Research Article

Zonal Heterogeneity for Expression in Human Pancreatic Carcinoma Toru Nakamura,1 Toshio Kuwai,1 Yasuhiko Kitadai,1 Takamitsu Sasaki,1 Dominic Fan,1 Kevin R. Coombes,2 Sun-Jin Kim,1 and Isaiah J. Fidler1

Departments of 1Cancer Biology and 2Biostatistics and Applied Mathematics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas

Abstract have been undertaken to predict disease outcome for individual patients (5–12) and to identify cancer patients who should receive Using Affymetrix HG-U133 Plus 2.0 array and laser capture microdissection techniques,we determined whether different chemotherapy (13). zones of the same pancreatic tumor exhibited differential Pretherapeutic gene expression profiling has been undertaken to expression of . Human L3.6pl pancreatic cancer cells were identify breast cancer patients who should receive chemotherapy implanted into the pancreas of nude mice. Three weeks later (13), to predict response to preoperative chemoradiotherapy in when tumors were 7 to 9 mm in diameter,gene expression rectal carcinoma patients (14), and to predict response to docetaxel patterns in tumor cells within the central and peripheral zones in breast cancer patients (15). A series of studies suggested that were compared,and 1,222genes showed statistically signifi- pretherapeutic gene expression profiles might predict the response cant differences. Bioinformatic functional analysis revealed to the treatment in various other tumors (10–16). For example, a that 346 up-regulated genes in the peripheral zone were specific gene expression pattern was correlated with recurrence in Dukes’ B colon carcinomas (16). None of these studies, however, related to cytoskeleton organization and biogenesis,cell cycle, cell adhesion,cell motility,DNA replication,localization, accounted for the biological heterogeneity of gene expression integrin-mediated signaling pathway,development,morpho- within a tumor. Tumor cells depend on multiple and redundant genesis,and I KB kinase/nuclear factor-KB cascade; 876 up- pathways for growth, survival, and adaptation to the host regulated genes in the central zone were related to regulation microenvironment (3, 17–19). Genetic instability of cancer cells of cell proliferation,regulation of transcription,transmem- frequently yields extensive intratumoral heterogeneity in the brane receptor tyrosine kinase signaling pathways, pattern of structural aberrations in highly malignant response to stress,small GTPase-mediated signal transduc- neoplasms (20–22). Three-dimensional tumor growth and uneven tion,hexose metabolism,cell death,response to external fractional dividing of tumor cells within a tumor mass cause stimulus,carbohydrate metabolism,and response to wound- biologically different zones, whether central or peripheral, within a ing. The reliability of the microarray results were confirmed single neoplasm. The microenvironment of the central zone differs by in situ hybridization analysis of the expression of two significantly from that of the peripheral zone, and zonal genes. Collectively,the data showed zonal heterogeneity for heterogeneity for several molecules in different tumor systems has been reported (23–30). Among these, basic fibroblast growth gene expression profiles in tumors and suggest that charac- terization of zonal gene expression profiles is essential if factor, matrix metalloproteinase (MMP)-2, and MMP-9 are highly microarray analyses of genetic profiles are to produce repro- expressed at the invasive edge of tumors (23–27), whereas the cell- ducible data,predict disease prognosis,and allow design of to-cell cohesion molecule, E-cadherin, was down-regulated at the periphery of the tumors (24–28). In fact, the ratio of expression of specific therapeutics. [Cancer Res 2007;67(16):7597–604] MMP-2 and MMP-9 to E-cadherin (MMP/E-cadherin ratio) at the periphery of the tumors correlated with metastatic potential and Introduction recurrent disease (23–27). The genetic instability of neoplastic cells and the consequent To provide reproducible data among different tumor specimens, biological heterogeneity of individual tumors has been well clinical or experimental, it may be necessary to standardize the documented (1). Tumor growth, prognosis, and metastasis are area zone of each specimen to be studied. The purpose of this study dependent on multiple interactions of tumor cells with homeo- was therefore to compare gene expression profiles of pancreatic static factors in the microenvironment (1–4). A better understand- cancer cells growing in the peripheral zone with expression in cells ing of the cross-talk between cancer cells and the organ in the central zone of the same tumor and to identify any zonal microenvironment involving multiple genes requires in the first heterogeneity in a genome-wide gene expression analysis. We instance identification of gene expression profiles within the compared gene expression profiles generated from microarray tumor. Among current methods for establishing these profiles, analysis of cancer cells in the peripheral and central zones of the microarray analysis is the most powerful, and several such studies same tumor by laser capture microdissection (LCM) and determined whether the resulting gene ontogeny of pancreatic cancer cells in the two zones correlated with presumptive functions Note: Supplementary data for this article are available at Cancer Research Online of cells at the periphery and the center of tumors. (http://cancerres.aacrjournals.org/). Requests for reprints: Isaiah J. Fidler, Department of Cancer Biology, The University of Texas M. D. Anderson Cancer Center, Unit 173, P. O. Box 301429, Houston, TX 77230-1429. Phone: 713-792-8580; Fax: 713-792-8747; E-mail: ifidler@ Materials and Methods mdanderson.org. I2007 American Association for Cancer Research. Human pancreatic cancer cell line. The human pancreatic cancer cells doi:10.1158/0008-5472.CAN-07-0874 from line L3.6pl (31) were maintained in Eagle’s minimal essential medium www.aacrjournals.org 7597 Cancer Res 2007; 67: (16). August 15, 2007

