Fluorescence in situ hybridization analysis with a tissue microarray: ‘FISH and chips’ analysis of pathology archives



This article is corrected by:

  1. Errata: Corrigendum Volume 60, Issue 9, 650, Article first published online: 15 August 2010

  • Declaration of conflicts of interest to declare: Shinichiro Kiyose is an employee of Jokoh Inc.

Haruhiko Sugimura, MD, PhD, Department of Pathology, Hamamamatsu University School of Medicine, 1-20-1, Handayama, Higashi-ward, Hamamatsu 431-3192, Japan. Email: hsugimur@hama-med.ac.jp


Practicing pathologists expect major somatic genetic changes in cancers, because the morphological deviations in the cancers they diagnose are so great that the somatic genetic changes to direct these phenotypes of tumors are supposed to be correspondingly tremendous. Several lines of evidence, especially lines generated by high-throughput genomic sequencing and genome-wide analyses of cancer DNAs are verifying their preoccupations. This article reviews a comprehensive morphological approach to pathology archives that consists of fluorescence in situ hybridization with bacterial artificial chromosome (BAC) probes and screening with tissue microarrays to detect structural changes in chromosomes (copy number alterations and rearrangements) in specimens of human solid tumors. The potential of this approach in the attempt to provide individually tailored medical practice, especially in terms of cancer therapy, is discussed.


Extreme copy number alterations (aneuploidy) are the norm in human solid tumors.1–3 Karyotyping solid tumors is so labourious4 that only limited information on chromosomal abnormalities in human solid tumors in situ was available until recently. The latest methodologies that involve the use of human genome information, however, have provided us techniques that make it possible to identify any locus-specific chromosomal changes in a tumor. Several examples of applications of these state-of-the-art methodologies are essential diagnostic tools in diagnostic laboratories to, for example, identify translocation in certain solid tumors.5–7

New information is being obtained every day in genetic research on human solid tumors (especially carcinomas). The high-throughput, ‘genome-wide’ approach to genetic changes in human tumors has been widely adopted in every branch of medicine, and it is now known that there are extensive somatic changes, including multiple point mutations,8,9 copy number alterations,10,11 and further complex rearrangements12 in every kind of tumor. Since most of these somatic changes have been identified in the analysis of the DNAs of advanced primary tumors and tumor cell lines, questions about when and where these genetic changes occur during cancer development in the human body remain to be answered by pathologists. Human pathology archives contain specimens of human tumors in various stages of development, from the incipient stage to the metastatic stage, and they are a treasure trove in the post-human-genome-sequencing era. The know-hows of two methods are important, especially for diagnostic pathologists: intensive application of bacterial artificial chromosome (BAC) clones as probes that have exact ‘addresses’ in the whole genome and construction of tissue microarrays (TMAs) which consist of hundreds of tissue specimens on a single slide. Using a combination of these two know-hows is a strategy that facilitates identification of changes at any genomic locus in several hundreds of tissue samples at once.

Use of some of the specific BAC probes has already acquired a niche in routine examinations in diagnostic laboratories as a means of verifying a diagnosis, selecting subjects for particular molecularly targeted therapies, and for predicting recurrence.13–20 Use of BAC probes by diagnostic pathologists, however, is still not widespread because of the difficulty of accessing and making the BAC probes for interests of their own. In this article we review the various facets of the latest advances in the application of BAC probes to diagnostic pathology and describe some of our own experiences with using many BAC probes to investigate pathology archives. We think that using numerous BAC probes will soon become a popular diagnostic practice, the same as the current use of monoclonal antibodies.

Actually, several ambitious pathology laboratories around the world that possess these methods in their arsenals, have started to propose an agenda of TMA-FISH (‘Fish and chips’) approaches to tumor DNA analysis.21–30 The recent observation of repositioning of chromosomal loci during carcinogenesis has further encouraged the analysis of human tumor specimen in various clinicopathological settings.31,32


