Tissue microarray analysis of cytoskeletal actin-associated biomarkers gelsolin and E-cadherin in urothelial carcinoma


  • JianYu Rao M.D.,

    Corresponding author
    1. Department of Pathology and Laboratory Medicine, University of California at Los Angeles, Los Angeles, California
    • Department of Pathology and Laboratory Medicine and Jonsson Comprehensive Cancer Center, University of California at Los Angeles, 10833 Le Conte Ave., Los Angeles, CA, 90095
    Search for more papers by this author
    • Fax: 310-825-7795

  • David Seligson M.D.,

    1. Department of Pathology and Laboratory Medicine, University of California at Los Angeles, Los Angeles, California
    2. Department of Human Genetics, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
    • D.S. and H.V. contributed equally to this article.

  • Harri Visapaa M.D.,

    1. Department of Pathology and Laboratory Medicine, University of California at Los Angeles, Los Angeles, California
    2. Department of Human Genetics, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
    • D.S. and H.V. contributed equally to this article.

  • Steve Horvath Sc.D., Ph.D.,

    1. Department of Human Genetics, University of California at Los Angeles, Los Angeles, California
    2. Department of Biostatistics, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
  • Mervi Eeva M.S.,

    1. Department of Human Genetics, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
  • Kia Michel M.D.,

    1. Department of Urology, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
  • Allan Pantuck M.D.,

    1. Department of Urology, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
  • Arie Belldegrun M.D.,

    1. Department of Urology, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author
  • Aarno Palotie M.D.

    1. Department of Pathology and Laboratory Medicine, University of California at Los Angeles, Los Angeles, California
    2. Department of Human Genetics, University of California at Los Angeles, Los Angeles, California
    Search for more papers by this author



Alterations of expression of the cytoskeletal proteins Gelsolin and E-cadherin have been implicated in urothelial carcinoma tumorigenesis. However, it is not clear how these altered expressions associate with tumor progression, nor is it clear how these protein markers provide prognostic value for urothelial carcinomas.


Primary urothelial carcinoma tissue microarrays were constructed for 146 patients with urothelial carcinoma. Where available, four replicate tissue samples of invasive tumor, adjacent dysplastic and in situ lesions, and benign tumors were arrayed for each case, resulting in a total of 1208 tissue spots. Immunohistochemical staining for Gelsolin, E-cadherin, p53, and Ki67 (MIB-1) was performed on the arrays. For each marker, the maximum staining intensity (Max), the percentage of positive staining (Pos), and the product of both Max and Pos (MaxPos) were analyzed.


Compared with the benign fields, the expression of both cytoskeletal proteins decreased in premalignant and malignant lesions. For Gelsolin, decreased MaxPos was seen in premalignant and preinvasive lesions. However, with an increase in tumor grade and stage, there was a gradual increase in Gelsolin (P < 0.05 for both). E-cadherin expression decreases mainly in high-grade lesions (carcinoma in situ and Grade 3 tumors). Univariate and multivariate analyses showed that Gelsolin Max was a strong independent predictor for the probability of tumor recurrence and for early tumor recurrence in high-grade or high-stage tumors, as well as a strong indicator for tumor progression.


Gelsolin and E-cadherin have distinctive expression patterns. Gelsolin, but not E-cadherin, provides independent prognostic information for high-grade urothelial carcinomas. Cancer 2002;95:1247–57. © 2002 American Cancer Society.

DOI 10.1002/cncr.10823

Urothelial carcinoma, or transitional cell carcinoma (TCC), is the most common cancer type in the lower urinary tract. Like other malignant neoplasms, urothelial carcinomas develop through multiple genetic and epigenetic alterations that lead to altered growth, differentiation, and apoptotic control.1, 2 They provide a useful model system to study carcinogenic processes because these tumors have a well defined progression of disease (e.g., from premalignant dysplasia to preinvasive carcinoma in situ [CIS], to superficial carcinoma, and, finally, to invasive carcinoma). In addition, it is relatively easy to access the entire organ system through urine cytologic and cystoscopic examinations.3 Urothelial carcinomas are a unique tumor type. Morphologically, low-grade papillary tumors, which are predominantly noninvasive at the time of initial presentation, have a high recurrence rate with over two-thirds of cancers recurring. In addition, high-grade tumors are more invasive at initial presentation, although not all high-grade tumors are invasive initially.4 Tumor grade alone may not be able to predict the behavior of an individual tumor. Therefore, additional biomarkers that can predict tumor recurrence and the invasive behavior of a tumor will be useful clinically.

