BRCA1 Expression and Outcome in Patients With EGFR-Mutant NSCLC Treated With Gefitinib Alone or in Combination With Olaparib
ABSTRACT
Introduction: DNA repair capacity, as exemplified by BRCA1 gene expression, is related with outcome to EGFR tyrosine kinase inhibitors in patients with EGFR-mutant NSCLC. Olaparib, a PARP inhibitor, reduces BRCA1 expression. Olaparib was tested in combination with gefitinib versus gefitinib single agent, as a first-line therapy for patients with EGFR-mutant NSCLC in the GOAL study (trial registration: NCT01513174). Here, we report the results of the biomarker-related prespecified secondary objectives of the GOAL study. Methods: We evaluated the impact of BRCA1 mRNA expression in 91 patients with EGFR-mutant NSCLC. Of those 91 patients, 51 were randomized to treatment with gefitinib and 40 were randomized to treatment with gefi- tinib plus olaparib. We explored in vitro whether BRCA1 mRNA levels are related with outcome to gefitinib plus olaparib. The expression levels of 53BP1, CtIP, and AXL were also explored and correlated with the treatment outcome. Results: Overall, as what happened in the GOAL study, no statistically significant difference was observed in median progression-free survival (PFS) between the two treatment arms, for the 91 patients of the present study (p = 0.2419). For patients with high BRCA1 mRNA expression (BRCA1- high group), median PFS was 12.9 months in the gefitinib plus olaparib arm, compared with 9.2 months in the gefitinib arm (p = 0.0449). In the gefitinib arm, median PFS was 9.1 months for the BRCA1-high group and 10.2 months for the BRCA1-low group (p = 0.0193). We observed a more pronounced synergism of gefitinib plus olaparib in cells with higher BRCA1 compared with those with low BRCA1
mRNA expression. Conclusions: High BRCA1 mRNA expression identified patients with NSCLC who benefited from gefitinib plus ola- parib in the GOAL phase 2 clinical trial.
Introduction
We identified that low BRCA1 gene mRNA levels were an independent favorable predictive marker to erlotinib in patients with EGFR-mutant advanced NSCLC.1,2 On the basis of the findings, we proposed a model for a BRCA1- dependent DNA repair of erlotinib-induced DNA damage through an H2AX-independent pathway. We speculate that DNA breakage caused by erlotinib is different from that caused by radiotherapy or platinum-based chemo- therapy. In addition, poly(ADP)-ribosylation of proteins by PARP1 is a rapid response to DNA lesions. PARP1 inhibitors down-regulate BRCA1 expression.3 We posit that BRCA1 by itself could be a predictive biomarker, and studies are warranted to use PARP inhibitors in combination with erlotinib in patients with elevated BRCA1 gene expression.2The GOAL study, performed by the Spanish Lung Cancer Group, was a phase 1B and 2B study to evaluate the efficacy and tolerability of gefitinib plus olaparib versus gefitinib alone as first-line therapy in patients with metastatic EGFR-mutant NSCLC (NCT01513174). In the phase 1B dose escalation part of the study, tolerance in the absence of pharmacokinetic interactions and the activity of gefitinib plus olaparib were confirmed in 22 patients with EGFR-mutant NSCLC.
