College of American Pathologists
Printable Version

  GI stromal tumors a model
  for what’s coming’


cap today

September 2004
Cover Story

Related Article:
Antibodies and kits for KIT
GIST prognosis based on histology

William Check, PhD

In the beginning, there were smooth muscle tumors
of the gastrointestinal tract. And darkness reigned over the face of these tumors. "Among the small number of pathologists who cared what they were," says Christopher Corless, MD, PhD, of Oregon Health and Science University, "no one had any idea how to classify them." During the 1970s and 1980s, these tumors were variously called leiomyomas (benign), leiomyosarcomas (malignant), or Schwannomas.

Then, in 1998, light shined forth over this "mass of confusion," when Japanese and Swedish investigators independently identified expression of the cell surface receptor KIT (also known as CD117) in up to 95 percent of a subset of GI tumors. "This was a real breakthrough," says Dr. Corless, who is professor of pathology and director, OHSU Cancer Pathology Shared Resource. "It was the first marker indicating that some of these tumors were distinct." The darkness was rolled back and a new category of GI tumors appeared, called gastrointestinal stromal tumors, or GISTs. "Detection of KIT became the new gold standard in making a diagnosis [among abdominal tumors]," Dr. Corless says.

Not only was KIT expression recognized in GISTs, but mutations in the kit gene were established as the initiating neoplastic events, says Christopher Fletcher, MD, FRCPath, professor and director of surgical pathology at Brigham and Women’s Hospital. Dr. Fletcher calls these findings "revolutionary."

"Prior to this discovery, the term GIST was used," he says, "but purely descriptively. It was a diagnostic wastebasket for many spindle cell tumors in the GI tract." Now the term GIST is specifically used for a subset of GI lesions that are almost always KIT-positive. "KIT provides us with an immunohistochemical marker that we can use to identify most of these tumors, which was a big coup for pathologists," Dr. Fletcher says.

Markku Miettinen, MD, chairman of the Department of the Soft Tissue Division and distinguished scientist at the Armed Forces Institute of Pathology, says, "We now know that almost all tumors previously considered gastric or intestinal smooth muscle tumors are actually GISTs."

Recognition of KIT expression in GISTs provided a huge benefit beyond classification. "Obviously what is most newsworthy is that KIT is a tyrosine kinase receptor," Dr. Fletcher says. "All of a sudden there was a way to treat [GISTs]." Imatinib (Gleevec, Novartis) had just been proved to be effective therapy for chronic myelogenous leukemia, or CML, which is also initiated by a runaway tyrosine kinase, in that case produced by a chromosome translocation. A researcher at Dana Farber Cancer Institute, David Tuveson, hypothesized that the KIT receptor was structurally and functionally similar to BCR/ABL in CML and would also be inhibited by imatinib, Dr. Fletcher says. Tuveson produced in vitro evidence for this idea, which was followed by successful clinical trials in patients with GISTs. "So we rapidly went from having a nice classification, which was good because we pathologists could diagnose GISTs more reproducibly and define them in a meaningful way, to the marvelous discovery that there was a way to inhibit this new mechanism in patients," Dr. Fletcher says.

"This is one of the fastest examples I know of a bench-to-bedside story," he adds. "It sets the paradigm for developing targeted therapeutics by specifically switching off the mechanism that drives tumor cells. Most cancer treatments smash up DNA. You hope you kill more tumor cells than normal cells. Now we are beginning to see targeted therapies."

Dr. Corless agrees that GISTs are "basically a model for what is coming." For example, the kinase inhibitor gefitinib (Iressa, AstraZeneca) is being developed for lung and colon cancer. "We have a lot better handle on the biology of those tumors than for many other cancers," he says. "The correlation between mutation testing and treatment is pointing to where a lot of other cancer therapies will be going in the future."

At the root of all this is immunohistochemistry, or IHC. Dr. Fletcher wrote in an editorial: "As other new pharmacologic agents are discovered by screening for small molecule inhibitors of oncoproteins, it is likely that there will be a rapid increase in the immunohistochemical stains that pathologists will be asked to perform to detect potential therapeutic targets" (Am J Clin Pathol. 2003; 119: 325-327).

Says Dr. Miettinen: "Pathologists have become very aware of GISTs because they represent a new type of diagnosis that has a special clinical correlation. There is a receptor tyrosine kinase activation that can be countered with a specific inhibitor drug." He sees KIT IHC as the key to GIST diagnosis: "GIST is one of not so many tumors that are almost always KIT-positive."

