Increased centrosome number in BRCA-related breast cancer specimens determined by immunofluorescence analysis.

BRCA-related breast carcinoma can be prevented through prophylactic surgery and an intensive follow-up regimen. However, BRCA genetic tests cannot be routinely performed, and some BRCA mutations could not be defined as deleterious mutations or normal variants. Therefore, an easy functional assay of BRCA will be useful to evaluate BRCA status. As it has been reported that BRCA functions in the regulation of centrosome number, we focused on centrosome number in cancer tissues. Here, 70 breast cancer specimens with known BRCA status were analyzed using immunofluorescence of γ-tubulin (a marker of centrosome) foci. The number of foci per cell was higher in cases with BRCA mutation compared to wild-type cases, that is, 1.9 (95% confidence interval [CI], 1.5-2.3) vs 0.5 (95% CI, 0.2-0.8) (P < .001). Specifically, foci numbers per cell in BRCA1 and BRCA2 mutation cases were 1.2 (95% CI, 0.6-1.8) and 2.2 (95% CI, 1.7-2.6), respectively, both higher than those in wild-type cases (P = .042 and P < .0001, respectively). The predictive value of γ-tubulin foci as determined by area under the curve (AUC = 0.86) for BRCA status was superior to BRCAPRO (AUC = 0.69), Myriad Table (AUC = 0.61), and KOHBRA BRCA risk calculator (AUC = 0.65) pretest values. The use of γ-tubulin foci to predict BRCA status had sensitivity = 83% (19/23), specificity = 89% (42/47), and positive predictive value = 77% (20/26). Thus, γ-tubulin immunofluorescence, a functional assessment of BRCA, can be used as a new prospective test of BRCA status.

INTRODUCTION

Breast cancer is the most common type of cancer and the second leading cause of cancer‐related death in women.1 The Cancer Genome Atlas study revealed that approximately 10% of breast cancer cases had deleterious germline mutations, of which over half were in the BRCA1 and/or BRCA2 genes.2

Genetic tests for BRCA mutation are important for identifying suspected cases of BRCA‐related breast and ovarian cancer syndrome so that prophylactic surgery and intensive follow‐up programs can be recommended to the patient. Furthermore, BRCA status predicts patient chemosensitivity to platinum or poly (ADP‐ribose) polymerase (PARP) inhibitor.3, 4 Widespread use of PARP inhibitors, thus, necessitates BRCA genetic tests. However, genetic tests are costly and have problems associated with privacy and genetic counseling. Furthermore, some BRCA mutations could not be defined as deleterious mutations or normal variants, and thus, are treated as a variant of unknown significance (VUS). It is therefore important to devise an inexpensive and simple test for predicting BRCA mutation status; functional evaluation of the BRCA protein is particularly desired.

Many risk estimation tools for detecting deleterious BRCA mutations based on clinicopathological information have been reported, such as BRCAPRO,5, 6 Myriad Table,7 and the Korean Hereditary Breast Cancer BRCA risk calculator (KOHCal).8 However, factors such as small family size or a small number of female relatives can prevent accurate assessment of risk.9

In this study, we focused on another BRCA function, controlling centrosome duplication;10, 11, 12 it is known that BRCA plays important roles in the DNA repair pathway. Cells normally have either one or two centrosomes. Suppression of BRCA1 or BRCA2 causes centrosome amplification.13, 14 We therefore speculated that an increase in the number of centrosomes may indicate BRCA mutations in breast cancer specimens. γ‐Tubulin, a centrosome component, could not be detected as foci by immunohistochemistry using 3,3‐diaminobenzidine (DAB),15 whereas immunofluorescence of γ‐tubulin was detected as foci in mammalian cells12 and human breast tissue.16, 17 We used this latter approach in the present study to determine whether the number of γ‐tubulin foci could predict BRCA status in clinical samples.

