The “Guardian of the Genome”—An Old Key to Unlock the ERCC1 Issue Free

In this issue of Clinical Cancer Research, Heyza and colleagues show how p53 is essential to induce checkpoint activity and ultimately cell death in Excision Repair Cross-Complementation Group 1 (ERCC1)–deficient cancer cells treated with cisplatin (1). Consequently, loss of p53 function should at least partially be considered as a confounding biomarker with ERCC1, because only ERCC1-negative/p53 wild-type cancer cells would be sensitive to cisplatin treatment (Fig. 1A).

Platinum-based chemotherapy is a standard treatment for numerous patients with cancer of the lung, head and neck, ovary, stomach, bladder, or testicles. There is a strong need to identify and evaluate biomarkers that can be used to select patients who are likely to benefit from cisplatin-based drug combinations and those who will resist. Cisplatin is a small molecule that binds covalently to DNA, thereby forming adducts either within the same strand [intrastrand adducts (ISA)] or between the two strands [interstrand cross-links (ICL)], which interfere with DNA replication and transcription to induce cancer cell death. This direct deleterious effect on DNA justifies that most studied candidate biomarkers predicting cisplatin efficacy are those involved in DNA-repair processes.

The main DNA-repair pathways thought to participate in the elimination of ISAs and ICLs are respectively nucleotide excision repair (NER) and the ICL repair/Fanconi anemia (FA). In ICL repair, the endonuclease heterodimer ERCC1/XPF allows the “unhooking” of platinum ICLs either in G₀/G₁- phase or in S-phase (FA proteins being expectedly active during S-phase only) and has recently been reviewed (2). Any default in ICL repair would lead to secondary double-strand breaks (DSB), mainly due to “collapsing” of blocked replication forks during S-phase. DSBs are generally repaired by additional pathways such as the high-fidelity homologous recombination repair (HRR) or a pathway that do not verify the genetic integrity such as nonhomologous end-joining (NHEJ) or other pathways that systematically introduce errors, such as microhomology-mediated end-joining (MMEJ/alternative-NHEJ) or single-strand annealing.

To date, the most advanced biomarker to predict cisplatin efficacy is ERCC1 expression. As confirmed in the work from Heyza and colleagues, modulation of ERCC1 expression alters the sensitivity of cells to cisplatin treatment in vitro. More than a decade ago, the expression of ERCC1 protein measured by IHC in formalin-fixed paraffin-embedded (FFPE) lung tumor samples appeared to be a predictive marker of platinum-based chemotherapy in the International Adjuvant Lung Trial (IALT) and several prospective randomized trials emerged. However, none of these clinical studies were able to validate the interest of ERCC1 as a predictive biomarker. Several explanations have been advanced, such as the reliability of ERCC1 antibodies and the presence of nonfunctional ERCC1 isoforms, the ERCC1-202 isoform being the only one to show functionality (in the sense of elimination of ISAs in vitro; ref. 3).

So, how does the work from Heyza and colleagues fit into the ERCC1 story? The introduction of a second biomarker (p53 status) to be considered in combination with ERCC1 can have a significant translational impact to select patients who will benefit the most from cisplatin-based chemotherapy from those who will resist. In addition, the measure of p53 status is easily determined by DNA sequencing or conventional IHC methods, which is in contrast to ERCC1 alone because its expression level is not robustly examined by a single antibody in FFPE samples. Further, using The Cancer Genome Atlas (TCGA) provisional lung adenocarcinoma cohort and the 2017 TCGA ovarian cancer dataset, Heyza and colleagues were able to stratify patients according to ERCC1 gene expression and TP53 mutational status to confirm their findings in vitro where they used CRISPR-Cas9 and lentiviral rescue experiments in cell lines.

