Factors That Influence the DNA Repair Capacity of Normal and Skin Cancer-affected Individuals

Human populations display a large variability in cancer susceptibility. The risk of developing a given cancer will be determined by the interplay of host constitutional susceptibility and exposure to environmental hazards. Genes involved in DNA repair play a critical role in protecting the cell genome against mutations that might lead to cancer. This was originally demonstrated in individuals affected by the rare cancer-prone disease xeroderma pigmentosum(1) and more recently in the relationship between defects in mismatch repair and familial colon cancer (2). However, the evidence that a reduced DRC² is a risk factor for cancer in the general population is still limited because of the paucity of molecular epidemiology studies that have addressed this question. The measurement of interindividual variation in DRC in a large number of subjects has become feasible since the development by Athas et al. (3) of an in vitro DNA repair assay based on a host cell reactivation assay in human peripheral lymphocytes. In this assay, a UV-treated expression vector is transfected into human T lymphocytes, and the host repair capacity is evaluated by measuring the plasmid encoded reporter gene product CAT. By using this assay, it has been shown that individuals with the mean repair capacity of 65–95% of the general population are usually more frequently in the cancer cohorts—in particular lung (4) and skin (5) cancer cohorts—than in the control cohorts. This variation in DRC has characteristics expected of cancer susceptibility genes, for instance, the association of reduced repair with cancer onset at a young age and with a family history of cancer (5).

However, several lines of evidence indicate that DRC can also be transiently induced or stimulated as a response to DNA damage. Mammalian cells can generate an SOS-like response to DNA damage (6), similar to that described previously in prokaryotic cells (7, 8). In fact, by using a UV-damaged CAT reporter vector, Protic et al. (9) showed that pretreatment of normal human cells with a chemical carcinogen or with UV light enhances the DRC of these cells. Interestingly, higher DNA repair levels were also detected among users of photosensitizing drugs and estrogens as compared with the control population in a molecular epidemiology study for BCC risk factors (10). All together, these findings raise the question of whether the measurement of DRC should be considered purely a marker of cancer susceptibility. On the basis of the current knowledge, the expectation is that, although a repair gene defect would lead to a decrease in DRC, the presence of DNA-damaging agents in the environment might affect the host DNA repair by producing the opposite effect of transiently increasing the host repair ability.

In this study, we examined the repair capacity of UV-damaged plasmid DNA of T lymphocytes from 49 BCC patients and 68 cancer-free controls. The analysis was focused on the identification of the factors that may affect host repair capacity.

The initial study population consisted of 76 patients affected by BCC and 88 controls belonging to a larger hospital-based case-control population enrolled for studying BCC risk factors.³ All of the participants were seen at the dermatology clinic, Istituto Dermopatico dell’Immacolata (Rome, Italy), where the subjects were examined and interviewed by means of a structured questionnaire by a dermatologist and had blood drawn for DNA repair assay on lymphocytes. The questionnaire included information on sun exposure behavior, history of skin reaction to sun-light, medical anamnesis, and recent treatment and life-style factors like smoking habits and alcohol consumption. The control group was cancer-free and had diagnoses of mild skin disorders such as nevomelanocytic nevus (33%), subacute eczematous dermatitis (13%), androgenetic alopecia (10%), common warts (9%), and seborrheic keratosis (8%).

Blood collection took place from March 1995 to June 1997. Each subject had 30 ml of blood collected. The isolation of lymphocytes was performed the same day as the blood collection in the laboratory of the dermatological clinic. Lymphocytes were stored in liquid nitrogen before being shipped in liquid nitrogen, to the laboratory of the Istituto Superiore di Sanità, where they were stored in liquid nitrogen up to their processing for DRC.

