Experimental and Molecular Pathology 95 (2013) 235–241

Contents lists available at ScienceDirect

Experimental and Molecular Pathology

j ourna l homepage: www.e lsev ie r .com/ locate /yexmp

TP53 mutations are common in all subtypes of epithelial ovarian cancerand occur concomitantly with KRAS mutations in the mucinous type

Markus Rechsteiner a, Anne-Katrin Zimmermann a, Peter J. Wild a, Rosmarie Caduff a, Adriana von Teichman a,Daniel Fink b, Holger Moch a, Aurelia Noske a,⁎a Institute of Surgical Pathology, University Hospital Zurich, Switzerlandb Department of Obstetrics and Gynaecology, University Hospital Zurich, Switzerland

⁎ Corresponding author at: Institute of Surgical PatholSchmelzbergstr.12, CH-8091 Zurich, Switzerland. Fax: +4

E-mail address: aurelia.noske@usz.ch (A. Noske).

0014-4800/$ – see front matter © 2013 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.yexmp.2013.08.004

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 7 August 2013Available online 18 August 2013

Keywords:Epithelial ovarian cancerDeep-sequencingTP53KRASBRAF

Aims:Epithelial ovarian cancer (EOC) can be classified into fourmajor types (serous, endometrioid, clear cell, mu-cinous). The prevalence of driver gene mutations in the different subtypes is controversial. High-grade serouscarcinomas show frequent TP53mutations, whereas KRAS and BRAF mutations are less common. In non-serousEOC, the relevance of these gene mutations remains to be elucidated.Methods:We investigated 142 formalin-fixed, paraffin-embedded EOC, including serous (n = 63), endometrioid(n = 29), clear cell (n = 25), mucinous (n = 14), and others (n = 11) for mutations in TP53 exons 5–8, KRASexons 2 and 3, and BRAF exon15 bypyro-sequencingusing theGS Junior 454 platform. Themutational statuswascorrelated with clinicopathological features and patient overall survival.Results:We identifiedmutations in the coding region of TP53 in 51.4% (73/142), and of KRAS in 9.9% (14/142) but

not of BRAF. TP53 mutations occurred frequently not only in high-grade serous carcinomas (58.7%), but also inmucinous (57%) and clear cell EOC (52%). TP53 mutations were associated with high-grade carcinomas (p =0.014), advanced FIGO stage (p = 0.001), intraoperative residual disease N1 cm (p = 0.004), as well as pooroverall survival (p = 0.002). KRASmutations were mainly identified inmucinous EOC (57%) and were concom-itantly with TP53 mutations in five mucinous carcinomas (36%).Conclusions: TP53 gene driver mutations are a common feature of all advanced ovarian cancer subtypes, whereasBRAFmutations seem to be a rare event in EOC.KRASmutationswith synchronous TP53mutations occur predom-inantly in low-grade mucinous carcinomas, suggesting a specific molecular background of this ovarian cancertype.

© 2013 Elsevier Inc. All rights reserved.

Introduction

Background

Ovarian cancer has the highest mortality rate among gynaecologicalmalignancies in the Western world (Siegel et al., 2012). Due to theabsence of early clinical symptoms, themajority of thepatients are diag-nosed with an advanced disease stage. Despite many efforts in ovariancancer research, reliable prognostic and predictive biomarkers are stillnot available. In the last decade, molecular studies have shown thatthe histological subtypes of ovarian cancer are associated with differentgenetic alterations and biological behaviour as well as response tochemotherapy (Bast et al., 2009; Prat, 2012).

Recent molecular studies have mainly focused on the commonserous histological subtype and little is known about the non-seroussubtypes. The Cancer Genome Atlas Research Network (TCGA) and

ogy, University Hospital Zurich,1 44 255 4416.

ghts reserved.

Ahmed et al. have demonstrated that high-grade serous cancer is char-acterized by TP53 mutations in up to 96% of the cases (Ahmed et al.,2010; Anon, 2011). Both studies identified exons 5–8 of TP53 as themost frequently mutated region with 89% and 81.5%, respectively(Ahmed et al., 2010; Anon, 2011). Previous studies have shown thatTP53 mutations are highly prevalent in high-grade serous carcinomaswhen compared to low-grade serous carcinomas or serous borderlinetumours (Salani et al., 2008; Singer et al., 2005). TP53 mutations werefurther reported in high-grade endometrioid and transitional cell carci-nomas (Cuatrecasas et al., 2009; Kolasa et al., 2006). In contrast, KRASand BRAF mutations were found in low-grade serous carcinomasand its precursors (Singer et al., 2003; Sieben et al., 2004). BRAFmu-tations seem to be restricted to low-grade serous neoplasms, where-as KRAS mutations were identified in mucinous carcinomas andits precursors as well as in endometrioid and clear cell subtypes(Auner et al., 2009; Mayr et al., 2006; Sieben et al., 2004). However,in these publications the TP53, KRAS, and BRAF mutation status wasnot assessed simultaneously.

Deep-sequencing technology allows the detection of mutations inmultiple target regions in parallel within the tumour population of a

Table 1Clinicopathological parameters of 142 ovarian carcinomas.

