interested in pursuing a large-scale biomarker program at a cooperative level. During the 4th Canadian Ovarian Cancer Research Conference (May 4th–7th, 2008; Montreal, QC), the community outlined the initial plan of a Canadian Ovarian Cancer Research Consortium (COCRC). “A Trans-Canadian Platform for the Development of Biomarker Driven Subtype Specific Management of Ovarian Carcinoma” was created in November 2008. In January 2009, members of the Canadian ovarian cancer community were invited to participate in the pan-Canadian Biomarker project funded by the Terry Fox Research Institute (TFRI, see www.tfri.ca/terryfox). Researchers were provided the executive summary of the proposal, as well as a work plan to address all facets of the proposed project. To become collaborators, COCRC members had to agree that all biospecimens would be openly shared within the COCRC, and the ultimate decision for access to the resource would be through a project management committee. This was central to the proposal, as biospecimens within the Canadian Ovarian Experimental United Resource (COEUR) would be exclusively reserved for the TFRI-funded program.

The Canadian consortium proposed to develop a pan-Canadian discovery and validation platform for biomarker research, and to use this biomarker platform to develop a molecular-pathology classification system for ovarian cancer integrated into clinical nomograms (decision making algorithms) for rational clinical management. Finally, the aim of the Canadian consortium is to integrate this correlative, preclinical initiative with prospective clinical trials in order to expedite the translation of research findings into clinical practice.

To ensure the success of the pan-Canadian ovarian cancer biomarker program, a series of actions was completed during the first year of the program that permitted it to be fully operational. Since the primary focus is on the correct pathological classification of specimens, coupled with clinical outcomes to identify samples belonging to both responders and nonresponders of first line therapy, an environmental scan of existing biobanks was conducted to identify samples to be reserved for the TFRI COEUR and to determine the means by which appropriate quality specimens and clinical data could be retrieved. Since the quality of research samples is critical to the success of the program, member biobanks were audited to ensure the quality of research materials using metrics relevant to anticipated future research assays. As a retrospective cohort of biobanked samples will be used, defining the level and variance in quality of existing materials is critical. As a first step, a test cohort of 135 high-grade serous ovarian carcinomas (HGSC) was assembled from samples obtained from each of the contributing biobanks. The quality assessment done in the preparatory phase of the repository collection is reported here.

Material and Methods

Participating biobanks

To date, the network consists of nine independent Canadian biobanks from four Canadian provinces that are committed to facilitate joint studies by working towards similar policies and quality assurance with the support of CTRNet (Canadian Tumor Repository Network, www.ctrnet.ca). These biobanks have all received ethics approval by their local ethics committees. General SOPs for all major biobanking activities are available at the CTRNet website. Additionally, the central laboratory work done for the COEUR study was approved by the Research Center review board of the Center Hospitalier de l'Université de Montréal (CHUM). Tumor samples were collected and banked, following appropriate written consent from patients undergoing surgery. Descriptions of participating biobanks are provided in Table 1.

Table 1.

Description of Biobanks Participating in the COEUR Program

Title	Location	Date of creation	Funding/support	Partner	Web address
Segal Cancer Center LDI biobank	McGill-University- Jewish General Hospital	2003	Fonds National de Recherche du Québec and the Montreal-Israel Cancer Research Foundation	OCC, CTRNet	www.rrcancer.ca
Banque de tissus et de données du Réseau de recherche sur le cancer du FRSQ	CHUM-Hopital Notre Dame	2000	Fonds de la recherche en santé du Québec	OCC, CTRNet	www.rrcancer.ca
Ottawa Ovarian Cancer Tissue Bank	Division of Gynecologic Oncology (Ottawa Hospital Research Institute)	1995	Ovarian Cancer Canada, Ottawa Hospital	OCC	N/A
OvCaRe Gynecologic Tissue Bank	BC Cancer Agency, Vancouver	2001	OvCaRe BC, Michael Smith Foundation for Health Research, and Ovarian Cancer Canada		ovcare.ca/research/platforms.php
Ovarian cancer specimen and data bank at CHUQ	Hopital Hotel-Dieu de Quebec, Quebec city	2000	Fonds de la recherche en santé du Québec	RR-Cancer provincial network and CTRNet	www.rrcancer.ca
Banque D'échantillons biologiques et de données de Sherbrooke	Faculté de médecine et des sciences de la santé, Université de Sherbrooke	1998	Fonds de la recherche en santé du Québec	CTRNet, OCC	www.rrcancer.ca
CBCF Tumor Bank	Alberta	2001	Canadian Breast Cancer Foundation	CTRNet, CBCF	www.abtumorbank.com/
BCCA Tumor Tissue Repository	BC Cancer, Agency, Vancouver Island Center, BC	2004	BC Cancer Foundation	CTRNet	www.bccrc.ca/dept/ttr
UHN Program in Biospecimen Sciences	PMH	2001	UHN Research Institutes (OCI, T&RI), UHN foundations	N/A	N/A