MATERIALS AND METHODS

Patients

From 2001 to 2014, 68 female Japanese breast cancer patients (including two patients with bilateral breast cancer) underwent breast cancer surgery and genetic testing for BRCA mutations at Hoshi General Hospital (Fukushima, Japan) and Ishinomaki Red Cross Hospital (Ishinomaki, Japan). In both hospitals, patients who met the criteria for BRCA1/2 testing according to the National Comprehensive Cancer Network (NCCN) guidelines were recommended to undergo genetic testing; however, the testing was not covered by Japanese national health insurance. One patient who did not meet the NCCN guidelines underwent genetic testing because of her desire to be tested. All participants were interviewed by experienced genetic counselors to determine the personal and family history of cancer (at least first‐ and second‐degree relatives). The study protocol was approved by the institutional review board at each institution and at Tohoku University Graduate School of Medicine (Sendai, Japan).

Mutation detection

Genomic DNA samples from study subjects at Hoshi General Hospital and Ishinomaki Red Cross Hospital were analyzed at Myriad Genetic Laboratories (Salt Lake City, UT, USA). Full sequencing analysis was carried out for probands, and single‐site testing for the family‐specific mutations (seven patients) was undertaken for relatives of mutation‐positive probands.

Estimation of BRCA1/2 mutation probability using available prediction models, BRCAPRO, Myriad Tables, and KOHCal

Our analysis of the predictive value of BRCA status compared to pretests was restricted to patients who did not have a family history of deleterious BRCA mutations, as seven patients who had such a history were likely to have undergone BRCA genetic testing. One patient was excluded from the analysis because of sufficient familial history.

The BRCAPRO model calculated the probability of a BRCA1 and/or BRCA2 mutation from patients’ personal and first‐ and second‐degree relatives’ history of breast and ovarian cancers.5, 6 We used the version implemented in the BayesMendel 2.1‐2 package of R statistical software (R Foundation, Vienna, Austria). The Myriad prevalence tables provided a probability of detecting BRCA mutations and are based on observations of deleterious mutations in Myriad Genetics Laboratories databases of clinical testing services (http://d1izdzz43r5o67.cloudfront.net/brac/brca-prevalence-tables.pdf).7 KOHCal was constructed using a logistic regression model based on the Korean Hereditary Breast Cancer study.8

Immunodetection of γ‐tubulin

Experiments were carried out at the Pathology Department of Tohoku University Hospital on unstained tissue specimens (4‐μm thickness) mounted on slides, which were provided by Hoshi General Hospital and Ishinomaki Red Cross Hospital. Antigen retrieval was carried out by heating the slides in an autoclave at 120°C for 5 minutes in citric acid buffer (2 mmol/L citric acid and 9 mmol/L trisodium citrate dehydrate, pH 6.0). After treatment with Block Ace (Dainippon Sumitomo Pharmaceutical, Osaka, Japan) for 30 minutes at room temperature, samples were incubated with an antibody against γ‐tubulin clone GTU‐88 (1:600; Sigma, St Louis, MO, USA) followed by Cy‐3‐conjugated Affinipure F(ab’)2 fragment donkey anti‐mouse IgG (H + L) (1:100; Jackson Immunoresearch, West Grove, PA, USA) for immunofluorescence or biotinylated rabbit anti‐mouse antibody (1:100; Nichirei Bioscience, Tokyo, Japan) for DAB staining for 1 hour at room temperature. Nuclei were stained with DAPI (Vector Laboratories, Burlingame, CA, USA).

Microscopy

The number of γ‐tubulin foci per cell was counted using a BZ9000 all‐in‐one fluorescence microscope (Keyence, Osaka, Japan) with an integrated camera and BZ‐Analyzer software (Keyence). Foci were counted with Hybrid Cell Counter software in nine to 12 visual fields at 100× magnification. γ‐Tubulin foci were captured from 10 lower to higher depth fields with a 0.4‐μm pitch for Z‐axis scanning. Composite images were generated with the image‐joining function of the software and Z‐axis data were extracted at a single focal point from multiplane images to obtain a fully focused image using the quick full‐focusing function of the software.

