• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Hum Pathol. Author manuscript; available in PMC Mar 1, 2010.
Published in final edited form as:
PMCID: PMC2746414

Anterior Gradient-2 is Over Expressed by Fibrolamellar Carcinomas


Anterior Gradient-2 expression is critical in normal embryonic development. Aberrant expression of Anterior Gradient-2 in adult tissues has been linked to breast, prostate, esophageal, and pancreatic carcinoma. To define the role of Anterior Gradient-2 in primary hepatocellular neoplasms, we used tissue microarrays and examined protein expression in typical hepatocellular carcinomas (N=44), fibrolamellar carcinomas (N=12), and hepatic adenomas (N=9). In non-neoplastic liver tissues, Anterior Gradient-2 was expressed in the septal-sized bile ducts and weakly in zone 3 hepatocytes in 11/61 (18%) of cases. In tumors, Anterior Gradient-2 was over expressed by only 1/44 (2%) of hepatocellular carcinomas. In contrast, 6/8 (75%) of fibrolamellar and 3/4 (75%) metastatic fibrolamellar carcinomas were positive. All 9 hepatic adenomas were negative. Further analysis of mRNA in fibrolamellar carcinomas identified 2 novel splice variants, but expression levels were very low. Sequencing of the Anterior Gradient-2 gene in fibrolamellar carcinomas identified several polymorphisms (refSNP Ids: rs6842, rs8071, rs1051905) but no mutations. In conclusion, Anterior Gradient-2 is over expressed in the majority of fibrolamellar carcinomas but only rarely is over-expressed in hepatocellular carcinomas.


Anterior Gradient-2 (AGR2) is located on Chr:7p21.3 [1] and encodes for the human homologue of a secreted protein first identified in Xenopus laevis. In Xenopus, the AGR2 homologue is critical in forming anterior structures during embryonic ectoderm development [2]. The function of AGR2 in human tissue is unknown, but AGR2 mRNA remains expressed in the mature stomach, small intestine and colon of mice [3]. In the small intestine of mice, expression is found predominately in paneth, neuroendocrine, and goblet cells [4]. Mice with mutant AGR2 have functional abnormalities of intestinal goblet cells and develop diarrhea [5].

Aberrant AGR2 expression has been found in an esophageal carcinoma cell line[4], and in primary breast, lung, and prostate carcinomas [4,68]. AGR2 can inhibit the function of P53[9] and over-expression of AGR2 in breast epithelial cell lines leads to metastases in an animal model [10]. However, the ways in which AGR2 contributes to neoplasia is not restricted to over-expression, as loss of AGR2 expression has also been associated with the dysplasia-to-carcinoma sequence in colonic polyps [11].

The potential role for AGR2 in hepatocellular carcinoma has not been investigated. In this study, we sought to explore the normal expression pattern of AGR2 in the non-neoplastic liver and to characterize AGR2 expression in primary hepatocellular neoplasms including typical hepatocellular carcinomas, fibrolamellar carcinomas, and hepatic adenomas. AGR2 was found to be over-expressed in fibrolamellar carcinomas and to further study mechanisms that may contribute to over-expression, the coding region of the gene was sequenced and expression of splice variants was studied.

Materials & Methods

This study was performed with appropriate Institutional Review Board approval on de-identified tissues.

Protein expression and tissue microarrays

To investigate the expression of AGR2 protein in hepatic tissues, previously described tissue microarrays [12] were used. Following heat antigen retrieval, five micron sections were immunostained with an antibody to AGR2 (Abcam, Cambridge, MA, polyclonal IgG, 1:250 dilution). The Dako EnVision+ Peroxidase kit was used for immunostaining. Cases were scored as positive when at least 5% of hepatocytes were immunolabeled. Staining distributions for positive cases were scored on a scale of 0–3: 0 (from 0–4% of cells positive), 1 (from 5–25%), 2 (26–50%), 3 (51–100%). Intensity was graded on a scale of 0–3. Benign colon served as a positive control.

