Abstract
Antibodies recognizing tissue-specific antigens are widely used to identify the histological origin of tumors. Here we tested the fidelity of selected tissue markers on all 167 solid tumor-derived continuous cell lines in the DSMZ cell lines bank. Most lines had an intermediate filament content consistent with the tumor type from which they were derived. Thus, 93% of all carcinoma cell lines expressed keratin filaments. With certain antibodies, some subclassification was possible. For example, the CK7 keratin 7 antibody can differentiate between colon and pancreas-derived carcinoma cell lines. Cell lines derived from non-carcinomas, in general, did not express keratin but were vimentin-positive. Four of 10 glioma/astrocytoma cell lines expressed GFAP, five of six neuroblastoma cell lines expressed neurofilaments, and the TE-671 rhabdomyosarcoma cell line expressed desmin. When other tissue markers were tested, 12/16 melanoma-derived cell lines expressed HMB-45, while PSA, CA125, and thyroglobulin were less useful. These results demonstrate that cell lines retain some but not all markers typical of the original tumor type and identify certain markers useful in characterizing the histological origin of cell lines. Our data question the identity of some cell lines submitted to the bank in the past. The immunoprofiles of 167 solid tumor-derived and 131 hematopoetic cell lines can be found at www.dsmz.de.
T
Recently, using DNA fingerprinting and chromosome analysis, we have shown that 18% of the cell lines obtained directly from the originators were falsely named or misidentified before arrival at the DSMZ (MacLeod et al. 1999). These methods, especially when they include sensitive methods such as fluorescence in situ hybridization (FISH), are powerful new techniques able to detect many “false” cell lines. Nevertheless, the positive identification of a given line is still the exception. The optimal use of both techniques mentioned above relies on preconditions not given in most instances. Authentication by DNA fingerprinting requires a positive identity control (DNA from the original tumor or well-characterized subclones of the cell line). Chromosome analysis, on the other hand, is most valuable for cell identification in cases in which chromosomal aberrations can be detected that are typical for the originating tumor or when previously reported marker chomosomes are present. These considerations suggest that the proportion of “false” cell lines may be actually higher than the 18% found in the MacLeod et al. (1999) study. Therefore, a major concern is to find additional ways to characterize the properties of the cell lines in the cell bank and, in particular, to provide independent confirmation of the cell type of origin.
All hematopoetic cell lines submitted to the bank are routinely tested for the expression of cell surface markers with a panel of well-characterized monoclonal antibodies (Drexler et al. 2001). In general, our data show that in the vast majority of hematopoetic cell lines the expression pattern of the cell surface markers reflects that of the cell type of origin. Those rare cases expressing inappropriate markers proved to be misidentified or cross-contaminated cell lines. Therefore, for hematopoetic cells, immunocytochemical analysis is a valuable tool not only for cell characterization but in some cases also for cell authentication. This is the first study showing the results of immunophenotypic analyses for the full complement of solid tumor-derived cell lines from a major cell line bank. This study was designed to review the applicability of various monoclonal antibodies used to analyze the origin of primary tumor material for the histological characterization of solid tumor-derived cell lines. Intermediate filament proteins have been shown to be useful markers of cell type in histology and cytology. In general, tumors retain the intermediate filament types characteristic of the cell of origin (e.g., Altmannsberger et al. 1985; Osborn et al. 1986a; Moll et al. 1987; Domagala et al. 1989; Goddard et al. 1991; for review see Osborn and Domagala 1991). Carcinomas and all cell lines derived from carcinomas are expected to be keratin-positive (Debus et al. 1982; Moll et al. 1982,1987; Gown and Vogel 1984; Cooper et al. 1985; Moll 1998). In contrast, most but not all cell lines derived from non-carcinomas are expected to be keratin-negative. In a first set of experiments, all cell lines in the DSMZ collection listed as derived from solid tumors were tested to see whether they were keratin-positive or -negative. Other monoclonal antibodies were then employed to identify markers useful to further characterize the cell type of origin. For the keratin-positive group, antibodies specific for individual keratin polypeptides were used. For the keratin-negative group, antibodies specific for the other intermediate filament types, i.e., vimentin, neurofilaments, desmin, and glial fibrillary acidic protein (GFAP), as well as antibodies to the melanoma surface marker HMB-45 and antibodies to CD 31, were used.
