Original Article
Loss of BAP1 Protein Expression Is an Independent Marker of Poor Prognosis in Patients With Low-Risk Clear Cell Renal Cell Carcinoma Richard W. Joseph, MD1; Payal Kapur, MD2; Daniel J. Serie, BS3; Jeanette E. Eckel-Passow, Ph. D4; Mansi Parasramka, Ph.D5; Thai Ho, MD, Ph.D6; John C. Cheville, MD6; Eugene Frenkel, MD7; Dinesh Rakheja, MD9; James Brugarolas, MD, Ph.D7,8; and Alexander Parker, PhD3
BACKGROUND: The majority of patients diagnosed with clear cell renal cell carcinoma (ccRCC) have low-risk disease with a < 10% chance of ccRCC-specific death. DNA sequencing revealed that mutations in BAP1 (BRCA1 associated protein-1) occur in 5% to 15% of ccRCC cases and are associated with poor outcomes. The vast majority of BAP1 mutations abolish protein expression. In this study, we used a highly sensitive and specific immunohistochemistry (IHC) assay to test whether BAP1 expression is an independent marker of ccRCC-specific survival, particularly in patients with low-risk disease. METHODS: BAP1 expression was assessed, using IHC, in 1479 patients who underwent nephrectomy to treat clinically localized ccRCC. A centralized pathologist dichotomized patients as either BAP1-positive or BAP1-negative. The authors employed Kaplan-Meier and Cox regression models to associate BAP1 expression with cancer-specific survival. RESULTS: A total of 10.5% of tumors were BAP1-negative, 84.8% of tumors were BAP1-positive, and 4.6% of tumors had ambiguous staining for BAP1. Patients with BAP1-negative tumors have an increased risk of ccRCC-related death (hazard ratio [HR] 5 3.06; 95% confidence interval [CI] 5 2.28-4.10; P 5 6.77 3 10214). BAP1 expression remained an independent marker of prognosis after adjusting for the UCLA integrated staging system (UISS) (HR 5 1.67; 95% CI 5 1.24-2.25; P 98%.8 Using this IHC assay, we sought to determine if BAP1 protein expression is an independent marker of ccRCC-related prognosis, especially in those patients with low-risk disease as defined by individual pathologic indices (ie, stage and grade) and our own institution’s multivariable prognostic algorithms that account for tumor stage, size, grade, and necrosis (SSIGN score14,15). Finally, in an exploratory analysis, we assessed whether BAP1 expression remained an independent marker of prognosis after adjusting for other biomarkers that are associated with ccRCC prognosis (ie, PDL1, Ki67, survivin).
Data Collection
MATERIALS AND METHODS Patient Selection
Previous Published ccRCC Nomograms and Assessment of PDL1, Ki-67, and Survivin
After Institutional Review Board approval, we identified 1479 patients treated with radical nephrectomy or nephron-sparing surgery for unilateral, sporadic, noncystic ccRCC between 1990 and 2006 from the Mayo Clinic Rochester Nephrectomy Registry with representative paraffin-embedded tissue blocks available for IHC staining and data on RCC-specific death. Of these 1479 patients, we successfully stained 1454 (98.3%) for BAP1, and 25 slides were defective or did not stain.
