Tissue microarrays: leaping the gap between research and clinical adoption
Tissue microarrays (TMAs) have become a main stay in preclinical and translational research, especially for the development of bio-marker assays for characterization of disease. They allow for ana lysis of extremely small amounts of tissue, thus preserving valuable tis-sue blocks. At the same time, they dramatically increase the efficiency and cost–effectiveness of performing tissue-based studies [1,2] by enabling the examination of 10s to 100s of different patient samples on the same slide. This approach also leads to greater reproducibility and compa-rability of patient samples since all samples are exposed to identical processing conditions [3]. By linking patient sample data to pathological and clinical data, including disease follow-up and treatment response, TMAs have expanded translational research, enabling large-scale bio-marker and disease progression studies, leading to more informed and improved clinical hypoth-eses. Coupling TMAs with high-throughput and objective biomarker expression ana lysis with robust imaging technologies has led to further improvement in biomarker data that will ensure the continued use of TMAs, especially in the translational setting.
TMA constructionPlacement of multiple tissues from multiple patients/organs in a single block/slide were described first in 1986 by Hector Battifora who termed it the ‘sausage’ tissue block [4], fol-lowed by description of another approach by
Wan et al. in 1987 [5], but the process was labo-rious and time consuming. The use of TMAs did not become popular until Kononen et al. developed a mechanized version of TMA con-struction [6]. The use of TMAs has increased since then, with over 800 TMA-related pub-lications in 2011 [7]. TMA construction has been reviewed previously [2,3,7–9]. However, in brief, TMAs are constructed by first selecting formalin-fixed paraffin-embedded (FFPE) tissue blocks of interest to be included in the TMA block to be constructed. Once cases have been selected, a pathologist reviews a hematoxylin and eosin (H&E) section from each sample block to identify (i.e., place circles around) regions of interest. For example, for tumors, the patholo-gist can select areas of invasive or noninvasive tumor (i.e., ductal carcinoma in situ in breast cancer), adjacent normal tissue, stroma, or all three, differentially, depending on the intended application. The circled H&E is then used as a guide to identify the coring location on the donor block. Once the core has been removed from the donor block, it is placed in the recipi-ent block in a specific row/column orientation. Maintenance of this row/column information is absolutely vital for performing studies to enable linkage of pathology and clinical information [10]. Furthermore, although one to two cores can potentially be representative of the whole tumor [11,12] it is good practice to include multiple cores (two to four) [13] from a single case to not only ensure disease representation, but also to
The use of tissue microarrays (TMAs) in the preclinical and translational research settings has become ubiquitous as they allow for high-throughput in situ biomarker ana lysis of hundreds of patient samples, with time and cost efficiency. Coupled with advanced imaging and image-ana lysis technologies that allow for objective and standardized biomarker expression assessment, TMAs have become critical tools for the development and validation of clinically meaningful biomarker diagnostic assays. However, their diagnostic use in the clinical laboratory setting is limited due to the need for conventional whole-section tissue assessment used for routine diagnostic purposes. In this article, after reviewing TMA basics and their translational and clinical research applications, we will focus on the use of TMAs for robust assay development and quality control in the clinical laboratory setting, as well as provide insights into how TMAs may serve well in the clinical setting as assay performance and quantification controls.
KEYWORDS: assay development n biomarker n clinical application n diagnostic testing n digital pathology n quality control n tissue microarray
Mark D Gustavson1, David L Rimm2 & Marisa Dolled-Filhart*3
1Genoptix Medical Laboratory, Carlsbad, CA, USA 2Yale University School of Medicine, New Haven, CT, USA 3Clinical Development Laboratory, Merck & Company, Office: RY50-1E-144, Maildrop RY 50-100, 126 Lincoln Avenue, Rahway, NJ 07065, USA *Author for correspondence: Tel.: +1 732 594 0587 Fax: +1 732 594 7801 [email protected]
enable characterization of disease or biomarker heterogeneity, and to increase the likelihood of maintaining a large number of samples for ana-lysis due to differential core thickness [14]. Two comprehensive reviews by Rimm et al. [10] and Franco et al. [15] discuss, in detail, considerations for TMA construction and utility from selection of blocks to statistical ana lysis considerations.
�n Advantages & disadvantages of TMAsIn addition to the clear advantage of TMAs with respect to the amount of tissue used and thus preservation of valuable tissue due to the relatively small amount of tissue required for construction, TMAs have advantages in sev-eral other key areas including reproducibility, ana lysis time, costs and applicability. As TMAs contain small cores representing all samples on a single slide, assay conditions are uniform across all samples, leading to greater reproducibility of results and reduced assay ana lysis time than individual slide ana lysis of each sample, and reagent costs are kept at a minimum since only one (or few) slides need to be analyzed. Addi-tionally, tissue ana lysis methods that can be per-formed on whole tissue sections can be applied to TMAs, including immunohistochemistry (IHC) [16], immunofluorescence [17,18], FISH [19], chromogenic in situ hybridization [20], mRNA in situ hybridization [21,22] and miRNA in situ hybridixation [23]. Furthermore, recent advances have enabled efficient extraction of DNA and/or RNA from TMA cores, enabling TMA technology to be coupled with advanced molecular testing [24].
With respect to assay development, quality assurance and, ultimately, clinical adoption, the primary subject of this review, these advantages are critical. Implicit with these advantages, TMAs enable efficient ana lysis across a wide range of biomarker expression levels. Advance-ment of quantitative methods in tissue will rely on the rapid and efficient ana lysis of many samples of differential expression levels.
Although the use of TMAs in the clinical laboratory as tools for primary diagnosis will most likely never be realized, several applica-tions of TMAs from assay development through to quality assurance and assay control can be employed. Additionally, TMAs face the same challenges that FFPE-based sample ana lysis face, which includes how preanalytical variables (time to fixation, time in fixation and antigen oxidation/hydrolysis) affect sample quality, and, hence, ana lysis. However, TMAs are a
great tool for assessment of the specific effects of these variables.