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Cancer Research supplemented with 10% fetal bovine serum (FBS), sodium pyruvate, (1 Â 106/50 AL of HBSS) were injected under the capsule. The abdominal nonessential amino acids, L-glutamine, a 2-fold vitamin solution (Life wound was closed in one layer with wound clips (Autoclip, Clay Adams). Technologies, Inc.), and a penicillin-streptomycin mixture (Flow Laborato- The animals tolerated the surgical procedure well, and no procedure- ries). Adherent monolayer cultures were maintained on plastic and related deaths occurred. incubated at 37jC in a mixture of 5% CO2 and 95% air. The cultures were After 3 weeks, the pancreatic tumors reached 7 to 9 mm in diameter. At free of Mycoplasma and the following pathogenic murine viruses: reovirus this time, specimens were harvested, immediately embedded in ornithine type 3, pneumonia virus, K virus, Theiler’s encephalitis virus, Sendai virus, carbamyl transferase compound (Miles), rapidly frozen in liquid nitrogen, minute virus, mouse adenovirus, mouse hepatitis virus, lymphocytic and stored at À80jC. choriomeningitis virus, ectromelia virus, and lactate dehydrogenase virus LCM and RNA preparation. The frozen tissues from three different (assayed by M.A. Bioproducts). The cultures were maintained for no longer pancreatic tumors growing in three different mice were cut into 8-Am than 12 weeks after recovery from frozen stocks. sections and stained with H&E for histologic examination. LCM was Animals. Male athymic nude mice (NCI-nu) were obtained from the done by using Pixcel II LM200 (Arcturus). A total of 180 frozen tissue Animal Production Area of the National Cancer Institute Frederick Cancer sections were prepared from three independent pancreatic tumor samples Research and Development Center (Frederick, MD). The mice were (60 sections each), and f20,000 tumor cells were captured from either the maintained under specific pathogen-free conditions in facilities approved central or peripheral zones of each sample. The sample for the peripheral by the American Association for Accreditation of Laboratory Animal Care zone was taken from the edge, and the sample for the central zone was and in accordance with current regulations and standards of the U.S. taken at the center, at least 3.5 mm from the tumor edge. Total RNA was Department of Agriculture, Department of Health and Human Services, and extracted using an RNeasy mini kit (Qiagen) with on-column DNase I NIH. The mice were used in accordance with institutional guidelines when treatment done according to the manufacturer’s protocol. they were 8 to 12 weeks old. Affymetrix GeneChip hybridization. The Affymetrix Orthotopic pancreatic cancer model. To produce pancreatic tumors, U133 Plus 2.0 GeneChip arrays were used for microarray hybridizations. L3.6pl cells were harvested from subconfluent cultures by a brief exposure This GeneChip carries 54,675 probe sets. For microarray hybridization, we to 0.25% trypsin and 0.02% EDTA. Trypsinization was stopped with medium followed the protocol described in the Affymetrix GeneChip eukaryotic one- containing 10% FBS, and then the cells were washed once in serum-free cycle target preparation protocol (Affymetrix). In short, 200 ng of total RNA medium and resuspended in HBSS. Only single-cell suspensions of >90% were used to prepare antisense RNA. A single-stranded cDNA was viability (trypan blue exclusion) were used for injection into the pancreas. synthesized using a T7-oligo(dT) promoter primer followed by RNase The mice were anesthetized by i.m. injection of Nembutal (0.5 mg/kg). H-facilitated second-strand cDNA synthesis. The cDNA was purified and A small left abdominal flank incision was made, the spleen was exter- served as a template for the subsequent in vitro transcription. The second iorized, and the pancreas was identified in a region just beneath the in vitro transcription reaction was carried out in the presence of T7 RNA spleen. A 30-gauge needle was inserted into the pancreas and tumor cells polymerase and a biotinylated nucleotide analogue/ribonucleotide mix for

Figure 1. Mice were killed when pancreatic tumors reached 7 to 9 mm in diameter. The specimens were processed for H&E staining (A). Under the microscope, viable cancer cells were collected from the peripheral zone (B) and central zone (C) of the tumor mass by LCM.

