Introduction

Some notions in cancer biology and oncology may seem conflicting. For example, avoidance of apoptosis is a hallmark of cancer.1, 2, 3, 4, 5, 6 On the other hand, it is often presumed that anticancer drugs should induce apoptosis preferentially in cancer cells. Even inhibitors of universally vital targets such as of histone deacetylases and the proteasome are asserted to be selectively toxic to cancer cells. On the one hand, it is well recognized that genetic instability (e.g., through defects in DNA repair) predisposes to cancer and accelerates tumor progression.7, 8 On the other hand, inhibition of DNA repair assists in treating cancer.9, 10

Although carcinogens and anticancer drugs are viewed as opposites, the two groups of agents highly overlap (Figure 1):

  1. a

    Alkylating agents are classic carcinogens (methylnitrosourea, 7,12-dimethylbenzanthracene, diethylnitrosoamine) and also they are classic anticancer drugs (melphalan, chlorambucil, busulfan, cyclophosphamide, the nitrosoureas, cisplatinum and carboplatin).11, 12, 13, 14

  2. b

    Radiation is both a carcinogenic agent and an effective anticancer therapy. As a striking example, radioactive iodine causes thyroid cancer15 and is the most effective therapy for thyroid cancers that retain the ability to accumulate iodine.16

  3. c

    Arsenic, a classic cytotoxic agent and carcinogen, recently was re-discovered as one of the most promising anti-leukemia and anticancer drugs.17, 18

  4. d

    Tamoxifen, an antiestrogen, which is effective in therapy of breast cancer, can cause endometrial cancer19, 20 and malignant mixed mullerian tumors.21

  5. e

    For 25 years, it was well known that phorbol ester (PMA/TPA) inhibits proliferation of leukemia cells. Yet, for many years, TPA was not considered for clinical use, because TPA is carcinogenic. In pioneering clinical trials (initiated by Allan Conney), TPA caused clinical remissions in patients who were refractory to chemotherapy.22, 23, 24 Combinations of TPA with other experimental therapeutics are under preclinical investigations.25 Similarly, anticancer drugs, especially topoisomerase inhibitors, cause secondary cancers.26, 27, 28, 29 To complicate matters further, chemopreventive agents may promote cancer in some conditions.30 In certain mice, sulindac, a nonsteroidal anti-inflammatory drug, prevents tumors in the small intestine but increases the incidence of tumors in the large intestine.31 Retinoic acid prevents papillomas but accelerates progression from papillomas to carcinomas.32 Tamoxifen can prevent breast cancer,33 while increasing the risk of endometrial cancer.34 In a chemoprevention trial, beta-carotene was unexpectedly found to increase both the risk of lung cancer among high-risk groups35 and colorectal adenomas.36 Genistein, a chemopreventive agent,37 increases carcinogen-induced mammary tumorigenesis in female rat offspring.38 In women, even such chemopreventive agents as aspirin can increase the risk of pancreatic cancer.39 Selenium, a chemopreventive agent, can increase squamous cell carcinomas.40

Figure 1
figure 1

Anticancer drugs, carcinogens and chemopreventive agents overlap. The same agent (arsenic, TPA, alkylating agents, radiation, topoisomerase inhibitors) can cause and treat cancer. Similarly, chemopreventive agents can cause cancer in one setting and prevent it in other setting. Some genes (e.g., E2F) are both oncogenes and tumor suppressors. Certain hormones can prevent, cause and treat cancers

Full size image

To reconcile these seemingly conflicting data, this review links tumor progression to cancer therapy to chemoprevention. Since each field is enormously complex by itself, this review cannot discuss them in detail. Instead, it is focused on common underlying principles such as selection for resistance to cytotoxic agents. The notion that carcinogenesis is associated with acquiring resistance to cytotoxic carcinogens is not particularly novel.41 Yet, these ideas were viewed as an alternative to the somatic mutation theory. They were not integrated in the concept that cancer results from accumulation of genetic and epigenetic alterations in oncogenes and tumor suppressors.42 This article combines both concepts. We distinguish oncogenic resistance (due to oncogenic mutations that alter cell cycle and apoptotic pathways) and nononcogenic resistance (e.g., detoxifying enzymes, drug pumps43) and envision that the cytotoxic activity of carcinogens is the driving force for ‘accumulation of mutations in oncogenes and tumor suppressors’. (For brevity, the term ‘oncogenic mutations’ includes both genetic and epigenetic alterations.) Similarly, by inhibiting proliferation and causing apoptosis, cytotoxic anticancer drugs and chemopreventive agents can select for oncogenic resistance, thus driving tumor progression. While therapy-induced killing of cancer cells results in therapeutic response and cancer remission, preferential killing of drug-sensitive cancer cells simultaneously selects for resistance that may be accompanied by tumor progression.

