Review Tissue microarrays: Potential in the Indian Article subcontinent

Venkataraman Girish, Ananthanaranayanan Vijayalakshmi* Department of Pathology, Loyola University Medical Center, Maywood, IL60153 and *Department of Preventive Medicine, Robert H.Lurie Center, Northwestern University Feinberg School of Medicine, Chicago IL60611

Correspondence to: Dr. Girish Venkataraman, E-mail: [email protected]

Abstract

Tissue microarrays (TMAs) are a means of combining hundreds of specimens of tissue on to a single slide for analysis simultaneously. The evolution of this technology to validate the results of cDNA microarrays has impacted tremendously in accurately identifying prognostic indicators significant in determining survival demographics for patients. TMAs can be generated from archival paraffin blocks, combined with sophisticated image analysis software for reading TMA , and a staggering amount of useful information can be generated in terms of the biomarkers useful in predicting patient outcome. There is a wide range of uses for the TMA technology including profiling of specific proteins in cancerous tissues and non-cancerous tissues. Given the wide variety of tissue resources available in India, investment in a dedicated TMA facility will be of immense use in the research arena in India. This review article discusses the basics of TMA construction, design, the software available for the analysis of this technology and its relevance to Indian scientists. A potential workflow structure for setting up a TMA facility is also included.

Key Words: Tissue microarrays, histopathology

Introduction institutions across the world have taken advantage of this fact and incorporated TMA facilities exclusively for In the current world of proteomics, the use of high this purpose. A thorough understanding of the potential density Tissue Microarrays (TMAs)—also called “tissue advantages of this technology coupled with a unified chips”—has been progressively increasing in the last system for generating TMA blocks using paraffin blocks decade ever since its inception in 1998. Arguably, it from high-histopathology-volume institutions in India might be touted as the most noteworthy development will help in establishing a good tissue database for the in histopathology techniques in the last decade.[1,2] Indian pathologists. The basics of TMA construction, While DNA microarrays permit expression analysis of hardware involved and scoring options available are thousands of genes from one tissue specimen on a discussed here. single array, TMAs make it possible to analyze hundreds or thousands of tissue specimens in a single experiment Basics of TMA construction using a single gene or antibody probe. Given the tremendous pace at which novel genes implicated in TMAs are a method of relocating tissue from standard cancer are being discovered using gene chips and histological paraffin blocks such that tissue from expression microarrays, TMAs hold an immense multiple donor blocks can be placed on the same potential to validate this genomic data across multiple recipient block. Therefore, the first step in the process tumor types in a limited time frame. Many pathology of creating a good TMA block is to locate the most

Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1 5 Venkataraman et al: Potential of TMAs representative area on each H/E slide and careful sufficient clinical data, an important issue that merits identification of the same area on the corresponding consideration is the number of cores per case to put paraffin block of the slide. Once this is done, the block into the TMA block while evaluating any biomarker is positioned underneath the TMA puncher such that expression by immunohistochemistry. Although this the representative area is directly under the punching number is variable, for most tumor TMAs, three to pins. Thereafter, small “core needle biopsies” of these four cores/block adequately represent a biomarker’s representative tissues are punched out directly from ability to predict survival outcomes.[8] Often two cores donor paraffin blocks and re-embedded onto a new are taken from the donor block; one core is taken from recipient TMA paraffin block. Many such cores can be the center of the donor block and the other close to embedded in such a “master-arrayer” block employing the periphery of the block. It is important to include this technique.[2] Using a 0.6 mm diameter puncher, appropriate positive and negative controls besides nearly 600 or more tissue cores can be arrayed on a orientation cores. It would be useful to chart out an standard glass slide in a precise manner defined by X-Y MS Excel worksheet identifying the numbers of the coordinates. Although TMA block generation is a labor- various cores in the same pattern in which the cores are intensive and time-consuming process, the ability of this embedded in the TMA block. In addition, inclusion of high throughput technology to generate data from a irrelevant tissue at defined X-Y coordinates can help in staggering number of cases vastly improves statistical accurate orientation of the cores. Often cores from precision and power. tumors of same T size, stage or grade or tissue of origin can be clustered together on the TMA block. A TMA types, design and data handling options typical low-density TMA block and slide is shown in Figure 1. Last but not the least, it is important to Having such a large number of spots on one single ensure that all blocks used for TMA construction have slide calls for a precise organization of the tissue spots been fixed and processed similarly because some at three levels—firstly, in the design and placement of immunohistochemical markers may not work when the the tissue cores in the TMA block; secondly, in the tissue is fixed in a different fixative. linking of clinical and pathological data to the correct core; and lastly, a validated means of scoring and The technology used to generate TMA can range from analyzing any biomarker data so that a meaningful manual to fully automated systems. One of the more statistical analysis can be done. prominent companies include Beecher Instruments (San Prairie, Wisconsin, USA), which features manual, semi- Types of TMAs automated as well as fully automated systems. The cost can vary between $10,000 -$42,000 (USD) depending The nature of tissues that can be used on TMAs is on the type of the equipment. Cost-effective alternative varied, ranging from totally normal tissues from various methods of generating TMA blocks can also be adapted organs to non-neoplastic (like diabetes) and neoplastic ones and sometimes even cell lines.[3,4] The Cooperative Human Tissue Network (CHTN), a division of the National Cancer Institute (NCI) can provide investigators with a wide range of normal tissues placed on a TMA slide so the expression profile of a single protein in many tissues can be assessed simultaneously. The information from such TMAs can provide valuable information regarding the biology of diseases in which these proteins are altered. TMAs based on neoplastic tissues also termed as tumor TMAs, are broadly classified into three types—multi-tumor arrays, progression arrays (based on stage of tumor) and prognostic arrays where tumors with known clinical end points are arrayed.[5-7]

Figure 1: A typical low-density TMA block containing 30 cores Design of TMAs (0.6mm each). The corresponding H/E slide with appropriate ID is depicted. Investigators musts be aware that individual cores can get depleted and the hence the need for duplicate representation An integral part of constructing TMAs is the design. of each core on the TMA block. Orientation cores have not been After deciding on the optimum number of cases with included here because of the low-density of cores on the block

6 Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1 Venkataraman et al: Potential of TMAs if 60 or fewer number of cores are to be imprinted into worksheet is created and grid locations of the cores are the TMA block. A standard microscope fitted with a noted in rows and columns in the same pattern as in holing needle or a blunted 16G bone marrow trephine the TMA slide. Thereafter, hyperlinks are inserted for needle can be adapted to design TMA blocks at a each grid such that clicking on the hyperlink fraction of the price of commercial instruments.[9] corresponding to a specific core in the Excel worksheet will open the corresponding image of the same core for Data acquisition, analysis and integration scoring. This method is especially useful if the number of cores on a slide is not going to exceed 150 cores. Digital scanning and analysis For managing high-density TMAs with more than 600 The next issue of importance is reading the TMA spots cores per slide, an excellent set of software tools for in an orderly, reproducible and reliable manner. Of high-throughput analysis has been developed at Stanford course, the intuitive option would be to read one core University.[11] These investigators used the BLISS after another manually under a bright-field microscope. System mentioned previously for generating a database Even so, keeping track of the precise position of each of images. Immunohistochemical staining results are spot can become a painstaking process if there are more recorded into an MS Excel Worksheet and this Excel than 300 tissue spots on any single slide. Consequently, data is re-formatted by a program called “TMA- digital options are popular for analyzing biomarker Deconvoluter” which converts the Excel data into a text expression on TMA. Systems like the BLISS Imaging file so that a Hierarchical Cluster Analysis can be done (Bacus Labs, Lombard, IL) and ACIS system of using the “Cluster” and “TreeView” software—this Chromavision (www.chromavision.com) can scan and analyzes the relatedness within tumor subsets depending acquire images of all spots on the slide in one go on the immunohistochemical biomarker profile. The without the user having to manually focus each spot TreeView software generates dendrograms akin to the separately before acquiring. Tracking and recording the ones seen with cDNA microarray data. Free access to scores from these tissue cores can be done using TMA-Deconvoluter is possible at the Stanford TMA commercially available proprietary software. Bacus Website (http://genome-www.stanford.edu/TMA) while Laboratories also markets a TMA analysis software the other two software programs are available at the called ‘TMAscore’. With this, TMA slides are scanned website of Michael B. Eisen’s lab (http://rana.lbl.gov/ into virtual slides and placed on a network. Thereafter, EisenSoftware.htm). multiple people can collaborate, and score each core manually or by using the aforementioned proprietary Another noteworthy open-source java-based software software. called “TMAJ” is available from the website of the Johns Hopkins University TMA core facility (http:// Cost-effective alternatives for analysis tmaj.pathology.jhmi.edu/). This helps in recording pathology data as well core tracking and scoring. A Given the logistics involved in using proprietary license is however required for users with potential software, there is a need for exploring other economical commercial interests. means of doing spot scoring and analysis. Some noteworthy cost-effective software alternatives are Data integration available for recording patient data, details and coordinates of each core for scoring, some of which are As long as TMA slides are read by a single pathologist, described below. This still requires an image-capturing it is possible to manage data with just MS Excel and system to create an archived database of images for statistical analysis software. When multiple investigators further analysis. The analysis can however be done read the same slides, systematic data integration can using an ‘open-source’ free software available over the become overwhelming. TMAs generate an immense Internet for academic use. amount of data and so collaborating pathologists analyzing the same TMAs in different institutions need In the first of these, a novel relational database for to be aware that they have to record the generated data TMA analysis has recently been described using Adobe in a uniform and standardized format. This is crucial Photoshop (an imaging software) for image-editing and because Excel data files generated by multiple Microsoft Excel for recording core coordinates and pathologists across the country scoring the same TMA performing scoring.[10] The authors of this study explain slides, can be integrated for statistical analysis only if that a library of digital images of individual cores is the scoring format is uniform between all the stored in a specified folder of the computer. An Excel investigators. Addressing this issue, a TMA data

Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1 7 Venkataraman et al: Potential of TMAs exchange software specification language for free use has and discovered 2 genes namely, IGFBP2 and HSP27 been developed in XML (eXtensible Markup Language) that were upregulated in xenografts of hormone recently.[12,13] The basic pre-requisite of this data refractory prostate cancer (HRPC). They followed this integration language is that all investigators need to up with immunohistochemistry for both markers on a have their Excel data scoring files with the same TMA and found that 100% of HRPC cases were column headings viz. Patient ID, Age, core ID etc in positive for IGFBP2 while none of the normal prostate the same order. A PERL (Practical Extraction and cores were positive for the same marker. Report Language) (another computer language) script will export all these Excel data in the final XML Criticism is often directed towards the fact that the database accurately only if the column headings are the small size of TMA cores may not adequately represent same in all Excel files. If XML is difficult, collaborating biomarkers (like ER) that are distributed pathologists can also use MS Excel exclusively and heterogeneously in the tissue. Addressing this problem configure it to concatenate data from multiple of spatial heterogeneity, Camp et al found that just two investigators for analysis.[14] 0.6 mm cores from the same block provided equivalent information on ER indices as that of whole sections.[20] Advantages, validation and caveats of TMAs Thus, intra-tumoral heterogeneity will not be an issue of significant concern. There are a number of plus points to using TMAs instead of standard histology sections. Firstly, a large Caveats with TMAs: With TMAs, investigators also number of cases can be analyzed simultaneously and need to be able to decide if constructing a TMA in this in turn, improves the power and precision of any any situation is going to give additional information statistical analysis considerably. Schraml et al analyzed for the extra price being paid; if you have only 30 the oncogenes CCD1, CMYC and ERBB2 in 397 donor blocks of a prostatic adenocarcinoma, it might tumor spots by Fluorescent In-Situ Hybridization in work out better doing conventional one slide-one just one week attesting to the rapidity with which section IHC rather than constructing a TMA. The TMAs can generate results.[15] cost-benefit ratio is high only if the array is valuable (i.e. if it contains rare to find tissues and lesions), Secondly, immunohistochemistry on a single TMA slide the number of cores on the array is very high or if is much more time-saving and economical in terms of there is a need to assess many proteins at one time. the amount of antibodies and reagents used as If too many spatially close cores (>6) are marked for compared to paraffin sections while ensuring uniform punching on the same block, the technologist will treatment of all cores on a slide—no batch to batch have a hard time locating these on the block. Lastly, variability needs to be accounted for as with standard it is necessary to do a Quality Assurance (QA) with sections.[16] Van de Rijn et al have described the utility every TMA slide to see if the representative tissue is of a 29-case mini TMA of breast cancer blocks as a present at the pre-defined X-Y coordinate on the slide. useful control for Estrogen Receptor. Instead of a single The representative tissue of interest may often get strong positive control slide, the TMA includes a depleted in deeper sections (e.g. breast lobules or spectrum of positivity from weak to strong.[17] As a areas of HGPIN in blocks from radical result, any change in the strength of the detection can prostatectomy). In such cases, an auxiliary TMA be easily detected. Besides, inter-laboratory quality block using the same tissues should be constructed control for immunohistochemistry can also be assessed. using larger tissue cores.

On the other hand, TMAs can be used to test new Resources: The National Cancer Institute (NCI, prognostic candidate biomarkers evolving from cDNA Bethesda, MD USA) has set up a Tissue Array microarray studies. Natkunam et al were able to test a Research Program, TARP (http://ccr.cancer.gov/ new marker of plasmacytic differentiation called tech_initiatives/tarp/) to facilitate access to multi-tumor MUM1/RF4 against 1335 different human tumors and TMA slides to investigators interested in procuring pre- demonstrate it to be a sensitive albeit non-specific designed TMA slides. There are two other divisions of marker to the extent that even melanocytic tumors the NCI namely, CBCTR and CPCTR (Central Breast/ stained positive.[18] Prostate Cancer Tissue Resource databases respectively) which have a repository of pre-arrayed TMA slides. Yet another study has demonstrated the elegant way in Access to TMA slides containing prostate and breast which TMAs validated the results of a cDNA cancer cores is available from them at a nominal price microarray analysis.[19] The authors screened 5184 genes after approval of the proposed intent of use.

8 Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1 Venkataraman et al: Potential of TMAs

On the other hand, commercial companies will also blocks received from the peripheral centers (Form-B); make TMA blocks if they are provided with the Form-C and D for scoring recording the clinico- standard paraffin blocks (www.petagen.com) although pathologic data of the array. All outgoing forms can be the costs can get prohibitive if high-density arrays are included as multiple Excel worksheets in one single planned. In a recent article, Mobasheri et al have Excel file that makes data handling and integration an provided a comprehensive list of the commercial easier job. A useful thing for pathologists participating companies constructing TMAs.[21] in such TMA facilities would be to learn to use MS Excel in a proficient manner so that they can record Potential of TMA in India data efficiently. For a start, programmers can design the aforementioned Excel guide sheets based on the Given the immense potential of this technology, it is CHTNs Excel sheets available from their website http:/ clearly evident that TMAs will play a major role in the /faculty.virginia.edu/chtn-tma/2002N1/ foreseeable future in the arena of evaluating potential CHTN2002N1X.xls. biomarkers. There are a number of high-volume histopathology institutions within India like AIIMS, On the logistic front, a manual tissue microarrayer often PGIMER, JIPMER and Tata Memorial Hospital. Any comes at the price of a high-end microtome and the one of these institutes can operate a ‘TMA core facility’. monetary savings from using TMAs will definitely offset All other institutes could send in their blocks along the initial costs of establishment, in the long run. For with relevant clinico-pathological information for TMA partial funding of core center technologists, 20% of the construction. In return, these institutes would then have TMA slides from tumor arrays may be allocated to access to the TMA slides generated from the blocks at a foreign researchers at a nominal price to generate nominal price that would hopefully cover for the pay of revenue. the technicians and computer programmers working at the core facility. TMA slides with relevant information could be used as thesis material by PhD and post-graduate pathology Two technical staff (one for receiving and the other for residents who are required to complete a thesis work in arraying blocks), one computer programmer (for partial fulfillment of their residency program. Currently, computerized cataloguing of the TMA database and pathology residents often have a difficult time accessing generating template MS Excel data files) and one archival paraffin blocks, slides and clinical information pathologist for marking out areas to be punched are in the process of working on their thesis. Unfortunately, sufficient within such a core facility. Figure 2 depicts a many medical and even pathology postgraduates are potential hierarchy to streamline the process in our often ignorant about this technology. Students, faculty country. Sample forms that will be useful for recording and technicians at all levels should be educated about array design (Form-A), clinico-pathological data of TMA technology. At the medical school level, the pathology curriculum should probably try to include a chapter on recent medical technology like TMA and Peripheral 4 3 Centre others that have impacted significantly on the research Filled front. At the postgraduate level, invited talks by experts Send Blocks AND Clinical Data Excel Forms-C from Unstained Slides Peripheral Centres in the field at national conferences like the IAPM would 1 Attach Relevant Excel be a good way to introduce this technology at all levels Forms CORE CENTRE 5 in the scientific community. Another potential advantage TECHS (x2) On request PATHOLOGIST PROGRAMMER is that a single TMA slide with all possible grades of • Array Design • Block Receipt • Create Excel any single tumor type (like prostate cancer) will help • Mark Areas for • ID Assigning/Label Forms A-D coring • TMA Block/slide • Integrate XML pathology post-graduates in mastering the fine art of • Analyze integrated generation data for statistical data with from Both inputs analysis tumor grading. This in turn will help in reducing inter- Biostatistician observer variability while grading these tumors in daily 2 Archive blocks and unstained slides surgical pathology practice.

Figure 2: Flowchart depicting the flow of blocks and slides in On the other hand, there will be abundant material for a potential TMA setup in order, going through steps 1-5. The Peripheral centres send in their blocks with relevant clinical the faculty to conduct research without having to use information to the Core Centre. The Core Centre designs the up large amounts of costly antibodies. On the research array and creates TMA blocks and slides. It is important for a front, there will be ample opportunities to collaborate coordinating computer programmer at the Core Centre to design standardized forms for receipt of blocks, clinical data with overseas institutions where access to histology and array design tissues is relatively difficult owing to bureaucratic issues.

Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1 9 Venkataraman et al: Potential of TMAs

