Applications of tissue microarray technology in immunohistochemistry: A study on c-kit expression in small cell lung cancer

Applications of Tissue Microarray Technology in Immunohistochemistry: A Study on C-Kit Expression in Small Cell Lung Cancer VALENTINA DONATI, MD, PINUCCIA FAVIANA, MD, MATTEO DELL’OMODARME, MARIA CRISTINA PRATI, TIZIANO CAMACCI, KATIA DE IESO, RICCARDO GIANNINI, PhD, MARCO LUCCHI, MD, ALFREDO MUSSI, MD, RAFFAELE PINGITORE, MD, FULVIO BASOLO, AND GABRIELLA FONTANINI, MD Tissue microarray technology allows the immediate evaluation of molecular profiles of numerous different tissues, with savings of money and time. It was created for rapid, large-scale molecular studies, and the main concern regarding its possible broad acceptance is that the analysis of tissue microarrays instead of whole tissue sections may lead to false negative or positive results because of tissue heterogeneity. In the present study, we analyzed in 54 small cell lung cancers, by immunohistochemistry, the expression of the antigen c-kit, which seems to be important in these neoplasms’ tumorigenesis, and compared the staining obtained on whole sections with that of the corresponding tissue microarrays. Although c-kit expression of the whole sections agreed with that of the corresponding biopsies in many cases, the correlation between whole sections and all the companion nonlost single cores or their mean value turned out to be highly significant only if the 36 double negatives (ie, both whole sections and companion tissue microarrays negative) were included

(P < 0.0001). In fact, if only cases positive to at least 1 of the tests (ie whole sections or corresponding tissue microarrays positive) were considered, the correlation was not significant (P ⴝ 0.055). Tissue microarrays showed a good specificity (94.2% for all single cores and 92.3% for their mean value) but a rather poor sensitivity (respectively, 69.4% and 71.4%). Moreover, a high percentage (13.4%) of cores was lost, and this loss was not random. To sum up, in our experience, tissue microarray technology cannot be a substitute for whole sections in clinical diagnosis of individual cases. HUM PATHOL 35:1347-1352. © 2004 Elsevier Inc. All rights reserved. Key words: tissue microarray, small cell lung cancer, c-kit, immunohistochemistry. Abbreviations: TMA, tissue microarray; IHC, immunohistochemistry; SCLC, small cell lung cancer; SCF, stem-cell factor; WS, whole tissue section.

Tissue microarray (TMA) technology is a new technique, devised by Kononen and colleagues in 1998,1 that has many advantages compared with the conventional methodologies because it makes it possible to analyze at once molecular targets, whether at the DNA, RNA, or protein level, in many normal and pathologic tissues by using archival formalin-fixed, paraffin-embedded specimens, with a concomitant saving of time, money, materials, and technical resources. The innovative aspect of TMAs is the simultaneous analysis of tissue from many specimens of different type (tumors, xenografts, or cell lines) by using a variety of techniques and staining procedures (hematoxylin and eosin, immunohistochemistry [IHC], in situ hybridization)1-4 under highly standardized conditions, because all specimens

are processed at the same time and in an identical way.5 Moreover, it permits the use of archival specimens in high-throughput molecular analysis (unlike other highthroughput techniques, such as cDNA microarray analysis, serial analysis of gene expression (SAGE), or proteomics screens) and the preservation of precious and rare tissue, with a low consumption of “donor” specimens.6 However, this technology, born as a populationlevel research tool, has some obvious defects and theoretical objections, the greatest of which is that core biopsies could be too small to capture all the information from large and heterogeneous tumors.5 The most common application of TMA technique is in the study of the molecular profile of different tumors by detecting protein antigens, by using IHC with antibodies against a single or several different molecular targets. Small cell lung cancer (SCLC) represents about 20% of all lung cancer types. These neoplasms have an aggressive clinical behavior with high growth rate, early and widespread metastases, and poor prognosis, with a life expectancy without therapy of less than 4 months7 and a 5-year survival rate after chemotherapy between 3% and 7%.8,9 For these reasons, there is an obvious need for fuller comprehension of the molecular properties of SCLC to project new therapeutic strategies.

