Δ133p53 is an independent prognostic marker in p53 mutant advanced serous ovarian cancer

Background: We aimed to evaluate the clinical relevance of p53 and p73 isoforms that modulate the function of p53. Methods: This prospective multicentre study included 154 patients with stage III and IV serous ovarian cancer. A functional yeast-based assay and subsequent sequencing were performed to analyse the p53 mutational status. Expression of p53 and p73 isoforms was determined using RT–qPCR. Results: Δ133p53 expression constituted an independent prognostic marker for recurrence-free (hazard ratio=0.571, P=0.016, 95% CI: 0.362–0.899) and overall survival (hazard ratio=0.365, P=0.004, 95% CI: 0.182–0.731) in patients with p53 mutant ovarian cancer (n=121). High Δ40p53 expression was associated with favourable tumour grading (P=0.037) and improved recurrence-free survival (33.4 vs 19.6 months, P=0.029), but not overall survival (43.1 vs 33.6 months, P=0.139), in patients with p53 wild-type cancer (n=33). Neither the p53 mutational status nor p73 isoform expression possessed prognostic significance in the examined ovarian cancer cases. Conclusion: Δ133p53 expression was associated with prognosis in the vast majority of ovarian cancer cases, that is, patients with p53 mutant advanced serous carcinomas. Thus, our findings underline the importance of considering the complex p53 regulatory network.

Recently, the N-terminally truncated p53 isoforms D40p53 and D133p53 were added to the complex network that modulates the function of the tumour suppressor p53. D40p53 is generated by alternative splicing of intron 2 or altered initiation of translation in exon 4 (Courtois et al, 2002;Yin et al, 2002;Ghosh et al, 2004), whereas D133p53 is derived from an alternative promoter located in intron 4 (Bourdon et al, 2005). D40p53 lacks the first 40 amino acids, but retains the second transactivation domain (Murray-Zmijewski et al, 2006). By contrast, D133p53 is devoid of both transactivation domains and part of the DNA-binding domain ( Figure 1). Both p53 isoforms contain the C-terminal tetramerisation domain, allowing their incorporation into p53 tetrameres. The exact function of D40p53 and D133p53 has not yet been fully characterised. The present consensus is that their ratio to FLp53 determines functional outcome . Low D40p53 and D133p53 levels have been reported to act as potent dominantnegative inhibitors of FLp53, suppressing transcription of genes under the control of a p53-binding element as well as p53-induced apoptosis (Yin et al, 2002;Bourdon et al, 2005;Graupner et al, 2009). Contrarily, high D40p53 and D133p53 levels have resulted in enhanced transcription of FLp53 target genes (Chen et al, 2005;Ohki et al, 2007;Powell et al, 2008). Studies analysing the expression patterns of N-terminally truncated p53 isoforms in distinct types of carcinomas are rare and have not shown a clinical relevance of p53 isoforms (Boldrup et al, 2007;Avery-Kiejda et al, 2008;Ebrahimi et al, 2008;Chen et al, 2009;Song et al, 2009).
Similarly, the p53 family member p73 gives rise to multiple isoforms. Alternative splicing of the P1 promoter transcript generates both the full-length TAp73 as well as DN'p73, whereas an alternative P2 promoter in intron 3 produces DNp73. Importantly, DN'p73 and DNp73 transcripts encode the same protein product lacking the transactivation domain. This N-terminally truncated p73 protein DNp73 acts as a powerful dominant-negative inhibitor of both wild-type p53 and TAp73, either by direct competition for DNA binding sites or by the formation of heterocomplexes (Bailey et al, 2011). DNp73 has been an independent prognostic marker in distinct types of carcinomas (Uramoto et al, 2004;Müller et al, 2005;Liu et al, 2006;Vilgelm et al, 2010). In ovarian cancer, we previously reported that expression of N-terminally truncated p73 isoforms has a role in response to platinum-based chemotherapy and constitutes an independent prognostic marker in patients with p53 mutant ovarian cancer (Concin et al, 2005).
Most existing studies in ovarian cancer comprise patient groups heterogeneous for well-known prognostic factors such as FIGO stage and histological subtypes. While 5-year survival rates can be as high as 80 -95% among patients with early-stage disease (stage I or II), patients with advanced carcinomas (stage III or IV) have survival rates of 10 -30% (Cannistra, 2004). Recently, the histological subtype was reported to be an independent prognostic marker in patients with stage III ovarian cancer. Mucinous and clear cell carcinomas possessed an impaired prognosis as compared with serous ovarian carcinomas, whereas endometrioid carcinomas showed favourable prognosis (Winter et al, 2007). In addition, histological subtypes display different biomarker expression profiles, thus further supporting the hypothesis that histological subtypes represent different disease entities. For instance, serous carcinomas show WT1, Mesothelin, oestrogen receptor and CA125 expression in 475%, while the mucinous subtype displays frequent expression of Matriptase, and endometrioid carcinomas express high rates of estrogen and progesterone receptor and CA125 (Köbel et al, 2008). Heterogeneous patient groups might substantially confound results of biomarker studies and constitute an important reason for the difficulty experienced in confirming potential prognostic markers in subsequent studies.
Thus, a homogeneous cohort of patients with primary advanced serous ovarian cancer was prospectively recruited within the multicentre study OVCAD (OVarian CAncer Diagnosis of a silent killer). We aimed to evaluate the clinical relevance of p53 (D133p53, D40p53, and FLp53) and p73 isoforms (TAp73 and DTAp73) that modulate the function of p53 in patients with advanced serous ovarian cancer.

