PD-1, PD-L1 and cAMP immunohistochemical expressions are associated with worse oncological outcome in patients with bladder cancer

Purpose In this study, we aimed to identify prognostic factors of cancer mortality in patients who received radical cystectomy and to identify genomic alterations in a sub-cohort of patients with locally advanced (pT3-4) and/or positive lymph nodes bladder cancer (BC). Methods We collected 101 BC samples from 2010 to 2018 who previously received radical cystectomy. Immunohistochemical slides were evaluated for PPAR, cAMP, IMP3, Ki67, CDK4, POU5F1, Cyclin E and MDM2, p65, CD3, CD4, CD8, CD20, CD68, CD163, FOXP3, PD-1 and PD-L1 expression. We calculated a prognostic score (PS) based on the positivity to PD-1, PD-L1 and of cAMP (final score ranging from 0 to 3). DNA of each sample have been used for sequencing by NGS in a sub-cohort of 6 patients with locally advanced (pT3-4) and/or positive lymph nodes BC. Results PD-1 + (HR [hazard ratio] 2.59; p = 0.04), PD-L1+ (HR = 6.46; p < 0.01) and cAMP+ (HR 3.04; p = 0.02) were independent predictors of cancer-specific mortality (CSM). Increase of PS (score = 0 as reference) was associated with CSM, 0.81 (p = 0.80), 4.72 (p = 0.01) and 10.51 (p < 0.0) for PS 1, 2 and 3, respectively. ERBB2 was the gene most frequently mutated. Conclusion BC exhibited heterogenous protein expression and variable genomic features. Identification of expression of PD-1, PD-L1 and cAMP could help in predicting oncological outcomes. Supplementary Information The online version contains supplementary material available at 10.1007/s00432-022-04262-0.


Introduction
Bladder cancer (BC) represents the fourth most frequent neoplasm in males and the eight cause of cancer mortality in 2021 (Siegel et al. 2021).
Nevertheless, bladder cancer appears to be a heterogeneous disease since it is characterized by the presence of various pathological prognostic factors. However, the underlying mechanisms are still being discovered and, therefore, BC characterization studies are critical.
Interestingly, Wang et al. found that expression of altered cell cycle regulators (p53, Rb, and p21) was associated with increased risk of cancer recurrence .
For this reason, nowadays, several studies aimed to develop BC characterization, through cytogenetic examination, PCR, hybridation, DNA extraction and consequent genome profiling. Among these, also immunohistochemistry (IHC) represents the most widely used evaluation technique and especially when is coupled to tissue microarray (TMA) technology, they may obtain rapid and reliable results (Broggi et al. 2021a).
The identification of different prognostic factors has in the past been used to predict the prognosis of BC, but even today some pathological mechanisms are not clear and further information is needed to implement current clinical practice (Matsushita et al. 2011).
Giorgio Ivan Russo and Nicolò Musso have equally contributed.
* Giorgio Ivan Russo giorgioivan.russo@unict.it Extended author information available on the last page of the article Together with these considerations, genomic characterization of BC may help to identify alterations associated with prognosis and novel therapeutic targets and biomarkers to predict outcomes. Ross et al. demonstrated a high frequency of clinically relevant genomic alterations among urothelial carcinomas of the bladder, with 93% of tumor specimens exhibiting at least one actionable alteration with a mean of 2.6 potentially actionable alterations per tumor. The most common potentially clinically relevant genomic alterations found include the genes, CDKN2A (34%), FGFR3 (21%), PIK3CA (20%), and ERBB2 (17%), a finding similar to the TCGA experience with somewhat different frequencies (Ross et al. 2016).
Based on all these premises, we aimed to identify prognostic factors in patients who received radical cystectomy and to identify genomic alterations in a random sub-cohort of patients with locally advanced (pT3-4) and/or positive lymph nodes disease BC.

