Characterizing Metastatic HER2-Positive Gastric Cancer at the CDH1 Haplotype

The CDH1 gene, coding for the E-cadherin protein, is linked to gastric cancer (GC) susceptibility and tumor invasion. The human epidermal growth factor receptor 2 (HER2) is amplified and overexpressed in a portion of GC. HER2 is an established therapeutic target in metastatic GC (mGC). Trastuzumab, in combination with various chemotherapeutic agents, is a standard treatment for these tumors leading to outcome improvement. Unfortunately, the survival benefit is limited to a fraction of patients. The aim of this study was to improve knowledge of the HER2 and the E-cadherin alterations in the context of GC to characterize subtypes of patients that could better benefit from targeted therapy. An association between the P7-CDH1 haplotype, including two polymorphisms (rs16260A-rs1801552T) and a subset of HER2-positive mGC with better prognosis was observed. Results indicated the potential evaluation of CDH1 haplotypes in mGC to stratify patients that will benefit from trastuzumab-based treatments. Moreover, data may have implications to understanding the HER2 and the E-cadherin interactions in vivo and in response to treatments.


Introduction
Gastric cancer (GC) is a serious health problem worldwide. This year 28,000 new cases, with approximately 10,960 related deaths, are expected in the United States [1]. Even though surgery is the primary treatment option for early stage GC, diagnosis is often late in Western countries. This is probably due to the lack of proper screening programs and a lack of symptoms for a long time. About 35% of patients present with "de novo" metastatic GC (mGC) and approximately 70% that underwent surgery for the primary tumor will have disease recurrence or develop distant metastases, with a median survival of about one year, despite palliative chemotherapy [2][3][4].
In selected small tumors (i.e., stage Tis or T1) endoscopic resection may be performed, mainly in experienced centers. However, complete tumor resection with adequate margins and lymph node dissection remains the only potentially curative therapy for patients with non-metastatic GC. Furthermore, perioperative chemotherapy can improve survival outcomes for patients with operable disease [5].
In metastatic or recurrent disease, chemotherapy is the standard treatment, although it is not curative. However, despite the introduction in clinical practice of new drugs and chemotherapeutic schedules, only little progress has been made in recent years. The most important advance came from the international (24 countries) phase 3 randomized "ToGA" study (NCT01041404) [6]. In this trial, the addition of the trastuzumab, a monoclonal antibody, to the chemotherapy with cisplatin and capecitabine was compared to the same chemotherapy combination alone, in a population of HER2-positive metastatic gastric or gastro-esophageal junction cancers. The HER2 is a transmembrane protein with tyrosine kinase activity implicated by its interaction with epidermal growth factor (EGF) family in cell growth and differentiation ( Figure 1). In the ToGA study the median overall survival (OS) was found to be higher in patients who received trastuzumab plus chemotherapy compared with those who received the only chemotherapy (13.8 vs. 11.1 months). Immunohistochemical (IHC) scoring evaluates both the HER2 membranous staining (absent, weak or detected in only one part of the membrane, moderate/weak complete or basolateral membranous staining and strong) and the percentage of the tumor cells staining (<10% or ≥ 10% of cells). A greater survival benefit was detected in patients whose tumors were IHC-positive (score 3+) or IHC-equivocal (score 2+), but in situ hybridization-positive (16 vs. 11.8 months; hazard ratio (HR) = 0.65). In addition, it was seen that the addition of the trastuzumab to chemotherapy did not compromise patient quality of life. Unfortunately, in a substantial proportion of mGC patient who progress after initial response to chemotherapy the death occurs in a few months [4]. Furthermore, the prognostic significance of the HER2 expression in GC remains to be elucidated.
Breakthroughs in the GC biology are currently changing the landscape of GC. More specifically, the understanding of molecular mechanisms underlying the different pathological features has led to new GC classifications [7,8]. Originally GC was categorized according to anatomical presentation [9,10], and to histological classes (WHO classification and Lauren classification) [11,12]. More recently, the characterization of GC includes the HER2 status and the HER-positive disease is reported in about 18% (range 4.4% to 53.4%) of patients [13].