Two Types of Resistance to Cytotoxic Agents

Resistance to cytotoxic agents can be divided into nononcogenic and oncogenic. Nononcogenic resistance includes expression of drug transporters, metabolism of cytotoxic agents and mutations of targets. The resistance is due to prevention and repair of the damage caused by cytotoxic agents. For example, the enzyme O6-methylguanine-DNA methyltransferase (MGMT) repairs alkylation of DNA caused by alkylating agents.10 Overexpression of MGMT reduces the risk of carcinogenesis.44, 45 Therefore, MGMT-deficient cells may preferentially undergo transformation. In turn, low levels of MGMT render cancer cells vulnerable to DNA-damaging drugs and can be exploited for therapy.46 The line between nononcogenic and oncogenic resistance can be blurred. For example, amplification of drug target, if the target is an oncogenic kinase, can confirm both nononcogenic and oncogenic resistance. In oncogenic resistance, a cell may ignore the damage. Although cytotoxic agent interacts with its target, downstream pathways of apoptosis and/or growth arrest are blocked (oncogenic resistance). For instance, despite DNA damage, a cell that overexpresses oncoproteins (e.g., ErbB) or lacks tumor suppressors (p53, Rb, p16) continues to proliferate.42, 47, 48 Inhibitors of apoptosis (cIAP-1, cIAP-2, XIAP, survivin) and oncoproteins (Bcl-2, BclxL, Ras, MEK, Akt, Bcr-Abl) and especially the Akt–TOR pathway can block apoptosis, rendering cancer cells resistant to chemotherapy.1, 42, 49, 50, 51, 52, 53, 54, 55 In other words, oncogenic resistance results from alteration in regulation of cell cycle and apoptosis and is associated with the resistance to physiological inhibitors. The main feature of oncogenic resistance is oncogenic translation of ambivalent signaling: namely, due to mutations in oncogenes and tumor suppressors, ambivalent signals (which can cause either growth arrest, proliferation, apoptosis or differentiation) are translated as mitogenic and/or antiapoptotic.

Disabled apoptosis can be linked to genomic instabilities, which are engaged in a positive amplification loop.56 During tumor progression, cancer cells accumulate alterations (genetic and epigenetic) in oncogenes and tumor suppressors.57 Genetic instability and mutagenic agents provide mutation repertoire.7, 8 But what is the driving force for accumulation of mutations? Conditions (agents) that restrict proliferation and survival of cells lacking mutations select for adaptive mutations that confirm resistance to these agents (Figure 2). Some carcinogens (e.g., radiation) also cause mutations, thus accelerating selection for carcinogen resistance. But many carcinogens and tumor promoters (nickel, arsenic, TPA) and carcinogenic environments (high cell density, hypoxia, inflammation, TGF-β) are not mutagenic. What most carcinogens and tumor promoters have in common is cytotoxicity against certain normal cells, thus selecting for oncogenic mutations that confirm resistance (Figure 2). Similarly, by killing drug-sensitive cells, anticancer drugs select for cells with mutations that confirm resistance. In other words, the driving force is the cytotoxicity of anticancer drugs. Although drug resistance is traditionally divided into intrinsic and acquired, both types of resistance are actually acquired (Figure 3). Intrinsic resistance is acquired during carcinogenesis and before therapy.

Figure 2
figure 2

Selection for resistance in carcinogenesis and cancer therapy. Cytotoxic agent inhibits proliferation and/or survival of sensitive cells. Cytotoxic carcinoagent (carcinogen or anticancer drug) selects for resistant cells with mutations (m). This is the selection step (from sensitivity to resistance). Multiple steps can follow, as shown in Figure 3

Full size image
Figure 3
figure 3

Multistep selection driven by carcinoagents (cytotoxic carcinogens and anticancer drugs). Growth-limiting conditions such as cytotoxic carcinogens can select for resistant (m) cells. A multistep selection both limits and drives carcinogenesis.8, 57 Carcinogens and tumor promoters inhibit proliferation and/or survival of normal cells selecting for oncogenic mutations (m, mm). Accumulation of mutations results in cancer. Anticancer drugs further promote tumor progression. Under cytotoxic pressure, cell population can either become extinct (effective cancer therapy) or acquire further resistance due to mutations (tumor progression). From an initial selective step caused by carcinogens through selection steps caused by self-inflicted conditions to selection for drug resistance caused by anticancer drugs, cells acquire oncogenic resistance

Full size image

Carcinogens (tumor promoters) select for resistance

It is generally assumed that carcinogens stimulate cell proliferation and block differentiation, because proliferation and dedifferentiation are common characteristics of cancer cells. In contrast to this assumption, carcinogens (e.g., DNA-damaging agents and phorbol ester) inhibit proliferation and induce differentiation. Carcinogenic factors are restrictive (for normal cells) not permissive. While inhibiting proliferation of normal cells, carcinogens do not inhibit (or even stimulate) premalignant cells: namely, cells with oncogenic mutations (see Wille et al.58). For example, the TPA tumor promoter is cytostatic in certain normal cells.59, 60, 61 Loss of Rb, overexpression of c-myc, and other oncogenic changes can render cells resistant to TPA.62 Then, TPA inhibits normal epithelial cells but their malignant counterparts, selectively favoring their proliferation.63

