Epigenomic Alterations and Gene Expression Profiles in Respiratory

Home , ARHGDIB, ARMCX6, Caballero (cigarette), EMP3, Lexington (cigarette), TSPAN7

Oncogene (2010) 29, 3650–3664 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc ORIGINAL ARTICLE Epigenomic alterations and gene expression proﬁles in respiratory epithelia exposed to cigarette smoke condensate

F Liu1, JK Killian2, M Yang1, RL Walker2, JA Hong1, M Zhang1, S Davis2, Y Zhang1, M Hussain1,SXi1, M Rao1, PA Meltzer2 and DS Schrump1

1Thoracic Oncology Section, Surgery Branch, Center for Cancer Research, National Cancer Institute, Bethesda, MD, USA and 2Genetics Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA

Limited information is available regarding epigenomic events Introduction mediating initiation and progression of tobacco-induced lung cancers. In this study, we established an in vitro system to Mounting evidence implicates aberrant expression/ examine epigenomic effects of cigarette smoke in respiratory activity of epigenetic regulators of gene expression in epithelia. Normal human small airway epithelial cells and the initiation and progression of lung cancers, the cdk-4/hTERT-immortalized human bronchial epithelial cells majority of which are directly attributable to cigarette (HBEC) were cultured in normal media with or without smoking (Lee et al., 2005; D’Alessio and Szyf, 2006; cigarette smoke condensate (CSC) for up to 9 months under Lin et al., 2007; Schrump et al., 2007). For example, potentially relevant exposure conditions. Western blot increased DNA methyltransferase (DNMT) expression analysis showed that CSC mediated dose- and time- coincides with progression to malignancy in murine dependent diminution of H4K16Ac and H4K20Me3, while pulmonary adenomas induced by tobacco carcinogens increasing relative levels of H3K27Me3; these histone (Belinsky et al., 1996). Over-expression of DNMT1 and alterations coincided with decreased DNA methyltransferase DNMT3b, as well as methyl-binding domain 2, correlates 1 (DNMT1) and increased DNMT3b expression. Pyrose- with hypermethylation of tumor suppressor genes in quencing and quantitative RT–PCR experiments revealed tobacco-induced lung cancers, and diminished survival of time-dependent hypomethylation of D4Z4, NBL2, and patients with these neoplasms (Kim et al., 2006; Lin et al., LINE-1 repetitive DNA sequences; up-regulation of H19, 2007; Xing et al., 2008). Aberrant activities of histone IGF2, MAGE-A1, and MAGE-A3; activation of Wnt lysine methyl transferases and histone deacetylases signaling; and hypermethylation of tumor suppressor genes enhance oncogenic transformation of bronchial epithelial such as RASSF1A and RAR-b, which are frequently cells (Watanabe et al., 2008), and induce global alterations silenced in human lung cancers. Array-based DNA methyla- in the histone code, which correlate with advanced stage tion profiling identified additional novel DNA methylation of disease and poor prognosis in lung cancer patients targets in soft-agar clones derived from CSC-exposed (Barlesi et al., 2007; Van Den et al., 2008; Seligson et al., HBEC; a CSC gene expression signature was also identified 2009). Over-expression of polycomb proteins such as Bmi in these cells. Progressive genomic hypomethylation and 1 and Ezh2 facilitates epigenetic silencing of tumor locoregional DNA hypermethylation induced by CSC suppressor genes, and enhances stem cell signaling in coincided with a dramatic increase in soft-agar clonogeni- lung cancer cells (Dovey et al., 2008; Vrzalikova et al., city. Collectively, these data indicate that cigarette smoke 2008; Hussain et al., 2009; McCabe et al.,2009). induces ‘cancer-associated’ epigenomic alterations in cul- Epigenetic alterations during malignant transforma- tured respiratory epithelia. This in vitro model may prove tion seem analogous to those regulating gene expression useful for delineating early epigenetic mechanisms regulating during gametogenesis and stem cell development (Simp- gene expression during pulmonary carcinogenesis. son et al., 2005; Mathews et al., 2009). For example, Oncogene (2010) 29, 3650–3664; doi:10.1038/onc.2010.129; tumor suppressor genes such as p16 and p19/ARF, published online 3 May 2010 which mediate replicative senescence and apoptosis in response to oncogene signaling (Agherbi et al., 2009), Keywords: tobacco smoke; lung cancer; epigenetics; are frequently observed to be silenced in cancer cells respiratory epithelial cells exhibiting coordinate de-repression of germ-line re- stricted genes such as NY-ESO-1 and MAGE family members, several of which physically interact and suppress p53-mediated apoptosis (Cho et al., 2006; Correspondence: Dr DS Schrump, Thoracic Oncology Section, Monte et al., 2006; Yang et al., 2007). These observa- Surgery Branch, Center for Cancer Research, National Cancer tions suggest that epigenetic perturbations in lung Institute, 10 Center Drive, Building 10, Room 4-3940, Bethesda, MD cancer cells are not merely manifestations of random 20892-1201, USA. E-mail: [email protected] stochastic events, but instead reflect strong selective Received 14 September 2009; revised 3 February 2010; accepted 17 pressure to reactivate/maintain stem cell gene expression March 2010; published online 3 May 2010 during multistage pulmonary carcinogenesis. Epigenetics of tobacco smoke exposure FLiuet al 3651 Despite the unequivocal association between cigarette CSC (1%) mediated time-dependent decreases in smoking and lung cancer, the epigenetic mechanisms by H4K16Ac, as well as H4K20me3 levels; reduced levels which tobacco smoke initiates and promotes pulmonary of H4K16Ac—but not H4K20Me3—were observed in carcinogenesis have not been fully delineated. In HBEC 5 days after initiation of CSC exposure. particular, epigenetic events associated with initiation Decreased H4K20Me3 levels were evident in HBEC 1 of tobacco-induced lung cancers have not as yet been month after initiation of CSC exposure. More pro- elucidated. This study was undertaken to characterize longed exposure to 1% CSC virtually abolished epigenomic alterations in cultured human respiratory H4K16Ac, and dramatically reduced H4K20Me3 levels epithelial cells mediated by cigarette smoke. without significantly diminishing total H4 levels in HBEC; these histone alterations coincided with a 2.6- fold relative increase in H3K27Me3, an 80% reduction in DNMT1, as well as B2–3-fold increase in EZH2 and Results DNMT3b levels, respectively. Growth inhibitory effects of cigarette smoke condensate in cultured cells DNA methylation changes mediated by CSC Preliminary experiments were initiated to examine the Additional experiments were performed to ascertain effects of cigarette smoke condensate (CSC) in cultured whether CSC exposure altered global DNA methylation respiratory epithelia and lung cancer cells to define status in HBEC. Initial pyrosequencing experiments appropriate exposure conditions for subsequent studies. focused on NBL2-seq3 and D4Z4 DNA repeats, as well As shown in Figure 1a, CSC mediated dose-dependent as LINE-1 elements, as these have been shown to be growth inhibition, which coincided with morphologic demethylated in cancer cells exhibiting decreased levels changes in cultured respiratory epithelia; after CSC of H4K16Ac and H4K20Me3 (Fraga et al., 2005). exposure, cdk-4/h-Tert immortalized human bronchial Preliminary analysis showed modest demethylation of epithelial cells (HBEC), and to a lesser extent normal these sequences in untreated immortalized HBEC human small airway epithelial cells (SAEC) became less relative to primary cultures of SAEC or NHBE cells, polygonal and appeared more elongated and rectangu- although not to the extent observed in A549 lung cancer lar in shape. Interestingly, under these exposure condi- cells (Figure 2a). Subsequent experiments revealed a tions, the effects of CSC were less apparent in fully dose-dependent demethylation of NBL2 and D4Z4 transformed A549 lung cancer cells. subtelomeric DNA repeats, and to a lesser extent, LINE-1 elements in HBEC after 5-month continuous ‘Cancer-associated’ histone alterations mediated by CSC CSC exposure (Figure 2b). Malignant transformation is associated with global loss Quantitative RT–PCR experiments were performed of monoacetylated H4K16 as well as decreased to ascertain whether hypomethylation of the aforemen- H4K20me3 levels (Fraga et al., 2005; Van Den et al., tioned repetitive DNA sequences coincided with activa- 2008). As such, western blot experiments were per- tion of imprinted loci, as well as cancer-testis–X formed to ascertain whether CSC exposure could induce chromosome (CT-X) genes that are frequently de- these histone changes in respiratory epithelial cells. repressed in primary lung cancers through epigenetic Briefly, SAEC and HBEC were cultured in normal mechanisms. As shown in Figure 2c (left panel), CSC media with or without CSC at various concentrations mediated dose-dependent up-regulation of IGF2 and and exposure durations. As shown in Figure 1b (upper H19 imprinted loci, as well as several CT-X genes, panel), within 24 h of exposure, CSC mediated a dose- including MAGE-A1 and MAGE-A3. Interestingly, dependent decrease in H4K16Ac levels, without NY-ESO-1 (cancer-testis antigen 1), a CT-X gene appreciably changing total H4 expression in HBEC. A frequently de-repressed in lung cancer cells (Schrump similar phenomenon was observed after 24 or 48 h CSC et al., 2007), did not seem to be activated in HBEC by exposure. Densitometry analysis revealed no appreci- CSC under these exposure conditions. able change in H4K20Me3 levels under these exposure Additional experiments were performed to determine conditions. Additional analysis using indirect immuno- whether enhanced expression of IGF2 and H19 was due fluorescence techniques (Figure 1b, lower panel) to mono-allelic up-regulation of these loci, or loss of confirmed results of aforementioned western blot imprinting. Genomic sequencing (Figure 2c, right panel) experiments. Short-term CSC exposure mediated dose- revealed polymorphisms that could distinguish maternal dependent decreases in H4K16Ac levels without appear- and paternal IGF2 and H19 alleles in HBEC; additional ing to alter H4K20Me3 levels in short-term normal pyrosequencing experiments indicated that only the human bronchial epithelial (NHBE) cells as well as paternal IGF2 and the maternal H19 alleles were SAEC (Figure 1c), suggesting that the acute effects of expressed in untreated as well as CSC-exposed HBEC. CSC exposure on the histone code were not unique to These data suggested that increased IGF2 and H19 immortalized HBEC. expression in HBEC after 5-month CSC exposure was Additional experiments were performed to examine due to mono-allelic up-regulation of the normally the effects of prolonged CSC exposures in immortalized expressed alleles, rather than loss of imprinting. These HBEC. Representative results of this analysis are findings are consistent with the data reported by Kaplan depicted in Figure 1d. Relatively low concentrations of et al. (2003) after exposure of NHBE to cigarette smoke.

