In this paper, we have presented an analysis of published data collected from a large number of studies and analyzed using the GAC database. In performing this analysis, we have combined data from individual studies, some of which were small, to look for overall trends in gene alterations in tumors while minimizing the effect that single study anomalies have on the conclusions.
Current understanding of the molecular pathways of cancer development describes a multitude of genetic and epigenetic changes that accumulate as tumors develop (32
). To understand the roles played by different genes in this process, we have characterized some of the genetic changes that occur in tumors from different topographical sites. Several mechanisms are known to activate or inactivate genes; we have presented data for two that are commonly studied: mutation and gene LOH. Other types of alterations are known to affect genes including homozygous deletion, methylation and chromosome LOH; these undoubtedly also play a vital role in tumor development but have not been evaluated here.
Greenman et al.
) recently analyzed 518 protein kinase genes from 210 diverse cancers for mutations. The authors concluded that the majority of the somatic mutations detected in these genes were ‘passenger’ mutations that did not directly effect tumor development. Similarly, Sjoblom et al.
) sequenced 13 023 different genes in breast and colorectal cancers and showed that on average ~90 mutations per tumor were observed and that 11 genes per tumor were mutated at a significant frequency. However, these studies did not take into account other types of genetic alteration such as LOH that also play a significant role in tumor formation. The data we have presented in this paper demonstrates that in oral tumors and tumors of the digestive tract and lung, genes are more frequently affected by allele loss than by mutation. Gene LOH was reported in each tumor group analyzed for allelic loss and the incidence was >30% in over two-thirds of the studies. These data probably represent an underestimate of the total frequency of allelic loss because analysis was restricted to intragenic markers and markers tightly linked to a gene. Chromosome markers located in the vicinity of a gene or gene locus that could conceivably extend into the gene were not considered in these analyses of gene LOH.
When gene mutations are considered in parallel with allele loss, it is evident that TP53 was one of the few genes inactivated at similar frequencies by allele loss and mutation; this occurred at all cancer sites examined. The only other example of this occurring was in the APC gene and this was restricted to gastric and colorectal cancers. Although the datasets defining the mutation and LOH data are distinct and cannot be used to infer bi-allelic inactivation, it is tempting to conclude that this data show TP53 and in some instances APC fulfill Knudson’s theory of gene inactivation in cancer development. Although several other tumor suppressor genes have been studied, they typically demonstrated very low levels of gene mutations in association with higher levels of LOH. Whether these genes perform a role in cancer development and what that role is remains unclear but possibly gene inactivation occurs by an alternative mechanism: gene methylation or germ line mutation. Alternatively, single allele inactivation may exert a dominant oncogenic effect or it may simply confer a predisposition to cancer. The number of genes and topographical sites where this occurred might suggest this as an important avenue of future research.
Data for alterations of oncogenes were less numerous and largely restricted to mutation studies of the RAS and BRAF genes of the MAPK/extracellular signal-regulated kinase pathway. No conclusions could be drawn concerning the role LOH plays in activation of these genes because of the lack of information but mutations appeared to play site-specific roles in the development of lung (KRAS), oral (HRAS), esophageal (BRAF) and colorectal (BRAF and KRAS) cancers. In the case of lung and oral cancers, it is possible that these were the result of site-specific exposures to environmental carcinogens (tobacco smoke or chewing tobacco). Still with the addition of allelic loss data for these genes, the specificity of gene activation for specific topographies may be negated.
When considered by topography, it is evident that alterations in cell signaling play a role in cancer formation: in colorectal and gastric cancer through genes of the Wnt pathway and the cell cycle genes, particularly TP53, at all sites of the digestive tract, oral and lung cancers. The potential for cell cycle disruption was particularly evident in esophageal cancers where LOH was detectable in a number of different genes. Constitutive transcriptional activation of genes involved in signal transduction in the MAPK/extracellular signal-regulated kinase pathway occurred most frequently in colorectal and lung cancers. All topographies but particularly lung cancers demonstrated alterations in FHIT and DCC, suggesting that inhibition of apoptosis was common to many tumors.
Some of the most well documented gene alterations in these data-sets were KRAS
mutations in lung cancers. The reason for this abundance of data is that it represents an instance where a common environmental exposure, tobacco smoke, evokes a specific genetic response (35
). When data for tumors with known exposure to tobacco smoke are compared with those that were unexposed, the incidence of KRAS
mutations was significantly higher in smokers. Likewise, the incidence of allelic loss of the FHIT
gene in smokers was significantly higher than in non-smokers. Other gene alterations have not been reported in sufficient numbers to draw conclusions about the effect of smoking on their incidence. This analysis does demonstrate that the range of genes activated in smokers and non-smokers was identical, suggesting that one mechanism by which tobacco can give rise to lung cancer is through an increased frequency of genetic alterations. Of course, these analyses are limited by the scope of the published literature; studies tend to focus on the same, well-characterized genes resulting in an absence of information for many genes and genetic pathways. In future, studies of a broader spectrum of genes would help to ascertain whether tumorigenesis in smoking-related lung cancers is the result of alterations in genetic pathways unique to tobacco smoke exposure and distinct from tumors with no exposure to tobacco.
There are >60 carcinogenic components of tobacco smoke which include polycyclic aromatic hydrocarbons, aromatic amines, aldehydes and nitrosamines (15
); direct exposure to these carcinogens occurs not only in lung but also in the upper aerodigestive tract, particularly the mouth and larynx. In the lung, it has been shown that preferential binding of benzo[a
]pyrene and acrolein to guanine residues leads to an increased incidence of GC → TA transversions (25
). From this it follows that GC → TA transversions do not frequently occur at sites not directly exposed to tobacco smoke (37
). Indeed, our analysis of oral tumors from non-smokers demonstrated this; GC → TA substitutions were present as only 2% of the mutations. Surprisingly, this was not the case for lung cancers from non-smokers where 22% of alterations were GC → TA transversions. It is possible that this number was an overestimation of the true rate due to the inclusion of a potentially anomalous study. Gao et al.
) reported a 66% mutation rate in lung tumors of non-smokers compared with < 36% for all other studies. Each tumor in this study demonstrated multiple mutations such that 10 of 15 mutated tumors had 48 mutations comprising 31% of the total mutations for the group. This undoubtedly influenced the results for this group, yet even if this study is excluded, the percentage of mutations that were GC → TA in lung tumors (16% excluding the Gao study) is significantly higher than oral tumors (2%). Why this should occur is not clear but it is possible that lung exposure to polycyclic aromatic hydrocarbons, aromatic amines and nitrosamines derives from alternative sources that are not pertinent to oral tumorigenesis. One of these might be side-stream tobacco smoke for passive/involuntary smokers. Qualitatively, sidestream smoke has the same chemical constituents as mainstream smoke and could result in exposure in non-smokers (39
). Another possible exposure that might present an environment risk to non-smokers is the complex chemical mixtures of diesel fuel and vehicle exhaust fumes (40
). Despite the differences in the lung and oral non-smoking groups, our data show an increase in the percentage of GC → TA transversions in tobacco smokers; this was most apparent in oral tumors but was also evident in lung tumors. The increase in GC → TA transversions was associated with an accompanying decrease in the percentage of GC → AT transitions; this was seen in both lung and oral tumors and was mirrored in laryngeal tumors of smokers. The fact that this decrease occurred particularly at CpG sites supports the view that it was the consequence of a chemical exposure (42
Efforts made to analyze some of the individual chemical components of tobacco smoke in lung tumors of laboratory rodents have supported these findings; GC → TA transversions were prevalent in lung tumors of mice exposed to benzo[a
). Interestingly, 4-(N
-nitrosomethylamino)-1-(3-pyridyl)-1-butanone also found in a variety of tobacco products (chewing tobacco, snuff, cigarettes and cigars) primarily causes GC → AT transitions in exposed mice (16
). Although exposure to 4-(N
-nitrosomethylamino)-1-(3-pyridyl)-1-butanone undoubtedly occurs in smokers, our analysis of the human data in the literature did not demonstrate any increase in the percentage of GC → AT transitions in lung or oral cancers.
Unlike other carcinogens such as aflatoxin B1, the mutagenic specificity of tobacco smoke exposure is defined at the base level by G → T transversions rather than at the codon level (50
). Still comparison of positional spectra and incidences of GC → TA transversions for lung, larynx and oral tumors from smokers define several codons as hot spots of mutations in TP53
, all of which occur in the DNA-binding domain of the protein. For both lung and larynx, these were codons 157, 245, 248 and 273; this agrees with in vitro
studies that show benzo[a
]pyrene derivatives bind with greatest strength to codons 157, 248 and 273 (28
). In contrast, mutational hot spots in oral tumors were restricted to codons 175 and 248 when the types of base change occurring at these codons was considered codon 248 showed 0–25% incidence of G → T substitutions. This would suggest that the benzo[a
]pyrene component of tobacco smoke did not play a prominent role in mutation at this site.
Codons 175 and 245 are commonly mutated in different types of cancer and are not usually associated with specific exposures. Despite this, lung and larynx (codon 245 only) tumors frequently demonstrate G → T transversions at these sites, suggesting that tobacco smoke may play a role in their mutation in some instances. In oral cancer, this did not occur; codon 175 and 245 mutations were not G → T transversions. Of course, this does not preclude tobacco smoke from playing a role in their mutation but it is unlikely that benzo[a]pyrene is involved; instead 4-(N-nitrosomethylamino)-1-(3-pyridyl)-1-butanone may play a role. Moreover, the sites at which the highest incidences of G → T substitutions occurred were not mutational hot spots. It is unclear why the differences in mutational hot spots exist for these topographies; there was a difference in tumor morphology between oral (all squamous cell carcinomas) and lung (non-small cell and small cell carcinomas) cancer but this was unlikely to be one of the factors because laryngeal tumors were all squamous cell carcinomas. Perhaps, they are a reflection of prominent roles played by other factors such as human papilloma virus (HPV) or alcohol in the development of oral tumors. Further studies are needed to resolve these possibilities.
In conclusion, this paper has highlighted how study of peer-reviewed literature using the GAC database (http://dir-apps.niehs.nih.gov/gac/
) can bring about a new understanding of the carcinogenic processes. We show that generally the same genes are part of multiple pathways to cancer in all target sites examined (lung, oral, esophagus, stomach, colon/rectum) with changes commonly occurring in TP53, Ras family and CDK family genes. What distinguished gene changes at a one particular target site versus another was the per cent incidence of mutation or LOH for a particular gene. Moreover, environmental exposure defined a further level of specificity with the spectra of substitutions being characteristic of exposure. Compiling information on gene changes in cancer identifies target gene changes that can be used to develop biomarkers for the cancer disease process and strategies for disease prevention.