Ki-67 expression score correlates to survival rate in gastrointestinal stromal tumors ( GIST )

PURPOSE: To evaluate the immunohistochemical expression of p16, Ki-67, p53 and Bcl-2 proteins in gastrointestinal stromal tumors (GIST); to assess the possible association between these variables and clinical and histopathological factors of cancer; and to check for prognostic value of these variables (survival and recurrence). METHODS: A sample of 55 patients treated surgically for GIST in three hospitals was studied. The surgically excised tumors were confirmed as GIST by KIT, vimentin, desmin S100 protein, CD117, 1A4 and CD34 assessment in paraffin blocks. RESULTS: Only 9 (16%) cases of GIST were positive for p53, p16 was positive among 43.6%; 80% of GISTs showed staining for Bcl-2. The proliferative index (expressed as the proportion of positive cells) assessed by immunohistochemical expression of Ki-67 was high in 49% of cases. Elevated Ki-67 scores were associated to high histological grade (p=0.0026) and mitosis index, MI (p=0.0001). High Ki-67 index was associated to death. Expression of p53, p16 and Bcl-2 did not correlate to morphological or clinical variables. CONCLUSIONS: Ki-67 immunohistochemical evaluation should be included in preoperative evaluation of GIST biopsies or surgical specimens as a prognostic tool for clinical staging; and all other proteins studied (Bcl-2, p53 and p16) did not play a role in GIST metabolic or carcinogenic process, remaining without prognostic value.


Introduction
Gastrointestinal stromal tumors (GIST) are considered a rare entity, encompassing for 2.2% of gastric neoplasias, 13.9% in small intestine and only 0.1% of colon tumors.However, in the United States 3000 to 5000 new GIST cases are diagnosed yearly 1 .Before 1986 this entity was encompassed among gastrointestinal mesenchimal tumors.With the identification of KIT protein expression in these lesions, a novel malignant tumor referred as GIST was identified 2,3 .
In the 2001 consensus, immunopositivity was defined as a key element for the diagnosis of GIST, and the classification of malignancy is based on tumor size and mitotic count [3][4][5] .However, lack of uniform diagnostic criteria leaded to controversial epidemiological and survival data.Some tumors previously considered GIST could be diagnosed as other particular GI sarcomas (as leiomiosarcomas or schwannomas) 3,6 .Differential diagnosis among GIST and other muscular or neural tumors require therefore immunohistochemical assessment with standard profile including S100 protein, CD34, 1a4, desmin, and CD117 (c-KIT) antibody, determinant of GIST diagnosis 3 .
Only 5% of GISTs are sensitive to classic chemotherapy 7 and radiotherapy is ineffective in this particular entity.Surgical resection was the main therapeutic approach to GIST until mesylate imatinib (Glivec), a molecular target drug that blocks KIT signalization (STI-571) 8 , was available for adjuvant therapy in advanced or irressecable GIST cases 7,9 .However, it is still difficult to determine which cases will metastasis or present an aggressive behavior based uniquely in morphological variables.Several potential prognostic variables were assessed in GIST, such as DNA ploidy, protein expression (S100, CD34, VEGF, S-VEGF) and genetic alterations (deletion of exon 11) without reaching statistical significance and, more important, with lack of technical reproducibility.Recently, we evaluated the role of p53 and Ki-67 immunohistochemically with preliminary favorable results 10 .
The objectives of this study are: 1) To evaluate the immunohistochemical expression of p16, Ki-67, p53 and Bcl-2 proteins in 55 cases of GIST; 2) to assess the possible association between these variables and clinical and histopathological factors of breast cancer; and 3) to check for prognostic value of these variables, relating them to survival and recurrence.

Immunohistochemical assessment
Representative tumor paraffin-embedded tissue blocks were cut into 5 micra slides.Immunohistochemistry analysis was performed on each tumor slide block for p53, p16, Bcl-2 e Ki-67, with standard streptavidin ABC methodology.Two minutes of pressure cooking at pH 6.0 was used for antigen retrieval with p53 clone (1:x) (BioGenex Inc., San Ramon, CA, USA).Immunohistochemical essay with the tumor markers was performed using the antibodies and conditions shown in Table 1.For p53, p16, and Ki-67, nuclear expression was assessed by identification of nuclear brown staining of neoplastic cells for further analysis at the nuclear markers [11][12][13][14] .A semiquantitative method for nuclear immunohistochemical grading was used 15 .
Bcl-2 showed a cytoplasmatic diffuse staining.A lesion was designated as positive if at least 10% of true neoplastic tumor cells expressed the antibody studied [15][16][17] .In some tumors virtually every cell expresses the marker and others may show a focal or variable percentage of positive cells.