Quantification of γ‐tubulin foci per cell

Non‐cancer cells were excluded from the analysis by serial H&E staining, and cancer cells were counted based on DAPI staining by the cell (hybrid cell count) separation method (Keyence) with manual correction. γ‐Tubulin foci were counted using color extraction and size selection (0.2‐20.0 μm2). The total number of γ‐tubulin foci divided by DAPI‐positive cells in fully focused composite images is expressed as foci/cell.

Statistical analysis

The relationship between clinicopathological parameters and BRCA status was analyzed with Fisher's exact test. The Kruskal–Wallis test was used to evaluate the relationship between family history of breast or ovarian cancer and BRCA status. The predictive value of foci per cell and pretests for BRCA status was analyzed with the Wilcoxon test. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were estimated. To determine optimal thresholds, the point on the ROC curve with a maximum Youden index (sensitivity−[1−specificity]) and the point with the shortest distance value from the 0.1 point of the ROC curve were calculated. Statistical analyses and AUC estimates were carried out using JMP pro11 software (SAS Institute, Cary, NC, USA). Statistical significance was defined as P < .05.

RESULTS

Patient characteristics and BRCA mutation status

Patient characteristics by BRCA status are summarized in Table 1. Of 68 patients, including two with bilateral breast cancer, seven (10.3%) had BRCA1 mutations, 17 (25.0%) had BRCA2 mutations, and one (1.5%) had mutations in both genes (Table S1). Only one patient had a normal variant, BRCA1 G275D, which was classified as wild‐type BRCA as in previous reports,18, 19 and two BRCA1 intronic mutations (one patient with bilateral breast cancer: exon 9‐62, 1 bp deletion). Variant of unknown significance was not observed. All of the patients with suspected BRCA‐related cancer underwent BRCA genetic testing without coverage by national health insurance as a matter of clinical practice; therefore, nearly all patients (96.6%) met BRCA1/2 testing criteria of NCCN guidelines (https://www.nccn.org/professionals/physician_gls/pdf/genetics_screening.pdf). Among these patients, familial history of breast or ovarian cancer was the only statistically significant factor related to BRCA mutation status (P < .001).

Table 1

Patient characteristics by BRCA status

	BRCA status						P‐value
	Non‐carriersa(n = 45)		BRCA1 mutation carriersb (n = 7)		BRCA2 mutation carriersb (n = 17)
	No. of patients	%c	No. of patients	%c	No. of patients	%c
Age at diagnosis, years	48.7		42.8		52.3		.4200
NCCN HBOC testing criteria
Met	42	93.3	7	100.0	17	100.0
Unmet	3	6.7	0	0.0	0	0.0
BRCA sequencing
Full sequencing	44	97.8	6	85.7	12	70.6
Single‐site	1	2.2	1	14.3	5	29.4
Pathological type
Invasive ductal carcinoma	38	84.5	6	87.5	13	76.5
Invasive lobular carcinoma	1	2.2	0	0	0	0
Ductal carcinoma in situ	6	13.3	1	14.3	4	23.5
Invasive size, mm	22.8		38.3		22.3		.0600
LN status
Negative	27	61.4	4	57.1	8	47.1	.4400
Positive	17	38.6	3	42.9	9	52.9
Stage
0	6	13.3	1	14.3	4	23.5	.6600
1	15	33.3	1	14.3	1	5.9
2	15	33.3	4	57.1	8	47.1
3	8	17.8	0	0.0	3	17.6
4	1	2.0	1	14.3	1	5.9
ER
Negative	17	40.5	5	71.4	3	17.6	.1900
Positive	25	59.5	2	28.6	14	82.4
Unknown	3		0		0
PR
Negative	20	47.6	5	71.4	10	58.8	.5100
Positive	22	52.4	2	28.6	7	41.2
Unknown	3		0		0
HER2/neu
Negative	22	81.5	5	71.4	14	87.5	.9800
Positive	5	18.5	2	28.6	2	12.5
Unknown	18		0		1
ER and HER2 status
ER+, HER2−	14	56.0	3	42.9	11	68.8
ER+, HER2+	0	0.0	0	0.0	2	12.5
ER−, HER2+	2	8.0	2	28.6	0	0.0
ER−, HER2−	9	36.0	2	28.6	3	18.7
Unknown	20		0		1
Familial history of breast or ovarian cancer
None	4	8.9	2	28.6	0	0.0	<.0001
1	17	37.8	0	0.0	0	0.0
2	17	37.8	1	14.3	1	5.9
≥3	7	15.5	4	57.1	16	94.1