Previous studies have demonstrated that alterations of cytoskeletal actin remodeling play an important role in various aspects of urothelial carcinoma carcinogenesis, including cellular differentiation,5 transformation,6, 7 and apoptotic control.8 Monitoring the changes of actin polymerization status provides a biomarker to predict tumor recurrence.9 Because the cytoskeleton actin network also plays an important role in cell motility and adhesion, it has been generally assumed that alterations of the actin network may also be involved in tumor invasiveness. Studying several actin-associated proteins involved in different aspects of actin function simultaneously in actual tumor samples may improve our overall understanding of how actin is involved in the carcinogenic process. In turn, it may also determine whether an actin-based molecular profiling analysis may assist traditional histopathologic markers, such as tumor grade and stage, in predicting tumor behaviors such as tumor invasion and recurrence.

Using our newly established urothelial tumor tissue microarray analysis (TMA), the primary purpose of this study was to examine the expression patterns of two important cytoskeleton actin-related proteins, Gelsolin and E-cadherin, in conjunction with two other well established markers, Ki67 and p53, and to correlate their expressions with tumor progression, invasion, and recurrence. Gelsolin is a major actin regulatory protein involved in regulating the actin polymerization process by severing and capping F-actin, the polymerized filamenteous form of actin,10 whereas E-cadherin is an actin-binding protein associated with cell adhesion.11 In vitro studies showed that the loss of Gelsolin expression played a critical role in urothelial carcinoma carcinogenesis12 and that Gelsolin is also involved in regulating cellular apoptotic processes by functioning as an important candidate for caspase activity.13 However, no study has been performed to examine the relationship of Gelsolin expression with clinical variables and, in turn, the clinical value of Gelsolin as an independent marker for outcomes prediction in patients with urothelial carcinoma. Although a number of studies have observed decreased E-cadherin expression in patients with urothelial carcinoma, results concerning the independent clinical value of E-cadherin immunostaining are controversial.14–17 In this study, we analyzed the protein level expression patterns of Gelsolin, E-cadherin, Ki67, and p53 using immunohistochemical methods on 202 tissue samples of urothelial carcinoma from 146 patients.


Urothelial Carcinoma TMA Construction

Our TMA included archived bladder carcinoma tissue samples from cases dated between 1985 and 1995. The material was obtained from the Department of Pathology at the University of California, Los Angeles (UCLA) Medical Center, following approval by the UCLA Institutional Review Board. The original hematoxylin and eosin (H&E)-stained case slides were reviewed by one pathologist (J.Y.R.), utilizing the 1997 TNM classification.18 During the review, slides containing tumor, adjacent dysplasia, and distant benign fields were selected and marked as such by designated colored ink.

TMA blocks were constructed following the technique described by Kononen et al.19 Where available, four representative tissue samples (a “core set”) from each selected area were included in the arrays. Three arrays contained tissue samples from 172 patients, comprising 232 samples and 1363 tissue spots. The analysis was limited to urothelial carcinoma cases, including 140 from the bladder (2 of which showed small cell differentiation, 1 showed signet ring features, and 1 accompanied concomitant renal pelvis urothelial carcinoma), 3 from the renal pelvis alone, and 3 from the ureter alone. Metastatic tumors and tumors showing exclusive squamous cell carcinoma and adenocarcinoma differentiation were excluded from the analysis. Final analysis included 146 patients, 202 cases, and 1208 tissue spots.

Immunohistochemistry on TMA Sections

H&E-stained array sections were evaluated histopathologically by two anatomical pathologists (J.Y.R., D.S.) in a blinded fashion to validate the diagnostic morphology of each array spot. Commercial antibody preparations were utilized for the analyses: p53 (Dako, Corporation, Carpenteria, CA); E-cadherin (Zymed, San Francisco, CA); Ki67 (Dako); and Gelsolin clone GS-2C4 (Sigma, St. Louis, MO).