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3652 SAEC HBEC A549 SAEC 20X 20X 20X CSC 1 0% Control 0.31% 0.75 NHBE 72h SAEC 72h 0.63% 0.5% CSC 0.5 1.25% CSC 0% 0.5% 1% 0% 0.5% 1% 0.25 H4K16Ac 1.0% CSC 0 1.0 1.0 0.5 1.0 0.9 0.5 0123456 Day H4K20Me3 HBEC A549 CSC CSC 1.00.9 1.0 1.00.9 1.0 1.5 0% 1.5 0% 0.31% 0.31% Total H4 1 0.63% 1 0.63% 1.25% 1.25% OD OD OD 0.5 0.5 5 days 1 month 5 months CSC 0%1% 0% 1% 0% 1% 0 0 0123456 0 123456 H4K16Ac Day Day 1.0 0.4 1.00.4 1.0 0.1 H4K20Me3 HBEC 1.0 1.2 1.0 0.6 1.0 0.1 4h 24h 48h Total H4 CSC 0% 1%5%0% 1% 5% 0% 1% 5% H3K27Me3 H4K16Ac 1.0 1.1 1.0 1.1 1.0 2.6 1.0 0.8 0.6 1.0 0.7 0.3 1.0 0.8 0.6 Total H3 H4K20Me3 1.0 0.9 1.1 1.0 0.9 0.9 1.0 1.4 1.1 DNMT1 Total H4 1.0 1.2 1.0 0.8 1.0 0.2 DNMT3B H4K16Ac H4K20Me3 1.0 1.1 1.01.2 1.0 2.9

HBEC EZH2 control 1.0 1.2 1.0 1.2 1.0 2.1 Actin 48h 1% CSC

Figure 1 (a) Morphology and proliferation of SAEC, HBEC, and A549 cells exposed to CSC. Dose-dependent alterations in cell morphology seemed to coincide with growth inhibition after CSC exposure. CSC-mediated morphologic changes and growth inhibition in A549 lung cancer cells were considerably less than those observed in short-term SAEC or immortalized HBEC. (b) Upper panel: Western blot analysis of H4K16 acetylation and H4K20 trimethylation in HBEC after exposure to CSC. A dose-dependent decrease in H4K16 acetylation was observed within 4 h of CSC exposure; this phenomenon persisted after exposures of 24 and 48 h. No appreciable change in H4K20 trimethylation was observed after short-term CSC exposure. Lower panel: Immunoﬂourescence analysis of acetylated H4K16 and trimethylated H4K20 expression in HBEC exposed to CSC for 3 days. The results reveal a decrease in H4K16 acetylation in HBEC without signiﬁcant reduction in H4K20 trimethylation in HBEC exposed to 1% CSC. These results are consistent with western blot data in upper panel. (c) Western blot analysis of H4K16Ac, and H4K20Me3 and total H4 levels in cultured NHBE and SAEC exposed to CSC for 72 h. Short-term CSC exposure seemed to decrease H4K16 acetylation, without altering H2K20Me3 in both types of normal respiratory epithelia. (d) Western blot analysis of histone alterations as well as DNMT and EZH2 expression in HBEC exposed to CSC. Densitometry analysis revealed time-dependent decreases in H4K16 acetylation and H4K20 trimethylation without appreciable alterations in total histone H4 levels. These alterations coincided with a relative increase in H3K27 trimethylation as well as a modest increase in EZH2, the polycomb protein mediating this repressive mark. These phenomena, which were evident after 5-month CSC exposure, coincided with a decrease in DNMT1:DNMT3b protein ratios.

Additional qRT–PCR and pyrosequencing experi- diminution of RASSF1A as well as Dickkopf-1 (Dkk-1) ments were performed to examine whether CSC expression; repression of RASSF1A, but not Dkk-1, exposures sufﬁcient to induce global DNA demethyla- coincided with promoter hypermethylation. The results tion also mediated repression of tumor suppressor genes pertaining to Dkk-1 repression in HBEC after CSC typically silenced in human lung cancers by DNA exposure were consistent with recent data showing that hypermethylation mechanisms. As shown in Figure 2d, cigarette smoke induces polycomb-mediated repression CSC exposure did not appear to alter expression or of Dkk-1 in lung cancer cells (Hussain et al., 2009). methylation status of p16, E-cadherin, DAP kinase, or Interestingly, CSC mediated dose-dependent hyper- MGMT. However, CSC mediated dose-dependent methylation without appreciably diminishing expression

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3653 * P<0.01 * P<0.05 ** P<0.01 100 * 100 * * * ** ** 80 * 80 * ** 60 60

40 40 % Methylation % Methylation 20 20

0 0 NBL2D4Z4 Line-1 NBL2D4Z4 Line-1 NHBE SAEC HBEC A549 Control 0.5% CSC 5 mo 1% CSC 5 mo

40 IGF2 H19 Genotype mRNA Genotype mRNA 30 HBEC control A/G G/G C/T T/T 0.5% CSC 5 mo A/G G/G C/T T/T 20 1% CSC 5 mo A/G G/G C/T T/T A549 A/G A/A C/C C/C

Fold change 10 H2028 A/G A/G C/T C/T

0 IGF2H19 MAGEA1 MAGEA3 NY-ESO-1 actin

Control 0.5% CSC 5 mo 1% CSC 5 mo

2.0 1.6 1.2 0.8 0.4 Fold change 0.0 p16 E-cad DAPK MGMT RASSF1A DKK1 RARβ

Control 0.5% CSC 5 mo 1% CSC 5 mo 50 40 30 20 10 % Methylation 0 p16 E-cad DAPK MGMT RASSF1A DKK1 RARβ control 0.5% CSC 5 mo 1% CSC 5 mo Figure 2 Pyrosequencing analysis of methylation status in repetitive DNA sequences before and after CSC exposure. (a) Pyrosequencing analysis of NBL2, D4Z4, and LINE-1 methylation in untreated respiratory epithelia and A549 lung cancer cells. Relative to SAEC and NHBE, immortalized HBEC exhibited somewhat reduced methylation in these DNA repeats. This phenomenon was even more pronounced in A549 lung cancer cells. (b) Pyrosequencing analysis of NBL2, D4Z4, and LINE-1 sequences in HBEC exposed to CSC for 5 months. A dose-dependent decrease in DNA methylation was observed, suggestive of global DNA demethylation mediated by cigarette smoke. (c) Left panel: Quantitative RT–PCR and correlative RT–PCR analysis of imprinted and cancer-testis gene expression in HBEC after CSC exposure (upper and lower panels, respectively). The results of RT–PCR depicted as fold change relative to untreated HBEC. CSC dose-dependently increased expression of H19 and IGF-2 imprinted loci, as well as MAGE-A1 and MAGE-A3. No increase in NY-ESO-1 expression was observed. The results of RT–PCR analysis were consistent with qRT–PCR data. Right panel: Pyrosequencing analysis of genomic DNA and cDNA pertaining to IGF2 and H19 in control HBEC as well as HBEC exposed to CSC for 5 months. Genotyping revealed polymorphisms that could distinguish between maternal and paternal alleles of IGF2 and H19. After prolonged CSC exposure, only the maternally imprinted IGF2 and the paternally imprinted H19 loci were expressed. These results suggest that enhanced expression of these loci was attributable to mono-allelic up-regulation rather than loss of imprinting. (d) qRT–PCR with representative RT–PCR, and pyrosequencing analyses of tumor suppressor gene loci in HBEC after prolonged CSC exposure (upper and lower panels, respectively). A progressive, dose-dependent increase in promoter methylation was observed in RASSF1A and RAR-b promoter regions. No apparent hypermethylation was observed in several tumor suppressor genes such as p16, E-cadherin, DAP kinase, and MGMT, which are frequently silenced in tobacco-associated lung cancers by epigenetic mechanisms. DAP kinase expression in untreated and CSC-exposed HBEC was exceedingly low. Dkk-1 was not methylated, but was repressed, a phenomenon possibly attributable to polycomb repressor complexes (Hussain et al., 2009). Diminution of RASSF1A expression coincided with hypermethylation of the RASSF1A promoter. Hypermethylation of RAR-b seemed insufﬁcient to silence this gene under exposure conditions used for these experiments. of RAR-b, which is frequently silenced through Activation of Wnt signaling by CSC epigenetic mechanisms in tobacco-associated lung can- As CSC exposure silenced Dkk-1, additional experi- cers (Toyooka et al., 2003, 2006). ments were performed using focused qRT–PCR arrays