Cancer Res 2007; 67: (16). August 15, 2007 7598 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Zonal Heterogeneity for Gene Expression

Table 1. analysis of gene expression patterns in pancreatic tumor cells from central versus peripheral zone

Zone GO category term No. genes P

Peripheral Cytoskeleton organization and biogenesis 20 2.7e-05 Cell cycle 28 3.4e-05 Cell adhesion 22 0.0031 Cell motility 10 0.015 DNA replication 9 0.017 Localization 63 0.026 Integrin-mediated signaling pathway 5 0.029 Development 41 0.030 Morphogenesis 17 0.033 InB kinase/NF-nB cascade 6 0.039 Central Regulation of cell proliferation 28 6.8e-06 Regulation of transcription 112 2.3e-04 Transmembrane receptor PTK signaling pathway 13 0.0025 Response to stress 57 0.0029 Small GTPase-mediated signal transduction 19 0.0093 Hexose metabolism 13 0.012 Cell death 32 0.013 Response to external stimulus 29 0.015 Carbohydrate metabolism 29 0.017 Response to wounding 23 0.023 Response to hypoxia 3 0.062

NOTE: Gene ontology was based on the DAVID database. Abbreviations: GO, gene ontology; PTK, protein tyrosine kinase.

cRNA. The biotinylated cRNA targets were then cleaned up and fragmented. specific probes for each gene were as follows: 5¶-CTTGCACCTCCC- The fragmented cRNA was used for hybridization to the U133 Plus 2.0 chip GACTTCCTTTCC-3¶ (AMIGO2)and5¶-GACAGACACGGGCTCG- at 42jC for 16 h. The chips were washed and stained using Affymetrix GAGCGGGC-3¶ (RARRES1). GeneChip Fluidic and scanned and visualized using a GeneArray scanner The preservation of mRNA in the samples was confirmed by the (Hewlett-Packard). hybridization of the samples with a poly(dT)24 probe. The specificity of the Preparation of samples for fluorescence in situ hybridization. hybridization signal was confirmed by the negative controls pretreated with Frozen sections (4 Am) were cut on silane-treated ProbeOn slides. The RNase or hybridized with a sense probe instead of the antisense probe. slides were fixed for 15 min with 4% paraformaldehyde in PBS and then Negative controls yielded no signal or a markedly decreased signal. Water rinsed four times for 1 min each time with 50 mmol/L Tris buffer (pH 7.6). treated with diethyl pyrocarbonate (Sigma) was used in the preparation of The tumor samples were then hybridized as described below. all solutions to preserve RNAs. Fluorescence in situ hybridization. Fluorescence in situ hybridization Image analysis to quantify intensity of FISH. For quantification of the (FISH) was done as described previously (32–34) with modification. The FISH reaction intensity, the absorbance of AMIGO2 and RARRES1 from FISH analysis was carried out by using the microprobe manual staining three of different orthotopic human pancreatic tumors were measured at system (Fisher Scientific; ref. 35). The samples were hybridized with the Â200 magnification using the Optimas image analysis software (version 5.2; probes at 45jC for 45 min, washed for 2 min with 5Â SSC three times at refs. 25, 31). For each section, we determined the absorbance in 0.5 Â 45jC, and then washed with PBS for 1 min three times at room 0.5–mm zones located at the center and periphery of the tumors. We temperature. Nuclei were stained with Hoechst. Samples were then rinsed quantified the fluorescence intensity in six different fields in the peripheral three times with PBS and covered with a glass coverslip (Fisher Scientific). region and the central region of each of the three tumors. The samples for Mounting medium consisted of 90% glycerol, 10% PBS, and 0.1 mol/L propyl quantification were not nuclei stained, so that the absorbance was gallate. attributable solely to the product of the FISH reaction. The intensity of Oligonucleotide probes. Specific antisense DNA oligonucleotide probes staining was standardized to that of the integrated absorbance of were designed to consist of sequences complementary to the mRNA poly(dT)24. To analyze these data, we used a linear mixed effects model transcripts based on the UniGene database of the National Center for with location as a fixed effect and mouse as a random effect to account for Biotechnology Information (NCBI). The oligonucleotide sequence was the repeated measurements. (A more general model with an additional initially confirmed with the NCBI basic local alignment search tool. All random effect for location within the orthotopic tumor was tested by the oligonucleotide sequences showed 100% homology with the target gene Akaike Information Criterion and found to be equivalent to the simpler sequences, and any homology with nonspecific mammalian gene sequences model.) Mixed effects models were fit by maximum likelihood using the was minimal. All DNA oligonucleotide probes, fluorescence labeled (Alexa nlme package in the R statistical programming environment; significance 594 or Alexa 488) on the 5¶-prime end, were commercially synthesized by was assessed by an F test. Invitrogen custom DNA oligonucleotides. The probes were reconstituted Statistical analysis. Microarray data were loaded into the DNA chip to a stock concentration of 1 Ag/AL with TE buffer [10 mmol/L Tris base, analyzer (dChip 2006)3 for quality control and analysis. Data were 1 mmol/L EDTA (pH 7.5)]. The final dilution of probes in the reac- tion mixture was 1:1,000 for polydeoxythymidylic acid [poly(dT)] 24 and 1:400 for adhesion molecule with Ig-like domain 2 (AMIGO2) and retinoic acid receptor responder (tazarotene-induced) 1 (RARRES1). The sequences of 3 http://biosun1.harvard.edu/complab/dchip www.aacrjournals.org 7599 Cancer Res 2007; 67: (16). August 15, 2007

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Cancer Research

Table 2. Top 20 up-regulated genes in peripheral zone and central zone of human pancreatic cancer growing in the pancreas of nude mice

Zone Probe set Gene Fold difference

Peripheral 216405_at Lectin, galactoside-binding, soluble, 1 (galectin 1) 8.83 201852_x_at Collagen, type III, a1 (Ehlers-Danlos syndrome type IV) autosomal dominant 8.04 214416_at Transcribed , strongly similar to NP_033799.1 amylase 2, pancreatic [Mus musculus] 6.55 215076_s_at Collagen, type III, a1 (Ehlers-Danlos syndrome type IV, autosomal dominant) 6.45 217683_at ESTs moderately similar to B chain B, crystal structure of deoxy-human hemoglobin b6 6.26 222108_at Adhesion molecule with Ig-like domain 2 4.84 216442_x_at Fibronectin 1 4.39 204619_s_at Chondroitin sulfate proteoglycan 2 (versican) 4.34 209312_x_at MHC, class II, DRb1 4.29 212233_at Microtubule-associated protein 1B 4.28 211719_x_at Fibronectin 1 4.26 226084_at Microtubule-associated protein 1B 4.1 239176_at Nebulin-related anchoring protein 4.1 210926_at Similar to FKSG30 4.07 210495_x_at Fibronectin 1 3.8 224833_at v-ets erythroblastosis virus E26 oncogene homolog (avian) 3.78 226535_at FLJ23083 fis, cl. LNG06541, highly similar to IR2005735 3.7 Homo sapiens mRNA full-length insert cDNA clone EUROIMAGE 2005735 239430_at Insulin growth factor-like family member 1 3.63 233149_at cDNA clone IMAGE:4750272, containing Frame-shift errors 3.57 1552411_at Defensin, b106 3.55 Central 212977_at Chemokine orphan receptor 1 13.7 209395_at Chitinase 3-like 1 (cartilage glycoprotein-39) 9.75 209396_s_at Chitinase 3-like 1 (cartilage glycoprotein-39) 8.14 206392_s_at Retinoic acid receptor responder (tazarotene-induced) 1 7.86 212143_s_at Insulin-like growth factor binding protein 3 6.84 202917_s_at S100 calcium binding protein A8 (calgranulin A) 6.08 223278_at Gap junction protein, b2, 26-kDa (connexin 26) 5.98 206697_s_at Haptoglobin 5.83 218960_at Transmembrane protease, serine 4 5.75 217109_at Mucin 4, tracheobronchial 5.74 205476_at Chemokine (C-C motif) ligand 20 5.61 209821_at Chromosome 9 open reading frame 26 (NF-HEV) 5.53 221872_at Retinoic acid receptor responder (tazarotene-induced) 1 5.34 206391_at Retinoic acid receptor responder (tazarotene-induced) 1 5.26 203921_at Carbohydrate (N-acetylglucosamine-6-O) sulfotransferase 2 4.64 202948_at Interleukin-1 receptor, type 1 4.6 213432_at Mucin 5, subtype B, tracheobronchial 4.28 204637_at Glycoprotein hormones, a polypeptide 4.23 222830_at Transcription factor CP2-like 2 4.13 213693_s_at Mucin 1, transmembrane 4.06 242868_at Hypoxia-inducible factor 2a 2.06 202986_at Aryl-hydrocarbon receptor nuclear translocator 2 1.82 202221_s_at E1A binding protein p300 1.42

normalized to a sample of median brightness using the invariant set discovery (DAVID).4 We used the Functional Annotation Tool program method and then quantified using the perfect match–only model-based and reported only GOTERM-BP (Biological Process) that had corrected expression indices (36). All arrays passed quality control checks based on P values of <0.05. percentage present calls (>30%) and outliers (<5%). To identify differentially expressed genes, the confidence interval estimates produced by dChip were Results used. Specifically, the criteria were that the lower bound of the 90% confidence interval on the fold change should be at least 1.2-fold, and the Microarray analyses. We did six microarray experiments using difference in expression should be at least 100. human U133 Plus 2.0 GeneChip microarrays that contained 54,675 Gene ontology analysis. Up-regulated genes in the central and probe sets. Samples of human pancreatic cancer cells growing in peripheral zones of each tumor were analyzed for the functional the pancreas of nude mice were microdissected to obtain tumor categories using the database for annotation, visualization, and integrated cells from the central and peripheral zones of the tumor. Because necrosis, especially in the central zone of a growing tumor, could affect gene expression, nonnecrotic tissues were carefully selected 4 http://niaid.abcc.ncifcrf.gov/ under the microscope and dissected by the LCM (Fig. 1).

