View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Available online at www.sciencedirect.com

Genomics 90 (2007) 559–566 www.elsevier.com/locate/ygeno

Identification of associated with tumorigenesis of meibomian cell carcinoma by microarray analysis ⁎ Arun Kumar a, , Syril Kumar Dorairaj b, Venkatesh C. Prabhakaran b, D. Ravi Prakash b, Sanjukta Chakraborty a

a Department of Molecular Reproduction, Development, and Genetics, Indian Institute of Science, Bangalore 560012, India b Minto Ophthalmic Hospital, Bangalore, India

Received 15 May 2007; accepted 20 July 2007 Available online 21 September 2007

Abstract

Meibomian cell carcinoma (MCC) is a malignant tumor of the meibomian glands located in the eyelids. No information exists on the cytogenetic and genetic aspects of MCC. There is no report on the expression profile of MCC. Thus there is a need, for both scientific and clinical reasons, to identify genes and pathways that are involved in the development and progression of MCC. We analyzed the profile of MCC by the microarray technique. Forty-four genes were upregulated and 149 genes were downregulated in MCC. Differential expression data were confirmed for 5 genes by semiquantitative RT-PCR in MCC tumors: GTF2H4, RBM12, UBE2D3, DDX17, and LZTS1. We found dysregulation of two major pathways in MCC: MAPK and JAK/STAT. Clusters of genes on 1, 12, and 19 were dysregulated in MCC. The data presented here will facilitate the identification of specific markers and therapeutic targets for the treatment of MCC patients. © 2007 Elsevier Inc. All rights reserved.

Keywords: Expression profile; Meibomian cell carcinoma; MAPK signaling pathway; JAK/STAT pathway; Microarrays

Meibomian glands are responsible for the maintenance of the Although microarray analysis has been used extensively to health and integrity of the ocular surface. They promote stability provide a vast amount of knowledge on etiology and and prevent evaporation of the preocular tear film through the biomarkers for therapeutic applications by profiling the synthesis and secretion of lipids at the lid margin [1]. First transcriptome of many , no microarray analysis has described by Meibomius, these oil-producing sebaceous glands been done previously on this tumor. The main aim of this study number around 30–40 in the upper and 20–30 in the lower was to identify genes involved in the tumorigenesis of this eyelids [2]. Meibomian cell carcinoma (MCC), also referred to tumor. We report the first gene expression profile of this tumor as sebaceous carcinoma, is a malignant tumor that occurs most using cDNA microarray analysis. commonly on the eyelids [3]. Although rare, MCC is the fourth most common tumor among eyelid tumors after basal cell Results carcinoma, squamous cell carcinoma, and cutaneous melanomas [4]. Local recurrence and regional lymph node metastasis are Differential gene expression in MCC common in MCC. In addition, this neoplasm has also been reported to spread to the trigeminal ganglion and cerebellum [5]. Microarray analysis of MCC yielded a data set with 193 The only known therapy for this cancer is the surgical excision of differentially regulated genes. Of these, 44 genes were the tumor, but it reappears after some time following the surgery. upregulated and 149 were downregulated (Supplemental Table Nothing is known about the cytogenetic and genetic events 1S). These included genes involved in various cellular functions involved in the development and progression of this neoplasm. such as and cell proliferation, signal transduction, transcription and translation, cell death and , enzyme ⁎ Corresponding author. Fax: +91 80 2360 0999. metabolism, and cell adhesion. Some of the selected genes are E-mail address: [email protected] (A. Kumar). listed in Table 1. Our analysis showed dysregulation of two

0888-7543/$ - see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ygeno.2007.07.008 560 A. Kumar et al. / Genomics 90 (2007) 559–566

Table 1 Selected differentially regulated genes with known function in MCC Gene Symbol UniGene Expt 1 Expt 2 Average location ratio Cell cycle/cell proliferation DEAD (Asp-Glu-Ala-Asp) box polypeptide 17 DDX17 Hs.528305 22q13.1 0.278 0.201 0.240 , putative tumor suppressor 1 LZTS1 Hs.521432 8p22 0.453 0.463 0.458 CDK2-associated 1 CDK2AP1 Hs.433201 12q24.31 0.374 0.598 0.486 Metallothionein 3 (growth inhibitory factor) MT3 Hs.73133 16q13 0.449 0.532 0.491 Glycoprotein (transmembrane) nmb GPNMB Hs.190495 7p15 0.400 0.583 0.492 DEAD (Asp-Glu-Ala-As) box polypeptide 19 DDX19B Hs.221761 16q22.1 2.197 4.079 3.138

Signal transduction/ Signal transducer and activator of transcription 2 STAT2 Hs.530595 12q13.2 0.301 0.543 0.422 Casein 1, ε CSNK1E Hs.474833 22q13.1 0.273 0.606 0.440 Serum/glucocorticoid-regulated kinase SGK Hs.510078 6q23 0.393 0.567 0.480 Tyrosine 3-mono-oxygenase β polypeptide YWHAB Hs.651212 20q13.1 0.470 0.513 0.492 Neuronal PAS domain protein 3 NPAS3 Hs.659456 14q12–q13 0.418 0.590 0.504 Centaurin, β1 CENTB1 Hs.337242 17p13.1 0.578 0.431 0.505 Amyloid β (A4) precursor protein-binding, APBB2 Hs.479602 4p14 0.424 0.590 0.507 family B, member 2 Dual-specificity phosphatase 19 DUSP19 Hs.132237 2q32.1 0.402 0.614 0.508 Serine/threonine kinase 38 STK38 Hs.409578 6p21 0.568 0.451 0.510 Phospholipase C, δ4 PLCD4 Hs.632528 2q35 0.539 0.494 0.517 MAP/microtubule affinity-regulating kinase 4 MARK4 Hs.34314 19q13.32 2.577 1.773 2.180 Interleukin 23, α subunit p19 IL23A Hs.98309 12q13.2 1.664 5.726 3.700

