Examples of staining of gastointestinal stromal tumors (GISTs). A, Histologic appearance of primary GIST demonstrated by hematoxylin-eosin staining. The GIST is composed of sheets of spindle cells. B, The same tumor stained with CD117 with polyclonal rabbit antihuman antigen shows the strong positive staining for CD117. C, This is the negative control for CD117 with polyclonal rabbit antihuman antigen staining. D, The same GIST stained for ras p21 with monoclonal mouse antihuman ras p21 demonstrates strong positivity for ras protein. E, This is the negative control for ras p21 with monoclonal mouse antihuman ras p21.
Blair SL, Al-Refaie WB, Wang-Rodriguez J, Behling C, Ali M, Moossa AR. Gastrointestinal Stromal Tumors Express ras OncogeneA Potential Role for Diagnosis and Treatment. Arch Surg. 2005;140(6):543-548. doi:10.1001/archsurg.140.6.543
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
Gastrointestinal stromal tumors (GISTs) constitute the largest category of nonepithelial neoplasms of the gastrointestinal tract. Histologically, they have a spindle cell appearance but stain by immunohistochemistry for the proto-oncogene, c-kit (CD117). There is some evidence that phosphorylation of these receptors leads to a cascade that may activate the ras/mitogen-activated protein kinase pathway, which may, in turn, allow other oncogenes to become active.
Immunohistochemical staining pattern of GISTs will aid in their differentiation from other spindle cell tumors and predict clinical outcome in patients.
Design and Setting
Retrospective review of patient records and paraffin block specimens of spindle cell tumors.
We have identified 65 patients with spindle cell tumors of the gastrointestinal tract at our institution in the past 10 years. Tumors were diagnosed by their morphology as leiomyomas, leiomyoblastomas, or leiomyosarcomas.
Main Outcome Measures
CD117 and ras p21 were stained by immunohistochemistry on formalin-fixed, paraffin-embedded sections of normal and tumor tissues.
Of the 65 patients, there were 23 patients diagnosed as having GIST confirmed by CD117 expression and 42 patients without GIST. Gastrointestinal stromal tumor samples of 17 (77%) of 22 patients stained positive for ras protein compared with 0 of 27 patients with leiomyomas (P<.001).
To our knowledge, this study is the first to demonstrate that GISTs stain positive for ras p21. This molecular trait may be a useful diagnostic tool in addition to the c-kit (CD117) to separate GISTs from leiomyomas and leiomyosarcomas. In the future, ras inhibitors may potentially be a therapeutic to treat GISTs.
Gastrointestinal stromal tumors (GISTs) are uncommon tumors representing only 1% to 3% of all gastrointestinal tract malignancies, but they are still the largest category of primary nonepithelial neoplasms of the gastrointestinal tract. Previously, they were believed to arise from smooth muscle cells and were designated as leiomyomas, leiomyosarcomas, or leiomyoblastomas. More recently, the interstitial cell of Cajal, intestinal pacemaker cells, are believed to be the cell of origin of most of these tumors. These tumors are readily identifiable because they stain for CD117, the c-kit proto-oncogene. Kit is a 145-kDa transmembrane glycoprotein that is a product of the c-kit gene. A member of the receptor tyrosine kinase subclass III family, kit is normally expressed by the interstitial cells of Cajal and by hematopoietic progenitor cells, mast cells, and germ cells.1
The clinical behavior of GISTs is varied, which poses difficulties in determining the degree of aggressiveness and prognosis; however, certain factors were found to be determinants of outcome. The overall 5-year disease-free survival rates in patients with malignant GISTs is between 20% and 60%. Surgical resection remains the mainstay of treatment, but these tumors have a high propensity for local recurrence. Traditional chemotherapy and radiation treatment have not been shown to improve disease-free survival. Large tumor size, cell necrosis, and high-grade and incomplete surgical resection have all been shown to decrease disease-free survival.2- 5 Controversy exists whether the staining of a tumor for CD117 translated into a malignant phenotype or whether there exists benign CD117-positive tumors that clinically behave like leiomyomas with no propensity to recur.
