Transcriptomic and Protein Expression Analysis Reveals Clinicopathological Significance of Bloom Syndrome Helicase (BLM) in Breast Cancer

Bloom syndrome helicase (BLM) is a key member of the RecQ family of DNA helicases and essential for the maintenance of genomic stability. BLM is an ATP-dependent 3′–5′ DNA helicase involved in unwinding a variety of DNA substrates that can arise during DNA replication and repair (1–5). BLM has important roles in the initiation and regulation of homologous recombination (HR) repair of double-strand breaks (DSB). In addition, BLM is required for Holliday junction dissolution during the terminal stages of HR. To accomplish its various biologic functions, BLM interacts with several DNA repair factors, including topoisomerase III, hRMI1, hRMI2, and Rad51. BLM is also part of the BRCA1-associated genome surveillance complex (BASC), which contains BRCA1, MSH2, MSH6, MLH1, ATM, PMS2, and the RAD50–MRE11–NBS1 protein complex (6). In addition to its DNA repair function, BLM is involved in the processing of stalled replication forks during replication and in telomere maintenance in cells (1–5).

Bloom syndrome is a rare disorder caused by germ-line mutation in the BLM gene. Bloom syndrome is characterized by cancer predisposition, growth retardation, immunodeficiency, sunlight hypersensitivity, and impaired fertility (7). BLM germ-line mutation results in dramatic reduction in BLM mRNA levels and BLM protein expression, leading to extensive chromosomal instability manifested classically as excessive frequency of sister chromatid exchanges (SCE) in Bloom syndrome cells (1–5). Patients with Bloom syndrome are prone to develop leukemia, lymphomas, and a variety of epithelial cancers, including breast cancers (7). Interestingly, polymorphisms in the BLM gene have been associated with increased risk of development of sporadic breast cancers (8). In preclinical models, depletion of BLM by shRNA not only reduced proliferation in cells (9) but also sensitized them to chemotherapeutic agents such as camptothecins, cisplatin, 5-fluoruracil, and hydroxyurea treatment (1–5, 7). BLM is an attractive anticancer drug target and small-molecule inhibitors of BLM are currently under preclinical development (10). However, target validation studies including prognostic and/or predictive significance of BLM in human sporadic tumors have not been reported and therefore remain largely unknown. We hypothesized that BLM may be dysregulated in sporadic breast cancers and influence clinical outcomes in patients. In this study, we present the first and the largest comprehensive study providing compelling evidence that altered BLM expression has prognostic and predictive significance in patients. Our data suggest that BLM is a rational target in breast cancer.

METABRIC (Molecular Taxonomy of Breast Cancer International Consortium) cohort was evaluated for BLM gene expression. The METABRIC study protocol, detailing the molecular profiling methodology in a cohort of 1,980 breast cancer samples, is described by Curtis and colleagues (11). Patient demographics are summarized in Supplementary Table S1 of supporting information. Estrogen receptor (ER)–positive (ER⁺) and/or lymph node–negative patients did not receive adjuvant chemotherapy. ER-negative (ER⁻) and/or lymph node–positive patients received adjuvant chemotherapy. RNA was extracted from fresh-frozen tumors and subjected to transcriptional profiling on the Illumina HT-12 v3 platform. The data were preprocessed and normalized as described previously (11). BLM expression was investigated in this dataset. There was only one probe for BLM (BLM probe id: ILM_1709484) in the Illumina HT-12 v3 platform. This probe has a perfect quality score as no repeat regions were targeted by the probe. The χ² test was used for testing association between categorical variables and a multivariate Cox model was fitted to the data using breast cancer–specific death as an endpoint. Recursive partitioning was used to identify a cutoff in gene expression values such that the resulting subgroups have significantly different survival courses.

The external validation was done using bc-GenExMiner v3.0 (Breast Cancer Gene-Expression Miner v3.0) online dataset (http://bcgenex.centregauducheau.fr) comprising previously published gene expression datasets from 15 independent breast cancer studies totaling 2,413 tumors and summarized in Supplementary Table S2. The bioinformatics tool is composed of two statistical mining modules. The first module is a “prognostic module,” which offers the possibility to evaluate the in vivo prognostic informativity of genes of interest in breast cancer, and the second module is a “correlation module,” which permits to compute correlation coefficients between gene expressions or to find lists of correlated genes in breast cancer. We used the prognostic module in this external validation. Statistical analyses were performed by means of survival statistical tests (Cox model, Kaplan–Meier, and Forest plots). Supplementary Table S2 summarizes individual cohorts where BLM mRNA expression was investigated.

