The Adenocarcinoma-Associated Antigen, AGR2, Promotes Tumor Growth, Cell Migration, and Cellular Transformation

Introduction

AGR2 is a secreted protein initially described in Xenopus laevis, from which it was identified in a screen for differentially expressed genes in neural development. AGR2 serves an essential role in Xenopus development by inducing the formation of the forebrain and the mucus-secreting cement gland ( 1). In humans, AGR2 was first identified in studies focused on differentially expressed genes in estrogen receptor–positive breast cancers ( 2). Subsequent studies showed elevated AGR2 expression in adenocarcinomas of the esophagus, pancreas, and prostate ( 2– 8). The clinical effect of AGR2 expression in tumors is unclear as the implications for prognosis are mixed in breast and prostate cancer ( 7, 9, 10). Evidence that AGR2 may influence tumor biology stems from studies in which overexpression in a rat nonmetastatic breast tumor cell line were associated with increased metastasis when propagated as xenografts in nude mice ( 11).

AGR2 expression in Barrett's esophagus, a premalignant lesion characterized by intestinal metaplasia, is elevated >70-fold compared with normal esophageal epithelia. Esophageal AGR2 expression alone is sufficient to distinguish Barrett's esophagus from normal esophageal epithelia ( 8). Barrett's esophagus increases the risk of developing esophageal adenocarcinoma by 30-fold ( 12).

AGR2 was chosen for further investigation for several reasons. Its universal expression in all premalignant and malignant esophageal adenocarcinomas suggests that it serves an important role in disease pathogenesis. Second, multiple highly conserved genes important in development, such as those belonging to the Wnt and Hedgehog pathways, have been found to significantly influence tumor development ( 13– 15). Because AGR2 is established as an essential gene in Xenopus development and is expressed in human tumors, we chose to further explore its potential role in esophageal cancer using well-established assays in tumor biology.

Materials and Methods

Cell lines and antibodies. SEG-1 ( 16) cells (a generous gift from Dr. David G. Beer, University of Michigan, Ann Arbor, MI) were grown in 5% CO₂ in DMEM supplemented with 4.5 g/L glucose, l-glutamine (Cellgro, Mediatech, Inc.), penicillin (100 units/mL), streptomycin (100 units/mL), and 10% (v/v) fetal bovine serum. NIH3T3 cells were grown in the same culture conditions.

RNA interference. RNA interference was achieved using microRNA-adapted short hairpin RNA (shRNA) as previously described ( 17). Three constructs were employed that included the following AGR2 sequences (underlined) with intervening sequences representing short hairpins: KD1, CTGATTAGGTTATGGTTTAATAGTGAAGCCACAGATGTATTAAACCATAACCTAATCAG; KD2, CCCACACAGTCAAGCTTTAATAGTGAAGCCACAGATGTATTAAAGCTTGACTGTGTGGG; and KD3, CAACAAACCCTTGATGATTATAGTGAAGCCACAGATGTATAATCATCAAGGGTTTGTTG. Each construct was subcloned into the MSCV-LTRmiR30-PIG retroviral vector, which was used to transduce SEG-1 cells according to the manufacturer's protocol (OpenBiosystems, Inc.). SEG-1 control cells were transduced with only the viral vector without the shRNA. Successful incorporation of the retroviral vector was confirmed with the expression of green fluorescent protein that was contained within the vector in cis. Transduced cells were selected with 2 μg/mL of puromycin (MP Biomedicals, Inc.) for a period of 2 weeks. The resultant cells for each construct were assessed for AGR2 expression with protein immunoblotting and quantified with conjugated second antibodies that emit in the IR spectrum (Odyssey Infrared Imaging Systems; LiCor Biosciences).

NIH3T3 cell transfection. NIH3T3 cells were cotransfected with the plasmids pCMV-SPORT6.0 containing the full-length AGR2 sequence (Openbiosystems, Inc.) and pcDNA3.1-GFP that carries a neo selective marker. The transfection was performed using Fugene 6.0 (Roche Diagnostics) followed by drug selection with G418 at 0.8 mg/mL.