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3654 Gene Fold Change WNT10A 135.5 WNT6 28.9

-0.18 WNT5A 28.7 DKK1 18 fold FOXN1 18.4 -1.18 WNT3A 10.7 -2.18 WNT2 10.5 TCF7 4 fold -3.18 WNT5A 29 fold RHOU 7.9

HBEC normal TLE2 5.6 -4.18 Log10 (Count Group) WNT10A 136 fold TCF7 4.4 -5.18 SFRP1 -3.2 WNT7A -4.6 -6.18 -5.18 -4.18 -3.18 -2.18 -1.18 -0.18 Log10 (Group 1) WISP1 -9.0 HBEC 1.0% CSC – 5 months WNT16 -11.7 WNT5B -14.4 DKK1 -18.2

Gene Fold Change

UGT1A4 101.8 EPHX2 80.5 -0.05 SERPINE1 5 fold CYP1A1 69.6 -1.05 CCL4 19.7

-2.05 Casp10 6 fold HMOX1 10.0 PTGS1 8.7 -3.05 CYP1A1 70 fold FMO1 8.5 HBEC normal -4.05 Log10 (Control GrouP) UGT1A4 102 fold CCL3 6.4 -5.05 GSR 6.2 CYP7A1 5.7 -6.05 -5.05 -4.05 -3.05 -2.05 -1.05 -0.05 GDF15 5.7 Log10 (Group 1) HBEC 1.0% CSC – 5 months MDM2 5.3 SERPINE1 -5.2 CASP10 -5.8

Figure 3 (a) The results of Wnt superarray analysis showing marked up-regulation of a variety of Wnt ligands and down-regulation of antagonists of Wnt signaling including SFRP1 and Dkk-1. (b) The results of superarrays showing CSC-mediated changes in expression levels of numerous genes regulating cellular response to oxidative stress.

to further examine the effects of cigarette smoke on Wnt signaling, including FOXN1 and TCF7. In addition, signaling in HBEC. The results of this analysis are CSC exposure markedly decreased expression of several summarized in Figure 3. Prolonged CSC exposure Wnt antagonists, including SFRP1 and Dkk-1(Winn dramatically enhanced expression of several Wnt ligands et al., 2006; Katoh and Katoh, 2007). Whereas implicated in invasion and metastasis of cancer cells as epigenetic mechanisms may have contributed to en- well as maintenance of cancer stem cells, including Wnt hanced Wnt signaling perhaps through repression of 10a, Wnt 6, Wnt 5a, and Wnt 2 (Huang et al., 2005; Le SFRP1 and Dkk1, which are known to be silenced in et al., 2005; Katoh and Katoh, 2007; Huang and He, human lung cancers by DNA methylation and poly- 2008), and up-regulated downstream targets of Wnt comb repressor complexes (Licchesi et al., 2008; Hussain

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3655 800 Control CSC 5 months CSC 9 months

600

400

200 Number of clones

0 Control 1.0% CSC- 1.0% CSC- 5 months 9 months

* * P<0.001 80 * Control * 1.0% CSC 6 months 60 1.0% CSC 9 months 40

20 % methylation 0 NBL2 D4Z4 Line -1 P16 DAPK E-Cad CDH13 RASSF1A RAR

90 80 Spontaneous clone 1-5 70 1.0% CSC 9 months clone 1-6 60 50 40 30

% methylation 20 10 0 NBL2 D4Z4 P16 DAPK E-Cad CDH13 RASSF1A RAR Figure 4 (a, b) Clonogenic potential of HBECs after prolonged CSC exposure. Soft-agar assay showing a time-dependent increase in clonogenic potential of HBEC after prolonged CSC treatment. (c) Pyrosequencing analysis of DNA methylation within repetitive DNA sequences and tumor suppressor genes in HBEC after prolonged CSC exposure. A progressive time-dependent increase in RASSF1A and RARb promoter methylation was observed without an apparent increase in methylation of p16, DAPK, E-cadherin, or CDH13. Increased clonogenic potential also coincided with a time-dependent decrease in methylation status within DNA repeats, particularly NBL2 and D4Z4. (d) Pyrosequencing analysis of DNA methylation in soft-agar clones emerging spontaneously from untreated HBEC and representative soft-agar clones derived from HBEC exposed to CSC for 9 months. The CSC-exposed clones showed markedly diminished methylation status within NBL2 and D4Z4, findings which were consistent with results observed in bulk cultures. Once again, there was no change in methylation status within p16, DAPK, E-cadherin, H-cadherin, or MGMT. However, clones derived from CSC-exposed HBEC exhibited markedly increased methylation status of RASSF1A and RARb. These data suggest that CSC induced alterations, which contributed to enhanced clonogenicity rather than selecting for outgrowth of cells exhibiting aberrant methylation profiles in bulk untreated HBEC. et al., 2009), the data do not exclude the possibility that and D4Z4, and hypermethylation of RASSF1A and perturbation of Wnt signaling was in part a manifesta- RAR-b, but not p16, DAPK, E-cadherin, or CDH13 tion of cellular response to oxidative stress (Hoogeboom (Figure 4c). Five soft-agar clones arising spontaneously and Burgering, 2009) (Figure 3b). from control HBEC and B30 clones derived from HBEC exposed to CSC for 9 months were expanded for further analysis. Pyrosequencing experiments indicated Clonogenicity and tumorigenicity of HBEC after CSC that methylation profiles in spontaneous and CSC- exposure derived clones reflected those observed in untreated and Additional experiments were performed to ascertain CSC-exposed cell populations, respectively (Figure 4d). whether CSC exposure was sufficient to transform The fact that consistent methylation profiles were HBEC. As shown in Figures 4a and b, CSC exposure observed in CSC-derived relative to control clones dramatically increased soft-agar clonogenicity of HBEC strongly suggests that enhanced clonogenic potential of in a time-dependent manner, a phenomenon that CSC-treated HBEC was a manifestation of selection coincided with progressive demethylation of NBL2 pressure mediated by CSC on the HBEC epigenome,

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3656 rather than outgrowth of cells exhibiting aberrant findings strongly suggest that DNA methylation DNA methylation already present in untreated pooled changes induced by prolonged CSC treatment were HBEC. Despite the dramatic increase in soft-agar quite stable and heritable. Several CpG targets, includ- clonogenicity of CSC-treated HBEC, no tumors were ing RASSF1A, LOX, and CALCA, that exhibited DNA observed in 100 nude or NOD.SCID\IL-2Rg null mice hypermethylation in HBEC after CSC exposure were inoculated in contralateral flanks with control HBEC also noted to be methylated in cultured lung cancer and one of the six CSC-derived soft-agar clones (data lines; in contrast, other targets such as MEG-3, not shown). HOXA9, TFF-2, and GNAS that were methylated in several cultured as well as primary lung cancers did not appear to be methylated in HBEC after CSC exposure Array analysis of DNA methylation and gene expression (data not shown). These results suggest that either more mediated by CSC prolonged CSC exposures, or cigarette smoke compo- Illumina array techniques were used to more compre- nents not present in the condensates are necessary to hensively examine DNA methylation and gene expres- further shift DNA methylation profiles in HBEC toward sion profiles in HBEC exposed to CSC. Briefly, sodium those typically observed in cultured non-small cell lung bisulfite-modified DNA samples derived from untreated cancers. and CSC-exposed HBEC, as well as soft-agar clones A variety of CpG targets observed to become derived from control and CSC-exposed HBEC, were relatively hypo- or hypermethylated in HBEC by CSC analyzed using the Illumina Infinium methylation assays exposure have not been typically associated with lung that measure DNA methylation at over 27 000 CpG loci cancer development. For example, heat shock protein- distributed among 414K genes. Also profiled were binding protein 1 (HSPB1) and phospholipase D-5 bisulfite-modified DNAs derived from several estab- seemed to become hypermethylated in HBEC after lished lung cancer lines. Partial results of this analysis prolonged CSC exposure (Figure 5a; Table 1). Quanti- are depicted in Figure 5. Short-term (5 days) CSC tative RT–PCR and pyrosequencing experiments re- exposure had negligible effects in SAEC and HBECs vealed that CSC exposure diminished expression of (data not shown). However, using an absolute value HSPB1 coincident with increased methylation of the DiffScore 425, which corresponded to Po0.05, 48 HSPB1 promoter (Figures 5b and c). This phenomenon targets appeared to become demethylated, whereas 56 was not observed for phospholipase D-5 (data not targets exhibited increased methylation in HBEC after shown). Subsequent chromatin immunoprecipitation 9-month CSC exposure (Figure 5a; Table 1). The (ChIP) experiments revealed that CSC exposure dimin- Illumina methylation assays were non-informative for ished H3K9Ac, while increasing H3K27Me3 levels global methylation, but rather provided locus-specific within the HSPB1 promoter (Figure 5d); no appreciable methylation measurements; however, pyrosequencing change in H4K16Ac level within the HSPB1 promoter data were consistent with global genomic demethylation was observed in these ChIP experiments (data not occurring in the context of locoregional hypermethyla- shown). Analysis of a panel of cultured respiratory tion. Overall, methylation profiles in spontaneous soft- epithelial lines revealed that HSPB1 levels in immorta- agar clones derived from untreated HBEC were virtually lized bronchial epithelial cells and numerous lung cancer identical to those observed in pooled, control HBEC. lines appeared lower than those detected in SAEC. Furthermore, methylation changes observed in soft-agar Interestingly, HSPB1 expression in spontaneous clones clones from CSC-exposed HBEC expanded under CSC appeared lower than those observed in untreated exposure were remarkably similar to those observed in controls (but higher than in CSC-treated cells/clones). the same clones expanded in the absence of CSC. These Collectively, these data raise the possibility that HSPB1

