Concomitant Tumor Expression of EGFR and TATI/SPINK1 Associates with Better Prognosis in Colorectal Cancer

Background Epidermal growth factor receptor (EGFR) activation plays a role in colorectal cancer (CRC) carcinogenesis, and anti-EGFR drugs are used in treatment of advanced CRC. One of the EGFR ligands is tumor-associated trypsinogen inhibitor TATI, also called serine protease inhibitor Kazal type1 (SPINK 1), which we recently showed to be an independent prognostic marker in CRC. Methods We studied the prognostic value of immunohistochemical expression of EGFR and concomitant expression of EGFR and TATI/SPINK1 in a series of 619 colorectal cancer patients. Results Of the samples, 92% were positive for EGFR. EGFR+/TATI+ was seen in 62.8%, EGFR+/TATI− in 29.5%, EGFR−/TATI+ in 4.9%, and EGFR−/TATI− in 2.7% of patients. EGFR expression correlated with WHO grade (p = 0.040). In univariate analysis, EGFR expression correlated with favourable survival (p = 0.006). EGFR+/TATI+ patients showed better survival than did those with other combinations (p<0.001). In multivariate analysis, EGFR+/TATI+ was an independent prognostic factor of favourable prognosis (p<0.001). Conclusion Concomitant positivity of EGFR and TATI/SPINK1 predicts favourable prognosis in CRC.

Prognosis of patients with local CRC is good, 5-year survival being 80-90%, for node-positive tumors it is 60-70%, while for tumors with distant metastases it is less than 10%) [4][5][6]. Patients with stage III-IV (Dukes' C and D) disease usually receive adjuvant chemotherapy. In stage II (Dukes' B) disease, chemotherapy is not routinely used although some of these patients obviously would benefit from adjuvant therapy. To identify patients at high risk requires additional prognostic factors like biomarkers.
The epidermal growth factor receptor (EGFR), a target for treatment of advanced colorectal cancer, belongs to a transmembrane glycoprotein of the ErbB tyrosine kinase receptor family. Ligand-receptor interaction and dimerization of the receptor leads to tyrosine autophosphorylation, activating an intracellular signal pathway that promotes cell division and migration, inhibition of apoptosis, and angiogenesis [7]. The monoclonal antibodies cetuximab and panitumumab bind to EGFR and disable the activation of tyrosine-kinase and downstream signalling pathways.
Mutations in the molecular pathways activated via EGFR can contribute to carcinogenesis. In CRC, the most frequent mutations concern the KRAS gene occurs in about 40% of CRC cases [8], [9]. The mutations deactivate guanosine triphosphatase (GTPase) activity, leading to accumulation of activated KRAS. These KRAS mutations lead to lack of response to anti-EGFR therapy [10], [11].
We have recently shown that tissue expression of TATI/ SPINK1 is an indicator of favourable prognosis in colorectal cancer patients [15]. In the present study we evaluated the relationship between EGFR and TATI expression and its possible prognostic value in colorectal cancer.  [16]. The study has been approved by local ethics committee and complies with the Declaration of Helsinki (Dnro HUS 226/E6/06) and the National Supervisory Authority for Welfare and Health. Clinicopathological characteristics of the patients are described in Table 1.

Tissue samples and preparation of TMA blocks
Formalin-fixed and paraffin-embedded surgical tissue samples were collected from the archives of Department of Pathology, University of Helsinki. Histopathologically representative regions of tumour specimens were defined and marked on H&E slides. Three cores from each tumour block were sampled with 1.0 mm punchers by use of a semiautomatic tissue microarrayer (Tissue Arrayer 1, Beecher Instruments Inc., Silver Spring, MD, USA). Three parallel serial blocks were constructed, all including one sample from each patient. From each block, 4 mm thick sections were cut for immunohistochemistry.

Immunohistochemistry
The Lab Vision Autostainer TM 480 (LabVision, Fremont, CA, USA) was used for immunostaining. Tissue sections were deparaffinized in xylene and rehydrated through graded alcohol series. To retrieve antigens, samples were heated in the pretreatment module of the autostainer in pre-heated TRIS-EDTA pH 9.0 buffer for 20 min at 98uC. The samples were incubated for 5 min in DAKO REAL Peroxidase-Blocking Solution (DAKO, Glostrup, Denmark) for inactivation of endogenous peroxidases. The sections were incubated for 60 min with primary monoclonal NCL-EGFR antibody (clone 113) against the extracellular domain, which stains both membrane and cytoplasma (Novo Castra, Newcastle, UK, dilution 1:10). The sections were reacted for 30 min with HRP link and for 30 min with HRP enzyme (anti-mouse-anti-rabbit labelled polymer) (DAKO, Carpinteria, CA, USA). Between each of the steps the sections were rinsed with Tween 20-PBS (phosphate-buffered sodium chloride buffer, pH 9.0). Then the samples were incubated for 10 min in the DAB + EnVision System (DAKO, Glostrup, Denmark), and stained with hematoxylin for 1 minute. Finally the samples were rinsed in running tap water. Immunostaining of TATI was performed as described (Koskensalo 2011). Scoring EGFR cytoplasmic immunopositivity in tumour cells was scored by two independent investigators (S.K. and J.H.) without knowledge of clinical outcome. Cytoplasmic EGFR immunopositivity was evaluated by percentage of positively stained cells. Positivity in over 50% of cells was scored as 3, 10-50% as 2, and less than 10% as 1 ( Figure 1). Absence of positivity was scored as 0. Tissue spots without tumour cells were excluded. The highest score was used for each patient. For statistical analysis, the patients were divided into two groups: EGFR2 (score 0) and EGFR+ (scores 1-3).

Statistical analysis
We analyzed separately EGFR staining alone and together with TATI/SPINK1 staining (EGFR+/2, TATI+/2 or EGFR2 TATI+/2). The association between immunoexpression and clinicopathological variables was assessed by the x 2 test or Fisher's exact test in case of low expected frequencies. Survival was analyzed by the Kaplan-Meier method. Statistically significant covariates in univariate analysis were analyzed in multivariate survival analysis by the Cox proportional hazards model. A p,0.05 was considered significant. Statistical analyses were performed with IBM SPSS Statistics 20.0 software.
EGFR immunoreactivity associated with tumour grade; positivity was detected more often in highly and moderately differentiated than in undifferentiated tumours (p = 0.040). EGFR expression showed no association with Dukes' stage, but positivity
EGFR+/TATI+ correlated with histology occurring more often in adenocarcinomas than in other histological tumour types (p = 0.005), and varied by WHO grade, being most often present in highly and moderately differentiated tumours (p,0.001) ( Table 2).

