Abstract
Background: Tissue microarrays (TMAs) are an attractive alternative to analysis of whole sections (WS). For breast carcinomas, the recent recommendations for cut-offs (i.e. Ki67, H-score) have necessitated the re-evaluation of TMAs. Materials and Methods: TMA results of immunohistochemistry (IHC) and Fluorescence in situ hybridization (FISH) testing for Estrogen receptors (ER), Progesterone receptors (PgR), Ki67 and HER2 were compared against the results of WS for 88 breast carcinomas. Results: We found excellent agreement between the two methods for ER and PgR IHC evaluation, using the H-score (Kappa coefficient 0.972 and 0.9, respectively). There was also excellent correlation for HER2 IHC (Kappa coefficient 1) and amplification (Kappa coefficient 0.933). Furthermore, scoring of Ki67 was highly-correlated between TMAs and WS (Kappa coefficient 0.954). The latter excellent correlation has not, to our knowledge, been previously reported. Conclusion: For breast cancer, TMAs are an efficient and reliable alternative to the use of WS, using the currently recommended markers, evaluation protocols and cut-off values.
Cancer development represents manifestation of eight essential alterations in cell physiology that control malignant transformation and progression: self-sufficiency in growth signals; insensitivity to growth-inhibitory signals; evasion of programmed cell death; limitless replicative potential; sustained angiogenesis; activation of tissue invasion and metastasis; re-programming of energy metabolism; and evasion of immune destruction (1, 2). Several of the proteins involved in carcinogenesis have also been used as prognostic or predictive markers in the eternal quest for optimal treatment.
Biomarkers in breast carcinoma have been studied for years to such an extent as to be compared to the long-lasting theatrical play “The Mousetrap” (3). It has been shown that immunohistochemical (IHC) expression of estrogen (ER) and progesterone (PgR) receptors is a weak prognostic, but a strong predictive marker for response to specific (endocrine) therapy (4). HER2 status has been shown to have several uses, as a prognostic factor, since HER2 positivity is associated with worse prognosis, but also as a predictive factor for several systemic therapies (5). The high levels of inaccuracy in testing have hampered the general acceptance of the value of these biomarkers. It has been estimated that up to 20% of IHC determinations of ER and PgR, and approximately the same percentage of current HER2 testing (either by immunohistochemistry or in situ hybridization) may be inaccurate (3, 4). In order to improve the accuracy of testing, guidelines were developed through collaboration between the American Society of Clinical Oncology and the College of American Pathologists (4, 5). Studies on accuracy and subsequent guideline development were largely based on published reports by large national or multi-national cooperative groups, where results between local versus central reference laboratories were compared. Often in these studies, local laboratories used the standard whole-section slides for IHC, while the reference laboratories used tissue microarrays (TMAs) (4).
Multi-core TMAs have been proposed as an alternative to the use of whole sections for the assessment of prognostic factors for various types of neoplasms (6-8). Validation of this method is absolutely necessary since the use of TMAs is constantly increasing, not only for the initial evaluation of biomarker expression, but also for external proficiency testing and training, if necessary (9).
Limitless replicative potential is one of the originally-described hallmarks of cancer (1). The most common method of its assessment in tissues is the estimate of the IHC expression of Ki67, an antigen expressed in all but the G0 phase of the cell cycle. Recently, Ki67 has drawn attention as a prognostic and predictive marker in breast cancer and relevant recommendations for its assessment have been published (10). Today, there are no systematic comparisons of the Ki67 score in TMAs vs. whole sections, and the notion is that scores in TMAs are possibly lower (10). Given the recent attention paid to the importance of Ki67 in invasive breast cancer, it is important to establish whether TMAs are a valid alternative to the use of whole sections, specifically when dealing with the current 14% cut-off.
In the present study, we compared the TMA results of IHC and Fluorescence in situ Hybridization (FISH) testing for breast biomarkers, ER, PgR, Ki67 and HER2, with the results obtained via traditional examination of whole sections.
Materials and Methods
Formalin-fixed paraffin-embedded (FFPE) primary tumor tissue samples were selected from archival tissue material from the Hellenic Cooperative Oncology Group Tumor Repository. Selection criteria were: histologically-confirmed invasive breast cancer, absence of infiltrated axillary lymph nodes, adequate tumor tissue for histopathological examination and known clinicopathological parameters. The protocol was approved by the Institutional Review Board of Papageorgiou General Hospital.