The study was performed in a consecutive series of 1,650 patients with primary invasive breast carcinomas who were diagnosed between 1986 and 1999 and entered into the Nottingham Tenovus Primary Breast Carcinoma series. Patient demographics are summarized in Supplementary Table S3. This is a well-characterized series of patients with long-term follow-up that have been investigated in a wide range of biomarker studies (12–20). All patients were treated in a uniform way in a single institution with standard surgery (mastectomy or wide local excision) with radiotherapy. Before 1989, patients did not receive systemic adjuvant treatment. After 1989, adjuvant treatment was scheduled on the basis of prognostic and predictive factor status, including Nottingham Prognostic Index (NPI), ER status, and menopausal status. Patients with NPI scores of <3.4 (low risk) did not receive adjuvant treatment. In premenopausal patients with NPI scores of ≥3.4 (high risk), classical cyclophosphamide, methotrexate, and 5-flurouracil (CMF) chemotherapy was given; patients with ER⁺ tumors were also offered endocrine therapy. Postmenopausal patients with NPI scores of ≥3.4 and ER positivity were offered endocrine therapy, while ER⁻ patients received classical CMF chemotherapy. Median follow-up was 111 months (range, 1–233 months). Overall survival data were maintained on a prospective basis. Breast cancer–specific survival (BCSS) was defined as the number of months from diagnosis to the occurrence of BC-related death. Survival was censored if the patient was still alive at the time of analysis, lost to follow-up, or died from other causes. We also evaluated 20 tumor-associated normal breast tissue for BLM expression.

Tumor Marker Prognostic Studies (REMARK) criteria, recommended by McShane and colleagues (21), were followed throughout this study. Ethical approval was obtained from the Nottingham Research Ethics Committee (C202313).

Tumors were arrayed in tissue microarrays (TMA) constructed with two replicate 0.6-mm cores from the center and periphery of the tumors. The TMAs were immunohistochemically profiled for BLM and other biologic antibodies (Supplementary Table S4) as previously described (12–20). Immunohistochemical (IHC) staining was performed using the Thermo Scientific Shandon Sequenza chamber system (REF: 72110017), in combination with the Novolink Max Polymer Detection System (RE7280-K: 1,250 tests), and the Leica Bond Primary Antibody Diluent (AR9352), each used according to the manufacturer's instructions (Leica Microsystems). The tissue slides were deparaffinized with xylene and then rehydrated through five decreasing concentrations of alcohol (100%, 90%, 70%, 50%, and 30%) for 2 minutes each. Pretreatment antigen retrieval was performed on the TMA sections using sodium citrate buffer (pH 6.0) and heated for 20 minutes at 95°C in a microwave (Whirpool JT359 Jet Chef 1000W). A set of slides were incubated for 18 hours with the primary anti-BLM antibody (NBP1-89929; Novus Biologicals), at a dilution of 1:100. Negative and positive (by omission of the primary antibody and IgG-matched serum) controls were included in each run. The negative control ensured that all the staining was produced from the specific interaction between antibody and antigen.

The tumor cores were evaluated by two scorers (T.M.A. Abdel-Fatah and A. Arora) and the concordance between the two scorers was excellent (k = 0.79). Whole field inspection of the core was scored and intensities of nuclear staining were grouped as follows: 0 = no staining, 1 = weak staining, 2 = moderate staining, 3 = strong staining. The percentage of each category was estimated (0%–100%). Histochemical score (H score; range, 0–300) was calculated by multiplying intensity of staining and percentage staining. A median H score of ≥50 was taken as the cutoff for high BLM nuclear and cytoplasm expression. Not all cores within the TMA were suitable for IHC analysis as some cores were missing or lacked tumor (<15% tumor).