The tissue arrays contained eight primary fibrolamellar carcinomas from six women and two men with an average age at resection of 27± 13 years. In addition, 4 metastatic fibrolamellar carcinomas were studied from two men and two women. Two of the metastasis corresponded to primary tumors in the study while the other two did not have the primary tumor available. All individuals in which fibrolamellar carcinomas arose were Caucasian. The background livers showed no significant fibrosis or inflammation, as is typical of fibrolamellar carcinomas.

The tissue microarray slides contained hepatocellular carcinomas that arose in the setting of cirrhosis. However, because fibrolamellar carcinomas arise in the setting of non-cirrhotic livers [13], we also chose to use tissue microarray slides that were enriched for typical hepatocellular carcinomas that arose in non-cirrhotic livers. All together, the arrays for this study contained 44 primary hepatocellular carcinomas with paired non-neoplastic tissues. The tumors were from primary liver resections in 28 men and 16 women with an average age at resection of 57.9±16.5 years. The underlying liver diseases were available in 35 individuals and included no known underlying liver disease with no significant fibrosis (N=16), chronic viral hepatitis C cirrhosis (N=10), chronic hepatitis B cirrhosis (N=2), cryptogenic cirrhosis (N=4), and alcohol related cirrhosis (N=3).

As a second control group, tissue microarrays of hepatic adenomas were also studied. These primary liver tumors also arise in non-cirrhotic livers with no evidence of background liver disease. This group also served as a control for any potential association with estrogen receptor expression and AGR2 positivity, an association that has been identified in breast carcinomas[7,10]. Hepatic adenomas typically arise in the setting of excess estrogen exposure and can express estrogen receptors[14]. Nine hepatic adenomas were studied and all arose in women with no back ground liver disease and had an average age of 34.2+9.0 years at tumor resection.

Identification of CpG islands in AGR2 promoter

The promoter of AGR2 [5] was analyzed for evidence of a CpG island using Methprimer (http://www.urogene.org/methprimer/index1.html), a publicly available web based program that can identify CpG islands.

Tissues for mRNA studies and for sequencing of the coding region of AGR2

For examination of mRNA, additional cases were studied based on available tissues. Fresh tissues were obtained at the time of surgery from primary liver neoplasms and adjacent non-neoplastic liver tissues. All tissues were obtained between 1999 through 2007 from routine surgical cases at the Johns Hopkins Hospital. Tissues were frozen in liquid nitrogen, and stored at −80 C prior to use. The pathological diagnoses were confirmed in all cases. The studied frozen tissues were from primary fibrolamellar carcinomas and metastases. Two cell lines were also examined, HepG2 and Hep3B. These cell lines were purchased (American Type Culture Collection, Manassas, VA) and grown in minimal essential medium under routine conditions. To normalize mRNA expression in the two cell lines, we used the average expression levels of five non-neoplastic, non-inflammed and non-cirrhotic livers.

RNA was extracted using TRIzol (Invitrogen life technologies, Carlsbad, CA, USA) followed by precipitation with isopropyl alcohol. cDNA was prepared using one μg of RNA with oligo dT primers and the Superscript First –Strand synthesis system for RT-PCR (Invitrogen life technologies, Carlsbad, CA, USA).

Primers were designed using the Genebank cDNA transcript NM_006408 to amplify exons 1–8 from cDNA. All primers are shown in Table 1.

Table 1
Primers used to study AGR2.