It is important to notice that these findings are not simply correlative data on two “old” biomarkers that together predict clinical effect of cisplatin in lung adenocarcinomas and ovarian carcinomas in public databases. Indeed, their findings confirm that cisplatin cytotoxicity is mainly due to ICLs rather than ISAs, which has been a subject of debate. The authors also suggest that DNA-repair pathways other than HRR, such as NHEJ or the systematically error-prone MMEJ do resolve secondary cisplatin-induced DNA damages (DSB) in ERCC1-deficient cells when p53 function is simultaneously lost.

Perhaps more importantly, the data give new insight into these ICL-repair events in the context of cell-cycle regulation. According to the data, cisplatin-induced DNA cross-linking rapidly leads to G₂–M-phase arrest, unless ERCC1 proficiency allows the cells to recover by NER and ICL unhooking with subsequent cell survival. In contrast, ERCC1-deficient cells seem to accumulate DSBs and the cells slowly recover from the G₂–M arrest moving to G₁-phase where the p53-dependent checkpoint activity leads to cell death (Fig. 1B). However, when p53 function is lost, the cells are allowed to continue into S-phase where alternate DNA-repair pathways (excluding HRR) seem to allow ICL tolerance and cell survival. Overall, the data provide a mechanistic explanation for an exquisite synthetic viability relation between ERCC1 and p53 deficiencies leading to cisplatin resistance in lung, ovarian, and potentially other carcinomas.

From there, several points will need further attention. First, the in vitro investigations would benefit from more direct DNA-repair functional tests, such as vector-based substrates of DSB repair. Such molecular tools would clarify the respective roles of NHEJ versus MMEJ in the observed ICL tolerance. Second, the clinical data were collected from publically available databases and should be independently confirmed in existing ERCC1-dedicated studies and cohorts such as the TASTE, ET, and ERCC1/RRM1 prospective clinical studies. However, that task might be challenging because the statistical power required to validate biomarkers in two-by-two subgroups is not easily achieved, as well as determining a universal ERCC1 positivity cutoff. Indeed, other reports have previously highlighted the potential of p53 expression alone to predict efficacy of cisplatin-based chemotherapy as in the JBR10 trial, although it was not confirmed in the LACE-Bio study. The p53 biomarker alone was also studied in the IALT series showing a potential predictive interest but was not investigated in combined subgroups (e.g., p53 plus ERCC1). Third, combined ERCC1 and p53 effect on chemosensitivity during short-term treatment in vitro might be different from the effect observed in vivo. For instance, one ERCC1-deficient/TP53-deficient K-RAS–driven murine lung adenocarcinoma model was reported as being highly sensitive to cisplatin initially, but when treated mice relapsed, the tumors had acquired sensitivity to etoposide (4). However, the context of DNA-repair deficiency and its consequences on genetic instability in murine cells are somehow different compared with human cells (5). Finally, because ERCC1-deficient/p53-deficient tumors probably harbor high mutation rates, the identification of a common genomic signature could be of interest. Such investigations would be warmly welcomed in the current era of high-throughput next-generation sequencing that is used to characterize individual tumor mutational loads.

In conclusion, wild-type p53 expectedly favors cisplatin-induced cell death in repair-deficient (i.e., loss of ERCC1 expression) tumor cells. In contrast, loss of p53 function promotes tolerance to cisplatin-induced ICLs and DSBs through G₁ checkpoint abrogation with subsequent S-phase entry that permits activation of alternative DNA-repair pathways and tumor cell survival. If these data are validated in independent clinical cohorts, ERCC1 and p53 status should be considered concomitantly when selecting patients with cancer for cisplatin-based chemotherapy.

J.-C. Soria is senior vice president of AstraZeneca, reports receiving speakers bureau honoraria from AstraZeneca, Astex, Clovis, Gamamabs, Lilly, MSD, Mission Therapeutics, Merus, Pfizer, PharmaMar, Pierre Fabre, Roche/Genentech, Sanofi, Servier, Symphogen, and Takeda, and holds ownership interest (including patents) in AstraZeneca and Gritstone. L. Friboulet, J.-C. Soria, and K.A. Olaussen are coinventors of a patent on ERCC1 (patent number: 9702875). No other potential conflicts of interest were disclosed.