Hepatocyte Nuclear Factor (HNF) 4 [alpha] Expression Distinguishes Ampullary Cancer Subtypes and...

ORIGINAL ARTICLE

Hepatocyte Nuclear Factor (HNF) 4α Expression DistinguishesAmpullary Cancer Subtypes and Prognosis After Resection

Florian Ehehalt, MD,∗ Petra Rummele, MD,† Stephan Kersting, MD,∗ Corinna Lang-Schwarz, MD,†Felix Ruckert, MD,∗ Arndt Hartmann, MD,‡ Wolfgang Dietmaier, MD,† Luigi Terracciano, MD,§

Daniela E. Aust, MD,¶ Beatrix Jahnke,∗ Hans-Detlev Saeger, MD,∗ Christian Pilarsky, PhD,∗ andRobert Grutzmann, MD∗

Objective: To investigate biological differences and prognostic indicators ofdifferent ampullary cancer (AC) subtypes.Background: AC is associated with a favorable prognosis compared withother periampullary carcinomas. Aside from other prognostic factors, thehistological origin of AC may determine survival. Specifically, the pancreato-biliary subtype of AC displays worse prognosis compared with the intestinalsubtype. However, knowledge of inherent molecular characteristics of differ-ent periampullary tumors and their effects on prognosis has been limited.Methods: Gene expression profiling was used to screen for differential geneexpression between 6 PDAC cases and 12 AC cases. Among others, hepatocytenuclear factor 4α (HNF4α) mRNA overexpression was observed in AC cases.Nuclear HNF4α protein expression was assessed using tissue microarraysconsisting of 99 individual AC samples. The correlation of HNF4α expressionwith clinicopathological data (n = 99) and survival (n = 84) was assessed.Results: HNF4α mRNA is 7.61-fold up-regulated in AC compared with that inPDAC. Bioinformatics analyses indicated its key role in dysregulated signalingpathways. Nuclear HNF4α expression correlates with histological subtype,grading, CDX2 positivity, MUC1 negativity and presence of adenomatouscomponents in the carcinoma. The presence of HNF4α is a univariate predictorof survival in AC mean survival (50 months versus 119 months, P = 0.002).Multivariate analysis revealed that HNF4α negativity (HR = 17.95, 95% CI:2.35–136.93, P = 0.005) and lymph node positivity (HR = 3.33, 95% CI:1.36–8.18, P = 0.009) are independent negative predictors of survival.Conclusions: Immunohistochemical determination of HNF4α expression isan effective tool for distinguishing different AC subtypes. Similarly, HNF4α

protein expression is an independent predictor of favorable prognosis in carci-noma of the papilla of Vater and may serve for risk stratification after curativeresection.

(Ann Surg 2011;254:302–310)

O riginating at the border between the duodenal and pancreato-biliary epithelium, ampullary adenocarcinomas (ACs) (ie, ade-

nocarcinomas of the papilla of Vater) display a favorable prognosis

From the *Department of General, Thoracic and Vascular Surgery, UniversityHospital Dresden, Germany, †Department of Pathology, University HospitalRegensburg, Germany, ‡Department of Pathology, University Hospital Erlan-gen, Germany, §Department of Pathology, University Hospital, Basel, Switzer-land and ¶Department of Pathology, University Hospital Dresden, Germany.

F. Ehehalt and P. Rummele contributed equally to this work.C. Pilarsky and R. Grutzmann shared senior authorship.Supported by the MedDrive Program (Faculty of Medicine, University of Technol-

ogy Dresden).Supplemental digital content is available for this article. Direct URL citations

appear in the printed text and are provided in the HTML and PDF versions ofthis article on the journal’s Web site (http://links.lww.com/SLA/A136).

Reprints: Robert Grutzmann, Fetscherstraße 74, 01307 Dresden, Robert. E-mail:[email protected].

Copyright C© 2011 Wolters Kluwer Health | Lippincott Williams & WilkinsISSN: 0003-4932/11/25402-0302DOI: 10.1097/SLA.0b013e31821994a8

compared with other periampullary tumors (eg, ductal adenocarci-noma of the pancreatic head and carcinoma of the common bileduct). This prognostic difference can be partially attributed to theearly presentation of obstructive jaundice and easy diagnosis by up-per gastrointestinal endoscopy and simultaneous histological verifi-cation. However, some data suggest that tumor-inherent biologicaldifferences may determine the prognostic superiority of AC.1–5

Frequently, AC contains residues of a preexisting adenomatouscomponent. Although mechanistic information of tumor progression(eg, genetic alterations and epigenetic modification) is limited, thepresence of an adenoma–carcinoma sequence similar to that in col-orectal cancer is repeatedly hypothesized.6–11

Different AC subtypes can be distinguished by histomorphol-ogy, for example, pancreatobiliary, intestinal, intestinal-mucinous,invasive-papillary, and poorly differentiated adenocarcinomas.12 Theintestinal and intestinal-mucinous subtypes display a favorable prog-nosis compared with pancreatobiliary, invasive-papillary and poorlydifferentiated carcinomas. Coarse histological subtype referring tothe epithelial origin of the tumor (ie, intestinal vs. pancreatobiliary)is likely to be one of the most important prognostic factors of overallsurvival (OS) in resected AC.13,14

Histological subtype can be determined by morphology andoptional immunohistochemistry. Pancreatobiliary subtypes preferen-tially exhibit positive immunoreactions for cytokeratins (CKs) 7 and17 and apomucins (MUC) 1 and 5a, whereas intestinal subtypes ex-press CK20, MUC2, and caudal homeobox gene transcription factor(CDX) 2.15–19

