Loss of Antigenicity in Stored Sections of Breast Cancer Tissue Microarrays

Immunohistochemical characterization of tumor tissues in epidemiological studies is a promising approach to identify breast cancer subtypes with distinct etiology. An important limitation of previous studies has been the small size of most studies and the use of nonstandardized immunohistochemical assays performed at different study centers in larger studies (Althuis et al. “Article title,” submitted for publication). Tissue microarrays (TMAs), which consist of recipient paraffin blocks that contain small tissue cores removed from targets of many individual donor paraffin blocks, offer an attractive method for obtaining standardized, rapid, and cost-effective immunohistochemical characterization of many tumors (1). Before histological sections of paraffin-embedded tissues can be mounted on slides for immunostaining, tissues must be repeatedly sectioned to achieve a flat surface (“facing the block”), which results in loss of specimen. Cutting and storing multiple sections every time blocks are faced conserves tissue and allows the rapid return of loaned blocks but assumes that immunoreactivity is stable over time.

Recent clinicopathological studies have reported loss of staining intensity for several markers in stored whole sections prepared from standard tissue blocks (2–5). This problem is likely to be worse in TMAs because they contain only limited tissue representation of each case. Accordingly, patchy loss of antigenicity in stored sections of TMAs may not only reduce the intensity and extent of staining but also result in unacceptably numerous false-negative results. However, repeated sectioning of the same TMA block on multiple occasions may exhaust targets, requiring exclusion of cases with inadequate material remaining in the TMA block and reducing the value of the TMA for future assays.

To assess this concern, we performed an immunohistochemical study that compared results for sections of a single TMA block from 125 invasive breast carcinomas that were cut and stored for 6 months prior to staining with sections that were cut and stained on the same day. The cases were collected in a population-based case-control study conducted in Poland sponsored by the National Cancer Institute (Bethesda, MD).

Breast cancer cases were participants in a population-based case-control study of breast cancer in Poland, conducted by the National Cancer Institute in collaboration with the Cancer Center and M. Sklodowska-Curie Institute of Oncology in Warsaw and the Institute of Occupational Medicine in £ódz. Appropriate institutional review boards in the United States and Poland approved the study. Eligible cases for the case-control study were women aged 19–80 years residing in Warsaw and £ódz who had been newly diagnosed with pathologically confirmed in situ or invasive breast cancer between February 1, 2000 and January 31, 2003. The participation rate for cases was about 80%, with tissue blocks obtained from over 90% of enrolled cases. This report analyzed a random sample of 125 invasive breast carcinoma specimens obtained from a population of patients who had not received chemotherapy, radiation, or hormonal treatment prior to tumor resection.

The tissues included in this study were obtained from mastectomy (n = 110), quadrantectomy (n = 6), lumpectomy (n = 8), or other (n = 1) resections. Pathologists at four hospitals in Poland collected research samples of tumor that were fixed in formalin and prepared as paraffin-embedded blocks according to standard protocols. We microscopically reviewed one H&E-stained section per block to identify foci of tumor suitable for TMA construction. The tumors were classified as invasive ductal (n = 80), lobular (n = 25), mixed ductal carcinoma in situ and invasive ductal (n = 6), medullary (n = 2), tubular or cribriform (n = 3), papillary (n = 2), mucinous (n = 1), and other (n = 6).

We reembedded blocks in fresh paraffin and prepared TMAs using a manual arraying instrument (Beecher Instruments, Silver Spring, MD) at the Advanced Technology Center (National Cancer Institute, Gaithersburg, MD). We selected targets for arraying by marking the corresponding areas on H&E-stained sections of each block. Three tissue cores (diameter 0.6 mm, height 3–4 mm) were transferred from each “donor” block to the “recipient” TMA block (45 × 20 mm, 0.7 mm center), which was designed to contain 375 cores from 125 cases in 15 subarrays (three columns of five subarrays each containing 25 cores). One of three cores per case was placed in each column to permit independent assessment. Two additional 5 × 5 subarrays of normal tissues including breast, liver, ovary, salivary gland, kidney, and breast cancer cell lines were included in the TMA block as negative and positive controls. Following TMA construction, 50 sections (5 μm) were cut using the adhesive-coated tape system and transferred to adhesive-coated slides (Instrumedics, Hackensack, NJ) and stored at room temperature for 6 months prior to immunostaining (“stored sections”). During the intervening 6 months, the TMA block was stored at room temperature. Then, sections from the TMA block were cut and immunostained on the same day (“fresh sections”) in the same staining batch as the stored sections.

