Introduction

TP53 is a well-known tumor suppressor gene that has been extensively studied over these past decades in cancers. TP53 encodes p53, a transcription factor regulating cell cycle to prevent proliferation of genetically damaged cells with oncogenic properties. In addition, TP53 is involved in modulation of multiple cell functions that are pivotal in cancer biology. Indeed, mutant p53 participates to tumor cells invasion, proliferation and survival [1]. Somatic TP53 alterations are frequent in most human cancers, ranging from 5–80 % depending on type and stage of tumors. Most of these alterations are missense mutations (~75 %) leading to complete or partial loss of p53 functions (http://p53.iarc.fr/).

TP53 mutation is an early event in gliomagenesis [2–4]. In diffuse gliomas, and particularly in astrocytomas, TP53 is frequently mutated (~50 %). TP53 is also found mutated in oligoastrocytomas and oligodendrogliomas, albeit at a lower rate (~40 % and ~10 % of cases, respectively) [5]. Most studies investigating TP53 in diffuse gliomas were focused on mutational hotspots (exons 5–8 sparing exons 1, 2, 3, 4, 9, 10 and 11) (Supplementary Data 1). However, it has been shown that in primary glioblastoma that frequent TP53 mutations are also located outside these classic mutational hotspots within the transactivation domain, the prolin-rich domain and the splice donor sites [6].

Although the threshold of p53-positive tumor cells predicting TP53 mutation is not consensual, p53 expression determined by immunohistochemistry (IHC) is used as a surrogate marker of TP53 mutations in tumors. Indeed, many TP53 mutations result in p53 stabilization and an increased percentage of stained cells using IHC. However, this parallel is not appropriate for null mutation since there is no protein expression and for other gene alterations not leading to p53 accumulation [7].

The prognostic value of TP53 mutations remains uncertain, especially in diffuse low grade gliomas (WHO grade II, LGG). Heterogeneity of cohorts and techniques used to assess TP53 or p53 statuses may explain, at least partly, the conflicting results reported in the literature [3, 5, 7–11].

These data prompt us to sequence all the TP53 coding-exons using high-throughput sequencing method and to assess in a large retrospective series of LGG: (i) the prevalence of TP53 mutations, (ii) the correlation between TP53 mutation and p53 expression and (iii) the pathological and clinical values of TP53 mutation and p53 expression.

Materials and methods

Material

Sixty-one (n = 61) patients and tumors operated between July 1987 and December 2011 in the groupe hospitalier Pitié Salpêtrière were included in the present study. They were selected in our database based on the following inclusion criteria: (i) age at diagnosis of 18 or above (ii) diagnosis of grade II astrocytoma, oligodendroglioma or oligoastrocytoma, (iii) available tumor DNA or frozen tissue, (iv) previously determined chromosome arms 1p/19q statuses, (v) documented clinical outcome and (vi) no contrast enhancement on brain MRI. The patients have signed a consent form for molecular analysis.

TP53 sequencing

Tumor DNA was extracted from frozen samples using a standard protocol (Qiagen, QIAmp DNA minikit). A first amplification of all exons (1–11) was performed using high-fidelity Fast-start polymerase (Roche®). Primers’ sequences and the conditions of the touchdown polymerase chain reaction (PCR) are reported in the Supplementary Data 2 and 3 respectively. After amplification, primary amplicons were normalized to equimolar concentrations and pooled for groups of 11 amplicons (1 group = 1 patient). Each group was then labeled using a combination of two different Multiplex IDentifiers (MIDs), according to the manufacturer’s specifications (Roche/454 Life Sciences®).

Sequencing, emulsion PCR and pyrosequencing steps were conducted using the GS Junior instrument (Roche/454 Life Sciences®) according to the manufacturer’s instructions. Data analysis was conducted using Genomic Workbench (CLCBio) and Amplicon Variant Analysis (Roche®) softwares. The sequence used as TP53 reference was NM_000546. The mean coverage depth for all exons was 99X. Every mutation found was then validated using Sanger’s method (direct sequencing).

