Predictive pathology: a patholoGIST’s perspective



Peter MacCallum Cancer Centre
Melbourne, VIC

The history of how gastrointestinal stromal tumours (GISTs), an obscure soft tissue tumour of the bowel wall, came to occupy the center stage of oncology during 2001/2002 best illustrates the “Current face of cancer pathology in Australia”, the forum subject of this edition of Cancer Forum.

GISTs arise predominantly in the stomach and small intestine, but also occur in the rectum, oeophagus and a variety of other locations including the mesentery. They are relatively rare tumours with an estimated incidence of 100 new cases per year in Australia. The majority of GISTs behave in a benign manner but larger tumours and those with a higher mitotic index can be aggressive. These malignant GISTs follow a characteristic clinical course that includes local recurrence at the site of resection, intra-abdominal spread on serosal surfaces and the development of liver metastases. Notably, radiation therapy and chemotherapy, used singly or in combination, are ineffective in the treatment of advanced disease.

For many decades GISTs were thought to be neoplasms of smooth muscle origin and were therefore classified as leiomyomas, leiomyosarcomas and leiomyoblastomas. When electron microscopy and immunohistochemistry were applied to these tumours however, there was little evidence of smooth muscle differentiation, and their histogenesis became uncertain. In the early 1990s it was discovered that most GISTs were positive for CD34, but this gave no clue to their cell of origin1. In 1997, Drs Seichi Hirota (Osaka University) and Lars-Gunnar Kindblom (University of Gothenburg) each independently observed that GISTs express the receptor tyrosine kinase KIT (CD117). This observation provided the clue that GISTs originate from the interstitial cell of Cajal (ICC). These inconspicuous dendritic-like cells are widely distributed throughout the muscularis propria of the GI tract where they play an important role in gut motility. Like GISTs, ICCs express CD34 and KIT. A landmark paper the following year from Hirota et al2 transformed the field. They showed that GISTs not only express KIT but also contain mutations in the KIT gene that result in activation of the tyrosine kinase. Many laboratories have since established that KIT mutations are present in >85% of GISTs, primarily in exon 11 and less commonly in exons 9 and 13. The mutations are invariably in-frame deletions and the mutant isoforms, when cloned and expressed in vitro, have constitutive kinase activity. These mutations are thought to favour spontaneous dimerisation of KIT in the absence of its natural ligand, stem cell factor, leading to constitutive phorphoryation and activation of the tyrosine kinase.

Meanwhile, Dr Brian Druker (Oregon Health and Science University), in collaboration with Novartis Pharmaceuticals, was working on molecules that block the binding of ATP to tyrosine kinases. He discovered that STI571 (imatinib mesylate; Glivec) inhibited the kinase activity of ABL and the BRC-ABL fusion gene product of the Philadelphia chromosome. Glivec received worldwide attention in 2000 for its effectiveness against chronic myelogenous leukaemia (CML) which is driven by the constitutive activation of the BRC-ABL fusion gene. More than 85% of chronic phase CML patients taking one oral dose per day achieve a complete haematological response and many reach a complete cytogenetic remission. Glivec received FDA approval in the US for the treatment of CML in May 2001.

Glivec however, is not perfectly specific and inhibits tyrosine kinases that are closely related to ABL, including ARG (ABL-related kinase), PDGRFA and PDGRFB. Dr Michael Heinrich (Oregon Health and Science University) demonstrated that Glivec is also a potent inhibitor of KIT in vitro and Dr Jonathan Fletcher (Brigham & Women’s Hospital, Boston) showed that the drug could inhibit the growth of a GIST cell line. Remarkably, Novartis declined suggestions that it would be worthwhile conducting a clinical trial to explore the effectiveness of Glivec in GISTs. The remarkable efficacy of this drug in patients with these tumours may have gone undiscovered had it not been for the persistence of the husband of a Finnish woman who had a locally advanced GIST with metastases to the liver. He, apparently, bought a large number of Novartis shares and demanded to discuss his wife’s dilemma with the company’s president. Compassionate use of Glivec was granted to his wife in March 2000. Within a matter of weeks the liver metastasis showed an overall reduction in size of 75% and six of 28 liver metastases were no longer detectable in CT scans after eight months of therapy. The drug was well-tolerated and the patient reported that all cancer-related symptoms had disappeared. The clinical response correlated with near complete inhibition of [18F] fluorodeoxyglucose uptake on PET scan. A post-treatment biopsy showed a marked decrease in tumour cellularity and extensive myxoid degeneration. The publication that followed3 sent a shock wave through the oncology community around the world. The long-awaited promise of effective targeted therapy against a solid tumour had finally arrived. In it lay the hope that patients with incurable tumours could now live in long-term remission with their tumours without the need for toxic chemotherapy and radiotherapy.

This ‘proof of principle’ case quickly convinced Novartis to conduct multicentre clinical trials in the US and Finland (STIB2222), Europe (EORTC Soft Tissue and Sarcoma Group) and Australia (Australian Gastro Intestinal Trials Group). The rate of patient recruitment into these trials was unprecedented. The B2222 trial showed that 75% of patients had a partial response or had stable disease4. Most patients reported almost immediate improvement in well-being and in several cases PET activity was markedly reduced within 24 hours of commencing treatment. On the basis of these results Glivec was approved in the US by the FDA for the treatment of unresectable and metastatic GISTs in February 2001, a record eight months after the NEJM publication. Novartis also sponsored another multicentre trial (STI571B2225) to determine whether Glivec was effective in non-GIST tumours that express KIT or related tyrosine kinases. Recruitment into the B2225 trial was on the basis of a tumour being strongly KIT positive by immunohistochemistry or a likely biological rationale such as known involvement of PDGRFA or B in the biology of the tumour.

