| Home > Zentren > Zentrum für Diagnostik > Institut für Pathologie > Tissue Microarrays (TMA)
Advances in molecular medicine and biology have fundamentally changed biomedical research strategies. The rate of discovery of new genes involved in cancer and other diseases has increased heavily. Evaluation of the clinical importance of promising candidate genes is optimally performed on histological sections of diseased and non-diseased patient tissues. This not only allows a distinction of positive and negative cases, but also a definition of positive cell types and the subcellular localization of the gene product in question. However, using traditional methods of molecular pathology, such studies would lead to a massive workload in involved laboratories. This is not only because of the high number of candidate genes but also because hundreds or better thousands of tissue samples, optimally with attached follow-up data, must be analyzed in order to yield statistically sound results. Moreover, the analysis of multiple genes resulted in a critical loss of precious tissue material since the number of conventional tissue sections that can be taken from a tumor block does usually not exceed 200 - 300.
To overcome the limitations of conventional tissue analysis, we developed the tissue microarray (TMA) technology in a collaborative project with the laboratory of Olli Kallioniemi (at that time located at the National Human Genome Research Institute in Bethesda).
Preparatory work
It is important to realize that most of the work (approximately 95%) devoted to TMA manufacturing is traditional pathology work that cannot be accelerated by machinery. This preparatory work is similar to what is needed for traditional studies involving "large" tissue sections, apart from the number of tissues involved, which can be an order of magnitude higher in TMA studies than in traditional projects. The preparatory work includes:
1. Identification of potentially suited tissues.
2. Review of all sections from all candidate specimens to select the optimal slide.
3. Definition of the structure (outline) of the TMA, including generation of a "punch file" that contains the identification numbers of the tissues together with their locations and real coordinates.
Recipient block preparation (paraffin embedded tissues)
In contrast to normal paraffin blocks, tissue microarray blocks are cut at room temperature. Therefore, a special type of paraffin is recommended with a melting temperature between 55 and 58°C ("Peel-A-Way" paraffin, Polysciences Inc., PA, USA). The paraffin is melted at 60°C, filtrated, and poured in a stainless steel mold (e.g. 30 x 45 x 10 mm). A slotted plastic embedding cassette (as used in every histology lab) is then placed on the top of the warm paraffin.
TMA manufacturing equipment
We use multiple versions of simple homemade arraying instrument of which we feel that is significantly more effective (and cheaper) than any existing commercial instrument for both paraffin embedded and frozen tissues. Our instruments main components are an automated XY stage and a drill replacing the initially suggested needle for preparing holes in the recipient block. Improved tissue arraying devices can be constructed from existing commercial arrayers (home made upgrade) or de novo. In addition we have manufactured a variety of small gadgets facilitating the arraying process including needles and multiblock holders. For information on our TMA instrument upgrade program (Taucherinsel-Neuenburg@t-online.de).
We do not recommend the use of automated tissue arrayers as we strongly feel that these machines are:
This recommendation is based on our experience in the manual arraying of more than 20'000 tissues in >700 TMA blocks (>350'000 punches) and our interaction with users of automated arrayers. 
Bild anzeigen
TMA manufacturing (paraffin embedded tissues)
The TMA manufacturing is a simple three step procedure that is repeated for each sample placed on the TMA:
1. making a hole in an empty (recipient) paraffin block
2. taking a cylindrical sample from the tissue sample (donor) paraffin block
3. placing the cylindrical tissue sample in the premade hole in the recipient block
Exact positioning of the tip of the tissue cylinder at the level of the recipient block surface is crucial for the quality and the yield of the TMA block. Placing the tissue too deeply into the recipient block results in empty spots in the first sections taken from the TMA block. Positioning the tissue cylinder not deep enough causes empty spots in the last sections taken from this TMA. If tissue cylinders protrude, they may be gently pressed deeper into the pre-warmed TMA block (40°C for 10 minutes) using a glass slide. 
Bild anzeigen
TMA block sectioning (paraffin embedded tissues)
TMAs can be sectioned like standard paraffin blocks using regular microtomes. A special tape sectioning kit (Instrumedics Inc., NJ, USA) may be utilized to facilitate cutting, especially for large TMAs. The tape system leads to highly regular non-distorted sections (which are ideal for automated analysis) and helps to prevent arrayed samples from floating off the slide during incubation and washing steps. In addition, the loss of tissue during the sectionning is minimized. The tape system has the disadvantage that a slight background can occur between the arrayed samples. Also, the slightly irregular section surface appears to prevent an ideal distribution of reagents for some automated immunostainings(especially Ventana machines including the Discovery system). This may result in an uneven staining of TMAs in automated immunostainers. 