For immunohistologic staining of Gelsolin, a standard two-step indirect avidin-biotin complex (ABC) method was used (Vector Laboratories, Burlingame, CA). Tissue array sections (4-μm thick) were cut immediately before staining. They were heated to 56 °C for 20 minutes, followed by deparaffinization in xylene. The sections were rehydrated in graded alcohols and endogenous peroxidase was quenched with 3% hydrogen peroxide in methanol at room temperature. The sections were placed in 95 °C solution of 0.01 M sodium citrate buffer (pH 6.0) for antigen retrieval. Protein blocking was accomplished through application of 5% normal horse serum for 30 minutes. Endogenous biotin was blocked with sequential application of avidin D, then biotin (A/B blocking system). Primary mouse anti-Gelsolin monoclonal IgG1 antibody was applied at a 1:750 dilution for 60 minutes at room temperature. After washing, biotinylated horse antimouse IgG was applied for 30 minutes at room temperature. The ABC complex was applied for 25 minutes and diaminobenzidine (DAB) was used as the chromogen. Phosphate-buffered saline (10 mM), pH 7.4, was used for all intermediate wash steps and a moist humidity chamber was used for prolonged incubations. The sections were counterstained with Harris hematoxylin, followed by dehydration and mounting. For staining of p53, E-cadherin, and Ki67, the DakoEnvision biotin-free dextran peroxidase staining system was used as was the Dako Autostainer automated staining system.

Tissue samples with known expression for each marker were used as positive controls (Ki67 and p53, high-grade breast carcinoma tissue sections; Gelsolin and E-cadherin, normal prostate tissue sections). Negative controls were sections treated the same as described above, but with the primary antibody replaced with pooled nonimmune mouse IgG of the same concentration (for Gelsolin) or with the primary antibody omitted as a null slide (for Ki67, p53, and E-cadherin). The sections were analyzed with a BX-40 brightfield microscope (Olympus, Tokyo, Japan) using the ×10–20 objectives. When there were questions concerning tissue morphology, H&E-stained sections were reviewed for confirmation.

Scoring of antibody staining was performed by pathologists who were blind to clinicopathologic variables. The intensity of DAB brown chromogen staining was graded. For the nuclear staining markers p53 and Ki67, a 0–3 scale was used (0, negative; 1, weak staining; 2, moderate staining; 3, strong staining). For nonnuclear staining using the cytoskeletal markers Gelsolin and E-cadherin, a 0–4 scale was used (0, negative; 1, weak staining; 2, weak but distinct staining; 3, moderate staining; 4, strong staining). Gelsolin has a granular cytoplasmic staining pattern, whereas E-cadherin has a membrane staining pattern. Metrics included both maximum intensity of staining (Max) and the proportion of the analyzed cells staining positively (Pos). Scoring procedures were done by J.Y.R. For each marker, the Max, Pos, and the product of both (MaxPos) were determined. The median value of repeated core spots representing each area of interest for each sample was used for the final analysis.

Clinical and Pathology Database

Detailed demographic, pathologic, and clinical information including treatment and follow-up data for at least 5 years was incorporated into a correlative database linked to the tissue specimens. In addition, data from the original pathology reports were utilized for analysis. Tumor registry data including treatment, recurrence, and survival data were obtained from the UCLA Cancer Program of the Jonsson Comprehensive Cancer Center.

Statistical Analysis

The association of marker expression versus clinicopathologic parameters was demonstrated in several ways. First, mean ± standard error of MaxPos was analyzed against progressive field changes, tumor grade, and stage. Student t test, analysis of variance (ANOVA), and the nonparametric Kruskal–Wallis tests were used to test whether the immunoreactivity of each marker differed between different groups defined by the clinicopathologic parameters of grade and stage. To analyze the recurrence risk of multiple clinicopathologic and marker expression variables, multivariate logistic regression was used. The adjusted odds ratios (relative risks) and their respective P values were determined. Kaplan–Meier curves were used to estimate recurrence-free time curves and the log rank test was used to test whether the curves differed between groups. To assess which covariates affect recurrence-free time, we used the Cox proportional hazards model. For each covariate, the relative hazard rate and the associated P value were reported. For all analyses, a P value of less than 0.05 was accepted as significant and the analyses were performed with the software package R (url:http://cran.r-project.org/).


Clinicopathologic Characteristics of Patient Samples

Table 1 shows the clinicopathologic data of the 146 patients included in the final analysis. For patients with multiple cases, data from either the patient's first procedure of any type or their cystectomy case, if performed within 1 month of the first procedure, were used for all subsequent analyses. The patients ranged in age from 33 to 94 years, with a mean age of 67 years. The male-to-female ratio was 3.6:1. There were 57 Tis/Ta/T1 noninvasive or superficially (lamina propria) invasive tumors and the remaining 89 were deeply invasive, including 44 T2, 35 T3, and 10 T4 tumors. There were 8 CIS, 6 Grade 1, 40 Grade 2, and 92 Grade 3 tumors. Therefore, there was a wide range of TCC tumors represented.