Aside from the histological subtype, other independent nega-tive prognostic markers for OS have been reported in different cohorts(eg, age, CDX2 negativity, high UICC stage, lymph node invasion,high histological grade, and microsatellite stability).5,15,20,21

The primary scope of this work was to identify candidate genesthat are differentially expressed between pancreatic ductal adenocar-cinoma (PDAC) and AC to shed light on the differences in the tumorbiology of these both entities. Among the 630 gene products foundto be differentially expressed, bioinformatics analyses revealed thatHNF4α overexpression is a hub of dysregulated intracellular signal-ing pathways in AC, making it an interesting candidate for furtherevaluation. Finally, immunohistochemical analysis of nuclear HNF4αexpression in 99 AC cases using tissue microarrays (TMAs) produceda data set for evaluating interrelations among histological subtype,nuclear HNF4α expression, and survival.

METHODS

Patient Characteristics and Tissue Sampling forGene Expression Profiling

Freshly frozen tissue samples of PDAC (n = 6) and tumorsof the ampulla of Vater (n = 12) were obtained from surgical spec-imens from patients who underwent surgery at the Department ofGeneral, Thoracic, and Vascular Surgery, University Hospital Carl

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

302 | www.annalsofsurgery.com Annals of Surgery � Volume 254, Number 2, August 2011

Annals of Surgery � Volume 254, Number 2, August 2011 HNF4α Expression and Ampullary Carcinoma

Gustav Carus, Technical University of Dresden (Germany), and theDepartment of Surgery at the University of Regensburg (Germany)between 2000 and 2007. Clinicopathological data from these patients(ie, gender, age at diagnosis, histological subtype, and pTNM sta-tus) are shown in the supplementary data (see Supplemental DigitalContent 1: Supplementary Data Table 1, available at: http://links.lww.com/SLA/A136).

Informed consent, approved by local ethics committees, wasobtained from all patients before surgery. Immediately after surgical

removal, the specimens were sectioned and macroscopically evalu-ated. Suitable samples of tumor tissue and normal tissue were snap-frozen in liquid nitrogen and stored at − 80◦C until further process-ing.

Gene Expression ProfilingMicrodissection, RNA isolation, and array hybridization were

performed according to Grutzmann et al.22 with minor changes.Briefly, total RNA was prepared from microdissected samples using

TABLE 1. List of 100 Differentially Expressed mRNAs (50 Most Down- and 50 Most Up-regulated) in AC Compared to PDAC

Up-regulated mRNA Down-regulated mRNA

Affymetrix ID Gene Code Fold Change Affymetrix ID Gene Code Fold Change

213953 at KRT20 132.07 204748 at PTGS2 − 13.96206149 at CHP2 27.23 210794 s at MEG3 − 9.63206430 at CDX1 19.21 205666 at FMO1 − 6.80210390 s at CCL14 /// 15 17.91 210119 at KCNJ15 − 6.15232428 at MOGAT2 16.57 221009 s at ANGPTL4 − 6.12240509 s at GREM2 13.6 213524 s at G0S2 − 6.01232321 at MUC17 13.38 202627 s at SERPINE1 − 5.80217238 s at ALDOB 12.16 232113 at — − 5.42205929 at GPA33 12.06 237151 s at hCG 17324 − 5.24206312 at GUCY2C 11.65 206336 at CXCL6 − 5.22211916 s at MYO1A 10.7 232544 at — − 5.20225667 s at FAM84A 10.69 203868 s at VCAM1 − 5.14229215 at ASCL2 10.61 229170 s at TTC18 − 4.86231734 at RBP2 10.57 230341 x at ADAMTS10 − 4.83238103 at — 10.14 213906 at MYBL1 − 4.80204705 x at ALDOB 10.07 219873 at COLEC11 − 4.76210159 s at TRIM31 9.49 241866 at SLC16A7 − 4.751552870 s at C1orf125 9.39 223395 at ABI3BP − 4.55231632 at — 9.11 202628 s at SERPINE1 − 4.53208383 s at PCK1 8.87 212977 at CXCR7 − 4.48243669 s at PRAP1 8.59 219519 s at SIGLEC1 − 4.48217207 s at BTNL3 8.39 219737 s at PCDH9 − 4.341552281 at SLC39A5 7.65 229307 at ANKRD28 − 4.30230914 at HNF4A 7.61 207057 at SLC16A7 − 4.27219545 at KCTD14 7.48 203180 at ALDH1A3 − 4.24233604 at FLJ22763 7.1 223467 at RASD1 − 4.23207202 s at NR1I2 7.07 206115 at EGR3 − 4.15222257 s at ACE2 6.93 243332 at — − 4.131554436 a at REG4 6.92 205328 at CLDN10 − 4.08233059 at — 6.83 244026 at — − 4.0741469 at PI3 6.79 242915 at ZNF682 − 4.02236496 at DEGS2 6.74 219550 at ROBO3 − 3.97205506 at VIL1 6.73 208037 s at MADCAM1 − 3.94205311 at DDC 6.7 206315 at CRLF1 − 3.93218756 s at MGC4172 6.62 227399 at VGLL3 − 3.90229158 at WNK4 6.61 223533 at LRRC8C − 3.89203953 s at CLDN3 6.57 202436 s at CYP1B1 − 3.87211357 s at ALDOB 6.47 223741 s at TTYH2 − 3.82203691 at PI3 6.39 236222 at C3orf15 − 3.82229889 at C17orf76 6.36 204150 at STAB1 − 3.78228912 at VIL1 6.35 229947 at PI15 − 3.76207158 at APOBEC1 6.28 223121 s at SFRP2 − 3.74219962 at ACE2 6.10 202435 s at CYP1B1 − 3.73217073 x at APOA1 6.08 222162 s at ADAMTS1 − 3.72215444 s at TRIM31 6.06 238919 at — − 3.72223597 at ITLN1 5.89 1556039 s at GPR173 − 3.71211603 s at ETV4 5.85 202992 at C7 − 3.7058916 at KCTD14 5.78 203066 at GALNAC4S − 3.651553559 at TMEM171 5.66 215388 s at CFH /// CFHR1 − 3.65205517 at GATA4 5.57 204604 at PFTK1 CFHR2 − 3.61236055 at DQX1 5.52 206910 x at CFHR2 − 3.6

