Stromal marker fibroblast activation protein drives outcome in T1 non-muscle invasive bladder cancer

Fibroblast activation protein-α (FAP) is a transmembrane peptidase and a surrogate marker for cancer-associated fibroblasts (CAFs). FAP has been linked to worse prognosis and therapy resistance in several cancers. We hypothesised that FAP might have a prognostic 3biomarker potential to stratify patients with high-grade (HG) T1 non-muscle-invasive bladder cancer (NMIBC). We selected 30 patients with HG T1 NMIBC that progressed to ≥T2 disease which were pair-matched based on CUETO progression score variables with 90 patients that did not progress. After revision a final cohort of 86 patients was retained. Slides were stained for FAP, the luminal marker GATA3 and the basal marker CK5. All HG T1 tumour regions of interest (ROIs) within each patient were annotated, analysed and scored using image analysis software. FAP expression in HG T1 ROIs was significantly higher in progressors vs. non-progressors and was prognostic for recurrence-free survival, progression-free survival, cancer-specific survival, and overall survival. FAP expression in HG T1 ROIs remained strongly prognostic for these outcomes in a bivariable model corrected for adequate BCG per FDA definition. Expression of GATA3 and CK5 did not differ between progressors vs. non-progressors, and were not prognostic for these outcomes. FAP might serve as an easily applicable prognostic biomarker to risk-stratify patients with HG T1 NMIBC if these results are prospectively validated in a larger series.


Introduction
Due to a significant heterogeneity in clinical behaviour of stage T1 non-muscle-invasive bladder cancer (NMIBC), there is an unmet need for prognostic biomarkers to aid in patient counseling and treatment decisions [1]. Fibroblast activation protein-α (FAP) is a transmembrane peptidase expressed on stromal cells and acts as a surrogate marker for cancer-associated fibroblasts (CAFs) which play a major role in the tumour microenvironment [2,3]. FAP overexpression in CAFs has been linked to resistance to therapy in several solid cancers, such as colorectal, pancreatic, gastric and ovarian cancer, highlighting the importance of stromal components in determining outcome [2][3][4][5]. Furthermore, FAP overexpression in colorectal, pancreatic and ovarian cancer is associated with an increased risk of lymph node metastasis and worse prognosis [2,6]. We hypothesised that FAP expression might have a prognostic biomarker potential in patients with T1 NMIBC. To that end, we designed a retrospective study with high grade (HG) T1 NMIBC and pair-matched patients, who did experience progression to �T2 disease, with patients who did not progress. Given the emerging data on the correlation between molecular subtypes and clinical outcome in bladder cancer (BC), we also included CK5 and GATA3 as surrogate immunohistochemical (IHC) markers for basal and luminal differentiation, respectively [7,8].

Materials and methods
From our institutional database of HG T1 NMIBC, we retrospectively selected 30 patients who progressed to �T2. Those were pair-matched using the Club Urologico Español de Tratamiento Oncologico (CUETO) progression score variables with 90 patients who never progressed despite having even longer follow-up [9]. All patients completed at least 5/6 doses of their initial BCG induction course. Haematoxylin slides and the corresponding formalin-fixed paraffin-embedded blocks from the transurethral resection of bladder tumour of the first diagnosis of HG T1 of the selected patients were retrieval from our institutional pathology archives. All procedures were performed between March 1993 and April 2015. After single pathologist revision of all slides and tissue quality assessment, we retained a final cohort of 86 patients (64 non-progressors and 22 progressors) (S1 Fig). There were no significant differences in clinical and pathological variables between progressors and non-progressors, except for the presence of carcinoma in situ (CIS) (p = 0.02) ( Table 1). Slides were stained for FAP and CK5-GATA3 (double stain), using validated protocols on the Ventana Discovery Ultra and Ventana Benchmark XT platform (S1 Table). Stained slides were scanned and all HG T1 tumour regions within each patient were digitally annotated on a serial H&E-stained slide by an expert uropathologist (TG). Multiple tumour stages (T1, Ta or CIS) and/or regions of interest (ROIs) were often present within one patient, but only HG T1 ROIs were used for our analyses. Digital annotations of ROIs were transferred to the image analysis software Visiopharm, and stains were scored via a quantitative method per ROI. CK5 and GATA3 were analysed in the tumour area of HG T1 ROIs, and were measured as relative marker area (RMA) (formula: IHC-positive tumour area/total tumour area). FAP was analysed in the stromal area of HG T1 ROIs, and was also measured as RMA (formula: IHC-positive stromal area/total stromal area; Fig 1). Ethical approval was obtained at the Ethics Committee Research UZ/KU Leuven (internal study number S59191) in accordance with the ethical standards of the Declaration of Helsinki of 1964 and its later amendments, with a waiver of informed consent due to the retrospective nature and the secondary use of tissue.