Xenograft tumor model. Ten-week-old male BALB/c nude mice were obtained from Taconic Farms, Inc.; 1 × 10⁶ cells were injected s.c. The resultant tumors were measured with calipers (Westward, Grainger International, Inc.) and the volume was calculated using the formula (length × width² / 2) as previously described ( 18). Animals were sacrificed when the maximal allowable tumor size was achieved or after observation for 21 days.

Microscopy. Immunohistochemistry was performed using Dako Cytomation Envision Plus following the manufacturer's instructions. Tissue sections were obtained from formaldehyde-fixed paraffin-embedded mouse intestine. Rabbit anti-AGR2 antibodies (Imgenex Corp.) were incubated overnight at room temperature at a dilution of 1:100 in PBS. Rabbit anti-chromogranin A antibodies (ImmunoStar, Inc.) were incubated overnight at room temperature at a dilution of 1:200 in PBS and 0.3% (v/v) Triton X-100. Rat anti-mouse anti-Ki67 (Dako Cytomation, Inc.) was incubated for 1 h at room temperature at a dilution of 1:25 in PBS. Rat anti-mouse musashi-1 monoclonal antibodies were a kind gift of Dr. H. Okano (Keio University, Tokyo, Japan) and were incubated at a dilution of 1:500 for 16 h at 4°C ( 19). Fluorescent images were visualized with a Nikon Eclipse E600 microscope equipped with fluorescence filters for Texas red, FITC, and 4′,6-diamidino-2-phenylindole. A monochrome image was acquired with a Spot RT Slider digital camera (Diagnostic Instruments, Inc.) for each color channel and merged with Adobe Photoshop 7.0 (Adobe Systems, Inc.).

Migration assay. The cell migration assay used to assess metastatic potential was performed using Matrigel-coated transwells according to the manufacturer's protocols (BD BioCoat growth factor reduced Matrigel invasion chamber, BD Biosciences; ref. 20); 1 × 10⁵ cells were plated in each well. The inside chamber within the transwell was incubated with 500 μL of unconditioned cell culture media containing DMEM supplemented with 4.5 g/L of glucose, l-glutamine (Cellgro, Mediatech, Inc.), and 1% (w/v) bovine serum albumin (Sigma-Aldrich). The bottom chamber outside the transwell was filled with 750 μL of DMEM supplemented with 4.5 g/L of glucose, l-glutamine medium, and 10% (v/v) FCS that was conditioned for 24 h with either SEG-1:KD1 cells or SEG-1 control cells. The cultures were incubated at 37°C in 5% CO₂ for 18 h, followed by fixation in 100% methanol for 15 min and staining in 0.2% (w/v) crystal violet, and 2% (v/v) ethanol for 15 min. The colony number was quantified by counting directly with a microscope.

Cell proliferation. Cell proliferation was determined by plating 10⁵ cells per 6 cm dish in duplicate for each time point and at serum concentrations from 2% to 10%. Every 2 days following the initial plating, the cells were harvested and manually counted using a hemocytometer.

Miscellaneous methods. The assays for focus-formation and anchorage-independent growth were performed as previously described ( 21). Culture plates containing foci or colonies in soft agar were stained with crystal violet, scanned on a flatbed scanner, and quantified using ImageJ. ⁴

Real-time PCR used SYBR Green for quantification (Bio-Rad Laboratories, Inc.). PCR primer sequences used for AGR2 included 5′-ATGAGTGCCCACACAGTCAA-3′ and 5′-GGACATACTGGCCATCAGGA-3′; for β-actin, 5′-CGGGAAATCGTGCGTGACATTAAG-3′ and 5′-TGATCTCCTTCTGCATCCTGTCGG-3′; and for GFI1, 5′-TCCACACTGTCCACACACCT-3′ and 5′-CTGGCACTTGTGAGGCTTCT-3′.

Additional analysis of ATOH1 and GFI1 gene expression from the data set in ref. 22 was explored using the Gene Expression Omnibus at the GEO web site ⁵ (accession no. GDS1321).