Table 1. Characteristics of 146 Patients According to Age, Gender, Tumor Grade (WHO Classification), Stage (UICC TNM Classification), and Surgical Procedures (TUR vs. Cystectomy)
CharacteristicsNo. of patients (%)
  1. WHO: World Health Organization; UICC: International Union Against Cancer; CIS: carcinoma in situ; TUR: transurethral resection.

 Male114 (77.7)
 Female32 (22.3)
Age (yrs)33–94 (mean, 67)
 16 (4.1)
 240 (27.4)
 392 (63.0)
 CIS8 (5.5)
 Tis8 (5.5)
 Ta, T149 (33.6)
 T244 (30.1)
 T335 (24.0)
 T410 (6.8)
 TUR58 (39.7)
 Cystectomy88 (60.3)

For each case, we attempted to obtain not only tumor areas, but also adjacent dysplastic areas and distant benign fields. These field samples, which include progressive changes from benign, to adjacent dysplasia, and finally CIS, provided a mechanism to study how marker expressions are altered in the early stage of the malignant process. The benign fields were obtained from the ureter or urethral resection margins (for cystectomy specimens) or from benign-appearing urothelium at least 5 mm distant from the tumor that was sampled (for transurethral resection). A total of 81 benign field, 8 adjacent dysplasia, and 40 informative CIS samples were included in our analysis.

Association of Marker Expression with Tumor Histopathologic Features

Figure 1 presents representative images of individual markers stained by immunohistochemistry. Ki67 and p53 showed an exclusively nuclear staining pattern, Gelsolin had a cytoplasmic staining pattern, and E-cadherin had a membrane staining pattern. Gelsolin not only stained the urothelium, but also stromal tissue including some smooth muscle cells and endothelium. Figure 2 shows the mean MaxPos of each marker in relation to progressive field changes and tumor grade. Both Ki67 and p53 showed a similarly progressive increased MaxPos from benign, to adjacent dysplasia, to CIS, and from low-grade to high-grade tumors. Gelsolin MaxPos decreased in the field lesions including dysplasia and CIS compared with the benign urothelium (P < 0.05 by ANOVA). Although still decreased in tumor samples compared with the benign urothelium, there was a trend of increased Gelsolin MaxPos with an increase in tumor grade (P < 0.05 by ANOVA). E-cadherin MaxPos also decreased in the adjacent dysplastic and CIS lesions compared with the benign urothelium and there was a gradual decrease in MaxPos with an increase in tumor grade, contrary to the trend seen with Gelsolin. This decreasing trend was also statistically significant at a P value less than 0.05 by ANOVA. It is noteworthy that E-cadherin expression, but not Gelsolin expression, was lower in CIS lesions compared with the Grade 1 lesions (P < 0.05 by Student t test). This is consistent with the notion that CIS is a high-grade noninvasive lesion, which is supported by the fact that it also had a higher Ki67 and p53 MaxPos compared with low-grade tumors.

Figure 1.

Tissue microarray analysis (TMA) photomicrographs. Representative images of individual markers stained by immunohistochemistry. Ki67 and p53 show an exclusively nuclear staining pattern, Gelsolin has a cytoplasmic staining pattern, and E-cadherin has a membrane staining pattern. Gelsolin not only stains the urothelium, but also stromal tissue including some smooth muscle cells and endothelium.

Figure 2.

Distribution of marker expression patterns (mean ± 1 standard error of maximum staining intensity and percentage of positive staining [MaxPos]) versus urothelial histopathologic category. Range, 0–400 for Gelsolin and E-cadherin and 0–300 for p53 and Ki-67.

Table 2 compares the MaxPos of each marker in different stages of tumor, from noninvasive (Ta/Tis), to superficially invasive (limited to the lamina propria, T1), to deeply invasive (invasion to or beyond the muscularis propria, T2 and above). One hundred cases had informative data across all markers. Both Gelsolin and p53 showed a significantly increased MaxPos with progression of tumor invasion (P < 0.05 by ANOVA). E-cadherin showed a marginal association (P < 0.1 by ANOVA) and Ki67 did not show a significant association with tumor stage. Correlation matrix analysis for the interrelationship of these markers using the MaxPos value showed that, except for some associations between Ki67 and p53 (r = 0.28), all other marker matrices had an r value less than 0.20, indicating that these markers were expressed rather independently in individual patient samples.