The requirement of these trials that tumours be strongly KIT positive by immunohistochemistry effectively placed the onus for deciding patient eligibility for treatment on pathologists. Several pathologists, including myself, were caught unprepared for this new role and it quickly became apparent that there was significant inter-laboratory variation in KIT immunohistochemistry reporting. With hindsight it is now apparent that most laboratories had optimised their KIT immunohistochemistry only for the purpose of diagnosing GISTs and not for identifying a therapeutic target. The optimal concentrations of the antibodies and the specificity and sensitivity of the test had not been validated against the wide range of tumours that were included in the B2225 trial. It is worth remembering that the predictive value of any test is dependent upon the prevalence of the disease. In other words, for any given sensitivity and specificity, the false-positive and negative rates of a test will vary depending upon the pre-test probability of the disease. In the context of an intra-abdominal spindle cell tumour (where there is a very high pre-test probability of the tumour being a GIST), a positive or negative KIT immunoassay result effectively confirms or excludes a GIST, respectively. Outside this context the false positive and negative rates will rise. Accordingly, several patients with KIT positive tumours were enrolled in the trial and treated with Glivec but their tumours were subsequently shown to be KIT negative by the reference laboratory that had carefully validated the assay for this purpose.

The results of the B2225 trial were interesting for several reasons. It quickly became clear that many tumours expressed KIT but almost all, except GISTs, were non-responsive to Glivec. Three tumour types however, did show a consistent clinical response, namely myelomonocytic leukaemia, dermatofibrosarcoma protuberans and hypereosinophilic syndrome. All were either known or were later found to contain chromosomal translocations or mutations that involve the PDGFRA or B genes, which result in constitutive activation of PDGFR kinase activity in the same manner as KIT and ABL. Glivec responsiveness, therefore, requires the presence of an activating mutation in a target tyrosine kinase.

Glivec is currently under consideration by the Pharmaceutical Benefits Advisory Committee for treatment of GISTs. It appears likely that patients with KIT-positive GISTs will soon be eligible for initial treatment with Glivec. Accordingly, patient eligibility will be determined routinely by pathologists on the basis of KIT immunohistochemistry. Given the cost of the drug (approximately $50,000 per patient per annum) and its effectiveness, it is important for all laboratories to carefully validate their KIT immunohistochemistry tests. In a series of 28 cases with a prior diagnosis of malignant GIST, reviewed at the Peter MacCallum Cancer Centre during the past two years, eight cases (30%) proved not to be GISTs on the basis that they were CD34 negative and had no detectable mutation in the KIT gene. Five of these cases had been previously reported as KIT positive but the staining pattern was different from the strong membranous and cytoplasmic staining typically seen in GISTs5. The weak cytoplasmic staining seen in the five tumours was similar to that seen in reactive fibroblasts when too high an antibody concentration is used. In addition, two of the 20 true GISTs proved to be KIT-negative and these tumours probably contain mutations in other genes such as PDGFRA. Our series was undoubtedly biased in favour of diagnostically difficult cases, but it highlights that there is a significant error rate for the diagnosis of malignant GIST. Several patients are likely to receive an ineffective treatment while others will be denied a potentially effective treatment. Clearly, as with HER-2 immunohistochemistry in metastatic breast cancer, there ought to be a second line test to help sort out the indeterminate cases. The two options are KIT and PDGFRA mutation detection (which can be performed on DNA extracted from paraffin-embedded tissues) or Western blots using anti-phosphotyrosine antibodies (which would require fresh tissue). A strong case could also be made to refer all difficult and non-responsive cases to a central referral laboratory in the same manner as HER-2 FISH.

The development of new targeted therapies is creating a new role for pathologists who increasingly will be required to identify therapeutic targets. This will necessitate the inclusion of pathologists from the earliest stages of drug development and clinical trial design to ensure that the development and evaluation of a new drug occurs in parallel with the diagnostic reagents and standardised tests required to identify the target. There are apparently several hundred drugs, such as Glivec, being developed by pharmaceutical companies. These will undoubtedly usher in a new era of predictive pathology where therapeutic target identification will be an essential tool in the patholoGIST’s ‘kit’.


1. CD Fletcher, JJ Berman, D Corless, F Gorstein F, et al. “Diagnosis of gastrointestinal stromal tumours: A consensus approach.” Human Pathology, 33 (2002): 459-65.

2. S Hirota, K Isozaki, Y Moriyama, K Hashimoto, et al. “Gain-of -function mutations in c-kit in human gastrointestinal stromal tumours.” Science, 279 (1998): 577-80.

3. H Joensuu, PJ Roberts, M Sarlomo-Rikala, LC Andersson, et al. “Effect of the tyrosone kinase inhibitor STI571 in a patient with a metastatic gastrointestinal stromal tumour.” NEJM, 344 (2001): 1052-6.

4. GD Demetri, M Von Mehren, CD Blanke, AD Van den Abbeele, et al. “Efficiacy and safety of imatinib mesylate in advanced gastrointestinal stromal tumours.” NEJM, 347 (2002): 472-3.

5. M Sabah, M Leader, E Kay. “The problem with KIT: clinical implications and practical difficulties with CD117 immunostaining. Appl Immunohistochem Mol Morphol, 11 (2003): 56-61.

Be the first to know when a new issue is online. Subscribe today.