Italic typeface highlights: cytokeratin 20 (gene symbol KRT20) as the most over-expressed mRNA in AC and HNF4α with a 7.61-fold overexpression in AC.


C© 2011 Lippincott Williams & Wilkins www.annalsofsurgery.com | 303

Ehehalt et al Annals of Surgery � Volume 254, Number 2, August 2011

an RNeasy Mini Kit (Qiagen, Hilden, Germany) according to themanufacturer’s instructions. The quality of the prepared RNA wascontrolled using an Agilent RNA 6000 Pico Kit (Agilent, Boblingen,Germany). RNA samples with a concentration >3 ng/μL and a RINvalue >4 were subjected to further analysis. From each sample, 100 ngof RNA was used for cDNA synthesis and in vitro transcription-basedamplification using an Affymetrix GeneChip Two-Cycle Target La-beling and Control Kit (Affymetrix, Santa Clara, USA). Hybridiza-tion and detection of labeled aRNA on the Affymetrix GeneChipHuman Genome U133 Plus 2.0 (Affymetrix) were performed as rec-ommended by the manufacturer. Affymetrix raw probe level intensity(CEL) files were background-corrected, normalized, summarized, andanalyzed using the software dChip.23

Raw data are available via ArrayExpress accession: E-MEXP-2894 (specified release date: December 12, 2010).

To identify relevant signal transduction pathways, the resultinglist of differentially expressed genes was evaluated using the PathwayAssist software (Agilent).

Patient Characteristics and Tissue Sampling forImmunohistochemistry

Between 1985 and 2007, tissue samples from 99 margin-negative resected carcinoma cases with unequivocal origin from theanatomic structures forming the ampulla of Vater were obtained bysurgical resection. Formalin-fixed, paraffin-embedded tissue blocks(containing representative tumor tissue) were retrieved from thearchives of the Institute of Pathology at the University of Regensburg(Germany), the Institute of Pathology (Nurnberg, Germany), the Insti-tute of Pathology (Kassel, Germany), the Anatomic Pathology Unit inthe Department of Human Morphology at the University of Insubria,Varese (Italy), the Department of Pathology at the University of Kiel(Germany), the Institute of Pathology at the University of TechnologyDresden (Germany), and the Institute of Pathology at the UniversityHospital Basel (Switzerland). All tumors were staged according totumor-node-metastases (TNM) classification for AC (UICC 2002).

The histology of all tumors was reviewed by 3 experiencedpathologists (L.T., P.R., and D.E.A.) without knowledge of clinicaland molecular data, as previously described.5 Briefly, the histopatho-logic tumor type, the presence of coexisting adenomas, the histo-logical grade, and the percentage of mucinous tumor differentiationwere noted during review. All of the AC cases were histologicallyclassified according to the criteria published by Albores-Saavedraet al.24 In addition, the intestinal-mucinous subtype was defined andcharacterized by the presence of mucinous differentiation. Grade wasassigned in accordance with the World Health Organization Classi-fication of Tumors of the Digestive System (2000). Patient data, in-cluding follow-up information, were obtained through retrospectivereview of medical records and the regional clinical tumor registries, orby direct contact of the referring physicians. This study was approvedby the Review Boards of the institutions involved.

TMA ConstructionFor immunohistochemical analyses, a total of 6 arrays with

up to 529 cores were constructed from formalin-fixed, paraffin-embedded archival tumor tissues using a semi-automated tissuearrayer25,26 or a manual device (Beecher Instruments, Silver Springs,MD, USA) as previously described.27 Briefly, hematoxyline/eosin-stained slides of all tumors were evaluated by 1 pathologist (P.R.)with subspecialty expertise in gastrointestinal pathology; representa-tive regions were chosen from each sample. For each tumor sampleand the corresponding non-neoplastic intestinal or gastric mucosasample, a tissue core with a diameter of either 0.6 or 1.6 mm was

used (1–4 cores for each case). Subsequently, immunohistochemicalstudies were performed on 4 μm tissue sections of TMAs.