Elizabeth Montgomery, MD, associate professor of pathology and director of clinical GI pathology at the Johns Hopkins Medical Institutions, says, "When confronted with a spindle cell tumor in the GI tract, the two main points are that stromal tumors should always be in the differential diagnosis and that CD117 IHC should virtually always be performed." Anatomic pathologists are now so aware of GISTs that they sometimes overdiagnose them, though, for the most part, she says, "they do an excellent job with GISTs."

Beyond the simple fact of doing IHC for KIT in suspected GISTs lie a number of complex and challenging issues:

  • How do mutant KIT oncoproteins induce tumors?
  • How does imatinib kill KIT-positive tumors?
  • What histological features tip off the pathologist that a tumor is a possible GIST?
  • Is there anything to help choose among antibodies or assays for KIT?
  • What should be done when the KIT IHC turns up negative?
  • What non-GIST tumors can give positive KIT results?
  • Are all KIT-positive tumors candidates for imatinib therapy?
  • Is it possible to determine the likelihood of an individual GIST responding to imatinib?
  • How does the pathologist predict malignant potential in a GIST?

Clear answers are available to some of these questions, while answers to others are evolving.

In its wild-type state, kit is a cell surface receptor whose natural ligand is stem cell growth factor. At the same time that the Japanese scientists recognized KIT on the surface of GISTs, they found that mutations in the gene coding for this receptor activate the kinase constitutively. Mutant isoforms of KIT become autonomous—they phosphorylate tyrosine residues on all types of signaling proteins that bind to or interact with them in the absence of the growth factor. Mutant KIT receptors also phosphorylate each other, called autophosphorylation. "Assessing KIT autophosphorylation is a commonly used approach in assessing the effectiveness of inhibitors like imatinib," Dr. Corless says.

"Mutant KIT escapes normal regulation," Dr. Miettinin adds. It sends signals into the interior of the cell along the signal transduction pathway that ultimately leads to the nucleus to promote cell proliferation and inhibition of apoptosis.

Imatinib inhibits autophosphorylation of KIT. Dr. Corless says that several groups, including his, proposed to Novartis to treat GIST with imatinib. "Their response," he recalls, "was ’What the hell is a GIST?’ Outside pathology, few people had heard of GISTs." Eventually, a patient in Finland was given imatinib for metastatic GIST to the liver in a compassionate-use protocol. "The patient responded beautifully," Dr. Corless says. "Clinical trials with imatinib subsequently showed that about 80 percent of patients have a response or significant stabilization of their disease" (Demetri GD, et al. N Engl J Med. 2002;347: 472-480). Imatinib is now approved for treating advanced GISTs, which make up about one-third of the approximately 5,000 new cases of GIST diagnosed each year in the United States.

"Inhibitor treatment is now given to patients who have metastatic or unresectable GISTs," Dr. Miettinen says. "Neoadjuvant imatinib has also been used as a precautionary measure immediately following surgery even without evidence of metastasis, mostly in clinical trials."

Already a second kinase in hib it or, SU-11248 (Pfizer), is in phase three trials for GIST patients who show imatinib resistance. About 65 percent of such patients have had either partial response or stabilization on this new agent, Dr. Corless says. "Based on mutation analysis, we suspect it might do even better as a first-line drug for some patients," he proposes.

What we now know as GISTs were initially considered smooth muscle tumors simply because they occur in the gastric or intestinal wall amidst smooth muscle, Dr. Miettinen explains. "However, it has turned out that these tumors are closely related to interstitial cells of Cajal [ICC], which are cellular intermediates between nerves and smooth muscle cells," he says. GISTs either arise from ICC or share a common stem cell with them. What is key to the GIST story is that ICC express kit.

Even before staining for KIT, a pathologist will typically suspect a GIST based on histology. "You always know," Dr. Montgomery says. GISTs have a characteristic morphological pattern and histologic features somewhat akin to—but not identical to—smooth muscle tumors. "When one is confronted with a tumor that looks almost, but not quite, like a classic muscle tumor," she says, "one says, ’Aha! This is a GI stromal tumor.’"

Dr. Miettinen says Cajal cells are normally dispersed around the myenteric plexus and in the muscle layer of the GI tract. "Whenever you see these cells in masses," he says, "it is either Cajal cell hyperplasia or a GIST." Another clue is that, in the stomach, GIST is by far the most common mesenchymal tumor. In the intestine the proportion of GISTs to other tumors is less striking, but still the majority of mesenchymal tumors in the intestine are GISTs as well.