Figure 5 (a) Illumina array analysis of DNA methylation in control bulk HBEC, control spontaneous soft-agar clones, and soft-agar clones derived from HBEC exposed to CSC for 9 months. Treatment groups were as follows—1: control untreated HBEC; 2: control spontaneous soft-agar clones; 3: CSC-derived soft-agar clones expanded in the absence of CSC; 4: CSC-derived soft-agar clones expanded in the presence of CSC; 5: pooled HBEC exposed to CSC for 9 months. Each lane pertains to an individual clone or pooled sample. Spontaneous clones exhibited methylation profiles remarkably similar to those observed in untreated HBEC cells. In contrast, CSC-exposed clones showed considerable alterations in methylation status across numerous gene loci. Additional short-term CSC exposure did not seem to further alter methylation profiles in CSC-derived soft-agar colonies. Heat map represents a differential score 425 for genes hypo- or hypermethylated by CSC treatment. These data are summarized in Table 1. (b) Quantitative RT–PCR analysis of HSPB1 expression in untreated control and CSC-treated HBEC, as well as spontaneous and CSC-derived soft-agar clones. Data from two independent experiments revealed that CSC exposure diminished HSPB1 expression in HBEC. Interestingly, HSPB1 expression in spontaneous soft-agar clones was decreased relative to untreated controls, albeit to a lesser extent than CSC-treated HBEC. (c) Pyrosequencing analysis of HSPB1 promoter methylation in untreated and pooled CSC-treated HBEC, as well as spontaneous and CSC-derived soft-agar clones. CSC exposure mediated a progressive increase in HSPB1 promoter methylation. No apparent increase in HSPB1 promoter methylation was observed despite diminished HSPB1 expression in spontaneous clones relative to untreated HBEC controls. (d) ChIP analysis showing decreased H3K9Ac (activation mark) with increased H3K27Me3 (repression mark) within the HSPB1 promoter in HBEC exposed to CSC. These findings are consistent with CSC-mediated down-regulation of HSPB1. (e) Quantitative RT–PCR analysis of HSPB1 expression in a panel of immortalized respiratory epithelia and lung cancer lines. The results are depicted as fold relative to SAEC. HSPB1 expression in immortalized respiratory epithelia (HBEC and BEAS) and numerous lung cancer lines was lower than that observed in SAEC.

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3657 may be a novel target of aberrant silencing by epigenetic been shown recently to correlate with diminished mechanisms in cigarette smoke induced lung cancers. survival of lung cancer patients (Malusecka et al., Whereas loss of intratumoral HSPB1 expression has 2008), the signiﬁcance of our ﬁndings regarding CSC-

ABS DIFF SCORE > 25: 9MO CSC RX VS. CNTRL

Exp 1 Exp 2

2111 11 223344 3 335555544555344 1.00 GAD1 1.0 C2orf10 FAM105A 0.75 ZNF467 HOXA9 0.9 0.50 MOBKL2A FLJ31196 NOS3 0.25 MEG3 0.8 GPR97 Fold change BCNP1 0.00 RLN1 GNAS 0.7 1234 1234 RAB11FIP2 ARR3 MC2R Exp 1 Exp 2 MSR1 0.6 ZNF264 XKRX 100 NMBR ADORA2A 0.5 CCRL2 75 UNQ6411 CCND1 DNM1L 0.4 50 DGCR2 VHL ITM2A 25 IL1RAPL2 0.3 BAGE % Methylation 0 CHRFAM7A UROS 0.2 PPP3R2 1234 1234 FXN ABCC6 PRRG2 0.1 IgG H3K9Ac H3K27Me3 input SUMF1 SLC26A3 123412341234 1234 VHL DNTTIP2 0.0 FAM71C HSPB1 TFF2 TBC1D21 MYR8 MyoD GHRH FLJ38451 HEM1 GAPDH CCND1 CFLAR ATP10A 1. Control RASA4 VAV3 NELL2 2. 1% CSC 5 months TGFBR1 UBE2G1 3. Spontaneous clone GATA4 PRG-3 4. 1% CSC 9 months clone TUBB2B ST3GAL4 ADAMTS17 1.5 NKX2-2 CALCA IGF2AS STON1 1.0 FLJ13391 DRD1 DRD5 PRRX1 0.5

MOXD1 fold/SAEC CPT1A PCSK6 0.0 TRPA1 PTN KIF1A PTGDR H157 H358 H460 H596 A549

C1orf65 H1299 H1734 H1975 H2122 H2126 H3255 H2087 Calu-6 MEGF11 CLSTN2 SK-LU-1 LOX NAALAD2 GRID2 1.5 KCNJ8 C1QTNF1 NMUR1 HSPB1 1.0 IRXL1 ABHD7 SLC26A11 SOCS1 0.5

RARB fold/SAEC TSSK3 CILP2 WNT3A 0.0 INA CCDC37

PLD5 H82

KCNJ8 H123 H128 H345 H520 H526 H889 H841 BEAS HBEC PNMA2 H1618 H1688 H2028 H2721 DOCK2 RASSF1 C1QTNF1 LOC221091 C10orf59 INSR

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3658 Table 1 Differentially methylated targets in CSC-derived soft-agar clones Hypomethylated in CSC-treated vs controls Hypermethylated in CSC-treated vs controls

Symbol Illumina index ABS (DiffScore) Symbol Illumina index ABS (DiffScore)

ABCC6 2617 25.6 ABHD7 13 946 25.3 ADORA2A 21 571 29.4 ADAMTS17 7955 25.9 ARR3 11 432 25.4 C10orf59 6803 28.3 ATP10A 26 040 26.1 C1ord65 27 143 30.2 BAGE 7940 26.1 C1QTNF1 8298 27.3 BCNP1 6275 27.3 C1QTNF1 24 812 28 CCND1 990 27.6 C2ord10 13 038 25.9 CCND1 7397 28.2 CALCA 6269 26.1 CCRL2 5667 26.1 CCDC37 914 28.3 CFLAR 18 131 26.1 CILP2 10 236 25.9 CHRFAM7A 6289 25.9 CLSTN2 14 499 25.9 DGCR2 16 686 26.4 CPT1A 19 599 28 DNM1L 27 322 28.3 DOCK2 21 201 26.1 DNTTIP2 22 772 26.1 DRD1 16 186 25.6 FAM105A 23 585 25.2 DRD5 9862 27.1 FAM71C 13 272 26.5 FLJ13391 10 049 26.1 FLJ31196 24 709 32 GAD1 947 28 FLJ38451 20 517 28 GATA4 14 108 28 FXN 7155 27.3 GRID2 14 015 29.4 GHRH 4495 26.1 HSPB1 27 358 26.1 GNAS 25 261 28.4 IGF2AS 13 734 25.2 GPR97 8320 30 INA 25 743 25.5 HEM1 17 514 29.4 INSR 10 575 32.9 HOXA9 26 483 27.1 IRXL1 14 619 25.9 IL1RAPL2 23 356 29.4 KCNJ8 1234 30 ITM2A 10 794 26 KCNJ8 2039 26.1 MC2R 11 493 25.4 KIF1A 21 303 30 MEG3 25 815 29.4 LOC221091 25 613 28 MOBKL2A 6887 28.3 LOX 22 807 30 MSR1 1695 25.2 MEGF11 26 400 31.5 MYR8 14 385 27.6 MOXD1 7539 30.2 NMBR 17 286 25.9 NAALAD2 5511 27.1 NOS3 3781 34.8 NELL2 14 911 25 PPP3R2 15 807 26.3 NKX2-2 17 058 25.9 PRRG2 4235 28.3 NMUR1 158 25.2 RAB11FIP1 17 823 25.1 PSCK6 20 210 26.1 RLN1 20 739 25.9 PLD5 12 552 30.2 SLC26A3 22 266 28 PNMA2 2174 28.3 SUMF1 18 839 34.4 PRG-3 16 937 26.1 TBC1D21 12 658 25.9 PRRX1 10 632 30 TFF2 11 073 28.3 PTGDR 9454 26.3 UNQ6411 16 017 34.8 PTN 11 461 27.1 UROS 19 336 26.1 RARB 26 099 28 VHL 16 908 27.1 RASA4 10 198 26.1 VHL 24 046 27.6 RASSF1 804 25.9 XKRX 11 593 27.1 SLC26A11 24 809 28.3 ZNF264 16 697 30.2 SOCS1 6181 27.1 ZNF467 23 709 28.3 ST2GAL4 1242 25.9 STON1 25 043 26.3 TGFBR1 15 523 28 TRPA1 1629 25.9 TSSK3 21 138 26.3 TUBB2B 3481 26.3 UBE2G1 21 711 25.9 VAV3 34 624 30 WNT3A 1334 26.1