TMA applicationsTMA applications are reviewed extensively in several published articles [8,15,25]. Although the goal of this review is to focus on the clinical labo-ratory application of TMAs, it is important to review some of the basic applications of TMAs. The most widely used application of TMAs is the large-scale cohort ana lysis of biomarker expression as a function of pathological characteristics and clinical outcome, looking at a single biomarker or a few biomarkers in a large number of cases [26] or biomarkers, based on literature searches [27].
However, as the genome and proteome become more studied and understood with ever-increas-ing complexity of regulation (e.g., miRNA), the importance of dissecting these pathways will become ever more critical for the development of meaningful biomarker assays and diagnostic tests. With this complexity comes the necessity for inte-gration of multiple high-throughput technologies, enabling a funnel-down approach to biomarker assay development. TMAs serve as an invalu-able tool where data from very high-throughput experiments (i.e., gene chip, mass spectrometry or reverse-phase protein arrays) can be validated in actual human tissue samples. A good example of this systems biology approach is provided by Faratian et al., where both in silico (mathemati-cal modeling) and reverse-phase protein array approaches were used to dissect receptor tyrosine kinase pathways to identify several critical path-way members [28]. They were then validated on a TMA cohort consisting of 122 breast cancer cases that had been treated with trastuzumab.
Large co-operative research groups frequently take advantage of TMAs by placing samples from their large clinical trial cohorts. There are several examples of this, including the Breast Cancer International Research Group use of large TMAs to assess accuracy and reproducibility in large-scale clinical trials [29] and the ATAC trial. The ATAC trial employed TMAs to examine prog-nostic significance in 1125 patient samples using the 21-gene recurrence assay (OncotypeDx®, Genomic Health, CA, USA) [30], as well as with a four-protein IHC assay [31]. The latter was sub-sequently validated on TMAs in the TEAM trial [32] and serves as an example of how validation of biomarker assays on TMAs translate to the clinical setting.
Beyond these typical oncology-focused applications, there have also been publications describing cell pellet-based TMAs [33], which have
Personalized Medicine (2013) 10 (5)442 future science group
Review Gustavson, Rimm & Dolled-Filhart
applications for screening of cell lines, and for control inclusion, biopsy TMAs (such as bone marrow core biopsies or prostate needle core biopsies, reviewed by Datta and Kajdocsy-Balla [34]) as well as nononcology applications such as TMA-based ana lysis of Alzheimer’s disease [35] and ana lysis of xenografts [36]. Of particular interest is construction of cell-line TMAs as a discovery tool, specifically for interrogation of pathways related to drug response, since cell lines can be treated and highly controlled [37]. Further, cell-line TMAs can serve as assay con-trols for either assurance of assay performance or as calibration controls (see ‘Future perspective’ section).
Requirement for automated & objective image-analysis toolsThe development of next-generation tissue-based diagnostic tests in the clinical laboratory will be critically reliant upon automated imaging plat-forms coupled with robust and objective image-analysis tools, both for the whole-tissue clinical sample as well as TMAs. Likewise, the enhanced adoption of TMAs in the clinical setting will also be reliant upon such objective tools since manual assessment of TMA data is not only time-consuming, but also highly subjective, leading to issues with reproducibility [38,39].
There has been significant growth of semi-automated and automated objective methods for scoring of tissue (and unique programs geared for scoring of TMAs as well) in order to reduce intra- and inter-observer variability, and improve repro-ducibility of biomarker ana lysis. There are a vari-ety of image-ana lysis software programs available for use in brightfield applications (H&E staining, chromagen-based IHC and chromogenic in situ hybridization, among others), as well as some for fluorescence applications (e.g., FISH and immuno fluorescence among others), with some systems including both the image scanning/digi-tization and the algorithms for scoring that are described in more detail in a review by Rojo et al. [40] and listed in Table 1. A small number of image-analysis algorithms, such as for estrogen recep-tor (ER), progesterone receptor and/or human EGF receptor 2 (HER2) ana lysis (by Aperio [CA, USA] or Ventana [CA, USA]) have been US FDA-cleared for use in breast cancer. A more complete assessment of digital microscopy solu-tions beyond image-analysis has been reviewed by Rojo et al. [41] and Mulrane et al. [42,43].
Critical to all image-analysis systems is the ability to identify and/or differentiate regions of interest (image segmentation). For example, for
assessment of ER expression, the tumor would have to be differentiated from surrounding nor-mal and stromal elements. Once these regions have been identified, additional ana lysis, includ-ing morphological (i.e., positive nuclear identi-fication) and protein expression ana lysis (IHC) can be performed. There are several different ways to accomplish image segmentation. The first, and most subjective, is to perform manual identification of regions by an operator circling those regions of interest through an imaging interface. The second is through feature-based or contextual compartmentalization whereby the image-analysis software is trained, based on morphological and/or histological features, to recognize areas of tumor and subsequent subcellular compartments (i.e., the nucleus). Three examples of this approach are Aperio’s PRECISION Image-analysis tools, Perkin Elmer’s (MA, USA) inFORM™ and Definiens (Germany) Tissue Studio™. The third approach is through molecular identification of regions of interest (e.g., cytokeratin to identify areas of epithelium and differentiate from stromal com-ponents). Molecular identification of regions of interest does not require user training, but rather relies on objective image-analysis to differentiate molecular signal from background to define a compartment of interest (i.e., tumor nuclei). An example of this approach is Genoptix’s AQUA® technology.
Additionally, adoption of TMAs into the clini-cal setting will require standardization of imaging
Table 1. Tissue image ana lysis software programs for immunohistochemistry- and/or immunofluorescence-stained slides.
Manufacturer (location) Image ana lysis software Ref.