All patients after 2004 signed an informed consent allowing the use of biological material for future research purposes. For patients treated before 2004, waiver of consent was provided by the Board.
TMA construction. After evaluation of haematoxylin-eosin tissue sections from each case, representative neoplastic areas were marked and the corresponding paraffin block was retrieved. From each selected block, after acquiring 3-μm tissue sections for whole-section IHC and 5-μm sections for FISH, two tissue cores, 1.5 mm in diameter were obtained and TMA blocks were constructed, using a manual arrayer (Model I; Beecher Instruments, San Prairie, WI, USA), as described previously (11-14). Each TMA block contained 26-57 tissue cores, and multiple neoplastic and non-neoplastic tissue cores were obtained as assay controls.
Immunohistochemistry. IHC staining was performed according to standard protocols. IHC for ER (clone 6F11; Dako, Glostrup, Denmark), PgR (clone 1A6; Novocastra, Leica Biosystems, Newcastle Upon Tyne, UK), HER2 (A0485 polyclonal; Dako) and Ki67 (clone MIB-1; Dako) was performed on 3-μm thick sections (whole sections and TMAs), using a Bond Max™ autostainer (Leica Microsystems, Wetzlar, Germany), as previously described (14). All stained sections were evaluated independently by two experienced pathologists (AB and DT). When disagreement arose, the pathologists examined the sections together and reached a consensus.
Interpretation of the IHC results. The ER and PgR protein expression was evaluated using the H-score method, where the percentage of immunopositive neoplastic cells and the intensity of staining are recorded (15). Sections (whole or TMA cores) in which fewer than 1% of tumor cells exhibited staining regardless of intensity were considered negative. HER2 protein expression was scored from 0 to 3+ according to guideline recommendations (5). For Ki67 evaluation, the number of immunopositive cells (nuclear staining) was divided by the total number of neoplastic cells in the evaluated area. Protein expression of Ki67 was categorized as low (<14%) and high (≥14%), as reported previously (14).
Fluorescence in situ hybridization. Due to logistic/organizational and technical barriers for the FISH analysis, FFPE neoplastic tissue from 70 patients was examined. TMA sections or whole-tissue sections (5 μm thick) were used for FISH analysis, using the ZytoLight® SPEC HER2/CEN17 dual color probe (Zytovision, Bremerhaven, Germany), as previously described (10, 11). TMA slides were analyzed at the FISH unit of the Laboratory of Molecular Oncology of the Hellenic Foundation for Cancer Research, Aristotle University of Thessaloniki School of Medicine, using a Nikon 80i fluorescence microscope with a motorized four-slide stage, equipped with Plan Apo X100/1.4 oil objective lens (Nikon, Melville, NY, USA), an appropriate four filter set (DAPI, doublePath FIRC/TRITC, ZyGreen, ZyOrange; all from Chroma Technology Corp, Rockingham, VT, USA) and an ultrasensitive black and white camera (QImaging, Surrey, BC, Canada). As a source of fluorescence illumination, an X-cite 120 (EXFO Photonic Solutions Inc., ON, Canada) equipped with a long-life 120-watt metal halide short arc lamp was used. The system was controlled by the cytogenetic software XCyto-Gen, (Alphelys, Plaisir, France). For all probes, sequential digital images were captured by a stack motor for DAPI (one or two planes at 0.5 μm), ZyGreen (five planes at 1.0 μm) and ZyOrange (five planes at 1.0 μm). The resulting images were reconstructed with blue and, green and orange pseudo-colors.
Fluorescence images from whole sections were analyzed at the Laboratory of Pathology at the Medical School of the University of Thessalia using a Zeiss Axio Imager Z1 microscope (Carl Zeiss Light Microscopy, Göttingen, Germany), equipped with a Plan Neofluar ×100/1.3 oil objective lens and the appropriate filter settings for DAPI, SpectrumGreen and Spectrum Orange. Images were recorded using a monochrome progressive scan CV-M4/CL camera (Applied Imaging Corp., San Jose, CA, USA). The complete setup was controlled by CytoVision software, 3.93.1 release, (Applied Imaging Corp.). A z-stack function was used to capture signals in different focal planes. The image stacks consisted of 16 planes with an axial distance between two subsequent slices of 0.34 μm.
