Monoallelic Up-Regulation of the Imprinted H19 Gene in Airway Epithelium of Phenotypically Normal Cigarette Smokers¹

Genomic imprinting is an epigenetic form of gene regulation in which the two parental alleles are differentially expressed, resulting in monoallelic expression with either the maternal or paternal copy silenced in normal somatic tissues (1, 2). Imprinting is thought to play an important role in embryonal development and growth regulation (1, 2). Alterations of the normal imprinting pattern (i.e., LOI³) have been observed in a variety of human tumors, including lung cancer (3, 4, 5, 6). Most examples of LOI in human cancers involve three genes: IGF2, a potent mitogen (6); p57^KIP2, a cyclin-dependent kinase inhibitor (4); and H19, an untranslated RNA with possible oncofetal or tumor suppressor function (3, 7, 8).

H19 was the first human gene recognized to be paternally imprinted (i.e., maternally expressed; Ref. 9). The 2.7-kb human H19 gene, located on the short arm of chromosome 11, band 15.5, contains five exons and four small introns (9). As in a number of other imprinted genes, H19 is transcribed by RNA polymerase II, capped, spliced, and polyadenylated as a typical mRNA, although it generates no known protein product (9). In both mice and humans, H19 is located within a cluster of at least five imprinted genes, including IGF2 (10, 11). H19 and IGF2 are reciprocally imprinted and coordinately regulated by an intergenic imprinting center and a common enhancer region (10, 12, 13). H19 is normally expressed during embryogenesis at high levels in many organs (14). In humans, there is biallelic expression of H19 in the placenta at <10 weeks of gestation, but expression becomes monoallelic after 18–20 weeks (15). In adults, H19 expression remains monoallelic, with expression primarily in skeletal muscle, thymus, heart, and lung (7). The function of H19 is unknown, but it is postulated to function as a regulator of translation, and possibly as a tumor suppressor or oncofetal gene (7, 8, 9, 16, 17).

The focus of this study is the status of the expression and imprinting of the two parental H19 alleles in the airway epithelium of individuals with a history of mild to moderate (average, 20 pack-years) cigarette smokers but who are otherwise phenotypically normal. The data reveal that these individuals have a dramatic elevation of H19 RNA levels in the airway epithelium compared with that of nonsmokers. Importantly, the smoking-induced airway epithelial overexpression of the H19 gene is not because of LOI since H19 expression remains monoallelic, with up-regulation of only the active H19 allele. Given the close association of cigarette smoking with lung cancer (18, 19) and lung cancer with LOI of H19 (3), it is likely that monoallelic up-regulation of H19 is an early response to smoking and that this progresses to LOI of the paternal H19 gene as the burden of smoking increases, and the airway epithelium progresses from normal to premalignant to frank neoplasia. If this concept is correct, elevated levels of airway epithelial H19 expression and eventual LOI may represent early markers in progression of the epithelium toward lung cancer.

Under an Institutional Review Board-approved protocol, volunteers were solicited and subjected to a phone survey to establish eligibility to enter the study. To be considered eligible to enter the study, individuals had to be male or nonpregnant female nonsmokers or smokers, with a minimum 10 pack-year smoking history, in overall good health, and without a history of either chronic lung disease and either recurrent or recent acute pulmonary disease within the preceding 3 months before entry into the study. Of those interviewed, 47% were brought in for screening, which included general assessment, blood chemistry, complete blood count, drug screen, chest X-ray, pulmonary function testing, and urine chemistry. Those who did not meet eligibility criteria and/or did not have a normal physical examination and diagnostic tests mentioned above were excluded from participation in the study. Of those screened, 39% passed the inclusion/exclusion criteria and were scheduled for fiberoptic bronchoscopy and airway brushing. Of those, 93% completed the procedure. Satisfactory microarray data were obtained in 88% of subjects who underwent bronchial brushing. When possible, two independent samples of airway epithelium were obtained from each subject at the time of each bronchoscopy from the third branching of the bronchi in both the left and right lower lobes. The demographic characteristics of the nonsmokers and smokers were similar with respect to sex (P = 0.1, χ²) and race (P = 0.5, χ²), as well as age (P = 0.3, two-tailed t test; Table 1).

Fiberoptic bronchoscopy was performed for collection of airway epithelial cells using methods developed in our laboratory to ensure high quality RNA for analysis (20, 21, 22). The bronchoscope was positioned in the desired segment of the bronchial tree, and under direct visualization, a 1-mm disposable brush (Wiltek Medical) was advanced through the channel of the bronchoscope. Airway epithelial cells were obtained by gently gliding the brush back and forth on the desired airway three to five times/brush, using 10 brushes for each area. The cells were suspended by flicking the brush in a 5-ml tube containing LHC8 medium (Life Technologies, Inc.) and maintained at 4°C until processed for isolation of DNA and RNA (22). An aliquot of 0.5 ml was kept at 23°C for differential cell count. Cell viability was evaluated by Trypan Blue exclusion and expressed as a percentage of the total cells recovered. Cells were counted on a hemocytometer and differential cell count (epithelial versus inflammatory cells) was assessed on sedimented cells prepared by cytocentrifugation (Cytospin 11; Shandon Instruments, Pittsburgh, PA) and stained with DiffQuik (Baxter Healthcare, Miami, FL).

The HuGeneFL Affymetrix array, containing probes for ∼6800 human genes, was used for analysis of mRNA levels in all airway epithelium samples. All samples were prepared as specified by Affymetrix. Total RNA was extracted from the brushed cells by the TRIzol (Life Technologies, Inc., Carlsbad, CA) method followed by RNeasy (Qiagen, Valencia, CA) to remove residual DNA, a procedure giving a yield of 2–4 μg/10⁶ cells. First strand cDNA was synthesized using the T7-(dT)₂₄ primer [sequence 5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-(dT)₂₄-3′, high-performance liquid chromatography purified] and converted to double-stranded cDNA using Superscript Choice system (Life Technologies, Inc.). Double-stranded cDNA was purified by phenol-chloroform extraction and precipitation and the size distribution examined by agarose gel electrophoresis. Biotinylated RNA was synthesized using the BioArray High Yield reagents (Enzo), purified by the RNeasy kit (Qiagen), and fragmented immediately before use. The microarrays were processed on the fluidics station under the control of the Microarray Suite software and read. All RNA samples were hybridized to test chips as indicated by Affymetrix protocols and they all passed quality control (ratio of 3′–5′ controls < 3:1). The data were analyzed using the GeneSpring software (Silicon Genetics, Redwood City, CA). Normalization was carried out using GeneSpring software sequentially as: (a) per microarray sample, dividing the raw data by the 50^th percentile of all measurements; and (b) per gene, dividing the raw data by the median of the expression levels for the gene, over all samples. For additional data analysis, only expressed genes were taken into account (Microarray Suite, Absolute Call “P” in at least 40% of the samples). Data were obtained for all 3020 of these genes, including the imprinted genes H19 on chromosome 11 (ch 11), IGF2 (ch 11), the IGF2 receptor (ch 6), pleiomorphic adenoma gene-like 1 (ch 6), the growth factor receptor-bound protein 10 (ch 7), mesoderm-specific transcript homologue (ch 7), necdin (ch 15), imprinted in Prader-Willi syndrome (ch 15), and small nuclear ribonucleoprotein polypeptide N (ch 15; Ref. 11), as well as nonimprinted control genes, nonimprinted small nuclear ribonucleoprotein polypeptide E, actin, and GAPDH. For those individuals where two microarrays were available (i.e., one for each lung), the normalized values for both lung lobes were used. These values were compared against each other to assess the correspondence among the H19 levels at two sites within the same individual.

H19 expression levels were also measured using real-time quantitative RT-PCR (TaqMan), using the same samples that were assessed by microarray analysis. H19 RNA levels were measured relative to GAPDH and adult human lung RNA (Stratagene) by real-time quantitative PCR with fluorescent TaqMan chemistry (using the ΔΔ Ct method; PE Biosystems, Instruction Manual). The H19 TaqMan reactions were optimized and shown to have an equal amplification efficiency as the GAPDH amplification. The H19 forward primer (TGCTGCACTTTACAACCACTG) lies in exon 4 of the human H19 gene and the reverse primer (ATGGTGTCTTTGATGTTGGGC) lies in exon 5. The TaqMan probe (TCGGCTCTGGAAGGTGAAGCTAGAGGA) spans the junction of exons 4 and 5, thus providing specificity to the RNA and not the genomic DNA.

Imprinting status was determined using a known RsaI polymorphism present in exon 5 of the human H19 gene (15). Only informative samples (i.e., heterozygous individuals) were used for this analysis. Allele frequency was determined by PCR amplification of H19 exon 5 using genomic DNA extracted from blood mononuclear cells. Genomic DNA samples extracted from airway epithelium or blood mononuclear cells, and airway epithelial RNA samples were compared by assessing the presence or absence of the polymorphic RsaI site in exon 5, by PCR (for genotype) and RT-PCR (for expression; Ref. 15). Primers were designed to amplify either the DNA or the RNA specifically. The following oligonucleotides were used for amplifying a 1153-bp fragment of exon 5 from genomic DNA: forward primer 5′-ACTGCCCCGACCTCTGTCTTCTAC; reverse primer 5′-CCCCATCCCCCTTTTCATGTA. If the RsaI site was present, digestion with RsaI yielded three fragments, 787 and 346 bp in length, plus an undetectable 20-bp fragment produced by an additional RsaI site in that location. To specifically amplify the RNA, the forward primer (5′-GGCTCTGGAAGGTGAAGCTAGAGG) was designed to span the exon 4-exon 5 junction, and the reverse primer (5′-TTTTTTTTTTGCTGTAACAGTGTT) was complementary to the end of the RNA and beginning of the polyadenylic acid tail. The resulting fragment was 637 bp, and if the RsaI site was present, was digested by that enzyme into fragments of 375 and 262 bp.

Primary HBE cells (gift of Dr. Michael Welsh, University of Iowa) were cultured on Millicell-PCF membrane inserts (Millipore, Bedford, MA) in 1:1 DMEM:Ham’s F12 media supplemented with 2% Ultroser G and antibiotics (100 units/ml penicillin, 100 μg/ml streptomycin, 50 μg/ml gentamicin, Life Technologies, Inc.; 2 μg/ml fluconazole, Diflucon, Pfizer; and 1.25 μg/ml amphotericin B, Sigma, St. Louis, MO) at 37°C in a 5% CO₂ atmosphere. Viability of the cells was confirmed before each experiment by measurement of transepithelial resistance. CSE was prepared using a modification of the method used by Wyatt et al. (23). Four research grade cigarettes (2R4F, University of Kentucky) were bubbled into 50 ml 1:1 DMEM:Ham’s F12 medium using a vacuum pump apparatus. Five percent of CSE was then prepared from this stock solution, and the pH of the resultant solution was adjusted to 7.40 using 1 n HCl. The 5% CSE was filtered through a 0.22-μm filter to remove particles and bacteria before use. RNA was isolated from both experimental and control groups at 24, 48, and 72 h after CSE exposure using TRIzol (Life Technologies, Inc.) initially and was purified using RNeasy (Qiagen). Samples were obtained in triplicate for each time point. H19 cDNA was prepared by RT-PCR: forward primer 5′-TTGAATCCGGACACAAAACCCTC; reverse primer 5′-CAGGAGCCCTGGACTCATCA. TaqMan PCR was then performed using the primers and TaqMan probe described above. H19 expression levels relative to GAPDH were assessed. Each data point was generated from triplicate wells, and a total of six separate data points were obtained at each time interval.

Comparisons of the ages of the smoking and nonsmoking subjects and the yield, percentage of cell types, and viability of epithelial airway samples were performed by Student’s two-tailed t test. The H19 expression levels in the right and left lungs of the same individuals and the correlation between expression levels measured by microarray and TaqMan analysis were performed using linear regression. The Ps for all expressed genes were calculated comparing the nonsmokers to the smokers, using the Wilcoxon-Mann method with the Benjamini-Hochberg correction for False Discovery Rate (GeneSpring software).

The observations on H19 overexpression in the airway epithelium of smokers emerged from an ongoing study in our laboratory to identify changes in gene expression patterns in the airway epithelium of disease-free smokers (∼20-pack-year smoking history) compared with a matched group of nonsmokers. The overall aim of the study is to identify potential susceptibility factors for the development of lung disease in smokers, including early markers for progression to lung cancer. The study is nonbiased in that it is designed to identify all genes up-regulated or down-regulated in a significant fraction of healthy smokers relative to nonsmokers without bias to classes of genes or presumed pathogenic processes.

Affymetrix HuGeneFL microarrays were used to evaluate mRNA levels in the airway epithelium of nonsmokers and smokers. The HuGeneFL array contains probes for assessing the expression level of ∼6800 full-length human genes, most of which are well characterized. The data were analyzed for a total of 40 microarrays from 21 individuals, including 7 nonsmokers and 14 smokers (Table 1). For the majority of individuals, RNA obtained separately from the right and left lungs was assessed independently, with samples obtained from only one lung in a few individuals. These 40 microarrays passed quality control as assessed by the GeneSpring software (Silicon Genetics). Analysis of the epithelial cell samples used for these 40 microarrays demonstrated that smokers and nonsmokers were comparable with respect to yield (P = 0.7) and percentage of nonepithelial cells (P = 0.6). The cells derived from smokers had on average 12% lower viability compared with cells recovered from nonsmokers (P < 0.005; Table 1). To eliminate genes not expressed in airway epithelium or expressed at low levels, those genes that were called absent by the Microarray Suite software (Affymetrix) in >60% of the 40 microarrays were discarded before additional analysis. The number of genes remaining (i.e., called present in >40% of the microarrays) was 3020. Using this subset of genes, nonparametric statistical methods (GeneSpring software) were used to identify genes that were expressed at a higher or lower level in a significant number of smokers versus nonsmokers. Of the 3020 genes that were expressed, there were a total of 47 genes that were significantly (P < 0.01) up- or down-regulated in smokers compared with nonsmokers. As expected, there were some genes whose expression is known to be affected by cigarette smoke such as those involved in antioxidant mechanisms (e.g., glutathione S-transferase A2 and glutathione peroxidase 2), those that participate in xenobiotic detoxification and activation (e.g., CYP1B1), and those whose expression is altered as part of the normal airway epithelial response to cigarette smoke, (e.g., MUC5AC). However, there were also many other genes, which have not been previously linked to tobacco exposure, the expression of which was found to be significantly altered in smokers versus nonsmokers. These include those involved in cellular intermediary metabolism, i.e., transaldolase, transketolase, phosphogluconate dehydrogenase, and glucose 6-phosphate dehydrogenase, all enzymes of the pentose phosphate pathway, which contribute to the ultimate generation of NADPH and subsequent regeneration of glutathione in its reduced state. There were also a number of other genes such as the glycoproteins hevin and tenascin-C, as well as the transcription factor pirin, which have a possible link to carcinogenesis (pirin, with proto-oncogene Bcl3; hevin and tenascin-C with lung cancer) but whose link to cigarette smoke has not yet been established.

Among the genes significantly up-regulated in smokers was H19. A 5-fold up-regulation of H19 expression was observed in the smokers’ samples compared with the normalized median expression level for nonsmokers (P < 0.00001; Fig. 1). The data are specific to this imprinted gene, i.e., other imprinted genes were expressed at a similar level in the airway epithelium of smokers and nonsmokers. For example, IGF2, located in the same imprinted cluster as H19 on chromosome 11 and reciprocally expressed under some circumstances, was expressed at similar levels in smokers and nonsmokers (P = 0.8). The other imprinted genes, the expression of which was unchanged in smokers, included: the IGF2 receptor (P = 0.9) and pleiomorphic adenoma gene-like 1 (P = 0.2) on chromosome 6; the growth factor receptor-bound protein 10 (P = 0.6) and mesoderm-specific transcript homologue (P = 0.9) on chromosome 7; necdin (P = 0.7), imprinted in Prader-Willi syndrome (P = 0.6); and small nuclear ribonucleoprotein polypeptide N (P = 0.6) on chromosome 15. Similarly, most nonimprinted genes were expressed at the same level in smokers and nonsmokers with the exception of the 47 genes expressed at statistically significant different levels in smokers and nonsmokers. For example, the (nonimprinted) small nuclear ribonucleoprotein polypeptide E (P = 0.3) and the commonly used control genes cytoplasmic actin (P = 0.4) and GAPDH (P = 0.9) were all expressed at the same level in smokers and nonsmokers (Fig. 1). As an additional control, H19 levels in the airway epithelium were assessed in both the left and right lungs of the same individuals, including both nonsmokers and smokers (Fig. 2). The data demonstrated a good correlation in the H19 levels at different sites of the airways of the same individuals (r² = 0.86, P < 0.0001).

To confirm the microarray data showing overexpression of H19 in the airway epithelium of smokers, H19 expression levels were assessed by an independent method using the same RNA samples studied by microarray analysis. TaqMan real-time quantitative PCR was used with a primer design that specifically amplified H19 RNA but not genomic DNA. The data showed a good correlation between the expression levels measured by microarray and TaqMan methods, reinforcing the validity of the observation with the microarray analysis that H19 mRNA levels are markedly elevated in the airway epithelium of smokers compared with nonsmokers (r² = 0.59, P < 0.001; Fig. 3). Given the established validity of using microarrays as a means of evaluating alterations in gene expression, as well as the limited supply of RNA from volunteer subjects and costs of normal volunteers undergoing bronchoscopy, the verification of every gene identified as differentially expressed by microarray technique using additional methods such as TaqMan PCR was not undertaken.

Up-regulation of H19 expression may result from a number of mechanisms, including LOI, resulting in biallelic expression (3). To assess the imprinting status of H19 in the airway epithelium samples, a known RsaI polymorphism in exon 5 of the H19 gene was used (15). With appropriate primer design, specific amplification of either the H19 RNA or DNA containing exon 5 could be achieved, and the PCR product digested with RsaI to show the presence or absence of the site (Fig. 4,A). As expected, when PCR-amplified genomic DNA showed homozygosity for the RsaI polymorphism, either for the presence (cutter) or absence (noncutter) of the site, the amplified RNA from the same subjects showed the same pattern of RsaI digestion. These individuals are not informative regarding their imprinting status because it is not possible to use homozygosity to determine whether either one or both of the H19 alleles is expressed. Of a preliminary screen of 45 random blood lymphocyte DNA samples, the presence of the RsaI site was observed at a frequency of 82%, i.e., there are fewer alleles in the population as a whole that do not have the RsaI site at that position, accounting for the paucity of homozygous noncutters. For the heterozygotes in which airway epithelium RNA was assessed, four of five individuals expressed the RsaI cutter allele and one expressed the RsaI noncutter allele. When the four informative heterozygote smokers were examined, the patterns from the RT-PCR analysis of airway epithelial H19 RNA showed consistently that only one H19 allele was expressed (Fig. 4, B and C; Table 2). None of the heterozygotes expressed both alleles. These individuals had smoking histories of 10, 20, 27.5, and 10.5 pack-years, compared with the average of 22.8 pack-years for all subjects. H19 expression was up-regulated (average of left and right) by 4.4-, 1.4-, 5.4-, and 2.4-fold, respectively, for the informative subjects versus an average 5.0-fold for all smokers. Thus, in individuals with an average of 20 pack-year smoking history who have increased airway epithelial expression levels of H19, imprinting is retained and the overexpression of H19 most likely results from increased expression from the active allele.

Primary HBE cells were exposed to CSE in vitro in an attempt to additionally examine the role of cigarette smoke in H19 gene expression suggested by the up-regulation observed in airway epithelium from healthy volunteers. HBE cells were used because they most closely mimic airway epithelial cells in their natural environment in vivo (24). Approximately 2-, 2.5-, and 4.5-fold up-regulation of H19 expression at 24, 48, and 72 h, respectively, was observed in HBE cells exposed to CSE compared with the control group cultured in 1:1 DMEM:Ham’s F12 media without CSE (Fig. 5).

H19 is a paternally imprinted gene postulated to play a role in embryonic development and oncogenesis (7). Although its function is not known, LOI with expression of both parental genes has been observed in a variety of neoplasms, including lung cancer (3, 5, 8). This study demonstrates that expression of the H19 gene is markedly up-regulated in the airway epithelium of cigarette smokers. The observed smoking-induced up-regulation of H19 gene expression in airway epithelium is not due to an alteration of the normal imprinting pattern, as assessed by allele-specific expression analysis using a known H19 transcribed polymorphism. On the basis of these observations and the known association of lung cancer with H19 LOI in previous studies (3), the observations in this study lead to the interesting hypothesis that cigarette smoking initially induces up-regulation of the normally active, maternal H19 allele, with eventual temporal progression to H19 LOI as the burden of smoking increases, paralleling the sequential pathologic transition in the epithelium itself from normal to neoplastic. To prove this hypothesis that monoallelic increase in H19 gene expression in the airway epithelium of healthy smokers may prelude the loss of H19 imprinting and alteration of the gene’s expression in lung cancer, however, would necessitate that the same group of individual smokers be tracked over a number of years. This would be a prohibitively complex task, far beyond the scope of the current study. In this context, we, therefore, believe that the data we have accumulated can be used instead as a point of reference with respect to findings about H19 and its potential relevance to lung cancer. If proven correct, over time, assessment of the levels of expression and eventual LOI of H19 could be early markers in the progression of airway epithelium toward malignant transformation and could ultimately serve as a useful adjunct in risk stratification and early screening for lung cancer.

The functional nonequivalence of the parental genomes is due to an epigenetic phenomenon, genomic imprinting, by which the expression of specifically marked or imprinted genes depends on whether they are inherited from the mother or father (1, 2). The parental imprinting marks are determined by patterns of DNA methylation and are erased and reestablished during gametogenesis and maintained somatically (1, 2, 21). To date, 41 imprinted genes have been described in human and mouse (11).⁴ Imprinted genes are unusually rich in CpG islands and often have sequence elements that are methylated on only one of the two parental alleles (25, 26). These allele-specific differentially methylated regions play an important role in control of expression of imprinted genes (1, 2, 26). Approximately 80% of imprinted genes are physically linked in clusters where they may be under the control of imprinting control centers (2, 11, 25).

H19 was the first human gene recognized as being expressed solely from the maternal allele, i.e., to be paternally imprinted (9, 15, 27, 28). Studied initially in murine models, the gene was isolated from differentiating myoblasts as a gene regulated by Raf, involved in α-fetoprotein gene expression and from embryonic stem cells differentiating to embryoid bodies (29, 30). The human H19 RNA was subsequently identified in hepatoma cells and cytotrophoblasts (9, 31). The 2.7-kb human H19 gene is located on chromosome 11p15.5 and contains five exons and four small introns (9). There is an estimated 77% sequence homology between the murine and human H19 genes (7, 25). In both species, the H19 gene is located in a cluster containing at least five additional imprinted genes (11), including: (a) p57^KIP2, a cyclin-dependent inhibitor that causes G₁-S arrest (32); (b) KvLQT1, a voltage-gated potassium channel (33); (c) TSSC3, homologous to mouse TDAG51, implicated in Fas-mediated T lymphocyte apoptosis in mice (34); (d) TSSC5, a putative transmembrane protein-encoding gene (35); and (e) IGF2 (36). The H19 and IGF2 genes, located within 200 kb, are reciprocally imprinted and coordinately regulated by an intergenic imprinting control center and a common enhancer region downstream of the H19 gene (10).

The human H19 gene shares a feature associated with many imprinted genes: it generates no known protein product (2, 9). Unlike other noncoding RNAs, it is transcribed by RNA polymerase II, capped, spliced, polyadenylated, and transported to the cytoplasm, where the 2.3-kb mature transcript is associated with 28S cytoplasmic particles, but not with polyribosomes (9). The function of the H19 RNA has not been elucidated. It is postulated to act as a ribo-regulator, similar to the 3′-UTR of the tropomyosin mRNA during muscle cell differentiation and Xist/XIST during X-inactivation (2, 7). Homozygous murine knockouts with a deletion spanning the 2.7-kb H19 gene plus 10-kb length of 5′ flanking sequence showed no deleterious phenotype. Heterozygous animals inheriting the deletion from the father had no demonstrable phenotype but those inheriting the deletion from the mother had increased birth weights, perhaps, because of the constitutive expression of IGF2 (12). H19 expression occurs at much lower levels in adult mouse and human tissues than in fetal tissues; in adults, H19 is expressed primarily in skeletal muscle, thymus, heart, and lung (7). In human fetuses, H19 expression is variable; there is biallelic expression in placenta at <10 weeks of gestation, but expression is monoallelic after 18–20 weeks of gestation, as observed in many tissues, including lung, adrenal gland, spleen, thymus, and placenta (15).

The imprinted genes most frequently altered in human cancer are IGF2, p57^KIP2, and H19. Imprinted genes are hypothesized to be involved in carcinogenesis in several ways (5, 7, 8): (a) loss of heterozygosity or UPD in an imprinting control center may result in deletion or absence of the only functional copy of a tumor suppressor gene; (b) LOI or UPD of an imprinted gene that promotes cell growth may result in inappropriate overexpression of a potential oncogene; and (c) mutational inactivation of an imprinting control region could result in abnormal expression of both tumor suppressors and growth promoter genes present within one cluster.

H19 overexpression has been documented in a variety of adult tumors, including breast, lung, and esophageal cancers, as well as bladder carcinoma and choriocarcinoma cell lines (8, 27). Assessment of the imprinting status of the H19 gene in 79 surgical specimens of lung cancer demonstrated that 38% of the heterozygous (informative) samples showed LOI accompanied by overexpression (3). Hypomethylation of the promoter region of the abnormally expressed paternal allele was observed in some of the specimens.

H19 has also been implicated in human genetic diseases associated with increased cancer risk (5, 7, 8, 37). For example, BWS (5, 25) is linked to chromosome 11p15 where the H19/IGF2 cluster is located. BWS is a prenatal and postnatal overgrowth (hemihypertrophy, macroglossia, and visceromegaly) syndrome, frequently associated with a predisposition for childhood tumors, in particular Wilms’ tumors and rhabdomyosarcomas (37). The rate of Wilms’ tumor formation in BWS patients is 100-fold higher than in the normal population (37). A common molecular event in those patients is biallelic expression of the IGF2 gene, often accompanied by aberrant imprinting patterns of H19 (38).

Imprinting of the IGF2/H19 locus results in expression of H19 from the maternal allele and expression of the IGF2 gene from the paternal allele (10). The regulation of the imprinting and the expression of both genes is linked. In the mouse, expression of IGF2 and H19 is regulated by two enhancers downstream of the H19 gene (12, 13). In the paternal chromosome, the use of these enhancers is biased toward IGF2 expression because the 5′ flanking sequences of H19 are methylated on the paternal allele, silencing the H19 promoter (39). A differentially methylated region located 90 kb downstream of the IGF2 gene and 2 kb upstream of the H19 gene functions as an imprinting control region (1, 2). This region contains several binding sites for the chromatin insulator factor CTCF (1) that function to block the action of a downstream enhancer on the IGF2 promoter, preventing its expression from the maternal allele. In the paternal allele, methylation of the imprinting control region abolishes binding of the insulator factor, and the IGF2 gene is expressed (1). The 5′ flanking region immediately adjacent to the human H19 gene contains a TATA-less promoter (40). Little is known about the cis- and trans-factors regulating the expression of the human H19 promoter, although a minimal promoter region has been mapped within 823 bp upstream from the transcription start site (40). There is a critical putative CCAAT box and two putative C/EBP (CCAAT/enhancer binding protein) binding sites located between nucleotides -229 and -114 (41), possibly related to the balance between cell proliferation and growth arrest during terminal differentiation (42).

The fact that H19 RNA levels are increased in the airway epithelium of smokers compared with nonsmokers, whereas the expression level of the reciprocally imprinted adjacent IGF2 gene is comparable in both smokers and nonsmokers, is interesting because imprinting of the H19 and IGF2 genes is under the control, in part, of a shared differentially methylated imprinting control region located between the two genes (1, 10). The data in this study suggest that at least in airway epithelium of 20 pack-year smokers, the H19 and IGF2 genes are not coordinately regulated. Uncoupling of control of expression of H19 and IGF2 has also been observed in somatic cell cultures from mice with UPD of distal chromosome 7, where the H19/IGF2 cluster is located in the mouse, apparently because of de novo methylation of sites upstream of IGF2 and within the H19 promoter (43). Uncoupling of H19 and IGF2 expression has also been observed in invasive cervical carcinomas, breast cancers (44, 45), hepatoblastoma and hepatocarcinoma cell lines, human hepatocellular carcinoma, and fibroblast cell lines from patients with BWS (35, 38, 46).

In vitro experiments in which cultured human airway epithelial cells were exposed to cigarette smoke support the investigators’ hypothesis that cigarette smoke influences H19 gene expression patterns. Primary HBE cells were chosen for examination in attempts to both mimic as closely as possible the in vivo airway epithelial architecture and function, as well as to avoid any unwanted preexisting alterations in gene expression or regulation that might be observed with transformed cell lines. The up-regulation of H19 gene expression in primary HBE cells, although modest, suggests a direct causality between cigarette smoke exposure and alteration in H19 gene expression. Moreover, it appears that the up-regulation of H19 expression in cultured primary HBE cells increases progressively over time, suggesting a possible dose-dependent effect or chronicity involved in the interaction between H19 and the effects of cigarette smoke on this gene’s regulation. This data, together with the observations in this study, make it likely that the two imprinted genes, IGF2 and H19, share common regulatory elements, whereas other independent factors control H19 expression, including the response to environmental stresses such as cigarette smoking.

We thank Mary Harris and Charleen Hollmann for assistance in recruitment of study subjects; Jenny Xiang for advice on microarray data analysis; and Nahla Mohamed for help in preparing this manuscript.