P53 and Ki67 expression using immunohistochemistry

Paraffin-embedded tissue slices, 3 μm in thickness, were immunostained. Monoclonal antibody anti-p53 (clone DO-7) from DAKO (Trappes, France) was used at a titer of 1:100 using the BenchMark XT automate (Ventana®). The number of tumor cells with strong p53 expression (out of at least 500 nuclei) was quantified through manual counting by two operators. p53 immunostaining was performed in 59/61 samples. Ki67 expression was investigated using Ki67 antibody (clone Mib1, 1/100, Dako®) as previously reported [12].

Chromosome arms 1p/19q deletion assessment

Loss of heterozygosity (LOH) on chromosome arms 1p and 19q was studied using microsatellite analysis as described before [13].

IDH1/2 mutation statuses

IDH1/2 mutation statuses were assessed as described elsewhere [14].

Statistical analysis

Overall survival (OS) was evaluated by the time from radiological or histopathological diagnosis to death. Survival curves were drawn using Kaplan–Meier method and compared using log-rank test. A p value < 0.05 was considered as significant. The correlation between TP53 mutation or p53 expression and other parameters (gender, tumor type, chromosome arms 1p/19q status, TP53 mutation or p53 expression status) were performed using χ2 test. Means’ ages between two groups were compared using Student test. Comparison of the distribution of Ki67 index across groups was performed using Mann–Whitney test.

Results

Characteristics of tumors and patients

The current study included 61 LGG: (i) 32 WHO grade II oligodendrogliomas, (ii) 8 WHO grade II astrocytomas and (iii) 21 WHO grade II oligoastrocytomas. Chromosome 1p/19q co-deletion was present in 34.4 % (21/61) of cases including 16 oligodendrogliomas.

Median age at diagnosis was 39.3 years (Interquartile Range—IR−, 19.8 years). Gender ratio of the cohort was 1.2 (33 males/28 females). Median follow-up and median overall survival of the cohort was 8.3 years [CI (5.4–11.3)] and 3.6 years [CI (2.5–4.7)] respectively (Table 1).

Table 1 Characteristics of the patients and the tumors
Full size table

Most of the patients underwent partial or total resection of the tumor (85.7 %); the others underwent diagnostic biopsy (14.3 %).

TP53 sequencing

In our series of 61 LGG, 38 mutations were detected in 32 tumors (52.4 %). Six tumors exhibited two mutations within TP53. There were 30 transitions and 8 transversions, leading to 31 missense mutations, 5 nonsense mutations, and 2 aberrant splicing mutations, as shown in Table 2.

Table 2 TP53 mutations detected in our series of diffuse low grade gliomas
Full size table

TP53 mutations were detected in exon 4 (n = 4), exon 5 (n = 8), exon 6 (n = 5), exon 7 (n = 4), exon 8 (n = 14) and exon 10 (n = 1). No mutation was detected in exons 1, 2, 3, 9 and 11. The location and the frequency of TP53 mutations in our series are reported in Fig. 1.

Fig. 1
figure 1

Distribution of TP53 mutation in our series of low grade gliomas. Although sequenced, exon 1 is not represented because it is a non-coding exon

Full size image

The demographic and molecular data according to TP53 mutational status are reported in Table 3. Patients with TP53-mutated tumor are younger than patients with TP53 wild-type LGG (p = 0.01), but there is no difference considering gender. TP53 mutations are significantly less frequent in oligodendrogliomas (34.4 %) compared to oligoastrocytomas (71.4 %) and astrocytomas (75 %) (p = 0.01). Chromosome 1p/19q co-deletion is rare in TP53 mutated LGG compared to TP53 wild-type LGG (p = 2.7.10−5).