We compared the impact of BAP1 expression on several previously published and validated ccRCC nomograms including the 2002 American Joint Committee on Cancer (AJCC) TNM stage groupings,16 the UCLA Integrated Scoring System (UISS),17 nomograms from Memorial Sloan-Kettering Cancer Center,18,19 and the Mayo Clinic SSIGN Score.14,15 In a subset of patients and as previously described, we assessed the expression by IHC of programmed death ligand 1 (PDL1), Ki-67, and survivin.20
1060
Follow-up data (ie, date of RCC death, date of last follow-up) and clinicopathologic covariates were abstracted from the Registry at Mayo Clinic. Briefly, these data are routinely updated and maintained through a combination of active (mail-out questionnaires) and passive (medical record, linkage to national databases) surveillance by experienced clinical coordinators. Pathologic features were analyzed in a standardized fashion by one urologic pathologist (J.C.C.) who centrally reviewed the microscopic hematoxylin and eosin–stained slides from all specimens without knowledge of patient outcome. BAP1 Protein Expression by IHC
IHC assay for BAP1 was performed as described.8 Positive staining in the background stromal cells and intratumoral lymphocytes served as internal positive control. A pathologist (P.K.), who was blinded to the clinicopathological variables, reviewed all immunostained slides, and a second pathologist (D.R.) reviewed all cases that lacked diffuse strong nuclear staining. Pathologists did not agree on a total of 6 (0.4%) samples. Tumors were categorized as BAP1-negative when tumor cells showed diffuse absence of nuclear BAP1 staining (previously shown to correlate with BAP1 mutation),8 and BAP1-positive when tumor cells demonstrated diffuse nuclear staining with intensity equal to or stronger than the surrounding stromal cells and lymphocytes. In a small subset of tumors, BAP1 expression was unclear. Within this unclear group, some tumors were “BAP1 heterogeneous” with some cells staining positive for BAP1 and some cells negative (range of 1%-99% cells staining negative for BAP1). Also within the group that was unclear for BAP1 staining, there were tumors expressing BAP1 uniformly but weakly. Because of the uncertainty of BAP1 expression, we excluded both the “BAP1 heterogenous” and “diffusely weak” cohorts from our primary analysis.
Cancer
April 1, 2014
BAP1 IHC in ccRCC Associated With Poor Prognosis/Joseph et al
Statistical Methods
Clinical and pathologic data were compared between patients with BAP1-positive and BAP1-negative tumors using t tests, Fisher’s exact tests, and Chi-square tests, as appropriate. Unless noted otherwise, SSIGN and UISS scores were modeled as continuous variables. Cox models were used to assess the association of BAP1 expression, dichotomized as negative versus positive expression, with RCC-specific death after adjusting for age. In addition, the Kaplan-Meier method was used to estimate RCCspecific death, adjusting for the competing risk of death due to other causes. Subsequently, BAP1 was further evaluated in multivariable models adjusting for other known prognostic variables of ccRCC outcome, one at a time, in addition to age. We calculated concordance (c) index values to assess predictive ability. All reported c indexes were generated using the bootstrap methodology proposed by Harrell et al and represent optimism-corrected estimates of concordance.21 Statistical analyses were performed using the R programming language, version 2.15. All tests were 2-sided and P values < .05 were considered statistically significant. RESULTS Association of BAP1 With Clinical and Pathologic Characteristics
