Predicting outcome in minimally invasive (T1a and T1b) urothelial bladder carcinoma using a panel of biomarkers: a high throughput tissue microarray analysis

Authors

  • Paulette Mhawech-Fauceglia,

    1. Department of Pathology and Laboratory Medicine at Roswell Park Cancer Institute, Buffalo, NY, USA, and Department of Rehabilitation and Geriatrics at Geneva University Hospitals, Geneva, Switzerland.
    Search for more papers by this author
  • Gabor Fischer,

    1. Department of Pathology and Laboratory Medicine at Roswell Park Cancer Institute, Buffalo, NY, USA, and Department of Rehabilitation and Geriatrics at Geneva University Hospitals, Geneva, Switzerland.
    Search for more papers by this author
  • Victor Alvarez Jr,

    1. Department of Pathology and Laboratory Medicine at Roswell Park Cancer Institute, Buffalo, NY, USA, and Department of Rehabilitation and Geriatrics at Geneva University Hospitals, Geneva, Switzerland.
    Search for more papers by this author
  • Arsalan Ahmed,

    1. Department of Pathology and Laboratory Medicine at Roswell Park Cancer Institute, Buffalo, NY, USA, and Department of Rehabilitation and Geriatrics at Geneva University Hospitals, Geneva, Switzerland.
    Search for more papers by this author
  • François R. Herrmann

    1. Department of Pathology and Laboratory Medicine at Roswell Park Cancer Institute, Buffalo, NY, USA, and Department of Rehabilitation and Geriatrics at Geneva University Hospitals, Geneva, Switzerland.
    Search for more papers by this author

Paulette Mhawech-Fauceglia, Roswell Park Cancer Institute, Department of Pathology, Elm and Carlton Street, Buffalo, NY 14263, USA.
e-mail: pmhawech1@yahoo.com

Abstract

OBJECTIVE

To evaluate the protein expression of fibroblast growth factor receptor-3 (FGFR3), hamartin, 14-3-3σ, Aurora-A, and E-cadherin using immunohistochemistry (IHC) in a series of human bladder carcinomas and to evaluate their value in distinguishing T1a from T1b tumours and in predicting their behaviour, as T1 urothelial bladder tumours present great diagnostic and therapeutic challenges to pathologists and clinicians.

PATIENTS, MATERIALS AND METHODS

Tissue microarrays were constructed from 94 patients (Ta 20, T1a 31, T1b 14, and T2 29 patients) using tissue obtained at first disease presentation.

RESULTS

FGFR3 and 14-3-3σ were the only markers that were significantly associated with tumour grade and 14-3-3σ was significantly associated with tumour stage. Furthermore, none of these markers could help in distinguishing T1a from T1b tumours. After adjusting for the E-cadherin expression, FGFR3 expression was a significant factor in predicting the time to recurrence in T1a/T1b. Furthermore, among all the clinical variables, grade and depth of invasion were the only ones that had a significant value in predicting T1a/T1b tumour progression.

CONCLUSIONS

Even though the staging of T1 to T1a/T1b is not a common practice and it is not included in the Tumour-Node-Metastasis classification, our data clearly confirmed the importance of a proper sub-staging of T1 tumours whenever feasible.

Abbreviations
UBC

urothelial bladder carcinoma

FGFR3

fibroblast growth factor receptor-3

IHC

immunohistochemistry

TUR

transurethral resection

BCG

bacillus Calmette-Guérin

MM

muscularis mucosae

H&E

haematoxylin and eosin

HR

hazard ratio.