Cancer Res 2007; 67: (16). August 15, 2007 7600 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Zonal Heterogeneity for Gene Expression

A different level of expression between the central and peripheral We examined AMIGO2 and RARRES1 because AMIGO2 was highly zones was identified for 1,222 genes, of which 346 were up- expressed in the peripheral zone and RARRES1 was highly regulated in the peripheral zone and 876 were up-regulated in the expressed in the central zone according to microarray analysis central zone (Supplementary Table S1). To identify statistically (see Table 2). The staining intensity with AMIGO2 probes was significant functional categories by gene ontology, the up-regulated strong in the peripheral zone of the tumor (Fig. 2B) but weak in the genes were placed into DAVID (Table 1). Many categories with center of the tumor (Fig. 2D) and vice versa with RARRES1 probes statistically significant differences were identified. The patterns (Fig. 3B and D). A poly(dT) probe was stained as a control for up-regulated in the peripheral zone were enriched for cytoskeleton mRNA (Fig. 2C and E;Fig.3C and E). We quantified the FISH data organization and biogenesis, cell cycle, cell adhesion, cell motility, by measuring the fluorescence intensity in six different fields in DNA replication, localization, integrin-mediated signaling pathway, both the peripheral regions and the central regions of three development, morphogenesis, and InB kinase/nuclear factor-nB different orthotopic pancreatic tumors. The quantification of the (NF-nB) cascade (Table 1). The patterns up-regulated in the central FISH reaction intensity was summarized in Table 3. For the zone were enriched for regulation of cell proliferation, regulation of RARRES1 gene, the intensity was significantly lower in the peri- transcription, transmembrane receptor protein tyrosine kinase pheral zone by 51.8 units (F = 914; P < 0.0001). For the AMIGO2 signaling pathway, response to stress, small GTPase-mediated gene, the intensity was significantly higher in the periphery by signal transduction, hexose metabolism, cell death, response to 51.5 units (F = 903; P < 0.0001). The results of FISH analysis were external stimulus, carbohydrate metabolism, and response to coincident to those of microarray analysis showing the heteroge- wounding. Three hypoxia-related genes were identified in the neity of gene expression in the same tumor. patterns up-regulated in the central zone, but the differences were not statistically significant (P = 0.062; Table 1). These data suggest that gene expression of cells growing in the Discussion peripheral zone of the tumor were optimized for cell proliferation In this study, we clearly showed zonal heterogeneity for gene and invasion, whereas the gene expression of the cells in the central expression profiles of cells in the same tumor depending on the zone of the same tumor were related to stress, a unique cascade of zones where they are growing (dividing). Hundreds of genes showed signaling genes whose expression helps the cell to adapt to or differential expression in the central and the peripheral zones of undergo programmed cell death. The top 20 overexpressed genes in pancreatic tumor. Differences in gene expression profiles result from each category are shown in Table 2. the processes of selection, adaptation, or both. Clonal populations Verification of expression of genes by FISH. FISH analysis was expressing genes suited for specific microenvironments may survive done for confirmation of data obtained from the microarray (selection) or certain gene expression patterns may be induced by analysis. We did the FISH analysis on human pancreatic cancer interactions between the tumor cells and a different microenviron- L3.6pl cells growing in the pancreas of nude mice (Figs. 2 and 3). ment (adaptation; ref. 1). Characterization of gene expression

Figure 2. FISH analysis of AMIGO2. Pancreatic tumors 7 to 9 mm in diameter were analyzed. AMIGO2 probe was labeled for red fluorescence, and poly(dT) (control for mRNA) probe was labeled for green fluorescence. A, H&E staining of specimen showing the zones from which images of FISH were captured. Expression of AMIGO2 was high in the peripheral zone (B) and low in the central zone (D). Preservation of mRNA was confirmed by hybridization with poly(dT) probe (C and E). www.aacrjournals.org 7601 Cancer Res 2007; 67: (16). August 15, 2007

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Cancer Research

Figure 3. FISH analysis of RARRES1. Pancreatic tumors 7 to 9 mm in diameter were analyzed. RARRES1 probe was labeled for red fluorescence, and poly(dT) (control for mRNA) probe was labeled for green fluorescence. A, H&E staining of specimen showing the zones from which images of FISH were captured. Expression of RARRES1 was high in the central zone (B) and low in the peripheral zone (D). Preservation of mRNA was confirmed by hybridization with poly(dT) probe (C and E). profiles is very important for predicting the prognosis of the disease By contrast, the central zone had many genes involved in regu- and whether the patient may respond to therapy. lation of cell proliferation, response to stress, hexose metabolism, To produce reproducible and reliable data, the collection of carbohydrate metabolism, and cell death (Table 1). We expected viable cells from a specific zone of a tumor and extraction of fresh that genes related to response to hypoxia would appear on the top mRNAs were the most important and critical steps of this study. of the list for up-regulated genes in the central zone of the tumor; Tumors of similar size were selected, anoxic time was minimized however, only three genes were included and the ratio was low by immediate extraction of mRNAs, and cells were collected by (Table 2), falling below the cutoff for statistical significance in our laser microdissection from the corresponding zones of each tumor. study (P = 0.062; Table 1). Either surviving cells depend on other Microscopic dissection allowed us to exclude necrotic areas. pathways (adaptation) or cells that are independent of this pathway Our results from the gene ontology analysis were biologically survive (selection). relevant in that the cells at the edge of the tumor expressed genes We confirmed the reliability of microarray data by FISH analysis related to cell division, attachment, and cell migration because for AMIGO2 and RARRES1. AMIGO2 is the novel cell adhesion tumor cells at the periphery are dividing, attaching to the next molecule related to AMIGO, a leucine-rich repeat superfamily tissue, and changing their shape in preparation for invading. In member expressed in fiber tracts of neuronal tissues (37). AMIGO2 particular, overexpression of MMP-2 in the peripheral zone agrees is also associated with tumorigenesis and cell adhesion signaling in with the previous data showing the biological function of MMP-2 in gastric cancer (38). Overexpression of AMIGO2 in pancreatic cancer cells within the invasive edge of the tumors (23–27). is a novel finding in this report. Further study of it may lead to

Table 3. FISH analysis showing intratumoral heterogeneity in orthotopic pancreatic cancer

Intratumoral region No. tumors Gene expression median (range)

TIG-1 AMIGO2

Center 3 73.6 (61.8–83.0)* 18.6 (16.0–32.4)* Periphery 3 19.7 (15.1–31.3)* 74.9 (51.2–85.3)*

NOTE: Median (range) fluorescence intensities of each field. Intensity was measured in six different fields in both the peripheral and central regions of three different orthotopic tumors. *P < 0.0001, F test.

Cancer Res 2007; 67: (16). August 15, 2007 7602 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Zonal Heterogeneity for Gene Expression development of novel and effective therapeutic strategies against largest published studies that have attempted to predict prognosis pancreatic cancer. of cancer patients based on DNA microarray analysis and Decreased expression of RARRES1, a potential tumor suppressor concluded that five of the seven studies did not classify patients gene for human prostate cancer, was associated with an increase better than chance. To minimize intratumoral heterogeneity for in malignant characteristics of both prostate cancer cell lines gene expression among different cervical cancer patients, and tissues (39). RARRES1 may be related to the ability of cell to Bachtiary et al. (43) recommended analyzing multiple biopsies react to retinoid stimulation (39). Strong association of loss of of each tumor. RARRES1 expression with RARRES1 promoter hypermethylation In summary, the present results showing regional differences in was observed in half of the cases tested in various cancer cell lines gene expression between the peripheral zone and central zone of and primary tumor samples (40). tumors are not surprising. Whereas earlier data derived from non- L3.6pl, the human pancreatic cancer cell line used in this study, dissected tumor samples differed among different studies (41–43), has an aggressive phenotype (31). Whether endowing a highly our results clearly indicate the absolute necessity of using LCM metastatic phenotype and candidate genes producing an anti- from specific locations of a neoplasm for collection of reproducible proliferation phenotype were expressed depended on the zonal data. microenvironment of the tumor. Further study to reveal whether those differences arise from adaptation or selection would be required to understand the mechanism of disease progression and Acknowledgments allow development of treatment modalities. Received 3/6/2007; revised 6/1/2007; accepted 6/12/2007. Collectively, our present results predict that unless the zone Grant support: Cancer Center Support Core grant CA16672 and National Cancer Institute, NIH Specialized Programs in Research Excellence in Prostate Cancer grant of the tumor used to analyze gene expression is standardized, CA90270. the results will be heterogeneous. Indeed, Moroni et al. (41) The costs of publication of this article were defrayed in part by the payment of page examined human colorectal cancers and reported intratumoral charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. heterogeneity in EGFR gene amplification as correlated with We thank Walter Pagel for critical editorial review and Lola Lo´pez for expert protein level. Michiels et al. (42) reanalyzed data from the seven preparation of this manuscript.

