Oncogene (2004) 23, 5084–5091 & 2004 Nature Publishing Group All rights reserved 0950-9232/04 $30.00 www.nature.com/onc

Distinctive expression of human lung adenocarcinomas carrying LKB1 mutations

Paloma Fernandez1, Julian Carretero1, Pedro P Medina1, Ana I Jimenez1, Sandra Rodriguez- Perales2, Maria F Paz3, Juan C Cigudosa2, Manel Esteller3, Luis Lombardia4, Manuel Morente5, Lydia Sanchez-Verde6, Teresa Sotelo7 and Montserrat Sanchez-Cespedes*,1

1Lymphomas and Lung Cancer Laboratory, Spanish National Cancer Centre (CNIO), Madrid, Spain; 2Cytogenetics Unit, Spanish National Cancer Centre (CNIO), Madrid, Spain; 3Cancer Epigenetics Laboratory, Spanish National Cancer Centre (CNIO), Madrid, Spain; 4Genomic Analysis, Spanish National Cancer Centre (CNIO), Madrid, Spain; 5Tumor Bank, Spanish National Cancer Centre (CNIO), Madrid, Spain; 6Immunohistochemistry and Histology Units, Spanish National Cancer Centre (CNIO), Madrid, Spain; 7Pathology Department, Hospital Universitario, 12 de Octubre, Madrid, Spain

