Prognostic Significance of Lymphocyte Infiltration and a Stromal Immunostaining of a Bladder Cancer Associated Diagnostic Panel in Urothelial Carcinoma

We set out to expand on our previous work in which we reported the epithelial expression pattern of a urine-based bladder cancer-associated diagnostic panel (A1AT, ANG, APOE, CA9, IL8, MMP9, MMP10, PAI1, SDC1, and VEGFA). Since many of the analytes in the bladder cancer-associated diagnostic signature were chemokines, cytokines, or secreted proteins, we set out to report the stromal staining pattern of the diagnostic signature as well as CD3+ (T-cell) cell and CD68+ (macrophage) cell staining in human bladder tumors as a snapshot of the tumor immune landscape. Immunohistochemical staining was performed on 213 tumor specimens and 74 benign controls. Images were digitally captured and quantitated using Aperio (Vista, CA). The expression patterns were correlated with tumor grade, tumor stage, and outcome measures. We noted a positive correlation of seven of the 10 proteins (excluding A1AT and IL8 which had a negative association and VEGFA had no association) in bladder cancer. The overexpression of MMP10 was associated with higher grade disease, while overexpression of MMP10, PAI1, SDC1 and ANG were associated with high stage bladder cancer and CA9 was associated with low stage bladder cancer. Increased tumor infiltration of CD68+ cells were associated with higher stage disease. Overall survival was significantly reduced in bladder cancer patients’ whose tumors expressed eight or more of the 10 proteins that comprise the bladder cancer diagnostic panel. These findings confirm that the chemokines, cytokines, and secreted proteins in a urine-based diagnostic panel are atypically expressed, not only in the epithelial component of bladder tumors, but also in the stromal component of bladder tumors and portends a worse overall survival. Thus, when assessing immunohistochemical staining, it is important to report staining patterns within the stroma as well as the entire stroma itself.


Image Analysis
The slides were scanned using the Aperio Scanscope Cs (Aperio Technologies, Vista, CA, USA) as high-resolution images (20× objective). Images were visualized using Image Scope (Aperio, Vista, CA, USA). The immunoreactivity staining patterns within the stroma were noted. A prescribed algorithm developed by Aperio was used to assess staining intensity of the tissue and percent of cells/tissue staining for each target within the stromal only. Notably, we have previously used an algorithm to report the staining within the epithelial component of the tumor [28]. For statistical purposes, the data for each target was divided into quartiles: 1st quartile having the lowest staining intensity (0-10%) and the fourth quartile having the highest staining intensity (>50%). The data and slides were then reviewed and corroborated independently by an experienced pathologist (OTMC).

Statistical Analysis
The relationship between the immunoexpression of the 12 targets within the stroma and clinicopathological features were analyzed by Chi-square or Fisher's test, and all tests were two-tailed. Kaplan-Meier curves were constructed using the log-rank test to estimate and compare disease-specific survival based on the immunoexpression of stroma and epithelial. Multivariate analysis using Cox proportional hazards models for overall survival was performed to evaluate the influences of age, sex, race, tumor grade, tumor stage and immunostaining on disease-specific survival. All statistical tests were two-sided with significance set at a p value of <0.05. SAS V9.4 (Cary, NC, USA) was used to perform statistical analyses.

Demographics of the Patients and Tumor Characteristics
The age of the bladder cancer subjects ranged from 30 to 94 years (mean ± SD, 71.8 ± 11.9), while all of the control subjects were less than 65 years. Seventy-two percent of bladder cancer subjects were male and 76.5% of the cancer patients were Caucasian, while 78% of control subjects were male. Seventy patients (33%) with bladder cancer had a history of bladder cancer (recurrence). Twelve percent of bladder cancer subjects had tumors > 5 cm, while 23% had tumors < 2 cm. All tumors were confirmed to be urothelial carcinoma. The tumors were classified as either low-grade (26; 12.2%) or high-grade (176; 82.6%) as well as non-muscle invasive bladder cancer (Ta, T1, and Tis; carcinoma in situ) 132 (52%) and muscle invasive bladder cancer (T2-T4, N+, M+) 70 (32.9%). Accurate stage and grade assessment could not be performed in 11 patients (Table 1), i.e., stage was not reported in the medical records and limited pathologic specimen inhibited pathologists from grading.  Table 2. The expression of seven of the 10 bladder cancer -associated diagnostic signature (MMP9, MMP10, PAI1, CA9, APOE, SDC1, and ANG) showed a positive association with bladder cancer diagnosis, while IL8 and A1AT showed a negative correlation with cancer diagnosis. In particular, we found cancer cases expression levels in the third and fourth quartiles for MMP9 (65.4% vs. 12.2% of control), MMP10 (64% vs. 16.3% of control), PAI1 (65% vs. 12.5% of control), CA9 (63.8% vs. 11.5% of control), APOE (61.1% vs. 22.6% of control), SDC1 (64.3% vs. 12.9% of control) and ANG (68.9% vs. 0% of control) to be significantly increased compared to control. Age and race did not correlate with the expression levels of these 10 bladder cancer -associated diagnostic signature (data not shown).  The association between immunophenotype for each of the 10 targets and tumor grade is summarized in Table 3. We found expression of MMP10 in high-grade disease compared to lowgrade disease (i.e., 66.7% vs. 60.4%, respectively p = 0.016). The association between immunophenotype for each of the 10 targets and tumor stage is noted in Table 4. High stage disease was associated with increased expression (i.e., more third and fourth quartile immunostaining) for  The association between immunophenotype for each of the 10 targets and tumor grade is summarized in Table 3. We found expression of MMP10 in high-grade disease compared to low-grade disease (i.e., 66.7% vs. 60.4%, respectively p = 0.016). The association between immunophenotype for each of the 10 targets and tumor stage is noted in Table 4. High stage disease was associated with increased expression (i.e., more third and fourth quartile immunostaining) for MMP10 (60.5% Ta vs. 69.1% T2 vs. 77.7% >T2, p = 0.010), PAI1 (59.6% Ta vs. 71.5% T2 vs. 77.2% >T2, p = 0.013) and ANG (58.7% Ta vs. 75.9% T2 vs. 77.8% >T2, p = 0.003). With CA9, high stage disease was associated with reduced expression (i.e., more first and second quartile immunostaining) 42.2% Ta vs. 37.9% T2 vs. 47.4% >T2, p = 0.019. Immunoexpression of SDC1 was noted to be inversely association with tumor stage (80.9% Ta vs. 58.6% T2 vs. 40% >T2, p < 0.0001). This is similar to what our group previously reported [20].   3.3. CD3 + and CD68 + Immunohistochemical Results Figure 2 shows representative expression status for CD3 + T cells and CD68 + histiocytes in the stroma of a high-grade non-muscle invasive tumor. The relationship between immunophenotype for CD3 + and CD68 + and tumor grade is summarized in Table 3. Neither CD3 + nor CD68 + were Diagnostics 2020, 10, 14 9 of 15 associated with tumor grade. The relationship between CD68 + , not CD3 + , was associated with higher tumor stage (Table 4). High stage disease correlated with increased expression level (i.e., more third and fourth quartile immunostaining) for CD68 + (30% Ta vs. 69.5% T2 vs. 61.7% > T2). Figure 2 shows representative expression status for CD3 + T cells and CD68 + histiocytes in the stroma of a high-grade non-muscle invasive tumor. The relationship between immunophenotype for CD3 + and CD68 + and tumor grade is summarized in Table 3. Neither CD3 + nor CD68 + were associated with tumor grade. The relationship between CD68 + , not CD3 + , was associated with higher tumor stage (Table 4). High stage disease correlated with increased expression level (i.e., more third and fourth quartile immunostaining) for CD68 + (30% Ta vs. 69.5% T2 vs. 61.7% >T2).

Figure 2.
Representative expression status for CD3 + and CD68 + in high-grade non-muscle invasive bladder tumor. All images were captured at 400× magnification.

Immunophenotype and Survival
The follow-up period for the cohort ranged from 1 to 82 months (median six months), and the mean survival time was 16 months. Univariate analysis indicated that only high stage disease predicted worse overall survival. Similarly, with multivariate analysis, tumor stage < T2 (HR 0.093, 95% CI 0.036-0.235, p < 0.0001) independently predicted a worse overall survival (Table 5). Using the Kaplan-Meier survival analysis with the log-rank test, we found significantly reduced overall

Immunophenotype and Survival
The follow-up period for the cohort ranged from 1 to 82 months (median six months), and the mean survival time was 16 months. Univariate analysis indicated that only high stage disease predicted worse overall survival. Similarly, with multivariate analysis, tumor stage < T2 (HR 0.093, 95% CI 0.036-0.235, p < 0.0001) independently predicted a worse overall survival (Table 5). Using the Kaplan-Meier survival analysis with the log-rank test, we found significantly reduced overall survival in subjects whose tumors expressed ≥8 of the targets in the diagnostic signature vs. <8 targets in the diagnostic signature (p = 0.0271) (Figure 3).

Discussion
Briefly, our bladder cancer -associated diagnostic signature was derived from voided urine samples. The urine was directly subjected to molecular profiling, proteomic [29,30] and genomic [31,32] for biomarker discovery. This strategy was chosen to avoid the potential drop-out of biomarkers that can occur when translating from tissue-based discovery studies [33][34][35] to biological

Discussion
Briefly, our bladder cancer -associated diagnostic signature was derived from voided urine samples. The urine was directly subjected to molecular profiling, proteomic [29,30] and genomic [31,32] for biomarker discovery. This strategy was chosen to avoid the potential drop-out of biomarkers that can occur when translating from tissue-based discovery studies [33][34][35] to biological fluids for assay development. A number of tissue-based biomarkers have translated to urinalysis [36][37][38], but translation can be affected by secretion rate, enzymatic breakdown, or the stability of the protein in the dilute media. Next, four large and unique independent cohorts (totaling 409 bladder cancer subjects and 880 controls) were used to confirm the signature [11][12][13][14]. Then, this diagnostic signature was incorporated into an electrochemiluminescence multiplex assay and validated in three large cohorts [39,40]. The bladder cancer associated diagnostic signature achieved 85% sensitivity, 81% specificity, 93% PPV, and 63% NPV (AUROC 0.8925) for non-invasive detection of bladder cancer [39]. Lastly, pooled data from 1173 patients were analyzed. The log OR for each biomarker was improved by 1.5 or greater with smaller 95% CI in a meta-analysis of the overall cohort compared with each analysis of an individual cohort. The combination of the ten biomarkers showed a higher log OR (log OR: 3.46, 95% CI: 2.60-4.31) than did any single biomarker irrespective of histological grade or disease stage of tumors [41]. Thus, our results justify further advancement of this innovative bladder cancer associated diagnostic signature. In this study, we expanded on our previous work in which we not only confirmed the presence of this urine-based diagnostic signature in epithelial of human bladder tumors. However, we also noted poor prognostic significance when human bladder tumors expressed both IL8 and A1AT.
Maintenance of both normal epithelium and neoplasm is supported by the host's stroma. The stroma mainly consists of the basement membrane, fibroblasts, extracellular matrix, immune cells, and vasculature. Although most host cells in the stroma possess certain tumor-suppressing abilities, the stroma changes during tumorigenesis. In fact, there is significant crosstalk between tumor epithelial and tumor stroma leading to tumor growth, invasion, and metastasis [42]. A better understanding of the host's stromal contribution to cancer progression will increase our knowledge about the growth-promoting signaling pathways and hopefully lead to novel therapeutic interventions targeting the tumor stroma. Of the 10 biomarkers in the diagnostic signature, seven biomarkers (MMP9, MMP10, PAI1, CA9, APOE, SDC1 and ANG) were overexpressed in bladder tumor tissue compared to control. Interestingly, IL8 and A1AT were overexpressed in control tissue compared to tumor. Many of the control tissues were associated with benign inflammatory conditions of the bladder (e.g., chronic cystitis, cystitis cystica and cystitis); thus, stromal elevation of IL8 is not surprising. Similar to the results in tumor epithelia, VEGFA expression was not associated with tumor [28]. Additionally, we found that elevated levels of ≥8 of the biomarkers were noted to be associated with a reduced 5-year overall survival (51%) compared to five-year overall survival of 89% when <8 biomarkers were present.
A number of molecular changes have been associated with development and progression of bladder cancer. Such molecular changes include: (1) upregulation of luminal-infiltrate genes, (2) upregulation of basal-like genes, (3) upregulation of luminal-like genes, (4) alterations in expression and regulation of the receptor tyrosine kinases, fibroblast growth factor receptor 3, and members of the epidermal growth factor receptor family, (5) functional down-regulation of the tumor suppressors, p53, pRb, and p16 through deletion, mutation and/or silencing, and (6) upregulation of signaling through RAS and phosphatidylinositol 3-kinase/AKT pathways [9,[43][44][45]. Consequently, better understanding the molecular mechanisms associated with initiation, promotion, and progression of bladder cancer is pivotal to effectively prevent, diagnose, and treat bladder cancer.
Since many of the biomarkers in the bladder cancer-associated diagnostic panel are known immunomodulators, we assessed the bladder tumors for the presence of CD3 + cells (pan-T cell) and CD68 + cells (macrophage) as a snapshot of the tumor immune landscape. High CD3 + cells were found to be a good prognostic factor in both NMIBC and MIBC. We could not confirm the results of Sjodahl et al. [46], who demonstrated that increased infiltration of CD3 + was associated with improved outcomes in NMIBC and MIBC. Furthermore, we noted CD68 + cells to be associated with higher stage disease. This is in agreement with the study by Sjodahl [46]. Unlike the study by Sjodahl et al., we could not confirm that a high ratio between CD68 + and CD3 + was associated with poor overall survival. We noted macrophage infiltration to be associated with tumor stage, suggestive of its importance in bladder cancer progression and metastasis. To confirm the importance of infiltrating T cells and macrophages, future studies should address the subtypes of T cells and macrophages present in the microenvironment of the bladder tumor since recent studies suggest that during carcinogenesis. For example, CD8 + T cells that infiltrate lung tumors were reported to be dysfunctional due to microenvironmental factors, which led to reduced numbers of effector CD8 + T cells [47]. Tumor macrophages may polarize to M1 (anti-tumorigenic) or M2 (contributing to carcinogenesis), thus demonstrating different effects on the tumor [48].

Conclusions
Bladder cancer management is hampered by lack of diagnostic or prognostic markers capable of (a) predicting treatment response and (b) predicting the likely disease course. The molecular characterization of bladder cancer by our diagnostic signature may risk stratify individuals with bladder cancer who have a poor prognosis. The prognostic value of our signature must be independently validated.