Autophagy is an evolutionarily conserved stress response mechanism that often occurs in apoptosis-defective cancer cells and can protect against cell death. In this study, we investigated how apoptosis and autophagy affect each other in cancer cells in response to chemotherapeutic treatment. We found that specific ablation of the proapoptotic function of cytochrome c, a key regulator of mitochondria-mediated apoptosis, enhanced autophagy following chemotherapeutic treatment. Induction of autophagy required Beclin 1 and was associated with blockage of Beclin 1 cleavage by caspase 8 at two sites. To investigate the role of Beclin 1 cleavage in the suppression of autophagy and cell survival, a caspase-resistant mutant of Beclin 1 was knocked into HCT116 colon cancer cells. Beclin 1 mutant knockin resulted in markedly increased autophagy and improved long-term cell survival after chemotherapeutic treatment but without affecting apoptosis and caspase activation. Furthermore, Beclin 1 mutant tumors were significantly less responsive to chemotherapeutic treatment than were wild-type tumors. These results show that chemotherapy-induced apoptosis inhibits autophagy at the execution stage subsequent to cytochrome c release through caspase 8–mediated cleavage of Beclin 1. If apoptosis fails to execute, autophagy is unleashed due to lack of Beclin 1 cleavage by caspases and can contribute to cancer cell survival and therapeutic resistance. Therefore, Beclin 1 may be a useful target for inhibiting autophagy to sensitize chemotherapy. Cancer Res; 71(10); 3625–34. ©2011 AACR.
Apoptosis is a major cytotoxic mechanism of chemotherapy, and its deregulation contributes to therapeutic resistance (1, 2). Stress-induced apoptosis in mammalian cells often proceeds through the intrinsic pathway mediated by mitochondria and the Bcl-2 family proteins. Once cells are committed to apoptosis, a number of downstream events are triggered to execute cell death, including permeabilization of the mitochondrial outer membrane, release of several apoptogenic mitochondrial proteins such as cytochrome c, and activation of the caspase cascade (3, 4).
Autophagy, a catabolic degradation process, has recently emerged as an important stress response and cell death regulatory mechanism (5). It is characterized by the formation of double membrane autophagosomes, which enclose cytosolic materials and organelles, and then fuse with lysosomes, leading to degradation of their luminal content by lysosomal proteases (6). Autophagy is controlled by a group of evolutionarily conserved autophagy-related genes (ATG), among which Bcl-2-interacting protein 1 (Beclin 1) is a key regulator of autophagy (7). Beclin 1 binds to VPS34, a class III phosphoinositide 3-kinase, to drive autophagosome formation (6). A number of studies indicate a functional role of Beclin 1 in cancer (7–9).
Apoptosis interconnects with autophagy in a number of ways (10). Morphologic features of autophagy are often observed in dying cells in which caspases are not sufficiently activated (11). Autophagy can be inhibited by antiapoptotic Bcl-2 and Bcl-XL (12), which interact with Beclin 1 through its Bcl-2 homology 3 (BH3) domain (13). Several ATG proteins, including ATG5 and Beclin 1, can be cleaved by calpain or caspases that are activated during apoptosis (14, 15). Inhibition of autophagy is known to enhance chemotherapy-induced apoptosis, and autophagy-modulating drugs are being tested in a number of combination chemotherapy trials (16). However, it has remained unclear how apoptosis and autophagy affect each other in chemotherapy-induced apoptosis and whether autophagy is simply an innocent bystander that is suppressed during the execution of apoptosis (17).
In this study, we used a genetic approach to dissect when and how autophagy is suppressed during chemotherapy-induced apoptosis. Our results revealed autophagy suppression through caspase 8–mediated cleavage of Beclin 1, following cytochrome c release. Specific ablation of Beclin 1 cleavage restored autophagy in apoptotic cells and led to enhanced cancer cell survival and therapeutic resistance.
Materials and Methods
Human colorectal cancer cell lines, including HCT116 (obtained before 2002), Caco2 (obtained in 2008), RKO (obtained in 2010), and LoVo (obtained before 2002), were from American Type Culture Collection. Cell lines were last tested and authenticated for absence of Mycoplasma, genotypes, drug response, and morphology in our laboratory in October 2010.