LKB1, a tumor-suppressor gene that codifies for a serine/ kinase activation. In conclusion, our results reveal threonine kinase, is mutated in the germ-line of patients that several important factors contribute to LKB1- affected with the Peutz–Jeghers syndrome (PJS), which mediated carcinogenesis in LADs, confirming previous have an increased incidence of several cancers including observations and identifying new putative pathways that gastrointestinal, pancreatic and lung carcinomas. Regard- should help to elucidate the biological role of LKB1. ing tumors arising in non-PJS patients, we recently Oncogene (2004) 23, 5084–5091. doi:10.1038/sj.onc.1207665 observed that at least one-third of lung adenocarcinomas Published online 12 April 2004 (LADs) harbor somatic LKB1 gene mutations, supporting a role for LKB1 in the origin of some sporadic tumors. To Keywords: LKB1/STK11; Peutz–Jeghers syndrome; characterize the pattern of LKB1 mutations in LADs lung adenocarcinomas; cDNA microarrays further, we first screened for LKB1 gene alterations (gene mutations, promoter hypermethylation and homozygous deletions) in 19 LADs and, in agreement with our previous data, five of them (26%) were shown to harbor mutations, Introduction all of which gave rise to a truncated protein. Recent reports demonstrate that LKB1 is able to suppress cell Germ-line mutations of the LKB1 gene, also named growth, but little is known about the specific mechanism STK11, cause Peutz–Jeghers syndrome (PJS) (Hemmin- by which it functions. To further our understanding of ki et al., 1998; Jenne et al., 1998), an autosomal LKB1 function, we analysed global expression in lung dominant genetic disorder that predisposes to a wide primary tumors using cDNA microarrays to identify variety of malignancies predominantly of the gastro- LKB1-specific variations in . In all, 34 intestinal tract and also the genitourinary tract, breast transcripts, 24 of which corresponded to known , and lung (Jeghers et al., 1949; Giardiello et al., 1987). differed significantly between tumors with and without We recently detected somatic LKB1 gene-inactivating LKB1 gene alterations. Among the most remarkable mutations in one-third of sporadic lung adenocarcino- findings was deregulation of transcripts involved in signal mas (LADs) and in half of lung cancer cell lines of the transduction (e.g. FRAP1/mTOR, ARAF1 and ROCK2), LAD histological type, indicating that LKB1 plays a cytoskeleton (e.g. MPP1), transcription factors (e.g. relevant role in an extremely frequent cancer type such MEIS2, ATF5), metabolism of AMP (AMPD3 and as is that of the lung (Sanchez-Cespedes et al., 2002; APRT) and ubiquitinization (e.g. USP16 and UBE2L3). Carretero et al., 2004). Real-time quantitative RT–PCR on 15 tumors confirmed The LKB1 protein is a serine/threonine kinase. the upregulation of the homeobox MEIS2 and of the Closely related orthologues of LKB1 have been found AMP-metabolism AMPD3 transcripts in LKB1-mutant in mouse (Lkb1), Xenopus (Xeek1), Drosophila and tumors. In addition, immunohistochemistry in 10 of the Caenorhabditis elegans (Par4) (Su et al., 1996; Smith lung tumors showed the absence of phosphorylated et al., 1999; Watts et al., 2000; Martin and St Johnson, FRAP1/mTOR protein in LKB1-mutant tumors, indicat- 2003). Lkb1 orthologues in Xenopus and C. elegans are ing that LKB1 mutations do not lead to FRAP1/mTOR involved in early embryonic development (Su et al., 1996; Watts et al., 2000) and the Lkb1-homologue in Drosophila is required for early anterior–posterior axis *Correspondence: M Sanchez-Cespedes, Lymphomas and Lung formation and epithelial polarity (Martin and St Cancer Laboratory, Spanish National Cancer Centre (CNIO), Spain; Johnson, 2003). The relevance of Lkb1 in early E-mail: [email protected] Received 27 May 2003; revised 18 February 2004; accepted 18 February mammalian development has been demonstrated further 2004; published online 12 April 2004 by the use of Lkb1-deficient mice (Lkb1À/À), which die Expression profile of LKB1 mutant tumors P Fernandez et al 5085 early in gestation due to severe neural-tube defects and to unravel the biological function and to identify the numerous vascular abnormalities (Ylikorkala et al., upstream/downstream targets of the LKB1 protein. 2001). These observations are consistent with the global expression profile of embryonic fibroblasts from Lkb1À/À mice, which have abnormal expression levels of factors Results associated with angiogenesis, extracellular remodeling, cell adhesion and inhibition of ras transformation Characteristics of LKB1 gene alterations in LADs (Bardeesy et al., 2002). Taken together, these results herald an essential role of LKB1 during normal LKB1 gene somatic mutations were detected in five out embryonic development. Interestingly, Lkb1-heterozy- of 19 primary tumors (26%). Interestingly, and in gous mice (Lkb1 þ /À) are viable and mimic the cancer agreement with our previous observations (Sanchez- predisposition phenotype of PJS, developing multiple Cespedes et al., 2002; Carretero et al., 2004), all gastrointestinal hamartomatous polyps and various mutations generated a truncated or incomplete LKB1 tumors such as hepatocellular carcinomas (Miyoshi protein. The mutations detected were a Tyr to Stop at et al., 2002; Nakau et al., 2002). codon 60 (tumor 158), Glu to Stop at codon 120 (tumor Human LKB1 encodes a 436-amino-acid protein with 129), G insertion at codon 279 (tumor 136), Glu to Stop a kinase domain (residues 50–319), similar to the SNF1/ at codon 65 (tumor 272) and a complete deletion of exon AMP-activated family, and a putative 8 (tumor 227) that abrogates part of the kinase domain carboxyl-terminal regulatory domain with a CAAX- of the protein (Figure 1a and b) (Table 1). The box, a consensus sequence for prenylation. Two homozygous nature of all mutations in the primary potential nuclear localization signals are located be- tumors was confirmed by FISH analysis. As detailed in tween amino acids 38–43 and 81–84 (see reference Yoo the methodology section, a tumor was considered et al., 2002, for a review). LKB1 protein has been positive for LOH at the LKB1 locus (located at the detected in both the cytoplasm and nucleus of endogen- short arm of 19, 19p13.3) when more than ous and transfected cells (Nezu et al., 1999; Collins et al., 35% of the cells showed a single signal for the LKB1 2000; Karuman et al., 2001). LKB1-mutant protein probe and two control signals coming from the normal devoid of nuclear localization signal still has the copies of the long arm of . According to ability to inhibit cell growth, suggesting that LKB1 this definition, 10 of the 18 tumors available for FISH cytoplasm retention may be critical for its tumor- analysis (55%) carried LOH at this locus. We also suppression function (Tiainen et al., 2002). Several detected one tumor carrying 6–8-fold amplification at such as p53, BRG1 and LIP1 (Karuman et al., the LKB1 gene. No homozygous deletions spanning 2001; Marignani et al., 2001; Smith et al., 2001) LKB1 gene, or gene promoter hypermethylation were are thought to bind LKB1. LKB1 has been implicated detected in any of the tumors analysed, according to our in different pathways such as p53-dependent apoptosis, data based on FISH analysis and MSP, respectively. p21-mediated cell growth arrest, VEGF-signaling Table 1 lists all LKB1 gene alterations that we have pathway and ATM response to ionizing radiation found to date in primary LADs. As can be seen, (Tiainen et al., 1999; Karuman et al., 2001; Bardeesy mutations are scattered throughout almost the entire et al., 2002; Sapkota et al., 2002). Moreover, LKB1 coding region of LKB1 gene. However, mutations in the activity may be regulated by p90RSK phosphorylation first two exons account for almost half of the tumors, and there is also evidence that LKB1 physiologically whereas no alterations were detected in exon 9. In interacts with the kinase-specific hetero- addition, of the LADs carrying LKB1 mutations, dimer CdC37 and , thereby playing a role in simultaneous KRAS and p53 mutations occurred in regulating the cellular stability of LKB1 (Sapkota et al., four (36%) and six (54%) of the tumors, respectively 2001). (Table 1). We have recently shown that overexpression of ectopic wild-type-LKB1 protein triggers cell growth Differences in gene expression depending on the presence suppression in the A549 LAD cell line. In addition, or absence of LKB1 mutations analysis of global expression patterns in A549 cells following LKB1 ectopic expression revealed the dereg- To determine the global expression profile in our set of ulation of transcripts involved in apoptosis, cell lung tumors, we used 30–35 mg of total RNA extracted adhesion and , including several from 19 primary tumors and equal amounts of reference members of the PTEN/PI3K/AKT pathway (Jimenez RNA. cDNA from lung tumors and from the reference et al., 2003). pool were labeled with Cy5 (red) and Cy3 (green), To investigate the role of LKB1 in lung cancer respectively, and applied in competitive hybridization further, we looked for LKB1 gene inactivation in on the CNIO OncoChip v.1.1c microarray (see Materi- primary LADs and segregated them according to the als and methods section). Fluorescence–intensity ratios presence or absence of LKB1 gene mutations. Global were calculated and gene-expression profiles were expression profiling was performed using cDNA micro- generated for each sample. To assess interexperimental arrays to identify those genes whose expression patterns variability and thus, the reproducibility of the results, significantly varied between lung tumors with and this assay was repeated for one of the primary tumors without LKB1 mutations. Our observations can help (no. 227). A reliable and reproducible result was