INTRODUCTION

Most patients (≈ 75%) with urothelial bladder carcinoma (UBC) are diagnosed at first presentation with noninvasive (Ta) or minimally invasive (T1) tumours. A small percentage (≈ 20%) present with muscular invasive (T2) and 5% present with a metastatic tumour. The management of Ta and T2 tumours is well established; Ta tumours are treated by transurethral resection (TUR) with or without bacillus Calmette-Guérin (BCG) or intravesical chemotherapy and T2 tumours are treated by cystectomy with or without pre- or postoperative chemotherapy. The management of T1 tumours is far more complicated [1]. As no consensus on the best treatment for these patients has been reached, they can be treated as Ta or as T2 tumours [1,2]. This issue became more challenging when T1 tumours were further sub-classified into T1a and T1b tumours by the discovery of the muscularis mucosae (MM) in the lamina propria. MM is defined as scattered, discontinuous smooth muscle fascicles adjacent to intermediate size blood vessels in the lamina propria [3]. Based on the MM T1 has been further sub-classified into two groups: T1a (minimally invasive tumours) where tumours cells invade the lamina propria but are still located above the level of the MM and T1b (invasive tumours) where tumours cells are seen invading beyond the MM but with no involvement of the muscularis propria [3,4]. As T1b tumours have been proven to behave more aggressively than T1a tumours, this sub-classification became an important factor in patients’ prognosis and management [4–8]. Subsequently, recommendations to treat patients with T1a tumours very conservatively and patients with T1b more aggressively started to emerge [2,9]. However, poor orientation, cautery artefact, and tumour necrosis, all impedes the ability of a pathologist to sub-classify T1 tumours. Subsequently, these factors account for most of intra- and inter-pathologists variation and disagreement [10–12]. Due to these difficulties and despite the proven impact of this sub-classification on patients’ management, it is still not a common practice among pathologists and even more it was not included in the TNM classification. Thus, finding a biomarker that can help distinguishing T1a from T1b is of major importance. Furthermore, finding a parameter that is capable of predicting T1a/T1b tumours recurrence and progression, so patients with tumours that are more prone to progression, are treated more aggressively, is needed. Long after the MM had been discovered, and compelling evidence indicated the importance of T1 sub-classification, this subject has been forgotten with few published studies addressing this problem [13–17].

The objectives of the present study were to evaluate the expression of five biomarkers in T1a/T1b tumours. Fibroblast growth factor receptor-3 (FGFR3), 14-3-3σ, hamartin, Aurora-A, and E-cadherin were chosen because they had a proven role in UBC pathogenesis [18]. None of these biomarkers have been studied in T1a and T1b tumours. Well-selected cases of T1a and T1b tumours were chosen for a tissue microarray and FGFR3, 14-3-3σ, hamartin, Aurora-A, and E-cadherin protein expressions were evaluated by immunohistochemistry (IHC). This study aimed to determine whether: (i) these biomarkers are associated with tumour grade and stage, (ii) any of these biomarkers can distinguish T1a from T1b tumours, and (iii) any of these biomarkers can predict tumour outcomes such as progression and recurrence.

PATIENTS, MATERIALS AND METHODS

This retrospective study was conducted over a 3-year period, using TUR specimens. Criteria for inclusion were tumours from untreated patients, tumours that could be accurately staged, and an adequate follow-up. The specimen had to have sufficient tumour, well orientated, including the visualizing of the MM and the muscularis propria. Most of the specimens were selected from a previous study; where a thorough evaluation was conducted to accurately sub-classify tumours into T1a and T1b [19]. Tumour grading was based on the 2003 WHO classification, and specimens were stratified into two groups, low grade and high grade [20]. Tumours were reviewed by two pathologists to confirm tumour histological grade. The follow-up data was retrieved from the patients’ files. Progression was defined as disease progression to a higher tumour stage or metastasis, with the progression confirmed by histology and reviewed by the same pathologists. Treatment consisted of repeated TUR with or with no intravesical administration of BCG therapy or intravesical chemotherapy and cystectomy. Patients with Ta and T2 tumours were also evaluated to determine the association between the biomarkers with tumour grade and tumour stage.