Cells were lysed in the presence of a protease inhibitor cocktail (Roche Diagnostics) with RIPA radioimmunoprecipitation assay buffer (for probing LC3 and p62), a mixture of 0.25 mol/L sucrose, 10 mmol/L HEPES (pH 7.4), and 1 mmol/L EGTA (for probing endogenous Beclin 1), or 2× Laemmli sample buffer (for probing all other proteins). Western blotting was carried out as previously described (18). Antibodies included those against N-terminal and C-terminal Beclin 1 (Sigma), caspase 8 (Cell Signaling Technology), glutathione S-transferase (GST; GE Healthcare), V5 (Invitrogen), cytochrome oxidase subunit IV (Cox IV; Invitrogen), α-tubulin (BD Biosciences), LC3 (19), and p62 (Novus Biologicals).
Analysis of apoptosis
After treatment, attached and floating cells were harvested at different time points. Apoptosis was analyzed by counting cells with condensed chromatin and micronucleation following nuclear staining with Hoechst 33258 (Invitrogen; ref. 20). A minimum of 300 cells were analyzed for each sample. Annexin V staining was carried out as previously described (21). Activation of caspases 3, 8, and 9 was analyzed by Western blotting. Long-term cell survival was determined by colony formation assay (22). Briefly, an equal number of treated cells were diluted and seeded in 12-well plates. As a control, untreated cells were also plated at 1:100 dilutions relative to treated cells. After 14 days, cell colonies were visualized by crystal violet staining and counted. Results of nuclear staining and colony formation assays were expressed as means ± SD of 3 independent experiments.
Analysis of autophagy
Cells were treated by camptothecin (CPT) with (for LC3) or without (for p62) the lysosomal inhibitors E64d (10 μg/mL) and pepstatin A (10 μg/mL). LC3II accumulation and p62 degradation were analyzed by Western blotting. For analysis of green fluorescent protein (GFP)-LC3 puncta formation, cells were transfected with pEGFP-LC3 (19) prior to treatment to induce autophagy. GFP-LC3 puncta signals were detected by confocal microscopy using a Zeiss LSM Pascal 5 and the Plan-Apochromat 63×/1.40 Oil DIC objective lenses. Images were acquired using the Zeiss LSM Image Browser Software. GFP-LC3 puncta signals were quantified by counting at least 300 cells for each sample and expressed as means ± SD of 3 independent experiments. Autophagic vesicles, including autophagosomes and autolysosomes, were analyzed by transmission electron microscopy as described (19). Electron microscopy sections were imaged using a JEOL JEM 1011 (Peabody) at 80 V fitted with a bottom mount AMT 2k digital camera (Advanced Microscopy Techniques).
Knockin of mutant Beclin 1
The Beclin 1 targeting vector was constructed using the pUSER-rAAV (recombinant adenoassociated virus) System (23). Briefly, 2 1-kb homologous arms flanking the fourth intron of Beclin 1 were inserted between 2 USER sites in the AAV shuttle vector pTK-Neo-USER. The coding sequence for Beclin 1 double mutant (DM; D133A/D146A) was introduced into the right arm using the QuickChange XL Site-Directed Mutagenesis Kit (Agilent Technologies). For gene targeting, HCT116 cells were infected with the targeting rAAV and selected by G418 (0.5 mg/mL; Mediatech) for 3 weeks. G418-resistant clones were pooled and screened by PCR for targeting events using the primer pairs listed in Supplementary Table S1. To target the second allele, Neo flanked by 2 LoxP sites was excised from a heterozygous clone by infection with an adenovirus expressing Cre recombinase (Ad-Cre). The same targeting construct was used in the second round of gene targeting. After the second round, Neo was excised by Ad-Cre infection and targeting was verified by sequencing of genomic DNA and Western blotting.
Xenograft tumor experiments
All animal experiments were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Xenograft tumors were established by s.c. injecting 3.5 × 106 wild-type (WT), or 7.0 × 106Beclin 1-KI HCT116 cells into both flanks of 5- to 6-week-old female Nu/Nu mice (Charles River). Mice were treated with CPT-11 (irinotican; APP Pharmaceuticals) as previously described (24). Briefly, treatment was initiated on day 5 by intraperitoneal injection of 40 mg/kg of CPT-11 in 20 mmol/L sodium citrate buffer (pH 6.0) and repeated on days 9, 13, 17, and 21. Tumor growth was monitored by calipers, and tumor volumes were calculated according to the formula: 0.5 × length × width2.