Results

AGR2 RNA interference in SEG-1 cells. We first explored AGR2 function in esophageal cancer by reducing its expression in the esophageal adenocarcinoma cell line, SEG-1 ( 16). Stable AGR2-deficient SEG-1:KD1 (knockdown) cells were produced by transformation with retrovirus containing microRNA-adapted shRNAs that interfere with AGR2 expression through RNA interference. Real-time PCR showed that AGR2 mRNA levels decreased by 85% in the SEG-1:KD1 cells. An associated decrease in protein expression was confirmed with immunoblotting ( Fig. 1A ).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Effects of AGR2 suppression in SEG-1 cells. A, AGR2 expression in SEG-1 cells suppressed with shRNAmir. SEG-1 cells were transduced with shRNAmir retroviral constructs (KD1, KD2, KD3) followed by selection with puromycin. SEG-1 control cells were transduced with the retroviral vector alone (Vector). The chart displays the cell lines assayed by real-time PCR for AGR2 mRNA normalized to β-actin mRNA; and compared with SEG-1 control cells. Inset, AGR2 protein expression assayed with protein immunoblotting. All remaining experiments used the SEG-1:KD1 cells. B, anchorage-independent growth as measured by colony growth in soft agar at different initial plating densities. SEG-1:KD1 cells showed an 82% reduction in colonies with the plating of 1,000 cells. Columns, mean; bars, 1 SD. C, SEG-1 control cells (♦) and SEG-1:KD1 cells were grown as tumor xenografts in nude mice (▪). Points, mean from five mice; bars, 1 SD. D, migration assay. SEG-1:KD1 cells were grown in filter chambers bathed in conditioned culture media derived from either SEG-1 control cells (AGR2+) or SEG-1:KD1 cells (AGR2−). Columns, mean fold change from three experiments performed in duplicate and normalized to the experiments using AGR2-deficient conditioned media; bars, 1 SE. The number of cells which migrated for the AGR2+ media ranged from 133 to 264 cells.

In vitro and in vivo assays often used to characterize malignantly transformed cells were employed to assess the effects of reducing AGR2 expression in SEG-1 cells. Anchorage-independent growth in soft agar is an in vitro characteristic displayed by many transformed cells, including wild-type SEG-1 cells ( 21, 23). SEG-1:KD1 cells with reduced AGR2 expression formed 82% fewer colonies in soft agar than SEG-1 control cells ( Fig. 1B). The in vivo assay consisted of the same cells implanted s.c. in BALB/c nude mice and grown as tumor xenografts. Twenty-one days after implantation, tumors formed by AGR2-deficient cells were on average 60% smaller than tumors established from SEG-1 control cells ( Fig. 1C). Thus, reduction in AGR2 expression compromised SEG-1 growth in two assays of malignant transformation.

AGR2 possesses a signal peptide and is secreted from cells ( 1, 24). We explored whether secreted AGR2 could affect cells in an in vitro migration assay in which cells are induced to move across a filter support in response to components added to the bathing culture media. Conditioned media derived from SEG-1 cells that express high levels of AGR2 enhanced SEG-1:KD1 cell migration at 2.7-fold higher rates than conditioned media from AGR2-deficient SEG-1:KD1 cells ( Fig. 1D). The results support a model in which SEG-1 cells are able to respond to secreted AGR2, which results in increased migration.

