2.1 Study population and clinical samples

Three hundred ninety-eight patients were enrolled in the MITO16A/MaNGO OV-2 study.⁷ Formalin-fixed paraffin-embedded (FFPE) chemo-naïve tissue specimens were collected at clinical collaborating centers and sent to the INT G. Pascale of Naples, which centralized them. Sample quality check, tissue microarray (TMA) building, and nucleic acid extraction were performed by the INT G. Pascale of Naples, following recently published procedures.¹³

The flowchart reporting the causes of sample exclusion and the final number of tissues analyzed for p53 IHC and TP53 mutations is provided in Figure S1.

2.2 Determination and classification of TP53 mutations

Up to 200 ng of DNA obtained from FFPE specimens was used to construct next-generation sequencing (NGS) libraries targeting the TP53 gene. To this purpose, an amplicon-based mutation detection panel was designed for the TP53 gene (QIAGEN, USA), covering all coding and UTR exons, padding the exon-intron junctions by 10 bp to also detect splicing mutations. Briefly, up to 200 ng of tumor DNA was fragmented, barcoded, and enriched with the mentioned custom panel for the TP53 gene following the manufacturer's instructions (QIAseq targeted DNA, QIAGEN, USA). Libraries obtained were then amplified and checked for quality (4200 Tape Station System, Agilent Technologies, USA) and amount (Qubit dsDNA HS assay kit, Thermo Fisher, USA). Finally, the barcoded libraries were pooled and sequenced on an Illumina NextSeq 500 (Illumina, USA) sequencer, to reach an approximate coverage of 2500× for each sample. DNA extracted from blood samples was used as a control and subjected to the same preparation. Following sequencing, raw data from the instruments were aligned to the human reference genome (hg19) with “bwa”¹⁴ and de-duplication to remove artifacts using Universal Molecular Identifiers was performed with the “fgbio” software suite (https://github.com/fulcrumgenomics/fgbio). Seventy blood samples with high-quality data and no known germline variants were then pooled to create a panel of normal (PoN) with MuTect2. Variant calling of somatic TP53 variants was performed by running MuTect2 and VarDict^{15, 16} for each tumor sample against the previously prepared PoN. Variant effect and severity were calculated with the Variant Effect Predictor.¹⁷ Variants were then filtered to remove low-quality calls (depth <200×, variant allele fraction <10%) and known benign variants (general population frequency >1%, known benign variants in ClinVar). A second filtering step was performed to only include high-quality mutations with potential or known deleterious effects: known deleterious TP53 mutations were annotated with the IARC TP53 database (version 19, https://tp53.isb-cgc.org/) and additional missense mutations were screened for tumor driver potential with the Cancer Genome Interpreter (https://www.cancergenomeinterpreter.org/2018),¹⁸ only keeping mutations with a predicted driver value of “Tier 1” (high confidence) or “Tier 2” (medium confidence). The sequencing coverage and quality statistics for each sample are summarized in Table S1.

TP53 mutations were classified into three categories based on the type of mutation: (1) missense mutations, when a single nucleotide change codes for a different amino acid; (2) frameshift insertion or deletion or nonsense mutations resulting in a truncated, unstable protein; and (3) undetermined mutations, including splice site or intronic mutations or in-frame insertions and deletions. Moreover, TP53 mutations were divided into groups with respect to function: (1) missense mutations that result in “gain of function” that, as a consequence, not only abolish the tumor-suppressor activity but also endow the protein with oncogenic activities (oncomorphic mutations); (2) loss of function, frameshift insertion or deletion, or nonsense mutations resulting in a complete loss of tumor-suppressing activity; and (3) unclassified missense mutations that refer to single base substitutions not fulfilling oncomorphic criteria and that currently have an unknown function. “Gain of function” missense mutations (P151S, Y163C, R175H, L194R, Y220C, R248Q, R248W, R273C, R273H, R273L, and R282W) were defined according to Brachova et al.^{19, 20} and confer to the mutated protein an oncogenic phenotype, as demonstrated in in vitro and in vivo preclinical models.

2.3 TMA building, p53 IHC, and scoring

MITO16A/MaNGO OV-2 TMAs were created by selecting representative areas for each case from hematoxylin and eosin stained sections. The seven TMAs were constructed using an automated TMA (Galileo CK3500 TMA, ISENET, Milan, Italy) and contain a total of 358 tissue cores from 236 primary and 122 metastatic OC samples. Three 1-mm cores have been collected from different areas of each tumor block in order to overcome sample heterogeneity and the possible loss of tissue due to cutting. About 4-μm-thick sections were cut from TMAs and stained with anti-p53 (clone DO-7, 1:2, Agilent) antibody on a Bond Max automatic immunostainer (Leica Biosystems). Immunoreaction was revealed by using Novolink Polymer (Leica Microsystems), followed by diaminobenzidine as a chromogen and hematoxylin as a nuclear counterstain. Stained slides were acquired using an Aperio C2 digital scanner and ScanScope software, and images were taken as a snapshot by using ImageScope, RRID:SCR_014311. Nuclear immunoreactivity was evaluated by two independent investigators (MB and LA), blinded for clinicopathological characteristics and p53 gene mutational data, and an established scoring system was applied.²¹ In cases of discrepant scoring results, a consensus meeting with a third investigator (EK) was performed to solve the discordance. In detail, an abnormal p53 pattern includes (i) tumors with “overexpression” (OE), which was attributed to cases with 60%–100% of tumor cells showing intense nuclear p53 staining, or (ii) tumors completely negative for p53 in the presence of positive internal controls, defined as having a pattern of “complete absence,” so-called “null” (CA). Strong and diffuse (OE) p53 immunoexpression plus the CA of staining are interpreted as aberrant/abnormal p53 expression, as established previously.²¹ The normal or wild-type (WT) pattern was characterized by heterogeneous p53 staining, an alternating patchy pattern with negative staining, more commonly weak and focal.

2.4 Statistical analysis

Quantitative variables were described as median and interquartile range, while qualitative variables were reported as absolute numbers and relative frequency. The associations between biomarker (TP53 mutation and type/function classification) and the clinical prognostic factors were investigated using a chi-squared test. The prognostic effect of TP53 (mutation, type, and function) was evaluated using PFS and overall survival (OS) as endpoints. PFS was defined as the time elapsing from inclusion in the study to the first occurrence of either death for any cause or disease progression. OS was defined as the time elapsing between inclusion in the study and death for any cause. Kaplan–Meier curves were drawn for PFS and OS, the medians of survival were calculated and comparison was performed by a two-sided log-rank test. Furthermore, the prognostic role of TP53 on both PFS and OS was investigated using univariable and multivariable Cox proportional hazard models. Proportionality assumption was assessed using the Schoenfeld residuals. For each model, hazard ratios (HRs) and 95% confidence intervals (CIs) were reported. A multivariable model was performed using as covariates: Eastern Cooperative Oncology Group performance status (PS) (0 vs. 1–2), residual disease (none; ≤1 cm; >1 cm + not operated), The International Federation of Gynecology and Obstetrics (FIGO) stage (III vs. IV), and tumor histology (high-grade serous vs. other). Covariates were chosen according to the model defined in the manuscript reporting the clinical results of the trial.⁷ Data were analyzed using R software version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

3.1 Patient cohort

Among 398 patients enrolled in the MITO-16A/MaNGO-OV2 study, the following patients were excluded from the final p53 IHC evaluation: 13 patients because the samples were not available for research; 12 patients who underwent neo-adjuvant chemotherapy and could not provide treatment-naïve tissue; 15 because of inadequate tumor sample; 42 because the samples had inadequate morphology for TMA; 5 because tumor samples were exhausted after nucleic acid extraction. Samples from 311 patients were used for TMA construction and underwent p53 immunohistochemical staining (Figure S1).

For NGS, 25 out of 311 patients were excluded because the tumor sample was inadequate for DNA extraction. Of 286 remaining patients, the TP53 mutational status was investigated in 202 patients by means of NGS (Figure S1).

3.2 TP53 mutational analysis

For 197 out of 202 patients, TP53 mutational status was successfully determined. This cohort did not show any significant difference in terms of clinicopathological characteristics from the whole trial population (n = 398) (Table S2). Among 197 patients, 180 (91.3%) had tumors harboring TP53 mutation, while the remaining 17 (8.7%) had TP53 WT neoplasms. Table 1 describes the different types of TP53 mutations detected, with missense mutations representing the most common (n = 119, 66.1%), followed by truncating mutations (n = 51, 28.3%), mostly frameshift insertions and deletions (n = 34, 18.8%), and nonsense mutations (n = 16, 8.9%). Finally, there were 10 undetermined mutations, consisting of in-frame insertions and deletions, splice site, and intronic mutations. A total of 146 out of 180 (81.1%) hotspot mutations were observed, with the most common occurring at codons R273 (15 out of 146, 10.3%), R248 (11 out of 146, 7.5%), Y220 (8 out of 146, 5.5%), and R175 (6 out of 146, 4.1%). A detailed description of all detected TP53 mutations (in terms of variant types, function, domain, and allele frequency) with information on hotspot mutations is reported in Table S3.