3D Histech (Budapest, Hungary)
MIRAX HistoQuant™ [102]
Aperio (CA, USA) Aperio Image Analysis Toolbox/Spectrum™/PRECISON Image Analysis tools
[103]
Caliper (a Perkin Elmer company; MA, USA)
inForm™ [104]
DAKO (CA, USA) ACIS® III [105]
Definiens (Munich, Germany)
Tissue Studio® [106]
Hamamatsu (NJ, USA) NDP®.analyze [107]
Genoptix (CA, USA) AQUA® technology [108]
Leica Microsystems (Wetzlar, Germany)
Ariol®, Tissue IA 2.0 [109]
Ventana (AZ, USA) VIAS™ [110]
Visiopharm (Hoersholm, Denmark)
TissuemorphDP™ [111]
www.futuremedicine.com 443future science group
Tissue microarrays: leaping the gap between research & clinical adoption Review
and image-analysis such that assay results provide precise day-to-day and laboratory-to-laboratory results [44]. Without standardization, the use of TMAs and, thus, advanced assay technologies in the clinical laboratory will fall short.
Assay development�n Quantitative optimization
Tissue-based assay optimization is typically per-formed using a few (<10) representative whole-tissue samples to ensure assay specificity and sensitivity. This is typically a subjective process where the operator decides, by eye, the optimal assay conditions. However, with the advent of TMAs coupled with quantitative and objec-tive tissue-based methodologies, the ability to optimize assays objectively and quantitatively has been realized [Christiansen J et al. Quantita-
tive and objective methodologies for immunohisto-
chemistry assay development and quality control
(2013), Manuscript in preparation]. Although this approach was developed using AQUA technol-ogy, it is applicable to all quantitative and robust image-analysis approaches. As discussed earlier, the distinct advantage of TMAs is the ability to array hundreds of different tissue samples on a single slide, with these hundreds of samples representing not only a wide range of analyte expression, but also a range of tissue types and histologies. The goal of any assay optimization is to maximize specificity and sensitivity; in this application we strive to maximize the dynamic range of the assay between signal (sensitivity) and noise/background (specificity) such that the assay is in the linear range for as many samples as possible. In much the same way that quantita-tive immunoassays such as ELISA are optimized, the use of TMAs together with quantitative assay approaches enable the same process to be performed in tissue.
An idealized example is presented in Figure 1 where multiple assay conditions (i.e., antibody concentration for quantitative IHC assays) can be tested using multiple serial sections of the same TMA block [Christiansen J et al. Quantitative
and objective methodologies for immunohistochemis-
try assay development and quality control (2013), Manu-
script in preparation]. As TMAs represent a range of differentially expressing samples, a dynamic range can be calculated whereby the high-level expressing cases are compared with the low level or negative-expressing cases (red dashed line in Figure 1). By selecting the assay condition that pro-vides the optimal dynamic range, it is ensured that the assay will provide the greatest resolu-tion for quantitatively differentiating samples.
Interestingly, the assay condition that provides the visually ‘best’ condition is different (typically 0.5–2 logs differential in assay concentration) to that for the quantitatively ‘best’ condition. Furthermore, depending on TMA design, the data obtained can be further used for disease characterization.
�n Performance & precisionOnce assays have been objectively and quanti-tatively optimized, TMAs afford the ability to assess performance and precision of an assay since a population of samples can be assessed for day-to-day reproducibility with time and cost effi-ciency in a highly reproducible format. As dis-cussed earlier, the future use of TMAs will be for quantitative assays rather than categorical type assays (i.e., traditional IHC) and, as such, assess-ment of performance and precision assessment of tissue-based assays can be achieved on a more rigorous and quantitative level. With continuous ‘scale’-type data, linear regression ana lysis can be performed for assessment of assay performance through comparisons of correlation coefficients and slope of the line. For assay precision, the per-centage coefficient of variation can be determined across a single sample for multiple assays runs and days to provide an overall assay precision assess-ment. Performance and precision criteria can be established whereby assays can be qualified as strongly, moderately or weakly reproducing assays [Christiansen J et al. Quantitative and objective meth-
odologies for immunohistochemistry assay development
and quality control (2013), Manuscript in preparation] by assigning cut-offs to correlation coefficients (i.e., Pearson’s R >0.9 for strong performance), slope (i.e., 1.9–1.1 for strong performance) and the percentage coefficient of variation (i.e., <10% for strong precision).
�n Other TMA applications for assay validationTissue analytes are differentially susceptible to preanalytical variables that affect tissue-based assays. TMAs serve as great tools for assessment of analyte stability and provide an opportunity to develop novel biomarkers for assessment of the effects of preanalytical variables. TMAs can also be utilized for exploring the impact of dif-ferent tissue processing and fixation conditions, in addition to ana lysis of inter-laboratory and -observer variability [45–47]. Standardization of fixation conditions is imperative to ensure the highest quality and accurate assay data [48], and TMAs have been used to examine the effects of time to fixation [49,50]. This TMA tool can
Personalized Medicine (2013) 10 (5)444 future science group
Review Gustavson, Rimm & Dolled-Filhart
be used to validate assays for analyte stability as a function of time to fixation, as well as pro-vide the opportunity to develop normalization and/or go/no-go controls for clinical labora-tory tissue samples, for general or specific assay requirements.
Furthermore, the quality of a TMA slide (or any FFPE sample slide) may decrease due to the time between FFPE sample sectioning and stain-ing, which can result in potential false negatives [51–53], but not due to long-term storage of samples in paraffin blocks [54]. However, some studies have found that while the percentage of positive cases decreases, most associations were found in both fresh and older sections [55]. This also may impact ana lysis of mRNA in situ hybridization where there may be issues with sensitivity of the probe based on sample and/or section age, and RNA quality, but is less of a concern as far as ana-lysis of DNA by FISH, since DNA is more stable, although protocol enhancements for TMAs may be utilized [56,57].