Table 3 Clinical, pathological and molecular features of tumor with or without TP53 mutation
Full size table

p53 expression using immunohistochemistry

Strong positive p53 immunostaining (Fig. 2) was observed in 62.7 % of LGG (37/59). A ROC curve correlating the percentage of p53-positive tumor cells and TP53 mutational status of the tumor was established (Fig. 3). The best couple sensitivity/specificity was found for a threshold of 10 % [respectively 77.4 and 78.6 %; Predictive Positive Value (PPV) = 80.0 % and Predictive Negative Value (PNV) = 75.8 %]. Using 10 % of p53 positive tumor cells as threshold, a consistent correlation between p53 overexpression and TP53 mutation was detected (p = 2.3 × 10−5).

Fig. 2
figure 2

p53 expression and TP53 mutation. a and b a diffuse low grade glioma with missense TP53 mutation and p53 overexpression. c and d, a diffuse low grade glioma with a non missense (stop) TP53 mutation and low p53 expression (arrow indicates a positive cells). a, c hematoxylin eosin staining; b, d p53 staining. ×100 magnification

Full size image
Fig. 3
figure 3

Receiver operating characteristic (ROC) curve of p53 immunopositive cells percentage predicting TP53 mutation. Area under ROC curve (AUC) was 0.81 (*p = 3.5 × 10−5). The best combination of sensitivity (77.4 %) and specificity (78.6 %) was provided for a threshold of 10 % p53-positive tumor cells

Full size image

The characteristics of the patients and tumors according to p53 expression status are reported in Table 4. No association was found between p53 expression and age or gender. However, similarly to TP53 mutation, p53 overexpression is less frequent in oligodendrogliomas compared to astrocytomas and oligoastrocytomas (p = 0.002). Chromosome 1p/19q co-deletion and p53 overexpression were mutually exclusive (p = 9.2 × 10−4). However, in 4 samples both alterations were detected. Interestingly, in these cases the Ki67 index was significantly higher compared to tumors without this molecular pattern (p = 0.03).

Table 4 Clinical, pathological and molecular features of tumors with and without p53 overexpression
Full size table

Prognostic value of TP53 mutation and p53 expression

Considering OS, TP53 mutation and p53 expression are not associated with patients’ outcome (p = 0.49 in Fig. 4a and p = 0.77 in Fig. 4b respectively). Similarly, TP53 mutations located in exons 5–8 have no prognostic value in the present series. Finally when considering both 1p/19q and TP53 statuses, no prognostic significance was detected (i.e. 1p/19q intact and TP53 wild-type LGG versus 1p/19q co-deleted and TP53 wild-type LGG versus 1p/19q intact and TP53 mutated LGG) (Supplementary Figure).

Fig. 4
figure 4

Prognostic significance, in terms of overall survival, of TP53 mutation (a) and p53 expression (b) in diffuse low grade gliomas. WT wild type, mt mutated, IHC− <10 % p53-positive tumor cells, IHC+ ≥10 % p53-positive tumor cells

Full size image

Discussion

TP53 is a critical gene in cancer including glial tumors. Indeed, TP53 mutations have been detected in approximately 50, 40 and 10 % of astrocytomas, oligoastrocytomas and oligodendrogliomas respectively [5, 14]. The vast majority of published works focused on classic mutational hotspot exons and did not investigated the exons 1, 2, 3, 9, 10 and 11 [3, 5, 8, 10, 11, 15–46]. Only eight studies have explored non hotspot exons but in limited cohorts of LGG (Supplementary data 4). Therefore we have analyzed all TP53 exons in a retrospective cohort of 61 LGG.