Materials and methods
We collected 101 BC samples from 2010 to 2018 who previously received radical cystectomy at the Urology section of the University of Catania. This study received consent from the participants and it was conducted in accordance with the Declaration of Helsinki. The study was approved by the Local Ethics Committee (#05/2020). From each patient, we have collected clinical and pathological features. Highrisk BC was considered as the presence of locally advanced (pT3-4) and/or positive lymph nodes disease while low-risk BC as the presence of organ confined (pT1-2) and negative lymph nodes, as reported by the European Association of Urology Guidelines on Muscle-invasive and Metastatic Bladder Cancer ).

Tissue microarray construction
All specimens were stained by hematoxylin and eosin to find the representative cores for the tissue microarray construction. For the TMA construction, the Galileo TMA CK3500 (Integrated System Engineering, Milan, Italy) was used as previously described. This instrument is associated with an X-Y-Z automated stage that allows one to directly place selected tissue cores in the recipient TMA block containing premade holes, ensuring not only a significant reduction in the array construction time but also an extreme alignment accuracy (Cardano et al. 2013).
We evaluated the immunopositivity for cAMP, IMP3, Ki67, CDK4 (Cyclin-dependent kinase4), POU5F1, Cyclin E, MDM2, p65, FOXP3 and PD-L1 on bladder tumor cells, while the immunoreactivity for PD-1, CD3, CD4, CD8 and CD20 has been assessed on tumor infiltrating lymphocytes. Finally, immunostainings for CD68 and CD163 have been evaluated on tumor-associated macrophages. Cells were considered as positive if brown chromogen was detected within cytoplasms/cell membranes. DNA analysis in a random sub-cohort of patients (n = 6) with locally advanced (pT3-4) and/or positive lymph nodes BC Samples were delivered in the form of paraffinized tissue samples. For each patient, a portion of healthy tissue and a portion of tumor were removed.
The deparaffinization was carried out following the protocol provided by Qiagen with modifications. The paraffin drums containing the samples were initially treated in such a way as to eliminate any excess of pure paraffin. They were then incubated in 400 µL of Xylene for 8 min and vortexed. The suspension obtained was centrifuged for 90 s at maximum speed. The supernatant was removed, and the procedure was repeated a second time. The samples were immersed for 5 min in 400 µL of Absolute Ethanol, subjected to vertexing and centrifuged at maximum speed for 90 s, after which the supernatant was removed. This step was also repeated a second time. The last deparaffinization step involves immersion in 400 µL of Nuclease Free Water for 5 min, vertexing and centrifuging at maximum speed with subsequent removal of the supernatant. Repeated this step, the recovered tissue was stored at − 80° C.
DNA was extracted from the tissue obtained with Deparaffinization Procedure using the QIAamp DNA Mini Kit (Ref. 51304, QIAGEN, 40724 Hilden, Germany) according to the manufacturer's instructions. DNA was quantified using the fluorimeter Qubit dsDNA BR Assay Kit (Invitrogen, Carlsbad, CA, USA; cod. 32850). The results of quantification are shown in Suppl. Table 1.

NGS sequencing
Eighty nanograms of DNA of each sample have been used for sequencing by NGS. The NGS sequencing was carried out according to the manufacturer's instructions provided in QIAseq Targeted

Statistical analysis
The mean and median of tissue protein stainings, expressed as a percentage of immunoreactive cells, will be calculated. Continuous variables will be presented as median and interquartile range (IQR) and were compared by the Studentindependent t test or the Mann-Whitney U test based on their normal or not-normal distribution, respectively (normality of variables' distribution was tested by the Kolmogorov-Smirnov test). Categorical variables will be tested with the Chi-square test.
CSM was defined as time from random assignment to death resulting from bladder cancer, and overall mortality as time from random assignment to death resulting from any cause (Kamat et al. 2016).
Kaplan-Meier curves were applied to calculate cancerspecific mortality and log-rank test was used to verify statistical significance between curves.
Cox regression analysis has been performed to test independent variables associated with cancer-specific mortality. All statistical analyses were completed using Stata software, version 14 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP.).
Mutational analysis was performed using QIAGEN CLC Genomics Workbench 21 software and following the User Manual for CLC Genomics Workbench 21.0.3, released on January 25 2021 (QIAGEN Aarhus, Silkeborgvej 2 Prismet DK-8000 Aarhus C, Denmark). In this step, the fast data generated by the Miseq run are processed, aligned to the reference genome hg39, filtered for quality parameters and exported as non-annotated CSV.
Further information on the mutations obtained, the amino acid changes, the impact on the protein structure and the dbSNP_IDs, was used the online tool PROVEAN®, to provide predictions for any type of protein sequence variations including the following.