Data analysis was performed using SPSS (SPSS, version 17). Where appropriate, the Pearson χ², Fisher exact, Student t, and ANOVA one-way tests were used. Cumulative survival probabilities were estimated using the Kaplan–Meier method, and differences between survival rates were tested for significance using the log-rank test. Multivariate analysis for survival was performed using the Cox proportional hazard model. The proportional hazards assumption was tested using standard log–log plots. Hazard ratios (HR) and 95% confidence intervals (95% CI) were estimated for each variable. All tests were two-sided with a 95% CI and a P < 0.05 was considered significant. For multiple comparisons, P values were adjusted according to Benjamini-Hochberg method (22).

MCF-7 (ER⁺/PR⁺/HER2⁻, BRCA1 proficient), MDA-MB-231 (ER⁻/PR⁻/HER2⁻, BRCA1 proficient), MDA-MB-468 (ER⁻/PR⁻/HER2⁻, BRCA1 proficient), and MDA-MB-436 (ER⁻/PR⁻/HER2⁻, BRCA1 deficient) were used in this study. All cell lines were purchased from the ATCC and authenticated by the ATCC. Cells were grown in RPMI (MCF-7, MDA-MB-231) or DMEM (MDA-MB-468 and MDA-MB-436) medium with the addition of 10% fetal bovine serum and 1% penicillin/streptomycin. Cell lysates were prepared and Western blot analysis was performed. Primary anti-BLM antibody (NBP1-89929; Novus Biologicals) was incubated overnight at room temperature at a dilution of 1:1,500. Primary anti-β actin antibody (1:10,000 dilution; Abcam) was used as a loading control. Infrared dye-labeled secondary antibodies (Li-Cor) [IRDye 800CW Mouse Anti-Rabbit IgG and IRDye 680CW Rabbit Anti-Mouse IgG] were incubated at a dilution of 1:10,000 for 1 hour. Membranes were scanned with a Li-Cor Odyssey machine (700 and 800 nm) to determine protein expression.

Total RNA was extracted from MCF-7, MDA-MB-231, MDA-MB-468, and MDA-MB-436 cells using RNeasy Mini kit (Qiagen). The quantification of the extracted RNA was done using a NanoDrop 2000c Spectrophotometer (Thermo Scientific). The cDNA was synthesized from 0.5 μg of total RNA using RT² first-strand kit (Qiagen). qPCR was performed using SYBR Green PCR Master mix (Applied Biosystems) with primer set (BLM QuantiTect Prier Assay; cat. no. QT00027671; Qiagen) targeting BLM gene. The glyceraldehyde-3-phosphate dehydrogenase housekeeper gene was used as an internal control (GAPDH QuantiTect Prier Assay; cat. no. QT00079247; Qiagen). The real-time PCR for each RNA sample was performed in triplicate. No template control (NTC) was used to rule out cross-contamination of reagents and surfaces. NTC included all the RT-PCR reagents except the RNA template. Minus reverse transcriptase (−RT) control was used to rule out genomic DNA contamination.

BLM mRNA level was investigated in the METABRIC (Molecular Taxonomy of Breast Cancer International Consortium) cohort comprising 1,980 breast tumors. High BLM mRNA expression was highly significantly associated with aggressive clinicopathologic features (Table 1), including high histologic grade, larger tumor size, high-risk NPI (NPI > 3.4), Her2 overexpression, ER⁻, progesterone receptor (PR)–negative, and triple-negative phenotypes (ps < 0.0001). High BLM mRNA expression was also found to be significantly associated with previously described molecular phenotypes in breast cancer: PAM50.Her2 (P < 0.0001), PAM50.Basal (P < 0.0001), and PAM50.LumB (P < 0.0001), Genufu subtype (ER⁻/Her2⁻), Genufu subtype (ER⁺/Her2⁻/high proliferation), and Genufu subtype (Her2-positive) breast tumors. However, PAM50.LumA tumors and Genufu subtype (ER⁺/Her2⁻/low proliferation) were more likely to express low levels of BLM mRNA (ps < 0.0001). Similarly, BLM mRNA level was significantly associated with the various biologic subgroups [labeled integrative clusters (intClust) 1–10] described in the METABRIC study, which was based on gene copy number changes and gene expression data (11). High BLM mRNA expression was significantly associated with intClust.1 (P < 0.0001), intClust.5 (P < 0.0001), intClust.9 (P < 0.0001), and intClust.10 (P < 0.0001), which had the worst clinical outcome in the METABRIC study (11). Low BLM mRNA expression was associated with intClust.3 (P < 0.0001), intClust.4 (P < 0.0001), intClust.7 (P = 0.003), and intClust.8 (P < 0.0001), which had intermediate to good prognosis in the METABRIC study (11).