Materials and Methods
Cell Culture
The continuous cell lines were taken from the DSMZ cell lines bank (Drexler et al. 2001). The cells were maintained in the appropriate media (Life Technologies; Karlsruhe, Germany) with 5–20% fetal bovine serum (FBS; Sigma, Deisenhofen, Germany). All cell lines were free of contamination with mycoplasma and were harvested in the logarithmic growth phase with viabilities exceeding 85% as determined by trypan blue dye exclusion. References describing the origin of all cell lines as well as the culture conditions used are summarized in the DSMZ catalogue (www.dsmz.de).
Antibodies
Sources. The following monoclonal antibodies were obtained from DAKO (Hamburg, Germany): the broad specific keratin antibodies MNF-116, 34βE12, and AE1/AE3; the human keratin 17 monoclonal antibody (clone E3); the keratin 18 monoclonal antibody (clone DC10); antibodies against melanoma (HMB-45), CD31 (JC/70A), vimentin (V9), desmin (D33), neurofilament (2F11), thyroglobulin (DAK-Tg6), CA125 (M11), and prostate specific antigen (ER-PR8). Anti-GFAP monoclonal antibodies (N358 and GF5) were purchased from Amersham (Braunschweig, Germany) and DPC Biermann (Bad Nauheim, Germany), respectively. The anti-GFAP (GA5), neurofilament (NN18 and NR4), broad specificity keratin (Lu-5), keratin 7 (CK7), and keratin 8 (a4.1) monoclonal antibodies are also commercially available from a variety of sources.
Specificity. MNF-116 is reported as showing a broad spectrum of reactivity with human epithelial tissues (Moll et al. 1982). In immunoblotting it reveals a number of discrete 45–56.5-kD keratin polypeptides, including keratins 5, 6, 8, 17, and probably 19. 34βE12 was generated against keratin from stratum corneum (Gown and Vogel 1984). It characteristically decorates epidermis and duct-type epithelium. In immunoblotting it is reported as recognizing keratins 1, 5, 10 and 14. AE1/AE3 is a mixture of two monoclonal antibodies raised against human epidermal keratin (Cooper et al. 1985). AE3 is reported as recognizing keratins 1–6, 8 in immunoblots. AE1 recognizes keratins 10, 14, 15, 16, and 19. Lu-5 reacts with a wide variety of epithelial tissues and reacts with a variety of keratins, including 1, 5, 6, 7, 8, 10, 11, and 18, in immunoblotting (Gigi et al. 1982).
Immunocytochemical Analysis
Adherent cells were detached from the flask using trypsin/EDTA. Alternatively, cells were grown on slides before acetone fixation. Cytospin preparations were air-dried for 2–18 hr before acetone fixation (4C, 10 min). The preparations were allowed to dry for 20 min and were stored at — 80C. After thawing, the cells were fixed again with acetone. Staining was performed in a two-step procedure. The antibodies were added to the cells at the appropriate concentrations in a volume of 40 μl (24C, 30 min). The cytospin preparations were washed thoroughly and stained with fluorescein isothiocyanate (FITC)-conjugated anti-mouse Ig antibodies (24C, 30 min). After a further wash, they were mounted either in PBS/glycerol (1:1) or, more recently, in Mowiol 4–88 (Hoechst; Frankfurt, Germany). Antigen expression was analyzed by fluorescence microscopy.
A carcinoma cell line was defined as keratin-negative when less than 2% of the cells were positive in immunofluorescence analysis and as positive when more than 90% expressed the antigen. Only 3/121 carcinoma cell lines exhibited percentages in between these borders (all three had more than 25% keratin-positive cells and were also evaluated as keratin-positive). In some cases, scored as negative for a given antigen, occasional single cells stained brightly positive.