Transcript expression by reverse transcriptase-PCR

cDNA was synthesized using 1 μg of RNA as input. One μl of cDNA was used then used as input for PCR. Primers were designed using the Genebank cDNA transcript NM_006408, with the forward primer in the exon 1 and the reverse primer in exon number 8 (Table 1). With this assay, an un-spliced transcript would produce a 653 bp product and any internal splice variation would be seen as smaller bands on agarose gels. An additional splice variant, termed the “long-form”, that uses alternative codons at the 5 primer end has been reported [5] and reverse transcriptase-PCR was also performed for this transcript (Table 1). PCR cycling conditions were 10 minutes at 95°C followed by 45 cycles of 95°C for 20 seconds, 55°C 25 seconds, and 72°C for 1 minute. The amplified products were analyzed by ethidium bromide in 1% agarose gels. Experiments without RT polymerase showed no amplicons. Selected bands were purified from the gel and directly sequenced.

AGR2 expression analysis by real time PCR

Two smaller transcripts were identified in this study by reverse transcriptase-PCR (termed AGR2Δ4-6 and AGΔ6). The relative expression levels of these transcripts were further studied by real time PCR with assays that amplified only the smaller sized transcripts. To ensure specificity of the real time PCR assays for the smaller splice variants, the full length transcript was cloned and the real time PCR assays for the smaller sized transcript were negative using the full length transcript as template (data not shown).

Real time PCR was performed with the SmartCycler system (Cepheid) using the Fast Start SYBR green master mix (Roche) with 1μg of cDNA and cycling conditions of 95°C for 10 minutes followed by 40 cycles of 95°C for 20 seconds, 55°C 30 seconds, and 72°C for 30 seconds. The specificity of PCR products were ascertained by melt curve analysis and agarose gel electrophoresis. Expression levels were normalized to beta-glucuronidase using previously described primers[15].


AGR2 protein expression in the non-neoplastic livers of the tissue microarrays

In the non-neoplastic livers, septal-sized bile ducts were typically positive (Figure 1A), while smaller ducts were negative. In addition, 11/61 (18%) of the benign livers showed weak immunostaining, generally limited to zone 3 hepatocytes (Figure 1B).

Figure 1
Panel A, 100X original magnification. AGR2 showed weak expression in zone 3 hepatocytes in the background livers of 18% of cases. Panel B, 100X original magnification. Septal sized bile ducts were also positive for AGR2 in the non neoplastic liver tissues. ...

AGR2 protein expression in hepatocellular tumors

In tumors, AGR2 was positive in 6/8 (75%) of primary fibrolamellar carcinomas (Figure 1C) and 3/4 (75%) metastatic fibrolamellar carcinomas. The staining was cytoplasmic with no nuclear accumulation evident. The pale bodies present in fibrolamellar carcinomas were often strongly stained by AGR2 (Figure 1C). The median distribution of staining in fibrolamellar carcinomas was 2, and the median intensity of staining was 2. The staining patterns were as follows for the distribution (D) and intensity (I): D1I1 (N=1), D1I3 (N=1), D2I3 (N=2), D3I3 (N=2), D3I2 (N=1). For the two cases with paired primary and metastases, both the primary and metastases were AGR2 positive. In addition, five of the primary tumors had paired non-neoplastic tissues and in three of these cases the background livers showed weak hepatocellular staining as described above.

In one interesting case, a fibrolamellar carcinoma was present adjacent to a separate nodule of a more typical hepatocellular carcinoma in the same liver of a 26 year old woman with no background liver disease. AGR2 was positive only in the fibrolamellar tumor and negative in the separate tumor with typical hepatocellular carcinoma morphology.

In contrast to the fibrolamellar carcinomas, most typically hepatocellular carcinomas were negative for AGR2 (Figure 1D), while only 1/44 (2%) hepatocellular carcinoma was positive for AGR2 protein expression. The AGR2 positive hepatocellular carcinoma showed a typical compact growth pattern and arose in a 40 year old woman with no underlying liver disease. In one additional case, a hepatocellular carcinoma arose in the setting of cryptogenic cirrhosis and showed an unusual mixed hepatocellular-cholangiocellular morphology: the cholangiocellular component was positive for AGR2 while the hepatocellular component was negative. All of the nine hepatic adenomas were negative for AGR2 expression.

