Smoking and Lung Cancer: Future Research Directions

Abstract

The aim of future research in this area is to provide the mechanistic understanding and the tools for effective prevention, early diagnosis, and therapy of lung cancer. With the established causal link between cigarette smoking and the risk of developing lung cancer, the most effective prevention is certainly not to smoke. A much better mechanistic understanding of lung cancer and its variability will support the development and evaluation of potentially reduced risk products for those who maintain smoking as well as for the development of early diagnostic tools and targeted therapies. Because of the complexity of lung cancer and the long duration for its development, nonclinical and clinical research efforts need to complement each other. Recent promising advances in this research area are the understanding of the interaction between genotoxic and epigenetic effects of smoking, the development of laboratory animal models for lung tumorigenesis by smoke inhalation, the unraveling of molecular pathways and signatures in clinical lung cancer research useful for developing diagnostic tools and therapeutic approaches, and the first successful therapy for lung cancer—although less suitable for smokers. The above—in combination with emerging data sets from explorative non-clinical and clinical studies as well as improved modeling approaches—are setting the stage for accelerated progress towards developing successful early diagnostic tools and therapies as well as for the assessment of new consumer products with potentially reduced risk.

The major goals with regard to smoking and lung cancer are the prevention of the disease as well as its early diagnosis and successful therapy. Because smoking is the most prevalent cause of lung cancer, the most effective prevention is not to smoke. This imperative of not smoking, however, does not directly eliminate the further need for diagnosis and therapy—for those who had previously smoked or maintain smoking and for those who may acquire lung cancer with other or mixed etiologies.

The causal association between smoking and lung cancer has been established by epidemiology (e.g., International Agency for Research on Cancer 2004). Nevertheless, an agreed-upon mechanistic understanding of the molecular pathways linking individual smoke constituents or classes thereof to the process of pulmonary tumorigenesis and disease outcome is still missing. This mechanistic understanding would be the basis for the development of (early) diagnostic or prognostic tools and targeted therapies, which are urgently needed in view of a current overall survival of just 5 years after diagnosis. In addition, such improved mechanistic understanding would support the development and evaluation of chemopreventive means as well as of novel products, which might potentially be less risky than conventional lit-end tobacco products, as a potential alternative for those who maintain using tobacco products.

The concept of tobacco harm reduction via reduced risk products has been addressed by a committee of the U.S. Institute of Medicine under contract for the U.S. Food and Drug Administration (US Institute of Medicine 2001). The committee concluded that “for many diseases attributable to tobacco use, reducing risk of disease by reducing exposure to tobacco toxicants is feasible”: the dose makes the poison. Although this is trivial in principle, it is particularly difficult from a scientific and societal point of view to evaluate any novel product and substantiate the potentially reduced lung cancer risk for a given consumer or the reduced harm for the population at large, which includes effects on smoking behavior and smoking prevalence (Warner 2002; Hatsukami et al. 2002). In addition, responsible risk communication by the manufacturer and the public health community is needed in conjunction with reasonable governmental oversight.

With the causal association between smoking and lung cancer being established, the gold standard for the evaluation of any novel product or approach for potential risk reduction will be epidemiology. However, due to the long latency of smoking-related lung cancer development as well as the complexity of the exposure and the disease, it seems to be unlikely that in short- or mid-term there would be enough consumers having switched long enough to such novel product to enable a comparative epidemiological study for this purpose. Thus, as a default, the toxicological evaluation of the risk reduction potential of a new product or approach will most probably include similar principles as the safety and efficacy assessment used for new chemicals or drugs, such as preclinical and clinical evaluations. The selection, design, and evaluation of these studies should be based on the mechanistic understanding of smoking-related lung tumorigenesis. The same holds true for molecular epidemiological studies, which may eventually improve the overall timeline.

This review is focusing on examples of current research and potential future directions that might enable the assessment of new products for their potential for lung cancer risk reduction, including nonclinical surrogate assays, clinical research towards developing biomarkers for early diagnosis, and learnings from recent therapeutic successes. Progress in imaging technologies is not covered. Also, no effort is made in differentiating the various histological types of lung cancer.

NONCLINICAL SURROGATE ASSAYS

Cigarette smoke is a complex combustion mixture with more than 4000 chemical constituents having been identified (Dube and Green 1982). Thus, any evaluation of a new product must include a comprehensive chemical analysis of the aerosol (or solution in the case of noncombustible tobacco products) generated. Different selections of smoke constituents have been suggested for this purpose (Vorhees et al. 1997; Borgerding et al. 1998; Stabbert et al. 2003), which are not principally different to each other and try to include representatives of as many as possible chemical classes with potential toxic impact. Regulatory bodies have required data on similar selections of smoke constitutions for the comparative assessment of ingredients or cigarette brands (e.g., Health Canada 2000). As long as there is a general trend for reduction in the yields of all constituents of an aerosol, a toxicological assessment based on chemical composition is relatively straightforward. However, in cases of mixed responses with increased and decreased yields of individual smoke constituents, the toxicological assessment of simple yield data needs to be augmented by considering the toxic potency of the individual smoke constituents (Rustemeier et al. 2002; Stabbert et al. 2003; Fowles and Dybing 2003; Euchenhofer et al. 2003; Rodgman and Green 2003). Because there are neither validated methods to analyze all several thousands of smoke constituents on a routine basis nor sufficient corresponding toxic potency data for smoking-related diseases, an assessment solely based on this approach can never be complete, but it still can serve as valuable contribution to an overall risk assessment of novel products.

In vitro and in vivo assays provide the possibility of a more comprehensive inclusion of aerosol components for toxicological testing, although selections for parts and modifications of the aerosol are being introduced, e.g., by trapping of the aerosol for in vitro exposure or by losses and aging during in vivo exposure. Nonclinical surrogate assays for the evaluation of the risk reduction potential of a novel product need to be sensitive to smoke in a dose-dependent manner, and at best, observed effects should be reversible upon cessation of exposure. Most importantly, these assays should be mechanistically relevant to the development of lung cancer in human smokers.

Although epidemiological studies may not be readily available for the assessment of potential risk reduction, the available epidemiological knowledge of smoking and lung cancer may still be informative for the selection of relevant surrogate assays. It has long been established that the mortality risk for smoking-related lung cancer decreases with half-times of 5 to 10 years upon smoking cessation (Kahn 1966; Peto et al. 2000). The decreasing mortality risk with years of smoking cessation can be interpreted as a strong indication of the role of nongenotoxic or epigenetic effects of cigarette smoke in the process of lung cancer development. Age-dependent multistage modeling of existing epidemiological data has recently corroborated the relatively low contribution of (genotoxic) initiation events to lung cancer risk compared to (epigenetic) promotion events (Hazelton et al. 2005). The most important contributor, however, was the interaction term of initiation and promotion. Thus, besides the commonly used genotoxicity assays with cigarette smoke condensate, such as the Salmonella reverse mutation assay or mammalian cell mutagenicity assays (Lee et al. 1990; Tewes et al. 2003; Schramke et al. 2006), surrogate assays for the assessment of the non-genotoxic or promoting activity of cigarette smoke are required. Few application examples for the nongenotoxic assessment of potentially reduced-risk tobacco products are available. The smoke from a cigarette prototype that primarily heats tobacco showed less activity to inhibit gap-junctional intercellular communication in cultured primary bronchial/tracheal cells compared to that of conventionally burning cigarettes (McKarns et al. 2000). Intercellular communication is considered essential for the growth control of cells.

Subchronic inhalation studies with cigarette smoke in rodents reveal hyperplastic or metaplastic changes in respiratory tract epithelia. These morphological changes have been interpreted as adaptive responses to repeated irritation, which are reversible upon cessation of smoke exposure (Burger et al. 1989). In principle, such morphological changes could be considered as preneoplastic changes, if they prevail long enough for other procarcinogenic effects to occur as well. Studies on novel products, such as cigarettes that mainly heat tobacco, revealed lower hyperplastic or metaplastic potential on a per cigarette basis than conventional cigarettes (Coggins et al. 1989; Terpstra et al. 2003). Assays on cell proliferation and apoptosis in the lungs of cigarette smoke-exposed laboratory rodents would also qualify as examples for nongenotoxic endpoints related to tumor promotion (e.g., Lemjabbar et al. 2003). Pulmonary inflammation is another nongenotoxic response of cigarette smoke inhalation in smokers and laboratory rodents (Friedrichs et al. 2006). Smokers switching to a cigarette-like nicotine delivery device that primarily heats tobacco had significantly reduced signs of inflammation assessed by bronchoscopy (Rennard et al. 2002).

The concept of initiation and promotion has originally been developed in the mouse dermal carcinogenicity model. Although this model is restricted to the assessment of the particulate phase of an aerosol, it is straightforward to use it to assess the promoting potency and maybe even the synergism of initiation and promotion of cigarette smoke condensate (Slaga 1986; Rubin 2002). Using 7,12-dimethylbenz(a)anthracene as an initiator, the condensate of a cigarette prototype that primarily heats tobacco expressed less promoting activity than that of a conventional lit-end reference cigarette (Meckley et al. 2004).

The gold standard for nonclinical surrogate models for cigarette smoke-induced lung tumors should be inhalation exposure, because it involves the most comprehensive part of the cigarette smoke aerosol and the most correct route of administration, although laboratory rodents are obligatory nose breathers. Over the past decades, though, it has been difficult to establish such models (Coggins 2002; Schleef et al. 2006), for reasons not completely understood. Only recently, life-time exposures of rats and mice to cigarette mainstream smoke (MS) were reported to produce significantly elevated rates of lung tumors in rats (Mauderly et al. 2004) and mice (Hutt et al. 2005) at maximum tolerable daily doses, although so far only in the female gender. The strain A mouse has been successfully used in the past ten years with an environmental tobacco smoke surrogate (ETSS) (Witschi 2005a). Previously, several attempts with MS inhalation to enhance lung tumor formation in strain A mice had failed (Finch et al. 1996; D’Agostini et al. 2001) or were inconclusive (Curtin et al. 2004). However, a concentration-dependent increase in lung tumor multiplicity was recently demonstrated after MS inhalation for 5 months followed by a postinhalation period of 4 months (Stinn et al. 2006), similar in design and effect to that seen with an ETSS in the same laboratory (Stinn et al. 2005). In this latest MS inhalation study with strain A mice, the lung tumorigenic activity mainly resided in the particulate phase (Stinn et al. 2006). This is at odds to the observation with the ETSS in the same model, which showed tumorigenic potential in the gas/vapor phase (Witschi et al. 1997a; Witschi 2005b). This apparent discrepancy may provide important clues regarding the constituents in tobacco-derived combustion aerosols active in this lung tumor model, if investigated further. Interestingly, the MS-induced inflammatory response in rats is also associated with the inhalation of the particulate phase, whereas the gas/vapor phase does not provoke such effects (Friedrichs et al. 2006). It remains to be established to what extent these differential inflammatory and tumorigenic responses are also applicable to the human smoker, who is not filtering the smoke by the nasal turbinates.

The classic rodent inhalation models reviewed all suffer from certain disadvantages. There is no initiation-promotion model by inhalation; ETSS did not act as an initiator or promoter in the strain A mouse, when promoted with the classic promoter in this model, i.e., butylated hydroxytoluene, or when initiated with classic initiators for this model, i.e., urethane and 3-methylcholanthrene, respectively (Witschi et al. 1997b). Pulmonary tumors that have been found so far in MS-exposed rodents are limited to adenoma and adenocarcinoma and lack other histological types of lung tumors found in cigarette smokers (Schleef et al. 2006). The study designs employed so far do not reflect the switching or cessation scenarios and the reduced lung tumor-related mortality risk after cessation. Finally, applications of these models to potentially reduced risk products are not available.

The question arises to what extent transgenic mouse models would be more useful than the classic models described so far. If selected properly, they might reflect genetic susceptibilities in developing lung cancer from smoking, advance the mechanistic understanding of smoking-related lung tumorigenicity, and/or provide leads for the development of early lung cancer detection markers (Shaw et al. 2005). Hypothetically, smoke inhalation might promote lung tumorigenesis in mice that already carry an important genotoxic initiating insult, such as a mutation in K-ras (Fig. 1). So far, only few models have been applied to cigarette MS inhalation for the investigation of lung tumorigenicity, with limited success. Among them are models carrying mutations in the K-ras gene, an effect also observed in about a third of human adenocarcinomas of the lung associated with tobacco smoke (Rodenhuis et al. 1988; see also below). In mice, K-ras has been associated with the pulmonary adenoma susceptibility locus 1 (To et al. 2006; Liu et al. 2006). The rasH2 transgenic mouse, which harbors a mutated human K-ras gene, only produced numerical and statistically nonsignificant increases in lung tumor multiplicity at the end of an exposure regimen with MS similar to the one described above for the strain A mouse (Curtin et al. 2004). An explorative inhalation study with MS inhalation on rasH2 and p53+/− mice was also negative (Vanscheeuwijck et al. 2006). Strain A mice with additional heterozygotic mutations in K-ras2, p53, or p16 showed no advantage over wild-type mice in the typical strain A exposure regimen using ETSS (Wang et al. 2005); only the combined mutation in p16 and p53 was considered to be statistically significantly more susceptible than wild-type strain A mice.

There are many more transgenic mouse models, whether targeted for lung cancer research or not, which might be useful in assessing tobacco smoke–associated lung tumorigenicity (Meuwissen and Berns 2005). Lung adenocarcinoma are growing within weeks in conditional transgenic mice carrying mutated K-ras genes under control of doxycycline in alveolar type II cells (Fisher et al. 2001). Within days of switching the transgene off, those tumors regress by apoptosis and related mechanisms. Other transgenic models have also shown the high penetrance of K-ras mutations in the development of lung tumors in mice (Guerra et al. 2003; Kim et al. 2005; Jackson et al. 2005; Politi et al. 2006). No studies have been published to date investigating the response of these models to tobacco smoke inhalation. K-ras-dependent tumorigenic effects may indeed be a valuable model for studying initiation-promotion or genotoxic-epigenetic interactions, because the ratio of the transcribed gene doses of wild-type and mutant genes seems to be an important determinant for murine lung tumor susceptibility, initiation, and promotion (To et al. 2006). The actual relevance of these K-ras–based mouse models for the human lung tumorigenic process, in particular with its various histological subtypes, remains to be established. Recent studies revealed 97% accuracy in classifying human samples as either normal or tumor by comparing gene expression patterns from mouse lung tumors initiated by mutant K-ras with those of human tissue samples (Sweet-Cordero et al. 2005). This finding supports the relevance of this kind of transgene mouse model for research and surrogate assay development for human smoking-related lung tumorigenesis.

CLINICAL RESEARCH TOWARDS BIOMARKER DEVELOPMENT

Beyond the development and application of relevant nonclinical surrogate assays, the evaluation and substantiation of any novel product for potentially reduced risk will only be possible with pivotal clinical trials for exposure and effect assessment. Also, nonclinical assays will provide only limited support to the other above-stated goals of developing early diagnostic or prognostic tools as well as targeted therapies. Although not perfect, exposure studies are already using a comprehensive set of markers for cigarette smoke exposure, representing the aerosol particulate and gas/vapor phases as well as various chemical classes of toxic smoke constituents (Hatsukami et al. 2004; Roethig et al. 2005; Hatsukami et al. 2005). Substantiated reduced exposure is with a certain probability indicative of reduced risk. However, due to the unproportional changes of aerosol constituents in novel versus conventional products and the lack of etiological understanding, there is no single smoke constituent, nor any combination thereof, that could be used to determine that reduced exposure would be linearly related to reduced risk. Consequently, biomarkers of effect need to complement biomarkers of exposure in relevant clinical studies. However, so far “no existing biomarkers have been demonstrated to be predictive of tobacco-related disease, which highlights the importance and urgency of conducting research in this area” (Hatsukami et al. 2006), which is particularly true for lung cancer. The suggested determination of urinary mutagenicity (Smith et al. 1996; Roethig et al. 2005) and sister-chromatid exchange in peripheral lymphocytes as genotoxic end points may qualify as markers of genotoxic effect. More sophisticated approaches for the discovery of early lung cancer diagnostics base on the concept that molecular lesions precede morphological transformation and growth to a tumor, which would only be visible by current imaging technologies (Hu et al. 2002; Brambilla et al. 2003). Abnormalities in processes involved in cell cycle, senescence, apoptosis, repair, and differentiation may be unraveled and found suitable as biomarkers. For the future development of effect markers for lung tumorigenesis, some hope is based on explorative technologies, including proteome, methylome, or other global array-based analyses. As for nonclinical assays, the requirements for valid biomarkers in this regard are a dose-response relationship with smoking and change with cessation or reduced use. In addition, clinical assays should be as little invasive as possible, e.g., by collecting and analyzing body fluids, and should go through some formal or better regulatorily controlled validation process.

Some serum-based protein assays, such as on prostate-specific antigen for prostate tumorigenesis or carcinoembryonic antigen (CEA) for colon have been already introduced into medical practice for screening and monitoring (Ludwig and Weinstein 2005). CEA and other specified serum protein markers have also been investigated for diagnosis and prognosis of lung cancer (e.g., Nieder et al. 2003), although it seems that most protein markers currently discussed occur rather late in the serum of patients with established lung cancer (Lee and Chang 2005; Sheu et al. 2006a; Mizuguchi et al. 2007), increase in concentration with developing histological stage (Kulpa et al. 2002; Fujiwara et al. 2005), and may therefore be rather used for prognosis than for diagnosis. Several studies have identified candidate sets of serum proteins to distinguish between lung cancer patients and controls using various proteomic technologies (Gao et al. 2005; Yang et al. 2005; Fujii et al. 2005). So far, it is not clear whether these approaches are useful for the diagnosis of lung cancer earlier than with current imaging methods or whether their use will eventually be restricted to classification and prognosis.

The detection of cancer-specific mRNA from circulating cancer cells presents another possibility to develop biomarkers for lung cancer, because mRNA detection is very sensitive. The results obtained in such studies indicate a correlation with histological tumor stage (Sheu et al. 2006b), outcome prediction (Guo et al. 2006; Sun and Yang 2006; Chen et al. 2007), and the potential for use in prognosis and rapid assessment of therapeutic response (Sher et al. 2005). In addition, microRNA array analyses are being introduced for potential diagnosis and prognosis of lung cancer (Yanaihara et al. 2006). No indication of an application to early diagnosis is available to date.

Because DNA is more stable than RNA, analyzing circulating DNA shed from tumor tissue or obtained from circulating tumor cells seems to be another promising approach for the detection of lung cancer. Previous research focused on microsatellite DNA (Chen et al. 1996) or gene-specific DNA fragments with protumorigenic mutations, such as p53 (Gonzalez et al. 2000). With current technologies, broader ranges of tumor-specific DNA fragments can be simultaneously analyzed. However, whether the detection of tumor-related genes in the blood can be used to diagnose lung cancer at an early stage has not been fully evaluated to date (Sonobe et al. 2004; Bremnes et al. 2005), although it would be straightforward being able to screen for fragments of mutated genes with understood role in lung tumorigenesis, such as the epidermal growth factor receptor (EGFR) (Kimura et al. 2006) or K-ras (Camps et al. 2005; Neri et al. 2006) (Fig. 1). Over the years, mutations in K-ras and p53 have also been analyzed in bronchoalveolar lavage fluid (BAL) and sputum samples of patients with lung cancer and achieved considerable sensitivity and specificity in detecting lung adenocarcinoma, sometimes preceding the actual diagnosis of the disease (Mao et al. 1994; Mills et al. 1995; Yakubovskaya et al. 1995; Ahrendt et al. 1999; Kersting et al. 2000; Zhang et al. 2003; Destro et al. 2004).

The time frame of occurrence and the stability of methylated DNA promoter fragments in body fluids may render them more suitable for the development of early diagnostic markers. Methylation prevalence of individual promoters might be relatively low in a given tumor tissue and variable between patients. Because of this variability, pattern analysis is recommended for DNA promoter methylation in order to increase accuracy for diagnosis, prognosis, and therapeutic control. For example, methylation analysis of 14 genes could distinguish tissues with adenocarcinoma or mesothelioma from normal lung tissue (Tsou et al. 2005).

Specific fragments of methylated DNA have been found in serum at increased prevalence already at stage I lung cancer, in particular promoter fragments of O ⁶-methylguanine DNA methyltransferase (MGMT), the tumor suppressor gene p16^INK4a, ras association domain family 1A (RASSF1A), death-associated protein kinase (DAPK), and the retinoic acid receptor beta (RARβ) (Fujiwara et al. 2005). Methylation prevalences were found up to ninefold higher in the blood of lung cancer patients compared to non-malignant patients. Methylation in p16 (Fig. 1) seems to be one of the best-investigated candidates for epithelial cancer markers. In normal respiratory epithelium bronchoscopically collected from non-smokers, p16 has been found to be unmethylated (Belinsky 2005). In morphologically ‘normal’ respiratory epithelium from smokers, however, there is already an approximately 20% prevalence of p16 promoter methylation, which increases with morphological development through metaplasia, carcinoma in situ, to squamous cell carcinoma with a prevalence of approximately 60%. Also in 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK)-induced rat lung tumorigenesis, p16 was found to be methylated at high prevalence already in preneoplastic hyperplasia (Belinsky et al. 1998). p16 promoter methylation has also been reported to be dependent on the cumulative smoke dose, i.e., number of years smoked, and to be approximately 40% lower in ex-smokers, who had quit smoking for an average of seven years (Kim et al. 2001). In addition to p16, methylation prevalence of RASSF1A and adenomatous polyposis coli promoter 1A (APC) was also reported to be closely related to smoking status (Zochbauer-Muller et al. 2001; Toyooka et al. 2004). Increased promoter methylation in blood and in brushings of ‘normal’ bronchial epithelium of smokers compared to non-smokers was reported for the E-cadherin gene and DAPK, but not for p16 (Russo et al. 2005). Promoter methylation in the tumor suppressor in lung cancer-1 (TSLC1) gene in tumor tissue was reported to be significantly associated with pack-years smoked and cigarettes per day (Kikuchi et al. 2006). The use of DNA methylation as a biomarker in the assessment of potential reduced-risk products would be further qualified by the demonstration of a reduction in controlled smoking cessation studies.

Apart from blood, DNA promoter methylation has also been investigated in sputum (Destro et al. 2004; Shivapurkar et al. 2007), which might be more specific for respiratory tract effects than blood. In prognostic studies, aberrant p16 and MGMT methylation in sputum samples could be detected from 5 to 35 months before clinical detection of squamous cell cancer (Palmisano et al. 2000). Similarly, promoter methylation of 14 genes was analyzed using sputum samples taken from a cohort of high risk patients who were later diagnosed with lung cancer (Belinsky et al. 2006). Those patients who developed lung cancer within 3 to 18 months after sampling showed adjusted odds ratios (ORs) for correct prediction of up to 2.2 (p16), whereas the ORs in those samples collected within 19 to 72 months prior to lung tumor diagnosis were inconclusive. The short time between successful marker-based diagnosis and occurrence of the tumor may still be a progress for clinical cancer screening, but it seems to be too short to be useful as a marker for clinical trials investigating potentially reduced risk products. With current technologies, however, many gene promoters can be analyzed in parallel, and the inclusion of just four genes into a panel indeed increased the adjusted OR of correctly predicting later lung cancer at 3 to 18 months prior to diagnosis to 6.5 (Belinsky et al. 2006). So far though, no difference in methylation prevalence in sputum samples has been observed by comparing current versus former smokers (Belinsky et al. 2002). In BAL samples of lung tumor patients, a pattern of promoter methylation was found similar to that in tumor tissue, but different from that in BAL of control patients (Topaloglu et al. 2004).

Due to the complexity and variability of the disease and with developing array technologies, the use of panels of molecular markers seems to be the most promising approach to achieve the required sensitivity and specificity for a biomarker assay. Combining different end points into a panel should not be restricted to one source or one technology. Classic histological and cytological diagnoses can be combined with modern molecular markers to improve the predictive power. Bronchial aspirates from a cohort of patients with suspected lung cancer were retrospectively analyzed for promoter methylation (Schmiemann et al. 2005). The sensitivity for correctly predicted lung cancer was 53% with a panel of only three genes (p16, RASSF1A, and APC) analyzed for promoter methylation compared to 44% and 59% for cytology and histology, respectively. Combining the three diagnostic methods, the sensitivity increased to 81%. This accuracy is still not satisfying but underlines the value of combining diverse markers.

Tumor suppressor genes can be silenced by promoter methylation but also by mutation or allelic loss of small or larger parts of a chromosome. Loss of heterozygosity (LOH) has been investigated in several studies in dependence of the development of tissues towards lung tumors. LOH at chromosomes 8p, 9p, 11q, and 13q has been associated with early tobacco smoke exposure-related events leading to cancer precursor cells (Pan et al. 2005). These authors also suggest that LOH at 3p and 17p would trigger towards the development of squamous cell carcinoma, while LOH at 5q and 18q would be indicative of adenocarcinoma development. Chromosomal locations 9p and 13q harbor the tumor suppressor genes p16 and retinoblastoma (Rb), respectively, whereas 5q harbors APC and 3p the fragile histidine triade (FHIT) gene. LOH could also be detected in blood samples. There is, however, no general agreement on the association of individual LOHs to certain stages in lung tumorigenesis. Allelic loss in 3p is nearly universal in lung cancer pathogenesis and involves multiple, discrete, 3p LOH sites that often show a “discontinuous LOH” pattern in individual tumors (Wistuba et al. 2000). In the latter study, LOH at 3p occurred in preneoplastic/preinvasive lesions in smokers with and without lung cancer. Others consider LOH at 3p an early event in lung tumorigenesis as well (Osada and Takahashi 2002; Brambilla et al. 2003), and because it seems to be smoking related (Wistuba et al. 2000; Pan et al. 2005) and more pronounced in current than in former smokers (Mao et al. 1997), LOH analyses thus may qualify as a candidate for early diagnosis (Allan et al. 2001). This notion has been supported by a recent report on the combined analysis of deletions in FHIT and HYAL2 genes in sputum samples, suggesting 76% sensitivity and 92% specificity for the detection of stage I lung cancer (Li et al. 2007). Even in cancer-free patients, such deletions were only found in the sputum of smokers and not in healthy nonsmokers. Again, the combined use of markers based on various technologies is required to improve accuracy of marker sets, as shown for a combination of sputum LOH and DNA promoter methylation analyses (Wang et al. 2006).

THERAPEUTIC RESEARCH AND SUCCESS

The increasing mechanistic understanding of molecular pathways involved in cancer of various types should be the basis for the development of diagnostic tools and in particular of targeted therapies, which are superior compared to previous broadly acting chemotherapies. In a variety of lung tumor tissue samples, specific mutations in the tyrosine kinase (TK) domain of the EGFR have been found, which generate a condition of permanent activity of the receptor-mediated downstream growth signaling (Sordella et al. 2004) (Fig. 1). These mutations were predominantly found in adenocarcinoma of female never-smokers of East Asian ethnicity (Gazdar et al. 2004; Shigematsu et al. 2005). The therapeutic use of small molecules inhibiting this TK activity in the mutated EGFR resulted in dramatic remission of lung adenocarcinoma, improving the survival of the respective patients. However, these remissions are typically short-term, and drug resistances develop that might be associated with secondary mutations in the EGFR (Haber and Settleman 2005). Other small molecules were developed that may overcome this resistance (Kwak et al. 2005). Based on the importance but also the possibility for successful therapy, the EGFR was termed the “Achilles heal” of lung cancer (Gazdar et al. 2004). Genomic hybridization analysis confirms the principle difference between lung tumors bearing EGFR mutations or the respective wild-type receptor gene (Shibata et al. 2005). Furthermore, in one cluster of tissue samples with EGFR mutations, the EGFR gene was found to be amplified, which might be considered as a positive feedback loop related to this mutation. EGFR gene amplification was not observed in adenocarcinoma tissue with wild-type EGFR. EGFR mutations can be detected in histologically normal respiratory epithelium in lung cancer patients within the tumor and adjacent sites, thus suggesting an at least localized field phenomenon and that EGFR mutations might constitute a relatively early event in this type lung tumorigenesis (Tang et al. 2005).

Although EGFR-targeting therapies can be considered a breakthrough, this therapy has only limited success in smokers. Tumor tissue from smokers has a different molecular footprint. The OR of finding EGFR mutations in lung tumor samples was found to decrease with smoking status from never smoking (1.00) via light (0.37) to heavy smoking (0.11), whereas the OR for mutations found in the proto-oncogene K-ras increased with smoking status (1.00, 2.08, and 5.61, respectively) (Toyooka et al. 2006). Similarly, decreasing EGFR mutation rates with increasing tobacco exposure were reported by others (Soung et al. 2005; Tam et al. 2006; Pham et al. 2006). Again, K-ras mutations were significantly associated with smoking status in these and other studies (Husgafvel-Pursiainen et al. 1993; Noda et al. 2001). Mutations in K-ras are exclusive to mutations in the EGFR with no exception so far (Shigematsu et al. 2005; Shibata et al. 2005; Tam et al. 2006). Two distinct pathways were thus postulated for adenocarcinoma formation in smokers and nonsmokers (Gazdar et al. 2004), with no therapy in sight for smoking-associated tumors, which are the majority of cases.

SIGNAL TRANSDUCTION PATHWAYS

The signaling cascade downstream of the EGFR was shown to be important for lung tumorigenesis in nonsmokers (Fig. 1). Among the downstream effectors of the EGFR pathway are K-ras, signal transducers and activators of transcription (STAT), and protein kinase B/Akt (Fig. 1). Activation of Akt can induce a survival signal via inhibition of the tumor suppressor gene p53. In parallel, K-ras can initiate a cascade of mitogen-activated protein kinases (MAPK) eventually triggering the inhibition of the tumor suppressor gene retinoblastoma (Rb), thus inducing a proliferative signal. These downstream pathways can also be triggered by effects independent of the mutational activation of EGRF but related to smoke exposure, such as K-ras or p53 mutations (see above), p16 promoter methylation (see above), or activation of Akt. Recently, several studies have reported that nicotine may elicit anti-apoptotic characteristics in normal airway epithelial cells (West et al. 2003) and lung cancer cells (Maneckjee and Minna 1994; Jin et al. 2004; Tsurutani et al. 2005; Zhang et al. 2006). This effect seems to be mediated by nicotinic acetylcholine receptors (nAChR) and mainly linked to the above pathways by activation of Akt. The relevance of this nicotine activity is not known to date, but it might be important in tumorigenesis as well as in attenuating the responsiveness to chemotherapeutic treatment (Dasgupta et al. 2006). Akt can also be activated via suppression of its negative effector phosphatase and tensin homologue deleted from chromosome 10 (PTEN) mediated by nuclear factor kappa B (NFκB) activation as a consequence of inflammatory stimuli (Vasudevan et al. 2004; Tang et al. 2006), which are a hallmark of early smoking-related pulmonary effects.

Tobacco smoke can also impact signaling via binding of constituents to the aryl hydrocarbon receptor (AhR) (Kitamura and Kasai 2006). Among other effects, this activation leads to the induction of phase I and II xenobiotics–metabolizing enzymes. In a second capacity and probably in the absence of exogenous ligands, the AhR is also interacting with several pathways involved in cell cycle regulation and growth, such as EGFR and Akt, in a manner not satisfactorily understood to date (Marlowe and Puga 2005; Wu et al. 2007). Activation of such signal transduction pathways by other smoke constituents or smoke fractions has been shown repeatedly in in vitro and in vivo studies (Muller et al. 1997; Muller and Gebel 1998; Lemjabbar et al. 2003; Gensch et al. 2004; Zhong et al. 2005; Gebel et al. 2006) and may provide additional non-clinical surrogate assays complementing the available genotoxicity assays.

There are several mechanisms leading to the activation of EGFR and other members of the ErbB receptor family (Normanno et al. 2001; Osada and Takahashi 2002; Hynes and Lane 2005): apart from mutation, gene amplification, increased gene expression, increased ligand expression, and increased ligand activation provide several levels of control including positive feedback responses by autocrine, juxtacrine, and paracrine mechanisms (Fig. 1). ErbB receptors are activated in response to ligand peptide binding, involving ErbB receptor homo- and heterodimerization. Several members of the ErbB receptor family including EGFR were found overexpressed in lung cancer. Amplification of the EGFR gene was found in a subclass of EGFR-mutated tumors (Shibata et al. 2005) and may thus increase the impact of such mutations. There are three ErbB ligand groups, including the epidermal growth factor, transforming growth factor α(TGF-α), amphiregulin, or heregulin (Harris et al. 2003). Their proligands are proteolytically cleaved and thereby released for signaling (Normanno et al. 2001) by members of a disintegrin and metalloprotease (ADAM) family (Blobel 2005). In particular, ADAM17 has been suggested to be a major ErbB proligand protease, which has been reported to be stimulated by cigarette smoke in vitro (Lemjabbar et al. 2003). ADAM8 was found increased in lung tumor tissue and serum of patients (Ishikawa et al. 2004). TGF-α and amphiregulin were found overexpressed predominantly in lung tumor tissue of male smokers and their respective serum levels and were suggested to be useful as prognostic markers (Ishikawa et al. 2005). Finally, several of the above pathways are dependent on the cellular redox status, in particular on reduced sulfhydryl functions. Cigarette smoke can influence signal transduction pathways via its oxidative activity in aqueous solutions (Muller and Gebel 2006).

Yet another level of control for this pathway, which has only recently been discovered, is provided by the expression of non-coding microRNAs (mir) that can interact with the processing of mRNA. Mir-let7a can inhibit the proper splicing of the K-ras mRNA and thus prevent the translation to the K-ras protein (Yanaihara et al. 2006). The relevance of this interference was demonstrated by the significantly improved survival of patients who have high expression of the mir-let7a precursor hsa-let7a. To date, it is not known whether there is also a regulation of mir expression by smoke exposure.

Overall, this research into signal transduction pathways involved in lung tumorigenesis and potentially triggered by cigarette smoke exposure underscores the importance of considering both genotoxic and epigenetic effects of smoking, such as gene activation or silencing by mutation, LOH, amplification, promoter methylation, regulatory effects induced by oxidative and other stressors, through phosphorylation or dephosphorylation. It seems that a certain critical level of changes or a certain penetrance needs to be achieved in a cell in order to proceed through pro-tumorigenic survival and proliferative pathways. This may involve just one change, which has a high penetrance for tumorigenesis, such as K-ras mutations in murine models (Fisher et al. 2001). In humans, other effects may need to accompany K-ras mutations, such as p16 promoter methylation (Toyooka et al. 2006). EGFR mutations in the TK domain seem to exert high penetrance, otherwise treatment with a TK inhibitor would not be as successful as observed (Gazdar et al. 2004), although this mutation is sometimes accompanied by EGFR gene amplification (Shibata et al. 2005). EGFR agonists such as TGF-α and amphiregulin were found increased particularly in such tumors that normally do not carry EGFR mutations, such as in male smokers’ non-adenocarcinoma (Ishikawa et al. 2005). Other changes in the EGFR downstream pathways may cooperate to achieve the required penetrance, such as activation of Akt by nicotine (West et al. 2003) or inflammatory stimuli (Vasudevan et al. 2004; Tang et al. 2006), or by oxidative stress-mediated activation of pathways, such as EGFR ligand shedding (Lemjabbar et al. 2003). This requires escalating from individual molecular effects to a systems biology approach and moving from the current cancer stage–specific to systematic modeling of the underlying molecular pathways. It will also inform the search for the ideal set of early diagnostic markers and the determination of their respective biological relevance. At optimum, the available set of markers will combine diverse technologies and mirror the most relevant tumorigenic pathways.

SUMMARY

The major goals in view of smoking and lung cancer research are to allow better prevention by replacement or chemopreventive products, more accurate and earlier diagnosis and prognosis, and improved targeted therapies. All of these goals can only be achieved on a solid basis of improved mechanistic understanding of the disease.

There are many recent promising advances towards these goals. The interaction of genotoxic and epigenetic effects has been highlighted, with major emphasis on signaling pathways leading to proliferation and survival signals. In the nonclinical research area, laboratory animal surrogates for lung tumorigenesis by tobacco smoke inhalation have been presented, although the final validation and general agreement on the relevance of such assays has not been obtained so far. The existence of a first successful therapy is a hallmark breakthrough for lung cancer research, although it is limited to a small percentage of lung tumors. It generates the optimism that similar breakthroughs can eventually be achieved for lung tumors induced by smoking.

In the future, more in vitro assays need to be developed trying to map the current understanding of epigenetic effects by smoking, such as intercellular communication, cell proliferation and apoptosis, or assays based on individual steps of the above-described growth signal transduction pathways. Molecular signatures of clinical tumor and biofluid samples need to be further investigated using explorative array and bioinformatic technologies in order to allow the development of early diagnostic and prognostic tools based on panels of heterogeneous molecular markers. The progress that needs to be achieved can be used for the evaluation of novel products with potentially reduced risk, but more importantly for the general prevention, diagnosis, and therapy of lung cancer.