Clinical implications of AGR2 in primary prostate cancer: Results from a large‐scale study

Human anterior gradient‐2 (AGR2) has been implicated in carcinogenesis of various solid tumours, but the expression data in prostate cancer are contradictory regarding its prognostic value. The objective of this study is to evaluate the expression of AGR2 in a large prostate cancer cohort and to correlate it with clinicopathological data. AGR2 protein expression was analysed immunohistochemically in 1023 well‐characterized prostate cancer samples with a validated antibody. AGR2 expression levels in carcinomas were compared with matched tissue samples of adjacent normal glands. AGR2 expression levels were dichotomized and tested for statistical significance. Increased AGR2 expression was found in 93.5% of prostate cancer cases. AGR2 levels were significantly higher in prostate cancer compared with normal prostate tissue. A gradual loss of AGR2 expression was associated with increasing tumour grade (ISUP), and AGR2 expression is inversely related to patient survival, however, multivariable significance is not achieved. AGR2 is clearly upregulated in the majority of prostate cancer cases, yet a true diagnostic value appears unlikely. In spite of the negative correlation of AGR2 expression with increasing tumour grade, no independent prognostic significance was found in this large‐scale study.

In 2020, prostatic cancer (PCa) was the most frequent cancer among men with more than 1.4 million new cases as well as the fifth leading cause of cancer death among men [1].Still, due to the fairly indolent nature of this neoplasm, most patients diagnosed with prostate cancer will not die from it.However, a certain population will eventually succumb, which is socio-economically and epidemiologically relevant due to the high frequency of the disease with incidence rates expected to continuously increase over the next years [2].Therefore, it is of major importance to delineate these groups of patients and to provide individual information concerning the prognosis of each patient.If possible, this would allow a well-founded decision on the various therapeutic options available for this disease.Tumour grade (Gleason score [GS]/ISUP grade groups), TNM staging, prostate-specific antigen (PSA) levels and tumour margin status are well established for the prognostic stratification of prostate cancers [3][4][5].However, patients with comparable histological and clinical prostate cancer parameters can demonstrate strikingly varied presentations of clinical outcomes [3].Therefore, prognostic biomarkers that add information about prognosis or even therapeutic prediction are highly warranted.
A candidate biomarker for prostate cancer, others and we identified in former studies, is Anterior Gradient 2 protein (AGR2) [4][5][6][7][8][9].This is a hormoneinducible secreted protein localized on chromosome 7p21 that was first identified as an orthologue of the Xenopus laevis gene XAG-2 [5].It is a member of the protein disulphide isomerase family of endoplasmic reticulum-resident proteins, exhibiting basic features of pro-oncogenic proteins in humans [10][11][12].Nonetheless, several studies have documented the role of AGR2 in physiological as well as pathological processes [10][11][12][13].Generally, endoplasmic reticulum stress is seen as an inducer of AGR2 production through unfolded protein response (UPR) [10,13] and it has been shown that AGR2 can inhibit the function of p53, which gives it a possible prooncogenic function [10,12].
In humans, AGR2 expression has been reported in various adenocarcinomas and other tumours (reviewed in [14]).While most of the published studies have focused on the role of AGR2 in breast and prostate cancer, it has also been described in malignant tumours of the lung, liver, ovaries, intestine and urinary bladder [15][16][17][18][19][20].In most of these entities, AGR2 was found upregulated in comparison to normal tissue.Several studies have demonstrated a correlation between increased AGR2 expression and poor prognosis in lung, breast and gastric cancer [14,[20][21][22][23][24][25][26].Controversially, some studies discovered a protective effect of AGR2 upregulation on tumour progression [15,16,27].In breast cancer, AGR2 expression is associated with estrogen receptor positivity and it is thought to enhance metastasis in the subgroups of estrogen receptor-positive and tamoxifen-treated tumours [20,28].Most studies advocated a pro-invasive, pro-metastatic or adverse prognostic effect of AGR2 over-expression in the respective tumour.
In an earlier study, we found AGR2 overexpressed in 89% of prostate carcinomas, but no prognostic value [7].Since, numerous studies have confirmed the overexpression of AGR2 in prostate cancer [4-6, 8, 29-31].The prognostic value of AGR2 expression in prostate cancer remains controversial to date, as some studies reported no prognostic value, while others reported a better or a worse outcome of AGR2-positive tumours.These conflicting findings prompted us to clarify the prognostic value of AGR2 in prostate cancer by a large-scale expression analysis of representative prostate cancer cohorts following radical prostatectomy.

