Circulatory MIC-1 as a Determinant of Prostate Cancer Racial Disparity

Simple Summary African American men are diagnosed with more aggressive prostate cancer and have worse outcomes than Caucasians. This study examined the role of MIC-1 as a risk factor and demonstrated a conceptual observation for the differential level of MIC-1 in circulation (serum and urine) and tumor tissues from prostate cancer patients of racial disparity. The circulatory MIC-1 levels in serum and urine are significantly higher in prostate cancer patients of African American ethnicity, with higher sensitivity and specificity than Caucasians. The validation of circulatory MIC-1 in a larger cohort of patients may help identify high-risk prostate cancer patients and develop race-oriented therapies to reduce the observed cancer outcome gaps between the races. Abstract In this study, we investigated the potential of MIC-1 (macrophage inhibitory cytokine-1) on the severity of prostate cancer between African American men and Caucasians. Differences between the races were examined using Mann–Whitney tests for continuous variables and Fisher’s exact tests for categorical variables. Pearson’s correlation coefficient was used to identify associations between continuous measures across all samples and within each race. Analysis of variance, including clinical parameters, was used to identify differences in serum and urine MIC-1 levels between races. We found significant differences between the two races for age (p = 0.01), Gleason scores (p = 0.01), and stage of disease (p = 0.03). African American men in the study had higher Gleason scores (mean = 6.9) than Caucasians (mean = 6.5), during earlier stages of the disease. In Caucasian men with prostate cancer, serum MIC-1 expression was positively associated with age (r = 0.7, p < 0.01). However, African American men had highly expressed MIC-1 and high Gleason scores (r = 0.16, p = 0.3). Interestingly, the urine MIC-1 level was significantly higher in African American men with prostate cancer than in Caucasian patients. It appeared to be more sensitive and specific for African Americans (AUC = 0.85 vs. 0.56). Thus, high circulatory MIC-1 in prostate cancer patients may indicate MIC-1 as a potential biomarker to improve the diagnostic ability of an aggressive stage of prostate cancer in African American men. However, a larger cohort of sample analysis is required to validate these observations.


Introduction
Prostate cancer is the most common cancer and is the second leading cause of cancer-related deaths among men, accounting for almost 30,000 deaths annually in the United States [1]. Unfortunately,

Univariate Analyses
The primary question addressed in this analysis is whether MIC-1 provides any predictive capability for prostate cancer severity at the time of initial diagnosis. Our data consists of information on 50 African American and 55 Caucasian men between the ages of 43 and 77 years (Median = 61 years). Because the African American race is a well-known risk factor for prostate cancer development, the severity of disease at diagnosis, and poor disease prognosis, we report the results both for the entire sample and by race ( Table 1). The p-values in Table 1 were generated from the untransformed variables using nonparametric tests that do not assume normality of the data. Highly significant differences between the two races were found in age (p = 0.02), Gleason scores (p = 0.01), and stage of disease (p = 0.03). On average, African American men were 7 years younger than Caucasian men and had higher Gleason scores (mean = 6.9, SD = 0.7) than Caucasians (mean = 6.5, SD = 1.1), and earlier stage of the disease. PSA levels were also 23% higher (p = 0.09) in African American men (mean = 8.6, SD = 5.3) than in Caucasians (mean = 7.0, SD = 4.0). African American men, overall, had higher serum MIC-1 expression (median = 1152.2 pg/mL, range = 329.1 to 8682.5) compared to Caucasian men (median = 809.9 pg/mL, range = 126.3 to 4664.4). Serum MIC-1 levels in healthy subjects from both races were similar (n = 20 African American, median = 865.7 pg/mL, range = 329.1 to 1667.1; n = 20 Caucasian, median = 747.0 pg/mL, range = 364.5 to 3311.9; p = 0.8). However, serum MIC-1 from African American prostate cancer patients (median = 1393.9 pg/mL, range = 512.2 to 8682.5) was significantly higher as compared to African American healthy donors (median = 865.7 pg/mL, range = 329.1, 1667.1; p < 0.001) (Figure 1). This trend was not observed in Caucasian men (p = 0.3) (Figure 1).   pg/mL, range = 329.1, 1,667.1; p < 0.001) (Figure 1). This trend was not observed in Caucasian men (p = 0.3) (Figure 1).  individual data points between healthy and prostate cancer patients in Caucasians (Cauc) and African Americans (AA). All samples include both races (healthy: n = 40; Cancer: n = 105). Statistical significance assessed by using Tukey-adjusted p-values from ANOVA demonstrates p < 0.01 (**), p < 0.001 (***) for starred comparisons, and ns represents non significant.

Correlation Analyses
Upon examination of the data, MIC-1 and PSA levels were found to be highly skewed. Due to the non-normality of these data, we performed natural log transformations on MIC-1 and PSA levels to meet the parametric assumptions required by correlation and regression analyses. Pairwise correlations were used to identify linear relationships between variables. Pearson's correlation coefficient is provided in Table 2 for age, log(PSA), and log(MIC-1). Because the Gleason score is ordinal over the range 6 to 9, Spearman's nonparametric correlation coefficient was used to calculate with individual data points between healthy and prostate cancer patients in Caucasians (Cauc) and African Americans (AA). All samples include both races (healthy: n = 40; Cancer: n = 105). Statistical significance assessed by using Tukey-adjusted p-values from ANOVA demonstrates p < 0.01 (**), p < 0.001 (***) for starred comparisons, and ns represents non significant.

Correlation Analyses
Upon examination of the data, MIC-1 and PSA levels were found to be highly skewed. Due to the non-normality of these data, we performed natural log transformations on MIC-1 and PSA levels to meet the parametric assumptions required by correlation and regression analyses. Pairwise correlations were used to identify linear relationships between variables. Pearson's correlation coefficient is provided in Table 2 for age, log(PSA), and log(MIC-1). Because the Gleason score is ordinal over the range 6 to 9, Spearman's nonparametric correlation coefficient was used to calculate bivariate correlations that included the measure. Among African Americans, no significant correlations were found among any Cancers 2020, 12, 3033 5 of 12 of the variables. However, in Caucasians log(MIC-1) serum was found to be positively associated with log(PSA) (r = 0.26) and age (r = 0.67), while the Gleason score was positively associated with log(PSA) (r = 0.21) and age (r = 0.20). Thus, higher levels of MIC-1 and higher Gleason scores are associated with older patients when limiting our sample to Caucasians. However, this was not true in African Americans, as both older and younger patients have highly expressed MIC-1 (r = 0.35) ( Figure 2A) and high Gleason scores (r = −0.08) ( Figure 2B). This phenomenon was not observed when examining PSA expression levels by age among African Americans and Caucasians. Though it appears that African Americans have slightly higher PSA levels than Caucasians when accounting for age, there is no age association of PSA in either of the groups ( Figure 2C).

Multivariable Regression Models
To further examine these differences at the levels of the categorical variable such as race, we used analysis of covariance (ANCOVA), which combines regression and analysis of variance by introducing the continuous variable, age, into the ANOVA model. Because we noticed that the slopes of the regression lines within the two groups were not parallel for serum MIC-1 (Figure 2A), we used a model that included an interaction term with the race to determine whether the slopes of the lines differ significantly by race. The slope of the regression line for Caucasians was significantly more positive in the log(MIC-1) when compared to that of African Americans (p < 0.01, Figure 2A); that is, the association of age with MIC-1 is significantly greater in Caucasians than in African Americans. Additionally, from a centered-age model, we found that Caucasians at 61 years of age have significantly reduced log(MIC-1) expression levels (mean = 6.72, 95% CI = 6.58, 6.87) when compared to African Americans of the same age (mean = 7.33, 95% CI = 7.17, 7.48) (p < 0.001). We built a similar model to adjust the comparison of log serum MIC-1 between races for Gleason score. The centered-Gleason Score model indicated that Caucasians with a Gleason Score of 6.7 have significantly reduced

Multivariable Regression Models
To further examine these differences at the levels of the categorical variable such as race, we used analysis of covariance (ANCOVA), which combines regression and analysis of variance by introducing the continuous variable, age, into the ANOVA model. Because we noticed that the slopes of the regression lines within the two groups were not parallel for serum MIC-1 (Figure 2A), we used a model that included an interaction term with the race to determine whether the slopes of the lines differ significantly by race. The slope of the regression line for Caucasians was significantly more positive in the log(MIC-1) when compared to that of African Americans (p < 0.01, Figure 2A); that is, the association of age with MIC-1 is significantly greater in Caucasians than in African Americans. Additionally, from a centered-age model, we found that Caucasians at 61 years of age have significantly reduced log(MIC-1) expression levels (mean = 6.72, 95% CI = 6.58, 6.87) when compared to African Americans of the same age (mean = 7.33, 95% CI = 7.17, 7.48) (p < 0.001). We built a similar model to adjust the comparison of log serum MIC-1 between races for Gleason score. The centered-Gleason Score model indicated that Caucasians with a Gleason Score of 6.7 have significantly reduced log(MIC-1) expression levels (mean = 6.81, 95% CI = 6.63, 6.98) when compared to African Americans with the same Gleason Score (mean = 7.29, 95% CI = 7.10, 7.48).
To further understand the association of serum MIC-1 level with Gleason scores (GS), we categorized these samples into two groups of GS ≤ 6 and GS ≥7 and performed ANOVA. Based on Tucky-adjusted p-values for multiple comparisons for all samples, the serum MIC-1 level was significantly higher (p = 0.02) in patients grouped in GS ≥ 7 (n = 67; mean = 1669.57 pg/mL) compared to GS ≤ 6 (n = 38; mean = 1079.71 pg/mL). Within the races, differences in the serum MIC-1 level between two GS groups were not statistically significant (AA: p = 0.41; Cau: p = 0.23). However, the comparison between the races for GS ≥7 showed significantly higher serum MIC-1 (p = 0.031) in African American prostate cancer patients than Caucasians. A similar ANCOVA model for Gleason score, including the factors of age and race, showed no differences in the relationships between age and Gleason score for Caucasians and African Americans (p = 0.14, Figure 2B). Gleason scores differed significantly by race (p = 0.01), however, with Caucasians having lower Gleason scores (mean = 6.46, 95% CI = 6.23, 6.70) when compared to African American men of a similar age (mean = 6.91, 95% CI = 6.66, 7.17). One possible explanation for the lack of age-Gleason score association among African Americans is the lack of variability in Gleason scores for the African Americans in our samples. The range of scores varied only from 6 to 9, with greater than 80% of subjects having a 7 or higher score. An ANCOVA model utilizing a parallel-slope assumption was also used for log(PSA) based on the parallel lines found in Figure 2C. From this model, a significant racial difference was found in PSA levels (p = 0.04), with African Americans having higher PSA levels (mean = 7.54, 95% CI = 6.49, 8.85) than Caucasians across all ages (mean = 5.87, 95% CI = 5.05, 6.89). Additionally, no effect of age was found, indicating that there is no significant association between age and PSA levels, as was previously demonstrated (p = 0.18).

Analysis of urine MIC-1
To further examine the role of MIC-1 for its predictive utility in the disparity of prostate cancer, we had access to the limited number of urine samples collected at the time of diagnosis from prostate cancer patients (African American = 10; and Caucasians = 15), whereas healthy donors (African American; n = 14) served as the control (Figure 3). A comparative analysis among urine samples showed significantly higher urine MIC-1 level in prostate cancer patients of African American race (median = 5959.4, range = 1564.0 to 19,411.8) as compared to Caucasian patients (median = 3263.6, range = 356.8 to 10,481.8; p = 0.03) or healthy African American donors (median = 3601.8, range = 94.9 to 5273.7; p = 0.01). Similar to serum analysis, we used pairwise correlations to examine linear relationships between clinical variables (age, PSA, Gleason score, and ln(MIC-1)). In both races, no significant association of urine ln(MIC-1) between clinical variables was observed ( Table 2). This lack of correlation could be due to the limited number of urine samples.
(p = 0.14, Figure 2B). Gleason scores differed significantly by race (p = 0.01), however, with Caucasians having lower Gleason scores (mean = 6.46, 95% CI = 6.23, 6.70) when compared to African American men of a similar age (mean = 6.91, 95% CI = 6.66, 7.17). One possible explanation for the lack of age-Gleason score association among African Americans is the lack of variability in Gleason scores for the African Americans in our samples. The range of scores varied only from 6 to 9, with greater than 80% of subjects having a 7 or higher score. An ANCOVA model utilizing a parallel-slope assumption was also used for log(PSA) based on the parallel lines found in Figure 2C. From this model, a significant racial difference was found in PSA levels (p = 0.04), with African Americans having higher PSA levels (mean = 7.54, 95% CI = 6.49, 8.85) than Caucasians across all ages (mean = 5.87, 95% CI = 5.05, 6.89). Additionally, no effect of age was found, indicating that there is no significant association between age and PSA levels, as was previously demonstrated (p = 0.18).

Analysis of urine MIC-1
To further examine the role of MIC-1 for its predictive utility in the disparity of prostate cancer, we had access to the limited number of urine samples collected at the time of diagnosis from prostate cancer patients (African American = 10; and Caucasians = 15), whereas healthy donors (African American; n = 14) served as the control (Figure 3). A comparative analysis among urine samples showed significantly higher urine MIC-1 level in prostate cancer patients of African American race (median = 5959.4, range = 1564.0 to 19,411.8) as compared to Caucasian patients (median = 3263.6, range = 356.8 to 10,481.8; p = 0.03) or healthy African American donors (median = 3601.8, range = 94.9 to 5273.7; p = 0.01). Similar to serum analysis, we used pairwise correlations to examine linear relationships between clinical variables (age, PSA, Gleason score, and ln(MIC-1)). In both races, no significant association of urine ln(MIC-1) between clinical variables was observed ( Table 2). This lack of correlation could be due to the limited number of urine samples.

ROC Curve Analysis of Circulatory MIC-1
To examine the potential of MIC-1 as a diagnostic tool for prostate cancer between the races, we compared the serum and urine MIC-1 observations using the area under the receiver operating curve (AUC-ROC). The cutoff value was chosen based on the highest Youden's Index calculated based on the predicted values of sensitivity and specificity from the ROC analysis. Using a cutoff value of 904.6 pg/mL, the overall sensitivity and specificity of serum MIC-1 was 67.6% and 65%, respectively for serum MIC-1. For Caucasians, the predicted MIC-1 cutoff was 819.5 pg/mL with 54.5% sensitivity (95% CI = 40.5%, 68.0%) and 65% specificity (95% CI = 40.8%, 84.6%). In African Americans, the

MIC-1 Protein Expression in Prostate Tumor Biopsies
We further examined the expression of MIC-1 protein in archival prostate tumor specimens from both races (n = 24) using IHC. Out of 24 Caucasian prostate tumor specimens, six lacked cancer regions, while in African American specimens, three samples did not have cancer regions. Therefore, we scored MIC-1 immunostaining intensity in 18 specimens for both races ( Figure 5). Interestingly, we observed a significantly higher (p = 0.0001) MIC-1 nuclear staining in African American specimens (intensity score 11.67 ± 0.24) as compared to Caucasians (intensity score 6.06 ± 1.15), while MIC-1 cytoplasmic level was similar in both races (6.89 ± 0.54 vs 6.44 ± 0.57; p = 0.429). In African American specimens, 85-100% of cells showed MIC-1 staining. In contrast, in Caucasians, the percentage of nuclear staining cells was highly variable (ranged from 2% to 100%), and almost 100% of the cells showed weak to moderate MIC-1 cytoplasmic staining. In Caucasian specimens, 14/24 lacked benign tissue; therefore, comparison among races for the MIC-1 protein in adjacent benign tissue was omitted. For urine samples, a cutoff value of 4518 pg/mL yielded an overall sensitivity and specificity of 56% and 85.7%, respectively. In Caucasians, the MIC-1 cutoff was 5088 pg/mL with sensitivity and specificity of 26.7% (95% CI = 7.8%, 55.1%) and 92.9% (95% CI = 66.1%, 99.8%), respectively. In African Americans, the cutoff value of MIC-1 was 4969 pg/mL which provided a sensitivity of 80% (95% CI = 44.4%, 97.5%) and a specificity of 92.9% (95% CI = 66.1%, 99.8%) ( Figure 4B). A comparison of AUC-ROC analysis between the races showed higher sensitivity and specificity of MIC-1 as a potential diagnostic tool for African American patients than for Caucasian prostate cancer patients.

MIC-1 Protein Expression in Prostate Tumor Biopsies
We further examined the expression of MIC-1 protein in archival prostate tumor specimens from both races (n = 24) using IHC. Out of 24 Caucasian prostate tumor specimens, six lacked cancer regions, while in African American specimens, three samples did not have cancer regions. Therefore, we scored MIC-1 immunostaining intensity in 18 specimens for both races ( Figure 5). Interestingly, we observed a significantly higher (p = 0.0001) MIC-1 nuclear staining in African American specimens (intensity score 11.67 ± 0.24) as compared to Caucasians (intensity score 6.06 ± 1.15), while MIC-1 cytoplasmic level was similar in both races (6.89 ± 0.54 vs. 6.44 ± 0.57; p = 0.429). In African American specimens, 85-100% of cells showed MIC-1 staining. In contrast, in Caucasians, the percentage of nuclear staining cells was highly variable (ranged from 2% to 100%), and almost 100% of the cells showed weak to moderate MIC-1 cytoplasmic staining. In Caucasian specimens, 14/24 lacked benign tissue; therefore, comparison among races for the MIC-1 protein in adjacent benign tissue was omitted.

Discussion
Without considering race, several studies have proposed serum MIC-1 as a prognostic and predictive biomarker. Based on a prospective cohort study of Swedish men with a confirmed diagnosis of prostate cancer, elevated serum MIC-1 predicted poor cancer-specific survival [15]. A p-Chip-based immunoassay on 70 serum samples showed that MIC-1, combined with PSA, improved the specificity of prostate cancer diagnosis [16]. The current study showed a differential level of MIC-1 both in circulation and in tumor tissues from prostate cancer patients of African American and Caucasian races and proposes circulatory MIC-1 as a biomarker to improve clinical sensitivity in the diagnosis of prostate cancer in African American men.
MIC-1 level both in tissues and in plasma is usually low under normal conditions. However, it increases during inflammation or malignancy, including in the pathogenesis of cancer [9,17,18]. A tissue microarray-based immunostaining of prostate tumors showed a stage-wise decrease in MIC-1/GDF15 expression in African American men, but there was no change in MIC-1/GDF15 expression by higher grade. The overall MIC-1/GDF15 was elevated in prostate cancer compared to benign controls [19]. Interestingly, an increase in NF-κB was observed in association with increasing grade of prostate tumors in African American men. Although this study did not differentiate the nuclear and cytoplasmic staining of MIC-1/GDF15 and NF-κB, we observed a significantly higher nuclear MIC-1 protein in prostate tumors of African American men compared to Caucasians, while the cytoplasmic MIC-1 protein remained similar in both races. Increased MIC-1 level has been linked with NF-κB activation in the prostate cancer cell lines model [20]. During the transition of disease from prostatic intraepithelial neoplasia to prostate cancer, translocation of NF-κB to the nucleus was

Discussion
Without considering race, several studies have proposed serum MIC-1 as a prognostic and predictive biomarker. Based on a prospective cohort study of Swedish men with a confirmed diagnosis of prostate cancer, elevated serum MIC-1 predicted poor cancer-specific survival [15]. A p-Chip-based immunoassay on 70 serum samples showed that MIC-1, combined with PSA, improved the specificity of prostate cancer diagnosis [16]. The current study showed a differential level of MIC-1 both in circulation and in tumor tissues from prostate cancer patients of African American and Caucasian races and proposes circulatory MIC-1 as a biomarker to improve clinical sensitivity in the diagnosis of prostate cancer in African American men.
MIC-1 level both in tissues and in plasma is usually low under normal conditions. However, it increases during inflammation or malignancy, including in the pathogenesis of cancer [9,17,18]. A tissue microarray-based immunostaining of prostate tumors showed a stage-wise decrease in MIC-1/GDF15 expression in African American men, but there was no change in MIC-1/GDF15 expression by higher grade. The overall MIC-1/GDF15 was elevated in prostate cancer compared to benign controls [19]. Interestingly, an increase in NF-κB was observed in association with increasing grade of prostate tumors in African American men. Although this study did not differentiate the nuclear and cytoplasmic staining of MIC-1/GDF15 and NF-κB, we observed a significantly higher nuclear MIC-1 protein in prostate tumors of African American men compared to Caucasians, while the cytoplasmic MIC-1 protein remained similar in both races. Increased MIC-1 level has been linked with NF-κB activation in the prostate cancer cell lines model [20]. During the transition of disease from prostatic intraepithelial neoplasia to prostate cancer, translocation of NF-κB to the nucleus was associated with biochemical recurrence of prostate cancer [21]. Pancreatic adenocarcinoma study also revealed that the activation of NF-κB directly regulates the MIC-1/GDF15 expression in tumor development [22]. Therefore, the observed high level of nuclear MIC-1 in African American prostate tumors may drive aggressive prostate cancer in association with NF-κB activation implicating a biological role of MIC-1 in the racial disparity of prostate cancer.
This study is the first exploratory report demonstrating a differential level MIC-1 in circulation (serum and urine) and tumor tissues from prostate cancer patients of racial disparity. The circulatory MIC-1 levels in serum and urine are significantly higher in prostate cancer patients of African American ethnicity, with a higher AUC-ROC than in Caucasians. These observations suggest that the combined utility of circulated MIC-1 may add to the diagnostic assessment of aggressive prostate cancer, as is often reported in African American men. The association of the PSA and Gleason scores with the clinical state of the disease has been well characterized. However, statistical analysis of age with the PSA and Gleason scores is often neglected in prostate cancer studies. A study involving 110 Brazilian Indian triable men of Macuxi ethnicity found a strong association of serum PSA (free-PSA and total-PSA) with the age groups of 60-79 year old patients [23].
Similarly, multiple studies analyzing different disease conditions, including cancer, showed MIC-1 association with age [24][25][26]. Previously, it was reported that the MIC-1 level increases in response to injury or inflammation and that the age-associated inflammation may lead to high MIC-1 in the circulation [17,27]. We also found an elevated MIC-1 level correlated with age and PSA in Caucasian prostate cancer patients. However, no correlation was observed when restricted to African American samples, suggesting MIC-1 as an independent predictor of prostate cancer in African American men. This study is limited to a small number of patient samples, specifically, urine-based observations. If validated in a large cohort of samples, circulatory MIC-1 may help to identify high-risk prostate cancer patients and to develop treatment strategies to reduce the observed cancer outcome gaps between the races.

Conclusions
Our observations suggest that serum and urine MIC-1 are significantly higher in African Americans with prostate cancer than Caucasians. Among African Americans, Gleason scores and MIC-1 levels were found to be independent of age. A comparison of serum MIC-1 level in patients within the category of GS ≥ 7 between the races showed a significantly high MIC-1 (p = 0.031) in African American patients. While an increase in nuclear MIC-1 protein in prostate tumors of African American patients suggests its biological role, high AUC-ROC (serum and urine), sensitivity, and specificity indicate the use of circulatory MIC-1 as a potential biomarker to improve the clinical diagnosis of prostate cancer in African American men. Another critical observation to be examined further is whether accumulated nuclear MIC-1 helps predict the aggressive prostate cancer for its subsequent biochemical recurrence. However, a larger cohort of study is needed to establish a cutoff range of MIC-1 as a diagnostic tool to improve the clinical sensitivity of prostate cancer.