Loss of PTEN is associated with elevated EGFR and HER2 expression and worse prognosis in salivary gland cancer

Background: Activity of the tumour-suppressor gene PTEN is reduced in different types of cancer and implicates non-responsiveness to targeted therapy. This study evaluates the gene and protein status of PTEN in salivary gland carcinomas. Methods: A total of 287 carcinomas of the major and minor salivary glands were investigated for phosphatase and tensin homologue located on chromosome 10 (PTEN) deletion and loss of PTEN expression using fluorescence in situ hybridisation (FISH) and immunohistochemistry (IHC), respectively. Results were correlated to clinicopathological parameters, long-term survival, epidermal growth factor receptor (EGFR) and human epidermal growth factor receptor 2 (HER2) (IHC and FISH) status of the tumours. Results: Hemizygous deletions of PTEN were found in 35 out of 232 (15.1%) carcinomas, while homozygous deletions were observed in 17 out of 232 (7.3%) tumours. Phosphatase and tensin homologue located on chromosome 10 deletion was common in certain histological subtypes and especially homozygous deletion was associated with high-grade malignancy, lymph node metastases and unfavourable long-term prognosis (P<0.001). Loss of PTEN expression was present in 59 out of 273 (21.6%) carcinomas and was significantly correlated to genomic PTEN deletion, high-grade malignancy (P<0.001), increased tumour size (P=0.036), lymph node metastases (P=0.007) and worse disease-specific survival (P=0.002). Genomic PTEN deletion, in particular homogenous deletion (P<0.001) predominantly occurred in tumours with increased gene copy number of EGFR (60.0%) and/or amplification of HER2 (63.6%). Loss of PTEN expression was frequently found in tumours overexpressing EGFR (28.6%) and/or HER2 (52.6%). Conclusion: PTEN function is reduced in different types of salivary gland cancer indicating unfavourable prognosis. Its association with EGFR and HER2 signalling might affect targeted therapy.

Salivary gland carcinomas are rare tumours with an annual incidence rate between 0.4 and 2.6 per 100 . As the current World Health Organisation (WHO) classifies 24 different malignant subtypes with varying clinical courses and prognoses, diagnosis and treatment of these tumours represent a special challenge for both pathologists and surgeons. Recent attempts of treatment for unresectable or metastatic disease comprise targeted therapies against the epidermal growth factor receptor (EGFR, ErbB1) and the human epidermal growth factor receptor 2 (HER2, ErbB2) (Haddad et al, 2003;Agulnik et al, 2007;Locati et al, 2008). To date, however, response rates are rather disappointing.
Phosphatase and tensin homologue located on chromosome 10 (PTEN) is a tumour-suppressor gene frequently lost on chromosome 10q23 in different types of cancer (Li et al, 1997).
In the nucleus, PTEN regulates genomic stability, cell cycle progression, differentiation and gene expression (Carracedo et al, 2011). Beside these tumour-suppressive activities, PTEN functions as a negative regulator of the PI3K pathway by dephosphorylating the 3 0 position of phosphoinositide 3,4,5-triphosphate (PIP3). Stimulated by EGFR or HER2, PI3K activates the Ser/Thr kinase AKT, thus promoting cell survival and tumour growth (Courtney et al, 2010). Hemi-or homozygous deletions of PTEN are frequent events in prostate cancer and highgrade gliomas with unfavourable impact on prognosis (Korshunov et al, 2004;Yoshimoto et al, 2007). Loss of PTEN seems associated with unresponsiveness to targeted therapy against EGFR in colorectal cancer and glioblastomas (Mellinghoff et al, 2005;Frattini et al, 2007;Negri et al, 2010), moreover it predicts resistance to trastuzumab in HER2-positive breast cancer (Nagata et al, 2004;Pandolfi, 2004). The purpose of this study was to evaluate the gene and protein status of PTEN in a representative cohort of salivary gland carcinomas and to correlate the results to clinicopathological parameters, to long-term survival, and to the gene and protein status of EGFR and HER2.

Patients and treatment modalities
The study comprised 287 patients with carcinomas of the major and minor salivary glands, diagnosed at the Departments of Pathology of the University Hospital Regensburg, the University Hospital Erlangen-Nuremberg and the Hospital Clinic Nuremberg between 1984 and 2008. Clinical and follow-up data were obtained from the clinical tumour registries of Regensburg and Erlangen-Nuremberg in accordance with the Research Ethic Guidelines of the medical faculties. At diagnosis, patients were staged according to the TNM system of the Union for International Cancer Control (UICC) (Sobin et al, 2009). Tumour surgery was performed at the Departments of Oral and Maxillofacial Surgery and Otorhinolaryngology of Regensburg University Hospital, Erlangen-Nuremberg University Hospital and Nuremberg City Hospital. A total of 287 patients underwent primary surgery; lymph node dissection was performed for 205 (71.9%) patients. Postoperative radio-or radiochemotherapy was applied in 170 (59.2%) cases with highgrade malignancy, positive resection margins, lymph node metastases or distant metastasis.