Table 2. Comparison of Marker Expression (MaxPos, Mean ± SE) among Noninvasive (Ta/Tis), Superficially Invasive (T1), and Deeply Invasive (T2 and above) Tumors (P value calculated by ANOVA) (n = 130)
BiomarkersTa/Tis (n = 33)T1 (n = 18)T2 and higher (n = 79)P
  1. ANOVA: analysis of variance. MaxPos: combination of the maximum staining intensity and the percentage of positive staining.

Gelsolin90.0 ± 9.5126.5 ± 12.0153.0 ± 13.9< 0.05
E-cadherin138.7 ± 13.8161.0 ± 14.6120 ± 10.8> 0.05
p5332.1 ± 4.638.6 ± 4.368.8 ± 8.5< 0.05
Ki6735.7 ± 5.740 ± 5.231.5 ± 8.0> 0.05

Relationship of Marker Expression Versus Tumor Recurrence

The association between marker expression and tumor recurrence was analyzed using both univariate and multivariate approaches. Table 3 shows the results of univariate analysis, in which the mean of labeling parameters (Max, Pos, and MaxPos) of each marker was compared between patients with tumor recurrence versus those without tumor recurrence, after stratification by tumor grade (Grade 1–2 vs. Grade 3). We analyzed the information on tumor/no tumor recurrence for 113 patients, 38 and 75 of whom had Grade 1/2 and Grade 3 tumors, respectively. Patients with Grade 1 and Grade 2 tumors who experienced subsequent tumor recurrence had significantly higher Ki67 MaxPos and p53 Max, as well as lower E-cadherin MaxPos, compared with patients who did not have tumor recurrence (P < 0.05 for all). For patients with Grade 3 tumors, however, the Gelsolin Max value was significantly higher in patients with tumor recurrence compared with patients with no tumor recurrence (P < 0.05).

Table 3. Mean of Individual Marker Expression (Max, Pos, and MaxPos) vs. Recurrence Status Stratified by Tumor Grade (Grade 1/2 vs. Grade 3)
Recurrence statusNo.Ki67p53GelsolinE-cadherin
  • Max: maximum staining intensity; Pos: percentage of positive staining; MaxPos: combination of Max and Pos.

  • a

    P value determined by Student t test.

Grade 1/2             
 No recurrence121.212.320.30.417.330.91.656.1124.02.381.0192.0
 Pa   < 0.05< 0.05       < 0.05
Grade 3             
 No recurrence391.315.928.71.229.758.91.465.1135.01.765.3139.0
 Pa       < 0.05     

Multivariate analysis was performed for those who had complete follow-up results and informative data across all markers (n = 68). Initial analysis showed that for each marker analyzed, including p53 and Ki67, Max outperformed Pos or MaxPos in predicting tumor recurrence. Table 4 shows the results of logistic regression analysis. For the multivariate analysis, the significant predictors for the probability of tumor recurrence were low tumor grade (Grade 1–2 of 3) and higher Gelsolin and p53 Max. Tumor stage and the expression of Ki67 and E-cadherin were not associated significantly with the probability of tumor recurrence. The association of tumor recurrence with low tumor grade is not surprising because low-grade tumors tend to recur. In addition, patients with low-grade tumors are less likely to be treated with definitive therapy (e.g., cystectomy) compared with patients with high-grade tumors. In this study, 12% of patients with low-grade tumors received a cystectomy, compared with 88% of patients with high-grade tumors (P < 0.001 by chi-square test). Further analysis of tumor grade and tumor recurrence after stratification by treatment methods revealed no significant differences among patients with low and high-grade tumors treated with cystectomy.

Table 4. Multivariate Analysis: Significant Factors Associated with Recurrence Status by Logistic Regression Analysis (n = 68)
FactorsOR95% CIP
  • OR: odds ratio; CI: confidence interval; Max: maximum staining intensity.

  • a

    Bold numbers indicate significant factors.

Ki67 Max1.210.58–2.510.62
p53 Max2.661.19–5.950.017
Gelsolin Max2.601.22–5.510.013
E-cadherin Max1.040.55–1.950.914

We further analyzed whether marker expression correlated with tumor recurrence-free time using the Cox regression model, in which all four markers (Max value) were examined in conjunction with the patient's age, tumor grade, and stage. All tumor markers reached statistical significance (Table 5). Again, similar findings were seen when the Pos or MaxPos value was used (data not shown). For tumor markers, the strongest association was increased Ki67, followed by increased Gelsolin, increased p53, and decreased E-cadherin expression, in a decreased order of significance. Patient's age and tumor grade were associated negatively with early tumor recurrence.