Immunostaining for estrogen receptor (ER-α), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) was performed on the DAKO Autostainer (DAKO Corp., Carpinteria, CA) using the avidin-biotin peroxidase method (6) in one batch using a standard protocol. Sections were deparaffinized through two changes of xylene (3 min), rehydrated through graded alcohols to distilled water, and subjected to antigen retrieval using DAKO Target Retrieval Solution and steam for 20 min. After blocking endogenous peroxidase activity with hydrogen peroxidase, primary antibodies were applied: mouse monoclonal for ER (Novocastra, Newcastle Upon Tyne, UK, NCL-ER-6F11/2; 1:40 for 60 min), mouse monoclonal for PR (Novocastra NCL-PGR-312; 1:100 for 60 min), and rabbit polyclonal for HER2 (DAKO A0485; 1:125 for 30 min). Biotinylated goat anti-mouse and anti-rabbit antibody (DAKO Biotinylated Multi-Link) followed by streptavidin-horseradish peroxidase were used as secondary and tertiary reagents. Sections were visualized with diaminobenzidine (DAKO) followed by light counterstain with hematoxylin.

Each core was evaluated for the percentage of invasive tumor cells stained on a continuous scale (0–100%) and for intensity on a three-point scale (none = 0, weak = 1, moderate = 2, strong = 3). Cores in which tumor occupied less than 10% of the total area that yielded negative immunostaining results were considered unsatisfactory and were excluded from the analysis. Stains for ER and PR were evaluated for specific nuclear staining and stains for HER2 were assessed for complete membranous staining. TMA slides were evaluated masked to time to staining (stored versus fresh).

Results for cases having at least one satisfactory core in both stored and fresh sections were compared in paired analyses. A case was categorized as positive for a marker if staining was identified on any core; the average percentage of cells stained was calculated considering all satisfactory cores whether positive or negative (up to three cores).

Immunoreactivity as the percentage of positive cases in stored and fresh sections was compared in paired 2 × 2 tables using McNemars' test. A case with positive staining in either stored or fresh sections is considered to be a true positive case; therefore, we also calculated the percentage of positive cases for stored and fresh sections using the total number of positive cases according to either stored or fresh sections as the denominator (i.e., number of stored positives/total number of positives or number of fresh positives/total number of positives). Percentage agreement was calculated as the number of pairs concordantly categorized (negative or positive) in stored and fresh sections divided by the total number of pairs multiplied by 100. Agreement between stored and fresh sections was also assessed using unweighted κ statistics; κ ≥ 0.4 was considered a moderate and κ ≥ 0.8 was considered a strong level of agreement (7). We compared the percentage of cells stained per case (based on the average of up to three cores) in stored and fresh sections using the Wilcoxon signed rank test. Staining intensity was compared for stored and fresh sections by calculating the sum of intensity scores for all satisfactory cores divided by the total number of such cores converted to a categorical variable: 0–1.0 (weak), 1.1–2.0 (moderate), and 2.1–3.0 (strong). Categorical classifications of intensity were compared as percentage agreement, unweighted κ statistics, and by calculating the ratio of cases showing greater staining intensity in stored as compared with fresh sections. Percentages and κ values are shown with corresponding 95% confidence intervals (CIs). The SAS Statistical Package, Version 8.1 was used for all analyses (SAS Institute, Cary, NC).

Table 1 shows the cross-classification of stored and fresh sections stained for ER, PR, and HER2. For ER, a similar percentage of cases were classified as positive in stored and fresh sections (59.0% versus 59.0%, P = 1.0). Significantly fewer cases were classified as PR positive in stored versus fresh sections (56.3% versus 64.1%, P = 0.01); this disparity was more pronounced for HER2 (45.5% versus 64.4%, P < 0.001). When the total number of positive cases by either stored or fresh sections (ER, n = 65; PR and HER2, n = 67) was considered as the denominator, the percentage of positive cases in stored sections and fresh sections was 90.8% (95% CI = 83.8–97.8) in both stored and fresh sections for ER, 86.6% (95% CI = 78.4–94.8) and 98.5% (95% CI = 95.6–100) for PR, and 68.6% (95% CI = 57.5–79.7) and 97.0% (95% CI = 93.0–100) for HER2.

The percentage agreement between stored and fresh sections was higher for ER (88.0%, 95% CI = 81.6–94.4) and PR (90.3%, 95% CI = 84.6–96.0) than for HER2 (77.2%, 95% CI = 69.0–85.4; Table 1). The κ statistic indicated moderate to strong agreement for ER (κ = 0.75, 95% CI = 0.62–0.88) and PR (κ = 0.80, 95% CI = 0.68–0.92) and slightly lower agreement for HER2 (κ = 0.56, 95% CI = 0.41–0.70).