For the construction of the tissue microarrays, paraffin wax-embedded tissues from 94 patients’ samples were used. The tissue microarray was constructed as described previously by Kononen et al. [21]. Briefly, after carefully choosing the morphologically representative region from the haematoxylin and eosin (H&E) stained section, 0.6-mm cores were punched from the individual paraffin wax-embedded blocks (donor blocks), and transferred to the receiver paraffin wax-embedded block (receiver block). To overcome tumour heterogeneity, core biopsies were taken from five different areas of each tumour. One section was stained with H&E to confirm the presence of the tumour by light microscopy.

For the IHC, 4-µm sections were cut, deparaffinized with xylene, and washed with ethanol. The characteristics of the five antibodies used in this study are summarized in Table 1. Endogenous peroxidase was blocked with 0.3% hydrogen peroxidase for 5 min Antigen retrieval was carried out as described in Table 1. Sections were incubated with either FGFR3, 14-3-3σ, hamartin, Aurora-A, or E-cadherin at room temperature. The biotin-free horseradish peroxidase enzyme-labelled polymer of the Envision plus detection system (Dakocytomation, CA, USA) was used as a secondary reagent. The diaminobenzidine complex was used as the chromogen. In negative controls, a normal goat serum was used instead of the primary antibody. Evaluation of the IHC slides was done semiquantitatively by two pathologists (P.M., G.F./V.A.) with a double-headed microscope. The pathologists were ‘blinded’ to the original histological diagnosis. The IHC evaluation was done twice and each review was separated by a 2-week interval and the scores compared. Whenever a discrepancy between the first and the second readings occurred, the two pathologists discussed the case and reached a consensus agreement for the final scoring. Disagreement was not very frequent between the first and the second pathologists and occurred for ≈5% of the specimens. In addition, whenever any of the antibodies was not expressed in a homogenous manner in the five different core biopsies, we took the final results as the expression of this particular biomarker as seen in most of the cores. However, heterogeneity was not a big issue as maybe one core out of five would be different but most of the cores were equal in expressing a particular antibody.

Table 1.  Summary of the characteristic of the five antibodies used
AntibodyClonalityPretreatment (min)DilutionIncubation time, minSource
  • *

    This antibody was a gift from Dr Ramesh at Massachussetts General Hospital, MA, USA.

FGFR3MonoclonalHigh pH, steamer-cooker (3)1 : 10060Santa-Cruz, CA, USA
HamartinPolyclonalCitrate pH 6, microwave (3)1 : 2510Home made*
Aurora-APolyclonalSteamer-cooker (3)1 : 10030Bethyl, TX, USA
14-3-3σPolyclonalMicrowave (3)1 : 2510IBL, Japan
E-cadherinMonoclonalHigh pH, steamer-cooker (3)1 : 10030Novacastra, Newcastle, UK

Cytoplasmic staining for FGFR3 (Fig. 1.), hamartin, and Aurora-A, membranous for E-cadherin and cytoplasmic and nuclear for 14-3-3σ (Fig. 2), was considered positive. Tumours were segregated into two groups depending on the intensity and the percentage of cells stained. Group I contained tumours with no staining, weak intensity, strong intensity but in <10% of cells or weak intensity in >10% of cells. Group II containing tumours with moderate/strong intensity in >10% of cells; 10% was a rough estimation and this scheme has been used in other published studies [22,23]

Figure 1.

An example of positive cytoplasmic staining for FGFR3;. the tumour cells are strongly positive (×20).

Figure 2.

An example of positive cytoplasmic and nuclear staining for 14-3-3σ; the tumour cells are strongly positive (×20).

