Intersex-like (IXL) is a cell survival regulator in pancreatic cancer with 19q13 amplification. Cancer Res 67

Pancreatic cancer is a highly aggressive disease characterized by poor prognosis and vast genetic instability. Recent microarray-based, genome-wide surveys have identified multiple recurrent copy number aberrations in pancreatic cancer; however, the target genes are, for the most part, unknown. Here, we characterized the 19q13 amplicon in pancreatic cancer to identify putative new drug targets. Copy number increases at 19q13 were quantitated in 16 pancreatic cancer cell lines and 31 primary tumors by fluorescence in situ hybridization. Cell line copy number data delineated a 1.1 Mb amplicon, the presence of which was also validated in 10% of primary pancreatic tumors. Comprehensive expression analysis by quantitative real-time reverse transcription-PCR indicated that seven transcripts within this region had consistently elevated expression levels in the amplified versus nonamplified cell lines. High-throughput loss-of-function screen by RNA interference was applied across the amplicon to identify genes whose down-regulation affected cell viability. This screen revealed five genes whose down-regulation led to significantly decreased cell viability in the amplified PANC-1 cells but not in the nonamplified MiaPaca-2 cells, suggesting the presence of multiple biologically interesting genes in this region. Of these, the transcriptional regulator intersex-like (IXL) was consistently overexpressed in amplified cells and had the most dramatic effect on cell viability. IXL silencing also resulted in G(0)-G(1) cell cycle arrest and increased apoptosis in PANC-1 cells. These findings implicate IXL as a novel amplification target gene in pancreatic cancer and suggest that IXL is required for cancer cell survival in 19q13-amplified tumors.


Introduction
Pancreatic adenocarcinoma is a highly aggressive malignancy with extremely poor prognosis.In the United States, pancreatic cancer is the fifth leading cause of cancer death, accounting for f30,000 deaths per year (1).Pancreatic cancer is characterized by rapid progression, invasiveness, and profound resistance to treatment (2).Apart from surgery, there is practically no effective therapy; typically, the disease is diagnosed at an advanced stage when surgical resection is no longer possible.Consequently, the 5-year survival rate for pancreatic cancer is <5% and the median survival is <6 months (2,3).Even for patients who undergo potentially curative resection, the 5-year survival rate is only f20% (2).
Aneuploidy and increased genetic instability manifesting as complex genetic aberrations, such as losses, gains, and amplifications, are common features of pancreatic cancer (4,5).These genetic alterations are likely to conceal genes involved in disease pathogenesis, and uncovering such genes might thus provide targets for the development of new diagnostic and therapeutic tools.In particular, gene amplification is a common mechanism for activating oncogenes, and other growth-promoting genes in cancer and amplification of target genes, such as ERBB2 in breast cancer and MYCN in neuroblastomas, have been shown to have clinical significance as diagnostic and prognostic markers as well as therapeutic targets (6).We recently did a microarray-based copy number analysis in pancreatic cancer cell lines and identified an f2.9 Mb amplicon at 19q13 (7).This result has since been confirmed in multiple subsequent studies (8)(9)(10)(11)(12) and the same amplicon has also been observed in other tumor types, including ovarian, breast, cervical, gastric, and small-cell lung cancer (13)(14)(15)(16)(17)(18)(19)(20).Unfortunately, the microarray used in our previous study did not provide a complete coverage of the 19q13 region and thus did not allow direct elucidation of putative target genes.Here, we present results from the comprehensive evaluation of the 19q13 amplicon in pancreatic cancer, including detailed copy number and expression analyses as well as high-throughput loss-of-function screen using the RNA interference (RNAi) technology.
Genomic clones.Public genome databases (National Center for Biotechnology Information3 and University of California Santa Cruz Genome Bioinformatics4 ) were used to select 15 bacterial artificial chromosome (BAC) clones evenly distributed over the 2.9 Mb amplicon at 19q13.These BAC clones were obtained from Invitrogen (Carlsbad, CA).In addition, a BAC clone (RP11-345J21, a kind gift from Mariano Rocchi, University of Bari, Bari, Italy) adjacent to the chromosome 19 centromere on 19q was used as a control.The list of all BAC clones is shown in Supplementary Table S2.