Abbreviation: CSC, cigarette smoke condensate.

mediated down-regulation of HSPB1 has not been Figure 6. Gene expression profiles in control sponta- firmly established. neous soft-agar clones were remarkably similar to those Additional correlative Illumina array experiments in pooled untreated HBEC. In contrast, gene expression were performed to examine gene expression profiles profiles in six randomly selected soft-agar clones derived mediated by CSC, using RNA harvested from cells from HBEC exposed to CSC for 9 months (Figure 4d), providing DNA for the aforementioned DNA methyla- and expanded in the absence of CSC were considerably tion arrays. The results of this analysis are depicted in different than those observed in controls, and seemed to

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3659 1234 CSC(-) CSC(+)

NRF2-mediated Oxidative Stress Response

Coagulation System

Aryl Hydrocarbon Receptor Signaling

Xenobiotic Metabolism Signaling

VDR/RXR Activation

TREM1 Signaling

Wnt/Beta-Catenin Signaling

P53 Signaling Airway Pathology in Chronic Obstructive Pulmonary Disease PI3K/AKT Signaling

Human Embryonic Stem Cell Pluripotency

Eicosanoid Signaling

mTOR Signaling 1. Control 2. Control spontaneous clones VEGF Signaling 3. 1.0% CSC 9 months clones without CSC 4. 1.0% CSC 9 months clones with CSC Figure 6 (a) Illumina array analysis of gene expression in untreated HBEC, and spontaneous clones from control HBEC emerging from soft agar, as well as clones exposed to CSC and then expanded with or without additional tobacco smoke exposure. Gene expression profiles in untreated controls as well as spontaneous control clones were remarkably similar. Interestingly, array analysis revealed that gene expression profiles in clones derived from CSC-treated HBEC that were subsequently expanded without further CSC exposure were markedly altered; additional CSC exposure further modulated gene expression in HBEC, reflecting acute effects of CSC. (b) Ingenuity analysis of Illumina gene expression arrays revealing major pathways modulated in CSC-derived soft-agar clones expanded in the absence (CSCÀ) or presence (CSC þ ) of CSC. represent stable sequelae of epigenetic (and presumably expanded in the absence or presence of CSC relative to genetic) alterations induced by CSC. Profiles in CSC- control (spontaneous) soft-agar clones. Fifteen of the derived soft-agar clones expanded in the absence of smoke top 50 genes (30%) activated in CSC-derived clones were markedly different from those detected expanded in the absence of CSC were among the top 50 in the same clones after 6-day CSC exposure (Figure 6a, genes activated in these clones after additional CSC lanes 3 and 4). Using criteria of Xtwofold relative to exposure. In contrast, 23 of the 50 most repressed controls and Po0.05, 85 genes were up-regulated, genes (46%) in CSC-derived soft-agar clones expanded whereas 109 genes were repressed in CSC-derived soft- in the absence of CSC were among the top 50 most agar clones expanded in the absence of CSC. Using repressed genes in these clones after CSC exposure. similar criteria, 230 genes were induced, whereas 295 were Although some exceptions were noted, the magnitude repressed in these CSC-derived soft-agar clones expanded of repression of genes simultaneously down-regulated in under CSC exposure. Interestingly, 44% of differentially CSC-derived soft-agar clones expanded in the absence modulated genes were up-regulated, whereas 56% were or presence of CSC (that is TXNIP, DUSP23, VIM) was repressed in CSC-derived soft-agar clones expanded under relatively comparable. In contrast, CSC seemed to either treatment condition. Ingenuity pathway analysis enhance the magnitude of activation of genes such as revealed that networks involving antigen presentation/ SPANXB1, KLK6, and KYNU that were up-regulated immune response, organ development, tissue morphol- in CSC-derived clones expanded under both treatment ogy/cancer, DNA replication/repair, and inflammation conditions. Overall, these data raise the possibility that were preferentially modulated in CSC-derived soft-agar repression was more stable than activation of gene clones expanded in the absence of CSC. In contrast, expression in HBEC after prolonged CSC exposure. networks preferentially modulated in these soft-agar clones expanded in the presence of CSC included cellular movement, small molecule biochemistry/gene expression, Discussion and genetic disorders (data available on request). Several pathways preferentially modulated in these clones are A number of studies have been performed to examine depicted in Figure 6b. the effects of individual components of tobacco smoke, Table 2 contains a list of the top 100 genes such as nicotine or NNK in cultured respiratory differentially modulated in CSC-derived soft-agar clones epithelia (West et al., 2003; Ho et al., 2005; Jin et al.,

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3660 Table 2 Genes most signiﬁcantly induced/repressed in soft-agar clones expanded in the presence or absence of CSC Top 50 genes up-regulated in Top 50 genes up-regulated in Top 50 genes down-regulated in Top 50 genes down-regulated in soft-agar clones without soft-agar clones with soft-agar clones without soft-agar clones with CSC vs control CSC vs control CSC vs control CSC vs control

Symbol (ÀCSC) ( þ CSC) Symbol ( þ CSC) (ÀCSC) Symbol (ÀCSC) ( þ CSC) Symbol ( þ CSC) (ÀCSC)

