Characterizing TP53 mutations in ovarian carcinomas with and without concurrent BRCA1 or BRCA2 mutations

Abstract

Objectives:

Mutations in the TP53 tumor suppressor gene are common in ovarian carcinoma (OC) but their impact on outcomes is controversial. We sought to define the relationship of TP53 mutations to cancer outcomes and their interactions with co-occurrent BRCA1 or BRCA2 (BRCA) mutations, comparing three different TP53 mutation classification schemes.

Methods:

We performed next generation sequencing on 393 cases of OC prospectively followed for survival. TP53 mutations were classified according to three schemes termed Structural, Functional, and Hotspot. Mutation distribution was compared between cases with and without BRCA mutations. In a subset of 281 cases of high grade serous carcinoma (HGSC), overall survival was compared using Kaplan-Meier curves, logrank testing, and multivariate Cox regression analysis, both stratified and adjusted for BRCA mutation status. Multivariate logistic regression was used to analyze the effects of TP53 mutation type on platinum resistance.

Results:

TP53 mutations were identified in 76.8% of the total cohort (n=302/393) and 87.9% of HGSC (n=247/281). Cases with BRCA mutations demonstrated significantly higher TP53 mutation frequency overall (n=84/91, 92.3% vs. n=218/302, 72.2%, p<0.001). TP53 mutations were not associated with overall survival, even when stratified by BRCA mutation. TP53 mutations were associated with platinum sensitivity, even after adjusting for BRCA mutation status (OR 0.41, p=0.048). The choice of TP53 mutation classification scheme was not found to alter any significant outcome.

Conclusions:

BRCA mutations significantly co-occur with TP53 mutations. After adjusting for BRCA mutations, TP53 mutations are associated with platinum sensitivity, and this effect is not dependent on TP53 mutation type.

Introduction

Mutations in TP53 are found in greater than 80% of epithelial ovarian cancers; specific rates depend on the histological distribution of the cohort, with the highest frequency in high grade serous carcinoma (HGSC) [1–3]. The corresponding protein p53 functions as a tumor suppressor by mediating the transcription of genes involved in cell cycle arrest, senescence, DNA damage repair, or apoptosis [4]. Although rarely found in the germline (as part of Li Fraumeni syndrome), the vast majority of TP53 mutations in cancer occur somatically [5]. Prior studies into the frequency and effects of TP53 mutations in ovarian carcinoma have often used immunohistochemistry studies (IHC) to predict mutational status [6]. While this has recently been optimized to increase specificity of testing, the gold-standard remains tumor sequencing [7].

More research is needed to understand the prognostic value of TP53 mutations in ovarian cancer. Prior studies have begun to investigate associations between TP53 mutations and ovarian carcinoma outcomes such as survival and platinum-resistance, however results have been discordant [1, 8–10]. Disparate findings may be due to population or cohort differences but are also likely related to varying methods used to classify and describe TP53 mutations.

There is no single accepted classification scheme for defining different types of TP53 mutations, thus limiting cross-literature comparisons. Studies have used classification schemes based on frequency found in cancer (often termed “hotspot” mutations), functional activity, or structural protein changes both to describe the mutations seen in ovarian carcinoma and to investigate effects on outcome such as survival or chemoresistance [2, 10]. There has been increasing interest in mutations that confer oncomorphic or gain-of-function (GOF) activity on p53, rather than just nullification of the tumor suppressor function, but there is not consensus on what defines a GOF mutation [9, 11]. There is also no clear definition of what constitutes a hotspot mutation, with previous studies defining hotspots based on codon location, individual mutation, specific cancer or all-cancer frequency [10, 12]. Differences in classification schemes may have impacted the conflicting conclusions of whether TP53 mutations are associated with outcomes in ovarian carcinoma.

Decades of research show that mutations in BRCA1 and BRCA2 (BRCA), and other genes involved in homologous recombination, are major prognosticators in survival and response to therapy in ovarian carcinoma [13, 14]. It is thus critical to study the interactions between TP53 and BRCA mutations in any analysis of molecular determinants of outcomes, but previous studies have been limited in doing so based on sample size [2, 15]. We aimed to better characterize the landscape of TP53 mutations and its effects on ovarian cancer outcomes while addressing these limitations by using a large, centrally-reviewed cohort with mutational status of both TP53 and BRCA confirmed through next-generation sequencing.