Specimen collection and storage

Specimens were collected by trained staff in each participating center. Resected specimens were immediately transported from the operating room to the pathology department, where a pathologist provided an appropriate specimen to the biobank staff for immediate processing. When possible, time of the resection in the operating room and time of tissue processing by the biobank staff were recorded. CTRNet recommends that frozen tissue samples be placed into an empty cryovial, sealed, and immediately submerged in liquid nitrogen for at least 30–60 sec. Samples are then stored long-term in −80°C/-150°C freezers or liquid nitrogen. Formalin fixed-paraffin embedded (FFPE) samples were collected directly from samples processed within hospital pathology departments. A random selection of 15 HGCS cases per center was requested by the central site (Hôpital Notre Dame, CRCHUM), which were shipped on dry ice in accordance with the COEUR SOPs that were developed in collaboration with CTRNet to ensure uniformity in shipping of specimens. For the present study, a minimum size of 9 mm³ was requested for frozen tissues. The inclusion criteria for FFPE samples was a minimum tumor tissue size of 4 mm². Upon receipt, frozen tissues were immediately stored at −80°C and matching paraffin blocks were stored at room temperature in a secure locker. All histologic slides were reviewed to ensure presence of available tumor tissue. An overall summary description of the collected samples is provided in Table 2. A description of how the specimens were handled and stored in each biobank is provided in Table 3.

Table 2.

Summary Description of Collected Samples

Age median in years (range)	59 (36–85)
¹Stage 1	12
Stage 2	8
Stage 3	79
Stage 4	10
Stage unknown	1
²Mean follow-up time in months (range)	18 (0–98)
Number of deaths	40
Number of patients with progressive disease	68
Frozen tissues in OCT	15
Frozen tissues in RnaLater	11
Flash frozen tissues	109

Samples were categorized by Stage according to FIGO classification.

Follow-up time is the time from the date of primary resection of ovarian tumor until the event of death or last contact.

Table 3.

Description of Tissue Handling Procedures ^1,2

Site #	³ FT storage media	FT storage temperature	Fixation time
1	none	−80	24 h/72 h
2	OCT	⁴LN	>24 h
3	none	−80	24 h/72 h
4	none	−80/LN	>24 h
5	none	−80	>24 h
6	RNALater	−80	>24 h
7	none	LN	>24 h
8	none	−150	24 h/72 h
9	none	LN	>24 h

Sites are not listed in the same order as in Table 1.

All specimens were flash frozen and fixed with formalin buffer.

FT, frozen tissue; ⁴LN, liquid nitrogen.

RNA extraction and analysis

Total RNA was extracted from homogenized tumor tissue by a phenol-based extraction method with TRIzol^tm reagent (Gibco/BRL, Life Technologies Inc., Grand Island, NY). RNA concentration was measured with the NanoDrop ND-1000 spectrophotometer (ThermoScientific, Willmington, MD). Quality was assessed with a 2100 Bioanalyzer with a RNA 6000 Nano LabChip kit (Agilent Technologies, Mississauga, ON, Canada) to determine the ratio of ribosomal 28S/18S RNA on denaturing 1% agarose gels. If the ratio is greater than 2.0 it may indicate the presence of sheared single-stranded genomic DNA; if this ratio is less than 1.0, it indicates probable degradation of RNA. Quality was also estimated using the RNA Integrity Number (RIN). RIN values between 1 and 4 are suitable only for PCR assays with short regions of amplification. RIN values between 4 and 7 may be suitable for quantitative PCR applications and gene expression profiling, and RIN values above 7 are suitable for most applications (www.asterand.com/Asterand/human_tissues/Asterand_RIN.pdf).²

DNA extraction and analysis

Total DNA was extracted in accordance with CTRNet SOP# 8.2.004 and processed in accordance with SOP #5.1.002 (www.CTRNet.ca). Briefly, tissues were cut into small pieces with sterile scissors and incubated overnight at 55°C with 10 mM Tris pH 7.9, 2 mM EDTA pH 8, 40 mM NaCl, 0.8% SDS, and 0.4 mg/mL proteinase K. Tris-saturated phenol was then added and the solutions were mixed for 10 min on a rolling rack. Samples were centrifuged and DNA-containing supernatants were extracted with phenol, followed by phenol/chloroform/isoamyl alcohol (25:24:1) and chloroform/isoamyl alcohol (24:1), and then precipitated with 95% cold ethanol. The DNA pellets were resuspended in Tris/EDTA (TE) buffer and quantified. DNA quantification was assessed with the Nanodrop ND-1000 Spectrophotometer. DNA concentrations were estimated from absorbance at 260 nm. The A₂₆₀/A₂₈₀ ratios were also measured. Values for the A₂₆₀/A₂₈₀ ratio normally ranged between 1.8 and 2.1. After quantification, DNA was diluted in TE buffer to a working concentration of 100 ng/μl.