NUAK2 8.73 5.77 KYNU 42.39 3.42 KRT5 À2.63 À3.5 SPARC À3.54 À1.2 FABP4 6.82 1.43 GPX2 37.67 2.12 BACE2 À2.64 À3.72 ARMCX6 À3.6 À3.6 SERPINA5 5.11 2.44 S100A9 24.14 1.49 EFEMP1 À2.65 À5.74 CDH4 À3.6 À1.81 ESM1 4.76 À1.33 ANXA10 22.24 1.29 VLDLR À2.73 À4.01 ITGA6 À3.7 À1.45 TGM2 4.41 1.21 IL8 19.58 2.84 DDIT4 À2.77 À1.52 G0S2 À3.71 À1.54 EEF1A2 4.18 6.12 CYP1B1 18.5 À1.02 CAMP À2.8 1.28 HNT À3.71 2.73 KLK5 4.12 13.02 S100A8 17.83 À1.6 FAT2 À2.84 À1.65 BACE2 À3.72 À2.64 CDH2 3.9 À1.06 HSD17B2 13.72 1.01 MYO5C À2.87 À3.06 MARCH4 À3.73 1.75 ADAM19 3.46 1.73 ALDH3A1 13.08 À1.53 CCNG2 À2.88 À2.63 CD59 À3.75 À1.44 TIMP2 3.44 1.93 KLK5 13.02 4.12 DLL3 À2.89 À2.92 LMCD1 À3.84 À2.41 KYNU 3.42 42.39 AKR1C3 11.74 À2.2 LXN À2.98 À1.73 CTGF À3.86 À1.4 CSF3 3.39 3.54 AKR1C4 11.23 À2.44 RPL39L À3.02 À3.07 LAMA3 À3.9 À1.37 NTSR1 3.32 3.17 PTGES 10.61 3.03 PTGS2 À3.06 À1.94 CAV1 À4.01 À1.46 PDPN 3.3 1.03 SPANXB1 10.15 2.37 TMEM47 À3.06 À3.06 VLDLR À4.01 À2.73 MRGPRX4 3.25 À1.07 SPANXC 9.72 2.22 UST À3.16 À1.97 CDC42EP5 À4.04 À3.82 KLK6 3.21 8.95 GDF15 9.67 1.36 ANGPTL4 À3.32 À4.1 DBN1 À4.08 À1.71 VGLL1 3.2 1.53 MGC59937 9.64 2.27 SLC2A3 À3.35 À2.43 ANGPTL4 À4.1 À3.32 PTGES 3.03 10.61 AKR1B10 9.3 À1.47 NRCAM À3.37 À1.44 F3 À4.13 À1.52 CD83 3.02 1.95 KLK6 8.95 3.21 MGMT À3.41 À2.71 PRSS23 À4.25 1.07 TNFRSF10D 2.99 1.88 TSPAN7 8.73 À1.29 COL8A1 À3.44 À4.29 SRPX À4.25 À1.91 GFPT2 2.93 3.04 DHRS3 8.24 2.14 GJA1 À3.59 À2.49 COL8A1 À4.29 À3.44 XYLT1 2.86 À3.01 S100P 8.05 À1.33 ARMCX6 À3.6 À3.6 AOX1 À4.34 À2.14 IL8 2.84 19.58 GCNT3 7.87 2.62 PTHLH À3.6 À4.85 CA9 À4.42 À4.44 MAL2 2.79 5.03 CYP1A1 6.86 1.3 CRIP2 À3.68 À2.7 TXNIP À4.61 À5.13 SLC19A1 2.75 1.74 FTH1 6.81 À1.33 CKMT1B À3.77 À2.43 ADAMTS6 À4.62 À1.26 HNT 2.73 À3.71 PIR 6.21 À1.01 CDC42EP5 À3.82 À4.04 TMEM16D À4.62 À4.07 CGI-96 2.73 2.33 EEF1A2 6.12 4.18 C1orf24 À3.82 À2.63 ARHGDIB À4.63 1.33 STARD10 2.72 5.83 SPANXE 6.09 1.94 CLCA2 À4.01 À2.01 BNIP3 À4.7 À4.58 E2F2 2.66 2.52 OLR1 5.92 1.33 FBN2 À4.02 À8.92 PLOD2 À4.72 À1.33 LTB 2.66 2.62 STARD10 5.83 2.72 CBS À4.04 À2.04 EMP3 À4.84 À4.98 MDM2 2.66 2.08 UGT1A6 5.82 1.08 TMEM16D À4.07 À4.62 FLRT2 À4.85 1.09 FAM43A 2.63 4.45 NUAK2 5.77 8.73 STOM À4.22 À2.44 PTHLH À4.85 À3.6 GCNT3 2.62 7.87 CSF2 5.72 2.24 HSPB1 À4.32 À2.22 SLC38A5 À5.04 À5.12 TFAP2C 2.6 4.26 KLK9 5.27 1.53 CA9 À4.44 À4.42 PCDH7 À5.14 À1.96 SEMA4D 2.53 1.72 ABCC3 5.15 À1.93 SNCA À4.55 À3.07 AEBP1 À5.22 À5.16 RAB3IL1 2.53 2.98 MAL2 5.03 2.79 BNIP3 À4.58 À4.7 GSPT2 À5.72 À5.35 CATSPER1 2.5 1.72 FTL 4.67 1.46 CKMT1A À4.94 À2.51 EFEMP1 À5.74 À2.65 CYP26B1 2.46 1.37 GCLM 4.62 1.35 PLD5 À4.96 À9.07 DKK1 À5.88 À1.42 C14orf147 2.46 4.28 IMPA2 4.55 1.5 EMP3 À4.98 À4.84 CYR61 À5.93 À1.58 APOBEC3B 2.44 4.11 FAM43A 4.45 2.63 SLC38A5 À5.12 À5.04 LPHN2 À6.34 À6.14 POLDIP3 2.42 2.9 KRT23 4.35 À1.99 TXNIP À5.13 À4.61 KCNMA1 À6.46 À1.88 LHX1 2.42 2.4 CLIC3 4.32 1.28 AEBP1 À5.16 À5.22 MFAP5 À6.51 1.44 SH3KBP1 2.41 2.82 IL13RA2 4.3 À1.22 NDRG1 À5.24 À3.16 KRT14 À6.89 À6.41 SERPINA1 2.41 1.95 C14orf147 4.28 2.46 GSPT2 À5.35 À5.72 THBS1 À7.78 À1.42 SPANXB1 2.37 10.15 HMOX1 4.28 1.23 LPHN2 À6.14 À6.34 SERPINE1 À8.71 À1.45 PACSIN2 2.37 2.58 TFAP2C 4.26 2.6 KRT14 À6.41 À6.89 FBN2 À8.92 À4.02 PLCG2 2.36 2.79 APOBEC3B 4.11 2.44 DUSP23 À11.23 À10.46 PLD5 À9.07 À4.96 ITM2C 2.32 1.59 G6PD 4.01 1.14 GJB2 À13.17 À2.13 DUSP23 À10.46 À11.23 HOXB5 2.32 1.17 ZP3 3.91 1.4 PRSS3 À13.37 À11.11 PRSS3 À11.11 À13.37 KHDRBS3 2.32 1.58 EPHX1 3.87 1.48 VIM À17.34 À36.44 VIM À36.44 À17.34

Abbreviation: CSC, cigarette smoke condensate.

2008; Wu et al., 2009). Although informative, these and NNK, or vapor phase such as formaldehyde or studies are inconclusive as the effects of single tobacco ethyl carbamate are used for the exposures (Sexton components are unlikely to represent those mediated by et al., 2008). Additional studies using CSCs have the complex mixture of organic as well as inorganic produced variable results depending on the type of carcinogens detected in tobacco smoke (Smith et al., condensates and treatment conditions used for the 2003; Shin et al., 2009). Indeed, genetic responses of experiments, including supplementation of culture cultured respiratory epithelia to tobacco carcinogens media with S9 microsomal fractions to enhance activa- vary considerably depending on which components tion of tobacco carcinogens in respiratory epithelial cells from air/liquid interphase such as nicotine, cadmium, (Jorgensen et al., 2004; Fields et al., 2005).

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3661 In this study, we focused our efforts on characterizing showing no correlation between DNA methylation and the epigenomic effects of CSC in SAEC and HBEC, gene expression profiles in follicular lymphomas (Killian acknowledging that neither was perfect for this analysis. et al., 2009). For example, SAEC, which theoretically contain true The results of our studies vary considerably from target cells of tobacco carcinogens, spontaneously those reported by Damiani et al. (2008), who observed senesce after several passages in culture. On the other that the tobacco carcinogens methyl-nitrosourea and hand, immortalization may have subtly perturbed benzo (a) pyrene-diolepoxide increased DNMT1 chromatin structure, and altered susceptibility of HBEC levels, and induced hypermethylation of E-cadherin, to tobacco carcinogens. CDH13, proto-cadherin 10, GATA5, and PAX5 in Despite these potential limitations, our experiments HBEC; these epigenetic alterations coincided with yielded several interesting and potentially relevant enhanced soft-agar clonogenicity, but not tumorigeni- findings. For example, CSC exposure dramatically city in nude mice. Notably, in Damiani’s model as decreased acetylation of H4K16, and subsequently well as ours, p16 was neither silenced nor hypermethy- diminished H4K20 trimethylation in cultured respira- lated by carcinogen exposure, despite the fact that tory epithelia. These histone modifications have been inactivation of this tumor suppressor gene is believed to observed in several malignancies, including lung cancers be a frequent, early event during tobacco-induced (Fraga et al., 2005; Van Den et al., 2008). Given the pulmonary carcinogenesis (Toyooka et al., 2006). In highly dynamic nature of histone modifications in our experiments, CSC exposure seemed insufficient to response to environmental stimuli including tobacco inactivate E-cadherin and CDH13, yet mediated pro- carcinogens (Hussain et al., 2009; Jensen et al., 2009; gressive DNA hypermethylation of RASSF1A and Zhou et al., 2009), it is not surprising that modulation of RAR-b, which are known to be targets of aberrant the histone code seemed to precede aberrant DNA DNA methylation in human lung cancers (Licchesi methylation typically observed in lung cancers (Toyoo- et al., 2008; Seng et al., 2008); these alterations seemed ka et al., 2003, 2006). Moreover, our data suggest that to coincide with a markedly increased DNMT3b/ demethylation of repetitive DNA sequences, mono- DNMT1 protein ratio. Discrepancies regarding results allelic up-regulation of H19 and IGF2, as well as reported by Damiani et al. (2008) and our current de-repression of several CT-X genes such as MAGE-A1, findings may be attributable (at least in part) to MAGE-A3, and SPANXB1, precedes DNA methyla- exposure conditions including the use of purified, tion-mediated silencing of tumor suppressor genes, such activated carcinogens, serum supplementation and as p16, MGMT, DAPK, E-cadherin, and cdh13 during avoidance of oxidative stress by Damiani et al. (2008) tobacco-induced pulmonary carcinogenesis. These find- rather than CSC, serum-free media and oxidative stress ings suggest that global DNA demethylation is an early used for our experiments. event during—not a consequence of—malignant trans- From a clinical biomarker discovery perspective, the formation. Interestingly, progressive global DNA de- predictive value of the model described in this manu- methylation with consistent hypermethylation of script seems to be relatively low regarding identification RASSF1A-a tumor suppressor frequently silenced by of differentially methylated genetic targets during epigenetic mechanisms in tobacco-induced lung cancers cigarette smoke-induced pulmonary carcinogenesis. (Richter et al., 2009), and activation of Wnt signaling Conceivably, additional CSC exposures will induce coincided with increased clonogenicity of HBEC. The epigenetic alterations in HBEC more typically observed fact that CSC-exposed HBEC were not tumorigenic in in lung cancer cells; these experiments are ongoing in murine hosts suggests that additional epigenetic altera- our laboratory. Despite potential limitations, this model tions that silence other tumor suppressors or activate may enable investigation of a variety of epigenetic stem cell genes such as Oct4, SOX, and NANOG-a mechanisms contributing to tobacco-induced lung can- phenomenon not observed in our present studies cers, and evaluation of novel agents for the treatment (data not shown), may be necessary to fully transform and prevention of these neoplasms. these cells. Despite the fact that DNA methylation and gene expression profiling could readily discriminate between Materials and methods control and CSC-exposed HBEC, little correlation was observed between DNA methylation and gene expres- Cell culture sion profiles for any given treatment condition. These All lung cancer cell lines were obtained from American Type discrepancies may be attributable to the fact that many Culture Collection (Manassas, VA, USA) and maintained in of the CpG targets on the methylation arrays are not RPMI media supplemented with 10% FBS and 10 mM within promoter regions; furthermore, modulation of glultamic acid. SAEC and NHBE were obtained from Lonza, gene expression by CSC may occur through mechanisms Inc. (Frederick, MD, USA), and cultured in appropriate media independent of DNA methylation (Kang et al., 2007; (SAGM and BEBM, respectively) according to vendor instructions. HBEC were generously provided by John D Hussain et al., 2009). Overall, these findings are Minna (U-T Southwestern, Dallas, TX, USA), and cultured in consistent with data indicating that hypermethylation complete Keratinocyte-SFM media (Invitrogen, Carlsbad, CA, of CpG islands in cancer cells frequently involves genes, USA) supplemented with 5 mg/l epidermal growth factor and which typically do not mediate proliferation or tumor- 50 mg/l bovine pituitary extract. Cell proliferation was assessed igenicity (Keshet et al., 2006), as well as a recent report by MTT techniques.

