Prognostic impact of gastrointestinal bleeding and expression of PTEN and Ki-67 on primary gastrointestinal stromal tumors

Gastrointestinal stromal tumors (GISTs) constitute the most common mesenchymal neoplasms of the gastrointestinal (GI) tract, and the incidence has increased significantly over the past three decades [1, 2]. Gain-of-function mutations in c-kit and, less commonly, the platelet derived growth factor receptor α (PDGFRA) oncogene are believed to be the driving force of GISTs [3, 4]. All GISTs have malignant potential, varying from small lesions to aggressive sarcomas. Disease relapse is not uncommon, even where tumors are R0 resected. The application of imatinib mesylate (IM), a small-molecule tyrosine kinase inhibitor, has dramatically promoted the disease-free survival of GISTs. However, side effects and resistance to IM pose new challenges in the management of GISTs. Thus, an accurate risk classification scheme has become increasingly crucial for selecting patients who are most likely to benefit from adjuvant IM therapy. National Institutes of Health (NIH) consensus criteria [5], Armed Forces Institute of Pathology (AFIP) criteria [6], and modified NIH consensus criteria [7] are used frequently to estimate the risk of recurrence after surgery in GISTs. However, even the latest risk stratification system may still have room for improvement.

Most GISTs initially manifest as GI bleeding. We have noted in clinical practice that patients presenting with GI bleeding have appeared to fare worse than their counterparts. Few, if any, studies have addressed the prognostic implication of this important manifestation and its correlation with risk classification.

The PTEN (phosphatase and tensin homolog deleted from chromosome 10) gene is considered ‘the most highly mutated tumor-suppressor gene in the post-p53 era’ [8]. By dephosphorylating phosphoinositol 3,4,5-triphosphate (PIP3), PTEN negatively controls the activation of the PI3-kinase/Akt pathway, which has been found to be a crucial survival cell signaling transduction in GISTs [9], and thus functions as a proapoptosis factor. Also, PTEN plays a role in the control of the cell cycle and cell migration [10, 11]. Loss of PTEN protein expression or function has been reported in many human cancers, including ovarian, endometrial, and prostate carcinoma, and breast cancer [12].

Ki-67, a nuclear protein universally expressed among proliferating cells and absent in quiescent cells [13], is one of the most frequent biomarkers investigated on GISTs. Although previous studies have shown that Ki-67 labeling index (LI) could be used in predicting the malignant potential of GISTs [14], conflicting opinions have challenged whether its expression level provides better prognostic utility than mitotic rate [15].

By using tissue microarray (TMA)-based immunohistochemistry (IHC), we aimed to analyze the impact of aforementioned factors, including clinicopathological parameters, PTEN expression, and Ki-67 LI, on the recurrence risk of R0 resected primary GISTs to identify potential new indicators that might better predict their clinical behavior and prognosis.

In this retrospective study based on follow-up data, we have assessed the impact of a set of clinicopathological parameters and the role of PTEN and Ki-67 on the prognostic of primary GIST patients treated with surgery alone. Our main findings were as follows: (1) GI bleeding, a common manifestation for primary GISTs, was an independent predictor of worse prognosis for RFS; (2) the IHC underexpression of PTEN was associated with worse RFS at univariate analysis; (3) a Ki-67 LI >1% was associated with worse RFS at univariate analysis; (4) GI bleeding could be considered an additional prognostic indicator in the GIST risk classification scheme.

Gastrointestinal bleeding is one of the most common clinical symptoms of GIST [16]. In a long-term follow-up of 1,765 cases of gastric GISTs [17], 54.5% patients presented with GI bleeding, ranging from chronic insidious hemorrhage with or without anemia to acute melena or hematemesis. Although we could not reproduce this high incidence, our data still illustrate an incidence of GI bleeding as high as 28.6%. Ulceration and mucosal invasion are the most common causes. Ulceration of the mucosa is commonly manifest in GISTs (39.6%), and is a statistically significant predictor for progressive disease (P < 0.0003) [16]. Although mucosal invasion is rare, it has been shown in a surprisingly high proportion of progressive GISTs (22 of 24 cases) [18]. Research regarding the prognostic implication of GI bleeding on RFS in GISTs is scant; in addition to one study [19] that found a statistical significant prediction value on univariate analysis but not on multivariate analysis. Our data demonstrated that GI bleeding was an independent predictor of poor prognosis for RFS. As shown in Figure 2E, the curves of the ‘very low’ and ‘low’ groups merged totally, which might suggest that these two groups in our proposal risk classification system can be incorporated into the same group. In that case, there would still be four groups, while the ‘very high’ group highlights very well that GI bleeding is a negative prognostic indicator. Moreover, Kaplan-Meier curves illustrated that patients whose risk category upgraded due to GI bleeding in our re-modified risk stratification scheme are more likely to suffer from disease relapse, suggesting that GI bleeding, a clinical feature worth recording, might be added to the risk classification system in further studies.