Of the total cohort of 1479 slides, we successfully stained BAP1 in 1454 (98.3%). An additional 44 samples were excluded from the analysis secondary to inadequate follow up or duplicate ID leaving a total of 1410 (95.3%) for analysis. Of the 1410 successfully stained slides with adequate clinical follow up, 1344 were classified as either BAP1 negative (n 5 148, 10.5%) or BAP1 positive (n 5 1196, 84.8%), and the rest were BAP1 unclear (n 5 66, 4.7%). Of the 66 BAP1 unclear, 33 (50%) had BAP1 heterogenous staining with some cells staining positive and others staining negative, and 33 (50%) were diffusely weak for BAP1. The 66 BAP1 unclear were removed from the primary analysis. The list of clinical characteristics of the BAP1 negative and positive (n 5 1344) are listed in Table 1. Representative examples of BAP1 staining are demonstrated in Figure 1. A total of 1092 patients were alive at last follow-up, with a mean duration of follow-up of 8.3 years (median, 7.7 years; range, 0.003 5 22.1 years). Only 20 (1.4%) patients had fewer than 2 years of follow-up. There was no difference in age, sex, history of smoking, or history of alcohol use between patients with BAP1 negative and positive tumors. Tumors that were BAP1-negative by IHC possessed a larger tumor size, TNM stage, nuclear grade, and presence of coagulaCancer
April 1, 2014
tive tumor necrosis (all with P < .0001).3,4,8 BAP1negative tumors also had significantly higher SSIGN and UISS scores. Notably, all BAP1-negative tumors had a nuclear grade of 2 or higher and in fact, 90% of the BAP1mutant tumors were of high (grade 3-4) Fuhrman grade. Univariate Analysis BAP1 Expression With ccRCC-Specific Death
At last follow-up, a total of 562 (42%) deaths and 252 (19%) ccRCC-specific deaths occurred. Patients with BAP1-negative tumors were 3 times more likely to experience RCC specific death compared with patients with BAP1-positive tumors (HR 5 3.06; 95% CI 5 2.284.10; P 5 6.77 3 10214) (Fig. 2). In secondary analyses, we interrogated the 66 patients who had unclear BAP1 tumor staining, and of these 66, 62 had adequate follow-up. We were unable to detect a difference in ccRCC-specific death between the 30 patients with heterogeneous BAP1 expression and patients with BAP1-positive tumors (HR 5 1.25; 95% CI 5 0.55-2.81; P 5 .59). However, the 32 patients with “diffusely weak” BAP1 expression had an increased risk of ccRCC death (HR 5 2.00; 95% CI 5 1.09-3.67; P 5 .025) when compared to patients with BAP1-positive tumors (Supporting Fig. 1; see online supporting information). Multivariable Analysis of BAP1 With ccRCCSpecific Death
We performed a multivariable analysis to assess if BAP1 expression was independently associated with ccRCCspecific death after adjustment for known prognostic variables (Table 2). BAP1 was no longer statistically significant after adjusting for nuclear grade and SSIGN score, but retained its statistical significance after adjusting for TNM (HR 5 1.69; 95% CI 5 1.25-2.27; P 5 .0006), presence of necrosis (HR 5 1.57; 95% CI 5 1.16-2.12; P 5 .0036), tumor size (HR 5 1.90; 95% CI 5 1.402.57; P < .0001) and UISS (HR 5 1.67; 95% CI 5 1.242.67; P 5 .0008). In these multivariable models, we also noted improvements in the c index when the BAP1 variable was added to a model containing known predictors of ccRCC outcome. Impact of BAP1 Expression on ccRCC-Specific Death in Low-, Intermediate-, and High-Risk Groups
As noted above, BAP1 was not an independent prognostic variable after adjusting for SSIGN score. Next, we evaluated whether there was a statistically significant interaction between BAP1 expression and SSIGN score and 1061
Original Article TABLE 1. Clinical and Pathological Information for Patients in the BAP1 Analyses BAP1-Negative N 5 148 (11%) Sex Female Male Age at surgery Mean Median Range BMI Mean Median Range Cigarette use Missing Never Current Past Alcohol use Missing Never Current Past Tumor size Mean Median Range Time to last follow-up Mean Median Range Death No Yes Death from RCC No Yes TNM stage Missing 1 2 3 4 Nuclear grade 1 2 3 4 Coagulative tumor necrosis No Yes SSIGN category Missing 0-3 4-7 81 UISS score Missing 1 2 3 4 5 PDL1 expression Missing Negative Positive
