Modulation of Biologic Endpoints by Topical Difluoromethylornithine (DFMO), in Subjects at High-Risk for Nonmelanoma Skin Cancer¹ Free

More than one million new skin cancers are diagnosed yearly in the United States accounting for ∼40% of all of the new cancer diagnoses. The incidence of skin cancer has been increasing dramatically, and this increase is expected to continue as the population ages and larger amounts of UV radiation reach the surface of the earth because of depletion of the ozone layer (1). Nonmelanoma skin cancers make up the majority of skin cancers. Approximately 80% of nonmelanoma skin cancers are basal cell carcinomas, and 20% are SCCs (2). Basal cell carcinoma is a slow growing tumor that rarely metastasizes but can cause tremendous morbidity. In contrast, SCCs are more likely to metastasize. SCC was responsible for ∼1200 deaths in 1998 (a number equivalent to the yearly mortality attributed to Hodgkin’s lymphoma; Refs. 2, 3, 4). SCC has a premalignant precursor, AK (5, 6, 7). The rate of SCC in the population is much lower than the rate of the premalignant AKs demonstrating that not all AKs progress to SCC (6).

Effective primary preventive and chemopreventive strategies are necessary to reduce this high incidence of skin cancer. It is hypothesized that cancer chemoprevention could prevent or delay cancer in high-risk subjects, such as those with AK, using dietary and/or chemical interventions (8). Clinical endpoints such as tumor incidence and number are often used in chemoprevention studies. However, measures that reflect basic cellular events (histological change, cell proliferation, and death or apoptosis), molecular targets (p53 mutation/overexpression), or biochemical pathways (polyamine synthesis) can function as surrogate endpoint biomarkers in chemoprevention studies and aid in determining the mechanism(s) by which agents act to alter the carcinogenesis process.

DFMO, an irreversible inhibitor of ornithine decarboxylase, suppresses increased polyamine synthesis and inhibits skin tumors in UVB- and chemically induced animal models of skin carcinogenesis. Consequently, DFMO appears to be a high priority candidate for chemoprevention trials in high-risk populations. DFMO is an ornithine analogue and irreversible inhibitor of ornithine decarboxylase, the first enzyme in the polyamine pathway. The polyamines putrescine, spermidine, and spermine are ubiquitous polycations that are essential for normal cellular proliferation and differentiation (9).

We reported previously that topical DFMO reduced the number of AK and the level of spermidine in biopsies of AK (10). In the current study, we determined whether topical DFMO can alter additional surrogate endpoint biomarkers, including cellular proliferation (i.e. PCNA), apoptosis, p53 protein expression, and the number as well as type of p53 mutations in biopsies from participants with multiple AK.

Males and females with 10 or more clinical AKs on their forearms were recruited from Pima and adjoining Southeastern Arizona counties. Institutional informed consent was obtained. Details of the study design and methods have been published previously (10). A 1-month placebo (hydrophilic cream twice daily) run-in was performed before randomization to placebo or DFMO cream on the right or left arm, so that one arm served as a control for the contralateral arm. An equal number of right and left arms were treated (10).

Baseline measures and biopsies were done at the end of the 1-month placebo run-in and again at the end of the 6-month placebo or DFMO treatment. AKs were circled (with washable ink) and counted by a dermatologist or dermatological surgeon. Skin punch biopsies were obtained for polyamine levels, and shave biopsies were obtained for PCNA and apoptotic indices, p53 protein expression, and p53 mutations. All of the biopsies were obtained from AKs on the dorsal surface of forearms between knuckle and elbow.

The area of skin to be biopsied was anesthetized with 1% xylocaine with epinephrine (Ekins-Cinn Inc., Cherry Hill, NJ). Procedures for measurement of polyamine levels in skin have been published previously (10). Shave biopsies were obtained with a #15 scalpel, yielding primarily epidermis and a minimal amount of dermis. Immediately after removal, biopsies were placed into 10% formalin (apoptosis) or 70% ethanol until histological processing (p53 and PCNA immunohistochemistry) or snap frozen (p53 mutations).

The number of apoptotic cells based on morphology (i.e., condensed and/or pyknotic nuclei, eosinophilic cytoplasm, formation of apoptotic bodies) on H&E were expressed as the number of apoptotic cells/mm of skin. The entire tissue section was included in the count. We demonstrated previously that in situ terminal deoxynucleotidyl transferase (Tdt)-mediated nick end labeling and morphological evaluation of apoptotic cells on H&E yield essentially identical results in skin biopsies (11).

After deparaffinization of 3-μm tissue sections through a series of alcohols, immunohistochemical staining was performed using a streptavidin-biotin peroxidase system with a 3,3′-diaminobenzidine chromagen and a hematoxylin counterstain (Ventana Medical Systems, Tucson, AZ) on an automated VMS 320 immunostainer (Ventana Medical Systems). Anti-PCNA PC10 and p53 (Oncogene Science, Uniondale, NY) were used at a 1:200 dilution in antibody diluent (Ventana Medical Systems). Negative controls (to assure lack of nonspecific staining) were run on each sample by substituting antibody diluent for the antibody. Tonsil tissue (PCNA) or the T47D breast cancer cell line (p53; American Type Culture Collection, Manassas, VA) was included in each run as a positive control to assure proper staining and to assess variability of staining intensity. Tissue sections were measured on an RPW Image Analysis System (Roche Image Analysis Systems, Burlington, NC). The percentage of positive nuclear area per ×40 field was determined for both PCNA and p53-labeling index calculations.

Skin biopsies were digested with lysis buffer [1 m NaHCO₃, 1 m Na₂CO₃, 0.5 m EDTA, N-lauroyl sarcosine, proteinase K (pH 8.0)] overnight at 50°C. Samples were then extracted with phenol-chloroform-isoamyl alcohol. DNA was resuspended in sterile water and stored at −20°C.

Radioactive PCRs were performed using primers for exons 5–8 (Clontech Laboratories, Inc., Palo Alto, CA; Ref. 11). The radioactive PCR procedure for each exon was as follows: 2 μl of genomic DNA, 36.5 μl of sterile filtered water, 11.5 μl of PCR mixture (5 μl PCR buffer, 2 μl of 2.5 mm nucleotide triphosphates, 2 μl of the up and downstream primer at 10 μm, 1 μl Taq for each sample, and various amounts of P³² cytidine triphosphate depending on the activity). A mock tube was also run with no DNA added. Samples were overlayed with 50 μl of mineral oil and heated to 94°C before adding the mixture (11).

PCR products were diluted in loading buffer (95% formamide, 5% EDTA, 10 mm NaOH, and 0.025% xylene cyanol blue), heat denatured, and loaded onto 6% PAGE 49:1 with 5% glycerol gels, run at room temperature at 10 W for 4 h (exon 7, 8), or without 5% glycerol (exons 5, 6) run at 4°C, 20 W for 3–4 h (12). Positive controls were run for each exon: U266 for exon 5; T47D for exon 6; Ovcar-3 for exon 7; and 8226 for exon 8. DNA from normal human lymphocytes was used as negative controls. Positive samples were those showing a bandshift compared with control.

Samples that showed a bandshift on single-strand conformational polymorphism gels were subjected to direct DNA sequencing in both directions after nonradioactive reamplication of genomic DNA (exons 5–8; Clontech Laboratories). PCR product was run on a 2% agarose gel to assure amplification, and PCR products were prepared for sequencing with spin columns. Samples were sequenced with a Sanger’s dideoxy termination method on an Applied Biosystems model 373A DNA sequencer (Applied Biosystems, Foster City, CA). Sequencher (Gene Codes Corp., Ann Arbor, MI) was used for mutation analysis of chromatograms. Mutations had to be present in both directions to be considered positive.

DFMO effects were assessed by fitting regression models for change data. The outcome measure was the difference in the change (of a factor) on the DFMO treated arm minus the difference on the untreated arm. The explanatory factor was the change in the untreated arm, centered at its mean so that the regression intercept had the interpretation as the (adjusted) DFMO effect (10). Initial values were used to adjust for regression to the mean, whereas the change on the untreated arm was used to adjust for person-specific effects. Additionally, all of the models initially included dummy variables for whether the right or left arm was treated. A nonparametric Wilcoxon signed-rank test was used to determine whether there were any differences between right and left arms at baseline for endpoint measures.

