Background
Bladder cancer (BCa) is causing about 213 000 deaths every year [
1]. 25% of patients present with muscle-invasive bladder cancer (MIBC) at the time of diagnosis, which is associated with 5 year overall survival rates of about 50%. Radical cystectomy with lymphadenectomy is the standard therapy for patients with MIBC. Perioperative platin-based chemotherapies are recommended by guidelines to improve survival rates [
2]. However, response rates to chemotherapies are limited and besides pathological staging and platin eligibility no markers exist to select patients for the application of perioperative chemotherapy [
2]. Promising disease-free survival (DFS) rates for an adjuvant treatment with the immune checkpoint inhibitor Nivolumab led to its approval for patients with locally advanced and lymph node positive bladder cancer or patients not responding to neoadjuvant chemotherapy, when tumors show PD-L1 expression [
3]. However, the application of adjuvant atezolizumab failed to show an improvement in DFS in a similar patient cohort [
4]. In the latter study, high immune cell infiltration and basal-squamous features were associated with lower risk of recurrence [
5]. Considering novel therapeutic approaches and the progress in precision medicine, further tumor classifications and markers besides clinical and pathological factors are necessary to select patients for either chemo- or immunotherapy in the adjuvant setting. The advances in immunomodulating therapies have increased the focus on the tumor microenvironment (TME) of MIBC, which can influence the response to therapy [
5‐
12]. For example, the presence of tumor-infiltrating immune cells has been proposed as a positive prognostic factor and to be predictive for the response to perioperative chemotherapy as well as immunotherapy [
8,
13,
14]. Moreover, tumor-associated macrophages (TAMs) are abundant in solid tumors and can promote tumor progression [
15,
16]. In bladder cancer, high presence of TAMs is associated with poor prognosis [
16‐
19].
Here, we investigated the presence and the impact of subsets of tumor-associated immune cells in MIBC on survival using multiplex immunohistochemistry.
Cohort
Tissue/tumor samples and patient data used in this study were provided by the University Cancer Center Frankfurt (UCT). Written informed consent was obtained from all patients and the study was approved by the institutional review boards of the UCT and the ethical committee at the University Hospital Frankfurt (project-number: SUG-6-2018 and UCT-53-2021) which was conducted according to local and national regulations and to the Declaration of Helsinki.
A total of 145 formalin-fixed, paraffin-embedded (FFPE) tissue samples from patients with MIBC treated at the department of Urology, University Hospital Frankfurt from 2010 to 2020 were retrieved from the Dr. Senckenberg Biobank (SBB) at the Senckenberg Institute of Pathology.
Histopathology of all cases was systematically re-reviewed by two genitourinary pathologists according to current WHO-criteria [
20]. Histological subtypes were reported if at least 10% of tumor showed subtype histology including pure and mixed tumors.
Immunohistochemical (IHC) analysis
The construction of the tissue microarray (TMA) has been described before [
21]. In brief, tissue cores (1 mm) of representative tumor areas were transferred to a new block using the TMA Grandmaster (3DHISTECH, Budapest, Hungary). Hematoxylin and eosin stainings (H&E) were performed automatically (Sakura Finetek, Torrance, CA, USA) and IHC was performed using the DAKO Omnis staining system (Agilent, Santa Clara, CA, USA). We performed staining for Anti-CD68 (Clone KP1; DAKO, Agilent, Santa Clara, CA, ready to use) and CD163 (Clone MRQ-26; Cell Marque, Rockin, CA, USA, 1:100). TMA cores with absence of tumor, tissue or artefacts were excluded from the analysis. We used the open-source software QuPath (
https://qupath.github.io) to quantify cells and DAB positive cells on TMA cores and calculate the percentage of positive cells [
22]. The settings for the positive cell detection are displayed in Additional file
1: Figure S1.