Methods

This was a prospectively collected cohort of 393 women diagnosed with primary ovarian, peritoneal, and fallopian tube carcinoma, collectively termed ovarian carcinoma (OC), undergoing surgery at the University of Washington Medical Center between 1996 and 2017. The included women provided informed consent for collection of blood and surgical tissue and clinical information, including long term follow-up, through an Institutional Review Board-approved protocol at the University of Washington. Pathology was centrally reviewed. Patients with recurrent OC, cancer diagnosed at time of risk-reducing salpingo-oophorectomy or borderline neoplasm were excluded. Demographic and clinical characteristics including primary histology, stage, grade, and platinum response were collected along with subsequent disease status and survival information on follow-up. The International Federation of Gynecology and Obstetrics (FIGO) ovarian cancer staging classification from 1988 was used to define stage given the timeline of included cases. Overall survival was measured from time of diagnosis. Cancers were classified as platinum sensitive if time to progression after completion of adjuvant platinum therapy was greater than six months or, if that information was unknown, the time between surgical diagnosis and disease progression was greater than fifteen months. All other cases were termed platinum resistant (including those that were platinum refractory). DNA sequencing was performed on the primary neoplastic sample using BROCA, a targeted capture, highly sensitive next generation sequencing platform that was developed at the University of Washington, which identifies all classes of mutations and targets the entire TP53 gene including intronic sequence [16–18].

TP53 mutation classification

Three different TP53 mutation classification schemes (Structural, Functional, and Hotspot) were compared in the study analysis in order to determine whether one classification better associated with outcomes (Table 1) [19]. The full list of GOF mutations used is outlined in Supplementary Table S1.

Table 1:

Mutation classification schemes

	Structural Classification:	Functional Classification:	Hotspot Classification:
Definition	Standard definitions of mutation types based on output from Seshat (https://p53.fr/TP53-database/seshat), a web service which performs TP53 annotation using data from the UMD TP53 database [19]. In-frame mutations were considered missense.	Nonsense/frameshift/splice mutations were considered loss-of-function mutations. Missense mutations were further subdivided into loss-of-function or gain-of-function based on experimental studies demonstrating an oncogenic phenotype, primarily interference with p73 activity, transactivation of genes repressed by wild-type p53 protein, or cooperation with oncogenes for transformation of rat or mouse embryonic fibroblasts (see Supplementary Table 1 for details and sources).	The nine most frequent codons mutated in ovarian cancer in the UMD TP53 database (https://p53.fr/tp53-database) [5], each found in greater than 1.5% of ovarian cancer samples, were termed “hotspots”. Missense mutations in these codons (273, 248,175, 220, 245, 195, 282, 179, 241) were separated from other mutations.
Classes	Wildtype Missense Nonsense/Frameshift/Splice	Wildtype Loss-of-function Gain-of-function	Wildtype Hotspot Missense Other mutations

All but two women had germline BRCA testing using either BROCA or another testing platform with information extracted from their medical records. The two women without germline sequencing were wildtype for BRCA mutations based on tumor sequencing results. Cases were classified as carrying a homologous recombination deficiency (HRD) mutation if testing revealed deleterious germline or somatic BRCA1 and BRCA2 mutations or mutations in other homologous recombination genes (ATM, BARD1, BRIP1, NBN, PALB2, RAD51C, and RAD51D) [13, 14, 16]. Deleterious mutations were characterized using the ClinVar database [20].

For confirmatory analysis, distribution of TP53 class was also investigated using data from The Cancer Genome Atlas (TCGA) and the PanCancer Atlas, accessed through cBioPortal [21–23]. Available mutational information from cases of ovarian serous carcinoma were downloaded and the distributions of TP53 mutations and BRCA mutations were analyzed.

Outcomes and statistical analysis

The distributions of TP53 mutations were described based on structural, functional, and hotspot classification schemes for the total cohort and within high-grade serous carcinoma (HGSC) only (Table 1). Comparative testing with Fisher’s exact test was used to examine TP53 mutation distribution and associations with clinicopathologic factors.