Follow-up studies
In total, 137 male and 150 female patients showed a mean age of 60.6 (11 -99) years at diagnosis. The mean follow-up of all patients was 4.75 (range 0.1 -21.2) years. Recurrence was observed in 74 (26.8%) patients. Disease-related deaths occurred in 73 (25.4%) cases. The 5-and 10-year disease-specific survival rates of all patients were 73.2% and 68.1%, respectively.
Squamoid variants of MECs were thoroughly sorted out (Schwarz et al, 2011). Grading was based on a three-tiered grading system (Therkildsen et al, 1998;Jouzdani et al, 2010). The ACCC, BCAC, EMC, CAC and PLGA were considered low grade (G1) with the exception of dedifferentiated tumours, which were classified high grade (G3). In contrast, SDC, SQCC, MMT, OCC, UC and LCC were classified high grade (G3). Mucoepidermoid carcinomas were graded according to the criteria proposed in the current WHO classification (Barnes et al, 2005). Adenoid cystic carcinomas were divided into predominantly tubulo-cribriform (G2) and predominantly solid (G3) tumours. Grading of ACNOS and MYEC was based on nuclear pleomorphism and mitotic activity similar to the Elston and Ellis grading of breast cancer (Elston and Ellis, 1991). The 27 cases of carcinoma ex pleomorphic adenoma were classified and graded according to the malignant component of the tumour.

Immunohistochemistry
A tissue microarray (TMA) with 2.0 mm diameter punch cores was constructed from formalin-fixed paraffin-embedded tissue blocks of all patients as previously described (Milanes-Yearsley et al, 2002). Haematoxylin -eosin-stained TMA sections were used for reference histology.
Homozygous deletion of PTEN was defined by the simultaneous lack of both PTEN locus signals and by the presence of centromere signals in 420% of nuclei. Hemizygous deletion of PTEN was defined as 430% of tumour nuclei containing either one PTEN locus signal and X2 centromere signals or 2 PTEN locus signals and X4 centromere signals (relative deletions) (Sircar et al, 2009). Representative examples are shown in Figures 1D -F (centromere signals in red).
For EGFR, samples were grouped as normal disomy, p2 centromere signals in X50% of cells; low polysomy/trisomy, X3 centromere signals in X40% of cells, excluding cases with high polysomy or gene amplification; high polysomy, X4 centromere signals in X40% of cells, excluding cases with gene amplification; and gene amplification, ratio of gene/chromosome X2 or clusters of probes (410 copies per tumour cell) in X40% of cells. Disomy and trisomy/low polysomy were grouped as FISH negative, while high polysomy and amplification ( Figure 1J) were classified as FISH positive or copy number gain (CNG) of EGFR in dichotomisation (Chung et al, 2006;Pectasides et al, 2011).
Human epidermal growth factor receptor 2 was evaluated referring to the guideline recommendations for HER2 testing in breast cancer (Wolff et al, 2007). Gene amplification was assessed by a FISH ratio (HER2 gene signals to chromosome 17 signals) of 42.2 ( Figure 1I).

