Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Several inherited genetic polymorphisms in carcinogen-metabolizing enzyme genes have been hypothesized as contributing to lung cancer susceptibility. A number of studies have examined carcinogen-metabolizing enzyme activity, such as aryl hydrocarbon hydroxylase activity, as well as several variant cytochrome P450 1A1 (CYP1A1) alleles as they relate to lung cancer. Initial human studies suggested a positive correlation between aryl hydrocarbon hydroxylase inducibility of peripheral blood lymphocytes when stimulated and assayed ex vivo and presence of lung cancer. However, actual CYP1A1 coding sequence polymorphisms, examined in case-control molecular epidemiology studies, have yielded inconsistent results across studies of Euro-American populations (Hirvonen et al., 1992; Shields et al., 1993; Drakoulis et al., 1994; Sugimura et al., 1994; Spivack et al., 1997). Similar inconsistencies apply to the accumulated literature for polymorphisms of CYP2D6 and CYP2E1 genes and their correlation with the presence of lung cancer (Kato et al., 1992; Hirvonen et al., 1993; Agundez et al., 1994; Hamada et al., 1995; Shaw et al., 1995; Bouchardy et al., 1996; Spivack et al., 1997). Therefore, with respect to tobacco-induced lung cancer susceptibility, the identity of the key relevant phase I carcinogen metabolizing genes, and the key polymorphism(s) in those genes, remains unclear.

CYP1B1 bioactivates many of the same exogenous procarcinogens as the well studied CYP1A1, including polycyclic aromatic hydrocarbons, nitro-aromatics, and arylamines (Sutter et al., 1994; Shimada et al., 1996; Kim et al., 1998). While substrate affinity and kinetics differ for the individual xenobiotic for CYP1B1 versus CYP1A1 (Kim et al., 1998), both enzymes are induced by tobacco smoke, and their expression is aromatic hydrocarbon receptor (Ahr¹)-mediated at one or more xenobiotic response elements (Spink et al., 1998a,b; Eltom et al., 1999). Tobacco-induced expression of CYP1B1 in human lung has been suggested to vary widely interindividually, over several orders of magnitude, as assessed in endobronchial mucosal biopsies taken from active cigarette smokers as compared with nonsmokers (Willey et al., 1997). Thus, variable interindividual expression patterns in human lung are suspected to exist.

There is considerable support in the epidemiologic literature for the position that women are at higher risk than men for lung cancer at any given level of smoking (Ernster, 1994; Ryberg et al., 1994; Zang and Wynder, 1996). The estrogen hormonal environment is thought to synergize with the mutagenicity of inhaled tobacco components. CYP1B1 readily metabolizes 17-estradiol, with its primary hydroxylase activity at C-4 (Hayes et al., 1996; Spink et al., 1998b). In animal models, 4-hydroxyestradiol is carcinogenic (Liehr et al., 1986; Li and Li, 1987). Mechanisms underlying the interplay of hormonal, genetic, and environmental factors are largely speculative at this time.

In an effort to lay the groundwork for assessments of the role of CYP1B1 in lung carcinogenesis, we have assessed CYP1B1 expression at mRNA and protein levels in human lung, used CYP1A1 expression for comparison, and correlated CYP1B1 expression levels with tobacco smoke exposure, gender, lung tissue of origin (tumor versus nontumor), and histologic diagnosis.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Subjects. A group of 16 individuals undergoing lung resectional surgery for suspected carcinoma consented to participate in the study and provided lung tissue for analysis. Subjects came from the Albany, NY region, and surgery was performed at either a tertiary care center (Albany Medical Center) or a large community hospital (St. Peter's Hospital). The study was conducted under the auspices of the respective institutional review boards. Precise mainstream or sidestream tobacco exposure history, self-reported down to the cessation date if applicable, along with exposures to other inhaled toxicants and medications, and other medical history were obtained preoperatively in a direct interview by a trained research nurse.

Tissue Handling. Lung tissue was surgically resected for clinical indications, and if otherwise not needed for diagnostic purposes, the lung sample was quickly and manually divided by the pathologist into "involved" (tumor) versus "uninvolved" (nontumor) tissue on the basis of gross visual appearance, flash-frozen in liquid isopentane or nitrogen within 15 min of blood supply ligation, and placed into the 80°C tissue bank until analyzed. Speed and proper preservation for RNA analyses were preeminent considerations. Blood (30 ml) was collected preoperatively at the time of interview from each subject and stored briefly at room temperature, the plasma fraction was frozen, and lymphocytes were isolated by a Ficoll gradient technique for viable storage in 50% fetal bovine serum, 8% dimethyl sulfoxide in RPMI media and banked at 80°C for future studies.

mRNA Analysis. RNA was extracted from approximately 100 mg of fresh-frozen human lung tissue using a thiocyanate guanidinium-based method (TRI Reagent protocol, Molecular Research Center, Inc., Cincinnati, OH). Great care was taken to keep lung tissue frozen throughout fractionation and pulverizing, by using a liquid N₂-immersed mortar and pestle until the moment of immersion in the guanidinium-containing solution. Yield was generally 1 to 5 µg of total RNA/mg of lung. Multiple replicate gene expression trials for a given individual were performed using the same RNA isolate from a single lung sample. Qualitative reverse transcription (RT)-PCR was performed by oligo-dT isolation of mRNA and RT using Superscript II Reverse Transcriptase (Life Technologies, Gaithersburg, MD) precisely according to the manufacturer's package insert instructions with the exception that dNTP concentration was augmented 8-fold over protocol. PCR of cDNA was performed using two sets of CYP1B1 primers: DS primers (generously provided by Dr. David Spink, Wadsworth Center, NYSDOH) spanned the 3032 bp intron 3 (CYP1B1 GenBank entry HSU56438), allowing for kinetics favorable to the amplification of 1B1 cDNA versus any contaminating genomic DNA [1B1DSF 5'-GCCACTATCACTGACATCT-3', 1B1DSR 5'-CTTGCCTCTTGCTTCTTATT-3', product length = 684]. A second primer set (G1F/G2R) also spanned intron 3 and was run for confirmation on a subset of samples [1B1G1F 5'-GGACGCCTTTATCCTCTCTG-3', 1B1G2R 5'-AAGCAGCACAAAAGAGGAAC-3']. For the comparison transcript CYP1A1, two sets of CYP1A1 cDNA primers were applied, designed either to individually span exon/exon junctions and therefore not amplify contaminating genomic DNA [1A1JWF 5'-CATCCCCCACAGCACAACAAG-3' (Willey et al., 1997), 1A1G2R 5'-AATCACCTTCTCACTTAACACC-3', product length = 1434] or to span multiple exons leading to a kinetically favorable cDNA amplification over any contaminating genomic DNA [1A1G5F 5'-TCCCTGATCCTTGTGATCCC-3', 1A1G2R 5'-AATCACCTTCTCACTTAACACC-3', product length = 190]. A two-stage nested PCR strategy for CYP1A1 cDNA used first stage primer set 1A1G5F, 1A1G2R for 30 cycles, followed by PCR using primer set 1A1JWF, 1A1G2R for 30 cycles, which proved to be a sensitive and specific approach. -actin primers -actin1F 5'-CCACGAAACTACCTTCAACTCC-3', -actin1R 5'-TCATACTCCTGCTGCTTGCTGATCC-3', product length = 270 (Fasco et al., 1993) were used in separate reactions as housekeeping genes. No reactions were multiplexed.

Protein Analysis. Microsomal preparation was performed from human lung tissue by a standard technique (Fasco et al.,1993). Briefly, 100 mg of tissue was pulverized in a liquid N₂-immersed mortar and pestle apparatus and immersed in 1.0 ml of microsomal preparation buffer (0.2 mM phenylmethylsulfonyl fluoride, 1.0 mM dithiothreitol, 1 mM EDTA, 20 mM Tris acetate, 0.14 M KCl). Samples were then sonicated for 15 s and centrifuged at 12,000g for 20 min at 4°C. The supernatant was ultracentrifuged at 100,000g for 60 min at 4°C, and the pellet was resuspended in 0.5 ml of microsomal storage buffer (50 mM Tris acetate, 1.0 mM EDTA, 20% glycerol, 0.2 mM phenylmethylsulfonyl fluoride, 1.0 mM dithiothreitol). Microsomal protein was quantified by BCA Protein Assay kit (Pierce, Rockford, IL). Multiple replicate protein expression trials for an individual were performed on the same microsomal isolation from a single sample.

Nicotine and Cotinine Analysis. Plasma nicotine and cotinine levels were measured using a modification of a standard procedure (Davis, 1986). Five hundred microliters of plasma spiked at 100 ng/ml with deuterated nicotine and cotinine internal standards was diluted with 500 µl of 5 M sodium hydroxide and extracted with methylene chloride. The organic layer was concentrated to dryness, and the solvent was changed to 2-propanol and reduced to 10 µl. Gas chromatography/mass spectrometry was used to separate and detect nicotine and cotinine using the deuterated internal standards to quantitate the amount of nicotine and cotinine present in plasma. The limit of detection for nicotine was 5.0 ng/g of plasma, and that for cotinine was 10.0 ng/g of plasma.

Data Analysis. The univariate analyses explored the factors, taken one at a time, that correlated with a subject ever having been positive for expression of either of the two genes among multiple replicate trials (Table 1, Fig. 2). This ever-positive criteria, where one positive among multiple replicate experimental trials is sufficient to label the subject as positive for that gene's expression at the mRNA or protein level, is standard analysis. Fisher's exact test (Agresti, 1992) was used to evaluate the fraction of the 16 subjects "ever-positive" for one measure of expression (e.g., 1B1 mRNA tumor tissue) versus the fraction ever-positive for another marker of expression (e.g., 1B1 mRNA nontumor tissue). p values express the probability that, for the given sample sizes, the observed difference between the compared fractions ever-positive was a random occurrence. Bonferroni adjustment (Seber, 1997) for multiple comparisons was used in this analysis to establish conservative significance levels at 0.05/n, where n = number of comparisons made in the particular analysis.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

For our 16 subjects, self-described tobacco exposure was distributed as follows: light exposure (n = 2), former smokers (n = 7), recent smokers (within 1-2 weeks, n = 2), and current mainstream smokers (as of preoperative interview, n = 5). There was no detectable nicotine present in any of the plasma samples, implying tobacco exposure preceded the collection of blood at interview by at least several hours and by inference the harvest of lung tissue by several days. There were measurable cotinine levels in two of the four current mainstream smokers, consistent with this inference. The group was comprised of 4 females and 12 males and included an assortment of lung cancer and noncancer histologic diagnoses (Table 1).

The frequency of CYP1B1 expression at mRNA and immunoreactive protein levels is summarized in Table 1 and Figs. 2 and 3, and representative gels are displayed in Fig. 1, A and B. CYP1A1 expression was assayed for reference. The presence of CYP1B1 message was confirmed by both Southern blotting of the PCR product and by direct PCR product sequencing of a subset of samples that excluded any detectable CYP1A1 in the PCR product. All mRNA and protein experiments were run with appropriate positive and negative controls for both tissue quality and assay integrity.

View larger version (48K): [in this window] [in a new window]   Fig. 1.   A, qualitative RT-PCR of CYP1B1 and CYP1A1 in human lung; B, human lung immunoblot for CYP1B1 and 1A1 protein in the same subjects. A, CYP1B1 cDNA was subject to 40 cycles, and CYP1A1 cDNA was subject to two stages of PCR, at 30 cycles each, for maximum sensitivity. Top panel is CYP1A1 PCR product electrophoresed with -actin, middle panel is CYP1A1 electrophoresed alone for clarity, and bottom panel is CYP1B1 electrophoresed with -actin. Dioxin-treated MCF-7 RNA subjected to RT-PCR was used as positive control. T, tumor; NT, nontumor tissue. See Materials and Methods for details. B, positive standards are lymphoblast-expressed CYP1B1 and CYP1A1 at 50.2 and 10.2 fmol. Negative control lanes using 10.2 fmol of the reciprocal protein (CYP1A1 in the case of anti-CYP1B1 primary antibody) are displayed. See Materials and Methods for details.

Data were analyzed using two approaches. First, the univariate analyses examined factors one-at-a-time and correlated these individual factors with gene expression. In this analysis, gene expression was determined by traditional "ever-positive on replicate experiments is positive" criteria to categorize subjects as expressors or nonexpressors. CYP1B1 mRNA and protein were consistently expressed across >75% of subjects for both tumor and nontumor tissue (Fig. 2). Significantly more subjects expressed CYP1B1 than CYP1A1 mRNA and protein; this was true in both tumor and nontumor tissue (p < 0.0019, Fig. 2). Gender, historical smoking status, biomarker smoking status, and tumor histology did not contribute to gene expression by this analysis (p > 0.05, data not presented).

View larger version (23K): [in this window] [in a new window]   Fig. 2.   Fraction of subjects displaying CYP1B1 mRNA and protein expression, using ever-positive criteria, where any replicate trial positive defines a "positive expressor" subject. Standard error bars are shown. CYP1B1 mRNA and protein was consistently expressed across >75% of subjects for both tumor and nontumor tissue. Significantly more subjects expressed CYP1B1 than CYP1A1 mRNA and protein; this was true in both tumor and nontumor tissue (p < 0.0019). The CYP1B1 mRNA and protein expression did not differ between tumor and nontumor tissue (p > 0.05) by this univariate analysis.

The second approach analyzed multiple factors that were associated with gene expression. In this analysis, the fraction of replicate experiments that showed positive gene expression was modeled as the dependent variable. Plausible clinical or biological factors that might impact on the expression of CYP1B1 were modeled against the fraction of trials that were positive. There was substantial variability between individuals in the expression of CYP1B1 message and protein (Fig. 3), not easily explained by multivariate modeling for confounders (Table 2). For example, pooling expression data from both tumor and nontumor tissue, subjects 2 and 5 had a higher fraction of replicate trials demonstrating CYP1B1 mRNA; subject 11 appeared to have a lower fraction of replicate trials demonstrating CYP1B1 mRNA than the group as a whole. For CYP1B1 protein, subjects 2 and 8 demonstrated a higher fraction of replicate trials that were positive, whereas subject 6 was lower than the group as a whole. Discordant mRNA and protein results for several individuals were notable and suggested differing kinetics for the two measures of gene expression. Another predictive factor explaining some of the CYP1B1 mRNA expression frequency variance was tissue source; CYP1B1 mRNA expression frequency in tumor tissues exceeded that demonstrated in nontumor tissue (p = 0.0003). When integrating Bonferroni multiple comparisons corrections for statistical significance (p < 0.002), CYP1B1 protein expression was not higher in the tumor (p = 0.03). Similarly, plasma cotinine levels did not linearly increase with CYP1B1 protein (p = 0.027). Self-described smoking status displayed no correlation with these measures of P450 expression. Gender and tumor histology did not correlate with CYP1B1 expression in this group of subjects.

View larger version (28K): [in this window] [in a new window]   Fig. 3.   CYP1B1 mRNA and protein expression, with each individual replicate experimental trial counted as an observation. Since multiple replicate assays were conducted for each individual, there was a fraction of those trials positive for the expression marker of interest. Each individual's fraction of replicate trials that were positive is represented by a filled circle (). The shaded box represents the 25th to 75th percentile of subjects, the middle half of the data, for the expression marker of interest. White bars denote the median fractions positive. Overlapping data points are slightly separated for clarity.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

We have demonstrated the expression of the carcinogen- and estrogen-bioactivating enzyme CYP1B1 in human lung tissue, across individual subjects, at the mRNA and protein levels. We have, to the best precision available in an observational study, defined the tobacco exposure of the individuals, historically and by plasma biomarkers, and related that to gene expression levels. We have analyzed CYP1B1 expression frequency at mRNA and protein levels, using the common "ever-positive is a positive" standard, by traditional univariate analyses and also used multivariate techniques in the analysis of our replicate experiments.

The occurrence of CYP1B1 and protein at the threshold of detectability in these human specimens makes it impractical to perform quantitative analysis by conventional techniques. Univariate analysis using an ever-positive criteria does not make use of the expression difference between a subject with nine of nine (e.g., subject 5) replicates positive versus another subject with two of nine (e.g., subject 7) replicates positive. Thus, in addition to the traditional univariate primary analysis for our qualitative data, we applied a more detailed statistical approach to correlate biological and clinical factors and gene expression. As part of this secondary, multivariate analysis, we counted each replicate experimental trial on a sample as an observation, analyte presence or absence. This approach avoids arbitrary cutpoints on what fraction of trials is an appropriate threshold to summarize all of them as concordantly positive or "negative", allows examination of individual experimental trial dates as possible confounders, and increases the power of the multivariate analyses without losing data in such trial-set summaries as positive or negative. Thus, in Table 1, the four appropriately controlled trials of CYP1B1 RT-PCR for the tumor tissue from subject 9 were counted as two trials positive and two negative, a total of four observations, rather than applying an arbitrary judgment as all positive or negative. Working at the limits of sensitivity for the respective assays, where trial-to-trial variability is observable, despite proper positive and negative controls performing their respective functions with each trial adequately, we believe this was a conservative and rigorous way to analyze the data. Most of our data on any given individual were concordant among that individual's trials. The numbers of replicate trials performed for any individual differed were based on several factors: a) tissue availability; b) inclusion of only valid replicate experiments, judged by adequacy of positive and negative controls; c) concordance of results of previous trials; and d) historical smoking status. The multiple replication of trials, to a maximum of 10 in those with provocative "current" or "recent" cigarette smoke exposure did not confound results, as there was no relationship between historical smoking status and any of the biomarkers of gene expression studied.

Our raw data, and univariate and multivariate results, differ from those of Hakkola et al. (1997), who reported no measurable CYP1B1 mRNA in a small number of "pooled" human postsurgical lung samples. That same group did detect CYP1B1 mRNA in alveolar macrophages lavaged from volunteers, as well as in nonlung organs. RNA degradation is always suspect when tissues yield negative results. We assume the pooling of different subjects' tissues in that study did not confound the analysis and that their negative results are not due to individual trial data summaries that reduce experimental power, as discussed above. We speculate that their use of a DNase technique to exclude contaminating genomic DNA from the RNA isolation and subsequent cDNA amplification in their study may be problematic. The use of an RNase-free DNase treatment regimen to address genomic DNA contamination of RNA extracts can result in dose-dependent degradation of RNA (Huang et al., 1996), despite strict precautions. Certainty of expression for human lung CYP1B1 mRNA by RT-PCR in our study was conferred by a) the dichotomous nature of a run (present or absent); b) the design of the CYP1B1 primers to span the intron between exons 2 and 3 that was kinetically unfavorable and visually recognizable for contaminating genomic DNA PCR (product length 3716 bp) compared with the cDNA PCR product (684 bp); c) the agreement of the CYP1B1-DS primer set results with a second designed primer set; d) product size from any possible CYP1A1 cDNA product not observed and discordant from the CYP1B1 cDNA product size from this primer set; e) the trial repetition of three to nine times per tumor or nontumor sample; and f) in a subset of samples from several subjects, confirmation by both Southern blotting and direct sequencing of the PCR product.

The presence of human lung CYP1B1 protein identified by immunoblot in our series of patients is in agreement with other, smaller studies (Murray et al., 1997; Tang et al.,1999). Certainty of protein expression for our study lies in the specificity of the anti-peptide primary antibody used (Tang et al., 1999), as well as in our own current data (Fig. 1B). A primary antibody omission study confirmed this specificity; an absence of primary CYP1B1 antibody extinguished the CYP1B1 signal in both experimental and control lanes.

The explanation for interindividual heterogeneity of CYP1B1 expression is not mechanistically addressed by this study. The regulatory control of members of the cytochrome P450 superfamily is complex, varying across individual gene and cell type. Control of CYP1B1 and CYP1A1 expression is thought to involve the binding of a polycyclic aromatic hydrocarbon ligand to Ahr, binding of hsp-90 to the Ahr-ligand complex and translocation to the nucleus, heterodimerization with Arnt, release of hsp90, and interaction with enhancer regions within the 5' regulatory region (Whitlock et al., 1996; Rowlands and Gustafsson, 1997). One can therefore envision a variety of sites and mechanisms resulting in polymorphic CYP1B1 transcription, and by extension, translation. There are probably non-Ahr-dependent factors, some of them cell-type specific, regulating expression of these genes as well (Hakkola et al., 1997; Spink et al., 1998b). Thus, our homogenized lung tissues contain epithelial and nonepithelial cell types, the precise ratio uncontrolled across samples containing identical quantities of RNA or protein, and thus varying across individuals, possibly spuriously introducing heterogeneity of P450 expression (Christou et al., 1995).

Frequency of expression of CYP1B1 at message (p = 0.0003) and possibly protein (p = 0.03) levels was statistically greater in tumors than nontumors, by the multivariate analysis. This finding is consistent with an immunostaining study in a broad range of tissues (Murray et al., 1997) and in breast tumors (McFayden et al., 1999). We did not compare the expression of CYP1B1 in the noninvolved lung tissue of those with lung cancer versus those without lung cancer (there was only one patient without a chest malignancy in this series). Therefore the significance of "underexpression" of CYP1B1 in nontumor tissue with respect to carcinogenesis is not clear. However, this finding of elevated tumor CYP1B1 expression could potentially be used therapeutically for preferential bioactivation of anticancer prodrugs.

Several factors did not contribute to expression at the mRNA level of CYP1B1 including self-reported smoking history. Self-reported smoking history was assessed several days before surgery, during the preoperative interview and phlebotomy session, thus probably unlinking smoking history from tissue gene expression. Furthermore, given identical cigarette pack-year exposure histories, there is wide variation in the literature on resulting CYP1A1 expression, and we infer the same is true for CYP1B1, largely because proximate tobacco exposure is probably more relevant than total lifetime tobacco exposure for P450 expression measurements. A previous study by Willey et al. (1997), using bronchial mucosa cells and biopsies from healthy current smokers and nonsmokers, suggested induction of these genes' expression was generally detectable by quantitative PCR, although there was considerable interindividual variation across several orders of magnitude. We confirmed smoking status by plasma nicotine and cotinine levels. Plasma cotinine levels in our subjects was weakly associated in the multivariate model with protein level expression of CYP1B1. While cotinine is not known to be a direct inducer of either CYP1B1 or CYP1A1, its turnover kinetics may parallel those of the respective carcinogen-metabolizing enzyme induced by other components of the tobacco smoke mixture. Neither gender nor tumor histology correlated with the expression pattern for CYP1B1 or CYP1A1 by either univariate or multivariate analyses within the statistical power of the study.

We conclude that CYP1B1 is commonly expressed in human lung at both mRNA and protein levels. Given that previous studies suggest this enzyme has carcinogen-metabolizing activity for many tobacco-smoke constituents, and has substantial 17-estradiol metabolism capacity with resultant carcinogenicity, one may speculate on a link between these respective metabolic roles and the human epidemic of lung cancer among females. Further study of the regulatory control of this gene is under way in this and other laboratories.