Human Pathology (2014) xx, xxx–xxx

www.elsevier.com/locate/humpath

Original contribution

Spindle assembly checkpoint protein expression correlates with cellular proliferation and shorter time to recurrence in ovarian cancer☆,☆☆ Barbara McGrogan BSc, MSc, PhD, FIBMS a,b,d,⁎,1 , Sine Phelan MD, FRCPath a,b,c,1 , Patricia Fitzpatrick MD, MPH, FFPHMI, FRCPI e , Aoife Maguire MB, BCh, BAO, MRCPI, FRCPath a , Maria Prencipe BSc, PhD a,b , Donal Brennan MD b,f , Emma Doyle MD g , Anthony O'Grady MSc, PhD h , Elaine Kay MBBCHBAO, FRCSI, FRCPATH, MD h , Fiona Furlong BSc, MPH, PhD i , Amanda McCann BA Mod (Micro), PhD a,b a

UCD School of Medicine and Medical Science, University College Dublin, UCD, Belfield, Dublin 4, Ireland UCD Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Belfield, Dublin 4, Ireland c Department of Histopathology, University Hospital Galway, Co. Galway, Ireland d National Cancer Control Programme, King's Inns House, Dublin 1, Ireland e UCD School of Public Health, Physiotherapy & Population Science, University College Dublin, Belfield, Dublin 4, Ireland f National Maternity Hospital, Dublin 2, Ireland g Rotunda Hospital, Dublin 1, Ireland h Royal College of Surgeons in Ireland (RCSI) and Department of Histopathology, Beaumont Hospital, Dublin 9, Ireland i School of Pharmacy, Queens University Belfast, Belfast, BT97BL, Northern Ireland, UK b

Received 9 November 2013; revised 4 March 2014; accepted 12 March 2014

Keywords: Ovarian cancer; Mitotic arrest deficient 2 (MAD2); BUB1-related protein kinase (BUBR1); Spindle assembly checkpoint (SAC); Ki-67; Recurrence-free survival (RFS); Chemotherapeutic agents

Summary Ovarian carcinoma (OC) is the most lethal of the gynecological malignancies, often presenting at an advanced stage. Treatment is hampered by high levels of drug resistance. The taxanes are microtubule stabilizing agents, used as first-line agents in the treatment of OC that exert their apoptotic effects through the spindle assembly checkpoint. BUB1-related protein kinase (BUBR1) and mitotic arrest deficient 2 (MAD2), essential spindle assembly checkpoint components, play a key role in response to taxanes. BUBR1, MAD2, and Ki-67 were assessed on an OC tissue microarray platform representing 72 OC tumors of varying histologic subtypes. Sixty-one of these patients received paclitaxel and platinum agents combined; 11 received platinum alone. Overall survival was available for all 72 patients, whereas recurrence-free survival (RFS) was available for 66 patients. Increased BUBR1 expression was seen in serous carcinomas, compared with other histologies (P = .03). Increased BUBR1 was significantly associated with tumors of advanced stage (P = .05). Increased MAD2 and BUBR1 expression also correlated with increased cellular proliferation (P b .0002 and P = .02, respectively).

☆

Competing interests: The authors confirm that there is no conflict of interest in this submission. Funding/Support: The authors would like to acknowledge the Pathological Society of Great Britain and Ireland (Small grant scheme) and Cancer Research Ireland (CRF08FUR) for the funding of this project. ⁎ Corresponding author. Research Fellow, UCD Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Belfield, Dublin 4, Ireland. E-mail address: [email protected] (B. Mc Grogan). 1 Joint first authors. ☆☆

http://dx.doi.org/10.1016/j.humpath.2014.03.004 0046-8177/© 2014 Elsevier Inc. All rights reserved.

2

B. McGrogan et al. Reduced MAD2 nuclear intensity was associated with a shorter RFS (P = .03), in ovarian tumors of differing histologic subtype (n = 66). In this subgroup, for those women who received paclitaxel and platinum agents combined (n = 57), reduced MAD2 intensity also identified women with a shorter RFS (P b .007). For the entire cohort of patients, irrespective of histologic subtype or treatment, MAD2 nuclear intensity retained independent significance in a multivariate model, with tumors showing reduced nuclear MAD2 intensity identifying patients with a poorer RFS (P = .05). © 2014 Elsevier Inc. All rights reserved.

1. Introduction Ovarian carcinoma (OC) is the most lethal of the gynecological malignancies, often presenting at an advanced stage, with intraperitoneal dissemination and distant metastases. Treatment of advanced OC is hampered by eventual tumor resistance to chemotherapy. Up to 30% of patients do not respond to conventional chemotherapy, and a high proportion of initial responders (22%-59%) develop acquired drug resistance [1]. Most deaths result from metastases, which are refractory to conventional chemotherapy. Paclitaxel (Taxol) is a first-line chemotherapeutic agent effective in the treatment of OC and is used in combination with platinum agents, following optimal surgical debulking [2]. Paclitaxel is a microtubule stabilizing agent that functions primarily by interfering with spindle microtubule dynamics causing cell cycle arrest and apoptosis [3]. The mechanisms underlying its action have not, however, been fully elucidated [4]. The spindle assembly checkpoint (SAC) is a regulatory mechanism present in all eukaryotes, which prevents chromosome mis-segregation during mitosis, thereby preventing aneuploidy [5]. From a chemotherapeutic perspective, it is the checkpoint through which microtubule inhibitory drugs, such as paclitaxel, exert their effects, and reduced expression of SAC proteins such as BUB1-related protein kinase (BUBR1) and mitotic arrest deficient 2 (MAD2) are associated with acquired paclitaxel resistance in ovarian carcinoma cell lines [6,7]. The components of the SAC were first described in budding yeast and include mitotic arrest deficient proteins 1 to 3 (MAD1, MAD2, and MAD3, the latter now known as BUBR1 in humans) and “budding uninhibited by benzimidazole proteins,” BUB1-3 [8-11]. A functional SAC is activated following treatment with antimicrotubule agents, such as paclitaxel due to interference with spindle assembly [12]. In this situation, a functioning SAC can impede cell cycle progression and promote apoptosis. In vitro studies have shown that alterations in certain SAC proteins compromise this function and thereby confer resistance to drugs such as paclitaxel. Specifically, BUBR1 is required for a sustained mitotic arrest in human cancer cells treated with microtubule targeting agents [13]. Importantly, small interfering RNA (siRNA)–mediated suppression of MAD2 and BUBR1 expression in paclitaxel-treated cancer cells abolishes checkpoint function and results in paclitaxel resistance in vitro [14].

BUBR1 exists in 2 forms within the cell—kinetochorebound insoluble BUBR1, reported to be required for checkpoint function, and a putatively nonessential soluble form [15]. Previous immunohistochemical (IHC) analyses have reported cytoplasmic BUBR1 expression in a variety of normal tissues such as skin, colon, and testis [16] and nuclear expression in malignancies such as squamous carcinoma and adenocarcinoma of the pancreas [16]. Cytoplasmic BUBR1 has also been reported in urothelial cancers [17]. Positive cytoplasmic BUBR1 expression has been associated with reduced recurrence-free survival (RFS), advanced stage, high grade, and serous histology in OC [18]. Furthermore, our group has recently identified the independent association of decreased MAD2 intensity with reduced progression free survival in ovarian carcinomas of the papillary serous histologic subtype [19]. IHC localization of MAD2 has demonstrated nuclear expression in some squamous cancers and in adenocarcinoma of the pancreas and colon cancer [16]. A study of hepatocellular carcinoma demonstrated nuclear and cytoplasmic expression of MAD2 [20]. From other data, MAD2 localizes to the cytoplasm in normal gastric epithelium but to the nucleus in gastric carcinoma [21]. This is in contrast to testicular tissue, where MAD2 is overexpressed in the cytoplasm in seminoma but is more commonly localized to the nucleus in normal testicular parenchyma [22]. In relation to cellular proliferation, overexpression of BUBR1 in bladder cancer has been associated with high Ki67 expression [17]. In addition, overexpression of spindle checkpoint proteins, including MAD2, has been associated with increased Ki-67 expression in colorectal mucosa [16]. In this study, we performed an IHC analysis for BUBR1, MAD2, and Ki-67 on a tissue microarray (TMA) constructed from a cohort of 72 OC patient tumor samples prospectively collected and displaying a variety of histologic subtypes including papillary serous, endometrioid, and clear cell carcinomas. One case had insufficient material for the quantification of MAD2. Overall, our objectives were as follows: (a) specifically characterize and quantify the immunolocalization patterns of key SAC proteins MAD2, BUBR1, and the proliferation marker Ki-67, in different subtypes of OC using automated (Aperio Technologies, Vista, CA) and manual assessment; (b) use the weighted κ measure of agreement to compare manual and automated scores to validate the automated scoring outputs; (c) investigate the relationship between the automated scoring

Q5

SAC protein expression in OC recurrence

3

outputs and tumor proliferation as determined by Ki-67; (d) associate the IHC expression of MAD2, BUBR1, and Ki-67 as determined by automated image analysis with clinicopathological variables and patient outcome for women undergoing surgical resection and systemic chemotherapy for OC.

2. Materials and methods 2.1. Patient samples and histopathologic data The patients had all undergone surgical resection of an ovarian carcinoma between 1988 and 2002. A total of 72 cases were studied, including 37 (51.4%) serous, 11 (15.3%) endometrioid, 1 (1.4%) clear cell, 2 (2.7%) transitional, 3 (4.2%) mucinous, and 18 (25%) mixed carcinomas. This ovarian cancer cohort is a previously published cohort of sequential patients [23] treated at the National Maternity Hospital, Dublin, Ireland. All tissues were formalin-fixed and paraffin-embedded (FFPE). Median age of patients at the time of diagnosis was 52 years (range, 32-77 years). All clinicopathological variables are outlined in Table 1. All tumors were graded Table 1 Clinicopathological characteristics of primary invasive OCs (n = 72) Patient and tumor characteristics

Total OC cohort

No. of case Age (y), median (range) Histologic subtype Serous Endometrioid Transitional Clear cell Mucinous Mixed Histologic grade Well differentiated Moderately differentiated Poorly differentiated Stage I II III IV Outcome Alive Dead Disease-free postchemotherapy Responder Nonresponder Residual disease No ≤2 cm N2 cm

72 52 (32-77) 37 (51.4%) 11 (15.3%) 2 (2.7%) 1 (1.4%) 3 (4.2%) 18 (25%) 11 (15.3%) 27 (37.5%) 34 (47.2%) 0 20 (27.8%) 51 (70.8%) 1 (1.4%) 28 (38.9%) 44 (61.1%) 56 (77.8%) 16 (22.2%) 7 (14.6%) 15 (31.2%) 26 (54.2%)

by a single pathologist (E. D.) using the universal grading system, also known as the Shimizu-Silverberg system [24], and staged using the International Federation of Obstetrics and Gynaecology system. Clinical information was available including age, systemic chemotherapy (platinum-based therapy alone or paclitaxel and platinum), and optimal debulking categorized as no residual disease, 2 cm or less residual disease, and greater than 2 cm residual disease. Information on optimal debulking was available for 48 of the total cohort of 72 patients. Of the 72 patients, 61 were adjuvantly treated with a combined platinum-paclitaxel regimen, and the remaining 11 were treated with platinum only. None of the patient cohort received neoadjuvant chemotherapy. Overall survival data were available for all 72 patients. Data on RFS were available for 66 of the 72 patients (with 57 patients treated with combined platinum-paclitaxel and the remaining 9 were treated with platinum only). The primary end point was overall survival as measured from 6 weeks postchemotherapy to the time of death. RFS rates are measured from 6 weeks postchemotherapy to the date of recurrence. Patients with 6 months or greater disease-free interval were classified as responders, whereas those whose disease recurred within 6 months were classified as nonresponders, in line with other studies in the area [25] and in line with local practice at the National Maternity Hospital, Holles St, Dublin, Ireland. A TMA was constructed using a manual tissue arrayer (MTA-1, Beecher, WI), from FFPE tissue retrieved from the pathology archives of the National Maternity Hospital, Holles St, Dublin, Ireland, following approval by the National Maternity Hospital Ethics Committee and represents a previously published cohort of patient samples [23,26]. The array contained 4 cores per patient, 2 × 1.0 mm cores having been taken from each of 2 donor blocks and assembled into the recipient block. In general, cores were taken from the peripheral part of the tumor in cases where the tumor had well-defined borders. In more diffusely growing tumors, areas with the highest tumor cell density were primarily targeted, avoiding necrotic tissue.

2.2. IHC staining methodology IHC staining for BUBR1 and Ki-67 was performed on a fully automated platform (Ventana Systems, Tucson, AZ) applied to 5-μm sections, following standard optimization steps. BUBR1 staining was performed using a mouse monoclonal antibody (1:300 dilution, EDTA pH 8 antigen retrieval; Abcam, Cambridge, UK), with an invasive ductal carcinoma of the breast used as a positive control. Ki-67 staining was carried out using a mouse monoclonal antihuman MIB-1 clone (1:100 dilution, EDTA pH 8 antigen retrieval; DakoCytomation, Glostrup, Denmark), and a section of normal appendix was used as the positive control. In the case of MAD2, immunostaining was performed using a mouse monoclonal antibody (1:100 dilution, EDTA antigen retrieval;

4 BD Biosciences, San Diego, CA), on an automated platform (Bond system; Leica MicroSystems, Newcastle, UK), with normal testis serving as the positive control.

2.3. Manual quantitative IHC of MAD2, Ki-67, and BUBR1 The manual evaluation of MAD2, Ki-67, and BUBR1 was performed using a standard light microscope on highpower field (×200) magnification. Staining results for each antibody were interpreted by 2 independent observers (B. M. and A. M.) and (B. M. and S. P.) and scored for the presence

B. McGrogan et al. of nuclear or cytoplasmic immunolocalization as described below for each antibody. Discordant cases were reviewed, and consensus was reached before statistical analysis. All scores from multiple cores were averaged. MAD2 and Ki-67 immunostaining was manually quantified by 2 independent observers and scored for the presence of nuclear immunolocalization (B. M. and A. M.), respectively. Manual scoring of nuclear MAD2 and Ki-67 IHC was similar to a previously published scoring method combining intensity and percentage IHC distribution of positively stained cells [27]. Briefly, immunostaining intensity was scored as negative, weak (1+), moderate (2+), and strong (3+). The

Fig. 1 Aperio automated imaging of the ovarian cancer TMA. Immunostaining for MAD2 (A), BUBR1 (B), p53 (C), and Ki-67 (D) in a moderately differentiated serous carcinoma. Ovarian TMA core (objective magnification ×5) (i), annotated region (×10) (ii), markup images using the IHC Nuclear algorithm (version 8.001) and the Colour Deconvolution algorithm (version 9) (×10) (iii). The markup images of the Nuclear IHC algorithm (version 8.001) show negative nuclei in blue and positively stained nuclei in yellow (weak 1+), orange (moderate 2+), and red (strong 3+). The markup images of the Colour Deconvolution algorithm (version 9) show negative areas in blue and positively stained areas in yellow (weak 1+), orange (moderate 2+), and red (strong 3+).

Q5

SAC protein expression in OC recurrence percentage of positive tumor cell IHC distribution with nuclear localization of the antigens was determined on each core and assigned the following values: negative (cutoff threshold for MAD2 b10%, Ki-67 [b5%] [27]), 33% or less (1+), 34% to 66% (2+), and greater than 66% (3+). For each core, the total (T) score reflects the combined percentage distribution and intensity scores. For example, a value of 3 for percentage IHC distribution and 2+ for intensity score equals T score of 5/6 (strong expression). Using this approach, T score was categorized as negative, weak expression (score ≤2/6); moderate expression (score N2 b 4/6); strong expression (score ≥4/6). The cores for each case were averaged to produce a final T score. BUBR1 immunostaining was manually quantified by 2 independent observers and scored for the presence of cytoplasmic immunolocalization (B. M. and S. P.), respectively. Manual scoring of cytoplasmic BUBR1 was also based on a combination of intensity and percentage IHC distribution of positively stained tumor cells as previously described [28]. Similar to MAD2, 10% was used as the cutoff threshold for scoring of cytoplasmic BUBR1 [29].

2.4. Automated quantitative IHC of MAD2 and Ki-67 Immunostained slides were scanned at objective magnification ×20 onto Spectrum using the Aperio ScanScope XT Slide Scanner. Tumor areas on the digital images were annotated to exclude stromal tissue. The IHC Nuclear (version 8.001) algorithm (Aperio Technologies) was used for the automated analysis of the nuclear proteins MAD2 and Ki-67. The input parameters of the algorithm were adjusted to account for variations in staining intensity and features of cell morphology including cell size and shape. A pseudocolor markup image was generated for each nuclear protein (Fig. 1), allowing confirmation of the reliability of the algorithm's result. The poor quality of 1 set of the ovarian tumor cores resulted in its exclusion from subsequent MAD2 analyses. This algorithm generated the following outputs: The percentage distribution positively stained nuclei as a continuous variable—this value was then specifically subcategorized as 0, less than 33% (1+), 34% to 66% (2+), greater than 66% (3+); the average nuclear staining intensity—this output generates continuous variable values ranging from 0 (black) to 255 (white). These values will vary depending on the particular stain, and therefore, the algorithm is trained for each individual stain giving thresholds of negative (0), weak (1+), moderate (2+), and strong (3+). The T score reflects the combined percentage distribution and intensity outputs. The maximum value for this will therefore be 6. For example, 3+ percentage distribution and 3+ intensity equals 6, whereas 2+ percentage distribution and 3+ intensity equals 5. Using this approach, total automated staining was subsequently categorized as negative, weak expression (score ≤2/6); moderate expression (score N2 b 4/6); strong expression (score ≥4/6). These differential outputs are

5 referred to as percentage distribution, intensity, and T score. OC tumors were considered positive for MAD2 if greater than 10% of tumor cells demonstrated nuclear expression as previously described [29]. The cutoff threshold for positive expression of Ki-67 in OC tumors was 5%.

2.5. Quantitative IHC of BUBR1 The Colour Deconvolution (Version 9) algorithm (Aperio Technologies) was used for the automated quantification of cytoplasmic and nuclear BUBR1 expression. The algorithm was applied to the annotated digital image, and a pseudocolor markup image of tumor cell DAB intensity (1+, 2+, and 3+) was generated. The output parameters of this algorithm were percentage distribution of positive staining and intensity variable values of 1+, 2+, and 3+, which lie between the light intensity ranges of 0 (black) to 255 (white). The percentage distribution score was classified as 33% or less (1+), 34% to 66% (2+), or greater than 66% (3+), and the average positive intensity values were classified as negative (0), weak (1+), moderate (2+), or strong (3+). An overall total score ranging from 0 to 6, designated the BUBR1 T score, was generated. This represents a combined nuclear and cytoplasmic score but was, in fact, predominantly cytoplasmic and is referred to as BUBR1 (T). OC tumors were considered positive for BUBR1 if greater than 10% of tumor cells demonstrated cytoplasmic expression as previously described [29].

2.6. Concordance between manual and automated scoring outputs Automated scoring outputs were validated by comparing automated scoring outputs to manual scores using the weighted κ measure of agreement. The κ agreement scores and concordance rates for MAD2 intensity, percentage distribution, and T scores were 0.61 and 69%, 0.61 and 62%, and 0.75 and 77.5%, respectively. This represents substantial agreement between manual and automated scoring outputs for MAD2. The κ agreement scores and concordance rates for Ki-67 intensity, percentage distribution, and T scores were 0.74 and 72.2%, 0.87 and 92.1%, and 0.74 and 73.6%, respectively. This represents substantial agreement between manual and automated scoring outputs Ki-67 intensity and T scores and almost perfect agreement for percentage distribution between manual and automated scoring of Ki-67. The κ agreement scores and concordance rates for BUBR1 intensity, percentage distribution, and T scores were 0.50 and 64.1%, 0.1 and 12.5%, and 0.13 and 19.4%, respectively. This represents moderate agreement between the manual and automated scoring output for BUBR1 intensity. There was slight agreement between the manual and automated scoring outputs for BUBR1 percentage distribution and T score. Therefore, this study only reports on associations for automated BUBR1 using intensity scores.

6

2.7. Statistical methods The χ2 test was used for comparison of proportions. The WINPEP1 statistical package was used to calculate the weighted κ measure of agreement. This is available online and can be accessed at http://www.brixtonhealth.com/. This statistical method compared manual versus automated scoring of IHC markers and compared intensity, percentage distribution and T scores for each marker. Kaplan-Meier (KM) methods were used to construct overall survival and RFS curves, and log-rank tests are presented. Patients who died of other causes or were still alive were censored at the last date of follow-up. Patients who did not recur were also censored at last date of follow-up. The effect of IHC staining outputs on disease-free interval was assessed using multivariate Cox regression analyses to adjust for stage, tumor grade, and optimal debulking. Results were considered to be significant for P values less than .05. The statistical analysis was conducted using SAS version 9 (SAS, Cary, NC).

3. Results 3.1. BUBR1 immunolocalization and association with histologic subtype, stage, and grade Immunohistochemically, BUBR1 was predominantly expressed in the cytoplasm (Fig. 2A). Of the 72 cases, 9 (12.5%) of 72 were negative for BUBR1 intensity, 22 (30.5%) of 72 were weakly positive, 41 (57%) of 72 were moderately positive, and no tumors were strongly positive for BUBR1 intensity. BUBR1-negative tumors were heter-

B. McGrogan et al. ogenous and included serous, endometrioid, and mixed types. There was a significant difference in the proportions of cancer types with increased BUBR1 intensity scores. Specifically, a higher proportion of serous cancers showed increased BUBR1 intensity of expression compared to the endometrioid or other histologies (P = .03) (Table 2). Moreover, serous type tumors were more likely to have an increased BUBR1 intensity compared to endometrioid tumors (P = .01) (Table 2). In addition, OC tumors from stage III or IV patients were significantly more likely to be positive for BUBR1 intensity than stage II patients (P = .05). There were no stage I tumors in this cohort. There was no association between BUBR1 expression and tumor grade or response to treatment. There was no association between BUBR1 intensity and residual disease defined as no residual disease versus 2 cm or less versus greater than 2 cm (Table 2). For 12 (17%) of 72 of our OC tumors, a distinctive punctate nuclear expression of BUBR1 was also seen (Fig. 2B), of which 7 were of the serous type. In terms of tumor grade, 8 of the tumors with this punctate nuclear BUBR1 expression were grade 3, 3 were grade 2, and 1 was grade 1.

3.2. MAD2 and Ki-67 immunolocalization and association with histologic subtype, stage, grade, and residual disease In the case of MAD2, this SAC protein was predominantly expressed in the nucleus, with distinctive perinuclear accentuation (Fig. 2C). One case was insufficient for quantification of MAD2 (n = 71). In total, therefore, 5 (7%) of 71 tumors were negative for MAD2 (T) expression,

Fig. 2 BUBR1 and MAD2 immunostaining patterns in ovarian carcinomas. A, Cytoplasmic BUBR1 in a moderately differentiated transitional carcinoma. B, Punctate nuclear BUBR1 in a moderately differentiated serous carcinoma. C, Perinuclear MAD2 immunostaining in a moderately differentiated clear cell carcinoma. D, Cytoplasmic and nuclear MAD2 immunolocalization in a moderately differentiated serous carcinoma (objective magnification ×40).

Q5

SAC protein expression in OC recurrence

7

Table 2 Association of BUBR1 (n = 72), MAD2 percentage distribution (n = 71), Ki-67 intensity (n = 72) with clinicopathological variables in the ovarian cancer cohort Total number, n

BUBR1 intensity Neg

Pos b

Total number, n

11 27 34

0 4 5

11 23 29

NS

11 27 33

0 6 8

11 21 25

20 52

5 4

15 48

.05 ⁎

20 51

3 11

15 48

NS

15 56

Grade 1 2 3 Stage II III and IV Response to treatment Nonresponder Responder Histologic subtypes Serous Endometrioid Other d

16 56

1 8

37 11 24

1 36 3 8 5 19 BUBR1 intensity Negative/ Moderate weak

Residual disease e None ≤2 cm N2 cm

7 15 26

3 6 11

4 9 15

P

MAD2 percentage distribution a

Total number, n

Ki-67 intensity Neg

Pos c

NS

11 27 34

6 12 9

5 15 25

NS

17 40

NS

20 52

10 17

10 35

NS

4 10

11 46

NS

16 56

5 22

11 34

NS

.03 ⁎

36 11 24

7 1 6

29 10 18

NS

37 11 24

11 5 11

26 6 13

NS

NS

7 15 25

0 2 5

7 13 20

NS

7 15 26

4 5 8

3 10 18

NS

Negative/ weak

P

Moderate/ strong

P

NOTE. When comparing serous versus endometrioid subtypes with BUBR1 intensity categorized as negative versus positive, there was still statistical significance (P = .01). Abbreviation: Neg, negative; Pos, positive; NS, not significant. a The total number of cases assessed for MAD2 IHC are 71 and not 72 due to insufficiency in tumor material for one of the cases. b Weak/moderate (positive). c Weak/moderate/strong (positive). d Other histologic subtypes include transitional, clear cell, mucinous, and mixed. e Information on residual disease was available for 48 of the total number of 72 patients; hence, there were insufficient numbers to statistically analyze BUBR1 intensity categorized as negative versus positive against residual disease; therefore, the categories negative/weak and moderate BUBR1 intensity were compared with residual disease defined as none versus 2 cm or less and greater than 2 cm. ⁎ Significant results, P b .05 using χ2 test.

9 (13%) of 71 showed weak expression, 31 (44%) of 71 showed moderate expression, and 26 (36%) of 71 showed strong expression of nuclear MAD2. Thirty-four percent (24/71) OC tumors also displayed cytoplasmic expression (Fig. 2D). None of the automated MAD2 or Ki-67 outputs associated with any specific OC histologic subtype stage, tumor grade, or residual disease (Table 2).

3.3. Association of IHC outputs with each other BUBR1 and MAD2 IHC staining were significantly associated with each other (P = .007) (Table 3). Specifically, tumors, which had increased expression of BUBR1 intensity, also expressed high levels of MAD2 (T). Increased BUBR1 intensity was significantly associated with increased cellular proliferation as quantified by Ki-67 IHC (P = .02). Increased MAD2 percentage distribution was also strongly associated with increased Ki-67 (T) expression (P b .0002) (Table 3).

3.4. Association of IHC outputs with patient outcome 3.4.1. BUBR1 None of the automated IHC BUBR1 outputs associated with patient outcome. Interestingly, 10 (84%) of 12 of tumors demonstrating punctuate nuclear BUBR1 staining (Fig. 2B) showed a response to chemotherapy, defined as disease-free status at greater than or equal to 6 months postchemotherapy. 3.4.2. MAD2 Univariate KM survival distributions showed that reduced nuclear MAD2 intensity associated with a poorer RFS (P = .03), in ovarian tumors of differing histologic subtype in the 66 patients on whom RFS was available (Fig. 3). The subgroup who received combination paclitaxel and platinum chemotherapy (n = 57) were also analyzed separately to the platinum-only group (n = 9). In this subgroup, reduced

8

B. McGrogan et al. Table 3

Associations between the expression of MAD2, BUBR1, and Ki-67 IHC scores MAD2 (T)

BUBR1 intensity Negative Weak Moderate MAD2 a percentage distribution Negative/weak Moderate Strong

P

Ki-67 (T)

P

Negative

Weak

Moderate

Strong

Negative

Weak

Moderate

Strong

2 2 1

4 3 2

3 9 19

0 8 18

8 9 10

1 8 14

0 5 15

0 0 2

11 14 2

3 7 13

0 10 11

.007 ⁎

.02 ⁎

b.0002

a

The total number of cases assessed for MAD2 IHC are 71 and not 72 due to insufficiency in tumor material for one of the cases. ⁎ Significant results, P b .05 using χ2 test.

MAD2 nuclear intensity associated with a reduced RFS (P = .007). In a multivariate Cox regression analysis adjusting for stage, grade, and optimal debulking (model 1; Table 4), all variables failed to reach significance. When debulking was removed from the model (model 2; Table 4), reduced MAD2 intensity of expression was significantly associated with a poorer RFS, adjusted for stage and grade (P = .05).

4. Discussion The taxanes, paclitaxel (Taxol) in particular, are important first-line chemotherapeutic agents in the treatment of ovarian cancer. Up to 30% of tumors, however, display intrinsic or acquired chemoresistance, and overall 5-year survival rates remain at less than 50% [1]. There is currently no effective method to assist clinicians in predicting the likelihood of chemotherapeutic response. In general terms, IHC is a robust and cost-effective test and represents a possible future source

Fig. 3

of predictive biomarkers of chemoresistance and outcome in advanced OC. BUBR1 and MAD2, essential components of the SAC, play a key role in the cellular response to chemotherapeutic agents, such as taxanes and platinum agents [11-13]. We have demonstrated that both markers are commonly overexpressed in clinical OC samples. Increased BUBR1 intensity expression was associated with the serous phenotype (P = .03) and advanced stage (P = .05) in our study. Increased cytoplasmic BUBR1 IHC expression has been described as an independent prognostic factor in predicting recurrence in ovarian carcinomas [18]. Interestingly, increased cytoplasmic BUBR1 expression also occurred more commonly in serous tumors than in other histologic subtypes in that study [18]. BUBR1 additionally displayed a distinctive punctuate staining in 12 (17%) of 72 of our patient cohort, the majority (10/12, 84%) of which were of the serous subtype, with expression occurring in cell nuclei morphologically in interphase. Nuclear expression was observed in previous studies [16,30].

KM distribution of MAD2 nuclear intensity and RFS. Reduced MAD2 intensity score predicts shorter RFS in ovarian cancer (n = 66).

Q5

SAC protein expression in OC recurrence Table 4

9

Prognostic factors in Cox proportional hazards for models 1 and 2 Variables

Model 1 MAD2; intensity 3 + 4 (high) vs 1 + 2 (low) Tumor grade; 2/3 (high) vs low Tumor stage; 3/4 (high) vs 2 (low) Debulking—residual disease remaining; 2, N2 cm (high) vs 0, no; 1, ≤2 cm (low) Model 2 MAD2; intensity 3 + 4 (high) vs 1 + 2 (low) Tumor grade; 2/3 (high) vs low Tumor stage; 3/4 (high) vs (low)

Hazard ratio Univariate 95% CI

P

n

Hazard ratio Multivariate P 95% CI

n

0.541

0.304-0.961 .04 ⁎ 66 0.507

0.234-1.099 .09

66

1.497 2.259 1.014

0.669-3.350 .33 66 1.199 1.034-4.933 .04 ⁎ 66 1.394 0.501-2.053 .97 48 1.102

0.458-3.141 .71 0.559-3.478 .48 0.526-2.309 .80

66 66 48

0.541 1.497 2.259

0.304-0.961 .04 ⁎ 66 0.555 0.669-3.350 .33 66 1.201 1.034-4.933 .04 ⁎ 66 2.224

0.308-1.00 .05 ⁎ 66 0.526-2.744 .66 66 1.007-4.913 .05 ⁎ 66

Automated scoring of MAD2 intensity IHC where 1 is negative; 2, weak; 3, moderate; 4, strong. Abbreviations: CI, confidence interval. ⁎ Significant results, P b .05.

Overexpression of the BUBR1 gene has been seen in renal clear cell carcinomas, as compared with normal controls, using real-time quantitative polymerase chain reaction [31]. In the same study, comparative genomic hybridization analysis of renal clear cell carcinomas showed a significant correlation between BUBR1 overexpression and increased copy number alterations [31]. Increased BUBR1 IHC expression has also been shown to associate with shorter RFS and progression-free survival in bladder cancer and with chromosomal instability and aneuploidy in those tumors [17]. High-grade serous neoplasms have increased chromosomal instability when compared with low grade and other histologies [32]. We postulate that overexpression of BUBR1 may indicate a molecular shift in this protein during tumor progression. MAD2 was most commonly localized to the nucleus in our OC cohort, in all histologic subtypes, with perinuclear accentuation and less frequently cytoplasmic expression. Cytoplasmic BUBR1 and nuclear MAD2 correlated with each other, with high nuclear MAD2 expressers also showing strong BUBR1 intensity staining (P = .007). Nuclear MAD2 associated with increased cellular proliferation, as demonstrated by IHC assessment of the proliferation marker Ki-67 IHC (P b .0002). Overexpression of the BUBR1 gene has previously been shown to be present in 68% of gastric cancers and to correlate with increased Ki-67 messenger RNA levels in these tumors [33]. MAD2 expression has also been associated with increased Ki-67 by IHC in colorectal mucosa [16]. We have demonstrated a robust association between the IHC expression of nuclear MAD2 and cellular proliferation in OC, as measured by Ki67 IHC. Cytoplasmic BUBR1 intensity also associates with cellular proliferation (P = .02). Tumors with a high proliferation rate probably have an increased risk of acquired genetic instability due to accumulated mutations, but by default, this may make them more amenable to the use of antimitotic chemotherapeutics such as the taxanes. Low MAD2 IHC intensity has recently shown by our group to be independently associated with reduced progression-free survival in papillary serous ovarian cancers [19].

Another study found that high expression of MAD2 IHC is associated with increased sensitivity to taxane/carboplatin therapy and reduced recurrence (P = .023) [34]. In the current study, we assessed a TMA platform, which included a variety of histologic subtypes. The univariate analysis showed that reduced nuclear MAD2 IHC intensity was associated with a shorter RFS (P = .03). Importantly, this was found in all histologic subtypes irrespective of treatment. In addition, we have shown in this study that tumors expressing reduced MAD2 intensity continue to be independently associated with a shorter RFS after adjustment for grade and stage in a cohort of mixed ovarian cancer histologies (P = .05). There is undoubtedly a strong relationship between MAD2 expression and intrinsic and acquired taxane resistance [3,19,35]. However, the precise nature of that relationship remains unclear. Fu et al [6] demonstrated that a weakened SAC with reduced BUBR1 but not MAD2 protein expression was significantly associated with acquired paclitaxel resistance in ovarian cell lines. In another study, suppression of MAD2 and BUBR1 using siRNA in paclitaxel-treated Michigan Cancer Foundation-7 (MCF-7) breast cell line abolished checkpoint function, conferring resistance [14]. Moreover, miR-433 was identified as a down-regulator of MAD2, which made cells less sensitive to paclitaxel in A2780 ovarian cell lines [19]. The SAC has a role in cell cycle arrest and apoptosis after DNA damage. Reduced MAD2 expression in nasopharyngeal carcinoma cell lines correlates with increased resistance to cisplatin compared with MAD2 overexpressing cell lines [36]. Other data have shown that induction of double-stranded breaks in mitosis caused hyperphosphorylation and association of BUBR1 with kinetochores in mammalian cells [37]. Moreover, siRNA targeted knockdown of BUBR1 abrogated mitotic delay in response to chromosomal damage [37]. In all, this suggests that the SAC is activated following treatment with DNA-damaging agents, and those checkpoint proteins such as MAD2 and BUBR1 play a role in response to platinum-based chemotherapeutic agents as well as taxanes. In conclusion, in this study, we have accurately described the immunolocalization patterns of MAD2 and BUBR1 in

10 OC, in a variety of histologic subtypes and confirmed their association with cellular proliferation. Interestingly, MAD2 and BUBR1 are also associated with each other. Specifically, tumors with high MAD2 (T) IHC expression displayed high BUBR1 intensity IHC expression. We have confirmed the previously reported association between BUBR1 overexpression and serous histology, in addition to its association with advanced stage. We have also demonstrated a significant association between IHC nuclear intensity of MAD2 and RFS for women presenting with OC, with those women displaying reduced mitotic arrest deficient 2 nuclear expression having a shorter RFS not confined to any particular histologic subtype or treatment. Furthermore, reduced MAD2 intensity continued to be independently associated with a poorer RFS after adjustment for grade and stage in a cohort of varying OC histologies (P = .05).

4.1. Concluding remarks Because most cancer cell line studies have shown that decreased MAD2 messenger RNA and protein expression is associated with defects in SAC functioning and paclitaxel and cisplatin resistance [14,36], it could be postulated that overexpression of MAD2, as demonstrated in this study, may be a predictive marker for response to the chemotherapeutic agents paclitaxel and cisplatin. Furthermore, this study may help elucidate the complex role of the SAC and cell checkpoint proteins in response to microtubule inhibitory and DNA-damaging chemotherapeutic agents in the treatment of ovarian cancer.

Acknowledgments We would like also to acknowledge Dr Michael Foley and his colleagues at the National Maternity Hospital Dublin for allowing us access to the ovarian cancer database.

References [1] Harries M, Gore M. Part 1: chemotherapy for epithelial ovarian cancer —treatment at first diagnosis. Lancet Oncol 2002;3:529-36. [2] McGuire WP, Hoskins WJ, Brady MF, Kucera PR, Partridge EE, Look KY. Cyclophosphamide and cisplatin compared to paclitaxel and cisplatin in patients with stage III and stage IV ovarian cancer. N Engl J Med 1996;334:1-6. [3] McGrogan B, Gilmartin B, Carney DN, McCann A. Taxanes, microtubules and chemoresistant breast cancer. Biochim Biophys Acta 2008;1785:96-132. [4] Mielgo A, Torres VA, Clair K, Barbero S, Stupack DG. Paclitaxel promotes a caspase 8–mediated apoptosis through death effector domain association with microtubules. Oncogene 2009;28:3551-62. [5] Li M, Fang X, Wei Z, York JP, Zhang P. Loss of spindle assembly checkpoint–mediated inhibition of Cdc20 promotes tumorigenesis in mice. J Cell Biol 2009;185:983-94.

B. McGrogan et al. [6] Fu Y, Ye D, Chen H, Lu W, Ye F, Xie X. Weakened spindle checkpoint with reduced BubR1 expression in paclitaxel-resistant ovarian carcinoma cell line SKOV3-TR30. Gynecol Oncol 2007; 105:66-73. [7] Hao X, Zhao Z, Ye S, et al. Effect of MAD2 on paclitaxel-induced cell death in ovarian cancer cells. Huazhong Univ Sci Technology Med Sci 2010;30:620-5. [8] Li Y, Benezra R. Identification of a human mitotic checkpoint gene: hsMAD2. Science 1996;274:246-8. [9] Jin DY, Spencer F, Jeang KT. Human T cell leukaemia virus type 1 oncoprotein Tax targets the human mitotic checkpoint protein MAD1. Cell 1998;93:81-91. [10] Taylor SS, Ha E, McKeon F. The human homologue of Bub3 is required for kinetochore localisation of Bub1 and a Mad3/Bub1related protein kinase. J Cell Biol 1998;142:1-11. [11] Chan GK, Jablonski SA, Sudakin V, Hittle JC, Yen TJ. Human BUBR1 is a mitotic checkpoint kinase that monitors CENP-E functions at kinetochores and binds at the cyclosome/APC. J Cell Biol 1999;146:941-54. [12] Musacchio A, Salmon ED. The spindle-assembly checkpoint in space and time. Nature Reviews Mol Cell Biol 2007;8:379-93. [13] Greene LM, Campiani G, Lawler M, Williams DC, Zisterer DM. BubR1 is required for a sustained mitotic spindle checkpoint arrest in human cancer cells treated with tubulin-targeting pyrrolo-1,5-benzoxazepines. Mol Pharmacol 2008;73:419-30. [14] Sudo T, Nitta M, Saya H, Ueno NT. Dependence of paclitaxel sensitivity on a functional spindle assembly checkpoint. Cancer Res 2004;64:2502-8. [15] Malureanu LA, Jeganathan KB, Hamada M, Walilewski L, Davenport J, Van Deursen JM. BubR1 N terminus acts as a soluble inhibitor of cyclin B degradation by APC/C (Cdc20) in interphase. Dev Cell 2009;16:118-31. [16] Burum-Auensen E, De Angelis PM, Schjølberg AR, Kravik KL, Aure M, Clausen OP. Subcellular localisation of the spindle proteins Aurora B, MAD2 and BUBR1 assessed by immunohistochemistry. J Histochem Cytochem 2007;55:477-86. [17] Yamamoto Y, Matsuyama H, Chochi Y, et al. Overexpression of BUBR1 is associated with chromosomal instability in bladder cancer. Cancer Genet Cytogenet 2007;174:42-7. [18] Lee YK, Choi E, Kim MA, Park PG, Park NH, Lee H. BubR1 as a prognostic marker for recurrence-free survival rates in epithelial ovarian cancer. Br J Cancer 2009;101:504-10. [19] Furlong F, Fitzpatrick P, O'Toole S, et al. Low MAD2 expression levels associate with reduced progression free survival in patients with highgrade serous epithelial ovarian cancer. J Pathol 2012;226:746-55. [20] Zhang SH, Xu AM, Chen XF, Li DH, Sun MP, Wang YJ. Clinicopathologic significance of mitotic arrest defective protein 2 overexpression in hepatocellular carcinoma. H UM P ATHOL 2008;39:1827-34. [21] Wang L, Yin F, Du Y, Du W, Chen B, Zhang Y. MAD2 as a key component of mitotic checkpoint: a probable prognostic factor for gastric cancer. Am J Clin Pathol 2009;131:793-801. [22] Fung MK, Cheung HW, Wong HL, Yuen HF, Ling MT, Chan KW. MAD2 expression and its significance in mitotic checkpoint control in testicular germ cell tumour. Biochim Biophys Acta 2007;1773:821-32. [23] Brennan DJ, Ek S, Doyle E, et al. The transcription factor Sox11 is a prognostic factor for improved recurrence free survival in patients with high-grade serous epithelial ovarian cancer. Eur J Cancer 2009;45:1510-7. [24] Silverberg SG. Histologic grading of epithelial ovarian carcinoma: a review and proposal. Int J Gynae Path 2000;19:7-15. [25] Österberg L, Levan K, Partheen K, Staat J, Sundfeld K, Horvath G. High-resolution genomic profiling of carboplatin resistance in earlystage epithelial ovarian carcinoma. Cytogenet Genome Res 2009;125:8-18. [26] Doyle EM, Foley M, Kelehan P, Mooney EE. Histological grading of epithelial ovarian cancers. J Obstet Gynaecol 2007;27:71-4.

Q5

SAC protein expression in OC recurrence [27] Halliday AI. Spindle checkpoint protein hMAD2 and histone H3 phophoserine 10 mitosis marker in pediatric solid tumors. Anticancer research 2006;26:4687-94. [28] Kandalaft PL, Chang KL. Prognostic significance of the immunohistochemical analysis of cathepsin-D in low-stage breast cancer. Cancer 1993;71:2756-63. [29] Hisaoka M, Matsuyama A, Hashimoto H. Aberrant MAD2 expression in soft-tissue sarcoma. Pathol Int 2008;58:329-33. [30] Du J, Du Q, Zhang Y, et al. Expression of cell-cycle regulatory proteins BUBR1, MAD2, Aurora A, cyclin A and cyclin E in invasive ductal breast carcinomas. Histol Histopathol 2011;26:761-8. [31] Pinto M, Vieira J, Ribeiro F, et al. Overexpression of the mitotic checkpoint genes BUB1 and BUBR1 is associated with genomic complexity in clear cell kidney carcinomas. Cell Oncol 2008;30: 389-95. [32] Kuo KT, Guan B, Feng T, et al. Analysis of DNA copy number alterations in ovarian serous tumors identifies new molecular genetic

11

[33]

[34]

[35]

[36]

[37]

changes in low-grade and high-grade carcinomas. Cancer Res 2009;69:5267. Grabsch H, Takeno S, Parsons WJ, et al. Overexpression of the mitotic checkpoint genes BUB1, BUBR1 and BUB3 in gastric cancerassociation with tumour cell proliferation. J Pathol 2003;200:16-22. Nakano Y, Sumi T, Teramae M, et al. Expression of the mitotic-arrest deficiency 2 is associated with chemotherapy resistance in ovarian serous adenocarcinoma. Oncol Rep 2012;28:1200-4. Wu YC, Yen WY, Yih LH. Requirement of a functional spindle checkpoint for arsenite-induced apoptosis. J Cell Biochem 2008;105:678-87. Cheung HW, Jin DY, Ling MT, Wong YC, Wang Q, Tsao SW. Mitotic arrest deficient 2 expression induces chemosensitization to a DNA-damaging agent, cisplatin, in nasopharyngeal carcinoma cells. Cancer Res 2005;65:1450-8. Choi E, Lee H. Chromosome damage in mitosis induced BubR1 activation in prometaphase arrest. FEBS letters 2008;582:1700-6.