The development and modifications of the FISH procedure, especially for use in formalin-fixed-paraffin-embedded (FFPE) tissues have been extensively reviewed.30 Equivalent hybridization efficiency of probes for the arrayed pieces of tissue after different fixation times and storage methods is necessary to correctly evaluate copy number amplification. In many studies, the FISH procedure has been performed as a means of validation, that is, to verify amplification data generated by other methodologies, such as by quantitative PCR, array-based comparative genomic hybridization (aCGH), and single nucleotide polymorphism (SNP) arrays,33 and comparisons between methods and the interpretations of the results obtained by each method have sometimes been a matter of controversy.34,35 FISH analysis, especially of FFPE tissues, is often technically demanding, and standardized quality control, which is very important in practical settings, has just begun. There are large inconsistencies between the prevalence of amplification of well-known and familiar genes that we consider clinically useful and that are routinely used in practice without rigorous quality control guidelines.34,35 Thousands of BAC clones are commercially available, and, in theory, any of them can be used as FISH probes. The BAC clones or labeled probes can be ordered from at least two Japanese companies (Advanced GenoTechs Co., Tsukuba, Japan; GSP laboratory, Kawasaki, Japan). When we use these BAC clones for FISH procedures in paraffin-embedded tissue sections, several steps must be carefully performed including labeling and hybridizing them to DNA. The BAC clone must be confirmed to be the correct one, because assignments of BAC clones often change to reflect the daily process of refining the human genome database. The information on exact location of each BAC probe according to the most recent Build (Build 37 in March, 2010) of the human genome is necessary. Although the reason is usually unclear, some BAC clones hybridize with multiple sites (more than 4) in normal interphase cells, and logically they cannot be used to evaluate human tumors. Thus, commercial BAC probes must be tested to determine whether they are hybridized to the two corresponding sites (or two pairs of the signals on the sister chromatids) in the metaphase chromosome spread before they are applied to human tissues containing cancer cells (Fig. 1). Sequencing of part of the BAC probes is of some help in further confirming the correctness of the BAC probes.

Figure 1.

A metaphase spread (top) for testing a bacterial artificial chromosome (BAC) probe. (a) Two signals (green) with the corresponding centromere probe (red signals) are seen in the same chromosome. (b) Red and green signals are seen in different chromosomes, although they were supposed to be in the same chromosome according to the information in the database. Interphase cells exhibit two (pairs of) signals each (bottom).

In addition to the above-mentioned hurdles to obtaining the right BAC probes, there is another stumbling block to completion of a FISH procedure: the labeling step. Several labeling methods are available, and some are commercially available and packaged in the form of a kit. Sufficiently efficient labeling is sometimes achieved only in an heuristic manner.

The following limitations in interpretation must be considered when using a FISH procedure to enumerate chromosomes in paraffin-embedded tissue sections. The signals can be weak for many reasons. Clinical practice has been standardized only for the system for detection of HER2 amplification in breast cancer cases.36 The merits of protease treatment, microwave treatment, heating, and other treatments such as using various detergents have been debated. Some ‘pre-treatment’ kits are commercially available, but retrieval efficiency usually depends on the condition of the specimen, and individual adjustments must be made each time in each laboratory. For example, the recommended pretreatment to augment signal strength in the two kits available, the Hercep test (Abbott, Tokyo, Japan) and the HISTRA (Jokoh, Tokyo, Japan) are different.37 Based on our own experience, one technical tip for generating stable, sensitive signals in pathology archives that have been fixed by various methods and stored for a long period is appropriate, careful pretreatment with protease.

Since overlapping cells and cells whose nuclei are partially cut cause miscounting of the numbers of signals, cut-off values must be set based on preliminary evaluation of the signals in several non-tumorigenic tissues.38,39 Several quality controls are necessary before applying the new probes to clinical uses the same as for the HER2 probe.


The preparation of FISH probes is a tedious task that includes several hurdles described in the previous section that must be overcome. Many investigators have constructed tissue microarrays for efficient use of probes they had laboriously prepared, especially in retrospective studies. The idea of embedding many pieces in a single block existed in the early days of anatomical pathology, but several embedding instruments for this purpose recently became popular, and technical refinements are under way. One well-circulating brand of microarray instruments is Beecher Instruments (Beecher Instruments, Inc. Sun Prairie, WI, USA). Their models have 0.6 mm, 1 mm, 2 mm cylinders, and the Azumaya model KIN-1 and model FIN-3 (Azumaya Cooperation, Tokyo) have wider cores that are 2 mm, 3 mm, 5 mm, and 7 mm in diameter. There are pros and cons in regard to using the smaller cores, and several problems encountered in using the instruments with various sized-cores are addressed in the instructions included with each of the instruments. A validation study in regard to possible sampling error when small core specimens are collected was performed and the results were published.40 Very recently, donor blocks containing multiple slots and an apparatus for making them have become commercially available (Fig. 2) (Patent Application 2009-028167), and many other variations will become available commercially. In addition to genomic and immunohistochemical studies, a proteomics approach by imaging mass spectrometry on a TMA platform is also feasible.41

Figure 2.

