Paradoxical Effects of Antioxidants on Cancer

Introduction

There is a tendency for researchers to view specific biomolecules and drugs as health promoting, health neutral, or health damaging. Unfortunately this perspective is a gross simplification. For example, treatments that initially may help prevent or slow cancer can actually promote the creation or selection of more aggressive cancer cells at later stages of the illness.

Antioxidants are a particularly apt example of a class of substances that are thought to be beneficial. antioxidants are believed by many to slow aging or the diseases associated with aging, such as cancer. It has been widely believed that antioxidants could block reactive oxygen species (ROS)-mediated DNA damage that may contribute to cancer initiation or progression. However, there is limited evidence to support this idea. Studies have shown that antioxidants have, at best, no effect. β-Carotene was actually associated with a small increased risk of lung cancer. ¹ At the molecular level, it has been known that ectopic over-expression of oncogenic proteins, such as RASV12 mutants, causes increased ROS levels. Increased ROS levels were thought to not only increase a transformed cell's rate of mutation, but also to drive proliferation. However, at least one group reports that oncogenic RAS only increases ROS levels when over-expressed, and actually decreases ROS when expressed at physiological levels. A large body of work has accumulated suggesting that in the correct context, significant ROS levels actually protect animals from cancer and conversely antioxidants support cancer growth (see ref. 28).

antioxidants Can Promote Tumorigenesis

In an enlightening paper, Sayin et al. ² show that treatment with antioxidants can actually accelerate lung cancer progression in mice. Their results stand in contrast to some earlier experiments suggesting that high ROS levels play a key role in tumorigenesis. For example, the anti-oxidant N-acetylcysteine (NAC) prevented lymphomas in mice in which the genome integrity master regulator p53 had been knocked out. ³ By contrast, other data show that stimulation of ROS levels by pro-oxidants or agents that stimulate ROS, such as elesclomol, inhibit or kill tumors. ⁴ Expression of the activated oncogenes KRAS, B-RAF, or Myc at physiological levels increases expression of the anti-oxidant master regulator Nrf2, which lowers ROS levels. Reduction of Nrf2 inhibits KRAS-induced proliferation and tumorigenesis in mice. ⁵

Savin et al. used mice that were genetically engineered to express the Kras^G12D oncogene when infected with adenovirus expressing the site-specific recombinase Cre by rearranging the DNA so that the Kras^2G12D coding region was juxtaposed with a transcriptional promoter. Within 1 week, mice developed a range of tumors, from mild epithelial hyperplasia to aggressive adenocarcinoma. Treatment with water-soluble NAC or fat-soluble vitamin E resulted in ∼2.8-fold higher tumor burden (% tumor area per lung area) than untreated controls. To determine whether the pro-cancer effect of the antioxidants was limited to tumors triggered by mutant KRAS, a similar experiment was performed in mice with a Cre-inducible B-RAF^V600E mutation. B-RAF^V600E mice developed a greater number of tumors than the Kras^G12D mice, although the tumors were less aggressive. Treatment with NAC or vitamin E increased tumor burden ∼3.4-fold in the B-RAF^V600E mice and survival time was reduced 50%–60%. ² Keep in mind that the KRAS and BRAF signaling pathways are linked and are known to interact with each other in a variety of epithelial-derived cancers. Therefore, these models are not as independent as the authors suggest.

As expected, treatment with NAC or vitamin E reduced ROS in the lungs of the Kras^G12D and B-RAF^V600E mice, as determined by the fluorescent redox-sensitive probe CM-H₂₂DCFDA. Of interest is that the ROS levels in the lungs were already reduced by about 30% in the mutant mice compared to wild-type control mice, an example of decreased oxidative stress associated with tumorigenesis. Consistent with the reduced ROS, increased ratios of reduced to oxidized glutathione were observed. Reduced amounts of DNA containing 8-oxoguanine, which typically results from oxidative damage, meant that anti-oxidant–treated Kras^G12D and B-RAF^V600E mice had less DNA damage in their lung cells than untreated animals. Interestingly, wild-type animals had more DNA damage in their lung cells than either of the oncogene mutant mice. In tumor extracts, levels of phosphorylated H2AXSer139 and ataxia telangiectasia model (ATM)Ser1981, which correlate with DNA damage, were also decreased after anti-oxidant treatment. Of particular interest is that cell proliferation increased in the lungs of both strains of mutant mice treated with the antioxidants, as assessed by staining for bromodeoxyuridine (BRDU) incorporation (DNA replication) or phosphorylated histone H3 levels (mitosis). Removing antioxidants 1 week before assaying proliferation reduced DNA proliferation levels to close to those seen in the untreated mutant mice. In primary fibroblasts cultured from the mice expressing either activated RAS or BRAF, treatment with NAC or a water-soluble vitamin E (Trolox) resulted in increased cell proliferation with no change in apoptosis. ² In summary, anti-oxidant treatment resulted in increased proliferation despite reduced ROS and DNA damage levels.

The authors hypothesized that the activity of p53, a key regulator of genomic integrity and cell proliferation, which is known to be activated by ROS, might be reduced. Indeed, p53 protein levels were reduced in extracts from cultured fibroblasts or lung tumors treated with antioxidants. To more rigorously test their hypothesis, mice with either Cre-activatible Kras^G12D or B-RAF^V600E were crossed with mice with Cre-conditional p53 knockout alleles.

After simultaneously activating either KRAS or BRAF and inactivating p53, the antioxidants (NAC or Trolox) no longer affected the degree of cell proliferation in culture or in vivo. In a preliminary attempt to demonstrate that p53 plays a similar role in human tumors, antioxidants were observed to increased proliferation in p53⁺ human lung cancer cell lines. When p53 expression was knocked down using small hairpin RNA (shRNA), antioxidants no longer had any effect on cell proliferation. These experiments link p53 to the tumor-promoting activity of antioxidants and support a simple model by the authors that reduction of oxidative stress reduces DNA damage, in turn reducing p53. With decreased p53 surveillance, tumor cell proliferation increases. ² Although DNA damage was reduced, it would be interesting to examine genomic stability in these lung tumors, because lack of p53 is known to result in increased risk of chromosomal instability.

Treatment with “Anti-Oxidant” Ascorbate Sensitizes Tumors to Chemotherapy

Ascorbate, also called vitamin C, is an anti-oxidant with a long history of claims touting health benefits, among them fighting cancer. Nobel Prize–winning chemist Linus Pauling and his medical colleague Ewan Cameron reported anti-cancer effects of high doses of ascorbate delivered both intravenously or orally. ^6,7 However, larger clinical trials with oral ascorbate failed to confirm the earlier studies. ^8,9 Later, there were several reported case histories describing the potential benefit of high doses of intravenous vitamin C in patients with terminal cancer diagnoses. ^10,11 Unfortunately, uncontrolled case reports, while interesting, are of limited utility in determining bona fide efficacy.

Fortuitously, Mark Levine's group recently reported that ascorbate kills cancer cells in vitro and in mice by acting as a pro-oxidant through generation of hydrogen peroxide (H₂O₂). ¹² In a recent phase I/IIa clinical trial that was not designed to show efficacy, high-dose intravenous ascorbate treatment reduced toxicity associated with chemotherapy. Ma et al. ¹² hypothesized that failures of vitamin C in the large clinical trials in the late 1970s were caused by the route of delivery. It turns out that oral supplementation of ascorbate cannot achieve plasma concentrations greater than 200 μM due to limited adsorption, transport, and excretion by the kidneys. On the other hand, intravenous delivery of vitamin C can achieve a peak 10 mM plasma concentration in humans for several hours. This is 50-fold higher than oral delivery. ^13,14 In several earlier papers, Levine and colleagues showed that injection of sufficient ascorbate to attain millimolar plasma levels leads to the formation of ascorbate radicals and H₂O₂ in extracellular spaces, but not in whole blood of rats and mice. ^{15 –17} Ascorbate levels of 10 mM appear to have selective toxicity for several different kinds of cancer cell lines in culture and slow the growth of xenografts of human glioblastoma and ovarian and pancreatic cancer in mice. ¹⁷

Ma et al. showed that seven cultured ovarian cell lines were sensitive to 0.3–3.5 mM ascorbate (at least 99% loss of viability), while HIO-80, an immortal non-tumorigenic ovarian epithelial cell line, was nearly insensitive (about 25% loss of viability at 3.5 mM ascorbate, and almost no loss of viability at lower concentrations). Ascorbate-mediated production of H₂O₂ is necessary for cell death, as addition of peroxide-scavenging enzyme catalase abrogated cytotoxicity. In at least one ovarian cancer cell line (SHIN3), ascorbate induced significant DNA damage as assessed by increased levels of phosphorylated histone H2AX, which binds to DNA with double-stranded breaks, and the Comet assay, a electrophoresis-based method to detect fragmented DNA. ¹² Presumably, DNA damage is due directly or indirectly to the H₂O₂.

DNA damage could be increased in an ovarian cell line by treatment with both ascorbate and DNA alkylating agent carboplatin. An even greater amount of DNA damage could be induced by using olaparib to block DNA repair via inhibition of poly(ADP ribose) polymerase (PARP), an enzyme that plays a key role in repair of single-stranded DNA breaks. Adding catalase prevented DNA damage by ascorbate, demonstrating that H₂O₂ is necessary for ascorbate-mediated cancer killing. Alone carboplatin or olaparib induced only minor DNA damage at the same experimental concentrations where they cooperated with ascorbate to kill cancer cells effectively. ¹²

Although the results from the Comet assay might suggest that tumor cells die via an apoptotic process, previous work suggests that caspases are not involved and death is by necrosis and is associated with increased autophagy. ^{18 –20} Although Ma et al. did not assess caspase activation in this latest work, they did observe a sharp (60%) drop in adenosine triphosphate (ATP) levels in the tumor cell line compared to a smaller drop (20%) in the non-transformed ovarian cell line, which is consistent with autophagy playing a role in the tumor-specific death. The authors speculate that the tendency of tumor cells to rely on glycolysis for ATP production, i.e., the Warburg effect, may sensitize the tumor cells to ascorbate by reducing ATP production. Glycolysis is far less efficient at ATP production than oxidative phosphorylation. Low ATP levels can cause metabolic stress, and indeed biochemical analysis of key stress and growth regulatory factors ATM, 5′ AMP-activated protein kinase (AMPK), and mammalian target of rapamycin (mTOR) suggest that the cancer cell line, but not the normal cell line, is metabolically stressed. Ascorbate treatment results in phosphorylation and activation of ATM within 15 min. Downstream AMPK is phosphorylated and thereby activated by ATM, which in turn results in decreased expression of mTOR and phosphorylated mTOR, possibly through activation of tuberous sclerosis 2 (TSC2). ²¹ Decreased expression of mTOR is known to increase autophagy. ²²

It should be pointed out that there are significant unknowns in the proposed mechanism of action for high ascorbate acting as a pro-oxidant. The most glaring problem is the lack of identification of a specific mechanism by which high levels of ascorbate lead to formation of the ascorbate radicals and in turn produce H₂O₂ in cultured cells and in the extracellular fluid. It is important to understand that intracellular levels of ascorbate are tightly controlled by transporters, so the generation of free radicals takes place outside of the cells. The absence of ascorbate radicals and peroxide in the blood is likely due to the presence of reducing enzymes in red blood cells. Furthermore, the mechanism of cell death is somewhat murky, because the relative roles of DNA damage and ATP depletion have not been determined. For example, Ma et al. did not try to increase intracellular ATP during ascorbate treatment to untangle the relative contributions of these effects. Also, the percentage of cells with DNA damage in culture was only about 30% with ascorbate, and 40% with the combination of ascorbate, olaparib, and carboplatin.

A drug combination study was performed in culture using constant ratio design analysis, ²³ which allows determination of synergism, additivity, or antagonism between multiple drugs at constant ratios of drugs. In two ovarian cell lines (OVCAR5 and SHIN3), an additive to the synergistic effect was shown for ascorbate and carboplatin at all combination ratios, and at a high ascorbate ratio in the third cell line. Carboplatin did affect normal cells, but independently of ascorbate. Together, this suggests that adding ascorbate to carboplatin allows reduction of carboplatin doses to achieve a similar amount of killing.

In a xenograft model, ascorbate reduced tumor burden. The combination of ascorbate and carboplatin reduced tumor burden more than either alone. The combination of ascorbate and paclitaxel, an anti-cancer drug that targets microtubules, also reduced tumor burden more than the individual compounds. The best combination was ascorbate/paclitaxel/carboplatin, which reduced tumor burden by 94%. Paclitaxel and carboplatin are frequently used in combination as standard therapy, so if ascorbate had similar effects in humans, it could augment current standard therapy. At the doses used in this study, none of the drug combinations, nor ascorbate alone, demonstrated any toxicity to the liver, kidney or spleen. ¹²

In a phase I/IIa clinical trial lasting 5 years, 12 patients received paclitaxel and carboplatin and 13 patients received that combination with intravenous ascorbate. Toxicity was graded on a scale from 1 to 5, where 5 inidcated death. The addition of ascorbate reduced grade 1 and 2 toxicities. There was a statistically insignificant trend toward improvement of survival, with mean time of progression increased by 8.5 months. However, this study was too statistically underpowered to truly test for efficacy. The authors suggest that a larger clinical trial designed to test efficacy be undertaken, and we believe that would be a good idea, especially in the context of reducing toxicity of conventional chemotherapy, the mechanism of which remains unidentified. ¹² Nevertheless, every clinical trial using intravenous vitamin C so far has failed to show significant clinical effect compared to preclinical rodent models. ^24,25 There are various potential problems—the short half-life of ascorbate in the body, development of resistance by cancer cells, and differences between rodents and people.

Medical Implications

There is no evidence demonstrating that anti-oxidant supplementation extends life span or prevents cancer, at least in studies of large populations (Bjalekvic 2009). In fact, there are a few studies suggesting that some antioxidants are associated with a small increased risk for cancer.

That vitamin E or NAC could spur lung tumor progression, even in an animal model, raises an alarm bell. That vitamin C may have anti-cancer activity, not as an anti-oxidant but rather as a pro-oxidant at pharmacological concentrations in extracellular fluids, does not diminish the possibility that antioxidants could play a malignant role in some cancers. ²⁸ However, it is important to recognize that the anti-oxidant lung cancer story is based only on one study in murine models. Only two antioxidants were examined; it would be useful to confirm the results using other antioxidants. Unfortunately, direct confirmation in humans will be difficult and will require an indirect epidemiological approach.