In 2014, the Cancer Genome Atlas (TGCA) project [14] subdivided GC according to different molecular biology tests in four subgroups: EBV-positive (about 8% of all GC), Microsatellite Instable (MSI, about 22%), Chromosomal Instable (CIN, about 50%), and Genomically Stable (GS, about 20%) cancers. Additionally, in 2015 the Asian Cancer Research Group (ACRG) [15] proposed an alternative molecular classification due to the different biological characteristics of Asian patients. This classification divided GC in four subgroups represented by: Microsatellite Stable/epithelial-to-mesenchymal transition (MSS/EMT), Microsatellite Stable TP53-positive (MSS/TP53+, somehow overlapping with EBV type of TCGA classification), Microsatellite Stable TP53-negative (MSS/TP53−, similar to CIN by TCGA), and Microsatellite Instable (MSI). The E-cadherin cytoplasmic tail presents two regions: the catenin-binding domain and the juxtamembrane domain. β-catenin binds to the E-cadherin domain and this complex via α-catenin connects and regulates E-cad interaction with the actin cytoskeleton. p120-catenin binds the CDH1 juxtamembrane domain and stabilizes E-cad expression at the cell surface. (A) Activation of the HER2 by inducing the phosphorylation of β-catenin directs the dissociation of β-catenin from the E-cad complex, thus leading to a decrease of E-cad-mediated cell adhesion, facilitate epithelialmesenchymal transition (EMT), and the translocation of β-catenin to the nucleus where it acts as a transcriptional regulator of genes involved in cell growth and the EMT process; (B) HER2 activation increases metalloproteinase (MP) activity, which leads to an increased production of soluble Ecadherin (sE-cad) through the cleavage of E-cad. Metalloproteinase also cleaves HER2 into a cytoplasmic tail domain, p95HER2, and a shaded soluble HER2 fragment. The p95HER2 fragment maintains the phosphokinase activity, thus favoring the dissociation of the β-catenin/E-cad complex leading to GC progression and metastasis. The production of the sE-cad causes a reduction in cell adhesion and, by its diffusion into the microenvironment, acts as a paracrine/autocrine signaling molecule that regulates numerous signaling pathways implicated in tumor progression, including a key role in the HER2 interaction/activation and phosphorylation of β-catenin. The E-cadherin cytoplasmic tail presents two regions: the catenin-binding domain and the juxtamembrane domain. β-catenin binds to the E-cadherin domain and this complex via α-catenin connects and regulates E-cad interaction with the actin cytoskeleton. p120-catenin binds the CDH1 juxtamembrane domain and stabilizes E-cad expression at the cell surface. (A) Activation of the HER2 by inducing the phosphorylation of β-catenin directs the dissociation of β-catenin from the E-cad complex, thus leading to a decrease of E-cad-mediated cell adhesion, facilitate epithelial-mesenchymal transition (EMT), and the translocation of β-catenin to the nucleus where it acts as a transcriptional regulator of genes involved in cell growth and the EMT process; (B) HER2 activation increases metalloproteinase (MP) activity, which leads to an increased production of soluble E-cadherin (sE-cad) through the cleavage of E-cad. Metalloproteinase also cleaves HER2 into a cytoplasmic tail domain, p95HER2, and a shaded soluble HER2 fragment. The p95HER2 fragment maintains the phosphokinase activity, thus favoring the dissociation of the β-catenin/E-cad complex leading to GC progression and metastasis. The production of the sE-cad causes a reduction in cell adhesion and, by its diffusion into the microenvironment, acts as a paracrine/autocrine signaling molecule that regulates numerous signaling pathways implicated in tumor progression, including a key role in the HER2 interaction/activation and phosphorylation of β-catenin.
In the present study, our attention is focused on tumors with CDH1 mutation, which could be included in the GS subtype of TCGA classification, mostly represented by GC of diffused histotype widely distributed to all the anatomical sites of the stomach and tending to a metastatic process linked to EMT [14]. CDH1 encodes the E-cadherin (E-cad), a transmembrane glycoprotein especially abundant in epithelial tissues that mediate calcium-dependent adhesion between epithelial cells. Several CDH1 mutations with a reduced activity/expression of the E-cad [16], as well as the HER2 overexpression [17], have been associated with shorter GC patient survival. More recent evidence points to β-catenin as a common link between the HER2 overexpression and the E-cad repression in influencing EMT, the metastatic process, and outcome [18].
A deep understanding of molecular characterization of patients with mGC focusing on both the HER2-positive and the CDH1 polymorphisms could provide the scientific background to develop modern clinical trial protocols in order to maximize the benefit of novel biological agents in a proper patient population [7]. The present study was designed to characterize CDH1 in mGC subtypes according to the HER2-expression and to evaluate the association between the CDH1 and the prognosis.