Statistical analysis (SYSTAT, V10) showed that AGR2 positivity was strongly associated with fibrolamellar carcinomas (Pearson Chi-square, p< 0.0001).

Promoter and mRNA analysis

In silico analysis of AGR2 promoter

Because AGR2 is normally expressed during embryogenesis, we investigated whether this gene was potentially inactivated by methylation in normal adult liver tissues. Many human genes are inactivated by methylation of CpG dinucleotides that cluster into CpG islands and can be located within gene promoters. The 2000 bases upstream of AGR2, which are reported to contain the promoter[5], showed no evidence for CpG islands.

mRNA analysis

As expected, AGR2 was over expressed in fibrolamellar carcinomas (Table 2). The AGR2 mRNA transcripts were next studied by reverse transcriptase PCR and two different splice variants were identified (Table 2). Sequencing of the variants showed one lacked exon 6 (Δ6 transcript) and one lacked exons 4–6 (Δ4–6 transcript) (Figure 2). The Δ6 transcript was present in the majority of the tumors and metastases while the Δ4–6 transcript was found in a single case (Table 2). Both used alternative splice sites. The 20 base pairs surrounding the Δ4–6 transcript spliced splice site are as follows: exon 4 5′-AGACATATGAAGAAGCTCTA/spliced/ATGTTTGTTGACCCATCTCT-3′ exon 7; while the 20 base pairs surrounding the Δ6 transcript are as follows: exon 5 5′-TTGTCCTCCTCAATCTGGTT/spliced/ACCCATCTCTGACAGTTAG-3′ exon 7. Quantitation by real time PCR showed expression levels of both transcripts were low, less than 1% of the total AGR2 expression levels (data not shown).

Figure 2
The full length reference cDNA and splice variants of AGR2 are shown. WT, full length reference transcript; LF, long form (see materials and methods); Delta 4–6, transcript missing exons 4 through 6; Delta 6, transcript missing exon 6.
Table 2
Clinicopathological information and slice variant analysis of fibrolamellar carcinomas and control cell lines

An additional splice variant, termed the “long form” has been previously reported [5]. PCR analysis for this form were negative in all tissues.

Sequencing of AGR2 exons

To further understand why AGR2 was over-expressed in fibrolamellar carcinomas, we next sequenced the exons 1–8 of AGR2. Five primary fibrolamellar carcinomas, 5 metastases, and the HepG2 and Hep3B cell lines were examined. Several polymorphisms were found (Table 3) but no mutations were identified.

Table 3
Base pair changes found in Exons 1 through 8 of AGR2 in fibrolamellar carcinomas and cell lines.rs6842refSNP ID: rs8071refSNP ID: rs1051905


In this study, AGR2 protein expression was strongly associated with fibrolamellar carcinomas but not typical hepatocellular carcinomas. This observation underscores the differences between fibrolamellar carcinomas and hepatocellular carcinomas and links fibrolamellar carcinomas to a unique biomarker that is not shared by most hepatocellular carcinomas.

To further understand why AGR2 was over-expressed in fibrolamellar carcinomas, mechanisms relevant to gene expression were studied. We investigated whether this gene had a CpG island within its promoter region, for which differential methylation could explain its aberrant expression. However, no CpG island was identified. Aberrant splicing of genes has been associated with hepatocellular carcinoma [15] and to investigate whether over-expression of AGR2 was related to changes in splicing, reverse transcriptase PCR was performed. Novel splice variants were found, but subsequent quantitative analysis showed expression of these splice variants was at very low levels, even though they were unique to tumors. Whether splice variants could also potentially be detectable in typical hepatocellular carcinomas was not further investigated because we found little evidence of protein expression by tissue microarrays. We also sequenced the coding regions of AGR2 and identified several polymorphisms but found no mutations. Together, these findings suggest AGR2 is most likely over-expressed because it is a target of a larger signaling/metabolic pathway. However, other possibilities such as translation dysregulation or abnormalities in protein degradation remain to be investigated.