Fluorescence in situ hybridization.BAC clone DNA was labeled with SpectrumOrange-dUTP (Vysis, Downers Grove, IL) using random priming.Chromosome 19 reference probe (RP11-345J21) was labeled with fluorescein-12-dUTP (Perkin-Elmer, Boston, MA).Dual-color interphase fluorescence in situ hybridization (FISH) to pancreatic cancer cell lines was done as described (21).Hybridization signals were evaluated using Olympus BX50 fluorescence microscope (Olympus, Tokyo, Japan).Forty intact nuclei were scored for each probe, and relative copy numbers were calculated as ratios of mean absolute copy number of the test probe versus reference probe.FISH on tissue microarray was carried out as described (22).Three adjacent, partly overlapping BAC clones (RP11-67A5, RP11-256O9, and CTC-488F21) were combined to increase signal intensity and hybridization efficiency.Control experiments on normal lymphocytes verified that this probe combination gave a single hybridization signal.The RP11-345J21 probe was again used as a reference.Hybridization signals were scored and evaluated as described above.
Quantitative real-time reverse transcription-PCR.Gene expression analyses were done using either the Light Cycler equipment (Roche, Mannheim, Germany) or the ABI 7900HT Fast Sequence Detection System (Applied Biosystems, Foster City, CA).Total RNA was isolated from cell lines using RNeasy Mini total RNA extraction kit (Qiagen, Valencia, CA) and reverse transcribed into first-strand cDNA using Superscript III reverse transcriptase (Invitrogen) and random hexamers or iScript cDNA synthesis kit (Bio-Rad, Inc., Hercules, CA).Normal human pancreatic RNA was obtained from Ambion (Cambridgeshire, United Kingdom).For the Light Cycler analyses, primers and probe sets were obtained from TIB MolBiol (Berlin, Germany), and Light Cycler software (Roche) was used for data analysis as described (23).Expression levels of the target genes were normalized against a housekeeping gene TBP (TATA box binding protein; ref. 23).For the ABI system, TaqMan Gene Expression Assays were obtained from Applied Biosystems and ABI 7900HT software (Applied Biosystems) was used for data analyses.Expression levels of the target genes were normalized against an endogenous reference gene GAPDH (glyceraldehyde-3-phosphate dehydrogenase).Primer and probe sequences are shown in Supplementary Table S3.
Loss-of-function RNAi screen.Loss of function screening was done by high-throughput RNAi using a focused small interfering RNA (siRNA) library targeting all 19 known genes within the amplicon that were expressed in PANC-1 cells (Supplementary Table S4).Two genes, SAMD4B and EID2B, were hypothetical proteins at the time of the experiment and were thus not included.Four siRNAs were designed for each gene using previously described criteria (24,25) and were obtained from Qiagen.A nonsilencing siRNA (Qiagen) was used as a negative control.High-throughput RNAi was done using siRNA reverse transfection of cells.Briefly, siRNAs were printed in quadruplicate wells of a 384-well Costar microtiter plate (Fisher Scientific, Hampton, NH).Diluted Oligofectamine (Invitrogen) was added to the wells to allow for the complexing of siRNA and transfection reagent.After a 30-min incubation period, cell suspensions of either MiaPaca-2 or PANC-1 cells were added to give a final concentration of 1,000 per well.Cells were grown at 37jC with 5% CO 2 for 96 h.Total cell number was analyzed using Cell Titer Blue (Promega, Madison WI), and the plate was read at excitation 544 nm/emission 560 nm using a EnVision plate reader (Perkin-Elmer, Wellesley, MA).Reduced cell viability to 50% or less compared with nonsilencing siRNA-treated cells was considered a significant change.
