**Survival Outcomes of Patients with Pathologically Proven Positive Lymph Nodes at Time of Radical Cystectomy with or without Neoadjuvant Chemotherapy**

**Guillaume Ploussard 1,\*, Benjamin Pradere 2,3, Jean-Baptiste Beauval 1, Christine Chevreau 4, Christophe Almeras 1, Etienne Suc 5, Jean-Romain Gautier 1, Anne-Pascale Laurenty 5, Mathieu Roumiguié 6, Guillaume Loison 1, Christophe Tollon 1, Loïc Mourey 4, Ambroise Salin 1, Evanguelos Xylinas <sup>7</sup> and Damien Pouessel <sup>4</sup>**


Received: 26 May 2020; Accepted: 22 June 2020; Published: 23 June 2020

**Abstract:** Background: To compare overall survival (OS) outcomes in pN1-3 disease at the time of radical cystectomy (RC) for muscle invasive bladder according to the neoadjuvant chemotherapy (NAC) status. Materials and Methods: This multicenter study included 450 consecutive patients undergoing RC for muscle-invasive urothelial bladder cancer with pN1-3 pM0 disease from 2010 to 2019. NAC consisted in platinum-based chemotherapy. The primary endpoint was the comparison between NAC and non-NAC in terms of death from any cause. OS was assessed using the Kaplan–Meier method and multivariate Cox proportional hazards regression was used to estimate adjusted hazard ratios. Results: Median age was 69 years. Patients receiving NAC were younger (*p* = 0.051), and more likely had downstaging to non-muscle invasive disease (10.7% versus 4.3%, *p* = 0.042). Median OS was 26.6 months. NAC patients had poorer OS compared with those who did receive NAC (Hazard ratio (HR) 1.6; *p* = 0.019). The persistence of muscle-invasive bladder in RC specimens was also significantly associated with OS (HR 2.40). In the NAC cohort, the two factors independently correlated with OS were the number of positive lymph nodes (*p* = 0.013) and adjuvant chemotherapy (AC) (HR 0.31; *p* = 0.015). Conclusions: Persistent nodal disease in RC specimens after NAC was associated with poor prognosis and lower OS rates compared with pN1-3 disease after upfront RC. In this sub-group of NAC patients, AC was independently associated with better OS.

**Keywords:** bladder cancer; nodal disease; pN1; radical cystectomy; neoadjuvant; adjuvant; chemotherapy

#### **1. Introduction**

Muscle-invasive bladder cancer is a highly aggressive disease with poor oncologic outcomes in case of lymph node involvement. Neoadjuvant chemotherapy (NAC) prior to radical cystectomy (RC) has proven to improve survival outcomes in localized muscle-invasive bladder [1–3]. Level I evidence demonstrates a survival advantage of 5% as well as complete response on both primary and nodal tumor tissues [3]. The pN0 rate after NAC in cN+ patients has been evaluated as high as 48% in a retrospective series of 304 patients [4]. However, in spite of this proven overall survival (OS) advantage, a certain proportion of patients did not respond to NAC and exhibited aggressive patterns at the time of deferred RC, including pN1-3 disease. Despite NAC, up to one-fifth of the patients harbored nodal disease involvement at the time of RC [5]. However, the differential outcomes of pN1-3 patients stratified by the use or not of NAC is not well established. Moreover, there is little evidence and no firm recommendation on how to treat patients with positive lymph nodes after RC, especially after NAC administration [6]. In that setting, the use of adjuvant chemotherapy (AC) and of platinum-based regimens could be limited by potential tumor cells resistance and cumulative toxicity. Thus, whereas the impact of NAC on survival outcomes of cN1-3 patients prior to RC has been assessed in retrospective trials, to our knowledge, no series has compared OS between NAC and non-NAC patients harboring pN1-3 disease at the time of RC, and therefore the potential benefit of AC administration in that setting [4,7]. Studies comparing oncologic outcomes of pN1-3 disease according to the NAC status are biased by the selection, in the NAC group, of patients who did not respond to chemotherapy given persistent or progressing node disease after NAC. This selection bias based on resistance to neoadjuvant therapy has to be considered but helped to understand the need for aggressive post-RC treatment or monitoring in case of NAC failure.

#### **2. Materials and Methods**

#### *2.1. Patients*

We included 450 consecutive patients that underwent radical cystectomy (RC) for muscle-invasive urothelial bladder cancer with pathologically proven nodal disease from 2010 to 2019 at two institutions. After institutional review board approval (IRB number: 00006477 2017-016; review board: CEERB Paris Nord), all patients gave their written informed consent to participate in the prospective assessment of the outcomes (personal data collection and analysis). All RC were planned for cT2-4 cM0 disease, and we only included patients with pN1-3 disease. Clinical stage showed cT3 and cT4 disease in 31% and 20% of NAC patients, and 30% and 13.8% of non-NAC patients, respectively (48.2% of missing data for that variable). Patients with distant metastases (pM1a-b) on the pre-operative computerized tomography (CT) scan were excluded from analysis. The CT scan was systematically performed at the time of diagnosis. RC was performed less than 6 weeks after the diagnosis or less than 6 weeks after the last cycle of NAC. In case of NAC, another CT scan was performed before RC to confirm the absence of progression during NAC which would contra-indicate surgery. NAC and AC consisted of platinum-based chemotherapy. All patients treated by NAC received MVAC (methotrexate-vinblastine-doxorubicine-cisplatin) or GC (gemcitabline-cisplatin) regimen. AC was defined as a chemotherapy regimen given after RC before any sign of post-surgery progression, and platinum-based chemotherapy was the regimen of choice in the absence of contra-indication. Chemotherapy regimen and number of cycles were administered at clinician discretion in accordance with institutional standards and on individual decision-making. Patients treated with adjuvant radiotherapy or a combination of radiation and chemotherapy were excluded. All pathology data, including TNM stage, tumor grade, presence of positive soft tissue margin, total number of removed lymph nodes (LN), and number of LN+ were obtained from the pathological reports. Clinicopathological characteristics, surgical and adjuvant treatments, and follow-up data were collected in medical records. The chemotherapy status (NAC, AC) was recorded.

#### *2.2. Primary and Secondary Endpoints and Statistics*

The primary endpoint was the comparison between NAC and non-NAC in terms of death from any cause. Overall survival (OS) was assessed from the date of surgery until the date of death. OS was estimated using the Kaplan–Meier method and was compared using log-rank analysis. OS rates were calculated with 95% confidence intervals. Multivariate Cox proportional hazards regression was used to estimate adjusted hazard ratios with 95% confidence interval. The limit of statistical significance was defined as *p* < 0.05. The SPSS 22.0 (IBM, Chicago, IL, USA) software was used for analysis.

#### **3. Results**

#### *3.1. Clinical and Pathological Features of the Entire Cohort (n* = *450)*

Median age was 69 years with 73.1% male patients (Table 1). Downstaging to non-muscle invasive disease in RC specimens was 5.0%. Lymphovascular invasion and concomitant carcinoma in situ (CIS) were reported in 67.1% and 40.2% of cases, respectively. Soft tissue surgical margins were positive in 12.9% of the specimens. Median lymph node yield and positive lymph nodes were 16 and 2, respectively. Overall, 12.4% and 54.2% of patients received NAC +/− AC, and AC only, respectively. Among the overall cohort, 4.4% of patients received both chemo regimens. Approximately, half of patients died after a mean follow-up of 23 months. Distant systemic progression (bone and/or visceral metastases) was reported in 41.8% of patients.


**Table 1.** Overall cohort clinical and pathological characteristics (*n* = 450).

IQR = interquartile range, CIS = carcinoma in situ.

#### *3.2. Comparisons of Clinical and Pathological Features Stratified by NAC Administration*

Clinical and pathological features of both cohorts were compared (Table 2). Patients receiving NAC were younger (65 versus 68 years, *p* = 0.051), and more likely had downstaging to non-muscle invasive disease (10.7% versus 4.3%, *p* = 0.042). No significant difference was seen regarding CIS, lymphovascular invasion, positive lymph nodes, and soft tissue margin. Non-NAC patients were more frequently treated by AC (56.9% versus 35.7%, *p* = 0.003) and developed fewer systemic progression (39.1% versus 60.1%, *p* = 0.002).


**Table 2.** Comparisons between neoadjuvant chemotherapy (NAC) and non-NAC patients.

NAC = neoadjuvant chemotherapy, CIS = carcinoma in situ.

#### *3.3. Survival Analysis in the Overall Cohort*

The OS curve of the overall cohort is shown in Figure 1A. Median OS was 26.6 months. The 1-, 2-, and 5-year OS rates were 75.9% (±2.1), 54.3% (±2.7), and 29.2% (±3.2) in the overall cohort.

NAC patients had poorer OS compared with those who did not receive NAC (log rank test: *p* = 0.019, Figure 1B). The 1-, 2-, and 5-year OS rates were 66.8% (±7.3), 34.6% (±8.3), and 16.3% (±7.7) in the NAC cohort, versus 76.9% (±2.2), 56.3% (±2.8), and 30.5% (±3.5) in the non-NAC cohort. Median OS in the NAC and non-NAC cohorts was 16.7 and 28.8 months, respectively.