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5086

Figure 1 LKB1/STK11 genetic alterations in LADs. (a) Deletion of 866-bp at exon 8 in tumor no. 227 (T) but not in the corresponding normal DNA (N), (b) Another example of point mutation in a lung primary tumor (T) compared with matched normal DNA (N). The arrow indicates the mutation

Table 1 Somatic mutations in sporadic LADs Identification Exon (codon) Mutation type Predicted effect Other mutations (KRAS or p53) Reference

997-JHH Promoter Hypermet. No protein p53 Previous work 1012-JHH 1 (37) cag-tag Gln to Stop KRAS/p53 Previous work 758-JHH 1 (44) aag-tag Lys to Stop p53 Previous work 898-JHH 5 (210) tgc-tga Cys to Stop Previous work 946-JHH 5 (223) gag-tag Glu to Stop p53 Previous work 1122-JHH 6 (279) 1C del. Frameshift KRAS Previous work 136-SP 1 (52-63) 1G ins. Frameshift KRAS Present work 158-SP 1 (60) tac-tag Tyr to Stop Present work 129-SP 2 (120) gaa-taa Glu to Stop KRAS Present work 227-SP 8 866-bp del. Frameshift p53 Present work 272-SP 1 (65) gag-tag Glu to Stop p53 Present work Frequency of LKB1 alterations in primary tumors: 26%

Overall, 10 mutations and one promoter hypermethylation were detected among the 39 primary LADs analysed: 20 tumors from Johns Hopkins Hospital (JHH) (Sanchez-Cespedes et al., 2002) and 19 tumors from Spain (SP) (present work)

obtained in two independent hybridizations (correlation tively, in LKB1-mutation-positive tumors relative to coefficient r ¼ 0.75) (Figure 2a). their expression in LKB1-mutation-negative tumors To identify genes whose expression level may be (Figure 2b). Figure 2b also depicts all genes differen- linked to the LKB1 gene status, we determined the tially expressed in the two groups of tumors. The P- distinctive patterns of gene expression in tumors with values (Po0.01) are indicated in each case. Examination and without LKB1 gene alterations. Our statistical of individual genes from the list reveals that many of analysis (see Materials and methods section) revealed them belong to signal transduction pathways, such as that 34transcripts, including 24known genes, were ARAF1 and FRAP1/mTOR, implicated in the signaling differentially expressed in the five tumors carrying LKB1 through KRAS and PTEN/PI3 K, respectively. Other gene mutations, compared with the 14remaining common deregulated cell processes were cell adhesion/ tumors. Overall, our results showed that 11 and 13 of cell structure (MPP1, FAT2), transcription factors the known genes were up and downregulated, respec- (MEIS2, ATF5, PLAGL1), ubiquitinization (e.g.

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5087

Figure 2 Classification of the gene expression profiles of primary LADs according to the LKB1status. (a) Representative scatter-plot of cDNA microarray analysis showing the correlation of a duplicate (tumor 227). Each of the 4505 transcripts available for the study after processing of the data (see Materials and methods) is represented as a single dot. The two diagonal lines in the graph represent the twofold ratios of expression levels. (b) Supervised clustering (above) and detail on the expression levels of some selected transcripts. Blue and orange bars indicate tumors with and without LKB1 mutations, respectively. Relative increased and decreased expression is depicted in red and green, respectively. The absence of data is indicated in gray (c). Real-time (Taqman) RT–PCR analysis of the genes selected for the cDNA microarray data validation. The analysis for the MEIS2 and AMPD3 genes was performed in 10 LKB1-wild- type and in five LKB1-mutant tumors. All data were standardized by TBP (DCt, cycle difference). The black dot and the error bars represent the mean 7s.d. The fold change was relative to the LKB1-wild-type group. (d) Representative examples of immunohistochemistry for the activated FRAP1/mTOR protein used to validate differences in gene expression among groups. As shown, high levels of FRAP1/mTOR protein in two different lung tumors wild type for LKB1 (original magnification, Â 400) and loss of FRAP1/mTOR protein in a lung tumor carrying an LKB1 gene mutation (original magnification, Â 400)