The significance of the polymorphisms identified in the fibrolamellar carcinomas is uncertain. The polymorphism T498C (refSNP ID: rs:6842) was present in all but one of the fibrolamellar carcinomas and the majority of these cases were homozygous with a C/C genotype. This observation may be of interest, given that the frequency in fibrolamellar carcinomas appears higher than that in the NCBI database for the general Caucasian population, where 16.7% are reported to have a C/C genotype. However, the small numbers of cases in this study preclude drawing any conclusions on this point. Also, no change in amino acid sequence is predicted, making this SNP of uncertain relevance to AGR2 expression.

Within the hepatic tumors we studied, AGR2 expression was strongly associated with fibrolamellar carcinomas. Little is known about the signaling or metabolic pathways that are associated with AGR2 expression, but MAPK/ERK signaling can regulate AGR2 expression in cell lines in the context of physiological stress [16]. This observation may be relevant to fibrolamellar carcinomas in that we have previously shown that MAPK signaling is a feature of the fibrolamellar carcinoma transcriptome [17]. However, more definitive data on how AGR2 fits into the pathogenesis of fibrolamellar carcinomas are lacking at this time. In addition to data showing MAPK and PI3K signaling in fibrolamellar carcinomas[17], recent studies have also shown that the mTOR pathway can be dysregulated in a subset of these tumors[12]. There is data from other cell types that these pathways can “cross-talk” [18], but the ways in which these pathways interact in fibrolamellar carcinomas are unclear. It may be of interest to note, however, that both the PI3K and MAPK pathways can converge on mTOR in the context of autophagy signaling.

The potential role for AGR2 role in maintaining neuroendocrine lineage in small intestinal cells [4] is particularly interesting given that fibrolamellar carcinomas can express chromogranin and other neuroendocrine markers [13]. A previously identified protein over-expressed in fibrolamellar carcinomas, neurotensin [19], has also been linked to neuroendocrine cells of the intestine [20].

In normal adult tissues, AGR2 is expressed in gastrointestinal organs in paneth cells, endocrine cells, and goblet cells. The role of AGR2 expression in these cell types is poorly understood, but in the small intestine a role for determining cell lineage fate has been suggested [4]. In this study, we extend the known expression pattern of AGR2 to include the septal sized bile ducts of the liver. We have not studied the largest of the intrahepatic bile ducts nor the extrahepatic biliary tree so the full expression pattern of AGR2 in the biliary system remains to be investigated. In addition, we found AGR2 expression in zone three hepatocytes in a subset of cases. One intriguing possibility would be that expression is related to maintaining the zonal gradients in the liver. However, another possibility would be that expression was stimulated due to microenvironmental stressors, as studies have shown that AGR2 can be up-regulated in response to stressors such as nutrient and oxygen deprivation[16].

In conclusion, AGR2 can be expressed in non-neoplastic liver tissues including septal-sized bile duct and zone 3 hepatocytes. The majority of fibrolamellar carcinomas (both primary and metastatic) over-express AGR2 protein, while hepatocellular carcinomas and hepatic adenomas are typically negative. Over expression of AGR2 protein does not appear to be associated with genetic or epigenetic changes within the AGR2 gene and is thus most likely over-expressed as part of a larger pathway.


Supported by K08DA016370 (MT), R01DK078686 (MT).


This data was presented in part at the United States and Canadian Academy of Pathology, March 2007, San Diego, California.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


1. Petek E, Windpassinger C, Egger H, et al. Localization of the human anterior gradient-2 gene (AGR2) to chromosome band 7p21.3 by radiation hybrid mapping and fluorescencein situ hybridisation. Cytogenet Cell Genet. 2000;89:141–2. [PubMed]
2. Aberger F, Weidinger G, Grunz H, et al. Anterior specification of embryonic ectoderm: the role of the Xenopus cement gland-specific gene XAG-2. Mech Dev. 1998;72:115–30. [PubMed]
3. Komiya T, Tanigawa Y, Hirohashi S. Cloning of the gene gob-4, which is expressed in intestinal goblet cells in mice. Biochim Biophys Acta. 1999;1444:434–8. [PubMed]
4. Wang Z, Hao Y, Lowe AW. The adenocarcinoma-associated antigen, AGR2, promotes tumor growth, cell migration, and cellular transformation. Cancer Res. 2008;68:492–7. [PubMed]
5. Zheng W, Rosenstiel P, Huse K, et al. Evaluation of AGR2 and AGR3 as candidate genes for inflammatory bowel disease. Genes Immun. 2006;7:11–8. [PubMed]
6. Fritzsche FR, Dahl E, Dankof A, et al. Expression of AGR2 in non small cell lung cancer. Histol Histopathol. 2007;22:703–8. [PubMed]
7. Fritzsche FR, Dahl E, Pahl S, et al. Prognostic relevance of AGR2 expression in breast cancer. Clin Cancer Res. 2006;12:1728–34. [PubMed]
8. Zhang JS, Gong A, Cheville JC, et al. AGR2, an androgen-inducible secretory protein overexpressed in prostate cancer. Genes Chromosomes Cancer. 2005;43:249–59. [PubMed]
9. Pohler E, Craig AL, Cotton J, et al. The Barrett’s antigen anterior gradient-2 silences the p53 transcriptional response to DNA damage. Mol Cell Proteomics. 2004;3:534–47. [PubMed]
10. Liu D, Rudland PS, Sibson DR, et al. Human homologue of cement gland protein, a novel metastasis inducer associated with breast carcinomas. Cancer Res. 2005;65:3796–805. [PubMed]
11. Lee S, Bang S, Song K, et al. Differential expression in normal-adenoma-carcinoma sequence suggests complex molecular carcinogenesis in colon. Oncol Rep. 2006;16:747–54. [PubMed]
12. Sahin F, Kannangai R, Adegbola O, et al. mTOR and P70 S6 kinase expression in primary liver neoplasms. Clin Cancer Res. 2004;10:8421–5. [PubMed]
13. Torbenson M. Review of the clinicopathologic features of fibrolamellar carcinoma. Adv Anat Pathol. 2007;14:217–23. [PubMed]
14. Torbenson M, Lee JH, Choti M, et al. Hepatic adenomas: analysis of sex steroid receptor status and the Wnt signaling pathway. Mod Pathol. 2002;15:189–96. [PubMed]
15. Kannangai R, Sahin F, Adegbola O, et al. FHIT mRNA and protein expression in hepatocellular carcinoma. Mod Pathol. 2004;17:653–9. [PubMed]
16. Zweitzig DR, Smirnov DA, Connelly MC, et al. Physiological stress induces the metastasis marker AGR2 in breast cancer cells. Mol Cell Biochem. 2007;306:255–60. [PubMed]
17. Kannangai R, Vivekanandan P, Martinez-Murillo F, et al. Fibrolamellar carcinomas show overexpression of genes in the RAS, MAPK, PIK3, and xenobiotic degradation pathways. Hum Pathol. 2007;38:639–44. [PubMed]
18. Campbell M, Allen WE, Sawyer C, et al. Glucose-potentiated chemotaxis in human vascular smooth muscle is dependent on cross-talk between the PI3K and MAPK signaling pathways. Circ Res. 2004;95:380–8. [PubMed]
19. Collier NA, Weinbren K, Bloom SR, et al. Neurotensin secretion by fibrolamellar carcinoma of the liver. Lancet. 1984;1:538–40. [PubMed]
20. Evers BM. Neurotensin and growth of normal and neoplastic tissues. Peptides. 2006;27:2424–33. [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...