The OS curves were then stratified according to the type of primary chemotherapy received (Figure 1C): no chemotherapy, NAC, or AC. Patients treated by AC had better OS outcomes compared with those receiving NAC or no chemotherapy (log rank test: *p* < 0.001). Median OS was 33.6 months, compared with 22.0 and 16.7 months in the no chemotherapy and NAC cohorts, respectively. Survival curves did not differ significantly between patients who did not receive any chemotherapy and NAC patients, in spite of a trend toward better outcomes during the first 18 months after RC (*p* = 0.557). Curves crossed at this time point with better long-term outcomes in patients without any neoadjuvant or adjuvant chemotherapy regimens.

**Figure 1.** (**A**) Overall survival (OS) curve in the overall cohort; (**B**) OS stratified by the use of neoadjuvant chemotherapy (NAC); (**C**) OS stratified by the type of primary chemotherapy: neoadjuvant chemotherapy (NAC), adjuvant chemotherapy (AC), no chemotherapy.

#### *3.4. Multivariable Analysis of Predictive Factors for OS in the Overall Cohort*

Cox regression model confirmed that NAC was independently associated with overall mortality (Table 3). NAC patients had a 1.6-fold higher risk of death compared with non-NAC patients (*p* = 0.018; 95% confidence interval: 1.09–2.47). The persistence of muscle-invasive bladder in RC specimens was also significantly associated with OS (HR 2.40; 95% confidence interval: 1.06–5.44). This negative effect of NAC (*p* = 0.072) failed to reach significance when AC was taken into the multivariable model. AC was then positively and independently correlated with improved OS (HR 0.56; 95% confidence interval: 0.42–0.73; *p* < 0.001).



HR = hazard ratio; CI = confidence interval; CIS = carcinoma in situ.

#### *3.5. Stratified Survival Analysis in NAC Cohort*

Among NAC cohort, the administration of adjuvant chemotherapy was correlated with improved OS, without significant difference (Figure 2; *p* = 0.099). Median OS was 16.5 versus 31.7 months in patients receiving AC after NAC. The one-year OS rates were 61.9% (±9.7) versus 75.0% (±10.8) comparing patients who received AC and those who did not.

**Figure 2.** Survival curves for overall survival (OS) in the neoadjuvant chemotherapy (NAC) cohort stratified by the use of adjuvant chemotherapy (AC).

#### *3.6. Multivariable Analysis of Factors Associated with Overall Mortality in the NAC Cohort*

Cox regression analysis was performed in the subgroup of NAC patients (Table 3). Given the low number of patients (*n* = 56) and consequently the low number of events, we only included three factors which were the most correlated with overall mortality in univariable analyses. In the NAC cohort, the two factors independently correlated with overall mortality were the number of positive lymph nodes (>3 nodes; *p* = 0.013) and the administration of AC. AC was independently associated with a lower risk of overall mortality (HR 0.31; 95% confidence interval: 0.12–0.80; *p* = 0.015).

#### **4. Discussion**

NAC prior to RC has proven to improve survival outcomes in localized and locally advanced muscle-invasive bladder [1–3]. However, a non-negligible proportion of patients did not respond to NAC and exhibited aggressive patterns at the time of deferred RC including one-fifth of patients with nodal disease [5].

To date, there is little evidence on how to treat patients with positive lymph nodes after NAC and RC [6]. In a recent UK survey, 45% of oncologist responders would not give AC in patients with node disease after NAC and RC. Due to several factors, such as post-operative complications, impaired renal function, and poor performance status, the delivery of AC may be challenging even if an OS benefit is achieved [8]. Thus, the feasibility of re-challenging this group of NAC patients with AC is currently not well established, and patients are often offered salvage chemotherapy only at time of disease progression for palliation. A previous study of 37 patients with node positive disease after NAC previously suggested that patients who have persistent nodal disease have a very poor prognosis [9]. The two-year OS survival rate was 20%. The findings of this single-arm retrospective study highlighted a potential benefit from adjuvant chemotherapy. As reported in our series, there was a trend toward improved OS when AC was used.

While the rate of pT0 disease after NAC has been well assessed in the literature (approximately 30%), the complete response rate in node cannot be accurately evaluated due to the inaccuracy of preoperative evaluation. Indeed, node staging is currently performed by CT scan or pelvic magnetic resonance imaging (MRI). Both procedures are limited by poor sensitivity and specificity. In a series of clinical node-positive patients prior to NAC, Hermans et al. suggested that the rate of complete post-NAC response in pelvic lymph nodes (pN0) was 31% and 19% in cN1 and cN2-3 patients, respectively [7]. A complete response in lymph nodes has been evaluated at 48% in another retrospective study [4]. We were unable to assess this node downstaging rate given that we only included pN1-3 patients. However, even in patients having an aggressive disease with positive nodes at RC, our study suggests a positive impact of NAC on tumor tissue given that the pT0-1 rate was 10.8% in the NAC cohort, versus 4.3% only in non-NAC patients (*p* = 0.042). Unfortunately, given the limitations already evoked, the potential difference of response between primary cancer and metastatic nodal tissue cannot be relevantly evaluated.

The poorer OS achieved by NAC versus non-NAC patients with pN1-3 disease confirmed the need for adapting post-RC treatment in this high-risk sub-population. These patients will more frequently develop post-RC systemic progression (60.7% versus 39.1%) and die prematurely. Our findings suggest that the use of AC could be beneficial even after NAC. Indeed, OS was improved when AC was given, and AC was an independent protective factor in multivariable analysis, after taking into account positive lymph node burden and pT stage.

Consistently with French habits, MVAC was regarded in our experience as the first-line treatment of choice [10]. The pathological complete response rate achieved by dose dense MVAC appeared better than GC in retrospective studies [11]. Few patients received GC which could be preferred in other centers and/or countries due to a better toxicity profile. Comparable efficacy of GC has been emphasized, but in the metastatic setting [12]. Preliminary data from the VESPER trial (NCT01812369), comparing GC and MVAC as NAC, were presented recently, and the mature publication is awaited.

The role of AC after RC remains controversial. The main data come from underpowered trials due to poor recruitment, or from studies suffering from methodological issues. The advent of NAC before RC has also had a negative impact on enrollment in such trials [13]. The European Organisation for Research and Treatment of Cancer (NCT 30994) evaluated four cycles of immediate adjuvant chemotherapy versus six cycles of deferred chemotherapy at the time of relapse [14]. The benefit in OS was only seen in a small sub-group of pN0 patients (*n* = 86). Meta-analyses tend to confirm the reduction in the risk of death with AC (approximately 23%) [15,16]. Thus, although AC is no longer recommended, evidence suggests that it could be efficient, but mainly in chemotherapy-naive patients with locally advanced bladder cancer (pT3-4, pN0/pN +, pM0). Until now, no prospective trial has compared the sequence NAC versus NAC plus AC in patients with persistent locally advanced bladder cancer or lymph node involvement at the time of RC.

We did not report the detailed chemotherapy regimens in terms of number of cycles, toxicity data, palliative chemotherapy, and number of subsequent lines. The OS we showed could be impacted by all these parameters. Subsequent therapies for metastatic disease, that may have affected OS rates, were not available for all patients. Until recently, the only licensed second-line chemotherapy was vinflunine, which has demonstrated a three-month survival benefit with toxicity. However, the therapy landscape of advanced bladder cancer rapidly evolves. It is also worthy to note that this cohort was followed before the approval of immunotherapy regimens in advanced bladder cancer. The implementation of immunotherapy in the metastatic as well as in the neoadjuvant setting may modify the response to neoadjuvant treatment, as well as progression-free and overall survival [17]. In this study, we found that NAC patients treated by AC after RC achieved better OS outcomes compared with patients receiving only palliative chemotherapy. However, only one-third of NAC patients received AC due to poor performance status, post-operative complications, cumulative toxicity or various reasons. The possibility to change AC for adjuvant immunotherapy could increase the number of NAC patients eligible for adjuvant therapy and offer life-prolonging drug options in that particular setting of pN1-3 NAC patients.

The combination of therapy could also be an interesting option in pN1-3 disease. Zaghloul et al. recently demonstrated in a phase II study that the addition of radiotherapy to AC could improve the locoregional recurrence-free survival [18]. The trend reported in terms of OS has to be confirmed in larger phase III trials. The GETUG-AFU 30 trial (NCT03333356) is ongoing to evaluate the benefit of adjuvant radiotherapy in high-risk cancers in terms of pelvic recurrence-free survival as primary endpoint, and OS as secondary endpoint.

It seemed worthy to note that we only included in this study NAC patients who did not respond to chemotherapy given persistent or progressing node disease after NAC. This sub-group selection based on first therapy resistance explained the worse prognosis of NAC patients compared with non-NAC patients who were not selected by any type of treatment resistance. This selection bias has to be considered and helps to understand the need for aggressive post-RC treatment or monitoring in case of NAC failure.

Finally, the main limitation was the difficulty to draw any firm conclusion based on a retrospective study. In addition to potential selection biases in the selection of patients for NAC, for surgery and for AC, our results could have also been limited by the relatively small sample size. Currently, it is not possible to establish with absolute certainty what is the best sequence of perioperative treatments. However, to our knowledge, this study was the first to directly compare contemporary outcomes after RC in pN1-3 patients treated or not with NAC, and it confirmed the potential of AC even in patients already treated by NAC.

#### **5. Conclusions**

Persistent nodal disease in RC specimens after NAC is associated with poor prognosis and lower OS rates compared with pN1-3 disease after upfront RC. In this sub-group of NAC patients, AC was given to one-third of NAC patients and was an independent predictive factor for better OS outcomes. Larger prospective data as well as studies assessing the impact of other adjuvant therapies such as immunotherapy or radiotherapy are awaited.

**Author Contributions:** Conceptualization, G.P. and M.R.; data curation, G.P., B.P., J.-B.B., A.-P.L., A.S., E.X., and D.P.; formal analysis, G.P., B.P., J.-B.B., C.C., C.A., E.S., J.-R.G., M.R., C.T., L.M., A.S., E.X., and D.P.; investigation, J.-B.B., E.S., and G.L.; methodology, B.P., J.-B.B., J.-R.G., and G.L.; supervision, C.C., A.-P.L., G.L., C.T., and L.M.; validation, C.C., C.A., E.S., J.-R.G., A.-P.L., M.R., G.L., C.T., L.M., A.S., E.X., and D.P.; visualization, C.A.; writing—original draft, G.P., B.P., E.X., and D.P.; writing—review and editing, C.C., C.A., E.S., J.-R.G., A.-P.L., M.R., C.T., L.M., and A.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Di**ff**erential Expression and Clinicopathological Significance of HER2, Indoleamine 2,3-Dioxygenase and PD-L1 in Urothelial Carcinoma of the Bladder**

#### **Donghyun Kim 1,2, Jin Man Kim 1,2, Jun-Sang Kim 3,4, Sup Kim 4,\* and Kyung-Hee Kim 1,2,5,\***


Received: 6 March 2020; Accepted: 24 April 2020; Published: 27 April 2020

**Abstract:** Purpose: Evasion of the immune system by cancer cells allows for the progression of tumors. Antitumor immunotherapy has shown remarkable effects in a diverse range of cancers. The aim of this study was to determine the clinicopathological significance of human epidermal growth factor receptor 2 (HER2), indoleamine 2,3-dioxygenase (IDO), and programmed death ligand-1 (PD-L1) expression in urothelial carcinoma of the bladder (UCB). Materials and Methods: We retrospectively studied 97 patients with UCB. We performed an immunohistochemical study to measure the expression levels of HER2, IDO, and PD-L1 in UCB tissue from these 97 patients. Results: In all 97 cases, the PD-L1 expression of tumor-infiltrating immune cells (ICs) was significantly correlated with higher pathologic tumor stage (pT). In pT2–pT4 cases (*n* = 69), higher levels of HER2 and IDO expression in invasive tumor cells (TCs) were associated with shorter periods of disease-free survival (DFS). Conclusion: These results imply that the expression of PD-L1 in ICs of the UCB microenvironment is associated with cancer invasion and the expression of HER2 or IDO in the invasive cancer cell and suggestive of the potential for cancer recurrence. We suggest that the expression levels of IDO, HER2, and PD-L1 could be useful as targets in the development of combined cancer immunotherapeutic strategies.

**Keywords:** human epidermal growth factor receptor 2; indoleamine 2,3-dioxygenase; programmed death ligand-1; urothelial carcinoma; urinary bladder; immunotherapy

#### **1. Introduction**

Urothelial carcinoma of the bladder (UCB) remains one of the most common malignant cancers of the genitourinary tract [1]. Among UCB patients, approximately 30% will have muscle invasion at diagnosis, show rapid progression to metastatic disease, and succumb to their disease [2]. Although there are several different treatment regimens, very poor treatment outcomes have been reported in locally advanced and metastatic UCB patients, and this trend has remained unchanged in the last few decades [3,4]. Therefore, further studies are required to better understand the molecular mechanisms of tumor aggressiveness in UCB.

One barrier limiting the efficacy of classic cancer therapies is the interactions of cancer cells with their microenvironment, which ultimately determine whether the primary tumor is eradicated, metastasizes, or establishes dormant micrometastases [5]. Furthermore, the tumor microenvironment can also determine treatment outcome and resistance [6]. Thus, future anticancer treatment strategies should not only act directly on the proliferative processes of transformed cells but also interrupt the crosstalk circuits established by tumor cells with the host microenvironment [7].

Tumor immunogenicity is simply defined as the ability to induce adaptive immune responses [6]. Although most tumors carry particular substances which can induce an immune response, such as antigens or epitopes, the immunogenicity of cancer varies greatly between cancer types. It has been reported that tumor immunogenicity relies on its own antigenicity and several immunomodulatory mechanisms that render tumor cells less sensitive to immune system attack, or create a highly immunosuppressive tumor microenvironment [8]. Classic cancer therapies, such as chemotherapy and radiotherapy, reduce the tumor burden by killing cancer cells. Furthermore, during apoptosis and necrosis, antigens and damage-associated molecular patterns (DAMPs) stimulate an antitumor immune response which induces immunogenic cell death [9,10]. However, cancer cells can escape immune surveillance and progress through modulating immune checkpoint molecules that suppress antitumor immune responses [8].

Considering their high immunogenicity, the expression levels of immune checkpoint-related proteins have been measured and linked to the clinicopathological features and treatment outcomes in UCB. Many studies have reported that expression of immune checkpoint-related proteins, such as programmed cell death 1 (PD-1)/programmed death ligand-1 (PD-L1) and indoleamine 2,3-dioxygenase (IDO) show prognostic significance in UCB [11–13]. Additionally, various cancers have responded to treatment with immune checkpoint inhibitors, including UCB [14].

In breast cancer, Filippo et al. highlighted the role of innate and adaptive immune responses in HER2-targeted drugs [15]. This article has prompted investigations into the interaction of immune checkpoint proteins with HER2 targeted therapies. Recently, human epidermal growth factor receptor 2 (HER2) signals have been found to potentially regulate the infiltration of tumor microenvironment immune cells, and to have a role in the expression of PD-L1 in breast and gastric cancers [16,17]. Similarly, increased IDO expression was observed in a subset of HER2+ breast tumors (43.1%), which could be used to develop a combination treatment regimen [18]. These results suggest that immune-escape genes could be used to develop a combination treatment regimen in HER2 overexpression UCB patients. However, the clinical significance of immune checkpoint-related molecules in the context of HER2-positive and -negative UCB have not yet been fully evaluated.

We hypothesized that information on the expression of HER2 and immune-escape genes could be useful in the development of therapeutic strategies. This study aimed to evaluate the expression levels of HER2 and immune-escape genes by immunohistochemistry (IHC) in 97 cases of UCB. Therefore, we first evaluated the influence of immune cell infiltration on UCB survival using the Tumor IMmune Estimation Resource (TIMER) database. Then, to identify immunomodulatory genes, correlations between CD8+ T cell infiltration and candidate genes were analyzed by TIMER. Finally, we evaluated expression levels of HER2, IDO, and PD-L1 by immunohistochemistry (IHC) in 97 cases of UCB. The levels of these three protein expressions were correlated with various clinicopathological characteristics, including patient survival.

#### **2. Patients and Methods**

#### *2.1. Patients and Tissue Samples*

This study was approved by the Institutional Review Board of Chungnam National University Hospital (CNUH 2019-10-041). All formalin-fixed paraffin-embedded (FFPE) tissue samples for IHC and clinical data were obtained from the National Biobank of Korea at Chungnam National University Hospital. The requirement for informed consent for the retrospective comparison study was waived because the study was based on immunohistochemical analysis using FFPE tissue.

We conducted a review of the records of 97 patients with UCB between 1999 and 2014 at Chungnam National University Hospital in Daejeon, South Korea. The inclusion criteria were that the FFPE UCB tissues were available, and that the follow-up clinical data were sufficiently detailed. The exclusion criteria were as follows: (1) patients had a previous history of other cancers; (2) patients had received previous curative resection for any urinary tract tumor lesion; (3) patients had received preoperative chemotherapy or radiation therapy; or (4) patients had received any molecular targeted therapy. The tumor, node, and metastasis (TNM) staging and histologic grading for UCB were determined at the time of tumor resection, and were based on the 8th edition of the American Joint Committee on Cancer (AJCC) staging system [19].

The 97 UCB cases included 4 cases of noninvasive papillary urothelial carcinoma, 24 cases of pT1, 40 cases of pT2, 26 cases of pT3, and 3 cases of pT4. The 28 patients who underwent transurethral resection of the bladder (TUR-B) were in the pathologic tumor stage (pT) pTa–pT1; the 69 patients who underwent total or partial cystectomy were pT2–pT4. The histologic type of all 97 cases was conventional urothelial carcinoma. For the 69 cases of pT2–pT4, data were collected regarding their disease-free survival (DFS) and overall survival (OS) periods. Among the 69 cases, 29 patients underwent post-operative radiotherapy (PORT). DFS was determined as the time interval between the date of initial surgical resection and the date of UCB recurrence or metastasis. UCB recurrence or metastasis was determined via imaging and/or histological analysis. OS was defined as from the time of initial surgical resection to the date of death due to any cause. Without confirmation of death, recurrence, or metastasis, OS or DFS time was recorded based on the last known date that the patient was alive. We used representative FFPE whole-tissue samples of 97 UCB cases for immunohistochemistry (IHC).

#### *2.2. Immunohistochemical Staining Analysis*

Immunohistochemical staining of the FFPE tissue sample of UCB was conducted as previously described [20]. Target Retrieval Solution, pH 9 (catalog #S2368, Dako, Glostrup, Denmark), was used for antigen revitalization. The tissue sections were incubated for 30 min at room temperature with the following primary antibodies: rabbit polyclonal anti-human c-erbB-2 oncoprotein (1:200, catalog #A0485, Dako, Glostrup, Denmark), rabbit polyclonal anti-PD-L1 antibody (1:200, catalog #GTX104763, CD274 molecule, GeneTex, Irvine, CA, USA), mouse monoclonal anti-indoleamine 2,3-dioxygenase antibody, clone 10.1 (1:100, catalog #MAB5412, MERCK, Bellanca, MA, USA), CD8 (Ready-to-Use, catalog #IR623, Dako, Glostrup, Denmark), and CD43 (Ready-to-Use, catalog #IR636, Dako, Glostrup, Denmark).

We only scored HER2, IDO, and PD-L1 IHC stains for invasive urothelial carcinoma cells of 93 invasive UCB cases, while four cases of noninvasive papillary urothelial carcinoma were evaluated for intraepithelial dysplastic urothelial cells. We analyzed the cytoplasmic or cytoplasmic membrane expression of HER2 using the modified DAKO HercepTest TM Interpretation Manual—Breast Cancer Row version [21] (Staining scored 0, 1+, 2+ and 3+). Staining of 2+ or 3+ was regarded as high expression of HER2. The PD-L1 IHC staining was interpreted using the PD-L1 IHC 22C3 pharmDx Interpretation Manual—Urothelial Carcinoma [22] and VENTANA PD-L1 (SP142) Assay Interpretation Guide for Urothelial Carcinoma [23]. Any convincing partial or complete linear cytoplasmic membrane staining of viable tumor cells (TCs) exceeding 1% of the tumor cell proportion was defined as high expression of TC. Presence of discernible PD-L1, CD43, and CD8 staining of any intensity in the tumor-infiltrating immune cells (ICs) covering ≥1% of the tumor area was regarded as high expression of ICs. For CD43 and CD8, we only scored IHC staining of tumor microenvironment ICs in the muscularis propria of 61 cystectomized UCB cases among 67 cases of pT2–pT4. IDO cytoplasmic expression in TCs was scored using the method described by Allred et al. (score 0–8) [24]. A high expression of IDO was regarded as a median score or above (score ≥5). The results were examined separately and scored by Kim, K-H, and Kim, J-M, who were blinded to the patients' clinicopathological details. Any discrepancies in the scores were discussed to obtain a consensus.

#### *2.3. TIMER Database Analysis*

TIMER is a comprehensive resource for systematic analysis of immune infiltrates across diverse cancer types (https://cistrome.shinyapps.io/timer/) [25]. TIMER applies a deconvolution previously published statistical method to infer the abundance of tumor-infiltrating immune cells (TIICs) from gene expression profiles [26]. We investigated the relationship between tumor-infiltrating immune cells and UCB survival outcomes. Additionally, we analyzed the correlation of PDL1, IDO, CTLA4, CCL1, CCL2, and CCR2 expression with the abundance of CD8+ T cells.

#### *2.4. Statistical Analyses*

The correlations of the clinicopathological parameters with expressions of HER2, IDO, and PD-L1 were evaluated using Pearson's chi-square test and Fisher's exact test. The associations between HER2, IDO, PD-L1, CD43 and CD8 proteins were examined by Spearman rank correlation coefficients. Postoperative OS and DFS were determined using Kaplan–Meier survival curves and a log-rank test. The Cox proportional hazards model was applied for univariate and multivariate survival analyses. The mean values of absolute lymphocyte count (ALC), absolute neutrophil count (ANC), and neutrophil to lymphocyte ratio (NLR) were compared for the subgroups with HER2, IDO, PD-L1 (TCs), and PD-L1 (ICs) expression using an unpaired Student's t-test. Statistical significance was set at *p* < 0.05 (SPSS v.24; SPSS Inc., Chicago, IL, USA).

#### **3. Results**

#### *3.1. Association of Immune Cell Infiltration with Survival and Expression of Immune Escape Genes*

Even if there is evidence for the action of various immune cell populations in bladder cancer, a comprehensive landscape of the immune response to UCB and its driving forces is still lacking. Therefore, we tried to identify the correlation between immune cell infiltration of this cancer and survival by using the TIMER (Tumor IMmune Estimation Resource) database. In UCB, only the immune infiltrating level of CD8+ T cells was negatively correlated with survival (Figure S1). These results are in line with the tumorcidial function of CD8+ T in immune cells, which can be mitigated by the immune escape mechanism [27].

It was reported that various molecules may be involved in tumor-induced immune tolerance in UCB [28,29]. Therefore, we evaluated the correlation between CD8+ T cell infiltration of UCB and these molecules by using TIMER. Among various molecules, PD-L1 and IDO1 expression are most highly correlated with CD8+ T cell infiltration in UCB (Table S1).

#### *3.2. Association of Clinicopathological Characteristics with Expression of HER2, IDO and PD-L1*

The 97 UCB cases were evaluated using IHC to determine HER2, IDO, and PD-L1 levels. The clinicopathological characteristics of the 97 UCB patients associated with expressions of HER2, IDO, and PD-L1 are presented in Table 1. Most non-neoplastic urothelial epithelial cells or noninvasive urothelial carcinoma cells showed no expression of PD-L1, while HER2 and IDO were generally expressed with mild to moderate intensity in a large majority of reactive urothelial cells or noninvasive intraepithelial urothelial carcinoma cells, while there was no expression of IDO in normal urothelial epithelia. Invasive UCB cancer cells in lamina propria showed a relatively decreased expression of HER2 or IDO in comparison to the expression of reactive or dysplastic intraepithelial urothelial cells (Figure 1). Invasive UCB was scored using IHC stains of deeper invasive cancer lesions, except for intraepithelial lesion. However, the noninvasive papillary urothelial carcinomas were evaluated for intraepithelial dysplastic urothelial cells. Expression of HER2 or IDO in the 97 cases of UCB showed trends of decreased expression in pT2–pT4 compared to pTa–pT1 (*p* = 0.055 and *p* = 0.0007). However, PD-L1 expression of ICs was higher in pT2–pT4 than in pTa–pT1 (*p* = 0.001). HER2 expression in TCs was marginally associated with ALC /μL (*p* = 0.069). IDO expression in TCs was positively correlated with ALC /μL (*p* = 0.030) and was negatively correlated with ANC /μL (*p* = 0.007) and NLR (*p* = 0.050). PD-L1 expression in ICs was positively correlated with ANC /μL (*p* = 0.041) and NLR (*p* = 0.063) (Table S2).


**Table 1.** Correlations of HER2, IDO, and PD-L1 expressions with clinicopathological factors in 97 patients with urothelial carcinoma of the bladder.

\* Pearson's chi-square test or Fisher's exact test.

**Figure 1.** Representative images of HER2, IDO, and PD-L1 immunohistochemical staining in urothelial carcinoma of the bladder (UCB). (**A–C**) Invasive cancer cells with strongly positive expressions of HER2, IDO, and PD-L1. (**D**) Intermediate positive expression of HER2 in low-grade noninvasive urothelial tumor (left upper) and very weakly positive expression of HER2 in invasive cancer cells (right lower). (**E**) Intermediate positive expression of IDO in a low-grade noninvasive urothelial tumor (left) and strongly positive expression of IDO in a high-grade urothelial tumor (right). (**F**) Strongly positive expression of PD-L1 in intra-tumoral immune cells. (**G**) Weakly positive expression of HER2 in reactive urothelial epithelium. (**H**) Strongly positive in situ expression of IDO in urothelial carcinoma. (**I**) Negative expression of PD-L1 in reactive urothelium (scale bar = 20 μm).

#### *3.3. Correlation Between Expression of HER2, IDO, PD-L1, CD43 and CD8 Measured in Tumor Cells or Immune Cells*

The correlation between expression of the five proteins is presented in Table 2. CD43 is one of the major glycoproteins of thymocytes and T lymphocytes, suggesting a negative regulatory role in adaptive immune reactions as one of the positive markers of myeloid-derived suppressor cell phenotyping. The inverse correlation between PD-L1 expression in ICs and IDO expression in TCs was observed (*p* = 0.010). HER2 expression in TC was marginally associated with IDO expression in TCs (*p* = 0.058). There was significant positive correlation between the expression of PD-L1, CD43 and CD8 in ICs. There was a tendency to have a negative feedback phenomenon between the expression of IDO and HER2 in TC and the expression of PD-L1, CD43, and CD8 in ICs.


**Table 2.** Correlations between HER2, IDO, PD-L1, CD43, and CD8 expression according to immunohistochemical staining of urothelial carcinoma of the bladder.

\*\*, Correlation is significant at the 0.01 level (2-tailed); \*, Correlation is significant at the 0.05 level (2-tailed); TC, tumor cell; IC, immune cell.

It has been observed that expression of CD43 and CD8 in tumor microenvironment ICs is generally predominant in the lamina propria rather than the muscle layer. Since CD8 and CD43 expression showed various degrees according to the depth of tumor infiltration, intra-tumoral or contiguous peritumoral ICs in the muscularis propria and deeper layer were evaluated in 61 cases of pT2–pT4 (Figure 2).

#### *3.4. Expression of HER2 or IDO May Predict Shorter Disease-Free Survival Period in 69 Cases of pT2–pT4*

In pT2–pT4 cases (*n* = 69), we found that expression of HER2 or IDO in TCs was associated with a shorter DFS in both univariate Cox regression analysis (*p* = 0.028 and *p* = 0.048, respectively) (Table 3) and Kaplan–Meier survival curves (*p* = 0.022 and *p* = 0.040, respectively) (Figure 3). The expression of HER2 in TCs was also associated with shorter OS and DFS periods according to multivariate Cox regression analysis for HER2 expression, IDO expression, gender, age, pathologic tumor stage, and

radiation therapy after surgery (*p* = 0.031 and *p* = 0.019, respectively) (Table 4). The PD-L1 expression in TCs or ICs showed no correlation with survival outcome (Table 3 and Figure 3), even though the PD-L1 expression of ICs was higher in pT2–pT4 than in pTa–pT1 (*p* = 0.001). The expression of CD43 and CD8 in ICs showed no correlation with survival outcome. In 29 cases of pT2–pT4 with radiation therapy after surgery, the expression of HER2 or IDO in TCs showed an association with shorter DFS in Kaplan–Meier survival curves (*p* = 0.061 and *p* = 0.033) (Figure 4).

**Figure 2.** Representative images of CD43 and CD8 immunohistochemical staining in urothelial carcinoma of the bladder (UCB). Positive expression of CD43 and CD8 in intra-tumoral or contiguous peritumoral immune cells of lamina propria invasion (**A**,**B**) and muscularis propria (**C**,**D**) (scale bar = 20 μm; \*, muscularis propria; and \*\*, tumor cells).

**Figure 3.** Kaplan–Meier survival curves of disease-free survival in 69 patients with pathologic tumor stage pT2–pT4 urothelial carcinoma of the bladder according to expression of HER2 in tumor cells, IDO in tumor cells, PD-L1 in tumor cells, and PD-L1 in immune cells. (**A**) HER2; (**B**) IDO; (**C**) PD-L1 (TCs); (**D**) PD-L1 (ICs)).


**Table 3.** Univariate analysis of overall survival and disease-free survival in 69 patients with pathologic tumor stage pT2–pT4 urothelial carcinoma of the bladder.

\* univariate Cox regression analysis; HR, hazard ratio; CI, confidence interval; TC, tumor cell; IC, immune cell.

**Table 4.** Multivariate analysis of overall survival and disease-free survival in 69 patients with pathologic tumor stage pT2–pT4 urothelial carcinoma of the bladder.



**Table 4.** *Cont*.

\* multivariate Cox regression analysis; HR, hazard ratio; CI, confidence interval; TC, tumor cell; IC, immune cell.

**Figure 4.** Kaplan–Meier survival curves of disease-free survival in 29 cases with post-operative radiotherapy among 69 patients of pathologic tumor stage pT2–pT4 urothelial carcinoma of the bladder, according to expression of HER2 in tumor cells, IDO in tumor cells, PD-L1 in tumor cells, and PD-L1 in immune cells. (**A**) HER2; (**B**) IDO; (**C**) PD-L1 (TCs); (**D**) PD-L1 (ICs)).

#### **4. Discussion**

In this study, we evaluated the expression of HER2, IDO, and PD-L1 in 97 UCB cases. The three proteins showed a correlation with tumor progression or patient outcome, although they did not show the same trends for clinicopathological correlations. We demonstrated that PD-L1 expression in ICs was significantly higher in pT2–pT4 than in pTa–pT1. Increased HER2 and IDO levels in TCs of 69 pT2–pT4 cases were positively correlated with a shorter DFS period, and could be considered potential factors in poor disease outcomes.

The roles of HER2 and IDO protein in cancer initiation or progression are still poorly understood. The consistent association between the effects of anti-HER2 therapies and immune infiltration has been

reported in breast cancer and supports that an anti-tumor immune response can modulate the effect of anti-HER2 therapy [30,31]. In our study, the invasive UCB cancer cells showed a relatively reduced expression of HER2 or IDO in comparison to the expression of reactive or dysplastic intraepithelial urothelial cells. In pTa–pT1 UCBs, the expression of HER2 and IDO increased relative to that of pT2–pT4, apart from that, in pT2–pT4 cases, increased expressions of the two proteins are associated with reduced DFS expression. The altered expression of IDO or HER2 could be interpreted to be a different phase or play a different role for cancer immunoediting to the immune response against noninvasive UCB and invasive UCB [30,32,33]. Our data show a significant positive correlation between the expression of PD-L1, CD43 and CD8 in ICs. It has been observed that there is higher expression of CD43 and CD8 in lamina propria invasion in comparison to muscularis propria invasion. Moreover, there was a tendency to have a reverse correlation between the expression of IDO and HER2 in TCs and the expression of PD-L1, CD43 and CD8 in ICs. Cancer immunoediting describes a complex mechanism between ICs and TCs and has three phases: elimination, equilibrium and escape [34]. In the final escape phase, the expression of IDO in cancer cells inhibits the host immune protection. Paradoxically, IDO is elevated upon various immune molecules of adaptive or innate or tolerogenic immune cells. We speculate that elevated levels of IDO and HER2 in TC may reflect a tumor microenvironment immune reaction. And those immune-evasive transformed cancer cells may reduce IDO expression after down-regulation of immune response with a negative feedback mechanism [30,33,35]. It is predicted that in early cancer development, the expression of IDO or HER2 is upregulated in the majority of cancer cells stimulated by various immune molecules, including IFN-γ, IL-10, IL-27, CTLA4, TGF-β, cyclooxygenase-2 and prostaglandin E2, which are regulated by tumor antigen level or tolerogenic tumor microenvironment [33]. In advanced invasive cancer, the two proteins could be continuously expressed in a relatively reduced number of poorly immunogenic and immune evasive transformed cancer cells, which can lead to a poor prognosis [34]. Therefore, a spatial and periodic variety of cancer immunoediting phase could be in the same tumor mass.

In UCB, HER2 expression status has been evaluated since 1990, when overexpression of HER2 protein was first reported [36]. One study of high-grade UCB (pT2–pT4) ranked the *HER2* gene amplification as the third most significant in terms of associated genetic mutations [37]. Although the first study on the relationship of HER2 expression with clinical outcomes is confounding, a meta-analysis has indicated that its expression is associated with tumor grade, lymph node metastasis, and poor prognosis in UCB [38]. Even so, recent studies have not produced encouraging results for HER2 targeted therapy as a strategy against UCB [39–41]. The major scientific reasons for the failure of HER2 targeted therapy are a lack of standardization of HER2 testing and co-expression of other immunomodulatory molecules [42]. To overcome the poor results achieved thus far with anti-HER2 therapy, it is necessary to identify correlations between HER2 and immune checkpoint proteins in UCB. Our study reported that HER2 expression is marginally associated with IDO expression. To the best of our knowledge, this is the first study to correlate HER2 and immunosuppressive molecules in UCB.

Anti-HER2 therapy has revolutionized the treatment of malignant tumors, especially overexpressing breast cancer. Furthermore, with increasing concentrations of anticancer immunotherapy, the connection between HER2 expression and antitumor immunity has emerged as a possible target for combined oncological treatment. The whole-transcriptome profiling of HER2-positive breast carcinomas has revealed a remarkable enrichment in immune pathways [43]. HER2-positive trastuzumab-sensitive breast carcinomas have shown positive associations with chemokines involved in immune cell infiltration of the tumor microenvironment and the expression of PD-1 ligands in tumor cells [16,44]. HER2 expression has recently been found to suppress antiviral defenses and antitumor immunity as a result of HER2 signaling through its intracellular domain, which interferes with cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) pathway and prevents cancer cell death [45]. Therefore, innate and adaptive immune system responses are increasingly being acknowledged as important regulators of the effects of HER2 targeted therapy [46,47]. Based on previous research, in this study HER2 expression was scored in the cytoplasm as well as the cytoplasmic membrane to include

the immune systemic function of the intracellular domain of HER2 signaling. Considering the role of HER2 protein in interfering with antitumor immunity in the cytoplasm, the indications for HER2 targeted therapy are not limited to the cytoplasmic membrane expression of HER2 and we expect that they may also be extended to HER2 protein expression in the cytoplasm of cancer cells.

IDO, also referred to as IDO1, is one of the cytosolic enzymes that catalyzes the initial and rate-limiting steps of tryptophan to kynurenine [33,48]. IDO has been described as having immunosuppressive functions on host immune surveillance of tumor cells, with a focus on its potential immunotherapeutic targets [49]. The role of IDO has been implicated in immune tolerance related to the suppression of T-cell responses such as fetal tolerance, tumor resistance, chronic infections, and autoimmune diseases [50]. One study delineated the action of kynurenine to promote apoptosis in murine bone marrow-derived neutrophils, providing a possible mechanism for increased neutrophil accumulation in IDO-deficient mice [51]. Our results show that IDO expression is correlated with increased ALC and decreased ANC. These findings support previous studies on the immunomodulatory functions of IDO, although its effects or mechanisms in tumor progression remain unclear. IDO expression in TCs showed a negative correlation with ANC and positive correlation with ALC, while the PD-L1 expression in ICs was positively correlated with ANC in the 97 UCB cases.

Recently, phase II and preliminary phase III studies have shown that the application of a PD-L1 inhibitor in metastatic platinum-refractory NSCLC and urothelial cancer resulted in a significant improvement in the response rate and median overall survival [52]. Furthermore, PD-L1 tumor expression has emerged as a biomarker for patient stratification in immunotherapy targeting for the PD-L1/PD-1 pathway, particularly for NSCLC [53]. However, the prognostic impact of this molecule in tumor tissue is still controversial in various cancers, such as NSCLC and head and neck squamous cell carcinoma, because of the high discrepancies between PD-L1 expression and treatment outcomes [54,55]. Some studies have emphasized the significance of a comprehensive evaluation of PD-L1 expression on tumor and immune cells because its expression in immune cells, but not tumor cells, is a favorable prognostic factor for NSCLC and HNSCC [55–57]. However, our results show that PD-L1 expression in ICs is a significant poor prognostic factor in UCB.

Radiotherapy induces a host immune response by exposing tumor-specific antigens that make tumor cells detectable by the immune system, promoting the priming and activation of cytotoxic T cells [58]. Furthermore, radiation may have an impact on the tumor microenvironment by facilitating the recruitment and infiltration of immune cells [58–60]. Although radiotherapy acts as an in-situ tumor vaccine, it may be insufficient to sustain long-term antitumor immunity, resulting in later relapse [61]. Therefore, there are many studies identifying correlations between molecular regulators of tumor immune escapes and radio-resistance. PD-L1 positive cancer cells have been demonstrated to have a radio-resistant phenotype, inhibiting T cell signaling and T cell-mediated immunogenic cell death [62]. HER2 activation is a potential mechanism that may compromise the outcome of radiotherapy [63,64]. Additionally, in vitro and in vivo experiments blocking PD-L1 and IDO alongside radiation have successfully overcome rebound immune suppression [65,66]. Similarly, our data reveal that the expression of HER2 and IDO are significantly associated with DFS in UCB treated with radiotherapy after surgery (Figure 4).

#### **5. Conclusions**

The present study is the first to measure the expression levels of IDO, HER2, and PD-L1 and to analyze the correlation between these three proteins and clinicopathological values in UCB. The expression of IDO and HER2 in TCs and PD-L1 in ICs were positively correlated with poor prognostic factors in pT2–pT4 cases, including shorter DFS and OS periods or higher tumor stage. Our results suggest that the expression of IDO, HER2, and PD-L1 are useful as predictive prognostic factors and could potentially be utilized for the development of combined cancer immunotherapeutic strategies.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/5/1265/s1, Table S1: Candidate genes associated with CD8+ T cell infiltration in urothelial carcinoma of the bladder., Table S2: Correlations of HER2, IDO, and PD-L1 expressions with hematologic parameters in 97 patients with urothelial carcinoma of the bladder, Figure S1: Kaplan-Meier survival curves comparing the high and low infiltrating levels of CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells in urothelial carcinoma of the bladder.

**Author Contributions:** Conceptualization, K.-H.K.; Data curation, K.-H.K., D.K. and S.K.; Funding acquisition, K.-H.K. and S.K.; Investigation, K.-H.K., D.K. and S.K.; Methodology, S.K. and K.-H.K.; Project administration, K.-H.K.; Resources, J.-S.K., J.M.K., S.K. and K.-H.K.; Supervision, K.-H.K.; Validation, S.K. and J.M.K.; Writing–original draft, S.K. and K.-H.K.; Writing–review & editing, D.K., J.M.K., J.-S.K., S.K. and K.-H.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education, Science, and Technology (NRF-2016R1D1A1B01014311; K.-H. K. and NRF-2019M3E5D1A02068546; S. K.)

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in the design of the study; the collection, analyses, or interpretation of data; the writing of the manuscript, or in the decision to publish these results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **FOXA1 Gene Expression for Defining Molecular Subtypes of Muscle-Invasive Bladder Cancer after Radical Cystectomy**

**Danijel Sikic 1,\*, Markus Eckstein 2, Ralph M. Wirtz 3, Jonas Jarczyk 4, Thomas S. Worst 4, Stefan Porubsky 5, Bastian Keck 1, Frank Kunath 1, Veronika Weyerer 2, Johannes Breyer 6, Wolfgang Otto 6, Sebastien Rinaldetti 7, Christian Bolenz 8, Arndt Hartmann 2, Bernd Wullich <sup>1</sup> and Philipp Erben <sup>4</sup> on behalf of the BRIDGE Consortium e.V., Germany**


Received: 4 March 2020; Accepted: 31 March 2020; Published: 2 April 2020

**Abstract:** It remains unclear how to implement the recently revealed basal and luminal subtypes of muscle-invasive bladder cancer (MIBC) into daily clinical routine and whether molecular marker panels can be reduced. The mRNA expression of basal (KRT5) and luminal (FOXA1, GATA3, KRT20) markers was measured by reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR) and correlated to clinicopathological features, recurrence-free survival (RFS), disease-free survival (DFS), and overall survival (OS) in 80 patients with MIBC who underwent radical cystectomy. Additionally, the correlation of single markers with the basal and non-basal subtypes defined by a 36-gene panel was examined and then validated in the TCGA (The Cancer Genome Atlas) cohort. High expression of FOXA1 (*p* = 0.0048) and KRT20 (*p* = 0.0317) was associated with reduced RFS. In the multivariable analysis, only FOXA1 remained an independent prognostic marker for DFS (*p* = 0.0333) and RFS (*p* = 0.0310). FOXA1 expression (AUC = 0.79; *p* = 0.0007) was closest to the combined marker expression (AUC = 0.79; *p* = 0.0015) in resembling the non-basal subtype defined by the 36-gene panel. FOXA1 in combination with KRT5 may be used to distinguish the basal and non-basal subtypes of MIBC.

**Keywords:** FOXA1; GATA3; KRT20; molecular markers; mRNA; muscle-invasive bladder cancer; PCR; urothelial carcinoma

#### **1. Introduction**

Urothelial carcinoma of the bladder (UCB) is the 10th most common cancer worldwide, with an estimated 549,000 new cases and 200,000 deaths per year [1]. While the majority of patients have non-muscle-invasive bladder cancer (NMIBC), approximately 25% of patients with UCB have muscle-invasive bladder cancer (MIBC) or metastases at the time of diagnosis [2]. While radical cystectomy and platin-based chemotherapy have remained the therapeutic standard for MIBC and metastatic disease in the last few decades, the treatment and follow-up of MIBC continues to be very challenging [3,4]. Radical cystectomy is associated with high rates of perioperative morbidity and mortality [5], and approximately 50% of patients experience distant disease recurrence after cystectomy, mostly within the first two years, although later recurrences have also been reported [6–8]. The median survival of patients treated with cisplatin-based chemotherapy for metastatic disease ranges between nine and 26 months [3]. The need for lifelong surveillance as well as high treatment costs result in UCB being the most expensive cancer per patient from diagnosis to death in the US [9–11]. Given the high costs and poor outcome of patients with UCB, there is a high demand for novel molecular markers to improve diagnostics and serve as targets for new therapies.

In recent years, several independent groups have demonstrated the existence of distinct molecular subtypes in UCB comparable to the molecular subtypes in breast cancer [12–16]. It was also shown that these molecular subtypes were associated with different outcomes and responses to chemotherapy [15, 17]. While these findings bear great potential to improve the diagnostics and treatment of UCB in the future, there are still many uncertainties. Based on genetic expression patterns, most groups defined the basal and luminal subtypes of UCB by measuring the expression of hundreds of genes, which is not conveyable into daily clinical practice because of the high cost and effort [12–16]. The identification of relevant surrogate markers is necessary for the easy and feasible implementation of the molecular subtyping of UCB into daily clinical routine, as is the case in breast cancer [18].

Moreover, the exact number and definition of clinically relevant subtypes remain unclear. A recent consensus meeting agreed on the structure and features of a basal-squamous-like subtype, which is characterized by the high expression of the keratins KRT5/6 and KRT14 as well as the low expression of the transcription factors FOXA1 and GATA3 [19], which are suggested to drive luminal cell biology in bladder cancer [20]. However, to date, there has been no agreement on the definition of other non-basal subtypes or the markers necessary to define them.

Recently, using a 36-gene panel quantified by NanoString nCounter (NanoString Technologies Germany GmbH, Hamburg, Germany) in patients with MIBC treated with radical cystectomy, we were able to discriminate three prognostically distinct molecular subtypes (basal, luminal, and infiltrated) [21]. In an attempt to further reduce the required marker panel, we previously analyzed the prognostic relevance of the mRNA expression of KRT5 and KRT20 as surrogate markers for the basal and luminal subtypes of UCB, respectively [22,23]. However, it remains unclear if such a reduced marker panel adequately mirrors subtypes defined by larger marker panels.

In the present study, we investigated the association of the mRNA expression of suggested surrogate markers for the basal (KRT5) and luminal (FOXA1, GATA3, KRT20, androgen receptor (AR)) subtypes of MIBC with clinical and pathological characteristics and survival. Furthermore, the association of the surrogate markers with the subtypes defined by the previously established 36-gene panel was examined with the intent to reduce marker panels for the non-basal subtypes [21].

#### **2. Materials and Methods**

#### *2.1. Patient Population and Histological Assessment*

In this study, we retrospectively analyzed tissue samples and clinical data from 80 patients with MIBC (stage pT2–pT4) who were treated between 1998 and 2006 with radical cystectomy and bilateral lymphadenectomy at the Department of Urology of the Medical Faculty Mannheim (Mannheim, Germany). Only patients who were treated with curative intention were included. All patients with

metastases (*n* = 7) or unresectable (*n* = 1) tumors at the time of diagnosis were excluded, leaving a total of 73 patients to be included in this analysis. None of the patients received neoadjuvant or adjuvant therapy. The median follow-up time was 24 months (range: 1–184 months). All patients gave written informed consent. The study was approved by the relevant institutional review board at the Medical Faculty Mannheim under numbers 2013-517N-MA and 2016-814R-MA. Hematoxylin-eosin stained sections of the tumor samples were evaluated for pathological stage according to the 2010 TNM classification and were graded according to the common grading systems (WHO 1973, WHO 2016) by an experienced uropathologist (AH).

Expression data and clinicopathological information from the publicly available cancer genome atlas network (TCGA) cohort were used for validation (*n* = 406) [14]. Only patients with MIBC (T2–T4) were included, while all patients with no documented T or N stage and patients who received neoadjuvant therapy were excluded, leaving a total of 365 patients to be included in the analysis. Based on gene expression, the samples were clustered into five molecular subtypes (basal-squamous, luminal, luminal-papillary, luminal-infiltrated, and neuronal) [24].

#### *2.2. Assessment of mRNA Expression by RT-qPCR*

A reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR)-based assessment was used for the objective quantification of FOXA1, GATA3, and AR mRNA expression, as previously described and performed with KRT5 and KRT20 [22,25]. In brief, RNA was extracted from a single 10-μm section of formalin-fixed paraffin embedded (FFPE) routine tissue using a commercially available bead-based extraction method (Xtract® kit; STRATIFYER Molecular Pathology GmbH, Cologne, Germany). After a lysation and purification process, the nucleic acids were eluted and treated with DNase I. After the DNA was digested, the RNA eluates were stored at −80 ◦C until use.

One-step RT-qPCR was applied for the relative quantification of the mRNA expression of the genes of interest (FOXA1, GATA3 and AR) as well as the reference gene (Calmodulin 2 (CALM2)) by gene-specific TaqMan®-based assays using the SuperScript III PLATINUM One-Step, quantitative RT-PCR System (Invitrogen, Karlsruhe, Germany) on a Stratagene Mx3005p system (Agilent Technologies, Böblingen, Germany) with 30 minutes at 50 ◦C, two minutes at 95 ◦C, followed by 40 cycles of 15 seconds at 95 ◦C and 30 seconds at 60 ◦C as described previously [22,25].

Forty amplification cycles were applied, and the cycle threshold (Ct) values of the genes of interest and CALM2 for each sample were estimated as the mean value of the duplicate measurements. Ct values were then normalized against the mean expression levels of CALM2 using the 40-ΔCt method to ensure that normalized gene expression obtained by the test was proportional to the corresponding mRNA expression levels.

A set of 36 genes was previously quantified in 28 patients of this cohort using standard nCounter chemistry as previously described [21]. The nCounter assay was normalized using the geometric mean of six reference genes (CALM2, RPL37A, B2M, TUBB, GAPDH, and G6PD) and six internal positive controls, while negative background subtraction was conducted by eight negative internal controls, as previously described. Based on gene expression, urothelial carcinomas were assigned to one of three subtypes (basal, luminal, or infiltrated) [21]. Because of the small cohort size of 28 patients with available data on expression of the 36-gene panel, the subtypes were dichotomized into basal and non-basal subtypes.

The datasets for the TCGA cohort were downloaded as processed data from the open access cBioPortal database (http://www.cbioportal.org/study?id=blca\_tcga#clinical) provided by the Memorial Sloan Kettering Cancer Center (New York, NY, USA). Gene expression analyses were based on paired-end RNA-Seq analysis on an Illumina HiSeq. All RSEM (RNA-Seq by Expectation Maximization) values were log2 transformed [24].

#### *2.3. Statistical Methods*

Correlation between variables was investigated by Spearman's rank correlation coefficient, Wilcoxon/Kruskal–Wallis test, or Fisher's exact test, whichever was appropriate. In addition, the cohort was stratified into patients with high or low marker expression using the median mRNA expression of KRT5, KRT20, FOXA1, GATA3, and AR as objective cut-offs. Recurrence-free survival (RFS), disease-free survival (DFS), and overall survival (OS) were analyzed by the Kaplan–Meier method and log-rank test. Univariable and multivariable analyses were performed by a Cox proportional hazards regression model. Receiver operating characteristic (ROC) curve analyses were used to measure the correlation between the markers and molecular subtypes.

Statistical analysis was performed with JMP SAS 13.0 (SAS Institute, Cary, NC, USA) or Graph Pad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA). All *p*-values were two-sided, and a *p*-value <0.05 was considered to be significant.

#### **3. Results**

#### *3.1. Association of the Surrogate Markers with Clinicopathological Features*

The characteristics of the included patients are summarized in Table 1. The median mRNA expression of all analyzed markers is shown in Figure 1.


**Table 1.** Patient characteristics of the Mannheim cohort.

**Figure 1.** Distribution of normalized mRNA expression of KRT5, KRT20, FOXA1, GATA3, and androgen receptor (AR) in the Mannheim cohort (\*\* *p* < 0.01; \*\*\* *p* < 0.001).

Spearman correlation demonstrated a significant positive association of FOXA1 and GATA3 with the luminal marker KRT20 (Figure 2). All three luminal markers showed a significantly negative association with the basal marker KRT5. Moreover, AR also showed a significantly positive association with all luminal markers, while the association between AR and KRT5 was negative.


**Figure 2.** Spearman correlation of all measured markers (\* *p* < 0.05).

The Wilcoxon/Kruskal–Wallis test showed significant positive associations between grade and both KRT5 (0.0388) and GATA3 (*p* = 0.0133). There were no significant associations with tumor stage, nodal status, sex, or age.

#### *3.2. Association of the Surrogate Markers with Survival*

For survival analysis, the median mRNA expression levels of each marker were used as an objective cut-off to stratify patients with high and low marker expression. Kaplan–Meier analysis indicated that high KRT20 (*p* = 0.0317) and FOXA1 (*p* = 0.0048) expression was associated with significantly reduced RFS. GATA3 (*p* = 0.0629) and KRT5 (*p* = 0.0513) were not significantly associated with RFS. When analyzing the association with DFS and OS, only FOXA1 was significantly associated with reduced DFS (*p* = 0.0186) (Figure 3), while KRT5, KRT20, and GATA3 showed no associations with OS or DFS. AR mRNA expression showed no relevant associations with RFS, DFS, or OS.

**Figure 3.** Kaplan Meier analysis for recurrence-free survival (RFS) (**a**), disease-free survival (DFS) (**b**) and overall survival (OS) (**c**) based on FOXA1 mRNA expression within the Mannheim cohort.

In univariable Cox regression analysis, positive nodal status, tumor stage, and expression of FOXA1 and KRT20 were associated with worse outcome (Tables 2–4). In the multivariable analysis, of all examined markers, only FOXA1 remained an independent prognostic marker for DFS (*p* = 0.0333) and RFS (*p* = 0.0310) (Tables 2–4). When analyzing patients with pure urothelial carcinomas (*n* = 43) and patients with histologic variants (*n* = 30) separately, we found an improved survival for patients with low FOXA1 expression and pure urothelial carcinomas but not histologic variants, which might be attributed to the low number of 30 patients and high heterogeneity of the histologic variants (Figure S1).


**Table 2.** Univariable and multivariable Cox regression analyses for RFS (Recurrence-free survival), accounting for all five analyzed markers and clinicopathological features.

Multivariable analysis was performed only for significant parameters in univariable analysis (significant values in bold).


**Table 3.** Univariable and multivariable Cox regression analyses for DFS (Disease-free survival), accounting for all five analyzed markers and clinicopathological features.

Multivariable analysis was performed only for significant parameters in univariable analysis (significant values in bold).

There was no association between FOXA1 expression and DFS in the TCGA cohort (Figure S2).

#### *3.3. Correlation of Surrogate Markers with Molecular Subtypes Defined by Multigene Panels*

In the Mannheim cohort, a total of 28 patients were clustered into three molecular subtypes (basal, luminal, and infiltrated) according to the expression of a 36-gene panel previously quantified with nCounter [21]. Because of the small cohort with available data on molecular subtypes, we dichotomized the subtypes into basal and non-basal subtypes in the present study. The Mann–Whitney test showed that FOXA1 (*p* = 0.0028) and KRT20 (*p* = 0.011) expression was significantly higher in the non-basal subtype (Figure 4). The expression of KRT5 (*p* = 0.083) and GATA3 (*p* = 0.11) was not significantly different between the two subtypes in the Mannheim cohort. In the TCGA cohort, all three luminal markers had significantly higher expression in the non-basal subtype, while KRT5 was significantly higher in the basal subtype (each *p* < 0.0001).


**Table 4.** Univariable and multivariable Cox regression analyses for OS (Overall survival), accounting for all five analyzed markers and clinicopathological features.

Multivariable analysis was performed only for significant parameters in univariable analysis (significant values in bold).

Using the median marker expression as the cut-off, ROC analyses showed a high but not significant correlation of KRT5 with the basal subtype (AUC = 0.65; *p* = 0.097) in the Mannheim cohort. FOXA1 (AUC = 0.79; *p* = 0.0007) and KRT20 (AUC = 0.75; *p* = 0.0066) correlated significantly with the non-basal subtype, unlike GATA3 (AUC = 0.65; *p* = 0.097). The use of all three luminal markers combined showed no relevantly improved approximation to the non-basal subtype (AUC = 0.79; *p* = 0.0015) over the use of KRT20 or especially FOXA1 alone. Validation in the TCGA cohort showed that the use of FOXA1 alone (AUC = 0.77; *p* < 0.0001) achieved a high approximation to the non-basal subtype, similar to the use of all three markers combined (AUC = 0.79; *p* < 0.0001), while KRT5 achieved a close approximation to the basal subtype (AUC = 0.75; *p* < 0.0001). To exclude the possibility that the high association of KRT5 and FOXA1 with the basal and non-basal subtypes is mainly based on the central role of KRT5 and FOXA1 in the classification of subtypes using the 36-gene panel and the TCGA classification, we applied the BASE47 signature, which does not include KRT5 and FOXA1 for defining subtypes, on the TCGA cohort for validation [16]. This way, KRT5 and FOXA1 still showed a significantly higher distribution in the basal and luminal subtype, respectively (Figure S3). Furthermore, there was still

a high association between the basal subtype and KRT5 (AUC = 0.72; *p* < 0.0001) and the luminal subtype and FOXA1 (AUC = 0.77; *p* < 0.0001).

**Figure 4.** Distribution of FOXA1, GATA3, KRT20, and KRT5 within the basal and non-basal subtypes defined by a 36-gene panel within the Mannheim cohort showing significantly higher expression of FOXA1 and KRT20 in the non-basal subtype.

#### **4. Discussion**

With the advent of molecular subtyping in UCB, researchers and clinicians are faced with several problems, as was the case with molecular subtyping in breast cancer 15 years ago. First, apart from a basal squamous-like subtype, no consensus on the number and essential characteristics of other molecular subtypes has yet been reached [19]. As most groups have defined various numbers of luminal-like subtypes, these differences in labeling, subclassification and marker expression have hindered the general acceptance of these non-basal subtypes so far [12,14,16,21].

Second, it is necessary that subtypes are either prognostically or therapeutically relevant; otherwise, they are useless for daily clinical routine. For instance, analogous to breast cancer, a claudin-low subset of basal UCB was previously defined which demonstrated a similar outcome as regular basal UCB [16], therefore being of no interest for clinical routine.

Third, the analysis of hundreds of genes per patient currently used by most groups for the definition of their subtypes is too time-consuming and cost-intensive for easy transfer into a routine clinical setting, which is why a small set of surrogate markers per subtype has to be determined. In breast cancer, it was shown that the analysis of only four markers (estrogen receptor, progesterone receptor, HER2, and Ki-67) is enough to make a valid therapeutically relevant molecular classification [18,26]. In addition, a recent study in prostate cancer showed that the status of the PTEN gene alone matched a multigene panel to predict the risk of metastasis in patients treated with radical prostatectomy, allowing for more cost-saving diagnostics [27].

By analogy to previous findings in breast and prostate cancer, the goal of the present study was to identify surrogate markers for molecular subtypes with regard to their prognostic relevance and concordance with subtypes defined by multigene panels. Given that there is consensus about the basal subtype, we focused on non-basal subtypes. Therefore, we decided to analyze the mRNA expression of the two prominent luminal markers, FOXA1 and GATA3, together with the previously measured luminal marker KRT20 and the basal marker KRT5 [22], alongside AR as a potential target associated with the luminal subtype of MIBC [28].

Regarding the expression pattern, our results are in concordance with previous findings, as FOXA1 and GATA3 are strongly associated with KRT20 expression and show a negative association with KRT5 [12,16]. As before, we found AR to be associated with the luminal subtype in MIBC [28].

When comparing single-marker analysis to the subtypes defined by the 36-gene panel, FOXA1 and KRT20 expression was significantly higher in the non-basal subtype than in the basal subtype. Moreover, the analysis of only FOXA1 or KRT20 showed a similar high correlation with the 36-gene panel when compared to the analysis of FOXA1, GATA3 and KRT20 combined. The use of FOXA1 alone showed an almost identical AUC in the Mannheim (0.79) and TCGA (0.77) cohorts when compared to the use of all three luminal markers together (Mannheim 0.79; TCGA 0.76). FOXA1 is known to play a central role in urothelial differentiation. In addition, the low expression or loss of FOXA1 in basal tumors was described in the development of squamous cell carcinoma in preclinical models of bladder cancer, which is in concordance with the subtype association in the current study [29]. The nonsignificant correlation of GATA3 with the non-basal subtype and KRT5 with the basal subtype in the Mannheim cohort might be attributed to the small sample size of only 28 patients for whom the 36-gene panel results were available. The correlation of all markers with their respective subtype was confirmed in the TCGA cohort. The current results indicate that the measurement of one of the luminal markers, could be enough to determine a non-basal subtype, potentially allowing for more cost-saving diagnostics in daily clinical routine.

When analyzing the prognostic relevance, high KRT5 showed a non-significant trend for prolonged RFS, while FOXA1 was the only luminal marker that remained an independent prognostic marker for reduced RFS when all markers were accounted for, which suggests that a combined analysis of all three luminal markers does not provide any additional prognostically relevant information. These results are in contrast to several previous studies that found the basal subtype to be associated with worse outcome [15,17]. With regard to FOXA1 in particular, Yuk and colleagues reported higher FOXA1 expression to be associated with a positive prognostic outcome using immunohistochemistry on tissue microarray slides [30]. One possible reason for this discrepancy might be the high percentage (41%) of histologic variants within the analyzed Mannheim cohort, which are often associated with poorer prognosis than pure urothelial carcinomas and demonstrate a higher expression of luminal markers such as FOXA1 [20,31]. Furthermore, some studies show an association between the high expression of luminal markers such as KRT20 and high tumor stage, grade and micrometastasis, which are known to be associated with worse survival, which further indicates a luminal patient group with impaired survival [32,33]. On the other hand, no prognostic relevance for any of the luminal markers could be shown in the TCGA cohort. Tumor heterogeneity, which was not accounted for in the current study, might also be a factor for these contradictory results [34]. Further studies are necessary to clarify the prognostic role of molecular markers in MIBC.

Moreover, while the analysis of only one basal and one luminal marker seems to be enough to make a valid distinction between the basal and non-basal subtypes, additional markers still might be necessary for subclassification. In a comprehensive molecular analysis of MIBC within the TCGA cohort, Robertson et al. were able to identify three distinct luminal subclasses, with the luminal and luminal-infiltrated subtypes being associated with reduced survival compared to the luminal-papillary subtype [24]. All three luminal subtypes showed comparable expression patterns of FOXA1, GATA3, and KRT20 but differed with regard to FGFR3 mutations, lymphocytic infiltration, smooth muscle gene signatures, and uroplakin expression. Differences in luminal subclasses were not accounted for in our current study. The analysis of one or two additional markers (for instance, the epithelial–mesenchymal transition (EMT) markers TWIST1 or SNAI1) might be enough to draw clinically relevant conclusions. Moreover, a rare neuronal/neuroendocrine-like subtype associated with poor survival has previously been described in approximately 5% of patients with MIBC [24,35]. These tumors are mainly characterized by the

upregulation of genes of neuroendocrine origin, such as TUBB2B but can also express FOXA1 and GATA3. Markers to distinguish these neuronal/neuroendocrine-like subtypes from luminal subtypes still have to be defined.

As previously mentioned, with a total of 73 included patients, our cohort is relatively small compared to other multicentric studies. On the other hand, given that this is a single center study, we have exact information on treatment modalities, which is necessary to interpret the prognostic relevance of markers, although some data on salvage therapies is missing due to the retrospective nature of the study. However, this is also the case in the TCGA cohort, where no precise information on the treatment modality is provided, rendering statements about prognosis even more difficult.

#### **5. Conclusions**

In conclusion, we were able to demonstrate that the measurement of only one of the prominent luminal markers alongside KRT5 as the basal marker is enough to make a valid distinction between the basal and non-basal subtypes and potentially draw prognostically relevant conclusions. Given the closest concordance with subtypes defined by multigene panels as well as strongest prognostic relevance, FOXA1 seems to be the marker best suited as a surrogate marker to distinguish the non-basal subtypes from the basal subtypes. The measurement of FOXA1, GATA3, and KRT20 combined does not provide any additional relevant information. However, additional studies are necessary to further clarify the prognostic role of molecular markers in MIBC. Moreover, surrogate markers for the further subclassification of the luminal subtype still have to be defined.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/4/994/s1, Figure S1: Kaplan–Meier analysis for RFS, DFS, and OS based on FOXA1 mRNA expression within the Mannheim cohort in patients with pure urothelial carcinoma and patients with histologic variants. Figure S2: Kaplan–Meier analysis for disease specific survival based on FOXA1 mRNA expression within the TCGA cohort. Figure S3: Distribution of normalized mRNA expression of FOXA1 within the basal and luminal subtypes defined by a 47-gene panel (BASE47) within the TCGA cohort.

**Author Contributions:** Conceptualization: D.S., R.M.W., B.K., C.B., B.W., and P.E.; Methodology, D.S., M.E, R.M.W., J.J., S.P., V.W., S.R., A.H., and P.E.; Validation: M.E., T.S.W., F.K., V.W., J.B., W.O., S.R., C.B., and B.W.; Formal Analysis: D.S., R.M.W., A.H., and P.E.; Investigation, D.S., R.M.W., J.B., and P.E.; Data Curation: D.S., J.J., T.S.W., S.P., and P.E.; Writing—Original Draft Preparation: D.S., B.K., J.B., B.W., and P.E.; Writing—Review and Editing: D.S., M.E., J.J., T.S.W., F.K., V.W., J.B., W.O., S.R., C.B., A.H., and P.E.; Visualization: D.S., R.M.W., and P.E.; Supervision: T.S.W., S.P., B.K., F.K., W.O., S.R., C.B., A.H., and B.W.; Project Administration: D.S. and P.E. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** D.S. is supported by a Ferdinand Eisenberger grant from the German Society of Urology (Deutsche Gesellschaft für Urologie), grant ID SiD1/FE-16.

**Conflicts of Interest:** R.M.W. is a founder of STRATIFYER Molecular Pathology GmbH. The other authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