McNemars test of proportions was used for analysis of p53 mutations (Stata Corporation, College Station, TX). Apoptotic rates were normally distributed and consequently were analyzed using a paired t test (SPSS, Chicago, IL).

Alberts et al. (10) reported previously a significant reductions in the number of AK as well as in concentrations of spermidine in biopsies of AK in a randomized placebo-controlled topical study of DFMO in subjects with multiple AK. The posterior forearms of each subject were assigned to either DFMO or placebo so that one arm served as a control for the contralateral arm. The mean age of the 42 participants who completed the trial and contributed biopsies for endpoint measures was 69 years, and 76% were male (10).

The primary publication (10) reported a significant decrease of 6.1 AK (23.5% reduction; P = 0.001) on the DFMO-treated arms as illustrated in Fig. 1 A. DFMO also significantly reduced spermidine concentrations in skin by 26% (P = 0.04). Suppression of skin spermidine levels was interpreted as a measure of DFMO effect. Of interest, when the treated arms were analyzed separately, the significant affect of DFMO on AK number and spermidine concentrations was primarily limited to right arms (10).

Biopsies of AK from the original topical DFMO trial (10) were evaluated for surrogate endpoint biomarkers including analysis of p53 mutations, p53 expression measured by immunohistochemistry, PCNA, and apoptotic indices.

Individual p53 mutations are listed in Table 1, A (baseline biopsies), B (DFMO-treated biopsies), and C (placebo-treated biopsies). The headings for Table 1, A–C include the arm on which the p53 mutation was found, exon affected (5, 6, 7, 8), the specific p53 codon affected, the base sequence surrounding the p53 mutation, the specific base change, and the resulting amino acid change. Fig. 1,B and Table 2 summarize the p53 mutation results. DFMO did not affect the frequency of p53 mutations in AK lesions (baseline 21.4%, final DFMO treated 28.6%, and final placebo treated 26.2%; P = 0.61). At baseline, 80% (16 of 20) of p53 mutations occurred at dipyrimidine sites, with 65% (13 of 20) of these mutations were C-T substitutions, deletions of C or T, or additions of C or T. There were no CC-TT mutations in any of the biopsies. The frequency of p53 mutations at dipyrimidine sites or transitions of C-T were similar at baseline and after DFMO or placebo treatment (summarized in Table 3).

Table 4 summarizes the p53 mutation data across p53 exons 5–8. In baseline biopsies, 55% (11 of 20) of mutations were in exon 5, 5% (1 of 20) in exon 6, 20% (4 of 20) in exon 7, and 20% (4 of 20) in exon 8. The arm treated with placebo had 25% (3 of 12), 25% (3 of 12), 50% (6 of 12), and 8% (1 of 12) of the p53 mutations in exons 5, 6, 7, and 8, respectively. DFMO-treated arms had 13% (2 of 15) of p53 mutations in exon 5, 13% (2 of 15) in exon 6, 47% (7 of 15) in exon 7, and 27% (4 of 15) in exon 8.

There appeared to be fewer p53 mutations on DFMO-treated right arms (19%; 4 of 21) compared with DFMO-treated left arms (38%; 8 of 21) at the end of study (Fig. 1, B and C). This did not reach statistical significance.

The mean percentage of p53- and PCNA-positive cells by immunohistochemistry, and number of apoptotic cells on H&E/mm of AK lesion are shown in Fig. 2, A–C and in Table 5. Using a linear regression model, there was a significant reduction of 9.1% in p53 protein expression (P = 0.04) due to DFMO treatment or an approximate magnitude change of 22% (Fig. 2,A). The regression effects of DFMO treatment on p53 and PCNA expression are shown in Table 6. DFMO had no effect on PCNA expression (Fig. 2,B) or on the rate of apoptosis (Fig. 2 C). Apoptotic indices were normally distributed allowing comparison of the DFMO- and placebo-treated arms using a paired t test (P = 0.71). As discussed previously, the significant affect of DFMO was largely seen on right arms for reduction in AK number, spermidine, and a trend toward a decrease in the frequency of p53 mutations. This was not the case for p53 protein expression, PCNA expression, or apoptosis.