ER, estrogen receptor; HBOC, hereditary breast and ovarian cancer; HER2, human epidermal growth factor receptor 2; PR, progesterone receptor; LN, lymph node; NCCN, National Comprehensive Cancer Network.

Including one normal variant (BRCA1 G275D) and two intronic mutation patients.

Numbers of BRCA1 and BRCA2 mutation carriers included one patient with both mutations.

Calculated by excluding the number of unknown cases.

Increased number of γ‐tubulin foci with deleterious mutation in BRCA

Representative images of γ‐tubulin foci in wild‐type and mutant BRCA1/2 cases are shown in Figure 1. γ‐Tubulin foci could not be detected by DAB immunohistochemistry (Figure 1B,G,L), but were observed by immunofluorescence (Figure 1C,H,M). The magnified views of Figure 1(C,H,M) are shown in Figure 1(D,I,N) and the views for counting using a hybrid cell counter are shown in Figure 1(E,J,O), respectively. All of the areas of the foci/cell number were higher for BRCA1 L63X (foci/cell = 1.10, Figure 1H) and BRCA2 2423del4 (foci/cell = 6.52; Figure 1M) compared to wild‐type BRCA (foci/cell = 0.53; Figure 1C).

Figure 1

Representative cases for detection of breast cancer susceptibility gene by immunofluorescence analysis. (A‐C) Wild‐type breast cancer tissue sample with H&E staining (A), 3,3‐diaminobenzidine immunostaining of γ‐tubulin (B), and immunofluorescence (C). (D) Enlarged view of area in the yellow square in (C). (E) Numbers of γ‐tubulin foci (red) and DAPI‐stained cells (blue) were counted using a hybrid cell counter. (F‐J) L63X and (K‐O) 2423del4 breast cancer specimens with the same staining and views as those shown in (A‐E)

As shown in Figure 2, the number of γ‐tubulin foci/cell was higher in cases with BRCA1 and/or BRCA2 mutation than in wild‐type BRCA1 and BRCA2 (1.9 [95% confidence interval (CI), 1.5‐2.3] vs 0.5 [95% CI, 0.2‐0.8]) (P < .001). The number of foci/cell in cases of BRCA1 and BRCA2 mutations was 1.2 (95% CI, 0.6‐1.8) and 2.2 (95% CI, 1.7‐2.6), respectively, which was higher than that for wild‐type BRCA1 and BRCA2 (P = .042 and P < .0001, respectively).

Figure 2

Scatter plot of foci/cell by status of breast cancer susceptibility gene in breast cancer tissue samples. The number of γ‐tubulin foci stained per cell was counted using Hybrid Cell Counter software. CI, confidence interval SE, standard error

As shown in Figure 3, the number of γ‐tubulin foci/cell was higher in cases with BRCA1 and/or BRCA2 deleterious mutations than in cases with BRCA1 and BRCA2 normal variant mutations (2.0 [95% CI, 1.4‐2.7] vs 0.3 [95% CI, 0.02‐0.55]) (P = .049).