The ample autopsy material available within India can 2004;20:707-12. 8. Rubin MA, Dunn R, Strawderman M, Pienta KJ. Tissue microarray serve as a valuable source of tissues for TMA blocks sampling strategy for prostate cancer biomarker analysis. Am J while collaborating foreign institutions can supply Surg Pathol 2002;26:312-9. primary antibodies thus pooling resources. 9. Hidalgo A, Pina P, Guerrero G, Lazos M, Salcedo M. A simple method for the construction of small format tissue arrays. J Clin Pathol 2003;56:144-6. Final word 10. Shaknovich R, Celestine A, Yang L, Cattoretti G. Novel relational database for tissue microarray analysis. Arch Pathol Lab Med TMA technology is a powerful proteomic platform that 2003;127:492-4. 11. Liu CL, Prapong W, Natkunam Y, Alizadeh A, Montgomery K, Gilks plays an integral role in understanding protein CB, et al. Software tools for high-throughput analysis and archiv- expression patterns in a wide variety of normal and ing of immunohistochemistry staining data obtained with tissue neoplastic tissues. The technology is highly evolved and microarrays. Am J Pathol 2002;161:1557-65. 12. Berman JJ, Edgerton ME, Friedman BA. The tissue microarray data has abundant potential to bridge the chasm between exchange specification: A community-based, open source tool translational research and clinical therapeutics. Given the for sharing tissue microarray data. BMC Med Inform Decis Mak wealth of tissue resources available in our country, it 2003;3:5. only seems that the adopting this state-of-the-art 13. Berman JJ, Datta M, Kajdacsy-Balla A, Melamed J, Orenstein J, Dobbin K, et al. The tissue microarray data exchange specifica- technology will go a long way in benefiting students tion: Implementation by the Cooperative Prostate Cancer Tissue and researchers alike in the time to come. Resource. BMC 2004;5:19. 14. Manley S, Mucci NR, De Marzo AM, Rubin MA. Relational data- base structure to manage high-density tissue microarray data and Addendum: All the hyperlinks have been tested and the images for pathology studies focusing on clinical outcome: The websites are accessible as of December 29, 2004. prostate specialized program of research excellence model. Am J Pathol 2001;159:837-43. References 15. Schraml P, Kononen J, Bubendorf L, Moch H, Bissig H, Nocito A, et al. Tissue microarrays for gene amplification surveys in many different tumor types. Clin Cancer Res 1999;5:1966-75. 1. Wan WH, Fortuna MB, Furmanski P. A rapid and efficient method 16. Milanes-Yearsley M, Hammond ME, Pajak TF, Cooper JS, Chang for testing immunohistochemical reactivity of monoclonal anti- C, Griffin T, et al. Tissue micro-array: A cost and time-effective bodies against multiple tissue samples simultaneously. J Immunol method for correlative studies by regional and national cancer Methods 1987;103:121-9. study groups. Mod Pathol 2002;15:1366-73. 2. Kononen J, Bubendorf L, Kallioniemi A, Barlund M, Schraml P, 17. van de Rijn M, Gilks CB. Applications of microarrays to histopa- Leighton S, et al. Tissue microarrays for high-throughput molecu- thology. Histopathology 2004;44:97-108. lar profiling of tumor specimens. Nat Med 1998;4:844-7. 18. Natkunam Y, Warnke RA, Montgomery K, Falini B, van De Rijn M. 3. Mobasheri A, Marples D. Expression of the AQP-1 water chan- Analysis of MUM1/IRF4 protein expression using tissue nel in normal human tissues: A semiquantitative study using tis- microarrays and immunohistochemistry. Mod Pathol sue microarray technology. Am J Physiol Cell Physiol 2001;14:686-94. 2004;286:C529-37. 19. Bubendorf L, Kolmer M, Kononen J, Koivisto P, Mousses S, Chen 4. Moskaluk CA, Stoler MH. Agarose mold embedding of cultured Y, et al. Hormone therapy failure in human prostate cancer: Analy- cells for tissue microarrays. Diagn Mol Pathol 2002;11:234-8. sis by complementary DNA and tissue microarrays. J Natl Cancer 5. Nocito A, Kononen J, Kallioniemi OP, Sauter G. Tissue microarrays Inst 1999;91:1758-64. (TMAs) for high-throughput molecular pathology research. Int J 20. Camp RL, Charette LA, Rimm DL. Validation of tissue microarray Cancer 2001;94:1-5. technology in breast carcinoma. Lab Invest 2000;80:1943-9. 6. Shergill IS, Arya M. Tissue microarrays. Expert Rev Mol Diagn 21. Mobasheri A, Airley R, Foster CS, Schulze-Tanzil G, Shakibaei M. 2004;4:421-3. Post-genomic applications of tissue microarrays: Basic research, 7. Shergill IS, Shergill NK, Arya M, Patel HR. Tissue microarrays: prognostic oncology, clinical genomics and drug discovery. Histol A current medical research tool. Curr Med Res Opin Histopathol 2004;19:325-35.

10 Indian Journal of Cancer | January - March 2005 | Volume 42 | Issue 1