Quality assuranceFor the same reasons TMAs are great tools for assay development and validation, they are starting to become well-used tools for clinical laboratory quality assurance [58–61] by providing a much simpler means to assess intra- and inter-laboratory reproducibility on both a qualitative (positive/negative) and potentially quantitative level. The College of American Pathologists [101] is implementing the use of TMAs for labo-ratory proficiency testing. A specific study by Fitzgibbons et al. [60] and highlighted by the Col-lege of American Pathologists demonstrates the use of TMAs to assess reproducibility of HER2 IHC staining and interpretation, as well as FISH testing across 243 laboratories. Each laboratory received 80 cases across eight TMA blocks to stain by HER2 IHC and interpret using their own standard procedures for determination of HER2 positive/negative status. The study showed that “70% of the samples had 90% concordance among laboratories, suggesting [TMA’s] potential utility as consensus specimens for methodologic validation and controls” [60].
Another study by Terry et al. used TMAs as part of an implementation of the Canadian External Quality Assurance program for breast cancer testing [61]. The study included TMAs with 38 independent samples tested across 18 laboratories. The results of the study showed that laboratories had a strong positive agree-ment (>0.8 by kappa interobserver statistics) in 85% of cases. One of the advantages of TMAs,
as pointed out by this study, is rapid testing turnaround time and bioinformatics (statistical ana lysis).
The limitation in using TMAs for quality assurance purposes in the clinical laboratory is that it does not provide the real-world variabil-ity of whole-tissue section (WTS) ana lysis since comparison of single cores usually only entails evaluation of a single 20× field-of-view (FOV), whereas WTS evaluation can sometimes have hundreds of FOVs for evaluation. Typically, a pathologist will not evaluate all FOVs for a given clinical specimen, thereby introducing variance in terms of laboratory-to-laboratory assessment of WTS. Thus, although TMAs are a great tool for overall assessment of laboratory performance for a given assay or test, proficiency testing may also need to be performed using WTS.
Next-generation applications of TMAs in the clinical laboratoryTMAs have been established as a robust and dynamic research tool, especially in the transla-tional space, where the ability to assay hundreds of cases on a single slide is advantageous both in terms of time and cost savings, but also as an analytically superior method. Their use and adoption in the clinical laboratory, however, has
Dyn
amic
ran
ge
rati
o
Concentration (µg/ml)
Qu
anti
tati
ve s
core
100 33 10 13 0.33 0.1 0.033 0.01
16,000
14,000
12,000
10,000
8000
6000
4000
2000
0 0
2
4
6
8
10
12
14
16
18
High expression
Low expression
Dynamic range
Figure 1. Idealized example of using tissue microarrays to quantitatively optimize tissue-based biomarker assays. Quantitative assay data (primary y-axis) is compared across multiple assay conditions (x-axis) to determine the optimal dynamic range (secondary y-axis). The ‘low expression’ curve (light blue line) describes the background or low expression, and the ‘high expression’ curve (dark blue line) describes the maximum signal scored in the tissue microarrays. The ‘dynamic range’ curve is shown in red. The dashed circle represents the optimal assay condition.
www.futuremedicine.com 445future science group
Tissue microarrays: leaping the gap between research & clinical adoption Review
been quite limited, most likely due to TMAs not being sufficient for making a clinical diagnosis, since there is not enough tissue to ascertain con-textual features of diseased tissue or to capture the heterogeneity within a sample. Nonetheless, by taking advantage of the reproducibility, as well as time– and cost–effectiveness, TMAs serve well as tools to assess a clinical laborato-ry’s overall performance with respect to specific assays.
However, further adoption of TMAs into the clinical laboratory is likely to be driven by and, to some extent, dependent upon, two main fac-tors. First, the increasing need for and clinical utility of truly quantitative diagnostic tissue-based assays and, second, the adoption of digital pathology and robust objective image-analysis tools for assessment and quantification of assay results. Both of these factors have been reviewed here as drivers for TMA use due to the necessity
for cost- and time-effective robust assay and imaging controls.
�n TMAs as quantitative assay controls in the clinical laboratoryAs tissue-based assays become more quantitative and as these assays begin to be adopted into the clinical laboratory for objective assessments, there will be a critical need for quantitative assay con-trols as differentiated from qualitative controls (simply positive/negative). As outlined above, a distinct advantage of TMAs is that they provide a methodology for the assessment of tens to hun-dreds of samples that represent a wide range of expression and/or or pathological/clinical char-acteristics. In order to control for a quantitative assay, a wide range of expression levels will need to be analyzed. A simple positive or negative control will not suffice for a truly quantitative assay. For example, for immunoassays such as ELISA, a range of controls is used to establish a standard curve. As a first approximation, TMAs of small sample size representing a range of biomarker expression could be used as simple quantitative assay controls (Figure 2). The TMA could be placed on the actual clinical sample slide to not only serve as an assay control, but also as a reagent control to ensure equivalent and appropriate treatment of the individual clinical sample. The TMA samples could either be tissue or clonal cell lines. The advantage of clonal cell lines is the homogeneity of expression compared with tissue. Additionally, the use of pure protein is also a viable option.
This approach is outlined in Figure 2. As one level of control, these TMAs could serve as go/no-go decision points for the assay and/or individual clinical samples. Based on established criteria (e.g., correlation coefficient and slope of the line) compared with a reference standard, a decision could be made with respect to the validity of the assay. Two idealized examples are provided in Figures 2b & 2C as successful and failed assay runs, respectively. Notice for the successful assay, the correlation coefficient and slope both approach 1, indicating strong performance of the assay with respect to established data. However, for the failed control, although the correlation coefficient approaches 1, the slope of the line is 0.5, indicating a fundamental assay perfor-mance error, resulting in much higher scores than expected. The use of the clinical sample data could thus result in a false positive. As a second level of control and in the absence of a standardized imaging platform, this type of con-trol could also be used as a normalization control
25,000
20,000
15,000
10,000
5000
0
0
0
5000
10,0
00
15,0
00
20,0
00
25,0
00y = 0.9301x + 205.62R2 = 0.9682
y = 0.5311x + 440.13
R2 = 0.9259
Expression score (current run)
0
5000
10,0
00
15,0
00
20,0
00
25,0
00
30,0
00
35,0
00
Expression score (current run)
Exp
ress
ion
sco
re (
refe
ren
ce)
25,000
20,000
15,000
10,000
5000
Exp
ress
ion
sco
re (
refe
ren
ce)
Sample/controlslide
Tissue
Figure 2. Using tissue microarrays as assay controls for clinical sample testing. (A) Cartoon depiction of the use of a small assay control tissue microarray on the same slide as the clinical tissue sample, allowing for complete assay control. (B & C) Idealized regression ana lysis from tissue microarray control in (A) use of a sample/control slide as a reference standard, with data shown in Y-axis for (B) a successfully run assay (data plotted on x axis), versus (C) a failed assay (data plotted on x-axis). Notice the slope and R2 values in (B) are approaching 1, whereas although the R2 for the failed assay is strong, the failure is due to a complete quantitative shift in the assay, as depicted by the slope of the line.