Statistical analysis
Full clinical follow-up data, including presence of metastatic disease, were available for all GIST cases.Associations between p53, p16, Bcl-2 e Ki-67, immunoreactivity and clinic pathological parameters -including tumor size (as defined by disease-free survival (DFS), survival to metastases at specific sites, and time from first recurrence to death.DFS was defined as time to any type of recurrence, distant metastasis, or death from any cause.Survival curves were calculated using the Kaplan-Meier method.Tests of differences in survival between groups were performed using the log-rank test and were expressed as hazard ratios (HRs), which were estimated using Cox regression.
Cox regression analysis was also used to evaluate any independent effect of prognostic factors on DFS, OS, and survival from recurrence.Factors found to be significant (at a p value of less than 0.05) in the survival analysis, together with for p53, p16, Bcl-2 e Ki-67 as the factors of principal interest, were included in the regression analyses.All tests were two-tailed, and 95% confidence intervals were presented where appropriate.All analyses were carried out using software SAS 9.1 (Statistical Analysis System, Cary, NC, USA).

Clinical and morphological data
Among the 55 studied patients, 13 died from the disease (23.6%) and 42 were alive (76.4%) at the last follow up assessment.Among the survivors, 34 had no sign of disease (80.9%) and 8 were currently presenting recurrence (14.5%).All clinical epidemiological and morphological data are disposed in Table 2.The stomach was the most frequent affected organ (22 cases, or 40%).However, atypical locations as appendix, mesenterium, rectum and pelvis were represented with one case each (disposed as "others" in Table 2).Median tumor width was 12.78 cm (0.7 a 32 cm) with predominance of fusiform differentiation (32 or 58.1%).
More than 20% of the cases were composed of highgrade tumors ( ).These data were not significantly correlated to any morphological or clinical variables.
Only nine (16%) cases of GIST were positive for p53, and the proportion of neoplastic positive cells were largely variable among the positive cases.
p16, however, was positive among 25 reactive cases (43.6%); with positivity in the majority of the transformed cells.
According to the proliferative index, tumor cells were negative for Ki-67 in 27 cases (49%) and positive in the remaining 28.
Among the biological markers, Ki-67score reached significant correlation with other important prognostic variables.
Higher Ki-67 scores were associated to high histological grade (p=0.0026 with Fisher test, p=0.007 with Kruskal-Wallis test) and MI (p=0.0001).Tumors with high Ki-67 indexes presented necrosis (p=0.0257) and p53 positive results (p=0.0018) more frequently.Moreover, a high Ki-67 index was associated to death: 13 of 24 Ki-67-positive GIST patients died during the study and association to evolution was confirmed with Fisher and Kruskal-Wallis tests (p=0.049 and p=0.01, respectively).Figure 1 shows the survival curves of positive and negative Ki-67 cases.Expression of p53, p16 and Bcl-2 did not show any association and did not correlate to any morphological or clinical variables (Table 3).

Discussion
The current prognostic classification in GIST is based in tumor width and mitotic index as proposed by Fletcher et al. 3 However, the classical mitotic index evaluation requires 50 high power camps and usually small samples are insufficient to complete the analysis.Moreover, there are irregularities.For instance, fixation may shrink tumor size and mitoses may "go over" during fixation 18 .
Recently, diagnostic approach to GIST started to include ultrasound-guided fine needle aspiration or core biopsy.
Preliminary results from our group 10 and also from Ando et al. 19 suggested that Ki-67 expression is a powerful prognostic tool for easy assess in smaller samples.The new Food and Drug Administration (FDA)-approved indication of imatinib mesylate in the adjuvant therapy turns biopsy specimens assessment necessary.Currently, there is no consensus in literature about the most suitable methodology for proliferation assessment and Ki-67 expression evaluation in GIST.Besides the variation of reading and assessment methodologies, limit values for positive and negative results remain controversial [20][21][22][23][24] .
We used a previously described method for Ki-67 assessment (Allred method for estrogen receptor).By using a high score for intensity 13 , since it is assumed that it would be equivalent in all samples and not variable, adapted this method to proliferation assessment.Distribution remained with punctuation from 0 to 5, leading to an index from 4 to 8. The final result is expressed in crescent categories, very reproducible.One advantage is that it constitutes an easily understandable method among distinct observers, as in ER expression, and practical for clinical application.Another important feature of this approach is that it can be performed in biopsy specimens, therefore allowing the pathologist to have an idea of proliferation status before surgery.
The strong correlation to other important prognostic variables, such as histological grade (p=0.0026),mitotic index (p=0.001)and presence of necrosis (p=0.0257)endorsed the potential clinical role of this Ki-67 index.Finally, the association of Ki-67 index to evolution (p=0.0161) was evidenced, demonstrating that higher Ki-67 index is associated to poor survival rate.
The evaluation of cell cycle-related proteins in neoplasms by immunohistochemistry has been included in prognostic staging in central nervous system 25,26 and other tumors 21,[27][28][29][30] .We verified if these same proteins could also play a prognostic role in GIST 20,31 .
Bcl-2 protein is highly expressed in many neoplasms and also in GIST.Our results confirmed that GISTs are Bcl-2 expressing tumors.Classically, chemotherapy agents hit cells in an apoptosis pathway.In carcinomas and lymphomas, a higher contingent of apoptotic cells may represent a better response to chemotherapy 30,32 .Unlike many chemotherapeutic agents, imatinib mesylate, used in GIST is not based on a proliferating contingent of cells, but rather a blocking tyrosine kinase receptors and the further induction of cellular proliferation 33 .Therefore, Bcl-2 would not be useful to predict response to imatinib.Indeed, most of the studies could not associate the highly expressed Bcl-2 to other prognostic factors or to drug response in GIST [22][23][24]34 . We onfirmed that Bcl-2 assessment could not improve staging or prognosis assessment in GIST.The reason why Bcl-2 is so up regulated in GIST is not clear yet.One could speculate that uncontrolled proliferation in ultimately stable mesenchimal cells could trigger cell mechanisms to balance increasing population or cell transformation in GIST affects Bcl-2 gene regulation with protein stabilization and accumulation.Molecular assessment would be necessary to clarify this issue.
Gene mutations in TP53 tumor-suppressor gene are one of the most common genetic alterations in human solid tumors.Loss of p53 protein function may increase the risk for carcinogenesis 35,36 .There are dozens of publications on p53 expression in GIST with a large methodological variation among them.However, the series of patients are usually smaller than ours and the results are very heterogeneous, ranging from 0 to 100% of positive neoplastic cells.In five larger studies, comprehending more than 60 patients, p53 staining was present in less than 50% of neoplastic cells and prognostic value remained controversial 35,36 .
We found nine p53 positive GIST cases in 55 analyzed (16%), and p53 staining correlated to Ki-67 index, but not to any other clinical or morphological variable.The smaller contingent of p53 positive cells could mean that p53 alterations are not adamant to GIST early steps in carcinogenesis and may occur later in this process 37,38 .Also p53 deletion is not frequent and missense mutation seems confined to less than 50% of neoplastic cells 39 .
The assessment of p16 protein in GIST is still incipient and three previous studies [40][41][42] with positivity ranging from 34% and 62% assessed p16 by PCR, RNA and tissue microarray (TMA).We found 25 immunohistochemically positive cases (45.4%).Metilation of p16 gene was previously demonstrated by House et al. 42 with may partially explain the diversion of results.
Ki-67 protein is one of several factors involved in cell proliferation that may be assessed easily by immunohistochemistry.
Ki-67 index reflects the proportion of cycling cells in a given population.In GIST, Ki-67 expression has been previously associated to morphological variables and also survival, but results are hardly comparable due to methodological different approach 19,20,23,31 .

Conclusions
The Ki-67 immunohistochemical evaluation should be included in preoperative evaluation of GIST biopsies or surgical

A
total of 55 samples of surgically excised GIST, diagnosed between February of 1992 and September of 2006, were retrospectively identified from the archives of the Department of Surgical Pathology of three university Hospitals of Sao Paulo, Brazil, and were included in this study.Ethics Committees of all the three institutions approved this study.

FIGURE 1 -
FIGURE 1 -Survival curve of patients with GIST tumors and positive or negative for Ki-67 (Mib).

TABLE 1 -
Immunohistochemical markers used to assess cell proliferation and apoptosis in gastrointestinal stroma tumor.

TABLE 2 -
Clinical and demographic data on gastrointestinal stromal tumor (GIST) patients.

TABLE 3 -
Statistical significance (p value) of the correlations between the variables.