Staining by IHC for KIT provides confirmation for a specimen suspected from histology of being a GIST. A number of antibodies are available for this purpose. Most are analyte-specific reagents, or ASRs, but Ventana’s Pathway c-KIT assay has just been approved by the Food and Drug Administration. It is approved as an automated aid to diagnosis and as an aid for the selection of GIST patients who may qualify for Gleevec therapy.

Thomas M. Grogan, MD, Ventana’s chief medical officer, says this automated assay offers two advancements. First, he says, "the automation addresses the complex assay conditions including tight control of temperature and buffer pH to allow reliable results. Second, the rabbit monoclonal [clone 9.7] reduces the false-positivity found in polyclonal assays." Finally, Dr. Grogan says, their comparison of the Pathway c-KIT clone 9.7 rabbit monoclonal to mutational analysis shows an 88 percent concordance when mutations to exons 9, 11, 13, and 17 are examined. "Importantly, among the GIST cases that were c-KIT negative by IHC [10 percent of GISTcases], many of these [56 percent] demonstrated PDGFRA mutational change," he says.

DakoCytomation’s c-KIT pharmDx assay is at the FDA now awaiting approval.

Dr. Corless has worked with most available IHC methods. "I don’t have a position to take on any particular antibody," he says. "All of them work if used correctly." In his view, it is more important to titer the antibody, whether it is a prepackaged kit or an ASR. "It is important that the titer is set so that there is reasonable staining of GIST samples but no staining of smooth muscle elements or epithelial elements in control sections," Dr. Corless says.

Dr. Fletcher notes that one of the major commercial antibodies has been demonstrated to be less specific, though both antibodies tested were unreliable when used with heat-induced epitope retrieval and at low (1:50) dilution (Lucas DR, et al. Am J Clin Pathol. 2003;119: 339-345). "Most IHC is performed with antigen retrieval," says Dr. Fletcher. "It turns out if you use antigen retrieval with the most widely utilized anti-KIT antibody, which comes from Dako, that gives you false-positives in many other tumor types."

Kenneth Bloom, MD, medical director of ChromaVision, says workup of GIST involves more than one antibody. At the time of his interview with CAPTODAY, he was senior medical director of USLabs, where he used a panel consisting of antibodies to KIT, s100, CD34, PDGFRA, smooth muscle actin, desmin, and beta-catenin in the workup of most cases submitted to rule out a GIST. "Even if there is aberrant staining with one of the antibodies, that should not necessarily confuse the diagnosis," he says. The diagnosis can be made in most cases by the overall histologic pattern plus the immunophenotype.

Under optimal conditions, about 95 percent of GISTs stain positive for KIT. (About 85 percent of GISTs have KIT mutations, Dr. Corless explains; in some cases KIT appears to be activated by a mechanism other than mutation). Drs. Fletcher and Corless and others have shown that many KIT-negative GISTs (four to five percent of all GISTs) have mutations in another transmembrane signaling protein, platelet-derived growth factor receptor alpha, or PDGFRA (Heinrich MC, et al. Science. 2003;299: 708-710). These authors wrote: "Tumors expressing KIT or PDGFRA oncoproteins were indistinguishable with respect to activation of downstream signaling intermediates and cytogenetic changes associated with tumor progression." Importantly, some PDGFRA-positive GISTs respond to imatinib. "A small but clinically relevant fraction of GI tumors will not be stained by antibodies to KIT but are still GISTs, and these patients deserve a trial of imatinib," Dr. Corless says.

Dr. Montgomery agrees: "Pa thologists should be aware that [a small fraction] of GI stromal tumors don’t express CD117, but a subset of those may still respond to imatinib" (Medeiros F, et al. Am J Surg Pathol. 2004; 28: 889- 894). As a result, she says, there is a role for mutation analysis for KIT and PDGFRA in a subset of tumors. A small number of laboratories, including Dr. Corless’, offer mutation analysis.

In the original clinical trials of imatinib in GISTs, KIT staining was an absolute requirement. "Now that we have learned that a subset of GISTs are KIT-negative," Dr. Fletcher says, "we will have to go back to relying more on morphology, at least in the short term." Dr. Corless puts it more strongly. "If any company tries to say that you have to use their reagent to demonstrate GIST and prescribe imatinib, I would say that is [nonsense]," he says. "A diagnosis should come from a pathologist, not an antibody."