Personalized Medicine (2013) 10 (5)446 future science group
Review Gustavson, Rimm & Dolled-Filhart
whereby the equation of the line could be used to normalize the clinical sample score to a reference standard, provided the performance of the assay was within acceptable limits.
�n TMAs as quantitative calibration controls in the clinical laboratoryThe final level of control for tissue-based assays would be a control that enabled calibration to an
absolute concentration of the analyte based with respect to a gold standard. In tissue, directly obtaining an in situ absolute level of an analyte is theoretically not possible without grinding the tissue up to perform an in vitro assay such as an ELISA. However, by coupling tissue and/or cell-line TMAs with in vitro assays such as ELISA, western blot or reverse transcriptase-PCR, a calibration or standard curve could be derived
Highest negative control case Lowest positive control case
00
Figure 3. Example of using cell lines and tissue microarrays to construct a standard curve for the conversion of relative expression scores into absolute concentrations. (A) Depicts correlating western blot ana lysis of protein concentrations in cell lines with (C) quantitative immunofluorescence of cell lines in tissue microarray format (AQUA® scores [y-axis] and pg ER/µg total protein [x-axis]). With this correlation, an equation was derived, allowing for the conversion of tissue-based AQUA scores into protein concentrations. (D) Using test arrays, the lower limit of quantification was determined at 2 pg ER/µg total protein. (E) Visual confirmation of lower limit of quantification. Scale bars represent 100 µm. CK: Cytokeratin; DAPI: 4’,6-diamidino-2-phenylindole; ER: Estrogen receptor; rER: Recombinant estrogen receptor. Reproduced with permission from [62].
www.futuremedicine.com 447future science group
Tissue microarrays: leaping the gap between research & clinical adoption Review
such that assay scores from individual clinical tissue samples could be converted to absolute concentration or levels. Regarding blood-based clinical assays (i.e., glucose or cholesterol), there is an expectation of not only obtaining a quanti-tative objective number (i.e., concentration), but also that this number is commonly understood and standardized. This is not the case currently with tissue-based assays; however, the future of clinical tissue assays and TMAs could be the means to drive this achievement.
This concept was demonstrated by Welsh et al. (Figure 3) [62] and outlined in Figure 4. Welsh et al. used commercially available and engineered cell lines with increasing calibrated concentrations of estrogen receptor (a critical biomarker for the effective treatment of breast cancer [63]). With known concentrations (as determined by quantitative western blotting; Figure 3a), expression scores for these cell lines were calculated using AQUA technology in conjunction with a cell-line TMA (Figure 3b) and subsequently correlated to absolute protein concentration to derive a standard curve with a conversion equation (Figure 3C). Welsh et al. went on to determine the lower limit of quantifica-tion using a test series of tissues (Figure 3D) and validated their findings with respect to breast cancer outcome.
ConclusionTMAs are a potentially valuable tool that have grown in popularity over the last two
decades, especially in preclinical and transla-tional research. Clinical adoption of TMAs will occur on three levels. The first level is for assay development and validation, where TMAs can be used as initial tools to optimize and vali-date clinically relevant assays. The second level is quality assurance where TMAs are starting to be used to assess intra- and inter-laboratory assay reproducibility. The third and final level will be development of robust and quantita-tive assay controls. Driving clinical adoption of TMAs will depend on the development of more quantitative tissue-based assays that are clinically relevant, as well as adoption of digital pathology and objective image-analysis tools. Taken together, the use of TMAs in the clini-cal laboratory will become ever more valuable in the future as the science and technology of tissue-based assays progresses.
Future perspectiveAs TMAs have become the predominant tool in the translational and clinical research labora-tory for high-throughput assessment and inter-pretation of biomarker expression, we anticipate that TMAs will become more prevalent in the clinical laboratory space. This will coincide with the development of quantitative in situ assays and technologies that require more quan-titative-type controls that mimic the real-world clinical sample or tissue. It is this use that will enable TMAs to gain prominence in the clinical laboratory. Additionally, if measuring analytes in situ is to become like measuring analytes in blood or serum, sample-relevant quantitative calibrators will also have to be developed and the easiest medium for this development will be TMAs. Furthermore, due to their ease and effi-ciency of use, TMAs should gain adoption even for simpler assays (i.e., chromogen-based IHC) as laboratory-to-laboratory quality control tools that will enable the assurance of reproducible clinical results critical for the accurate diagnosis of patients.
Financial & competing interests disclosureMD Gustavson and DL Rimm are an employee and paid consultant, respectively, of Genoptix Medical Laboratory, which is the sole licensee of AQUA® technology, which is referenced and discussed in this review. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
Ass
ay s
core
30,000
25,000
20,000
15,000
10,000
5000
00.1 1 10 100 1000 10,000
LLQ
ULQ
Theoretical example
Quantitative assaycalibration slide
Dec
reas
ing
prot
ein
conc
entr
atio
n
Analyte concentration (pg/ml)
Figure 4. Using tissue microarrays as quantitative assay calibration controls for clinical sample testing. (A) Cartoon depiction of the use of a small assay calibration control tissue microarray representing samples of known increasing protein concentrations (cell lines or protein spots). This quantitative assay calibration would be run with each assay for each marker to convert clinical sample assay scores into absolute analyte concentrations using a standard curve (B). Derivation of the standard curve would enable analyte-specific LLQs and ULQs to define the linear range of the assay. LLQ: Lower limit of quantification; ULQ: Upper limit of quantification.