Immunostaining for MUC1, CK7, CK20, CDX2, andP1-HNF4α

Immunohistochemistry was performed for protein markersMUC1, CK7, CK20, and CDX2. As described previously, 4 μmformalin-fixed paraffin-embedded tissue sections of the TMAs weredeparaffinized and immunostained after microwave- or water bath-based antigen retrieval using the IView DAB Detection Kit on aVentana Nexes autostainer (Ventana, Tucson, AZ, USA) follow-ing manufacturer’s instructions28 or manually by standard indirectimmunoperoxidase procedures29 (ABC-Elite, Vector Laboratories,Burlingame, CA, USA; K4005, EnVision + System-HRP (AEC);DakoCytomation; Bond Vision Biosystems). The following primaryantibodies were used: MUC1 (1:4000; clone MA695; Novocastra,Newcastle, UK), CK7 (1:200; clone OV-TL 12/30, DAKO, Baar,Switzerland), CK20 (1:50; clone IT-Ks 20.8, DAKO); and CDX2(1:50; clone AMT28, Abcam, Cambridge, UK). P1-HNF4α expres-sion was evaluated using a mouse monoclonal antibody against P1-HNF4α (1:50; clone K9218, Abcam). Staining results were indepen-dently scored by 2 pathologists (P.R. and C.L.S.). Cytoplasmic and/ormembranous CK20 and CK7 and nuclear CDX2 and HNF4α werescored as the number of positive tumor cells. The labeling was semi-quantitatively graded according to both, the extent and the intensityof staining. A cutoff of >10% positive cells displayed at least weakstaining was used for all investigated proteins to consider a tumorpositive. Normal intestinal and gastric samples served as positivecontrols.

StatisticsStatistical correlation among clinicopathologic, morphologic,

and immunohistochemical staining results was analyzed using a 2-sided exact Fisher test or a χ 2 test as appropriate. Interobserver relia-bility of TMA data (HNF4α, MUC1, and CDX2) was estimated by κtesting. OS time was calculated according to Kaplan-Meier survivalcurves. OS was measured from the time of surgery and was comparedusing the 2-sided log-rank analysis. Differences were considered tobe significant if P < 0.05.

A Cox regression model was adjusted, testing the independentprognostic relevance of nuclear HNF4α presence. All univariatly sig-nificant parameters and age were included in the starting model. Agewas included to this model because it exhibited independent prognos-tic relevance in previous studies.15 Using backward elimination andlog-likelihood adjustment, parameters were retained if P values were<0.05 and removed if P values were >0.1.

The stability of the model was assessed by a nonparametricbootstrap resampling technique. This was used to estimate the likeli-hood that each of the final model variables would have been selectedin repeated samples.

All statistical analyses were conducted using SPSS (version18.0) statistical software program (SPSS, Chicago, IL).

RESULTS

Gene Expression Analysis in PDAC and AC Revealed630 Differential mRNAs

Although early diagnosis based on obstructive symptoms maycontribute to better survival in AC, how much (molecular) biologicaldifferences can affect overall prognosis is still a matter of debate.Therefore, we screened for differentially expressed candidate genesusing microdissected tumor materials from 6 PDAC patients and 12AC patients [9 intestinal (-mucinous) and 3 pancreatobiliary/othersubtypes].


304 | www.annalsofsurgery.com C© 2011 Lippincott Williams & Wilkins


We identified a total of 630 differentially expressed mRNAs,of which 300 were significantly down-regulated (≥2-fold) and 330were significantly up-regulated (≥2-fold) in AC compared with theirexpression in PDAC. Table 1 displays a list of the 50 most up-regulatedand 50 most down-regulated mRNAs.

As expected, the mRNA of CK20, a well-known marker ofintestinal subtype, was the most up-regulated gene product in ACs (a132-fold change). Although the arrays in use covered the majority ofthe known histological subtype markers in AC (MUCs 1 and 2; CKs7, 17, and 20; and CDX2), none of these markers, except CK20, wasfound differentially expressed between AC and PDAC.

Interestingly, HNF4α mRNA was strongly up-regulated (7.61-fold) in AC.

HNF4α is Central to DysregulatedPathways in AC

For data organization and visualization of regulatory gene net-works within the 630 differentially expressed gene products, we used“PathwayAssist” software.

Notably, in AC, HNF4α was evaluated as a shared regu-lator of a considerable number of differentially expressed genes(Fig. 1).

Taken together, the results of gene expression profiling andsubsequent bioinformatics analysis suggest that HNF4α is a key dys-regulated gene product in AC, and it may have significant effectson the tumor biology and oncologic behavior of these cancers. Totest this hypothesis, immunohistochemistry of P1-HNF4α (promoter

FIGURE 1. HNF4α-centered cut-out from the “PathwayAssist” calculation result of dysregulated signaling pathways based on the630 differentially expressed gene products identified by GeneChip analysis. Red and green colors refer to down- and up-regulatedgene products in AC compared with PDAC, respectively. The scheme illustrates that HNF4α is the hub of a remarkable numberof differentially expressed genes.




FIGURE 2. Immunohistochemical staining of CDX2,MUC1, CK 7, and CK20 in AC subtypes. A, Pancreato-biliary subtype with cytoplasmic MUC 1-positivity; B, In-testinal subtype with nuclear CDX2-positivity and charac-teristic cytokeratin pattern; C, CK20 positive and (D) CK7negative. Immunohistochemical staining of P1-HNF4αin normal duodenal mucosa and ampullary cancer sub-types: A nuclear positive reaction is always seen in thenuclei of normal duodenal epithelial cells (E; crypts andvilli) and in the majority of adenomatous components(F), intestinal (G) and intestinal-mucinous cancer (H) sub-types. In contrast, nuclear negative poorly differentiated(I) and pancreatobiliary (J) carcinomas. Original magnifi-cation 100 × (A–J).




P1 driven isoforms 1–6) was performed on a large group of well-characterized AC samples using TMAs.

HNF4α Expression Correlates with Grading,Histological Subtype, and the Presence ofPreexisting Adenomatous Components

For 99 AC specimens, TMA data of P1-HNF4α were obtained.For a random subset of this cohort, immunostaining against CDX2(n = 78), MUC1 (n = 77), CK7 (n = 78), and CK20 (n = 79) wasperformed (Fig. 2).

Univariate analysis revealed significant correlation of positivenuclear P1-HNF4α staining with intestinal and intestinal-mucinoushistological subtypes (n = 99, χ2 = 17.6, P = 0.001), the presence ofpreexisting adenomatous components (n = 99, χ 2 = 6.0, P = 0.015),CDX2 positivity (n = 78, χ 2 = 11.3, P = 0.002), MUC1 negativity(n = 77, χ 2 = 7.3, P = 0.008), and low histological grade (n = 99,χ 2 = 8.3, P = 0.016, Table 2).

CK7 (n = 78, χ 2 = 0.3, P = 0.54) and CK 20 (n = 79, χ 2

= 1.7, P = 0.79) immuno-reactivity, gender (n = 99, χ2 = 0.4, P =0.64), T-stage (n = 99, χ 2 = 3.3, P = 0.34), nodal involvement (n =99, χ 2 = 0.18, P = 0.82), and UICC-stage (n = 99, χ 2 = 9.1, P =0.11) were not associated with nuclear P1-HNF4α positivity.

Only 3 of 86 evaluated individual PDACs (3.5%) stained posi-tive for nuclear P1-HNF4α (data not shown). Therefore, PDACs wereexcluded from any further analysis.

Furthermore, interobserver reliability of immunohistochemi-cal data was highest for HNF4α (κ 0.78; 92% concordance) comparedwith CDX2 (κ 0.064; 55% concordance) and MUC1 (κ 0.67; 86%concordance).

Survival Analysis in ACComplete follow-up was available for 84 of 99 specimens and

univariate survival analysis was performed on this subset of patients.The median age at diagnosis was 65 years (ranging from 45 to 79years) with a mean follow-up of 41.3 months. Five-year survivalfor all patients was 41.8%, similar to previous survival rates aftercomplete resection (eg, 36.8% for a total of 1301 resected patients ina population-based analysis).30

The univariate survival analysis of all resected AC patientsrevealed significant differences in nuclear HNF4α expression, his-tological subtype (grouped for intestinal (-mucinous) ACs vs. pan-creatobiliary, poorly differentiated, and invasive-papillary ACs), pTstage (grouped for pT1/2 vs. pT3/4), nodal involvement, histologi-cal grade (grouped in G1/2 vs. G3), UICC stage (grouped in UICCIA/IB/IIA vs. IIB vs. III/IV), and the presence of adenomatous com-ponents within the carcinoma. The subgroup analysis of intestinal(-mucinous) ACs (n = 55) showed a significantly favorable survival inthe presence of nuclear HNF4α expression in these tumors. Variable-and group-specific Kaplan-Meier curves, mean survival time and Pvalues given by the log-rank analysis are shown in Fig. 3 and in thesupplementary data (see Supplemental Digital Content 2: Supple-mentary Data Fig. 1, available at: http://links.lww.com/SLA/A136).

CDX2, MUC1, CK7, and CK20 immunostaining were per-formed on a random subset of specimens, of which the univariate anal-ysis revealed that CDX2 and MUC1 were predictors for survival (forKaplan-Meier estimates, see Supplemental Digital Content 2: Supple-mentary Data Fig. 1, available at: http://links.lww.com/SLA/A136),whereas CK7 and CK20 were not associated with OS (data notshown).

TABLE 2. Univariate Analysis of Clinical and Pathological Characteristics of AC Patients in Correlation withNuclear HNF4α Expression (P values According to 2-Sided χ2 Test or Fisher’s Exact Test asAppropriate)

Variable Characteristics n P1-HNF4α positive P1-HNF4α negative P

Histological subtype (n = 99) Intestinal 48 18 30 0.001Intestinal-mucinous 11 6 5Pancreatobiliary 19 2 17Poorly differentiated 14 0 14Invasive-papillary 7 0 7

Gender (n = 99) Male 62 15 47 0.638Female 37 11 26

T stage (n = 99) pT1 11 5 6 0.344pT2 35 10 25pT3 42 8 34pT4 11 3 11

Histological grade (n = 99) G1 5 4 1 0.016G2 45 12 33G3 49 10 39

Lymph node(n = 99) pN0 46 13 33 0.423

pN1 53 13 40UICC stage (n = 99) IA 11 5 6 0.105

IB 18 5 13IIA 13 2 11IIB 45 10 35III 10 2 8IV 2 2 0

Tumor entity (n = 99) Carcinoma w/o adenoma 47 7 40 0.02Carcinoma with adenoma 52 19 33

CDX 2 (n = 78) Negative 46 4 42 0.002Positive 32 13 19

MUC 1 (n = 77) Negative 54 18 36 0.008Positive 23 1 22




FIGURE 3. Kaplan-Meier estimates and univariate analysis of the overall survival of AC patients (n = 84) according to: A, nuclearHNF4α expression in all AC cases (n = 84); B, subgroup analysis of intestinal and intestinal-mucinous ACs (n = 55); C, histologicalsubtype (intestinal vs. pancreatobiliary, poorly differentiated and invasive-papillary, n = 84); D, UICC stage (UICC IA/IB/IIA vs. IIBvs. III/IV). P, P values according to log-rank test; n, initial number at risk.

For multivariate analysis, the following parameters were ap-plied to a Cox regression model (stepwise backwards, likelihoodratios): nuclear HNF4α expression (negative vs. positive), his-tological subtype (pancreatobiliary, invasive-papillary and poorly-differentiated ACs vs. intestinal and intestinal-mucinous ACs),T-stage (pT3/4 vs. pT1/2), nodal involvement (positive vs. negative),histological grade (G3 vs. G1/2), tumors without vs. with an adenoma-tous component, CDX2 (negative vs. positive), and MUC1 (positivevs. negative) immunostaining and age.

Negative nuclear HNF4α expression (HR = 17.95, 95% CI:2.35–136.93, P = 0.005) and lymph node positivity (HR = 3.33, 95%CI: 1.36–8.18, P = 0.009) were identified as independent predictorsof poor prognosis, whereas histological grade and T stage reached thefinal model but with nonsignificant P values. In fact, increasing agewas also independently associated with prognosis (HR = 0.95, 95%

CI: 0.91–1.0, P = 0.046), but its relevance was highly questionablebecause the resulting hazard ratio was close to 1 (Table 3).

To assess the stability of the final Cox proportional hazardmodel, a bootstrap resampling technique of 1250 samples was per-formed with the parameters initially used. According to bootstrap-ping, HNF4α is the most stable variable in the before mentioned Coxproportional hazard model (HR 2.7, 95% CI: 1.4–14.8, P = 0.002(for details see Supplemental Digital Content 3: Supplementary DataTable 2, available at: http://links.lww.com/SLA/A136).

DISCUSSIONBecause completely resected AC displays a favorable prog-

nosis compared with carcinoma of the distal common bile ductand ductal carcinoma of the pancreatic head, intrinsic molecular




TABLE 3. Final model after multivariate analyses (stepwisebackwards Cox proportional hazard ratio) for univariatestatistically significant predictors of survival: the presence orabsence of adenomatous components within the ACs, grade(G1/2 vs. G3), histological subtype (pancreatobiliary,invasive-papillary, poorly differentiated vs. intestinal andintestinal mucinous), T stage (pT1/2 vs. pT3/4), nodalinvolvement (pN0 vs. pN1), increasing age, nuclear HNF4α,CDX2 and MUC1 expression (positive vs. negative) in allavailable AC cases (n = 58) were applied to the model.

Variables HR 95% CI P

Step 5HNF4α negativity 17.95 2.35 136.93 0.005Grade (G3 vs. G1/2) 2.01 1.00 4.06 0.051T stage (pT3/4 vs. pT1/2) 2.09 0.87 5.05 0.100Nodal involvement positivity 3.33 1.36 8.18 0.009Increasing age 0.95 0.91 1.00 0.046

differences affecting the oncological behavior of these tumors havebeen a matter of long-standing debate.

By definition, AC originates from the junction of the pan-creatobiliary and intestinal epithelium and, thus, is subclassified ac-cording to histomorphology that partially relates to epithelial origin.Although some researchers doubt the independent prognostic rele-vance of histological subtype (ie, pancreatobiliary vs. intestinal) andimmunohistological marker expression, in the majority of publishedcohorts, these subtypes have been associated (at least in univariateanalyses) with long-term survival.11,12,15,19,31 Remarkably, intestinalsubtypes show comparable long-term survival with duodenal carci-noma (ca. 60%), whereas the pancreatobiliary subtypes just meet themean survival of pancreatic cancer patients (ca. 20%).13 Thus, in-testinal and pancreatobiliary tumors have been hypothesized to beseparate entities.

In this study, we analyzed the differential gene expression be-tween AC and PDAC, providing the first concise comparison of geneexpression between these entities. Among a huge group of identi-fied differentially expressed genes, bioinformatics analyses revealedHNF4α as a key indicator to molecularly distinguish AC from PDAC.Furthermore, because most of the analyzed AC samples were of in-testinal subtypes, this result suggests that very fundamental biologicaldifferences are present between tumors of pancreatic and intestinalorigins.

HNF4α is a member of the nuclear receptor family and is atranscription factor with multiple functions in different organ sys-tems. The most well-established function of HNF4α is its pivotal rolein β-cell function and the pathogenesis of maturity-onset diabetes ofthe young type 1 (MODY 1).32 It is also known to be a key factorfor the function of the intestinal epithelium.33 Furthermore, it is anupstream transcriptional regulator of CDX2 in the differentiation pro-cess of the intestinal epithelium in rodents and humans.34–36 Proteinexpression analysis in human pancreas has revealed that no pancreaticcell type expresses P1-HNF4α (K9218; P1 promoter-driven isoforms1–6), whereas positive staining was found for P2-HNF4α (H6939;P2 promoter-driven isoforms 7–9) in the majority of pancreatic celltypes. Thus, in humans, the P1 promoter-driven isoforms of HNF4αare likely to be differentiation markers of intestinal epithelium be-cause they are not expressed in pancreatic cells.37

In our cohort of AC patients, we quantified nuclear P1-HNF4α protein expression by high-throughput TMA immunostain-ing. Positive nuclear P1-HNF4α staining was significantly associatedwith intestinal differentiation, low grade, CDX2 positivity, MUC1

negativity, and histological detection of adenomatous componentswithin the carcinomas (ie, tumors of adenomatous origin).

In the univariate analysis, HNF4α-positive tumors clearly dis-played a favorable prognosis compared with HNF4α-negative tumors.Even in the subgroup of intestinal AC cases, HNF4α indicated a fa-vorable prognosis.

Furthermore, according to the cohort presented in this study,HNF4α negativity and lymph node involvement are independent fac-tors associated with poor long-term prognosis after margin-negativeresection. These results strongly indicate that P1-HNF4α is a markerfor intestinal differentiation and low tumor grade.

Conclusively, our data suggest that although the term “am-pullary carcinoma” refers to the anatomic site of tumor origin, thedifferences between intestinal and pancreatobiliary subtypes of ACmay meet the criteria of different tumor entities. The oncological dif-ferences between the pancreatobiliary and intestinal subtypes wereexamined by the following: (1) gene expression profiling of AC andPDAC samples; (2) protein expression comparisons between the sub-entities of AC; and (3) survival follow-up after margin negative resec-tion. Long-term prognosis strikingly differed between intestinal andpancreatobiliary ACs and was associated with the expression of P1-HNF4α, CDX2, and MUC1. In the multivariate investigated cohort,nuclear P1-HNF4α expression was independently associated withbetter survival and predominated other established prognostic factors(except lymph node involvement). Therefore, P1-HNF4α immunos-taining may serve as a powerful tool for both subclassification andrisk stratification of AC after radical resection. After validation of theprognostic and predictive value of P1-HNF4α status in different co-horts and interventional trials, P1-HNF4α immunostaining will likelyplay a role in establishing a concept of differentiated (neo-)adjuvanttherapy in AC.

ACKNOWLEDGMENTSWe thank Peter-Heinz Wuensch, MD (Institute of Pathology,

Nurnberg), Ernst Heinmoeller, MD (Institute of Pathology, Kassel),Guenter Kloeppel, MD (Department of Pathology, UniversityHospital, Kiel), Daniel Baumhoer, MD (Institute of Pathology,University Hospital, Basel, Switzerland), and Fausto Sessa, MD(Anatomic Pathology Unit, Department of Human Morphology, Uni-versity of Insubria, Varese, Italy) for supporting this study by provid-ing formalin-fixed, paraffin-embedded tissue blocks of AC-resectionspecimen. We thank Anne Pietryga-Krieger, Doris Gaag, and MonikaKerscher for excellent technical assistance. We acknowledge the ser-vice of the “American Journal Experts” (Durham, US) editing thepaper.

REFERENCES1. Bouvet M, Gamagami RA, Gilpin EA, et al. Factors influencing survival after

resection for periampullary neoplasms. Am J Surg. 2000;180(1):13–17.2. Chang MC, Chang YT, Tien YW, et al. Distinct chromosomal aberrations

of ampulla of Vater and pancreatic head cancers detected by laser cap-ture microdissection and comparative genomic hybridization. Oncol Rep.2005;14(4):867–872.

3. Esposito I, Friess H, Buchler MW. Carcinogenesis of cancer of the papillaand ampulla: pathophysiological facts and molecular biological mechanisms.Langenbecks Arch Surg. 2001;386(3):163–171.

4. Friess H, Guo XZ, Tempia-Caliera AA, et al. Differential expression ofmetastasis-associated genes in papilla of Vater and pancreatic cancer corre-lates with disease stage. J Clin Oncol. 2001;19(9):2422–2432.

5. Ruemmele P, Dietmaier W, Terracciano L, et al. Histopathologic features andmicrosatellite instability of cancers of the papilla of Vater and their precursorlesions. Am J Surg Pathol. 2009;33(5):691–704.

6. Stolte M, Pscherer C. Adenoma-carcinoma sequence in the papilla of Vater.Scand J Gastroenterol. 1996;31(4):376–382.

7. Chung CH, Wilentz RE, Polak MM, et al. Clinical significance of K-ras onco-gene activation in ampullary neoplasms. J Clin Pathol. 1996;49(6):460–464.




8. Howe JR, Klimstra DS, Cordon-Cardo C, et al. K-ras mutation in adenomasand carcinomas of the ampulla of vater. Clin Cancer Res. 1997;3(1):129–133.

9. Kaiser A, Jurowich C, Schonekas H, et al. The adenoma-carcinoma se-quence applies to epithelial tumours of the papilla of Vater. Z Gastroenterol.2002;40(11):913–920.

10. Park S, Kim SW, Kim SH, et al. Loss of heterozygosity in ampullaof Vater neoplasms during adenoma-carcinoma sequence. Anticancer Res.2003;23(3C):2955–2959.

11. Fischer HP, Zhou H. Pathogenesis of carcinoma of the papilla of Vater. JHepatobiliary Pancreat Surg. 2004;11(5):301–309.

12. Kimura W, Futakawa N, Yamagata S, et al. Different clinicopathologic findingsin two histologic types of carcinoma of papilla of Vater. Jpn J Cancer Res.1994;85(2):161–166.

13. Westgaard A, Tafjord S, Farstad IN, et al. Pancreatobiliary versus intestinalhistologic type of differentiation is an independent prognostic factor in resectedperiampullary adenocarcinoma. BMC Cancer. 2008;8:170.