Characteristics Cases n (%)

Histological typeSerous 63 (44.4)Mucinous 14 (9.9)Endometrioid 29 (20.4)Clear cell 25 (17.6)Transitional 9 (6.3)Malignant müllerian mixed tumor (MMMT) 2 (1.4)

FIGO stage (n = 135)I 34 (25.2)II 12 (8.9)III 50 (37.0)IV 39 (28.9)

236 M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

patient. In a recent study,we assessed the TP53mutation status in endo-metrial carcinomas by ultra-deep-sequencing with input DNA that de-rived from formalin-fixed and paraffin-embedded (FFPE) tissue (Wildet al., 2012). Furthermore, we demonstrated the reliability of this meth-od for other targets such as BRAF using FFPEmaterial (Rechsteiner et al.,2013) and using DNA derived from fresh frozen tissue (Gerstung et al.,2012).

Consequently, we aimed to characterize a broad spectrum of humanepithelial ovarian carcinomas for specific cancer-causing mutations inTP53 (exons 5–8), KRAS (exons 2 and 3), and BRAF (exon 15) by ahigh-throughput deep-sequencing approach with DNA derived fromFFPE. We further compared the mutational profile among differenthistological subtypes, clinicopathological factors, andwith patient over-all survival.

Tumour grade1 17 (12.0)2 59 (41.5)3 66 (46.5)

Residual disease (n = 95)b1 cm 25 (26.3)N1 cm 70 (73.7)

Materials and methods

Ovarian cancer patients

Consecutive primary epithelial ovarian carcinomas (n = 142) diag-nosed at the Institute of Surgical Pathology, University Hospital Zurich(Switzerland) between 1995 and 2005 were studied. Borderline tu-mours of the ovary were excluded from this analysis. Tissue sampleswere fixed in 4% neutral buffered formaldehyde and embedded in par-affin. Routine haematoxylin and eosin sections were processed forhistopathological evaluation. All carcinomas were reviewed by two ex-perienced gynaecologic pathologists (RC, AN) and selected for theirsuitability for molecular analysis. Histological subtype and grade weredefined according to the WHO classification 2003. Histological typingwas primarily performedwith conventional microscopy and in ambigu-ous cases immunohistochemistry was used.

Microscopically, high grade serous carcinomas show a destructivestromal invasion and are characterized by a papillary or solid growthpattern. Tumour cells form slit-like spaces and show a high variationin nuclear size and mitotic activity (generally above 12 per 10 HPF). Incontrast, low grade serous carcinomas have a more uniform architec-ture with small papillae or nests and relatively uniform nuclei. Mucin-ous carcinomas show a complex glandular architecture with expansileor infiltrative stromal invasion. The epithelium resembles intestinal orendocervical epithelium and precursor lesions are often adjacent.Medical records were thoroughly reviewed to exclude metastases tothe ovary. All mucinous carcinomas included in this study originatefrom the ovary. Endometrioid carcinomas closely resemble the uterinecounterpart and the same criteria for grading were used. Clear cell car-cinomas show a tubulocystic, papillary and solid growth pattern with ahyalinised, fibroblastic or myxoid stroma. The usual cell type is theclear or hobnail cell but variations are possible. Even though, theWHO does not recommend a grading for clear cell cancer, weperformed it for clinical purpose according to the nuclear polymor-phism and mitotic activity (mitotic figures/10 HPF: 0–9 vs. 10–24vs. N25) resulting in G2 (n = 5) and G3 (n = 20) clear cell carcino-mas. Finally, transitional cell carcinomas are composed of cysticspaces, undulating macropapillae with smooth borders, transitionalepithelium, marked nuclear atypia and without any evidence of a be-nign or borderline Brenner component. The tumour stage wasassessed according to the International Federation of Gynaecologyand Obstetrics (FIGO) staging system.

The median patient age at the time of surgery was 60 years, rang-ing between 20 and 87 years of age. The mean follow-up time was56.5 months (range, 0.13–201 months). Data on adjuvant chemo-therapy was known for all patients. In the majority of the cases(67.8%), a platinum-based combination therapy was administered.The study was approved by the Cantonal Ethics Committee of Zurich(StV 27-2009). Characteristics of the study population are summarizedin Table 1.