Results
The 5-year progression-free survival (PFS) of the total population was 73% ( preparation of the manuscript. The specific roles of these authors are articulated in the 'CRediT author statement'.

Competing interests:
The authors have declared that no competing interests exist. The above does not alter our adherence to PLOS ONE policies on sharing data and materials. doses of a second induction course; yes vs. no) for recurrence-free survival (RFS), PFS, cancerspecific survival (CSS), and overall survival (OS). CUETO progression score was not included in the model as our primary cohort was pair-matched using this score, and CUETO progression scores remained balanced between groups following single pathologist revision of all slides and tissue quality assessment (p = 0.8) [9]. There was no significant difference in RFS (p = 0.08) or PFS (p = 0.8) when stratifying for presence of CIS (yes vs. no). Additionally, there was no correlation between FAP RMA in HG T1 ROIs and CUETO progression score (ρ = -0.02; p = 0.8), and CUETO progression score was also not prognostic for PFS (p = 0.5), and RFS (p = 0.2). Interestingly, FAP RMA in HG T1 ROIs was prognostic for RFS, PFS, CSS, and OS in these bivariable models (for PFS: S2 Fig). CK5 and GATA3 were not prognostic for these endpoints in comparable bivariable models replacing FAP with either CK5 or GATA3. Adequate BCG, as defined by the FDA, only impacted RFS, but not PFS, CSS, and OS, which stresses the fact that FAP is indeed an independent prognostic variable for all endpoints. To visualize the impact of FAP RMA in HG T1 ROIs, we dichotomized FAP as a categorical variable and compared low vs. high using log-rank test with the median FAP RMA as cut-off ( Fig  2). Additionally, we visualized PFS stratified by FAP RMA in T1 ROIs (low vs. high) and adequate BCG (yes vs. no), which suggested a potential predictive role of FAP (Fig 3). However, no significant differences in pairwise comparison were found in PFS in FAP low patients between adequate vs. inadequate BCG (p = 0.5), and in FAP high patients between adequate vs. inadequate BCG (p = 1). We did find a significant difference in RFS in FAP low patients between adequate vs. inadequate BCG (p = 0.01) on separate analyses for RFS and PFS for FAP high and FAP low patients (Fig 4). These findings suggest a potential predictive role with regard to RFS of FAP expression for adequate BCG treatment per FDA definition.

Discussion
This is the first time that a stromal factor is shown to be prognostic for progression of T1 NMIBC. Additionally, luminal and basal markers were not prognostic for clinical outcome in T1 NMIBC. The stromal marker FAP is an emerging biomarker and has been linked to impaired prognosis and therapy resistance in several cancer [2][3][4]. Moreover FAP-specific PET represents a promising imaging modality for radiotherapy planning [10], and various approaches to clinically target FAP are currently being studied in other cancers [4]. At present, the possible role of FAP in BC is largely understudied. The single arm, phase 2 ABACUS-trial with neo-adjuvant atezolizumab in patients with MIBC found that FAP expression remained high in relapsing tumours whereas a decrease was seen in responders, suggesting that FAP is linked to resistance to therapy [11]. Another study on a mixed MIBC/NMIBC population identified a FAP-dominant cluster prognostic for decreased OS [12], and an additional study on MIBC found that presence of FAP in CAFs implies worse CSS [13]. Our data, focussing exclusively on a HG T1 population, are in line with these findings, strongly suggesting a prognostic role for FAP expression. Clinical response to PD-L1 inhibitors in patients with metastatic UC is associated with a CD8+ immune phenotype in the tumour cells, whilst lack of response is associated with presence of CD8+ immune cells in the fibroblast-rich peritumoural stroma, excluded from tumour cells [14]. Future research regarding FAP in T1 disease should  PLOS ONE relatively small sample sizes, and absence of detrusor at re-TURBT in a limited number of patients (n = 9). FAP might serve as an easily applicable prognostic biomarker to risk-stratify patients with HG T1 NMIBC if these results are prospectively validated in a larger series.

Acknowledgments
• Tim Muilwijk and Murat Akand contributed equally and should therefore be considered the first co-authors. Thomas Gevaert and Steven Joniau also contributed equally should therefore be considered the senior co-authors.
• Steven Joniau is a senior clinical researcher of the Research Foundation of Flanders (FWO).