More importantly, these studies shed light on the potential paradoxical behavior of drugs and genes in cancer cells and the difficult problem faced in developing conventional therapeutics to fight cancer. For cancer cells, biological context becomes supremely important. In some situations, genes that are known to be tumor suppressors can actually promote tumor growth. Nrf2, a master regulator of the oxidative stress protection program, is a tumor suppressor gene that controls a set of anti-oxidant enzymes called “phase II detoxifying enzymes” that protect cells from ROS and may play a role in protecting cells from aging-associated oxidative damage. For example, these Nrf2-stimulated enzymes are known to inhibit the action of carcinogens like benzopyrene that can cause stomach cancer. Mice lacking Nrf2 have a greater incidence of cancer after treatment with benzopyrene (stomach cancer) ²⁹ or N-nitrosobutyl(4-hydroxybutyl)amine (BBN) (bladder cancer). ³⁰ Agents that protect mice from such cancers, such as oltipraz and sulforaphane, require Nrf2 to function. So it appears Nrf2 expression is beneficial; however, alternately Nrf2 is strongly activated in many cancers and actually plays a critical role in their tumorigenicity. Expression of physiological levels of oncogenic forms of KRAS, B-RAF, or Myc induces Nrf2, resulting in lowered ROS levels. ⁵ In a mouse model of pancreatic cancer, genetic inactivation of the Nrf2 pathway inhibits K-Ras^G12D-induced proliferation and tumorigenesis in vivo. In pancreatic cancer, Nrf2 has the potential to promote cancer. ⁵ It has been hypothesized that increased levels of phase II detoxifying enzymes may help confer resistance to chemotherapy associated with pancreatic cancer. So context may be of great importance. Even the ultimate tumor suppressor p53 may sometimes act to help cancer cells resist treatment. ³¹ Paradoxically, wild-type p53 is sometimes associated with a worse outcome in breast cancer. ³² Some breast cancer cells carrying wild-type p53 temporarily stop dividing after treatment with DNA-damaging drugs, such as doxorubicin, protecting them from death via mitotic catastrophe. ³³ In lung cancer, wild-type p53 can help protect cancer cells from metabolic stress induced by inhibition of glycolysis by 2-deoxyglucose. ³⁴ Paradoxically, p53 protects some cancer cells by making them somewhat more normal, allowing them to evade therapy.

Analogous to Nrf2, it is quite possible that antioxidants may suppress early stages of cancer initiation, but then promote tumor growth after cancer cells appear, at least in some situations. On the other hand, the presence of exogenous antioxidants has been reported to decrease ROS after exercise. ³⁵ Because decreased ROS down-regulates stress-protective pathways, Nrf2 expression is likely reduced. In the case of antioxidants, it is possible that we have the worst of both worlds—less oxidative damage protection from the potentially more effective phase II detoxifying enzymes with subsequent tumor promotion due to p53 down-regulation.

There are also examples where a beneficial anti-cancer drug can increase tumorigenesis. For example, the BRAF^V600E inhibitors vemurafenib and dabrafenib have been successful in treating melanoma bearing BRAF^V600 mutations. However, these drugs have the ability to paradoxically activate the MAPK pathway through dimerization of wild-type BRAF and there are reports of secondary malignancies in other tissues such as the colon resulting from the treatment. ³⁶ Even the anti-diabetic drug metformin, which is associated with a 40% decreased incidence of cancer in diabetic patients and has significant anti-cancer activity in a variety of preclinical systems, may act paradoxically on some cancers. For example, metformin has been reported to promote new blood vessel formation in the estrogen receptor-α (ERα)-negative MDA-MB-435 breast cancer model ³⁷ and increase vascular endothelial growth factor-A (VEGF-A) in several human and mouse melanoma cell lines. ³⁸

The bottom line is that distinguishing between beneficial and detrimental agents to prevent cancer may be a difficult task that will strongly depend on genetic predilections. Personal genomic analysis and personalized cancer genome sequencing may provide the means to obtain context to avoid paradoxes and obtain a successful outcome. However, with these recent data, rational cancer prevention has become more difficult.

Conclusions

Antioxidants have shown little ability to protect people from cancer. Recent work suggests that antioxidants may even support tumorigenesis in a mouse model of lung cancer. However, much work needs to be done to substantiate and generalize these results. The paradox that potentially helpful agents that prevent damage could actually be harmful in some contexts likely applies to many drugs and specific biomolecules/cellular regulators. Perhaps it is time that we recognize that biomolecules and therapeutics are in some sense beyond good and evil.