1. Introduction
A variety of powerful high-throughput technologies are available for the identification of potentially relevant genes in individual tissue samples. These tools generally identify more interesting genes than the resources that are available for their further analysis. In situ analyses methods are optimally suited for studying gene alterations. Technologies such as immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), and to some extent also RNA in situ hybridization (RNA ISH) allow a cellular and subcellular localization of the gene product of interest and a clear distinction between diseased and nondiseased cells. However, traditional slide-by-slide analysis cannot match the speed of modern high-throughput target-identification technologies. Tissue microarrays (TMAs) permit in situ analysis of hundreds of tissue samples on one microscope glass slide. Since studies have shown that minute tissue samples as small as 0.6 mm in diameter are sufficiently representative of their donor tissues (reviewed in Sauter et al., 2003), TMAs have become a widely used research tool.
2. TMA technology
Hundreds of cylindrical samples measuring up to 4 mm (usually 0.6 mm) in diameter are removed from formalin-fixed or frozen tissues and placed into one recipient block (Kononen et al., 1998). Simple commercial or homemade arrayers can be used for this purpose. Sections from such TMA blocks containing hundreds of different tissues are then placed on a glass slide. All methods for in situ tissue analysis can be performed on TMAs using similar protocols as for conventional large sections (Figure 1).
3. TMA applications
After its initial description in 1998 (Kononen et al., 1998), TMAs have become widely used in recent years. More than 150 publications published in 2003 have utilized TMAs. Some of the most powerful applications of TMAs are described below.
Figure 1 Tissue microarrays and TMA applications. (a) TMA block; (b) hematoxylin and Eosin (H&E)-stained section (5 |m) of the TMA; (c) magnification of a H&E-stained tissue spot (diameter 0.6 mm); (d) immunohistochem-istry of a breast cancer tissue spot. Brownish membranous staining indicates strong expression of the Her2 receptor protein. (e) FISH analysis. Magnification (1000x) of cell nuclei (blue staining) showing two green signals corresponding to the centromere of chromosome 17 and multiple red signals indicating Her2 gene amplification. (f) RNA in situ hybridization on a TMA spot from frozen tissue using an radioactively labeled oligonucleotide probe
3.1. TMAs for molecular epidemiology
Data on the expression of a given gene are often available from the literature. However, examples of extensively analyzed genes suggest that published data often lack practical value. For example, HER2, the target gene for Trastuzumab (Herceptin) has been analyzed in thousands of publications. Yet, the data are highly controversial. Published expression frequencies range from <10 to >90% for many important tumor types (reviewed in Sauter et al., 2003). The paramount impact of experimental parameters such as antibody selection, IHC protocol, tissue characteristics, and scoring criteria on the results of IHC analyses greatly reduce the comparability of IHC data derived from different groups. TMAs containing samples from all different tumor types make it possible to analyze a gene of interest under fully standardized conditions in one experiment. For example, we used a set of TMAs containing more than 3500 different samples from more than 120 different tumor categories to investigate target genes for established drug targets such as KIT/CD117 (Imatinib; Glivec) (Went et al., 2004) or epidermal growth factor (EGFR) (Tarceva, Iressa, Cetuximab, Erbitux) (Sauter et al., 2003). Such studies often show expression of a target gene in a wide variety of tumor entities. The analysis of a representative number of cases from various tumor types typically results in a representative ranking list of marker expression in different tumor entities (Sauter et al., 2003).
3.2. TMAs for prognosis evaluation
TMAs containing a large number of samples from one tumor type with attached clinico-pathological information allow an estimation of the biological importance of a gene alteration. Associations with advanced stage, high grade, presence of metastases, or poor clinical outcome argue for a role in tumor progression. TMAs are highly efficient in identifying associations between molecular features and prognosis. For example, significant associations were found between estrogen or progesteron expression (Torhorst et al., 2001) or HER2 alterations (Barlund et al., 2000) and survival in breast cancer patients, between vimentin expression and prognosis in kidney cancer (Moch et al., 1999), and between Ki67 labeling index and prognosis in urinary bladder cancer (Nocito et al., 2001), soft tissue sarcoma (Hoos et al., 2001a), or Hurthle cell carcinoma (Hoos et al., 2002).
3.3. TMAs for normal tissue evaluation
The expression pattern of genes of interest in normal tissues is often important. Often it allows conclusions on the potential biological role of a gene product. Normal tissue analysis is critically important for potential drug targets. Expression of a target gene in important normal cell types may predict the site of potential side effects. Cell type-specific expression analysis would not be possible if other methods than in situ techniques were used. For example, EpCam, a target for several anticancer therapies, is expressed in bile ducts of the liver. Since bile ducts constitutes are a very small (<1%) component of the liver, EpCam analyses by non-in situ techniques would suggest either no or low expression in liver (Figure 2). IHC analysis, however, reveals high EpCam levels in a small but vital liver compartment.
Figure 2 EpCam expression in a tissue spot from normal liver. Inset: 1000x magnification. EpCam expression if confined to small areas of bile ducts. A similar staining intensity, however, can be seen as well in cancer cells. This emphasizes the importance of in situ analyses for normal tissue evaluation. The use of disaggregated tissue for such an expression comparison (like Northern or Western blots) would suggest a negligible low expression level in normal liver
3.4. Important technical aspects
Although the technology is simple and easily applicable, some aspects of TMAs are controversially discussed and others are underestimated. Understanding the impact of tissue heterogeneity, the potential for automation, and the importance of pathology expertise is particularly important for the successful use of TMAs.
4. Tissue heterogeneity
The utility of TMAs for reliably finding associations between molecular and clinico-pathological features is the breakthrough discovery of the TMA field. Placing multiple tissues into one paraffin block has been described before (Wan et al., 1987). At this time, it was found that molecular analysis of small tissue samples would be insufficient for the identification of clinically relevant information. Technical improvements, therefore, aimed at the deposition of as large as possible tissues on multitissue blocks (Battifora, 1986). Remarkably, all studies comparing molecular features obtained on TMAs (with a spot diameter of 0.6 mm) with clinical or pathological data found the expected associations if these were previously well established (reviewed in Sauter et al., 2003). Results of these studies considerably weaken the practical importance of studies showing an increasing concordance between TMA data and corresponding large section results if 3-4 tissue samples are analyzed per tumor (Rubin et al., 2002; Engellau et al., 2001; Camp et al., 2000; Fernebro et al., 2002; Hoos et al., 2001b). It is obvious that TMA results will match large section results more closely if the highest possible number of cores per tumor block is analyzed. However, as shown in a p53 analysis in breast cancer, this has also disadvantages. Notably, the risk of false positivity is also increasing with the size of analyzed tissue. Torhorst et al. (2001) found p53 positivity to be prognostically relevant in four independent TMAs composed of one core each of >500 breast cancers but not in corresponding large sections. The use of TMAs containing only one core per tumor is therefore recommended for the vast majority of applications.
5. Pathology expertise and automation
Although TMA technology is often discussed in the context of other array techniques such as DNA or protein arrays, for which array construction, analysis, and data recording can easily be automated, the TMA method is fundamentally different. TMAs represent the ultimate miniaturization of molecular pathology and share the inherent strengths and limitations of pathology analyses. The level of pathology expertise is most decisive for the success of TMA studies, while classical array tools like automation are less relevant. Automated TMA manufacturing lacks importance, because sufficiently large ready-to-use “tissue libraries” are lacking.
Classical pathology skills are critical for the classification of arrayed tissues, experimental design, and staining interpretation. More than 100 studies can be based on one TMA block. As the molecular data will be compared with pathological information, the quality of “basic pathology data” is of highest importance. The difficulties related to the development of an adequate IHC protocol are greatly underestimated. Nonspecific positivity is a frequent problem and requires the use of multiple controls. A significant fraction of antibodies cannot be optimized for use on tissue sections. In general, IHC staining results are greatly dependent on antibody selection, antigen retrieval strategy, staining protocol, and on minor variables such as the section age (Mirlacher et al., 2004). For example, the use of three different antibodies for EGFR resulted – after state-of-the-art protocol development – in an up to fivefold difference in the rate of positivity (Sauter et al., 2003).
These inherent difficulties reduce the importance of automated TMA analyses. Although studies have demonstrated the feasibility of automated quantitation of brightfield or fluorescence signals on TMAs, automated systems have limited capability for recognition of staining artifacts, necrotic, crushed or damaged tissue elements, or for distinction of neoplastic versus nonneoplastic cells. Nevertheless, modern imaging tools have high value for image documentation and preliminary screening of staining results in a high-throughput scale. At the same time, manual (visual) analysis of the TMA staining by an experienced pathologist provides the greatest level of reading precision and will remain the gold standard for TMA staining interpretation. Attempts to improve automated TMA analysis are paralleled by an increased use of lysate arrays composed of protein extracts from tumor tissues (Nath and Chilkoti, 2003; Nishizuka et al., 2003). These are spotted on glass slides in a DNA array-like manner and can be analyzed with the same software used for DNA arrays. If morphologic information is not needed, lysate arrays may represent a better option than complicated and expensive systems for automated TMA analysis.