Effects of AGR2 expression in NIH3T3 cells. In view of AGR2's effects on SEG-1 cells, additional studies were performed to examine its role in tumorigenesis. Transfected NIH3T3 cells are commonly used in in vitro and in vivo assays to assess whether specific genes could promote cellular transformation ( 21, 23). The in vitro assays include measurements of foci formation, which reflect the loss of density-dependent growth, or anchorage-independent growth in soft agar as previously described. Transfected NIH3T3 cells that express AGR2 show significantly enhanced foci formation, as multiple foci are formed with AGR2 expression compared with almost no foci formed by the controls ( Fig. 2A ). Anchorage-independent growth, as assessed by colony growth in soft agar, resulted in an up to 7-fold increase in colonies for NIH3T3:AGR2 cells compared with NIH3T3 control cells ( Fig. 2B). Finally, NIH3T3 cells were also used in an in vivo assay in which xenografts were evaluated for their ability to grow as tumors in nude mice. AGR2-expressing NIH3T3:AGR2 cells grown as xenografts were 15-fold larger in size 21 days after implantation compared with controls ( Fig. 2C and D). Therefore, in vitro and in vivo assays with NIH3T3 cells implicate AGR2 in cellular transformation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Effects of AGR2 expression in NIH3T3 cells. A, focus-formation assay of AGR2-expressing NIH3T3 cells or control cells transfected with the expression vector alone. Top, images of the control cells (left) and a foci of NIH3T3 AGR2 cells (right). Bottom, chart depicting one of three representative experiments. B, anchorage-independent growth as measured by colony growth in soft agar after different initial plating densities of NIH3T3:AGR2 cells or control NIH3T3 cells. Columns, mean from one of two experiments performed in triplicate; bars, 1 SD. C, representative example of NIH3T3 control cells (right) and NIH3T3:AGR2 cells (left) grown as xenografts in BALB/c nude mice. Implantation site of the cells (dotted lines). D, chart depicting the growth of NIH3T3 control cells (♦) and NIH3T3:AGR2 expressing cells (▪) grown as tumor xenografts. Points, mean of five mice; bars, 1 SD.

Expression of AGR2 in the normal mouse intestine. We explored AGR2 expression in the normal mouse small intestine because Barrett's esophagus is characterized by intestinal metaplasia, which significantly increases the risk of developing adenocarcinoma. In addition, knowledge available with respect to intestinal development and cell growth may be capitalized to provide insights into AGR2's function in normal gastrointestinal cells and tumors ( 25, 26). The small intestine is composed of four cell types, absorptive enterocytes, goblet cells, Paneth cells, and enteroendocrine cells; with the latter three classified together as secretory cells. We performed immunohistochemistry on normal mouse intestine and detected AGR2 protein in small intestinal crypts and villi ( Fig. 3A ). The majority of cells in the intestinal villus, represented by absorptive enterocytes, did not label for AGR2. Granule-containing Paneth cells located in the base of intestinal crypts showed strong AGR2 labeling ( Fig. 3A, G–J). AGR2 staining was also observed in all cells that expressed chromogranin A (CHGA), a marker for enteroendocrine cells ( Fig. 3B–C). Although AGR2 staining was strongest for enteroendocrine cells in the intestinal villus, other positive-staining cells were also observed and found to stain with Alcian blue, which identifies mucus-containing goblet cells ( Fig. 3A). The ratio of AGR2-staining enteroendocrine cells to goblet cells was ∼1:8, which is consistent with the previously described ratio for these cells in the mouse ( 27). Thus, AGR2 is expressed in all three intestinal secretory cell lineages.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Immunohistochemistry of mouse small intestine. A, small intestinal section in which the goblet cells are labeled with Alcian blue (black arrows) and AGR2 is labeled with horseradish peroxidase stain (brown). Inset, Paneth cells at the bottom of a crypt that stain intensely for AGR2. Arrows, representative cells that stain for Alcian blue. B and C, immunofluorescence of adjacent serial intestinal sections stained for chromogranin A (CHGA) antibodies (B, green) or anti-AGR2 antibodies (C, green). Nuclei in both sections were stained with 4′,6-diamidino-2-phenylindole (blue). Cells that label for both chromogranin A and AGR2 (white arrows). D to F, triple labeling of an intestinal crypt for MSI1 (E and F, green), AGR2 (D and E, red), and nuclei (D, E, and F, blue). Arrows, cells labeled for AGR2 and MSI1. G to J, a single intestinal crypt labeled for AGR2 (G–J, red), the proliferation marker Ki-67 (H and I, green), and nuclei (I and J, blue). G, phase contrast image of the crypt with overlying AGR2 immunofluorescence. White arrows, cells that label for the nuclear antigen Ki-67 and AGR2, which labels outside the nucleus. Arrowheads, representative Paneth cells at the bottom of the crypt.