1062
BAP1-Positive N 5 1196 (89%)
Total (N 5 1344)
P .6468
48 (32.4%) 100 (67.6%)
415 (34.7%) 781 (65.3%)
463 (34.4%) 881 (65.6%) .3008
64.2 65.4 (35.9-90.0)
62.9 64.0 (19.8-90.2)
63.1 64.2 (19.8-90.2)
28.9 27.8 (17.9-61.3)
29.9 29.0 (15.9-60.8)
29.8 28.7 (15.9-61.3)
.0313
.2708 13 57 (42.2%) 25 (18.5%) 53 (39.3%)
77 507 (45.3%) 150 (13.4%) 462 (41.3%)
90 564 (45.0%) 175 (14.0%) 515 (41.1%)
12 70 (51.5%) 56 (41.2%) 10 (7.4%)
104 510 (46.7%) 477 (43.7%) 105 (9.6%)
116 580 (47.2%) 533 (43.4%) 115 (9.4%)
.5306
Loss of BAP1 Protein Expression Is an Independent Marker of Poor Prognosis in Patients With Low-Risk Clear Cell Renal Cell Carcinoma Richard W. Joseph, MD1; Payal Kapur, MD2; Daniel J. Serie, BS3; Jeanette E. Eckel-Passow, Ph. D4; Mansi Parasramka, Ph.D5; Thai Ho, MD, Ph.D6; John C. Cheville, MD6; Eugene Frenkel, MD7; Dinesh Rakheja, MD9; James Brugarolas, MD, Ph.D7,8; and Alexander Parker, PhD3
BACKGROUND: The majority of patients diagnosed with clear cell renal cell carcinoma (ccRCC) have low-risk disease with a < 10% chance of ccRCC-specific death. DNA sequencing revealed that mutations in BAP1 (BRCA1 associated protein-1) occur in 5% to 15% of ccRCC cases and are associated with poor outcomes. The vast majority of BAP1 mutations abolish protein expression. In this study, we used a highly sensitive and specific immunohistochemistry (IHC) assay to test whether BAP1 expression is an independent marker of ccRCC-specific survival, particularly in patients with low-risk disease. METHODS: BAP1 expression was assessed, using IHC, in 1479 patients who underwent nephrectomy to treat clinically localized ccRCC. A centralized pathologist dichotomized patients as either BAP1-positive or BAP1-negative. The authors employed Kaplan-Meier and Cox regression models to associate BAP1 expression with cancer-specific survival. RESULTS: A total of 10.5% of tumors were BAP1-negative, 84.8% of tumors were BAP1-positive, and 4.6% of tumors had ambiguous staining for BAP1. Patients with BAP1-negative tumors have an increased risk of ccRCC-related death (hazard ratio [HR] 5 3.06; 95% confidence interval [CI] 5 2.28-4.10; P 5 6.77 3 10214). BAP1 expression remained an independent marker of prognosis after adjusting for the UCLA integrated staging system (UISS) (HR 5 1.67; 95% CI 5 1.24-2.25; P 98%.8 Using this IHC assay, we sought to determine if BAP1 protein expression is an independent marker of ccRCC-related prognosis, especially in those patients with low-risk disease as defined by individual pathologic indices (ie, stage and grade) and our own institution’s multivariable prognostic algorithms that account for tumor stage, size, grade, and necrosis (SSIGN score14,15). Finally, in an exploratory analysis, we assessed whether BAP1 expression remained an independent marker of prognosis after adjusting for other biomarkers that are associated with ccRCC prognosis (ie, PDL1, Ki67, survivin).
Data Collection
MATERIALS AND METHODS Patient Selection
Previous Published ccRCC Nomograms and Assessment of PDL1, Ki-67, and Survivin
After Institutional Review Board approval, we identified 1479 patients treated with radical nephrectomy or nephron-sparing surgery for unilateral, sporadic, noncystic ccRCC between 1990 and 2006 from the Mayo Clinic Rochester Nephrectomy Registry with representative paraffin-embedded tissue blocks available for IHC staining and data on RCC-specific death. Of these 1479 patients, we successfully stained 1454 (98.3%) for BAP1, and 25 slides were defective or did not stain.