In Table 1, the different functions of TP53 mutations are displayed and classified as: (i) gain of function (n = 43, 23.9%), that either abolish p53 WT functions or confer novel oncogenic properties and are represented by most, but not all, missense mutations; (ii) loss of function (n = 57, 31.7%), that significantly abate p53 tumor suppressor activity and are represented by all nonsense and frameshift mutations; and (iii) unclassified (n = 80, 44.4%), mostly missense mutations with unknown or undefined function. Regarding mutation site, as expected, the majority of TP53 mutations (158 out of 180, 87.8%) were localized in the DNA binding domain, covering exons 4–8, while a minority of them were found in the amino- (5.6%) or carboxy-terminal domains (4.4%) or in the splice (1.6%) or intron regions (0.6%) (Table 1 and Figure S2).

3.3 Association of TP53 mutational status with clinical features and survival

TP53-mutated tumors were significantly associated with high-grade serous histology (p = .0025) and with no other clinicopathological characteristics (Table 2). Type of mutation (missense, truncating, and undetermined) and function (gain of function, loss of function, and unclassified) were also significantly correlated with tumor histology (Tables S4 and S5).

As a result of the univariable survival analysis, the presence of a TP53 mutation was significantly prognostic for a favorable outcome in our cohort of patients. In particular, a significant association with an increased OS was observed for TP53-mutated tumors compared to WT cases (HR = 0.48, 95% CI = 0.24–0.97, p = .039), as represented in Figure 1A (median OS, 41.2 vs. 30.6 months, respectively). The significantly positive prognostic value of mutated TP53 on OS was also maintained when a multivariable Cox survival analysis was applied, adjusted for PS, FIGO stage, residual tumor and histology (HR = 0.46, 95% CI = 0.22–0.96, p = .039, Figure 2A). Differently, no significant difference in PFS was observed between patients with TP53 mutated tumors compared to WT by univariable analysis (p = .21, Figure S3).

Then, the univariable survival analysis was performed considering the four groups: (1) WT, (2) missense mutations, (3) truncating mutations, and (4) undetermined mutations. We observed that the presence of a missense mutation was of borderline significance for improved OS (HR = 0.49, 95% CI = 0.24–1.01, p = .052, Figure 1B). This trend becomes statistically significant (HR = 0.46, 95% CI = 0.22–0.97, p = .043) in multivariable survival analysis, where missense mutations were associated with a favorable OS compared to WT (Figure 2B). Finally, we analyzed survival in patients harboring tumors with TP53 mutations classified according to function: (1) WT, (2) gain of function, (3) loss of function, and (4) unclassified. The presence of an unclassified mutation was indicative of an improved OS both in univariable (HR = 0.43, 95% CI = 0.20–0.92, p = .03, Figure 1C) and multivariable analysis (HR = 0.39, 95% CI = 0.18–0.86, p = .02, Figure 2C).

3.4 p53 immunohistochemical results

Out of 311 patients stained for p53 protein expression, 17 were excluded due to technical problems. Therefore, 294 patients with p53 immunostaining were available for further evaluation and analysis (Figure S1). Patients in the p53 cohort (n = 294) and in the whole MITO-16A/MaNGO-OV2 trial population (n = 398) did not show any difference in terms of clinicopathological characteristics (Table S2). p53 abnormal immunostaining was observed in 267 out of 294 (90.8%) OCs, while the remaining 27 (9.2%) showed a WT pattern (Figure 3A). Specifically, 206 cases showed an OE pattern and 61 tumors displayed CA staining (Figure 3B,C).

3.5 Association of p53 immunohistochemical results with clinicopathological features and survival

Abnormal p53 immunohistochemical expression was significantly associated with high-grade serous histology, with both OE and CA staining enriched in this subset of patients (p < 0.0001, Table S6). Applying a univariable survival analysis, no significant difference was observed in PFS according to immunohistochemical staining (Figure 4A). We observed an HR of 0.70 for OS in patients with p53 IHC overexpression compared to p53 WT patients, which however did not reach statistical significance (p = .31, 95% CI = 0.36–1.38), as displayed in Figure 4B.

3.6 Association of TP53 mutation status with p53 IHC

For 197 patients, both TP53 mutational status and p53 immunohistochemical staining results were available. As reported in the literature, tumors with truncating mutations generally showed p53 CA staining (72.5%), cases harboring missense mutations were predominantly p53 OE (96.6%), while WT tumors showed focal, weak p53 staining, as shown in Figure 3D. Overall, the degree of concordance between protein expression and mutational analysis was 91% (Table S7). Discordant cases were a limited number, only 18 out of 197 (9%), mainly present in the WT group (Figure 3D). The detailed list of 18 discordant cases is represented in Table S8.