Microarray analysis of nicotine-induced changes in gene expression in endothelial cells

SHAOLI ZHANG, IAN N. M. DAY and SHU YE

Human Genetics Research Division, School of Medicine, University of Southampton, Southampton, SO16 6YD, United Kingdom


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
Cigarette smoking causes vascular endothelial dysfunction and is a major risk factor for cardiovascular diseases. Nicotine, a major constituent of cigarette smoke, has been shown to alter gene expression in endothelial cells; however, the regulatory pathways involved remain to be defined. We hypothesized that there might be distinct pathways that could be identified by systematic transcriptome analysis. Using the cDNA microarray approach, we ascertained the expression of over 4,000 genes in human coronary artery endothelial cells and identified a number of nicotine-modulated genes encoding a protein involving in signal transduction or transcriptional regulation. Among these were phosphatidylinositol phosphate kinase and diacylglycerol kinase, which are regulators of the inositol phospholipid pathway. Changes were also detected for transcription factors cAMP response element binding protein and nuclear factor-{kappa}B, of which the activities of both have been previously shown to be altered in nicotine-stimulated cells. The data from this study are relevant to understanding the mechanisms underlying the pathophysiological effect of nicotine and smoking, particularly on endothelial function and pathogenesis of atherosclerosis.

microarray; inositol phospholipid pathway; cAMP response element binding protein; nuclear factor-{kappa}B


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
ENDOTHELIAL INJURY resulting from chemical, mechanical and/or biological insults is believed to be an important mechanism initiating the pathogenesis of atherosclerosis (16). Cigarette smoking, a major risk factor for atherosclerosis-related diseases, causes endothelial dysfunction (14). Although it has been shown that both cigarette smoke and its major constituent, nicotine, can alter the expression of a number of genes in endothelial cells (2, 3, 11, 21, 28), the regulatory pathways involved remain to be defined. We hypothesize that there may be distinct pathways that could be identified by systematic transcriptome analysis.

The cDNA microarray technique, in contrast with classic RNA methodologies such as Northern blot analysis, which study one gene at a time, enables systematic examination of gene expression. This holistic approach may provide clues to an understanding of complex pathways and their interactions in gene regulation under different conditions.

In this study, we used this systematic approach to ascertain nicotine-induced changes in gene expression in vascular endothelial cells. Gene expression profile in primary human coronary artery endothelial cells stimulated with nicotine (at the concentration of 10 µM) was compared with that in unstimulated cells. The results of the cDNA microarray analyses were verified by examining selected genes using the RT-PCR technique.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 

Cell culture.
Primary human coronary artery endothelial cells (HCAEC) were purchased from Cell Applications (San Diego, CA) and cultured in endothelial cell growth medium (Cell Applications) supplemented with trace elements, growth factors, and antibiotics at 37°C in a humidified atmosphere with 95% air-5% CO2. When cells reached subconfluence, (-)-nicotine (98–100% free base; Sigma, Poole, UK) was added to the medium to a final concentration of 10 µM, which is a concentration frequently used in in vitro physiological studies of nicotine effects (6, 12, 24, 26), although this is higher than levels (up to 0.6 µM) of nicotine found in arterial blood of smokers (9, 15) and might not completely resemble the in vivo situation. Cells were then cultured with nicotine for 24 h and harvested by trypsinization. Controls were cells cultured under the same conditions but were not exposed to nicotine. For both nicotine-treated and control cells, no change in morphology was observed by light microscopic examination, and viability of the cells was estimated to be 98–100% using the trypan blue exclusion method.

RNA extraction.
The total cellular RNA from cultured HCAEC cells was extracted using the SV Total RNA Isolation System (Promega, Southampton, UK) according to the manufacturer’s instruction. Briefly, cells were lysed by a lysis buffer containing guanidine thiocyanate. After removing cellular proteins from the lysate, RNA was precipitated, then immobilized in a column containing silica glass fibers and subjected to RNase-free DNase I digestion to remove genomic DNA. The RNA sample was further purified by washing and finally eluted from the column into distilled water. Integrity of the RNA sample was verified using formaldehyde agarose gel electrophoresis.

Preparation of cDNA probe.
Radiolabeled cDNA probes were generated by reverse transcription of RNA extracted from nicotine-treated and untreated (control) HCAEC cells, as follows. First, 1 µg of total RNA was mixed with 2 µg of oligo dT15 (Promega) and incubated at 70°C for 10 min. Then, 40 nmol each of dATP, dGTP, and dTTP, 200 U of Moloney murine leukemia virus (MMLV; RNase H-) reverse transcriptase (Promega) and 100 µCi (33 nmol) of [{alpha}-33P]dCTP (3,000 Ci/mmol; Amersham, Bucks, UK) were added, and the solution (40 µl of final volume) incubated at 37°C for 2 h. The labeled cDNA probe was then purified using a Bio-Spin 6 chromatography column (Bio-Rad, Herts, UK) to remove unincorporated radioactive nucleotide.

Microarray filters.
cDNA microarray nylon filters (catalog no. GF211) were purchased from Research Genetics. Each filter had dimensions of 5 cm (length) x 7 cm (width) and contained ~4,500 spots (details can be found from ftp://ftp. resgen.com/pub/genefilters/gf211_final_data_110498.txt). The spots with a center-to-center spacing of 750 µm were divided into 2 fields each consisting of 8 grids. Each grid consisted of 30 rows and 12 columns. Spotted on every other row in the first 12 and last 6 rows of the right-hand column and the first 6 rows of the second column were control DNA (total genomic DNA) for orienting the membrane during data analysis and for monitoring the homogeneity of hybridization. Each of the remaining spots contained 0.5 ng of insert DNA (~1,000 bases in length) from an IMAGE/LLNL cDNA clone (isolated, sequenced, and verified by Research Genetics) representing the 3' end of a gene. Each filter contained 4,132 spots for human genes with known function or those with sequence similarity to genes with known function in human or other organisms. Spotted on rows 7–18 of the right-hand column of each grid were "housekeeping" genes that were present in duplicates in the corresponding spots in the two fields.

Hybridization.
Prehybridization was carried out in a 5 ml MicroHyb buffer (Research Genetics) supplemented with poly-dA (1 µg/ml) and denatured human Cot-1 DNA (1 µg/ml) at 42°C for 2 h. A radiolabeled cDNA probe reverse transcribed from 1 µl of total RNA (see above) was denatured at 100°C for 3 min and then added into the above solution, followed by hybridization at 42°C for 16 h. The filters were then washed twice in a solution containing 2x SSC and 1% SDS at 50°C for 20 min and once in a solution containing 0.5x SSC and 1% SDS at room temperature for 15 min, followed by autoradiography. The autoradiographs were scanned, and the resultant TIFF images were analyzed using a bioinformatics package (Pathways 2, Research Genetics) developed specifically for analyzing data from the type of cDNA microarray nylon filters used. Using the normalization function of the software, the intensities of the spots were normalized to mean intensities of all spots on each filter. Normalized intensities of corresponding spots were compared between two filters. Differences were expressed as a ratio of normalized spot intensity in a filter hybridized with a cDNA probe generated from nicotine-treated cells over that in a filter hybridized with a probe from untreated cells. A positive value indicated a higher spot intensity in the former filter, whereas a negative value indicated a higher spot intensity in the latter filter.

Competitive RT-PCR.
For competitive RT-PCR of the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene, a RNA competitor corresponding to a 400-bp GAPDH cDNA sequence with a deletion of 100 bp in the middle was generated by PCR in which the forward primer contained sequences flanking the deletion (5'-taatacgactcactatagggtgaaggtcggagtcaacggatttgattccacccatgg-3'), followed by in vitro transcription. A series of dilutions of this RNA competitor was individually mixed with a fixed amount of total RNA extracted from HCAEC cells, then reverse transcribed using a hexamer primer (5'-gatacc-3'). The resultant cDNA mixtures were used as templates in subsequent PCR reactions using a forward primer and a reverse primer whose sequences were within the sequence of the competitor (Table 1). Each PCR produced two amplicons, one deriving from endogenous cDNA and the other deriving from the competitor cDNA; the latter (~400 bp) was shorter than the former (~500 bp), because of the presence of the deletion (100 bp) in the competitor. The PCR products were fractionated by agarose gel electrophoresis.


View this table:
[in this window]
[in a new window]
 
Table 1. PCR primers

 
Relative RT-PCR.
Each relative RT-PCR reaction contained two pairs of primers, one pair for a target gene and the other pair for the GAPDH gene to serve as an internal reference for sample normalization. cDNA was synthesized by reverse transcription from the mRNA fraction of total cellular RNA using an oligo-dT15 primer (Promega) in a 20-µl reaction containing 1 µg of total RNA and 200 U of MMLV reverse transcriptase (Promega) at 42°C for 1 h. The cDNA was diluted with equal amount of distilled H2O, and 1 µl of the diluted cDNA sample was used as template in a subsequent PCR reaction (25 µl of total volume) containing 1.5 mM of MgCl2, 200 µM of each dNTP, 2.5 U of Taq DNA polymerase, and primers for the target and GAPDH genes. The PCR cycle condition was as follows: 95°C for 30 s, 55°C for 30 s and 72°C for 45 s. For each target gene, initially, four PCR reactions with different cycle numbers, i.e., 20, 25, 30, and 35 cycles, were carried out to determine the exponential phase. Having found that 30 cycles were within the exponential phase for all genes studied, three additional experiments each with 30-cycle PCR reaction in duplicate were carried out for each gene; i.e., in each of the three experiments, two PCR reactions were performed for each of the nicotine-treated and untreated conditions.

All PCR primers (Table 1) for the target genes and the GAPDH gene were designed using computer software Primer3 [Rozen S and Skaletsky HJ (1998) Primer3, available at http://www-genome.wi.mit.edu/genome_software/other/primer3.html], with each pair of primers spanning at least one intron. The concentrations of each pair of primers used in the PCR reactions were adjusted according to the mRNA levels of the target gene, to achieve similar yields of PCR products for the target and GAPDH genes.

Quantitation of PCR products.
Products of RT-PCRs were subjected to agarose gel (2%) electrophoresis. The gels were then stained with ethidium bromide, and the intensities of the PCR product bands were measured using a FluorImager 595 (Molecular Dynamics, Sunnyvale, CA) and computer software ImageQuant (Molecular Dynamics). The ratios of intensity (after gel background subtraction) of the target gene band vs. that of the GAPDH gene band were compared between the nicotine-treated and control conditions.


    RESULTS AND DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
The study presented here used the cDNA microarray technique to interrogate simultaneously the expression of over 4,000 genes with known or inferred function, in nicotine-stimulated and unstimulated human coronary artery endothelial cells (Fig. 1). Correlation coefficients of intensities of duplicate spots within the same filter, representing 84 different housekeeping genes, range from r = 0.982 to 0.984 for filters hybridized with probes, respectively, from untreated and nicotine-treated cells. Replicate probe preparations from untreated cells yielded results with a correlation of spot intensities r = 0.908 (n = 4,132 spots for different human genes in each filter).



View larger version (118K):
[in this window]
[in a new window]
 
Fig. 1. cDNA microarray analysis. Image generated using the Pathways software, which compared data from two microarray filters hybridized, respectively, with cDNA from nicotine-treated (10 µM) human coronary artery endothelial cells (HCAEC) cells and with cDNA from untreated HCAEC cells. Upregulated genes are in orange; downregulated genes are in green.

 
Supplemental Table 2 lists those whose expression was found to be increased or decreased over 1.5-fold by nicotine treatment in this study samples (for Supplemental Table 2, please refer to the Supplementary Material1 for this article, published online at the Physiological Genomics web site). To verify these results, RT-PCR was carried out on selected genes including the insulin-induced protein (INSIG1) gene and the vacuolar H+-ATPase (ATP6E) gene. We first performed two pilot experiments to determine 1) the amplification curve and 2) whether the expression of a commonly used housekeeping gene, i.e., the GAPDH gene, was unaffected by nicotine treatment and thus could be used as a reliable internal reference for sample normalization. Having determined the amplification curve (Fig. 2) and confirmed that GAPDH mRNA levels did not differ between nicotine-treated and untreated cells (Fig. 3), we carried out three independent RT-PCR assays for each target gene, using the GAPDH gene as an internal reference and PCR conditions within the exponential phase. The findings of the RT-PCR assays were consistent with those from the cDNA microarray analysis, with both methods demonstrating an increase in mRNA levels of the INSIG1 gene and a decrease in ATP6E mRNA levels (Fig. 4). For INSIG1, an increase by 1.76-fold was detected by microarray analysis and 1.6-fold (95% CI = 1.1–2.2) by RT-PCR (P = 0.037; n = 6, from 3 independent RT-PCR experiments each with duplicate PCRs). For ATP6E, a decrease by 1.74-fold was detected by microarray analysis and 1.4-fold (95% CI = 1.2–1.5) by RT-PCR (P = 0.001; n = 6 ,from 3 independent RT-PCR experiments each with duplicate PCRs).



View larger version (15K):
[in this window]
[in a new window]
 
Fig. 2. Kinetics of RT-PCR. RNA from HCAEC cells was reverse transcribed, and the resultant cDNA were used as templates in subsequent duplex PCR reactions to amplify concurrently the GAPDH gene and the INSIG1 gene (A) or the ATP6E gene (B). Plotted are band intensities vs. PCR cycle numbers.

 


View larger version (17K):
[in this window]
[in a new window]
 
Fig. 3. Competitive RT-PCR of the GAPDH gene to determine whether nicotine alters its expression. Reverse transcription was carried out with 25 ng total RNA from HCAEC cells and various amount of a GAPDH RNA competitor (see MATERIALS AND METHODS), followed by PCR with primers annealing to cDNA from endogenous GAPDH mRNA and from competitor GAPDH RNA. PCR products were then subjected to ethidium bromide agarose gel electrophoresis. Lanes 1–5, untreated HCAEC cells; Lanes 6–10, nicotine (10 µM) treated HCAEC cells. Amounts of GAPDH RNA competitor used were as follows: lanes 1 and 6, 0.25 pg; lanes 2 and 7, 0.125 pg; lanes 3 and 8, 0.0625 pg, lanes 4 and 9, 0.031 pg; lanes 5 and 10, 0.0156 pg. Equal intensities of upper and lower bands in lane 3 (untreated) and lane 8 (nicotine treated), where 0.0625 pg of GAPDH RNA competitor was used in the RT-PCR, indicate that nicotine treatment does not alter GAPDH gene expression levels. MW, molecular weight marker.

 


View larger version (29K):
[in this window]
[in a new window]
 
Fig. 4. Comparison of results from microarray analysis and from RT-PCR. A: representative RT-PCR results. Lanes 1 and 2, untreated; lanes 3 and 4, nicotine (10 µM) treated. Bottom band, INSIG1 (lanes 1 and 3) or ATP6E (lanes 2 and 4); top band, GAPDH (all lanes). B: differences in mRNA level between untreated and nicotine-treated cells by RT-PCR. Data are mean values from 3 independent experiments. Each bar represents the ratio of intensities of PCR bands of a target gene vs. the GAPDH gene. C: comparison of results from microarray analysis and from RT-PCR.

 
The most well-established receptors for nicotine are the nicotinic acetylcholine receptors, a family of neurotransmitter-gated ion channels. Nicotine stimulation is thought to activate nicotinic cholinergic receptors, resulting in an initial cytosolic influx of sodium and creating membrane depolarization that leads to calcium influx through voltage-gated calcium channels. Subsequently, calcium activates protein kinase C and thereby triggers the activation of the mitogen-activated protein (MAP) kinase cascade, leading to the activation of transcription factors such as the cAMP response element binding protein (CREB) (25) (19).

However, noncholinergic nicotinic receptors have been reported. Garnier et al. (5) in 1994 showed that in frog melanotrope cells the action of nicotine was not mediated through plasma membrane ion channels but associated with activation of inositol phospholipid breakdown and mobilization of inositol trisphosphate (IP3)-dependent intracellular Ca2+ stores (5). Similar findings in mouse myogenic cells were reported by Giovannelli et al. (7) in 1991. These findings indicate that the inositol phospholipid pathway plays an important (or even a primary) role in mediating the effects of nicotine in certain types of cells, consistent with the observations of the present study (see below). In addition, it has been shown that nicotine can also activate {alpha}2- and ß2-adrenergic receptors, which are coupled to the adenylyl cyclase and cAMP pathway (17, 20, 22).

The observation in this study of an increase in the mRNA level of type II phosphatidylinositol-4-phosphate kinase (PIP kinase) and a decrease in diacylglycerol kinase mRNA level in nicotine-stimulated endothelial cells suggests that the inositol phospholipid pathway may be activated by nicotine and could play a primary role in mediating the effects of nicotine in endothelial cells. This signaling pathway is known to involve the activation of phospholipase C-ß which in turn hydrolyzes phosphatidylinositol bisphosphate (PIP2) to generate IP3 and diacylglycerol. IP3 increases cytosolic Ca2+ concentration by releasing Ca2+ from the endoplasmic reticulum, whereas diacylglycerol activates protein kinase C which can, in turn, activate the MAP kinase cascade to regulate the transcription of specific genes (1). Because PIP kinase catalyzes the biosynthesis of PIP2, and diacylglycerol kinase breaks down diacylglycerol, the increased expression of PIP kinase and reduced expression of diacylglycerol kinase in nicotine-stimulated endothelial cells may be involved in the activation of this pathway.

Among the transcription factors whose mRNA levels were affected by nicotine stimulation in this study were CREB and nuclear factor-{kappa}B (NF-{kappa}B). Activation of CREB as a downstream event in the MAP kinase cascade activated by nicotine has been well documented (24, 25). The present study demonstrates that nicotine stimulation can also enhance the expression of this important transcription factor. In contrast to CREB, activation and expression of NF-{kappa}B appears to be suppressed by nicotine, as it has been shown that nicotine reduced activity of NF-{kappa}B in macrophages (23), and the present study shows a decrease in its expression in nicotine-treated endothelial cells. Inhibition of NF-{kappa}B activity has been proposed to mediate the suppressive effect of nicotine on interleukin-1 (IL-1) and IL-8 expression and on immunoresponse (23).

In previous work using the RT-PCR method, we found that nicotine (0.1–10 µM) increased the mRNA levels of plasminogen activator inhibitor-1, von Willebrand factor, tissue-type plasminogen activator, vascular cell adhesion molecule-1, endothelial nitric oxide synthase, and angiotensin I converting enzyme, ranging from 1.4- to 2.6-fold. A similar trend was observed in the present study using the cDNA microarray technique, although the magnitude of the changes detected was generally smaller, suggesting the lower sensitivity of the particularly microarray procedure used.

This study analyzed only a portion (4,132 genes) of the genome, and it is likely that a considerable number of nicotine-responsive transcripts were not represented in the cDNA microarray filters used. However, most studies in the literature to date on the effect of nicotine and other chemicals in cigarette smoke have not been systematic (but, instead, investigating one or a few target genes at a time), and the present study represents the first attempt in a systematic description of differential gene expression in different conditions related to smoking. Although these assays could not distinguish primary (direct) and secondary (indirect) effects of nicotine, the data suggest that the inositol phospholipid pathway may be involved in mediating the actions of nicotine on vascular endothelium. This signal transduction pathway has previously been implicated in the pathogenesis of atherosclerosis, particularly in the response of endothelial and smooth muscle cells to endothelin-1, angiotensin II, homocysteine, fluid shear stress, and various cytokines (4, 8, 10, 13, 18, 27). Thus it appears that nicotine shares a common signaling mechanism with these atherogenic factors. The general approach extended to other constituents or derivatives of cigarette smoke, to other cell types and to denser gene arrays, should elucidate the pathways by which smoking and perhaps other insults cause coronary artery disease.


    ACKNOWLEDGMENTS
 
This work was supported by British Heart Foundation Grants PG/98192 and PG/98183.


    FOOTNOTES
 
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

Address for reprint requests and other correspondence: S. Ye, Human Genetics Research Division, Univ. of Southampton, Duthie Bldg. (Mailpoint 808), Southampton General Hospital, Tremona Road, Southampton SO16 6YD, United Kingdom (E-mail: Shu.Ye{at}soton.ac.uk).

1 Supplementary Material (Table 2) to this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/5/4/187/DC1. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 

  1. Alberts B, Bray D, Lewis J, Raff M, Roberts K, and Watson JD. Cell signaling. In: Molecular Biology of the Cell, edited by Alberts B, Bray D, Lewis J, Raff M, Roberts K, and Watson JD. New York: Garland, 1994, p. 721–785.
  2. Carty CS, Soloway PD, Kayastha S, Bauer J, Marsan B, Ricotta JJ, and Dryjski M. Nicotine and cotinine stimulate secretion of basic fibroblast growth factor and affect expression of matrix metalloproteinases in cultured human smooth muscle cells. J Vasc Surg 24: 927–934, 1996.[ISI][Medline]
  3. Cucina A, Corvino V, Sapienza P, Borrelli V, Lucarelli M, Scarpa S, Strom R, Santoro-D’Angelo L, and Cavallaro A. Nicotine regulates basic fibroblastic growth factor and transforming growth factor ß1 production in endothelial cells. Biochem Biophys Res Commun 257: 306–312, 1999.[ISI][Medline]
  4. Dalton ML, Gadson PF, Wrenn RW, and Rosenquist TH. Homocysteine signal cascade: production of phospholipids, activation of protein kinase C, and the induction of c-fos and c-myb in smooth muscle cells. FASEB J 11: 703–711, 1997.[Abstract/Free Full Text]
  5. Garnier M, Lamacz M, Tonon MC, and Vaudry H. Functional characterization of a nonclassical nicotine receptor associated with inositol phospholipid breakdown and mobilization of intracellular calcium pools. Proc Natl Acad Sci USA 91: 11743–11747, 1994.[Abstract/Free Full Text]
  6. Gerber SH, Haunstetter A, Kruger C, Kaufmann A, Nobiling R, and Haass M. Role of [Na+]i and [Ca2+]i in nicotine-induced norepinephrine release from bovine adrenal chromaffin cells. Am J Physiol Cell Physiol 269: C572–C581, 1995.[Abstract]
  7. Giovannelli A, Grassi F, Mattei E, Mileo AM, Eusebi F, and Giovannelli A. Acetylcholine induces voltage-independent increase of cytosolic calcium in mouse myotubes. Proc Natl Acad Sci USA 88: 10069–10073, 1991.[Abstract]
  8. Hajjar DP and Pomerantz KB. Signal transduction in atherosclerosis: integration of cytokines and the eicosanoid network. FASEB J 6: 2933–2941, 1992.[Abstract/Free Full Text]
  9. Henningfield JE, Stapleton JM, Benowitz NL, Grayson RF, and London ED. Higher levels of nicotine in arterial than in venous blood after cigarette smoking. Drug Alcohol Depend 33: 23–29, 1993.[ISI][Medline]
  10. Ishida T, Takahashi M, Corson MA, and Berk BC. Fluid shear stress-mediated signal transduction: how do endothelial cells transduce mechanical force into biological responses? Ann NY Acad Sci 811: 12–23, 1997.[Abstract]
  11. Kuo BS, Dryjski M, and Bjornsson TD. Influence of nicotine and cotinine on the expression of plasminogen activator activity in bovine aortic endothelial cells. Thromb Haemost 61: 70–76, 1989.[ISI][Medline]
  12. Lacor PN, Grayson DR, Auta J, Sugaya I, Costa E, and Guidotti A. Reelin secretion from glutamatergic neurons in culture is independent from neurotransmitter regulation. Proc Natl Acad Sci USA 97: 3556–3561, 2000.[Abstract/Free Full Text]
  13. Marcus AJ and Hajjar DP. Vascular transcellular signaling. J Lipid Res 34: 2017–2031, 1993.[Abstract]
  14. Michael PR. Cigarette smoking, endothelial injury and cardiovascular disease. Int J Exp Pathol 81: 219–230, 2000.[ISI][Medline]
  15. Rose JE, Behm FM, Westman EC, and Coleman RE. Arterial nicotine kinetics during cigarette smoking and intravenous nicotine administration: implications for addiction. Drug Alcohol Depend 56: 99–107, 1999.[ISI][Medline]
  16. Ross R. The pathogenesis of atherosclerosis: a perspective for the 1990s. Nature 362: 801–809, 1993.[ISI][Medline]
  17. Sanchez-Merino JA, Marin J, Balfagon G, and Ferrer M. Involvement of alpha 2-adrenoceptors and protein kinase C on nicotine-induced facilitation of noradrenaline release in bovine cerebral arteries. Gen Pharmacol 26: 827–833, 1995.[Medline]
  18. Schiffrin EL and Touyz RM. Vascular biology of endothelin. J Cardiovasc Pharmacol 32, Suppl 3: S2–S13, 1998.[ISI][Medline]
  19. Schuller HM, Jull BA, Sheppard BJ, and Plummer HK. Interaction of tobacco-specific toxicants with the neuronal alpha(7) nicotinic acetylcholine receptor and its associated mitogenic signal transduction pathway: potential role in lung carcinogenesis and pediatric lung disorders. Eur J Pharmacol 393: 265–277, 2000.[ISI][Medline]
  20. Schuller HM, Tithof PK, Williams M, and Plummer H III. The tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone is a beta-adrenergic agonist and stimulates DNA synthesis in lung adenocarcinoma via beta-adrenergic receptor-mediated release of arachidonic acid. Cancer Res 59: 4510–4515, 1999.[Abstract/Free Full Text]
  21. Su Y, Han W, Giraldo C, De Li Y, and Block ER. Effect of cigarette smoke extract on nitric oxide synthase in pulmonary artery endothelial cells. Am J Respir Cell Mol Biol 19: 819–825, 1998.[Abstract/Free Full Text]
  22. Suemaru K, Gomita Y, Furuno K, and Araki Y. Effect of beta-adrenergic receptor antagonists on nicotine-induced tail-tremor in rats. Pharmacol Biochem Behav 46: 131–133, 1993.[ISI][Medline]
  23. Sugano N, Shimada K, Ito K, and Murai S. Nicotine inhibits the production of inflammatory mediators in U937 cells through modulation of nuclear factor-{kappa}B activation. Biochem Biophys Res Commun 252: 25–28, 1998.[ISI][Medline]
  24. Tang K, Wu H, Mahata SK, Mahata M, Gill BM, Parmer RJ, and O’Connor DT. Stimulus coupling to transcription versus secretion in pheochromocytoma cells. Convergent and divergent signal transduction pathways and the crucial roles for route of cytosolic calcium entry and protein kinase C. J Clin Invest 100: 1180–1192, 1997.[Abstract/Free Full Text]
  25. Tang K, Wu H, Mahata SK, and O’Connor DT. A crucial role for the mitogen-activated protein kinase pathway in nicotinic cholinergic signaling to secretory protein transcription in pheochromocytoma cells. Mol Pharmacol 54: 59–69, 1998.[Abstract/Free Full Text]
  26. Tang K, Wu H, Mahata SK, Taupenot L, Rozansky DJ, Parmer RJ, and O’Connor DT. Stimulus-transcription coupling in pheochromocytoma cells. Promoter region-specific activation of chromogranin a biosynthesis. J Biol Chem 271: 28382–28390, 1996.[Abstract/Free Full Text]
  27. Touyz RM and Schiffrin EL. Signal transduction mechanisms mediating the physiological and pathophysiological actions of angiotensin II in vascular smooth muscle cells. Pharmacol Rev 52: 639–672, 2000.[Abstract/Free Full Text]
  28. Zidovetzki R, Chen P, Fisher M, Hofman FM, and Faraci FM. Nicotine increases plasminogen activator inhibitor-1 production by human brain endothelial cells via protein kinase C-associated pathway. Stroke 30: 651–655, 1999.[Abstract/Free Full Text]