We compared the impact of BAP1 expression on several previously published and validated ccRCC nomograms including the 2002 American Joint Committee on Cancer (AJCC) TNM stage groupings,16 the UCLA Integrated Scoring System (UISS),17 nomograms from Memorial Sloan-Kettering Cancer Center,18,19 and the Mayo Clinic SSIGN Score.14,15 In a subset of patients and as previously described, we assessed the expression by IHC of programmed death ligand 1 (PDL1), Ki-67, and survivin.20
1060
Follow-up data (ie, date of RCC death, date of last follow-up) and clinicopathologic covariates were abstracted from the Registry at Mayo Clinic. Briefly, these data are routinely updated and maintained through a combination of active (mail-out questionnaires) and passive (medical record, linkage to national databases) surveillance by experienced clinical coordinators. Pathologic features were analyzed in a standardized fashion by one urologic pathologist (J.C.C.) who centrally reviewed the microscopic hematoxylin and eosin–stained slides from all specimens without knowledge of patient outcome. BAP1 Protein Expression by IHC
IHC assay for BAP1 was performed as described.8 Positive staining in the background stromal cells and intratumoral lymphocytes served as internal positive control. A pathologist (P.K.), who was blinded to the clinicopathological variables, reviewed all immunostained slides, and a second pathologist (D.R.) reviewed all cases that lacked diffuse strong nuclear staining. Pathologists did not agree on a total of 6 (0.4%) samples. Tumors were categorized as BAP1-negative when tumor cells showed diffuse absence of nuclear BAP1 staining (previously shown to correlate with BAP1 mutation),8 and BAP1-positive when tumor cells demonstrated diffuse nuclear staining with intensity equal to or stronger than the surrounding stromal cells and lymphocytes. In a small subset of tumors, BAP1 expression was unclear. Within this unclear group, some tumors were “BAP1 heterogeneous” with some cells staining positive for BAP1 and some cells negative (range of 1%-99% cells staining negative for BAP1). Also within the group that was unclear for BAP1 staining, there were tumors expressing BAP1 uniformly but weakly. Because of the uncertainty of BAP1 expression, we excluded both the “BAP1 heterogenous” and “diffusely weak” cohorts from our primary analysis.
Cancer
April 1, 2014
BAP1 IHC in ccRCC Associated With Poor Prognosis/Joseph et al
Statistical Methods
Clinical and pathologic data were compared between patients with BAP1-positive and BAP1-negative tumors using t tests, Fisher’s exact tests, and Chi-square tests, as appropriate. Unless noted otherwise, SSIGN and UISS scores were modeled as continuous variables. Cox models were used to assess the association of BAP1 expression, dichotomized as negative versus positive expression, with RCC-specific death after adjusting for age. In addition, the Kaplan-Meier method was used to estimate RCCspecific death, adjusting for the competing risk of death due to other causes. Subsequently, BAP1 was further evaluated in multivariable models adjusting for other known prognostic variables of ccRCC outcome, one at a time, in addition to age. We calculated concordance (c) index values to assess predictive ability. All reported c indexes were generated using the bootstrap methodology proposed by Harrell et al and represent optimism-corrected estimates of concordance.21 Statistical analyses were performed using the R programming language, version 2.15. All tests were 2-sided and P values < .05 were considered statistically significant. RESULTS Association of BAP1 With Clinical and Pathologic Characteristics