Discussion
We recently showed that TATI is an independent prognostic factor in colorectal cancer, high tissue expression being associated with favourable prognosis. Here we show that concomitant immunoexpression of EGFR and TATI is an independent prognostic factor for favorable survival in colorectal cancer, and it is a stronger prognostic factor than EGFR or TATI/SPINK1 alone.
In concordance with previous findings, we observed EGFR overexpression in colorectal cancer samples, [17], [18]. EGFR immunoexpression correlated significantly with tumour grade,  being more often positive in high and moderately differentiated tumours, as reported previously [19]. Conversely, EGFR expression has also been shown to correlate with poor differentiation [18]. As in other studies, we found correlation with histology [20], [21] but no correlation between EGFR immunoexpression and Dukes' stage. This is in concordance with the study by Giralt [20], whereas Spano et al. reported stronger EGFR overexpression in T3 than in T4 colorectal tumours [17]. Deng et al. found, in 94 colorectal cancer patients, an association between high expression and high tumour stage [22].
Here we show that EGFR immunoexpression is an independent marker for favourable prognosis in colorectal cancer patients. Our results differ from those others' ; EGFR expression did not correlate with survival in a study of 249 [19] and 132 CRC patients [17] or in 87 Dukes' C CRC patients, [6]. However, in some studies specifically of colon cancer, EGFR expression correlated with poor prognosis. In a study of 149 patients, EGFR expression was an independent marker of adverse prognosis, but EGFR expression was observed in only 35.6% of the samples, which is less than usually reported [23]. Resnick et al. also showed an association between strong EGFR expression and poor prognosis in colon cancer [24]. In a subanalysis including only colon cancer, our study showed an association between EGFR expression and improved survival, but the difference was not significant (data not shown).
In a subgroup analysis of rectal cancer, we found significantly better survival rates (p = 0.001) in patients with EGFR positivity (data not shown). In 87 rectal cancer patients who had received preoperative radiation therapy, Giralt et al reported EGFR expression in pre-treatment biopsies-but not in surgical samplesto associate with adverse prognosis [20]. In a study by Fernebro et al. on 269 rectal cancer patients, no association was observed with metastasis-free survival [25]. In a study of 40 rectal cancer patients, Yasuda et al. found decreasing EGFR levels after chemoradiotherapy, but EGFR expression was not a prognostic marker [26]. Radiotherapy can both reduce and increase EGFR expression [27], [20]. We did not analyze EGFR expression in pretreatment biopsies, but no difference in EGFR expression emerged between rectal cancer patients receiving or not receiving preoperative radiotherapy (data not shown).
EGFR-targeted treatments have been used for metastasized colorectal cancers, but the value of EGFR immunoexpression to predict the efficiency of adjuvant treatment is controversial. Interestingly, the EGFR antagonist cetuximab has proven effective even against EGFR-negative tumours [28]. One explanation for this may be that anti-EGFR-treatment is targeted against metastasized tumors, and correlation between EGFR immunoexpression in the primary tumor and metastatic lesion is unclear and varies between studies [19], [29], [22]. In addition, EGFR-positive tumors do not always respond to cetuximab treatment [30]. Furthermore, tumors with KRAS mutations have been found to respond poorly to anti-EGFR therapy [10], [11], and the analysis of KRAS mutational status is recommended for all metastatic CRCs [31]. KRAS mutations in codon 12 correlates with adverse prognosis [31,32]. BRAF is the second step of the EGFRmediated pathway. Mutated BRAF correlates with adverse prognosis [33], but patients with BRAF mutation may benefit from EGFR-targeted treatments [31]. A novel method that appears to be superior to analysis of KRAS mutations to select patients for anti-EGFR therapy is combined analysis of KRAS mutations and EGFR gene copy number [34].
The PIK3CA (the phosphatidylinositol-4,5-bisphosphonate 3-kinase, catalytic subunit alpha polypeptide gene) pathway is another downstream pathway of EGFR signalling. Mutations of the PIK3CA gene can promote malignant transformation [35]. Use of aspirin blocks the PIK3CA pathway, and correlates with better prognosis among patients with a mutated gene [36]. Unfortunately data on aspirin use was not available in our patient records.
Ozaki et al. showed that EGFR and TATI/SPINK1 are coexpressed in pancreatic cancer and that SPINK1 stimulates the proliferation of pancreatic cancer cells through EGFR activation EGFR+TATI/SPINK1+ Improves Prognosis in Colorectal Cancer PLOS ONE | www.plosone.org [13]. Here we show that in roughly two out of three colorectal carcinoma samples EGFR and TATI/SPINK1 are co-expressed. We have shown earlier that high TATI/SPINK1 tissue expression correlates with better prognosis in CRC; here we show that concomitant expression of EGFR and TATI/SPINK1 is an even better marker for improved survival. It is known that there are positive and negative feedback loops within EGFR-mediated pathways depending on cell type, and one reason for better prognosis in colorectal cancer may be that binding of TATI leads to inhibition of cascades leading to malignant transformation. Also, TATI binding to EGFR may inhibit the binding of stronger activating ligands.
It is therefore plausible that TATI/SPINK1 and the interaction between TATI/SPINK1 and EGFR play different roles in CRC than in pancreatic cancer.Interestingly, expression of TATI/ SPINK1 in cancer tissue is also associated with good prognosis in gastric cancer [37] and loss of TATI/SPINK1 expression correlates with aggressive disease in bladder cancer [38]. This indicates that TATI/SPINK1 exerts different functions in different pathologies. TATI/SPINK1 is not only an activator of EGFR but also an effective trypsin inhibitor and trypsin is expressed by several tumors [38]. Thus it is not surprising if it also exerts different functions in different tumours. So far, EGFR immunohistochemistry is not a tool for prediction of which patients that are likely to benefit from EGFR-targeted adjuvant treatment. It would be interesting to study whether adding TATI/ SPINK1 tissue analysis plus KRAS mutation analysis and EGFR gene copy number assessment could further improve the promising results of Å lgars et al. [34].
In conclusion, to the best of our knowledge, we show for the first time that concomitant immunoexpression of EGFR and TATI/ SPINK1 is an independent prognostic marker for favourable prognosis in colorectal cancer patients. The combination is a stronger prognostic tool than either TATI/SPINK1 or EGFR alone. Further studies are necessary to better understand the association between TATI/SPINK1 and EGFR in colorectal cancer and for evaluating the potential use of this marker combination to predict treatment response.