DNA Methylation of PITX2 and PANCR Is Prognostic for Overall Survival in Patients with Resected Adenocarcinomas of the Biliary Tract

Biliary tract cancers (BTC) are rare but highly aggressive malignant epithelial tumors. In order to improve the outcome in this lethal disease, novel biomarkers for diagnosis, prognosis, and therapy response prediction are urgently needed. DNA promoter methylation of PITX2 variants (PITX2ab, PITX2c) and intragenic methylation of the PITX2 adjacent non-coding RNA (PANCR) were investigated by methylations-specific qPCR assays in formalin-fixed paraffin-embedded tissue from 80 patients after resection for BTC. Results were correlated with clinicopathologic data and outcome. PITX2 variants and PANCR showed significant hypermethylation in tumor vs. normal adjacent tissue (p < 0.001 and p = 0.015), respectively. In survival analysis, dichotomized DNA methylation of variant PITX2c and PANCR were significantly associated with overall survival (OS). Patients with high tumor methylation levels of PITX2c had a shorter OS compared to patients with low methylation (12 vs. 40 months OS; HR 2.48 [1.38–4.48], p = 0.002). In contrast, PANCR hypermethylation was associated with prolonged survival (25 vs. 19 months OS; HR 0.54 [0.30–0.94], p = 0.015) and qualified as an independent prognostic factor on multivariate analysis. The biomarkers investigated in this study may help to identify BTC subpopulations at risk for worse survival. Further studies are needed to evaluate if PITX2 might be a clinically useful biomarker for an optimized and individualized treatment.


Introduction
Biliary tract cancers (BTC) represent a heterogenous group of malignancies that include adenocarcinomas of the intra-and extrahepatic bile ducts, the ampullary region, and the gallbladder.
With less than 3% of all newly diagnosed malignancies in adults in the United states, BTCs are rare neoplasms [1] that share a particularly aggressive biological behavior. Even after treatment with curative intention, the prognosis is generally poor with 5-year survival rates of <30% depending on the BTC subtype and stage of disease [2][3][4]. In recent years, new insights into the molecular pathogenesis of BTC have been gained by high throughput 'omics' technologies. Exomic sequencing, for instance, has illuminated the molecular landscapes of cholangiocarcinoma and gallbladder cancer and their potential for novel diagnostic applications and targeted therapies [5,6].
DNA methylation is an epigenetic mechanism involved in various fundamental biological processes including cell differentiation and cell development [7,8]. Furthermore, aberrant DNA methylation has been shown to play a crucial role during carcinogenesis [9,10]. Due to the presence of cancer-specific epigenetic alterations and the considerable biostability of DNA methylation, the capacity of methylation as a biomarker has been strongly implicated [11]. The evolutionary conserved gene pituitary homeobox 2 (PITX2) encodes for transcription factors involved in pattern formation, genesis of several organs (e.g. heart, lungs, pituitary gland), and the determination of left-right asymmetry during embryonic development [12,13]. Mutations in PITX2 lead to Axenfeld-Rieger syndrome, which is associated with malformations of the anterior segment of the eye [14,15]. In humans, three predominant PITX2 isoforms transcribed from two alternative promoters sites (P1 and P2) have been identified [16,17] (Fig 1). The variants PITX2a and PITX2b are alternatively spliced transcripts originating from promoter P2, which is known to be regulated by the WNT pathway [18]. The PITX2c variant is transcribed from an alternative promoter P1, which is regulated by TGF-β family members [19]. Emerging evidence suggests an oncogenic role of PITX2 [20,21], and DNA methylation of the PITX2 gene locus has been established as a prognostic biomarker in various human malignancies. DNA methylation of the PITX2ab promoter region has been associated with outcome in lung cancer [22] and hormone-receptor positive breast cancer patients [17,[23][24][25]. Furthermore, PITX2ab and PITX2c methylation is associated with biochemical recurrence (BCR) in patients with prostate cancer after radical prostatectomy [26][27][28][29].

Ethical Approval
The present study has been approved by the Institutional Review Board of the University Hospital Bonn, which waived the need for written informed consent from the participants.