Faced with a KIT-negative tumor that looks like a GIST by morphology, an oncologist might say, "I’m going to give imatinib anyway and see what happens." Dr. Fletcher proposes an alternative: Send the case to a specialized center where they have seen very large numbers of GISTs over the past few years—as opposed to a few cases a year—and see if they think it is a GIST.

Says Dr. Corless: "This tumor has evolved from being a new diagnostic entity to a very complex family within the space of a few years. It is not fair or appropriate at this time to say something should be KIT-positive or -negative for diagnosis and treatment. We have to be a lot more sophisticated in our approach to these tumors."

Pathologists also need to be sophisticated about KIT-positive non-GISTs. Dr. Miettinen says a number of other tumor types show non-specific CD117 staining. "The one that causes the most trouble," he says, "is mesenteric fibromatosis. In my consult material, I have more than once received a specimen in which the referring physician was looking for a reason to use imatinib treatment and it has been desmoid."

Positive IHC for KIT can also occasionally be seen in mastocytosis, testicular and ovarian seminoma, sinonasal T-cell lymphoma, rare examples of thymic carcinoma, angiosarcoma, and Ewing sarcoma. All must be considered in the differential diagnosis of a KIT-positive tumor.

Dr. Montgomery agrees that KIT-positive desmoids are the most problematic, because they occur in the GI tract. "If you have a mass in the bowel wall, it isn’t going to be a seminoma," she says. "It will be a desmoid."

It is critical for pathologists to be aware of non-GIST tumors that can give positive KIT results. Oncologists often look for an excuse to use imatinib because it is relatively nontoxic and because patients have been bombarded by promotional material about it. "We get requests for KIT testing on everything because clinicians want to throw imatinib at it," Dr. Montgomery says. She calls such requests "frivolous and inappropriate." She and her colleagues do the test, "then tell the doctors that the result may be meaningless," she says.

Dr. Fletcher has had similar experiences. "It was believed for a while that any type of tumor that was KIT-positive would be a reasonable target for imatinib," he says. "So some oncologists were asking that almost any bad cancer—ovarian, breast—especially if it was metastatic, be stained for KIT." But KIT expression in non-GIST tumors doesn’t result from mutations in the kit gene, so it doesn’t have the same biological meaning as in GISTs. "Imatinib has no therapeutic effect on those other tumors," Dr. Fletcher says. "The key message is that unless there has been initial work to demonstrate that a given type of tumor has kit mutations, there is no basis for doing KIT immunostaining and trying imatinib empirically." His concern: "That this approach can destroy the value of a good drug by using it in tumors that were destined never to respond."

Some good has resulted from overtesting for KIT. "It has emerged in recent months, though it is not well understood, that other tumors that do not have kit mutations may respond to imatinib therapy," Dr. Fletcher says. For example, in some cases, dermatofibrosarcomas have activated platelet-derived growth factor receptor beta, which can be susceptible to imatinib.

Because the basic biology of GISTs has been so well elucidated, molecular characterization can be used to predict the likelihood of an individual tumor responding to imatinib. "Molecular characterization of mutations in the kit gene correlates very well with response to imatinib," Dr. Fletcher says. "Such a test was not available at all four to five years ago," he notes, "then it was developed as a research tool." Because it is so good at predicting treatment response, it has now been made available as a clinical test by Dr. Corless’ laboratory, he says, and other laboratories are setting it up.

This test, performed on paraffin-embedded tumor, is an analysis to determine the presence and exon location of mutations in kit or pdgfra. Dr. Corless says the correlation between the location of the mutation in the kit or pdgfra genes and response to imatinib came out in phase two trials. Tumors with mutations in exon 11 of the kit gene have the best response to drug and are, fortunately, most common. Those with exon 9 mutations do not respond as well overall. Interestingly, the small fraction of tumors that are wild-type for both kinases, about 12 percent, have the worst response, presumably because they have a different molecular pathway to the oncogenic state. Pediatric GISTs are virtually all wild-type, Dr. Corless says, as are GISTs associated with neurofibromatosis; however, imatinib response data are not yet available for these subsets of patients.

Despite its value, mutational analysis for GISTs is only available in a few centers and is not always understood by treating clinicians, according to Dr. Fletcher. "When advances go so fast," he says, "obviously there will be a lag time for people to get used to these new approaches."