Transcription/translation NSFL1 (p97) cofactor (p47) NSFL1C Hs.12865 20p13 0.304 0.326 0.315 Ribosomal protein S4, X-linked RPS4X Hs.446628 Xq13.1 0.280 0.599 0.440 Nuclear factor of activated T cells 5 NFAT5 Hs.371987 16q22.1 0.516 0.451 0.483 General IIH, polypeptide 1 GTF2H1 Hs.577202 11p15.1–p14 0.585 0.461 0.523 C9 HOXC9 Hs.658823 12q13.13 0.607 0.470 0.539 BarH-like homeobox 1 BARX1 Hs.164960 9q12 2.335 1.997 2.166 Activating signal cointegrator 1 complex subunit 2 ASCC2 Hs.517438 22q12.1 1.678 2.874 2.276 Polymerase (RNA) II (DNA-directed) POLR2E Hs.24301 19p13.3 2.022 3.212 2.617 polypeptide E General transcription factor IIH, polypeptide 4 GTF2H4 Hs.485070 6p21.3 2.549 3.901 3.225 U5 snRNP-specific protein, 116 kDa EFTUD2 Hs.151787 17q21.31 1.789 4.869 3.329 RNA binding motif protein 12 RBM12 Hs.246413 20q11.21 2.701 4.378 3.540

Cell death/apoptosis Baculoviral IAP repeat-containing 4 BIRC4 Hs.356076 Xq25 0.245 0.507 0.376 Interleukin 24 IL24 Hs.658964 1q32.1 0.289 0.590 0.440 Growth arrest and DNA-damage-inducible, γ GADD45G Hs.9701 9q22 0.437 0.607 0.522 Tumor necrosis factor superfamily, TNFRSF21 Hs.443577 6p21.1–p12.2 0.584 0.469 0.527 member 21 CD74 antigen CD74 Hs.694723 5q32 2.470 2.227 2.349

Enzyme/metabolism Ectonucleoside triphosphate diphosphohydrolase 2 ENTPD2 Hs.123036 9q34 0.301 0.210 0.256 Carbonic anhydrase XII CA12 Hs.210995 15q22 0.242 0.419 0.331 Betaine-homocysteine methyltransferase 2 BHMT2 Hs.114172 5q13 0.293 0.442 0.368 N-myristoyltransferase 2 NMT2 Hs.60339 10p13 0.206 0.596 0.401 Carboxypeptidase Z CPZ Hs.78068 4p16.1 0.385 0.489 0.437 Collectin subfamily member 12 COLEC12 Hs.464422 18pter–p11.3 0.302 0.609 0.456 Endo-β-N-acetylglucosaminidase FLJ21865 Hs.29288 17q25.3 0.490 0.456 0.473 Methylenetetrahydrofolate dehydrogenase MTHFD1 Hs.652308 14q24 0.373 0.597 0.485 Histidyl-tRNA synthetase 2 HARS2 Hs.432560 20p11.23 0.528 0.444 0.486 Scavenger receptor class A, member 3 SCARA3 Hs.128856 8p21 0.419 0.565 0.492 Pyruvate dehydrogenase kinase, isoenzyme 3 PDK3 Hs.658190 Xp22.11 1.757 2.907 2.332 Ubiquitin-conjugating enzyme E2D 3 UBE2D3 Hs.518773 4q24 2.139 3.282 2.711

Cell adhesion Keratin 1 (epidermolytic hyperkeratosis) KRT1 Hs.80828 12q13.13 0.195 0.612 0.404 Periplakin PPL Hs.192233 16p13.3 0.463 0.348 0.406 A. Kumar et al. / Genomics 90 (2007) 559–566 561