Table 5. Multivariate Analysis of Marker Expression Versus Recurrence-Free Time (OR and 95% CI) Derived from Cox regression analysisa
CharacteristicsAll tumors (n = 68)GradeStage
1–2 (n = 29)3 (n = 41)Ta/T0/T1 (n = 33)≥T2 (n = 37)
  • OR: odds ratio; CI: confidence interval.

  • a

    Bold numbers indicate P < 0.05; maximum values used.

Age0.73 (0.66–0.80)0.64 (0.52–0.78)0.74 (0.63–0.87)0.67 (0.55–0.81)0.75 (0.66–0.86)
Grade0.07 (0.02–0.21)0.30 (0.05–1.79)0.14 (0.02–0.88)
Stage2.07 (1.12–3.62)4.45 (1.42–14.0)2.58 (1.06–6.31)
Ki672.55 (1.45–4.47)2.28 (0.69–7.52)3.10 (1.11–8.67)3.76 (1.38–10.3)1.43 (0.57–3.55)
p531.84 (1.26–2.69)2.08 (1.06–4.11)1.58 (0.92–2.70)1.96 (0.92–4.20)1.38 (0.86–2.21)
Gelsolin2.08 (1.31–3.31)1.09 (0.37–3.21)3.21 (1.51–6.83)1.76 (0.73–4.25)2.17 (1.15–4.12)
E-cadherin0.62 (0.43–0.89)0.45 (0.24–0.85)1.15 (0.61–2.18)0.32 (0.29–1.02)1.00 (0.50–2.00)

Because the negative association between tumor grade and tumor recurrence-free time may be related to treatment method, as indicated above, we performed Cox regression analysis separately in cases stratified for low or intermediate grade (Grade 1/2) versus high grade (Grade 3) and for noninvasive or superficially invasive (Tis/T0/T1) versus deeply invasive (T2 or above) tumors. For the low to intermediate-grade tumors (Grade 1–2), both E-cadherin and p53 were significant markers in addition to patient's age and tumor stage. However, for the high-grade tumors, increased Gelsolin was the strongest indicator for early tumor recurrence (P = 0.0025, Table 5), followed by Ki67 (P = 0.031, Table 5). When superficial (Ta/T0/T1) and deeply invasive (T2 and above) tumors were analyzed separately, Ki67 was the only marker significant for the former group, whereas Gelsolin was the only marker significant for the latter group.

Figure 3 shows the Kaplan–Meier curve of tumor recurrence time versus staining intensity of Gelsolin in high-grade tumors. No tumor recurrence was seen in tumors with negative Gelsolin staining. The median recurrence time for weak (1+), moderate (2+), strong (3+), and strongest (4+) staining was 10 years, 7 years, 4 years, and less than 1 year, respectively. Our data show that increased Gelsolin staining intensity in tumor areas is the most important predictor for decreased tumor recurrence-free time in high-grade tumors.

Figure 3.

Kaplan–Meier curves of Gelsolin protein expression (maximum staining intensity [Max]) as a predictor for time to tumor recurrence in 41 Grade 3 tumors (0, no stain; 1, weak; 2, moderate; 3, strong; 4, very strong; P = 0.0061 by log rank test).

Longitudinal Association of Individual Marker Expression with Tumor Progression

To further test whether an abnormal marker expression in the primary tumor might predict the behavior of an individual tumor, we compared the expression level (MaxPos value) of markers in tumors with and without progression in a longitudinal fashion. Among 146 patients, we identified 21 patients who had more than one tumor specimen collected at least 1 month apart and had an adequate amount of tissue to be analyzed. These samples were included in our TMA analysis. For these 21 cases, paired tissue samples from the first and last visit were analyzed for the relationship of tumor progression and marker expression. Tumor progression was defined as an increase in either grade (e.g., from Grade 1 to Grade 3) or stage (e.g., from Stage 1 to 2). The patients with (n = 14) and without tumor progression (n = 7) had a similar length of follow-up time (an average of 25 and 21 months, respectively). Figure 4 compares the marker expression (MaxPos value) in the primary tumor samples of patients with and without tumor progression using Boxplot. Patients with tumor progression showed a significantly higher level of expression for Gelsolin than the patients without tumor progression in the primary tumor samples (P = 0.007), indicating that increased Gelsolin in a primary tumor sample predicts progressive behavior in patients with recurrent tumors. Although Ki67 and p53 expression was increased and E-cadherin was decreased in progressing tumors, these differences did not reach statistical significance. There was no statistically significant difference in marker expression between the primary and recurrent tumor samples of individual patients.