The recommended phase 2 dose was 250 mg of gefitinib once daily plus 200 mg of olaparib three times daily.4 In the phase 2B part of the GOAL study, between July 2013 and July 2016, a total of 186 patients with previously untreated metastatic EGFR-mutant NSCLC were included in 34 centers in Spain and one in Mexico.5 The intent-to-treat (ITT) analysis included 91 patients with EGFR-mutant NSCLC in the gefitinib arm and 91 patients with EGFR-mutant NSCLC in the gefitinib plus olaparib arm.5 Progression- free survival (PFS) and overall survival (OS) were eval- uated at the final data cutoff point on July 2017. The median follow-up time was 26.2 months (95% confi- dence interval [CI]: 20.3–27.8) for gefitinib and 21.2months (95% CI: 17.5–28.3) for gefitinib plus olaparib (p = 0.2858).5The primary end point of the phase 2B GOAL study,which was to determine whether the addition of olaparib to gefitinib improved PFS in previously untreated pa- tients with metastatic EGFR-mutant NSCLC, was notmet.5 Median PFS was 10.9 months (95% CI: 9.3–13.3)for gefitinib versus 12.8 months (95% CI: 9.1–14.7) for gefitinib plus olaparib (p = 0.1242; hazard ratio [HR] = 0.75, 95% CI: 0.52–1.08).5 Among 174 patients assess- able for response, the objective response rate was 68% in the gefitinib arm versus 71% in the gefitinib plus olaparib arm (p = 0.4873).5 No statistically significant differences were found between the two treatment armsin median OS.5
As far as safety concerns, there was an increase in hematological and gastrointestinal toxicities for gefitinib plus olaparib, compared with gefitinib alone.5Here, we report the results of the biomarker-related prespecified secondary objectives of the GOAL study. A secondary objective of the GOAL study was to evaluate whether the mRNA expression levels of BRCA1 may affect PFS in the two treatment arms.2 The correlation of the mRNA expression levels of other biomarkers, including 53BP1, CtIP,6 and AXL,7 with PFS, is also explored. All patients have provided written informed consent before being enrolled in the study. Preclinical in vitro evidence of the potential effect of BRCA1 mRNA expression levels on the outcome to gefitinib plus ola- parib is finally provided.Paraffin-embedded samples and slides and cell lines were processed as previously reported for gene mRNA expression.8,9 Tumor tissue was available from 91 pa- tients of the GOAL study, 51 of whom were randomized to treatment with gefitinib and 40 were randomized to treatment with gefitinib plus olaparib.5 The present gene expression study was conducted in the ISO 15189- certified Pangaea Oncology laboratory located in Hospi- tal Universitari Dexeus—Grupo Quirónsalud (Barcelona, Spain). Pangaea Oncology was the central laboratory for the GOAL study. The primer and probe sequences for BRCA1, 53BP1, CtIP, and AXL were designed using Primer Express 3.0 Software (Applied Biosystems) according to their reference sequence (http://www.ncbi.nlm.nih.gov/ LocusLink). b-actin was used as the endogenous gene.
Gene expression analyses were performed as previously described with quantitative real-time polymerase chain reaction.8,9PC9 (EGFR exon 19 deletion) cells were provided byF. Hoffmann-La Roche Ltd. The brain metastatic lung cancer cell line, PC9-BrM3, was generously provided by Professor Joan Massagué (Cancer Biology and Genetics Program, Memorial Sloan Kettering Cancer Center, New York, NY).10 Gefitinib was purchased from TocrisBioscience Company (Bristol, United Kingdom). Olaparib was purchased from Selleckchem (Houston, TX). Drugs were prepared in DMSO at a concentration of 10 to 100 mmol/liter stock solutions and stored at —20◦C. Further dilutions were made in culture medium to final con- centration before use, as previously described.7Cells were seeded on 96-well plates at the following densities: 2 × 103, 3 × 103, and 4 × 103 and incubated for 24 hours.7 Cells were treated with serial dilutions of the drugs administered at indicated doses. After 72 hours of incubation, 0.5 mg/mL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide reagent (Sigma-Aldrich, St. Louis, MO) was added to the medium in the wells for 2 hours at 37◦C. Formazan crystals in viable cells were solubilized with 100 mLDMSO and spectrophotometrically quantified using a microplate reader (Varioskan Flash; Thermo Fisher Sci- entific, Waltham, MA) at 550 nm of absorbance. Frac- tional survival was calculated as percentage to control cells. Data of combined drug effects were analyzed by the Bliss method.11The primary end point of this study was to examine the potential effects of gene mRNA expression levels onsurvival. PFS and OS were estimated by means of the Kaplan-Meier method and compared with a nonpara- metric log-rank test. Biomarker expression was assessed as a dichotomous estimate (low versus high using the median as the cutoff).7 A Cox proportional hazard model was applied with potential risk factors as covariates, obtaining HR and their 95% CI. Each analysis was per- formed with the use of a two-sided 5% significance level and a 95% CI. The statistical analyses were performed using SAS version 9.4. In vitro data were analyzed using unpaired t test (GraphPad Prism, GraphPad Software, Inc.). Values of p less than 0.05 were considered statistically significant. Proportion hazard regression analyses were generated using GraphPad Prism.