Kaplan-Meier and logrank testing was performed to compare survival based on TP53 mutation status. To decrease confounding, outcome analysis was limited to cases of HGSC. BRCA mutation status was further used to stratify the analysis to investigate specific effects of TP53 mutation. A multivariate Cox proportional hazards regression analysis was used incorporating a priori-defined clinicopathologic and genetic factors (age at diagnosis, stage, BRCA mutation status, primary platinum resistance, optimal cytoreduction, neoadjuvant chemotherapy). To test the proportionality of hazards we used Schoenfeld residuals. Association with primary platinum resistance was investigated as the secondary outcome of interest. Multivariate logistic regression was used to analyze the effects of TP53 mutation class within high-grade serous carcinoma cases using the same covariates. Stata version 14 [24] was used for all data analyses.

Results

Among the 393 women included in the total cohort, the majority had HGSC (n=281, 71.5%), grade 3 (n=355, 90.3%), and stage 3-4 (n=330, 84.0%) disease (Table 2). TP53 mutations were found in 76.8% of the total cohort (n=302), 87.9% of HGSC (n=247/281) and 49.1% of other histologies (n=55/112, p<0.001 comparing HGSC with other histologies). Of non-serous cases, the highest rates of TP53 mutations were found in adenocarcinoma not otherwise specified (94.7%), carcinosarcoma (85.7%) and endometrioid carcinoma (42.9%). Deleterious germline BRCA mutations were identified in 65 (16.5%) and somatic BRCA mutations were identified in 27 (6.9%). Further somatic or germline deleterious gene mutations conferring HRD (excluding BRCA mutations) were identified in 21 cases (5.3%) and included mutations in BRIP1 (n=5), RAD51C (n=4), RAD51D (n=4), ATM (n=3), BARD1 (n=2), PALB2 (n=2) and NBN (n=1), most of which have been previously described [16].

Table 2:

Cohort demographics

	N (total 393)	%
Mean age at diagnosis (SD)	59.9 (12.4)
Histology:
High grade serous	281	71.5%
Low grade serous	21	5.3%
Endometrioid	28	7.1%
Clear cell	21	5.3%
Adenocarcinoma	19	4.8%
Carcinosarcoma	14	3.6%
Other	9	2.3%
Grade:
1-2	35	8.9%
3	355	90.3%
Unknown	3	0.8%
Stage*:
I	34	8.7%
II	23	5.9%
III	275	70.0%
IV	55	14.0%
Unknown	6	1.5%
Platinum Response†:
Sensitive	219	68.2%
Resistant/refractory	82	31.8%
Unknown	92	55.7%
Cytoreduction‡:
Optimal	274	69.7%
Suboptimal	106	27.0%
Unknown	13	3.3%
Neoadjuvant chemotherapy:
Yes	48	12.2%
No	332	84.5%
Unknown	13	3.3%
BRCA1 or BRCA2§:
Wildtype	302	76.8%
Germline mutation	65	16.5%
Somatic mutation	27	6.9%
Other HRD genes|:
Wildtype	372	94.7%
Germline or somatic mutation	21	5.3%
TP53:
Wildtype	91	23.2%
Germline or somatic mutation	302	76.8%

^*Staged according to 1988 International Federation of Gynecology and Obstetrics (FIGO) ovarian cancer staging classification.

^†Platinum sensitive: no recurrence or progression within 6 months of platinum-containing primary adjuvant treatment.

^‡Optimal: less than 1 cm of residual disease.

^§One case with both somatic and germline mutation included in both categories.

^|HRD mutations (excluding BRCA1 or BRCA2) include ATM, BARD1, BRIP1, NBN, PALB2, RAD51C, and RAD51D.

Distribution of TP53 mutation types:

Using the structural classification scheme, the majority of cancers carried a missense TP53 mutation (n=220/393, 56.0%) (Figure 1A). Using the functional classification scheme, loss-of-function (LOF) TP53 mutations were found in 198 cancers (50.4%) while GOF TP53 mutations were found in 113 (28.7%). Missense mutations in the nine most frequently-mutated codons in ovarian cancer (termed “hotspot missense”) were found in 93 cancers (23.7%). GOF TP53 and hotspot missense mutations frequently overlapped; 85 (75.2%) of GOF TP53 mutations are also located in hotspot codons. HGSC had a significantly higher TP53 mutation frequency than cases with non-serous histology (n=247/281, 87.9% vs. 55/112, 49.1% respectively, p<0.001, Figure 1B).