Statistical analysis was conducted using GraphPad Prism IV software. P values were calculated by Student t test. Differences were considered significant if the probability of the difference occurring by chance was less than 5 in 100 (P < 0.05).
Induction of autophagy in apoptosis-defective cytochrome c mutant knockin cells
To determine how autophagy in apoptosis-defective cancer cells contributes to cell survival and therapeutic resistance, we dissected at which step autophagy occurs in apoptosis-defective cancer cells. The K72A mutant of cytochrome c (Cyt c), which disrupts its interaction with Apaf-1 but preserves its electron transport function (25), was knocked into HCT116 colon cancer cells (Wang and colleagues, unpublished data). WT and cytochrome c-knockin (Cyt c-KI) cells were treated with CPT, a DNA damaging agent whose derivatives are commonly used in cancer chemotherapy (26) to induce apoptosis. CPT-induced apoptosis was markedly reduced in Cyt c-KI cells (Fig 1A, left), although the mutant cytochrome c was still released into the cytosol following CPT treatment (Supplementary Fig. S1A). Activation of caspases 3, 8, and 9 was abrogated in Cyt c-KI cells compared with WT cells (Fig. 1A, right).
We then tested whether autophagy was induced in Cyt c-KI cells by analyzing commonly used autophagy markers including LC3 and p62. The level of lipidated LC3II was increased in Cyt c-KI cells compared with WT cells following treatment with CPT along with the lysosomal inhibitors E64d and pepstatin A (Fig. 1B). LC3II accumulation was also detected in Cyt c-KI cells treated with CPT in the presence of chloroquine, another lysosomal inhibitor, but not in cells treated without a lysosomal inhibitor (Supplementary Fig. S1B), indicating enhanced autophagosome formation but not reduced lysosomal degradation in these cells (27). The level of p62, a protein that is degraded by autophagy, was more reduced in CPT-treated Cyt c-KI cells than in WT cells (Fig. 1B). Confocal microscopic analysis showed enhanced GFP-LC3 puncta formation in Cyt c-KI cells transfected with GFP-tagged LC3 and then treated with CPT, relative to WT cells (Fig. 1C). Furthermore, transmission electron microscopy detected frequent autophagic vesicles, including double membrane autophagosomes, as well as autolysosomes with engulfed organelle structures (28), in CPT-treated Cyt c-KI cells but not in WT cells (Fig. 1D). These results show that autophagy can be induced when the initial execution step of apoptosis in Cyt c-KI cells is blocked.
Caspase-mediated Beclin 1 cleavage in apoptotic cells
We then tested whether Beclin 1, a critical regulator of autophagosome formation (7), was involved in CPT-induced autophagy in Cyt c-KI cells. CPT-induced GFP-LC3 puncta formation in Cyt c-KI cells was significantly decreased after siRNA knockdown of Beclin 1 (P < 0.01; Fig. 2A), indicating that autophagy in these cells was Beclin 1 dependent. Although blockage of apoptosis in Cyt c-KI cells resulted in a slight increase in cell survival in a long-term colony formation assay, knockdown of Beclin 1 significantly reduced the survival of Cyt c-KI cells following CPT treatment (Fig. 2B), suggesting that Beclin 1–mediated autophagy contributes to cell survival.
Although Beclin 1 expression was not significantly altered in Cyt c-KI cells compared with WT cells, 2 extra protein species were detected in WT HCT116 cells but not in Cyt c-KI cells following CPT treatment (Fig. 2C). Two similar fragments were found in cells transfected with V5-tagged Beclin 1 and then subjected to CPT treatment and detection of exogenous Beclin 1 (Supplementary Fig. S1C). Cyt c–dependent cleavage of Beclin 1 was also found in cells undergoing apoptosis induced by staurosporine (STS) or the BH3-only protein PUMA (Supplementary Fig. S1D). Beclin 1 cleavage was observed in 3 additional colon cancer cell lines following CPT treatment (Supplementary Fig. S1E). Furthermore, CPT-induced Beclin 1 cleavage was suppressed in BAX-knockout (BAX-KO) HCT116 cells (29) and in cells treated with CPT along with the pan-caspase inhibitor zVAD-fmk (Fig. 2D and Supplementary Fig. S1E). Therefore, Beclin 1 is likely to be cleaved by a caspase(s) that is activated through the intrinsic apoptotic pathway in response to anticancer agents.
