In AP/radiology venture, seeing is believing

September 2007
Feature Story

Karen Titus

When pathologists and radiologists and surgeons talk about cancer, their exchange, in its most elemental form, is about vision. What do they see?

By dint of their training and experience, each sees the same subject differently, each, often literally, through a different lens: stained cells, shadowy images, the tumor itself.

Now those views are starting to converge. Diagnostic medicine is experiencing a eureka moment, in which researchers are linking molecular pathology methods with radiology-based imaging techniques, creating fresh images for everyone. “We’re looking through a different set of glasses,” says Peter Choyke, MD, chief of the molecular imaging program at the National Cancer Institute.

They’re looking down the molecular road, where pathology has long been heading. Now radiology is, too. “Frankly, clinicians are moving that way as well,” says Dr. Choyke. “Because they’re going to need to use genomics and proteomics to help select the patients who undergo screening.” From screening to identifying lesions to pinpointing a diagnosis, it’s all part of medicine’s emerging molecular cascade.

In broad terms, the goal is to marry three-dimensional molecular measurements with three-dimensional imaging. “Imagine a noninvasive, high-throughput way that you can take the various measurements—the various ‘-omics’—and do that in a radiological method. That would be great,” says Michael R. Emmert-Buck, MD, PhD, clinical investigator and chief of the NCI’s Pathogenetics Unit.

That’s what he, Dr. Choyke, and others are pursuing in their work at the NCI. (Although they’re hardly alone—Related article: “In Diagnoses, a Tale of Two Specialties,” CAP TODAY, July 2007) As befits this melding of disciplines, pathologists, radiologists, and clinicians have no choice but to work in concert. This, in and of itself, shouldn’t be remarkable, at least not in theory. As Dr. Emmert-Buck notes, pathologists have long been encouraged to leave their labs and visit with their clinical colleagues. “Everyone says it, everyone knows it, but it doesn’t happen as often as perhaps it should,” he says.

Drs. Emmert-Buck and Choyke meet regularly with two urologic surgeons at the NCI, Peter Pinto, MD, and W. Marston Linehan, MD, and with Maria Merino, MD, chief of surgical pathology. Pathologists can talk all they want about the molecular targets they’re trying to develop, but talk only goes so far. “We get a real quick reality check from Pete Choyke, because he’ll say, ‘OK, but you need to show A, B, and D, or else that’s really just a research project,’” Dr. Emmert-Buck says. “And then the urologic surgeons will jump in and they’ll talk about their patients clinically, and how to refine our strategy to give them the information that they’re really going to need.”

Radiology is learning what pathologists have known for some time: how to harness varying wavelengths of light to do multichannel imaging. It’s impossible to do this kind of multiparametric imaging with MRI or CT, both of which are monochromatic. PET won’t work, either. And while there’s some discussion about whether radionuclides are up to the task, Dr. Choyke says the large amounts of radiation required, coupled with poor resolution, make their use unappealing at best.

Optical imaging, however, is quite feasible, and could transform current imaging methods, such as endoscopy. The current process—using an endoscope and white light to look for anatomic distortions—is imprecise, resulting in multiple biopsies, including those of normal tissue. Labeling the colon, for instance, with colored fluorophores that target cathepsins secreted by colon cancers or epidermal growth factor receptors expressed by the tumor would allow endoscopists using a fluorescence endoscope to identify in situ lesions that warrant biopsy because of their expression profile.

Dr. Choyke and his colleagues are also working closely with Dr. Emmert-Buck and his colleagues on approaches to multiplex analysis of histological tissue sections involving thin-film proteomic layering techniques, which allows them to identify the presence of a protein (or proteins) in a particular part of the tumor. “So there’s fidelity to the anatomic location, which ultimately we’ll understand,” he says. “Right now we don’t understand why tumors behave in certain ways. But having the three-dimensional anatomic localization will help us figure it out.”

Dr. Choyke is also intrigued by another connection between pathology and radiology—the need for live tissue biopsies. “We understand now that the genes and some proteins degrade so fast after removal from the body, the biopsy needs to be captured real-time.”

Dr. Emmert-Buck and his colleagues, for their part, have focused on tissue microdissection, specifically expression microdissection and layered expression scanning. Using these, along with promising methods being developed by others—serial immunostaining for one, quantum dots for another—pathologists will, “at the end of the day, be able to paint very precise, high-throughput measurements from tissue samples,” Dr. Emmert-Buck says.

Expression microdissection, or xMD, uses a targeting probe to obtain cells, eliminating microscopes and the need for a human to visualize cells. A detailed paper, published in Diagnostic Molecular Pathology (Tangrea MA, et al. 2004; 13: 207– 212), explains the process: An ethylene vinyl acetate polymer film coating is placed atop a conventional IHC tissue section (sans coverslip). The light energy passes through the EVA film (which does not contain a light-absorbing dye, unlike film used in laser capture microdissection, or LCM), becoming absorbed only where there are strong deposits of highly absorbing immunostain. The stain heats the overlying clear polymer, causing it to melt focally; the overlying film is unaffected. Thus, the thermoplastic bond is formed only with cells or organelles that exceed the stain threshold concentration. That means cells are selected by immunobased targeting alone.

The early promise of this approach has been shown in subsequent biological studies done by his group and others, says Dr. Emmert-Buck, who credits three of his colleagues at the NIH—Thomas Pohida, PhD, and Robert Bonner, PhD, who also are credited with co-inventing LCM in the 1990s with Lance Liotta, MD, PhD, now of George Mason University—and Michael Tangrea, PhD, with helping develop xMD. It increases dissection rate by several orders of magnitude and improves subcellular precision. “We see it coming out as a new commercial instrument,” he says. “We’re in pretty serious negotiations with one company in particular.”

From a research perspective, xMD is already proving valuable. While LCM has been the workhorse of genome and transcript studies, thanks to PCR amplification, it hasn’t been able to address the equally important proteome. Protein PCR doesn’t exist. And while newer mass spectrometry methods allow fairly deep proteome analysis, they don’t allow researchers to pinpoint specific subpopulations. “Even if you want to do extensive LCM dissection, to get enough protein is, typically, a gargantuan task,” Dr. Emmert-Buck says. Expression microdissection, on the other hand, is a highly efficient recovery process, allowing researchers to see more deeply and more precisely.

The method creates little molecular degradation in DNA and protein, Dr. Emmert-Buck reports. “For the mRNA, we’re less happy.” On the other hand, he says, mRNA may not be the best application for this particular technology anyway.

Nor has IHC staining intensity been an issue. “That’s been one problem we anticipated that hasn’t borne out to be true. And we’re happy about that,” he says.

“The rule of thumb is, there’s a threshold limit to get activation of the xMD film,” he continues. Anything that would be called IHC positive on a microscope slide, based on visual appearance and standard experience, will be procured by xMD. Material exhibiting the light background blush often seen with various antibodies will not be dissected. “If you have a decent antibody, pretty much anything that’s commercially available and used for immunohistochemistry will work for xMD.”

Dr. Emmert-Buck’s other chief area of interest, layered expression scanning, is being co-developed with 20/20 Gene Systems Inc. (www.2020gene.com) and has potential as a clinical tool as well as a research tool. In a nutshell, Dr. Emmert-Buck explains, the approach marries two-dimensional histopathology information on a slide with a third-dimensional molecular array, with its multiplex measurements. “So you’ve got your traditional slide, and then behind it, in the third dimension, are layers of membranes, each of which can measure a different molecular species.”

Without negating these important steps forward, Dr. Emmert-Buck foists some caution on all this work, citing the usual twin suspects of technological hurdles and validation issues.

It makes sense that new methods will require new tools, though this could be less of a problem than one might think. At the NCI, Dr. Choyke says, “We’re very, very fortunate to have highly talented optical engineers on site who have been waiting to help us, it seems.” In one case, researchers made a request for a new tool in November; by January, they had a working model.

Validation is the more overwhelming problem. “It’s incredibly difficult to validate everything at high confidence,” Dr. Choyke says. By the time pathologists process the tissue, much of the anatomic localization is either distorted or lost. “Then it becomes a little bit of guesswork about whether what they’re seeing on the image is actually what you’re seeing on pathology. So these validation steps between imaging and pathology are a major, major problem.”

Likewise, says Dr. Emmert-Buck, the molecular measurement work depends absolutely on bioinformatics. “This is pushing pathology and biology into a complicated statistical and mathematical arena,” he says. As an emerging literature is making clear, initial data interpretations don’t always hold up.

Yet another problem is remarkably empty of technical and scientific challenge. It’s just not easy to marry the two different cultures of pathology and radiology. “It was kind of minimized as a trivial thing, for many years, either that it was not important to address, or that it wouldn’t be a difficult problem to solve,” says Dr. Emmert-Buck. As it turns out, it’s tremendously tough. To wit: NIH grants cannot list two principal investigators. “You have to have one PI. The same with authorship on papers—you can’t have multiple groups with everybody sharing equally in the credit.” “We’ve got to change the rewards mechanisms, so that people who take the initiative to go outside their own lab and try to do this kind of work are helped rather than hurt by the system,” says Dr. Emmert-Buck.

None of this has stopped researchers from moving forward with their work, and all are envisioning a future that could be quite different from the present. Should the quixotic become quotidian, what will radiology and pathology practices look like?

Dr. Choyke predicts complementary roles for the two disciplines. “I don’t see in vivo imaging ever getting to the resolution that pathologists operate at every day, at the cell and subcellular level. But it’s useful in pointing or directing biopsies to specific areas, and increasing the yield of biopsies, and possibly improving the sampling.”

He’s not even sure the optically based tools will wind up in the hands of radiologists. They might be used by surgeons, internists, and oncologists in their offices, he suggests.

As for pathology, he says, “In medicine, the Supreme Court is the pathologist, and rightfully so, because they’re the only group that can inspect tissue at high granularity.” Final diagnoses will remain in the hands of pathologists; the new tools will improve what he calls the working diagnosis. For those who need it spelled out: “You can reassure your readers that the imagers are not going to take their business,” Dr. Choyke says.