The multiplex analysis was performed on 105 samples, which were stained with Opal 7‐Color Automation IHC Kits (Akoya Biosciences, Menlo Park, CA, USA) on a BOND‐RX Multiplex IHC Stainer (Leica, Wetzlar, Germany). We used the following primary antibodies on two panels: 1. T cell panel: anti‐CD163 (clone EPR19518; abcam, Cambridge, UK), anti‐CD4 (clone EPR6855; abcam, Cambridge, UK), anti‐CD3 (clone D7A6E; CellSignaling, Danvers, USA), anti‐PD-1 (polyclonal; Sigma, Darmstadt, Germany, HPA035981), anti‐CD8 (Clone C8/144B; DAKO, Agilent, Santa Clara, CA, USA), anti‐FoxP3 (clone 236A/E7; abcam, Cambridge, UK). 2. TME-panel: anti‐PD-L1 (clone SP142; abcam, Cambridge, UK), anti‐Pan-Cytokeratin (panCK) (clone C-11; abcam, Cambridge, UK) anti‐αSMA (clone 1A4; Sigma, Darmstadt, Germany), anti‐Ki67 (clone SP6; abcam, Cambridge, UK), anti-Vimentin (clone EPR3776; abcam, Cambridge, UK), and anti-CD45 (polyclonal; abcam, Cambridge, UK, ab10558). 4′,6‐diamidino‐2‐phenylindole (DAPI) (SouthernBiotech, Birmingham, AL, USA) was used for counterstaining of nuclei.
Corresponding secondary HRP-conjugated antibodies (Akoya Biosciences, Menlo Park, CA, USA, ARH1001A) and Opal fluorophores (Akoya Biosciences, Menlo Park, CA, USA, FP1500001KT, FP1487001KT, FP1488001KT, FP1495001KT, FP1497001KT, FP1501001KT) were used as described before [
23,
24]. Images were acquired with the PhenoImager HT imaging system (Akoya Biosciences, Menlo Park, CA, USA) and analyzed using the Phenotyping application of the inForm software V2.5 (Akoya Biosciences, Menlo Park, CA, USA). Briefly, after tissue (stroma/tumor) and cell segmentation, populations were identified with markers by using the phenotyping algorithm provided in the in Form software. The following cells were identified: double negative (DN) T-cells: CD3+ CD4-CD8-; T-helper cells: CD3+ CD4+ CD8-FoxP3-; cytotoxic T cells: CD3 + CD8+ CD4-; PD-1 positive cytotoxic T cells: CD3 + CD8 + PD-1 + ; Macrophages: CD163 + ; regulatory T cells (Tregs): CD3 + CD4 + FoxP3 + ; fibroblasts: αSMA + Vimentin + CD45-; CD45 + immune cells: CD45 + PD-L1-panCK-Vimentin-; tumor cells: panCK + CD45-Vimentin-; PD-L1 positive tumor cells: panCK + PD-L1 + Vimentin-CD45-; PD-L1 positive immune cells: CD45+ PD-L1+ panCK-Vimentin- proliferating immune cells: CD45+ Ki67+ ; proliferating tumor cells: panCK + Ki67 + Vimentin-CD45-. Since the vast majority of immune cells was detected in the assigned stroma area, we used cell counts of the whole tissue area to avoid incorrect tissue segmentation bias (Additional file
1: Figure S2). TMA cores with either absence of representative tumor tissue or presence of staining artifacts were excluded from the analysis.
Statistical analysis
We performed descriptive statistics of all data. Subgroup comparisons were tested with Mann-Whitney U test and Chi2 test for nonparametric variables, two-sided t-test and ANOVA for parametric variables.
For the visualization, the multiplex IHC data were standardized using JMP (SAS Institute Inc.) by the conversion of the data to a mean of 0 and a standard deviation (SD) of 1 (z-score: (z = xi − mean(x)/SD(x)).
For the survival analysis, only patients with radical cystectomy in “curative intent” were included: patients with metastatic disease or that received neoadjuvant chemotherapy were excluded from the analyses. We defined the overall survival (OS) as main endpoint of interest, which was defined as time interval between surgery and death.