The DRC of lymphocytes was monitored essentially as described in Athas et al. (3), and the assay was performed from March 1996 to November 1997. Briefly, this assay is based on the idea that the level of repair of the lymphocytes is a reflection of the repair capacity of the donor. A nonreplicating plasmid, pCMVCAT, which contains the bacterial CAT reporter gene is transfected into PHA-stimulated lymphocytes. The plasmid is previously damaged with 350 and 700 J/m² UV radiation (254 nm). To minimize experimental variation, the same batches of plasmid DNA (with or without irradiation) and of DEAE dextran for the transfection procedure were used for all of the samples in this study. The quick-thawed lymphocytes were resuspended at a final concentration of 1 × 10⁶ viable cells/ml (the viability was assessed using trypan blue exclusion) and incubated for 72 h with PHA at 37°C in a 5% CO₂ incubator. Cells were then transfected with undamaged and damaged DNA by the DEAE procedure and returned to the incubator for another 40-h incubation. Lymphocytes were then centrifuged, and protein extracts were made. The standard assay for CAT activity was quantified by measuring the formation of [³H]acetylchloramphenicol (expressed as radioactive counts, dpm) in protein extracts. The counts of the blank obtained on the day of the assay were subtracted from all of the values. The DRC was calculated as a percentage of CAT activity after the repair of damaged DNA compared with undamaged DNA (equal to 100%). The measures of CAT activity at UV doses of 350 and 700 J/m² were done, when possible (depending on the number of viable cells after PHA-stimulation) in triplicate. The assay was performed by two scientists (M. D. and A. C.) who processed the same samples but participated in two different phases of the assay: one scientist was involved in the transfection and cell culture procedure, and the other one in the determination of the CAT activity in protein extracts.

The statistical analyses were performed with the SPSS/Windows (11) and Stata (12) software.

Distributions of CAT activity values at 0, 350, and 700 J/m² as well as DRC levels were positively skewed. Group mean comparisons, ANOVA, and linear regression analyses were, therefore, carried out on square-root transformed values because they were approximately normally distributed. To allow direct comparisons with other studies, the nontransformed percent CAT activity values were also analyzed. In this case, group mean comparisons were conducted by nonparametric tests for paired (Wilcoxon signed-rank) and unpaired (Wil-coxon rank-sum) samples, whereas the relationship between continuous variables was estimated by the Spearman’s rank correlation coefficients.

Because there were only three UV dose levels (0, 350, and 700 J/m²) it was not possible to determine the mathematical function that would be the best descriptor of the UV dose/CAT activity relationship. A complete analysis of the repair data was, therefore, carried out at both the 350 and the 700 J/m² UV doses.

One-way random effects ANOVA model (Loneway procedure from Stata; Ref. 12) was performed to estimate the within-subject and the between-subject SD of the repeated CAT measurements taken at the three UV doses. The reliability of the assay was estimated, assuming the same model, by the intraclass (intrasubject) correlation coefficient.

Simple and multiple linear regression models were adopted to identify the best predictors of DRC.

Table 1 shows the variables reflecting pigmentary and constitutional traits as well as demographic and life-style characteristics of the study population, which consisted of 76 patients with histopathologically confirmed primary BCC and 88 cancer-free controls. Distribution of age, sex, educational level, skin type, and smoking habits and alcohol consumption were similar for BCC cases and control groups. BCC patients were more likely than controls to have a family history of BCC and actinic keratosis and a prolonged occupational sun exposure. Similar results were obtained in the larger case-control population enrolled for the epidemiological study.³

Blood samples were obtained from all of the subjects, and the DRC of isolated lymphocytes was determined by the plasmid/host-reactivation assay. Inspection of the assay results showed that the distribution of the blank radioactive counts recorded for each experiment was positively skewed, with a bimodal profile and median value of 313 dpm. Unusually high blank counts (>868 dpm) obtained on 5 different days were clustered in the right tail of this distribution. The 28 subjects who were analyzed during these days presented an estimated within-subject SD for repeated measurements at the three UV doses that was about 2-fold higher compared with that of subjects analyzed on days with low blank counts (≤868 dpm). These measurements were, therefore, excluded from the analysis. After these exclusions, there were 136 subjects (63 cases and 73 controls) of the original 164. Of these 164, 117 (49 cases and 68 controls) presented at least one measurement at all three of the UV doses. All of the statistical analyses were performed on this subgroup of subjects. The distribution of these individuals by the variables listed in Table 1 was similar to that of the original population.