Deep-sequencing

Only formalin-fixed, paraffin-embedded (FFPE) ovarian cancertissue blocks containing at least 70% tumour cells according tohaematoxylin and eosin staining were selected for punching, DNAextraction, and mutation analysis. Furthermore, the location ofpunching was precisely defined within the tumour area. Three tu-mour tissue cylinders (diameter 0.6 mm) were punched from eachFFPE tissue block. Genomic DNA was extracted using the DNeasyBlood & Tissue Kit 250 (Qiagen, Hilden, Germany). DNA quantityand quality were assessed with the NanoDrop. Average DNA qualityof the OD 260/280 ratio was 1.86 ranging from 1.68 to 1.97. Deep-sequencing and analysis protocols are described in Rechsteineret al. (2013). Briefly, 50 ng of DNA were used as input for PCR ampli-fication (AmpliTaq Gold, Roche, Switzerland) with fusion primers in-cluding multiplex identifiers (MID) targeting KRAS exons 2 and 3,BRAF exon 15, and TP53 exons 5–8 (Supplemental Table 1). Ampliconprocessing was performed as described by the Amplicon LibraryPreparation and emulsion PCR (emPCR) (Lib-A)Method GS Junior Ti-tanium Series manual from Roche. Samples were run on a 454 JuniorSequencer (Roche). The reads were aligned and variants were called tothe reference genome GRch37/hg19 using the 454 integrated softwareAmplicon Variant Analyser (AVA, version 2.5) with default settings.Only variants were included in the subsequent analysis which fulfilledthe following criteria: i) presence of forward and reverse reads, ii) atleast 50 reads containing the variant and iii) at least 0.38% variant fre-quency (Rechsteiner et al., 2013). To weigh the functional impact ofthe mutations only variants which fulfilled at least two of the followingcriteria were included: Mutation Assessor (http://mutationassessor.org/): high, medium; PolyPhen (http://www.ensembl.org/tools.html):probably_damaging, possibly_damaging; SIFT (http://www.ensembl.org/tools.html): deleterious; VEP (http://www.ensembl.org): non-synonymous coding missense mutations, STOP gained, STOP lost, com-plex indel, frame-shift coding; CHASM (http://www.chasmsoftware.org): p-value b 0.05 (Carter et al., 2009; Forman et al., 2009). Variantswere compared to entries in the Sanger Cosmic database (http://www.sanger.ac.uk/genetics/CGP/cosmic/), IARC database (R16, Nov.2012), and the dbSNP131.

Immunohistochemistry

Immunohistochemical staining of p53 was performed on a tissuemicroarray as previously described (Noske et al., 2011). Briefly, after

237M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

antigen retrieval with CC1 mild (TRIS buffer with a slightly increasedpH) for 30 min, the slide was incubated with a monoclonal p53 anti-body (Nr. M7001, clone DO-7, DAKO) which was diluted 1:80 inVentana dilution buffer. After incubation for 1 h at room temperature,the staining was further conducted with the Ventana Benchmark auto-mated system (Ventana Medical Systems, Tuscon, AZ, USA) using theUltraVIEW™ HRP detection kit. The expression analysis was blindlyperformed by two pathologists (AKZ, AN) and scored according to thepercentage of positive cells (0%, 1–60%, and N61%).

Statistical analysis

Deep-sequencing amplicon analysis resulting in coverage and vari-ant detection limits were calculated with Prism 5 (version 5.04) andare presented as median with interquartile range. Significance of corre-lation in Prism was calculated with linear regression. The statistical sig-nificance of the association betweenmutation status of TP53,KRAS, BRAFand clinicopathological parameters as well as immunohistochemicalp53 status was assessed using the Chi2 test for trends and Fisher'sexact test. The probability of overall survival as a function of time wasdetermined by the univariate regression analysis and the Kaplan–Meier method. Differences in the survival curves were compared bythe log rank test. Multivariate survival analysis was performed usingthe Cox model of proportional hazards. Generally, p-values b0.05were considered as significant. For statistical evaluation the softwareIBM SPSS Statistics Version 20 (SPSS Inc., Chicago, IL, USA) was used.

Results

Distribution of TP53, KRAS, and BRAF mutations in ovarian carcinomas

Isolated DNA of FFPE ovarian cancer samples (n = 144) were usedas input for PCR amplification of exons 5–8 of TP53, exons 2 and 3 ofKRAS, and exon 15 of BRAF. Due to fragmented DNA, exon 5 of TP53and exon 15 of BRAF were split into two separate amplicons and exon3 of KRAS into one short amplicon covering the first part of the exononly (Supplemental Table 1). This approach resulted in nine ampliconsper patient. Subsequently, nine patients were multiplexed and se-quenced simultaneously in one deep-sequencing run. In total, weperformed 16 deep-sequencing runs resulting in 1296 amplicons foranalysis. Twenty-nine (2.24%) amplicons were not covered by at least50 reads and did therefore not fulfil our filtering criteria as describedinmaterials andmethods. These amplicons were excluded from furtheranalysis. A coverage plot per amplicon is shown in Supplemental Fig. 1Awith amedian over all amplicons of 797 reads (dotted line). After apply-ing our filter criteria, a median variant detection limit of 6.28% wasachieved (dotted line; Supplemental Fig. 1B) ranging from 4.19%(TP53 exon 7) to 12.38 (TP53 exon 5 Fs2/R). The coverage or the detec-tion limit did not correlatewith amplicon size or GC% content. However,the ampliconwith lowest coverage andhighest detection limitwas TP53exon 5 Fs2/R which exhibited as well the highest GC% content (Supple-mental Fig. 1C–D). From the DNA probe set, two cases had to be exclud-ed from further analysis because of post-analytical errors (specimenidentification) resulting in 142 ovarian cancer samples in total for thisstudy.

The distribution and frequency of the TP53 mutations in the codingregion is depicted in Fig. 1A. Twenty-two mutations were found inexon 5, 12 in exon 6, 11 in exon 7, and 24 in exon 8. The hot-spot atamino-acid (AA) 273 was represented with the highest frequency(11 mutations). In total, we found 51.4% (73/142) of TP53 mutationsin the coding region of the primary ovarian carcinomas (Fig. 1B).