We then proceeded to survival analysis. High BLM mRNA expression in tumors was associated with adverse BCSS in the whole cohort (P < 0.0001; Fig. 1A). In ER⁺ subgroup, high BLM mRNA expression was associated with poor BCSS (P < 0.0001; Fig. 1B). In the ER⁺ subgroup that received adjuvant endocrine therapy, high BLM mRNA expression remains associated with poor BCSS (P < 0.0001; Fig. 1D). In ER⁻ subgroup, low BLM mRNA expression was associated with poor BCSS with borderline significance (P = 0.049; Fig. 1C). In the ER⁻ subgroup that received adjuvant chemotherapy, although there was a trend, BLM mRNA expression did not significantly influence outcome (P = 0.062; Fig. 1E) and was most likely due to limited number of patients in this cohort (n = 262). In multivariate Cox regression analysis that included other validated prognostic factors, such as lymph node stage, histologic grade, and tumor size, BLM mRNA expression was a powerful independent predictor for BCSS (P < 0.00001; Table 2). External validation was performed using bc-GenExMiner v3.0 (Breast Cancer Gene-Expression Miner v3.0) online dataset (http://bcgenex.centregauducheau.fr) comprising previously published gene expression datasets from 15 independent breast cancer studies totaling 2,413 tumors and summarized in Supplementary Data and Supplementary Table S2. The dataset provides information on metastasis relapse–free survival data. As shown in the Forest plot (Supplementary Fig. S1), low BLM mRNA expression was significantly associated with better metastasis relapse–free survival (Supplementary Fig. S1A and S1B). Taken together, the data provide the first compelling evidence that high BLM mRNA expression has prognostic and/or predictive significance in breast cancer.

BLM is a 1,417 amino acid protein with a highly conserved centrally located helicase domain. In addition, BLM has multiple domains involved in DNA-binding, ATPase activity, and interaction with other binding partners. The nuclear localization signal is present in the C-terminal region of the protein (1–5). BLM is primarily expressed in late S–G₂ phase of the cell cycle. Upon DNA damage, BLM localizes to the nucleus where it interacts with Rad51 and is intimately involved in HR repair that is operational during the S-phase of the cell cycle (23). In addition, BLM undergoes posttranslational modifications such as phosphorylation and SUMOylation that can affect intracellular localization and biochemical activity (1–5). Besides a role in HR repair, BLM is also known to interact with key factors involved in base excision repair (BER; e.g., FEN1) and nonhomologous end joining pathway (NHEJ; e.g., DNA-PKcs; refs. 1–5). Moreover, BLM also interacts with important players in DNA-damage signaling and cell-cycle regulation (ATM-Chk2 and ATR-Chk1 pathway), which ultimately dictate whether a cell initiates cell-cycle arrest to allow DNA repair or proceed to apoptosis (1–5). We therefore investigated BLM protein expression in breast cancer and correlated to expression of other markers associated the DNA-damage signaling, NHEJ, BER, cell-cycle regulation, and apoptosis.