At each UV dose tested, the within-subject SD of CAT activity was systematically lower than the corresponding between-subject SD (Table 2). Consequently, the intraclass correlation coefficient was high (range, 0.77–0.94) A dose-dependent decrease in CAT activity was observed among both cases and controls (test for nonparametric trend, P < 0.01). As expected, the level of transcriptional activity of the reporter CAT gene, by measuring host cell repair ability, is dependent on the amount of UV damage present on the target DNA (3).

The analysis of the association of a series of factors including the case-control status with CAT activity of undamaged plasmid is shown in Table 3. The DRC at 700 J/m2 was negatively associated with age among controls and with CAT activity at 0 J/m² among cases. Alcohol consumption was positively associated with age in the control population. Among cases, a negative association of this variable with CAT at 0 J/m² and a positive association with DRC at 700 J/m² was observed. The blastogenic rate correlated positively only with smoking habits among both cases and controls. These associations were further analyzed by means of uni- and multivariate regression analysis (see below).

Table 4 shows the mean values of DRC at both UV doses by age in cases and controls (cut-off = 40 years, according to the 33rd percentile of the study population distribution). For comparison, CAT activity of undamaged plasmid and blastogenic rates are also displayed.

The level of DRC decreased as a function of UV dose among both cases and controls independently of age group. Younger cases repaired less than their controls, although the difference was not statistically significant. On the contrary, older BCC patients showed a significant increase in DRC as compared with their controls at both 350 and 700 J/m² UV doses.

Cases who were older than 40 years presented a significantly lower CAT activity at 0 J/m² and a significantly lower blastogenic rate as compared with the controls of the same age group. These differences were not detected among younger individuals. Because blastogenic rate and CAT activity at 0 J/m² were not associated (Table 3) among older cases, we should conclude that lymphocytes from older BCC patients are characterized by both low percentage of blasting cells and low CAT activity of undamaged plasmid, but these two traits are independent of each other. Moreover, a negative and significant association between CAT at 0 J/m² and DRC of older BCC cases was observed (Spearman’s rank correlation coefficient = −0.44; P = 0.00; the negative association reported in Table 3 for the cases was indeed driven by the older BCC subjects).

Simple linear regression analysis revealed that the DRC of controls at 700 J/m² declined significantly with age (0.75% per year) from 20 to 70 years (Fig. 1,A). A similar decrease, although not significant (regression coefficient = −0.02; P = 0.104), was also observed at the UV dose of 350 J/m². An age-related decline in DRC was reported previously (5) by using the same assay in the control subjects of a study population from Baltimore. The impairment of DRC due to aging was observed only among reference subjects, whereas an age-related increase of DRC, although not significant, was detected in BCC patients at both UV doses. The data relative to the highest UV dose are displayed in Fig. 1 B.

The analysis of DRC data according to gender (Table 5) revealed that females affected by BCC repaired less than males at both UV doses, although the difference was significant only at the dose of 700 J/m². On the contrary, no sex-related difference in DRC was detected among the control subjects. Further stratification by age showed that the difference in DRC at 700 J/m² between genders among cases was driven by the DRC levels of males older than 40 years, who repaired more efficiently than females of the same age group (17.7 versus 9.0; P = 0.018). These data strongly suggest that the increased DRC of cases as compared with controls may be due to the high repair ability of males belonging to the older age group.