Cleavage of Beclin 1 at 2 sites by active caspase 8
To determine if and which caspase(s) can cleave Beclin 1, in vitro translated Beclin 1 was incubated with purified active caspases 3, 8, or 9. Beclin 1 could be strongly cleaved by active caspase 8, weakly by active caspase 3 but not by caspase 9 (Fig. 3A). This result was confirmed by using GST-tagged Beclin 1 (Supplementary Fig. S2A). Only one fragment was detected in the cell-free assays, probably due to partial activities of purified caspases. To test whether caspase 8 was responsible for CPT-induced Beclin 1 cleavage, caspase 8 was knocked down by short hairpin RNA in HCT116 cells (Supplementary Fig. S2B). Cleavage of both endogenous and transfected Beclin 1 was abrogated in caspase 8–knockdown (Casp 8-KD) cells treated with CPT or STS (Fig. 3B, left and Supplementary Fig. S2C and D). Similar to Cyt c-KI cells, Casp 8-KD cells were more prone to autophagy induction than WT HCT116 in response to CPT treatment, as indicated by enhanced LC3II accumulation, p62 degradation (Fig. 3B, right), and GFP-LC3 puncta signals (Supplementary Fig. S2E). These data indicate that caspase 8 is the major mediator of Beclin 1 cleavage in chemotherapy-induced apoptosis.
We then examined potential caspase cleavage sites of Beclin 1, focusing on residues 111 to 150 on the basis of the size of the cleavage fragments (Fig. 3C). Seven aspartic acids (D), each of which could be a potential caspase cleavage site (30), were noted. A battery of Beclin 1 mutants, including deletions and alanine (A) replacements (Fig. 3C), were generated. Upon transfecting these mutants and treating cells with CPT, we found that deletion of the residues 121 to 149 (D3) completely blocked Beclin 1 cleavage, whereas the microdeletion of the residues 146 to 149 (D2) and 2 point mutants, D133A (M4) and D146A (M6), each partially blocked Beclin 1 cleavage (Supplementary Fig. S3A and B), suggesting that both D133 and D146 contribute to Beclin 1 cleavage. Indeed, a DM with both sites altered (D133A/D146A) was completely resistant to cleavage by caspase 8 or 3 in cell-free assays (Fig. 3A and Supplementary Fig. S3C) or in CPT-treated HCT116 cells (Fig. 3D, left). Therefore, D133 and D146 of Beclin 1 are the 2 sites that are cleaved by caspases during apoptosis.
Specifically blocking Beclin 1 cleavage promotes autophagy in apoptotic cells
We then compared the autophagy-promoting activities of WT and DM Beclin 1, which differ only in their susceptibility to caspases, in cells undergoing apoptosis. Transfection of WT Beclin 1 significantly enhanced LC3II levels in untreated HCT116 cells but was unable to do so in CPT-treated cells (Fig. 3D, right), reflecting inhibition of autophagy by caspase-mediated Beclin 1 cleavage. In contrast, transfection of DM Beclin 1 not only increased LC3II levels in untreated HCT116 cells but also further enhanced LC3II accumulation and p62 degradation in CPT-treated cells (Fig. 3D, right). Therefore, specific blockage of caspase 8–mediated cleavage is sufficient to restore the autophagy-promoting function of Beclin 1 in cells that are undergoing apoptosis.
To further determine the effects of Beclin 1 cleavage by caspase 8, N- and C-terminal cleavage fragments of Beclin 1 were constructed (Fig. 3C) and transfected into HCT116 cells, followed by CPT treatment. Both Beclin 1 cleavage fragments lose much of the autophagy-promoting activity compared with the DM in CPT-treated cells (Supplementary Fig. S4A). Similar to WT Beclin 1, the transfected cleavage fragments did not affect CPT-induced apoptosis (Supplementary Fig. S4B) and were exclusively cytosolic in both untreated and CPT-treated cells (Supplementary Fig. S4C).