Statistical analysis
Data were analysed with SPSS for Windows, version 15.0 (SPSS, Erkrath, Germany). Relationships between parameters were examined using Pearson's w 2 -test (Po0.05) and Fisher's exact probability test (Po0.05) for dichotomised variables. Univariate survival curves were calculated by the Kaplan -Meier method and distributions were compared using the log-rank test. Diseasespecific overall survival was calculated from the date of diagnosis until disease-caused death or end of follow-up. Cox proportional hazards model (enter method) was used in multivariate analyses.
Deletion of genomic PTEN was strongly associated with immunohistochemical loss of PTEN expression (Po0.001). In all, 50.0% (24 out of 48) of the tumours with PTEN deletion (48.5% of hemizygous and 53.3% of homozygous deletion) also indicated loss of PTEN protein expression in contrast to 14.1% (24 out of 170) PTEN expression loss in non-deleted cases. Hemi-and homozygous deletion of PTEN were associated with lower diseasespecific survival rates in univariate Kaplan -Meier analysis revealing even worse prognosis for homozygous deletion (Figure 2). Prognostic significance of genomic PTEN was independent of grading, EGFR and HER2 status. Deletion of PTEN (hemi-and homozygous) harboured an unfavourable outcome in both EGFR-negative (Po0.001) and -positive cases (P ¼ 0.039). Similarly, loss of PTEN indicated worse prognosis in HER2negative (P ¼ 0.003) and HER2-positive (amplification and overexpression, P ¼ 0.027) tumours. With regard to grading, PTEN deletion went along with lower survival rates in both low/intermediate grade (P ¼ 0.002) and high-grade tumours (P ¼ 0.120), although not reaching statistical significance in the latter category. Moreover, deletion of PTEN (hemi-and homozygous) evolved as a strong negative predictor from multivariate Cox regression analysis (Table 2).
Loss of PTEN expression was also correlated to an increased gene copy number of EGFR, because 45.5% (15 out of 33) of the tumours with a CNG of EGFR presented a negative PTEN expression in comparison with 16.4% (30 out of 83) PTEN negativity in tumours without a CNG of EGFR (P ¼ 0.001). Negative PTEN expression more frequently occurred in carcinomas overexpressing EGFR protein (14 out of 49, 28.6%) than in those without EGFR overexpression (35 out of 188, 18.6%), although not reaching statistical significance (P ¼ 0.164). In all, 10 out of 19 (52.6%) carcinomas with an overexpression of HER2 protein showed an immunohistochemical loss of PTEN in comparison with 17.5% (38 out of 217) PTEN negativity in HER2-negative tumours (Po0.001). Phosphatase and tensin homologue located on chromosome 10 was more often absent in HER2-amplified tumours (4 out of 11, 36.4%) than in those without HER2 amplification (43 out of 214, 20.1%, P ¼ 0.247). In univariate analysis, loss of PTEN expression was significantly associated with unfavourable long-term survival (P ¼ 0.002, Figure 2).

