Poorly Differentiated Medullary Phenotype Predicts Poor Survival in Early Lymph Node-Negative Gastro-Esophageal Adenocarcinomas

Background 5-year survival rate in patients with early adenocarcinoma of the gastro-esophageal junction or stomach (AGE/S) in Caucasian patients is reported to be 60–80%. We aimed to identify prognostic markers for patients with UICC-I without lymph-node involvement (N0). Methods Clinical data and tissue specimen from patients with AGE/S stage UICC-I-N0, treated by surgery only, were collected retrospectively. Tumor size, lymphatic vessel or vein invasion, grading, classification systems (WHO, Lauren, Ming), expression of BAX, BCL-2, CDX2, Cyclin E, E-cadherin, Ki-67, TP53, TP21, SHH, Survivin, HIF1A, TROP2 and mismatch repair deficiency were analyzed using tissue microarrays and correlated with overall and tumor related survival. Results 129 patients (48 female) with a mean follow-up of 129.1 months were identified. 5-year overall survival was 83.9%, 5-year tumor related survival was 95.1%. Poorly differentiated medullary cancer subtypes (p<0.001) and positive vein invasion (p<0.001) were identified as risk factors for decreased overall—and tumor related survival. Ki-67 (p = 0.012) and TP53 mutation (p = 0.044) were the only immunohistochemical markers associated with worse overall survival but did not reach significance for decreased tumor related survival. Conclusion In the presented study patients with AGE/S in stage UICC-I-N0 had a better prognosis as previously reported for Caucasian patients. Poorly differentiated medullary subtype was associated with reduced survival and should be considered when studying prognosis in these patients.


Introduction
Adenocarcinoma of the esophago-gastric junction and stomach (AEG/S) is the fourth most common cancer worldwide in men and the fifth most common cancer diagnosis in women [1]. The prognosis of patients with AEG/S is poor, making it the second leading cause of cancer death worldwide [1].
Complete resection (R0) is considered the treatment of choice with curative intent, yet, in up to 40% of patients tumor recurrence occurs even after R0 resection [2]. An extended D2 lymphadenectomy as surgical procedure is well established, confirming the role of lymphatic invasion in this cancer entity [3,4]. Approximately 30% of patients with AEG/S exhibit stage UICC I at time of diagnosis [5] and despite early tumor-stage and radical tumor resection, this group of patients has a 5-year overall survival rate of only 60%-80% [6,7]. In these studies lymph node involvement was the main risk factor for tumor relapse. Due to higher incidence of this cancer type in East Asia, most studies on prognosis and survival describe patients from Japan and China. Solely Borie et al. [8] studied the survival of Caucasian patients stage UICC I stratified according to lymph node involvement and demonstrated a tumor related five year survival of 92% in patients without lymph node involvement.
These biomarkers have not yet been analyzed in this subgroup with AEG/S, UICC I without the main confounder lymph node involvement in order to refine the identification of patients at high-risk.
To this end we conducted this retrospective study with the aim to identify prognostic histopathological and immunohistochemical markers using tissue microarrays (TMAs) from patients with AEG/S stage UICC I-N0 and correlated these data with overall survival and tumor related survival. Refined biomarkers in this subgroup might not only be predictive for survival, but might guide treatment to improve survival in high-risk patients by possibly intensifying treatment by adjuvant chemotherapy or chemoradiotherapy.

Patients
Clinical data from patients with AEG/S stage pT1 and pT2 pN0M0, treated solely by surgery between September 1993 and May 2010 at Charité-University Medicine Berlin, were collected retrospectively.
The data including patient characteristics and follow-up information were retrieved from the patient management software (SAP1) and the regional population-based cancer registry (Gemeinsames Krebsregister). This study was approved by the Institutional Ethical Review Board of the Charité (EA4/115/10).
All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964 and later versions. No patient consent was necessary since this was a retrospective study. Patients' records were collected and anonymized, de-identified before the analysis.
Patients without recurrent disease were followed-up for a minimum period of 24 months. Overall survival, used as a measure of prognosis, was defined as time from diagnosis to death or last follow-up. Tumor related survival was defined as time from diagnosis to tumor related death or last follow-up.