Because AGR2 expression is enhanced in tumors, we explored whether it is expressed in proliferating cells in normal intestinal crypts. Musashi-1 (MSI1) is a RNA-binding protein that is expressed by intestinal stem cells or early progenitor cells ( 28). Coexpression of MSI1 and AGR2 was observed in the intestinal crypt ( Fig. 3D–F). MSI1 expression in the normal small intestine is rare, therefore, additional studies were performed with the proliferation-related Ki-67 antigen (MKI67; Fig. 3G–J). Consistent with previous studies, nuclear Ki-67 staining is restricted to intestinal crypts and excludes Paneth cells (ref. 29; Fig. 3H). AGR2 staining was observed in a small fraction of Ki-67–positive cells that often displayed intracellular inclusions characteristic of goblet cells ( Fig. 3H–J). The low number of Ki-67–positive cells that also stain for AGR2 is consistent with the previously published low proportion of proliferating goblet (6%) and enteroendocrine (0.6%) cells present in the mouse intestinal crypt ( 27). The data shows AGR2 expression by proliferating and nonproliferating intestinal cells of secretory lineage.

Effects of AGR2 expression on GFI1 expression. GFI1 is a basic helix-loop-helix transcription factor whose expression is important in cell lineage determination among intestinal secretory cells. GFI1^−/− mice exhibit increased numbers of enteroendocrine cells, a decline in intestinal goblet cells, and the absence of Paneth cells. GFI1 expression influences cell fate but does not influence cell proliferation ( 30). Enhanced GFI1 expression, however, is associated with tumor development ( 31, 32). Because AGR2 and GFI1 expression is dependent on ATOH1 expression (see Discussion), real-time PCR of SEG1:KD1 cells was performed to evaluate whether changes in AGR2 expression influences that of GFI1. Quantitative real-time PCR revealed no significant changes in GFI1 mRNA levels between SEG-1:KD1 and control SEG-1 cells ( Fig. 4A ).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Relationship of AGR2 expression to other intestinal cell fate determinants. A, real-time PCR of AGR2 and GFI1 mRNA in SEG1:KD1 and SEG-1 control cells. AGR2 and GFI1 mRNA levels were normalized to β-actin and represented as the fold-change to SEG-1 control cells. Columns, mean value of samples measured in triplicate (mean Ct values in SEG-1 control cells for AGR2 = 13.6 and GFI1 = 27). B, schematic of expression patterns for AGR2 in the context of basic helix-loop-helix transcription factors known to influence intestinal cell fate (derived from refs. 30, 33, 34, 38, 39). Whether each secretory lineage is derived from a dedicated progenitor cell remains to be determined. ISC, intestinal stem cell. Gene names used are those approved by the HUGO Gene Nomenclature Committee.

Discussion

We explored the functional effect of AGR2 expression on tumor growth because it is highly and universally expressed in premalignant Barrett's epithelium and esophageal adenocarcinomas. In addition, enhanced AGR2 expression has also been observed in numerous other common adenocarcinomas, including those derived from the pancreas, breast, and prostate. AGR2 is highly conserved and its essential role in X. laevis development enhanced its candidacy for further exploration because of the important role other developmental genes have served in tumor biology.

Previous work indicated that AGR2 enhances tumor metastasis based on the metastatic spread of xenografts derived from a mammary tumor cell line transfected to express AGR2 ( 7). Another study determined that AGR2 serves to attenuate the p53 response to cell stress ( 3). In these studies, both p53 and AGR2 were expressed in a lung cancer cell line that normally does not express either gene. In the present study, we chose to assess AGR2's role in tumorigenesis using two model systems, SEG-1 cells and NIH3T3 cells. SEG-1 cells provided an opportunity to perform studies in a cell line that expresses high levels of AGR2 protein and shows strong similarities in its overall gene expression profile to esophageal adenocarcinoma ( 8). Reduction of AGR2 expression with RNA interference permitted the assessment of its function in a cell line that naturally expresses the protein. NIH3T3 cells provided a second approach and was used because of the long-standing utility the cells have shown in characterizing genes capable of inducing cellular transformation ( 21, 23).