Figure 4.

Boxplot of marker expression (maximum staining intensity and percentage of positive staining [MaxPos]) of the primary tumors in 21 patients with tumor recurrence. The primary tumor samples are compared with the samples of tumor recurrence in both progressing (“yes”, n = 14) and nonprogressing (“no”, n = 7) recurrence groups.


Although much progress has been made in recent years in identifying molecular events that lead to the development of cancer, the exact mechanisms underlying the evolution of malignant progressive phenotypes remain poorly understood. For instance, the exact biochemical events leading to cancer invasion or recurrence have not been determined. Understanding these mechanisms may have an impact on developing markers that can be used to define an individual's risk for tumor progression and customizing appropriate management strategies for patients.

In general, it has been assumed that alterations of the cytoskeletal protein actin, a major structure and motility factor in the cell, play an important role in cell invasion and metastasis, even though the exact mechanisms remain to be fully elucidated. The actin network is a rather complex structural and functional system of all eukaryotic cells.20 The molecular mechanisms underlying actin remodeling involve several oncogenic signal transduction pathways, the most notable being the small GTPase of the Ras superfamily of proteins Rac, Rho, and Cdc42.21, 22 In addition, the actin polymerization process is regulated by numerous actin binding and regulatory proteins.23, 24 Many of these proteins (e.g., Gelsolin, E-cadherin, and Vinculin) have tumor suppressor functions individually.25–28 Therefore, a comprehensive and simultaneous analysis of all actin-related molecules, rather than one isolated molecule at a time, will be necessary for a more clear understanding of how actin is associated with malignant phenotypic changes.

The TMA technology provides a convenient high-throughput tissue-based tool for in situ gene dosage and protein expression studies.19 The technique has been utilized to profile rapidly a number of molecular markers and the results compare well with those obtained by standard methods.29–32 Using our urothelial carcinoma TMA, we evaluated the two most notable actin binding proteins, Gelsolin and E-cadherin, as markers for urothelial carcinoma recurrence and progression. We also compared their expression patterns with clinicopathologic characteristics of tumor progression and with the expression patterns of two other well studied markers, Ki67 and p53.

Gelsolin is an important regulator of actin cytoskeleton dynamics. It severs assembled actin filaments and caps the fast-growing plus end of a free or newly severed filament.10 It is a prominent substrate of caspase-3 activity in vitro and, therefore, is an important effector for cellular apoptosis.13 Gelsolin expression is down-regulated in 60–90% of breast,33 bladder,34 prostate, and lung carcinomas.35 Tanaka et al.12 demonstrated that the addition of retroviral Gelsolin cDNA constructs resulted in marked and reproducible tumor growth inhibition and prolonged survival time in the majority of animals tested. These findings demonstrate the tumor suppressor effect of this gene and the potential for treating human urinary bladder carcinomas with the Gelsolin gene. However, there has been no in vivo study of Gelsolin expression in actual human bladder carcinoma samples. Our study on urothelial carcinoma TMA revealed a distinctive biphasic protein expression pattern of the Gelsolin gene in urothelial carcinoma. Compared with the distant benign field, expression decreased in both premalignant and malignant lesions, as would be expected from in vitro observations. The decreased expression is more striking in the noninvasive lesions, including dysplasia, low-grade tumor, and even CIS. However, there is a trend of increased expression from noninvasive (Ta/T0), to superficially invasive (T1), and to deeply invasive (T2 plus) tumor. A higher Gelsolin labeling index (Max or MaxPos) is an independent marker for tumor recurrence and progression, particularly for high-grade tumors. Therefore, acquiring a certain level of Gelsolin expression may be necessary to convert a noninvasive tumor to an invasive tumor. This is consistent with the notion that cytoskeletal proteins play an important role in tumor cell motility and invasion.