For statistical analysis, the following baseline variables were considered for their prognostic value: age at presentation, gender, tumour stage and grade, tumour multifocality, treatment method and FGFR3, 14-3-3σ, hamartin, Aurora-A and E-cadherin expression (considered as binary variables). Associations between categorical variables were studied using the Fisher’s exact test. Comparisons of continuous variables were performed using either the Student t-test or the Mann–Whitney U-test. The date of diagnosis was considered as the time of origin. For recurrence- and progression-free intervals, we considered the date of first recurrence and the date of first progression, respectively, as events. Recurrence-free and progression-free survival were assessed using univariate and multiple Cox proportional hazards models and the results were expressed as hazard ratio (HR), which was derived from the estimated regression coefficients with their 95% CIs; P ≤ 0.05 was considered to indicate statistical significance.

RESULTS

In all, 94 patients were included in the study and the tumours were distributed as follows; Ta (20 patients), T1a (31), T1b (14), T2 (29). The patients were aged 46–92 years, 79 were male and 15 were female, and 58 were low-grade and 36 were high-grade tumours. To determine the association between the biomarkers expression and tumour grade and stage, all 94 patients were included in the first part of the study. However, to determine if any of these five biomarkers could help in distinguishing T1a from T1b tumours and if any had an independent value in predicting T1a and T1b outcomes, only the 45 T1 tumours were considered for the second part of the study. A summary of the histological and clinical data of the T1 cases are reported in Table 2.

Table 2. 
Summary of the clinical and histological data of the 45 patients with T1a/T1b tumours
CharacteristicT1aT1bTotal
Stage, n311445
Grade, n
 Low grade27 936
 High grade 4 5 9
Sex (M:F) ratio, n27:413:140:5
Multicentricity, n11 617
Therapy, n
 TUR23 932
 TUR + BCG 6 1 7
 Cystectomy 2 4 6
Recurrence, n 9 110
Progression, n 2 911
Age, years: mean (sd) range 71.0 (70.0)46.0–92.068.7 (69.5)70.0 (70.0)
Follow-up, months: median (sd) interquartile range13.0 (15.0)  3.0–77.010.0 (14.0)12.0 (13.0)

The association between the five biomarkers and tumour grade and stage are shown in Table 3. There was a significant association between strong FGFR3 expression and tumour grade (P = 0.017). Tumour grade has been shown to be an important prognostic factor. In the present study, low-grade tumours more frequently expressed FGFR3 than high-grade tumours. On the other hand, high 14-3-3σ expression was significantly associated with high-grade tumours and advanced tumour stage (P = 0.024 and P = 0.005, respectively). All five of the biomarkers (FGFR3, 14-3-3σ, hamartin, Aurora-A and E-cadherin) were expressed equally in T1a and T1b tumours. No single one was useful in distinguishing between those two tumours. 14-3-3σ, hamartin, Aurora-A, and E-cadherin proteins expression did not have significant value in predicting T1a and T1b tumours outcome. Likewise, FGFR3 expression by itself did not predict recurrence. However, by adjusting for E-cadherin expression, FGFR3 expression seemed to predict tumour recurrence (P = 0.024). This is mainly due to the few cases positive for FGFR3 expression with recurrences. Hence, tumours with high FGFR3 expression tend to have a longer recurrence period than those who had low FGFR3 expression. Therefore, FGFR3 expression might be useful to predict T1a/T1b recurrences but first it should be confirmed with a larger sample. In the present study, all five biomarkers failed to predict tumours progression. To investigate the value of clinical variables in predicting the outcome of T1a/T1b tumours, gender was not included as a variable because none of the women patients developed progressive disease. The clinical variables considered for analysis were age, tumour grade, depth of invasion, multicentricity, and treatment method. Grade and depth of invasion were the only variables to have a significant value in predicting tumour progression. Thus, patients with high-grade tumours were 6.6 times more likely (P = 0.002) to progress sooner than lower grade tumours. Patients with T1b tumours were 12.3 times more likely to progress sooner (P = 0.001) than those with T1a tumours (Table 4).

Table 3.  The association of the five biomarkers with tumour grade and tumour stage in 94 patients
BiomarkerStaining, − or +*GradeStage
LowHighPTaT1aT1bT2P
  • *

    −, no staining, weak intensity, or strong intensity but in <10% of cells; +, moderate/strong intensity and in >10% of cells.

N 5935 20311429 
FGFR337 (63)31 (89) 0.01712 (60)20 (65)1125 (86) 0.130
+22 (37) 4 (11)  8 (40)11 (36) 3 4 (14) 
Hamartin26 (44)13 (37) 0.20910 (50)10 (32) 514 (48) 0.491
+33 (56)22 (63) 10 (50)21 (68) 915 (52) 
14-3-3σ29 (49) 8 (23) 0.02413 (65)12 (39) 6 5 (17) 0.005
+30 (51)27 (77)  6 (30)19 (61) 824 (83) 
Aurora-A31 (53)13 (37) 0.24213 (65)13 (42) 612 (41) 0.343
+28 (47)22 (63)  7 (35)18 (58) 817 (59) 
E-cadherin15 (26)14 (40) 0.3635 (25) 7 (23) 413 (45) 0.275
+44 (75)21 (60) 15 (75)24 (77)1016 (55) 
Table 4.  The predictive value of the five biomarkers and patient clinical and histological data using univariate and multiple Cox propotional hazards regression models (45)
MarkerCrude HR (95% CI)PAdjusted full model HR (95% CI)PAdjusted mini model HR (95% CI)P
  • *

    Only five females, with 0 in the progression group resulting in convergence problem.

First recurrence
Male vs female0.31 (0.04–2.78)0.297    
Age1.05 (0.99–1.12)0.133    
Multicentricity1.02 (0.29–3.64)0.971    
Grade (low vs high)1.35 ([0.28–6.42)0.708    
Stage T1b vs T1a0.32 (0.04–2.52)0.277    
Markers:
 FGFR33.13 (0.90–10.86)0.0729.54 (1.36–67.11)0.0236.08 (1.26–29.3)0.024
 Hamartin0.69 (0.20–2.38)0.5522.57 (0.41–15.97)0.312  
 14-3-3σ0.38 (0.10–1.36)0.1370.46 (0.12–1.78)0.261  
 Aurora-A0.48 (0.14–1.74)0.2670.52 (0.11–2.37)0.396  
 E-cadherin0.48 (0.13–1.70)0.2540.22 (0.04–1.22)0.0830.21 (0.04–1.06)0.059
First progression
Male vs female*1.000    
Age0.94 (0.87–1.01)0.085    
Multicentricity0.87 (0.25–3.02)0.830    
Grade, low vs high6.55 (1.95–22.06)0.002    
Stage, T1B vs T1A12.29 (2.64–57.14)0.001    
Markers:
 FGFR30.28 (0.04–2.18)0.2240.32 (0.03–3.00)0.315  
 Hamartin2.68 (0.58–12.49)0.2091.47 (0.25–8.84)0.671  
 14-3-3σ1.03 (0.29–3.62)0.9620.79 (0.22–2.92)0.729  
 Aurora-A1.81 (0.47–6.95)0.3861.34 (0.30–5.90)0.701  
 E-cadherin2.99 (0.38–23.38)0.2973.56 (0.43–29.31)0.237  

DISCUSSION

In accordance with published data, strong FGFR3 protein expression seen in the present study was significantly associated with one of the most powerful prognostic factors in UBC, tumour grade [24,25]. Activating mutations of FGFR3, a tyrosine kinase receptor, have been implicated in the pathogenesis of UBC and in early stage papillary UBC [25,26]. There are few published studies using IHC to assess protein expression of FGFR3 in UBC [24,27,28]. We have previously reported that FGFR3 expression by IHC was associated with tumour grade and stage in Ta/T1 tumours, but FGFR3 expression was not an independent prognostic factor [24]. These T1 cases were not sub-classified into T1a/T1b and none of those cases were included in the present study. In the present study, FGFR3 protein expression predicted T1a/T1b tumour recurrence, after adjusting for E-cadherin protein expression. Therefore, patients with tumours with high FGFR3 expression tend to have a longer disease-free interval. The finding of a value for FGFR3 expression in T1a/T1b tumours only after adjusting for E-cadherin expression is mainly due to the few cases positive for FGFR3 expression with recurrences in our sample. Therefore, this finding should be confirmed in more cases.