To our knowledge, our study investigated the largest cohort of LGG for TP53 and p53 statuses. In agreement with previous studies, 52.4 % of the tumors exhibited TP53 mutation, and most of them (31/38) have been already described in LGG. Interestingly, six TP53 mutations have been detected in other tumors types (CNS and non-CNS) but never in LGG (Ser90Phe, Cys135Trp, Val143Met, Arg273Pro, Lys292X, 993+1G>A). Finally, one mutation has never been observed in cancer (according to IARCTP53 database). This mutation is a missense mutation located in exon 10, Pro359Leu. To assess the functional impact of this mutation we used Polyphen tool predicting impact of an amino acid substitution on the structure and function of a human protein using straightforward physical and comparative considerations [47]. To our concern, a Polyphen score 0 suggested a benign effect of this substitution on p53 functions. Overall, 86.8 % of the TP53 detected in our series are located in classic mutational hotspots supporting the proposed strategy to focus on these exons.

Our study confirmed already known data that TP53 mutation and p53 are associated with astrocytic phenotype and are mutually exclusive with 1p/19q co-deletion [9, 12, 15, 48], which is related to oligodendroglial features. In our cohort of low grade oligodendrogliomas, 1p/19q co-deletion was detected in 50 % which is within the boundaries reported in the literature (i.e. 39–70 %) [9]. In contrast, TP53 mutations are more frequent in our series of low grade oligodendrogliomas compared to data reported in the literature. A sampling bias might at least partially explain our results.

TP53 sequencing is time consuming and p53 expression is used as a surrogate marker in the setting of glioma diagnosis. Our study supports that immunohistochemistry, which is a straightforward technique that can be applied in routine histopathological assessment, may be used instead of molecular biology.

However, interpretation of p53 expression is heterogeneous across labs. The threshold of 10 % p53-positive tumor cells is debated [10–13, 43, 46, 48–51]. Indeed, p53 immunostaining is scored as positive or negative using varying thresholds [13, 46, 49–51] or using a “labeling index” [3]. In our study, a ROC curve demonstrated that the best cut-off to predict the mutational status of TP53 was 10 % of p53-positive tumor cells with a PPV of 80.0 %, and a PNV of 75.8 %.

Among the seven TP53-mutated tumors without p53 expression, five of them carry a nonsense mutation (codon stop or aberrant splicing). Indeed, this is a known limitation for the use of p53 expression as TP53 mutation sensor [7]. If we consider only TP53 missense mutations (81.6 % of all mutations), p53 immunolabeling detects 92 % of mutated tumors.

One of the largest TP53 study (n = 124) was conducted by Stander et al. [11]. Here, TP53 mutations were associated with a shorter survival, but not p53 overexpression. However, previous work conducted in the same research group failed to demonstrate any link between TP53 status and prognosis [10]. Similarly, Okamoto et al. [5]. showed a negative impact of TP53 mutation on OS in univariate, but not in multivariate analysis. These studies illustrate the conflicting reports on prognostic value of this marker. Based on our large cohort and comprehensive analysis, neither TP53 mutation nor p53 accumulation were found to be prognostic factors in LGG.

Finally, a recent study has shown the interest of combining IDH mutational, 1p/19q and p53 statuses in prognostic stratification of LGG [52]. Indeed, “triple-negative” LGG (i.e. IDH−, p53− and 1p/19q intact) exhibit dismal prognosis. We were unable to identify this prognostic impact in our cohort probably due to the limited number of “triple-negative” LGG (n = 7).

Conclusion

Our study is the largest investigating TP53 and p53 statuses in LGG. TP53 is mutated in 52.4 % of cases. TP53 mutations outside mutational hotspots (exons 4–8) are rare (2.6 %) supporting targeted TP53 sequencing in LGG. Interestingly, seven novel TP53 mutations have been discovered in LGG. TP53 mutations are associated with astrocytic phenotype, younger age and p53 overexpression. In contrast, they are mutually exclusive with 1p/19q co-deletion. Using a threshold of 10 % of p53-positive tumor cells, p53 expression is a good surrogate marker of missense TP53 mutation. However, it should be used with caution since it misses ~20 % of TP53-mutated tumors. TP53 and p53 statuses are not prognostic in our series of LGG. Further analyses of TP53 statuses in prospective series of LGG may overcome potential sampling biases observed in retrospective studies including the present work.