14. Heinrich S, Clavien PA. Ampullary cancer. Curr Opin Gastroenterol.2010;26(3):280–285.

15. Hansel DE, Maitra A, Lin JW, et al. Expression of the caudal-type home-odomain transcription factors CDX 1/2 and outcome in carcinomas of theampulla of Vater. J Clin Oncol. 2005;23(9):1811–1818.

16. Goldstein NS, Bassi D. Cytokeratins 7, 17, and 20 reactivity in pancreaticand ampulla of vater adenocarcinomas. Percentage of positivity and distribu-tion is affected by the cut-point threshold. Am J Clin Pathol. 2001;115(5):695–702.

17. Sessa F, Furlan D, Zampatti C, et al. Prognostic factors for ampullary adenocar-cinomas: tumor stage, tumor histology, tumor location, immunohistochemistryand microsatellite instability. Virchows Arch. 2007;451(3):649–657.

18. Chu PG, Schwarz RE, Lau SK, et al. Immunohistochemical staining in thediagnosis of pancreatobiliary and ampulla of Vater adenocarcinoma: appli-cation of CDX2, CK17, MUC1, and MUC2. Am J Surg Pathol. 2005;29(3):359–367.

19. Zhou H, Schaefer N, Wolff M, et al. Carcinoma of the ampulla of Vater:comparative histologic/immunohistochemical classification and follow-up. AmJ Surg Pathol. 2004;28(7):875–882.

20. Howe JR, Klimstra DS, Moccia RD, et al. Factors predictive of survival inampullary carcinoma. Ann Surg. 1998;228(1):87–94.

21. Talamini MA, Moesinger RC, Pitt HA, et al. Adenocarcinoma of the ampullaof Vater. A 28-year experience. Ann Surg. 1997;225(5):590–599; discussion599–600.

22. Grutzmann R, Pilarsky C, Ammerpohl O, et al. Gene expression profiling ofmicrodissected pancreatic ductal carcinomas using high-density DNA microar-rays. Neoplasia. 2004;6(5):611–622.

23. Li C. Automating dChip: toward reproducible sharing of microarray data anal-ysis. BMC Bioinformat. 2008;9:231.

24. Albores-Saavedra J, Henson DE, Klimstra DS. Tumours of the gallbladder,extrahepatic bile ducts, and ampulla of Vater. In: Atlas of Tumor Pathology, Vol.3rd series. Washington, DC: Armed Forces Institute of Pathology; 2000:171–174.

25. Bubendorf L, Nocito A, Moch H, et al. Tissue microarray (TMA) technology:miniaturized pathology archives for high-throughput in situ studies. J Pathol.2001;195(1):72–79.

26. Baumhoer D, Zlobec I, Tornillo L, et al. Immunophenotyping and onco-gene amplifications in tumors of the papilla of Vater. Virchows Arch.2008;453(6):579–588.

27. Simon R, Sauter G. Tissue microarray (TMA) applications: implications formolecular medicine. Expert Rev Mol Med. 2003;5(26):1–12.

28. Woenckhaus M, Merk J, Stoehr R, et al. Prognostic value of FHIT,CTNNB1, and MUC1 expression in non-small cell lung cancer. Hum Pathol.2008;39(1):126–136.

29. Lugli A, Tzankov A, Zlobec I, et al. Differential diagnostic and functionalrole of the multi-marker phenotype CDX2/CK20/CK7 in colorectal cancerstratified by mismatch repair status. Mod Pathol. 2008;21(11):1403–1412.

30. O’Connell JB, Maggard MA, Manunga J, Jr., et al. Survival after resectionof ampullary carcinoma: a national population-based study. Ann Surg Oncol.2008;15(7):1820–1827.

31. de PaivaHaddad LB, Patzina RA, Penteado S, et al. Lymph node involvementand not the histophatologic subtype is correlated with outcome after resectionof adenocarcinoma of the ampulla of vater. J Gastrointest Surg. 14(4):719–728.

32. Yamagata K. Regulation of pancreatic beta-cell function by the HNF transcrip-tion network: lessons from maturity-onset diabetes of the young (MODY).Endocr J. 2003;50(5):491–499.

33. Cattin AL, Le Beyec J, Barreau F, et al. Hepatocyte nuclear factor 4alpha, akey factor for homeostasis, cell architecture, and barrier function of the adultintestinal epithelium. Mol Cell Biol. 2009;29(23):6294–6308.

34. Babeu JP, Darsigny M, Lussier CR, et al. Hepatocyte nuclear factor 4alphacontributes to an intestinal epithelial phenotype in vitro and plays a partial rolein mouse intestinal epithelium differentiation. Am J Physiol Gastrointest LiverPhysiol. 2009;297(1):G124–G134.

35. Benahmed F, Gross I, Gaunt SJ, et al. Multiple regulatory regions control thecomplex expression pattern of the mouse Cdx2 homeobox gene. Gastroen-terology. 2008;135(4):1238–1247, 1247 e1–3.

36. Boyd M, Bressendorff S, Moller J, et al. Mapping of HNF4alpha target genesin intestinal epithelial cells. BMC Gastroenterol. 2009;9:68.

37. Takegoshi S, Jiang S, Ohashi R, et al. Protein expression of nuclear receptorsin human and murine tissues. Pathol Int. 2009;59(2):61–72.



Hepatocyte Nuclear Factor (HNF) 4 [alpha] Expression Distinguishes Ampullary Cancer Subtypes and...

Documents

Transcript of Hepatocyte Nuclear Factor (HNF) 4 [alpha] Expression Distinguishes Ampullary Cancer Subtypes and...