Personalized Medicine (2013) 10 (5)448 future science group
Review Gustavson, Rimm & Dolled-Filhart
Executive summary
Background � Tissue microarrays (TMAs) have become ubiquitous in the research setting for high-throughput in situ biomarker ana lysis due to the ability to analyze hundreds of samples on a single slide with cost– and time–effectiveness.
� Development of digital imaging and image ana lysis have allowed for improvements in biomarker assessments in the translational setting.
� Clinical diagnostic use of TMAs is limited due to the limited sample size; however, they have been used for quality assurance purposes in the clinical setting, such as inter- and intra-laboratory concordance.
TMA applications � Methods that can be applied to whole-tissue sections are applicable to TMAs including DNA-, RNA- and protein-based assessments.
� The most prevalent use of TMAs is large-scale cohort ana lysis of biomarker expression in the context of pathological characteristics and/or clinical outcome.
� TMAs are subject to the same limitations of other sample analyses with regard to variable preanalytic conditions of the samples used.
The requirement for automated & objective image ana lysis tools � Further adoption of TMAs into the clinical setting will likely rely on objective ana lysis to reduce subjectivity and to produce time–effectiveness, as well as to provide standardization and precise results.
� There are a variety of image ana lysis software programs available for brightfield as well as fluorescence applications for TMA ana lysis; image ana lysis is typically performed using methods of manual identification, feature- or contextual-based compartmentalization using morphological features, and/or differentiation of specific regions/compartments of interest.
Assay development � Optimization of immunohistochemistry assays has typically been a subjective, manual process, which can be improved using objective and quantitative approaches with TMAs.
� Assessment of assay reproducibility can be performed rapidly and efficiently using TMAs for intra- and inter-laboratory comparisons.
Quality assurance � TMAs have been adopted for use in clinical laboratory proficiency testing by organizations such as the College of American Pathologists and the Canadian External Quality Assurance program for breast cancer testing.
� Limitations to TMAs do exist as they lack heterogeneity seen in whole-tissue sections, yet they serve as a useful tool for laboratory assessment purposes.
Next-generation applications of TMAs in the clinical laboratory � Further adoption of TMAs into the clinical laboratory is likely to be driven by adoption of both digital pathology and/or the use of quantitative tissue-based assays.
� TMAs can serve as useful tools for quantitative assay controls in the future due to their ability to represent a wide range of biomarker expression and/or tissue characteristics on a single slide.
� Use of TMAs with concentration-based controls would bring further objectivity and standardization to tissue-based biomarker assays.
ReferencesPapers of special note have been highlighted as:n of interestnn of considerable interest
1 Rimm DL, Camp RL, Charette LA, Costa J, Olsen DA, Reiss M. Tissue microarray: a new technology for amplification of tissue resources. Cancer J. 7(1), 24–31 (2001).
n� Reviews tissue microarray (TMA) construction.
2 Rimm DL, Camp RL, Charette LA, Olsen DA, Provost E. Amplification of tissue by construction of tissue microarrays. Exp. Mol. Pathol. 70(3), 255–264 (2001).
nn� Reviews the advantages of TMAs for biomarker discovery.
4 Battifora H. The multitumor (sausage) tissue block: novel method for immunohistochemical antibody testing. Lab. Invest. 55(2), 244–248 (1986).
5 Wan WH, Fortuna MB, Furmanski P. A rapid and efficient method for testing immunohistochemical reactivity of monoclonal antibodies against multiple tissue samples simultaneously. J. Immunol. Methods 103(1), 121–129 (1987).
6 Kononen J, Bubendorf L, Kallioniemi A et al. Tissue microarrays for high-throughput molecular profiling of tumor specimens. Nat. Med. 4(7), 844–847 (1998).
7 Camp RL, Neumeister V, Rimm DL. A decade of tissue microarrays: progress in the discovery and validation of cancer biomarkers. J. Clin. Oncol. 26(34), 5630–5637 (2008).
8 Dolled-Filhart MP, Gustavson MD. Tissue microarrays and quantitative tissue-based
image-analysis as a tool for oncology biomarker and diagnostic development. Expert Opin. Med. Diagn. 6(6), 569–583 (2012).
nn� Reviews TMA applications, specifically the use of TMAs in translational biomarker assay development and robust statistical modeling.
9 Hassan S, Ferrario C, Mamo A, Basik M. Tissue microarrays: emerging standard for biomarker validation. Curr. Opin. Biotechnol. 19(1), 19–25 (2008).
10 Rimm DL, Nielsen TO, Jewell SD et al. Cancer and Leukemia Group B Pathology Committee guidelines for tissue microarray construction representing multicenter prospective clinical trial tissues. J. Clin. Oncol. 29(16), 2282–2290 (2011).
11 Sauter G. Representativity of TMA studies. Methods Mol. Biol. 664, 27–35 (2010).
www.futuremedicine.com 449future science group
Tissue microarrays: leaping the gap between research & clinical adoption Review
449www.futuremedicine.com
12 Camp RL, Charette LA, Rimm DL. Validation of tissue microarray technology in breast carcinoma. Lab. Invest. 80(12), 1943–1949 (2000).
13 Goethals L, Perneel C, Debucquoy A et al. A new approach to the validation of tissue microarrays. J. Pathol. 208(5), 607–614 (2006).