Cell cycle and apoptosis analyses.Oligofectamine reagent (Invitrogen) was used to transfect siRNAs in a final concentration of 100 nmol/L into PANC-1 cells according to the manufacturer's protocol.IXL 144 siRNA was used to determine the effect of inhibition of IXL expression; luciferase control siRNA, which targets the firefly luciferase gene (Genbank accession no.M15077), was used as a control (Supplementary Table S4).Experiments were done in triplicates using 24-well plates.After a 48-h transfection, cells were collected for cell cycle and apoptosis analyses as well as for parallel mRNA expression analyses (see above) to verify that efficient silencing was obtained.For the cell cycle analysis, trypsinized cells were centrifuged and suspended to 500 AL hypotonic staining buffer (0.1 mg/mL sodium citrate tribasic dehydrate, 0.03% Triton X-100, 50 Ag/mL propidium iodide, 2 Ag/mL RNase A) and the amount of propidium iodide incorporated was determined using flow cytometry (Coulter EPICS XL-MCL, Beckman Coulter, Inc., Fullerton, CA).Cell cycle distribution was analyzed using the Cylchred program. 5Annexin V FITC Apoptosis Detection Kit (Calbiochem) was used to detect apoptotic cells by flow cytometry (Beckman Coulter, San Diego, CA).

Results
We first did a systematic characterization of the extent of the 19q13 amplicon in a panel of 16 pancreatic cancer cell lines.To this end, FISH, using 15 BAC clones evenly distributed across the 2.9 Mb region (Supplementary Table S1), was applied.Greater than 2-fold copy number increases were observed in 3 of 16 (19%) cell lines, PANC-1, Su.86.86, and HPAC (Fig. 1).The PANC-1 cell line harbored massive amplification (up to 20-fold) that on metaphase chromosomes manifested as homogeneously staining region-like structures (Fig. 1E), whereas lower-level copy number increases (up to 5.3-and 2.6-fold, respectively) were observed in SU.86.86 and HPAC cells (Fig. 2).The relative copy numbers of PANC-1 cells varied considerably across the 2.9 Mb region (Fig. 1A-D), whereas SU.86.86 and HPAC cells showed essentially uniform copy number levels across the entire amplicon.Based on the PANC-1 copy number profile, we were able to delineate a 1.1 Mb amplicon core defined by six partly overlapping BAC clones from RP11-67A5 to CTC-425O23 (Fig. 2).Furthermore, three BAC clones within this 1.1 Mb amplicon core, RP11-67A5, RP11-256O9, and CTC-488F21, displayed 14-to 20-fold copy number increase in PANC-1 cells (Fig. 2), thus defining an 660 kb subregion of extremely high level amplification.
We also examined the presence of this amplicon in primary tumors using FISH to a tissue microarray containing 33 pancreatic cancer specimens.High-level amplification (relative copy number z5) was observed in 3 of 31 tumors (9.7%) with successful hybridizations (Fig. 3).All three tumors with amplification were moderately to poorly differentiated (grades 2-3) ductal adenocarcinomas and showed lymph node metastases (Supplementary Table S2).Two of the tumor samples also had extensions to peripancreatic soft tissue with perineural invasion.The metastasis status of the third tumor was not available.
To explore the consequences of amplification on gene expression, we first used public genome databases 6 to retrieve all transcribed sequences within the core 1.1 Mb amplicon.A total of 39 transcripts, including 27 known genes and 12 hypothetical or predicted proteins, were identified (Table 1).Sixteen of these were excluded as possible candidate genes because they represented (a) obvious pseudogenes, (b) predicted transcripts with no mRNA and expressed sequence tag evidence in public databases, or (c) transcripts with no or very low level expression in PANC-1 (Table 1).The last criterion was based on the expectation that any putative amplification target gene should be highly expressed in PANC-1 cells that show high-level amplification.The expression levels of the remaining 23 transcripts were then assessed in all 16 pancreatic cancer cell lines using quantitative real-time RT-PCR.Interestingly, 22 of the 23 genes were highly overexpressed in PANC-1 cells compared with the other pancreatic cancer cell lines (Fig. 4) and normal pancreas (data not shown).Only PLD3, located at the distal-most end of the amplicon, was not overexpressed in PANC-1.In contrast, HPAC and Su.86.86 cells showed variable expression levels from one gene to another (Fig. 4) thus allowing us to distinguish genes differentially expressed between amplified and nonamplified cell lines.Six known genes (GMFG, SAMD4B, IXL, SUPT5H, PSMC4, and MAP3K10) and one hypothetical protein (LOC284323) showed consistent overexpression in all three amplified cell lines (Fig. 4).Of these, IXL showed the most distinct differential expression pattern between the amplified and nonamplified cell line groups, with high-level expression occurring almost exclusively in the amplified cell lines.In contrast, GMFG and MAP3K10 were also highly expressed in additional cell lines with no copy number increase, indicating other activating mechanisms besides amplification.Interestingly, three of the overexpressed genes, PSMC4, MAP3K10, and LOC284323, are located outside the 660 kb amplicon maximum, thus making them less likely to be the main targets of the amplification.
