The biggest news these days in tissue microarrays (TMAs) may be that this former "next big thing" has become a standard tool for molecular profiling of disease. Some 682 papers have been published on TMAs, according to the ISI Web of Science, two-thirds of them on cancer. The technology has helped move at least one biomarker – racemase, for prostate cancer – into the clinic. Their usefulness for gene marker validation is broadly accepted, and some in the field see TMAs as platforms for multimarker discovery in their own right.
"What's most exciting for me as a pathologist who does a lot of molecular epidemiology-type studies, biomarker studies, is [that] the TMAs have just become accepted as a standard way of starting to do the testing and validation work," says Mark Rubin, associate professor of pathology at Harvard Medical School. In the early years, Rubin says, TMA advocates spent a lot of time defending the technology itself against criticism because of sampling issues.
Sampling problems have been largely resolved (though there remains no standard for core number or size), yet others persist. Back-end image capture and analysis, in particular, remain bottlenecks, as does the ability to tie a tissue sample with the clinical data that makes these arrays so valuable. Progress is being made on all these fronts, however.
BUILDING A BETTER TMA
The process of constructing TMAs – slicing thin sections from a block of paraffin injected with tens or hundreds of tiny tissue cores – hasn't changed much since Juha Kononen developed the technology in 1997 when he was a postdoc for the National Human Genome Research Institute. Automated array makers have hit the market, yet many researchers say they still prefer the old manual arrayer made by Beecher Instruments of Sun Prairie, Wis., where Kononen is now chief scientific officer.
Scientists usually make their own arrays or work with a few selected collaborators to assemble them, rather than buying them off the shelf. That's because for TMAs to be truly useful as a tool for clinical trials and molecular epidemiology, experimenters must be confident that the hundreds of patients represented on a slide have the same type of tumor and received the same type of treatment, says Martin L. Ferguson, senior vice president of bioinformatics at Lexington, Mass.-based Ardais, a company that makes highly annotated TMAs.
Other companies include such information to a greater or lesser degree. TriStar Technology Group of Bethesda, Md., includes detailed clinical follow-up information with its TMA products. Zymed Laboratories of South San Francisco has begun providing some patient information with its arrays, but does not yet include treatment information.
"It is just plain difficult, especially in these days of IRBs [institutional review boards] and HIPAA [Health Insurance Portability and Accountability Act], to legally and ethically produce rich clinical datasets, and to do it quickly, correctly, and without great cost," says David B. Seligson, director of the Tissue Array Core Facility at the University of California, Los Angeles. "Without clinical data, the ultimate power of the TMA – linking expression, pathology, and clinical data – is not realized."
Attesting to this difficulty, Ardais is getting out of the TMA-building business and shifting its focus to selling the software systems it has developed for collecting and annotating tissue. Ferguson says the company will continue to provide TMAs to customers who request them but will contract their actual construction out to other companies.
Ardais has developed protocols to ensure that tissues are collected in an ethical manner and patient privacy is safeguarded, key concerns with any human-tissue research. Anyone seeking to buy a TMA should ask the producer how the tissue was obtained, and be wary of companies that might be acting as intermediaries for unscrupulous collectors, says Stephen Hewitt, who runs the Tissue Array Research Program (TARP) at the National Cancer Institute. "I think there are some shady ones out there."
TARP provides its TMAs to collaborators at cost, as do several academic centers, while others prefer to keep these valuable resources in-house. Some academic centers, such as Yale University, will sell them either at cost or for profit.
Commercial TMAs, Rubin says, make good first-look screening tools. "You can use them as a sort of initial screening array to see what kinds of markers your sample's expressing," agrees UCLA's Seligson, who adds that distinct and complementary groups of tissue arrays are available – commercial arrays for first-tier work and academically produced arrays with detailed annotation for deeper research.
Researchers must consider several issues when purchasing a commercial array, however. Tissue fall-off during antigen retrieval and pretreatment can be a problem with some commercial TMAs, says Louise Yau-Shah, a spokesperson for Zymed. The company, recently acquired by Invitrogen of Carlsbad, Calif., uses a patented paraffin matrix technology to prevent tissue loss.
Tissue cores in Zymed's TMAs are 1.5 mm across, and its slides contain up to 96 cores in all. Researchers still cannot agree about the "best" core size or sample number to have on a slide. While 0.6 mm cores are popular because they conserve tissue, some researchers, including neurobiologists and neuropathologists, prefer larger cores that make it possible to see tissue architecture more clearly.