Western Blotting Analysis
A total of 4 × 106 cells were lysed in 100 μl SDS lysis buffer containing 15% glycerol, 125 mM Tris-HCl, pH 6.8, 5 mM EDTA, 2% SDS, 1% β-mercaptoethanol, and 0.1% bromophenol blue. The samples (10 μl/lane) were separated on a 9% gel and blotted onto nitrocellulose membranes by electroblotting (Trans-Blot; Biorad, München, Germany). Equal protein loading of the lanes was confirmed by Ponceau S staining. The membranes were labeled overnight with the appropriate antibodies and specific bands were visualized with the biotin-streptavidin-horseradish peroxidase system (Amersham) in combination with the “Renaissance Western Blot Chemoluminescence Reagent” protocol (NEN; Bad Homburg, Germany).
Results
Carcinoma-derived Cell Lines
A Total of 93% of All Carcinoma-derived Cell Lines Express Keratin. Epithelial cells contain keratins as intermediate filaments. Therefore, keratin antibodies can be used to identify tumors of epithelial origin (Moll et al. 1982). To ascertain whether continuous cell lines derived from carcinomas are keratin-positive, we tested keratin expression in the 121 cell lines derived from carcinomas in the DSMZ collection. We used the keratin monoclonal antibody MNF-116 because it recognizes a broad spectrum of keratin polypeptides and is commercially available (Goddard et al. 1991).
Of the carcinoma-derived cell lines, 93% (113/121) stained positively with this antibody (Figure 1). One hundred percent of the breast carcinoma cell lines (14/14), 100% of the colon carcinoma cell lines (12/12), and 100% of the squamous cell carcinoma cell lines (19/19), as well as the vast majority of cell lines derived from carcinomas at other sites, stained positively with the MNF-116 antibody (Table 1). Those cell lines evaluated here as positive for keratin have an intermediate filament content consistent with their derivation from carcinomas.
A total of 7% (8/121) of the carcinoma-derived cell lines were keratin-negative. They included one bladder cell line, one endometrium carcinoma cell line, one kidney cell line, three lung carcinoma cell lines, one ovary carcinoma cell line, and one thyroid carcinoma cell line (Tables 1 and 2).
Further Analysis of Keratin-negative Cell Lines Originally Reported as Carcinoma-derived. The carcinoma cell lines that did not express keratin in our first analysis were reexamined with MNF-116, by both immunofluorescence and Western blotting assays (Figure 2; Table 2). Again, all eight cell lines were negative in immunofluorescence with MNF-116 as well as with two other broad-specificity keratin antibodies (AE1/AE3 and 34βE12). The seven cell lines that were tested by Western blotting were keratin-negative also in this assay. In addition, the cell lines were tested with antibodies specific for single keratin polypeptides. Antibodies specific for either keratin 7, 8, 17, or 18, the keratins usually expressed in simple epithelia, were available. Six of eight of the cell lines that were negative with the pan-keratin antibodies were negative also with the single-chain specific keratin antibodies (Table 2). Only the bladder carcinoma cell line TCC-SUP and the kidney carcinoma cell line BFTC-909 displayed a weak dotted staining pattern, but not the usual filament pattern when the keratin 8 and 18 antibodies were applied, so we concluded that these cell lines do not express intact keratin filaments. It has been reported that AN3-CA cells do not express keratin 8 (Thie et al. 1995). To our knowledge, no studies on keratin expression in the other seven cell lines listed in Table 2 are published.
We next tried to gain further information on the origin of these cell lines by testing them with a variety of additional markers, including antibodies against other intermediate filament proteins, i.e., with antibodies against desmin, GFAP, neurofilaments, and vimentin. In addition, they were tested with antibodies against CD31 and HMB-45 (Table 2). Two of the keratin-negative cell lines (COLO-668 and COLO-699) stained positively with the melanoma marker HMB-45 and were also vimentin-positive. This strongly suggests that they are melanoma-derived (see below) and that the cells were mistakenly listed as being derived from carcinomas. Six cell lines remained that were vimentin-positive and negative not only for keratins but also for desmin, GFAP, neurofilaments, and CD31.
HeLa Subclones Show Variable Keratin Expression Patterns. Because the DSMZ cell line bank holds several HeLa subclones from different sources, we used these to check whether cells can downregulate keratin expression in vitro. The HeLa cell line has been reported as containing keratins 7, 8, 17, 18, and only trace amounts of keratin 19 (Moll et al. 1983). Studies with various HeLa subclones, identified by DNA fingerprinting, indicated that the intensity of keratin expression may change after in vitro passage. Thus, Western blotting analyses with MNF-116 showed that the HeLa subclone KB clearly expressed less keratin than the parental HeLa cell line (Figure 2). In addition, HeLa and SBC-2, but not the HeLa subclones BT-B, KB, and SBC-7, stained positively with the pan-keratin antibodies AE1/AE3 (not shown). Although this suggests that keratin polypeptides expressed in HeLa, and specifically recognized by AE1/AE3, might be lost completely in the HeLa subclones, all HeLa subclones expressed keratins 7, 8, 17, and 18 when tested with the antibodies specific for the individual keratin polypeptides. However, some subclones clearly displayed a lesser staining intensity with one or other of the chain-specific polypeptide antibodies than did the parental HeLa cell line itself. It is not clear why AE1/AE3 did not stain these cells, because AE3 is reported as recognizing keratin 8 at least in immunoblotting.
Keratin expression of thyroid carcinoma cell line ML-1. Cells were grown on slides and stained with pan-keratin antibody MNF-116 and FITC-labeled secondary antibody. DAPI was used as nuclear counterstain.
Characterization of Keratin-positive Carcinoma Cell Lines Using Additional Antibodies. Antibodies specific either for particular keratin polypeptides (e.g., Moll 1998) or for other markers can in some instances be used to recognize the histological origin of normal and transformed cells.
34βE12. 34βE12 is an antibody reported to recognize squamous cell carcinomas (SCCs) of lung, skin, nasopharynx and cervix, as well as some but not all adenocarcinomas in situ (Gown and Vogel 1984). Our results with the 34βE12 antibody showed that although this antibody recognized three of three SCC-derived lines it also reacted with two duct cell lines and 8/16 of the adenocarcinoma cell lines tested in this study (data not shown). Therefore, 34βE12 is not an appropriate antibody to distinguish cell lines derived from SCCs from those derived from adenocarcinoma, despite its reported specificity in immunoblots for keratins 1, 5, 10, and 14.
Keratin expression of carcinoma and non-carcinoma cell lines a
Keratin expression of each individual cell line was assessed by fluorescence microscopy applying the pan-keratin antibody MNF-116 with an FITC-conjugated secondary antibody. Differentiation marker profiles of all cell lines can be looked up in the DSMZ catalogue (www.dsmz.de).
Antibodies Specific for Keratin 8. Examination of keratins isolated from human tumors has shown that although keratin 8 expression is abundant in adenocarcinomas it is also found in certain types of squamous cell carcinoma, e.g., those from lung or cervix (Moll 1987). Expression of keratin 8 is also associated with hyperproliferation and was originally reported in cell lines as diverse as A-431, HeLa, and HT-29 (Moll et al. 1982). In our study, 20/21 epithelial cell lines were positive with the keratin 8-specific antibody a4.1. These included 17/18 adenocarcinoma-derived cell lines (including one from a benign prostate hyper-plasia and two duct carcinomas) and 3/3 squamous cell carcinomas. Therefore, keratin 8-specific antibodies are not useful for distinguishing between those cell lines derived from adenocarcinomas vs those derived from squamous cell carcinomas.
Antibodies Specific for Keratin 7. Pancreatic ducts in normal tissue as well as certain pancreatic tumors stain positively with the CK7 antibody, which is specific for keratin 7 in immunoblotting. Other gastrointestinal tumors, such as adenocarcinomas from stomach or colon, are negative when tested with the CK7 antibody (Osborn et al. 1986b). We therefore decided to test the 23 gastrointestinal carcinoma cell lines in the cell line bank for CK7 positivity: 8/9 (89%) pancreatic cell lines, 1/2 gastric cell lines, and 0/12 (0%) colon lines were positive with the CK7 antibody. CK7-positive are the pancreatic cell lines CA-PAN-1, CAPAN-2, DAN-G, HUP-T3, HUP-T4, PA-TU-8902, PA-TU-8988S, YAPC, and the gastric cell line 23132/87. CK7-negative are the colon cell lines CACO-2, COLO-206F, COLO-320, COLO-678, CX-1, DLD-1, HCT-15, HT-29, LOVO, SW-403, SW-480, SW-948, the pancreatic cell line PA-TU-8988T, and the gastric cell line MKN-45. Our data confirm results of earlier studies for HT-29, CAPAN-2, and CACO-2 (Moll et al. 1982; Rafiee et al. 1992). Keratin 7 expression in the gastric cell line 23132/87 was noteworthy because it had not been observed in the primary tumor and in early passages of the cell line (Vollmers et al. 1993).
Immunoprofile of keratin-negative carcinoma cell lines a
Expression of tissue markers was assessed by immunofluorescence microscopy (IF) or Western blotting (WB); -, negative; +, positive; (+) dot-like staining, no filaments; n.d., not done. According to the supplier's data sheet, the Abs recognize the following keratins: AE1/AE3: CK 1, 2, 3, 4, 5, 6, 8 10, 14, 15, 16, 19; 34βE12: CK 1, 5, 10, 14; MNF-116: CK 5, 6, 8, 17, and probably 19. Cell line HeLa was used as positive control.
Keratin expression of carcinoma cell lines. Upper panel shows keratin expression using monoclonal antibody MNF-116; lower panel shows vimentin expression. Melanoma cell line COLO-679 was included as negative control for keratin expression.
In the pancreas, keratin 7 expression is restricted to the pancreatic duct (Moll et al. 1983). Although most pancreatic carcinomas arise from the duct (Fernandez et al. 1994), keratin 7-negative pancreatic cell lines have been reported (Rafiee et al. 1992). Indeed our results suggest that keratin 7 may also be downregulated during cell culture. Therefore, whereas the pancreatic carcinoma cell line PA-TU-8988S expressed keratin 7, its genetically identical sister cell line PA-TU-8988T was clearly keratin 7-negative. PA-TU-8988T, but not PA-TU-8988S, has been described as exhibiting characteristics typical of polarized cells (Elsässer et al. 1992). Keratins (keratin 18 and 19) have been reported to be part of structures causing the polar organization of epithelial cells (Salas 1999). Whether downregulation of keratin 7 in PA-TU-8988T cells is paralled by upregulation of other keratins and whether this is related to the polar structure of the cells are interesting questions that deserve further study. Both cell lines were also positive with the pan-keratin antibody MNF-116.
HMB-45. This marker was negative on all 20 keratin positive carcinoma-derived cell lines on which it was tested (data not shown).
Other Antibodies. Twenty-one selected epithelial cell lines were tested with antibodies against prostate-specific antigen (PSA), CA125 (ovarian carcinoma), and thyroglobulin (thyroid carcinoma). The results of these tests suggested that these markers are not useful to further subdivide carcinoma cell lines (data not shown).
Vimentin. A total of 55% of the cell lines listed as carcinoma-derived co-expressed vimentin.