14-3-3σ is considered a very highly restricted epithelial antigen that functions as a tumour suppressor gene. Loss of 14-3-3σ is associated with failure to arrest the cell cycle at the G2-M phase checkpoint, and DNA damage can lead to increase G2-type chromosomal aberrations [29]. 14-3-3σ down-regulation by CpG methylation was detected in numerous cancer types including breast, lung, endometrium, ovary, prostate and hepatocellular carcinoma and its expression is up-regulated in pancreatic adenocarcinoma [29–31]. In one of our previous studies, we showed a close association between 14 and 3-3σ CpG island methylation detected by methylation-specific PCR and loss of protein expression using IHC [30]. In a separate study, we found that 14-3-3σ protein expression by IHC was a frequent event in UBC, where 97.8% of the UBC were positive for 14-3-3σ[32]. In the present study, we showed that 14-3-3σ was overexpressed in UBC. However, the exact role of 14-3-3σ in UBC is still widely unknown. We also found that 14-3-3σ overexpression is significantly associated with the two most powerful predictive factors; tumour grade and stage, but 14-3-3σ expression did not predict T1a/T1b tumour outcomes. However, more studies to explore the role of 14-3-3σ in UBC are ongoing in our laboratory.

In the present study, hamartin, Aurora-A/STK15 and E-cadherin biomarkers failed to show any association with tumour grade, stage and even outcome. Loss of tuberous sclerosis complex 1 gene, a tumour suppressor gene that encodes a protein termed hamartin has been implicated in bladder carcinogenesis [33,34]. It is expressed in low-grade, low-stage UBC, and its loss was attributed in Ta recurrence. Aurora-A/STK15 is a member of the serine/threonine kinase family and its overexpression was associated with advanced stage disease and poor prognosis [35–37]. E-cadherin is one of the proteins involved in cell–cell adhesion, and loss of its expression was significantly associated with UBC progression and poor prognosis [38,39].

Among all the variables that were evaluated in the present study including clinical data (age, sex, mutifocality, treatment methods), histological data (grade, stage) and protein expressions (FGFR3, 14-3-3σ, hamartin, Aurora-A, E-cadherin), tumour stage and tumour grade were the only independent factors in predicting tumour progression. Despite the difficulty in sub-staging T1 tumours, this sub-classification is once again proven to be the most important independent factor to predict disease progression, confirming previous reports [7,8,14]. The sub-classification of T1 tumours is not a common practice among pathologists worldwide and furthermore it is still not included in the TNM classification. This is mainly due to the difficulties that pathologists encounter during their histological interpretation of bladder tumours. In the present study, we attempted to find a biomarker that could help in distinguishing T1a from T1b tumours but none of the five biomarkers analysed were of any help. Finally, we should insist that to provide important prognostic information, and especially when it can be feasible by the pathologists, this sub-classification should be included in the final histology report and not ignored because it is not required.

In summary, the role of FGFR3 expression in predicting T1a/T1b tumour recurrence should be confirmed by larger studies. Tumour differentiation and depth of tumour invasion are the best independent predictive factors in T1a/T1b tumour progression. However, more studies are needed to understand the molecular mechanisms that explain the behaviour of T1a/T1b tumours.

ACKNOWLEDGEMENTS

We would like to thank Charles LeVea for his critical review and we also would like to thank the Core facility laboratory for their help. We also would like to thank Doug Nixon for his help in doing the illustration.

CONFLICT OF INTEREST

The authors state that there is no conflict of interest.

Ancillary