USP16 and UBE2L3) and, interestingly, two transcripts (TBP). In Figure 2c, it is shown the DCt difference and related to the metabolism of AMP (AMPD3 and the fold change in gene expression are relative to the APRT). LKB1-wild-type group. Both genes demonstrated con- sistent upregulation in LKB1-mutant tumors in agree- ment with the cDNA microarray observations. Validation of some deregulated transcripts: upregulation One of the downregulated transcripts detected in our of MEIS2 and AMPD3 and downregulation of active cDNA microarray analysis was FRAP1/mTOR,an mTOR/FRAP1 protein in LKB1-mutant tumors important mediator of the PI3K/PTEN pathway. To We verified the relative expression of two of these 24 further understand whether downregulation of FRAP1/ genes by real-time (TaqMan) quantitative RT–PCR in mTOR transcript correlated with decreased levels of 15 of the tumors (five LKB1-mutant and 10 LKB1-wild active (phosphorylated) FRAP1/mTOR protein, we type). The selected genes were MEIS2 and AMPD3.As performed immunohistochemistry using antibodies a control gene, we used the TATA-binding protein against the phosphorylated form of FRAP1/mTOR in

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5088 10 of the lung primary tumors with available paraffin- molecular pathway different from that of p53 or KRAS, embedded tissue (two LKB1-mutant and eight LKB1- and so cooperates during the carcinogenetic process of wild-type tumors). Following previously published LADs. criteria, levels of phosphorylated and active FRAP1/ To obtain further insights into the biological role of mTOR protein were categorized as 0, absence; 1, LKB1, we looked for genes that are differentially medium to low; and 2, high levels of protein expression expressed in LADs with and without LKB1 gene (Choe et al., 2003). Immunoreactivity for the active alterations. We previously demonstrated that ectopic form of the FRAP1/mTOR protein was observed in the wild-type LKB1 protein in A549 lung cancer cells leads cytoplasm. Representative examples of the immunos- to changes in the expression of more than 100 taining are shown in Figure 2d. The two LKB1-mutant transcripts, many of them involved in cell proliferation, tumors available for the analysis had absence of apoptosis and cytoskeleton/cell adhesion (Jimenez et al., activated FRAP1/mTOR protein, in agreement 2003). Overall, our current results share similarities to with the strong downregulation of the transcript our previous observations. It is of particular note that observed in the microarrays-based data. On the other LKB1-mutation-positive tumors show significantly dif- hand, among the LKB1-wild-type tumors five carried ferent levels of expression of genes involved in signal moderate-to-high levels and three had no detectable transduction pathways such as FRAP1/mTOR and expression of activated FRAP1/mTOR protein, in ARAF1 compared with those in LKB1-mutation- complete concordance with the cDNA microarray negative tumors. Indeed, FRAP1/mTOR belongs to observations. the PI3K/PTEN signal transduction pathway, which also turned out to be deregulated following ectopic expression of wild-type LKB1 in A549 cells (Jimenez et al., 2003). Given the similarity between Cowden’s Discussion disease (Hanssen and Fryns, 1995) caused by PTEN germ-line mutations and the PJS, it has been postulated The importance of LKB1 gene in the development of that PTEN and LKB1 may act in a common biochem- lung cancer is strongly supported by the common ical pathway (Yoo et al., 2002). However, others have presence of inactivating mutations (Sanchez-Cespedes not found alterations in PTEN or AKT-P protein levels et al., 2002; Carretero et al., 2004). One-third of primary when comparing cells harboring wild-type or mutant LADs harbored mutations at the LKB1 gene, but this LKB1 (Rossi et al., 2002). Our present observations percentage could be underestimated in primary tumors, indicate that activation of the FRAP1/mTOR protein most likely because the masking effect of contaminant kinase is important during lung carcinogenesis of most normal cells makes it difficult to detect large deletions LKB1-wild type but not LKB1-mutant tumors. Whereas and complex gene rearrangements. Thus, the complete the possible connexion between the PI3K/PTEN and exon 8- deletion in the 227-SP primary tumor would LKB1 pathways remains to be clarified, our data may have been masked if long-range PCR had not been help to discriminate those lung cancer patients resistant carried out. The summary of all LKB1 gene alterations to rapamycin or to other mTOR inhibitors, currently in lung tumors reveals a remarkable predominance of being developed by several pharmaceutical and biotech- nonsense/frameshift mutations (only one missense mu- nology companies. tation in a primary LAD had previously been reported) Interestingly, recent reports demonstrate that LKB1 is (Avizienyte et al., 1999), which tend to accumulate in the the upstream regulator of the AMPK protein (AMP- first half of the gene. This mutational pattern is similar activated protein kinase), together with its binding to that found in the PJS patients whose LKB1 mutations partners MO25 and STRAD (Boudeau et al., 2003; are mainly of the frameshift or nonsense type, giving rise Hawley et al., 2003). The AMPK is a sensor of cellular to truncated proteins (Hemminki et al., 1998; Jenne energy and is activated by the increase of the 50-AMP et al., 1998). Together, these observations imply that levels, which accompanies a fall in the ATP : ADP ratio complete abrogation of the LKB1 protein function is (Kemp et al., 1999). In addition to acting as a key required for LKB1-driven carcinogenesis. regulator of cell metabolism, active AMPK also inhibits In spite of the important tumor-suppressor role of cell proliferation (Kemp et al., 1999). Interestingly, we LKB1 gene in the Peutz–Jeghers syndrome and in observed that LKB1-mutant tumors undergo upregula- sporadic LADs, its biological function is still not fully tion of the AMPD3 transcript, involved in the metabo- understood. Several studies have demonstrated that lism of the AMP, which may be related to the lack of ectopic LKB1 can suppress the growth of LKB1- AMPK activation in these subsets of lung tumors. defficient tumor cells (Tiainen et al., 1999; Karuman There were other remarkable differences in gene et al., 2001; Marignani et al., 2001; Jimenez et al., 2003). expression such as MEIS2 (myeloid ecotropic viral Such growth suppression may arise through integration site 1 homologue 2), a transcript upregulated arrest at G1/S, mediated by p21 (Tiainen et al., 1999), in lung tumors carrying LKB1 mutations. This gene through p53-mediated apoptosis (Karuman et al., 2001) encodes a homeobox of the TALE family of proteins, or both. However, the coexistence of KRAS and p53 which are highly conserved transcription regulators, mutations with LKB1 mutations in several lung essential during development (Zhang et al., 2002) and (Sanchez-Cespedes et al., 2002) and pancreatic (Su highly overexpressed in some tumor types such as et al., 1999) tumors suggests that LKB1 functions in a neuroblastoma (Geerts et al., 2003).

