|From Laboratory to Clinic: Novel Skin Sampling Technique Simplifies Disease Detection|
For the first time, SLAS is collaborating with the Association for Molecular Pathology (AMP), whose clinical focus is on molecular diagnostic and prognostic medicine, to bring two speakers from the organization to SLAS2013.
One of those speakers is William Wachsman, M.D., Ph.D., an associate professor of medicine at the University of California, San Diego School of Medicine and a staff physician at the Veterans Administration San Diego Healthcare System. A hematologist/oncologist with genomics expertise, Wachsman has developed a novel method for distinguishing melanoma from benign pigmented skin lesions. The process by which he discovered and helped create this new tool is a story of science and serendipity, and includes two salient examples of how, as Pasteur said, “chance favors only the prepared mind.” As both a clinician and researcher, Wachsman works at the interface between basic research and clinical application. His current work, which began as an exploration of ways to facilitate biomarker discovery, could soon revolutionize the way skin cancer is detected, ultimately facilitating early diagnosis and saving time, effort, anxiety and money by substantially reducing the number of unnecessary biopsies.
For SLAS members, Wachsman’s story and his upcoming presentation provide an opportunity not only to appreciate a new technology, but also to “see the larger picture of where much of our work is going,” says John Robinson, Ph.D., chair of the SLAS2013 Diagnostics educational track. “Yes, we want to learn more about innovative technologies, but what we really need to ask is ‘how will they make a difference?’”
An estimated 76,250 new cases of invasive melanoma will be diagnosed in the United States in 2012, and an estimated 9,180 will result in death. Globally, about 132,000 new cases of the cancer are diagnosed annually—and the incidence is on the rise.
Against this backdrop, detection rates for the disease clearly are “suboptimal,” says Wachsman. Melanoma is curable if caught at an early stage, yet research suggests that only 3%-5% of biopsied skin lesions suspicious for melanoma actually contain the disease. Moreover, not all dermatologists biopsy the same lesions, which means a lesion could potentially be missed on initial presentation. “When I first heard this, I was shocked. Healthcare professionals aren’t just doing a lot of biopsies to detect a melanoma; they’re also leaving some melanomas undetected at the stage when they might still be curable.”
Limitations of Current Tools
Diagnosticians generally use visual criteria, including “ABCDE” (asymmetry, border, color, diameter, evolution) and results of optical imaging techniques such as dermoscopy, to distinguish benign from malignant lesions, Wachsman explains. But even with more sophisticated technologies such as digital epiluminescence microscopy, he says, “sensitivity may be very high but specificity is only in the range of 10%. Even histopathology—the current ‘gold standard’ for diagnosing melanoma—has limitations. Experts don’t always agree on what they see, and discordant pathology readings of excised tissue are said to occur in 10% to 35% of potential cases.” Moreover, pathologists look only at representative sections of suspicious lesions, and may miss melanoma due to a sampling error (i.e., they don’t look in sections that contain the melanoma).
A Chance Encounter
As a hematologist/oncologist, “I don’t see patients up front for melanoma. I see people with more advanced disease, who have already been diagnosed,” Wachsman says. “But as a genomics researcher, I had been asking myself whether we could find ways to sample differently—maybe find tumor markers in blood or in urine, or perhaps in the skin itself. I was chatting with a neighbor about this and his eyes suddenly widened. It turned out he was head of a small biotech company that was working on a tape-based method of sampling the surface of the skin and looking at genomic signatures in the cells.”
Within 72 hours of that discussion, Wachsman was meeting with his neighbor and others from Dermtech International, Inc. in La Jolla, CA. Together, they eventually decided to use the tape-based approach to target melanoma, first by assessing whether gene expression analysis of the cells pulled from the tape could distinguish malignant from benign lesions. This method of sampling cells from the stratum corneum—the outermost layer of the epidermis—by means of adhesive tape stripping is now known as “epidermal genetic information retrieval” (EGIR). The sampling process is simple: wearing gloves, a researcher wipes a patient’s skin with alcohol, lets it air dry, and then applies a nickel-sized (17 mm diameter) disc of the proprietary adhesive tape to the lesion. After rubbing the tape on and around the lesion, he/she peels it off.
“When we began our work, the belief was that melanoma does not exist in stratum corneum. In fact, if I were a dermatologist or a melanoma expert who wrote up our hypothesis for a National Institutes of Health grant, the grant would have been rejected. They would have said the idea is completely off the wall—that you can’t detect melanoma there by any method,” says Wachsman, an uncompensated member of Dermtech International’s Scientific Advisory Board. “So we used our own [Dermtech’s] money and forged ahead.”
The researchers knew that melanoma normally is found within the deeper epidermis. It spreads radially, and dives down through the basement membrane into the dermis, as it moves to a higher stage. Thus, when they sampled the skin overlaying the pigmented lesions using EGIR and amplified the RNA isolated from the tape, they were surprised by what they found.
“Anyone who has ever played with RNA realizes it's very labile, and prone to nuclease-mediated degradation. But it turned out that the nucleases in the RNA that came out of the stratum corneum were quiescent, and the RNA we were able to glean off the tape from that layer was only minimally degraded,” Wachsman explains. “As we began our analysis, we immediately started picking up signals that we could see a difference between melanoma and nonmelanoma.” (Their preliminary findings were first presented at the Society for Investigative Dermatology meeting in 2007; abstract 869.)