Results
From the 182 patients of the ITT analysis of the GOAL study,5 91 had sufficient tumor tissue for the gene expression analysis of BRCA1, CtIP, 53BP1, and AXL. The characteristics of the 91 patients, enrolled at 30 centers in Spain and one in Mexico, are illustrated in Table 1. Of those 91 patients, 51 were randomized to treatment with gefitinib and 40 were randomized to treatment with gefitinib plus olaparib. Patients allocated to the two arms were well balanced for baseline characteristics including age, sex, smoking history, Eastern CooperativeOncology Group performance status, and type of EGFRmutation (Table 1). The median age of all patients was67.5 years, and 69% of them were female. Most of the patients had Eastern Cooperative Oncology Group per- formance status 1 (68%). In both arms, almost two- thirds of the patients had EGFR exon 19 deletions and one-third had EGFR exon 21 L858R substitutions.BRCA1 mRNA Expression and Treatment OutcomeSimilarly to the whole population of the GOAL study,5 in this analysis of the 91 patients, there was no statisti- cally significant difference in median PFS between the twotreatment arms (p = 0.2419) (Fig. 1A). However, when we dichotomized the 91 patients into two groups, on thebasis of the median mRNA expression of BRCA1 (BRCA1- high group [N = 46] and BRCA1-low group [N = 45]), in the BRCA1-high group, a statistically significant longermedian PFS was found with gefitinib plus olaparib compared with gefitinib single agent.
Specifically, as illustrated in Figure 1B, in the BRCA1-high group, median PFS was 12.9 months (95% CI: 8.6–20.3) in the gefitinibplus olaparib arm, compared with 9.2 months (95% CI: 5.7–12.7) in the gefitinib arm, p = 0.0449 (HR for gefitinib versus gefitinib plus olaparib = 2.04, 95% CI: 1.00–4.13, p = 0.0492). In the BRCA1-low group, a longer median PFS of 14.5 months (95% CI: 9.3–16.7), which did notreach the statistical significance, was found in the gefitinib arm compared with the gefitinib plus olaparib arm, in which patients experienced a shorter median PFS of 10.9 months (95% CI: 7.7–36.4, p = 0.8293) (Fig. 1C).Furthermore, in the gefitinib treatment arm, theBRCA1-high group had a significantly shorter median PFS of 9.1 months (95% CI: 7.1–12.0) compared with10.2 months (95% CI: 9.2–16.7) for the BRCA1-lowgroup (p = 0.0193) (HR = 2.08, 95% CI: 1.11–3.91,p = 0.0223) (Table 2). In the gefitinib plus olaparib arm, the BRCA1-high group had a longer median PFS of 14.6months (95% CI: 8.6–20.3) compared with 10.9 months (95% CI: 7.2–36.4) for the BRCA1-low group, although this increase did not reach statistical significance (p = 0.8755) (Table 2). No statistically significant differenceswere observed in the responses in the BRCA1-high and BRCA1-low groups according to the treatment arm or in the two treatment arms according to the BRCA1 mRNA expression levels (Table 3).Overall, for the 91 patients, similar median OS was observed between the two treatment arms (23.1 mo, 95% CI: 15.1–28.5 for gefitinib versus 23.7 mo, 95% CI:16.4–not reached for gefitinib plus olaparib, p = 0.5385).
No differences in median OS were found between thetwo treatment arms, neither in the BRCA1-high group nor in the BRCA1-low group, although the patients in the BRCA1-high group experienced numerically shorter OS compared with those in the BRCA1-low group, inde- pendently of the treatment arm (Fig. 2A–C). This obser- vation was reinforced by data obtained from the R2: Genomics Analysis and Visualization Platform database (http://r2.amc.nl, Data Set Tumor Lung-Bild-114- MAS5.0-u133p2, EGFR-mutant). As presented in Figure 2D, patients with EGFR-mutant NSCLC with high BRCA1 mRNA expression have a significantly worseprognosis than those with low BRCA1 mRNA expression (p = 0.029).Our data indicate that patients with EGFR-mutantNSCLC with high BRCA1 mRNA expression may derive a better outcome with gefitinib plus olaparib compared with gefitinib single agent. High BRCA1 mRNA expression0.48 (range: 0.42–0.58) was found in the PC9-BrM3 cell line, compared with the parental PC9 cell line with me- dian combination index of 0.52 (range: 0.29–0.86).These preliminary and exploratory preclinical find- ings reinforce our clinical observations in the GOAL study, in which in the BRCA1-high group the combina- tion of gefitinib plus olaparib conferred a significantly longer PFS compared with gefitinib alone (p = 0.0449),whereas no differences between the two treatment arms were found in the BRCA1-low group.
Correlation of CtIP, 53BP1, and AXL mRNA Expression and Treatment OutcomeWe then explored whether the mRNA expression of other than BRCA1 biomarkers may have an impact on the treatment outcome, as prespecified in the protocol of theGOAL study. Similarly, to what we performed for the BRCA1 mRNA expression, the 91 patients were dichot- omized into two groups, on the basis of the median mRNA expression of each of the three biomarkers, CtIP, 53BP1, and AXL.In the case of CtIP mRNA expression, we obtained statistically significant associations with gefitinib treat- ment, and we were able to make some interesting ob- servations that merit further investigation. First, in the CtIP-low group, median PFS was numerically longer with gefitinib plus olaparib (12.5 mo, 95% CI: 5.4–notreached) compared with gefitinib alone (9.2 mo, 95% CI: 6.6–10.2) (p = 0.2479) (Fig. 4A). Second, in the gefitinib treatment arm, the CtIP-low group had a statisticallysignificant (Fig. 4A) shorter median PFS of 9.0 months (95% CI: 5.4–9.3) compared with 13.3 months (95% CI: 9.7–14.9) for the CtIP-high group (p = 0.0071; HR: 2.52,95% CI: 1.26–5.07, p = 0.0093). No statistically signifi-cant or clinically relevant correlations were found on thebasis of the expression levels of 53BP1 or AXL (Fig. 4A). In our study, no significant or relevant findings were found for OS on the basis of the mRNA expression levels of the three biomarkers or the treatment arm (Fig. 4B).
In the R2: Genomics Analysis and Visualization Platform database (http://r2.amc.nl, Data Set Tumor Lung-Bild- 114-MAS5.0-u133p2, EGFR-mutant), patients with EGFR- mutant NSCLC with low CtIP mRNA expression have a significantly worse prognosis than those with high CtIPmRNA expression (p < 0.001) (Fig. 4C). In our previous study, we were not able to find any significant correla- tion between CtIP mRNA expression and outcome toerlotinib.Collectively, these data point out that besides high BRCA1 mRNA expression, low mRNA expression of CtIP may be an indicator of worse outcome to single therapy with EGFR TKIs. Whether patients with EGFR-mutant NSCLC who are CtIP-low expressers may derive benefit from the combination of EGFR TKIs with PARP inhibitors requires further research.
Discussion
The biomarker-related prespecified secondary objective of the GOAL study was achieved. The combi- nation of gefitinib plus olaparib significantly improved PFS in the BRCA1-high group, as was predicted in our model (p = 0.0449).2 In addition, the BRCA1-low group had significantly longer PFS when treated with gefitinib alone compared with the BRCA1-high group (p = 0.0193). Nevertheless, no significant benefit of the combination of gefitinib plus olaparib was observed in the whole population of EGFR-mutant patients in the GOAL study.5 In our preclinical experiments, there was no difference in sensitivity to gefitinib between cells with higher BRCA1 mRNA expression (PC9-BrM3 cells) compared with those with lower BRCA1 mRNA expres- sion (parental PC9 cells). BRCA1 protein contains two BRCT (BRCA1 C-termi- nal) repeats, which form exclusive complexes with Abraxas, BACH1 and CtIP. These complexes are defined as BRCA1 A complex (RAP80 and Abraxas), B complex (BACH1), and C complex (CtIP and RAP80). The BRCA1 A complex is involved in DNA damage response; however, depletion of only BRCA1 has stronger effect on the various DNA damage response assays compared with depletion of either Abraxas or RAP80.13 Both the A and C complexes are required for the G2-M checkpoint. They are also implicated in transcription.13In this study, we observed that low CtIP mRNA levels predicted longer PFS in EGFR-mutant patients treatedwith gefitinib plus olaparib, in comparison with those treated with gefitinib single agent.
Noteworthy was the fact that, in the gefitinib treatment arm, the CtIP-low group had significantly shorter median PFS in compari- son with the CtIP-high group (HR = 2.52, p = 0.0093). Inthe case of 53BP1 mRNA expression,14 we were not ableto find any statistically significant correlation with the outcome to gefitinib or gefitinib plus olaparib. These findings are in concordance with our previous model that BRCA1-dependent repair of erlotinib-induced DNA damage occurs independently of the homologous recombination pathway.2 Although AXL overexpression is associated with intrinsic and acquired resistance to EGFR TKIs,7,15 in the current study, we did not find any significant association between AXL mRNA expression and treatment outcome. AXL has been noted to be associated to many components of the homologous recombination pathway.16In our previous study, BRCA1 mRNA expression predicted outcome to gefitinib or erlotinib in patients with EGFR-mutant NSCLC opening opportunities for alternative therapies, including PARP inhibitors or chemotherapy customization.2 BRCA1 regulation is associated with resistance to cisplatin17 and differential expression of BRCA1 supports the use of taxanes.18,19 In fact, intercalated combination of chemotherapy and erlotinib leads to a PFS of 16.8 months in patients with EGFR-mutant NSCLC, in comparison with 6.9 months in the chemotherapy group.
These differences also trans- late to significant improvement in median OS in the chemotherapy plus erlotinib group.20Finally, we noted that in the PC9-BrM3 cell line, a brain metastatic variant derived from the PC9 cell line, highly metastatic to the bones and brain, BRCA1 mRNA expression was higher, and a stronger synergistic effect was observed when gefitinib was combined with olaparib, compared with the parental PC9 EGFR-mutant cell line. Importantly, PC9 cells are sensitive to chemotherapy with etoposide or camptothecin through cyclic GMP-AMP synthase (cGAS) overexpression, which disrupts the for- mation of the PARP1-timeless complex for DNA repair by homologous recombination.21,22 Aberrant up-regulation of cGAS transcripts is often noted in NSCLC and could represent a biomarker to predict response to chemo- therapy. This mechanism of action is in contrast with the effect of PARP inhibitors, which prevent the translocation of cGAS from the cytoplasm to the nucleus.22The study has some shortcomings. Pretreatment tu- mor specimens for genetic analyses were available from only 91 of the 182 patients of the ITT analysis of the GOAL study. The relatively small sample size may have limited the statistical power of our results. The expres- sion of PARP1 and its effect on transcription regulation were not examined.23 PARP1 up-regulation has beenfound in models of castration-resistant prostate cancer and promotes cancer progression by DNA repair and transcriptional regulation.23 PARP1 interacts with and poly-ADP-ribosylates BRCA1.24 Finally, the fact that PARP1 enhances lung adenocarcinoma metastases im- plicates that PARP1 has a tumor progression effect independently of its role in DNA repair.
In conclusion, BRCA1 mRNA expression may poten- tially predict the benefit of combining gefitinib plus olaparib and usher the investigation of chemotherapy plus EGFR TKIs in patients with EGFR-mutant NSCLC. BRCA1 mRNA expression levels may be used for customizing therapy. The role of CtIP as a biomarker to predict the outcome to EGFR TKIs warrants further research.