Cancer chemoprevention involves the treatment and/or reversal of intraepithelial neoplasia (e.g., actinic keratosis) with chemical and/or nutrient interventions. Chemoprevention trials often include biological measures of carcinogenesis, markers of cancer risk, and/or surrogate endpoint biomarkers of the effect of an intervention.

AKs represent a discreet step in UV-induced skin carcinogenesis, as well as being a marker of chronic sun exposure and increased risk of SCC. As reported previously (10), we evaluated the ability of topical DFMO to reduce the number of AK lesions and inhibit polyamine synthesis in the skin. Secondary endpoints included p53 mutation frequency and type, p53 protein expression, PCNA expression, and apoptotic indices in AK lesions after DFMO treatment.

We found no significant difference in the frequency of p53 mutation with DFMO treatment, but there was a trend toward a shift in the exon distribution of mutations (e.g., reduction in exons 5 and 6 mutations and an increase in exon 8). Similarly, a mouse UV-induced skin carcinogenesis model found a change in exon distribution with no change in the frequency of p53 mutations after green tea treatment (13). This is very speculative and will need to be confirmed in studies designed to detect these subtle differences. Ultimately, the specific type of p53 mutation may have more meaning than the cumulative frequency, although it will take larger numbers of subjects to detect these kinds of differences.

We reported previously that p53 expression (median p53 expression value as a cut point for considering positive or negative) was in fairly good agreement with presence of a p53 mutation (11). It is reasonable to assume that p53 mutation and p53 protein expression are linked, although measurement of both may have value in skin chemoprevention studies. p53 expression was increased in sun-damaged skin and AK in a manner similar to p53 mutation frequency (11). In the current study of topical DFMO, p53 expression was significantly reduced by DFMO, whereas there was no significant change in the frequency of p53 mutation. This reduction in p53 protein expression with DFMO treatment may be attributable to the elimination of p53-positive cells. A similar phenomenon was documented by Hong et al. (14, 15) with retinoid treatment of premalignant lesions of the head and neck.

We detected fewer p53 mutations than had been reported previously for AK. There are a number of explanations including the lack of microdissection of samples and the reliance on the clinical diagnosis of AK that was not confirmed histologically. The biopsies of clinical AK for p53 mutation analysis were snap frozen, and the entire biopsy was extracted.

DFMO did not significantly alter PCNA expression, suggesting that DFMO does not affect proliferation but alternatively may inhibit cell growth. Another DFMO chemoprevention study in subjects with colorectal adenomas found colonic mucosal cell proliferation was unaffected by DFMO, whereas polyamine synthesis was inhibited (16). More recently, a number of studies suggest that DFMO or polyamine depletion can suppress cell proliferation and far more consistently increase the rate of apoptosis (17, 18, 19).

DFMO did not significantly alter the apoptotic rate and, consequently, in this setting does not appear to function through an up-regulation of apoptosis in AK. This lack of a DFMO effect on both cell proliferation (PCNA) and apoptosis is paradoxical, because we did see a significant reduction in the number of AK. Consequently, the question remains as to the mechanism by which DFMO reduces the number of AKs. It remains possible that DFMO does work via an apoptotic mechanism in those AKs that regressed but were not present for repeat biopsy. The effect of DFMO on cell proliferation and apoptosis may be dependent on a number of factors that are likely to include cellular context (in vivo versus in vitro studies), cell or tissue type, as well as presence of genetic alterations like mutation of the p53 gene.

A limitation of this study is that we were unable to biopsy a single AK before and after DFMO treatment. The size and focal nature of AK makes this approach unrealistic. Alternatively, we biopsied random AKs on forearms so that a different AK was removed before treatment and after treatment. This may limit our ability to see a change in markers because of the high rate of heterogeneity in AK lesions and the fact that we selected for AKs that survived DFMO treatment. Importantly, even with this limitation, we were able to detect a change in skin biopsy p53 protein expression and spermidine concentrations with reductions that were very similar to the reduction seen with the number of AKs.