Figure 1:

Distribution of TP53 mutations using different classification schemes. All cases had somatic tumor sequencing and mutations were classified according to three schemes based on structure, function, and hotspots (Table 1). A) Total cohort (n=393) B) Comparing high grade serous carcinoma to other histologies C) Comparing tumors with BRCA mutations (BRCAmut) and without BRCA mutations (BRCAwt) D) Confirmatory analysis using TCGA PanCancer Atlas cases of ovarian serous carcinoma (n=523), showing TP53 mutations by functional classification and BRCA status

OC with germline or somatic BRCA mutations demonstrated significantly higher TP53 mutation frequency overall when compared to OC that was BRCA wildtype (n=84/91, 92.3% vs. n=218/302, 72.2%, p<0.001, Figure 1C). There was no significant difference in distribution of specific TP53 mutations within each classification scheme (i.e. missense vs. nonsense/frameshift/splice, GOF vs. LOF) between cases with BRCA mutations and those without. Cases of HGSC with mutations in BRCA also demonstrated higher rates of TP53 mutations than HGSC without mutations in BRCA, though findings were not significant (n=66/72, 91.7% vs. 181/209, 86.6%, p=0.301). The full distribution of BRCA and TP53 mutations amongst HGSC and OC of other histologies is displayed in Supplementary Table S2. Even among non-high grade serous histologies, BRCA mutations tend to co-occur with TP53 mutations (Supplementary Table S3). In order to confirm these distribution findings, analyses were repeated using data from TCGA and PanCancer Atlas ovarian HGSC (Figure 1D). Cases with BRCA mutations similarly had higher incidence of TP53 mutations overall (n=32/33, 97.0% vs. n=341/490, 69.6%, p<0.001).

Mutations in TP53 (regardless of class) were more frequently found in carcinomas of patients older at diagnosis (ANOVA p<0.01) and in cases of stage III-IV disease (Fisher’s exact p<0.001, Table 3). Significant associations between mutation class and clinicopathologic factors are lost, however, when restricting analysis to within the 281 cases of HGSC only (Supplementary Table S4).

Table 3:

Comparison of TP53 mutation types between clinicopathologic subgroups.

	Wildtype (n=91)	Missense (n=220)	Nonsense/Frameshift/Splice (n=82)	P*	GOF (n=113)	LOF (n=189)	P*	Hotspot missense (n=93)	Other mutation (n=209)	P*
Mean age at diagnosis (SD)	55.3 (14.8)	60.9 (11.4)	62.2 (11.2)	0.002	61.4 (11.7)	61.2 (11.1)	<0.001	60.5 (11.4)	61.6 (11.3)	<0.001
Histology
High grade serous	34 (37.4%)	176 (80%)	71 (86.6%)	<0.001	92 (81.4%)	155 (82.0%)	<0.001	73 (78.5%)	174 (83.3%)	<0.001
Other carcinomas	57 (62.6%)	44 (20%)	11 (13.4%)		21 (18.6%)	34 (18.0%)		20 (21.%)	35 (16.8%)
Stage†
I-II	31 (35.2%)	17 (7.8%)	9 (11.1%)	<0.001	11 (9.8%)	15 (8.0%)	<0.001	11 (11.8%)	15 (7.3%)	<0.001
III-IV	57 (64.8%)	201 (92.2%)	72 (88.9%)		101 (90.2%)	172 (92.0%)		82 (88.2%)	191 (92.7%)
Platinum sensitive†	48 (71.6%)	119 (72.1%)	52 (75.4%)	0.881	58 (69.9%)	113 (74.8%)	0.699	48 (69.6%)	123 (74.5%)	0.718
Platinum resistant/refractory	19 (28.4%)	46 (27.9%)	17 (24.6%)		25 (30.1%)	38 (25.2%)		21 (30.4%)	42 (25.5%)
Optimal cytoreduction†,‡	41 (75.9%)	127 (64.1%)	48 (68.6%)	0.262	58 (62.4%)	117 (66.9%)	0.240	53 (64.6%)	122 (65.6%)	0.317
Suboptimal cytoreduction	13 (24.1%)	71 (35.9%)	22 (31.4%)		35 (37.6%)	58 (33.1%)		29 (35.4%)	64 (34.4%)
Germline or somatic BRCA mutation	7 (7.7%)	65 (29.6%)	19 (23.2%)	<0.001	33 (29.2%)	51 (27.0%)	<0.001	28 (30.1%)	56 (26.8%)	<0.001
Cases without BRCA mutation	84 (92.3%)	155 (70.4%)	63 (76.8%)		80 (70.8%)	138 (73.0%)		65 (69.9%)	153 (73.2%)
Germline or somatic HRD mutation §	12 (13.2%)	73 (33.2%)	24 (29.3%)	0.001	37 (32.7%)	60 (31.7%)	0.002	32 (34.4%)	65 (31.1%)	0.001
Cases without HRD mutation	79 (86.8%)	147 (66.8%)	58 (70.7%)		76 (66.3%)	129 (68.3%)		61 (65.6%)	144 (68.9%)