We proceeded to evaluation of BLM protein expression in breast cancers. We initially profiled a panel of breast cancer cell lines. As shown in Supplementary Fig. S2A, MDA-MB-231, MDA-MB-436, and MDA-MB-468 breast cancer cells have robust expression of BLM protein. In contrast, MCF-7 has low BLM expression. At the mRNA level, MCF-7 cells have low BLM mRNA compared MDA-MB-231, MDA-MB-436, and MDA-MB-468 cells. The data demonstrate differential BLM expression across different breast cancer cell lines. We then conducted IHC evaluation of BLM protein expression in the Nottingham Tenovus series comprising 1,650 breast tumors. Surprisingly, we observed complex subcellular localization of BLM protein in breast cancers, including tumors exhibiting nuclear staining only, cytoplasmic staining only, nuclear-cytoplasmic coexpression, or negative staining. We also evaluated 20 tumor-associated normal breast tissues for BLM expression. We observed strong nuclear staining in 19 of 20 normal breast tissue (mean H score = 235; Supplementary Fig. S2B1). One of 20 did not show any nuclear BLM staining. No cytoplasmic staining was observed in any normal breast tissue. The data confirm that nuclear expression is a common feature of normal breast tissue and altered subcellular localization is a feature of breast tumors.

Low nuclear BLM levels were seen in 54% of tumors (n = 682/1,253) and high nuclear BLM levels were observed in 46% of tumors (n = 571/1,253; Supplementary Fig. S2B4). As shown in Supplementary Table S5, low nuclear BLM level was significantly associated with larger tumors, high tumor grade, higher mitotic index, pleomorphism, and tumor type (P < 0.05). ER⁻, PR⁻, AR⁻, triple-negative, and basal-like phenotypes were more common in tumors with low nuclear BLM protein level (P < 0.01). BRCA1-negative, low XRCC1, low FEN1, low SMUG1, low APE1, low Polβ, low ATR, and low DNA-PKcs were significantly associated with tumors that have low nuclear BLM protein level. In addition, high p16, low p21, high MIB1, high p53, low Bcl-2, low Top2A, low nuclear pCHEK1, and low nuclear Chk2 were more common in tumors with low nuclear BLM protein level (P < 0.05).

High cytoplasmic BLM levels were seen in 53% of tumors (n = 642/1,212) and low cytoplasmic BLM levels were seen in 47% of tumors (n = 570/1,212; Supplementary Fig. S2B3). As shown in Supplementary Table S6, high cytoplasmic BLM level was significantly associated with pleomorphism, tumor type, high XRCC1, high FEN1, high APE1, high ATR, high DNA-PKcs, high MIB1, high Chk2, and high Bax levels.

Twenty-eight percent (333 of 1,253) of tumors were low nuclear/high cytoplasmic, 26.5% (332 of 1,253) were low nuclear/low cytoplasmic, 26.5% (333 of 1,253) were high nuclear/high cytoplasmic, and 19% (238 of 1,253) were high nuclear/low cytoplasmic (Supplementary Fig. S2B5). Clinicopathologic association are shown in Table 3 and Supplementary Table S7. Tumors with high cytoplasmic/low nuclear BLM levels were more likely to be high grade, high mitotic index, pleomorphism, IDC-NST tumor type, PR⁻, triple-negative, and basal-like phenotype tumors (P < 0.0001). High p16, low p21, high MIB1, high p53, and high Bax levels are more common in tumors with high cytoplasmic/low nuclear BLM levels. We also correlated BLM coexpression with various DNA repair factors and observed significant associations. BRCA1 negativity was observed in 24.6% of BLM n⁻/c⁻ tumors compared with 13.2% (BLM n⁺/c⁻ tumors), 20.3% (BLM n⁻/c⁺ tumors), and 17.3% (BLM n⁺/c⁺ tumors). Similarly, BLM n⁻/c⁻ tumors were more likely to exhibit low XRCC1 (25.6%), low FEN1 (83.8%), low SMUG1 (47.1%), low APE1 (66.8%), low pol β (50.9%), low ATR (75.9%), and DNA-PKcs (45.8%) compared with tumors that express BLM n⁺/c⁻, BLM n⁻/c⁺, or BLM n⁺/c⁺ coexpression (see Table 3).

A key interacting partner of BLM is Rad51 (24). Together BLM–Rad51 play an essential role in HR repair (1–5). We therefore conducted exploratory nuclear coexpression studies in breast cancer. As shown in Supplementary Table S8, we observed significant association between BLM⁻/Rad51⁻ tumors and NPI > 3.4, high grade, high mitotic index, pleomorphism, and tumor type. Interestingly, ER negativity was observed in 47.1% of BLM⁻/RAD51⁻ tumors compared with 30.5% (BLM⁺/RAD51⁻ tumors), 30.9% (BLM⁻/RAD51⁺ tumors), and 17.9% in BLM⁺/RAD51⁺ tumors. Similarly, PR negativity was observed in 64.4% of BLM⁻/RAD51⁻ tumors compared with 47.9% (BLM⁺/RAD51⁻ tumors), 42.9% (BLM⁻/RAD51⁺ tumors), and 35.3% (BLM⁺/RAD51⁺ tumors; see Supplementary Table S8).