Currently, with the advance of endoscopic ultrasonography, a biopsy specimen of even small submucosal tumors in the stomach may be obtained by minimally invasive techniques.6 With small specimens removed, it can be difficult to differentiate spindle cell tumors. Furthermore, at this point, we still recommend surgical resection for these tumors because we cannot always diagnose preoperatively the exact histologic features of spindle cell tumors or their propensity to progress locally or metastasize based on available pathologic factors. Many molecular markers have been used in other tumors to predict tumor aggressiveness besides standard histologic parameters. For example, researchers have studied immunohistochemical factors such as Ki67 and P53 with mixed results.7- 10
Many human tumors such as colorectal and lung cancers have been shown to express activated versions of the ras family of genes whose products are known as the ras p21. The ras p21 are guanosine triphosphate–binding proteins that have been shown to be involved in events downstream of tyrosine kinases. Likewise, kit is a tyrosine receptor; there is some evidence that phosphorylation of these receptors leads to a cascade that may activate the ras/mitogen-activated protein kinase pathway that may, in turn, allow other oncogenes to become active and accelerate tumor progression.11,12 We are unaware of any previous study of ras expression in GISTs. Therefore, the purpose of this study is to identify immunohistochemical evidence in GISTs and help predict the malignant behavior of these tumors.
We identified 65 patients having GISTs treated at the University of California at San Diego over the past 10 years. After obtaining institutional review board approval, we retrospectively obtained tumor specimens and reviewed the patients’ records for demographic information and outcome.
Routine staining in spindle cell tumors for CD117 began 3 years ago at our institution. Prior to that point, tumors were diagnosed by their morphology as leiomyomas, leiomyoblastomas, leiomyosarcomas, or as unspecified sarcomas. In this cohort of patients at the time of their initial diagnosis, the tumors of 42 patients were categorized as leiomyoma, 7 as leiomyosarcoma, 7 as sarcoma unclassified, 6 as leiomyoblastomas, and 3 as GISTs.
Paraffin tissues were cut into 5-μm-thick sections, mounted on slides, and air dried for more than 24 hours. Paraffin-embedded tissues were dewaxed by incubation at 60°C for 1 hour and submerged in xylenes (2×5 minutes), followed by a 100%, 95%, and 80% ethanol series (2×2 minutes each), and placed in deionized water for 2 minutes. Prior to tissue staining, tissue slides were treated with heat-induced, ×10-concentration Target Retrieval Solution (code No. S1700; DakoCytomation Inc, Copenhagen, Demark) for 75 minutes and rehydrated in a Tris-buffered saline solution. Standard immunohistochemical procedures were performed using hematoxylin counterstaining. A Tris-buffered saline solution was used between each washing step. Endogenous peroxidase activity was quenched with 3% hydrogen peroxide in a Tris-buffered saline solution for 20 minutes. All slides were applied with a protein block serum (Protein Block Serum-Free, code No. X0909; DakoCymation Inc) for 1 hour. Sections incubated with secondary antibodies alone served as negative controls. For color development, the slides were applied with a substrate-chromagen solution (DakoCytomation Inc) for 5 to 10 minutes then counterstained with hematoxylin. A Ki67-labeling index was calculated for each case. The percentage of Ki67-labeled cells was calculated 3 times for each tumor, and a mean of the 3 counts was used as the Ki67-labeling index for subsequent analysis. If 10% or more of tumor cells stained positive for Ki67 that was considered strongly positive staining. For CD117 staining, only increased (3+) staining was considered diagnostic for GISTs. A team of researchers including a pathologist (J.W.R.) and a surgeon (S.L.B.) confirmed staining results microscopically.
As primary antibody, monoclonal mouse antihuman p21ras (Code No. M0637; Dako Cytomation Inc) was diluted 1:50 with Antibody Diluent (S0809; DakoCytomation Inc). Polyclonal Rabbit Anti-Human CD117c-kit (code No. A4502; DakoCytomation Inc) was diluted 1:25 as primary antibody. After 1 hour of incubation, the slides were washed and incubated with horseradish peroxidase–labeled polymer conjugated secondary antibody from horseradish peroxidase (code No. K1392; L. V. Dako EnVision System, Copenhagen) for 30 minutes.