Table 1 (continued) Gene Symbol UniGene Chromosome Expt 1 Expt 2 Average location ratio Cell adhesion ARP3 actin-related protein 3 homolog (yeast) ACTR3 Hs.433512 2q14.1 0.470 0.426 0.448 Matrilin 1, cartilage matrix protein MATN1 Hs.150366 1p35.2 0.575 0.396 0.486 Sarcoglycan, ε SGCE Hs.371199 7q21–q22 0.613 0.378 0.496 Tubulointerstitial nephritis antigen TINAG Hs.127011 6p11.2–p12 2.129 1.995 2.062 Beaded filament structural protein 2, phakinin BFSP2 Hs.659862 3q21–q25 2.715 1.839 2.277 Thrombospondin 3 THBS3 Hs.169875 1q22 1.645 3.365 2.505 major pathways in MCC: mitogen-activated protein kinase DDX17, and LZTS1) in a panel of four MCC and two normal (MAPK) and Janus kinase/signal transducers and activators of eyelid tissues by semiquantitative RT-PCR analysis (Fig. 1). transcription (JAK/STAT) (Table 2). Nothing is known about the functional relevance of the To confirm the microarray results, we performed semiquan- GTF2H4 gene. It is a transcription factor that was found to be titative RT-PCR analysis for five genes in a panel of four MCC upregulated across MCC cases (Fig. 1). Another interesting and two normal eyelid samples: GTF2H4, RBM12, UBE2D3, gene that was also upregulated in MCC is RBM12 (Fig. 1). This DDX17, and LZTS1 (Table 1). The mean mRNA expression gene was first isolated from a human colon carcinoma cell line levels of GTF2H4 (0.53±0.01 in normal vs 2.03±0.29 in cDNA library and contains several RNA binding domains [6]. tumor, p=0.028), RBM12 (0.47±0.13 in normal vs 1.75±0.30 The functional relevance of this gene is unknown and further in tumor, p=0.049), and UBE2D3 (0.52±0.01 in normal vs analysis may lead to some interesting insights with relevance to 1.60±0.18 in tumor, p=0.017) were significantly greater in tumorigenesis. tumor samples compared with the expression levels in normal The ubiquitin–proteasome pathway (UPP) is the universal samples (Fig. 1). The mean expression levels of DDX17 (2.43± and highly substrate-specific housekeeping system involved in 0.39 in normal vs 0.93±0.24 in tumor, p=0.027) and LZTS1 the degradation of damaged, aberrant, and expired . A (1.99±0.25 in normal vs 0.67±0.15 in tumor, p=0.009) were variety of roles for UPP have been elucidated in cellular significantly lower in tumor samples compared to the expres- physiology and pathology, including apoptosis [7]. UBE2D3 sion levels in normal samples (Fig. 1). (ubiquitin-conjugating enzyme E2D3) is one of the components of the UPP pathway. Upregulation of UBE2D3 has been Significant chromosome association reported in oncocytic thyroid tumors by differential display analysis [8]. Interestingly, UBE2D3 was upregulated in MCC A careful analysis of our data showed clusters of tumors (Fig. 1). differentially regulated genes associated with certain chromo- The DDX17 (DEAD box polypeptide 17) gene is located on somes (Fig. 2). For example, we found clusters of genes chromosome 22 and encodes an RNA helicase [9]. The p72 associated with chromosomes 1, 12, and 19 to be differentially (DDX17) is a highly related member of the DEAD box family. regulated in MCC (Fig. 2). It has been implicated in growth regulation and also shown to be involved in both pre-mRNA and pre-rRNA processing [9]. Discussion Wilson et al. [9] have shown that p68 and p72 interact with HDAC1 and repress transcription in a -specific To the best of our knowledge, the microarray analysis of manner. DDX17 was found to be downregulated in MCC MCC has provided the first ever cataloguing of genes that are tumors (Fig. 1). dysregulated in this neoplasm. We have validated the micro- The LZTS1 (leucine zipper, putative tumor suppressor 1) array data for five genes (viz., GTF2H4, RBM12, UBE2D3, gene is located on chromosome 8p22 and encodes a protein with

Table 2 Major pathways dysregulated in MCC Pathway/gene Symbol UniGene Chromosome Expt 1 Expt 2 Average location ratio MAPK signaling Growth arrest and DNA-damage-inducible, γ GADD45G Hs.9701 9q22.1–q22.2 0.437 0.607 0.522 Sterile α motif and leucine zipper-containing kinase AZK ZAK Hs.444451 2q24.2 1.822 3.600 2.711 MAP/microtubule affinity-regulating kinase 4 MARK4 Hs.34314 19q13.32 2.577 1.773 2.175 Tumor necrosis factor receptor superfamily, member 21 TNFRSF21 Hs.443577 6p21.1–q12.2 0.584 0.469 0.527