In univariate analysis, in high-risk ER⁺ tumors that received no endocrine therapy, patients whose tumors had high nuclear/low cytoplasmic BLM had poor BCSS (P = 0.036), implying that altered expression has prognostic significance (Supplementary Fig. S3). In patients who received endocrine therapy, although low nuclear/high cytoplasmic BLM tumors have the worst survival status in breast cancer, there was no statistical significance. Similarly in ER⁻ tumors, BLM level did not significantly influence survival. When BLM (nuclear) and Rad51 (nuclear) were investigated together, BLM⁻/Rad51⁻ tumors have poor survival in the whole cohort and in the ER⁻ subgroup that received adjuvant chemotherapy (Supplementary Fig. S4). BLM/Rad51 expression did not influence survival in ER⁺ tumors (Supplementary Fig. S5). In multivariate analysis (Supplementary Table S9), nuclear BLM level independently influenced survival (P = 0.026). Tumor stage, grade, and HER2 expression were other factors independently associated with BCSS.

DNA helicases are molecular motors that unwind DNA, a process that is required during DNA replication, DNA repair, and telomere maintenance. RecQ family of DNA helicases includes RECQL1, RECQL4, RECQL5, WRN, and BLM. The critical role played by RecQ family of DNA helicases in genomic stability is underpinned by the fact that germ-line mutations in these genes result in genetic disorders characterized by premature aging and/or predisposition to cancers (1–5). RecQ helicases may also have a role in the pathogenesis of sporadic cancers. RECQL4 has been shown to be involved in prostate carcinogenesis (1–5). RECQL1 genetic polymorphisms have been linked to pancreatic cancer and RECQL1 overexpression has been demonstrated in head and neck and brain tumors (1–5). In this study, we have comprehensively investigated the role of BLM in breast cancer. We provide compelling evidence that high BLM mRNA expression is a strong prognostic and predictive biomarker in breast cancer. High BLM mRNA was linked to aggressive clinicopathologic phenotypes. High BLM mRNA was associated with aggressive molecular phenotypes, including PAM50. Luminal B, PAM50. Her2, and PAM50. basal molecular phenotypes. Given the role of BLM during replication and proliferation (25), it is perhaps not surprising that high BLM mRNA was more frequent in aggressive breast cancers. To further support this hypothesis, we also observed that low BLM mRNA expression was more common in PAM50. Lumina A and ER⁺/Her2-negative/low proliferation Genefu subtype tumors. Interestingly, BLM mRNA levels are also linked to biologically distinct integrative clusters reported in the METABRIC study (11). High BLM mRNA level was frequent in intClust 10 subgroup that is the most highly genomically unstable subgroup with basal-like features. Low BLM mRNA level was seen in intClust 3 subgroup that is characterized by low genomic instability. Together, the data suggest that BLM mRNA level may also inform genomic stability status in breast. In addition, high BLM mRNA level is also frequently seen in intClust 5 (HER2 enriched with worst survival), intClust 9 (8q cis-acting/20q–amplified mixed subgroup), and intClust 1 (17q23/20q cis-acting luminal B subgroup) subgroups that also manifest an aggressive phenotype. On the other hand, low BLM mRNA level is linked to intClust 4 (includes both ER⁺ and ER⁻ cases with a flat copy number landscape and termed the “CNA-devoid” subgroup with extensive lymphocytic infiltration), intClust 7 (16p gain/16q loss with higher frequencies of 8q amplification luminal A subgroup), and intClust 8 subgroups (classical 1q gain/16q loss luminal A subgroup; ref. 11). Of note, the data presented here is strikingly similar to the clinicopathologic associations we recently reported for FEN1 (flap endonuclease 1), a key player in long-patch BER and DNA replication, in the METABRIC cohort (15). Interestingly, BLM has been shown to stimulate FEN1 activity in a preclinical study (26). The functional interaction appeared to be independent of BLM helicase activity in that study (26).