14 Sauter G, Mirlacher M. Tissue microarrays for predictive molecular pathology. J. Clin. Pathol. 55(8), 575–576 (2002).
15 Franco R, Caraglia M, Facchini G, Abbruzzese A, Botti G. The role of tissue microarray in the era of target-based agents. Expert Rev. Anticancer Ther. 11(6), 859–869 (2011).
16 Simon R, Mirlacher M, Sauter G. Immunohistochemical ana lysis of tissue microarrays. Methods Mol. Biol. 664, 113–126 (2010).
17 Camp RL, Chung GG, Rimm DL. Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat. Med. 8(11), 1323–1327 (2002).
18 Dolled-Filhart M, Gustavson M, Camp RL, Rimm DL, Tonkinson JL, Christiansen J. Automated ana lysis of tissue microarrays. Methods Mol. Biol. 664, 151–162 (2010).
19 Kallioniemi A. DNA copy number ana lysis on tissue microarrays. Methods Mol. Biol. 664, 127–134 (2010).
20 Rummukainen JK, Salminen T, Lundin J, Joensuu H, Isola JJ. Amplification of c-myc oncogene by chromogenic and fluorescence in situ hybridization in archival breast cancer tissue array samples. Lab. Invest. 81(11), 1545–1551 (2001).
21 Veeck J, Dahl E. RNA expression ana lysis on formalin-fixed paraffin-embedded tissues in TMA format by RNA in situ hybridization. Methods Mol. Biol. 664, 135–150 (2010).
22 Bordeaux JM, Cheng H, Welsh AW et al. Quantitative in situ measurement of estrogen receptor mRNA predicts response to tamoxifen. PLoS ONE 7(5), e36559 (2012).
23 Hanna JA, Wimberly H, Kumar S, Slack F, Agarwal S, Rimm DL. Quantitative ana lysis of microRNAs in tissue microarrays by in situ hybridization. Biotechniques 52(4), 235–245 (2012).
24 Chung JY, Braunschweig T, Hewitt SM. Optimization of recovery of RNA from formalin-fixed, paraffin-embedded tissue.Diagn. Mol. Pathol. 15(4), 229–236 (2006).
25 Hewitt SM. Tissue microarrays as a tool in the discovery and validation of predictive biomarkers. Methods Mol. Biol. 823, 201–214 (2011).
nn� Reviews the use of TMAs as tools for the development and validation of predictive biomarkers.
26 Cantile M, Franco R, Tschan A et al. HOX D13 expression across 79 tumor tissue types.Int. J. Cancer 125(7), 1532–1541 (2009).
27 Schlomm T, Chun F, Erbersdobler A. From gene to clinic: TMA-based clinical validation of molecular markers in prostate cancer.Methods Mol. Biol. 664, 177–189 (2010).
28 Faratian D, Goltsov A, Lebedeva G et al. Systems biology reveals new strategies for personalizing cancer medicine and confirms the role of PTEN in resistance to trastuzumab. Cancer Res. 69(16), 6713–6720 (2009).
29 Press MF, Sauter G, Bernstein L et al. Diagnostic evaluation of HER-2 as a molecular target: an assessment of accuracy and reproducibility of laboratory testing in large, prospective, randomized clinical trials.Clin. Cancer Res. 11(18), 6598–6607 (2005).
30 Dowsett M, Cuzick J, Wale C et al. Prediction of risk of distant recurrence using the 21-gene recurrence score in node-negative and node-positive postmenopausal patients with breast cancer treated with anastrozole or tamoxifen: a TransATAC study. J. Clin. Oncol. 28(11), 1829–1834 (2010).
nn� Good example of the use of TMAs to study large clinical trial cohorts and the validation of robust prognostic tests.
31 Cuzick J, Dowsett M, Pineda S et al. Prognostic value of a combined estrogen receptor, progesterone receptor, Ki-67, and human epidermal growth factor receptor 2 immunohistochemical score and comparison with the Genomic Health recurrence score in early breast cancer. J. Clin. Oncol. 29(32), 4273–4278 (2011).
32 Christiansen J, Bartlett JM, Gustavson M et al. Validation of IHC4 algorithms for prediction of risk of recurrence in early breast cancer using both conventional and quantitative IHC approaches. ASCO Annual 2012 Meeting. Chicago, IL, USA, 30 May–3 June 2012 (Abstract 517).
33 Zhao S, Natkunam Y. Building ‘tissue’ microarrays from suspension cells. Methods Mol. Biol. 664, 93–101 (2010).
35 Kellner A, Matschke J, Bernreuther C, Moch H, Ferrer I, Glatzel M. Autoantibodies against beta-amyloid are common in Alzheimer’s disease and help control plaque burden. Ann. Neurol. 65(1), 24–31 (2009).
36 Bubendorf L, Kolmer M, Kononen J et al. Hormone therapy failure in human prostate cancer: ana lysis by complementary DNA and tissue microarrays. J. Natl Cancer Inst. 91(20), 1758–1764 (1999).
37 Braunschweig T, Chung JY, Hewitt SM. Tissue microarrays: bridging the gap between research and the clinic. Expert Rev. Proteomics 2(3), 325–336 (2005).
38 Paik S, Bryant J, Tan-Chiu E et al. Real-world performance of HER2 testing – National Surgical Adjuvant Breast and Bowel Project experience. J. Natl Cancer Inst. 94(11), 852–854 (2002).
39 Perez EA, Suman VJ, Davidson NE et al. HER2 testing by local, central, and reference laboratories in specimens from the North Central Cancer Treatment group N9831 intergroup adjuvant trial. J. Clin. Oncol. 24(19), 3032–3038 (2006).
40 Rojo MG, Bueno G, Slodkowska J. Review of imaging solutions for integrated quantitative immunohistochemistry in the pathology daily practice. Folia Histochem. Cytobiol. 47(3), 349–354 (2009).
41 Rojo MG, Garcia GB, Mateos CP, Garcia JG, Vicente MC. Critical comparison of 31 commercially available digital slide systems in pathology. Int. J. Surg. Pathol. 14(4), 285–305 (2006).