In all cases multiple digital photographs were taken from the invasive part of the tumor. A minimum of 20 neoplastic cells were counted for each case. In cases with values at or near the cut-off (1.8-2.2), an additional 20 or 40 nuclei were counted and the ratio was re-calculated. The HER2 gene was considered to be amplified when the ratio of the gene probe/centromere probe was >2.2, or the HER2 copy number was >6 (5). HER2 was considered to be positive if it was amplified by FISH or a HER2 score of 3+ was obtained by IHC. The signal evaluation was performed independently by two trained readers (MB and EK). When disagreement arose, the pathologists examined the digital photographs and data and reached a consensus.
Statistical analysis. Data on selected patient and tumor characteristics were obtained from medical records and are presented in a frequency table using descriptive statistics. Differences in the estimation of each marker using whole sections and TMAs are reported using two-way tables. Overall percentages of agreement, uncorrected for agreement by chance, between whole sections and TMA were calculated, while the degree of agreement corrected by chance was estimated using the Kappa coefficient. Kappa estimations were interpreted following the guidelines described by Landis and Koch: κ≤0, poor agreement; κ>0 but ≤0.20, slight agreement; κ>0.20 but ≤0.40, fair agreement; κ>0.40 but ≤0.60, moderate agreement; κ>0.60 but ≤0.80, substantial agreement; and κ>0.80 to 1.00, almost perfect agreement (16). Assuming that the result for the whole section was the gold standard method, the performance of the TMAs was examined by calculating diagnostic tools such as sensitivity, specificity, and positive and negative predictive value.
Analysis was conducted using the SPSS software for Windows (version 15.0; IBM Corporation, NY, USA).
Results
Clinicopathological characteristics. In the IHC study archival FFPE neoplastic tissues from 88, randomly selected, patients was included. The median age of the patients was 55 years (29-83 years), and they all suffered from invasive breast carcinoma (invasive ductal carcinoma not otherwise specified: 85.2%, invasive lobular carcinoma: 8%, and other types: 6.8%). The carcinomas were grade 1 in 15.9% of cases, grade 2 in 51.1% and grade 3 in 27.3% of cases (in 5.7% of cases the relevant data were missing). The ER status, based on whole-section immunostaining, was positive in 70.5% and negative in 28.4% of cases, while the PgR status was positive in 64.8% and negative in 35.2% of cases. The HER2 status was negative (score 0 and 1+) in 81.8%, equivocal (2+) in 4.5% and positive (3+) in 13.6% of cases.
Scoring of ER and PgR. The correlation between TMA and whole-section scoring of ER was excellent. Assuming that whole-section evaluation is the gold standard method, the statistics for ER-positive expression were the following: sensitivity 100%, specificity 96.2%, positive predictive value 98.4% and negative predictive value 100%. The Kappa coefficient (chance-corrected measure of agreement) was 0.972, a marker of excellent agreement. There was only one case which was scored as negative in whole-sections but positive in TMA. In this case, for the whole sections the ER staining for the total tumor area was <1%, but focally ranged from 1% to 5% (H-score: 2 to 5).
The correlation between TMAs and whole-section scoring of PgR was excellent. Assuming that the whole-section evaluation is the gold standard method, the statistics for the PgR-positive expression were the following: sensitivity 96.5%, specificity 93.5%, positive predictive value 96.5% and negative predictive value 93.5%. The Kappa coefficient (chance-corrected measure of agreement) was 0.9, again showing excellent agreement. In total, four discrepant cases were observed. Out of these, two cases were scored as positive in whole sections (with H-scores of 2 and 5) but were negative in TMA, and two cases were scored as negative in whole section, and exhibited very low expression of PgR in TMA sections, with the minimal H-score of 1.
HER2 score in whole sections (WS) and tissue microarrays (TMAs).