From the Department of Surgery, Division of Surgical Pathology, University of Pisa; the Department of Physics, Scuola Normale Superiore; INFN, Section of Pisa; the Department of Oncology, Transplants and New Technologies in Medicine, University of Pisa; and the Department of Cardio-Thoracic Surgery, University of Pisa, Pisa, Italy. Accepted for publication July 29, 2004. Supported by grants from the Associazione Italiana per la Ricerca sul Cancro (AIRC), Italy. Address correspondence and reprint requests to Gabriella Fontanini, MD, Department of Oncology, Transplants and New Technologies in Medicine, University of Pisa, via Roma 57, 56100 Pisa, Italy. 0046-8177/$—see front matter © 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.humpath.2004.07.016

1347

HUMAN PATHOLOGY

Volume 35, No. 11 (November 2004)

Many molecular markers are involved in pathogenesis and prognosis of SCLC10,11: One of them seems to be the c-kit proto-oncogene. It encodes a transmembrane tyrosine-kinase growth-factor receptor protein (145 kDa) similar in structure to the platelet-derived growth-factor receptor and the colony-stimulating factor-1 receptor. Binding of stem-cell factor (SCF), which is kit receptor’s ligand, to c-kit results in autophosphorylation on cytoplasmic tyrosines in the COOH-terminal tail region and consequent activation of its kinase activity and of several signaling cascades that control cellular activities such as proliferation, apoptosis, and motility.12 Kit is constitutionally expressed in hemopoietic cells, mast cells, germ cells, melanocytes, and interstitial cells of Cajal13-16 and is pathologically coexpressed with its ligand SCF in many tumors, such as myeloid leukemia, mastocytoma, neuroblastoma, PNET (peripheral neuro-ectodermic tumors), breast tumors, colon tumors, gynecologic tumors, testicular germ-cell tumors, and SCLC.14,17,18 c-Kit receptor expression in SCLCs appears to have clinical importance, because it has been observed that c-kit–positive tumors have a poor prognosis.19 For these patients, new therapies are urgently needed: One of these could be the inhibition of c-kit by humanized monoclonal antibodies, similar to the monoclonal antibody anti-extracellular domain of the HER/neu receptor (herceptin, transtazumab) in breast cancer. Another tyrosine-kinase inhibitor is STI571 (Gleevec), which blocks cellular Abl tyrosine-kinase activity and platelet-derived growth factor/c-kit receptor tyrosine kinases. On this basis, to determine whether small core biopsies can properly represent a conventional tissue section, yielding meaningful information regarding the expression of a particular antigen by a large tumor “donor” specimen, and, in the affirmative, to determine how many disks are required to obtain a result equivalent to the whole specimen, we built arrays with a variable number (from 1 to 7) of cores taken from each of the 54 cases of SCLC. Microarrays and the original blocks from which they were extracted were stained for c-kit tyrosine-kinase receptor. As stated earlier in this article, we chose to evaluate the expression of this antigen because it seems to have great importance in SCLC tumorigenesis and to offer opportunities for planning new therapies. MATERIALS AND METHODS Specimens Archival formalin-fixed, paraffin-embedded tissue sections from a series of 54 patients (50 males and 4 females; age range: 34-75 years) who had undergone surgical resection for small cell lung cancer at the Department of Thoracic Surgery of the University of Pisa in the period of 1976-1997 were selected.

Building TMAs A fresh section was cut from each donor block, stained with hematoxylin and eosin, and used as a guide to select the

morphologically most-representative regions of the tumor from which to sample the individual core needle biopsies (Fig 1A and B). A variable number, which ranged from 1 to 7, of 0.6-mm diameter cores were then punched from tumor areas of each donor tissue block and introduced into previously prepared recipient paraffin blocks of 30 ⫻ 25 mm with 0.3-mm spacing, after having made hosting holes in the blocks with a Tissue Microarrayer (Beecher Instruments, Silver Spring, MD). We constructed 9 recipient blocks with a maximum of 7 ⫻ 7 dots, building a grid with TMAs taken from the same tumor specimen on each single row. With a microtome, 5-␮m sections were cut from the TMA blocks to generate TMA slides for molecular analyses.

Immunohistochemistry Immunohistochemical staining patterns for c-kit were evaluated both in individual cores and in their companion whole sections (Fig 1C and D). Five-micrometer sections, cut from “donor” and corresponding TMA blocks, were stained with anti– c-kit antibody (Novocastra Laboratories, Newcastle upon Tyne, UK; NCLCD117 clone, dilution 1:20; 30 minutes) with a Ventana (Tucson, AZ) automated immunohistochemical stainer according to the manufacturer’s guidelines. To unmask the antigens, the slides were microwave treated in 10 mmol/L citrate buffer, pH 6, for a total of 10 minutes. C-kit expression was evaluated semiquantitatively by counting the percentage of tumor cells with cytoplasmic and membrane immunoreactivity for c-kit. The intensity of the staining was not taken into account because its evaluation is rather subjective and there is no clear evidence of its clinical relevance. In the case of whole tissue sections, c-kit expression was determined by counting the percentage of positive cells in a total of 1000 tumor cells (100 cells ⫻ 10 HPF); in the case of TMAs, c-kit expression was evaluated as the percentage of labeled cells over all tumor cells. We arbitrarily decided to consider as positive the sections with a number of immunoreactive tumor cells above the threshold value of 10%, consensually with the literature.19

Statistical Analysis The results of TMA technique were compared with those of the analysis of the whole tissue section (WS) by calculating specificity, sensitivity, predictive values of the positive and negative tests, and accuracy. A ␹2 goodness-of-fit test for the distribution of the number of lost cores was performed. A Spearman rank correlation between c-kit expression of TMA and of the conventional WS technique was performed. All these methods are described in Armitage et al.20 The analysis of a generalized linear model (logistic regression) was performed by using R 1.6.2,21 with WS c-kit expression as dependent variable and mean TMA c-kit expression as covariate. The model confidence intervals were calculated as stated in Hosmer and Lemeshow.22

RESULTS Histology A total of 216 cores were taken from the 54 cases included in the study. Upon microscopic examination, 29 cores (13.4%) were found to be lost. In Table 1, the frequency of cases with a fixed number of lost cores is displayed (in 16 cases, a variable number of cores ranging from 1 to 4 were missed). In 1 case, all 4 of the cores

1348

C-KIT EXPRESSION IN SMALL CELL LUNG CANCER (Donati et al)

FIGURE 1. A hematoxylin and eosin section taken from one block after the removal of a tissue biopsy (A) and the corresponding tissue core (B). Note that the full section is not damaged by the punching and that the core contains neoplastic tissue. Immunohistochemical staining for c-kit of a whole section (C) and of the tissue core taken from the corresponding area (D).

taken were lost. The distribution of the number of cases n(x) with x lost cores was found to agree with a negative binomial distribution of parameters, P ⫽ 0.496 and r ⫽ 0.520, by means of a ␹2 goodness-of-fit test (P ⫽ 0.062), whereas it was found to disagree with a Poisson distribution with mean ␮ ⫽ 0.537 (P ⫽ 0.005). Therefore, the cores were not randomly lost. There was a large number of cases with no cores lost, whereas in a few cases many cores were lost. It has been calculated that to have at least 1 nonlost core in 95% of the cases, one should take 3 cores from each case, whereas 4 cores are TABLE 1. Frequency Distribution of Cases With x Lost Cores Lost Cores (x)

No. of cases n(x)

0

1

2

3

4

38

10

1

3

2

needed to have at least 2 nonlost cores in 95% of the cases. All the cores present on the slides contained tumor cells. Immunohistochemistry In our study, the percentage of c-kit–positive tumors, based on conventional section staining, was 27.8% (15 of 54 cases). The pattern of staining of c-kit was heterogeneous in SCLC on whole sections. In most cases, we observed the presence of an internal control (labeled stromal cells and hemopoietic cells), although our control was outer and represented by a gastrointestinal stromal tumor. Comparing the c-kit expressions of WS test with those of TMA technique, one obtains the results collected in Table 2. A test was considered positive if its c-kit expression was above the threshold value of 10% positive tumor cells. All single-core results were compared with the corresponding WS. As one can see from

1349

HUMAN PATHOLOGY

Volume 35, No. 11 (November 2004)

TABLE 2. Comparison of the Results of the WholeSection (WS) Method and the Tissue Microarray (TMA) Technology Test Results Test WS/1 TMA WS/2 TMA WS/3 TMA WS/4 TMA WS/5 TMA WS/6 TMA WS/7 TMA WS/mean TMA

Cases (n)

⫹/⫹

⫹/⫺

⫺/⫹

⫺/⫺

Lost

Total

53 50 41 23 11 6 3

34 49 35 13 2 0 0

15 19 13 5 1 0 0

8 23 31 25 11 2 0

130 202 202 144 72 22 3

29 90 128 107 55 16 2

216 383 409 294 141 40 5

53

10

4

3

36

1

54

NOTE. In the first line, single cores are considered. In the second line, the mean value over couples of nonlost cores is compared with the WS result. In the following lines, the similar results for triples, quadruplets, and so on are presented. In the last line, the mean value over all cores that survived for each case is given. The threshold value of 10% of c-kit expression was chosen for all cases.

Table 3, this test has a rather poor sensitivity (S ⫽ 69.4%) and a good specificity (S' ⫽ 94.2%). The accuracy is calculated in 2 ways: A ⫽ number of cores agreeing with WS/number of nonlost cores (87.7%) and A' ⫽ number of cores agreeing with WS/number of all cores (75.9%; 95% confidence interval, 69.7%81.5%). In an analogous way, a test with couples of nonlost cores was considered whenever possible (50 clinical cases). A couple containing at least 1 lost core was considered lost. The mean value of the c-kit expression of the 2 cores was compared with the WS value. The indexes of this test are presented in the second line of Table 3, showing no improvement with respect to the single-core test. In this case, A' is defined as the number of couples agreeing with WS divided by the number of all couples. In the remaining lines of Table 3, tests with 3 and 4 nonlost cores were similarly treated. The last line concerns the case in which the mean value is taken over all nonlost cores of each clinical case. The indexes

of this test (S ⫽ 71.4%, S' ⫽ 92.3%, A ⫽ 86.8%, A' ⫽ 85.2%) are similar to those of the single-core test. A Spearman rank correlation between the c-kit expression of WS and the mean c-kit expression of all nonlost cores was found to be not significant (rs ⫽ 0.489, P ⫽ 0.055) if only cases positive to at least 1 of the tests were considered. The correlation turned out to be highly significant if the 36 double-negative cases were included (rs ⫽ 0.782, P ⬍ 0.0001). A generalized linear model with WS c-kit expression as dependent variable and mean TMA c-kit expression as covariate was found to be not significant (P value ⫽ 0.129 for the covariate). From Fig 2, one can see that 6 of 17 experimental points (35.3%) lay outside the 95% confidence interval of the model, that is, they are not interpreted by the model. Furthermore, for 11 points (64.7%), the WS c-kit expression is larger than the corresponding TMA value. DISCUSSION As we have stated, TMA technique is a new method that permits high-throughput immunohistochemical and molecular analyses with a saving of money and time, by evaluating a large number of cases on a single glass slide. This technique was born as a populationlevel research tool to study the expression profiles of wide populations of normal and pathologic tissue, neoplasms in particular. In fact, there are many applications of TMA technique for the analysis of tumor antigen expression, especially for the receptorial status of breast cancer.5,23-25 However, little is known about the feasibility and validity of TMAs in examining the molecular profile for the characterization of single individual cases or about the minimum number of cores needed to properly represent a whole section. Till now, few recent studies have investigated the relationship between TMAs and whole sections. Camp and colleagues5 affirmed that 2 needle core biopsies

TABLE 3. Indexes for the Tests of Table 2 Test Indexes (%)

Test

S

S’

PPV

NPV

A

A’ (95% Confidence Interval)

WS/1 TMA WS/2 TMA WS/3 TMA WS/4 TMA WS/mean TMA

69.4 72.1 72.9 72.2

94.2 89.8 86.7 85.2

81.0 68.1 53.0 34.2

89.7 91.4 94.0 96.6

87.7 85.7 84.3 84.0

75.9 (69.7–81.5) 65.5 (60.5–70.3) 57.9 (53.0–62.8) 53.4 (47.5–59.2)

71.4

92.3

76.9

90.0

86.8

85.2 (72.9–93.4)

NOTE. All of these indexes were calculated without taking into account lost cores, whereas A’ is the accuracy with lost cores included. Abbreviations: WS, whole section; TMA, tissue microarray; S, sensitivity; S’, specificity; PPV, predictive value of the positive test; NPV, predictive value of the negative test; A, accuracy.

FIGURE 2. Graph of the estimated mean tissue microarray (TMA) c-kit expression and 95% confidence interval as a function of the whole section c-kit expression. The diamonds represent the experimental results. On the dashed line, mean TMA value equals whole tissue section value.

1350

C-KIT EXPRESSION IN SMALL CELL LUNG CANCER (Donati et al)

adequately represent the immunohistochemical antigen expression (estrogen receptor [ER], progesterone receptor [PR], and HER2/neu) of a whole tumor tissue (breast cancer) section with ⬎95% accuracy, whereas other investigators concluded that triplet-core TMAs are sufficient for immunohistochemical profiling of individual cases and that in case of discordance between cores, it is useful to refer to the majority scores (ie, the majority of 3 cores determines the interpretation of the result). In fact, triplet-core TMAs led to a concordance rate with whole sections of 91%-98% in fibroblastic tumor resections by using 3 immunohistochemical stains (Ki-67, p53, and the retinoblastoma protein) in a study by Hoos et al,26 and this was a valid and efficient way to study endoscopic gastric cancer with a high correlation between biopsies and cores.27 Moreover, Jourdan et al28 concluded that in 30 colorectal carcinomas, the analysis of 3 disks per case was comparable to the analysis of whole sections in 99.6% (for p53), in 98.8% (for hMLH1), and 99.2% (for hMSH2) of cases. In the present research, we analyzed the expression of c-kit tyrosine-kinase receptor in SCLCs by the TMA method, in light of the putative role that this receptor may have in these neoplasms. In fact, some preclinical studies have shown a growth inhibition of SCLC cell lines by c-kit inhibitor such as STI-571. In our study, the best agreement between results of multicore arrays and whole sections was found for negative cases (specificity, S' ⫽ 92.3%), whereas for positive cases, the agreement was poorer (sensitivity, S ⫽ 71.4%). Although the WS c-kit expression and the corresponding TMA c-kit expression agree in many cases (A' ⫽ 75.9% for the single core test, A' ⫽ 85.2% for the test using the mean value of all nonlost cores), cores are not always representative of the conventional tissue section and cannot substitute for it in making a clinical diagnosis of individual cases. Gillett and colleagues29 arrived at a similar conclusion analyzing ER and PR immunohistochemical expression in 157 breast cancers: Despite a highly significant association in the ER and PR expression between whole sections and multicore systems, it was observed that the cores are not always fully representative of the diagnostic section because they underestimate or overestimate ER or PR status in some cases. In 5 of 53 clinical cases, cores had disagreeing c-kit expressions (positive and negative cores taken from the same block). These discrepancies are possibly due to the fact that the tumor has a heterogeneous c-kit expression. A high percentage of lost cores (13.4%) was observed, and therefore a more accurate technique for preparing cores is needed. The loss of TMA cores may be caused by 2 factors: First, that of TMA floating off during the IHC procedure and second, different thicknesses of donor blocks from which cores were taken. When the donor block is thin, loss is highly probable for most cores taken. This fact can also explain why cores are lost in clusters, as implied by negative binomial distribution. In 1 case, we missed all the cores taken from the original tumor; so if in that patient we

had decided to evaluate c-kit expression only in TMAs, we would have obtained no information. Unlike what was shown by the above-mentioned investigators (5, 26-28), who concluded that 2 or 3 needle core biopsies adequately represent the antigen expression on a whole tumor tissue section with ⬎95% accuracy, the lack of significance of both Spearman rank correlation and of the generalized linear model shows that for our tumors, TMA technology cannot replace the conventional technique. c-Kit expression on SCLC whole sections was positive only in 15 cases of 54 (27.8%), a percentage lower than those reported in the literature30-34 but similar to that (37%) observed by Micke et al19 in their recent work. Agreeing with Micke and colleagues, we think that this discordance could be explained by the use of different samples (cell lines and fresh-frozen sections in the studies with a higher percentage versus paraffinembedded tissue in our work and in that of Micke et al19), by the definition of positive immunostaining (we used a strict definition; in fact, we considered as positive only the tumors with cytoplasmic and clear-cut membrane staining), and by the number of patients evaluated (a small number in the studies with a higher percentage versus a large number in our study). To sum up, in our study, although the wholesection and the corresponding TMA c-kit expressions agree in many cases, cores are not always representative of the conventional tissue section and cannot substitute for it in making a clinical diagnosis of individual cases, also because of the high percentage of lost cores (13.4%) observed.

REFERENCES 1. Kononen J, Bubendorf L, Kallioniemi A, et al: Tissue microarrays for high-throughput molecular profiling of tumor specimens. Nat Med 4:844-847, 1998 2. Bubendorf L, Kolmer M, Kononen J, et al: Hormone therapy failure in human prostate cancer: Analysis by complementary DNA and tissue microarrays. J Natl Cancer Inst 91:1758-1764, 1999 3. Moch H, Schraml P, Bubendorf L, et al: High-throughput tissue microarray analysis to evaluate genes uncovered by cDNA microarray screening in renal cell carcinoma. Am J Pathol 154:981986, 1999 4. Mucci NR, Akdas G, Manely S, et al: Neuroendocrine expression in metastatic prostate cancer: Evaluation of high throughput tissue microarrays to detect heterogeneous protein expression. HUM PATHOL 31:406-414, 2000 5. Camp RL, Charette LA, Rimm DL. Validation of tissue microarray technology in breast carcinoma. Lab Invest 80:1943-1949, 2000 6. Mills SE, Fechner RE, Frierson HF, et al: Guardians of the wax and the patient. Am J Clin Pathol 110:758-764, 1995 7. Zelen M: Keynote address on biostatistics and data retrieval. Cancer Chemother Rep 4:31-42, 1973 8. Albain KS, Crowley JJ, Livingston RB: Long-term survival and toxicity in small cell lung cancer. Expanded Southwest Oncology Group experience. Chest 99:1425-1432, 1991 9. Sørensen JB, Østerlind K: Prognostic factors: From clinical parameters to new biological markers, in Van Houtte P, Klastersky J, Rocmans P (eds): Progress and Perspectives in the Treatment of Lung Cancer. Berlin, Springer, 1999, pp 1-21 10. Salgia R, Skarin AT: Molecular abnormalities in lung cancer. J Clin Oncol 16:1207-1217, 1998

1351

HUMAN PATHOLOGY

Volume 35, No. 11 (November 2004)

11. Rozengurt E: Autocrine loops, signal transduction, and cell cycle abnormalities in the molecular biology of lung cancer. Curr Opin Oncol 11:116-122, 1999 12. Ashman LK: The biology of stem cell factor and its receptor C-kit. Int J Biochem Cell Biol 31:1037-1051, 1999 13. Maeda H, Yamagata A, Nishikawa S, et al: Requirement of c-kit for development of intestinal pacemaker system. Development 116:369-375, 1992 14. Tsuura Y, Hiraki H, Watanabe K, et al: Preferential localization of c-kit product in tissue mast cells, basal cells of skin, epithelial cells of breast, small cell lung carcinoma and seminoma/dysgerminoma in human: Immunohistochemical study on formalin-fixed, paraffin-embedded tissues. Virchows Arch 424:135-141, 1994 15. Sanders KM: A case for interstitial cells as pacemakers and mediators of neurotransmission in the gastrointestinal tract. Gastroenterology 111:492-515, 1996 16. Natali PG, Nicotra MR, Sures I, et al: Expression of c-kit receptor in normal and transformed human nonlymphoid tissues. Cancer Res 52:6139-6143, 1992 17. Heinrich M, Blanke C, Druker B, et al: Inhibition of KIT tyrosine kinase activity: A novel molecular approach to the treatment of KIT-positive malignancies. J Clin Oncol 20:1692-1703, 2002 18. Ricotti E, Fagioli F, Garelli E, et al: C-kit is expressed in soft tissue sarcoma of neuroectodermic origin and its ligand prevents apoptosis of neoplastic cells. Blood 91:2397-2405, 1998 19. Micke P, Basrai M, Faldum A, et al: Characterization of c-kit expression in small cell lung cancer: Prognostic and therapeutic implications. Clin Cancer Res 9:188-194, 2003 20. Armitage P, Berry G, Matthews JNS: Statistical Methods in Medical Research. 4th ed. New York: Blackwell, 2002 21. Ihaka R, Gentleman R: A language for data analysis and graphics. J Comput Graphical Stat 5:299-314, 1996 22. Hosmer DW, Lemeshow S: Applied Logistic Regression. New York, Wiley, 2000 23. Zhang D, Salto-Tellez M, Do E, et al: Evaluation of HER-2/ neu oncogene status in breast tumours on tissue microarrays. HUM PATHOL 34:362-368, 2003 24. Zhang D, Salto-Tellez M, Putti TC, et al: Reliability of tissue

microarrays in detecting protein expression and gene amplification in breast cancer. Mod Pathol 16:79-85, 2003 25. O’Grady A, Flahavan CM, Kay EW, et al: HER-2 analysis in tissue microarrays of archival human breast cancer: Comparison of immunohistochemistry and fluorescence in situ hybridisation. Appl Immunohistochem Mol Morphol 11:177-182, 2003 26. Hoos A, Urist MJ, Stojadinovic A, et al: Validation of tissue microarrays for immunohistochemical profiling of cancer specimens using the example of human fibroblastic tumors. Am J Pathol 158: 1245-1251, 2001 27. Gulmann C, Butler D, Kay E, et al: Biopsy of a biopsy: Validation of immunoprofiling in gastric cancer biopsy tissue microarrays. Histopathology 42:70-76, 2003 28. Jourdan F, Sebbagh N, Comperat E, et al: Tissue microarray technology: Validation in colorectal carcinoma and analysis of p53, hMLH1, and hMSH2 immunohistochemical expression. Virchows Arch 443:115-121, 2003 29. Gillett CE, Springall RJ, Barnes DM, et al: Multiple tissue core arrays in histopathology research: A validation study. J Pathol 192:549-553, 2000 30. Krystal GW, Hines SJ, Organ CP: Autocrine growth of small cell lung cancer mediated by coexpression of c-kit and stem cell factor. Cancer Res 56:370-376, 1996 31. Hibi K, Takahashi T, Sekido Y, et al: Coexpression of the stem cell factor and the c-kit genes in small-cell lung cancer. Oncogene 6:2291-2296, 1991 32. Sekido Y, Obata Y, Ueda R, et al: Preferential expression of c-kit protooncogene transcripts in small cell lung cancer. Cancer Res 51:2416-2419, 1991 33. Papadimitriou CA, Topp MS, Serve H, et al: Recombinant human stem cell factor does exert minor stimulation of growth in small cell lung cancer and melanoma cell lines. Eur J Cancer 31:23712378, 1995 34. Rygaard K, Nakamura T, Spang-Thomsen M: Expression of the proto-oncogenes c-met and c-kit and their ligands, hepatocyte growth factor/scatter factor and stem cell factor, in SCLC cell lines and xenografts. Br J Cancer 67:37-46, 1993

1352