What are the factors possibly associated with high repair of older cases? The analysis of the determinants of DRC among cases by simple linear regression models (Table 6) revealed that the level of CAT activity at 0 J/m² and the gender were able to significantly predict 14.3 and 12.9%, respectively, of the overall variability. The high level of DRC detected in BCC patients was also significantly explained by life-style factors like smoking habits and alcohol consumption. Heavy smoking (>110,000 cigarettes in a lifetime, according to the 75^th percentile of the distribution of cases) and high alcohol consumption (>11,800 glasses of wine in a lifetime, according to the 75^th percentile of the distribution of cases) were significantly and positively associated with DRC and accounted for 12.6 and 16.9%, respectively, of the total variance of DRC measurements. When the analysis was restricted to subjects who were current smokers, an even stronger association was observed. Although this analysis concerned only 12 cases, smoking was able to predict as much as 43.3% of their DRC variability. Because there were only current drinkers in the study population, the same analysis could not be replicated for alcohol consumption. Interestingly, none of these factors except age were able to explain the repair capacity of control subjects.

By including in a multiple linear regression model (Table 6) the variables that significantly explained the DRC levels among cases (i.e., smoking, alcohol consumption, gender, and CAT activity at 0 J/m²) along with age, the determinants of DRC emerged. The variance explained was 38.0%.

Smoking was confirmed to be the best predictor of DRC levels. A residual effect on DRC levels of alcohol consumption and CAT activity at 0 J/m² was maintained in the multiple model. Conversely, the association of gender with DRC disappeared in the multiple model because it was driven by the combined effect on DRC of smoking and alcohol use.

Finally, as shown in Table 7, the increase in DRC of older cases as compared with their controls was largely reduced and not more significant after controlling for the variables included in the multiple regression model.

This study was designed to investigate whether a decreased DRC of UV photodamage is a susceptibility marker for nonmelanocytic skin cancer. Conflicting results were obtained by two previous studies (5, 13) that addressed the same question in a population from Baltimore and Western Australia, respectively.

The distribution of age, sex, and skin type were similar for BCC and control groups in our study population, whereas subjects with actinic keratosis, family history of skin cancer, and occupational sun exposure were represented more among cases than controls (these factors were also confirmed as major BCC risk factors in the larger case-control study).³ Pigmentary traits as well as sun sensitivity of the skin were poor risk indicators, which indicated that, in this ethnic group, constitutional features play a minor role in skin cancer susceptibility. Skin tumors were localized most frequently (73%) on the head and neck with a significant proportion (22%) localized also on the trunk. The majority of the cases were diagnosed with primary BCC (only 8% reported previous skin lesions).

The analysis of DRC revealed that the repair ability of patients affected by BCC was significantly higher than that of their controls. However, further analysis by age showed a striking difference between the two groups. The DRC of control subjects declined with age from 20 to 70 years, whereas no age-related decrease was recorded among cases. The effect of donor age on the processing of UV-damaged DNA has been described previously as a reduction in DRC and an increase in DNA mutability as a function of age (5, 14). The age-related down-regulation of the repair machinery would be consistent with an altered processing of DNA damage in elderly people associated with hypermutability by sunlight and, thus, with increased skin cancer risk (age is a risk factor for BCC). The analysis of the frequency of p53 mutations in skin cancer samples from a subset of patients belonging to this study showed, in fact, hypermutability of this tumor suppressor gene in elderly patients as compared with young cases in which p53 mutations were extremely rare (15). However, this is not reflected in the levels of DRC because older (>40-year-old) BCC patients presented a significantly higher repair ability as compared with younger (≤40-year-old) cases. These findings prompted us to investigate whether other factors could explain the high DRC of older BCC patients. First of all, differences in the physiology of the host lymphocytes were analyzed as a function of age. The T lymphocytes need to be cycling to express the CAT reporter gene whose activity is taken as an indirect measurement of the host repair ability. Lymphocyte stimulation by mitogens like PHA provides the most reproducible in vitro correlate of cell-mediated immunity and is therefore expected to reflect the host immune response. The clinical condition as well as the potential exposure to environmental hazards is expected to affect this response. In our study population, the mean blastogenic rate of cases was significantly lower than that of controls. This effect is of the same order as that reported by Wei et al. (5) between cases and controls. The decreased blastogenic rate of cases was due to the low percentage of blasting cells in older BCC patients lymphocyte cultures, which suggested that these patients might have impaired immune functions. Interestingly, older BCC patients also presented a significantly lower reporter gene expression (CAT at 0 J/m²) than their controls. These findings might indicate an alteration of the physiology of the target cells dependent on both age and case status. Moreover, the association found between the low level of expression of the undamaged CAT target gene and high alcohol consumption leaves open the possibility that the level of CAT at 0 J/m² may also reflect the functionality of the cell transcription apparatus, which may be transiently affected by exogenous agents.