Scoring of HER2. The correlation between TMAs and whole-section scoring of HER2 expression was also excellent. The observed agreement for the HER2 expression in whole-sections and TMA (score 0, 1+, 2+ or 3+) was 88.6% (Figure 1). Furthermore, the statistics for negative (0, 1+) or equivocal HER2 expression (2+) and positive expression (3+) gave the following results: sensitivity 100%, specificity 100%, positive predictive value 100% and negative predictive value 100%. The Kappa coefficient (chance-corrected measure of agreement) was 1 (Table I).
Scoring of Ki67. For technical reasons, evaluation of Ki67 expression was not possible in two cases, thus, for Ki67 scoring, the total number of cases was 86. The correlation between TMAs and whole-section scoring of Ki67 expression was excellent. Assuming that the whole-section evaluation is the gold standard method, the statistics for the Ki67 were the following: sensitivity 95.5%, specificity 100%, positive predictive value 100% and negative predictive value 95.5%. The Kappa coefficient (chance-corrected measure of agreement) was 0.954, showing of excellent agreement (Table II). There were two cases where expression was estimated as high in whole section and low in TMA (Figure 2). In both cases, the Ki67 stain was heterogenous with focal, highly-proliferative areas (hot spots). In TMA sections, the Ki67 index for the total area was 10% and 5%, and focally 30% and 20%, respectively. In whole sections the above cases had Ki67 indices of 28% and 14%, respectively, while in the hot spots, the Ki67 indices were 60% and 30%. If considering counting only in the hot spots both cases would belong in the highly-proliferative group.
Comparison between whole-section and tissue microarray (TMA) evaluation for HER2 expression.
Comparison between whole sections and tissue microarrays (TMA) for Ki67 expression
Comparison between whole sections (WS) and tissue microarrays (TMA) for HER2 amplification.
Association between fluoresence in situ hybridization (FISH) and immunohistochemistry (IHC) results for HER2 using whole sections.
Association between fluoresence in situ hybridization (FISH) and immunohistochemistry (IHC) results for HER2 using tissue microarrays (TMA)
FISH for HER2. The correlation for FISH analysis of HER2 amplification between TMAs and whole sections was excellent. The sensitivity was 88.9%, specificity 100%, positive predictive value 100% and negative predictive value 98.4%. The observed Kappa coefficient (chance-corrected measure of agreement) of 0.933 (98.6%) denoted excellent agreement (Table III). There was one case where amplification was detected in a whole section but not in TMA (Figure 3). In this case, 40 cells were counted in each approach. The HER2/CEN17 ratio was 2.23 in the whole section and 2.01 in TMA. Interestingly, the HER2 expression was overall 1+ in the tumor area overall in the whole section and TMA, but focally (<10%), areas of higher protein expression (2+ and 3+) were found. Should the case be interpreted using the former ratio of ≥2, then it would have been considered positive.
Associations between FISH and IHC results for HER2 in whole sections and TMAs. The correlation between HER2 expression scoring and HER2 gene status detected by FISH in whole sections revealed an excellent agreement of 96.4%, with Kappa coefficient (chance-corrected measure of agreement) of 0.849 (Table IV), while the above-described correlation for the TMAs gave a Kappa coefficient of 0.874 (97.3%) (Table V).
Discussion
When a pathologist examines a breast cancer specimen, the correct diagnosis, although being the most important aim, is just the first responsibility. In recent years the assessment of prognostic and predictive factors has become mandatory in routine pathology practice. The standard approach is to report the IHC expression of ER and PgR. However, it has been established that up to 20% of current IHC determinations of ER and PgR expression may be inaccurate (4, 17). Lack of accuracy and reproducibility are attributed to variation in pre-analytical variables, thresholds for positivity, and interpretation criteria. Regarding HER2, using an antibody against HER2, is currently the initial method for HER2 testing. However, approximately 20% of current testing may be inaccurate. Common reasons for inaccurate results are pre-analytical variables, sub-optimal interpretation (usually influenced by several technical factors), large inter-observer variation and marked heterogeneity observed in some carcinomas. It is generally recommended that HER2 immunostaining needs to be further validated by in situ hybridization (ISH) analysis. Even for ISH techniques, substantial intra- and interlaboratory variability still exists (18).