Patient Characteristics
Fifty-nine consecutive patients with mGC were enrolled in this study; patients meeting the criteria for hereditary diffuse GC have been excluded. In total, 44 patients were males and 15 were females, and the mean age at diagnosis was 60 years (range, 40-76 years). Twelve cases (20.3%; (N = 10 males, median age 56.5 years) were classified at diagnosis as HER2-positive by an IHC score of 2+/neu amplification or by an IHC score of 3. All mGC patients received the same chemotherapeutic regimen with the addition of trastuzumab in mGC HER2-positive tumors (mGC-HER2).
At a median follow-up time of two years, a trend for a better OS was showed in the mGC-HER2 positive group of patients although due to the limited number of cases the difference did not reach a statistical significance (468 days, standard deviation (SD) 389 vs. 584 days, SD 336; p = 0.20) ( Figure 2). These data are in accord with previous studies reported in the literature, which used targeted treatment [6,19]. A deep understanding of molecular characterization of patients with mGC focusing on both the HER2-positive and the CDH1 polymorphisms could provide the scientific background to develop modern clinical trial protocols in order to maximize the benefit of novel biological agents in a proper patient population [7]. The present study was designed to characterize CDH1 in mGC subtypes according to the HER2-expression and to evaluate the association between the CDH1 and the prognosis.

Patient Characteristics
Fifty-nine consecutive patients with mGC were enrolled in this study; patients meeting the criteria for hereditary diffuse GC have been excluded. In total, 44 patients were males and 15 were females, and the mean age at diagnosis was 60 years (range, 40-76 years). Twelve cases (20.3%; (N = 10 males, median age 56.5 years) were classified at diagnosis as HER2-positive by an IHC score of 2+/neu amplification or by an IHC score of 3. All mGC patients received the same chemotherapeutic regimen with the addition of trastuzumab in mGC HER2-positive tumors (mGC-HER2).
At a median follow-up time of two years, a trend for a better OS was showed in the mGC-HER2 positive group of patients although due to the limited number of cases the difference did not reach a statistical significance (468 days, standard deviation (SD) 389 vs. 584 days, SD 336; p = 0.20) ( Figure  2). These data are in accord with previous studies reported in the literature, which used targeted treatment [6,19].

Association between CDH1 P7-Haplotype and mGC-HER2
We investigated the association between mGC stratified by HER2 expression and the CDH1 haplotype resulting from the six polymorphic sites dispersed over the entire CDH1 gene. A total of 11 different CDH1 haplotypes were identified and their frequencies reported in Table 3. Haplotype P7 was exclusively present in six mGC-HER2 patients, one of them showing a P7 haplotype in homozygous cases (Table 4). Linkage disequilibrium (LD) analysis in mGC patients showed a tight association between rs16260 and rs1801552 (D = 1.0000, R 2 = 0.1731, χ 2 = 16.2759, p = 0.0001), which was not present in mGC-HER2 (D = 0.3555, R 2 = 0.0721, χ 2 = 1.7311, p = 0.1883). The contemporary presence of the rs16260-A allele and rs1801552-T-allele, both included in the P7 haplotype, was strictly associated with mGC-HER2 disease (Table 3).