^*ANOVA or Fisher’s exact testing. All groups are compared with the Wildtype patient group.

^†Cases with unknown covariate status excluded from analysis.

^‡Data shown for stage III-IV disease only (total n = 322).

^§Includes mutations any mutations in BRCA1, BRCA2, ATM, BARD1, BRIP1, NBN, PALB2, RAD51C, and RAD51D.

Overall survival by TP53 mutation type:

To decrease confounding, survival analyses were restricted to HGSC. Using the structural classification, median survival was 37 months in cases with wildtype TP53, 49 months with missense mutations, and 49 months with nonsense/frameshift/splice mutations. Overall survival did not significantly diverge between structural mutation class (logrank p=0.844, Figure 2A). Using the functional classification, cases with GOF TP53 mutations demonstrated a median overall survival of 44 months while LOF TP53 mutations demonstrated a median survival of 50 months (logrank p=0.488, Figure 2B). Cases with hotspot missense TP53 mutations had a median survival of 41 months (logrank p=0.305, Figure 2C).

Figure 2:

Overall survival compared between TP53 mutation class. Analysis restricted to high grade serous cancer only. A) Structural TP53 mutation classification (WT = wildtype) B) Functional classification (LOF = loss-of-function, GOF = gain-of-function) C) Hotspot classification

Having a germline or somatic BRCA mutation did confer a survival advantage (logrank p<0.01). Cases were stratified by the presence of BRCA mutations to further analyze potential contributions of TP53 mutations. For cases with germline and somatic BRCA mutations, there was no divergence of survival curves between cases with or without TP53 mutations (logrank p=0.248, Supplementary Figure S1A). Cases without BRCA mutations also showed no significant difference in survival based on the presence or absence TP53 mutations (logrank p=0.994, Supplementary Figure S1B). Similarly, stratifying by all HRD mutations including BRCA did not show a significant effect on survival with TP53 mutations (logrank p=0.761 for cases with HRD mutations, p=0.780 for cases without HRD mutations).

On multivariate Cox testing of the associations between TP53 mutations with overall survival, primary platinum resistance was found to violate the proportional hazards assumption, necessitating stratification on this variable in addition to adjusting for age, presence of a BRCA mutation, optimal cytoreduction, and neoadjuvant chemotherapy. The presence of any TP53 mutation was not independently associated with survival (OR 1.15, p=0.601), nor were any specific TP53 mutations once divided into structural, functional or hotspot classifications (see Supplementary Table S5 for details).

TP53 mutation effects on primary platinum resistance:

There was a similar distribution of structural, functional, and hotspot TP53 mutations between women with primary platinum sensitive and resistant OC both within the entire cohort (Table 4) and HGSC only (Supplementary Table S4). Cases with concurrent BRCA and TP53 mutations demonstrated similar rates of platinum resistance to BRCA mutant cases that were TP53 wildtype (10/69, 14.5% TP53 mutation vs. 1/6, 16.7% TP53 wildtype, p=1.0). Cases without BRCA mutations also showed similar platinum resistance rates despite TP53 mutation status (53/165, 32.1% TP53 mutation vs. 18/61, 29.5% TP53 wildtype, p=0.749). On multivariate logistic regression adjusted for BRCA mutation status, age at diagnosis, optimal cytoreduction, and neoadjuvant chemotherapy (with stage omitted due to collinearity) and restricted to cases of HGSC, the presence of any TP53 mutation significantly increased the odds of platinum sensitivity (OR 0.41, 95% CI 0.17-0.99, p=0.048). Individually classified TP53 mutations each trended toward an association with platinum sensitivity but did not meet statistical significance (see Table 4 for details). In BRCA-wildtype cases only, the presence of a TP53 mutation still trended toward an association with platinum sensitivity though the results were not statistically significant in that smaller population (OR 0.38, 95% CI 0.14-1.01, p=0.053).