42 Mulrane L, Rexhepaj E, Penney S, Callanan JJ, Gallagher WM. Automated image-analysis in histopathology: a valuable tool in medical diagnostics. Expert Rev. Mol. Diagn. 8(6), 707–725 (2008).
nn� Describes the importance of digital pathology and imaging ana lysis for the advancement of medical diagnostics.
43 Mulrane L, Rexhepaj E, Smart V et al. Creation of a digital slide and tissue microarray resource from a multi-institutional predictive toxicology study in the rat: an initial report from the PredTox group. Exp. Toxicol. Pathol. 60(4–5), 235–245 (2008).
44 Gustavson MD, Bourke-Martin B, Reilly D et al. Standardization of HER2 immunohistochemistry in breast cancer by automated quantitative ana lysis. Arch. Pathol. Lab. Med. 133(9), 1413–1419 (2009).
45 Mengel M, von Wasielewski R, Wiese B, Rüdiger T, Müller-Hermelink HK, Kreipe H. Inter-laboratory and inter-observer reproducibility of immunohistochemical assessment of the Ki-67 labelling index in a large multi-centre trial. J. Pathol. 198(3), 292–299 (2002).
46 De la Taille A, Viellefond A, Berger N et al. Evaluation of the interobserver reproducibility of Gleason grading of prostatic adenocarcinoma using tissue microarrays. Hum. Pathol. 34(5), 444–449 (2003).
47 Parker RL, Huntsman DG, Lesack DW et al. Assessment of interlaboratory variation in the immunohistochemical determination of
Personalized Medicine (2013) 10 (5)450 future science group
Review Gustavson, Rimm & Dolled-Filhart
estrogen receptor status using a breast cancer tissue microarray. Am. J. Clin. Pathol. 117(5), 723–728 (2002).
48 Yaziji H, Taylor CR, Goldstein NS et al. Consensus recommendations on estrogen receptor testing in breast cancer by immunohistochemistry. Appl. Immunohistochem. Mol. Morphol. 16(6), 513–520 (2008).
49 Neumeister VM, Sullivan CA, Lindner R et al. Hypoxia-induced protein CAIX is associated with somatic loss of BRCA1 protein and pathway activity in triple negative breast cancer. Breast Cancer Res. Treat. 136(1), 67–75 (2012).
50 Neumeister VM, Anagnostou V, Siddiqui S et al. Quantitative assessment of effect of preanalytic cold ischemic time on protein expression in breast cancer tissues. J. Natl Cancer Inst. 104(23), 1815–1824 (2012).
51 Jacobs TW, Prioleau JE, Stillman IE, Schnitt SJ. Loss of tumor marker-immunostaining intensity on stored paraffin slides of breast cancer. J. Natl Cancer Inst. 88(15), 1054–1059 (1996).
52 Divito KA, Charette LA, Rimm DL, Camp RL. Long-term preservation of antigenicity on tissue microarrays. Lab. Invest. 84(8), 1071–1078 (2004).
53 Bertheau P, Cazals-Hatem D, Meignin V et al. Variability of immunohistochemical reactivity on stored paraffin slides. J. Clin. Pathol. 51(5), 370–374 (1998).
54 Manne U, Myers RB, Srivastava S, Grizzle WE. Re: loss of tumor marker-immunostaining intensity on stored paraffin slides of breast cancer. J. Natl Cancer Inst. 89(8), 585–586 (1997).
55 Mirlacher M, Kasper M, Storz M et al. Influence of slide aging on results of translational research studies using
56 Chin SF, Daigo Y, Huang HE et al. A simple and reliable pretreatment protocol facilitates fluorescent in situ hybridisation on tissue microarrays of paraffin wax embedded tumour samples. Mol. Pathol. 56(5), 275–279 (2003).
57 Andersen CL, Hostetter G, Grigoryan A, Sauter G, Kallioniemi A. Improved procedure for fluorescence in situ hybridization on tissue microarrays. Cytometry 45(2), 83–86 (2001).
58 Gown AM. Current issues in ER and HER2 testing by IHC in breast cancer. Mod. Pathol. 21(Suppl. 2), S8–S15 (2008).
59 Wasielewski R, Hasselmann S, Ruschoff J, Fisseler-Eckhoff A, Kreipe H. Proficiency testing of immunohistochemical biomarker assays in breast cancer. Virchows Arch. 453(6), 537–543 (2008).
60 Fitzgibbons PL, Murphy DA, Hammond ME, Allred DC, Valenstein PN. Recommendations for validating estrogen and progesterone receptor immunohistochemistry assays. Arch. Pathol. Lab. Med. 134(6), 930–935 (2010).
nn� Outlines the use of TMAs for laboratory quality assurance processes in immunohistochemistry.
61 Terry J, Torlakovic EE, Garratt J et al. Implementation of a Canadian external quality assurance program for breast cancer biomarkers: an initiative of Canadian Quality Control in immunohistochemistry (cIQc) and Canadian Association of Pathologists (CAP) National Standards Committee/ Immunohistochemistry. Appl. Immunohistochem. Mol. Morphol. 17(5), 375–382 (2009).
62 Welsh AW, Moeder CB, Kumar S et al. Standardization of estrogen receptor measurement in breast cancer suggests false-negative results are a function of
threshold intensity rather than percentage of positive cells. J. Clin. Oncol. 29(22), 2978–2984 (2011).
n� Study showing the next-generation use of TMAs as quantitative controls for the assessment of absolute protein levels in tissue.
63 Hammond ME, Hayes DF, Dowsett M et al. American Society of Clinical Oncology/College of American Pathologists Guideline Recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. J. Clin. Oncol. 28(16), 2784–2795 (2010).
�n Websites101 The College of American Pathologists.
www.cap.org
102 3DHISTECH Kft. www.3dhistech.com
103 Aperio ePathology Solutions. www.aperio.com
104 Caliper. www.caliper.com
105 Dako, an Agilent Technologies company. www.dako.com