Bild anzeigen
TMAs from frozen tissues
TMAs may be manufactured from frozen tissues as well. Here, recipient blocks are made from OCT that is frozen down in a Tissue-Tek standard cryomold. The resulting OCT block (link to image) is mounted on top of a plastic biopsy cassette. During the arraying process, it is important to keep the tissue in the needle frozen during the procedure. This can be done by pre-cooling the needle with a piece of dry ice before punching and while dispensing the tissue core into the recipient block. 4 to 10#m sections of the whole block are cut from the array block using a cryostat microtome with or without a Tape Transfer System and slides. Simple home made tissue arrayers or modified commercial arrayers are highly suitable for frozen TMA manufacturing. Our frozen TMA blocks and sections typically consist of up to 300 spots. 
Bild anzeigen
Representativity of TMAs
An important concern connected to the TMA technology is the question whether or not "tiny" tissue cores measuring as less as 0.6 mm in diameter are representative of an entire tumor. To date, at least 20 studies have addressed this question, comparing IHC findings on TMAs with the corresponding traditional "large" sections. Most of these studies reported a high level of concordance of results, and concluded that inclusion more than one tissue core per donor block further increases the concordance. For example, Camp et al. studied expression of ER, PR, and Her2 in 2-10 tissue cores obtained from the same donor blocks in a set of 38 invasive breast carcinomas. They found that analysis of 2 cores was sufficient to obtain identical results as compared to the corresponding whole tissue sections in 95% of cases. 99% concordance was reached if 4 cores were analyzed, and analysis of additional cores did not result in a significant further increase of concordance.
However, all these studies are based on the assumption that classical large sections - the current gold standard for molecular tumor tissue analysis - is representative of an entire tumor. It is very well possible that this notion is not always true. In the optimal case, a "large" section will contain tumor tissue measuring 3 x 2 cm in diameter. Given a section thickness of 3#m the examined tumor volume is about 0.0018 cm3. This volume represents only 1/19,000 of a tumor with a diameter of 4 cm or a 1/150,000 of a tumor with a diameter of 8 cm. A TMA sample measuring 0.6 mm in diameter represents a tumor volume of 0.00000108 cm3 that is 1/1,600 of a 3 x 2 cm tumor area on a "large" section. Considering these numbers, the representativity problem is about 1,000 times greater between the entire tumor and a traditional "large" section than between a TMA sample and a "large" section. 
Bild anzeigen
More important than comparing individual results obtained on large sections and on TMA sections, therefore, is the question whether or not associations between molecular features and clinico-pathological parameters can be identified by either one of these not fully representative methods. In fact, all studies that we are aware of using TMAs to analyze known associations between molecular features and tumor phenotype or prognosis revealed the expected significant results. For example, expected associations with clinical outcome were found in TMA studies for the KI67 labeling index in urinary bladder cancer, for vimentin expression in kidney cancer, and for expression of estrogen and progesterone receptor proteins or HER-2 alterations in breast cancer patients. Our study comparing the prognostic significance p53 immunostaining in breast cancer on 4 different TMAs and large sections is most telling for judging the potential of TMAs for finding associations between molecular features and clinico-pathological associations. Although we found only about 20% positive cases in each TMA but almost 40% positivity in large sections (suggesting poor representativity of the TMA data) prognostic associations could only be found in all 4 TMA analyses but not in the large section analysis. In this study, irrelevant focal findings had been overrated by the pathologist on large sections. The greater objectivity and the highly standardized staining process had outweighted the theoretical advantage of our large section analysis.
Overall we believe that - at least for most biomarkers - TMAs are equally well or better suited to find associations between molecular features and clinico-pathological parameters than large sections. 
Bild anzeigen
Applications of TMA technology
In principle, all types of research requiring in-situ tissue analysis can be done in a TMA format. For example, TMA sections have been used for immunohistochemistry (IHC), fluorescence in situ hybridization (FISH) or RNA in-situ hybridization. As to yet, TMAs have mostly been used for cancer research but there are also many other applications. Depending on the aim of a particular analysis, oncology TMAs may be divided into four groups, namely prevalence TMAs, progression TMAs, prognostic TMAs, and TMAs composed of experimental tissues.
Prevalence TMAs are assembled from tumor samples of one or several types without attached clinico-pathological information. These TMAs are useful to determine the prevalence of a given alteration in tumor entities of interest. A typical example of a prevalence TMA has been published by Schraml et al. The TMA containing 4,788 different samples from 130 different tumor types has been used for the analysis of multiple different markers on the DNA and protein level, including FISH and IHC analysis of cyclin E amplification and overexpression. Cyclin E amplification was detected in 15 and Cyclin E protein accumulation in 48 different tumour types.