Patients
In order to evaluate the expression of AGR2 in prostate cancer, three previously published cohorts of radical prostatectomy (RP) cases were used.One cohort consisted of 640 prostate cancer patients, who underwent RP between 1999 and 2005 at the Charit e University Hospital in Berlin, Germany [7,32].The next cohort consisted of 238 patients having undergone RP for treatment of primary prostate cancer between 1999 and 2006 at University Hospital Zurich, Switzerland [33].The last cohort consisted of 300 patients, who underwent RP at the University Hospital Bonn between 2000 and 2008 [34].All cohorts were in tissue micro-array format, as described.We restricted the analysis to patients with clinical follow-up data, leaving 1023 patients in the analysis.The clinical-pathological parameters are given in Table 1.In this cohort, 212 patients (20.7%) experienced a biochemical recurrence (BCR), defined as a rising PSA value >0.2 ng/mL, after a median time of 50 months.BCR was used as an endpoint in the survival analyses.
The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.Appropriate ethic approvals have been obtained before the study begin from the respective local ethics committees (Berlin: EA1/06/2004; Zurich: StV 25-2007; Bonn: 071/14).

AGR2-siRNA transfection for transient gene knockdown
PC-3 cells were transfected using Lipofectamine (Invitrogen, Waltham, MA, USA) as a transfection reagent.Cells were transfected on the day of seeding with a final siRNA concentration of 10 nM (all siRNAs were purchased from Qiagen, Hilden, Germany).One day after transfection, the medium of PC-3 cells was changed.Knockdown efficiency was confirmed on the protein level using Western blot analysis.

Immunohistochemistry
Immunohistochemistry was carried out using freshly cut (3 lm) sections that were mounted on super frost slides (Menzel-Gl€ aser, Braunschweig, Germany).The AGR2/Gob-4 monoclonal mouse antibody (clone 1C3, dilution 1:500) was incubated using BondMax autostainers (Leica Microsystems, Wetzlar, Germany.Pretreatment Protocol H2-60, detection with Refine DAB Polymer-Kit).Finally, the slides were briefly counterstained with hematoxylin, dehydrated and mounted.For the peptide blocking experiment, one duplicate slide was incubated with (a) primary antibody and (b) primary antibody that had been pre-incubated with a 10 M excess of the respective immunogenic AGR2 peptide (Abnova Corp., Taipei City, Taiwan).As negative controls, four slides were processed without primary antibodies.

Evaluation of immunohistochemical stainings
Immunostaining of AGR2 was evaluated by two pathologists (MM, GK) using a four-tier intensity scoring system (0 negative, 1 weak, 2 moderate and 3 strong).In order to enhance intra-and interobserver conformity, a panel with four representative pictures with different expression levels as a benchmark was used (Fig. S1).

Analysis of TCGA and GTEx data sets
An in silico analysis of AGR2 mRNA expression levels from publicly available transcriptome databases [The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx), prostate cancer cohort] in the Xena browser software was used to complement the immunohistochemistry data [35].The endpoint used for survival analysis was biochemical recurrence (BCR).
Patients were dichotomized by the median expression of AGR2.Survival curves were visualized using the Kaplan-Meier plot with the difference between survival distributions assessed by the log-rank test.The Cox proportional hazards model was used to test the statistical significance of clinicopathological parameters in both a univariate and a multivariate setting.p Values of <0.05 were considered statistically significant.

Verifying specificity of the anti-AGR2 antibody (clone 1C3)
AGR2 expression on protein level was determined in PC-3 cells.A knockdown protocol based on RNA interference was established.Western blot analysis 72-h post-transfection showed a potent AGR2 knockdown on protein level with two different siRNAs, whereas one siRNA showed only a minor effect and one siRNA failed (Fig. 1A).Additionally, pre-incubation of the AGR2 antibody with an excess of its immunogenic peptide abolished immunoreactivity to background levels (Fig. 1B).

Association of AGR2 expression with clinical-pathological parameters
AGR2 expression in tumour tissues was significantly associated with younger patient age, negative margin status and lower ISUP grade groups.There was no significant association between AGR2 expression and tumour stage (pT-status) or PSA levels (Table 1).The associations with age and margins were confirmed in Spearman rank correlation analyses: Age: correlation coefficient (CC) À0.132, p = 0.001; Margins (R status) CC À0.101, p = 0.001.ISUP grade groups though dropped out (CC À0.049, p = 0.120), which clarifies that this association is weak.
Analysis of the prognostic value of low-level AGR2-mRNA using TCGA/GTEx databases Using the TCGA/GTEx RNA-seq data sets, we analysed the association between low/high mRNA levels of AGR2 or AGR2-splice variants in an independent virtual cohort of 497 PCa patients.A normal tissue cohort of 100 samples (GTEx) was used as a control.There was a statistically highly significant association between reduced mRNA levels of AGR2-203, the predominant AGR2 isoform (major allele/wild type) and reduced patient survival (p = 0.0223).The splice variants AGR2-201 and AGR2-204 showed a similar tendency, however, the correlation remained statistically insignificant (p = 0.0599 and p = 0.0668, respectively) (Fig. 3B,  C).AGR2-202 showed a significant correlation between high mRNA levels and a reduced progression-free interval (p = 0.0464) (Fig. 3D).AGR2-205, AGR2-206 and AGR2-207 were not studied, as they represent non-protein-coding AGR2 splice variants and in consequence do not affect IHC.

Prognostic value of clinicopathological parameters and AGR2 protein expression
First, univariate Cox analyses of the entire cohort were performed to demonstrate the prognostic value of clinicopathological parameters and hence confirm the representativity of this cohort.As expected, serum PSA, ISUP grade groups, pT categories and margin status could be verified as prognostic parameters in this cohort.Also, AGR2 expression in prostate carcinomas was demonstrated as a prognostic parameter, with tumours with lower levels showing an earlier biochemical disease progression (Table 2).
In a multivariate Cox regression, the conventional prognostic parameters remained highly significant, while AGR2 expression dropped out (Table 3).
In a Kaplan-Meier analysis stratified for ISUP grade groups (low grade = 1-2 vs. high grade = 3-5), AGR2 expression showed no separation of curves whatsoever in low-grade tumours (Fig. 4B), whereas in high-grade tumours, an AGR2 loss indicated a clear trend towards earlier biochemical relapse, but failed statistical significance (Fig. 4C, p = 0.104).

DISCUSSION
The protein AGR2 is a member of the protein disulfide isomerase (PDI) family and represents the human orthologue of the Xenopus laevis cement gland protein (XAG2) [36][37][38].According to the Human Protein Atlas, higher protein expression rates are seen physiologically in tissues of the respiratory tract, gastrointestinal tract including the gallbladder and pancreas, the urinary bladder and genital organs of both sexes including endometrium, cervix, fallopian tube, placenta, prostate, epididymis and seminal vesicle (https://www.proteinatlas.org/ENSG00000106541-AGR2/tissue)[39].The high expression seen in these organs indicates an involvement of AGR2 in mucus secretion but the functions of AGR2 appear to be much more complex.It plays a role in multiple signalling pathways including cell signalling, protein trafficking, proteostasis and in tumour chemoresistance and tumour cell dissemination [40].Dysregulation of AGR2 expression and/or function is widely implicated in human diseases.Preliminary investigations suggest that downregulation or loss of function of AGR2 is involved in inflammatory bowel disease, ulcerative colitis and Crohn's disease [41][42][43].Interestingly, AGR2 protein is often overexpressed and aberrantly secreted in a broad range of solid tumours [22].An overview of studies analysing the prognostic value of AGR2 in solid tumours is given in Table 4.There is experimental evidence that ARG2 can exert pro-oncogenic properties related to differentiation, migration/metastasis, proliferation, senescence, and chemoresistance when overexpressed or misplaced in the tumour niche/microenvironment, thus representing putative pro-oncogenic signalling intermediate in human cancer [9-11, 14, 44].Although in vitro studies reveal a complex functional role of AGR2 in tumourigenesis and tumour progression, its prognostic or clinical significance remains controversial for most tumour entities.
Aiming to clarify the prognostic value of AGR2 in PCa, a large-scale expression analysis of 3 representative prostate cancer cohorts with a total of 1023 patients was performed.Benign normal tissue was predominantly AGR2 negative, whereas the vast majority of prostate cancer cases were AGR2 positive, only 6.4% of the tumours were completely AGR2 negative.In tumour tissue, AGR2 expression was significantly associated with younger patient age, negative margins and lower ISUP grade groups.There was no significant association between AGR2 expression and tumour stage (pT status) or PSA levels.The prevalence of AGR2 in PCa compared with normal adjacent tissue is in agreement with previous studies in prostate cancer and other tumours [4,5,18,30].
Initial studies investigating the prognostic value of AGR2 expression in PCa yielded conflicting results.While Zhang et al. [8] associated increased AGR2 expression in PCa with high GS and lower overall survival, a study by Maresh et al. [4] demonstrated earlier PSA relapses in tumours with low AGR2 expression.This trend became significant in a small group (65/187) of patients with advanced stage III/IV tumours.While the Gleason score, preoperative PSA and pT stage were good prognosticators in the unrestricted cohort of 187 patients, only pT stage remained significant in the restricted cohort, questioning the relevance of this subgroup.The discrepancy of the studies mentioned above could arise from the use of different antibodies and highlights the necessity of thorough antibody controls (Western blotting, protein knockdown, peptide blocking experiments) to exclude unspecific binding.Although we can rule out unspecific binding of our antibody (Fig. 1), we do not know the precise epitope recognized by it.Several studies revealed the presence of AGR2 splice variants in a number of tumours, including prostate cancer [11,45,46].Without precise knowledge of the epitope recognized by the antibody, thorough discrimination of the various AGR2 forms is not possible and can affect the interpretation of IHC staining.
The absence of AGR2 immunoreactivity was observed in only 6.4% of PCa.To investigate a potential prognostic value of our immunohistochemical AGR2 data, tumour samples were dichotomized by the median expression of AGR2.Using Kaplan-Meier analysis, we found a significant association between reduced AGR2 expression and shortened patient survival (p = 0.015).This observation is consistent with several studies describing a gradual loss of AGR2 with increasing Gleason scores [6,19,30].Though it is tempting to speculate, that the negative prognostic value of AGR2 expression may be attributed to this correlation with tumour grade, we found in an analysis of AGR2 stratified for Gleason grade groups no prognostic value at all in low grade (GG1-2) tumour but a clear trend towards a prognostic value of AGR2 in higher-grade (GG3-5) tumours.This indicates, that the prognostic value of AGR2 in  prostate cancer, though it is expressed at lower levels in higher-grade tumours, is independent of tumour grade.In addition, we demonstrated an association between reduced AGR2-203 mRNA (AGR2 wild type) and lower patient survival in an independent virtual TCGA/GTEx-cohort of more than 490 PCa patients thereby confirming our IHC data (Fig. 3).As AGR2-202 is only weakly expressed across different cancers its putative protein is not expected to dramatically affect AGR2-203/major allele staining [11].Unfortunately, IHC studies on ARG2 are still hampered by our limited knowledge of the role of AGR2 splice products.Of particular interest, the mRNA of a new splice variant, designated AGR2 SV-H, was able to discriminate between benign and PCa in tissue samples and outperform serum PSA monitoring when detected in urinary exosomes [47].In cholangiocarcinoma, the same AGR2 variant (now referred to as AGR2vH) was associated with a more aggressive phenotype [46,48].Although AGR2vH mRNA is overexpressed in PCa and cholangiocarcinomas, the existence of specific AGR2vH translation products/proteins remains controversial, highlighting the need for splice variant-specific antibodies for future studies [47,49].There is experimental evidence that a functional AR is required for androgen-and estrogendependent regulation of AGR2 in PCa cell lines [6].Since both PSA and AGR2 are AR-regulated proteins, one might expect a correlation between AGR2 levels and PSA serum levels.However, the present study confirms previous findings that there is no significant association between AGR2 expression in tumours and serum PSA [29].
In summary, we found AGR2 expression significantly elevated in 93.5% of primary prostate cancer tissues.A gradual loss of AGR2 expression in tumour samples (protein/mRNA) was associated with increasing tumour grade (ISUP) and shortened patient survival.No associations of AGR2 expression with other clinical-pathological parameters like PSA were found in this large-scale study.We conclude that AGR2 is not suitable as an independent prognostic marker in prostate cancer but deserves further study as a target of therapy or a serum marker.
The excellent technical support by Silvia Behnke and Susanne Steiner (immunohistochemistry) is gratefully acknowledged.

Fig. 1 .
Fig. 1.Demonstration of AGR2 antibody specificity.(A) Transient knockdown of AGR2 by siRNA is seen in this Western blot for #5 and #6.(B) AGR2 Immunohistochemistry, demonstrating a predominantly cytoplasmic reaction in invasive epithelium (left).After pre-incubation with a 10 M excess of the immunogenic peptide, the immunoreactivity was reduced to background levels (right).

Fig. 2 .
Fig. 2. AGR2 Immunohistochemistry in prostate tissues.(A) AGR2 is only very weakly expressed in secretory epithelium of benign glands.(B) Strong AGR2 staining was seen in some cases of atrophic epithelium.(C) Prostate carcinoma without AGR2 expression.(D-F) Prostate cancer with weak (D), moderate (E) or strong (F) AGR2 positivity.Note the AGR2negative benign gland in the centre of (F).

Table 1 .
Clinical-Pathological parameters of the patients analysed

Table 2 .
Univariate Cox regression: prognostic value of clinicopathological parameters (BCR as endpoint)

Table 3 .
Multivariate Cox regression: prognostic value of clinicopathological parameters (BCR as endpoint)