We used the Kaplan-Meier method to estimate and illustrate survival probabilities as well as uni- and multivariate Cox’s proportional models to estimate the hazard ratio (HR) and the corresponding 95% confidence intervals. A significance level of α = 5% was used. Statistical analyses were performed using JMP (SAS Institute Inc.) Version 16.2.0 and R Studio (Version 2022.02.3). We used the cut-off finder to calculate the cut-off for the percentage of CD68 and CD163 positive cells with the minimum log-rank P-value method [
25].
Discussion
With the introduction of immune-checkpoint inhibitors as therapeutic option in muscle invasive and metastatic urothelial carcinoma, efforts have been made to improve management of patients [
3,
4,
26‐
29]. However, only a subset of patients responds to immunotherapies and/or are resistant to chemotherapies. Hopes were put into molecular subtypes to stratify patients, that respond from systemic therapies, but contradictory results and the diversity of molecular subtype taxonomy have hindered the use and clinical translation so far [
8,
30‐
34]. Molecular subtypes lack prospective validation and recent publications propose that high immune infiltration increases sensitivity to chemotherapy and might serve as predictive marker [
6‐
9].
By using an unbiased automated approach with 7‐color multiplex IHC to detect cells we were able to quantify different cell types and their association within a tumor. Our results clearly point out that high levels of TAMs are a negative prognostic marker in patients with MIBC undergoing radical cystectomy with and without adjuvant chemotherapy. Tumors with high macrophage and low Treg infiltration had less CD8 + T cells, but where rich in PD-1/PD-L1 expression of tumor and immune cells. This combination could cause poor survival and poor response to chemotherapy. On the other hand, Tregs, identified by FoxP3, were associated with high presence of CD8 + T cells and proliferating immune cells that have a positive effect on survival. Most patients were stratified into cluster 3, showing overall low immune cell infiltrates, having inferior OS than patients in the Treg rich cluster, but still superior to the macrophage rich cluster.
Higher stromal immune cell infiltration in MIBC has been shown to be associated with increased anti-tumor immune response, response to chemotherapy, and improved survival rates [
6‐
8,
12,
13]. While this effect is mostly attributed to cytotoxic T-cells [
7,
10,
11,
13,
18], anti‐inflammatory macrophages are considered to be tumor‐promoting and have been described as a negative prognostic predictor for multiple cancer entities [
15,
23,
35‐
37]. In bladder cancer several studies have shown TAMs to be associated with poor clinical outcome [
15,
18,
38,
39]. Recently it was shown that TAMs expressing IL10 are associated with inferior prognosis of MIBC, but these patients showed a superior response to chemotherapy [
37]. In another cohort, patients with the presence of Galectin-9 expressing TAMs had decreased OS and RFS, but an improved response to adjuvant chemotherapy [
39]. In our hands, patients in cluster 2, which is defined by high presence of macrophages, had poor outcome also when receiving adjuvant chemotherapy. It would be relevant to explore this feature in the setting of neoadjuvant chemotherapy and adjuvant immunotherapy. Furthermore, preclinical bladder cancer models suggest inhibition of macrophage recruitment via blockage of CCL2 can lead to a reduction of lymph node metastases and increased survival after chemotherapy [
40,
41]. For other tumor entities it was shown that TAM depletion by blocking colony-stimulating factor 1 receptor (CSF1R) promoted an anti-tumor immune response and lead to an enhanced response to anti PD-L1 treatment [
42‐
44].