We used p53 Protein Kit (code No. NCL-p53-Paraffin; NovoCastra Laboratories, Newcastle upon Tyne, United Kingdom) labeling for p53 protein and Ki67 Antigen Kit (Code No. NCL-Ki67-Paraffin; NovoCastra Laboratories) labeling for Ki67 antigen. After a 30-minute incubation, the slides were washed and incubated with biotinylated rabbit antimouse immunoglobulin secondary antibody at 1:500 for 30 minutes. After washing, slides were incubated with a mixture of avidin and biotinylated horseradish peroxidase for 30 minutes. Incubation time and the dilution factor needed for each reagent were carefully followed according to the kits’ protocols.
Groups were compared by χ2 or Fisher exact test wherever applicable.
There were 65 patients (17 women and 48 men) with spindle cell tumors of the gastrointestinal tract. The median age was 53 ± 15 years (age range, 28-77 years). Thirty-five patients presented with symptoms of either bleeding or pain (53%) and almost half were asymptomatic (46%) and their tumors were found incidentally on either computed tomographic scan or screening video gastroscopy or colonoscopy. Consistent with previous reports most of these tumors presented in the stomach. There were 44 patients (67%) with gastric spindle cell tumors, 8 patients (12%) with tumors in the small bowel, 9 patients (14%) with tumors in the colon, and 4 patients (0.6%) with tumors in the esophagus. Thirty-seven patients had small (<5-cm) tumors on pathologic review; 28 had tumors larger than 5 cm. Fifteen tumors had aggressive-appearing histologic features with more than 5 mitoses per high-power field (×400).
Prior to immunohistochemical staining there were 42 patients diagnosed as having leiomyoma; 7, leiomyosarcomas; 7, sarcomas unclassified; 6, leiomyoblastomas; and 3, GISTs. Twenty-two tumors (34%) stained positive for CD117 (Figure). Therefore, we reclassified 22 patients (34%) as having GISTs and 9 patients (14%) as having sarcomas; 7 patients (10.5%) remained with leiomyoblastoma, and 27 patients (41.5%) were classified as having leiomyomas. Twelve tumors were changed from leiomyomas to GISTs; 5 tumors were changed from sarcomas to GISTs.
A total of 7 tumors (10.5%) stained positive for p53. Five patients with GISTs and 2 patients with sarcoma were p53 positive. Strong Ki67-positive staining of 10% or more was present in 3 of 22 patients with GISTs and 3 of 8 patients with sarcoma. There was no statistical significance in ras p21 expression when compared with tumor size, mitotic figures, and Ki67 percentage (Table 1).
Significantly more patient tissue samples stained positive for ras p21 (Figure) with GISTs compared with either leiomyomas or sarcomas. The tissue samples of 17 (77%) of 22 patients with GISTs stained positive for ras p21, while 0 (0%) of 27 patients with leiomyomas and 1 (11%) of 9 patients with sarcomas (P = .004) stained positive for ras p21. There was no statistically significant difference in ras p21 expression in leiomyomas and sarcomas (Table 1 and Table 2).
At the time of this analysis we had a median follow-up of 38 months. Of the 22 patients with GISTs, 3 patients had recurrent disease and 1 of the 3 died of disease. Because our numbers are too small, formal survival analysis could not be performed, but all 3 patients who had recurrent tumors were strongly positive for Ki67 (Table 1). There were no disease-free patients with GISTs that were Ki67 positive.
Half the patients in this cohort were asymptomatic and their GISTs were found during studies for other reasons. Increasingly, we see GISTs present on screening endoscopy and especially for small tumors, diagnosis and treatment become a challenge. In the past, small submucosal tumors with a bland appearance would have been assumed to be leiomyomas, but as we are more aware of GISTs, a biopsy specimen of these tumors is obtained by at least cytology. It is even more important to be able to accurately separate GISTs from other nonepithelial tumors based on limited material. Recent reports have shown that cytology is about 80% accurate in diagnosing GISTs preoperatively.6 Therefore, adding additional immunohistochemistry staining such as ras p21 may increase this accuracy. In our study 77% of GISTs stained positive for ras p21 as opposed to no patients with leiomyomas. This study is the first to examine the presence of ras p21 in GIST tumors. Adding this test may increase the confidence of diagnosing GISTs, which will have a major effect on clinical management of asymptomatic tumors. Additional patients are needed for clinicopathologic correlation that predicts their malignant potential. We recommend surgical excision for all spindle cell tumors suspicious for GIST. Fortunately, most of these tumors present in the stomach and can be approached with minimally invasive techniques such as laparoscopy.