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5089 In summary, the biological and functional profile of LKB1 and the adjacent genomic area. For the deletion LKB1 mutant tumors is strongly characterized by an analysis, we used a control probe composed of two over- overactivation of signal transduction pathways, deregula- lapping bacterial artificial (BAC) that map to tion of the levels of adhesion molecules that may promote the long arm of chromosome 19. The three cosmid clones were cell invasion and migration and altered levels of transcripts pooled and labeled by nick translation with Spectrum Green (Vysis, Downer’s Grove, IL, USA). The control clones were involved in various cell functions, including transcriptional pooled and identically labeled with Spectrum Orange (Vysis, regulation, metabolism and ubiquitinization. Downer’s Grove, IL, USA). The frozen tissue was fixed onto the slides with 100% ethanol. The slides were boiled in a pressure cooker with 1 mM EDTA, pH 8.0 for 2 min, and incubated with pepsin at 371C for 30 min. The slides were then Materials and methods dehydrated. The samples and probes were codenatured at 751C for 2 min, then placed overnight for hybridization at 371Cina Patient selection and DNA extraction humid chamber. Slides were washed with 0.4X SSC and 0.3% Tumor and matched normal tissues from 19 patients NP40. diagnosed with lung cancer were provided by the CNIO The FISH analysis was performed by two investigators (SR Tumor Bank Network, CNIO (Madrid, Spain), with the and JCC) with no previous knowledge of the genetic or clinical collaboration of the following hospitals: Hospital Universitar- nature of the samples. Scoring of fluorescence signals was io 12 de Octubre and Hospital Puerta de Hierro in Madrid and made in each sample by counting the number of single copy Hospital Virgen de la Salud, in Toledo. Cases were selected on gene and control probe signals in an average of 180 (163–215) the basis of the availability of frozen tissue and were uniformly well-defined nuclei. The cut-off value for deletions was reviewed and classified as adenocarcinomas. established as the mean plus three times the standard deviation Representative sections from tissue used for DNA extrac- of the proportion of cells with one signal in the three normal tion were stained with H&E. Freshly frozen tissue from tumors nontumoral samples. This value corresponded to 35% of cells. was meticulously dissected to ensure that specimens contained Tumor samples with 435% of cells with one signal were al least 75% tumor cells. Approximately 10–20 mm sections considered to be subject to genomic loss involving the LKB1 were collected from normal and tumor samples and placed in genomic region. 1% SDS/proteinase K (10 mg/ml) at 581C overnight. Digested Hypermethylation at LKB1 gene promoter was assessed by tissue was then subjected to phenol–chloroform extraction and MSP (methylation-specific PCR) as previously described ethanol precipitation. (Esteller et al., 2000).