Tissue microarray gauges, prefabricated recipient blocks with holes, commercially available, and embedded blocks from top to bottom. The core diameters are 3 mm, 2 mm, and 1 mm in diameter (left to right).


Preparations of TMAs and requests to prepare them will become more frequent in both investigative and diagnostic pathology laboratories, and as members of institutional review boards (IRBs) pathologists are sometimes responsible for appropriate control of these TMA bioresources. The categories of pathology specimens are described in several documents and on several websites,42–44 and IRBs are required to facilitate research proposals of making or using TMAs to implement the research smoothly and ethically.

TMAs are a major component of tissue banks,45 which are tissue resources for future personalized medicine and national and international bio-bank systems are now being established (websites: http://www.stn.org.sg, http://www.ukbiobank.ac.uk, http://www.bbmri.eu, and http://www.src.riken.go.jp/english/project/person/index.html).


Data on copy number alterations in solid tumors deposited in databases and publications have rapidly accumulated since the introduction of aCGH led to the discovery of many tumor-specific and stage-specific gains or losses of particular regions of chromosomes.46–49 Much of the information generated by aCGH itself is used as a diagnostic or prognostic tool in pathology laboratories.50–52 Information on genome-wide genetic changes in cancer DNA are now viewed as academic knowledge that is only useful to the graduate students and researchers, but sooner or later it will be an essential tool of the diagnostic pathologist facing daily challenges in diagnosis and management. There are many issues in conventional pathology research and practice to which human genome data can be applied.53 Sano et al. conducted a chromosome-wide survey to the archives of adenomatous hyperplasia of the lung38 and proposed ‘adenocarcinoma in adenomatous hyperplasia’ as an early stage of carcinogenesis of lung adenocarcinoma. Although the tools were genetic, the story they told was morphological. Very recently, a more powerful system, an SNP array platform containing more than 500 000 SNP sites has come into widespread use, and copy number estimation by several algorithms has facilitated identification of copy number changes, such as loss of heterozygosity, uniparental disomy, and amplification, in many clinical tumors. Midorikawa et al.54 integrated the data based on pathological examination of ‘nodule in nodule’ in resected liver tissue with the results of a comprehensive copy number survey with the Affymetrix SNP array that were confirmed by FISH, and succeeded in clarifying genetic process in human hepatocarcinogenesis in detail.

Research on structural changes and balanced translocation of chromosomes in solid epithelial tumors is also a cutting edge area of research today.7,51,55–57 The numbers of candidate probes that should be investigated for clinical significance seem huge. Several points need to be addressed when interpreting the results generated when an aCGH array and SNP array are used to analyze a human tumor genome. The first point is that many platforms are available to analyze copy number alterations, and a few papers on the characteristics of each platform have been published.58–60 Furthermore, since many algorithms are available to enumerate copy numbers on the same platform, the characteristics of the platforms themselves and the benchmarks of the algorithms need to be known. Most algorithms for estimating the copy number of loci set the reference dosage of tumor autosomes as 2 (diplotype), but this reference number is not valid for most common epithelial malignant tumors. Ng et al.61 recently refined the protocol for ploidy-specific copy number estimation, and obtained a better threshold for detecting CNA in cell lines, and Suzuki et al. performed a benchmark test of two widely used algorithms and extensively characterized the features of the algorithms in terms of different formulas for setting the gain or loss thresholds of genetic loci.62 Because of the intrinsic limitations of each method, two or three methods need to be used simultaneously for the same tumor.


Only a few FISH kits have been authorized for clinical use, but many are available for use in research. Translocation detection kits are often used to confirm diagnoses.63,64 Mori et al. recently used tens of BAC probes to make the differential diagnosis between adrenal tumors.65 However, the clinical significance of copy number alterations warrants further accumulations of retrospective and prospective data. The rationale for the efficacy of molecularly targeted drugs varies with the mutation, overexpression, and genomic amplification of the target molecules, such as HER2 and EGFR. Fu et al. investigated copy number changes and expression of GATA-6 in pancreatic cancer and reported finding consistency between the results for overexpression and amplification of the genomic area of the GATA-6 locus,66 and they also validated their findings observation by FISH. Amplification itself, however, does not always imply activation of the molecules or pathways of the genes on that genomic locus. Actually the EGFR immunohistological findings in lung cancer cells are not always consistent with the FISH data,67 and borderline grades of immunostaining of HER2(2+) require FISH analysis to determine whether the HER2 gene has been amplified. Another receptor kinase gene, MET, has been evaluated as a potential target of tailor made therapy in the same manner as the EGFR gene and HER2 gene have, and in some studies MET amplification has been found to predict shorter patient survival after surgical resection of non-small cell lung cancer.68 Amplifications of PIK3CA is found in a considerable percentage of non-small cell lung cancers, and it and PIK3CA mutation are mutually exclusive.69 The list of the amplified segments continues to increase, although validation of their clinical significance awaits further study. The list of tumors in which amplification of certain gene product(s)can be indentified has been growing, meaning that the list of the promising targets of therapy is also growing. Comprehensive copy number analysis by large-scale sequence technology has revealed that a copy number gain of an unexpectedly high proportion of genes that encode kinases in cancers.11 We tested 100 BAC probes containing different kinase loci in a gastric, colorectal, and lung cancer detection sets (20 cases for each organ) by TMA-FISH technology, and found amplification of at least one kinase gene in a considerable number of cases, or, expressed another way, found that unexpected kinase loci were amplified in a significant proportion of human common solid tumors (Figs 3–5). The discovery blocks we used consisted of tumor tissues in both early and advanced stages, and various histological types. The observation above has also provided us with the following perspectives. Combinatory chemistry has already generated many drugs targeted to kinase genes or their products, thus amplifications of specific sites on certain kinase genes are amenable to pharmacological intervention which that will lead to the establishment of the target specific therapy. When observations like ours are validated and refined for clinical evaluation, the FISH diagnostic system with particular kinase probes may serve as another basis of tailor-made cancer therapy.

Figure 3.

Amplification of kinase loci detected in FFPE tissues from the undifferentiated carcinoma of the stomach. Symbol genes are FMS related tyrosine kinase 3 (FLT3) (a), Activated p21CDC42 kinase(ACK1) (b), V-SRC avian sarcoma (Schimidt-Ruppin A-2) viral oncogene (SRC) (c), and Cyclin dependent kinase 8 (CDK8) (d). The probes were labeled with Spectrum Orange (Abbott, Abbott Park, IL, USA), and the nuclei were stained with 4, 6-diamino-2-phenyl indole dihydrochloride (DAPI, Abbot). The method is described in detail in the previous literature.

Figure 4.

Distribution of the numbers of the loci amplified in any of the 60 cases (20 gastric cancer cases, 20 lung cancer cases, and 20 colon cancer cases) in a discovery set. More than half (51) of the 100 loci tested were not amplified in any of the 60 cases tested. Five or more loci were amplified in 5 (8%) of the 60 cases tested.

Figure 5.

Distribution of cases according to numbers of loci amplified (vertical axis). From 0 to 11 of the 70 or more (as many as 100) loci successfully tested were amplified. None of the 100 loci were amplified in 29 of the cases. Seven or more loci were amplified in 3 cases.

Major issues, however, remain to be resolved for before authorization of FISH-based diagnostic tools even if scientifically validated. Cost-benefit analysis of so-called targeted therapy is just starting in the tight-fisted health insurance environment, and there are gloom and doom forecasts that a bonanza of new authorized diagnostic kits is unlikely to arrive anytime soon. The time-line of the last few decades, however, in which many antibodies eventually became essential in pathology labs, evokes us a very different picture.


The basic knowledge required to perform the combination of TMA and FISH with many BAC probes is familiar to diagnostic pathologists, but that is different from actually running it (TMA-FISH with BACs) in real pathology practice. Obtaining an ample numbers of BAC probes, labeling, and expensive fluorescence microscopes may be hurdles for modestly equipped community hospitals. In the previous issue of Pathology International, Kato et al.70 have reported their experience with using of a commercialized product that applies chromogenic in situ hybridization, a friendlier method that allows the use of ordinary microscope.

Many DNA probes labeled ‘research use only’ are actually used in sarcoma diagnosis,20 and standardization and quality control of only a few FISH diagnostic systems have been achieved. Most of DNA probes are expensive, and there are few ‘generic’ diagnostic kits.

Over the coming decades, DNA probes will become a familiar diagnostic tool to the pathologists in community hospitals, and the information obtained by using them will suggest therapeutic guidance as well as the diagnosis. At the same time, the accumulation of the data generated by TMA-FISH approach will complement numerous OMICS data that have been accumulating in other disciplines of medicine. In other words, the TMA-FISH approach may be one of the smartest harvest (exit) strategies among OMICS projects related to human cancer, and many investments have been made in it over the last two decades.


This work was supported by a Grant-in-Aid for priority areas from the Japanese Ministry of Education, Culture, Sports, Science and Technology (20014007), from the Japanese Ministry of Health, and from the Smoking Research Foundation.