Non-carcinoma-derived Cell Lines
The non-carcinoma cell lines in the DSMZ collection are listed in Table 1. Again, we started by testing all cell lines in this group with the broad-specificity keratin antibody MNF-116. Among the non-carcinoma-derived cell lines in the collection, 80% (37/46) were keratin-negative when tested with the MNF-116 antibody (Table 1). These included all melanoma-derived cell lines (16/16). Both retinoblastoma cell lines were keratin-negative, as were 6/7 of the neuroblastoma cell lines and 8/10 of the glioma/astrocytoma cell lines. Altogether, nine cell lines in this group were keratin-positive—the seven cell lines listed in Table 3, CADO-ES1, and CAL-72. For the latter two cell lines no filamentous staining was seen. Whereas the seven lines listed in Table 3 were positive also with the broad-specificity Lu-5 antibody, CADO-ES1 and CAL-72 were not. In addition, CADO-ES1 has been described as keratin-negative (Kodama et al. 1991). Thus, 7/46 (15%) of the non-carcinoma cell lines were consistently keratin-positive and showed a filament pattern. To try to determine in which of these cell lines keratin expression was a “normal” feature and did not question the identity of the cell line, we have listed the reported origin of each cell line in Table 3. Human epitheloid sarcomas have been reported as keratin- and vimentin-positive, and therefore the finding of both types of intermediate filaments in the VA-ES-BJ cell line is not unexpected and indeed has been reported previously (Helson et al. 1995). SK-NM-C was originally reported as neuroblastoma-derived but more recently has been thought to be derived from Ewing's sarcoma. On tissue sections, keratin positivity is found in isolated and in clustered Ewing's tumor cells in the majority of cases, and desmoplakin staining has also been reported. Therefore, the finding that Ewing's sarcoma cell lines occasionally express keratin and vimentin is consistent with the results on tumor sections and would explain the keratin positivity of the SK-N-MC cell line.
We face a similar situation with the fibrosarcoma cell line HT-1080 and the thyroid sarcoma cell line S-117. Keratin-positive fibrosarcoma cells have been described (Banks et al. 1995). In addition, co-expression of epithelial (keratin) and mesenchymal (vimentin) markers in certain thyroid tumors has been reported (LiVolsi et al. 1987). Therefore, our results do not neccessarily question the identity of these two cell lines. The glioma cell lines 8-MG-BA and DBTRG-05MG, as well as the rhabdomyosarcoma cell line A-204, differ from the other cell lines mentioned here because only <5% cells react with the keratin antibodies. Our data on the glioma cell lines confirm earlier studies reporting that 5–7% (8-MG-BA) and less than 0.1% (DBTRG-05MG) cells express keratin (Macikova et al. 1999). We did not find GFAP expression in these cell lines, in agreement with Macikova et al. (1999), who reported that GFAP expression was negative at early time points after passaging the cells and weakly positive at later time points. In the case of the rhabdomyosarcoma cell line A-204, again only a low percentage of cells expressed keratin (Table 3) but, more importantly, the cells did not express desmin. Earlier reports on intermediate filament expression by this cell line were conflicting. It has been reported that A-204 cells express desmin filaments and that desmin expression is upregulated on stimulation with 1.5% DMSO (Melguizo et al. 1994). Later it was shown that this A-204 clone, like the first one obtained from the ATCC, did not express desmin filaments (Morton and Potter 1998). We found neither constitutive desmin expression nor an increase of desmin expression after stimulation with DMSO in the DSMZ clone or in a second clone, obtained from an independent source and identified by DNA fingerprinting. These results, taken together with data in the literature showing that rhabdomyosarcomas are desmin-positive (Altmannsberger et al. 1985), suggest that either A-204 is not rhabdomyosarcoma-derived or no longer retains important properties typical for this tumor type.
Keratin and HMB-45 antigen expression by non-carcinoma cell lines a
+, positive; (+), less than 5% positive cells; -, negative.
Most Melanoma Cell Lines, but None of the Carcinoma Cell Lines, Express the Antigen Detected by Antibody HMB-45. A total of 12/16 (75%) melanoma cell lines were positive when tested with the HMB-45 antibody (Table 4). In contrast, 0/20 carcinoma cell lines (data not shown) and 0/7 cell lines reported as being non-carcinoma/non-melanoma-derived (Table 3, CADO-ES1 and CAL-72) were HMB-45-positive. To our knowledge, only one of the melanoma cell lines tested here has been analyzed previously with HMB-45. IGR-39 was found to be HMB-45 negative (Kroumpouzos et al. 1994). In the same study, only 2/12 melanoma cell lines were shown not to express the antigen recognized by HMB-45. Western blotting analysis of lysates of the melanoma cell line IGR-37 showed that HMB-45 recognized a 30–35-kD protein (data not shown), supporting previous findings with the melanoma cell line HU-214 (Chiamenti et al. 1996). All melanoma-derived cell lines expressed vimentin, and in two cell lines additional expression of neurofilaments was noted.