JAK/STAT signaling Interleukin 24 IL24 Hs.658964 1q32.1 0.289 0.590 0.440 Interleukin 23, α subunit p19 IL23A Hs.98309 12q13.2 1.664 5.726 3.695 Signal transducer and activator of transcription 2, 113 kDa STAT2 Hs.530595 12q13.3 0.301 0.543 0.422 562 A. Kumar et al. / Genomics 90 (2007) 559–566

Previous study on gene expression profiles of cell lines has shown that chromosomal modifications can lead to regional bias in gene expression [13]. When we searched our microarray data for such regional biases, we found clusters of genes on chromosomes 1, 12, and 19 to be differentially regulated in MCC (Supplemental Table 1S, Fig. 2). In addition, clusters of genes in certain segments of these chromosomes were differentially regulated. For example, a cluster of five genes (viz., CD22, PAF1, MARK4, TBCB, and GRIN2D) on chromo- some segment 19q13.12–q13.32 was differentially regulated in MCC (Fig. 2). CD22, PAF1, and MARK4 were upregulated, whereas TBCB and GRIN2D were downregulated (Fig. 2). Huang et al. [14] have recently found aberrant expression of CD22 as a useful marker for the detection of monoclonal B cells admixed with numerous benign polyclonal B cells. PAF1/PD2 (pancreatic differentiation 2) is the human homolog of the yeast RNA polymerase II-associated factor 1 (hPaf1). PAF1 is a nuclear 80-kDa protein, which interacts with RNA polymerase II [15].Moniauxetal.[15] have recently demonstrated that overexpression of PAF1 in NIH 3T3 cells results in enhanced growth in vitro and tumor formation in vivo. Using a detailed FISH analysis of the breakpoints of structural 19q chromosome rearrangements in gliomas, Beghini et al. [16] have identified a 19q13.2 intrachromosomal amplification of the MARK4 (MAP/microtubule affinity-regulating kinase 4) gene in three primary glioblastoma cell lines. They observed that MARK4 is upregulated in both fresh and cultured gliomas and overexpressed in all of the above three glioblastoma cell lines. Taken together, the upregulation of CD22, PAF1, and MARK4 in MCC is not surprising. The status of the other two downregulated genes (TBCB and GRIN2D) in this cluster is not known in any other tumor. The transmission of extracellular signals to their intracellular targets is mediated by a network of interacting proteins that regulate a large number of cellular processes. One such pathway, which involves activation of several signaling molecules followed by a sequential stimulation of several cytoplasmic protein kinases, is the MAPK signaling cascade [17]. Our study found differential expression of a number of genes involved in this pathway (viz. GADD45G, ZAK, MARK4, and TNFRSF21). The GADD45G gene is a member of the MAPK signaling pathway. It is located at the commonly deleted region 9q22 in multiple tumors [18]. GADD45G (CR6), along Fig. 1. Semiquantitative RT-PCR analysis of five genes in two normal eyelids with GADD45A () and GADD45B (MyD118), and four MCC samples. Each triangle/square represents data from one sample. constitutes a family of evolutionarily conserved, small, acidic, nuclear proteins, which have been implicated in terminal a leucine-zipper region with similarity to the DNA-binding differentiation, growth suppression, and apoptosis [19].In domain of the cAMP-responsive activating transcription factor response to stress shock, GADD45G inhibits cell growth and 5 [10]. Mutations in this gene have been found in primary induces apoptosis [18]. Vairapandi et al. [19] have shown that esophageal cancers and in a cell line [10]. Ishii GADD45B and GADD45G specifically interact with the Cdk1/ et al. [11] have shown that LZTS1 inhibits cancer cell growth cyclin B1 complex, but not with other Cdk/cyclin complexes, in through regulation of mitosis and that its alterations result in vitro and in vivo. Interaction of GADD45B, GADD45G, and abnormal cell growth. Reduced expression of this gene was GADD45A with both Cdk1 and cyclin B1 results in inhibition observed in 27/98 non-small cell cancers [12]. The LZTS1 of the kinase activity of the Cdk1/cyclin B1 complex [19]. gene is inactivated in many cancers with 8p deletions, including Inhibition of Cdk1/cyclin B1 kinase activity by GADD45B and prostate, esophageal, gastric, bladder, and breast cancers [12]. GADD45A was found to involve disruption of the complex, As expected, it was downregulated in MCC tumors (Fig. 1). whereas GADD45G did not disrupt the complex [19]. Using A. Kumar et al. / Genomics 90 (2007) 559–566 563

Fig. 2. Associations of differentially expressed genes with chromosomes 1, 12, and 19. Genes are displayed according to their genomic locations. Average log-ratios (fold changes) are written next to the genes.

RKO lung carcinoma cell lines, which express antisense epigenetically in multiple tumors. Chung et al. [20] analyzed GADD45 RNA, Vairapandi et al. [19] showed that all three the expression of GADD45G in anaplastic thyroid cancer cell GADD45 proteins are likely to cooperate in the activation of S lines to test the potential utility of the GADD45G gene as a and G2/M checkpoints following exposure of cells to UV candidate for gene therapy in anaplastic thyroid cancer, which irradiation. Similar to other members of the other GADD45 is a highly aggressive and extremely lethal form of human family, GADD45G is ubiquitously expressed in all normal adult cancer. They found that GADD45G RNA was present at and fetal tissues [18]. However, it shows frequent down- significantly lower levels in anaplastic cancer cells compared regulation and promoter hypermethylation in tumor cell lines, with normal primary cultured thyrocytes. The adenovirus- including 11 of 13 (85%) non-Hodgkin lymphoma, 3 of 6 mediated reexpression of GADD45G significantly inhibited (50%) Hodgkin lymphoma, 8 of 11 (73%) nasopharyngeal the proliferation of cells in the anaplastic thyroid carcinoma carcinoma, 2 of 4 (50%) cervical carcinoma, 5 of 17 (29%) cell lines due to apoptosis. They suggested that this gene could esophageal carcinoma, and 2 of 5 (40%) lung carcinoma and be used as a candidate for gene therapy in anaplastic thyroid other cell lines, but not in any immortalized normal epithelial cancer. As expected, we found this gene to be downregulated cell line, normal tissue, or peripheral blood mononuclear cells in MCC (Table 2). [18]. The transcriptional silencing of GADD45G is reversed by The ZAK gene is another member of the MAPK signaling 5-aza-2′-deoxycytidine treatment or genetic double knockout pathway. It encodes a protein of 800 amino acids that contains a of DNMT1 and DNMT3B, indicating a direct epigenetic kinase catalytic domain, a leucine-zipper, and a sterile-α motif mechanism [18]. Aberrant was frequently detected [21]. Liu et al. [21] found that ZAK activates JNK/SAPK1 in primary lymphomas although less frequently in primary (MAPL8) and NFKB, and overexpression of ZAK resulted in carcinomas [18]. GADD45G could be induced by heat shock apoptosis in a human hepatoma cell line. We found this gene to or UV irradiation in unmethylated cell lines [18]. Ying et al. be upregulated in MCC (Table 2). [18] showed that ectopic expression of GADD45G strongly TNFRSF21 (tumor necrosis factor receptor superfamily, suppressed tumor cell growth and colony formation in silenced member 21) or death receptor-6 (DR6) is a member of a cell lines, suggesting that GADD45G can act as a functional subgroup termed death receptors in the tumor necrosis factor new-age tumor suppressor, but is frequently inactivated receptor superfamily. It contains a cytoplasmic death domain. 564 A. Kumar et al. / Genomics 90 (2007) 559–566

Death receptors and their ligands are known to regulate shown to be involved in oncogenesis, since they bring about the programmed cell death. Currently, there are six different death activation of cyclin D1 and of c- and Bcl-xl expression and receptors: tumor necrosis factor (TNF) receptor-1, CD95 (Fas/ are involved in promoting cell-cycle progression and cellular APO-1), TNF receptor-related apoptosis-mediating protein transformation and in preventing apoptosis [30,31,33]. How- (TRAMP), TNF-related apoptosis-inducing ligand (TRAIL) ever, the status of STAT2 in oncogenesis is not clear. STAT1 is receptor-1 and -2, and death receptor-6 (DR6) [22]. The considered as a potential tumor suppressor, since it plays an signaling pathways by which these receptors induce apoptosis important role in growth arrest and in promoting apoptosis [31]. are similar and rely on oligomerization of the receptor by death Interestingly, the STAT2 gene was found to be downregulated in ligand binding, recruitment of an adapter protein through MCC (Table 2). homophilic interaction of cytoplasmic domains, and subsequent MCC is a rare eyelid tumor. This is the reason we have been activation of an inducer caspase, which initiates execution of the able to collect MCC samples from only four patients, with ages cell death program [22]. Exogenous expression of DR6 induced ranging from 42 to 70 years, over the past several years. In apoptosis in untransformed or tumor-derived cells [23]. DR6 addition, the collection of normal age-matched eyelid tissue potently activates JNK [24]. This gene was downregulated in samples as controls was extremely difficult. Therefore, we MCC (Table 2). directed our effort in collecting the normal eyelid tissues from The JAK/STAT pathway is an important signaling cascade normal individuals who visited the hospital for eyelid surgery that transduces a multitude of signals leading to cell prolifera- for ectropion. With our tremendous effort over the past several tion, differentiation, cell migration, and apoptosis [25]. In this years, two normal eyelid tissue samples were collected from study, we found several genes (viz. IL24, IL23A, and STAT2) two individuals, ages 50 and 65 years, who underwent eyelid belonging to this pathway to be differentially regulated in MCC. surgery for ectropion. We believe that these could be the reasons The IL24 gene of this pathway is located on 1q32.1 and was that no similar studies have been attempted previously. As we cloned using a differentiation induction subtraction hybridiza- have used age-matched normal and tumor tissues, validated the tion strategy to identify and clone genes involved in growth differential expression data for five genes using two normal and control and terminal differentiation in human cancer cells [26]. four tumor samples, and functionally categorized the differen- This gene is also called as the melanoma differentiation- tially expressed genes, we believe that the present data are a true associated gene-7 (MDA7) and the suppression of tumorigeni- representative of the genes whose expression differs between city 16 (ST16). Jiang et al. [26] showed that overexpression of normal and tumor tissues. IL24 was growth inhibitory toward diverse human tumor cells. In summary, our microarray data provide interesting insights Huang et al. [27] suggested that it has potential utility for the into several genes and pathways that are dysregulated in MCC gene-based therapy of diverse human cancers. Su et al. [28] for the first time. Functional analysis of these genes will lead to showed that it selectively induces apoptosis in human breast a better understanding of the development and progression of cancer cells and inhibits tumor growth in nude mice. This gene this tumor. Furthermore, the data presented here will not only was downregulated in MCC (Table 2). provide important information about the genesis of this tumor, Another interleukin gene, IL23A, encodes a 19-kDa cytokine, but will also facilitate the identification of candidate genes that p19. IL23A and IL12 p40 (IL12B) constitute heterodimeric could be used as therapeutic targets for the treatment of patients IL23. Cytokines are soluble proteins that regulate cell with this tumor. proliferation and differentiation and effector functions of immune cells. Langowski et al. [29] have recently demonstrated Materials and methods that expression of the heterodimeric cytokine IL23 is increased in human tumors. As expected, IL23A was upregulated in MCC Sample collection (Table 2). Signal transducer and activator of transcription proteins are A total of four MCC samples were ascertained from Minto Ophthalmic latent cytoplasmic transcription factors [30]. STATs can be Hospital, Bangalore. The age of the MCC patients ranged from 42 to 70 years. MCCs were b10 mm in size and located in upper eyelids. Regional lymph node divided into two groups according to their specific functions. metastasis was present in one of the MCC patients. MCC is known to be One group is made up of STAT2, STAT4, and STAT6, which are associated with Muir–Torre syndrome, but none of the patients in the present activated by a small number of cytokines and play a distinct role study were diagnosed with this syndrome. Two normal eyelid tissue samples in the development of T cells and in IFN-γ signaling [31]. The were collected from two individuals, ages 50 and 65 years, who underwent other group consists of STAT1, STAT3, and STAT5, which are eyelid surgery for ectropion. Tumor and normal tissue samples were obtained either as surgically resected tissue or as biopsy samples. Normal and tumor activated in different tissues by means of a series of ligands and samples were stored in liquid nitrogen following surgery and then in a −80 °C involved in IFN signaling [31]. STAT2 is a transcription factor freezer for long-term storage. This research followed the tenets of the critical to the signal transduction pathway of type I interferons Declaration of Helsinki and informed consent for research was obtained from such as IFN-α [32]. The cytoplasmic STAT2 is tyrosine all individuals enrolled in the study. phosphorylated after IFN-α binds to cell surface receptors. The tyrosine-phosphorylated STAT2 rapidly localizes to the Microarray analysis nucleus and acquires the ability to bind specific DNA targets in Total RNA was isolated from normal and tumor tissues using the Tri reagent association with two other proteins, STAT1 and IFN regulatory (Sigma–Aldrich, St. Louis, MO, USA) according to the manufacturer's factor-9 (IRF-9) [32]. STAT1, STAT3, and STAT5 have been instructions. Total RNA was further purified on RNeasy Protect minikit A. Kumar et al. / Genomics 90 (2007) 559–566 565 columns (Qiagen, Hilden, Germany). RNA quality was checked using a a PTC-100 thermocycler (MJ Research, Inc., Waltham, MA, USA). PCR formaldehyde agarose gel as described by Sambrook and Russell [34]. RNAwas conditions were as follows: an initial denaturation at 95 °C for 2 min was quantitated spectrophotometrically by measuring the absorbance at 260/280 nm. followed by 18–25 cycles of denaturation at 95° C for 30 s, annealing at 55– Human 8K cDNA microarray slides (University Health Network, Toronto, ON, 60 °C for 30 s, extension at 72° C for 1 min, with a final extension at 72°C for Canada) were used for expression profiling of MCC. Human 8K arrays contain 5 min. RT-PCR primer sets for the GTF2H4, RBM12, UBE2D3, DDX17, ∼8000 human genes. cDNA synthesis, labeling, hybridization, and