ß-Globin PCR amplification

To assess DNA quality, four different length fragments of the housekeeping gene ß-globin were amplified. The maximum amplicon size positively correlates with DNA quality. PCR amplification was performed with 100 ng of tissue DNA using the EconoTaq DNA polymerase (Lucigen, Middleton, WI); the primers are shown in Table 4. PCR was performed using the following conditions: initial denaturation at 95°C for 3 min, followed by 40 cycles of denaturing at 95°C for 1 min, annealing at 55°C for 1 min, extension at 72°C for 1.5 min, and final extension at 72°C for 5 min. PCR products were analyzed on agarose gels.

Table 4.

Primers Used for β -Globin Gene Amplification by PCR

Amplicon	Forward primer	Reverse primer	Size (bp) ¹
I	GAAGAGCCAAGGACAGGTAC	CAACTTCATCCACGTTCACC	268
II	GCTCACTCAGTGTGGCAAAG	GGTTGGCCAATCTACTCCCAGG	536
III	ATTTTCCCACCCTTAGGCTG	TGGTAGCTGGATTGTAGCTG	989
IV	GGTTGGCCAATCTACTCCCAGG	TGGTAGCTGGATTGTAGCTG	1327

bp, base pair.

Tissue Microarray (TMA)

Following careful review of hematoxylin-eosin-stained slides, representative single cores (0.6 mm diameter) of each tissue sample were punched and arrayed in empty paraffin blocks. Tissue arrays, which were done in quadruplicate, were composed of 120 ovarian specimens per block. Fifteen cases were excluded because the tumor tissue size in the FFPE block was too small for four punches. The TMAs were then sectioned at 4 μm, stained with hematoxylin-eosin, and reviewed by a pathologist to confirm tumor content.

Immunohistochemistry (IHC)

The TMA slides were stained using the BenchMark XT automated stainer (Ventana Medical System Inc.). The antibodies used were anti-p53 (clone DO-1, Santa Cruz Biotechnology, Inc.), anti-WT1 (clone 6FH2, DAKO), anti-cytokeratin 7 (clone OV-TL12/30, NeoMarker), anti-Ki67 (clone SP-6, NeoMarker), and anti-E-cadherin (clone G10, Santa Cruz Biotechnology, Inc.). Antigen retrieval was carried out with Cell Conditioning 2 (VMSI; #950–123, Ventana Medical System) for anti-p53 and with Cell Conditioning 1 (VMSI; #950-124, Ventana Medical System) for all other antibodies. Prediluted antibodies were automatically dispensed, and the slides were incubated at 37°C for 30 min with the UltraView DAB detection kit (VMSI; #760–091, Ventana Medical System). Slides were counterstained with hematoxylin (VMSI; #760–2021, Ventana Medical System).

Staining quantification

TMA slides were scanned and converted into digital slides. Epithelial zones were scored according to the % of epithelial area stained. Each array was independently analyzed in a blind study by two independent observers. Inter-rating correlation was >75%. When strong differences in scoring between the two observers occurred, the core was re-evaluated to reach a consensus score.

Statistical analyses

The Pearson correlation test (two-tailed) was used to estimate the correlation coefficients. A Kruskall-Wallis test was used to estimate differences among the biobanks. The Mann-Whitney U test was used to evaluate the differences between one biobank and the others. All statistical analyses were done using Statistical Package for the Social Sciences software version 11.0 (SPSS, Inc.), and statistical significance was set at p<0.05. Hierarchical clustering analysis was carried out using Genespring™ software (Silicone Genetics) and a distance branch of 0.01 with a Pearson correlation as a similarity metric. Association between hierarchical cluster and disease recurrence interval was estimated using the Kaplan-Meier analysis and the log-rank test.

Results

DNA quality control

To evaluate the quality of the specimens collected for the COEUR repository and shipped to the central site, the quality of both DNA and RNA extracted from a set of HGSC frozen specimens was assessed. Each of the nine participating biobanks shipped 15 frozen specimens, and DNA and RNA extraction was performed on all 135 resulting specimens. The storage method, storage medium, processing time (time interval from surgical harvesting to storage in medium), and storage time in each site were recorded. One site provided specimens frozen in OCT medium (n=15, site #2), a second site provided some specimens stored in RNALater® solution (n=11, site #6), and the remaining sites provided specimens (n=109) that were flash frozen and stored at −80°C, −150° C, or in liquid nitrogen (see Tables 2 and 3). Since age has been reported as a factor affecting molecular quality, the effect of storage time on DNA/RNA integrity and immunohistochemistry staining was also evaluated. Most specimens were stored within 30 min after tissue resection (n=86, 83%), and were kept stored for more than 2 years prior to DNA and RNA extraction (n=64, 62%). After receipt of all frozen tissue specimens at the central site, samples were divided into two pieces for independent DNA and RNA extraction.