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3662 Generation of CSCs and cell line exposure obtained from SABiosciences (Frederick, MD, USA). Full CSCs were generated from Kentucky 2R4F research cigarettes details of these methods are submitted as Supplementary (University of Kentucky, Tobacco and Health Research Information. Institute, Lexington, KY, USA) using a Borgwaldt-LM1 smoking machine (Richmond, VA, USA) and standard Federal Trade Commission smoking conditions (35 ml puff Chromatin immunoprecipitation volume, 2.0 s duration, and 1 puff/min, 9 puffs/cigarette). The ChIP was performed using reagents and protocols contained in smoke condensates were trapped on Cambridge glass fiber the Millipore 17–295 ChIP kit (Billeria, MA, USA), and ChIP- filters, weighed, and dissolved in K-SFM, RPMI, SAGM, and grade antibodies recognizing H3K9Ac (ab4441–50; Abcam) or BEBM media, respectively, at a concentration of 1 mg/ml, H3K27me3 (07–449; Millipore, Billerica, MA, USA). Full which was defined as 10% CSC (Narayan et al., 2004). methods and primer sequences for ChIP PCR are submitted as Supplementary Information. CSC treatments NHBE, SAEC, HBEC, and A549 cells were plated in six-well Illumina methylation array 2 plates or 10 cm culture dishes in appropriate normal media Genomic DNA was extracted from cells using Qiagen DNeasy with or without CSC in the absence of microsomal supple- kit (Qiagen, Duesseldorf, Germany). Purified DNA was mentation. Media and CSC were changed daily. At appro- bisulfite treated with EpiTect Bisulfite kit (Qiagen). DNA priate times, cells were harvested, and processed for further methylation levels of bisulfite converted DNAs were measured analysis. using Illumina Infinium or GoldenGate methylation assays (Illumina, San Diego, CA, USA) according to the manufac- Genomic DNA isolation, bisulfite conversion, turer’s protocol. Methylation and differential methylation and pyrosequencing analyses were performed in BeadStudio (Illumina). In group Genomic DNA was isolated from cells using the DNeasy comparisons, differentially methylated targets were determined Blood and Tissue kit (Qiagen, Valencia, CA, USA) according based on abs (DiffScore)425, as described earlier (Killian to the manufacturer’s protocol. A total of 0.5 mg of DNA was et al., 2009). used for bisulfite conversion by using EpiTect Bisulfite kit (Qiagen). Bisulfite-modified DNA was dissolved in 30 mlH2O, and 2 ml of DNA template was used for pyrosequencing PCR Illumina gene expression array amplification. Dispensation order pyrosequencing reactions A total of 200 ng of total RNA was amplified and cRNA was and data analysis were performed using the Pyromark MD labeled with biotin using Illumina TotalPrep RNA amplifica- pyrosequencer with software provided by Biotage (Charlottes- tion kit (Ambion, Austin, TX, USA); 750 ng of biotin labeled ville, VA, USA). p16, MGMT, and Line-1 pyrosequencing cRNA was hybridized to Sentrix BeadChip Array for Gene primers were obtained from Biotage; additional pyrosequen- Expression HumanRef-8 V2 (Illumina) and incubated at 58 1C cing primer sequences are included in Supplementary Table 1. for 16–20 h in an Illumina hybridization oven with rocker speed at 5. Beadchips were washed and stained according to RNA extraction and quantitative real-time RT–PCR Illumina’s protocol. Arrays were scanned by Illumina chip Submitted as Supplementary Information. scanner, and images analyzed by Bead Studio (Illumina). Data were exported and processed using Genespring GX 7.31 (Agilent, Santa Clara, CA, USA). Differentially regulated Western blot genes between TSC-treated and control cells were identified by Immunoblotting of non-histone proteins was performed as statistical significance Po0.05 and fold change 41.5. Sig- described (Hussain et al., 2009); western blot analysis of nificantly regulated genes were subjected to further analysis histones was performed according to Millipore protocols, with using Ingenuity Pathway software (Ingenuity, Redwood City, minor modifications (detailed methods submitted as Supple- CA, USA). mentary Information). Antibodies used for immunoblotting included H4K16Ac (Abcam ab1762 Cambridge, MA, USA), H4K20Me3 (Upstate 07–463 Billerica, MA, USA), Soft-agar assays H3K27Me3 (Upstate 07–449), total H3 (Upstate 05–499), Submitted as Supplementary Information. and total H4 (Abcam ab10158), DNMT1 (Abcam ab16632– 100), DNMT3b (Cell Signal, 2161s, Boston, MA, USA), and EZH2 (BD, 612667, Woburn, MA, USA). Western blot signals Murine xenograft experiments were detected with the appropriate secondary-HRP-conju- Submitted as Supplementary Information. gated antibodies and ECL reagent (Amersham, Pittsburgh, PA, USA), followed by densitometric analysis.

RT–PCR superarrays Conﬂict of interest Wnt signaling RT–PCR superarrays (PAH-043A) and Human Stress and Toxicity Pathway Finder (PAH-003A) were The authors declare no conﬂict of interest.

References

Agherbi H, Gaussmann-Wenger A, Verthuy C, Chasson L, Serrano Barlesi F, Giaccone G, Gallegos-Ruiz MI, Loundou A, Span SW, M, Djabali M. (2009). Polycomb mediated epigenetic silencing and Lefesvre P et al. (2007). Global histone modiﬁcations predict replication timing at the INK4a/ARF locus during senescence. prognosis of resected non small-cell lung cancer. J Clin Oncol 25: PLoS One 4: e5622. 4358–4364.