The percentage of cells stained per case among positive cases was lower for stored sections compared with fresh sections for all markers (Fig. 1). Median percentages for stored versus fresh sections were 27.0% versus 38.0% (P = 0.001) for ER, 30.0% versus 70.0% (P < 0.0001) for PR, and 0% versus 30.0% (P < 0.0001) for HER2.

The percentage agreement for ER and PR staining intensity among positive cases was moderate (ER: 69.2%, 95% CI = 58.0–80.4; PR: 56.7%, 95% CI = 44.8–68.6) and stronger for HER2 (85.1%, 95% CI = 76.6–93.6). The κ statistic demonstrated moderate agreement for ER (κ = 0.52, 95% CI = 0.35–0.69), slight agreement for PR (κ = 0.29, 95% CI = 0.13–0.44), and stronger agreement for HER2 (κ = 0.66, 95% CI = 0.48–0.85). The ratio of cases showing stronger intensity in stored as compared with fresh sections was <1.0 for all markers (ER = 4/16, PR = 2/27, and HER2 = 1/9; Table 2; Fig. 2).

Our data indicate that sections of breast cancer TMAs stored for 6 months at room temperature show reduced immunostaining for ER, PR, and HER2 compared with freshly cut sections stained on the same day taken from the same TMA block. The greatest loss in the percentage of immunoreactive cases in stored sections was found for HER2; the percentage of cells stained and staining intensity were dramatically reduced for all markers in stored sections, with the most striking losses found for PR. Overall, extent and intensity of staining were reduced more by storage than classification of cases as negative or positive.

Previous investigations have observed a loss of staining intensity for ER following storage of sections (2–4) and have demonstrated that increased length and higher storage temperatures increase loss of antigenicity for PR and HER2. However, cold storage does not eliminate antigen degradation (5). Loss of immunoreactivity has been demonstrated with both light microscopic and computer-generated interpretation of stains. One report has suggested that nuclear and cytoplasmic stains are less affected by storage time and temperature than membranous stains (5). The cause of antigen degradation over time is unknown, but oxidation has been suggested as a mechanism and storage under nitrogen may preserve antigenicity.

For many markers, the etiological and clinical relevance of minimal staining is not entirely certain (8, 9) and cut points for scoring stains as positive have not been standardized. In this study, we evaluated staining of any extent and intensity because we could not exclude that immunoreactivity below cut points set for determining therapeutic options (i.e., Herceptin, tamoxifen, etc.) might prove to have etiological relevance. For example, had we set moderate intensity as a threshold for a positive HER2 stain, 29.6% of cases in fresh sections would have been classified as positive, similar to that reported for 2+ or 3+ staining using standardized clinical protocols.

Although we stained stored and fresh sections simultaneously using the standard protocol used in our institution, we recognize that results of immunohistochemical stains obtained with protocols used in other laboratories may differ. We also recognize that the paraffin that had been used in Poland differed from that used in the United States and that routine deparaffinization may not completely remove paraffin from tissue sections. However, we performed pilot work that demonstrated that residual paraffin was unrelated to our staining results (data not shown).

In our study, 17.6–20.0% of cases lacked a satisfactory core when the TMA was faced and sectioned a second time (fresh sections, cut after 50 sections) compared with 3.2–4.8% of cases at the initial sectioning (stored sections). This highlights the loss of tissue when a block is refaced and cut again. Similarly, losses of 10% to over 30% have been reported for different tissue types despite the routine inclusion of multiple cores per case (10). These data underscore the need to consider tissue access, tissue heterogeneity, redundancy of tissue core representation, and preparation of multiple TMA blocks in planning epidemiological investigations.

Efforts to develop improved microtome sectioning techniques that maximize tissue conservation and validation of improved methods for preserving immunoreactivity in stored tissue sections are urgently needed. Currently, we are working with storage under nitrogen and shipping in vacuum-packed bags. As high throughput quantitative techniques for assessing immunostaining assume greater importance in etiological and translational research, objective quality assurance guidelines will likely become a growing concern.

We thank the pathology department at the following institutions for the collection of tissue samples included in this study: Cancer Center and M. Sklodowska-Curie Institute of Oncology (Warsaw, Poland); Oncological Clinic of Medical Academy (£ódz, Poland); and Polish Mother's Memorial Hospital (£ódz, Poland). We also thank Kimberly Parker at Advanced Technology Center (Gaithersburg, MD) for technical help. Supported in part by a National Cancer Institute Breast Cancer Faculty Award (M.E.S).