Screening for mutations, deletions and promoter cDNA microarray analysis hypermethylation Total RNA extraction was carried out using the Trizol reagent In all, 100 ng of genomic DNA were used for exon amplifica- (Life Technologies, Inc., Grand Island, NY, USA), RNeasy kit tion. Cycle-sequencing reaction was performed according to (Qiagen Inc., Valencia, CA, USA) and digested with RNase- the manufacturer’s protocol (Perkin-Elmer, Roche Molecular free DNase I following the manufacturer’s instructions. A System). Sequencing conditions for LKB1 were: 951C30s, small aliquot of RNA was separated for quantification and for 561C 1 min and 721C 1 min for one cycle; and 951C30s,531C quality control. 1 min and 721C 1 min for 34cycles. Sequencing products were In all, 40 mg of total RNA was used for cDNA microarrays separated by electrophoresis through 6% polyacrylamide gels analysis. Fluorescent-labeled cDNA was synthesized and and exposed to a film for 24–48 h. Primers and specific hybridized to the CNIO OncoChip, as previously described conditions used for the screening of LKB1-point mutations (Tracey et al., 2002). The CNIO OncoChip is a cDNA have been described previously (Sanchez-Cespedes et al., microarray that has been specially designed for looking at 2002). Screening of KRAS mutations at codons 12 and 13 genes involved in cancer and includes a core of 2489 cancer- was performed in tumor DNA using a highly sensitive PCR- relevant genes in addition to genes involved in drug-response, RFLP assay (Schimanski et al., 1999). Mutant bands in the tissue-specific and control genes. There are a total of 6386 agarose gel were excised, purified (Ultraclean 15 Kit, MOBIO genes represented by 7237 clones (Tracey et al., 2002). Slides laboratories, CA, USA) and automatically sequenced on an were scanned for Cy3 and Cy5 fluorescence using Scanarray ABI PRISM 3700 DNA Analyser (PE Biosystems). Point 5000 XL (GSI Lumonics Kanata, Ontario, Canada) and mutations at p53, codons 5–9 were analysed by PCR quantified using the GenePix Pro 4.0 program (Axon instru- amplification of a 1800-bp fragment containing codons 5–9: ments Inc., Union City, CA, USA). Hybridizations of the primer forward 50-GCTTTATCTGTTCACTTGTG-30 and human universal reference total RNA from Stratagene (Cy3) reverse 50-CTACAACCAGGAGCCATTGTC-30 followed by against RNA from the different tumors (Cy5) were performed automatic sequencing of individual exons. in duplicate using different target preparations. To screen for deletions in the vicinity of exons 2–8, we used a long-range PCR strategy, as described previously (Sanchez- Data analysis and normalization Cespedes et al., 2002). Briefly, we amplified a 5-Kb fragment containing exons from 2 to 8 of the LKB1 gene using Elongase Fluorescence intensity from each array element was subtracted Enzyme Mix (Invitrogen). PCR products were loaded in 0.7– from the local background. To normalize the data, the Cy5/ 1% agarose gels. To ensure that all LKB1 mutations were Cy3 ratios were global median normalized. Substandard spots homozygous (loss of heterozygosity) and to detect the presence were manually flagged and discarded from further analysis. of homozygous deletions, we performed FISH analysis on The Cy5/Cy3 ratios of the duplicated spots in the array were frozen samples of the tumors. We used three cosmid clones: averaged. All ratio values were semilog transformed (base 2). LLNLR-252D12, 264D11 and 277D11, from the LLNL Inconsistent duplicates were discarded whereas uniform Human Chromosome library (MRC UK HGMP Resource duplicated spots were averaged. Gene profiles with less than Centre, Babraham, Cambridge, UK). These clones covered the 70% of available data were excluded from further analysis.