Tissue Samples
Tissue samples were collected from the pathological archive of the Charité-University Medicine Berlin. From 129 surgically treated patients 80 paraffin embedded tumor samples were available. A specialist in gastrointestinal pathology (H.B.) reanalyzed and reconfirmed the primary postoperative histological diagnosis and tumor stage and classified the histological architecture of gastric carcinoma using Lauren's, Ming and WHO classification [34]. According to the WHO classification for colon and rectum carcinoma we added the subtype of medullary tumors to our analysis [34][35][36]. We defined the poorly differentiated medullary cancer (PMC) subtype in congruence with Adachi et al. as a tumor characterized by sheets of monomorphic malignant cells showing solid growth and absence of glandular growth [35]. Tumor cells show only little pleomorphism, have round to oval, large nuclei and a high nuclear to cytoplasmic ratio, morphologically resembling medullary type colorectal cancer. The PMCs did not show a deficiency of the mismatch repair system, as tested by immunohistochemistry.
Additional data concerning tumor size, depth of invasion, and tumor invasion of veins or lymphatic vessels were extracted from the Charité-University Medicine patient management software.

TMA and Immunohistochemistry
Tissue samples were screened in a HE-stained section for representative areas of solid tumors. Two 1 mm-diameter tissue cores were punched out from each of the 80 available cases and were transferred to a recipient paraffin block. Tissue from a tonsil was used as positive staining control and was also transferred to the same paraffin block. After re-melting, sections (4 μmthick) were consecutively cut from each tissue microarray block. HE staining was performed on TMA sections for confirmation of tumor and non-tumor tissue.
Concerning Ki-67, tumors with a proliferation rate higher than 20% were regarded as highly proliferative tumors. For Survivin, all tumors with more than 5 percent positively stained cell nuclei were scored as Survivin positive samples [37]. A high number (75-100%) of TP53 positive stained cells with a strong nuclear staining intensity or tumors with no TP53 positive cells was used as a surrogate parameter for TP53 mutated samples [38,39].
Mismatch repair deficiency was scored, when a tumor showed specific loss of nuclear staining for at least one of the markers (MLH, PMS2, MSH2, MSH6) The count of CD3 positive cells was performed in 4 representative regions of the tumor. The average of counted cells was normalized in 0.1mm 2 of tumor area.
Each core was blindly scored by two independent evaluators (C.T. and S. D.) and reviewed by one pathologist (H.B.) and one biologist highly experienced in pathohistology (M.H.) and reconciled after intensive discussion.

Statistics
Quantitative values are expressed as mean± standard deviation, median, and range, and categorical values with absolute and relative frequencies (count and percent). Overall survival was evaluated in months from time of diagnosis until death or until the most recent follow-up using the Kaplan-Meier plots. Differences in the survival curves were evaluated by the logrank test for significance. The X 2 -test was used for comparison of frequencies. A pvalue < 0.05 was considered as statistically significant. IBM SPSS Version 21 (Ehningen, Germany) was used for statistical analysis.
Reporting recommendations for tumor marker prognostic studies (REMARK) were applied for this study whenever applicable [40].

Clinical Characteristics of Patients
Data of 129 patients (female = 48, age 29-90 years, median age 62.9 years) were retrieved for this study (see Table 1). The cohort consisted of 96 patients with gastric cancer (2 pts. upper third, 26 middle third, 55 lower third, 8 couplet stomach and 5 with no specified position in the stomach) and 33 patients with carcinoma of the gastroesophageal junction. A D2 lymph node dissection was performed in 97.6% (126 patients) of AEG and stomach resections (two patients had only a D1 dissection and in one patient lymph node dissection status was not reported.) The mean follow-up was 129.1 months and 21 patients (16.3%) died during the follow-up period, but only nine patients (7%) died from recurrence of AEG/S. The 5-year overall survival of our population was 83.9%, the 5-year tumor related survival was 95.1% (Fig 1). A relapse of AEG/S occurred in nine patients whereas 120 patients remained relapse-free in the observation period.

Histopathological Characteristics
From 129 patients with clinical data, 80 formalin-fixed paraffin embedded tissue samples were available and reassessed (  Fig 2A and 2B). Using the Ming classification, tumors were in 38 cases (47.5%) of the expansive and in 37 cases (46.3%) of the infiltrative type. In five cases (6.3%) a definite classification was not possible ( Table 2).
The correlation of histopathological characteristics with tumor related survival or overall survival showed significant differences between the overall survival and tumor related survival for the V0 and V1 subpopulation (5-year tumor related survival V0 96.7%, V1 60.0% p<0.001;
A multivariate cox regression model was not reasonable as the number of our cases, especially of relapsed cases, did not meet the critical number proposed by Peducci et al [41].

Immunohistochemistry
Although all cases were present in duplicate, between 12.5 to 20% (10-20) of the cases were not usable for analysis in the TMA. Reasons for exclusion were cores with less than 10% tumor tissue. In the analysis of the immunhistochemical markers BAX, BCL-2, CDX2, Cyclin E, E-Cadherin, Ki-67, TP53, TP21, SHH, Survivin, HIF1A, and TROP2 we could not find a significant correlation with marker expression and prognosis (see Table 3.).
In a more detailed analysis for positive and negative markers we detected a lower 5-year overall survival for Ki-67 high proliferative tumors (high: 79.9% vs. low: 100% p = 0.012) and for tumors with TP53 surrogate mutation pattern (mut 66.3% vs. wt 92.5%; p = 0.044). These markers showed a trend in the 5-year-tumor related survival analysis (Ki-67 high 90.2% vs.  Table 3. Immunohistochemical biomarker analysis: All immunohistochemical markers except Ki-67, TP53, and Survivin were analyzed using the semiquantitative scoring system described in the method section. low 100% and TP53 mut: 85.2% vs wt 95.7%), but did not reach level of significance (Ki-67 p = 0.075 and TP53 p = 0.057). The analysis of mismatch repair system deficiency and proficiency did not show any significant differences for overall survival or tumor related survival (5-year tumor related survival dMMR 100% vs. pMMR 92.1%, p = 0.461; 5-year overall survival dMMR 75.0% vs pMMR 82.0%, p = 0.842) With regard to the above identified histomorphological risk factors (poorly differentiated medullary cancer subtype, venous invasion) we found in PMC significantly more T2 tumors (p = 0.034), more undifferentiated tumors (p = 0.024) and more p53 mutated tumors (p = 0.045). None of the PMC showed a deficient MMR.

5-year 10-year
For detailed classification of the PMC we counted the CD3 positive lymphocyte per 0.1mm 2 tumor area. We detected less CD3 positive cells in the PMC compared to non-PMC tumors (120.5 per 0.1mm 2 vs 212.3 per 0.1mm 2 ). The difference did not reach level of significance (p = 0.1).

Discussion
The aim of this retrospective study was to identify prognostic histopathological and immunohistological markers for patients with AEG/S stage UICC I on top of the known main risk factor lymph node metastasis to identify potential high risk subgroups.
Detailed clinical data were retrieved for 129 patients and histopathological data were assessable in 80 cases in this study. This Caucasian UICCI/N0 patient cohort with FFPE samples is one of the largest of gastric cancer in Europe as the incidence is substantial lower and tumor stages are more advanced in comparison to Eastern Asia [42,43].
From 129 patients followed over 10 years we found 9 patients with tumor related death. This high 5-year tumor-related overall survival of 95.1% proves the effectivity of surgical treatment in this early stage disease without lymph node involvement. Similar favorable data are well known from Asian patients (97.8% 5-year tumor-related overall survival) [44]. Similar results in Caucasians were reported in a single study by Borie et al. with a 92% 5-year tumor- related overall survival [8]. The large German gastric cancer trial showed in about 250 patients recruited from 1986-1989 a 5-year overall survival depending on the number of resected lymph nodes from 81%-85% in T1N0 stages to 67% in T2N0 stages [6]. These differences in survival are probably attributable to increased surgical quality in the last decades and the decreased postoperative 30 day mortality rate (mean 30 day mortality rate of 0.8% in 18 trials between 1998-2014 [45] compared to the German gastric cancer trial with 5.2% [3]). For the histopathological markers in our study, venous infiltration was a significant predictor for tumor related death in our patient group (5 year-tumor related survival V0 96.7%, V1 60.0% p = 0.005). While these data are in accordance with Asian studies [10,11], we found a higher number of patients with poorly differentiated medullary cancer subtype, namely 11.3% (9/80) whereas the Asian data reported about 3.0% of PMC histology, defined as low differentiated lymphoepithelioma-like tumor type with a small amount of stromal tissue [35,36]. Although definition of medullary subtype is still a point of discussion, our defining criteria are easily reproducible. Similar to Asian patients, PMC subtype was associated with relapse and reduced prognosis in our patient population [35,36]. Although PMC phenotype (see Fig 2A  and 2B) is not part of the classification systems currently in use (WHO, Ming, Lauren), it seems to represent an atypical form of gastric cancer which should be studied further, considered as a separate entity, and included in histopathological classification systems [46].
According to our data, all other parameters analyzed, like Lauren and Ming classification, tumor size, grading, lymph vessel invasion, and signet ring cell type, did not show a prognostic relevance. As other authors found grading, lymph vessel invasion and tumor size as prognostic in early tumor stages mainly with lymph node involvement [10][11][12][13][14], employing a more detailed analysis (N0 versus N+) of the tumor, might explain these differences.
The conclusion from our immunohistochemical data for prognostic biomarkers is somewhat limited by the rather small number of cases, solely 58-70 from 129 included patients could be analyzed, and the number of patients with tumor related death among this group amounts to six patients (see Table 3.). A reduced overall survival for TP53 surrogate mutation pattern and for high proliferative (Ki-67>20%) tumors could be demonstrated. The identified histopathological risk factors vein invasion and PMC subtype were positively correlated with a TP53 mutation pattern. TP53 and Ki-67 did not reach the level of significance in correlation with disease relapse in contrast to correlation with death by any cause.
Yet, the results for TP53-mutating pattern and high proliferating tumors are in concordance with other studies. Over all stages of gastric cancer Ki-67 is associated with metastasis and disease progression in gastric cancer and is especially associated with lymph node involvement in early diseases [47,48].
Also TP53 mutation is known to be associated with reduced survival [15,49,50] and the mutation of TP53 plays a crucial role in the progression from intestinal metaplasia to gastric cancer [49,51]. Despite missing significance (p = 0.057) for this biomarker, the immunohistochemical TP53 surrogate mutation pattern seems to be a suitable molecular biomarker for early gastric cancer.
This retrospective study in a non-selected group of patients with early adenocarcinoma of the esophago-gastric junction and stomach corroborates older findings of a favorable prognosis in cases without lymph node involvement. We could confirm the negative prognostic role of vein invasion in our population despite the small number of patients with relapse. Furthermore, we could identify the subtype of poorly differentiated medullary cancer as a risk factor for disease recurrence in a Caucasian population.
The prognostic role of these findings has to be investigated in more detail in further adequately powered prospective studies.
Supporting Information S1 Table. Clinical data and IHC scoring value for each marker. (XLSX)