DNA was analyzed by the 260/280nm ratio and by PCR amplification of 4 fragments derived from the ß-globin gene (Fig. 1A and 1B). With the exception of one sample, DNA samples had a 260/280nm ratio ranging between 1.8 and 2.0. One sample showed a ratio of 2.09 (Fig. 1C), suggesting contamination during the phenol extraction. Following ß-globin amplification, the integrity of DNA was scored according to the number of amplified bands of increasing size from 1 to 4 (Fig. 1B). As seen in Figure 1D, 85.8% (n=115) of extracted DNA samples were of good quality with at least three amplified ß-globin bands of increasing size. Processing time from surgical resection to the biobanks had little effect on DNA integrity. Specimens processed within 30 min (81%) had at least three ß-globin gene bands, against 92% for specimens processed in more than 30 min (p=0.07). Similarly, time of storage had no significant effect on DNA integrity (p=0.16, Fig. 1E). The method of specimen storage either in OCT or in RNALater medium did not affect the quality of extracted DNA, although the temperature of storage did (data not shown). We noticed a higher number of samples with only one amplified ß-globin band in specimens stored at −80°C (n=14/69) compared to specimens stored at −150°C (n=1/19) or in liquid nitrogen (n=1/46) (p=0.001 Kruskal-Wallis test).

FIG. 1.

Integrity of DNA and RNA derived from epithelial ovarian cancer frozen tissues. (A) Representative image of gel agarose electrophoresis of genomic DNA from four different samples. (B) Representative image of DNA quality as assessed by PCR amplification of ß-globin gene fragments. A good DNA sample, such as sample IV, results in amplification of four PCR products in four separate PCR reactions. A poor DNA sample (Sample I) gives a low smear, and results in PCR amplification of only one ß-globin gene fragment. (C) Ratio of 260/280 nm by biobank site. Bars represent the average value for each biobank. (D) Frequencies of specimens with 1, 2, 3, or 4 ß-globin amplification fragments. (E) Number of ß-globin bands amplified according to the time of storage in the initial biobank. (F) Number of ß-globin bands amplified by biobank site.

While some variation was observed in the number of samples with good quality DNA from different sites, there was no statistically significant difference in the 260/280 nm ratios (Fig. 1C, p=0.593, Kruskal-Wallis test). However, the PCR amplification of ß-globin gene fragments was globally less efficient for specimens from site #1 (Fig. 1F, p<0.001 Mann-Whitney test) because of a lower proportion of specimens from which four fragments could be amplified and higher proportion from which three fragments could be amplified. The number of poor DNA samples with one or two amplified fragments was not different from one site to another (p=0.114, Kruskall Wallis test).

Altogether these results show that there was no major difference in quality of DNA extracted from specimens collected at different biobanks.

RNA quality control

RNA quality was evaluated by examining ratios of ribosomal 28S/18S RNA and RIN values obtained by gel electrophoresis (Fig. 2A and 2B). Analysis of the ribosomal 28S/18S RNA ratios showed that 82% of extracted RNA (n=112 out of 134) had ratios above 1, and 63% of those (n=85) had ratios between 1.5 and 2 (Fig. 3A). The average RIN value of the 135 samples was 7 (range 2.4–9.5). Sixty percent of RNA samples (n=80) had RIN values of 7 or more, 31% (n=41) had RIN values between 4 and 7, and only 10% (n=13) had poor RIN values below 4 (Fig. 3B); a RIN value was not obtained for one sample. There was no correlation between RNA and DNA integrity (p=0.25, Pearson's test). There were only two specimens with low RIN values and less than 3 amplified ß-globin bands.

FIG. 2.

RNA integrity analysis method. (A) Representative image of electrophoresis of four different samples. Ribosomal 28S (upper band) and 18S (lower band) RNAs are visible in samples 3 and 4 but degraded in samples 2 and 11. (B) Representative graphs of RNA samples with RIN value of 10 (upper panel) showing two peaks corresponding to ribosomal 28S and 18S RNAs, and an RNA sample with RIN value of 4 (lower panel) with absence of 28S and 18S RNAs.

FIG. 3.

RNA integrity results from frozen HGSC specimens. (A) Distribution of specimens (%) according to the ribosomal 28S/18S ratio. (B) Proportion (%) of specimens according to RIN values. Low=RIN<4, medium=4<RIN<7, high=RIN≥7. (C) Mean ratio of 28S/18S RNA according to time of storage. Graph represents the mean of each group±standard deviations. (D) RIN values according to time of storage for each sample. (E) Ratio of 28S/18S RNA by biobank site. Bars represent the average value for each biobank. (F) RIN values according to time of storage in each biobank. Bars represent the average value for each biobank.

Next, the effect of storage time on the ratio of ribosomal 28S/18S RNA and on RIN values was assessed. As seen in Figure 3C, a trend toward a higher 28S/18S ratio was observed for specimens stored for 6 months or more when compared to those stored less than 6 months (p=0.057, t-test). Similarly, when RIN values were compared between these sets of samples, the difference in integrity was statistically significant (p<0.001, Mann-Whitney U test) (Fig. 3D). Due to the relatively low number of specimens stored for the shortest period of time (n=16), this result needs to be confirmed on a larger set of specimens. We also considered the processing time interval between surgery and storage (<30 min vs.>30 min) and did not observe differences in the 28S/18S RNA ratios or RIN values (p=0.67 t-test; data not shown). There were also no differences observed for specimens stored in OCT or RNALater® compared to flash frozen specimens, although the low number of specimens in each group (n=11 for OCT specimens) precluded rigorous statistical analysis. Similarly, we did not notice any effect of temperature storage on RNA quality (p=0.238 Kruskal-Wallis test).

When the quality of RNA samples was evaluated according to the site of specimen origin, we noticed a lower average ratio of ribosomal 28S/18S RNA for site #4 (p=0.002, U-Mann-Whitney test, Fig. 3E). Although the average RIN value was slightly lower for site #4 compared to the other sites (RIN=6.5±0.18, p=0.039, U-Mann-Whitney test) (Fig. 3F), when the number of samples with low or medium RIN values (<4 and 4 to 7, respectively) is considered, there was no significant difference for site #4. In contrast, we observed a significant difference in the RNA quality of samples from site #9, with a lower average RIN value (RIN=4.71±0.6) due to a significantly higher number of samples with low RIN values (p<0.001, U-Mann-Whitney test).

Immunohistochemistry staining on paraffin embedded tissues

Blocks of paraffin-embedded tissues were collected from the nine participating biobanks, for a total of 135 blocks. All tissue specimens were analyzed for morphology by hematoxylin and eosin staining to confirm that the sample contained sufficient tumor tissue for further analysis. One-hundred twenty blocks (91%) with enough tumor tissue for the construction of tissue microarrays with four representative cores per case were selected. Tissue microarrays were then examined by IHC for staining with biomarkers that could be used for ovarian cancer subtype diagnosis: WT1, cytokeratin 7, Ki67, p53, and E-cadherin. After review, eight specimens were excluded based on an erroneous assignment of tumor type: three clear cell carcinomas, four low-grade serous carcinomas, and one high-grade endometrioid carcinoma.

A total of 87.5% of all cases showed nuclear WT1 staining in more than 50% of tumor cell nuclei and staining was completely negative in only three samples (Fig. 4A). There were no differences in the expression of WT1 among the different sites (p=0.57, Kruskal-Wallis test).

FIG. 4.

Immunohistochemistry results of TMA analysis with WT1, p53, Ki67, E-cadherin, and cytokeratin 7 (CK-7) stainings. Left panels show a graphical representation of the percentage of cells positively stained with antibodies; right panels show representative staining of positive cases: (A) WT1, (B) p53, (C) Ki67, (D) E-cadherin, (E) CK7. Each column represents a biobank site. (F) Unsupervised hierarchical clustering of immunostaining. Each column represents a specimen. Each row represents a marker stained as indicated on the right side. Blue color represents weak marker staining, orange represents moderate marker staining, and red represents strong marker staining. Branches and bottom ladder are colored according to a biobank site. *Poor quality sample with no staining of E-cadherin, WT1, p53, or CK7, and very weak staining for Ki67. A color version of this figure is available in the online article at www.liebertpub.com/bio.

Using a cut-off level of 50% of stained tumor cells as described elsewhere,^6,7 we observed that 73% of cases were positive for p53 staining (Fig. 4B). There were no significant differences in the percent of nuclear p53 positive cores among the different biobanks (p=0.07, Kruskal-Wallis).

Ki67 staining was highly heterogeneous across the entire cohort with an average expression of 66% (range 0 to 100%) of positive tumor tissues (Fig. 4C). No significant differences were observed among the different sites (p=0.80, Kruskal-Wallis test). A similar observation was made for the staining of E-cadherin, with an average of 70% of positive cells in the whole cohort (Fig. 4D). The heterogeneity of staining was observed in each set of specimens from the different biobanks, with the exception of specimens from site #9, where a more homogenous positive staining for E-cadherin was observed (range from 70% to 100%, p=0.011 Kruskal-Wallis).

There was no significant correlation between immunostaining results and the age of FFPE blocks (p>0. 05 in all cases, Pearson test). Only a weak trend toward a decrease of staining with age was noticed for expression of CK7 (r=−0.183, p=0.065, Pearson test). Similarly, there was no correlation among the different immunostaining observed with the exception of the two epithelial mesenchymal transition associated proteins, WT1 and E-cadherin, which correlated positively (p=0.028, r=0.21, Pearson's test).

We noticed only one specimen negative for E-cadherin, p53, CK7, WT1, and weak for Ki67 (indicated with a star on Fig. 4F). This result is likely due to an anomaly in the quality of this FFPE sample.

To assess whether some specimens were of a different quality for immunohistochemistry staining, we also performed a nonhierarchical clustering statistical analyses using staining results for the five biomarkers (Fig. 4F). We observed five different clusters of specimens. These clusters did not separate the samples by biobank site: Cluster #1 was obviously related to low p53 expression, whereas cluster #4 was related to low expression of CK7, and cluster #5 to low expresion of E-cadherin (Fig. 4F). While cluster #2 grouped specimens with low WT1 staining and high p53 staining, cluster #3 grouped specimens with low Ki67, high p53, high WT1, and high E-cadherin. Interestingly, the clustering was strongly associated with the progression status of patients (r=0.30, p=0.004 Pearson's test) with all patients in cluster #2 (n=20/20) exhibiting disease progression. Altogether these results show that there were no significant differences in immunostaining of paraffin tissues among specimens from different biobanks, and that the differences observed were most likely related to the clinical status of the patients from whom the specimens were obtained.

Discussion

The aim of the present study was to determine the quality of tissues collected from different sites contributing to the COEUR cohort. No particular project or methology was considered at the time of collection of the original samples but a general goal to validate biomarkers in a large cohort was predefined. The main goal of the pilot phase of the project was to identify and ensure the quality of research materials. The results of this pilot study, which consisted of 15 samples per collection site representing a total of 135 cases, have confirmed that overall, good quality samples are being collected by participating sites. A total of 85% of extracted DNA exhibited acceptable quality for amplification of long fragments by PCR. As expected, there was more variation in acceptable RNA quality, with eight of nine sites having more than 80% of RNA samples of acceptable quality, as measured by RIN values. However, only 58% of RNA samples exhibited sufficient quality for more sensitive analysis (RIN>7). Only two sites provided specimens with less than 50% of RNA having acceptable RIN values. We could not identify a major handling parameter responsible for the lower RIN values. Other biobanks have reported on the quality of RNA obtained from prostate and pancreatic cancer tissues with RIN values >7 for 40% to 73% of derived RNAs.^8,9 Even if all tissue types are not comparable due to their cellular content and other factors (eg, frequency of necrosis), our results show that cancer specimen collection across multiple centers should not create significant bias related to the quality of specimens or their derivatives.

The method of DNA and RNA extraction used here is simple and cost-effective, but may not be the most appropriate for all types of specimens. One parameter that was noted but not properly recorded, was the sample weight compared to the RNA integrity measured. It may likely be more difficult to extract good quality RNA from smaller samples. Tumor samples should be of a sufficient size to be processed for RNA using the Trizol reagent, and alternate methods of extraction should be employed when dealing with smaller samples (<10 mm³). A second parameter that is suspected to affect the quality of the specimens is the method of shipping, although each participating biobank followed the same specific SOPs. Either of these parameters may have caused the low quality of the RNA in the specimens from biobank #9. We did not observe any correlation between DNA and RNA integrity but suspect that DNA and RNA quality are sensitive to different handling or biobanking parameters. For example, one factor that was noted is the effect of storage temperature of biobanked tissues on DNA quality but not on RNA quality. While no significant effect was noticed on RNA quality, storage of tissues at −80°C was associated with lower DNA quality than tissues stored at −150°C or in liquid nitrogen.

Notably, we did not apply visual control of the frozen tumor specimens for cellularity. We think that a downstream measurement of DNA/RNA quality combined with a FFPE block from the same tumor area is a cost effective way for quality control. Less than 15% of samples were of insufficient DNA or RNA quality and/or did not have FFPE material for TMA construction. There was a substantial overlap (7%, n=9/135) in these cases, and those cases will be excluded from further analysis.

From our cohort of samples, DNA and RNA have been extracted from specimens stored for up to 9 years with satisfactory results (RIN value ≥7). We did not observe any negative effects on DNA, RNA, or immunostaining quality based on storage time. This observation has important implications for the future of COEUR activities, as our aim is to analyze specimens associated with clinical data, and a follow-up period of at least 2 years. However, it should be noted that our quality control assessment was limited to nucleic acid integrity and did not take into consideration changes in gene expression that may have occurred during tissue collection and processing. Indeed, external factors such as methodology of tissue collection, and methodology of tumor resection, including time of resection, may affect gene expression,^3–5 although this factor has been reported to have little effect in ovarian cancer tissues.^10–12

The result of a central pathology review of FFPE specimens showed a relatively low initial misclassification (5% of non-HGSC specimens). Considering the difficulty in the differential diagnosis of high-grade serous cell type from high-grade endometrioid cell type, this result was very encouraging.^13–16. Although the diagnosis of histological type is reproducible,^15,17 there are challenges to discriminate HGSC from low-grade serous carcinomas and clear cell carcinomas in rare instances.¹⁸ Knowing that clinical behavior of ovarian cancer patients is highly dependent on tumor subtype, an accurate diagnosis is necessary for more appropriate clinical patient management. This observation emphasizes the need for biomarkers in the accurate classification of epithelial ovarian cancers.¹⁹

Satisfactory results were also obtained on tissue integrity by immunohistochemical analysis. A relatively high quality of tissue was noted with our panel of five biomarkers. Only one case was antigen deficient for four of five markers, and insufficient for protein detection by IHC. There was no correlation between staining pattern of tissues and age of FFPE blocks or participating site, with the exception of CK-7, but rather staining distribution correlates with the outcome of patients (p=0.0016, Log Rank).

