TP53 mutations at codon 234 are associated with chlorambucil treatment in chronic lymphocytic leukemia

TP53 aberrations, including somatic mutations of TP53 gene and 17p deletion, are a major predictive factor of resistance to fludarabine-based chemotherapy in chronic lymphocytic leukemia (CLL) and remain an adverse prognostic factor in the chemofree era. The detection of deletion 17p and TP53 gene mutations has become an integral part of routine diagnostic and should be performed before any administration of treatment.1 TP53 mutations typically occur all along the DNA-binding domain of the p53 protein. However, the functional role or the effect of each TP53 variant remains elusive. In this study, we aimed at characterizing the profile of the TP53 variants in CLL, their distributions, and a correlation with clinical data. To this end, we retrospectively analyzed a large collection of 568 CLL-associated TP53 variants in 336 patients compiled from centers affiliated with the French Innovative Leukemia Organization-CLL (FILO). Fluorescence in situ hybridization analysis for del(17p) status was available for 207 patients, of which 108 (52%) harbored a 17p deletion. Based on the IGHV mutation status, most patients belong to a high risk group as 73% (172/236) were IGHV unmutated (Table S1). TP53 mutational status was assessed from blood samples using NGS technologies (either Illumina or Ion Torrent technologies) covering exons 2–11, with a sensitivity allowing the detection of subclones with variant allele frequency (VAF) >1%. Both amplicon- and capture-based approaches available in commercial kits or custom gene panels including the TP53 gene were used. Polymorphisms were carefully excluded using the gnomAD database and the new TP53 SNP data included in the most recent version of the UMD_TP53 database.2 Classification of TP53 variant pathogenicity was performed in accordance with ACMG criteria and based on population data and TP53-specific functional information previously described.3 Among the 568 TP53 somatic mutations, the majority were missense variants (n = 429, 75%) (Table S2). Using ACMG criteria from the UMD_TP53 database, 518 variants were classified as pathogenic, 42 were likely pathogenic, and 8 were variants of unknown significance. As expected, TP53 variants were predominantly located in the DNA-binding domain, with classical CpG-related hot spots at codons 175, 248, and 273 (Figure 1A). Nevertheless, an unusual mutation hotspot occurring at codon 234 was identified. A total of 25 single nucleotide variants at codon 234 was observed in 24 patients with one patient harboring 2 mutations: 22 mutations p.Tyr234Cys (c.701A>G), 2 mutations p.Tyr234Ser (c.701A>C), and 1 mutation p.Tyr234His (c.700T>C). It was the second most common missense variant of our cohort (25/568; 4.4%) and ranked fifth in the entire CLL subset of the UMD_TP53 database (Figure 1A, and Figure S1A). Mutations at codon 234 were found predominantly at a subclonal level (median of variant allele frequency VAF = 11% [1.8%–51%]) but with VAFs distributed in the same range as those observed for other hotspot codons (Figure 1B). Analysis of the UMD_TP53 database showed that the frequency of mutations occurring at hotspot positions 175, 248, and 273 was similar to that observed in other cancers, while the frequency of mutations at codon 234 reached 4% in CLL and was very low (less than 1%) in the other malignancies (Figure 1C and Figure S1B), suggesting that this mutation is highly specific of CLL cases. Both in silico predictive algorithms and structural studies suggested that variant p.Tyr234Cys was deleterious, which was confirmed by our functional analysis.4 Indeed, the overexpression of p.Tyr234Cys by transfection in the p53 null H1299 cell line was unable to induce growth arrest, similar to p.Arg175His, while wild-type p53 or p.Thr312Ser (a variant without loss of function) leads to a profound reduction of colony number (Figure S2A). On the other hand, in a luciferase assay, p.Tyr234Cys was unable to transactivate a reporter gene with the CDKN1 response element (Figure S2B). In line with these results, the large-scale analysis performed by Kotler et al. showing the relative fitness score for each TP53 variant (defining its capacity to induce growth arrest in H1299 cells) indicates that the majority of TP53 variants at position 234 are dysfunctional (Figure S2C).5 Altogether, these results indicate that TP53 variant p.Tyr234Cys is devoid of any transactivating and antiproliferative activities. In the present study, CLL patients frequently harbored multiple subclones with different TP53 mutations. A single TP53 variant was detected in 66% (n = 222) of the patients (SM-patients), whereas two, three, or more TP53 variants were found respectively in 17% (n = 57), 10% (n = 33), and 7% (n = 24) of the patients (polymutated patients) (Table S2 and Figure S3A). Interestingly, while mutations at classical TP53 hot spot positions (codons 175, 248, or 273) were observed both as single and associated mutations, we noticed that variants at codon 234 were found mostly in polymutated patients (82% cases), highlighting an important intratumoral heterogeneity in cases harboring this mutation (Figure 1D). Strikingly, analysis of the variants associated with mutations at codon 234 reveals a strong association with a splicing mutation c.673-2A (p < .0001, Fischer test) (Figure 1E). The frequency of this mutation was 2.6% (15/568) in the entire cohort and rose up to 37.5% (9 cases) in the 24 patients with mutation at codon 234. In contrast, it was associated with mutation at codon 248 or codon 273 in only one case. Consistently, the reanalysis of sequencing data from published cohorts gathering altogether 272 TP53 mutated patients showed a frequency of 3.3% of mutation at codon 234 (nine cases), among them 5 were associated with the mutation at position c.673-2A (62.5%).6-9 Mutation at position c.673–2A at the splice acceptor site in exon-7 generates a p53 truncated protein10 that functionally and molecularly resembles the naturally occurring alternative p53 splice variant, p53-psi. Shirole et al. have shown that “p53-psi like” proteins lack canonical p53 tumor suppressor activities but promote cancer cell proliferation, survival, and metastasis.11 These mutants can localize to the mitochondria where they promote tumor phenotypes by binding and activating the mitochondria inner pore permeability regulator, Cyclophilin D (CypD). However, the functional relationship between mutation at position c.673–2A>G and p.Tyr234Cys or whether one of these mutations promote the acquisition of the other remains elusive. Clinical data were available for 222 patients. TP53 sequencing was performed before treatment for 81 patients (37%). The type of treatment significantly impacted the nature of the TP53 variant and the number of associated mutations in patients. Indeed, variants at codons 175 and 248 were observed in both treated and untreated patients, whereas 100% of patients who harbored a variant at codon 234 had been treated (Figure 1F). Moreover, while most treatment-naïve patients had a single mutation (n = 72/81; 88%), nearly half of previously treated patients (n = 66/141; 47%) had more than one TP53 variant (from 2 to 11 mutations) (p < .0001), showing that the exposition to a therapy promotes the clonal evolution of TP53 (Figure S3B and C). Fifty-nine percent (53 of 89) of patients who received one line of chemoimmunotherapy (fludarabine, cyclophosphamide, and rituximab [FCR] or bendamustine and rituximab [BR]) had a single mutation, while those who received only continuous chlorambucil treatment (CLB) were mostly polymutated (32 of 44; 77%, p = .0012 with Yates correction), suggesting that CLB has a stronger capacity to drive clonal diversification (Figure 1G). Furthermore, variants at codon 234 were found exclusively in treated patient (24 of 141; 17%) and in 43% of patients treated with CLB (19 of 44). In contrast, only 5.7% of patients treated with FCR and/or BR harbored this mutation (5 of 89) (Figure 1H). Eventually, considering the association between mutation at codon 234 and c.673-2A, the nine patients displaying the two mutations were all treated with CLB, while this co-occurrence was not found in patients treated with fludarabine and/or BR. The co-occurrence of the two mutations was significantly associated with CLB treatment (p < .0001) with an odds ratio of 26 (95% CI: 10.7–69.3). Taken together, our data support the role of the alkylating agent CLB in the appearance and/or the selection of mutations at codon 234 and possibly the truncating mutation c.672-2a. No pre-CLB treatment sample was available for patients with Tyr234Cys mutations to test for TP53 mutation status. However, in contrast to the other hot spot mutations such as those at codon 175 or 248, the absence of variants at position 234 in any of the untreated patients in our database strongly supports the possibility that this mutation is associated with the mutagenic effect of CLB. Furthermore, the absence of this variant in a recent cohort of CLL patients in which none received CLB argues against the hypothesis of the specific selection of this mutation in CLL. This finding is reminiscent of the association between exposure to certain carcinogens (aflatoxin B1 in hepatocellular carcinoma and benzo(a)pyrene in lung cancer) and TP53 hot spot variants in codon 249 or 157.12 Whether or not chlorambucil adduct occurs at codon 234 or the splicing site 6.732-2A of the TP53 gene remains to be tested. This study was supported by Cancéropole Ille de France (convention n°2019-1-EMERG-22-INSERM 6-1) to Thierry Soussi and Fanny Baran-Marszak. Bernard Leroy for the management of the UMD_TP53 database. The authors declare no potential conflicts of interest. Conception and design: Florence Cymbalista, Fanny Baran-Marszak, and Thierry Soussi. Acquisition of data (acquired and managed patients, provided facilities, etc.): Grégory Lazarian, Floriane Theves, Myriam Hormi, Rémi Letestu, Virginie Eclache, Audrey Bidet, Pascale Cornillet-Lefebvre, Frédéric Davi, Eric Delabesse, Marie-Hélène Estienne, Pascaline Etancelin, Olivier Kosmider, Sophie Laibe, Laurence Lode, Marc Muller, Nathalie Nadal, Dina Naguib, Cédric Pastoret, Stephanie Poulain, Pierre Sujobert, Lauren Veronese, Samia Imache, Valérie Lefebvre, Florence Cymbalista, Fanny Baran-Marszak, and the members of the French Innovative Leukemia Organization. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Grégory Lazarian, Floriane Theves, Florence Cymbalista, Fanny Baran-Marszak, and Thierry Soussi. Writing, review, and/or revision of the manuscript: Grégory Lazarian Florence Cymbalista, Fanny Baran-Marszak, and Thierry Soussi. Study supervision: Florence Cymbalista, Fanny Baran-Marszak, and Thierry Soussi. Data S1. Supporting information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.