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3663 Belinsky SA, Nikula KJ, Baylin SB, Issa JP. (1996). Increased cytosine Kim H, Kwon YM, Kim JS, Han J, Shim YM, Park J et al. (2006). DNA-methyltransferase activity is target-cell-specific and an early Elevated mRNA levels of DNA methyltransferase-1 as an event in lung cancer. Proc Natl Acad Sci USA 93: 4045–4050. independent prognostic factor in primary nonsmall cell lung cancer. Cho HJ, Caballero OL, Gnjatic S, Andrade VC, Colleoni GW, Vettore Cancer 107: 1042–1049. AL et al. (2006). Physical interaction of two cancer-testis antigens, Le FN, Rivat C, De WO, Bruyneel E, Mareel M, Dale T et al. (2005). MAGE-C1 (CT7) and NY-ESO-1 (CT6). Cancer Immun 6: 12. The proinvasive activity of Wnt-2 is mediated through a non- D’Alessio AC, Szyf M. (2006). Epigenetic teˆte-a` -teˆte: the bilateral canonical Wnt pathway coupled to GSK-3beta and c-Jun/AP-1 relationship between chromatin modifications and DNA methyla- signaling. FASEB J 19: 144–146. tion. Biochem Cell Biol 84: 463–476. Lee SJ, Jeon HS, Jang JS, Park SH, Lee GY, Lee BH et al. (2005). Damiani LA, Yingling CM, Leng S, Romo PE, Nakamura J, Belinsky DNMT3B polymorphisms and risk of primary lung cancer. SA. (2008). Carcinogen-induced gene promoter hypermethylation is Carcinogenesis 26: 403–409. mediated by DNMT1 and causal for transformation of immorta- Licchesi JD, Westra WH, Hooker CM, Machida EO, Baylin SB, lized bronchial epithelial cells. Cancer Res 68: 9005–9014. Herman JG. (2008). Epigenetic alteration of Wnt pathway Dovey JS, Zacharek SJ, Kim CF, Lees JA. (2008). Bmi1 is critical for antagonists in progressive glandular neoplasia of the lung. lung tumorigenesis and bronchioalveolar stem cell expansion. Proc Carcinogenesis 29: 895–904. Natl Acad Sci USA 105: 11857–11862. Lin RK, Hsu HS, Chang JW, Chen CY, Chen JT, Wang YC. (2007). Fields WR, Leonard RM, Odom PS, Nordskog BK, Ogden MW, Alteration of DNA methyltransferases contributes to 50CpG Doolittle DJ. (2005). Gene expression in normal human bronchial methylation and poor prognosis in lung cancer. Lung Cancer 55: epithelial (NHBE) cells following in vitro exposure to cigarette 205–213. smoke condensate. Toxicol Sci 86: 84–91. Malusecka E, Krzyzowska-Gruca S, Gawrychowski J, Fiszer-Kierz- Fraga MF, Ballestar E, Villar-Garea A, Boix-Chornet M, Espada J, kowska A, Kolosza Z, Krawczyk Z. (2008). Stress proteins HSP27 Schotta G et al. (2005). Loss of acetylation at Lys16 and and HSP70i predict survival in non-small cell lung carcinoma. trimethylation at Lys20 of histone H4 is a common hallmark of Anticancer Res 28: 501–506. human cancer. Nat Genet 37: 391–400. Mathews LA, Crea F, Farrar WL. (2009). Epigenetic gene regulation Ho YS, Chen CH, Wang YJ, Pestell RG, Albanese C, Chen RJ et al. in stem cells and correlation to cancer. Differentiation 78: 1–17. (2005). Tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3- McCabe MT, Lee EK, Vertino PM. (2009). A multifactorial signature pyridyl)-1-butanone (NNK) induces cell proliferation in normal of DNA sequence and polycomb binding predicts aberrant CpG human bronchial epithelial cells through NFkappaB activation and island methylation. Cancer Res 69: 282–291. cyclin D1 up-regulation. Toxicol Appl Pharmacol 205: 133–148. Monte M, Simonatto M, Peche LY, Bublik DR, Gobessi S, Pierotti Hoogeboom D, Burgering BM. (2009). Should I stay or should I go: MA et al. (2006). MAGE-A tumor antigens target p53 transactiva- beta-catenin decides under stress. Biochim Biophys Acta 1796: 63–74. tion function through histone deacetylase recruitment and confer Huang CL, Liu D, Nakano J, Ishikawa S, Kontani K, Yokomise H resistance to chemotherapeutic agents. Proc Natl Acad Sci USA 103: et al. (2005). Wnt5a expression is associated with the tumor 11160–11165. proliferation and the stromal vascular endothelial growth Narayan S, Jaiswal AS, Kang D, Srivastava P, Das GM, Gairola CG. factor—an expression in non-small-cell lung cancer. J Clin Oncol (2004). Cigarette smoke condensate-induced transformation 23: 8765–8773. of normal human breast epithelial cells in vitro. Oncogene 23: Huang H, He X. (2008). Wnt/beta-catenin signaling: new (and old) 5880–5889. players and new insights. Curr Opin Cell Biol 20: 119–125. Richter AM, Pfeifer GP, Dammann RH. (2009). The RASSF proteins Hussain M, Rao M, Humphries AE, Hong JA, Liu F, Yang M et al. in cancer; from epigenetic silencing to functional characterization. (2009). Tobacco smoke induces polycomb-mediated repression of Biochim Biophys Acta 1796: 114–128. Dickkopf-1 in lung cancer cells. Cancer Res 69: 3570–3578. Schrump DS, Hong JA, Nguyen DM. (2007). Utilization of chromatin Jensen TJ, Wozniak RJ, Eblin KE, Wnek SM, Gandolfi AJ, Futscher remodeling agents for lung cancer therapy. Cancer J 13: 56–64. BW. (2009). Epigenetic mediated transcriptional activation of Seligson DB, Horvath S, McBrian MA, Mah V, Yu H, Tze S et al. WNT5A participates in arsenical-associated malignant transforma- (2009). Global levels of histone modifications predict prognosis in tion. Toxicol Appl Pharmacol 235: 39–46. different cancers. Am J Pathol 174: 1619–1628. Jin Q, Menter DG, Mao L, Hong WK, Lee HY. (2008). Survivin Seng TJ, Currey N, Cooper WA, Lee CS, Chan C, Horvath L et al. expression in normal human bronchial epithelial cells: an early and (2008). DLEC1 and MLH1 promoter methylation are associated critical step in tumorigenesis induced by tobacco exposure. with poor prognosis in non-small cell lung carcinoma. Br J Cancer Carcinogenesis 29: 1614–1622. 99: 375–382. Jorgensen ED, Dozmorov I, Frank MB, Centola M, Albino AP. Sexton K, Balharry D, BeruBe KA. (2008). Genomic biomarkers of (2004). Global gene expression analysis of human bronchial pulmonary exposure to tobacco smoke components. Pharmacogenet epithelial cells treated with tobacco condensates. Cell Cycle 3: Genomics 18: 853–860. 1154–1168. Shin HJ, Sohn HO, Han JH, Park CH, Lee HS, Lee DW et al. (2009). Kang Y, Hong JA, Chen GA, Nguyen DM, Schrump DS. (2007). Effect of cigarette filters on the chemical composition and in vitro Dynamic transcriptional regulatory complexes including BORIS, biological activity of cigarette mainstream smoke. Food Chem CTCF and Sp1 modulate NY-ESO-1 expression in lung cancer cells. Toxicol 47: 192–197. Oncogene 26: 4394–4403. Simpson AJ, Caballero OL, Jungbluth A, Chen YT, Old LJ. (2005). Kaplan R, Luettich K, Heguy A, Hackett NR, Harvey BG, Crystal Cancer/testis antigens, gametogenesis and cancer. Nat Rev Cancer 5: RG. (2003). Monoallelic up-regulation of the imprinted H19 gene in 615–625. airway epithelium of phenotypically normal cigarette smokers. Smith CJ, Perfetti TA, Garg R, Hansch C. (2003). IARC carcinogens Cancer Res 63: 1475–1482. reported in cigarette mainstream smoke and their calculated log P Katoh M, Katoh M. (2007). WNT signaling pathway and stem cell values. Food Chem Toxicol 41: 807–817. signaling network. Clin Cancer Res 13: 4042–4045. Toyooka S, Maruyama R, Toyooka KO, McLerran D, Feng Z, Keshet I, Schlesinger Y, Farkash S, Rand E, Hecht M, Segal E et al. Fukuyama Y et al. (2003). Smoke exposure, histologic type and (2006). Evidence for an instructive mechanism of de novo geography-related differences in the methylation profiles of non- methylation in cancer cells. Nat Genet 38: 149–153. small cell lung cancer. Int J Cancer 103: 153–160. Killian JK, Bilke S, Davis S, Walker RL, Killian MS, Jaeger EB et al. Toyooka S, Tokumo M, Shigematsu H, Matsuo K, Asano H, Tomii K (2009). Large-scale profiling of archival lymph nodes reveals et al. (2006). Mutational and epigenetic evidence for independent pervasive remodeling of the follicular lymphoma methylome. pathways for lung adenocarcinomas arising in smokers and never Cancer Res 69: 758–764. smokers. Cancer Res 66: 1371–1375.

Oncogene Epigenetics of tobacco smoke exposure FLiuet al 3664 Van Den BA, Brambilla E, Moro-Sibilot D, Lantuejoul S, Brambilla dependent activation of peroxisome proliferator-activated receptor C, Eymin B et al. (2008). Loss of histone H4K20 trimethylation gamma. J Biol Chem 281: 26943–26950. occurs in preneoplasia and inﬂuences prognosis of non-small cell Wu JP, Chang LW, Yao HT, Chang H, Tsai HT, Tsai MH et al. lung cancer. Clin Cancer Res 14: 7237–7245. (2009). Involvement of oxidative stress and activation of aryl Vrzalikova K, Skarda J, Ehrmann J, Murray PG, Fridman E, hydrocarbon receptor in elevation of CYP1A1 expression and Kopolovic J et al. (2008). Prognostic value of Bmi-1 oncoprotein activity in lung cells and tissues by arsenic: an in vitro and in vivo expression in NSCLC patients: a tissue microarray study. J Cancer study. Toxicol Sci 107: 385–393. Res Clin Oncol 134: 1037–1042. Xing J, Stewart DJ, Gu J, Lu C, Spitz MR, Wu X. (2008). Expression Watanabe H, Soejima K, Yasuda H, Kawada I, Nakachi I, Yoda S of methylation-related genes is associated with overall survival et al. (2008). Deregulation of histone lysine methyltransferases in patients with non-small cell lung cancer. Br J Cancer 98: contributes to oncogenic transformation of human bronchoepithe- 1716–1722. lial cells. Cancer Cell Int 8: 15. Yang B, O0Herrin SM, Wu J, Reagan-Shaw S, Ma Y, Bhat KM West KA, Brognard J, Clark AS, Linnoila IR, Yang X, Swain SM et al. (2007). MAGE-A, mMage-b, and MAGE-C proteins et al. (2003). Rapid Akt activation by nicotine and a tobacco form complexes with KAP1 and suppress p53-dependent carcinogen modulates the phenotype of normal human airway apoptosis in MAGE-positive cell lines. Cancer Res 67: epithelial cells. J Clin Invest 111: 81–90. 9954–9962. Winn RA, Van SM, Hammond M, Rodriguez K, Crossno Jr JT, Zhou X, Li Q, Arita A, Sun H, Costa M. (2009). Effects of nickel, Heasley LE et al. (2006). Antitumorigenic effect of Wnt 7a and Fzd chromate, and arsenite on histone 3 lysine methylation. Toxicol 9 in non-small cell lung cancer cells is mediated through ERK-5- Appl Pharmacol 236: 78–84.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene