α-Difluoromethylornithine Induction of Apoptosis: A Mechanism Which Reverses Pre-established Cell Proliferation and Cancer Initiation in Esophageal Carcinogenesis in Zinc-deficient Rats¹

The polyamines spermidine and spermine and their precursor putrescine are required for cell growth and proliferation (1, 2, 3). Their cellular content is tightly regulated and is primarily dependent on the activity of the enzyme, ODC³, which catalyzes the conversion of ornithine to putrescine, the first step in the synthesis of polyamines. High levels of ODC and polyamines have been found in tumors and proliferating cells (1, 3). Thus, the polyamine metabolic pathway is a target for chemoprevention. DFMO, an irreversible inhibitor of ODC, depletes cellular putrescine and spermidine levels but has limited effect on spermine levels. DFMO has exhibited potent antiproliferative function against many tumor cell types. In addition, numerous studies have demonstrated that DFMO inhibits tumorigenesis in animal cancer models of bladder, colon, mammary gland, liver, stomach, skin (reviewed in Ref. 4), and esophagus (5). This compound is currently being tested as a chemopreventive agent for specific human cancers including precancerous Barrett’s esophagus (4). DFMO suppresses polyamine contents in Barrett’s esophagus and, therefore, may prevent the development of adenocarcinoma of the esophagus in this patient group (6). However, the mechanism of cancer prevention by DFMO is not fully understood (7).

The ZD rat model of NMBA-induced esophageal cancer, which closely mimics human esophageal squamous cell carcinoma, is a valuable tool to investigate mechanism(s) underlying cancer initiation and prevention. In this in vivo model, esophageal cell proliferation is induced by a reduced dietary intake of an essential trace metal and tumor induction by NMBA, an agent strongly suspected of causing the human cancer (8, 9). In addition, NMBA produces morphologically similar lesions in rat and human (10), and nutritional zinc-deficiency (11, 12, 13) has been implicated in the etiology of esophageal cancer in several high-risk areas for the disease. Results from our studies (14, 15, 16) have provided evidence that increased cell proliferation plays a critical role in rat esophageal carcinogenesis. Using the cell proliferation-driven, single NMBA-dose, ZD rat esophageal cancer model (16), we first reported a role for apoptosis in the mechanism of cancer prevention by DFMO (5). Our results showed that simultaneous administration of DFMO and a ZD diet to weanling rats from the beginning inhibited the onset of zinc-deficiency-induced esophageal cell proliferation by stimulating apoptosis and of NMBA-induced esophageal tumorigenesis in ZD rats.

Because esophageal cancer initiation by NMBA is very rapid in this cell proliferation-driven esophageal cancer model (17), the present study determined whether DFMO, administered after the establishment of a carcinogenic environment, i.e., after weanling rats had been given a ZD diet for 5 weeks to establish sustained cell proliferation in the esophagus and then a single NMBA dose to initiate cancer formation, is effective in preventing esophageal carcinogenesis. This experimental design is relevant to the human esophageal squamous cell carcinoma situation because both increased esophageal cell proliferation (18, 19) and exposure to carcinogenic nitrosamines (8, 9) were reported in high-risk areas for esophageal cancer in China.

DFMO was purchased from Ilex Oncology (San Antonio, TX). NMBA was from Ash Stevens, Inc. (Detroit, MI). Custom-formulated, egg white-based, ZD diet was prepared by Teklad (Madison, WI). The zinc level in this diet was regularly monitored by atomic absorption spectroscopy in our laboratory and was 3–4 parts/million. This diet is identical to the nutritionally complete, zinc-sufficient diet used in other studies (5, 14, 15, 16, 17) except for the amount of zinc carbonate added.

Thirty-four weanling male Sprague Dawley rats (Taconic Laboratory, Germantown, NY) were given a ZD diet and deionized water for 5 weeks to establish increased cell proliferation in the esophagus. After 5 weeks (0 h), five rats were sacrificed for cell proliferation studies, and the remaining 29 animals each received a single intragastric dose of NMBA at 2-mg/kg body weight. One h after NMBA dosing, 20 ZD rats were switched to deionized water containing 1% DFMO, while nine control ZD rats remained on deionized water. All of the animals continued on the ZD diet and were killed 15 weeks later for tumor incidence analysis and related studies.

At sacrifice, the animals were anesthetized with isoflurane (AZ-Buck, Owings Mills, MD), blood was collected from the retroorbital venous plexus of each animal, and serum was prepared for zinc analysis by atomic absorption spectroscopy (15). Whole esophagi were excised and opened longitudinally. Esophageal tumors greater than 1 mm in diameter were mapped and counted.

A small portion of the esophagus (upper one-third) was cut, fixed in buffered formalin, and embedded in paraffin. Serial cross-sections (4 μm) were prepared and were stained with H&E or left unstained for immunohistochemical studies. Esophageal epithelium was prepared from the remaining esophagus using a blade to strip off the connective tissue layer. Samples containing only the esophageal epithelia were snap frozen in liquid nitrogen and stored at −80°C until polyamine analysis.

Polyamine content was determined after separation by ion-pair reversed-phase high-performance liquid chromatography using fluorescence detection after post-column derivatization with o-phthalaldehyde as described previously (20). The tissue samples were extracted using 0.1 n HCl and then deproteinized by the addition of 10% perchloric acid. After centrifugation to remove protein, the supernatant was filtered through a 0.22 μm filter, and aliquots were applied to a column (Beckman Ultrasphere octadecyl silane 5 μm; 4.6 mm × 25 cm protected by a 4.6 mm × 4 cm guard column of octadecyl silane-5S) equilibrated with a mixture of 70% buffer A [0.1 m Na acetate, 0.01 m Na octane sulfonate (pH 4.5)] and 30% buffer B [20 parts 0.2 m Na acetate, 0.01 m Na octane sulfonate, (pH 4.5); 6 parts acetonitrile; 3 parts methanol]. The column was then eluted with a linear gradient of 70% buffer A/30% buffer B to 100% buffer B over 40 min at a flow rate of l ml/min at 35°C (21). The results were expressed as nmol/mg protein. Protein was determined by the method of Bradford (22).

Monoclonal mouse anti-PCNA (Santa Cruz Biotechnology, Santa Cruz, CA) was used at 1:250 dilution, followed by incubations with biotinylated goat antimouse antibody and streptavidin horseradish peroxidase, as described previously (16). PCNA was localized by a final incubation with 3-amino-9-ethylcarbazole-substrate-chromogen system (DAKO Corp., Carpinteria, CA) and a light hematoxylin counterstain. Cells with red reaction product in the nucleus were positive for PCNA. PCNA analysis has the potential to identify cell cycle subpopulations (G₁, S, G₂, and M). Dark-staining nuclei represent S-phase cells, light staining nuclei represent G₁-S and G₂ cells, cells with cytoplasmic staining usually depict mitoses, and nonstaining nuclei represent quiescent (G₀) cells (23). Preliminary analysis (data not shown) demonstrated a good correlation between S-phase cells measured with PCNA or bromodeoxyuridine.

LI (percentage) was calculated by dividing the number of respective PCNA-labeled nuclei by the total number of cells counted/cross-section of an entire esophagus.

Both the TUNEL (in situ end labeling of DNA in apoptotic cells) and morphological methods using H&E staining were used.

The 3′-OH end labeling of apoptotic cell DNA was performed by using an ApopTaq Plus in situ peroxidase detection kit (Intergen Co., Purchase, NY). Briefly, after deparaffinization and rehydration in graded alcohols, tissue sections were incubated with proteinase K. Endogenous peroxidase was inhibited with 3% hydrogen peroxide. Terminal deoxynucleotidyl transferase enzyme was then applied to catalyze the addition of digoxigenin-labeled nucleotides to the 3′-OH ends of the fragmented DNA (37°C/l h). Subsequently, the slides were incubated with a horseradish peroxidase-conjugated anti-digoxigenin antibody. Staining was developed with DAB (Sigma Chemical Co.), and sections were counterstained with methyl green. Sections (Intergen) from normal female rat mammary gland, in which extensive apoptosis occurs, served as a positive control. Negative controls were run in which terminal deoxynucleotidyl transferase was omitted.

Apoptotic cells take on varied forms depending on their stage in the apoptotic process, as characterized by (a) diffuse staining of cytoplasm with only minimal nuclear condensation; (b) distinct apoptotic bodies resulting from nuclear disintegration; and (c) dense staining of nuclei retaining normal nuclear structure. All of the forms were considered equivalent for the purpose of analysis (24).

AI (percentage) is calculated by dividing the number of apoptotic cells by the total number of cells/cross-section of an esophagus.

After deparaffinization and rehydration in graded alcohols, the sections were heated in citrate buffer (0.01 m; pH 6.0) in a microwave (90°C; 10 min) before nonspecific binding sites were blocked with goat serum. Sections were then incubated overnight at 37°C in a humidified chamber with polyclonal rabbit bcl-2 antibody (Santa Cruz) at 1:3000 dilution or with polyclonal rabbit bax antibody (Santa Cruz) at 1:800 dilution, followed by incubation with biotinylated goat antirabbit antibody. Bcl-2 and bax expression was visualized by a final incubation in DAB.

After deparaffinization and rehydration in graded alcohols, the sections were heated in citrate buffer (0.01 m; pH 6.0) for Rb in a microwave oven (90–95°C; 3 × 5 min) before nonspecific binding sites were blocked with goat serum. The antigen retrieval procedure was not done on sections for p16^ink4a staining. Sections were incubated overnight at 37°C in a humidified chamber with respective primary antibodies: mouse anti-p16^ink4a monoclonal antibody (Santa Cruz) at 1:300 dilution; and mouse anti-Rb monoclonal antibody (PharMingen, San Diego, CA) at 1:20 dilution. Incubation with appropriate biotinylated secondary antibodies followed. Slides were then incubated with streptavidin horseradish peroxidase. Expression of p16^ink4a and Rb was localized by a final incubation with DAB and a light hematoxylin counterstain.

Western blotting analysis of bax and bcl-2 was conducted on esophageal epithelia from a previous study (5), in which weanling rats were given a ZD diet and 1% DFMO in the drinking water from the beginning for 5 weeks and then received a single dose of NMBA. This DFMO protocol reduced the incidence of esophageal tumors in ZD rats from 80 to 4%. Esophageal epithelia, isolated from DFMO-treated and DFMO-untreated ZD rats at end point for tumor analysis, were homogenized in a buffer containing 10 mm Tris-HCl (pH 7.4), 150 mm NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 5 mm EDTA, 100 μg/ml aprotinin, 50 μg/ml leupeptin, 1 mm benzamidine, 7 μg/ml pepstatin, and 1 mm phenylmethylsulfonyl fluoride. Debris was removed by centrifugation at 16,000 × g for 20 min. The protein concentration in the extract samples was measured using a Bradford protein assay kit (Bio-Rad, Hercules, CA). Proteins (100 μg) were separated by 14% SDS-PAGE and transferred onto Immobilon-P membranes (Millipore, Bedford, MA). After transfer, membranes were stained with Ponceau S (Sigma Chemical Co.) to test for equal loading of the samples and washed three times with PBS/Tween 20. Membranes were individually probed with goat polyclonal antibodies against bax and bcl-2. After antibody binding, membranes were incubated with appropriate horseradish peroxidase conjugate (Pierce Chemical Co., Rockford, IL). All of the incubations and washes were performed in PBS. Immunodetection was performed using the enhanced chemiluminescence method for Western blotting detection (Pierce Chemical Co.).

Data on polyamine contents, cell proliferation, and apoptosis were analyzed by one-way ANOVA using SAS statistical computer program (25). Tumor incidence differences were analyzed by Fisher’s exact test (26).

Continuous treatment of DFMO produces a substantial reduction in putrescine (85%) and spermidine (64%) contents in the esophagus of ZD/DFMO rats compared with ZD/H₂O animals at week 15 but an increase in spermine level (251%; Table 1). This increase has been seen in many cases where DFMO is used. The loss of putrescine formation renders the conversion of putrescine to spermidine the limiting step, and all of the available decarboxylated S-adenosylmethionine not used in this reaction can be used by spermine synthase to convert the residual spermidine into spermine (1, 2).

Also, a significantly higher putrescine level (P < 0.001) was found in ZD/H₂O animals at week 15 (1.9 ± 0.61 nmol/mg protein) compared with that (0.54 ± 0.26 nmol/mg protein) observed in DFMO-untreated ZD rats before NMBA treatment (5). This difference may be because of the fact that the esophagus of ZD/H₂O rats at week 15 has an average of 2.8 ± 2.2 tumors/esophagus, and tumors are known to contain high levels of ODC and polyamines (1, 3).

Consistent with our previous results (16), a single dose of NMBA produced a 89% (8/9) incidence of esophageal tumors in ZD rats with a multiplicity of 2.8 ± 2.2 tumors/esophagus at week 15. On the other hand, ZD rats that were switched to drinking water containing 1% DFMO after 5 weeks of a deficient diet exhibited a greatly reduced tumor incidence of 10% (2/20) with a markedly lower multiplicity of tumors/esophagus (0.15 ± 0.49) at the end point. These differences are significant at P < 0.0001 (two-tailed Fisher’s exact test) for tumor incidence and P = 0.008 for tumors/esophagus. In addition, similar levels of serum zinc content were found in ZD/H₂O rats (56 ± 16 μg/100 ml) and ZD/DFMO animals (58 ± 14 μg/100 ml); these levels were substantially lower than those reported for zinc-sufficient rats (5). Our results demonstrate that DFMO, given after the establishment of a carcinogenic environment by dietary zinc deficiency and administration of the carcinogen, effectively prevents esophageal cancer formation in ZD rats.

PCNA immunostaining showed DFMO, given after the animals had been fed a ZD diet for 5 weeks, effectively reversed the pre-established esophageal cell proliferation (Fig. 1, A–C). As illustrated in Fig. 1,A, esophagus from a ZD/H₂O animal at 0 h (before NMBA treatment) showed hyperplasia and a focal hyperplastic lesion with numerous PCNA-stained nuclei. At week 15 after NMBA treatment, abundant PCNA-positive nuclei were found in areas of hyperplasia, dysplasia, focal hyperplastic upgrowth (Fig. 1,B), and papillomas (data not shown). In stark contrast, esophagus of ZD/DFMO rats (Fig. 1 C) showed a very small number of PCNA-positive nuclei in basal cells of a restored esophageal epithelium that was two to three cells thick and resembled that reported for pair-fed, zinc-sufficient rats (15).

There are three separate indicators of epithelial cell proliferation: (a) the number of labeled cells (S-phase; G₁-S/G₂ cells)/cross-section of an esophagus averaged for the group; (b) the total number of cells for the group, both labeled and unlabeled; and (c) LI, the percentage of labeled cells for the group. At week 15, ZD/DFMO rats had markedly lower LI for S-phase only and S-phase and G₁-S/G₂ cells than ZD/H₂O animals at week 15 or 0 h (Fig. 2). In addition, the total number of cells/cross-section of an esophagus for ZD/DFMO rats (1389 ± 174) was substantially reduced compared with those of ZD/H₂O animals (2732 ± 139 at 0 h; 2543 ± 546 at week 15). The difference was significant at P < 0.001. These results affirm that DFMO treatment reverses and inhibits cell proliferation in the ZD esophagus.

At week 15, esophagus from ZD/H₂O rats displayed very reduced staining for p16^ink4a in focal hyperplastic lesions, hyperplastic areas (Fig. 3,C), and tumor areas (data not shown). However, strong staining for Rb was found in hyperplastic areas, involving basal cells (Fig. 3,A), and in focal hyperplastic lesions and papillomas (data not shown). These results are consistent with our previous findings (17), which demonstrated an inverse relationship between p16^ink4a and Rb expression in proliferative ZD rat esophagus. In contrast, strong expression of p16^ink4a involving basal as well as suprabasal cells was seen in the now restored esophageal epithelium from ZD/DFMO rat (Fig. 3,D). Rb expression in the esophagus from these animals was mostly localized in suprabasal cells (Fig. 3 B). Importantly, these data indicate that DFMO treatment restored the p16^ink4a/Rb control at G₁ to S, which was shown to be deregulated in ZD esophagus with uncontrolled cell proliferation (17).