Both commercial and academic TMA makers seem to be shifting toward producing slides containing fewer cores – dozens to hundreds, rather than a thousand – because they are easier to handle. Researchers agree that two or three cores from a single patient per slide are sufficient to account for patient heterogeneity, and some use only one core per sample.
ANALYSIS HURDLES AND HEADACHES
Selected TMA Suppliers & Resources
Cytomyx Research Products
Cooperative Human Tissue Network
Stanford Tissue Microarray Consortium
Tissue Array Research Program
TriStar Technology Group
UCLA Tissue Array Core Facility
Image and data analysis is another area in TMA research that has little standardization. Analyzing results of in situ hybridization and other staining techniques used on TMAs remains as much art as science, says Ardais' Ferguson. Indeed, developing an automated system that can capture the complex images on a TMA and then convert them into data that preserves this complexity remains a formidable, and unmet, challenge.
"There still isn't an intelligent system that's able to pinpoint the appropriate tissue types, pathologies, and cell compartments," says Seligson. Thus, even if image capture quality is good, a pathologist still needs to select regions to analyze "or at least make sure that the machine isn't generating data from regions you don't want," he adds.
"I don't see any future for automated tissue-array analysis," says Guido Sauter of the University of Basel, Switzerland, a coin-ventor of TMA technology who is on TriStar's scientific advisory board. No automated system rivals the trained human eye, says Sauter, who points out that using the automated systems can actually be more time consuming than manual analysis. He says lysate arrays, in which drops of protein from lysed frozen tissue samples are placed on a slide, make much more sense for automated analysis and deserve more attention. Such arrays would allow for relatively simple quantitative analysis of target proteins, similar to the way DNA microarrays are read.
Harvard's Rubin says he doesn't think TMA evaluation will ever really be fully automated. "I don't think we're ever going to get around having a pathologist at some point evaluate it," he says. However, he continues, the evaluation of biomarkers has the potential to become fully automatic. "There's some tumor types where we're getting close to being able to do that."
David Rimm of Yale University recently founded Histometrix to produce and market his TMA imaging and analysis system, AQUA (automated quantitative analysis), which reads samples stained with fluorescent-labeled antibodies, measuring co-localization of fluorescence signals within subcellular compartments.
"I'm biased but I still believe that it's head and shoulders above what's out there," says Rimm, who has been using the automated analysis technique in his own research for the past three years. Many of his colleagues, including Matt van de Rijn of Stanford University, agree. Nevertheless, van de Rijn and others are working on their own in-house systems for image capture and analysis.
Histometrix's imaging system is called "Professor Marvel." Other companies with technology for TMA imaging and analysis include Vista, Calif.-based Aperio Technologies, with its ScanScope system, and San Diego-based BioImagene. The ScanScope system, used in a number of academic institutions, is nonmicroscope based, instead using a fax-like scanning technology that captures images at 0.5- and 0.24-μm resolution per pixel. "This is as good as, if not better than a perfectly set up and functioning microscope lens at 20× and 40× magnification," says Aperio spokesperson Tim Marshall. Images are then loaded onto digital slides, and the company provides a software package for image processing and analysis.
While Aperio's system allows for easy sharing of images, and many researchers are developing their own accessible image databases, the subjective nature of reading immunohistochemistry (IHC) stains can make data sharing difficult. Seligson sees a big push ahead for standardization in IHC scoring as well as validation of the bioinformatics used to analyze TMA data. Higher-echelon journals already are beginning to require independent validation studies before publishing TMA research, he notes.
Another key goal of many TMA researchers is a searchable, open-source database of results, as well as an eBay-like TMA clearinghouse where investigators willing to have their samples studied could post their inventory and requirements for collaboration. Right now, Rubin points out, it's difficult for investigators to know what's out there except through word of mouth.
Small networks to share this information are already forming. UCLA participates in a shared pathology informatics network with the National Cancer Institute, Indiana University, Harvard, and Pittsburgh University, and has formed its own database of clinical, pathology, and tissue-array assay data, called TMAtrix, which handles more than 700 data fields. Stanford's van de Rijn has constructed a Web site with complete information from his team's research, including images of every single TMA core. The Rubin Lab at Harvard and the Tissue MicroArray Core at Johns Hopkins University have similar resources.
The next wave in TMAs, says NCI's Hewitt, is the xenograft microarray (XMA) and the cellular microarray. "For certain tissues, they're the perfect solution," he says. He and his colleagues already are distributing their first XMA, of pediatric tumors. The advantage here, says Hewitt, is that XMAs tend to have much more information attached, including information on cytogenetics and perhaps even drug response. What's more, he adds, they're a renewable resource.