Influence of Evolution in Tumor Biobanking on the Interpretation of Translational Research

Progress in translational cancer research has resulted in higher throughput technologies that are associated with increased analytic capacity to assess larger study cohorts. There is consequently an increasing need for human tissue biospecimens, both in terms of larger numbers as well as higher quality biospecimens. The realization of this increasing need has coincided with a shift in the approach to tissue biobanking. In Canada and elsewhere, until the late 1990s, the majority of tissue biobanking was accomplished by working within the usual operating procedures of the affiliated pathology department. These usual operating procedures were established to ensure optimal morphologic preservation, which does not necessarily correlate with optimal biomolecule preservation. The approach to biobanking changed around the year 2000 when many new biobanks were developed that deployed more stringent standard operating procedures (SOPs) to reduce the variability around the collection of tissue biospecimens for research. This change is reflected in the official recognition of the importance of human biospecimen banking by the U.S. National Cancer Institute in 2002 and subsequently National Cancer Institute's establishment of the Office of Biorepositories and Biospecimen Research (OBBR) to guide, coordinate, and develop biobanking activities to ensure availability of high-quality biospecimens for research. This widespread shift in biobanking methods has potential implications for biospecimen variability and, consequently, for interpretation of research data generated from cohorts collected in each era.

There are numerous variables around the process of collecting surgical tissues for research that may potentially influence subsequent gene expression analyses (preanalytic biospecimen variables; Table 1). These include clinical variables, such as type and duration of anesthesia (1) and extent and duration of ischemia (2-5). Additionally, there are nonclinical variables that may be introduced between the operating room and the biobank, such as transport time and mode (room temperature or on ice) between the operating room and the pathology department. Despite deployment of tissue collection SOPs, it is impossible to eliminate variability in the presence of competing clinical priorities. Although the importance of biomolecule quality is widely recognized within the laboratory, the recognition of the importance of biospecimen quality has only recently emerged. Consequently, few studies in the literature include documentation or normalization of these variables.

One variable widely recognized for its potential to affect biospecimen quality is the elapsed time from surgical excision of a biospecimen to its cryopreservation at or below −70°C (“biospecimen collection time”). There is a general consensus that biobanks should aim to achieve the shortest collection time possible to minimize biomolecular artifact (6-10). The often-cited gold standard for biospecimen collection time is 30 minutes (2). The potential effect of this variable is perhaps particularly high for RNA, the most labile biomolecule for which stability is influenced by multiple factors (11-13) and likely including each of the biospecimen collection variables listed in Table 1.

Most biobanks and translational researchers alike recognize that it is ideal to obtain rapidly collected tissue biospecimens. This was not achieved in most pre-2000 era tissue biobanks and, furthermore, despite the presence of SOPs in post-2000 era biobanks, it is not always possible to achieve a short collection time. In this study, we considered whether collection time varies between pre-2000 and post-2000 era biobanks and how this might affect the interpretation of gene expression data. Because current translational research is based on both pre-2000 and post-2000 era biobanks, we have examined whether the era in which biospecimens were collected might influence interpretation of gene expression studies.

The range of collection times of breast tumor biospecimens at two biobanks were compared. The Manitoba Breast Tumor Bank (MBTB) was selected as representative of pre-2000 era biobanks. The MBTB was established in 1993 within the Department of Pathology at the University of Manitoba and has relied on a largely passive mechanism involving unassisted pathologists to procure biospecimens. The operating protocol for tissue collection was not standardized other than logging of collection times. The British Columbia Cancer Agency Tumour Tissue Repository (BCCA-TTR) was selected as the representative of post-2000 era biobanks. The BCCA-TTR was established in 2003 and from its inception operated independently from affiliated pathology departments and deployed a stringent tissue collection SOP and biobank staff to assist pathologists to minimize biospecimen collection time and to standardize and document collection variables. These SOPs are documented elsewhere (14).

For MBTB and the BCCA-TTR, the collection time had been recorded for most biospecimens. Both of these biobanks defined the time of excision as the time of removal of the biospecimen from the operating table as documented in the operating room by a nurse. The time of cryopreservation was defined as the time that the biospecimen was stored at −70°C or in a liquid nitrogen vapor freezer.

Queries of the databases at MBTB and the BCCA-TTR were done by each biobank's informatics staff. These queries returned the excision time and cryopreservation time for each breast tumor biospecimen. Biospecimens for which either of these times were unknown were excluded from these analyses. Data were sorted into four groups: <30 min, 31 to 60 min, 61 to 120 min, and >120 min between excision and cryopreservation. These data were graphed using Microsoft Excel (Microsoft Corporation). Pearson χ² tests were done to compare the proportion of breast tumor biospecimens collected within the four collection time groups between MBTB and the BCCA-TTR (JMP statistical software, v7.0, SAS Institute).

An additional query was done to mimic the selection of a cohort of breast tissues to fulfill a researcher's request. From each biobank, a cohort of 50 breast cases was randomly selected with the following criteria: excision time available, cryopreservation time available, and positive estrogen receptor (ER) status. The collection times for each cohort were compared by t test (JMP statistical software, v7.0, SAS Institute).

Biopecimens collected by MBTB had associated ER protein expression data available for this part of the analysis. These data resulted from ligand binding assay (LBA) on tissue samples collected in parallel to the MBTB biospecimens. Using LBA, we were able to assess the level of the functional ER protein. This is in contrast to immunohistochemistry assays that indicate total protein. LBA was done by a single provincial steroid receptor assay laboratory and the cohort of MBTB breast tumor biospecimens considered in this analysis were collected from a single surgical pathology center.

A query of the MBTB database was done to find those cases that were ER positive by LBA (>3 fmol/mg) and for which tissue collection time and ER LBA results were available. A bivariate analysis (Spearman correlation) of collection time and ER protein level was done using the JMP statistical software (v7.0; SAS Institute).

Ten fresh breast carcinoma biospecimens were obtained for the mRNA portion of this study. These biospecimens were selected by the attending pathologist once they were determined to be suitable for the current study and to harbor sufficient gross tumor to enable sampling for clinical assessment, steroid receptor assay, and the MBTB collection. A standardized tissue collection protocol was used to minimize temperature variation and RNA degradation wherein biospecimens were removed from the surgical field, placed into a plastic bag, and directly immersed in a mixture of normal saline and slush ice. Biospecimens were then transported on ice from the operating room to the pathology department where they were dissected and examined as rapidly as possible. Where deemed appropriate by the attending pathologist, a portion of the tissue biospecimen was then removed for research purposes. Research biospecimens were homogenized rapidly into multiple fragments (each <1 mm³) by scalpel and equivalent portions were aliquoted into precooled cryovials to be stored on ice for 0, 3, 6, or 24 h before snap freezing at −70°C. The 0 time points corresponded to 1 h after removal from the operative surgical field. Samples were later retrieved from −70°C storage and RNA was extracted as described previously (15). A tissue block contiguous with the research sample was processed to create a paraffin block for histologic assessment (16).

To assess levels of c-myc and ER RNA in these tumor samples, competitive reverse transcription PCR (competitive RT-PCR) assays were developed and validated. Established protocols were used to create c-myc and ER competitive RT-PCR standards were subcloned into plasmids to facilitate in vitro transcription of synthetic cRNAs (17, 18). RT-PCR reactions were done as previously described (16) but with sample RNA and serial dilutions of synthetic cRNA standards. The c-myc competitive standard comprised a homologous sequence that could be amplified by primers myc-p1 (5′-ACCACCAGCAGCGACTCT-3′) and myc-p2 (5′-GTTCGCCTCTTGACATTCTC-3′) to yield a 261-bp PCR fragment in parallel with a 333-bp signal for the endogenous c-myc RNA. The ER competitive standard comprised a heterologous sequence that could be amplified by primers ER-p5 (5′-TGCTCCTAACTTGCTCTTGG-3′) and ER-p7 (5′-TCCAGAGACTTCAGGGTGC-3′) to yield a 276-bp PCR fragment in parallel with a 198-bp signal for the endogenous ER RNA.

Briefly, 100 ng total RNA from sample were reverse transcribed with varying amounts of synthetic standard in a volume of 2 μL of reverse transcriptase mix [1× reverse transcriptase; 200 units Moloney murine leukemia virus reverse transcriptase (BRL); 0.5 mmol/L each of dGTP, dATP, dTTP, dCTP; 1 μmol/L bovine serum albumin; 0.01 mol/L DTT; 1.25 mmol/L oligo d(T) primer; 5% DMSO] and incubated for 60 min at 37°C. PCR amplification was then done in an MJ Research PTC-100 thermal cycler. Each PCR reaction was done in a 50 μL volume using 2 μL of the completed reverse transcription reaction containing cDNA and the following reagents: 1× PCR buffer; 2 mmol/L MgCl₂; 1.1 units Taq polymerase (Promega); 200 mmol/L each of dGTP, dATP, dTTP, and dCTP; and 0.5 mmol/L PCR primers. The PCR protocol consisted of 5 min at 94°C, followed by 30 cycles of 45 s at 93°C, 45 s at 56°C and 90 s at 75°C, before one final incubation of 7 min at 72°C. After thermal cycling was completed, 1.5 μL of gel loading buffer were added to 15 μL of the PCR reaction and samples were electrophoresed on a 2% agarose gel, poststained with ethidium bromide, and viewed and captured as digital images under UV light. Band intensities were then determined by densitometry and these data were analyzed by linear regression plots to determine the point of equivalence. The approximate level of expression in samples was first identified in a range-finding experiment (using a wide range of competitor cRNA from 10² to 10⁻⁸ pg with 50 ng of sample RNA) and subsequently the level of expression was analyzed with a more restricted range of serial dilutions (typically from 10⁻² to 10⁻⁶ pg).

The proportion of breast tumor biospecimens collected within each of the four time groups (<30, 31-60, 61-120, and >120 minutes) at MBTB and BCCA-TTR are summarized in Fig. 1. Whereas only 32% (155 of 485) of biospecimens collected at MBTB were cryopreserved within 30 minutes of excision, 76% (305 of 400) of biospecimens collected at the BCCA-TTR were cryopreserved within the same timeframe. Conversely, MBTB had higher numbers of biospecimens collected in each of the other time groups (31-60 minutes: 33% MBTB, 22% BCCA-TTR; 61-120 minutes: 28% MBTB, 2% BCCA-TTR; >120 minutes: 6% MBTB, 1% BCCA-TTR). The difference in collection times between MBTB and the BCCA-TTR was statistically significant (χ²; P < 0.0001).

Because research studies using human tissue biospecimens may include a cohort of as few as 50 cases, we virtually selected 50-case cohorts of ER-positive breast biospecimens from MBTB and BCCA-TTR to address the question of whether the variability of collection times between a pre-2000 and post-2000 era biobank may be significant in a typical study cohort. Collection time in the MBTB 50-case cohort ranged from 5 to 95 minutes with a mean of 42 minutes, median 40 minutes. Within the BCCA-TTR 50-case cohort, collection time ranged from 9 to 52 minutes with a mean of 26 minutes and a median of 24 minutes. The difference in collection times between these cohorts was statistically significant (t test; P < 0.0001).

A query of the MBTB database returned 137 cases that met the criteria for this analysis: ER positive, collection time <120 minutes, and ER LBA results available. Within this 137-case cohort, collection time ranged from 20 to 120 minutes (mean 61 minutes; median 56 minutes). The functional ER protein level ranged from 3.2 to 307 fmol/mg (mean 44.1 fmol/mg; median 21 fmol/mg). When the collection time groupings considered above were applied to this cohort, the following mean ER protein levels were observed: <30 minutes—50.2 fmol/mg; 31 to 60 minutes—47.4 fmol/mg; and 61 to 120 minutes—28.9 fmol/mg.

A bivariate analysis (Spearman correlation) of collection time and ER protein level showed a statistically significant, inverse correlation between collection time and ER protein level (P = 0.025; r = −0.191; Fig. 2).

To validate the competitive RT-PCR assays for c-myc and ER, human breast cancer cell lines (MCF7 and MDA-MB-231) were grown and harvested as described previously (19). RNA was extracted and competitive RT-PCR assays were compared with the results of Northern blot assessment of c-myc and ER levels. The ER-positive MCF7 breast cell line subjected to treatment with estrogen for a range of times was used to generate RNA with different levels of c-myc expression. The results indicate that the c-myc competitive RT-PCR assay can accurately distinguish small differences (2-fold) by comparison with Northern blot analysis. This provided a calibration graph for analysis of c-myc expression in tumors (Fig. 3A). A similar analysis was done for the ER RT-PCR assay with test RNA samples composed of different ratios of RNA from ER-positive MCF7 cells or an ER-positive tumor sample (ER value of 211 fmol/mg of protein), mixed with RNA from the ER-negative cell line MDA-MB-231 (data not shown). The results indicate that the ER RT-PCR assay can accurately distinguish up to a 100-fold difference in ER level.

c-myc mRNA expression levels were assessed at different time points after collection in nine tumors collected and prepared for this study (note that case 3 was excluded due to the absence of tumor). There was a progressive decline in the level of c-myc mRNA with time of collection [mean^(SD) = 92%^(6%), 82%^(28%), and 75%^(23%) of initial levels at 3, 6, and 24 h] in all nine tumors (Fig. 3B; Table 2). The differences in the reduction of c-myc mRNA levels showed no obvious relationship with the different clinical and histopathologic features of these cases as assessed in adjacent paraffin sections. However, initial c-myc levels were 2- to 5-fold higher in progesterone receptor–negative tumors (t test; P = 0.0003).

In three of seven ER-positive cases where there was sufficient RNA, further analysis of both ER levels and RNA quality was done. These included case 8 that showed very little change in c-myc and cases 2 and 4 that showed marked decline in c-myc. However, only negligible reduction of ER mRNA level was observed in these three cases [mean^(SD) = 96%^(3%), 95%^(4%), and 94%^(4%) of the initial levels at 3, 6, and 24 h] and this change in mRNA level was significantly different from that of c-myc (t test; P = 0.025; Fig. 3B; Table 2).

Northern blot analyses of RNA from cases 2, 8, and 9 were carried out to assess RNA integrity by visual assessment of 28S and 18S rRNA bands (data not shown). No significant degradation of RNA up to 24 h was observed. This suggests that the reduction of c-myc mRNA levels, which was considerably higher in cases 2 and 9 compared with case 8, is probably not attributable to differences in RNA quality.

Our underlying premise for this study was that recent, fundamental changes in the approach to tissue biobanking will influence the characteristics of biospecimens held within biobanks and the molecular data obtained from cohorts derived from older, pre-2000 versus newer, post-2000 era biobanks. Biobanks established before the year 2000 have typically been populated by biospecimens collected without the resources or mission to apply standard procedures or to document collection variables. By contrast, newer biobanks have begun to deploy more active approaches to achieve rapid and standardized harvesting of biospecimens, governed by SOPs and documentation.

In 2005, the National Cancer Institute established the OBBR. The OBBR is charged with guiding biobanking within the National Cancer Institute to ensure that high-quality human biospecimens are available for cancer research. The OBBR builds on preexisting organizations such as the International Society of Biological and Environmental Repositories and the Cooperative Human Tissue Network and is paralleled by organizations such as the Canadian Tumour Repository Network. These organizations aim to improve and unify biobanking practices and have published biobanking “Best Practices” (6, 20) and SOPs (14). However, despite the growing awareness of the activities of the OBBR and similar biobanking organizations and the publication of biobanking Best Practices, as well as the increased attention that biobanking has received since the year 2000, there remains to be a dearth of scientific research and awareness around preanalytic biospecimen collection factors.

Collection time is likely to be an important variable influencing molecular data concerning levels of gene expression. We have shown that the notion of <30 minutes collection time for biospecimens that is often quoted in research publications may be undermined by the realities of clinical priorities. There is typically a range of collection times associated with tumor biospecimens accrued to biobanks and this can be significantly different between biospecimen cohorts held within an older, pre-2000 era versus a newer, post-2000 era tumor biobank. We have also confirmed by analysis of mRNA for two important breast cancer genes, c-myc and ER, and analysis of ER protein expression, that expression levels can decline in relation to tissue collection times. The decline in expression levels may be relatively small over the typical (<2 h) collection time periods if biospecimens are collected, under a standardized protocol, cooled on ice. However, differential rates of decline may occur between different genes.

A small number of previous studies in the literature have examined the effects of preanalytic biospecimen collection variables on analysis of gene expression. For autopsy-derived biospecimens, agonal factors such as hypoxia and hydration, tissue factors such as pH, and RNA integrity can all contribute to variability in gene expression profiles (21). For surgery-derived biospecimens, surgical manipulation, ischemia, anesthesia, and RNA stability have also been shown to influence gene expression (1-5, 22). Although some studies have suggested that collection times ranging up to several hours may be a minor factor for some tissues such as breast (23) and lung (24), these studies focused on the assessment of relatively stable housekeeping genes and limited numbers of normal tissue biospecimens. In contrast, a larger study of prostate biospecimens obtained during and after surgery showed significant effects on 8 of 91 tumor-associated genes studied by microarray expression analysis, attributable to both intraoperative factors and collection time (25). Many of the genes manifesting variability corresponded to relatively tightly regulated acute phase and stress response genes such as JunB and p21, whereas expression of housekeeping genes was unaltered.

The level of mRNA is determined by the gene transcription apparatus and mRNA stability and is typically coordinated and influenced by cell signaling that responds to both internal and external signals. These signals include hormones, cytokines and growth factors, and external factors such as hypoxia (11, 13) and other forms of cellular stress (e.g., starvation; ref. 26). Our initial assumption was that tissue collection time is an important variable to control for reliable analysis of gene expression in clinical samples, particularly where genes with tightly regulated or unstable mRNAs are involved. Our results show that when using a standardized collection protocol to minimize RNA degradation, a relatively small decay is observed in c-myc levels over the first 3 h that is unlikely to significantly affect the results of any comparison between tumors, particularly given the much larger scale of variation between tumors. However, the level of the relatively unstable c-myc mRNA (27) declines significantly over 24 h in comparison with the relatively stable ER mRNA, even when biospecimens are collected on ice. It should be noted that our protocol did not include any procedure to separate minced tissue fragments from degradative enzymes that may have been released by mechanical disruption, and this may be a factor contributing to decay of gene expression levels. However, we observed no significant change in overall RNA quality as assessed by Northern blot. This finding suggests that documentation of biospecimen collection time may be important for comparison of gene expression levels between certain genes. Furthermore, it is possible that biospecimens collected at room temperature or retrospectively from archival clinical paraffin blocks, with no knowledge of the original collection method or time, may show more variability (28). Due to specimen availability, we have not tested this on the present series.

Surgical tissue biospecimen collection times vary and the degree of variability may be higher in pre-2000 versus post-2000 era biobanks. Collection times for individual tissues can influence protein and mRNA expression levels and the magnitude of this effect can vary between genes. The extent to which this variability might be significant in comparison with the scale of biological differences between biospecimens remains to be determined. Future studies are also required to determine the degree to which collection time and other biospecimen collection variables may affect the translation of molecular assay data based on older, retrospective collections to data based on SOP-governed, prospective collections. We conclude that the effects of tissue biospecimen collection time on gene expression needs to be better understood and integrated into research.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Seminal work for this project was done by Rekha Singh and Linda Curtis.