In our series, the biopsy specimens of the 3 patients who had a recurrence were Ki67 positive. These numbers are too small for statistical analysis, but it suggests that Ki67 staining may be an important prognosticator for patients with GISTs. Other investigators have examined the prognostic role of Ki67 in outcomes of patients with GISTs. For example, Rudolph et al13 examined 52 GISTs and found on multivariate analysis that Ki67 labeling and atypical mitosis were independent predictors of overall survival. Other investigators such as Wong et al7 who stained 108 GISTs found on multivariate analysis that the mitotic count was the best predictor of mortality, and Ki67 immunohistochemistry did not provide additional information. Most authors advocate considering a panel of histologic factors in making clinical management decisions in patients with GISTs. Larger series are needed to validate these interesting preliminary results.7- 10,13
Unique to the treatment of GISTs there are drugs that target the c-kit receptor. The first such drug approved for use is imatinib mesylate (Gleevec; Novartis, East Hanover, NJ). Phase 1 and 2 trials have shown imatinib to be well tolerated and have a relatively good response rate. Phase 3 trials are ongoing in patients with metastatic GISTs. In the future, trials may focus groups of patients at high risk for recurrence and may benefit from adjuvant treatment of agents that block the c-kit receptor. In trials of patients with unresectable metastatic disease, series showed a 50% to 70% partial response rate to imatinib therapy. Therefore, a significant proportion of patients did not respond. Furthermore, this drug was originally developed for chronic myeloid leukemia. In chronic myeloid leukemia resistance to imatinib therapy in patients with advanced blast crisis has been associated with BCR-ABL gene amplification or development of new mutations in the kinase domain. These events presumably diminish the binding ability of imatinib to the c-kit enzymatic site and result in treatment failure.14,15 In the future, since these tumors are resistant to traditional cytotoxic chemotherapy and irradiation, a cocktail of specific target-blocking molecules may be used to treat these tumors. We already are using a specific target molecule to the c-kit receptor, but we may need to add targets to other tyrosine kinase receptors such as ras to completely eradicate these tumors in the future. This study is the first, to our knowledge, to test ras expression in GISTs. In the future ras inhibitors may potentially be a therapeutic target to treat GISTs as well.
Correspondence: Sarah L. Blair, MD, Department of Surgery, University of California at San Diego, 9500 Gilman Dr, San Diego, CA 92093-0987 (firstname.lastname@example.org).
Accepted for Publication: January 21, 2005.
Previous Presentation: This paper was presented at the 112th Scientific Session of the Western Surgical Association; November 9, 2004; Las Vegas, Nev; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.
Anton Bilchik, MD, Santa Monica, Calif: Gastrointestinal stromal tumors represent only 2% of gastrointestinal tract malignancies. These relatively uncommon tumors include most gastrointestinal mesenchymal tumors that were previously described as leiomyomas, leiomyoblastomas, or leiomyosarcomas. Most GISTs occur in the stomach and approximately one third are asymptomatic. Identification of the interstitial cell of Cajal as the cell of origin has made GIST a prototypic neoplasm for development of novel diagnostic tests and specific targeted therapy. The interstitial cells of Cajal express kit (a tyrosine kinase receptor) protein, a proto-oncogene marker for histogenesis, and a substrate for targeted drug therapy. Activation of kit protein stimulates the proliferation of GIST tumor cells and may inhibit apoptotic cell death. Most leiomyosarcomas have been reclassified as GIST tumors.
In a 10-year review of patients treated at the University of California at San Diego, Dr Al-Refaie and his colleagues identified 65 patients with spindle cell tumors of the gastrointestinal tract. Of the 42 tumors that had been diagnosed strictly by morphology during the first 7 years of the study, only 3 were initially identified as GIST tumors. When paraffin sections from all 65 tumors were examined by immunohistochemical staining, 22 specimens stained positive for CD117; these tumors were rediagnosed as GIST. Of the 22 GIST tumors, 16 (77%) stained for the ras p21 protein. Interestingly, none of the leiomyomas expressed ras protein. The authors conclude that the ras oncogene might be useful for diagnosis and targeted therapy of GIST tumors.
