Prediction of breast cancer sensitivity to neoadjuvant chemotherapy based on status of DNA damage repair proteins

Recent advances in chemotherapy have significantly improved the prognosis of breast cancer patients. However, prediction of tumor sensitivity to chemotherapy has not reached a high level of confidence, whereas determining sensitivity to hormone therapy or trastuzumab is relatively more established. Estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor (HER)2/ErbB2 are practical benchmarks to exclude non-responding patients, and tailoring treatment based on gene status significantly optimizes the response rate of hormone therapy and trastuzumab, respectively. Prediction of chemosensitivity with equivalent accuracy is currently anticipated to further improve breast cancer prognosis.

Anthracycline-based regimens, such as epirubicin plus cyclophosphamide (EC), and taxanes represent the major chemotherapeutic agents used in the breast cancer field [1, 2]. Of these, anthracycline-based chemotherapy induces DNA double-strand breaks (DSBs) [3, 4], the most cytotoxic DNA lesion, that leads cells into apoptosis especially when relevant repair pathways (represented by homologous recombination (HR) repair) are perturbed [5]. It is important to note that DNA damage repair competence varies among individual breast tumors and closely correlates with chemosensitivity. For example, secondary mutations of BRCA1 or BRCA2 (essential factors in the HR pathway) caused by chemotherapy using cisplatin or poly(ADP-ribose) polymerase inhibitor in BRCA1/2-mutated cancers restore the wild-type reading frame and, therefore, the tumor acquires resistance to these drugs [6‐8]. These facts indicate that chemosensitivity of BRCA-associated cancers could be strongly affected by DNA damage repair capability. Based on this evidence it has been suggested that HR competence could be a potential biomarker for chemosensitivity [9]. Rad51, a protein that plays a direct role in HR, especially reflects the HR competence of cells. Therefore, knowing its status is likely to be valuable when assessing HR competence in tumor cells in order to instruct therapeutic decisions [9].

The HR pathway for DSB repair is executed by sequential recruitment of repair proteins to chromatin around DNA lesions. Accumulation of the proteins is regulated by complex mechanisms that utilize phosphorylation and ubiquitination modifications mediated by kinases including ataxia telangiectasia mutated (ATM), and at least four ubiquitin E3 ligases, RNF8, RNF168, Rad18, and BRCA1 [10‐17]. The Mre11-Rad50-Nbs1 complex first recognizes DSBs and recruits ATM. ATM then phosphorylates the histone variant H2AX (γH2AX) [18, 19] that triggers accumulation of the downstream E3 ligases RNF8 [11‐13, 20] and RNF168 [14, 15]. Lysine 63-linked polyubiquitin chains built at the sites of DNA damage by these E3 ligases next recruits the BRCA1-Abraxas-RAP80 complex through the RAP80 component, a protein that contains ubiquitin interacting motif domains [21‐23]. BRCA1 is then essential in order to recruit repair effector proteins, including Rad51, that perform HR through sister chromatid exchange [24, 25]. Depletion of any one of these proteins results in HR deficiency accompanied by loss of Rad51 focus formation, causing cells to become hypersensitive to DSB-inducing agents.

In the present study using human tumor specimens we show for the first time that DNA repair competence may predict breast cancer sensitivity to DNA damage-inducing chemotherapy. We selected γH2AX, conjugated ubiquitin, BRCA1, and Rad51, proteins in the DSB repair cascade, to assess DNA repair competence because accumulated evidence demonstrates that inactivation of genes in the DSB repair pathway results in cellular sensitivity to DNA damage-inducing chemotherapy [16, 29, 31, 36‐38]. In our study, these repair proteins dramatically responded to EC treatment. The conjugated ubiquitin response was especially dramatic as approximately half of the cases analyzed formed conjugated ubiquitin foci, compared with undetectable foci formation prior to treatment. This suggests that ubiquitination occurs in vivo during the DNA damage response in an early stage after chemotherapy. However, in spite of the dramatic response, we did not find any significant correlation between conjugated ubiquitin foci formation and tumor response. The reason is currently unknown. One possibility is that this could be attributed to the fact that ubiquitination is also involved in DNA damage response pathways other than for DSBs.

