Skip to main content
Premium Trial:

Request an Annual Quote

The Awkward Adolescence of Arrays


As far as microarrays go, gene chips might as well be considered the new kid on the block. All sorts of samples can be put down on a slide in an array format, and the longstanding tissue microarray is among the most useful to the field of clinical pathology. While these arrays are primarily used to probe large numbers of tissue samples for cancer biomarkers, many pathologists still haven't made the jump to truly high-throughput biology — they eyeball each sample individually.

The first tissue array was created more than 20 years ago, but it's taken until the last decade for automated arraying and imaging to propel the technique into the realm of large-scale biology. Still, automation has yet to be widely adopted, leaving many pathologists looking for more objective analysis. Few vendors offer such standardized, quantitative readouts, and in many cases, the automated platforms are just not affordable.

Most labs create their own tissue microarrays using archived patient samples. There are at least 20 vendors that sell prefabricated arrays; however, most of these aren't that useful because they lack archival clinical annotation. "There's a very wide range of tissue and questions and possibilities with formalin-fixed, paraffin-embedded material," says David Rimm, director of Yale's tissue array core facility. "In fact, one of the great resources that I think many institutions have … is the pathology archives. [They] represent our oil well, so to speak."

Building blocks

The process of creating a tissue microarray is labor-intensive, whether one has automated arrayers on hand or not. Using a hollow needle, tissue cores as small as 0.6 mm in diameter are extracted from paraffin-embedded clinical biopsy samples, and then deposited into a master block in an evenly spaced array pattern. From this master block, sections are cut and mounted on a slide. Most often, researchers use immunohistochemistry to see which protein is being expressed, how much is there, and its specific location within the section. These arrays can also be used to visualize nucleic acids through fluorescent in situ hybridization.

Tissue arrays have found a strong footing in clinical research labs, whether for biomarker discovery and validation or for drug development studies. "Users are looking at tissue biomarkers, or looking at trying to understand protein function better. So they purified a new protein X from some gene screen and now they want to know, is it expressed predominantly in primary tumors or in metastases? Or is it expressed only in low-grade tumors or low-grade and high-grade? Or is it expressed in prostate cancer but not in breast cancer?" says Rimm. "Those are the kinds of questions they can ask."

In clinical trials, associating patient drug information with gene expression patterns is critical to making a successful drug, so the arrays are often put to use in dosage and toxicology studies. "It's condensed pathology," says Stephen Hewitt, who runs the National Cancer Institute's tissue core facility and produces arrays for both NCI and external researchers. "Instead of screening a few cases, you're screening large numbers of cases." His team tends to focus on validating biomarkers, both prognostic (Will a patient's tumor spread?) and predictive (Will the patient's tumor respond to therapy?). While much published work has focused on localizing specific markers for disease mechanism of action, Hewitt notes that arrays have been a boon for molecular epidemiologists studying risk factors in large populations. Though the field is currently being driven by SNPs, he says, "people really do want to look at environment and protein interactions. … This is a platform where you can look at those questions — the questions you just can't answer at the DNA level."

Translational research

As a translational research tool or, as Hewitt says, "the translation point from basic science towards clinical medicine through pathology," tissue microarrays put broad population samplings to good use.  NCI's Mark Sherman uses tissue arrays to classify tumors according to expressed protein biomarkers. He can then go back to find which factors impose more of a risk than others, and whether or not these can be related to all tumors. "Eventually we want to develop a risk model for breast cancer where we have factors that relate," Sherman says, in order to track how breast cancer develops, how it might be prevented, and how it could be detected more effectively.

His studies typically include thousands of patients. In one project, he is investigating 2,500 Polish women with breast cancer to correlate risk with different factors, including low penetrance genes. "If we had to look at one slide per patient, it would cost us a prohibitive amount of money," Sherman says. Not only do tissue arrays make these studies affordable, but they also increase reproducibility and accuracy, especially in light of the fact that staining many individual slides by hand can introduce error. "For efficiency, speed, standardization — very large studies have tried out this approach," he adds.

Enabling much of this research are standard assays, like IHC and FISH. For accreditation purposes, many clinical labs make use of the high-throughput nature of tissue microarrays for internal validation. At the Cleveland Clinic, Ziad Peerwani, a third-year resident in the department of pathology, is working on a project with the College of American Pathologists to standardize their IHC runs across the clinic — and thereby maintain its CAP accreditation — with specially created tissue arrays. Moreover, pre-fabricated arrays can also be used as internal experimental controls "for positive staining control to ensure that your techniques are working properly," Peerwani says.
With such widespread clinical use, it's hard not to wonder if tissue microarrays have potential diagnostic value. Most experts say that they predict little use as a diagnostic tool, simply because clinical diagnoses can't wait for hundreds of patient samples to come in before an IHC stain is run. Yale's Rimm sees it as basically a high-throughput research tool. However, echoing Peerwani, he's already seeing clinical use for control tissue. "Instead of putting on a single control tissue, you can now put on 20 or 30 or 50 control tissues with every slide," he says. "I expect that we'll see an increase in usage in that context."

Images and standards

Technological advances during the last decade have produced platforms for automatically arraying and digitally visualizing arrays, which has cut down on time and subjective error. However, most experts interviewed still create and visualize their arrays manually.

"The big 800-pound gorilla in the room is not so much applying these materials, but actually trying to do it in a very validated, standardized fashion and then reading them," says Sherman. Most of the analysis is still done by eye, as IHC itself is difficult to standardize. IHC is not very reproducible and is very quantitative. Additionally, pathologists continue to struggle against tissue heterogeneity — trying to create an array that accurately and completely represents protein expression patterns that might differ across a tumor mass.

Rimm and colleagues have confronted these problems head-on. While automating the arraying process hasn't been that successful — tissues vary not only in protein expression patterns but also in thickness and amount of fat content — building out a platform to read the arrays in a more standard fashion has. In 2000, Rimm began working on a measurement tool that reads absolute protein concentration from an IHC stain, and it's now licensed exclusively to HistoRx under the name of Aqua. "The biggest limitation in my own mind, which is what we started working on in 2000 and now have overcome, is the reading of them," he says. "Back when they were invented, the only way to read them was to have a pathologist look at them and issue an opinion. While I am a pathologist and I respect pathologists' opinions, they will never be objective."

When using IHC to visualize protein concentration, pathologists use a subjective approach, estimating what they think might be the intensity of the spot and then grading it on a gross range. HistoRx's Aqua technology measures absolute quantity and assigns each spot a standardized score. While a much less subjective route than eyeballing, it's still difficult to get statistically relevant information from an IHC stain. "I think the biggest hurdle is capturing the true statistical power that comes from analyzing large populations on a tissue microarray," says John Tonkinson, VP of business development at HistoRx. With IHC, "you're using a very subjective, qualitative way of saying, there's a little bit there, there's a lot there — and now you're trying to apply that in a quantitative fashion to generate statistical value. The two are not compatible."

While HistoRx's is the only system that does absolute concentration, Rimm says, a number of other vendors offer digital scoring tools. However, Tonkinson notes, despite a move toward digital pathology, most people are still "doing it by eye." The equipment is, for the most part, only available to drug development companies or researchers with deep pockets.

Says Hewitt at NCI, "The image analysis field is something that is continuing to evolve. There's not going to be a winner; there's going to be a number of high-quality solutions." He's used several different systems, and acknowledges that the technology is not only expensive, but also a work in progress. "It's still a challenge. Only recently have I been willing to take my bigger projects into full, automated analysis," he says.

A major roadblock is database infrastructure. For example, Hewitt says that for each tissue array, there are 400 to 500 cores, with up to 20 gigs in images for each core. With such immense data sets, the analysis part tends to lag behind, simply because the data has more complexity and depth, he says. "It's not trivial bioinformatics."

Peerwani at the Cleveland Clinic knows that automation won't take away the need for pathological analysis — when it comes to how tumors act in living tissue and how they interact with surrounding structures, looking at tissue samples on a slide probably won't cut it. But for now, he says, "I think that the newer technologies are a very big boon to the research itself and to the art and science of using tissue microarrays." He adds, "I think the problem is their expense. As those come down, it will be more practical to utilize it."


Hunting for Diagnostic Uses

Most pathologists are pessimistic about tissue arrays ever being used as a diagnostic tool, but try telling that to the British Columbia Cancer Agency's David Huntsman. An associate professor at the University of British Columbia, Huntsman recently finished a pilot study that tested whether arrays could be used in actual breast cancer patients to determine risk subgroups. "We considered whether we could use tissue arrays for the delivery of breast cancer biomarkers," Huntsman says. "We've done a pilot study which showed that the tissue arrays gave results which were basically similar to whole section data." While it would work in a clinical environment, he says, it would only work for laboratories where there's a very high throughput of samples being analyzed. "In the US, where the regulatory bodies are quite different, the use is likely to remain in quality assurance," he says.

Huntsman sees more immediate challenges to the field. He thinks miRNAs and other noncoding RNAs have a lot of potential as biomarkers, "but to my knowledge there isn't an effective way of studying these using tissue arrays. And that's going to be a challenge." Another challenge is not just identifying and localizing proteins, but determining whether they're active post-translation. For this task, "there are real issues with looking at phospho-specific antibodies in archival specimens," he says.