Fax: (011) 34 96 3864173
Cytokeratin expression patterns in low-grade papillary urothelial neoplasms of the urinary bladder
Version of Record online: 1 APR 2003
Copyright © 2003 American Cancer Society
Volume 97, Issue 8, pages 1876–1883, 15 April 2003
How to Cite
Ramos, D., Navarro, S., Villamón, R., Gil-Salom, M. and Llombart-Bosch, A. (2003), Cytokeratin expression patterns in low-grade papillary urothelial neoplasms of the urinary bladder. Cancer, 97: 1876–1883. doi: 10.1002/cncr.11265
- Issue online: 1 APR 2003
- Version of Record online: 1 APR 2003
- Manuscript Accepted: 30 DEC 2002
- Manuscript Revised: 5 DEC 2002
- Manuscript Received: 23 SEP 2002
- bladder neoplasms;
- 34βE12 antigen;
The differential expression patterns of cytokeratin 20 (CK20) and 34βE12 antigen in low-grade papillary urothelial tumors of the bladder are discussed.
A retrospective study of 120 patients with low-grade papillary bladder tumors (45 neoplasms of low malignant potential and 75 low-grade WHO G1 carcinomas) was performed. All tumors were graded in accordance with the 1998 World Health Organization/International Society of Urological Pathology (WHO/ISUP) and 1999 WHO classifications. The mean follow-up was 76.6 months (range, 36–168 mos), considering for prognostic purposes the time to first recurrence, or relapse-free interval (RFI), and the total number of recurrent patients. Immunohistochemically, normal or abnormal CK20 and 34βE12 antigen expression patterns were determined for each patient. CK20 (clone IT-Ks) and a high-molecular weight cytokeratin (clone 34βE12) were the monoclonal antibodies used in the immunohistochemical study.
Seventy-seven of 120 patients (64.2%) experienced a recurrence during follow-up. In recurrence prediction, the differential expression pattern of both cytokeratins showed a high sensitivity (76.6% for CK20 and 80.5% for 34βE12 antigen) and a high positive predictive value (85.5% for CK20 and 75.6% for 34βE12 antigen), although specificity was higher for CK20 (76.7%) than it was for 34βE12 antigen (53.4%). Independent of adjuvant intravesical chemotherapy, these 2 markers showed a strong statistical correlation (p < 0.001) in univariate studies with both the prediction of disease recurrences and RFI.
CK20 and 34βE12 antigen have proved to be strong predictive markers of disease recurrences when considering different topographic expression profiles, and, in the authors' opinion, these profiles could be incorporated into follow-up clinicopathologic strategies. Cancer 2003;97:1876–83. © 2003 American Cancer Society.
A new World Health Organization/International Society of Urological Pathology (WHO/ISUP) consensus classification of urothelial bladder neoplasms recently was proposed in an attempt to clarify the morphologic aspects of histologic grade as and to better correlate subjective morphologic features with patient outcome.1 In accordance with this classification system, low-grade bladder neoplasms were divided into papillary neoplasms of low malignant potential (cases with a low recurrence rate and lack of tumor progression) and low-grade (WHO, G1) papillary carcinomas (cases with a high recurrence rate and a slight but nonnegligible possibility of progression). Nevertheless, the distinction between these two subgroups of neoplasms is often subtle, and tumor recurrences have been reported in a significant percentage of both low-grade tumors that initially were diagnosed as papillary neoplasms of low malignant potential (LMP) and low-grade (WHO, G1) papillary carcinomas.2–6 In these studies, a variable risk of tumor progression also was reported, not only in low-grade carcinomas but also in some isolated neoplasms of LMP.2–6 For the assessment of histologic grade, the influence of tumor cell heterogeneity also has been studied.7 As a result, other biologic markers besides histologic grade currently are being investigated. In particular, it is necessary to emphasize the emerging role of diverse cytokeratin epitopes and their topographic distributions within the tumor urothelium; the expression patterns of some cytokeratins, such as CK20 and 34βE12 antigen, have proved to be helpful in predicting the biologic behavior of papillary urothelial bladder lesions.8–15
The cytokeratin family consists of a broad spectrum of intermediate filaments that are expressed by a number of human epithelial and mesothelial cells.8 In the specific study of the bladder epithelium (urothelium), several cytokeratin epitopes have been immunohistochemically identified. These cytokeratins are expressed by different cell layers within the normal urothelium.8–15 Specifically, CK20 normally is expressed by superficial cells and some isolated intermediate cells,9–13 whereas 34βE12 antigen expression is limited to the basal cell layers.14, 15 In their papillary neoplastic counterparts, these normal expression patterns may be maintained, lost, or altered as the tumor grade or stage increases, which could be indicative of not only the level of epithelial maturation8 but also patient outcome. The predictive value of these two markers (CK20 and 34βE12) recently has been reported in recurrence prediction.9, 11, 15 However, these studies were based on heterogeneous tumor populations, which included superficial (pTa–pT1) tumors, low-grade neoplasms, and urothelial dysplasias.9–15
We studied the comparative immunoexpression profiles of CK20 and 34βE12 antigen in a series of 120 patients with low-grade papillary urothelial neoplasms of the bladder in order to clarify which marker better predicts the risk of recurrence and might consequently facilitate the clinical management of these patients.
MATERIALS AND METHODS
The study group consisted of 120 patients with noninvasive (Ta, 116 patients) and undetermined-stage (Tx, 4 patients) papillary bladder tumors (TNM classification). Infiltration of the lamina propria was not considered if tumor nests were invaginated within the lamina propria without an inflammatory response around the tumor nests. The integrity of the lamina propria could not be ascertained for four patients (Tx) in whom a dense inflammatory reaction was observed. In these instances, a minimal depth of infiltration could not be discarded absolutely, and subsequently these cases were included in our series as likely noninvasive or microinvasive carcinomas. All cases were collected from January 1984 to November 1995. To minimize variability among different observers, the same pathologist assigned the histologic grade twice using criteria from the 1998 WHO/ISUP1 and 1999 WHO classifications. All cases were labeled as papillary neoplasms of low malignant potential (LMP) or low-grade (WHO, G1) papillary carcinomas.
Clinicopathologic data were obtained from cystoscopic, histopathologic, and operative reports. All histologic samples were initial biopsies with no previous treatments administered. At initial diagnosis all patients underwent transurethral resection (TUR) of the lesion; 64 patients were treated with TUR only, 35 were treated with TUR + mitomycin C (MMC), 20 were treated with TUR + bacillus Calmette–Guerin (BCG), and only 1 patient was treated with TUR + triethylenethiophosphoramide (thiotepa). Administration of intravesical agents did not follow any specific therapeutic protocol. However, 6 out of 20 patients who were treated with BCG and 9 out of 35 who were treated with MMC were not treated with intravesical chemotherapy until the first tumor recurrence. Consequently, a group of 79 patients was considered to be the TUR alone cohort, whereas the remaining 41 patients were considered to be the adjuvant chemotherapy cohort.
Patients were monitored with urinary cytologies, cystoscopies, and cystoscopies with multiple random biopsies. The first cystoscopic examination was performed at 3, 6 or 9 months after the initial diagnosis. Follow-up cystoscopies were performed every 3–6 months during the first 2 years, and every 6–12 months for the next 2. If patients remained free of disease, control cystoscopies were done every 12 months. However, the follow-up program was restarted if a tumor relapse was confirmed.
Tumor recurrence was defined as a positive finding on cytology, cystoscopy or biopsy specimens. Cases deemed positive by cytologic examination required confirmation by cystoscopy or biopsy reports. Time to first recurrence was referred to in our study as relapse-free interval (RFI). At the time of analysis, 77 patients (64.2%) had experienced recurrence, whereas the remaining 43 patients (35.8%) were alive with no evidence of disease (censored). Tumor progression was considered to be an increase in tumor grade or stage to high-grade or muscle-invasive carcinoma. During follow-up, three patients died as a result of disease, and eight patients died as a result of other causes.
Immunostaining Procedure and Evaluation
The immunohistochemical study was carried out on 4-micron-thick sections following the avidin-biotin complex (ABC) peroxidase method. Primary monoclonal antibodies were CK20 (clone IT-Ks 20.8; DAKO, Glostrup, Denmark) and a high-molecular weight cytokeratin specific for cytokeratins 1, 5, 10, and 14 (clone 34βE12; DAKO). The optimal dilution for primary antibodies was 1:100 for CK20 and 1:25 for 34βE12 antigen. All samples previously were treated for antigen retrieval heating in an autoclave (1.5 atm, 3 min) with a pH 6.0 sodium citrate buffer.
The immunoreaction was topographically evaluated and scored by the same pathologist. In accordance with previous reports, the CK20 staining pattern was considered normal when an intense reaction was observed in the apical cells together with a lack of expression or mild expression in the remaining cell layers, whereas diffuse or absent CK20 immunostaining was considered an abnormal staining pattern (Fig. 1A,B). 34βE12 antigen expression was considered normal when the reaction was limited to the basal cell layers, whereas diffuse staining was referred to as abnormal (Fig. 1C,D). With regard to 34βE12 antigen expression, the discovery of both a normal and abnormal staining pattern within the same lesion was not uncommon. Such instances were labeled as “mixed” 34βE12 expression patterns. The positive controls were non-neoplastic bladder mucosae from ten patients who had undergone surgery for nonmalignant conditions.
In univariate studies, the chi-square and Kruskal-Wallis tests were used to contrast differences between factors. Kaplan–Meier estimates then were generated to calculate the correlation of the two variables (CK20 and 34βE12) with relapse-free interval (RFI). A log rank test subsequently was used to contrast differences between factors. A specific study of progression was not feasible because only 11 out of 120 patients (9.26%) experienced progression in either grade (7 patients) or stage (4 patients) according to the WHO/ISUP and TNM classification systems. All statistical studies were performed with commercially available software (SPSS/PC 10.0 program package; Idessa Consulting, Valencia, Spain). For all statistical tests a significance level of 0.05 was adopted, with P < 0.001 considered to be the highest level of significance.
The mean follow-up for the whole series was 76.6 months (range, 36–128 mos), in contrast to censored patients, whose mean follow-up was 65.8 months (range, 48–99 mos). The mean patient age was 63.6 years (range, 29–93 yrs), and the male-to-female ratio was 5:1. Regarding histologic grade, 45 patients were diagnosed with papillary neoplasms of LMP, and the remaining 75 patients were diagnosed with low-grade (WHO, G1) papillary carcinomas. Undetermined-stage tumors (Tx, four patients) were initially diagnosed as low-grade (WHO, G1) papillary carcinomas; only one progressed to high-grade muscle-invasive carcinoma, whereas the other three recurred but progressed in neither grade nor stage.
In our study, no statistical differences (P = 0.497) were found between the TUR alone and intravesical adjuvant therapy cohorts regarding the total percentages of recurrent and nonrecurrent cases (Table 1).
|Nonrecurrent cases (%)||Recurrent cases (%)||Total (%)||Pa|
|TUR alone cohort||30 (38)||49 (62)||79 (100)|
|Chemotherapy cohort||13 (31.7)||28 (68.3)||41 (100)|
|Total||43 (35.8)||77 (64.2)||120 (100)||0.497|
Topographically, the assessment of a diffuse or absent CK20 staining pattern (Fig. 1A,B) corresponded to a higher percentage of recurrences (94.3% and 76.5%, respectively) than the percentage observed for patients with an apical cell distribution (35.3%). As a result, we consider this apical cell staining to be the normal CK20 expression pattern (64.7% nonrecurrences), and we consider the absent or diffuse staining to be the abnormal CK20 expression pattern (85.5% recurrences) (Table 2). With respect to 34βE12 antigen expression, the observation of basal or mixed staining patterns corresponded to a lower recurrence rate than the rate observed for patients with a diffuse distribution throughout all urothelial cell layers (Fig. 1C,D). We therefore consider both the basal and mixed patterns to be normal 34βE12 antigen expression (39.5% recurrences, compared with 75.6% recurrences for patients showing a diffuse [abnormal] 34βE12 antigen expression pattern [Table 2]). In univariate studies, the differential expression pattern of both cytokeratins demonstrated a strong statistical association (P < 0.001) with recurrence prediction. However, the assessment of histologic grade did not correlate (P = 0.731). In fact, similar percentages of cases diagnosed as neoplasms of LMP (62.2%) and low-grade carcinomas (65.3%) recurred during follow-up (Table 2). Regarding cytokeratin immunoexpression, only abnormal 34βE12 antigen expression appeared to be statistically correlated (P = 0.039) to the administration of adjuvant intravesical agents (Table 2).
|Nonrecurrent cases||Recurrent cases||Total||Pa||TUR alone cohort||Chemotherapy cohort||Pa|
|Normal (“1”)||33 (64.7%)||18 (35.3%)||51 (100%)||< 0.001||32||19|
|Abnormal (“2” + “3”)||10 (14.5%)||59 (85.5%)||69 (100%)||47||22||< 0.540|
|Normal (“1” + “2”)||23 (60.5%)||15 (39.5%)||38 (100%)||< 0.001||30||8|
|Abnormal (“3”)||20 (24.4%)||62 (75.6%)||82 (100%)||49||33||< 0.039|
|LMP neoplasms||17 (37.8%)||28 (62.2%)||45 (100%)||< 0.731||31||14|
|LG carcinomas||26 (34.7%)||49 (65.3%)||75 (100%)||48||27||< 0.585|
|Total||43 (35.8%)||77 (64.2%)||120 (100%)||79||41|
Both cytokeratins showed a strong statistical correlation (P < 0.001) to mean RFI (Table 3). However, no statistical differences in disease-free survival between absent and diffuse CK20 expression patterns (41 and 29 mos, respectively) were encountered. This combined absent plus diffuse (abnormal) CK20 staining pattern group exhibited a mean RFI of 34 months (Table 3; Fig. 2A,B). Similarly, no statistical differences were found between predominantly basal and mixed 34βE12 antigen expression patterns (71 and 66 mos, respectively). This combined basal plus mixed (normal) 34βE12 antigen staining pattern group exhibited a mean RFI of 68 months (Table 3; Fig. 2C,D).
|Expression pattern||Mean RFI||Rank||Expression pattern||Mean RFI||Rank|
|Apical (“1”)||73 mos||(65–82)||Basal (“1”)||71 mos||(54–87)|
|Diffuse (“2”)||29 mos||(21–37)||Mixed (“2”)||66 mos||(48–83)|
|Absent (“3”)||41 mos||(30–53)||Diffuse (“3”)||44 mos||(37–51)|
|Normal (“1”)||73 mos||(65–82)||Normal (“1” + “2”)||68 mos||(57–84)|
|Abnormal (“2” + “3”)||34 mos||(25–46)||Abnormal (“3”)||44 mos||(37–51)|
The combined expression profile of both cytokeratins simultaneously (Table 4) was also taken into account; only 11 patients showed apical cell immunostaining for CK20 together with a basal cell distribution for 34βE12 antigen. Of these 11 patients, only 1 (9.09%) recurred. Likewise, if we considered both (a predominantly basal plus “mixed”) distribution of 34βE12 antigen staining as being the normal expression pattern, a similar percentage of cases to that observed with only the basal cell distribution recurred (10.53%). In contrast, 28 patients showed a diffuse expression pattern for both cytokeratins simultaneously, and 96.5%; of these patients presented at least 1 tumor recurrence during follow-up. Similarly, being a diffuse plus absent distribution of CK20 staining considered the abnormal CK20 expression pattern, along with a diffuse distribution with 34βE12 antigen, then up to 92% (46 out of 50) of these patients recurred (Table 4). In univariate analyses, the combined expression pattern of both cytokeratins demonstrated a strong statistical correlation (P < 0.001) to both the prediction of recurrences and the mean RFI, or disease-free survival (Tables 4, 5; Fig. 3).
|CK20 apical||CK20 diffuse||CK20 absent||Total|
|34βE12 basal||11 (1)||3 (3)||7 (4)||21 (8)|
|34βE12 mixed||8 (1)||3 (2)||6 (4)||17 (7)|
|34βE12 diffuse||32 (16)||29 (28)||21 (18)||82 (62)|
|Total||51 (18)||35 (33)||34 (26)||120 (77)|
|CK20 normal/ 34βE12 normal (%)||CK20 normal/ 34βE12 abnormal (%)||CK20 abnormal/ 34βE12 normal (%)||CK20 abnormal/ 34βE12 abnormal (%)|
|Nonrecurrent cases||17 (89.5)||16 (50)||6 (31.6)||4 (8)|
|Recurrent cases||2 (10.5)||16 (50)||13 (68.4)||46 (92)|
|Total||19 (100)||32 (100)||19 (100)||50 (100)|
Otherwise, CK20 proved to be a highly sensitive and specific prognostic marker (76.6% and 76.7%, respectively), whereas 34βE12 antigen was also a highly sensitive marker (80.5%) but demonstrated remarkably lower specificity (53.4%) (Table 6). It is noteworthy that the high positive predictive values (PPV) recorded for both markers (85.5% for CK20 and 75.6% for 34βE12 antigen) were significantly greater than the negative predictive values (NPV) that were observed (64.7% for CK20 and 60.5% for 34βE12 antigen), which emphasizes these markers' major utility in predicting tumor recurrences rather than predicting nonrecurrences. Moreover, the statistical significance of these two variables increases in the study of the combined expression patterns of both cytokeratins simultaneously. For this combined analysis, the positive and negative predictive values (PPV and NPV) for CK20 and 34βE12 antigen were 92% and 89.5%, respectively (Table 5).
|Antibody||Sensitivity (%)||Specificity (%)||PPV (%)||NPV (%)|
|CK20||0.766 (76.6)||0.767 (76.7)||0.855 (85.5)||0.647 (64.7)|
|34βE12||0.805 (80.5)||0.534 (53.4)||0.756 (75.6)||0.605 (60.5)|
The natural history of papillary urothelial tumors of the bladder is difficult to predict. Indeed, a high recurrence rate generally has been reported in low-grade bladder lesions (papillary neoplasms of low malignant potential and low-grade WHO, G1 papillary carcinomas) when long-term follow-up programs are performed.2–7 Similarly, in our study, which exclusively focused on low-grade bladder tumors, more than 60% of patients (77 out of 120) presented at least 1 tumor relapse during a mean follow-up of 76.6 months. Specifically, 62.2% of 45 neoplasms of LMP and 65.3% of 75 (WHO, G1) low-grade carcinomas recurred in our series independently of adjuvant intravesical chemotherapy. Consequently, we agree with many authors in believing that, in addition to histologic grade, other markers need to be investigated so as to better predict the risk of disease recurrences and, therefore, to better determine in which subgroup of low-grade bladder lesions the number of control follow-up cystoscopies may be reduced.2–7
Although diverse cytokeratin epitopes are immunohistochemically expressed by tumor cells in papillary bladder lesions, the presence of apical cell expression of CK20,9–13 along with basal cell expression of 34βE12 antigen,14–15 has been reported to be a strong predictor of non-recurrent disease, especially in low-grade tumors. In these lesions, these topographic immunoreactions may be indicative of not only a maturative status but also a premalignant condition.8 In addition, these two staining patterns, referred to as the normal expression profile, overlap with the patterns observed in the nontumoral urothelium of the bladder.
Desai et al.9 recorded a 44.9% rate of nonrecurrence when a normal CK20 expression pattern was ascertained, in contrast to the 63.2% of recurrences related to an abnormal staining profile. Similarly, Alsheikh et al.10 found in low-grade bladder tumors a 75% rate of nonrecurrence when normal CK20 immunoexpression was seen, although up to 50% of patients showing the abnormal staining pattern did not experience recurrence in their series. Likewise, Harden et al.11 observed a 57.1% (8 out of 14 patients) rate of nonrecurrence when the restricted apical expression pattern of CK20 was present. In another study,12 Harden et al. also found a 100% rate of nonrecurrent disease in the presence of a normal CK20 immunoprofile. However, this latter study was performed in a series of 58 (G1–2, pTa) papillary bladder tumors, with only 10 patients exhibiting the normal CK20 expression pattern and with a mean follow-up of 18 months. In the current study, we observed an apical cell distribution of CK20 (normal expression pattern) in 51 of 120 patients (42.5%), among whom a high percentage (64.7%, 33 of 51 patients) did not recur during a mean follow-up of 65.8 months (censored). Nevertheless, in 35.3% of the 51 patients showing this normal CK20 expression pattern, at least 1 tumor relapse was recorded, which constitutes a higher percentage than what had been observed in previous investigations.9–13
Helpap et al.14–15 studied 122 pTa/pT1 papillary bladder tumors, including low-grade papillary neoplasms and some urothelial papillomas. In these reports,14–15 a close relation between an abnormal 34βE12 antigen expression pattern and cell proliferation index (mean MIB-1 score) was demonstrated. Furthermore, the relation between 34βE12 antigen expression pattern and recurrence was determined for a total of 28 patients who experienced recurrence. Helpap et al. found that a basal 34βE12 distribution indicated a lower relapse rate in comparison with the rate associated with a diffuse (abnormal) expression profile.14–15 Seven recurrent patients (25%) exhibited a basal 34βE12 staining pattern, whereas the remaining 21 (75%) stained throughout all urothelial cell layers.14–15 In our series, only 21 out of 120 patients (17.5%) stained exclusively in the basal cell layers of the urothelium, and 61.9% of these 21 remained free of disease. However, another 17 patients (14.2% of the series) exhibited 2 different expression patterns within the same lesion. In addition, similar percentages of patients exhibiting this “mixed” distribution (58.8%) and patients with a predominantly basal distribution (61.9%) showed no recurrence. As a result, we consider these two staining patterns (basal plus mixed) to be the normal 34βE12 immunoprofile.
In agreement with the reviewed literature, we found that many patients who exhibited a normal staining pattern for CK20 (64.7%) and 34βE12 antigen (60.5%) had no evidence of disease for more than 5 years of follow-up. In contrast, the prediction of tumor relapse on the basis of an abnormal staining pattern was notably higher in our series (85.5% for CK20 and 75.6% for 34βE12) in comparison with previous reports. These data might indicate that, in low-grade bladder neoplasms, the assessment of abnormal expression patterns for both CK20 and 34βE12 antigen could be a better predictive marker than the assessment of a normal staining. In this regard, we have also demonstrated a strong statistical correlation (P < 0.001) between cytokeratin immunoprofile and both the prediction of recurrences and the mean RFI. However, no statistical differences were encountered between diffuse and absent immunostaining for CK20 expression (P = 0.156) or between basal and mixed immunostaining for 34βE12 antigen expression (P = 0.232). Thus, from a statistical point of view, these two combined immunostainings could be considered the abnormal CK20 and normal 34βE12 antigen expression patterns. In regard to the influence of adjuvant intravesical agents, only the abnormal expression of 34βE12 antigen appeared to be more frequently associated with the subgroup of patients treated with adjuvant chemotherapy compared with the TUR alone cohort. However, CK20 expression patterns were not statistically influenced by the instillation of intravesical agents, a finding that reinforces the predictive value of this marker.
This study was the first to examine the prognostic value of analyzing the combined expression patterns of CK20 and 34βE12 antigen simultaneously. In this regard, 89.5% of patients who exhibited the combined normal expression pattern did not recur in our series, although only 19 patients (15.8% of the series) presented this particular staining pattern. Most notably, we found that 92% of the 50 patients showing a combined abnormal expression pattern had recurrences. This information reveals anew the importance of assessing this abnormal immunoprofile in low-grade bladder neoplasms. The importance of making this assessment is closely related to the high sensitivity and high positive predictive value of these two variables considered alone or in combination. These results regarding sensitivity and positive predictive value disagree with the reviewed literature,9–15 which stresses the importance of these two immunohistochemical markers in predicting nonrecurrent disease rather than recurrent disease.
In conclusion, our study emphasizes that, independently of the administration of adjuvant intravesical chemotherapy, the assessment of abnormal expression patterns for CK20 and 34βE12 antigen is feasible and important in the prediction of disease recurrences and disease-free survival for patients with low-grade papillary bladder neoplasms. Moreover, regardless of the high sensitivity and high positive predictive value of these two immunohistochemical markers, a combined study with both cytokeratins is necessary for better prediction of disease recurrence during follow-up. Therefore, the demonstration in these lesions of simultaneous abnormal staining patterns for CK20 and 34βE12 antigen will require, in our opinion, repeat follow-up cystoscopies. However, if simultaneous normal expression patterns are observed, then the number of control cystoscopies may be reduced. Nevertheless, further studies that include larger series of patients and take into account the administration of intravesical adjuvant chemotherapy are necessary to corroborate with absolute confidence the predictive value of cytokeratin expression patterns in low-grade papillary neoplasms of the urinary bladder.