Results
Supplementary Table 2 shows the IHC positivity of each prognostic factor, while supplementary Fig. 1 shows the expression of markers in bladder cancer tissue.
We found that patients with positive expression of PD-1, PD-L1 and of cAMP had worse cancer-specific mortality ( Fig. 1) and these findings were also present in the subcohort of patients with locally advanced (pT3-4) and/or positive lymph nodes disease BC (Fig. 2). Supplementary  Fig. 2 shows Cancer-specific mortality according to CD163, b-catenin, CD8, CD4, CD68, CD20, PPAR-gamma, FOXP3, CDK4 and IMP3 in the whole cohort.
We calculated a prognostic score (PS) based on the positivity to PD-1, PD-L1 and of cAMP, assigning 1 point based on the positivity on the immunostaining using and then summed, to calculate final score ranging from 0 to 3.
Slides for POU5F1, Cyclin E and MDM2 have been not considered because we did not find positivity in any tumor.

Genome sequencing
Sequencing Quality Analysis are shown in Supplementary  Table 2.
All the non-synonymous mutations sequenced with the available panel that have satisfied the filters in terms of coverage and frequency are shown in Suppl. Table 3. The frequency of the sequenced mutations for which there is a reference in the literature is instead reported in Suppl. Table 4.
The sequencing in several regions of the 23 genes of interest showed 8 of them often mutated in both cancer and mucosa. These mutations were found in the following genes: KRAS, KIT, ESR1, EGFR, ALK, PIK3CA, TP53 and ERBB2, from the lower to the higher number of mutations reported in each gene, respectively. The analysis output exhibited ERBB2 gene as the most mutated, in particular P1170A, c.3508C > G, rs1058808, was presented in every sample except for patient 2 which showed a departed group of mutations.
Of all found SNP mutations in this sub-cohort, 36 have a known dbSNP_ID, which is a unique identifier assigned to a single nucleotide polymorphism (SNP) when it is submitted to the SNP database. The most frequent mutations are those identified with the dbSNP_ID rs1042522 and rs1058808, with a frequency of 83.3% (5 out of 6 patients). Mutations rs2227983, rs7766585 and rs1042522 were found in 50% of patients. The rs61764370 mutation has a frequency of 33.3% while all remaining mutations have a frequency of 16.7%, found in individual patients (Suppl. Table 5).