Figure 3

Scatter plot of foci/cell by type of mutation of breast cancer susceptibility gene. The number of γ‐tubulin foci stained per cell was counted using Hybrid Cell Counter software. CI, confidence interval SE, standard error

Comparison of pretests for BRCA status: BRCAPRO, Myriad Tables, KOHCal, and γ‐tubulin immunofluorescence

γ‐Tubulin immunofluorescence and existing pretests were compared in terms of their predictive value for BRCA mutation status. We analyzed 60 patients who did not have a family history of deleterious BRCA mutations. γ‐Tubulin immunofluorescence had superior predictive value (P < .0001), followed by BRCAPRO (P = .0073) (Table 2).

Table 2

Predictive value of available pretests and immunofluorescence of γ‐tubulin for BRCA status

	BRCA status		P‐value
	Non‐carriersa (n = 44)	BRCA1/2 mutation carriers (n = 16)
γ‐Tubulin foci/cell
Mean (95% CI)	0.6 (0.4‐0.9)	1.7 (1.2‐2.2)	<.0001
Standard error	0.13	0.21
BRACAPRO
Mean (95% CI)	12.8 (5.6‐20.0)	42.5 (30.5‐54.5)	.0073
Standard error	3.6	6
Myriad table
Mean (95% CI)	10.3 (7.6‐13.0)	16 (11.6‐20.4)	.1167
Standard error	1.3	2.2
KOHcal
Mean (95% CI)	29.9 (3.1‐23.7)	42.5 (5.1‐32.2)	.0984
SE	3.1	5.2

CI, confidence interval; KOHcal, KOHBRA (The Korean Hereditary Breast Cancer) BRCA risk calculator.

Including one normal variant (BRCA1 G275D) and two intronic mutation patients.

We next analyzed ROC curves to determine the appropriate cut‐off point for predicting BRCA mutation status (Figure 4). γ‐Tubulin immunofluorescence had the highest AUC (0.86; Figure 4D). The AUC (0.97; Figure 4F) of the BRCA2 mutation was higher than that of the BRCA1 mutation (0.79; Figure 4E). The cut‐off value for γ‐tubulin immunofluorescence in BRCA1/2 mutations was 0.95 according to the maximum Youden index, with a sensitivity of 83% (19/23), specificity of 89% (42/47), and positive predictive value of 77% (20/26).

Figure 4

Receiver operating characteristic (ROC) curves of γ‐tubulin immunodetection and existing pretests for predicting mutation status in breast cancer tissue samples. BRCAPRO (A), Myriad Table (B), The Korean Hereditary Breast Cancer risk calculator (KOHCal) (C), and γ‐tubulin foci/cell (E) were evaluated using ROC curves for mutations. γ‐Tubulin foci/cell for (E) and (F) were evaluated. The cut‐off point was determined by the maximum Youden index. AUC, area under the ROC curve

DISCUSSION

The ability to identify BRCA‐related cancer patients among many breast cancer patients is important, because these patients are suitable for targeted therapy and prophylactic surgery as well as intensive follow‐up to prevent breast cancer‐related death.4, 20, 21, 22, 23 γ‐Tubulin immunodetection could be a useful pretest for patients who are considering BRCA genetic testing, as it is inexpensive and convenient and can be undertaken at any general hospital. The AUC (0.86) of γ‐tubulin immunofluorescence for predicting BRCA mutation status was superior to that of BRCAPRO (0.69), Myriad Tables (0.61), and KOHCal (0.65); these tests were previously reported to be more accurate with AUCs of 0.76, 0.71‐0.72, and 0.76, respectively.9, 24, 25 These pretests that use clinicopathological information might be more affected by ethnic differences than our method.26

The immunohistochemical analysis of breast cancer specimens, especially of estrogen receptor and human epidermal growth factor receptor 2 expression, is important for determining the appropriate treatment by targeted therapy. These overexpressed proteins are easy to evaluate by immunohistochemistry; on the contrary, evaluation of tumor suppressor proteins such as p53 and BRCA is challenging. We previously focused on protein function to determine gene status by immunohistochemistry, and have reported combined immunohistochemical detection of p53 and its downstream proteins for predicting TP53 mutation.27 Shimomura et al reported that loss of BRCA1 expression was associated with an increase in the number of centrosomes in breast cancer specimens, as determined by immunofluorescence analysis;17 they also found that centrosome numbers were higher in three breast cancer patients with BRCA1 mutation compared to those with normal BRCA1 expression. Martins et al reported that the centrosome number was higher in normal and cancerous breast tissue from BRCA1 mutation carriers than in non‐carriers.16 However, there have been no studies evaluating the predictive value of centrosome numbers for BRCA mutation status using immunofluorescence analysis.

BRCA genetic testing is rapidly spreading worldwide because of the low cost of next‐generation sequencing, availability of targeted therapy for BRCA‐related cancer, and BRCA patent invalidity. However, many problems exist with the genetic testing, especially regarding the evaluation of VUS mutations. The Breast Cancer Information Core (https://research.nhgri.nih.gov/bic/), the ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), and the Leiden Open Variation Database – International Agency for Research on Cancer (http://www.lovd.nl/2.0/index_list.php) are open access public databases for genome mutations. However, the databases still contain many VUS mutations. For this reason, a functional assessment of BRCA protein is needed for assisting the judgment of a VUS as a deleterious or normal variant mutation. In this study, three cases with BRCA normal variant mutations did not have increased centrosome numbers, the same as cases without BRCA mutations. However, further study is needed for this issue.

Tumors with the molecular features of BRCA mutation are known as having “BRCAness”.28 BRCAness is not only the inactivation of a protein related to DNA repair pathways, but also methylation of BRCA1 or BRCA2.29 The Cancer Genome Atlas study showed 60% of basal‐like breast cancers involved BRCA inactivation, including methylation and mutation.2 BRCAness evaluation is important for determining individualized treatment regimens that can include PARP inhibitors or DNA cross‐linking agents. Several studies have attempted to identify BRCAness using transcriptional biomarkers30, 31 or by assaying homologous recombination.32, 33 The homologous recombination deficiency score is highly correlated with defects in BRCA1/2 and is associated with the response to platinum therapy,33 but this scoring system is complex. In this study, some cases of increased centrosome number in BRCA wild‐type individuals (23% of false‐positive results) might be considered to represent BRCAness.

Centrosome amplification often occurs in human cancers, especially in breast cancer.34 Triple negative breast cancers strongly correlate with centrosomal amplification using immunofluorescence analysis of γ‐tubulin, the same as our method.35, 36 Based on their biological functions, both BRCA1 and BRCA2 play important roles in the proper control of centrosome duplication: BRCA1 is a ubiquitin ligase of γ‐tubulin that regulates centrosome number;13 and BRCA2 has a centrosomal localization signal that functions in centrosome positioning, amplification, and cohesion.37, 38, 39 Numerous studies have revealed that inhibition of BRCA function causes centrosome aberrations in human breast cell lines, mice, and human breast tissue.13, 17, 40 However, many proteins function to control centrosome duplications.41 Therefore, the 23% false‐positive rate in our study may involve disruption of other centrosomal proteins.

There is a debate as to whether all deleterious mutations of BRCA1/2 genes equally disrupt both functions: controlling centrosome duplication and homologous recombination.42 We and Cochran et al have reported that most BRCA deleterious mutations abrogate centrosome duplication and are well correlated with DNA repair function (Table S2).8, 10, 43 However, the 17% false‐negative rate in our study might retain controlling centrosome duplication with BRCA mutation.

Our study has two limitations. First, although the foci/cell were counted using an objective evaluation method and two previous reports support an increased number of specimens with BRCA mutations among breast cancer specimens,16, 17 our findings need to be validated by additional data. Second, centrosomes are regulated by multiple factors, including BRCA1/2.11 We need to investigate the other factors that contribute to increased centrosome number with wild‐type BRCA.

In conclusion, an increased number of centrosomes can be used to predict the presence of BRCA mutation, thus potentially assisting patients in deciding whether to pursue genetic testing.

CONFLICT OF INTEREST

The authors have no conflict of interest.

Click here for additional data file.

Click here for additional data file.