A similar Gelsolin expression pattern has been seen in other carcinoma types, such as non-small cell lung carcinoma,35 breast carcinoma,36 and clear cell carcinoma of the kidney (unpublished data). Thor et al.36 showed that overexpression of Gelsolin, epidermal growth factor receptor (EGFR), and erbB-2 significantly predicted poor clinical outcome by univariate and multivariate analyses. In lymph node-positive patients, coexpression of all three markers was associated with a 3-year disease-specific survival (compared with erbB-2-positive, EGFR-positive, and Gelsolin-negative patients who had a median survival of 6 years). The biphasic expression pattern of Gelsolin is intriguing. First, it suggests that epigenetic mechanisms may be involved in the regulation of Gelsolin gene expression. Second, the increased expression from a noninvasive to an invasive tumor supports the hypothesis that cytoskeletal actin-associated motility machinery plays an important role in tumor invasion. Third, in view of the positive association between increased Gelsolin expression and tumor progression found in this study, caution should be taken in the effort of developing therapeutic interventions based on Gelsolin regulatory strategies.

Our study confirmed observations from a number of recent studies that decreased expression of E-cadherin occurs in urothelial carcinoma, primarily in the high-grade tumors.14–17 However, our study also showed that decreased expression of E-cadherin is associated with tumor recurrence in low to intermediate-grade tumors. Bornman et al.14 showed that one of the potential mechanisms for decreased expression of E-cadherin in urothelial carcinoma is related to DNA methylation. Altered methylation is evident in the nontumor areas of older patients. However, in their study, altered methylation of E-cadherin did not correlate entirely with decreased expression in the nontumor areas. This later finding suggested that DNA methylation may not be specific for the down-regulation of E-cadherin in bladder carcinoma. In many of these previous studies, abnormal E-cadherin expression was defined arbitrarily by the complete absence of immunoreactivity or by the heterogeneous staining of tumor areas. In our study, semiquantitative Max and MaxPos values were used. Nevertheless, the overall findings were similar.

The purpose of this study was to examine the association between the expression of Gelsolin and E-cadherin, two important actin-associated proteins, and urothelial carcinoma progression. A wide spectrum of lesions were analyzed. However, a small sample size for some of these lesions (e.g., only six Grade 1 tumors were analyzed in this study) precluded a comprehensive analysis. Although this was a limitation of the study, our analysis provided strong support for the use of Gelsolin as a potential important prognostic marker for high-grade urothelial carcinomas. In future analyses, we will focus specifically on high-grade superficial urothelial carcinomas. Most bladder tumors (70–80%) are classified as superficial.37 Between 50 and 70% of patients with superficial tumors will develop new superficial TCC, often within 12 months of diagnosis. In 10–20% of these patients, these tumors will progress to infiltrate the muscular propria (T2 plus).38 Because only the progressive tumors will eventually lead to cancer-related death, it is crucial to identify these patients early so that more aggressive treatment methods can be applied. In future studies, we will determine the optimal cutoff threshold for each marker to distinguish between less aggressive and more aggressive tumors, as well as the optimal combination of marker profiles for prognostic indication of these tumors.

When Ki67 and p53 staining was converted to a positive/negative schema using the threshold reported in the literature (20% of cells staining for either marker),39, 40 Ki67 was positive in 75% of CIS lesions and in 17% of Grade 1, 55% of Grade 2, and 50% of Grade 3 tumors. p53 was positive in 38% of CIS lesions and in 17% of Grade 1, 38% of Grade 2, and 53% of Grade 3 tumors. These findings were similar to those reported in the literature. However, when marker expression was analyzed against tumor recurrence, the Max value of each marker provided stronger associations (small P values) than Pos or MaxPos values. In the subsequent analysis of marker expression versus tumor recurrence, we presented the results based on Max value, even though similar findings were seen when Pos or MaxPos values were used. This finding is slightly different from the results of other investigators.39, 40

In summary, we demonstrated that the expression patterns for the cytoskeletal proteins Gelsolin and E-cadherin are distinctive from Ki67 and p53. Both of these protein markers show an overall decreased expression compared with corresponding benign fields. However, Gelsolin is decreased mainly in premalignant and preinvasive lesions. In tumors, Gelsolin expression generally increases with increasing grade and stage. Elevated expression of Gelsolin relative to premalignant lesions is a strong indicator for the probability of tumor recurrence, early tumor recurrence, and tumor progression, particularly for high-grade tumors. Conversely, E-cadherin decreases as the tumor grade increases. This finding provides in vivo evidence to support the theory that distinctive patterns of actin family gene expression may be observed in different stages of the malignant transformation process. Whereas decreased expression of Gelsolin occurs at the early stages of malignant transformation, increased Gelsolin expression probably plays a critical role in converting the less aggressive superficial tumor to a more aggressive invasive tumor. Our study shows that it is worthwhile to perform further studies to examine the utility of these markers as prognostic indicators for urothelial carcinomas.