Discussion
Herein, we reported that the expression of the PD-1, PD-L1 and cAMP was associated with a worse prognosis and in particular with a greater mortality in patients with locally advanced (pT3-4) and/or positive lymph nodes disease. In fact, although radical surgery is performed with curative intent, more than 50% of patients with pathological evidence of cancer invading through the muscularis propria or involving the regional lymph nodes will have lethal metastatic recurrence .
These results, although preliminary, may offer new future scientific insights with the aim of highlighting possible predictors of prognosis in particular patients with worse adverse pathology. These findings, in fact, may serve for several aspects including the use of these markers in the context   Many recent evidences, together with several clinical trials, have highlighted the importance of immune system receptors for the prognosis and progression of bladder cancer.
The molecular basis for the intrinsic control of immune effector cells is based on groups of activatory (for example, CD28 and CD137) and inhibitory (for example, PD-1, cytotoxic T lymphocyte antigen 4 (CTLA-4) and B and T lymphocyte attenuator (BTLA; also known as CD272)) receptors on the cell surface (Schneider et al. 2019).
Moreover, the impact of tumor infiltrating lymphocytes (TILs) on survival was confirmed in various cancer types but the some controversies are still present (Huang et al. 2018). The evidences suggest that TILs are the representation of the relationship between our immune system and tumor microenvironment and it includes macrophages, neutrophil granulocytes, dendritic cells, mast cells, natural killer cells, naive and memory B lymphocytes and effector T cells (T helper cells; regulatory T cells; and cytotoxic T cells) (Fridman et al. 2012). Huang et al. in fact reported that patients with intense TILs were independently associated with better OS (Huang et al. 2018).
In contrast, we evaluated the expression of CD3 + and CD8 + TIL subpopulations together with FOXP3 for regulatory T cells and CD68 for macrophages and we reported that CD3 + patients were associated with worse outcomes as reported in Fig. 1.
In a study from Wang et al., that included a cohort of patients with non-muscle invasive bladder cancer (NMIBC) or muscle invasive bladder cancer (MIBC), authors demonstrated that high level of PD-L1 expression on tumourinfiltrating immune cells was an independent predictor of worse survival (Wang et al. 2019). Importantly, high PD-1 or PD-L1 expression levels in whole tumour samples might reflect strong immune cell infiltration and activation in the tumour microenvironment, which might be linked to improved patient outcomes (Schneider et al. 2019).
Due to the poor prognosis of patients with advanced BC and in consideration that, no standard adjuvant systemic therapies have been shown to improve outcomes, more and more evidences are suggesting a putative role of adjuvant immunotherapy in this category of patients. Two recent phase 3 randomized controlled trials (RCTs)-Checkmate-274 (Bajorin et al. 2021) and IMvigor010 (Bellmunt et al. 2021)-evaluated adjuvant immune checkpoint inhibitor therapy for patients with resected high-risk MIBC.
In the intent-to-treat population of CheckMate 274, disease-free survival was 21 months in patients who received nivolumab vs about half that-10.9 months-for those assigned to placebo, a 30% reduction in recurrence or death favoring nivolumab (p < 0.001).
In the PD-L1-positive (≥ 1%) population, the diseasefree survival benefit with nivolumab was even more robust, with median disease-free survival not reached for the nivolumab-treated patients compared with 10.8 months with placebo, representing a 47% reduction in risk of recurrence or death favoring nivolumab (p < 0.001) (Bajorin et al. 2021).
On the contrary, IMvigor010 did not meet its primary endpoint of improved disease-free survival in the atezolizumab group over observation (Bellmunt et al. 2021). However, in the exploratory analysis, patients who were positive for ctDNA had improved disease-free survival and overall survival in the atezolizumab arm versus the observation arm (disease-free survival hazard ratio = 0.58; p = 0.0024, overall survival hazard ratio = 0.59). No difference in disease-free survival or overall survival between treatment arms was noted for patients who were negative for ctDNA (Powles et al. 2021).
Importantly, patients those with positive ctDNA showed correlation with high tumour mutational burden and PD-L1 from the primary tumor. Thus, the biology of tumor is strongly important to predict outcomes and response to therapy.
Our findings on cyclic AMP (cAMP) are justified by previous researches that demonstrated that cyclic nucleotides regulate a myriad of biological processes such as cell growth and adhesion, energy homeostasis, neuronal signaling, and muscle relaxation and also cancer growth (Zhang et al. 2020).