Patients and Sample Collection
The patient cohort included 80 individuals who had undergone surgical resection for gallbladder carcinoma, intrahepatic cholangiocarcinoma (CC), perihilar CC, and extrahepatic CC at the Department of General, Visceral, Thoracic, and Vascular Surgery, University Hospital Bonn between 12/1989 and 05/2014 (clinicopathological parameters are summarized in Table 1). Archival formalin-fixed and paraffin-embedded tissue (FFPET) blocks were obtained from each patient and representative hemotoxylin and eosin (HE) slides taken from each tumor. Subsequently, an area of adenocarcinoma and if available a region with entirely normal adjacent tissue (NAT) was identified by experienced gastrointestinal pathologists at the Institute of Pathology, University of Bonn. A semi-automatic Tissue Microarrayer instrument with 0.6 mm stainless steel needle punchers was used to obtain tissue from FFPET blocks. Two punch biopsies of 0.6 mm diameter were taken from the most representative area of both tumor and NAT tissue.

Sample and Calibrator DNA Preparation
FFPET cores were deparaffinized, lysed, and bisulfite-converted using the innuCONVERT Bisulfite All-In-One kit (Analytik Jena) according to the manufacturer's recommendations [34]. The DNA concentration was determined by UV spectrophotometry using 33 as multiplication factor for single stranded bisulfite DNA (Nanodrop1 ND-1000 spectral photometer [Thermo Scientific]). A calibrator sample of bisulfite-converted artificially methylated human DNA was prepared as described previously [22].
PCR (qMSP). qMSP primers cover CpG-sites and specifically amplify methylated alleles. The quantification of total DNA was achieved with primers targeting a locus that does not contain CpG-sites and therefore amplifying alleles irrespective of the methylation status. In contrast, PANCR methylation was determined using a quantitative methylation (QM) assay with primers that do not contain CpG-sites within the target region and therefore amplify methylated as well as unmethylated sequences. The CpG-sites located in between the primer binding sites are probed with two hydrolysis probes which specifically and competitively detect methylated or unmethylated alleles, respectively. The relative DNA methylation of PITX2ab, PITX2c, and PANCR was quantified in triplicate qPCR reactions using an AB 7500 Fast Real-Time PCR System (Life Technologies Corporation). The PITX2ab-assay was conducted as previously described [35]. The amplification was carried out using following temperature profile: 15 min at 95°C of initial denaturation followed by 50 cycles with 75 sec at 60°C (for PITX2c-assay) or 60 sec at 56°C (for PANCR-assay; for both 100% ramp rate), and 15 sec at 95°C (75% ramp rate).

Analyzed Genomic Locations within PITX2 Gene Region and Analytical Assay Performance
DNA methylation within the PITX2 gene and its adjacent ncRNA PANCR were determined at three distinct genomic locations: promoters of the transcript variants PITX2a/b and PITX2c as well as in an intragenic GC-rich region located in an annotated promoter flanking site within the PANCR gene body (Fig 1). The analytical performance of the PITX2c and PANCR DNA methylation assays were verified using mixtures of methylated and unmethylated bisulfite-converted DNA from sperm, which is known to be unmethylated at many loci [37]. A plot of measured methylation values vs. amount of methylated DNA input (100% -0%) demonstrated high linearity for both PITX2c-(R 2 = 0.99) and PANCR-assay (R 2 = 0.91). The assays were sensitive down to 0.8% methylated DNA (equivalent to 40 pg) in a total of 5 ng bisulfite-converted DNA (Fig 2). The analytical performance of the PITX2ab-assay has been described in detail earlier [35].

Association of PITX2 and PANCR DNA Methylation with Clinicopathological Parameters
With respect to clinicopathologic parameters, only PITX2c methylation differed significantly depending on tumor location (extrahepatic CC vs. gallbladder [p = 0.029] and vs. intrahepatic CC [p = 0.015], respectively) ( Table 1). Additional statistically significant associations were not found.

Association of PITX2 and PANCR DNA Methylation with Overall Survival
As PITX2 and PANCR methylation were increased in tumor tissue, their potential as prognostic factor for OS was investigated by means of univariate Cox regression and Kaplan-Meier analysis.
In addition, differences in OS outcome were revealed for the clinicopathological parameters nodal status (pN) and surgical margin (R) in univariate Cox regression analysis ( Table 2). In multivariate Cox regression analysis, however, PITX2c DNA methylation lost its prognostic power as opposed to PANCR methylation and the known predictive parameters pN and R.