KRASmutations in the coding regionwere observed in 9.9% (14/142)(Fig. 1B). Thirteenmutationswere detected in exon 2 and one in exon 3.In exon 2, 11 mutations were at position 12 (#5 G12D, #3 G12V, #1G12S, #1 G12R, and #1 G12A) and two mutations at position 13(G13D). The mutation in exon 3 was at position 61 (Q61H). In six

cases, mutations in both genes TP53 and KRAS were simultaneouslyfound. No BRAF mutation was found in the coding region of any of thecarcinoma samples (Fig. 1B). In 61 ovarian carcinomas, all three geneswere wild-type (wt).

Furthermore, we assessed the mutations at the intron/exon bound-aries defined as splice site (3–8 bp into the intron) and essential splicesites (1–2 bp into the intron). For TP53, we detected two and for KRASone mutation at essential splice sites. One mutation in the splice siteregion was also detected for BRAF. All the types of mutations are sum-marized in Fig. 1C. The most prominent mutations in TP53 (60/73)and KRAS (14/14) were missense mutations leading to an exchange ofthe coding amino-acid. TP53 additionally exhibited frame-shift dele-tions (2/73), in frame deletions (2/73), and nonsense mutations gener-ating STOP codons (9/73). The number of exonicmutations in thewholecohort as well as for serous and mucinous carcinoma subtypes is indi-cated (Fig. 1D–F). Both frame-shift deletions and one in frame deletion(del (9) S149) were unknown in the IARC and Cosmic database. Fur-thermore, the STOP mutation p.R206* was not present in the IARC andCosmic database. All othermutationswere known and conferred a func-tional impact as described in materials and methods. All mutations de-tected are summarized in Supplemental Table 2.

Association of TP53 and KRAS mutations with pathological features

TP53 mutations occurred within all histological subtypes except formalignant müllerian mixed tumor (MMMT) while KRAS mutationswere only found in serous, mucinous, and endometrioid carcinomas.To further evaluate the association ofmutational status andpathologicalfeatureswe performed a statistical analysis. Due to the low frequency oftheMMMT subtype in our cohort, the two caseswere excluded from theevaluation.

As indicated in Table 2, TP53mutations were significantly more fre-quent in advanced ovarian carcinomas (FIGO stage III–IV), in high-gradecarcinomas as well as in cases with residual tumour disease N1 cm. Incontrast, KRAS mutations were associated with better differentiatedcarcinomas (all types) and predominantly identified in the mucinoussubtype.

We further performed a subgroup analysis for each histological type.In serous carcinomas, TP53 mutations were significantly associated(Table 3) withmoderately and poorly differentiated (high-grade) carci-nomas (p = 0.039) as well as advanced FIGO stage (p = 0.015). Inter-estingly, three KRAS mutations were found in high-grade (G2 and G3)cancer (all FIGO IV) but not in low-grade serous carcinomas (n = 3).In one of these high-grade serous carcinomas, both TP53 and KRASmu-tationswere detected. In contrast, the two remaining high-grade serouscarcinomas with only KRAS mutations displayed a p53 expression inonly 0% and 5% of the tumour cells assessed by IHC. This observationsuggests a loss of p53 expression due to promoter hypermethylationor instability.

Fifty-seven percent of themucinous carcinomas (8/14) showedmu-tations in TP53 or KRAS. From these 14 cases, five cases exhibited muta-tions in both genes simultaneously. TP53 mutations were present inseven out of 10 well differentiated carcinomas (G1) and KRAS in sixcases, respectively. Both mutations were predominantly found in mu-cinous carcinomas FIGO stage I.

In the endometrioid subtype, TP53mutations were significantlymorepresent inmoderately and poorly differentiated carcinomas (p = 0.022),whileKRASmutationswere observed inG1- andG2-carcinomas (1/4, and2/16, respectively). In clear cell cancer, we found TP53 mutations in 52%(13/25) which were associated with advanced FIGO stage III and IV(p = 0.005), and residual disease N1 cm (p = 0.039). All clear cell carci-nomas were KRAS wt. Finally, the transitional cell subtype showed TP53mutations in 78% (7/9), but was KRAS wt and did not show any associa-tion with pathological parameters. The distribution of TP53 mutationsin relation to histological type and grading are given in Table 4.

Fig. 1. Deep-sequencing analysis of epithelial ovarian carcinomas. A: Frequency and distribution of TP53 exons 5–8. B: Frequency of mutation in the coding region and wild-type (wt) inTP53, KRAS, and BRAF. C:Mutations found in the coding region and splice sites of TP53, KRAS, and BRAF. D: Number of carcinomaswith single and doublemutations in all histological types.E: Mutations found in serous carcinomas. F: Mutations found in mucinous carcinomas.

238 M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

Association of TP53 and KRAS mutations with overall survival

In our cohort, we found that established prognostic factors for ovar-ian cancer such as FIGO stage (p = 0.0001, HR 2.1 CI 1.7–2.7), histology(p = 0.004, HR 1.2 CI 1.0–1.4), grading (p = 0.004, HR 1.6 CI 1.1–2.2),intraoperative residual tumour (p = 0.0001, HR 4.7 CI 2.3–9.6), and pa-tient age at diagnosis (p = 0.0001, HR 2.3 CI 1.5–3.6) reached a signifi-cant prognostic impact on the univariate logistic regression analysis.Wefurther observed that TP53mutationswere associatedwith poor overallsurvival (p = 0.002, HR 1.9 CI 1.3–3.0) in the univariate logistic regres-sion analysis and Kaplan–Meier analysis (log rank, p = 0.002) asshown in Fig. 2A. In a multivariate COX regression analysis (includingTP53 mutation status, histology, grading, FIGO stage, intraoperativeresidual tumour, and patient age) only residual tumour and patientage reached independent prognostic impact on patient overall survival.

