DNA hypermethylation is an important event in lung carcinogenesis. However, it is currently unknown whether changes in DNA methylation are early events, occurring in previously normal lung tissue or whether they are late changes that occur only in overt tumor cells [30
]. To attempt to answer these questions, we tested DNA hypermethylation in lung tissues from subjects without cancer—both smokers and nonsmokers—using a panel of 19 genes which we had previously found to be hypermethylated in some nonsmall cell lung cancers [35
]. This unique study design allowed us, for the first time, to characterize the DNA hypermethylation profile of nonsmokers' lung tissues and to compare this profile to that of smoke-exposed lung.
Importantly, we observed that CCND2, which is known to be frequently hypermethylated in lung cancer tissue [35
], was hypermethylated more frequently in ever-smokers (26%) than in never-smokers (3%). Also, as predicted, in ever-smokers, CCND2 was hypermethylated more frequently in samples from the upper lobes, which are known to suffer far more negative effects from cigarette smoke, such as lung cancer and emphysema [36
]. These findings support the conclusion that CCND2 reflects an early, precancerous change in the lung, caused by cigarette smoke.
CCND2 encodes cyclin D2, a protein involved in cell cycle progression that is thought to act as a regulator of cyclin dependent kinase 4 and cyclin dependent kinase 6 in the transition from G1 to S-phase [39
]. CCND2 hypermethylation appears to be common in many cancers. In breast cancer, where it has been studied most extensively, CCND2 hypermethylation is detected frequently, though it appears to be rarely detected in normal breast tissue [48
]. Interestingly, while CCND2 hypermethylation (and therefore low CCND2 protein expression) has been associated with poor prognosis in epithelial ovarian cell cancer [55
] and recurrence of hepatocellular carcinoma [56
CCND2 expression has been associated with poor prognosis in diffuse large B-cell lymphoma [57
In the lung, CCND2 hypermethylation has been found in 40–56% of NSCLCs [35
]. In noncancerous lung tissue, whereas Virmani et al. found CCND2 hypermethylation in 0 of 18 samples [45
], our previous investigation found CCND2 hypermethylation in 24% of noncancerous lung tissues from patients with NSCLC [35
]. This closely matches the rate observed in the present study, in cancer-free ever-smokers (26%). Possibly, our group observed a higher rate of CCND2 hypermethylation in both cancerous and cancer-free lung tissues because we used MethyLight assays instead of methylation-specific PCR (MSP), which was used by Virmani et al. Thus, we may have detected low levels of hypermethylated genes in cancer-free tissues which were not detected by MSP. Discrepancies may also be due to the somewhat different primers and probes used in analyses, which indicate different sequence regions investigated. In addition, Kubo et al. did not observe any CCND2 hypermethylation in 30 matched noncancerous lung tissues but it should be noted that in this study, 70% of subjects were nonsmokers who would not be expected to have significant rates of CCND2 hypermethylation [46
Combined, these results reveal a progression in the rate of CCND2 hypermethylation in the lung, corresponding with the risk for developing lung cancer. While CCND2 hypermethylation was very infrequent (3%) in our current study's low-risk group of 30 never-smokers, it was more frequent in a high-risk group of ever-smokers (24–26% in our current and previous studies), and most frequent in overt NSCLC tissue (40–56%). This risk-stratified progression in lung tissues suggests that CCND2 hypermethylation may truly reflect an early precancerous change in the lung, en route to overt cancer, which may be due to the effects of smoking.
Still, our findings regarding CCND2 should be regarded as preliminary at this time, for several reasons. In multivariate analysis, the effect of smoking status on CCND2 hypermethylation was reduced to trend-level significance after taking into account the effects of sample location (upper versus lower lobe) and subject age. This likely occurred because in our sample, the majority of smokers underwent lung surgery for emphysema and represented a significantly older group, more likely to contribute samples from upper lobes (where emphysema is most prominent). In contrast, nonsmokers were younger and underwent lung resection for a variety of diseases. With such significant correlation of these factors, multivariate analysis may not have reliably separated each factor's relative contribution to gene hypermethylation. Thus, observed differences in the rate of CCND2 hypermethylation could be attributable to any of these factors or others that differed between ever and never-smokers. Emphysema, for example, made CCND2 hypermethylation more likely although significant rates of CCND2 hypermethylation were also found in smokers with other diagnoses. While CCND2 hypermethylation could be part of the unique pathophysiology of emphysema, it more likely arose because emphysema reflects severe smoking-induced lung damage. The effect of age on CCND2 hypermethylation has not been studied previously in noncancerous lung, although several genes have been reported to undergo increased rates of hypermethylation with age, in various body tissues, including CDH1 and DAPK1 in the lung [58
]. In noncancerous breast epithelium [59
] and in peripheral blood samples from cancer-free subjects [60
], advanced age was not observed to correlate with CCND2 hypermethylation. Thus, the relationship between age and CCND2 hypermethylation remains unknown at this time. In weighing the relative contributions of age, sample location, and emphysema status on CCND2 hypermethylation, it is worth noting that smoking history was by far the strongest single predictor of CCND2 hypermethylation in univariate analysis (OR = 6.9, 95% CI = 1.6–29.8). One limitation of the present study was that despite our overall large number of 151 subjects, only 30 were never-smokers. This occurred because never-smokers far less frequently undergo lung resections which produce tissue. This may have been part of the reason why in multivariate analysis, we observed only trend-level significance for smoking's effect on CCND2 hypermethylation. We were able to improve our statistical power somewhat by using generalized estimating equations (GEE) for our univariate and multivariate analyses, allowing us to enter multiple tissue blocks per subject when available (multiple observations), without biasing the results. However, future studies should seek to verify the low rate of CCND2 hypermethylation we observed in never-smokers. An additional limitation of our study design was that all subjects had an underlying non-cancer pulmonary diagnosis that necessitated lung surgery. Thus, while observed gene hypermethylation was unrelated to cancer, it cannot definitely be said to represent healthy lung. Finally, due to our study design, we only provide indirect evidence of interaction between smoking and CCND2 hypermethylation. Future studies utilizing animal models may be useful to elucidate the potential causal relationship between smoking and CCND2 hypermethylation.
In our current and previous studies, CDKN2A (p16) was hypermethylated in 26% of cancer tissues [44
] but was rarely hypermethylated in noncancerous lung tissues, regardless of smoking status [35
]. However, CDKN2A hypermethylation has previously been characterized as an early event in lung carcinogenesis [28
], and hypermethylation of CDKN2A has been commonly detected in sputum samples from heavy smokers without lung cancer [32
]. Overall, a very wide range of hypermethylation rates for CDKN2A has been reported in the literature, for noncancerous lung tissues. Along with other researchers who observed low rates of CDKN2A hypermethylation in noncancerous lung tissues, our results suggest that CDKN2A hypermethylation may actually represent a later change in carcinogenesis [62
]. However, the surprisingly large discrepancies between studies may be related to differences in assay methodology (including PCR primers and specific CpG islands) or patient populations.