To identify which of the amplified targets are functionally important, we did a targeted high-throughput RNAi screen across the defined 1.1 Mb amplicon.This loss-of-function survey was applied to study the effect of silencing of 19 known genes from the amplified region.Four different siRNAs were designed for each gene (Supplementary Table S4), and their effect on cell viability was examined in PANC-1 and MiaPaCa-2 cells 96 h after transfection.Down-regulation of IXL resulted in statistically highly significant reduction in cell viability (P < 0.001) in the amplified PANC-1 cells but not in the nonamplified MiaPaCa-2 cells (Fig. 5A; Supplementary Table S5).This effect was observed with two independent siRNAs in three repeated experiments.Similar reduction in cell viability was detected after LRFN1 and PLEKHG2 silencing as well but only with a single siRNA.Finally, down-regulation of GMFG and SUPT5H also led to reduction in cell viability but to a lesser extent (Fig. 5A).Other genes examined did not have a statistically significant effect on cell viability.
Further functional characterization concentrated on IXL because it was located within the 660 kb amplicon maximum and was implicated in both the expression survey and the RNAi viability  screen.To explore the functional role of IXL overexpression, we transfected siRNAs targeting IXL into amplified PANC-1 cells and assessed the effects on cell cycle and apoptosis 48 h after transfection.A parallel mRNA expression analysis was done to verify that sufficient silencing was obtained (Fig. 5B).All functional experiments were done in triplicates and repeated twice.Cell cycle analysis using flow cytometry showed an increased fraction of G 0 -G 1 phase cells after IXL siRNA (66%) transfection compared with nontransfected cells (37%) or cells transfected with a control siRNA (39%; Fig. 5C).Also, the percentage of apoptotic cells was increased in IXL siRNA-treated cells (8%) compared with control siRNA (1.6%) or untreated cells (2.5%; Fig. 5D).Similar phenotypic changes were not observed after IXL siRNA transfection to nonamplified MiaPaCa-2 cells (data not shown).

Discussion
Gross chromosomal aberrations, including losses, gains, and amplifications, are frequent in pancreatic cancer (3).Gene amplification is a common mechanism for solid tumors to upregulate the expression of genes involved in tumor progression (26,27).Although multiple amplified regions have been documented in pancreatic carcinoma, the putative target genes activated by these aberrations are largely unknown.Identification of novel amplification target genes is extremely important because it will not only advance our knowledge on pancreatic cancer pathogenesis but it might also provide new tools for the clinical management of this highly aggressive disease.
Previously, we did a genome-wide cDNA microarray-based copy number analysis to identify localized DNA amplifications in pancreatic cancer and recognized a novel amplified region at 19q13 spanning 2.9 Mb (7).However, due to incomplete clone coverage on the microarray used, this previous study did not permit exact definition of the amplicon boundaries or direct identification of possible amplification target genes.Therefore, we now carried out a systematic evaluation of this amplified region to achieve these objectives.Three cell lines, PANC-1, SU86.86, and HPAC, harbored the 19q13 amplification, with PANC-1 cells demonstrating a massive, up to 20-fold, amplification.In addition, the copy number profile of PANC-1 allowed us to narrow down the amplicon core to 1.1 Mb, which also contained a 660 kb subregion of extremely high level amplification.Previous studies have indicated that amplification target genes are likely to be located at or near the center of the amplification maximum; that is, the region with highest copy number increase (28,29).We thus believe this 660 kb subregion to be of particular interest in pinpointing the actual target genes of this amplicon.
Because the delineation of the amplicon core was accomplished using established pancreatic cancer cell lines, we next sought to validate the presence of the 19q13 amplicon in primary pancreatic tumors.Evaluation of a set of 31 pancreatic tumors revealed amplification in a f10% of the cases, thus confirming that this aberration is also present in actual human tumors and is not a cell culture-derived artifact.In general, oncogene amplification has been shown to be linked to advanced disease (30,31), and, in ovarian carcinoma, 19q13 amplification has been associated with less differentiated and more aggressive tumors (14).In our series, all three tumors with amplification were moderately to poorly differentiated.All three patients had nodal metastases and two of them also had local metastases, whereas this information was missing from the third patient.Based on these tumor characteristics, the 19q13 amplification seems to be associated with advanced disease in pancreatic cancer as well.However, the number of analyzed tumors was limited; therefore, the amplification frequency as well as the possible clinicopathologic associations need to be confirmed.
A comprehensive expression analysis was subsequently done across the 1.1 Mb region to identify genes whose expression levels are elevated through 19q13 amplification.This approach was based on the well-established concept that amplification leads to increased expression of the putative target gene.Our data revealed a very distinct expression profile in PANC-1 cells demonstrating high-level expression of all but one gene throughout the entire amplicon.We hypothesize that the extremely high-level amplification in PANC-1 cells leads to complete deregulation of transcriptional control across the amplicon and thereby increased expression of all genes within this region.In contrast, SU.86.86 and HPAC cells displayed more variable expression patterns with consistent overexpression in a subset of seven genes compared with nonamplified cells, with IXL having the strongest association with amplification.Four of these seven genes, GMFG, SAMD4B, IXL, and SUPT5H, are located within the 660 kb amplicon maximum and were therefore considered the most likely targets.
Because the expression analysis did not explicitly pinpoint a single putative amplification target gene, we chose to perform a targeted high-throughput RNAi screen across the entire amplicon to identify functionally relevant genes.Combination of data from the expression analysis and the RNAi screen allowed us to rapidly and systematically identify genes whose expression levels were elevated through amplification, and, at the same time, whose down-regulation resulted in phenotypic changes.This strategy highlighted IXL as a gene that is activated by the 19q13 amplification and whose down-regulation resulted in most dramatic reduction in cell viability in amplified PANC-1 but not in nonamplified MiaPaCa-2 cells.LRFN1 and PLEKH2G knockdowns also affected cell viability but not as consistently as IXL.However, these genes were not systematically overexpressed in all amplified cell lines.Down-regulation of SUPT5H and GMFG also depressed cell viability; however, the decrease was not as significant as for IXL.Based on these data, IXL seems to be the strongest candidate for the amplification target gene; however, we cannot exclude the involvement of some of the other genes in the region.
Further functional assays showed that IXL silencing resulted in increased apoptosis and G 0 -G 1 arrest again in PANC-1 cells but not in MiaPaCa-2 cells, suggesting that IXL affects cell cycle regulatory mechanisms that control the G 1 -S transition as well as induction of apoptosis.These data implicate that IXL is required for cell cycle progression and cell survival in 19q13-amplified pancreatic cancer cells.IXL is a homologue to Drosophila melanogaster intersex, a transcriptional regulator involved in female somatic sex determination (32).The protein is broadly conserved during evolution (33), suggesting its importance in transcriptional regulation also in other species.Indeed, mammalian IXL has been recognized as a subunit of Mediator, a multiprotein complex that transduces regulatory signals from DNA-binding transcription factors to RNA polymerase II and thereby regulates mRNA synthesis (33,34).The Mediator complex is required for transcriptional activation and thus controls key cellular processes.The exact function of human IXL remains elusive, although it was recently proposed to be involved in mitogen-activated protein kinase signaling pathway (33).The overexpression of IXL in pancreatic cancer may thus lead to inappropriate activation of several critical cellular processes, such as those regulating cell growth.