In the subgroup of high-grade serous carcinomas, univariate survivalanalysis revealed better overall survival for TP53 wt carcinomas whencompared to mutated TP53 (p = 0.024 HR 2.0 CI 1.1–3.8 and log rank,p = 0.022, Fig. 2B). However, this could be not confirmed in amultivar-iate analysis. Subgroup analyses for the non-serous carcinomas were

not performed because of the limited sample number. For the KRASmu-tation status, no relation to the patient overall survival (p = 0.2, HR 0.6CI 0.2–1.3) was observed.

Comparison of TP53 mutations with p53 immunohistochemical staining

Recently, Yemelyanova et al. reported that immunohistochemistryof p53 can serve as a surrogate marker for TP53 mutations in a cohortof 57 ovarian carcinomas (Yemelyanova et al., 2011). They showedthat two different p53 immunoexpression patterns (N60% positivecells and complete absence of p53) are commonly associated withTP53 mutations. Thus, we compared TP53 mutation status with immu-nohistochemical staining of p53 which was scored according to thiswork. This score is based on the percentage of positive p53 cells anddoes not consider staining intensity. We found a significant association(p b 0.0001) of the TP53mutational status and theprotein expression ofp53 as shown in Table 5. In carcinomas with complete lack of p53immunoexpression, nonsense mutations generating a stop-codonwere the predominant type (67%). Missense mutations (94%) were sig-nificantly associatedwith carcinomas showing strong expression of p53

Table 2Association of KRAS and TP53 mutations with pathological parameters.

Characteristics All cases(n = 140)

KRAS mutationn (%)

p-Value TP53 mutationn (%)

p-Value

Histological type n.s. n.s.Serous 63 3 (4.8) 37 (58.7)Mucinous 14 8 (57.1) 8 (57.1)Endometrioid 29 3 (10.3) 8 (27.6)Clear cell 25 0 (0) 13 (52.0)Transitional cell 9 0 (0) 7 (77.8)

FIGO stage (n = 133) n.s. 0.001I 34 7 (20.6) 9 (26.5)II 11 0 (0) 5 (45.5)III 49 3 (6.1) 29 (58.0)IV 38 4 (10.5) 26 (68.4)

Tumour grade 0.001 0.0141 17 7 (41.1) 7 (41.1)2 58 6 (10.3) 24 (41.3)3 64 1 (1.6) 42 (64.6)

Residual disease(n = 93)

n.s. 0.004

b1 cm 25 4 (16.0) 8 (32.0)N1 cm 67 4 (5.9) 45 (66.2)

Table 4

239M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

(chi2 for trends, p = 0.002). In the univariate regression analysis, theimmunohistochemical p53 status did not reach a prognostic signifi-cance for the overall survival (p = 0.12).

Discussion

In this study, we investigated 142 primary epithelial ovarian car-cinomas with different histological subtypes for mutations in TP53,KRAS, and BRAF using deep-sequencing technology. We demonstratean association between TP53 mutations with high differentiationgrade, advanced FIGO stage, residual disease, as well as poor overallsurvival. Synchronous mutations of KRAS and TP53 in early tumourstages were shown to be characteristic for the mucinous subtype.

Based on a high-throughput approach,wemultiplexed nine patientswith each nine amplicons resulting in a total of 81 target regions perrun. In total, we performed 16 deep-sequencing runs resulting in 1296amplicons for analysis fromwhich only twenty-nine (2.24%) ampliconsdid not fulfil our filtering criteria. Thus, the establishedmethod allowedus to screen the whole ovarian carcinoma cohort in a fast and efficientmanner.

Given that TP53 exons 5–8 account formutations in above 80% of theovarian carcinomas, we focused on these exons. With this approach, weidentified TP53 mutations in 58% of the serous carcinomas which is inline with on-line databases such as Cosmic (http://www.sanger.ac.uk/genetics/CGP/cosmic/;65%). In contrast, The Cancer Genome Atlas(TCGA) and Ahmed et al. reported TP53 mutations in exons 2–11 in upto 97% of the high-grade serous carcinomas (Ahmed et al., 2010;Anon, 2011). They identified TP53 mutations in exons 5–8 in 89%(TCGA) and 81.5% (Ahmed et al.). An explanationmight be that a highernumber of grade 3 serous carcinomas were studied than in our analysis.Additionally, we confirm for TP53 the high frequency (11 mutations) ofthe hot-spot at amino-acid (AA) 273 within our cohort.

In the Cosmic database, KRAS and BRAF are depicted to bemutated inovarian carcinomas at a frequency of 11% and 5%, respectively. We

Table 3Distribution of KRAS and TP53 mutations in serous ovarian cancer (n = 63).