Table 4:

Multivariate logistic regression examining effects of TP53 mutation type on platinum resistance in high grade serous carcinoma cases by classification scheme. Models were adjusted by age, presence of a BRCA1 or BRCA2 mutation, optimal cytoreduction, and neoadjuvant chemotherapy. Stage was omitted from model due to collinearity.

	Odds Ratio	95% Confidence Interval	Wald P
All TP53 mutations	0.41	0.17-0.99	0.048
Structural classification
Wildtype	1.00 (Ref)
Missense	0.42	0.17-1.05	0.064
Nonsense/Frameshift/Splice	0.37	0.13-1.06	0.064
Functional classification
Wildtype	1.00 (Ref)
Loss-of-function	0.40	0.16-1.00	0.051
Gain-of-function	0.43	0.16-1.15	0.093
Location classification
Wildtype	1.00 (Ref)
Hotspot missense	0.44	0.16-1.24	0.121
Other mutation	0.39	0.16-0.98	0.046

Discussion

This is the largest single-institution cohort of OC with centralized pathology review and tumor sequencing which allowed for careful analysis of the distribution of TP53 mutation classes within the context of other clinicopathologic factors. Cases with germline or somatic BRCA mutations had significantly higher rates of TP53 mutations in general (regardless of classification), with few cases (7.7%) being TP53 wildtype. Our study is one of the largest prospective cohorts specifically examining the distribution of co-occurrent mutations in these two driver pathways and findings were further replicated using TCGA and PanCancer Atlas data. In ovarian and breast carcinoma, p53 IHC analyses have previously demonstrated more frequent aberrant expression reflective of missense mutations in BRCA-mutation carriers [6, 25–27]. Our data suggest that TP53 mutations may be a pre-requisite in most BRCA-associated carcinogenesis. The requirement for p53 protein dysfunction has been previously suggested in in vitro models as well as in genetically engineered mouse models [28–30]. These authors hypothesized that losing p53-mediated tumor suppression allows for the cells that exhibit loss of heterozygosity in BRCA to escape cell surveillance and apoptosis. The addition of our data further strengthens the argument that there is selection towards the co-occurrence of these mutations in early BRCA-associated malignant transformation. TP53 mutations have been thought to be nearly universal in HGSC, but this does not appear to be the case in BRCA wildtype HGSC, with a TP53 mutation rate of only 86.6% in our study and 69.6% in TCGA.

TP53 mutations were associated with primary platinum sensitivity in HGSC, regardless of type of mutation, after adjusting for covariates including BRCA mutation status. Association with platinum sensitivity has also been observed in triple negative breast cancer [31]. Conversely, studies in head and neck and lung cancers suggest that certain TP53 mutations may associate with platinum chemoresistance [32, 33]. There is scant literature in ovarian cancer regarding the relationship between TP53 mutations and platinum response, with inconsistent findings from smaller studies neither of which accounted for BRCA mutation status [1, 8]. Laboratory studies suggest that chemoresistance and sensitivity depend less on one mutated gene, but a careful interplay between different signaling pathways, transcription factors, and enzymes, and are reliant on the wildtype or mutated function of various genes [34]. While our results show that TP53 mutations may be used to prognosticate response to platinum, full understanding of chemoresistance likely relies on the analyses of multiple signaling pathways. As BRCA-mutated OC are often disproportionately sensitive to platinum-containing chemotherapy, further studies into the effect of TP53 must take these interactions into account.