References Effectiveness of gene expression profiling for response expression of type IV collagenase, E-cadherin, and prediction of rectal adenocarcinomas to preoperative vascular endothelial growth factor/vascular permeabi- 1. Fidler IJ. Critical factors in the biology of human chemoradiotherapy. J Clin Oncol 2005;23:1826–38. lity factor in prostatectomy specimens distinguishes cancer metastasis: twenty-eighth GHA Clowes memorial 15. Chang JC, Wooten EC, Tsimelzon A, et al. Gene organ-confined from pathologically advanced prostate award lecture. Cancer Res 1990;50:6130–8. expression profiling for the prediction of therapeutic cancers. Clin Cancer Res 2000;6:2295–308. 2. Fidler IJ. The pathogenesis of cancer metastasis: the response to docetaxel in patients with breast cancer. 27. Herbst RS, Yano S, Kuniyasu H, et al. Differential ‘seed and soil’ hypothesis revisited (Timeline). Nat Rev Lancet 2003;362:362–9. expression of E-cadherin and type IV collagenase genes Cancer 2003;3:453–8. 16. Wang Y, Jatkoe T, Zhang Y, et al. Gene expres- predicts outcome in patients with stage I non-small cell 3. Talmadge JE, Wolman SR, Fidler IJ. Evidence for the sion profiles and molecular markers to predict recur- lung carcinoma. Clin Cancer Res 2000;6:790–7. clonal origin of spontaneous metastasis. Science 1982; rence of Dukes’ B colon cancer. J Clin Oncol 2004;22: 28. Ohmori H, Fujii K, Sasahira T, et al. Determinants for 217:361–3. 1564–71. prediction of malignant potential by metalloproteina- 4. Fidler IJ. Critical determinants of metastasis. Semin 17. Fidler IJ. The organ microenvironment and cancer se:E-cadherin ratio in prostate core needle biopsy. Cancer Biol 2002;12:89–96. metastasis [review]. Differentiation 2002;70:498–505. Pathobiology 2006;73:98–104. 5. Rosenwald A, Wright G, Chan WC, et al. The use of 18. Fidler IJ. Angiogenic heterogeneity: regulation of 29. Hedley D, Pintilie M, Woo J, et al. Carbonic anhydrase molecular profiling to predict survival after chemother- neoplastic angiogenesis by the organ microenviron- IX expression, hypoxia, and prognosis in patients with apy for diffuse large-B-cell lymphoma. N Engl J Med ment. J Natl Cancer Inst 2001;93:1040–1. uterine cervical carcinomas. Clin Cancer Res 2003;9: 2002;346:1937–47. 19. Liotta LA, Kohn EC. The microenvironment of the 5666–74. 6. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene tumour-host interface [review]. Nature 2001;411:375–9. 30. Sigalotti L, Fratta E, Coral S, et al. Intratumor expression profiling predicts clinical outcome of breast 20. Gisselsson D, Pettersson L, Hoglund M, et al. heterogeneity of cancer/testis antigens expression in cancer. Nature 2002;415:530–6. Chromosomal breakage-fusion-bridge events cause ge- human cutaneous melanoma is -regulated 7. van de Vijver MJ, He YD, van’t Veer LJ, et al. A gene netic intratumor heterogeneity. Proc Natl Acad Sci U S A and functionally reverted by 5-aza-2¶-deoxycytidine. expression signature as a predictor of survival in breast 2000;97:5357–62. Cancer Res 2004;64:9167–71. cancer. N Engl J Med 2002;347:1999–2009. 21. Griffin CA, Hruban RH, Morsberger LA, et al. 31. Bruns CJ, Harbison MT, Kuniyasu H, Eue I, Fidler IJ. 8. Beer DG, Kardia SL, Huang CC, et al. Gene expression Consistent chromosome abnormalities in adenocarci- In vivo selection and characterization of metastatic profiles predict survival of patients with lung adenocar- noma of the pancreas. Cancer Res 1995;55:2394–9. variants from human pancreatic adenocarcinoma by cinoma. Nat Med 2002;8:816–24. 22. Gorunova L, Hoglund M, Andren-Sandberg A, et al. using orthotopic implantation in nude mice. Neoplasia 9. Pomeroy SL, Tamayo P, Gaasenbeek M, et al. Cytogenetic analysis of pancreatic carcinomas: intra- 1999;1:50–62. Prediction of central nervous system embryonal tumour tumor heterogeneity and nonrandom pattern of chro- 32. Bucana CD, Radinsky R, Dong Z, Sanchez R, Brigati outcome based on gene expression. Nature 2002;415: mosome aberrations. Genes Cancer 1998; DJ, Fidler IJ. A rapid colorimetric in situ mRNA 436–42. 23:81–99. hybridization technique using hyperbiotinylated oli- 10. Iizuka N, Oka M, Yamada-Okabe H, et al. Oligonu- 23. Kitadai Y, Ellis LM, Takahashi Y, et al. Multipara- gonucleotide probes for analysis of mdr1 in mouse cleotide microarray for prediction of early intrahepatic metric in situ messenger RNA hybridization analysis to colon carcinoma cells. J Histochem Cytochem 1993;41: recurrence of hepatocellular carcinoma after curative detect metastasis-related genes in surgical specimens 499–506. resection. Lancet 2003;361:923–9. of human colon carcinomas. Clin Cancer Res 1995;1: 33. Radinsky R, Bucana CD, Ellis LM, et al. A rapid 11. Barrier A, Boelle PY, Roser F, et al. Stage II colon 1095–102. colorimetric in situ messenger RNA hybridization cancer prognosis prediction by tumor gene expression 24. Kitadai Y, Ellis LM, Tucker SL, et al. Multiparametric technique for analysis of epidermal growth factor profiling. J Clin Oncol 2006;24:4685–91. in situ mRNA hybridization analysis to predict disease receptor in paraffin-embedded surgical specimens of 12. Chen CN, Lin JJ, Chen JJ, et al. Gene expression profile recurrence in patients with colon carcinoma. Am J human colon carcinomas. Cancer Res 1993;53:937–43. predicts patient survival of gastric cancer after surgical Pathol 1996;149:1541–51. 34. Kitadai Y, Bucana CD, Ellis LM, Anzai H, Tahara E, resection. J Clin Oncol 2005;23:7286–95. 25. Kuniyasu H, Ellis LM, Evans DB, et al. Relative Fidler IJ. In situ mRNA hybridization technique for 13. Bogaerts J, Cardoso F, Buyse M, et al. Gene signature expression of E-cadherin and type IV collagenase analysis of metastasis-related genes in human colon evaluation as a prognostic tool: challenges in the design genes predicts disease outcome in patients with carcinoma cells. Am J Pathol 1995;147:1238–47. of the MINDACT trial [review]. Nat Clin Pract Oncol resectable pancreatic carcinoma. Clin Cancer Res 1999; 35. Reed JA, Manahan LJ, Park CS, Brigati DJ. Complete 2006;3:540–51. 5:25–33. one-hour immunocytochemistry based on capillary 14. Ghadimi BM, Grade M, Difilippantonio MJ, et al. 26. Kuniyasu H, Troncoso P, Johnston D, et al. Relative action. Biotechniques 1992;13:434–43. www.aacrjournals.org 7603 Cancer Res 2007; 67: (16). August 15, 2007