Association between the CDH1 P7 Haplotype and the Survival of mGC-HER2 Patients
The relationship between the P7 haplotype and OS was statistically analyzed by using Cox regression curves. The P5, P6, and P8-P11 haplotypes with less than two cases in mGC-HER2 (Table 3) were combined in a unique group, and designed them as haplotype "matched". Figure 3A shows the OS of all GC cases (independent of the HER2-expression) stratified on haplotype-based approaches. The haplotype P7 was associated with better outcome (median OS: 1037 days) compared to the other haplotypes (median OS: 312 to 448 days). Notably, a significantly worse prognosis was observed with the "matched" haplotype (median OS: 312 days; HR 2.79, 95% CI 1.032-7.548) and the haplotype P1 (median OS: 419 days; HR 2.54, 95% CI 1.049-6.169). The haplotype P7 was present only in the mGC-HER2 group where it distinguished patients with better survival compared to those with the "matched" haplotype and haplotype P1 (HR 4.33, 95% CI 1.033-7.548; HR 2.58, 95% CI 0.328-20.275, respectively). The "matched" haplotype and haplotype P1 were also associated with poorer prognosis in mGC group when compared to haplotype P3 (HR 1.412, 95% CI 0.717-2.779). Figure 3B indicates the OS distribution according to the restricted GC haplotype (i.e., rs16260 and rs1801552), specifically associated with the mGC-HER2 patients as reported above. Data confirmed the better outcome was associated with the restricted AT haplotype (median OS: 1037 days, 95% CI 371-1037) which, in our series, is supported by the observation of a statistically significant difference in the OS compared to the CC haplotype (median OS: 420 days, 95% CI 312-500; HR 2.374, 95% CI 1.077-5.229).  Abbreviations: HER2, the human epidermal growth factor receptor 2; OR, odds ratio; is the relative measure of the number of an allele or a genotype in the mGC-HER2 group relative to the comparison of the number of allele/genotype in the mGC, by considering as the reference the most frequent allele in the mGC-HER2. If the OR is >1 the allele or genotype having the greatest frequency in the mGC-HER2 is higher than that found in the mGC group. 95% CI (confidence interval) is the probability that the confidence interval contains the true odds ratio. Statistically significant p values are reported in bold type. Abbreviations: HER2, human epidermal growth factor receptor 2; mGC, metastatic gastric cancer; OR, odds ratio; is the relative measure of the number of the mGC haplotype relative to the comparison of the number of the same haplotype in the mGC-HER2, by considering as the reference the most frequent haplotype in the mGC. If the OR is >1 the haplotype having the most frequency in the mGC is higher than that found in the mGC-HER2 group. 95% CI (confidence interval) is the probability that the confidence interval contains the true odds ratio. Statistically significant p values are highlighted and reported in bold type.

Discussion
Targeted HER2 therapy for mGC works differently due to the heterogeneity of the tumor. Positivity for the HER2 status (by IHC or by fluorescence in situ hybridization) is a prerequisite for the HER2 targeted therapy, but it is not sufficient to predict the treatment response. In the present study, we found an association between specific CDH1 polymorphisms with a subset of HER2-positive mGC that, in turn, are associated with distinct prognosis behavior ( Figure 3A). The association between the haplotype P7 with the better OS is not due to the presence of additional mutations to the P7-related CDH1 mutations since, among the HER2 patients carrying haplotype P7,

Discussion
Targeted HER2 therapy for mGC works differently due to the heterogeneity of the tumor. Positivity for the HER2 status (by IHC or by fluorescence in situ hybridization) is a prerequisite for the HER2 targeted therapy, but it is not sufficient to predict the treatment response. In the present study, we found an association between specific CDH1 polymorphisms with a subset of HER2-positive mGC that, in turn, are associated with distinct prognosis behavior ( Figure 3A). The association between the haplotype P7 with the better OS is not due to the presence of additional mutations to the P7-related CDH1 mutations since, among the HER2 patients carrying haplotype P7, only one had an additional non-polymorphic CDH1 mutation (i.e., pt P296 with a missense variant); of note, this patient experienced a poor prognosis (median OS: 164 days). Notably, analysis of LD in mGC showed a specific association with two polymorphisms (i.e., CDH1 rs16260 and rs1801552) that were not found in the HER2-positive mGC. In fact, these polymorphisms showed minor allele frequency (MAF) variants in mGC-HER2: rs16260-A and rs1801552-T alleles. By using these two polymorphisms (i.e., rs16260/rs1801552) Cox regression analysis was simplified from 11 to four haplotypes and confirmed the association of haplotype P7, including the rs16260-A and the rs1801552-T alleles, with the mGC-HER2 subtype with the best OS ( Figure 3B). Functionally, the rs16260-A polymorphism located in the CDH1 promoter region had been associated with an alteration in the CDH1 transcriptional efficiency. In vitro testing using luciferase reporter gene revealed that the rs16260-A allele decreased CDH1 transcriptional activity by 68% compared to the C-allele. However, the effect of rs16260-A variant on CDH1 expression in vivo is still unknown [21] and a significant decrease of CDH1 production in the peripheral blood cells of mutation carrier patients compared to that produced in the control-cohort was not found [22]. Additionally, the potential contribution of rs16260-A allele to GC risk remains controversial [23,24], while the potential prognostic value of this variant in breast cancer and in metastatic colon cancer did not reach statistical significance [25,26].
With regard to the second polymorphism, the rs1801552-T, a protective association between this variant in homozygous and non-syndromic cleft lip has been reported [27]. Recent studies indicated that patients with cleft lip had a higher incidence of tumors than the general population [28,29] and, moreover, family members with pathogenetic CDH1 mutation showed a higher incidence of cleft lip/palate than the general population (6-7% versus about 0.1%) [28,30,31]. In addition, a previous study by using rs1801552 heterozygous individuals with GC compared to controls demonstrated a CDH1 allelic expression imbalance in hereditary GC family members with an increase of the ratio of the CDH1 RNA rs1801552 T-allele/C-allele that was not found in cancer-free individuals [32]. Results had suggested, for some unknown reason, the reduction of CDH1 C-allele-specific expression in patients at risk for GC.
In our series we found relationship between the CDH1 variants and patient survival in the setting of mGC, particularly in the HER2-positive disease. Previous studies supported a functional interaction between the HER2 and the E-cad. Briefly, the β-catenin binds the C-terminal cytoplasmic domain of the E-cad and this complex via the α-catenin connects and regulates the E-cad interaction with the actin cytoskeleton, while association of the p120-catenin with the juxtamembrane domain of the E-cad stabilizes the E-cad expression at the cell surface ( Figure 1). Activation of the HER2 by inducing the phosphorylation of the β-catenin directs the dissociation of β-catenin from the E-cad complex, thus leading to a decrease of the E-cad-mediated cell adhesion facilitating tumor cells invasion and migration. In addition, the dissociation of the β-catenin from the E-cad complex causes the translocation of the β-catenin to the nucleus which drives the transcription of various target genes associated with cell survival, proliferation and metastasis. E-cad can also be solubilized (sE-cad) by membrane E-cad cleavage and be released into the extracellular environment [33] (Figure 1). The production of sE-cad not only undermines adherence junctions, but, by its diffusion into the micro environment, it regulates numerous signals implicated in tumor progression, including a key role in HER2 interaction/activation and phosphorylation of the β-catenin [18]. Furthermore, HER2 activation was known to increase metalloproteinase activity and, thus, it further leads to high increased production of the sE-cad by the cleavage of the E-cad. Through the specific cleavage of its cytoplasmic tail domain into the p95HER2 fragment, it maintains the phosphokinase activity of the HER2 favoring the dissociation of the β-catenin/E-cad complex that, overall, both promote GC progression and metastasis ( Figure 1).
Overall, these findings support a possible functional role of the E-cad in response to the anti-HER2 treatment in mGC-HER2 subtype. However, the role of the CDH1 mutations in this context requires elucidation through further studies. Interestingly, in human transfected breast epithelial cell lines over-expressing the HER2 receptor resulted in inhibition of E-cad expression [34]. Lapatinib resistance in HER2-positive breast cancer cells was also associated with a EMT and the EMT-related down-regulation of E-cad [35]. Moreover, treatment with the HER2-specific tyrosine kinase inhibitors (AG285 and HER2 siRNA) was found to produce a down-regulation of the E-cad in ovarian cancer cells [36]. Overall, these studies demonstrate the important connection between HER2 and E-cad in human cancers.
The determination of the HER2 status in patients with advanced GC is crucial in order to select patients who may benefit from the new anti-HER2 agents; the results of this study, if confirmed in a prospective larger series, could improve the understanding of molecular interactions between HER2 and E-cad and define their role as predictive factors for targeted therapy.

Study Population
For the present study, between 2011 and 2017 we enrolled 59 consecutive patients with confirmatory of metastatic GC. Forty-four were males and 15 were females, and the mean age at diagnosis was 60 years (range, 40-76 years). Patients were grouped according to confirmed HER2-status at diagnosis. Patients agreed to participate to the study and provided informed consent (CRO-2011-2012 Code EUDRACT: 2011-001720-37).

