Chemotherapy has become an essential part of breast cancer treatment in the adjuvant or neoadjuvant settings.1 Response to chemotherapy is one of the most important prognostic factors, especially when pathological complete response is achieved.2, 3 Apoptosis,4 as well as senescence,5–8 are major treatment-induced cellular events that can determine tumor sensitivity.
p53 regulates cell fate in response to various stresses, either genotoxic or not genotoxic.9 Among many possible effects, p53 can induce apoptosis, cell cycle arrest or senescence. Senescent cells are enlarged, flattened and granular, positive for SA-β-gal.10 They often overexpress p53, p21CIP1 and p16INK4a.11–14 The senescent state is characterized by an irreversible cellular growth arrest physiologically induced by telomere shortening (replicative senescence). In pathological conditions such as cellular stress, oncogene activation15 or DNA damages, cells can develop a senescent phenotype, now recognized as “accelerated” senescence,11–16 leading to complete growth arrest in only a few divisions, as compared to the 30–50 mitoses occurring before induction of replicative senescence. Accelerated senescence is believed to be a mechanism of protection against tumors.17 It can be induced by DNA-damaging chemotherapy18 and often requires functional p53.19
We have previously shown that, in locally advanced breast cancer patients treated by high doses epirubicin/cyclophosphamide chemotherapy, pathologically assessed complete responses were observed only in TP53 mutated tumors, while all TP53 wild type tumors had incomplete response.20, 21 We hypothesized that breast tumor cells with wild-type TP53 could enter a prolonged, but reversible, cell cycle arrest state and/or senescence rather than apoptosis, precluding tumor clearance. These changes would then protect noncycling tumor cells against the effects of cyclophosphamide and allow a restored proliferation after discontinuation of chemotherapy. Conversely, complete responses observed in TP53 mutated tumors may be explained by the accumulation of genetic abnormalities, leading to p53-independent cell death, for example through mitotic catastrophes.
To test these hypotheses in vivo, we studied the early cellular and molecular events induced by epirubicin and cyclophosphamide in 2 TP53 mutated and 1 TP53 wild type human breast cancer tumors xenografted in nude mice.
Pre-treatment features of tumors xenografts
Three breast cancer xenograft mouse models obtained initially at Curie Institute22 were analyzed. To assess TP53 status, we used a yeast functional assay (Fig. 1a) which tests the function of the p53 protein and detects 90% of reported mutations.23 Tumors were considered TP53 mutant when (i) more than 15% of the yeast colonies were red and (ii) analysis using the split versions of the test could identify the defect in the 5′ or 3′ part of the gene, confirming the initial determination and (iii) sequence analysis from mutant yeast colonies could identify an unambiguous genetic defect (mutation, deletion, insertion, splicing defects). We selected 2 tumors with a TP53 mutation (TP53mut1 and TP53mut2) and 1 tumor with normal TP53 (TP53wt), because tumors with normal TP53 rarely grow in mice. TP53mut1 had a deletion on exon 4, codon 102 (del C) and TP53mut2 had a mutation on exon 5, codon 157 (V > F). The type of mutation found in TP53mut1 is rarely observed in the DNA-binding domain while the missense mutation found in TP53mut2 is more common. Using immunohistochemistry, TP53mut2 showed a strong and diffuse p53 nuclear immunostaining likely due to lack of MDM2-induced proteosomal degradation of the p53 protein, while TP53mut1 showed no staining (due to a truncated protein). TP53wt had a fully functional p53 pathway with the yeast assay and showed no p53 accumulation on immunostaining (Fig. 1b).
Histological types and grades as well as ER, PR, ERBB2, CK5/6 and CK17 status are shown in Table I. Briefly, TP53wt tumor was ERBB2 positive, while the 2 other tumors were of the basal phenotype, as defined by triple negative status (ER-PR-ERBB2-) and overexpression of basal cytokeratins.24
Table I. Characteristics of Tumors
Immunohistochemical tests were performed with Dako or Novocastra mouse monoclonal antibodies on formalin fixed and paraffin embedded samples. Percentage of positive tumor cells and staining intensity (+ to +++) are noted except for ERBB2 for which 3+ score was defined by strong complete membrane staining of more than 30% cells.
Treatment-induced senescence-like phenotype is dependent on TP53 status
These in vivo models were treated with epirubicin (Farmorubicine®, Pharmacia S.A.S, France) 2 mg/kg and cyclophosphamide (Endoxan®, Baxter, France) 33 mg/kg and subsequently sampled after treatment at days D1, D3, D5 and D7. For each tumor, 5 mice were used for each post-treatment day and 10 nontreated mice were used as controls. For these short treatments, no significant reduction in tumor size was observed and mice survival was not examined.
Mitotic counts and Ki67 immunostainings were performed on tumor sections. They were not statistically different before and after treatment (Figs. 2a-2c, 3a–3c and 4a–4c). SA-β-gal staining was negative before treatment in all 3 tumors. After treatment, 5–10% tumor cells in TP53wt showed SA-β-gal cytoplasmic staining, as soon as Day 1 (Figs. 2d–2f). The 2 TP53 mutant tumors remained negative for SA-β-gal at all time points after treatment (Figs. 3d–3f and Figs. 4d–4f). Before treatment, p21CIP1 immunostaining was negative or found in less than 20% tumor cells. In the TP53 wild-type tumor, p21CIP1 immunostained cells increased starting at Day 3 with a mean number of stained cells over 50% (Figs. 2g–2i). This clear increase was also demonstrated by Western blot (Fig. 2j) and QRT-PCR analyses (Fig. 5a). In contrast no significant change for p21CIP1 immunostaining or mRNA expression was observed in the 2 TP53 mutant tumors (Figs. 3g–3i, 4g–4i and 5d, 5g). p16INK4a immunostaining (data not shown) remained unchanged at the different time points after treatment. Such early post-treatment senescence-like phenotype in a TP53 wild type tumor extends in vivo previous in vitro findings that showed early induction of cell cycle arrest and a senescent phenotype in MCF7 cells treated by methotrexate.13
Treatment-induced mitotic catastrophe and apoptosis
Apoptotic cells were quantified on H&E, semi-thin and TUNEL-stained sections. Apoptosis increased after treatment in all tumors, reaching 20–30% of cells at Day 7 or Day 14 (Figs. 2k–p, 3j–o and 4j–o). We thus looked for the expression of proteins involved in apoptosis. Immunostaining for the p53 target pro-apoptotic gene BAX was negative before treatment in the 3 tumors. In TP53wt, there was BAX cytoplasmic staining at Day 3 in 10–20% tumor cells (data not shown) with a significantly increased BAX mRNA at D3 and D5 (Fig. 5b). In contrast, no significant change in Bax immunostaining (data not shown) or Bax mRNA was found in TP53mut1 and TP53mut2 (Figs. 5e and 5h). Similarly PUMA showed a strong mRNA overexpression at D3 and D5 in TP53wt, but no change in mutant tumors (Figs. 5c, 5f and 5i).
The number of abnormal mitoses on semi-thin sections significantly increased in TP53 mutated tumors, between Day 3 and Day 5 for both tumors (Figs. 3p–3r and Figs. 4p–4r). In contrast, abnormal mitoses were not found in TP53wt at any time point. In vitro, following a genotoxic stress, TP53 mutated tumor cells may undergo mitotic catastrophe.25, 26
In the TP53 wild-type tumor, post-treatment apoptosis is mostly induced through p53 target genes, while apoptotic cells observed in TP53 mutated tumors likely rely on other apoptotic pathways, including mitotic catastrophe. Similar to our results, a recent in vivo study in ovarian carcinoma has suggested that p53 inactivation could lead to significant DNA damage and to eventual cell death and tumor response either by way of a p53 independent apoptotic pathway or by mitotic catastrophe.27
A p53-dependent evolution of treated cancers
Although these results were obtained in a small number of tumors, they are fully in line with our previous results20, 21: in noninflammatory locally advanced breast cancers treated with this dose-intense epirubicin-cyclophosphamide association, no complete pathological response was observed in tumors without TP53 mutation. Interestingly, the 2 TP53 mutated tumors we studied were triple negative (ER-, PR-, ERBB2-) and overexpressed basal cytokeratins, therefore corresponding to the basal subtype of breast cancer.24 This subtype was particularly sensitive to this drug association in our previous study.21
The observations reported here suggest that, in our settings, the lack of response in TP53 wild-type tumors may be due to the induction of cell cycle arrest, allowing tumor cells to reinitiate proliferation at the end of chemotherapy. Middelburg et al.28 also studied early molecular events induced by chemotherapy in vivo and showed overexpression of the proapoptotic protein PUMA, consistent with our findings. However, no correlation with the TP53 status was done in this work.
Conversely, in TP53 mutated tumors in patients,21 pathological complete responses were observed in 60% cases. This may be explained by accumulation of genetic alterations29 leading to cell death through mitotic catastrophe, as observed here. Acquisition of additional genetic defects could also contribute to resistance to treatment in the nonresponsive tumors.
Such opposite effects of p53 (chemosensitivity/chemoresistance) depending on the type of drug used30 have already been reported in vitro. In the present in vivo model of human breast cancer, p53 is both a pro-apoptotic factor that may induce some cell death in response to drug-induced DNA damages, but is predominantly an inducer of early cell cycle arrest, protecting tumor cells from further cytotoxic damages, especially for high-dose epirubicin-cyclophosphamide regimen. Indeed, that in patients tumor clearance was only observed in TP53 mutant cases suggests that, in TP53 wild-type tumors, p53-induced apoptosis is insufficient to initiate tumor eradication, implying that the tumor compartment undergoing transient cell-cycle arrest is more important than the one undergoing apoptosis.
We thank Ms. Claire Bocquet, Ms. Laurence Françoise, Ms. Martine Legrand, Ms. Evelyne Wittmer, Ms. Stéphanie Belhadj and Ms. Helena Pisonero for their technical assistance, for the QRT-PCR and FASAY assays, Ms. Stéphanie Lefrançois and Mr. Sylvain Arien for semi-thin sections. This work was supported by PHRC, ARC, ANRT and PNES.