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Cancer Research

36. Li C, Wong WH. Model-based analysis of oligonucle- effects on ploidy, chromosomal stability, cell adhesion/ number for epidermal growth factor receptor (EGFR) otide arrays: expression index computation and outlier migration, and tumorigenicity. Oncogene 2004;23:5056–67. and clinical response to anti-EGFR treatment in detection. Proc Natl Acad Sci U S A 2001;98:31–6. 39. Jing C, El-Ghany MA, Beesley C, et al. Tazarotene- colorectal cancer: a cohort study. Lancet Oncol 2005;6: 37. Kuja-Panula J, Kiiltomaki M, Yamashiro T, Rouhiainen induced gene 1 (TIG1) expression in prostate carcino- 279–86. A, Rauvala H. AMIGO, a transmembrane protein mas and its relationship to tumorigenicity. J Natl Cancer 42. Michiels S, Koscielny S, Hill C. Prediction of cancer implicated in axon tract development, defines a novel Inst 2002;94:482–90. outcome with microarrays: a multiple random valida- protein family with leucine-rich repeats. Cell Biol 2003; 40. Youssef EM, Chen XQ, Higuchi E, et al. Hyper- tion strategy. Lancet 2005;365:488–92. 160:963–73. methylation and silencing of the putative tumor 43. Bachtiary B, Boutros PC, Pintilie M, et al. Gene 38. RabenauKE, O’Toole JM, Bassi R, et al. DEGA/ suppressor tazarotene-induced gene 1 in human can- expression profiling in cervical cancer: an exploration of AMIGO-2, a leucine-rich repeat family member, differ- cers. Cancer Res 2004;64:2411–7. intratumor heterogeneity. Clin Cancer Res 2006;12: entially expressed in human gastric adenocarcinoma: 41. Moroni M, Veronese S, Benvenuti S, et al. Gene copy 5632–40.

Cancer Res 2007; 67: (16). August 15, 2007 7604 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research. Zonal Heterogeneity for Gene Expression in Human Pancreatic Carcinoma

Toru Nakamura, Toshio Kuwai, Yasuhiko Kitadai, et al.

Cancer Res 2007;67:7597-7604.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/67/16/7597

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2007/10/29/67.16.7597.DC1

Cited articles This article cites 43 articles, 18 of which you can access for free at: http://cancerres.aacrjournals.org/content/67/16/7597.full#ref-list-1

Citing articles This article has been cited by 3 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/67/16/7597.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/67/16/7597. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2007 American Association for Cancer Research.