Transduced caudal-type homeobox (CDX) 2/CDX1 can induce growth inhibition on CDX-deficient gastric cancer by rapid intestinal differentiation

2.1 Kaplan-Meier plot analyses

Publicly available KM plotter23 and TCGA (The Cancer Genome Atlas) data set at the cBioPortal24, 25 were used to plot disease-free survival curves and overall survival curves of gastric cancer patients. Regarding the TCGA data, the patients were divided into two groups based on the level of CDX2 and CDX1 expression which was shown as Z-scores (>0.65 for “CDX high” group and ≤0.65 for “CDX low” group). According to the log-rank test, P values <0.05 were considered statistically significant.

2.2 Cell culture, retrovirus vectors, and cell proliferation assay

For the stable transduction of CDX genes, we used VSVG-pseudotyped pMXs-IRES-puro retrovirus vectors.10 To evaluate cell proliferation, we used MTT assay. Detailed information of cell lines used and cell-related experimental procedures are described in Doc S1.

2.3 Western blot analysis and reverse transcriptase-PCR analysis

Western blotting and RT-PCR were carried out as we previously reported.26 Antibodies used and primer sequences of 16 gene transcripts are described with detailed experimental procedures in Doc S1.

2.4 Tumor samples and immunohistochemistry

We randomly selected 89 gastric adenocarcinoma samples surgically resected at the Fujita Health University Hospital. This study was approved by the ethics committees of the University of Tokyo and the institutional ethical review board for human investigation at Fujita Health University. Immunohistochemistry to examine expression of CDX2 and CDX1 is described in detail in Doc S1.

2.5 Cell cycle analysis

MKN7 and TMK1 cells were grown semi-confluent in 100-mm culture dishes and infected with CDX-expressing retrovirus vectors (CDX2, CDX1, and mock) at appropriate MOI. After confirming the morphological changes in both cells infected with CDX expression vectors (6 days post-infection for MKN7 and 9 days post-infection for TMK1), staining with propidium iodide was carried out (using Cycletest plus DNA reagent kit; Becton Dickinson Immunocytometry Systems, San Jose, CA, USA). Cell cycle was then analyzed using FACS Calibur flow cytometer (Becton Dickinson Immunocytometry Systems) and ModFit LT version 3.0 software (Verity Software House, Topsham, ME, USA).

2.6 Microarray gene expression analysis and selection of gene probes for CDX2/CDX1 common signatures, CDX2 signature, and CDX1 signature

We examined total RNA obtained from MKN7 cells at 6 days post-infection and from TMK1 cells at 8 days post-infection with CDX-expressing retrovirus vectors. After extraction with RNeasy Plus Mini Kit (Qiagen, Venlo, the Netherlands), total RNA was analyzed on GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's protocols. After normalization of the raw data, fold changes in gene expression in CDX-transduced cells were calculated relative to the mock-infected control cells. Raw data are available in NCBI's Gene Expression Omnibus (GEO; series accession number is GSE102208).

To decide signature gene sets for CDX2/CDX1 combined, CDX2, and CDX1 expression, genes were selected based on more than threefold upregulation or downregulation compared with mock-infected control.

2.7 Latest comprehensive gene expression data of 132 gastric tissues

Latest comprehensive gene expression data of 111 gastric cancer tissues and 21 noncancerous gastric tissues were obtained from GSE54129 from NCBI's GEO.

3.1 Gastric cancer patients with a higher level of CDX expression have significantly better clinical outcomes

Using the data set from KM plotter23 and TCGA,24 clinical outcomes of gastric cancer patients were analyzed focusing on the expression of CDX2 and CDX1 (Figure 1). According to the data of KM plotter,23 overall survival curves and disease-free survival curves showed that gastric cancer patients with higher expression of CDX have significantly longer overall survival and disease-free periods compared to those without (Figure 1A,B). According to the data of TCGA,24 disease-free survival curves also showed that gastric cancer patients with higher expression of CDX have significantly longer disease-free periods (Figure 1D). For overall survival curves based on TCGA data (Figure 1C), gastric cancer patients with a high level of CDX expression have better survival, although we could not detect a significant association between overall survival and CDX expression. In total, higher expression of CDX in gastric cancer generally shows a tendency to be associated with better prognosis.

3.2 Exogenous expression of CDX2 and CDX1 induces dramatic morphological change and severe growth inhibition in CDX-deficient gastric cancer cells

To evaluate the effect of CDX transcription factors on gastric cancer cells, we introduced human CDX2 and CDX1 genes separately into MKN7 and TMK1 cells which are both deficient in CDX expression.10, 22 After infection with retrovirus vectors and selection with puromycin, we discovered that transduction of either CDX2 or CDX1 gave rise to dramatic morphological change on these CDX-deficient gastric cancer cells (Figure 2A). At 7 days post-infection, both MKN7 and TMK1 showed a flatter morphology accompanied by marked suppression of cell growth. Transformation of MKN7 occurred earlier and more drastically compared with that of TMK1. Namely, we observed complete growth arrest with no surviving cells for CDX-transduced MKN7 cells. On the contrary, a very small fraction of TMK1 cells survived 10 or more days after transduction of CDX, but the growth speed of surviving cells was much slower than for mock-infected cells. Such CDX-induced arrest or severe delay of cell growth was validated by MTT assay (Figure 2B), in which the effects of CDX2 and CDX1 were almost equivalent in both cell lines. Such suppressive effects on cell growth could not be detected in CDX-transduced MKN74 (Figure 2B), AGS, SW480, and LoVo cells (data not shown), all of which originally express CDX.10

3.3 CDX-induced growth inhibition in gastric cancer cells occurs through G0-G1 cell cycle arrest

To elucidate the growth suppression in CDX-transduced MKN7 and TMK1, cell cycle analysis was carried out by flow cytometry and western blotting. As shown in Figure 2C, exogenous expression of CDX2 and CDX1 caused a substantial decrease in S-phase populations and a concomitant increase in G0/G1-phase populations, indicating that the transduced CDX gene led to G0-G1 growth arrest in both cell lines. Western blot analyses of cell cycle-related proteins showed that expression of cyclin E2, CDK4 and p27/Kip1 was markedly decreased in both CDX-transduced cell lines (Figure 2D). In contrast, reduction of cyclin D3, CDK2 and p21/Waf1/Cip1 was observed in MKN7 only, and reduction of cyclin A2 and cyclin B1 was observed in TMK1 only (Figure 2D). Although there was some difference between the two CDX-deficient gastric cancer cell lines, expression of several cell cycle-related proteins was mostly decreased after exogenous expression of CDX. This was probably as a result of suppressed intracellular metabolism and reduced protein production, which simultaneously occurred with G0/G1 growth inhibition. We speculate that these are the reasons why p27 in CDX-transduced MKN7/TMK1 and p21 in CDX-transduced MKN7 were reduced, both of which usually increase along with G0/G1 growth arrest.

We further carried out a senescence-associated β-galactosidase assay (Figure S1) and two independent assays to detect apoptosis (cleaved caspase-3 immunohistochemistry in Figure S2A and TUNEL assays in Figure S2B), but we could not detect any difference between the CDX-transduced MKN7/TMK1 and the mock-infected control cells. It is indicated that severe growth suppression observed in CDX-transduced MKN7 and TMK1 was neither due to apoptosis nor rapid senescence.

3.4 Differentiation status is considerably changed in CDX-deficient gastric cancer cells after transduction with CDX2/CDX1 expression vector

Reverse transcription-PCR was next carried out to evaluate transcription of several genes related to intestinal differentiation.12, 27-30 As shown in Figure 2E, LI-cadherin and KRT20 (cytokeratin 20) transcripts were strongly upregulated in CDX-transduced MKN7 and TMK1. Induction of these typical intestinal marker genes indicates that exogenous expression of CDX2 and CDX1 can lead to intestinal differentiation in gastric cancer cells. Furthermore, we also found that other known downstream genes of CDX such as HEPH,27 MDR1,29 and DCS2 (desmocollin2)30 were similarly transactivated by exogenous expression of CDX2 and/or CDX1 (Figure 2E). Although we could not detect activation of MUC2 transcription,31 we concluded that retrovirally transduced CDX2 and CDX1 certainly functioned as transcription factors12, 14, 28 and induced intestinal differentiation through transactivation of gut-associated genes.

We further carried out RT-PCR to evaluate expression of genes related to stemness and gastric differentiation. Expression of LGR5 and DCLK1 (DCAMKL1), both of which are well-known marker genes of intestinal stemness,32-34 was induced in CDX-transduced MKN7 cells (Figure 2F). Although such induction was not detected in CDX-transduced TMK1 cells, possible transactivation of intestinal stemness-related genes by CDX should be worthy of note. Furthermore, we found that expression of several stemness-related genes such as Sall4,35 Satb2,36 and Satb136 was upregulated in response to exogenous expression of CDX (Figure 2F), although the meaning of induced stemness in gastric cancer cells remains unclear. Regarding expression of gastric marker genes, TFF2/SP (trefoil factor 2)37, 38 was downregulated in MKN7 cells, and SPARC (osteonectin)39, 40 was downregulated in TMK1 cells. Changed status of differentiation in CDX-transduced gastric cancer cells must have some association with the observed severe G0/G1 growth arrest in them.

To investigate the altered gene expression accompanied by CDX-induced growth arrest more broadly, mRNA microarray analyses using CDX-transduced MKN7 and TMK1 cells were carried out (Tables S1 and S2 for upregulated genes; Tables S3 and S4 for downregulated genes). From the view of gut differentiation, microarray data expectedly showed that exogenous CDX induced several intestinal marker genes such as KRT20,28 DSC2,30 TSAPN8,41 CEACAM6,42 and MME (CD10)43 (Tables S1 and S2). Microarray data also showed that several gastric marker genes such as TFF2/SP,37, 38 TFF1/pS2,44 KCNJ15,45 SPARC,39, 40 and CDA (cytidine deaminase)46 were contrastively downregulated (Tables S3 and S4). These results reinforced our hypothesis that rapidly induced “intestinal differentiation” accompanied by weakened “gastric differentiation” is a critical mechanism underlying the G0-G1 growth arrest of CDX-transduced MKN7 and TMK1.

3.5 Signature genes related to CDX expression based on our transcriptome analysis clearly categorize gastric tissues into non-cancerous mucosa, “CDX signature”-positive cancer, and “CDX signature”-negative cancer

To examine the meaning of CDX expression in gastric cancer, the signatures related to CDX2/CDX1 expression were analyzed using the latest comprehensive gene expression data of 132 gastric tissues (111 gastric cancer tissues and 21 noncancerous gastric tissues). Using our microarray gene expression data of MKN7 and TMK1 with exogenous CDX2 and CDX1 expression, 18 common signature genes were selected based on more than threefold upregulation or downregulation (common for CDX2 and CDX1) compared with mock-infected control. The 18 selected signature genes for CDX2/CDX1 expression were then applied to the gene expression data of a total of 132 total gastric cancerous/non-cancerous tissues (Figure 3A). Hierarchical clustering showed that this “CDX2/CDX1 (common) signature” can clearly distinguish gastric cancer from non-cancerous gastric mucosa. In addition, 111 gastric cancer lesions were plainly categorized into two categories: positive for “CDX2/CDX1 signature” and negative for “CDX2/CDX1 signature.” The gene expression pattern of non-cancerous gastric mucosa is obviously close to that of “CDX2/CDX1 signature”-positive gastric cancer rather than “CDX2/CDX1 signature”-negative” gastric cancer (Figure 3A).

We additionally selected 112 signature genes for CDX2 expression and 118 signature genes for CDX1 expression based on more than threefold up/downregulation in CDX-transduced MKN7 and TMK1 cell lines and carried out hierarchical clustering in the same way (Figures S3 and S4). Similar to the “CDX2/CDX1 (common) signature” (Figure 3A), both “CDX2 signature” and “CDX1 signature” can clearly distinguish gastric cancer from non-cancerous gastric mucosa and can also clearly categorize the 111 gastric cancer cases into “CDX signature”-positive and “CDX signature”-negative tumors.

3.6 Gene set enrichment analysis indicates a significant association between cancer-associated properties and expression status of CDX in gastric cancer

We further carried out gene set enrichment analysis (GSEA) to understand the molecular background of “CDX signature” in gastric cancer.47 For “CDX2/CDX1 signature”-positive gastric cancer (left panel of Figure 3B,C), we identified several pathways related to tumor-suppressive function including p53_PATHWAY, KRAS_SIGNALING_DN, G2M_CHECKPOINT, and DNA_REPAIR. For “CDX2/CDX1 signature”-negative gastric cancer (right panel of Figure 3B,D), we conversely identified several oncogenic pathways such as WNT_BETA_CATENIN_SIGNALING, IL2_STAT5_SIGNALING, EPITHELIAL_MESENCHYMAL_TRANSITION, ANGIOGENESIS, NOTCH_SIGNALING, TGF_BETA_SIGNALING, IL6_JAK_STAT3_SIGNALING, and KRAS_SIGNALING_UP. These results could partly explain the meaning of CDX2/CDX1 expression in gastric cancer, which was suggested by hierarchical clustering analyses (Figure 3A, Figures S3 and S4). In short, CDX-positive gastric cancer tends to have fewer malignant features, and CDX-negative gastric cancer conversely tends to have more malignant features.

3.7 Approximately half of gastric cancer lesions surgically resected neither express CDX2 nor CDX1

To examine expression of CDX in gastric cancer, immunohistochemistry was carried out using 89 randomly selected gastric cancer specimens surgically resected. As shown in Figure 4A, 50.6% of gastric cancer lesions did not express CDX2, and 66.3% of them did not express CDX1. Concerning CDX2, our result was similar to the previous report showing that about half of gastric cancer tissues lack CDX2 expression.20 Expression of CDX1 has not been adequately evaluated so far, but our result indicated that the rate of CDX1-deficient gastric cancer is similarly high (Figure 4A, Table S5). Our immunohistochemical analysis also showed that decrease or absence of CDX expression was observed not only in intestinal-type gastric cancer but also in diffuse-type gastric cancer (Figure 4A, Table S5).

Our data also showed that expression of CDX2 and CDX1 tends to be concomitantly reduced in gastric cancer lesions regardless of histological type (Figure 4B). Of the total 89 gastric cancer cases, as many as 40 lesions (44.9%) expressed neither CDX2 nor CDX1. Correlation coefficients of CDX2 and CDX1 expression were rather high (Figure 4B; all P values <0.001), which indicated that coexpression or codeficiency of CDX2 and CDX1 is frequently observed in gastric cancer. The high prevalence of CDX2/CDX1 double-negative tumors suggested that induced intestinal differentiation may be a novel strategy against gastric cancer in the future.

3.8 Clinicopathological and molecular features of gastric cancer focusing on CDX2/CDX1 expression

Using the TCGA data set, we further analyzed CDX2-positive (N = 55), CDX2-negative (N = 175), CDX1-positive (N = 29), and CDX1-negative (N = 201) gastric cancer patients. As shown in Figure 4D, we analyzed associations between the expression status of CDX and 24 various variables and found four clinicopathological/molecular features significantly associated with expression of CDX in gastric cancer.

First, Epstein-Barr virus (EBV)-positivity is negatively associated with CDX-positivity with statistical significance, probably because CDX-positive gastric cancer mostly originated from H. pylori-induced chronic gastritis. Second, DNA hypermethylation status is frequently observed in CDX-positive gastric cancer compared with CDX-negative cancer, probably reflecting the fact that chronic infection of H. pylori induces intestinal metaplasia of gastric mucosa with high levels of aberrant DNA methylation.48 Third, microRNA expression clusters denote significantly different patterns between CDX-positive gastric cancer and CDX-negative gastric cancer, possibly reflecting the transcription of miRNAs regulated by CDX2/CDX1. And, finally, lymphocyte infiltration was more frequently observed in CDX-negative gastric cancer, probably reflecting the abundant lymphocyte infiltration of CDX-negative and EBV-associated gastric cancer.25