However, it is important to underline that histological heterogeneity of BC is associated with different response to systemic therapies. A typical example of the need for comprehensive biomarker assessments is represented by the role of FGFR3 alterations, which do not correlate with response to preoperative pembrolizumab in muscle invasive bladder cancer (Necchi et al. 2020). Interestingly, Necchi et al. have previously reported that FGFR3 alterations are enriched in FGFR3-active tumors based on RNA expression signature, which in turn tend to have lower immune activity, and lower PD-L1 and PD-L2 expression, suggesting that FGFR3 activity may provide a potential tool for further discriminating the mechanisms underlying the response and resistance to neoadjuvant checkpoint inhibition (Necchi et al. 2019).
Data on FGFR3 are, however, conflicting. Isharwal et al. sequenced 454 BC patients using next generation sequencing assay (MSK − IMPACT). Patients whose primary tumors harbored TP53, RB1, TP53/MDM2, and cell cycle pathway alterations more frequently presented with advanced disease whereas those with tumors containing FGFR3, and RTK/ RAS/RAF pathway alterations generally presented with NMIBC (Isharwal et al. 2018).
Importantly, Yang et al. reported that receptor tyrosine kinases (ERBB2, FGFR3, and PIK3CA) were more commonly altered in the responders (p < 0.01) compared to the non-responders to neoadjuvant chemotherapy (Yang et al. 2018). Although we did not perform receptor tyrosine Fig. 3 Cancer specific mortality according to prognostic score expression in locally advanced (pT3-4) and/or positive lymph nodes disease 1 3 kinases in IHC, we found high frequency of ERBB2 mutations in our sub-cohort of patients with locally advanced (pT3-4) and/or positive lymph nodes disease.
Despite the heterogenicity of the mutational pattern of bladder cancer, it was possible to identify the particular genes involved: HER2, TP53, PIK3CA, HRAS as well as EGFR.
Receptors tyrosine kinase (RTK) expressed by nonneoplastic cells may also become an attractive therapeutic target, which is considered as an emerging class of innate immune checkpoints (Akalu et al. 2017). In particular, immune microenvironment is also modulated by RTK and interestingly its inhibitors are less effective in patients with a tumor highly infiltrated by CD8+ lymphocytes and associated with a high level of programmed death-ligand 1 (PD-L1) (Matsumoto et al. 2019). These findings strengthen the importance of an interaction between RTK and immune checkpoints (Pottier et al. 2020).
Furthermore, there are preliminary hypothesis that that anti-PD-1 antibody given by intravesical route will effectively treat bladder tumors while having a very good safety profile with limited systemic effects (Kirschner et al. 2021). In particular, Anti-PD-1 increases CD8 cell infiltration in tumors, particularly when administered intravesically.
Before concluding, we would underline some limitations. First, we were not be able to verify if our findings were related to BCG response before radical cystectomy. Second, a subset analysis in patients receiving immunotherapy during the followup was not feasible due the low number of events. Third, we did not perform RNA analysis investigating the cross link between genomic alterations and tissue protein expression.
Fourth, although almost of our patients did not receive neoadjuvant chemotherapy, we were not able to verify its impact on assessed biomarkers and long-term survival. We also limited the NGS analysis in few patients and not in the whole cohort, but further and more complex analysis will be performed in the same study cohort. Finally, as concerning the role of TILs and its differentiation between stromal and intra-tumoral, the TMA slides are not properly correct for this evaluation since they do not represent in some cases the area in which TIL are most represented rather the full slides that are more appropriate.
On the other hand, we reported a comprehensive analysis in BC patients that demonstrated that PD + , PD-L1 + and cAMP + were associated with worse prognosis and that the most frequent mutations were those identified with the dbSNP_ID rs1042522 and rs1058808, with a frequency of 83.3%.

Conclusions
Through our immunohistochemical studies carried out on bladder cancer samples, we have shown that the expression of PD-1, PD-L1 and cAMP is able to predict mortality at 3 years in patients with locally advanced (pT3-4) and/or positive lymph nodes disease. Genetic analyses carried out in NGS also revealed how somatic mutations can involve locus of different genes, and in particular the ERBB2 alteration proved to be the most prevalent. We think that these findings give deep knowledges in urothelial cancer strengthening the importance of the use of blood-based biomarker or tissue biomarker to select patients at high risk of relapse since they are linked with cancer biology.