Original researches performed in murine models, suggesting that PTEN might be involved in the aberrant activation of cell signaling transduction in GISTs tumorigenesis; however, the results are inconsistent. By using a Kit^K641E knock-in mouse model of GIST, Deneubourg et al.[20] indicated that the upregulation of PTEN might act as a negative feedback mechanism to limit PI3-kinase activation downstream of Kit in a context of oncogenic mutation. On the other hand, in a panel of GIST xenograft mouse models, Floris and colleagues [21] illustrated that homozygous PTEN loss was identified in the GIST-PSW (GIST biopsy tumor cell with KIT exon 11 mutation) and GIST882Ly (GIST cell line harboring homozygous KIT exon 13 mutation) xenografts, while heterozygous PTEN loss was observed in a GIST48 (GIST cell line carrying both KIT exon 11 and 17 mutation) xenograft. Importantly, they also confirmed the lack of PTEN protein expression in the GIST-PSW and GIST882Ly xenografts. PTEN protein expression has previously been investigated in clinical GISTs tissue samples by IHC, and a clear relationship between low or absent PTEN expression and poor prognosis was revealed, either alone or in combination with other biomarkers [22, 23]. Our study, adopting a PTEN IHC H-score method, provides further support for the role of absent or low expression (H-score, 0 to 4) of PTEN (in 52.4% GISTs) as a predictor of recurrence in patients with primary GIST, although the result was not statistically significant on multivariate analysis.

Additionally, diffuse cytoplasmic or both cytoplasmic and nucleic PTEN immunostaining was observed, while pure nuclei staining was not observed in this study. However, exclusively nuclear PTEN immunoreactivity in GIST was previously reported in two studies [22, 23]. A possible reason for this discrepancy could be the different anti-PTEN antibodies used in IHC analysis. Our findings are consistent with Deneubourg and his colleagues, who revealed PTEN subcellular expression in the cytoplasm and partly in the nucleus by immunofluorescence in human GIST882 cells [20]. PTEN has been shown to shuttle between the cytoplasm and nucleus. Its different subcellular localization might imply different molecular functions. It is worth noting that negative nuclear PTEN immunostaining, unlike cytoplasmic PTEN expression, is an independent prognostic factor for survival in esophageal squamous cell carcinoma [24]. Accordingly, further studies are warranted, to define the prognostic implication of different subcellular PTEN expression in GISTs.

Although Ki-67 shows the most potential as a prognostic indicator for GISTs in the literature, the exact threshold value of Ki-67 LI for malignant predicting remains debatable. In documented research, different cut-off points have been suggested: 0.82% [25], 2.7% [26], 3% [27], 5% [15, 23], 6% [28], or 10% [29]; however few of these cut-off points were established based on ROC analysis. In our series, a 1% LI was confirmed as the best threshold value for relapse in ROC analysis, with 71.7% sensitivity and 64.5% specificity. Whether Ki-67 LI could replace mitotic counting in prognostication remains controversial [30]. Some scholars hold that the mitotic index reflects the M stage of mitosis only; but, because Ki-67 LI can be used to recognize most of the proliferating cells in stages G1, S, and G2, it is considered to be more appropriate as an objective index of the malignancy of GISTs [31]. Molenaar et al.[32] also demonstrated higher observer reliability of Ki-67 than mitotic counting in assessment of mitotic activity. On the contrary, Wong et al.[15] advocated Ki67 LI to be inferior to mitotic index as a prognostic marker. Our study shows that a Ki-67 LI level higher than 1% is associated with a higher relapse potential of primary GISTs, on univariate survival analysis but not on multivariate analysis, and that mitotic index remains a stable prognostic indicator, on both univariate and multivariate analysis. We propose that if the mitotic index is based on counting, fewer than 50 high power fields (that is in a small biopsy sample), Ki-67 immunohistochemical analysis could be an alternative for mitotic index.

The NIH consensus criteria [5], AFIP criteria [6], and modified NIH consensus criteria [7] are frequently used to estimate the risk of recurrence after surgery in GIST, and their prognostic accuracy is generally similar [33]. NIH consensus criteria stratify risk based on tumor size and mitotic counting, whereas AFIP criteria also include tumor site. Modified NIH criteria encompass four factors: tumor size, mitotic index, site, and tumor rupture, which have been well established as important independent prognostic factors for GIST recurrence [34, 35] (Table 1). Our data regarding these parameters reach a general agreement with prior reports, showing that larger tumor size and high mitotic index are steadily indicative of poor RFS for primary GISTs, in addition to tumor site and rupture, which were not identified as independent prognostic factors in multivariate analysis. In a pooled analysis of ten population-based series including 2,459 GIST patients treated with surgery alone [33], the estimated 5-year and 15-year RFS rates were 70.5% and 59.9%, respectively. In line with this study, the overall outcome after surgery in our cohort was presented as 1, 3, and 5-year RFS rates (98.8%, 79.8%, and 70.2%, respectively).

There are certain limitations in our study. First, this is a retrospective study with a limited sample size of study population and a large proportion of censored cases. Secondly, the study was conducted in one single tertiary hospital. Thus, a larger scale, multicenter prospective study with longer-term follow-up investigation is warranted. Finally, the inability of TMA approach to detect all information of tumor samples might cause bias [36]. Despite these caveats, it appears that our findings can contribute to the knowledge of prognostic biomarkers in patients with primary GISTs who are treated with surgery alone.

The current NIH consensus criteria remain a stable and practical predicting tool in primary GISTs risk classification. We found that GI bleeding was an independent prognostic indicator for poor RFS in patients with primary GISTs treated with surgery alone. We proposed that the current NIH consensus criteria could be improved by including this important clinical manifestation. Whether absent or low expression of PTEN, and Ki-67 LI higher than 1% could add independently to the prognostication of GISTs should be further investigated in a larger series.