TMA technology is being increasingly used for high throughput analysis of tumor markers. It has the advantages of significantly reducing the reagent cost, the number of slides and the human workload. When using TMA, there is always the concern of intra-tumoral heterogeneity and variable staining due to fixation and other technical issues (leading for example to increased staining at the periphery and reduced staining at the center of the tissue sections). Despite the large number of published studies utilizing the TMA technology, there is still little evidence, or guidance, on the optimum manufacturing, use and assessment of TMAs (19). Thus, studies assessing the degree of agreement between the results obtained from whole sections and TMAs are needed. Furthermore, since categorization cut-offs have changed recently, with the most notable example being the Ki67 assessment, it is important to gather validation data based on the most recent guidelines.
In the present study, we examined the accuracy of TMAs as an alternative to whole-tissue sections for studying ER, PgR, HER2 gene expression and amplification, and proliferative activity by Ki67.
We have shown that there is excellent agreement between whole-section and TMA IHC assessment for hormone receptors (ER and PgR). This is consistent with previous published studies (6, 20-22). Camp et al. demonstrated that analysis of two cores, with a diameter of 0.6 mm, is comparable to analysis of whole sections in more than 95% of cases (20). In the study of Sapino et al. in which at least four cores/section and the “quick score” for evaluation were used, there was also good correspondence between TMAs and whole sections (8). Bhargava et al. constructed TMAs using also 0.6-mm tissue cores with four sampled cores per tumor (22). In their study, ER and PgR were considered positive when nuclear staining was detected in more than 10% of tumor cells. In our study, we used the currently recommended H-score and have shown that two cores 1.5 mm thick are sufficiently representative of the neoplasm for assessing ER and PgR.
Regarding IHC expression of HER2, we also showed excellent agreement between TMAs and whole sections, particularly when comparing negative (score 0 and 1+) and positive (score 2+ and 3+) cases. The relevant data in the literature are rather limited. Sapino et al. reported variable results with IHC evaluation of HER2 (8), which resulted in the study with FISH of every TMA in routine practice. The cores used in their study had a diameter of 1 mm. Results of TMA FISH testing on the other hand proved to correlate highly with those for whole sections (8, 22). Bhargava et al. have demonstrated less overall concordance for HER2 status by IHC between whole sections and TMA cores than by FISH (22). In our study, the correlation between TMAs and whole section for FISH analysis of HER2 amplification was also excellent. The greater agreement between IHC and FISH data in our study can be attributed to the larger size of cores (1.5 mm) used in this study. There is only one study in the literature comparing the core size and the expression of biomarkers (23).
The reported relevant data on Ki67 expression are also limited (8). Sapino et al. used ≥20% positive nuclei as the threshold for highly proliferating tumors and showed good agreement between TMAs and whole sections. Karlsson et al., using only 10 breast cancer cases and categorization into three classes (<10%, 10-20% and >20% positive cells), showed excellent agreement between TMAs and whole sections (24). Given the recently-established importance of Ki67 assessment in breast cancer, it is imperative to gather more data from systemic studies addressing the issue of agreement between TMAs and whole sections (10, 25). In our study, we used the recently proposed cut-off of 14% and have shown that there is a good agreement between the results from TMAs and whole sections. The data from our study do not fully-support that Ki67 scores are generally lower in TMAs, since in our series, there were only two cases with high expression in whole section and low in TMA (10). In fact, our data agree with the experience reported by Camp et al. that Ki67, although generally thought to be heterogenous, shows excellent core-to-core correlations (in renal cell carcinoma) (26).
Ki67 expression in whole sections (WS) and tissue microarrays (TMA). The two cases in red were estimated as having ‘high expression’ in WS but ‘low’ in TMA.
In summary, regarding breast cancer, we have shown that the use of TMAs with two cores of 1.5-mm diameter is an efficient and reliable approach for routine assessment of prognostic and predictive factors, e.g. hormone receptor status, HER2 amplification and Ki67 presence, even for those tissues that were considered heterogeneous. TMAs can replace whole-tissue sections in routine practice for laboratories with increased workload, and can be used not only for quality control but also for research purposes.
Fluoresence in situ hybridization (FISH) ratio for HER2 in whole sections (WS) and tissue microarrays (TMA). The case marked in red was evaluated as having a ratio >2.2 in WS and <2.2 ratio in TMA.
- Received February 25, 2013.
- Revision received April 3, 2013.
- Accepted April 4, 2013.
- Copyright© 2013 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
Referenses
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.