The influence of life-style factors on DRC is highlighted by this study. Smoking is strongly associated with the increased DRC levels of older cases. The presence of factors that stimulate the host repair capacity is not surprising because the induction of SOS-type responses including inducible repair has been described in mammalian cells after genotoxic stress. In vitro experiments that used host-cell reactivation of damaged reporter genes have shown that cells treated with a carcinogen (9) or with small DNA fragment thymidine dinucleotide (dThd; Ref. 6) would repair and, thereby, reactivate the damaged reporter gene to a greater extent than cells that were not pretreated. The strong association between heavy current smoking and high DRC among older cases is likely to reflect the stimulation by tobacco smoke carcinogens of the repair ability of the host lymphocytes. In agreement with this hypothesis, a low prevalence of heavy smokers was observed among older controls (5 of 41 versus 12 of 35 among cases), who are characterized by significantly lower DRC levels as compared with cases of the same age group. It would be also interesting to verify whether high-repair BCC patients present cancer susceptibility genes (for instance, metabolic polymorphisms) that might enhance their response to environmental carcinogens as compared with controls. Polymorphism in CYP and GST genes have been shown to influence susceptibility to BCC (16, 17). The high DRC of older cases may, thus, indicate the selection of a population at high risk not only for BCC but also for exposure to environmental carcinogens. In this context, it is interesting to recall that population-based data indicate that patients with BCC are at increased risk for not only new skin cancer but also for various types of noncutaneous cancer (18).

The novel findings of this study that high DRC is associated and likely due to the exposure to environmental carcinogens, although limited by the relatively small number of subjects analyzed in the multiple regression model (n = 41), raise the question of whether DRC should be considered purely a marker of genetic susceptibility to cancer. Previous studies (5, 4) provide evidence that there is a risk of cancer for the general population associated with an inherited trait such as defective repair ability. However, because endogenous and exogenous factors can modify the cell’s DRC, caution should be taken when interpreting DRC data.

Because the risk of a given cancer represents the end-product of genetics and environmental exposure, the characteristics of the study population may play a key role in the results of molecular epidemiological studies that use this biomarker. In particular, BCC risk is strongly associated with a family history of skin cancer, and this trait is likely to be due to inherited low DRC (5). Subjects with a family history of skin cancer were over-represented in the population of the Baltimore study (5; 37% of BCC cases and 17% of controls) and the prevalence was particularly high among young cases (45% of the BCC cases were <44 years old). In our study, subjects analyzed for DRC with a family history of skin cancer were very rare (7% of the cases distributed mostly in the older age group). This may explain why the DRC of young BCC cases (age, ≤40 years) in our study was not significantly different from that of their controls (although relatively lower), whereas in the Baltimore study, reduced DRC was significantly associated with both early onset of BCC and family history of this disease (5). In the study conducted by Hall et al. (13), the finding of an increase in the DRC of BCC cases as compared with their controls, which resembles what we have observed, could not be further analyzed because of the lack of information about life-style habits and family history of cancer of the study population.

In conclusion, the significance of the DRC assay should be further verified in larger, well-designed epidemiological studies in which the interactions between life-style factors, inherited predisposition, DNA repair, and other factors that may influence DNA repair are evaluated.

We thank L. Gargano for technical assistance.