Proteomics and Immunohistochemistry Define Some of the Steps Involved in the Squamous Differentiation of the Bladder Transitional Epithelium: A Novel Strategy for Identifying Metaplastic Lesions¹

Bladder cancer comprises a broad spectrum of tumors, including TCCs,⁴ SCCs, adenocarcinomas, small cell carcinomas, and leiomyomas (1). TCCs are, by far, the more prevalent tumors and represent nearly 90% of all bladder cancers in the Western Hemisphere. SCCs, on the other hand, encompass a small percentage (2–3%) of all bladder lesions diagnosed in Europe and America but are very frequent (80%) in areas of Africa and the Middle East, where Schistosoma hematobium, a parasite that induces bladder SCCs in humans, is prevalent (2). SCCs often arise in patients who have a history of many years of chronic inflammation, keratinizing squamous metaplasia, and bladder stones. These tumors are highly malignant, and therefore, the success of treatment relies heavily on early detection.

SCCs closely resemble epidermal keratinocytes both in morphology and in proteome expression profiles (3, 4) ⁵ and exhibit distinct squamous features such as “pearl” formation and keratohyalin bodies. Grading of these tumors is subjective and takes into consideration the degree of nuclear polymorphism, nuclear:cytoplasmic ratio, and chromatin clumping as well as the number of mitotic cells (5). These parameters, however, are difficult to evaluate with precision, and as a result, it is often impossible to distinguish poorly differentiated TCCs with areas of squamous differentiation from highly undifferentiated SCCs. In addition, different areas from the same tumor often show variable degrees of differentiation and/or metaplasia.

The histogenesis of SCCs is at present unknown, although there is some preliminary evidence indicating that these lesions may arise from: (a) extensive squamous differentiation of TCCs, e.g., from CIS or high-grade papillary TCCs; and (b) direct neoplastic transformation, on the basis of squamous metaplasia, of the bladder urothelium (6, 7, 8). Knowledge of the molecular mechanisms underlying the conversion of transitional to stratified epithelium as well as of the various steps leading to premalignant transformation is expected to accelerate the development of tests for the early detection of these lesions and may provide new targets for intervention.

Here, we present a novel strategy for dissecting some of the steps involved in the squamous differentiation of the bladder urothelium. The approach first makes use of proteomic technologies (9, 10) ⁵ to reveal and identify proteins that are differentially expressed in SCCs and normal urothelium. Thereafter, specific antibodies against the differentially expressed proteins are use to stain serial cryostat sections (immunowalking) of biopsies obtained from SCC-bearing patients who have undergone removal of the bladder due to invasive disease (cystectomy). Because bladder cancer is a field disease (11), i.e., a large part of the bladder lining is at risk of developing disease, we surmised that the urothelium of these patients may exhibit a spectrum of abnormalities ranging from early stages of metaplasia to invasive disease. Besides dissecting some of the steps involved in the squamous differentiation of the bladder urothelium leading to SCCs, our studies have shown that it is feasible to apply powerful proteomic technologies to the analysis of complex biological samples under conditions that are close as possible to the in vivo situation.

One hundred fifty fresh tumor samples and a few cystectomy samples, taken at Skejby Hospital, Aarhus, Denmark, were placed immediately on ice and transported to the Department of Medical Biochemistry at The University of Aarhus. Tumor pieces for cryostat sections were kept at −80°C. The three SCC tumors reported in detail in this study (864-1, 884-1, and 896-1) were invasive and did not exhibit urothelial components, as judged from the pathologist’s reports. The rest of the tumors corresponded mainly to TCCs, and a few exhibited extensive squamous differentiation.

Tumors that were clear of clots and contaminating tissue were minced in small pieces with the aid of a scalpel and were homogenized (with a glass homogenizer) in lysis solution (12) prior to electrophoresis. In most cases, one or two tumor pieces were labeled with [³⁵S]methionine for 14–16 h in a 10-ml sterile plastic conical tube containing 0.2 ml of modified Eagle’s medium lacking methionine and containing 2% dialyzed (against 0.95% NaCl) FCS and 100 μCi of [³⁵S]methionine (Amersham-Pharmacia-Biotech, Hørsholm, Denmark; Ref. 13). At the end of the labeling period, the pieces were centrifuged at 2000 × g for 2 min, resuspended in 0.3–0.4 ml of lysis solution, and homogenized in a small glass homogenizer.

Normal bladder urothelium was dissected with the aid of an scalpel and labeled with [³⁵S]methionine as described above (13). The purity of the preparations was assessed by monitoring for the presence of vimentin (connective tissue), desmin (smooth muscle), and T-plastin (WBCs).

Two-dimensional PAGE was performed essentially as described by Celis et al. (13).

Proteins were identified using a combination of procedures that included microsequencing and mass spectrometry (14, 15, 16), two-dimensional gel Western immunoblotting, and enhanced chemiluminescence detection (17) as well as by comparison with the master two-dimensional gel image of human keratinocyte proteins (10, 18).⁵

Monoclonal antibodies specific for keratins 7, 8, 10, 13, 14, 18, and 19 were purchased from Cappel (Durham, NC). The mAb against keratin 20 was purchased from Cymbus Biotechnology (Hants, United Kingdom). The specificity of these antibodies was determined by two-dimensional PAGE Western immunoblotting (isoelectric focusing; nonequilibrium pH gradient electrophoresis). mAb BG3C8, which reacts specifically with epidermal basal cells and myoepithelial cells, has been described previously (19). Rabbit polyclonal antibodies reacting with unknown antigen(s) in the suprabasal layers of the epidermis were prepared in the laboratory. Monoclonal antibodies specific for psoriasin were prepared as described previously (7).

Samples from tumor biopsies, random biopsies, and cystectomies, frozen in liquid nitrogen and kept at −80°C, were sectioned on a Reichert-Jung cryostat. Eight-μm sections were placed on coverslips, washed three times with HBSS, and treated for 10 min with 4% formaldehyde. After being washed extensively with HBSS, the coverslips were covered with 20 μl of the primary antibody and incubated for 45 min at 37°C in a humid environment. The coverslips were washed several times with HBSS and covered with 20 μl of rhodamine-conjugated secondary antibody (diluted 1:50 in HBSS). After 45 min of incubation at 37°C in a humid environment, the coverslips were washed thoroughly with HBSS and mounted in fluorescent mounting medium from DAKO A/S (Glostrup, Denmark). Observations were made on a Leica photomicroscope equipped with epifluorescence.

One hundred fifty fresh biopsies from bladder tumors resected at Skejby Hospital were analyzed by two-dimensional PAGE in combination with autoradiography and/or Coomassie Brilliant Blue staining. Of these, only three (SCCs 864-1, 884-1, and 896-1) resembled keratinocytes in their proteome expression profiles, were invasive, and exhibited no urothelial components according to the pathologist’s report. The rest of the tumors corresponded mainly to TCCs, and a few exhibited extensive squamous differentiation. The latter were not included in this study, given the uncertainty about their origin.

Fig. 1 shows two-dimensional PAGE proteome expression profiles of fresh biopsies from normal urothelium and SCC 884-1 labeled with [³⁵S]methionine. Visual inspection of the autoradiograms showed that many of the proteins were shared by both sample biopsies, although several major polypeptides were differentially expressed (Fig. 1). The identity of the latter was determined by a single procedure or a combination of procedures, including Edman degradation peptide sequencing, mass spectrometry, two-dimensional PAGE immunoblotting, and comparison with the keratinocyte database (14, 15, 16, 17).⁵ Proteins marked in red correspond to normal urothelial polypeptides not expressed by the SCC (keratins 7, 8, 13, 18, 19, and 20 and adipocyte-type FABP), whereas those indicated in blue were expressed only by the tumor (keratins 10, 14, and 16, psoriasin, and MRP 14). Polypeptides indicated with green were highly dysregulated in the SCC (aldose reductase, down-regulated; psoriasis-associated-FABP, up-regulated).

Like tumor 884-1, SCCs 864-1 and 896-1 did not express keratins 7, 8, 18, and 20 and only SCC 864-1 expressed keratins 13 and 19 (results not shown). The levels of keratins 10, 14, psoriasis-associated-FABP, psoriasin, and migration inhibitory factor-related protein 14 varied significantly among the three SCCs, a fact that reflects differences in their degree of differentiation (results not shown; Refs. 4 and 7).

To confirm the results obtained by two-dimensional PAGE, we performed immunofluorescence staining of cryostat sections of normal urothelium and of tumor 884-1 using antibodies specific for the keratins expressed by normal urothelium (keratins 19, 20, 7, and 13) as well for the keratinocyte associated proteins keratin 14, the BG3C8 basal cell antigen (present in the basal cells of stratified squamous epithelia as well as myoepithelial cells; Ref. 19), and psoriasin. As depicted in Fig. 2, the immunohistochemical analysis revealed a very good correlation between the antibody staining and the proteome expression profiles obtained by two-dimensional PAGE. Immunofluorescence pictures of sections of SCCs 864-1 and 896-1 are shown for reference in Fig. 2. Interestingly, SCC 864-1 exhibited cells that were positive and negative for keratin 13 and 19, a fact that could not be anticipated from the analysis of the two-dimensional gels.

Next, we analyzed by means of immunofluorescence serial sections of cystectomies of SCC-bearing patients with invasive disease as the urothelium of these patients was expected to exhibit a spectrum of abnormalities ranging from early stages of metaplasia to invasive disease. Two additional antibodies were included in the battery, one reacting with the differentiation associated marker keratin 10, and the other with a yet unknown antigen (JC-1) expressed by suprabasal cells in the epidermis.⁶ The results presented below are based on the immunohistochemical analysis of several sections of two cystectomies from patients 884-1 and 864-1.

Immunofluorescence staining of several areas of cystectomy 884-1 that were contiguous to the tumor revealed a consistent lack of reaction of the urothelium with antibodies to keratins present in the normal transitional epithelium (keratins 7, 8, 18, and 20; data not shown), suggesting the presence of metaplastic lesions, a fact that was further confirmed by the pathologist’s report. At least three types of metaplastic lesions (types 1–3; Table 1) could be readily distinguished based on the staining with keratin 19 antibodies. Type 1 lesions showed staining of all cell layers in the epithelium (with differences in the staining intensity of the basal compartment; Fig. 3, a and g, left parts of the sections), whereas type 2 lesions exhibited mainly basal cell staining (Fig. 4, a and c). Type 3 lesions, on the other hand, did not stain with keratin 19 antibodies (Fig. 3, a and g, right parts of the sections) and exhibited the same immunostaining phenotype as the SCC (see Fig. 2).

Most areas analyzed displayed well-defined regions in which both basal and suprabasal cells stained with the keratin 19 antibody (Fig. 3, a and g, left sides of the sections; Table 1). These regions showed marked differences in the staining intensity of the basal cells (Fig. 3, a and g), but none of these lesions expressed keratins 7, 8, 18, and 20 (not shown). Areas showing strong keratin 19 staining of the basal cell layer exhibited prominent basal staining with keratin 14 (Fig. 3,h) and the BG3C8 basal cell antigen (Fig. 3,i), and the basal compartment was negative for keratin 13 (Fig. 3,j). Most likely, these lesions (Fig. 3,g) are more advanced in the metaplastic process than those exhibiting a more homogeneous staining with keratin 19 (Fig. 3,a) because the latter stained weakly with the keratin 14 (Fig. 3,b) and the BG3C8 basal cell antigen antibodies (Fig. 3,c). None of these lesions, however, expressed the suprabasal JC-1 marker (Fig. 3, e and k) or keratin 10 (data not shown).

A minority of the areas analyzed by immunofluorescence exhibited mainly basal localization of keratin 19 (Fig. 4, a and c). These cells stained strongly with the keratin 14 and the basal cell antigen antibodies (data not shown) and exhibited weak but distinct suprabasal expression of keratin 10 (Fig. 4,b). In very few areas, the keratin 19-positive basal cells were followed by a string of adjacent keratin 19-negative basal cells (Fig. 4 c, arrows).

These lesions were less frequent than types 1 and 2. Two such metaplastic lesions are depicted in the right side of the serial sections shown in Fig. 3, a and g. Like the tumor (Fig. 2), these lesions expressed keratin 14 (basal and suprabasal staining; Fig. 3, b and h), the BG3C8 basal cell antigen (basal and suprabasal staining; Fig. 3, c and i), the suprabasal antigen JC-1 (suprabasal staining; Fig. 3, e and k), psoriasin (basal and suprabasal staining; Fig. 3. f and l), and keratin 10 (suprabasal staining; data not shown). Weak staining with keratin 13 was observed in the suprabasal layers, whereas the basal cells were negative (Fig. 3, d and j), suggesting that the tumor arose from the expansion of the basal cells.

Cystectomy 864-1 showed CIS of the SCC type (flat lesion, see Fig. 5), and as in the previous cases, the immunofluorescence analysis revealed a consistent lack of reaction of the epithelia with antibodies to keratins present in the normal transitional urothelium (keratins 7, 8, 18, and 20; data not shown). All three types of metaplastic lesions were observed (Table 1), although the type 2 lesions were, by far, the most common in this SCC that showed no urothelial components. Areas containing type 2 and 3 lesions (Fig. 5, a–c) reacted with antibodies to keratins 14 (data not shown), the BG3C8 basal cell antigen (Fig. 5,c), keratin 13 (data not shown), psoriasin (data not shown), and keratin 10 (Fig. 5,b). The suprabasal cells in the type 2 lesions, however, were more differentiated as judged by their stronger expression of keratin 10 (Fig. 5,c), the suprabasal antigen JC-1, and psoriasin (data not shown). It is possible that expansion of both keratin 19-positive and -negative basal cells may have given rise to the tumor because the latter was composed both of keratin 19-positive and -negative cells (Fig. 2).

Areas of the sections exhibiting type 1 lesions were less frequent (Fig. 5, d–f, stronger staining of the basal cells), and like the other two types of metaplastic lesions, cells in these lesions expressed keratin 14 (basal staining; data not shown), the BC3C8 basal cell antigen (basal staining; data not shown), psoriasin (data not shown), and keratin 13 (basal and suprabasal staining; Fig. 5,f). The intensity of the keratin 10 staining, however, was similar to that of the suprabasal cells in type 3 lesions (compare Fig. 5, b and e). Interestingly, the type 1 lesions in cystectomy 864-1 expressed keratin 10.

The presence of areas of extensive squamous metaplasia in bladder SCCs is often associated with poorly differentiated and invasive tumors and is considered to be an unfavorable prognostic factor (6). The question remains, however, as to which of these metaplastic lesions are premalignant because both clinical and histological assessments are clearly insufficient to define those abnormalities that may be associated with the invasive and metastatic potential. Identification of protein biomarkers specific for these lesions will not only lead to a better understanding of the mechanisms underlying squamous metaplasia but will also provide with landmarks to subdivide the various steps involved in the development of SCCs.

Toward this aim, we have presented here a novel strategy for dissecting some of the steps involved in the squamous differentiation of the bladder urothelium leading to SCCs. As a first approximation and in an effort to simplify the interpretation of the results, we analyzed biopsies from SCCs bearing patients showing no urothelial components because these tumors may have originated from the direct neoplastic transformation, on the basis of squamous metaplasia, of the bladder urothelium (6, 7). The approach makes use first of proteomic technologies such as two-dimensional PAGE, protein identification techniques, and databases (9, 10, 14, 15, 16, 17),⁵ to reveal and identify proteins that are differentially expressed in pure SCCs and normal urothelium, followed by immunohistochemistry of serial cryostat sections (immunowalking) of cystectomies obtained from SCC patients that have undergone cystectomy due to invasive disease. Because the urothelium of these patients is expected to exhibit a spectrum of abnormalities ranging from early metaplastic stages to invasive disease (8, 11), it provides with a unique material to define the various steps involved in the squamous differentiation of the transitional epithelium. We choose the keratins, in combination with a few other markers, to characterize the metaplastic lesions because (a) the former are known to be expressed in a tissue-specific manner (20) and (b) specific antibodies, an absolute requirement for this type of analysis, were readily available. In the future, however, phage antibody libraries (21) may prove to be instrumental for the rapid production of large number of specific antibodies against differentially expressed proteins, a fact that will make the approach easily applicable to the study of other metaplastic conditions, such as glandular, columnar, goblet, and signet ring cell as well as nephrogenic metaplasia. At present, the approach is being used systematically to study TCC progression as well as SCCs derived from them as a result of extensive squamous differentiation.

The analysis of cystectomies 884-1 and 864-1 revealed at least three types of nonkeratinizing metaplastic lesions that did not express the urothelial keratins 7, 8, 18, and 20 and that could be readily distinguished based on the staining with keratin 19 antibodies. Type 1 lesions showed staining of all cell layers in the epithelium (with differences in the staining intensity of the basal compartment), whereas type 2 lesions exhibited mainly basal cell staining. Type 3 lesions, on the other hand, did not stain with keratin 19 antibodies. In the case of cystectomy 884-1, only type 3 lesions resembled the tumor in their immunostaining phenotype and may, therefore, be regarded as direct precursors to the tumor. Cells in the basal compartment of these lesions did not express keratin 13, implying that the tumor, which was also keratin 13 negative, may have arisen as a result of the expansion of these cells. Type 3 lesions were often observed adjacent to type 1 and 2 metaplasias, suggesting that they may be derived from the expansion of the basal cell compartment in these lesions. As far as we can tell, type 1 lesions correspond to the earliest type of metaplasia detected in cystectomy 884-1; however, those showing a more distinct basal cell staining with the keratin 19 antibody are most likely more advanced in the metaplastic process because they expressed significant levels of keratin 10. Similar results were observed in the case of cystectomy 864-1, which showed CIS of the SCC type, although in this case, the suprabasal cells expressed keratin 10 indicating that there may be phenotypic differences within this type of metaplastic lesion. This tumor was composed of both keratin 19-negative and -positive cells, suggesting that it may have arisen from the expansion of the basal cell compartment of both type 2 and 3 lesions. Clearly, some additional changes must take place because the tumor cells are both keratin 13 positive and negative (Fig. 2), whereas all three types of metaplastic lesions are keratin 13 positive. Interestingly, basal cells in type 2 and 3 lesions in cystectomy 884-1 were keratin 13 negative, whereas those in 864-1 were positive, further indicating that there may be phenotypic differences within each type of metaplastic lesions. This is not surprising because it may reflect subtle changes in their degree of differentiation.

Clearly, there is still much work to be done to characterize all of the steps involved in the squamous differentiation of the urothelium as well as to assess their impact on prognosis. Our studies have identified nonkeratinizing metaplastic lesions that may represent only some of the steps involved in the development of bladder SCCs, and we do not have any firm data as to whether any of the metaplastic lesions identified thus far are premalignant. Clearly, the analysis of serial cryostat sections of involved cystectomies using an expanded battery of antibodies may prove instrumental for identifying additional changes that may help to pinpoint those lesions that will become premalignant. Only when these goals are achieved will it be possible to search for specific protein biomarkers, either in the urine pellet and/or in the urine, that will identify patients at risk very early during the disease course. Although we are still far from these goals, our studies have opened new possibilities for the systematic search for such biomarkers and have shown that it is feasible to apply powerful proteomic technologies to the analysis of complex biological samples under conditions that are as close as possible to the in vivo situation.

We thank P. Gromov and I. Gromova for their helpful discussions.