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5090 This process yielded 4503 transcripts that were suitable for (Applied Biosystem, Foster City, CA, USA) to perform real- bioinformatic analysis. time quantitative PCR analysis (Taqman Technology) on the MEIS2 and AMPD3 transcripts. Briefly, 1 mg of RNA was Bioinformatic analysis reverse transcribed (Reverse Transcription System, Promega, WI, USA) using a random hexamer primer. For Taqman The SOTA clustering program (http://bioinfo.cnio.es/cgi-bin/ PCR assay, cDNA from 20 ng of total RNA was used to tools/clustering/sotarray) was used for hierarchical clustering measure gene expression with an ABI Prism 7700 sequence analysis and the GeneCluster/Treeview program was used to detection system and the reagent TaqMan Universal view gene expression (Herrero et al., 2001). To identify the PCR Master Mix (Applied Biosystems). Target cDNAs genes that are important for distinguishing subgroups defined were amplified in duplicate using the following cycling by the presence of LKB1 mutations, we carried out an conditions: 10 min at 951C and then 40 cycles of amplification ANOVA test with 50 000 random permutations for P-value (951C for 30 s and 601C for 30 s). The TBP (Human TBP computation. Genes with unadjusted Po0.01 were considered Pre-Developed Taqman Assay Reagents, Applied Biosystems) the best candidates for identifying differential expression was used to normalize variations in the input cDNA. The among subgroups. These procedures are implemented in threshold cycle (Ct) was determined and the relative gene http://bioinfo.cnio.es/cgi-bin/tools/multest/multest.cgi. expression was calculated as follows: fold change ¼ 2ÀD(DCt), where DCt ¼ CttargetÀCtTBP (cycle difference) and D(DCt) ¼ FRAP1/mTOR immunohistochemistry CtLKB1-mutantÀCtLKB1-wild type. Immunohistochemical staining of FRAP1/mTOR was per- formed on 5-mm sections from paraffin-embedded tumors. Phospho-mTOR (Ser2448) polyclonal antibody (Cell Signal- Acknowledgements ing, Technology, Berkeley, MA, USA) was diluted at 1 : 50 and We thank all collaborators at the Tumor Bank Network, after incubation for 60 min at room temperature, immunode- especially Alicia Maroto for her meticulous technical support tection was performed with EnVision-HRP (DAKO, Copen- in providing normal and lung tumor tissues. We also thank hagen, Denmark) and peroxidase activity was developed using Ana Diez, Maria Jesus Acun˜ a, Raquel Pajares and Carmen 3,3-diaminobenzidene (DAB) chromogen as substrate. Sec- Martin from the Immunohistochemistry and Cytogenetics tions were counterstained with hematoxylin. A positive control Units at the CNIO for technical help. The research was was included with each batch of staining to ensure consistency supported by a grant from the Spanish Ministerio de Ciencia y between consecutive runs. Tecnologı´ a (SAF2002-01595) and by a grant from the Comunidad de Madrid (CAM 08.1/0032/2003 1). J Carretero is a postdoctoral fellow of the Excmo. Ayuntamiento Quantitative RT–PCR for MEIS2 and AMPD3 de Madrid and MS-C is supported by the Ramon y To validate the differences in gene expression, we used Cajal Program from the Ministerio de Ciencia y Tecnologı´ a, commercially available Assay-on Demand probe-primer sets Spain.

References

Avizienyte E, Loukola A, Roth S, Hemminki A, Tarkkanen Hanssen AM and Fryns JP. (1995). J. Med. Genet., 32, 117– M, Salovaara R, Arola J, Butzow R, Husgafvel-Pursiainen 119. K, Kokkola A, Jarvinen H and Aaltonen LA. (1999). Am. J. Hawley SA, Boudeau J, Reid JL, Mustard KJ, Udd L, Makela Pathol., 154, 677–681. TP, Alessi DR and Hardie DG. (2003). J. Biol., 2, 28. Bardeesy N, Sinha M, Hezel A, Signoretti S, Hathaway NA, Hemminki A, Markie D, Tomlinson I, Avizienyte E, Roth S, Sharpless NE, Loda M, Carrasco DR and De Pinho RD. Loukola A, Bignell G, Warren W, Aminoff M, Hoglund P, (2002). Nature, 419, 162–167. Jarvinen H, Kristo P, Pelin K, Ridanpaa M, Salovaara R, Boudeau J, Baas AF, Deak M, Morrice NA, Kieloch A, Toro T, Bodmer W, Olschwang S, Olsen AS, Stratton MR, Schutkowski M, Prescott AR, Clevers HC and Alessi DR. de la Chapelle A and Aaltonen LA. (1998). Nature, 18, (2003). EMBO J., 22, 5102–5114. 184–187. Carretero J, Medina PP, Pio R, Montuenga LM and Sanchez- Herrero J, Valencia A and Dopazo J. (2001). Bioinformatics, Cespedes M. (2004). Oncogene, (Advance online publication, 17, 126–136. Jeghers H, McKusic VA and Katz KH. (1949). N. Engl. J. 15 March 2004). Med., 241, 1031–1036. Choe G, Horvath S, Cloughesy TF, Crosby K, Seligson D, Jenne DE, Reimann H, Nezu J, Friedel W, Loff S, Jeschke R, Palotie A, Inge L, Smith BL, Sawyers CL and Mischel PS. Mulle O, Back W and Zimmer M. (1998). Nature Genet., 18, (2003). Cancer Res., 63, 2742–2746. 38–44. Collins SP, Reoma JL, Gamm DM and Uhler MD. (2000). Jimenez AI, Fernandez P, Dominguez O, Dopazo A and Biochem. J., 345, 673–680. Sanchez-Cespedes M. (2003). Cancer Res., 63, 1382–1388. Esteller M, Avizienyte E, Corn PG, Lothe RA, Baylin SB, Karuman P, Gozani O, Odze RD, Zhou XC, Zhu H, Shaw R, Aaltonen LA and Herman JG. (2000). Oncogene, 19, 164– Brien TP, Bozzuto CD, Ooi D , Cantley LC and Yuan J. 168. (2001). Mol. Cell, 7, 1307–1319. Geerts D, Schilderink N, Jorritsma G and Versteeg R. (2003). Kemp BE, Mitchelhill KI, Stapleton D, Michell BJ, Chen ZP Cancer Lett., 197, 87–92. and Witters LA. (1999). Trends Biochem. Sci., 24, 22–25. Giardiello FM, Welsh SB, Hamilton SR, Offerhaus GJ, Marignani P, Kanai F and Carpenter CL. (2001). J. Biol. Gittelsohn AM, Booker SV, Krush AJ, Yardley JH and Chem., 276, 32415–32418. Luk GD. (1987). N. Engl. J. Med., 316, 1511–1514. Martin SG and St Johnson D. (2003). Nature, 421, 379–384.

Oncogene Expression profile of LKB1 mutant tumors P Fernandez et al 5091 Miyoshi H, Nakau M, Ishikawa TO, Seldin MF, Oshima M Smith DP, Spicer J, Smith A, Swift S and Ashworth A. (1999). and Taketo MM. (2002). Cancer Res., 62, 2261–2266. Hum. Mol. Genet., 8, 1479–1485. Nakau M, Miyoshi H, Seldin MF, Imamura M, Oshima M Su JY, Erikson E and Maller JL. (1996). J. Biol. Chem., 271, and Taketo MM. (2002). Cancer Res., 62, 4549–4553. 14430–14437. Nezu J, Oku A and Shimane M. (1999). Biochem. Biophys. Su GH, Hruban RH, Bansal RK, Bova GS, Tang DJ, Shekher Res. Commun., 261, 750–755. MC, Westerman AM, Entius MM, Goggins M, Yeo CJ and Rossi DJ, Ylikorkala A, Korsisaari N, Salovaara R, Luukko Kern SE. (1999). Am. J. Pathol., 154, 1835–1840. K, Launonen V, Henkemeyer M, Ristimaki A, Aaltonen LA Tiainen M, Vaahtomeri K, Ylikorkala A and Makela TP. and Makela TP. (2002). Proc. Natl. Acad. Sci. USA, 99, (2002). Hum. Mol. Genet., 11, 1497–1504. 12327–12332. Tiainen M, Ylikorkala A and Makela TP. (1999). Proc. Natl. Acad. Sci. USA, 96, 9248–9251. Sapkota GP, Deak M, Kieloch A, Morrice N, Goodarzi AA, Tracey L, Villuendas R, Ortiz P, Dopazo A, Spiteri A, Smythe C, Shiloh Y, Lees-Miller SP and Alessi DR. (2002). Lombardia L, Rodrı´ guez-Peralto JL, Ferna´ ndez-Herrera J, Biochem. J. , 368, 507–516. Herna´ ndez A, Fraga M , Dominguez O, Herrero J, Alonso Sapkota GP, Kieloch A, Lizcano JM, Lain S, Arthur JS, MA, Dopazo J and Piris MA. (2002). Am. J. Pathol., 161, Williams MR, Morrice N, Deak M and Alessi DR. (2001). J. 1825–1837. Biol. Chem., 276, 19469–19482. Watts JL, Morton DG, Bestman J and Kemphues KJ. (2000). Sanchez-Cespedes M, Parrella P, Esteller M, Nomoto S, Trink Development, 127, 1467–1475. B, Engles JM, Westra WH, Herman JG and Sidransky D. Ylikorkala A, Rossi DJ, Korsisaari N, Luukko K, Alitalo K, (2002). Cancer Res., 62, 3659–3662. Henkemeyer M and Makela TP. (2001). Science, 293, 1323– Schimanski CC, Linnemann U and Berger MR. (1999). Cancer 1326. Res., 59, 5169–5175. Yoo LI, Chung DC and Yuan J. (2002). Nat. Rev., 2, 529–535. Smith DP, Rayter SI, Niederlander C, Spicer J, Jones CM and Zhang X, Friedman A, Heaney S, Purcell P and Maas RL. Ashworth A. (2001). Hum. Mol. Genet., 10, 2869–2877. (2002). Genes Dev., 16, 2097–2107.

Oncogene