WT1 is considered a marker of serous cell lineage. It is positive in 80% of serous tissues, usually negative in clear cell carcinomas and focally positive in some endometrioid tissues.^6,12,20–22 Low WT1 expression has been associated with a poor outcome of serous patients.^6,23 In our cohort, using the same cut-off value previously used by others,^6,23 95% of tissues were positive for WT1, with very little variability among tissues from different biobanks, possibly indicating an increased sensitivity for our WT1 nuclear positive immunohistochemical protocol. A cut-off of 68% of WT1-positive nuclei was associated with disease recurrence (p=0.004, Log Rank), confirming a previously reported observation.⁶

We observed 73% of cases positive for p53 nuclear signal, which is similar to a previous report (58% in 502 HGSC from multiple sources).¹⁵ The staining of p53 was not associated with disease recurrence as previously reported by others in a high-grade serous cohort,⁶ suggesting that the present cohort is representative of a larger cohort and has not been biased by the selection process. We noticed a more homogeneous staining of E-cadherin in specimens from site #9. A deeper analysis should be conducted to determine the parameter influencing this result. In view of a recent report,²⁴ we can speculate that time of fixation may be involved. Indeed, Khoury et al.²⁴ showed that delay of fixation of more than 2 h induces a reduction of E-cadherin staining, but this reduction is not observed for p53 or Ki67. While it is often difficult to collect information about fixation time in a retrospective cohort of specimens, the result obtained here indicates that this parameter should be recorded whenever possible. Despite differences in staining levels, we did however note that in this pooled cohort high expression of E-cadherin was associated with a better prognosis of patient (Log Rank=4.53, p=0.033), as previously reported by others.^25–28

Altogether these results demonstrate the idea that construction of a large, homogeneous cohort from multiple centers is feasible if stringent entry controls are applied. The freezing and storage of ovarian cancer specimens in Canadian biobanks is now conducted following established SOPs developed by CTRNet to support biobanking activities and to provide good quality samples. However, there is significant variation in the details of pre-acquisition protocols for paraffin embedded-tissues used in different institutions, especially in pathology departments. The present study shows that, even when using a combination of collection and storage methods, good yields of high quality specimens can be obtained when appropriate protocols are followed. In addition we found that collection of specimens from several biobanks instead of from one site is time and cost effective and may allow collection of specimens with longer follow-up in a shorter period of time, facilitating predictive and prognostic studies. However, it remains possible that for a particular downstream biomarker assay, site-specific effects may be observed. The ability to analyze data by site, and to control for any site-specific bias, will be incorporated into all studies associated with the resource.

A review of the recent literature shows that quality control of specimens from multicenters is not reported and may not be assessed. However, our results suggest that this is an essential step to estimate the overall quality of the collection and the outcome that could be obtained. It is an essential prerequiste before initiating a research project to know the nature of the biological materials being accessed and whether this material is appropriate for the techniques employed during the research project.

Footnotes

Acknowledgments

We thank the Gynecology-Oncology and Pathology services of all participating centers for tumor procurement. We are also grateful to all COCRC and laboratory members for tissue processing and thoughtful discussions. Tumor banking was supported by each provincial organization described in . Clinical specimens from the province of Quebec were provided by the Banque de tissus et de données of the Réseau de recherche sur le cancer of the Fonds de recherche du Québec–Santé which is affiliated with the Canadian Tumour Repository Network. This work was supported by a grant from the Terry Fox Research Institute.

Disclosure Statement

No competing financial interests exist.

References

Michener

, Belinson

. Modern management of recurrent ovarian carcinoma. A systematic approach to a chronic disease. Oncology (Williston Park), 2005; 19:1277–1285discussion 1285, 1288, 1293.

Opitz

, Salinas-Riester

, Grade

et al. Impact of RNA degradation on gene expression profiling. BMC Med Genomics, 2010; 3:36.

Bray

, Paulin

, Fong

et al. Gene expression in colorectal neoplasia: Modifications induced by tissue ischaemic time and tissue handling protocol. Histopathology, 2010; 56:240–250.

Barnes

, Parisien

, Murphy

, Watson

. Influence of evolution in tumor biobanking on the interpretation of translational research. Cancer Epidemiol Biomarkers Prev, 2008; 17:3344–3350.

De Cecco

, Musella

, Veneroni

et al. Impact of biospecimens handling on biomarker research in breast cancer. BMC Cancer, 2009; 9:409.