Grade 1(n = 3)

Grade 2(n = 31)

Grade 3(n = 29)

p-Value

KRAS 0 (0%) 2 (6.5%) 1 (3.4%) n.s.TP53 0 (0%) 17 (54.8%) 20 (69%) 0.039

detected KRAS mutations at a similar frequency (9.9%), but could notidentify any BRAFmutation in the coding region. The lack of BRAFmuta-tions in our cohort is consistent with recent studies (Mayr et al., 2006;Wong et al., 2010) suggesting that BRAF alterations are rare in ovariancancer.

We found the highest prevalence of TP53 mutations in high-gradeserous carcinomas. This has been previously proposed in a dualisticmodel of ovarian cancer pathogenesis (Kurman and Shih, 2011; Singeret al., 2005). This model distinguishes two types of progression path-ways. Type I low-grade ovarian carcinomas (low-grade serous cancer,mucinous, endometrioid, and clear cell cancer) are characterized bymutations in genes like KRAS, BRAF, ERBB2, PTEN, PI3KCA, CTNNB1,ARID1A, and PPP2R1A but rarely harbour TP53, whereas aggressivetype II high-grade ovarian carcinomas (especially high-grade serouscancer) exhibit a high frequency of TP53 mutations (Ahmed et al.,2010; Anon, 2011; Singer et al., 2005). In our cohort, moderately (G2)and poorly differentiated (G3) serous carcinomas displayed a similarfrequency of TP53mutations. Thus, according to our TP53 data, G2 carci-nomas should be categorized as high-grade. This finding supports arecent proposal to use a two-tiered instead of a three-tiered gradingsystem for serous ovarian cancer (Ayhan et al., 2009). Consistent withother studies, we observed frequent TP53 mutations in high-gradeendometrioid (G2 and G3) and in transitional cell carcinomas(Cuatrecasas et al., 2009;Wu et al., 2007). We further frequently identi-fied TP53mutations in clear cell (52%) andmucinous (57%) carcinomas.To date, only limited data about the p53 status exist for these subtypes.In previous reports, TP53 mutations occurred in 10–12% in clear cellcarcinomas (Esther et al., 2001; Leitao et al., 2004). Interestingly, wefound TP53mutations even in well differentiated (G1) mucinous carci-nomas. Five of these cases (36%) had a simultaneous KRASmutation. Sofar, genetic studies analysing the TP53 mutation status in mucinous

TP53 mutation and differentiation grade.

Subtype TP53 mutationn

Grade 1n

Grade 2n

Grade 3n

Serous 37 0 17 20Mucinous 8 7 1 –

Endometrioid 8 0 3 5Clear cell 13 – 1 12Transitional cell 7 – 2 5

Fig. 2. Kaplan–Meier analysis. A: Overall survival analysis for all histological subtypes.Ovarian cancer patients with wild-type TP53 carcinomas show a significant better overallsurvival (log rank, p = 0.002) when compared to mutated TP53 carcinomas. B: Subgroupanalysis for serous carcinomas. Significant better overall survival of patients with carcino-mas harbouring TP53 wt than a mutation (log rank, p = 0.022).

240 M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

carcinomas are rare. Vereczkey et al. investigated the p53 proteinexpression by immunohistochemistry inmucinous borderline and inva-sive tumours, but found no over-expression indicative of mutations(Vereczkey et al., 2011). Fallows and colleagues observed TP53 muta-tions in 44% of ovarian mucinous carcinomas (Fallows et al., 2001).Simultaneous TP53 and KRAS mutations were also recently identifiedin three out of nine (33%) ovarian mucinous carcinomas (Ryland et al.,2013). On the basis of the dualistic EOC progressionmodel, the frequen-cy of TP53mutations in clear cell and mucinous EOC was not expected.

Table 5TP53 mutation and immunohistochemistry.

TP53mutation p53 immunohistochemistry(% positive cells)

p-Value

0% 1–60% N61%

Wild-type 35 26 3Mutation 12 9 51 b0.0001

Further, we confirmed an association between TP53mutation and theimmunohistochemical p53 expression pattern. As recently reported,immunohistochemistry of p53 can serve as a surrogate marker for TP53mutations in ovarian carcinomas (Yemelyanova et al., 2011). Our dataconfirm that not only a strong protein expression (N60% positive cells)but also a complete absence of p53protein is indicative of a p53mutation.However, we observed that only the TP53mutational status was a prog-nostic factor for overall survival but not p53 protein expression. Severalprevious studies have shown conflicting results when correlating p53mutation/IHC status with clinical outcome, but it is difficult to comparethese studies due to different immunohistochemical protocols, cut-offs,etc. Our findings indicate that the prognostic value of the TP53mutationstatus seems to be superior compared to p53 immunohistochemistry.