Blocking Beclin 1 cleavage in vivo induces autophagy in apoptotic cells
To determine the effects of blocking Beclin 1 cleavage in vivo, we replaced the WT Beclin 1 locus in HCT116 cells with the caspase-resistant DM by homologous recombination (Fig. 4A). After 2 rounds of gene targeting, several desired recombinant clones were identified by genomic PCR analysis of drug-resistant clones. DNA sequencing revealed that the DM sequence was correctly targeted into both alleles of the Beclin 1 genomic locus (Fig. 4B). Mutant Beclin 1 was expressed at a similar level as WT protein (Fig. 4C). CPT-induced Beclin 1 cleavage observed in WT cells was abolished in Beclin 1-KI cells (Fig. 4D), indicating that endogenous Beclin 1 was cleaved at D133 and D146.
Similar to Cyt c-KI and Casp 8-KD cells, Beclin 1-KI cells were found to have higher levels of LC3II accumulation and p62 degradation than WT cells after CPT treatment (Fig. 5A, left). The increase in LC3II levels was due to enhanced autophagosome formation (Supplementary Fig. S5A). Furthermore, GFP-LC3 puncta formation was significantly enhanced by 2- to 3-fold in CPT-treated Beclin 1-KI cells relative to WT cells (Fig. 5B). Transmission electron microscopy detected frequent autophagosome and autolysosome structures in CPT-treated Beclin 1-KI cells, whereas these structures were rarely found in WT cells (Fig. 5A, right). Autophagy was also enhanced in Beclin 1-KI cells in response to other treatments including serum starvation, exposure to the mTOR inhibitor rapamycin, and amino acid starvation (Supplementary Fig. S5B). Therefore, specifically blocking the cleavage of Beclin 1 in vivo recapitulated autophagy induction in apoptosis-defective cells including Cyt c-KI (Fig. 1B and D), Casp 8-KD (Fig. 3B), and BAX-KO cells (31).
Blocking Beclin 1 cleavage does not affect apoptosis but results in chemoresistance
WT and Beclin 1-KI HCT116 cells were then compared for apoptosis induction in response to several stimuli including CPT, STS, PUMA expression, or sulindac sulfide (SUL), a nonsteroidal anti-inflammatory drug used in cancer chemoprevention (32). Analysis of apoptosis by nuclear staining, Annexin V staining, or caspase activation showed that blocking Beclin 1 cleavage in vivo did not affect the execution of apoptosis induced by these stimuli (Fig. 5C and data not shown).
We then determined whether blocking Beclin 1 cleavage affects response to chemotherapy using colony formation assay. Following CPT treatment, Beclin 1-KI cells formed significantly more colonies compared with WT cells (Fig. 5D), suggesting that suppression of Beclin 1 cleavage is sufficient to protect against cell death. Therefore, Beclin 1 mutant knockin, which specifically ablates caspase-mediated Beclin 1 cleavage, recapitulated protection against cell death seen in Cyt c-KI (Fig. 2B), and BAX-KO cells (29).
To investigate the effect of blocking Beclin 1 cleavage on chemotherapeutic response in vivo, tumors were established using WT and Beclin 1-KI HCT116 cells in nude mice, followed by CPT-11 treatment (24). Beclin 1-KI tumors were found to be significantly less responsive to CPT-11 treatment compared with WT tumors (Fig. 6A). The degradation of p62 was enhanced in Beclin 1-KI tumors compared with WT tumors (Fig. 6B). TUNEL (terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling) staining revealed that the levels of apoptosis induction were similar in WT and Beclin 1-KI tumors following CPT-11 treatment (Fig. 6C). These observations suggest that autophagy induction in vivo as a result of blocking Beclin 1 cleavage can protect tumors from chemotherapy-induced cell death.
Our results show that chemotherapy-induced apoptosis suppresses autophagy at the execution stage following cytochrome c release, at least in part, through caspase 8–mediated Beclin 1 cleavage at D133 and D146. Caspase 8 not only plays a key role in death receptor–mediated apoptosis but also contributes to the execution of DNA damage–induced apoptosis through the intrinsic pathway (22). Specifically blocking Beclin 1 cleavage by knocking in a caspase-resistant mutant of Beclin 1 recapitulated autophagy induction seen in apoptosis-deficient cells. On the other hand, autophagy induction in Beclin 1-KI cells did not affect apoptotic events but protected against cell death. These results provide mechanistic insight into how chemotherapy-induced apoptosis suppresses autophagy and explain why autophagy is generally observed in cells in which caspase activation is blocked.
Several recent studies using different cell types and stimuli also described caspase-mediated cleavage of ATG proteins in apoptosis (15). Cho and colleagues found caspase-dependent cleavage of Beclin 1 in HeLa cells treated with a death receptor ligand (33). Luo and Rubinsztein reported caspase 3–mediated Beclin 1 cleavage at D149 in apoptosis induced by Bax overexpression (34). Wirawan and colleagues showed caspase cleavage of Beclin 1 at D133 and D149 during apoptosis induced by IL-3 depletion in murine hematopoietic cells (35). ATG5 and ATG4D were also found to be cleaved by caspases in apoptotic cells (14, 36). Furthermore, caspase 8 has been implicated in autophagy suppression in mouse L929 fibroblast and human U937 monocytoid cells (37, 38) and also in proliferating T cells (39). Our knockin data indicate that specific blockage of endogenous Beclin 1 cleavage alone is sufficient to restore autophagy in apoptotic cells, at least in colon cancer cells undergoing chemotherapy-induced apoptosis, and possibly in other scenarios as well. Differences in cell types and apoptotic stimuli may explain why D146 was required for Beclin 1 cleavage in our study, but D149 was necessary for Beclin 1 cleavage in other studies.
Autophagy occurring subsequent to cytochrome c release is likely to be trigged by mitochondrial outer membrane permeabilization (MOMP) and is therefore mitophagy, a recycle process in which mitochondria are captured and degraded (40). Although Beclin 1 can interact with antiapoptotic Bcl-2 family members through its BH3 domain, it does not seem to significantly affect apoptosis initiation and caspase activation. Nonetheless, Beclin 1–mediated autophagy may modulate chemotherapeutic response by removing terminally damaged mitochondria with MOMP, which may help to sustain energy metabolism of cancer cells and allow them to tolerate and recover from the damages induced by chemotherapeutic agents. It has been shown that mitophagy mediated by ATG12 and metabolic changes mediated by glyceraldehyde-3-phosphate dehydrogenase (GAPDH) preserve the survival of STS-treated cells in which cytochrome c release had occurred but caspase activation was blocked (41). Therefore, autophagy may not simply be an innocent bystander (17) but rather an important cell survival pathway that has to be turned off for full execution of cell death.
A remaining question is how Beclin 1 cleavage during apoptosis affects its functions. Beclin 1 cleavage may not abolish its interaction with VPS34, which is mediated by a conserved region far apart from its cleavage sites (42). The cleavage of Beclin 1 also leaves its BH3 domain intact. Only a fraction of Beclin 1, as detected by Western blotting, was subject to caspase cleavage in CPT-treated cells, which seemed to suggest a gain of function by Beclin 1 cleavage fragments. However, both N- and C-terminal Beclin 1 cleavage fragments remained localized in the cytosol, and their overexpression did not significantly affect apoptosis, consistent with intact apoptosis induction in Beclin 1-KI cells. Therefore, caspase cleavage of Beclin 1 may serve primarily as an inhibitory mechanism of autophagy and cell survival.
Autophagy is a survival pathway utilized by cancer cells to tolerate metabolic stress. Most, if not all, anticancer agents can induce autophagy in cancer cells (16). Autophagy inhibitors, such as hydroxychloroquine, are being tested in a number of clinical trials as chemosensitizers (16). Our results suggest that Beclin 1, which is cleaved by caspases for switching off autophagy in chemotherapy-induced apoptosis, may be a useful target for more specific and effective pharmacologic inhibition of autophagy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
The authors thank Drs. Monica E. Buchanan, Crissy Dudgeon, and Brian Leibowitz for critical reading and comments; and Xiaohong Wang for technical assistance.
This work is supported by NIH grants CA106348, CA121105, American Cancer Society grant RSG-07-156-01-CNE, and the American Lung Association/CHEST Foundation (L. Zhang), NIH grants CA129829, U01DK085570, and American Cancer Society grant RSG-10-124-10-CCE (J. Yu), NIH grants CA132385, GM087798, and CA148629 (R.W. Sobol), and the UPCI Cancer Center Support grant (CA047904).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.