Tissue preparation for gene expression profiling of colorectal carinoma: three alternatives to laser microdissection with preamplification

Tissue preparation for gene expression profiling of colorectal carinoma: Three alternatives to laser microdissection with preamplification ROLAND S. CRONER, KLAUS GUENTHER, THOMAS FOERTSCH, RENATE SIEBENHAAR, WOLFGANG M. BRUECKL, CHRISTIAN STREMMEL, FALK HLUBEK, WERNER HOHENBERGER, and BERTRAM REINGRUBER ERLANGEN, REGENSBURG, and HENNIGSDORF, GERMANY

Colorectal-carcinoma specimens are heterogeneous and include areas of nonmalignant mucosal and connective tissue. For those study designs in which laser microdissection and RNA preamplification are impracticable, the optimal yield of genuine cancer RNA is a key factor in gene-expression analysis. In this study we compared alternative methods of tissue purification. Three contiguous 0.5-cm3 samples taken from an advanced primary adenocarcinoma of the sigmoid colon were processed immediately after surgery with the use of the following methods: (1) cryotomy after manual dissection (CMD), (2) microscopically assisted manual dissection (MAMD), and (3) tumor-cell isolation with the use of Ber-EP4 antibodies and Dynabeads (Dynal Biotech GmbH, Hamburg, Germany; technique abbreviated as DB). We generated gene-expression profiles with the use of GeneChip technology (Affymetrix, Santa Clara, Calif) and recorded preparation times, costs, and RNA quantity and quality. CMD took 60 minutes, MAMD 180 minutes, and DB 90 minutes to isolate 22, 8, and 23 ␮g of RNA, respectively. Expenses for materials amounted to $41, $23, and $91 (US) for CMD, MAMD, and DB, respectively. The 3'/5' ratio, as determined with the GeneChips, for GAPDH/␤-actin was 1.01:1.03 for CMD, 1.13:1.28 for MAMD, 1.43:1.68 for DB, K-ras, APC, smad 2, transforming growth factor-␤, and p53 were marked as present in all cases, with the exception of APC, which was graded as marginal on DB. The correlation values of gene-expression profiles were 91% (CMD/DB), 93% (CMD/MAMD), and 97% (DB/MAMD). All 3 methods provided enough RNA, of sufficient quality, for gene-expression microarray analysis in colorectal carcinoma. Cross-methodologic analyses of array data should not be performed uncritically. (J Lab Clin Med 2004;143:344-51) Abbreviations: CMD ⫽ cryotomy after macroscopic manual dissection; DB ⫽ tumor-cell isolation with the use of Ber-EP4 antibodies and Dynabeads; GAPDH ⫽ glyceraldehyde-3phosphatase; H&E ⫽ hematoxylin and eosin; LCM ⫽ laser-capture microdissection; MAMD ⫽ microscopically assisted manual dissection; PBS ⫽ phosphate-buffered saline solution; TGF ⫽ transforming growth factor

From the Departments of Surgery, Internal Medicine I, and Department of Pathology, University of Erlangen; and the Department of Surgery, University of Regensburg. Sponsored by the German Federal Ministry for Education and Science within the National Genome Science Network and Europroteome AG, Hennigsdorf, Germany. Drs Croner and Guenther have contributed equally to this work. Submitted for publication June 26, 2003; revision submitted February 17, 2004; accepted March 5, 2004.

344

Reprint requests: Roland S. Croner, MD, Chirurgische Universita¨tsklinik Erlangen, Krankenhausstrasse 12, 91054 Erlangen, Germany; e-mail: [email protected]. © 2004 Elsevier Inc. All rights reserved. 0022-2143/$ – see front matter doi:10.1016/j.lab.2004.03.003

J Lab Clin Med Volume 143, Number 6

Gene-expression profiling of cancer is a new method used for the characterization of specific tumor types.1–5 This technique enables the investigator to analyze simultaneously thousands of genes from a single tissue sample in 1 experiment and to characterize the biologic behavior of the tumor.6,7 In the future it may be used as a tool with which to identify malignancies8 and predict their courses and susceptibilities to various treatment strategies. However, tissue samples must be harvested under standardized conditions, with prolonged warm ischemia specifically avoided, for RNA of sufficient quality to be yielded and to minimize bias.9 The harvested samples usually consist of a mixture of tissue types, including connective tissue, noncancerous parenchyma, and cancer tissue. To reduce the influence of nonmalignant portions of specimens in gene-expression analysis, methods of purification and enrichment of cancer tissue have been developed. Laser-capture microdissection (LCM) has proved useful and has become an method of isolating specific cell populations from a section of complex, heterogeneous tissue.10 –12 However, this method of tissue purification is time-consuming, requires fixation of sample slices on microscope slides, and permits the collection of only small amounts of RNA at a time. Laser-microdissected cancer tissue generally requires preamplification to yield sufficient quantities of RNA for conventional commercial microarray analysis.10,12 In light of the shortcomings of molecular-biologic enrichment, we evaluated 3 alternative methods of cancer-tissue purification on the basis of their usefulness in gene-expression profiling of colorectal carcinoma. METHODS The ethics committee of our university approved the study, consent was obtained from the patient from whom samples were taken, and our research was conducted in accordance with the principles of the Declaration of Helsinki. Tissue was taken from a single patient with polypoid adenocarcinoma of the sigmoid colon (G2, pT3, pN1(2/23) L1, V0, M0, R0, stage III) who underwent surgery without neoadjuvant treatment. Immediately after surgery, 1 macroscopically homogeneous tumor segment was excised from the surgical-resection specimen by the pathologist and divided into 3 equal portions of 0.5 cm3 each. Two samples were shock-frozen in liquid nitrogen and subjected either to CMD or MAMD. The third sample was subjected to DB. CMD. The tissue was inserted into a cryotube (Roth, Karlsruhe, Germany) together with Tissue-Tek (Zakura, Zoeterwoude, the Netherlands) and immediately shock-frozen in liquid nitrogen. With the use of a cryotome, a first control slice (7 ␮m) was dissected and stained with H&E. Connective tissue and healthy mucosa were identified by a pathologist and then removed from the Tissue-Tek– embedded specimen. In a second control slice, the purity of the carcinoma tissue

Croner et al

345

Fig 1. Tissue before (A) and after (B) CMD.

was checked again; this procedure ws repeated until the carcinoma portion of the Tissue-Tek– embedded specimen was judged to be greater than 80%. Next, a continuous series of 10 slices (40 ␮m each) was dissected and collected with the use of the first buffer of an RNA isolation kit (buffer RLT, RNeasy-Kit; Qiagen, Hilden, Germany). Each series of 10 slices was collected separately in marked microtubes for further workup and followed by another control H&E-stained slice to reveal tissue composition. The corresponding series was discarded if the subsequent H&E-stained control showed a proportion of carcinoma tissue of less than 80%. A total of 30 slices, 40 ␮m each, were collected in this fashion (Fig 1). MAMD. Tissue was inserted into a cryotube with TissueTek and immediately shock-frozen in liquid nitrogen. With the use of a cryotome, 60 15-␮m slices were dissected and mounted on microscopy slides. To inactivate the RNAses, we dehydrated the slices by immediately immersing them successively, for 30 seconds each, in 70%, 95%, and 100% ethanol and finally in xylol. Under microscopic control (magnification 10⫻–20⫻), we abraded the cancer tissue manually and collected it in the first buffer of the RNA isolation kit (buffer RLT, RNeasy-Kit) (Fig 2).

346

Croner et al

J Lab Clin Med June 2004

Fig 2. Tissue before (A) and after (B) MAMD. Fig 3. DB. A, Dynabead-captured carcinoma cells. B, Tissue cylinder consisting of a mixture of tissue types. DB. The tissue was incubated in a solution consisting of RNAlater (Qiagen) and PBS (Bio Whittaker, Verviers, Belgium) at a 1:1 ratio, benzamidine 100 mmol/L (Sigma-Aldrich, Taufkirchen, Germany), and EDTA 0.5 mol/L (pH 7.4; RNAlater ⫹ PBS). The tissue was immediately fragmented through a mesh (200-␮m pore size).13 The homogenized tissue was collected in the RNAlater ⫹ PBS solution. We next centrifuged the tissue at 2630g for 10 minutes at 4°C. The supernatant was discarded. The tissue was rediluted in 5 mL of RNAlater ⫹ PBS. Ber-Ep4 antibodies (Dynabeads; Dynal Biotech GmbH, Hamburg, Germany) were prepared in accordance with the manufacturer’s protocol. Dynabeads are magnetic microspheres with unique high-stability, high-uniformity paramagnetic properties; low particle-particle interaction; and high dispersibility. These immunomagnetic microspheres can be used to minimize cellular heterogenity and to isolate cells with particular surface-membrane phenotypes in differentiation studies. In this study, the microspheres were coated with antibodies against epithelial cell antigen Ber-Ep4. The prepared antibody solution (80 ␮L) was added to 5 mL of sample solution. Antibodies were incubated for 30 minutes with gentle rotation in the Dynal Sample Mixer (Dynal Bio-

tech GmbH). With a the use of magnetic-particle concentrator (Dynal Biotech GmbH), the antibody-marked tumor cells were isolated from the supernatant during a 3-minute period. The supernatant was discarded and the isolated tumor cells were rediluted in RNAlater ⫹ PBS. This washing procedure was repeated 3 times. Finally the solution was distributed to several test tubes and a pellet was formed with the use of centrifugation at the maximal speed of 14,000 rpm (Sigma 2K15; Sigma Laboratory, Osterode/Harz, Germany). Again the supernatant was discarded, and the pellets were shockfrozen and stored in liquid nitrogen until further processing could be performed. All solutions and tubes were cooled (4°C) during the procedure (Fig 3). RNA Isolation. We conducted RNA isolation in the same way for all 3 tissue samples, with the use of commercial kits (RNeasy-Kit) applied in accordance with the manufacturer’s protocol. As part of this procedure, DNAse digestion (Qiagen) was included, in keeping with the manufacturer’s suggestion. RNA quantity and quality. RNA quality and quantity were determined with the use of the “lab on a chip” method

J Lab Clin Med Volume 143, Number 6

Croner et al

347

Table I. Time and material costs for sample preparation Expenditures

CMD

MAMD

DB

LCM4

Preparation time (min) Personnel expenditure ($50 per 60 min) Material costs/sample ($ US) Basic material and reagents/sample ($ US) H&E staining/sample ($ US) RNA isolation/sample ($ US) RNA preamplification/sample ($ US) Total costs/sample ($ US)4

60 50 41.08 1.33 31.43 8.32 — 91.08

180 150 23.44 10.63 4.49 8.32 — 173.44

90 75 90.76 82.44 — 8.32 — 165.76

960 800 340.48 2.36 89.80 8.32 240 1140.48

(Bioanalyzer 2100; Agilent Technologies, Palo Alto, Calif), in keeping with the manufacturer’s instructions.14 The 3'/5' ratios for the housekeeping genes glycerinaldehyde-3-phosphatase (GAPDH) and ␤-actin obtained with the use of the GeneChip were used as further measures of RNA quality and as a means of ruling out partial degradation. We regarded any 3'/5'-ratio of less than 3 as an indicator of good RNA quality, in accordance with the protocol set forth by the manufacturer (Affymetrix).15 RNA microarrays. We examined gene expression using GeneChip technology. Biotin-labeled complementary RNA was generated through the use of in vitro transcription and hybridized to the GeneChips (HG-U95A; Affymetrix, Santa Clara, Calif) in accordance with the manufacturer’s instructions.16 The resulting data were analyzed with the use of Microarray-Suite software (Affymetrix). We compared the correlation (Pearson) of the detected signals on the Gene Chips between the gene-expression profiles and the signals, detection calls, and detection P values of ␤-actin, GAPDH, and 5 other genes that were described in the Vogelstein model of colorectal carcinoma (k-ras, APC, smad 2, TGF-␤, and p53).17 A P value of .04 to 0.06 was graded as marginal (M). In keeping with the manufacturer’s instructions, we graded genes in samples with values of less than .04 and greater than .06 as present (P) and absent (A), respectively. The time required for tissue preparation and the expenses generated by each method were calculated, described, and compared against the efforts necessary for LCM. The costs of LCM were evaluated in cooperation with research groups using this technique for purposes of tissue isolation. The time and material costs required to generate 100 ng of RNA, the least amount necessary for preamplification, from isolated carcinoma tissue was calculated. We translated prices from euros to US dollars and estimated personnel expenditures at a cost of $50 US per 60 minutes (Table I). RESULTS CMD. From the 30 slices collected, 22 ␮ of total RNA of good quality was isolated (Fig 4, A). Gene Chip hybridization was used without any difficulties. The 3'/5'-ratio was 1.01 for GAPDH and 1.03 for ␤-actin on the GeneChip. Dissection, collection, and control of composition with the use of intermittent H&E staining of the cancer slices took about 60 minutes. Material costs were $41 US, with H&E staining being the most

expensive part ($31 US/sample) of procedure (Table I). The housekeeping genes GAPDH and ␤-actin, as well as the 5 selected genes,17 were graded as present on the Gene Chip, with detection P values of less than .01 (Table II). MAMD. From the 60 slices fixed on microscope slides, 8 ␮g of total RNA of good quality was isolated (Fig 4, B). GeneChip hybridization was performed without any problems. The 3'/5'-ratio was 1.13 for GAPDH and 1.28 for ␤-actin on the GeneChip. The total tissue-preparation time was 180 minutes (manual abrasion of the carcinoma tissue from the microscopy slides alone took 120 minutes). The material cost was $23 US per sample (Table I). The bulk of the costs involved material and reagents, including microscopy slides and alcohol-fixation fluids ($11 US/sample). GAPDH, ␤-actin, and the 5 selected genes17 were graded as present (P ⬍ .01; Table II). DB. The tissue pellet that resulted after the DB procedure contained 23 ␮g of total RNA of good quality (Fig 4, C). GeneChip hybridization was performed without any problems. The 3'/5' ratio was 1.43 for GAPDH and 1.68 for ␤-actin on the GeneChip. Tissuepreparation time was 90 minutes. The cost was $91 US per sample (Table I). In this case, most of the money was spent for Dynabead separation (antibodies and reagents $82 US/sample). The housekeeping genes GAPDH and ␤-actin and only 4 of the selected genes17 were graded as present (P ⬍ .01; Table II). The APC gene was graded as marginal (P ⫽ .044; Table II). Gene-expression profiles. The HG-U95A GeneChip contains 12.625 probe sets. The expression of a single gene can be detected on the basis of the signal values of one probe set. Using the CMD method, we found that 7.987 genes were present, 4.360 were absent, and 278 genes were marginal. Using MAMD, we found that 7.082 genes were present, 5.259 were absent, and 284 were marginal. Using the DB method, we found that 6.740 genes were present, 5.612 were absent, and 273 genes were marginal. The correlations were calculated with the use of Pearson’s correlation coefficient, including the signals of all 12.625 investigated probe sets

348

J Lab Clin Med June 2004

Croner et al

Fig 4. RNA control after CMD (A), MAMD (B), and DB (C) with the Bioanalyzer 2100.

Table II. Performance of CMD, MAMD, and DB Signal Gene

␤-Actin GAPDH k-ras APC smad 2 TGF-␤ p53

Detection call

Detection P value

CMD

MAMD

DB

CMD

MAMD

DB

CMD

MAMD

DB

48.392 40.876 804 733 927 569 438

40.210 39.865 1.138 .702 693 348 555

49.799 40.547 862 591 381 443 370

P P P P P P P

P P P P P P P

P P P M P P P

.000044 .000044 .000219 .002228 .000219 .001602 .000468

.000044 .000044 .000219 .009985 .000266 .006532 .000959

.000044 .000044 .000388 .043968 .000266 .000562 .000805

on the GeneChips. We calculated gene-expression values with respect to the background signal of each GeneChip. The correlation between CMD and DB was 91%, that between CMD and MAMD was 93%, and that between DB and MAMD was 97%.

DISCUSSION

In this study we demonstrated that cancer-tissue purification with the use of CMD, MAMD, and DB provided enough RNA of sufficient quality for gene-ex-

J Lab Clin Med Volume 143, Number 6

Croner et al

349

Fig 5. Comparison of costs for different samples sizes.

pression profiling of colorectal carcinoma. The 3⬘/5⬘ratios of the housekeeping genes GAPDH and ␤-actin, as determined on the basis of control areas on the GeneChips, were markedly lower than 3.0 in every procedure, implying that no relevant degradation of RNA had occurred.15 According to the manufacturer’s protocol, 5 ␮g of total RNA is needed for 1 GeneChip. None of our methods required preamplification of the isolated RNA to quantitatively hybridize a GeneChip. This is important because the proven nonlinear fashion of RNA amplification results in a distortion of the quantities of the hybridized genes and, subsequently, loss of comparability between analyses with and without preamplification.18 Preamplification is obligatory when LCM is used for gene-expression profiling. Even though LCM remains the gold standard of tissue preparation in terms of tissue purification, and even though sufficient RNA could theoretically be isolated for a GeneChip hybridization in this fashion, the necessity of preamplifying the minimal yield of RNA in the normal routine situation limits comparability of data sets with those derived through the use of an identical method. The financial costs supersede those of all other methods by a factor of about 7 to 13 (Fig 5). In the selection of the optimal gene-expression profiling technique for routine clinical use, time is a factor that must be considered because of the risk of RNA degradation and the economic costs. The time required to purify carcinoma tissue ranged from 60 and 180 minutes, far less than that needed to collect equal amounts of RNA by LCM.

Requiring just 60 minutes, CMD was the fastest procedure and therefore, despite its somewhat higher expenses with regard to reagents, the most cost-effective. Furthermore, CMD required the least amount of tissue preparation likely to produce the gene-expression profile closest to natural biology. However, tissue composition can change within a series of sections, and subsequent slices may contain different proportions than those found in the first or last H&E stained slices. For this reason, control staining should be performed at regular intervals (in our study, every 10 sections) and the slices dissected between each control slice should be collected separately in marked microtubes for further workup. If the subsequent H&E-stained controls show inadequate proportions of carcinoma tissue (in this study, ⬍ 80%), the corresponding series should be excluded. The MAMD method facilitates tissue purification under microscopic control. The degree of purity is somewhat dependent on the technician’s experience. Furthermore, if carried out as meticulously, this method is the most time-consuming but, with regard to the reagents used, the cheapest procedure of all. Losses of carcinoma tissue are considerably fewer than in CMD because even small portions of tumor can be collected from the slide and used for RNA isolation. The DB method was originally developed for proteomics studies in colorectal carcinoma,13 but RNA analyses have been performed on those samples as well. Although the enrichment of cancer cells cannot reach 100% because the Ber-Ep4 antibody is specific

350

J Lab Clin Med June 2004

Croner et al

not for carcinoma but for epithelial cells in general,19 the DB method has been successfully used for differential display analyses in which colorectal carcinoma is compared with normal mucosa (unpublished results). Indeed, among the tumor cells, we were also able to microscopically detect nonmalignant epithelial cells and even several fragments of other material (Fig 3, B). One possible disadvantage is the fact that DB can only be used with fresh tissue processed immediately after surgery. Furthermore, DB requires the most expensive reagents; however, with regard to preparation time, it was as cost-effective as MAMD, which was the most time-consuming method. Comparing costs for tumorcell isolation of more than one sample only during DB personnel costs could be reduced because 2 or 3 samples can be worked up with this method at the same time. For all other methods, costs increase linearly with the amount of the prepared samples (Fig 5). Although we detected no RNA degradation in any of the methods, the comparability of the gene-expression profiles was limited. The correlation between CMD and DB was 91%, that of CMD and MAMD was 93%, and that of DB and MAMD was 97%. These findings led us to the assumption that comparability of gene-expression profiles from materials that had been processed with the use of different methods must be regarded critically. Each method yields a varying purity of cancer tissue as a result of contamination with nonepithelial tissue and healthy mucosa and depending on the investigator’s experience and skill. Moreover, each method carries its specific bias with regard to geneexpression analysis. Signals are calculated in relation to the specific background of each GeneChip. The geneexpression profile of a single carcinoma may be heterogeneous. Hence signal values for 1 investigated gene within a single tumor-tissue sample (eg, smad 2; Table II) may vary. Nevertheless signals, detection calls and detection P values for GAPDH, ␤-actin, and the 5 important genes within the Vogelstein model of colorectal carcinogenesis17 were comparable on the GeneChips. Only under DB purification was the APC gene graded as marginal. The best correlation was observed between the MAMD and DB methods (97 %). It may therefore be assumed that both procedures, which involve visual and immunologic control, respectively, yield equivalent sample purities. We conclude that despite differences in details, all 3 methods are suitable for tissue processing for the purpose of gene-expression profiling in colorectal carcinoma, if they are applied under standardized conditions with regard to ischemia, cold storage, and material preparation. However, absolute comparability of their

results, as well as cross-methodologic comparability of expression data, is limited. REFERENCES