Concerning the presence of regulatory T cells (Tregs), ambiguous results have been published. Horn et al. did not observe a correlation when using FoxP3 as a single marker for Tregs, but described a slightly shorter OS in patients with an increased FoxP3/CD3-ratio [
45]. In contrast, a small study by Winerdal et al. found higher infiltration of FoxP3 + T cells to be correlated with better survival [
46]. Although Tregs seem to have negative impact on patient survival across a number of tumor entities, this prognostic effect varies between different cancer types [
47]. In the present study, tumors in the Treg rich cluster were enriched with other immune cells like CD3+ , CD8+ , proliferating immune cells and PD-L1 expressing immune and tumor cells. Correlations of Treg and T cell scores have also described by others [
15]. Although FoxP3 is often considered a specific marker for Tregs, it has also been described as a marker of activated T cells without regulatory or immunosuppressive functions [
48‐
50]. Thus, FoxP3 positive cells might function as representatives of high immune-cell infiltration, which overall leads to favorable survival. A multiomic study by Taber et al. showed similar results regarding the spatial immune cell analysis claiming that higher PD-L1 expression leads to better response rates of neoadjuvant chemotherapy in MIBC, and immune infiltrated and excluded tumors have better treatment outcome compared to immune desert tumors [
8]. High expression of PD-1/PD-L1 can represent T cell dysfunction and is often correlated to poor prognosis in urothelial cancer [
51‐
53]. However, a lower risk of recurrence and better prognosis for patients with high PD-1 and PD-L1 has also been described [
54,
55]. Since a significant number of immune cells with PD-L1 expression was observed in cluster 1 as well as cluster 2, the underlying biology of cells expressing immune checkpoint molecules and how this influences survival needs to be further investigated in detail.
Other studies focused either on stromal or tumor infiltrating immune cells or separated tumors into immune excluded, infiltrated or immune desert [
8,
10,
18,
38]. We also separately analyzed Tregs and macrophages in tumor- and stroma-areas. Both cell types were predominantly located in the stroma and had prognostic influence when being present in the stroma, but not in the tumor (see Additional file
1: Table S3 and Additional file
1: Figure S2). However, the predictive value was higher when analyzing the whole tissue area. Furthermore, a non‐biased approach of a representative TMA core or tumor area could be more reproducible.
The application of multiplex IHC and multispectral imaging is a powerful tool to image and characterize tumors and the cellular subsets. However, it is not applicable in clinical routine and the reproducibility is dependent on staining, scanning and the cell detection algorithm. Using routine IHC-staining and digital analysis with QuPath of CD163 in an extended cohort confirmed the prognostic value of the marker. This result had been confirmed by other groups before [
15,
19,
56]. In a study by Taubert et al., CD68 was also proposed to be predictive to stratify patients, which is in line with our results, even though CD68 as a macrophage marker did not reach significance in our cohort (HR 1.5, 95% CI 0.97–2.4; p = 0.06) [
19]. This might be an issue of sample size. CD163 is a pan‐macrophage marker, which can be induced under M2 conditions, but is still detectable at protein level in other macrophages [
23,
57,
58]. It has been described as a marker of alternatively activated and therefore, presumably, tumor‐promoting macrophages [
35,
36]. Although the correlation between CD68 and CD163 was high, some cells expressed CD163 without CD68. Adding further markers of macrophage subsets might be instrumental to better characterize, e.g., M1/M2 polarization and is planned for future experiments. Moreover, further research will focus on a more detailed analysis of macrophage subsets, such as resident versus recruited macrophages or subsets expressing markers such as CD206, and their functions [
23,
59].
The application of two multiplex IHC panels followed by the algorithm-based cell analysis enabling a morphological analysis with single cell resolution together with a well characterized MIBC cohort are strengths of this study. The data allow detailed analyses of the TME to confirm TAMs as an important predictor of poor survival. However, our study has important limitations. The retrospective design and limited number of patients receiving chemotherapy limits the predictive value. Using TMAs might not reflect bladder cancer heterogeneity sufficiently. Furthermore, only one marker (CD163) was used to identify macrophages. The antibodies used for the multiplex IHC were not the same as used in pathological routine. This can explain differences between the multiplex IHC cohort and the confirmation cohort. Especially regarding the PD-L1 expression, different antibodies can cause variable results and are not transferable from multiplex IHC to clinical routine [
60]. Finally, the mode of action of the Treg rich cluster and PD-1/PD-L1 expressing cells remains to be revealed.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.