detection of LZTS1, and GAPDH genes amplified 184-, 168-, 174-, 188-, 166-, and 458-bp signals were carried out using the Micromax TSA Labeling and Detection Kit PCR fragments, respectively. For each gene, the PCR protocol was optimized to (Perkin–Elmer Life Sciences, Boston, MA, USA) according to the manufac- obtain amplification in a linear range. Expression of GAPDH served as a turer's instructions. Due to the collection of very small tissue samples, which normalizing control. Following electrophoresis of RT-PCR products in 2% yielded minute quantities of RNA samples, and the rarity of the tumor and the agarose gels containing ethidium bromide, images of gels were acquired with a availability of only two normal eyelid tissue samples, we were able to perform Kodak CCD camera and quantification of the band intensity was performed by only two microarray hybridizations. Each microarray hybridization experiment densitometric analysis using the Kodak Digital Science Image Station Imaging was carried out using RNA samples from one normal and one tumor sample. In software version 3.6.1. Final data were expressed in arbitrary units (relative this way, we used two different normal and two different tumor samples for two expression) as a ratio of normal/GAPDH and tumor/GAPDH and plotted using hybridization experiments. After signal amplification, hybridized microarray the GraphPad Prism software version 3.00 (GraphPad Prism software, San slides were dried and scanned using the GenePix Personal Scanner 4100A Diego, CA, USA). The significance of differences in mRNA expression levels (Molecular Devices, Union City, CA, USA). Image analysis and data extraction between normal and tumor samples for each gene was assessed by two-tailed were carried out using the GenePix Pro 3.0 program (Molecular Devices). Raw unpaired Student's t test and the results were expressed as means±SEM [37].A data were normalized by grid-wise normalization [35]. Normalization was done probability value of p b 0.05 was considered significant. Semiquantitative RT- using the following formulae [36]. For each grid, calculation of normalization PCR for each gene was performed in duplicate. factor=(sum of channel 1 intensity)/(sum of channel 2 intensity) (excluding control and non-array-specific spots), calculation of raw ratio for each spot= (channel 1 intensity)/(channel 2 intensity), and calculation of normalization ratio Acknowledgments for each spot=(raw ratio)/(normalization factor). Likewise, ratio was calculated for spots in each grid separately [36]. Genes were considered to be upregulated We are very grateful to the patients and normal individuals ≥ when the mean of the normalized ratios was 2. Genes were considered to be for taking part in this study. We also thank Mr. Aiyaz Mohamed, downregulated when the mean of the normalized ratios was ≤0.5. Genes were selected as significantly altered if they showed a similar pattern of expression Mr. V. Madavan, and Dr. S. Khare for technical help. This work across two different experiments. For example, an upregulated gene showed was financially supported by a grant from the ICMR, New upregulation in both microarray experiments. A downregulated gene showed Delhi. We also thank three anonymous reviewers and the downregulation in both microarray experiments. Differentially regulated genes associate editor, Professor Sandrine Dudoit, for their valuable were taken along with the fold change values across the experiments. The scatter suggestions to improve the manuscript. plot of the log-ratios of experiment 1 and experiment 2 was generated using the Microsoft Excel program to evaluate the performance and the concordance of the two hybridization experiments (Supplemental Fig. 1S). Biological Appendix A. Supplementary data significance of differentials was computed and functionally classified using the GeneSpring GX software (Agilent Technologies, USA) on the basis of gene Supplementary data associated with this article can be found, ontology. The pathways were obtained using enrichment analysis based on gene in the online version, at doi:10.1016/j.ygeno.2007.07.008. ontology categories using the GeneSpring GX software. GeneSpring GX software depicts the biologically significant ontology for any given gene list as follows: number of genes in the array known to be present in the ontology References category vs number of genes in the array that are differentially regulated. [1] K.L. Krenzer, M.R. Dana, M.D. Ullman, J.M. Cermak, D.B. Tolls, J.E. Semiquantitative RT-PCR Evans, D.A. Sullivan, Effect of androgen deficiency on the human meibomian gland and ocular surface, J. Clin. Endocrinol. Metab. 85 (2000) We used semiquantitative RT-PCR analysis to validate the microarray data 4874–4882. for five genes in tumor samples using first-strand cDNA samples as templates. [2] J.M. Tiffany, Physiological functions of the meibomian glands, Prog. First-strand cDNAwas synthesized with 1 μg total RNA from each sample using Retinal. Eye Res. 14 (1995) 47–74. the Revertaid H Minus First Strand cDNA Synthesis kit (MBI Fermentas, [3] J.A. Shields, H. Demirci, B.P. Marr, R.C. Eagle, C.L. Shields, Sebaceous Burlington, ON, Canada) according to the manufacturer's instructions. RT-PCR carcinoma of the ocular region: a review, Surv. Ophthalmol. 50 (2005) primer sets were designed for each gene using gene sequences from the 103–122. GenBank database. Forward and reverse primers were selected from two [4] M.W. Swanson, G. Cloud, A retrospective analysis of primary eye cancer different exons of a gene to rule out the possibility of amplification of at the University of Alabama 1958–88. Part 2. Eyelid tumors, J. Am. contaminating genomic DNA. The GenBank accession numbers for the Optom. Assoc. 62 (1991) 820–823. GTF2H4, RBM12, UBE2D3, DDX17, LZTS1, and GAPDH genes are [5] J. Bryant, Meibomian gland carcinoma seeding intracranial soft tissues, BC004935, AJ289772, U39318, AK024985, BC075006, and AF261085, Hum. Pathol. 8 (1977) 455–457. respectively. Primer sequences for these genes are as follows: GTF2H4, F, 5′- [6] C. Stover, G. Gradl, I. Jentsch, M.R. Speicher, R. Wieser, W. Schwaeble, GGACTGAACCGAGTACACCTACAATG-3′,R,5′-CCTGTGGCAAAGGC- cDNA cloning, chromosome assignment, and genomic structure of a TGCTCCAGAA-3′; RBM12, F, 5′-AGGATTGACTTGGCCTCTACTTCTTG- human gene encoding a novel member of the RBM family, Cytogenet. Cell 3′,R,5′-CATCAGGAATGGTCAATCCAGAGAAG-3′; UBE2D3, F, Genet. 92 (2001) 225–230. 5′-TGACAGCCCATATCAAGGCGGTGTA-3′,R,5′-ATTGTTAAAG- [7] X.-B. Qiu, S.L. Markant, J. Yuan, A.L. Goldberg, Nrdp1-mediated CAGGCGACCACTGTGA-3′; DDX17, F, 5′-GTGTTGAGATCTGCATAGC- degradation of the gigantic IAP, BRUCE, is a novel pathway for triggering CACTCC-3′,R,5′-ACTCCACATCAGTGTCTGCCTATCA-3′; LZTS1, F, apoptosis, EMBO J. 23 (2004) 800–810. 5′-TCATGCCCTTCTCCAATCAGCTAGA-3′,R,5′-TCCTGCTCCTTGG- [8] C. Jacques, O. Baris, D. Prunier-Mirebeau, F. Savagner, P. Rodien, V. GCTTGTCTGGA-3′; and GAPDH, F, 5′-GTCATCCATGACAACTTTG-3′, Rohmer, Two-step differential expression analysis reveals a new set of R, 5′-GTCATACCAGGAAATGAGC-3′. PCR was carried out in a total volume genes involved in thyroid oncocytic tumors, J. Clin. Endocrinol. Metab. 90 of 25 μl containing 1 μl of first-strand cDNA template, 1.5 mM MgCl2, 200 μM (2005) 2314–2320. each dNTP, 1× buffer, and 1 U of Taq DNA polymerase (Sigma–Aldrich) using [9] B.J. Wilson, G.L. Bates, S.M. Nicol, D.J. Gregory, N.D. Perkins, F.V. 566 A. Kumar et al. / Genomics 90 (2007) 559–566