In this study, we detected the highest KRASmutation rate in mucin-ous carcinomas (57%). Up to 75% (Auner et al., 2009; Kurman and Shih,2011; Sieben et al., 2004; Vereczkey et al., 2011)KRASmutations are fre-quent in this subtype, and characteristic for the type I cancer pathway(Kurman and Shih, 2011). In endometrioid carcinomas, we identifiedKRAS mutations in only 10% which is in line with the literature (Auneret al., 2009; Stewart et al., 2012). In clear cell cancer, we observed noKRAS mutations, whereas Auner et al. found KRAS mutations in 26%(5/19) using a lowdensity biochip platform (Auner et al., 2009). The du-alistic ovarian cancer progressionmodel was developedwithmoleculardata from low-grade serous carcinomas and precursor lesions aswell ashigh-grade serous carcinomas. In our study,we analysed a series of non-serous subtypes. This might be a potential reason for an underrepresen-tation of KRAS and/or BRAF mutations in our cohort, because thesemutations are mainly present in low-grade serous and mucinous carci-nomas aswell as in borderline tumours. According to our data, a specificpathway for mucinous EOC is conceivable, because of a significantlyhigher KRAS mutation rate in well differentiated (G1) carcinomas withsynchronous TP53 mutations. For all other EOC subtypes, TP53 muta-tions are potentially relevant for tumour progression, because themuta-tions were seen in advanced tumour stages with poor differentiationgrade.

In summary, we provide clear evidence that not only high-gradeserous EOC but also mucinous, clear cell, endometrioid, and transitionalcell carcinomas contain TP53mutations. Mucinous EOC is characterizedby KRAS mutations with frequent synchronous TP53 mutations alreadyin early tumour stages. Thus, a future molecular stratification of epi-thelial ovarian carcinomas should consider the mutational status ofall subtypes and may be elucidated by high-throughput methods asdeep-sequencing.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.yexmp.2013.08.004.

Conflict of interest statement

The authors declare that there are no conflicts of interest.

Acknowledgments

We like to thank Martina Storz, Susanne Dettwiler, Sonja Brun-Schmid, Annette Bohnert, and Silvia Behnke for their excellent technicalassistance. This workwas supported by a grant “Ida de Pottère-Leupold-Fonds zur Förderung der Krebsforschung” (to AN, MR) and from theSwiss National Science Foundation (Sinergia Grant to HM).

References

Ahmed, A.A., Etemadmoghadam, D., Temple, J., Lynch, A.G., Riad, M., Sharma, R., et al.,2010. Driver mutations in TP53 are ubiquitous in high grade serous carcinoma ofthe ovary. J. Pathol. 221 (1), 49–56.

Anon, 2011. Integrated genomic analyses of ovarian carcinoma. Nature 474 (7353),609–615.

Auner, V., Kriegshauser, G., Tong, D., Horvat, R., Reinthaller, A., Mustea, A., et al., 2009.KRAS mutation analysis in ovarian samples using a high sensitivity biochip assay.BMC Cancer 9, 111.

241M. Rechsteiner et al. / Experimental and Molecular Pathology 95 (2013) 235–241

Ayhan, A., Kurman, R.J., Yemelyanova, A., Vang, R., Logani, S., Seidman, J.D., et al., 2009.Defining the cut point between low-grade and high-grade ovarian serous carcino-mas: a clinicopathologic and molecular genetic analysis. Am. J. Surg. Pathol. 33 (8),1220–1224.

Bast Jr., R.C., Hennessy, B., Mills, G.B., 2009. The biology of ovarian cancer: new opportu-nities for translation. Nat. Rev. Cancer 9 (6), 415–428.

Carter, H., Chen, S., Isik, L., Tyekucheva, S., Velculescu, V.E., Kinzler, K.W., et al., 2009.Cancer-specific high-throughput annotation of somatic mutations: computationalprediction of driver missense mutations. Cancer Res. 69 (16), 6660–6667.

Cuatrecasas, M., Catasus, L., Palacios, J., Prat, J., 2009. Transitional cell tumors of the ovary:a comparative clinicopathologic, immunohistochemical, and molecular genetic anal-ysis of Brenner tumors and transitional cell carcinomas. Am. J. Surg. Pathol. 33 (4),556–567.

Esther, S.H., Chiung-Ru, L., Yeun-Ting, H., Jung-Ta, C., Ai-Jane, L., Man-Jung, H., et al., 2001.p53 mutation is infrequent in clear cell carcinoma of the ovary. Gynecol. Oncol. 80(2), 189–193 (Feb).

Fallows, S., Price, J., Atkinson, R.J., Johnston, P.G., Hickey, I., Russell, S.E., 2001. P53 muta-tion does not affect prognosis in ovarian epithelial malignancies. J. Pathol. 194 (1),68–75.

Forman, J.R., Worth, C.L., Bickerton, G.R., Eisen, T.G., Blundell, T.L., 2009. Structural bioinfor-matics mutation analysis reveals genotype-phenotype correlations in von Hippel-Lindau disease and suggests molecular mechanisms of tumorigenesis. Proteins 77 (1),84–96.

Gerstung, M., Beisel, C., Rechsteiner, M., Wild, P., Schraml, P., Moch, H., et al., 2012. Reli-able detection of subclonal single-nucleotide variants in tumour cell populations.Nat. Commun. 3, 811.

Kolasa, I.K., Rembiszewska, A., Janiec-Jankowska, A., Dansonka-Mieszkowska, A.,Lewandowska, A.M., Konopka, B., et al., 2006. PTEN mutation, expression and LOHat its locus in ovarian carcinomas. Relation to TP53, K-RAS and BRCA1 mutations.Gynecol. Oncol. 103 (2), 692–697.