Despite the association with platinum sensitivity, the presence of a TP53 mutation, regardless of mutational classification, was not associated with overall survival in HGSC in our univariate or multivariate analyses. Conclusions from several studies using the TCGA database of high-grade serous ovarian cancers have been mixed, depending on how TP53 mutations were classified [8, 9]. A study with an expanded panel of GOF mutations based on functional studies, more consistent with our methods, showed a difference between recurrence patterns in OC with wildtype and gain-of-function TP53 mutations, but no change in overall or progression-free survival [9]. Tuna and colleagues found no difference in survival outcomes between TP53 missense and nonsense or truncated mutations [10]. They tested hotspot mutations separately based on position, finding that G266, Y163C, and R282 mutations were associated with worsened overall survival compared with other mutations. GOF and hotspot mutations may thus need to be tested individually before being used as prognostic factors in OC, which would require large sample sizes for conclusive analysis. In all such analyses in OC, the confounding influence of BRCA mutation status on survival must be considered. In our study, TP53 mutations were not associated with survival when examined within strata of BRCA wildtype or mutated OC. Few prior studies have examined interactions between TP53 and BRCA mutations on survival, although these are the two most common mutations in HGSC. Mandilaras and colleagues [2] similarly found no association between overall survival and cases with concurrent TP53 and BRCA mutations compared to cases that were BRCA wildtype while Li and colleagues [15] found that the presence of TP53 and BRCA mutation conferred longer overall survival when compared with OC without either mutation. These analyses are limited by the rarity of BRCA-mutated but TP53-wildtype OC, thus making stratification of survival analysis difficult.

This study is the largest unselected cohort study of the effects of TP53 mutations on survival outcomes in OC. It is also the only study that compared different TP53 mutational classifications and their interactions with BRCA mutations. Our methods were strengthened by centralized pathologic review, rigorous determination of somatic mutations through next generation sequencing, and an enhanced list of defined GOF mutations based on currently available literature. Our cohort was not powered for subset analysis and our general understanding of GOF/LOF classifications is still evolving [35], but this provides valuable preliminary investigation into potential interactions of commonly seen somatic and germline mutations in OC. An understanding of patient-specific mutations and interactions is vital for personalization of cancer therapy. Our data suggest that BRCA mutations are a major confounder in associations of TP53 mutation status and outcome in OC.

Supplementary Material

1

Supplementary Figure S1: Overall survival of patients with high grade serous carcinoma, comparing those with and without TP53 mutations, stratifying by the presence of germline or somatic BRCA mutations. A) Restricted to carcinomas with BRCA mutations B) Restricted to carcinomas without BRCA mutations

2

Supplementary Table S1: List of gain-of-function TP53 mutations used for the Functional classification scheme [11, 36, 37]

Supplementary Table S2: Distribution of TP53 and BRCA mutations amongst cases of high grade serous ovarian carcinoma and other histologies.

Supplementary Table S3: Distribution of TP53 and BRCA mutations amongst non-high grade serous cases

Supplementary Table S4: Comparison of TP53 mutation types between clinicopathologic subgroups in high grade serous carcinoma cases only

* ANOVA or Fisher’s exact testing. All groups are compared with the Wildtype patient group.

^† Data shown for stage III-IV disease only (total n=253

^‡ Includes mutations any mutations in BRCA1, BRCA2, ATM, BARD1, BRIP1, NBN, PALB2, RAD51C, and RAD51D

Supplementary Table S5: Multivariate cox proportional hazards model examining effects of TP53 mutation type on overall survival in high grade carcinoma cases based on TP53 classification scheme. Models were stratified by platinum resistance to meet the proportional hazards assumption, and adjusted by age, presence of a BRCA1 or BRCA2 mutation, stage, optimal cytoreduction and neoadjuvant chemotherapy.

Highlights:

1. TP53 mutations were identified in 76.8% of ovarian carcinomas and 87.9% of high grade serous carcinomas in this cohort

2. BRCA1 or BRCA2 mutations co-occur with TP53 mutations in ovarian carcinoma

3. TP53 mutations are associated with platinum sensitivity, but not overall survival, in high grade serous ovarian carcinoma

4. TP53 mutations were identified in 76.8% of ovarian carcinomas and 87.9% of high grade serous carcinomas in this cohort

5. BRCA1 or BRCA2 mutations co-occur with TP53 mutations in ovarian carcinoma

6. TP53 mutations are associated with platinum sensitivity, but not overall survival, in high grade serous ovarian carcinoma