1. Hedenfalk I, Duggan D, Chen Y, Radmacher M, Bittner R, Simon P, et al. Gene-expression profiles in hereditary breast cancer. N Engl J Med 2001;344:539 – 48. 2. Lau WY, Lai PB, Leung MF, Leung BC, Wong N, Chen G, et al. Differential gene expression of hepatocellular carcinoma using cDNA microarray analysis. Oncol Res 2000;12:59 – 69. 3. Leethanakul C, Patel V, Gillespie J, Shillitoe E, Kellman RM, Ensley JF, et al. Gene expression profiles in squamous cell carconomas of the oral cavity: use of laser capture microdissection for the construction and analysis of stage-specific cDNA libraries. Oral Oncol 2000;36:474 – 83. 4. Notterman DA, Alon U, Sierk AJ, Levine AJ. Transcriptional gene expression profiles of colorectal adenoma, adenocarcinoma, and normal tissue examined by oligonucleotide arrays. Cancer Res 2001;61:3124 –30. 5. Hippo Y, Taniguchi H, Tsutsumi S, Machida N, Chong JM, Fukayama T, et al. Global gene expression analysis of gastric cancer by oligonucleotide microarrays. Cancer Res 2002;62: 233– 40. 6. Stremmel C, Wein A, Hohenberger W, Reingruber B. DNA microarrays: a new diagnostic tool and its implications in colorectal cancer. Int J Colorectal Dis 2002;17:131– 6. 7. DeRisi JL, Penland L, Brown PO, Bittner ML, Meltzer PS, Ray M, et al. Use of a cDNA microarray to analyse gene expression patterns in human cancer. Nat Genet 1996;14:457– 60. 8. Takemasa I, Higuchi H, Yamamoto H, Sekimoto M, Tomita N, Nakamori S, et al. construction of preferential cDNA microarray specialized for human colorectal carcinoma: molecular sketch of colorectal cancer. Biochem Biophys Res Commun 2001;285: 1244 –9. 9. Huang J, Qi R, Quackenbush J, Dauway E, Lazaridis E, Yeatman T, Graveel CR, et al. Effects of ischemia on gene expression. Expression profiling and identification of novel genes in hepatocellular carcinomas. Tissue microarrays for gene amplification surverys in many different tumor types. J Surg Res 2001;99: 222–7. 10. Emmert-Buck MR, Bonner RF, Smith PD, Duaqui RF, Zhuang Z, Goldstein SR, et al. Laser capture microdissection. Science 1996;274:998 –1001. 11. Imamichi Y, Lahr G, Wedlich D. Laser-mediated microdissection of paraffin sections from Xenopus embryos allows detection of tissue-specific expressed mRNAs. Dev Genes Evol 2001;211: 361– 6. 12. Rubin MA. Use of laser capture microdissection, cDNA microarrays, and tissue microarrays in advancing our understanding of prostate cancer. J Pathol 2001;195:80 – 6. 13. Reymond MA, Sanchez JC, Hughes GJ, Gunther K, Riese J, Tortola S, et al. Standardized characterization of gene expression in human colorectal epithelium by two-dimensional electrophoresis. Electrophoresis 1997;18:2842– 8. 14. Nachamkin I, Panaro NJ, Li M, Ung H, Yuen PK, Kricka LJ, et al. Agilent 2100 bioanalyzer for restriction fragment length polymorphism analysis of the Campylobacter jejuni flagellin gene. J Clin Microbiol 2001;39:754 –7. 15. Hsiao LL, Dangond F, Yoshida T, Hong R, Jensen RV, Misra J, et al. A compendium of gene expression in normal human tissues. Physiol Genom 2001;7:97–104. 16. Kurian KM, Watson CJ, Wyllie AH. DNA chip technology. J Pathol 1999;187:267–71.

J Lab Clin Med Volume 143, Number 6

17. Arends JW. Molecular interactions in the Vogelstein model of colorectal carcinoma. J Pathol 2000;190:412– 6. 18. Puskas LG, Zvara A, Hackler L Jr, Van Hummelen P. RNA amplification results in reproducible microarray data with slight

Croner et al

351

ratio bias. Biotechniques 2002;32:1330 – 4, 1336, 1338, 1340. 19. Latza U, Niedobitek G, Schwarting R, Nekarda H, Stein H. Ber-EP4: new monoclonal antibody which distinguishes epithelia from mesothelia. J Clin Pathol 1990;43:213–9.