Both cell lines were employed in well-established in vitro and in vivo assays used to identify genes with transforming potential. We established that AGR2 compromises SEG-1 anchorage-independent growth in vitro and the growth of xenograft tumors in vivo. Use of the NIH3T3 cells carries the advantage that genes capable of cellular transformation often result in a clear change in phenotype. As shown in this study, NIH3T3 cells expressing AGR2 clearly displayed significant changes in density- and anchorage-dependent growth in vitro, and as tumor xenografts in vivo. Thus, the results obtained with the SEG-1 and NIH3T3 cells support a significant role for AGR2 in tumor growth.

Well-established correlations between anatomic location with levels of proliferation and differentiation in the normal intestine provided an opportunity to obtain insights into AGR2 function in the adult mouse. Our immunohistochemistry of AGR2 expression in the mouse intestine is identical to the previously described expression pattern for ATOH1 ( 33, 34). ATOH1 (also known as MATH1 or HATH1) encodes for a basic helix-loop-helix transcription factor that is required for the development of intestinal secretory cells ( 34). ATOH1^−/− mice possess intact intestinal villi populated by absorptive enterocytes, but lack enteroendocrine, goblet, and Paneth cells. Analysis of ATOH1^−/− embryonic mouse intestines revealed that AGR2 expression is absent compared with controls ( 30). Immunohistochemistry, in the present study, established that AGR2 expression was identical to that previously reported for ATOH1 in the intestine, which is consistent with AGR2 serving as a downstream target of ATOH1 ( 33). Similar to AGR2, ATOH1 is expressed in proliferating and differentiated intestinal secretory cells. AGR2 and ATOH1 gene expression, however, are not similar in esophageal adenocarcinomas. Previous studies comparing the gene expression of normal esophagus, Barrett's esophagus, and esophageal adenocarcinoma revealed no differences for ATOH1, but enhanced expression for AGR2 in Barrett's esophagus and adenocarcinoma ( 22). In many other tumors studied, including those derived from the colon, small intestine, and pancreas, ATOH1 expression is lower in tumors compared with normal tissue ( 35). Development of esophageal adenocarcinoma may therefore require factors that promote AGR2 expression independent of ATOH1, which may include components of a signal transduction process active between ATOH1 and AGR2, or may be secondary to changes in AGR2 itself.

Cell culture media supplemented with normal and low serum levels failed to reveal any changes in cell proliferation rate after AGR2 knockdown in SEG-1 cells or overexpression in NIH3T3 cells (data not shown). Thus, AGR2's ability to affect tumor growth may be achieved through means other than those that directly affect cell proliferation. Similar to its role in X. laevis, AGR2 may serve a similar function in the gastrointestinal tract in determining cell fate. GFI1 is a basic helix-loop-helix transcription factor whose expression is also dependent on ATOH1, and is important in cell lineage determination among intestinal secretory cells. GFI1^−/− mice exhibit increased numbers of enteroendocrine cells, a decline in intestinal goblet cells, and the absence of Paneth cells. GFI1 expression influences cell fate but does not influence cell proliferation ( 30). Enhanced GFI1 expression, however, is associated with tumor development ( 31, 32). Quantitative real-time PCR determined that GFI1 expression is not influenced by AGR2, indicating that the growth-promoting properties of AGR2 are not mediated by changes in GFI1 expression. Consistent with these results, changes in GFI1 expression have not been associated with Barrett's esophagus or esophageal adenocarcinoma ( 22).

AGR2 coexpression with the Ki-67 antigen indicates a potential role in the growth and migration of normal intestinal cells of secretory lineage. Similar to its role in cement gland formation in X. laevis, AGR2 may participate in establishing cell fate and promoting growth in the gastrointestinal tract ( Fig. 4B). The expression of AGR2 in potential progenitor cells in the intestinal crypt and in adenocarcinomas support hypotheses linking stem or progenitor cells and cancer ( 13, 36, 37).

The results reported in this study support a role for AGR2 in cellular transformation and esophageal adenocarcinoma growth. Because enhanced AGR2 expression is universally observed in premalignant Barrett's esophagus and in malignant esophageal adenocarcinomas, AGR2 expression may be an early and essential prerequisite for cancer development. Enhanced AGR2 expression in many other adenocarcinomas suggests a potentially similar regulatory function. As a secreted protein, AGR2 is potentially accessible to therapeutic intervention and thus serves as a promising candidate for future studies.