Kurman, R.J., Shih, Ie M., 2011. Molecular pathogenesis and extraovarian origin of epithelialovarian cancer—shifting the paradigm. Hum. Pathol. 42 (7), 918–931.

Leitao, M.M., Soslow, R.A., Baergen, R.N., Olvera, N., Arroyo, C., Boyd, J., 2004. Mutation andexpression of the TP53 gene in early stage epithelial ovarian carcinoma. Gynecol.Oncol. 93 (2), 301–306.

Mayr, D., Hirschmann, A., Lohrs, U., Diebold, J., 2006. KRAS and BRAF mutations inovarian tumors: a comprehensive study of invasive carcinomas, borderline tu-mors and extraovarian implants. Gynecol. Oncol. 103 (3), 883–887.

Noske, A., Zimmermann, A.K., Caduff, R., Varga, Z., Fink, D., Moch, H., et al., 2011. Alpha-methylacyl-CoA racemase (AMACR) expression in epithelial ovarian cancer. VirchowsArch. 459 (1), 91–97.

Prat, J., 2012. Ovarian carcinomas: five distinct diseases with different origins, geneticalterations, and clinicopathological features. Virchows Arch. 460 (3), 237–249.

Rechsteiner, M., von Teichman, A., Ruschoff, J.H., Fankhauser, N., Pestalozzi, B., Schraml, P.,et al., 2013. KRAS, BRAF, and TP53 deep-sequencing for colorectal carcinoma patientdiagnostics. J. Mol. Diagn. 15 (3), 299–311.

Ryland, G.L., Hunter, S.M., Doyle, M.A., Rowley, S.M., Christie, M., Allan, P.E., et al., 2013.RNF43 is a tumour suppressor gene mutated in mucinous tumours of the ovary.J. Pathol. 229 (3), 469–476.

Salani, R., Kurman, R.J., Giuntoli II, R., Gardner, G., Bristow, R., Wang, T.L., et al., 2008.Assessment of TP53mutation using purified tissue samples of ovarian serous carcino-mas reveals a higher mutation rate than previously reported and does not correlatewith drug resistance. Int. J. Gynecol. Cancer 18 (3), 487–491.

Sieben, N.L., Macropoulos, P., Roemen, G.M., Kolkman-Uljee, S.M., Jan Fleuren, G.,Houmadi, R., et al., 2004. In ovarian neoplasms, BRAF, but not KRAS, mutations arerestricted to low-grade serous tumours. J. Pathol. 202 (3), 336–340.

Siegel, R., Naishadham, D., Jemal, A., 2012. Cancer statistics, 2012. CA Cancer J. Clin. 62 (1),10–29.

Singer, G., Oldt III, R., Cohen, Y., Wang, B.G., Sidransky, D., Kurman, R.J., et al., 2003. Muta-tions in BRAF and KRAS characterize the development of low-grade ovarian serouscarcinoma. J. Natl. Cancer Inst. 95 (6), 484–486.

Singer, G., Stohr, R., Cope, L., Dehari, R., Hartmann, A., Cao, D.F., et al., 2005. Patternsof p53mutations separate ovarian serous borderline tumors and low- and high-gradecarcinomas and provide support for a new model of ovarian carcinogenesis: a muta-tional analysis with immunohistochemical correlation. Am. J. Surg. Pathol. 29 (2),218–224.

Stewart, C.J., Leung, Y., Walsh, M.D., Walters, R.J., Young, J.P., Buchanan, D.D., 2012. KRASmutations in ovarian low-grade endometrioid adenocarcinoma: association withconcurrent endometriosis. Hum. Pathol. 43 (8), 1177–1183.

Vereczkey, I., Serester, O., Dobos, J., Gallai, M., Szakacs, O., Szentirmay, Z., et al., 2011.Molecular characterization of 103 ovarian serous and mucinous tumors. Pathol.Oncol. Res. 17 (3), 551–559.

Wild, P.J., Ikenberg, K., Fuchs, T.J., Rechsteiner, M., Georgiev, S., Fankhauser, N., et al., 2012.p53 suppresses type II endometrial carcinomas in mice and governs endometrialtumour aggressiveness in humans. EMBO Mol. Med. 4 (8), 808–824.

Wong, K.K., Tsang, Y.T., Deavers, M.T., Mok, S.C., Zu, Z., Sun, C., et al., 2010. BRAF mutationis rare in advanced-stage low-grade ovarian serous carcinomas. Am. J. Pathol. 177 (4),1611–1617.

Wu, R., Hendrix-Lucas, N., Kuick, R., Zhai, Y., Schwartz, D.R., Akyol, A., et al., 2007.Mouse model of human ovarian endometrioid adenocarcinoma based on somaticdefects in the Wnt/beta-catenin and PI3K/Pten signaling pathways. Cancer Cell11 (4), 321–333.

Yemelyanova, A., Vang, R., Kshirsagar, M., Lu, D., Marks, M.A., Shih Ie, M., et al., 2011.Immunohistochemical staining patterns of p53 can serve as a surrogate marker forTP53 mutations in ovarian carcinoma: an immunohistochemical and nucleotidesequencing analysis. Mod. Pathol. 24 (9), 1248–1253.