Patients and tissue samples
Between August 2005 and December 2008, 154 consecutive patients diagnosed with primary advanced serous ovarian cancer at the Departments of Gynecology and Obstetrics in Berlin (n ¼ 55), Leuven (n ¼ 50), Hamburg (n ¼ 26), Vienna (n ¼ 19), and Innsbruck (n ¼ 4) were enrolled in the OVCAD project, a Sixth Framework Program Project of the European Union (http:// www.ovcad.eu). The Ethics Committees of the respective centres approved the study protocol. All patients signed written informed consent before enrollment.

RNA isolation and real-time RT -PCR
Tissue samples were obtained at the time of diagnosis and immediately stored in liquid nitrogen. RNA isolation was done by 1600 nucleic acid prepstation (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions including a DNase digestion step. Eluted RNA was precipitated with ethanol and resuspended in RNase-free water. Reverse transcription was performed as described previously (Reimer et al, 2007).
Primer pairs and probes for p53 (D40p53, D133p53, and FLp53) and p73 isoforms (TAp73 and DTAp73, which recognises both DNp73 and DN'p73 transcripts) as well as the internal control TBP (TATA box-binding protein) were designed using Primer Express software (Applied Biosystems; Supplementary Table 1). Real-time TaqMan RT -PCR was performed using the ABI Prism 7900 Detection System (Applied Biosystems) according to the manufacturer's recommended protocol. Each reaction was performed in duplicate. To determine absolute copy numbers for all p53 and p73 isoforms, standard curves were generated as reported previously (Concin et al, 2004).

Analysis of p53 mutational status
To detect alterations in the p53 gene resulting in a functionally inactive protein, a functional yeast-based assay and subsequent sequencing were performed as described in detail previously (Concin et al, 2004). Throughout the manuscript the term 'mutant p53' is used synonymously for 'functionally inactivated p53' as determined by the yeast-based assay.

Statistical analysis
The Shapiro -Wilk test was performed to assess the normality assumption. As the distribution of p53 and p73 isoforms was non-Gaussian, the Mann -Whitney U-test was performed to compare isoform expression and age with p53 mutational status. The w 2 -test was performed to examine the relationship between the p53 mutational status and categorical clinicopathological parameters. For correlations among isoforms, Spearman's correlation coefficients were calculated.
Cases were divided by the 50th percentile of p53 and p73 isoform expression levels into two approximately same-sized groups (i.e., a high-expressing group and a low-expressing group of isoform expression). In addition, a TAp73/DTAp73 quotient was calculated and again divided by the 50th percentile. The w 2 -test was used to examine the relationships between isoform expression and clinicopathological parameters.
Survival probabilities were calculated with the product limit method of Kaplan and Meier. The Cox proportional hazards model was used for multivariate analysis to assess the independence of different prognostic factors. Statistical Package for the Social Sciences for Windows 18.0 software (SPSS, Inc., Chicago, IL, USA) was used for all analyses. P-values o0.05 were considered statistically significant.

Patient characteristics
In all, 154 patients diagnosed with serous ovarian cancer were enrolled. Median age was 57 years (range: 26 -83). Of the patients, 81.8% (126 of 154) presented with FIGO stage III and 18.2%  Table 1 provides the prognostic relevance of clinicopathological variables in univariate and multivariate analysis.
Median follow-up was 24.5 months (range: 1 -49). Of the patients, 24.8% (38 of 153) suffered treatment failure, defined as primary progression or early recurrence within 6 months after termination of primary platinum-based chemotherapy. In one patient, treatment response could not be determined due to short follow-up.

Clinical relevance of p53 mutational status
Of the ovarian cancer specimens, 78.6% (121 of 154) harboured p53 mutations and 21.4% (33 of 154) were wild-type p53. p53 mutant carcinomas were significantly associated with older age at diagnosis (median 58 vs 50 years, Po0.001) and adverse tumour grade (III vs I/II, Po0.001) compared with p53 wild-type carcinomas. Tumour stage, residual disease, and treatment response did not differ with respect to p53 mutational status (Table 2).
Patients with p53 mutant ovarian cancer showed impaired recurrence-free (mean 22.4 vs 25.7 months) and overall survival (mean 37.1 vs 39.0 months) as compared with patients with p53 wild-type cancer. Neither difference was statistically significant (P ¼ 0.223, P ¼ 0.291).
p53 and p73 isoform expression levels did not differ between p53 mutant and p53 wild-type ovarian cancer specimens. Median copy numbers stratified by p53 mutational status are provided in Supplementary Table 2. p53 isoforms influence prognosis depending on p53 mutational status In patients with p53 mutant ovarian cancer, high D133p53 expression was associated with significantly improved recurrence-free and overall survival (Table 3; Figure 2A and B). A mean recurrence-free survival of 27.0 (95% CI: 22.5 -31.5) months was observed in patients with high D133p53 expression as compared with 17.5 (95% CI: 14.6 -20.4) months in patients with low D133p53 expression (P ¼ 0.002). Mean overall survival was 41.1 (95% CI: 37.4 -44.7) months in patients with high D133p53 expression and 32.7 (95% CI: 28.4 -37.1) months in patients with low D133p53 expression (P ¼ 0.007). The median survival times were not determined as o50% of patients were dead at the time of analysis, thus means are provided.
In patients with p53 wild-type cancer, high D40p53 expression was significantly associated with improved recurrence-free (33.4 vs 19.6 months, P ¼ 0.029), but not with overall survival (43.1 vs 33.6 months, P ¼ 0.139; Table 3; Figure 2C and D). Due to small case numbers, multivariate analysis could not be performed.
In the entire group of serous ovarian cancer cases, p53 isoforms did not possess prognostic relevance. Detailed survival data are given in Supplementary Table 3.  High 133p53 Low 133p53

High 40p53
Low 40p53 Low 40p53 Figure 2 Kaplan -Meier survival graphs for p53 isoforms. High D133p53 expression was associated with improved recurrence-free (A) and overall survival (B) in 121 patients with p53 mutant ovarian cancer in univariate analyses. In patients with p53 wild-type cancer (n ¼ 33), high D40p53 expression levels predicted improved recurrence-free (C), but not overall survival (D) in univariate survival analyses. Cases were divided at the 50th percentile of p53 isoform expression levels into a high-and a low-expressing group. P-value was determined with the log-rank test.
D133p53 as independent prognosticator in ovarian cancer G Hofstetter et al

Correlations between p53 isoforms and clinicopathological parameters
In patients with p53 mutant cancer, p53 isoform expression was not associated with clinicopathological parameters. In patients with p53 wild-type cancer, D40p53 expression levels correlated with favourable tumour grading (grade I/II vs III, P ¼ 0.037). In the entire group of advanced serous ovarian cancer cases, no correlation was seen between p53 isoforms and clinicopathological parameters. Significant correlations between all p53 isoforms were observed in p53 mutant cases as well as the entire group of ovarian carcinomas, whereas the only correlation found in p53 wild-type cases was between D40p53 and FLp53 expression. Detailed information is provided in Supplementary Table 4. p73 isoforms lack prognostic significance Neither TAp73 nor DTAp73 expression was associated with prognosis in the examined ovarian cancer cases. Given the fact that DTAp73 acts in a dominant-negative manner on TAp73, we tested the hypothesis that patients might have different clinical outcomes depending on the TAp73/DTAp73 ratio. This ratio also did not possess prognostic relevance in the patients with advanced serous ovarian cancer. In patients with p53 wild-type ovarian cancer, TAp73/DTAp73 ratio correlated with favourable tumour grading (grade I/II vs III, P ¼ 0.011).
Correlations between p73 isoforms are provided in Supplementary Table 2.

DISCUSSION
We herein provide first evidence for a potential clinical role of N-terminally truncated p53 isoforms with respect to the p53 mutational status in ovarian cancer. Only one previous study has analysed the expression of p53 isoforms in ovarian cancer (Marabese et al, 2007). Marbese et al (2007) found that FLp53 was significantly elevated in early as compared with advanced disease. However, p53 isoforms lacked prognostic significance in their group of patients with ovarian carcinomas comprising various histological subtypes.
In the present study, D133p53 expression constituted an independent prognostic marker in patients with p53 mutant advanced serous ovarian cancer, which represents the vast majority of ovarian cancer cases. High D133p53 expression resulted in a 43% risk reduction for recurrence and a 64% risk reduction for death as compared with low D133p53 expression. To date, only scarce information is available on the function of this N-terminally truncated p53 isoform ). Existing in-vitro studies evaluated the function of D133p53 either alone or in the presence of wild-type p53 (Bourdon et al, 2005;Chen et al, 2005). D133p53 has been shown to form heterocomplexes with wild-type FLp53, resulting in the dominant-negative inhibition of FLp53 (Bourdon et al, 2005). The present clinical study, however, demonstrates a favourable role of D133p53 in p53 mutant carcinomas. It is unclear how D133p53 might exert a beneficial function in the presence of mutant p53. We hypothesise that D133p53 may interact with mutant p53, thereby abolishing negative effects of mutant p53. Mutant p53 has been reported to counteract wild-type FLp53 and TAp73 in a dominant-negative manner, and specific mutants possess oncogenic gain-of-function mutations (Oren and Rotter, 2010). Functional studies analysing the role of D133p53 in the presence of mutant p53 are highly warranted.
A possible explanation for differing D133p53 levels in cancer specimens observed in the present study might be provided by the presence of several polymorphisms within the internal promoter region located in the intron 4 of the p53 gene, which gives rise to D133p53 (Bellini et al, 2010). Bellini et al (2010) reported that these polymorphisms are associated with differences in promoter activity by changing the affinity for distinct transcription factors.
In the present study, D40p53 influenced recurrence-free but not overall survival in p53 wild-type ovarian cancer. We assume that the latter is related to the short follow-up period in the present study. Existing in-vitro studies suggest that D40p53 might enhance the function of wild-type FLp53 . D40p53 and FLp53 have been reported to readily form heterocomplexes (Courtois et al, 2002;Ghosh et al, 2004). As D40p53 lacks the MDM2-binding site, these heterocomplexes escape MDM2mediated degradation and therefore accumulate (Yin et al, 2002). In addition, D40p53 supports a conformation of FLp53 that is associated with a more active state (Powell et al, 2008). Furthermore, Powell et al (2008) reported that D40p53 alters the post-translational modification profile of FLp53. Post-translational modifications at the N-terminus of FLp53, for instance, might increase the recruitment of transcriptional co-activators such as p300 and PCAF, and thus be responsible for increased promoterbinding capacity of the heterocomplexes. Furthermore, D40p53 alone has been reported to induce apoptosis through the transcriptional activation of many apoptosis-related genes that are not induced by FLp53, such as TP53BP2 (tumour protein p53 binding protein 2) and TIAL1 (TIA1 cytotoxic granule-associated RNA binding protein-like 1) (Ohki et al, 2007). However, the small number of p53 wild-type cases has to be considered when interpreting our finding.
p53 mutational status, determined with a highly sensitive yeastbased assay, did not impact prognosis in the examined patients with advanced serous ovarian cancer. In our previous retrospective study including a heterogeneous group of various histological subtypes and stages, p53 mutational status constituted a significant prognostic marker in univariate, but not in multivariate analyses (Concin et al, 2005). A recent meta-analysis evaluated the role of p53 mutational status in ovarian cancer (de Graeff et al, 2009). In studies restricted to serous carcinomas, a modest effect of p53 on overall survival was present (HR ¼ 1.61, 95% CI: 1.09 -2.38). The present study agrees with the meta-analysis restricted to advanced disease, in that it found no prognostic value of p53.
Existing studies in ovarian cancer have reported inconsistent results on the clinical relevance of p73 isoforms (Buhlmann and Pützer, 2008). High levels of N-terminally truncated p73 isoforms were found to be independently associated with impaired prognosis in p53 mutant cases (Concin et al, 2005). In contrast, high DNp73 expression correlated with better overall survival D133p53 as independent prognosticator in ovarian cancer G Hofstetter et al irrespective of p53 mutational status in another study (Marabese et al, 2007). However, the present prospective study was not able to verify a clinical role of p73 isoforms in ovarian cancer. Despite the myriad existing studies investigating the extremely complex network regulating p53 function, many details remain to be explored. The generation of multiple p53 isoforms adds an additional layer of complexity to this fascinating protein. It is increasingly being recognised that the classification of carcinomas into 'p53 wild-type' and 'p53 mutant' is an oversimplification that does not acknowledge the actual activity of the p53 pathway. Our clinical findings suggest that neglecting interactions between p53 and its N-terminally truncated isoforms might constitute an important reason for the difficulties encountered when attempting to correlate p53 with prognosis and treatment response. In addition, these complex interactions also have to be considered in therapeutic efforts aiming to directly affect p53 function, such as p53-gene therapy, vaccinating against p53, and the use of MDM2 inhibitors to activate the p53 antitumour response and cytoprotective function (Cheok et al, 2011). In future, upregulation of D133p53 in cancer cells could be an elegant means of improving outcome in patients with p53 mutant ovarian cancer. Thus, the tumour suppressor p53 remains a highly dynamic and rapidly expanding area of research.
Our findings underline the importance of considering the complex p53 regulatory network in clinical studies. In p53 mutant advanced serous ovarian cancers, which represent the vast majority of patients, D133p53 expression levels discriminate between two groups of patients with substantially different clinical outcome. High D133p53 expression resulted in a 43% risk reduction for recurrence and a 64% risk reduction for death as compared with low D133p53 expression.