The AKT2 gene has previously been suggested as the target for the 19q13 amplification (15).AKT2 is the human homologue of the viral v-akt oncogene, which is responsible for leukemia in mice (19).Amplification of AKT2 was originally discovered in ovarian cancer (15) but has been later observed also in other cancer types (14).Our data show that AKT2 is indeed amplified in pancreatic cancer, although it is located at the distal-most end of the 19q13 amplicon.However, AKT2 was not consistently overexpressed in all amplified cell lines and its down-regulation did not affect cell viability.These findings indicate that AKT2 is clearly not the main target of the 19q13 amplification in pancreatic cancer although it cannot be ruled out that simultaneous activation of AKT2 with other candidate genes from this region might provide a growth advantage for the cancer cells.Recently, two other genes from this region, PAF1 and DYRK1B, were proposed to associate with pancreatic cancer development and cell survival, respectively (35,36).Yet, again, our data do not show evidence that these genes would be the key targets of 19q13 amplicon.In summary, our detailed characterization of the 19q13 amplicon in pancreatic cancer cell lines delineated a minimal region of amplification to a 1.1 Mb segment and further pinpointed a 660-kb amplification maximum.This amplicon was recognized in 19% of cell lines and 10% of primary pancreatic tumors.Expression profiling of genes residing in the amplicon revealed seven biologically interesting genes that were more strongly expressed in the amplified cell lines compared with the nonamplified ones.High-throughput loss-of-function screen by RNAi technology showed that down-regulation of IXL, and, to a lesser extent, GMFG and SUPT5H, resulted in decreased cell viability in the amplified PANC-1 but not in the nonamplified cells.Additionally, IXL knockdown was found to associate with G 0 -G 1 arrest and increased apoptosis.Our results reveal IXL as a novel amplification  target gene that is essential for the growth and survival of a subset of pancreatic carcinomas with 19q13 amplification.Thereby, IXL has a critical role in pancreatic cancer development and growth regulation and represents an ideal therapeutic target.However, it is possible that other genes in this region might also contribute to pancreatic cancer pathogenesis.Finally, this study shows that the combination of copy number and expression analysis together with targeted RNAi screen provides an efficient method for rapid identification of putative amplification target genes in cancer.

Figure 2 .
Figure 2. Amplicon mapping at 19q13 in pancreatic cancer cell lines.Relative copy number ratios for eight BAC clones around the minimal amplified region are shown for the three amplified cell lines, PANC-1, SU.86.86, and HPAC, as well as a representative nonamplified cell line, Capan-1.Vertical dashed lines, the 1.1 Mb core region of amplification (defined by clones RP11-67A5 and CTC-425O23) as well as the 660 kb subregion (defined by clones RP11-67A5 and CTC-488F21).

Figure 3 .
Figure 3. Copy number analysis in primary pancreatic tumors by FISH.A contig of three adjacent probes (RP11-67A5, RP11-256O9, CTC-488F21; red signals ) was hybridized together with the chromosome 19 pericentromeric control probe (green signals ) to a tissue microarray containing 33 pancreatic tumor samples.Examples of tumors with (A-B ) and without (C) amplification.

Figure 4 .
Figure 4. Schematic representation of mRNA expression levels of 23 genes within the 1.1 Mb amplicon at 19q13 in 16 pancreatic cancer cell lines.The expression levels were determined using quantitative real-time RT-PCR and were normalized against a housekeeping gene TBP .The relative expression levels for each gene were median centered and displayed as a pseudocolor gradient.The genes are arranged according to their chromosomal position (from centromere to telomere).Bottom, key to the color coding.Horizontal line above the figure, 660 kb amplicon core.

Figure 5 .
Figure 5. Functional evaluation of RNAi-based gene silencing in the PANC-1 cells.A, summary of data from the high-throughput RNAi viability screen.The number of cells were determined 96 h after siRNA transfection and compared with that of untreated control (UT ).Columns, mean of four independent wells shown for selected genes as well as nonsilencing control siRNA; bars, SD. *, P < 0.01; **, P < 0.001, statistically significant reduction in cell viability.The experiment was repeated thrice with similar results.Raw data from the entire viability screen are shown in Supplementary TableS5.B, quantitative real-time RT-PCR analysis of IXL mRNA expression levels in PANC-1 cells 48 h after transfection of luciferase control siRNA (siLUC) or IXL 144 siRNA.Results were normalized against untreated cells.Columns, mean of triplicate experiments; bars, SD.C, IXL down-regulation results in G 0 -G 1 arrest.Cell cycle analysis of untreated, luciferase siRNA-treated and IXL 144 siRNA-treated PANC-1 cells 48 h after transfection.Columns, mean of triplicate experiments; bars, SD.D, induction of apoptosis after IXL silencing.Percentage of early apoptotic, late apoptotic, and total apoptotic cells are shown for untreated, luciferase siRNAtreated, and IXL 144 siRNA-treated PANC-1 cells 48 h after transfection.Columns, mean of triplicate experiments; bars, SD.

Table 1 .
List of genes in the 19q13 amplicon *Not expressed in PANC-1.