In contrast to the high-tech nature of the test for predicting response to imatinib, the criteria for estimating malignancy and predicting recurrence are standard anatomic pathology measurements: tumor size and mitotic index. Says Dr. Miettinen, "I believe that these two factors are important enough to warrant inclusion in any pathology report on GISTs." In general, he says, tumors with five or fewer mitoses per 50?? field and those that are 5 cm or less in diameter are clinically favorable. "However," he notes, "now we know that gastric tumors of equal parameters tend to behave better than intestinal ones. Even large tumors that have very low mitotic activity tend to be more benign in the stomach." Dr. Miettinen estimates that less than 25 percent of gastric GISTs are malignant by clinical course, compared with 40 to 50 percent of intestinal GISTs. Gastric tumors make up 60 to 70 percent of GISTs.

Guidelines for prognosis are included in the publication from the 2001 National Institutes of Health consensus meeting on GISTs (Fletcher CDM, et al. Hum Pathol. 2002;33:459-465). "Those guidelines produce a framework for individual cases but are not perfect," Dr. Fletcher cautions. He agrees that mitotic index and tumor size are powerful predictors. But he adds, "The fact remains that there are still tumors that don’t live by those rules. There are well-characterized examples of very small tumors—some even less than 2 cm—that give rise to metastases, and examples of tumors greater than 2 cm with very few mitoses that give rise to metastases. These exceptions," he says, "remain a focus of great attention." It may be helpful to identify upfront GISTs that have the potential for aggressive behavior and possibly even consider neoadjuvant treatment with imatinib before there is evidence of spread, in Dr. Fletcher’s view. "There are clinical trials going on right now in which persons with primary tumors without evidence of spread are randomized to surgery followed by imatinib versus surgery alone," he says. Selecting the most aggressive tumors for such testing would be desirable.

More accurate identification of potentially malignant GISTs could have financial benefits as well. Imatinib costs $30,000 to $45,000 per year of treatment. "Greater understanding of the biological behavior of GISTs would help to optimize treatment and target it to the patient population most likely to benefit," Dr. Miettinen says.

He points out another curious feature about the course of GISTs. "Some low-grade GISTs metastasize 10 to 15 years after surgery," he says. The longest history in the AFIP files is a patient who is alive 10 years after total gastrectomy for a recurrence that happened 30 years after the primary tumor. Dr. Miettinen and his colleagues at the AFIP are conducting a followup study on gastric and intestinal GISTs to understand their natural history and to help optimize treatment.

In theory, Dr. Montgomery says, patients should probably be followed forever after treatment for GIST. An exception might be made for the smallest tumors. "If a tumor is less than 1 centimeter and completely excised, we can probably let those patients relax," she says.

Adherence to size and mitotic index for determining prognosis in GISTs is not universal. "Henry Appelman has published morphologic criteria for gastric GISTs that he believes allow separation of benign from malignant tumors," Dr. Montgomery says. (See "GISTprognosis based on histology.") "The trouble is that every once in a while a tumor with benign features will metastasize. Prediction based on morphology is not so easy for GISTs," she points out. "For adenomas or carcinoma of the colon, we are quite sure which ones are benign and malignant by looking through the microscope. For GISTs, we will probably need extra markers."

Summing up these exciting findings, Dr. Corless says: "This has been a fun story. We have discovered that imatinib is effective not only for GISTs and CML, but also for dermatofibrosarcoma protuberans, hypereosinophilic syndrome, and CMML [chronic monomyelocytic leukemia]." The common theme is that all five of these tumors have activation of a kinase pathway that is targeted by imatinib. Dr. Miettinen believes that the KIT story presages the future of this field.

"Because there are at least 60 known receptor tyrosine kinases," he predicts, "new connections between receptor mutations and disease will undoubtedly be made in the near future."

Dr. Corless notes a recent example. "It was just published that EGFR [epidermal growth factor receptor] mutations in non-small-cell lung cancer correlate with response to gefitinib, which is a completely analogous situation," he says (Lynch TJ, et al. N Engl J Med. 2004;350:2129-2139; Paez JG, et al. Science. 2004;304(5676):1497-1500. Published online April 29, 2004). As more discoveries of this type are made, he predicts, "we will see increasing emphasis on the use of mutation testing as a way of assigning patients into therapeutic categories."

William Check is a medical writer in Wilmette, Ill.