Kobel

, Kalloger

, Boyd

et al. Ovarian carcinoma subtypes are different diseases: Implications for biomarker studies. PLoS Med, 2008; 5:e232.

Lassus

, Leminen

, Lundin

, Lehtovirta

, Butzow

. Distinct subtypes of serous ovarian carcinoma identified by p53 determination. Gynecol Oncol, 2003; 91:504–512.

Dev

, Rickman

, Sooriakumaran

et al. Biobanking after robotic-assisted radical prostatectomy: A quality assessment of providing prostate tissue for RNA studies. J Transl Med, 2011; 9:121.

Rudloff

, Bhanot

, Gerald

et al. Biobanking of human pancreas cancer tissue: Impact of ex-vivo procurement times on RNA quality. Ann Surg Oncol, 2010; 17:2229–2236.

10.

Dumur

, Sana

, Ladd

et al. Assessing the impact of tissue devitalization time on genome-wide gene expression analysis in ovarian tumor samples. Diagn Mol Pathol, 2008; 17:200–206.

11.

Jochumsen

, Tan

, Dahlgaard

, Kruse

, Mogensen

. RNA quality and gene expression analysis of ovarian tumor tissue undergoing repeated thaw-freezing. Exp Mol Pathol, 2007; 82:95–102.

12.

Hecht

, Kotsopoulos

, Gates

, Hankinson

, Tworoger

. Validation of tissue microarray technology in ovarian cancer: Results from the Nurses' Health Study. Cancer Epidemiol Biomarkers Prev, 2008; 17:3043–3050.

13.

Gilks

. Subclassification of ovarian surface epithelial tumors based on correlation of histologic and molecular pathologic data. Int J Gynecol Pathol, 2004; 23:200–205.

14.

McCluggage

. My approach to and thoughts on the typing of ovarian carcinomas. J Clin Pathol, 2008; 61:152–163.

15.

Kobel

, Kalloger

, Baker

et al. Diagnosis of ovarian carcinoma cell type is highly reproducible: A transcanadian study. Am J Surg Pathol, 2010; 34:984–993.

16.

Gilks

, Ionescu

, Kalloger

et al. Tumor cell type can be reproducibly diagnosed and is of independent prognostic significance in patients with maximally debulked ovarian carcinoma. Hum Pathol, 2008; 39:1239–1251.

17.

Malpica

, Deavers

, Tornos

et al. Interobserver and intraobserver variability of a two-tier system for grading ovarian serous carcinoma. Am J Surg Pathol, 2007; 31:1168–1174.

18.

Han

, Gilks

, Leung

et al. Mixed ovarian epithelial carcinomas with clear cell and serous components are variants of high-grade serous carcinoma: An interobserver correlative and immunohistochemical study of 32 cases. Am J Surg Pathol, 2008; 32:955–964.

19.

Kalloger

, Kobel

, Leung

et al. Calculator for ovarian carcinoma subtype prediction. Mod Pathol, 2011; 24:512–521.

20.

Kobel

, Kalloger

, Carrick

et al. A limited panel of immunomarkers can reliably distinguish between clear cell and high-grade serous carcinoma of the ovary. Am J Surg Pathol, 2009; 33:14–21.

21.

Madore

, Ren

, Filali-Mouhim

et al. Characterization of the molecular differences between ovarian endometrioid carcinoma and ovarian serous carcinoma. J Pathol, 2010; 220:392–400.

22.

McCluggage

, Soslow

, Gilks

. Patterns of p53 immunoreactivity in endometrial carcinomas: 'All or nothing' staining is of importance. Histopathology, 2011; 59:786–788.

23.

Yamamoto

, Tsuda

, Kita

et al. Clinicopathological significance of WT1 expression in ovarian cancer: A possible accelerator of tumor progression in serous adenocarcinoma. Virchows Arch, 2007; 451:27–35.

24.

Khoury

. Delay to formalin fixation alters morphology and immunohistochemistry for breast carcinoma. Appl Immunohistochem Mol Morphol, 2012; 20:531–542.

25.

Darai

, Scoazec

, Walker-Combrouze

et al. Expression of cadherins in benign, borderline, and malignant ovarian epithelial tumors: A clinicopathologic study of 60 cases. Hum Pathol, 1997; 28:922–928.

26.

Faleiro-Rodrigues

, Macedo-Pinto

, Pereira

, Lopes

. Prognostic value of E-cadherin immunoexpression in patients with primary ovarian carcinomas. Ann Oncol, 2004; 15:1535–1542.

27.

Voutilainen

, Anttila

, Sillanpaa

et al. Prognostic significance of E-cadherin-catenin complex in epithelial ovarian cancer. J Clin Pathol, 2006; 59:460–467.

28.

Shim

, Yoon

, Cho

. Prognostic significance of paired epithelial cell adhesion molecule and E-cadherin in ovarian serous carcinoma. Hum Pathol, 2009; 40:693–698.