An Unexpected Tool
Having demonstrated that specimens gathered by EGIR could be used to discern melanoma, Wachsman and colleagues sought to develop a way to predict the risk that a pigmented lesion might be a melanoma rather than a nevus. “Development of a diagnostic is based on class prediction modeling, and here again, it was the coincidence of living in San Diego, and knowing that we needed to move quickly, that led to an unusual solution,” Wachsman says. “A company literally a few miles down the road from us sold risk-management software packages that were used mainly by the financial industry.” After testing several of the software packages, the team settled on TreeNet—a program that produces algorithms used for credit card fraud detection, credit scoring and financial risk management.
TreeNet “generates thousands of very small decision-making trees, and then provides a consensus that allows you to rank the key targets—in this case genes—that are the most important for your class prediction model,” Wachsman explains. “In addition, it allows you to reduce the number of genes and continue to re-test your model. And it allows you to use imperfect data input, which was perfect for what we were doing.”
To test the model, the team used EGIR to collect cells from the stratum corneum of 202 patients with suspicious lesions. The RNA they harvested from the tape strips was quantified, amplified and hybridized, and profiled on a gene chip microarray. Then they randomly divided the resulting microarray data into training datasets, which consisted of data from 37 melanomas and 37 nevi, and a test dataset made up of data from 39 melanomas and 89 nevi.
The training dataset identified 422 genes that were differentially expressed between melanomas and nevi. The TreeNet data mining algorithm yielded a classifier from the training set containing 168 of those genes, and these were shown to have 100% sensitivity and 95% specificity in detecting melanoma. Further modeling enabled the researchers to whittle down the number of genes needed to identify melanomas versus nevi to 17. The 17-gene classifier has 100% sensitivity and 88% specificity, compared with about 10%-30% specificity by current methods, Wachsman observes.
“What also struck us was that for our initial training sets, we would always use pristine data,” Wachsman says. “But when we started testing, we would occasionally have a sample we knew was a melanoma from the pathology result, and even though the microarray data were less than adequate, the algorithm would invariably call it correctly. We were stunned. Even with mediocre data, it consistently outperformed any other class prediction modeling strategy we tried.” In the initial tests, the 17-gene classifier yielded 13 false positives; however, upon further scrutiny, one of those lesions was re-reviewed and found to contain melanoma. (Learn more about class prediction modeling and its role in detection/diagnostics development in Wachsman’s presentation at SLAS2013.)
A New Standard?
Wachsman and colleagues at Dermtech now are working to bring EGIR technology and their gene classifier closer to the clinic. “We’re transitioning over to a quantitative PCR assay because it’s much less costly than using a gene chip, and we want to get costs down to a range that makes it competitive with a biopsy. If it costs more, people will say, ‘why not just do the biopsy?’” says Wachsman. “Yes, EGIR is less invasive and would be preferable from a patient’s point of view for the face or any other cosmetically sensitive area. However, the reality is that economics and third-party payers drive medical systems in the U.S. and in Europe, so cost is the key factor. And a federally-funded plan such as Medicare is unlikely to cover it if it’s more expensive than the current standard of care.”
If EGIR takes off even in a small percentage of the market, “we could potentially be doing hundreds of thousands of assays,” Wachsman adds. “Right now, about 20 biopsies are done for every melanoma, and about 120,000 invasive and in situ lesions are diagnosed each year. To deal with that volume, we will need a high-throughput platform, and a cost-effective, highly sensitive and highly reproducible assay.”
Moreover, the team has used the same tape-based methodology to develop a second multigene classifier that distinguishes invasive from in situ melanoma, and also may be used in melanoma staging. “Our findings suggest that the classifier could aid dermatologists in making clinical management decisions,” Wachsman says.
EGIR, which currently is positioned as a tool to assist in the decision of whether or not a suspicious lesion should be biopsied, could be available within a year “if it’s offered through a single CLIA (Clinical Laboratory Improvement Amendments) laboratory or through a company-based CLIA lab,” Wachsman suggests. “If it has to undergo Food and Drug Administration review as a medical device, that could take two or three years. We hope it comes out sooner, and that it would be offered in Europe and Australia, as well.”
While acknowledging that a lot of work remains to be done, Wachsman envisions a bright future for the use of EGIR technology in melanoma. “It's conceivable, depending upon how we design our validation studies for the quantitative PCR assays, that we could have as a secondary objective whether we're actually outperforming the current standard. Histology based on microscopic review of a biopsy is imperfect. If we were able to pick up melanoma that was being missed on standard-of-care dermatopathology, we just might have a new gold standard.”
More than Skin Deep
Looking at the bigger picture, Wachsman says “the skin is potentially a much bigger and broader source of biomarkers for disease detection than we’ve ever imagined. When it comes to specimen collection, everybody thinks about blood and urine, less so about stool, and for cancer, a tumor biopsy. But blood requires invasive sampling. And while I might be able to place a detector the size of a pea that would rattle around in the bladder and constantly sample urine, that probably wouldn’t be much fun for the patient. But with skin, you can go back to the same site over and over and over again. Using the tape-stripping, the stratum corneum—which is constantly turning over—will give you a snapshot of what’s going on.”
Wachsman further posits that the stratum corneum, like blood, may turn out to be an important source of information not only for skin diseases, but for systemic diseases, as well. “There probably are humoral and other cellular factors that impact how the skin behaves, and that behavior is reflected through the epidermis, which evolves within about two weeks to stratum corneum. So EGIR could eventually provide us with a novel means of picking up systemic signals for a wide range of disorders.”