We did not find certain trends of the combinations of responding repair proteins. Several reasons could account for this observation. First, the metabolism and pharmacokinetics of the agents could vary per patient. The ideal time to obtain the in vivo sample was, therefore, difficult to determine. The experimental design employed in this study was not very robust in this way. In cultured cells, γH2AX accumulates at sites of DNA damage just minutes after the damage occurs, whereas BRCA1 and Rad51 foci appear 30 minutes to several hours afterwards [11, 35, 39, 40]. In this study we harvested samples 18 to 24 hours after EC treatment because the agents were still expected to be present in patients and we also considered the patient's convenience. However, the ideal timing remains to be determined if biopsy after chemotherapy is required.

The second reason for the diversity of the DDR response could be attributed to the diversity of aberrations of the genes responsible for DSB repair in each breast cancer. Theoretically, defects in the recruitment of upstream repair proteins could result in loss of downstream proteins at sites of DNA damage, and this has been shown to be the case in many molecular biological studies using cultured cells [10‐15, 21‐23]. Furthermore, it was also shown that Rad51 nuclear expression is absent in tumors associated with BRCA2 mutation [41]. The positive correlation found between EC-induced BRCA1 and Rad51 foci in this study (Table 2) may also support this interpretation. In contrast, it was reported that overexpression of Rad51 restored Rad51 focus formation and rescued the sensitivity of BRCA1-deficient cells to x-rays and cisplatin [42]. Importantly, up-regulation of Rad51 was a common feature of BRCA1-deficient breast tumors [42]. These data suggest that the mechanism of DSB repair response in vivo is not simple and that assessment of DSB repair aberrations in each patient case is, therefore, unreasonable at present.

In an attempt to address this problem in our current study, we assessed the comprehensive capacity of DSB repair by incorporating multiple candidate factors into one DDR score. We found that foci of BRCA1, γH2AX, and Rad51 prior to treatment and EC-induced foci of Rad51 correlated with tumor response when compared either with the mean tumor volume reduction or the tumor response rate. When incorporating these four factors into one DDR score a significant correlation was observed with mean tumor volume reduction after EC, whereas no other factors correlated with the mean tumor volume reduction (Table 4 and Figure 3a). Although it was not statistically significant, the similar correlation was also observed between DDR score and tumor response rate (Table 3). These correlations became more significant after EC and DOC treatment (Tables 3 to 5 and Figure 3b) and the DDR score was an independent predictive factor of other factors including tumor subtype when evaluated with volume reduction using 50% of the PR/SD border (Table 6). Recent studies suggested that luminal tumors have low response rate to neoadjuvant chemotherapy, whereas basal-like and HER2+ tumors have higher response rates. For example, it has been reported that clinical response rate (CR and PR) to anthracyclin-based chemotherapy of luminal A was 39%, whereas that of basal-like, which has been implicated with BRCA1 dysfunction [43, 44], was 85% [45]. The response rates to EC treatment of luminal A (15 of 37 cases, 40.5%) and basal-like (4 of 6 cases, 66.7%) subtypes in the current study were not very different from the previous report. However, we could not find any correlation between subtype and DDR score while DDR score independently predicted the chemosensitivity. The result may reflect the fact that luminal A tumors also include DNA damage-sensitive tumors with defective HR pathways that can be counted by the DDR score. Supporting this it has been shown that tumors caused by BRCA2 deficiency mainly become luminal A tumors [44, 46, 47].

The reason why the correlation between the DDR score and tumor response after EC and DOC treatment became more significant than that after EC is not clear at present. As DOC does not induce DNA DSBs, the observed effect is not likely to be due to the sensitivity to DNA damage in those tumors. DOC might be more toxic for the cells with gross genomic aberration caused by the pretreatment with EC under the condition of being less HR competent. Alternatively it is possible that time length after EC treatment enhanced the difference of the outcome.

Interestingly, DDR score group 4 consisted of cases with poor tumor responses to chemotherapy when evaluated for both mean tumor volume reduction (Figure 3) and tumor response rate (Table 3). This result may lead to the possibility of using DDR status in the clinic to predict and exclude non-responders to EC treatment. It is noteworthy to point out that the HR repair cascade for DSB contains many essential proteins other than those tested in this study. By including select subsets of proteins for analysis, it may be possible to identify non-responders in order to avoid unnecessary chemotherapy. Ideally in such cases, the levels of baseline foci present prior to treatment would provide enough information to determine appropriate treatment, preventing the need for additional core needle biopsy after chemotherapy.

In conclusion, our results suggest the importance of evaluating DDR competence to predict breast cancer chemosensitivity and warrant further investigation into its effectiveness as a way to exclude non-responding patients.