At the protein level, low nuclear and/or high cytoplasmic expression was associated with aggressive phenotypes. Association with high cytoplasmic expression was surprising. In contrast, normal breast tissue showed only strong nuclear staining and no cytoplasmic staining. As cytoplasmic function of BLM has not been described previously, we speculate that cytoplasmic accumulation in a proportion of breast tumors probably reflects dysregulation of mechanisms involved in nuclear localization of BLM. Cytoplasmic accumulation along with low nuclear BLM expression could then increase genomic instability in tumors and promote a mutator phenotype characterized by aggressive biology. To support this hypothesis, we observed that low nuclear BLM levels with or without cytoplasmic expression were more likely to be high grade, high mitotic index, pleomorphism, IDC-NST tumor type, PR⁻, triple-negative, and basal-like phenotype tumors. In addition, low nuclear BLM with or without cytoplasmic expression was also associated with impaired expression of other DNA repair factors, including BRCA1 negativity, low XRCC1, low FEN1, low SMUG1, low APE1, low Polβ, low ATR, and low DNA-PKcs. Moreover, in multivariate analysis, nuclear BLM level independently influenced survival. As BLM and Rad51 are known to interact with each other for efficient HR repair (24), we also performed BLM–Rad51 coexpression studies. As expected, low nuclear BLM/low nuclear RAD51 tumors exhibited aggressive phenotype and associated with poor survival. In a previous small study in normal and neoplastic human cells, BLM protein expression was shown to be overexpressed in a panel of tumor tissue compared with normal tissue, including a cohort of nine breast tumors (27). Similar to our study, the authors observed a positive correlation between BLM and Ki67 but did not report any clinicopathologic associations (27). Another interesting observation in this study was that although BLM mRNA overexpression was categorically associated with aggressive tumors and poor outcomes, at the protein level, the association appeared more complex with low nuclear BLM protein level or low nuclear/high cytoplasmic BLM protein level being associated with adverse features. We speculate that either BLM mRNA is subjected to posttranscriptional regulation or posttranslational dysregulation of BLM protein expression/subcellular localization could, in turn, affect BLM mRNA expression through feedback loops. Detailed mechanistic studies are therefore required to understand the regulation of BLM in vivo. Data presented in this study also suggest that BLM could be a promising marker for personalization of therapy. As low BLM is a marker of impaired HR repair, we would argue that low BLM tumors could be targeted by synthetic lethality using inhibitors of BER such as those targeting PARP (28). Alternatively, high BLM tumors could be targeted by small-molecular inhibitors of BLM that are currently under development (10). In conclusion, we provide the first clinical evidence that BLM is a promising biomarker and a rational drug target in breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: A. Arora, T.M.A. Abdel-Fatah, M.A. Aleskandarany, G. Ball, I.O. Ellis, S. Madhusudan

Development of methodology: A. Arora, T.M.A. Abdel-Fatah, M.A. Aleskandarany, S. Madhusudan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Arora, T.M.A. Abdel-Fatah, D. Agarwal, P.M. Moseley, M.A. Aleskandarany, A.R. Green, A.T. Alshareeda, E.A. Rakha, S.Y.T. Chan, I.O. Ellis, S. Madhusudan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Arora, T.M.A. Abdel-Fatah, D. Agarwal, M.A. Aleskandarany, G. Ball, S.Y.T. Chan, S. Madhusudan

Writing, review, and/or revision of the manuscript: A. Arora, T.M.A. Abdel-Fatah, D. Agarwal, R. Doherty, P.M. Moseley, M.A. Aleskandarany, A.R. Green, G. Ball, A.T. Alshareeda, E.A. Rakha, S.Y.T. Chan, I.O. Ellis, S. Madhusudan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Arora, T.M.A. Abdel-Fatah, E.A. Rakha, S.Y.T. Chan

Study supervision: T.M.A. Abdel-Fatah, M.A. Aleskandarany, S.Y.T. Chan, S. Madhusudan

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.