Discussion
BTC is a rare but clinically challenging disease with a poor prognosis. Due to non-specific symptoms in approx. 90% of patients presenting with BTC, the disease is often diagnosed in a locally advanced or metastatic and unresectable state. Complete surgical resection is the only curative treatment and applicable in only 10% of patients with early stage disease [38]. Even patients with localized BTC, who have undergone resection with intent to cure, have a high risk for disease recurrence [39]. A combination chemotherapy of gemcitabine and cisplatin is the standard-of-care for advanced and recurrent BTCs. This regimen, however, only has a small survival benefit compared to untreated patients with overall survival rates and progression-free survival of approx. 12 months and 8 months, respectively [40].
The present study demonstrated aberrant DNA hypermethylation of PITX2 promoter regions in tumor specimens of BTC compared to normal tissue. Furthermore, patients with PITX2c promoter hypermethylation revealed a significantly reduced OS, albeit OS did not qualify as an independent prognostic factor. This observation, even if only detected in univariate analysis in our small cohort, might be supported by a report which demonstrated that PITX2 is functionally involved in the suppression of pancreatic cancer progression. In pancreatic ductal adenocarcinoma (PDAC), the TGF-β signaling factor SMAD4 was shown to stimulate PITX2 expression, and a strong immunohistochemical PITX2 staining was reported to be associated with a better prognosis, whereas no correlation with methylation was observed [41]. However, in this study methylation was not measured in PITX2 promoter regions. Nevertheless, it is generally assumed that hypermethylation of a promoter region leads to transcriptional gene silencing [42,43]. Vice versa, low methylation would lead to a normal level of gene expression. The finding that patients with improved survival revealed PITX2c promoter hypomethylation might be caused by elevated PITX2c expression compared to patients of the hypermethylation group. Furthermore, PITX2c promoter is known to be regulated by TGF-β signaling [19], and SMAD4 protein binding was also reported for PDAC [41]. As biliary tract and pancreatic lesions have similar molecular and histological features [44], it might be admissible to transfer these conclusion to BTC. This theory will ultimately have to be proven by future investigations of expression levels though.
In contrast to our findings on PITX2c, PANCR hypermethylation revealed a significantly improved OS and retained its predictive power in univariate as well as multivariate Cox proportional hazard analysis. However, the PANCR DNA methylation was examined adjacent to a CpG island within an intragenic region instead of a promoter region. A positive correlation of hypermethylated gene body with an increased gene expression has been reported by several groups [45][46][47]. One might infer that PANCR expression is elevated as well as PITX2c; besides, a positive correlation of expression was observed in human adult left atria tissue [30].
In spite of the presented possible biomarker capabilities, it becomes apparent that aberrant DNA methylation of PITX2 variants is not as prognostic for BTCs as it is for breast or prostate cancer [17,35]. In the latter, these markers may identify patients with an adverse clinical courses who might actually benefit from a radical treatment, while patients with a better prognosis might benefit from a more conservative treatment with fewer side effects. Due to delayed diagnoses and a generally poor prognosis of BTC, all patients included actually underwent a radical therapy (i.e. complete surgical resection, chemotherapy). For clinical decision-making, predictive biomarkers are desired to identify patient subpopulations that most likely benefit from chemotherapy or molecular targeting therapy. Last but not least, a basic science rationale for studying PITX2 and PANCR methylation was to gain new insight into the biology of BTC in order to find novel therapeutic implications in this lethal disease. The study has some potential limitations: The low incidence of BTC allows the inclusion of only a small cohort that was analyzed in a retrospective fashion, clearly limiting the power of statistical analysis as well as lowering the evidence level. Even if the biliary ductal system shares one embryonic background, previous studies have documented distinct histomorphological and molecular patterns referring to the location of the lesion within the biliary tree [5,6]. In addition, the present study does not unveil the biological significance of PITX2 and PANCR methylation during biliary tract carcinogenesis and disease progression. An association between DNA methylation of the analyzed loci with mRNA and protein expression of the respective gene products need to be conducted in order to elucidate any direct epigenetic control of the genes via DNA methylation. Furthermore, cell culture experiments with hypomethylating agents (i.e. 5-azacytidine) should be performed in order to investigate the functional role of PITX2 and PANCR. Eventually, the potential value of PITX2 and PANCR methylation as predictive biomarkers for particular therapies (i.e. targeted therapies, epigenetic therapies, immunotherapies, chemotherapies) needs to be tested in preclinical models.

Conclusions
Assessment of PITX2 and PANCR methylation in BTC revealed a possible new biomarker implication to help differentiate tumors from normal tissue as well as to identify distinct biologic behaviours of individual tumors for a possible future therapeutic stratification. The data have to be evaluated on a prospective level. Herein, an appropriate stratification into BTC subtypes seems inevitable.
Writing -review & editing: GK VB AS PS BR PL NS HG.