Bild anzeigen
Progression TMAs contain samples of different stages of one particular tumor type. They are instrumental to discover associations between tumor genotype and phenotype. For example, an ideal breast cancer progression TMA would contain samples of normal breast from patients with and without breast cancer history, several different non-neoplastic breast diseases, ductal and lobular carcinoma in situ, invasive cancers of all stages, grades and histologic subtypes as well as metastases and recurrences after initially successful treatment.
Prognosis TMAs contain samples from tumors with available clinical follow-up data. They represent a fast and reliable platform for the evaluation of the clinical importance of newly detected disease-related genes. Validation studies using prognosis TMAs readily reproduced all established associations between molecular findings and clinical outcome. For example, significant associations were found between estrogen or progesteron expression or HER-2 alterations and survival in breast cancer patients, between vimentin expression and prognosis in kidney cancer, and between Ki67 labeling index and prognosis in urinary bladder cancer, soft tissue sarcoma, and in Hurthle cell carcinoma.
Experimental TMAs may be constructed from tissues like cell lines or xenografts. Cell line TMAs are optimally suited for screening purposes, e.g. the rapid identification of cell lines with a specific genotype. Selected cell lines can than be grown and, for example, utilized for testing potentially inhibiting drug candidates. 
Bild anzeigen
TMA publications
Automation
TMAs are optimally suited for large scale expression profiling projects. Automation of the interpretation is the absolute key issue for such projects. In principle, TMAs are highly suited for automated analysis. The greatest difficulty for automated tissue analysis, the selection of an appropriate tissue area has been soleved to a large extent during TMA manufacturing. Several commercial systems enabling automated TMA analysis have recently been introduced. These can be successfully used for many antibodies. However, it is important for non-Pathologists to understand that you never know in advance what the staining pattern will be for a particular new antibody. Each new antibody can possibly stain structures in one or several compartments of cancer cells (nucleus, cytoplasm, membrane) and/or in the stroma. Stromal staining can again involve different cell types including for example blood vessels. Example images are shown for some special staining patterns. We use the following procedure to combine automated and manual interpretation of TMA sections.
1. Optimize staining protocol very carefully
2. Initial manual review of slides: check quality, staining patterns, unexpected results
3. Automated analysis with a simple home made low tech system (image database)
4. Use intelligent software to search for relevant cutoffs (such as our home made system "ThresholdFinder")
5. Perform manual reading if automated data analysis looks promising (all or some slides)
Obviously it is both feasible and desirable to generate TMA analysis systems that yield a more precise reading than simple signal intensity per spot analysis. We have not invested in the development of such systems because we are not yet convinced of their practical importance. Unfortunately, the influence of experimental parameters (antibody selection, antigen retrieval, staining protocol) on our results is so huge, that factors related to the experimental setup may will in almost all instances overwrite any precision added to the experiment during the phase of slide interpretation.
In a remarkable recent publication fluorescent dyes were used for immunostaining because of the better dynamic range as compared to peroxidase based systems and the potential for multicolor analyses (Rimm laboratory). These results were encouraging. One major disadvantage that we see for fluorescence based immunostaining of TMAs is the difficulty of manual re-evaluation of questionable stainings. Also, setting up reproducible staining protocols for multicolor applications is even more difficult than for single parameter analysis. Most commercial products for automated TMA analysis therefore concentrate on conventionally immunostained TMA sections. These are generally based on an automated microscope with a CCD camera. An overview image is generated used to identify the localization of each tissue spot in the TMA for subsequent high resolution scanning. 
Bild anzeigen
The Basel TMA archive
During the past years we have developed one of the smallest tissue archives of the world. Our TMA archive consists of >20'000 arrayed tumors for which >40'000 years of follow up data have been collected. The heart pieces of our TMAs are normal tissue TMAs (>70 different tissue types, >500 different cell types) and multi tumor TMAs (>120 tumor types; >3500 tissue samples). In addition we have constructed prognosis TMAs of the most important tumor types like breast cancer (n=2200), lung cancer (n=1500), colon cancer (n=1500), prostate cancer (n=500), bladder cancer (n=1000) or kidney cancer (n=300).
Comprehensive Molecular Epidemiology
We feel that having a large scale TMA resource in one laboratory has unique advantages. A large TMA resource is ideally suited for rapid analysis of the molecular epidemiology of genes of interest. We suggest a two step strategy for a comprehensive gene evaluation:
