Skip to main content

A panel of intestinal differentiation markers (CDX2, GPA33, and LI-cadherin) identifies gastric cancer patients with favourable prognosis

Abstract

Background

Gastric cancer is the fifth most common cancer and the third cause of global cancer mortality. CDX2 is an intestinal differentiation marker with prognostic value in gastric cancer and transcriptionally regulates the expression of glycoprotein A33 (GPA33) and liver intestine cadherin (LI-cadherin).

Methods

This study evaluated the clinical significance of the combined expression of CDX2 and its targets GPA33 and LI-cadherin in gastric cancer by fluorescence-based multiplex immunohistochemistry together with digital image analysis and chromogenic immunohistochemistry in 329 gastric cancer samples arranged in tissue microarrays. Additionally, publicly available RNA-seq expression data from 354 gastric cancer samples from the TCGA database were used to validate the immunohistochemistry results.

Results

Expression of the three markers (CDX2, GPA33, and LI-cadherin) was strongly correlated, defining an intestinal differentiation panel. Low or negative protein expression of the intestinal differentiation panel identified patients with particularly poor overall survival, irrespective of the methodology used, and was validated in the independent series at the RNA-seq level.

Conclusions

Expression of the intestinal differentiation panel (CDX2, GPA33, and LI-cadherin) defines a set of biomarkers with a strong biological rationale and favourable impact for prognostication of gastric cancer patients.

Introduction

Gastric cancer is the fifth most common cancer type and the third cause of cancer-related death worldwide [1]. Patients are primarily treated with surgery, radiotherapy, and chemotherapy depending on tumour stage and location [2], and disease aggressiveness depends on tumour differentiation, among other factors.

An intestinal differentiation hallmark is caudal-related homeobox transcription factor 2 (CDX2), a transcription factor involved in intestinal development and differentiation [3]. We demonstrated that CDX2 is expressed in intestinal metaplasia of the stomach and dysplasia [4,5,6,7]. More than half of gastric cancer cases are CDX2-positive and CDX2 correlates with the intestinal mucin MUC2 [4, 8]. Furthermore, negative or low CDX2 expression associates with poor survival in gastric cancer [9] and, in colorectal cancer, an impact on survival was reported in different disease stages [10, 11]. We recently generated a CDX2 knock-out colorectal cancer cell line and identified novel and known gene targets transcriptionally controlled by CDX2 [12, 13].

Glycoprotein A33 (GPA33), which we recently identified as a CDX2 target [13], is a cell-surface differentiation glycoprotein and a member of the immunoglobulin superfamily [14]. Although its function is unknown, it was suggested that GPA33 is implicated in cell–cell adhesion [15]. GPA33 is expressed in the intestinal mucosa and in tumours of the gastrointestinal tract. In colorectal cancer, GPA33 is expressed in more than 95% of primary and metastatic tumours [16], whereas in gastric cancer, it is expressed in about two-thirds of the tumours and is absent in normal epithelium [17]. Based on these observations, GPA33 is an attractive target for immunotherapy in both gastric and colorectal cancer [18, 19].

Liver intestine cadherin (LI-cadherin), encoded by CDH17, was shown to be regulated by CDX2 in normal and malignant colonic epithelium and intestinal metaplasia [20]. Several studies reported LI-cadherin expression in intestinal metaplasia [20,21,22,23], as well as gastric [21,22,23,24,25,26,27,28,29] and colorectal cancer [20, 29]. Data from gastric cancer cell lines suggest that CDH17 might be a novel oncogene, since CDH17 inhibition counteracts several hallmarks of cancer [30, 31] and affects critical signalling pathways [30, 32, 33].

Here, we analysed the degree and pattern of expression of CDX2, GPA33, and LI-cadherin in a series of gastric carcinomas, using digital and visual approaches. Their clinical value was assessed individually and in combination. We propose a three-marker panel with strong biological coherence and impact on gastric cancer survival.

Materials and methods

Patients and tumour samples

This study includes gastric adenocarcinomas from total or distal gastrectomy at Centro Hospitalar São João (CHSJ), Porto, Portugal, between January 2008 and December 2014 (n = 329, Supplementary Table S1). The series excludes patients lost to follow-up and without paraffin-embedded material (Supplementary Fig. S1). Clinicopathological, treatment, and follow-up data were obtained from the institutional records at CHSJ and the hospital-based cancer registry. Patients were observed every 3 months during the 2 years after surgery and every 6 months or annually in the subsequent years. Data were collected prospectively and analysed retrospectively. All patients were followed up throughout the study period. The use of retrospective samples from which informed consent cannot be obtained is authorised for research studies by the Portuguese law. Recent samples included in the biobank had written informed consent. The study was approved by CHSJ ethics committee.

The Cancer Genome Atlas (TCGA; https://cancergenome.nih.gov/) data set of gastric carcinomas, including 354 cases for which expression data (based on RNA-seq) and updated survival information were available, was used as a validation series. These cases had diverse geographical origins (America, Europe, Asia and unknown), mean age of 65.5 years, 35.3% females, and 64.7% males.

Construction of tissue microarrays (TMAs)

Histological material was collected from surgical specimens using a standard protocol and tissue was fixed in 10% buffered formalin. Representative tumour tissue areas were selected from central parts of the tumour based on visual evaluation of the hematoxylin–eosin-stained sections by one pathologist (LD). One tissue core (2 mm in diameter) was obtained from each specimen and deposited into a recipient paraffin block using an Arraymold Kit A (IHC World, Woodstock, Maryland, USA) and nine TMA blocks were constructed. In each block, one gastric mucosa core and one gastric cancer cell line core were included as controls, as well as a normal colonic mucosa sample for orientation purposes. To homogenise the paraffins from the receptor block and from the cores extracted from donor blocks, TMAs were kept at 37 °C overnight. Afterwards, they were placed at room temperature for 1 h, followed by two cycles of 1 h at 37 ºC plus 1 h at room temperature. After homogenisation, an H&E-stained section from each TMA block was reviewed to confirm the presence of morphologically representative areas of the original tumours. For 26 cases, two separate cores from the tumour were included on the TMA, and for another 23 cases, two cores from the primary tumour and the corresponding lymph-node metastasis were included.

Fluorescence-based multiplex immunohistochemistry (mIHC)

Fluorescence-based mIHC was performed using the Opal™ 4-Color Manual IHC Kit (PerkinElmer, Waltham, Massachusetts, USA) according to the manufacturer’s protocol. Deparaffinisation and antigen retrieval were completed in a PT link module (Dako/Agilent, Santa Clara, California, USA) according to the manufacturer’s recommendations for 20 min at 97 °C using the EnVision™ FLEX Target Retrieval Solution (3-in-1) pH 9 (Dako/Agilent), in 65 °C preheat mode. Tissue samples were stained for 30 min with the following primary antibodies: CDX2 (1:50, clone 88, Abcam, Cambridge, UK; detected by Opal 520 at 1:200), GPA33 (1:2000, clone EPR4240, Abcam; detected by Opal 570 at 1:100), and LI-cadherin (1:5000, clone EPR3997, Abcam; detected by Opal 620 at 1:400). Afterwards, the tissue was stained with an epithelial marker cocktail to allow epithelial segmentation by the digital image analysis algorithm (anti-pan Cytokeratin [1:1500, C-11, Abcam] and anti-pan Cytokeratin Type I/II [1:1000, AE1/AE3, Thermo Fisher Scientific, Waltham, Massachusetts, USA], plus anti-E-cadherin [1:16,000, Clone 36, BD Biosciences San Jose, California, USA]; all detected using Opal 690 at 1:200). DAPI (PerkinElmer) was used for counterstaining, following the manufacturer’s recommendations. Slides were mounted using ProLong Diamond Antifade Mountant (Invitrogen/Thermo Fisher Scientific). A negative control was performed by omitting primary antibody during the staining procedure; no signal above noise values was detected here. A separate single-plex stain was done for each fluorophore to create spectral libraries for unmixing of individual spectral signatures in the multiplex. Additionally, one slide was not probed with any fluorophore to create the spectral signature of the tissue autofluorescence. Additional details concerning antibody concentration choice and signal intensity use can be found in [34].

Image acquisition and digital image analysis (DIA)

Multispectral images were obtained at  × 20 magnification between 420 and 720 nm at 20 nm intervals for each fluorescence multiplex stain using the Vectra 3.0 Automated Quantitative Pathology Imaging System (Vectra software version 3, PerkinElmer).

inForm Image Analysis Software (version 2.3, PerkinElmer) was used for the mIHC stains multispectral image analysis. First, a representative set of training images was selected and spectrally unmixed using spectral libraries generated from the library stains performed for each fluorophore and the autofluorescence stain. Afterwards, a machine learning algorithm was trained using tissue annotations aided by the epithelial markers’ signal to segment tumour tissue, as well as individual cells using the DAPI signal. After batch processing, all images were reviewed and regions of disinterest, such as normal mucosa, necrotic tissue, tissue folds, and other artefacts were excluded from analyses.

Chromogenic immunohistochemistry

All IHC assays were performed on 4 µm-thick sections. After deparaffinisation, antigen retrieval was completed in a Steamer Set (Black & Decker, Towson, Maryland, USA) for 40 min using Citrate buffer (Thermo Fisher Scientific) for CDX2 and EDTA buffer (Thermo Fisher Scientific) for GPA33 and LI-cadherin. Staining was performed using the Dako REAL™ EnVision™ Detection System Peroxidase/DAB + Rabbit/Mouse Kit (Dako/Agilent). Tissue samples were stained overnight with the following antibodies: CDX2 (1:50, clone 88, Biogenex, Fremont, California, USA), GPA33 (1:1000, clone EPR4240, Abcam), and LI-cadherin (1:1000, clone EPR3997, Abcam). Following counterstaining using Mayer’s hematoxylin (Dako/Agilent), slides were dehydrated and mounted using Richard-Allan Scientific™ Mounting Medium (Thermo Fisher Scientific). Normal colonic mucosa was used as a CDX2, GPA33 and LI-cadherin positive control, and cases were considered positive when more than 5% of the tumour cells were stained [9].

GPA33 and LI-cadherin required EDTA buffer as the antigen retrieval solution, both in chromogenic immunohistochemistry and in fluorescence-based mIHC. However, for CDX2 a comparison of CDX2 staining using Citrate buffer (used in chromogenic immunohistochemistry) and EDTA buffer (used in mIHC) was performed in six cases and a demonstrative image is shown in Supplementary Fig. S2. The staining obtained in all cases was similar with overlapping positive areas, although EDTA produced a more intense signal.

Statistics

This study followed the REMARK guidelines [35] to report biomarker studies (Supplementary Table S2). The Student t test was used for age comparisons. The Chi-square (χ2) test was used when comparing for Laurén classification, TNM stage, and T stage, and the Fisher’s exact test was used when comparing for gender, growth pattern, N stage, resection margins, and vascular and perineural invasion.

The Kaplan–Meier method was used to generate 5-year disease-free (DFS) and overall survival (OS) plots, and the survival curves were compared using the log-rank test for trend. DFS was defined as the time from surgery to the first event of either locoregional recurrence or metastasis, or death from the same cancer. OS was defined as the time from operation to death from any cause. Cox proportional hazard model were used to calculate univariable and multivariable hazard ratios and confidence intervals for disease recurrence. The clinicopathological parameters included in the multivariable models were selected based on their individual clinical relevance and a priori knowledge in a full model approach with no other predictor selection. Patients with missing data were excluded from the respective parameter analyses in the multivariable models.

The fluorescence-based mIHC comparison of biomarkers expression between two cores from the same primary tumour or between primary tumour and lymph-node metastases was carried out with R software [36]. Ladder plots were performed with the “ladderplot” function from R package “plotrix” version 3.7-2, deposited in CRAN on 2018-05-27.

To determine the 30% cases with high and low biomarker expression in the fluorescence-based mIHC, the “quantile” function from R package “stats” was used with probabilities = 0.30 and probabilities = 0.70. R studio version 1.0.143 was used to run the R script. For the intestinal differentiation panel, four categories were defined: the category “one marker low” identified those samples that exhibited any of the markers as low (and the remaining as intermediate or high), “two markers low” as those with any two markers as low (and the other either as intermediate or high), “all markers low” and “no marker low”.

The expression data inferred by TCGA were used, in fragments per kilobase million (FPKM) units. The expression-based cut-off values to establish the low-expression and high-expression groups for each gene were obtained from The Human Protein Atlas (https://www.proteinatlas.org/), which chooses the FPKM values corresponding to the lowest log-rank P values of the survival outcomes in the 20th to 80th percentiles of FPKM distributions. Individual (per gene, CDX2, GPA33, and CDH17, comparing low-expression and high-expression groups) and overall (expression of the three genes taken together, by comparing four groups: “one marker low”, “two markers low”, “all markers low”, and “no marker low”) Kaplan–Meier OS plots were calculated in R, using survival and survminer packages.

Statistical analysis was performed in IBM SPSS Statistics version 24 (Chicago, Illinois, USA). In all analyses, differences were considered statistically significant whenever the p value was < 0.05.

Results

CDX2, GPA33, and LI-cadherin expression are strongly correlated

The expression of the three markers was strongly correlated using both fluorescence-based mIHC and chromogenic IHC (Fig. 1 and Table 1). Using fluorescence-based mIHC, the correlation coefficient between CDX2 and GPA33 was 0.66 (p < 0.01), whereas that between CDX2 and LI-cadherin was 0.55 (p < 0.01) and between GPA33 and LI-cadherin was 0.64 (p < 0.01). Using chromogenic IHC, the correlation coefficient between CDX2 and GPA33 was 0.37 (p < 0.01), while that between CDX2 and LI-cadherin was 0.43 (p < 0.01) and between GPA33 and LI-cadherin was 0.63 (p < 0.01). Using chromogenic IHC, CDX2 expression was detected in 38% of the cases, while GPA33 was positive in 57% and LI-cadherin in 68% of tumours. In some cases, the expression of CDX2, GPA33, and LI-cadherin was co-localised in restricted tumour areas (Figs. 2a, b 10–12).

Fig. 1
figure1

Association between CDX2, GPA33, and LI-cadherin expression. The expression of the three markers is significantly correlated, using fluorescence-based mIHC, where larger bubble size indicates higher CDX2 nuclear expression

Table 1 Associations between the protein expression of CDX2, GPA33, and LI-cadherin
Fig. 2
figure2

Patterns of staining of CDX2, GPA33, and LI-cadherin in gastric cancer. a Expression of merged expression of all markers (1), showing example of co-localisation and individual expression of CDX2 (2), GPA33 (3), and LI-cadherin (4). Data obtained using fluorescence-based mIHC; magnification × 200. b All markers are expressed in the intestinal type (1–3) and diffuse type (4–6) of gastric carcinomas, as well as in lymph-node metastasis (7–9). In some cases, there is co-localisation of CDX2, GPA33, and LI-cadherin expression (10–12) in the same tumour areas. Data obtained using chromogenic IHC; magnification:  × 200 (1–3),  × 400 (4–6),  × 100 (7–9),  × 40 (10–12) and 200 (insert)

CDX2, GPA33, and LI-cadherin show homogenous intra-tumour expression that is maintained in lymph-node metastasis

Fluorescence-based mIHC followed by DIA allows quantification and objective measurements of CDX2, GPA33, and LI-cadherin expression, and was used to assess gastric cancer heterogeneity in 26 patients from which two cores of the same tumour were available. No significant differences in the expression of CDX2, GPA33, and LI-cadherin were found in the two samples of individual patients (Supplementary Fig. S3a–S3c). The series comprised 23 pairs of primary tumours and the corresponding lymph-node metastases, where CDX2, GPA33, and LI-cadherin were expressed at similar levels (Supplementary Fig. S3d–S3f).

CDX2, GPA33, and LI-cadherin marker panel: associations to clinicopathologic characteristics, tumour stage, and survival

The clinicopathological features from patients/tumours included in this study are presented in Supplementary Table S1. These features were determined by visual inspection of H&E sections, and their association with CDX2, GPA33, and LI-cadherin chromogenic IHC expression is shown in Supplementary Tables S3–S5. The rationale for using chromogenic IHC to determine clinicopathological associations is that this method is more commonly used, is directly comparable to other studies, and allows dichotomisation of results, simplifying the analyses. All markers were significantly associated with early stages (I/II) of disease progression, the expression of CDX2 and GPA33 correlated with less advanced T stage and GPA33, and LI-cadherin expression was associated with tumour-free resection margins. Lack of GPA33 and LI-cadherin expression was associated with the presence of lymph-node metastasis (67% and 71% of carcinomas, respectively). CDX2 and LI-cadherin were significantly more expressed in the intestinal subtype of gastric tumours and GPA33 showed the same trend.

Since CDX2 is a transcription factor related to good prognosis in gastric cancer, we tested whether GPA33 and LI-cadherin, transcriptionally regulated by CDX2, could add prognostic value beyond CDX2. The biological rationale is that all markers belong to an intestinal differentiation programme [13], and even in the absence of CDX2, through other transactivating pathways [37,38,39,40], the presence of GPA33 and LI-cadherin may “rescue” a CDX2-independent intestinal phenotype with prognostic impact.

Survival analyses were performed using fluorescence-based mIHC combined with DIA. Since this is a quantitative method based on continuous data, three categories were defined for each marker: “high”, comprising the samples that exhibited the 30% highest levels of expression, “low”, including the samples with the 30% lowest levels of expression, and “intermediate” which encompassed the remaining samples. To obtain a combined differentiation panel, four categories were defined as described in “Materials and methods” section: “one marker low”, “two markers low”, “all markers low”, and “no marker low”. Using this method, it was observed that gastric cancer patients expressing intermediate and high levels of CDX2 had significantly better overall and disease-free survival than those with low levels of expression (Figs. 3a, 4a). A marginal difference was also observed for overall survival in cases expressing intermediate and high LI-cadherin levels (Fig. 3c). The same trends were observed for disease-free survival and GPA33, but the differences were not significant (Figs. 3b, 4b, c). When the expression of CDX2, GPA33, and LI-cadherin was combined into a panel, patients with one or no marker low had a better overall and disease-free survival (Figs. 4d, 5a). This association was supported by univariable Cox proportional hazard models for overall and disease-free survival (Tables 2, 3). A multivariable Cox proportional hazard model, including Laurén classification, growth pattern, TNM staging, resection margins, vascular invasion, and perineural invasion as covariates, which showed that the differentiation panel was not an independent prognostic factor (Tables 2, 3).

Fig. 3
figure3

Kaplan–Meier curves showing overall survival for gastric cancer patients according to the expression of CDX2, GPA33, and LI-cadherin (data obtained using fluorescence-based mIHC, chromogenic IHC, and RNA-seq). Using fluorescence-based mIHC, the high and intermediate expressions of CDX2 are associated with a better overall survival (a). No significant differences are observed for GPA33 (b) and LI-cadherin (c). Using chromogenic IHC, no significant differences are observed for CDX2 (d), but GPA33 (e) and LI-cadherin (f) are markers of good prognosis. In the validation series from The Cancer Genome Atlas (RNA-seq analysis), CDX2 (g) and GPA33 (h) are markers of good prognosis, but no significant differences are observed for LI-cadherin (i). For fluorescence-based mIHC and chromogenic IHC, overall survival analyses have been performed in stages I–IV (n = 325); n indicates the total number of individuals at risk at the initial timepoint (zero months), whereas the individuals at risk at the final timepoint (60 months) for each condition are indicated by the subtraction (n minus events). Cox regression was used to generate hazard ratios (HR) and 95% confidence intervals (CI)

Fig. 4
figure4

Kaplan–Meier curves showing disease-free survival for gastric cancer patients according to the expression of CDX2, GPA33, LI-cadherin, and the panel of all markers (data obtained using fluorescence-based mIHC). The high expression of CDX2 is associated with a better disease-free survival (a). No differences were observed for GPA33 (b), LI-cadherin (c), and the panel of markers (d). Disease-free survival analyses have been performed in stages I–III (n = 254); n indicates the total number of individuals at risk at the initial timepoint (zero months), whereas the individuals at risk at the final timepoint (60 months) for each condition are indicated by the subtraction (n minus events)

Fig. 5
figure5

Kaplan–Meier curves showing overall survival for gastric cancer patients according to the expression of the panel of all markers (data obtained using fluorescence-based mIHC, chromogenic IHC and RNA-seq). Using fluorescence-based mIHC, gastric cancer patients with two or all markers low have a worse overall survival (a). Using chromogenic IHC, the absence of expression of all markers is associated with worse prognosis (b). In the validation series from The Cancer Genome Atlas (RNA-seq analysis), gastric cancer patients with all markers low display worse prognosis, whereas patients with no marker low show a better overall survival (c). For fluorescence-based mIHC and chromogenic IHC, overall survival analyses have been performed in stages I–IV (n = 325); n indicates the total number of individuals at risk at the initial timepoint (zero months), whereas the individuals at risk at the final timepoint (60 months) for each condition are indicated by the subtraction (n minus events)

Table 2 Univariable and multivariable Cox regression analyses (overall survival) in gastric cancer obtained using fluorescence-based IHC data for the differentiation panel
Table 3 Univariable and multivariable Cox regression analyses (disease-free survival) in gastric cancer obtained using fluorescence-based IHC data for the differentiation panel

For the differentiation panel establishment in chromogenic IHC, the markers were grouped as “all negative”, “one positive”, “two positives”, and “all positive”. Similar survival results were found compared with the fluorescence-based mIHC observations (Fig. 5b and Supplementary Table S6). Gastric cancer patients exhibiting GPA33 or LI-cadherin expression displayed a better overall survival (Fig. 3e, f) and a similar trend was observed for CDX2 (Fig. 3d). Importantly, patients negative for all markers had the worst prognosis (Fig. 5b).

A publicly available data set (TCGA) was used as a validation series to evaluate CDX2, GPA33, and LI-cadherin prognostic value in gastric cancer, as well as their combination in a biomarker panel. The comparison of our series with the TCGA regarding clinicopathological features, such as tumour stage and Laurén classification, is shown in Supplementary Table S7. Our series is enriched in stage I samples and has a lower percentage of unclassified cases. The RNA-seq data obtained from the TCGA series confirmed our results, indicating that high expression of CDX2 and GPA33 was significantly associated with good prognosis (Fig. 3g, h) and the same trend was observed for LI-cadherin (Fig. 3i). When these markers were combined in a panel, patients with no marker low showed a better overall survival, whereas patients with all markers low displayed worse prognosis (Fig. 5c).

Discussion

In this study, we evaluated the co-expression of CDX2 and its transcriptional targets GPA33 and LI-cadherin, and their individual and joint prognostic value in gastric cancer. The panel defines a coherent intestinal differentiation programme that refines the individual markers clinical significance. GPA33 and LI-cadherin are regulated by CDX2 as previously demonstrated by our group [13], but other transactivating pathways may “rescue” a CDX2-independent intestinal phenotype, reinforcing that their combined expression may become a new tool to define a prognosis-relevant intestinal differentiation programme. In addition to CDX2, the expression of GPA33 was shown to be under the transcriptional control of GKLF (gut-enriched Krüppel-like factor) and CDX1 [37, 38]. Also, PPARγ indirectly induces GPA33 expression in colon cancer cell lines via a KLF4-dependent mechanism [41]. CDH17 transcription can be regulated by HNF1α (hepatic nuclear factor 1α) in hepatocellular carcinoma [40] and EGFR as well as HOXA13 (homeobox A13) in gastric cancer [39, 42]. The multi-biomarker thus overcomes the context-specific role of CDX2 epigenetic regulation in lineage plasticity [43].

Using two protein expression methodological approaches, we also assessed the biomarker panel robustness. We confirmed that the results partly depend on the chosen method. Since the fluorescence-based method has a longer linear range and a semi-automatic scoring, the correlations among the markers were more robust with fluorescence-based mIHC than with chromogenic IHC.

CDX2 is emerging as a promising biomarker in gastric [9] and colorectal carcinomas [10, 11]. Here, we show that gastric cancer patients exhibiting CDX2 expression have a better overall and disease-free survival, a result that agrees with our previous report [9]. Similarly, in colon cancer, CDX2 loss was associated with poor prognosis and benefit from chemotherapy [10, 11].

We observed that GPA33 was expressed in 57% of gastric carcinomas, which is in accordance with the only study that evaluated GPA33 expression in gastric cancer, reporting 63% of positive samples [17]. Importantly, we demonstrate for the first time that gastric cancer patients exhibiting GPA33 expression have a better overall survival. However, the prognostic value appears to depend on the detection method and on how GPA33 expression is grouped and needs confirmation in a larger series.

Data available regarding LI-cadherin expression and clinicopathological associations in gastric cancer are conflicting and should be cautiously interpreted. We found LI-cadherin expression in 68% of the tumours, which is within the range of results from other groups [22, 24, 27, 30, 31, 44] that observed expression in 60–75% of gastric cancer samples. In agreement with our results, it was reported that LI-cadherin was more frequently expressed in early gastric cancer stages (I/II) [27, 42]. More notably, we found that patients who display LI-cadherin expression have better overall survival. This agrees with the previous studies, using protein [28] or gene expression data [45]. However, discrepant results associate LI-cadherin expression with poor prognosis [24, 27, 32] and the presence of lymph-node metastasis [26, 27, 30, 31].

Our most interesting result is the strength in prognostic capacity of the intestinal differentiation panel using any of the two protein-based approaches or RNA-seq. The survival data reinforced the biomarker panel importance in comparison with the individual assessment of CDX2, GPA33, and LI-cadherin. Tumours displaying two or all markers with low expression by fluorescence-based mIHC associate with poor overall and disease-free survival. Similarly, absent expression of all the panel markers by chromogenic IHC indicates a poor prognosis. The reverse read-out of the results implies that the presence of an intestinal differentiation programme has a favourable impact on gastric cancer biological behaviour. In addition, the RNA-seq data further supported our survival results, clearly separating and significantly distinguishing a better overall survival for patients with high levels of expression of CDX2, GPA33, and LI-cadherin. Moreover, when cases with no marker low were compared with the combination of the other categories (one marker low, two markers low, and all markers low), the significant association with good prognosis is maintained (data not shown). This is the first demonstration that an intestinal differentiation programme in gastric cancer is associated with good prognosis, using three different methodologies. The use of RNA-seq is a surrogate for protein expression, since correlation between expression of mRNA and protein has been shown for the three intestinal differentiation markers. In the paper by Pinto et al. [13], we unequivocally demonstrated the strict relationship between RNA and protein expression for CDX2. Similar observations, showing that RNA and protein levels are associated, were demonstrated for GPA33 in a paper by Rageul et al. [41] demonstrating that upon treatment with a PPARγ agonist, there was an increase in GPA33 transcripts, followed by an upregulation of protein levels. Finally, several papers reported the simultaneous evaluation of CDH17/LI-cadherin expression by semi-quantitative/quantitative RT-PCR and immunohistochemistry (IHC)/Western blotting with similar RNA–protein associations [22, 23, 39, 40, 42]. Our results show that prognostic assessment using the biomarkers panel is more robust, since it is independent of the method used to evaluate expression.

Another aspect highlighted by our study is the similar expression profile of the three markers in different areas of the same tumour and in lymph-node metastasis. Likewise, CDX2 expression is concordant between primary colorectal carcinomas and the corresponding liver metastases [46]. Our observations that GPA33 and LI-cadherin expressions are similar in primary tumours and metastatic deposits support the notion that the intestinal differentiation programme, once established, is conserved during cancer progression. A similar observation, indicating stability of intestinal metaplasia, was previously shown by our group, by demonstrating that CDX2 is autoregulated by an effective loop and may explain the stability of human intestinal metaplasia in follow-up studies [47].

In conclusion, we demonstrate that CDX2 expression together with its targets, GPA33 and LI-cadherin, defines a panel of differentiation biomarkers with a strong biological rationale and a favourable impact for prognostication of gastric cancer patients.

References

  1. 1.

    Ferlay J, Colombet M, Soerjomataram I, Mathers C, Parkin DM, Pineros M, et al. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int J Cancer. 2019;144(8):1941–53.

    CAS  PubMed  PubMed Central  Google Scholar 

  2. 2.

    Smyth EC, Verheij M, Allum W, Cunningham D, Cervantes A, Arnold D. Gastric cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2016;27(suppl 5):v38–v49.

    CAS  PubMed  Google Scholar 

  3. 3.

    Beck F, Chawengsaksophak K, Waring P, Playford RJ, Furness JB. Reprogramming of intestinal differentiation and intercalary regeneration in Cdx2 mutant mice. Proc Natl Acad Sci USA. 1999;96(13):7318–23.

    CAS  PubMed  Google Scholar 

  4. 4.

    Almeida R, Silva E, Santos-Silva F, Silberg DG, Wang J, De Bolos C, et al. Expression of intestine-specific transcription factors, CDX1 and CDX2, in intestinal metaplasia and gastric carcinomas. J Pathol. 2003;199(1):36–40.

    PubMed  Google Scholar 

  5. 5.

    Barros R, Camilo V, Pereira B, Freund JN, David L, Almeida R. Pathophysiology of intestinal metaplasia of the stomach: emphasis on CDX2 regulation. Biochem Soc Trans. 2010;38(2):358–63.

    CAS  PubMed  Google Scholar 

  6. 6.

    Barros R, Freund JN, David L, Almeida R. Gastric intestinal metaplasia revisited: function and regulation of CDX2. Trends Mol Med. 2012;18(9):555–63.

    CAS  PubMed  Google Scholar 

  7. 7.

    Camilo V, Garrido M, Valente P, Ricardo S, Amaral AL, Barros R, et al. Differentiation reprogramming in gastric intestinal metaplasia and dysplasia: role of SOX2 and CDX2. Histopathology. 2015;66(3):343–50.

    PubMed  Google Scholar 

  8. 8.

    Mesquita P, Jonckheere N, Almeida R, Ducourouble MP, Serpa J, Silva E, et al. Human MUC2 mucin gene is transcriptionally regulated by Cdx homeodomain proteins in gastrointestinal carcinoma cell lines. J Biol Chem. 2003;278(51):51549–56.

    CAS  PubMed  Google Scholar 

  9. 9.

    Camilo V, Barros R, Celestino R, Castro P, Vieira J, Teixeira MR, et al. Immunohistochemical molecular phenotypes of gastric cancer based on SOX2 and CDX2 predict patient outcome. BMC Cancer. 2014;14:753.

    PubMed  PubMed Central  Google Scholar 

  10. 10.

    Dalerba P, Sahoo D, Paik S, Guo X, Yothers G, Song N, et al. CDX2 as a prognostic biomarker in stage II and stage III colon cancer. N Engl J Med. 2016;374(3):211–22.

    CAS  PubMed  PubMed Central  Google Scholar 

  11. 11.

    Bruun J, Sveen A, Barros R, Eide PW, Eilertsen I, Kolberg M, et al. Prognostic, predictive, and pharmacogenomic assessments of CDX2 refine stratification of colorectal cancer. Mol Oncol. 2018;12(9):1639–55.

    CAS  PubMed  PubMed Central  Google Scholar 

  12. 12.

    Pinto R, Barros R, Pereira-Castro I, Mesquita P, da Costa LT, Bennett EP, et al. CDX2 homeoprotein is involved in the regulation of ST6GalNAc-I gene in intestinal metaplasia. Lab Invest. 2015;95(7):718–27.

    CAS  PubMed  Google Scholar 

  13. 13.

    Pinto R, Hansen L, Hintze J, Almeida R, Larsen S, Coskun M, et al. Precise integration of inducible transcriptional elements (PrIITE) enables absolute control of gene expression. Nucleic Acids Res. 2017;45(13):e123.

    PubMed  PubMed Central  Google Scholar 

  14. 14.

    Heath JK, White SJ, Johnstone CN, Catimel B, Simpson RJ, Moritz RL, et al. The human A33 antigen is a transmembrane glycoprotein and a novel member of the immunoglobulin superfamily. Proc Natl Acad Sci USA. 1997;94(2):469–74.

    CAS  PubMed  Google Scholar 

  15. 15.

    van Niel G, Raposo G, Candalh C, Boussac M, Hershberg R, Cerf-Bensussan N, et al. Intestinal epithelial cells secrete exosome-like vesicles. Gastroenterology. 2001;121(2):337–49.

    PubMed  Google Scholar 

  16. 16.

    Garinchesa P, Sakamoto J, Welt S, Real F, Rettig W, Old L. Organ-specific expression of the colon cancer antigen A33, a cell surface target for antibody-based therapy. Int J Oncol. 1996;9(3):465–71.

    CAS  PubMed  Google Scholar 

  17. 17.

    Sakamoto J, Kojima H, Kato J, Hamashima H, Suzuki H. Organ-specific expression of the intestinal epithelium-related antigen A33, a cell surface target for antibody-based imaging and treatment in gastrointestinal cancer. Cancer Chemother Pharmacol. 2000;46(Suppl):S27–32.

    CAS  PubMed  Google Scholar 

  18. 18.

    Wu Z, Guo HF, Xu H, Cheung NV. Development of a tetravalent anti-GPA33/anti-CD3 bispecific antibody for colorectal cancers. Mol Cancer Ther. 2018;17(10):2164–75.

    CAS  PubMed  PubMed Central  Google Scholar 

  19. 19.

    Moore PA, Shah K, Yang Y, Alderson R, Roberts P, Long V, et al. Development of MGD007, a gpA33 x CD3-Bispecific DART Protein for T-Cell Immunotherapy of Metastatic Colorectal Cancer. Mol Cancer Ther. 2018;17(8):1761–72.

    CAS  PubMed  Google Scholar 

  20. 20.

    Hinoi T, Lucas PC, Kuick R, Hanash S, Cho KR, Fearon ER. CDX2 regulates liver intestine-cadherin expression in normal and malignant colon epithelium and intestinal metaplasia. Gastroenterology. 2002;123(5):1565–77.

    CAS  PubMed  Google Scholar 

  21. 21.

    Grotzinger C, Kneifel J, Patschan D, Schnoy N, Anagnostopoulos I, Faiss S, et al. LI-cadherin: a marker of gastric metaplasia and neoplasia. Gut. 2001;49(1):73–81.

    CAS  PubMed  PubMed Central  Google Scholar 

  22. 22.

    Ko S, Chu KM, Luk JM, Wong BW, Yuen ST, Leung SY, et al. CDX2 co-localizes with liver-intestine cadherin in intestinal metaplasia and adenocarcinoma of the stomach. J Pathol. 2005;205(5):615–22.

    CAS  PubMed  Google Scholar 

  23. 23.

    Dong W, Yu Q, Xu Y. Altered expression of a Li-cadherin in gastric cancer and intestinal metaplasia. Dig Dis Sci. 2007;52(2):536–42.

    CAS  PubMed  Google Scholar 

  24. 24.

    Ito R, Oue N, Yoshida K, Kunimitsu K, Nakayama H, Nakachi K, et al. Clinicopathological significant and prognostic influence of cadherin-17 expression in gastric cancer. Virchows Arch. 2005;447(4):717–22.

    CAS  PubMed  Google Scholar 

  25. 25.

    Motoshita J, Nakayama H, Taniyama K, Matsusaki K, Yasui W. Molecular characteristics of differentiated-type gastric carcinoma with distinct mucin phenotype: LI-cadherin is associated with intestinal phenotype. Pathol Int. 2006;56(4):200–5.

    CAS  PubMed  Google Scholar 

  26. 26.

    Park SS, Kang SH, Park JM, Kim JH, Oh SC, Lee JH, et al. Expression of liver-intestine cadherin and its correlation with lymph node metastasis in gastric cancer: can it predict N stage preoperatively? Ann Surg Oncol. 2007;14(1):94–9.

    PubMed  Google Scholar 

  27. 27.

    Ge J, Chen Z, Wu S, Yuan W, Hu B, Chen Z. A clinicopathological study on the expression of cadherin-17 and caudal-related homeobox transcription factor (CDX2) in human gastric carcinoma. Clin Oncol (R Coll Radiol). 2008;20(4):275–83.

    Google Scholar 

  28. 28.

    Lee HJ, Nam KT, Park HS, Kim MA, Lafleur BJ, Aburatani H, et al. Gene expression profiling of metaplastic lineages identifies CDH17 as a prognostic marker in early stage gastric cancer. Gastroenterology. 2010;139(1):213–25.e3.

    CAS  PubMed  PubMed Central  Google Scholar 

  29. 29.

    Panarelli NC, Yantiss RK, Yeh MM, Liu Y, Chen YT. Tissue-specific cadherin CDH17 is a useful marker of gastrointestinal adenocarcinomas with higher sensitivity than CDX2. Am J Clin Pathol. 2012;138(2):211–22.

    PubMed  Google Scholar 

  30. 30.

    Lin Z, Zhang C, Zhang M, Xu D, Fang Y, Zhou Z, et al. Targeting cadherin-17 inactivates Ras/Raf/MEK/ERK signaling and inhibits cell proliferation in gastric cancer. PLoS ONE. 2014;9(1):e85296.

    PubMed  PubMed Central  Google Scholar 

  31. 31.

    Li R, Yang HQ, Xi HL, Feng S, Qin RH. Inhibition of CDH17 gene expression via RNA interference reduces proliferation and apoptosis of human MKN28 gastric cancer cells. Int J Oncol. 2017;50(1):15–22.

    CAS  PubMed  Google Scholar 

  32. 32.

    Qiu HB, Zhang LY, Ren C, Zeng ZL, Wu WJ, Luo HY, et al. Targeting CDH17 suppresses tumor progression in gastric cancer by downregulating Wnt/beta-catenin signaling. PLoS ONE. 2013;8(3):e56959.

    CAS  PubMed  PubMed Central  Google Scholar 

  33. 33.

    Wang J, Kang WM, Yu JC, Liu YQ, Meng QB, Cao ZJ. Cadherin-17 induces tumorigenesis and lymphatic metastasis in gastric cancer through activation of NFkappaB signaling pathway. Cancer Biol Ther. 2013;14(3):262–70.

    CAS  PubMed  PubMed Central  Google Scholar 

  34. 34.

    Lopes N, Bergsland CH, Bjornslett M, Pellinen T, Svindland A, Nesbakken A, et al. Digital image analysis of multiplex fluorescence IHC in colorectal cancer recognizes the prognostic value of CDX2 and its negative correlation with SOX2. Lab Invest. 2020;100(1):120–34.

    CAS  PubMed  Google Scholar 

  35. 35.

    McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM. REporting recommendations for tumour MARKer prognostic studies (REMARK). Eur J Cancer. 2005;41(12):1690–6.

    PubMed  Google Scholar 

  36. 36.

    Team RC. R: A language and environment for statistical computing. 2017. https://www.R-project.org. Accessed Aug 2019.

  37. 37.

    Johnstone CN, White SJ, Tebbutt NC, Clay FJ, Ernst M, Biggs WH, et al. Analysis of the regulation of the A33 antigen gene reveals intestine-specific mechanisms of gene expression. J Biol Chem. 2002;277(37):34531–9.

    CAS  PubMed  Google Scholar 

  38. 38.

    Mao Z, Song S, Zhu Y, Yi X, Zhang H, Shang Y, et al. Transcriptional regulation of A33 antigen expression by gut-enriched Kruppel-like factor. Oncogene. 2003;22(28):4434–43.

    CAS  PubMed  Google Scholar 

  39. 39.

    Qu LP, Zhong YM, Zheng Z, Zhao RX. CDH17 is a downstream effector of HOXA13 in modulating the Wnt/beta-catenin signaling pathway in gastric cancer. Eur Rev Med Pharmacol Sci. 2017;21(6):1234–41.

    PubMed  Google Scholar 

  40. 40.

    Zhu R, Wong KF, Lee NP, Lee KF, Luk JM. HNF1alpha and CDX2 transcriptional factors bind to cadherin-17 (CDH17) gene promoter and modulate its expression in hepatocellular carcinoma. J Cell Biochem. 2010;111(3):618–26.

    CAS  PubMed  Google Scholar 

  41. 41.

    Rageul J, Mottier S, Jarry A, Shah Y, Theoleyre S, Masson D, et al. KLF4-dependent, PPARgamma-induced expression of GPA33 in colon cancer cell lines. Int J Cancer. 2009;125(12):2802–9.

    CAS  PubMed  PubMed Central  Google Scholar 

  42. 42.

    Sakamoto N, Oue N, Sentani K, Anami K, Uraoka N, Naito Y, et al. Liver-intestine cadherin induction by epidermal growth factor receptor is associated with intestinal differentiation of gastric cancer. Cancer Sci. 2012;103(9):1744–50.

    CAS  PubMed  Google Scholar 

  43. 43.

    Kumar N, Tsai YH, Chen L, Zhou A, Banerjee KK, Saxena M, et al. The lineage-specific transcription factor CDX2 navigates dynamic chromatin to control distinct stages of intestine development. Development. 2019;146(5):dev172189.

    PubMed  PubMed Central  Google Scholar 

  44. 44.

    Ryu KH, Shim KN, Jung SA, Yoo K, Joo YH, Lee JH. Significance of preoperative tissue levels of vascular-endothelial cadherin, liver-intestine cadherin and vascular endothelial growth factor in gastric cancer. Korean J Gastroenterol. 2012;60(4):229–41.

    PubMed  Google Scholar 

  45. 45.

    Matsusaka K, Ushiku T, Urabe M, Fukuyo M, Abe H, Ishikawa S, et al. Coupling CDH17 and CLDN18 markers for comprehensive membrane-targeted detection of human gastric cancer. Oncotarget. 2016;7(39):64168–81.

    PubMed  PubMed Central  Google Scholar 

  46. 46.

    Shigematsu Y, Inamura K, Mise Y, Saiura A, Rehnberg E, Yamamoto N, et al. CDX2 expression is concordant between primary colorectal cancer lesions and corresponding liver metastases independent of chemotherapy: a single-center retrospective study in Japan. Oncotarget. 2018;9(24):17056–65.

    PubMed  PubMed Central  Google Scholar 

  47. 47.

    Barros R, da Costa LT, Pinto-de-Sousa J, Duluc I, Freund JN, David L, et al. CDX2 autoregulation in human intestinal metaplasia of the stomach: impact on the stability of the phenotype. Gut. 2011;60(3):290–8.

    CAS  PubMed  Google Scholar 

Download references

Acknowledgements

The authors would like to acknowledge Prof. Fátima Carneiro for access to the Tissue Bank of Centro Hospitalar S. João.

Funding

This work was supported by FEDER—Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020—Operational Programme for Competitiveness and Internationalisation (POCI), Portugal 2020, and by Portuguese funds through FCT—Fundação para a Ciência e a Tecnologia/Ministério da Ciência, Tecnologia e Inovação in the framework of the project "Institute for Research and Innovation in Health Sciences" (POCI-01-0145-FEDER-007274), POCI-01-0145-FEDER-029503 and POCI-01-0145-FEDER-029017. This work was also financed by the project NORTE-07-0124-FEDER-000029, supported by Norte Portugal Regional Programme (NORTE 2020), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF). The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7-PEOPLE-2013-COFUND) under grant agreement no. 609020—Scientia Fellows (to RAL supporting NL as a post doc), the South-Eastern Norway Regional Health Authority (project number 2016123 to RAL supporting CHB as a PhD fellow), and the Research Council of Norway & University of Oslo (“Toppforsk”, project number 250993 to RAL).

Author information

Affiliations

Authors

Contributions

All authors participated in the execution or analysis of this study, critically revised the manuscript, and approved its final version. Formal analysis and investigation: NL, CB, JB, MB, AFV, PM, RP, BC, and EB; Methodology: NL and JB; Resources: RG; Writing of the original draft: NL; Funding acquisition: RAL, RA, and LD; Supervision: LP, RAL, RA, and LD.

Corresponding author

Correspondence to Leonor David.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

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. Informed consent to be included in the study, or the equivalent, was obtained from all patients.

Informed consent

The use of retrospective samples from which informed consent cannot be obtained is authorised for research studies by the Portuguese law. Recent samples included in the study had written informed consent. The study was approved by Centro Hospitalar São João ethics committee under the reference number CES 122-15.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 61 kb)

Supplementary file2 Supplementary Figure S1 – Flowchart of the gastric cancer cases included in the study (JPEG 348 kb)

10120_2020_1064_MOESM3_ESM.jpg

Supplementary file3 Supplementary Figure S2 – Comparison of CDX2 staining by chromogenic IHC in gastric cancer, using citrate buffer (a) or EDTA buffer (b) as the antigen retrieval solution; magnification 200x (JPEG 929 kb)

10120_2020_1064_MOESM4_ESM.jpg

Supplementary file4 Supplementary Figure S3 – Heterogeneity in the expression of CDX2, GPA33 and LI-cadherin in gastric cancer. There are no significant differences between two cores (s1 and s2) from the same tumour (a-c) or between primary tumours (pt) and the corresponding lymph nodes (ln) (d-f), for all the markers analysed. s1 – Sample one; s2 – Sample two. Data obtained using fluorescence-based mIHC (JPEG 396 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lopes, N., Bergsland, C., Bruun, J. et al. A panel of intestinal differentiation markers (CDX2, GPA33, and LI-cadherin) identifies gastric cancer patients with favourable prognosis. Gastric Cancer 23, 811–823 (2020). https://doi.org/10.1007/s10120-020-01064-6

Download citation

Keywords

  • Gastric cancer
  • Intestinal differentiation markers
  • Prognostic markers