At 0 h, a H&E-stained esophageal section of a ZD/H₂O rat typically showed a thickened epithelium with multiple deep folding and focal hyperplastic lesions, which are considered as neoplastic precursors (Fig. 4,A). At week 15 after a dose of NMBA, papillomas are commonly seen in the ZD/H₂O esophagus (Fig. 4,B). Remarkably, DFMO treatment produced a ZD esophageal epithelium that was two to four cells thick, covered by a thin keratin (Fig. 4,C). However, histopathological examination revealed numerous apoptotic cells in the basal and suprabasal cell layers of ZD/DFMO esophagus (Fig. 4, C and D). These results were confirmed by in situ TUNEL assay of adjacent esophageal sections (Fig. 4, G and H). On the other hand, apoptotic cells are only found sporadically in the proliferative esophagus from ZD/H₂O rats at either time point (Fig. 4, E and F). As stated above, the total number of cells/cross-section of an esophagus for ZD/DFMO rats at week 15 was substantially reduced compared with those of ZD/H₂O animals. However, the number of apoptotic cells/cross-section of an esophagus is significantly higher (P < 0.01) in ZD/DFMO rats than in ZD/H₂O animals (78.5 ± 22.5 versus 55.4 ± 11.3 at 0 h; and versus 54.1 ± 12.7 at week 15). As shown in Fig. 5, AI was significantly higher in ZD/DFMO esophagus (5.9 ± 1.7) compared with that in ZD/H₂O esophagus (2.1 ± 0.4 and 2.0 ± 0.2 at 0 h and week 15, respectively). In addition, AI varies greatly among esophagi from ZD/DFMO animals, ranging from 4.4 to 11.3%. These results provide evidence that DFMO reduces esophageal cell proliferation by way of apoptosis.

Fig. 6 shows typical immunostaining pattern of bax, a proapoptotic protein, and bcl-2, an antiapoptotic protein. At week 15, weak cytoplasmic expression of bax was found in areas of the ZD/H₂O esophagus showing hyperplasia and focal hyperplastic lesion (Fig. 6,A) and in papillomas (data not shown). On the other hand, strong cytoplasmic expression of bcl-2 (Fig. 6,C) was observed in these same lesions described for ZD/H₂O animals. In contrast, intense cytoplasmic staining of bax involving basal and suprabasal cell layers (Fig. 6,B) versus weak bcl-2 staining (Fig. 6 D) was often seen in the now restored epithelium of ZD/DFMO rats that also displayed an increased number of apoptotic cells (adjacent section; data not shown).

Immunoblotting results for bax and bcl-2 analysis are consistent with those obtained using immunohistochemistry. On the basis of equal loading of cellular protein, ZD/DFMO esophagi show strong expression of bax (Fig. 7,A, Lanes 1–4) compared with ZD/H₂O esophagi (Fig. 7,A, Lanes 5–7), whereas ZD/H₂O esophagi show overexpression of bcl-2 (Fig. 7,B, Lanes 6–9) compared with ZD/DFMO esophagi (Fig. 7 B, Lanes 1–5). Thus, our data show an inverse relationship between the expression of bax and bcl-2 in the ZD/DFMO esophagus that exhibits increased apoptotic activities and in the ZD/H₂O esophagus that displays hyperplasia and preneoplastic lesions.

Weanling rats given a ZD diet for 5 weeks develop sustained uncontrolled cell proliferation in the esophagus (15, 16) that converts a nontumorigenic dose (27) of NMBA into a highly tumorigenic one (16). In this study, continuous provision of DFMO after the establishment of unrestrained cell proliferation in the esophagus and NMBA dosing remarkably reverses esophageal cell proliferation (Figs. 1 and 2) and subsequent tumor development. At week 15, ZD/DFMO esophagi exhibited a restored epithelium showing overexpression of p16^ink4a (Fig. 3,D) and underexpression of Rb involving mostly the suprabasal cells (Fig. 3,B), relative to respective ZD/H₂O esophagi (Fig. 3, A and C). Importantly, these data indicate that DFMO treatment reinstates the p16^ink4a/Rb control at G₁ to S, shown to be deregulated in the highly proliferative ZD esophagus (17). In this regard, abnormal regulation of the p16^ink4a-cyclin D1/Cdk4/6-Rb pathway, which can lead to uncontrolled cell proliferation and tumorigenesis, is positively regulated by mitogenic growth factors and negatively regulated by cyclin-dependent kinase inhibitors including p16^ink4a and other members of the INK4 gene family (reviewed in Ref. 28).

On the other hand, Table 1 shows that ZD/DFMO rats weighed 18% less than ZD/H₂O animals at week 15. DFMO administration is known to cause reduced food consumption and, thus, decreased body weight in several rodent cancer models (5, 29). Our earlier work (15) has established that the profound effect of dietary zinc deficiency on esophageal cell proliferation/tumorigenesis in rats is specific and overrides the opposing effect of caloric restriction, simultaneously accomplished by a reduced intake of zinc. Thus, the dramatic effect of DFMO in reversing and inhibiting the cell proliferation-driven esophageal tumorigenesis in DFMO/ZD rats is specific and unlikely to be attributed to a further caloric restriction in these animals relative to ZD/H₂O rats.

To date, the ability of DFMO to enhance apoptosis in cancer prevention has only been demonstrated in a few in vivo models (5, 30). Importantly, the present results show that induction of apoptosis is a mechanism underlying the chemopreventive and antiproliferative activities of DFMO. Thus, morphological changes detected in epithelial cells (Fig. 4, C–D), DNA fragmentation detected by TUNEL assay (Fig. 4, G–H), overexpression of bax (Fig. 6,B), a proapoptotic protein, and underexpression of bcl-2 (Fig. 6,D), an antiapoptotic protein, all of which are characteristic of apoptosis, are observed in esophagi from ZD/DFMO rats at week 15. On the contrary, proliferative esophagi from ZD/H₂O animals at 0 h or week 15 only show sporadic occurrence of apoptotic cells (Fig. 4, A–B and E–F), accompanied by strong staining of bcl-2 (Fig. 6,C) and weak expression of bax (Fig. 6 A).

Because preneoplastic precursor lesions have been detected in ZD esophagus as early as 24 h after NMBA treatment, and tumor induction is rapid under the carcinogenic environment provided by uncontrolled cell proliferation (17), the results from this study suggest that effective cancer prevention by DFMO under the present experimental condition entails: (a) prompt induction of apoptosis to remove damaged cells and, thus, reverse esophageal cell proliferation; and (b) sustained inhibition of cell proliferation to annul the continued effect of dietary zinc deficiency. In this study, experiments to determine the onset of apoptosis after DFMO treatment have not been performed. However, it is possible that the reversal of cell proliferation brought about by apoptosis may occur rapidly or within hours after the application of the chemopreventive agent, a conclusion based on our recent results obtained with zinc replenishment in ZD rats (31).

Many studies have now demonstrated roles for polyamines in apoptosis, but the underlying mechanism for these effects is still unknown. Overaccumulation of polyamines may induce apoptosis in both oxidation-dependent (32, 33) and oxidation-independent processes (34, 35, 36). Spermine has been shown to trigger caspase-3 activation both in vivo (37) and in vitro as a result of enhanced release of mitochondrial cytochrome c (38), and polyamines are needed for the movement of plasma membrane phospholipids that accompanies apoptosis (39). In contrast, the ability of DFMO, which lowers overall polyamine levels, to induce apoptosis is clearly demonstrated in the current report and our previous studies (5) in the esophagus of ZD rats and in several other recent papers (30, 40) for other tissues. Thus, an increased AI was seen in a human gastric cancer model after exposure to DFMO (40), and DFMO increased the AI in rat colon adenomas induced by azoxymethane (30). The effects of alteration in polyamine levels on apoptosis may be biphasic with both elevated and reduced levels favoring cell death, but it is noteworthy that DFMO causes a major increase in the ratio of spermine to its precursors spermidine and putrescine. It is possible that this alteration has the same effect as gross elevation of polyamine content.

Many studies on experimental animals have confirmed the potential value of DFMO as a chemopreventive agent, and current clinical trials to examine whether these results are translatable to the clinic for the treatment of individuals at high risk from neoplastic disease should establish whether this drug is useful, provided that adequate protocols are followed (see Refs. 4, 7, 30) and references therein). Although it is clear that the effects of DFMO are related to the reduction in the ability to synthesize polyamines, a better understanding of the mechanism by which altered polyamines alter the development of neoplastic growth is needed both to optimize the conditions for the use of DFMO and to allow for the production of second generation agents with greater effectiveness and ease of use. The ZD rat esophageal cancer model provides a useful model system in which such mechanistic studies can be carried out.

We thank Karl Smalley for help with statistical analysis of data. Dr. Peter N. Magee died in February 2000. This paper is dedicated in his memory. We are very grateful to have the opportunity to work with Dr. Magee, an outstanding scientist and former Editor-in-Chief of Cancer Research.