Fuller-Pace, The p68 and p72 DEAD box RNA helicases interact with death: promises and failures of death receptors and ligands in cancer HDAC1 and repress transcription in a promoter-specific manner, BMC therapy, Leukemia 15 (2001) 1022–1032. Mol. Biol. 5 (2004) 11. [23] G.M. Kasof, J.J. Lu, D. Liu, B. Speer, K.N. Mongan, B.C. Gomes, M.V. [10] H. Ishii, R. Baffa, S.I. Numata, Y. Murakumo, S. Rattan, H. Inoue, M. Lorenzi, Tumor necrosis factor-alpha induces the expression of DR6, a Mori, V. Fidanza, H. Alder, C.M. Croce, The FEZ1 gene at chromosome member of the TNF receptor family, through activation of NF-kappaB, 8p22 encodes a leucine-zipper protein, and its expression is altered in Oncogene 20 (2001) 7965–7975. multiple human tumors, Proc. Natl. Acad. Sci. U. S. A. 96 (1999) [24] H. Zhao, M. Yan, H. Wang, S. Erickson, I.S. Grewal, V.M. Dixit, Impaired 3928–3933. c-Jun amino terminal kinase activity and differentiation in death [11] H. Ishii, A. Vecchione, Y. Murakumo, G. Baldassarre, S. Numata, F. receptor 6-deficient mice, J. Exp. Med. 194 (2001) 1441–1448. Trapasso, H. Alder, R. Baffa, C.M. Croce, FEZ1/LZTS1 gene at 8p22 [25] J.S. Rawlings, K.M. Rosler, D.A. Harrison, The JAK/STAT signaling suppresses cancer cell growth and regulates mitosis, Proc. Natl. Acad. Sci. pathway, J. Cell. Sci. 117 (2004) 1281–1283. U. S. A. 98 (2001) 10374–10379. [26] H. Jiang, J.J. Lin, Z.-Z. Su, N.I. Goldstein, P.B. Fisher, Subtraction [12] D. Nonaka, A. Fabbri, L. Roz, L. Mariani, A. Vecchione, G.W. Moore, L. hybridization identifies a novel melanoma differentiation associated gene, Tavecchio, C.M. Croce, G. Sozzi, Reduced FEZ1/LZTS1 expression and mda-7, modulated during human melanoma differentiation, growth and outcome prediction in lung cancer, Cancer Res. 65 (2005) 1207–1212. progression, Oncogene 11 (1995) 2477–2486. [13] A. Cromer, A. Carles, R. Millon, G. Ganguli, F. Chalmel, F. Lemaire, J. [27] E.Y. Huang, M.T. Madireddi, R.V. Gopalkrishnan, M. Leszczyniecka, Z. Young, D. Dembele, C. Thibault, D. Muller, O. Poch, J. Abecassis, B. Su, I.V. Lebedeva, D. Kang, H. Jiang, J.J. Lin, D. Alexandre, Y. Chen, N. Wasylyk, Identification of genes associated with tumorigenesis and Vozhilla, M.X. Mei, K.A. Christiansen, F. Sivo, N.I. Goldstein, A.B. metastatic potential of hypopharyngeal cancer by microarray analysis, Mhashilkar, S. Chada, E. Huberman, S. Pestka, P.B. Fisher, Genomic Oncogene 23 (2004) 2484–2498. structure, chromosomal localization and expression profile of a novel [14] J. Huang, G. Fan, Y. Zhong, K. Gatter, R. Braziel, G. Gross, A. Bakke, melanoma differentiation associated (mda-7) gene with cancer specific Diagnostic usefulness of aberrant CD22 expression in differentiating growth suppressing and apoptosis inducing properties, Oncogene 20 neoplastic cells of B-cell chronic lymphoproliferative disorders from (2001) 7051–7063. admixed benign B cells in four-color multiparameter flow cytometry, Am. [28] Z. Su, M.T. Madireddi, J.J. Lin, C.S.H. Young, S. Kitada, J.C. Reed, N.I. J. Clin. Pathol. 123 (2005) 826–832. Goldstein, P.B. Fisher, The cancer growth suppressor gene mda-7 [15] N. Moniaux, C. Nemos, B.M. Schmied, S.C. Chauhan, S. Deb, K. selectively induces apoptosis in human breast cancer cells and inhibits Morikane, A. Choudhury, M. Vanlith, M. Sutherlin, J.M. Sikela, M.A. tumor growth in nude mice, Proc. Natl. Acad. Sci. U. S. A. 95 (1998) Hollingsworth, S.K. Batra, The human homologue of the RNA polymerase 14400–14405. II-associated factor 1 (hPaf1), localized on the 19q13 amplicon, is [29] J.L. Langowski, X. Zhang, L. Wu, J.D. Mattson, T. Chen, K. Smith, B. associated with tumorigenesis, Oncogene 25 (2006) 3247–3257. Basham, T. McClanahan, R.A. Kastelein, M. Oft, IL-23 promotes tumour [16] A. Beghini, I. Magnani, G. Roversi, T. Piepoli, S. Di Terlizzi, R.F. Moroni, incidence and growth, Nature 442 (2006) 461–465. B. Pollo, A.M. Fuhrman Conti, J.K. Cowell, G. Finocchiaro, L. Larizza, [30] J. Turkson, STAT proteins as novel targets for cancer drug discovery, The neural progenitor-restricted isoform of the MARK4 gene in 19q13.2 is Expert. Opin. Ther. Targets 8 (2004) 409–422. upregulated in human gliomas and overexpressed in a subset of [31] V. Calo, M. Migliavacca, V. Bazan, M. Macaluso, M. Buscemi, N. Gebbia, glioblastoma cell lines, Oncogene 22 (2003) 2581–2591. A. Russo, STAT proteins: from normal control of cellular events to [17] R. Seger, E.G. Krebs, The MAPK signaling cascade, FASEB J. 9 (1995) tumorigenesis, J. Cell. Physiol. 197 (2003) 157–168. 726–735. [32] G. Banninger, N.C. Reich, STAT2 nuclear trafficking, J. Biol. Chem. 279 [18] J. Ying, G. Srivastava, W.S. Hsieh, Z. Gao, P. Murray, S.K. Liao, R. (2004) 39199–391206. Ambinder, Q. Tao, The stress-responsive gene GADD45G is a functional [33] R. Buettner, L.B. Mora, R. Jove, Activated STAT signaling in human tumor suppressor, with its response to environmental stresses frequently tumors provides novel molecular targets for therapeutic intervention, Clin. disrupted epigenetically in multiple tumors, Clin. Cancer Res. 11 (2005) Cancer Res. 8 (2002) 945–954. 6442–6449. [34] J. Sambrook, D. Russell, Molecular Cloning: A Laboratory Manual, Cold [19] M. Vairapandi, A.G. Balliet, B. Hoffman, D.A. Liebermann, GADD45b Spring Harbor Laboratory Press, New York, 2000. and GADD45g are cdc2/cyclinB1 kinase inhibitors with a role in S and [35] Y.H. Yang, S. Dudoit, P. Luu, D.M. Lin, V. Peng, J. Ngai, T.P. Speed, G2/M cell cycle checkpoints induced by genotoxic stress, J. Cell. Physiol. Normalization for cDNA microarray data: a robust composite method 192 (2002) 327–338. addressing single and multiple slide systematic variation, Nucleic Acids [20] H.K. Chung, Y.-W. Yi, N.-C. Jung, D. Kim, J.M. Suh, H. Kim, K.C. Park, Res. 30 (2002) e15. D.W. Kim, E.S. Hwang, J.H. Song, B.-J. Ku, H.J. Han, H.K. Ro, J.M. Kim, [36] S. Chakraborty, S. Khare, S.K. Dorairaj, V.C. Prabhakaran, D.R. Prakash, M. Shong, Gadd45-gamma expression is reduced in anaplastic thyroid A. Kumar, Identification of genes associated with tumorigenesis of cancer and its reexpression results in apoptosis, J. Clin. Endocrinol. Metab. retinoblastoma by microarray analysis, Genomics 90 (2007) 344–353. 88 (2003) 3913–3920. [37] C.L. Estilo, P. O-Charoenrat, I. Ngai, S.G. Patel, P.G. Reddy, S. Dao, A.R. [21] T.-C. Liu, C.-J. Huang, Y.-C. Chu, C.-C. Wei, C.-C. Chou, M.-Y. Chou, Shaha, D.H. Kraus, J.O. Boyle, R.J. Wong, D.G. Pfister, J.M. Huryn, I.M. C.-K. Chou, J.-J. Yang, Cloning and expression of ZAK, a mixed lineage Zlotolow, J.P. Shah, B. Singh, The role of novel oncogenes squamous cell kinase-like protein containing a leucine-zipper and a sterile-α motif, carcinoma-related oncogene and phosphatidylinositol 3-kinase p110 alpha Biochem. Biophys. Res. Commun. 274 (2000) 811–816. in squamous cell carcinoma of the oral tongue, Clin. Cancer Res. 9 (2003) [22] P.T. Daniel, T. Wieder, I. Sturm, K. Schulze-Osthoff, The kiss of 2300–2306.