DISCUSSION
This study investigated the role of the tumour-suppressor PTEN in a considerable and representative number of salivary gland cancer. By use of interphase FISH analysis, it revealed for the first time a hemi-or homozygous deletion of the gene on chromosome 10q23.3 in 420% of these tumours, especially in SDCs, ACNOS and SQCCs. Homozygous deletion of PTEN (7.3% of all tumours) was almost exclusively (94.1%) found in high-grade malignancies pointing towards a more aggressive growth pattern in comparison with hemizygous gene deletion. This observation is in accordance with the hypothesis that hemizygous loss of PTEN with haploinsufficiency of the remaining allele leads to genomic instability and cancer development as shown by in vivo studies on knockout PTEN mice (Di Cristofano et al, 1998;Kwabi-Addo et al, 2001;Shen-Li et al, 2011), while complete inactivation accelerates tumour dedifferentiation, progression and metastases (Verhagen et al, 2006;Schmitz et al, 2007). In our study, both hemi-and homozygous genomic deletion of PTEN strongly correlated with immunohistochemical loss of the protein expression. Loss of PTEN function, especially deletion of genomic PTEN independently indicated an unfavourable clinical course of patients with lymph node metastases, rapid tumour progression and worse overall survival. In line with the above mentioned hypothesis, homozygous deletion displayed an even worse prognosis compared with hemizygous deletion in the investigated salivary gland carcinomas. Downregulation of PTEN with negative impact on prognosis has already been described for prostate cancer, colorectal and breast cancer and high-grade gliomas (Korshunov et al, 2004;Fujita et al, 2006;Frattini et al, 2007;Yoshimoto et al, 2007;Laurent-Puig et al, 2009).
Different types of salivary gland carcinomas show overexpression and an increased gene copy number of EGFR and HER2 with negative impact on prognosis (Press et al, 1994;Ettl et al, 2008;Lujan et al, 2010;Williams et al, 2010). In this study, 16.3% of the carcinomas presented high polysomy or amplification of the EGFR gene and 19.7% showed overexpression of the EGFR protein.
Overexpression and amplification of HER2 occurred in 10.5% and 7.1% of cases, respectively, especially in SDCs. Both EGFR and HER2 were associated with negative clinicopathological parameters and unfavourable survival. Increased gene copy number of EGFR and overexpression/amplification of HER2 pose the precondition for anti-EGFR and anti-Her2 therapy in breast, colorectal and lung cancer (Wolff et al, 2007;Hirsch et al, 2008;Laurent-Puig et al, 2009). Moreover, in salivary gland cancer first phase-II trials on targeted therapies have been conducted and revealed that the use of anti-EGFR agents, such as cetuximab and gefitinib, failed to produce obvious response, although the majority of patients showed disease stabilisation (Glisson et al, 2005;Locati et al, 2009). These studies mainly comprised adenoid cystic carcinomas (68 -77%) and the EGFR status was characterised by IHC. In the investigation of Locati et al (2009), tumours were also retrospectively analysed by FISH, however, no amplification or high polysomy of EGFR was found in these tumours. This might have been a reason for missing response, as EGFR protein expression observed by IHC has been described as an unreliable predictor of responsiveness to EGFR inhibitors (Ciardiello and Tortora, 2008;Vidal et al, 2009). The use of trastuzumab (HER2inhibitor) and lapatinib (combined TKI against EGFR and HER2) did also show no obvious reactivity in ACCs (Haddad et al, 2003;Agulnik et al, 2007;Vidal et al, 2009). However, partial response to trastuzumab was reported for a mucoepidermoid carcinoma with HER2 3 þ immunostaining (Haddad et al, 2003) and prolonged tumour progression to lapatinib was documented in three non-ACC tumours with HER2 amplification and 3 þ staining for EGFR and HER2 (Vidal et al, 2009). Therefore, increased gene copy number of EGFR and amplification or overexpression of HER2 should be a precondition for an anti-EGFR and anti-HER2 therapy in salivary gland cancer, too.
In this study, we demonstrate that deletion of genomic PTEN, in particular homogenous deletion, predominantly occurs in salivary gland carcinomas with an increased copy number of EGFR or amplification of HER2. Moreover, loss of PTEN protein expression is frequently found in tumours with EGFR and HER2 overexpression. This coincidence of PTEN loss and EGFR/HER2 gene/expression gain most likely represents an accumulation of independent genetic disorders in poorly differentiated carcinomas during tumourigenesis rather than a functional relationship between EGFR, HER2 and PTEN. Owing to its function as a tumour suppressor, however, there is some evidence that PTEN has significant impact on the efficiency of targeted therapies by antagonising the activation of the PI3K -AKT pathway that is predominantly activated by EGF-/HER2-receptor tyrosine kinases in different tumour types. In successfully treated cancers, PI3K signalling is attenuated or even turned off by EGFR-or HER2targeting (using antibodies or receptor kinase inhibitors). If there is a loss of PTEN expression/function, however, PI3K signalling remains active in spite of anti-EGFR or anti-HER2 therapy (Courtney et al, 2010;Nahta and O'Regan, 2010). In breast cancer, (active), PTEN has been reported to be an essential parameter for responsiveness to trastuzumab treatment, while loss of PTEN rather predicts resistance to trastuzumab in HER2-positive patients (Nagata et al, 2004;Pandolfi, 2004;Fujita et al, 2006). On the other hand, the additional use of a PI3K inhibitor is reported to overcome the PTEN loss-induced trastuzumab resistance (Courtney et al, 2010). In colorectal cancer, high polysomy or amplification of EGFR was associated with response to cetuximab therapy, whereas loss of PTEN expression, (evaluated by IHC) rather is indicative for therapy resistance (Frattini et al, 2007;Laurent-Puig et al, 2009;Loupakis et al, 2009). The same observation was made in glioblastomas, where expression of PTEN was associated with clinical response to the EGFR kinase inhibitors gefitinib and erlotinib (Mellinghoff et al, 2005). For these tumours, evidence derived from in vitro analysis suggests that resistance can be overcome by coupling anti-EGFR-agents with mTOR inhibitors (Wang et al, 2006).
In anti-EGFR/Her2-treated salivary gland cancer patients, a retrospective analysis could elucidate the suggested correlation of the PTEN status with therapy response rates. As a consequence, PTEN analysis might find the way into routine diagnostics and might facilitate therapy decisions related to anti-EGFR or anti-HER2 targeting. Additional targeting of the PI3K pathway potentially enhances therapy efficiency in salivary gland cancer.
Despite the significant findings in this study it has to be kept in mind that the investigated salivary gland carcinomas we investigated comprise a variety of subtypes with inter-and intratumoural differences and characteristics. This fact aggravates further preclinical and clinical investigations of these rare tumours.
In conclusion, this is the first report on a reduced PTEN function in different types of salivary gland cancer indicating worse prognosis. We could demonstrate an association between PTEN loss and EGFR and HER2 signalling, which might influence response to targeted therapies.