Genotyping Analysis
Genomic DNA was extracted from each subject's peripheral blood lymphocytes using an EZ1 DNA blood kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions.
Screening for mutations of CDH1 exons and neighboring intronic sequences was performed using the polymerase chain reaction (PCR) with previously-described primers and the reaction conditions [37]. In short, 15 individual PCR reactions were performed on each sample for a full mutational screen of all 15 exons and splice junctions of the CDH1 gene. Amplified PCR products were sequenced on an Applied Biosystems 3130 automated sequencer (Applied Biosystems, Foster City, CA, USA) using the Big Dye v3.1 Terminator Cycle Sequencing Kit (Life Technologies, Monza, Italy) and sequence data were aligned and analyzed using CodonCode Aligner software.
The promoter, 5 UTR and the exon 1 polymorphisms were genotyped by PCR method. A 581-bp fragment containing this region was amplified with the following primers: the forward CDH1FPRO 5 -TCCCAGGTCTTAGTGAGCCA-3 and the reverse CDH1Exon1REV 5 -TGACGACGGGAGAGGAAG-3 . The amplification was performed in a programmable thermal cycler as follows: touchdown (hold (94 • C 4 min) eight cycles (94 • C 45 s; 58 • C 50 s; 72 • C 1 min; reducing the annealing temperature by 1 • C each cycle)), three cycles (94 • C 45 s; 53 • C 50 s; 72 • C 1 min), and a final cycle of 72 • C for 10 min. After amplification the PCR product was sequenced using the primer CDH1FPRO.
We screened each of the above samples for the c.-472delA CDH1 polymorphism using a new PCR method in place of current genotyping analysis by the PCR-RFLP method. DNA fragments containing the promoter region of interest were amplified with two distinct PCR reaction using the following primers: 1 • reaction (forward, CDH1G347proFor 5 -CAGCTTGGGTGAAAGAGTGAGC-3 ; reverse ECad347Rev 5 -GGCCACAGCCAATCAGCA-3 ); 2 • reaction (forward, CDH1GA347proFor 5 -CAGCTTGGGTGAAAGAGTGAGA-3 ; reverse ECad347Rev 5 -GGCCACAGCCAATCAGCA-3 ). The PCR conditions were set as follows: initial denaturation at 94 • C for 4 min; 10 cycles at 94 • C for 30 s, 65 • C for 1 min, 20 cycles at 94 • C for 30 s, 62 • C for 30 s and 72 • C for 1 min; and a final extension at 72 • C for 8 min, in a Veriti Thermal Cycler (Life Technologies). PCR products were analyzed on a 2.5% agarose gel stained with ethidium bromide and photographed under UV light. G/G homozygous cases were represented by DNA bands in the 1 • reaction, GA/GA homozygous cases were represented by DNA bands in the 2 • reaction. G/GA heterozygous cases display a combination of both bands. To validate this new PCR method the promoter region of same samples was amplified using the following primers: forward, 5 -GCCCCGACTTGTCTCTCTAC-3 ; reverse, 5 -GGCCACAGCCAATCAGCA-3 and PCR products were sequenced.

Immunohistochemistry
A formalin-fixed, paraffin-embedded tumor block was cut into 4-µm-thick sections for H and E and immunostaining. Immunohistochemistry was performed by using the rabbit monoclonal antibodies against HER 2 (clone 4B5, Ventana Medical System, Tucson, AZ, USA).

Statistical Analysis
The frequencies of allele and genotype were compared between mGC-HER2 and mGC patients by means of chi-squared test (VassarStats, http://faculty.vassar.edu/lowry/VassarStats.html).
The haplotype frequencies were analyzed with the SNPator [38] and Arlequin software [39]. Survival analysis was performed at the time of the first treatment by using Cox regression analysis.

Conclusions
In conclusion, the evaluation of a restricted CDH1 haplotype in mGC could help to select the patients that gain greater benefit from the anti-HER2 treatments. Furthermore, since a high level of the sE-cadherin may modulate sensitivity to RTK inhibitors, the evaluation of its serum concentration in different phases of treatment could have a role in monitoring the therapeutic response overtime.