Of the total cohort of 1479 slides, we successfully stained BAP1 in 1454 (98.3%). An additional 44 samples were excluded from the analysis secondary to inadequate follow up or duplicate ID leaving a total of 1410 (95.3%) for analysis. Of the 1410 successfully stained slides with adequate clinical follow up, 1344 were classified as either BAP1 negative (n 5 148, 10.5%) or BAP1 positive (n 5 1196, 84.8%), and the rest were BAP1 unclear (n 5 66, 4.7%). Of the 66 BAP1 unclear, 33 (50%) had BAP1 heterogenous staining with some cells staining positive and others staining negative, and 33 (50%) were diffusely weak for BAP1. The 66 BAP1 unclear were removed from the primary analysis. The list of clinical characteristics of the BAP1 negative and positive (n 5 1344) are listed in Table 1. Representative examples of BAP1 staining are demonstrated in Figure 1. A total of 1092 patients were alive at last follow-up, with a mean duration of follow-up of 8.3 years (median, 7.7 years; range, 0.003 5 22.1 years). Only 20 (1.4%) patients had fewer than 2 years of follow-up. There was no difference in age, sex, history of smoking, or history of alcohol use between patients with BAP1 negative and positive tumors. Tumors that were BAP1-negative by IHC possessed a larger tumor size, TNM stage, nuclear grade, and presence of coagulaCancer
April 1, 2014
tive tumor necrosis (all with P < .0001).3,4,8 BAP1negative tumors also had significantly higher SSIGN and UISS scores. Notably, all BAP1-negative tumors had a nuclear grade of 2 or higher and in fact, 90% of the BAP1mutant tumors were of high (grade 3-4) Fuhrman grade. Univariate Analysis BAP1 Expression With ccRCC-Specific Death
At last follow-up, a total of 562 (42%) deaths and 252 (19%) ccRCC-specific deaths occurred. Patients with BAP1-negative tumors were 3 times more likely to experience RCC specific death compared with patients with BAP1-positive tumors (HR 5 3.06; 95% CI 5 2.284.10; P 5 6.77 3 10214) (Fig. 2). In secondary analyses, we interrogated the 66 patients who had unclear BAP1 tumor staining, and of these 66, 62 had adequate follow-up. We were unable to detect a difference in ccRCC-specific death between the 30 patients with heterogeneous BAP1 expression and patients with BAP1-positive tumors (HR 5 1.25; 95% CI 5 0.55-2.81; P 5 .59). However, the 32 patients with “diffusely weak” BAP1 expression had an increased risk of ccRCC death (HR 5 2.00; 95% CI 5 1.09-3.67; P 5 .025) when compared to patients with BAP1-positive tumors (Supporting Fig. 1; see online supporting information). Multivariable Analysis of BAP1 With ccRCCSpecific Death
We performed a multivariable analysis to assess if BAP1 expression was independently associated with ccRCCspecific death after adjustment for known prognostic variables (Table 2). BAP1 was no longer statistically significant after adjusting for nuclear grade and SSIGN score, but retained its statistical significance after adjusting for TNM (HR 5 1.69; 95% CI 5 1.25-2.27; P 5 .0006), presence of necrosis (HR 5 1.57; 95% CI 5 1.16-2.12; P 5 .0036), tumor size (HR 5 1.90; 95% CI 5 1.402.57; P < .0001) and UISS (HR 5 1.67; 95% CI 5 1.242.67; P 5 .0008). In these multivariable models, we also noted improvements in the c index when the BAP1 variable was added to a model containing known predictors of ccRCC outcome. Impact of BAP1 Expression on ccRCC-Specific Death in Low-, Intermediate-, and High-Risk Groups
As noted above, BAP1 was not an independent prognostic variable after adjusting for SSIGN score. Next, we evaluated whether there was a statistically significant interaction between BAP1 expression and SSIGN score and 1061
Original Article TABLE 1. Clinical and Pathological Information for Patients in the BAP1 Analyses BAP1-Negative N 5 148 (11%) Sex Female Male Age at surgery Mean Median Range BMI Mean Median Range Cigarette use Missing Never Current Past Alcohol use Missing Never Current Past Tumor size Mean Median Range Time to last follow-up Mean Median Range Death No Yes Death from RCC No Yes TNM stage Missing 1 2 3 4 Nuclear grade 1 2 3 4 Coagulative tumor necrosis No Yes SSIGN category Missing 0-3 4-7 81 UISS score Missing 1 2 3 4 5 PDL1 expression Missing Negative Positive
1062
BAP1-Positive N 5 1196 (89%)
Total (N 5 1344)
P .6468
48 (32.4%) 100 (67.6%)
415 (34.7%) 781 (65.3%)
463 (34.4%) 881 (65.6%) .3008
64.2 65.4 (35.9-90.0)
62.9 64.0 (19.8-90.2)
63.1 64.2 (19.8-90.2)
28.9 27.8 (17.9-61.3)
29.9 29.0 (15.9-60.8)
29.8 28.7 (15.9-61.3)
.0313
.2708 13 57 (42.2%) 25 (18.5%) 53 (39.3%)
77 507 (45.3%) 150 (13.4%) 462 (41.3%)
90 564 (45.0%) 175 (14.0%) 515 (41.1%)
12 70 (51.5%) 56 (41.2%) 10 (7.4%)
104 510 (46.7%) 477 (43.7%) 105 (9.6%)
116 580 (47.2%) 533 (43.4%) 115 (9.4%)
.5306