Immunoprofile of melanoma cell lines a
Expression of tissue markers was assessed by immunofluorescence microscopy. None of the cell lines expressed cytokeratin (MNF-116 AE1/AE3), desmin, GFAP, or CD31.
Intermediate Filament Typing of Rhabdomyosarcoma and Neuroblastoma Cell Lines. The results of the analysis of rhabdomyosarcoma and neuroblastoma cell lines with antibodies against different intermediate filament proteins are shown in Table 5. Both human and rat rhabdomyosarcoma in situ are distinguished by the presence of the intermediate filament protein desmin. Vimentin positivity has been reported in 5–50% of the tumor cells in approximately half the cases studied (Altmannsberger et al. 1985). Rhabdomyosarcoma cell lines are therefore expected to be desmin-positive. Of the two cell lines tested, TE-671 is desmin- and vimentin-positive, which is consistent with its proposed derivation from a human rhabdomyosarcoma.
The human neuroblastoma cell lines are also of interest. Of the six cell lines shown in Table 5, five are positive with antibodies against neurofilaments and vimentin, consistent with a derivation from neuroblastoma (Osborn et al. 1986a). The sixth cell line is positive only for vimentin.
Other Markers. A total of 4/10 glioma/astrocytoma cell lines expressed GFAP. A total of 98% of the cell lines listed as non-carcinoma-derived were positive with vimentin antibodies. The only vimentin-negative cell line in this group was the retinoblastoma cell line WERI-RB-1.
Discussion
The first major conclusion from this study is that the vast majority of carcinoma-derived cell lines express keratin filaments. In this study, 93% (113/121) of the cell lines listed in the DSMZ catalogue as carcinoma-derived were positive in immunofluoresence with broad-specificity pan-keratin antibodies (Table 1). This result is consistent with data in the literature on carcinoma cell lines (e.g., Moll et al. 1982) and with the demonstrated keratin-positivity of carcinomas and metastases when tested in histology and cytology (Moll et al. 1982, 1987; Gown and Vogel 1984; Cooper et al. 1985; Osborn and Domagala 1991).
Intermediate filament expression in human cell lines derived from rhabdomyosarcoma and from neuroblastoma a
+, positive; (+), less than 5% positive cells; -, negative.
The finding that carcinomas and cell lines derived from them are in general keratin-positive calls into question the origin of keratin-negative cell lines that are listed as carcinoma-derived. Such cell lines should be screened for the expression of markers associated with tissues other than epithelia in an attempt to understand the true origin of the cell line in question. In our study 7% (8/121) of such cell lines did not display true keratin filaments when tested with a panel of broad-specificity and single chain-specific keratin antibodies. By testing the eight cell lines for other markers, we were able to show that two were positive with the melanoma antibody HMB-45 and were vimentin-positive. Therefore, these two cell lines are most likely melanoma-derived. DNA fingerprinting enabled us to identify one of these cell lines as the melanoma cell line COLO-679, which is also HMB-45-positive and keratin-negative and which was provided to the DSMZ by the same depositor as the other two cell lines. A further four cell lines were vimentin-positive and negative for other intermediate filament proteins as well as for HMB-45 and for CD31. These cell lines might be derived from the stromal cells rather than tumor cells in the carcinoma, i.e., a selection for the stromal cells took place during early passages in culture. A complete loss of keratin filaments during in vitro passage has been reported in the literature only under very special circumstances (Boyer et al. 1989). However, it is clear from the results shown here for HeLa subclones and from the literature that individual keratin polypeptides can be down- or upregulated during in vitro passage. Supporting this notion, we found that 2/8 carcinoma cell lines without keratin filaments showed a dot-like staining pattern with keratin 8/18 antibodies, possibly the result of an unsuccessful filament assembly. However, regardless of which explanation is correct, those cell lines that have lost keratin filaments are unlikely to be good in vitro models of carcinomas or their metastases.
To assess the feasibility of a further subclassification of carcinoma cell lines, we tested a panel of antibodies also used for classification of carcinomas. This effort was only partially successful. The expression of PSA (prostate carcinoma), CA125 (ovarian carcinoma), thyroglobulin (thyroid carcinoma), and keratin polypeptides that are typical for squamous cell carcinoma did not correlate sufficiently with the respective cell types. Interestingly, however, 8/9 pancreatic carcinoma cell lines but 0/12 colon cell lines reacted positively with the CK7 antibody. Therefore, as in primary tumors (Osborn et al. 1986b), CK7 analysis may help to distinguish between cell lines derived from pancreas and colon despite the differences in reactivity of different keratin 7 antibodies shown in a recent study (Schon et al. 1999).
The results of testing non-carcinoma-derived cell lines in the DSMZ collection showed that 85% (39/46) were negative when tested with broad-specificity keratin antibodies. Keratin positivity of most of the seven keratin-positive cell lines in Table 3 can be explained because they originate from tumors such as Ewing's tumor or epitheloid sarcoma, which are known to contain keratin-positive tumor cells. The analysis of the immunoprofile of the melanoma cell lines shows that all cell lines are vimentin-positive and that 75% of the melanoma cell lines stained positively for HMB-45. Neither carcinoma cell lines nor other solid tumor-derived cell lines were HMB-45-positive, confirming that this marker effectively detects melanoma-derived cells. However, our data showed that a substantial fraction (25%) of cell lines reported as being derived from melanoma did not express the antigen recognized by HMB-45 (cf. Shah et al. 1997). In addition, HMB-45-positive neuroblastoma cells have been reported (Sorrentino et al. 1999). Therefore, although neither positive nor negative expression of the antigen recognized by HMB-45 identifies unequivocally the histological background of a cell, the finding of a vimentin-positive, HMB-45-positive immunoprofile is consistent with derivation from a melanoma.
The combined data show that tissue marker analysis is a very useful technique for verification of the histological origin not only of histological and cytological tumor specimens but also of cell lines. However, the panel of antibodies applicable for characterization of cell lines is relatively small. Useful for the histological characterization of cell lines are pan-keratin antibodies (carcinoma vs non-carcinoma), the CK7 antibody raised against keratin 7 (pancreatic carcinoma vs colon carcinoma), desmin antibodies (rhabdomyosarcoma), neurofilament antibodies (neuroblastoma), and HMB-45 (melanoma vs non-melanoma).
It is very important to draw attention to the fact that not all tissue markers show the same expression pattern in cell lines as in the original tumors. For example, vimentin, which is primarily expressed in mesenchymal tumors (see Osborn and Domagala 1991), is also expressed in a large variety of cell lines of non-mesenchymal origin. Thus, in this study vimentin expression was found not only in almost all non-carcinoma cell lines (98%) but also in a substantial fraction of the carcinoma cell lines (55%). In addition, certain markers typically associated with differentiated cells in normal tissues or in tumors derived from these tissues are not reliably expressed in cell lines, e.g., PSA, CA125, or thyroglobulin.
Better methods for cell line identification and characterization are needed, as shown in this study and in a previous study in which a minimum of 18% wrong cell lines designated for cell bank entry were identified by other techniques (MacLeod et al. 1999). Optimal tissue verification of cell lines requires more than the few markers tested in routine screening programs. It is hoped that the further development of the microarray technique will allow a more accurate differentiation, not only among various tumors but also among the respective cell lines. In this context, it should be noted that cell identification would be an easy task if DNA fingerprinting of cell line and patient DNA were routinely included in the procedure of establishing a new cell line (Stacey et al. 2000).