This interesting study highlights a diagnostic controversy. Although CD117 expression has been reported in more than 90% of GISTs, some pathologists do not believe that CD117 positivity is required for diagnosis, and many consider clinical and morphological features to be more reliable. A National Institutes of Health Consensus Conference held in April 2001 concluded that a positive CD117 finding can confirm the diagnosis when morphologic and clinical features are consistent with GIST. In the study reported by Al-Rafaie et al, CD117 immunostaining was considered diagnostic for GIST only if staining intensity was 3+; hence, only 34% of patients were considered for this study. How many patients in this study had morphologic characteristics of GIST that may only stain 2+ or 1+? Of these, how many were positive for ras p21?
Only tumors thought to originate from the muscle were reexamined in the study. This may not include all GIST tumors. Some tumors that were originally thought to originate from nerve tissue have now been determined to be GIST. Can the authors explain why these were not included?
Ki67 labeling was used to evaluate tumor aggression. How was the percentage of Ki67-labeled cells determined? If only 6 tumors stained positive for Ki67, can statistical significance be reached for those that are also positive for ras p21? Did Ki67 expression correlate with other indicators of tumor aggressiveness, that is, the number of mitosis and size of tumor?
Finally, these investigators mention that they are already using imatinib therapy in patients with GISTs. Have they found any correlation between response to therapy and ras p21 expression?
Immunohistochemical and molecular assessment of GIST is a timely and important area of research. Further characterization of the pathways responsible for the pathogenesis of this unusual tumor will likely provide novel specific targets for effective therapies.
Dr Blair: Dr Bilchik did correctly point out that CD117 positivity is not an absolute indication for GISTs. There are rare GISTs that morphology have the appearance of a GIST that pathologists will commonly term “GIST-like tumors,” but these are rare. These tumors have lost their ability to express CD117 for unknown reasons. In designing this study in collaboration with our pathologist colleagues, we decided [on] the use [of] the criteria of 3+ CD117 positivity for several reasons. First, large multicenter clinical trials use CD117 positivity for eligibility for these trials such as a recently published EORTC [European Organization for Research and Treatment of Cancer] sarcoma group trial that examined the efficacy of 2 doses of imatinib in patients with metastatic GISTs. We wanted our data to be consistent with larger trials. Second, in this study, we are using surgical resected specimens, and we would not have the sampling errors that you might have if you use FNA [fine-needle aspiration] samples or small tissue, and we would not have the sampling error of small sample size. And third, we wanted to have a[s] homogeneous [a] population as possible. Since it is a rare phenomenon to have a CD117-negative GIST, we felt is was prudent to err on the side of the more conservative. Having said all that, when we reexamined our data, we found that only 3 tumors were CD117 with 1+ positivity, and none of these were ras positive. We felt that this low level of positivity was more likely because of older specimens and fixation artifact. Regarding the nerve sheath tumors, one of the weaknesses of the study is that it is a retrospective review and, in our search through our database, we did not identify any nerve sheath tumors from the gastrointestinal tract.
Regarding the Ki67 labeling, the way we defined this was we looked at the percentage of Ki67-labeled cells by counting 10 high-power fields 3 times and averaging them. If there were greater than 10% of the cells labeled for Ki67, we considered this positive. In the article, we did comment that 3 patients had local recurrence and a poor outcome; they were all Ki67 positive. However, this was underpowered to show a statistical difference. We also did not find a correlation between mitosis, tumor grade, or clinical pathological criteria, and I am sure it is because this is just a pilot study and we do not have enough power. What we plan to do in the future is to validate these results with larger numbers by collaborating with other high-volume institutions. But, in my review of the literature, this is the first study to examine ras staining in GISTs.
Regarding the last question, we are using imatinib therapy, as other institutions are, for metastatic GISTs. In the cohort of patients we reviewed, only 2 patients had been treated with imatinib, and we did not feel we had enough data to make any correlations.