The same alterations that render cells insensitive to TPA also render them insensitive to TGF-β, an endogenous cytostatic agent for normal epithelial cells. TGF-β inhibits proliferation of normal cells but not malignant cells.64, 65 And, finally, TGF-β stimulates growth and metastasis of the most aggressive late-stage cancers. Whereas TGF-β inhibits the formation of benign skin tumors, it enhances the progression to invasive spindle carcinomas.66 The final stages of carcinogenic progression may specifically involve changes that radically alter the cellular interpretation of the TGF-β signal.64, 67

Similarly, the tobacco smoke and other DNA-damaging carcinogens inhibit growth of normal cells but not cells with oncogenic mutations (e.g., mutant p53).60, 63, 68, 69, 70 In cells with mutant p53, low doses of DNA-damaging carcinogens may not cause growth arrest. While inhibiting the proliferation of normal bronchial epithelial cells, the tobacco smoke may not affect lung cancer cells.70 So, the tobacco smoke not only causes p53 mutations but also inhibits proliferation of normal cells, thus selecting cells with p53 mutations.71 Agents that simultaneously induce mutations and select for these mutations may be especially carcinogenic. For example, human skin contains clonal patches of p53-mutated keratinocytes. In addition to being a mutagen, sunlight may favor the clonal expansion of p53-mutated cells.72 UV is an ambivalent (both cytotoxic and mitogenic) signal, based on the induction of early-response genes and proto-oncogenes such as jun and fos. UV light is preferentially cytotoxic to normal cells, thus driving skin carcinogenesis.73

Mesothelioma cells are resistant to the cytotoxicity of asbestos, a carcinogen that causes mesothelioma.69 Cytotoxicity caused by cadmium selects alveolar epithelial cells for apoptosis avoidance.74 Also, aflatoxin, which causes hepatocarcinoma, is cytotoxic to normal liver but not tumor cells.68 The essence of tumor promoters (including DNA-damaging carcinogens) is that they inhibit proliferation of normal cells, thus selecting cells that can proliferate in such a carcinogenic environment. In other words, carcinogens select for cells with oncogenic mutations because oncogenic mutations render cells resistant to growth arrest and apoptosis caused by these carcinogens.

Therapy-induced cancer and tumor progression

Typically, anticancer drugs are toxic to both cancer and normal cells, causing therapeutic and side effects, respectively. We can predict that cytotoxicity to cancer cells may cause tumor progression, whereas the cytotoxicity to normal cells may result in secondary tumors. Specifically:

Tumor progression

By selecting for oncogenic resistance, anticancer drugs may drive tumor progression. In vitro, chemotherapy can increase oncogenic characteristics, such as invasiveness and metastasis.75, 76, 77, 78, 79 Immortalized endocervical cells selected for resistance to cisplatin are malignantly transformed.80

Testicular cancer and childhood leukemia cells are apoptosis-prone. Therefore, chemotherapy preferentially kills such cancer and leukemia cells causing remissions and even curative outcomes. Yet, if the disease relapses, it becomes resistant to chemotherapy, because all sensitive cells have been selectively eliminated. As another example, hormonal therapy of prostate cancer is very effective, but eventually selects for hormone independence. Even further, inhibitors of androgen receptor may stimulate rather than inhibit resistant cancer cell.81 In differentiated thyroid cancer, therapy with TSH and radioactive iodine causes high response and remission rate. Following remission, thyroid cancer may relapse as dedifferentiated carcinoma, which is highly resistant to chemotherapy.82 As another example, Gleevec induces remissions in Bcr-Abl-positive leukemia. By killing sensitive leukemia cells, Gleevec selects for cells with additional oncogenic alterations that may be responsible for disease progression in relapse.83, 84, 85, 86 Similarly, antiangiogenic therapy may select for hypoxia-resistant cancers.87

When anticancer drugs cause therapeutic response, they simultaneously select for drug resistance due to elimination of drug-sensitive cancer cells (Figures 2 and 3). Yet, therapy of most common cancers is not sufficiently effective, thus causing neither remissions nor tumor progression. To cause tumor progression, a therapy must be near-curative.87

Secondary cancers

It is believed that normal cells do not acquire resistance to anticancer drugs because resistant normal cells do not emerge. Thus, patients do not become tolerant to chemotherapy over time.

Yet, normal cells can acquire resistance. Just they are not normal anymore: resistance to anticancer drugs is manifested as therapy-induced cancers. By selecting for resistance and for genetic instability, anticancer drugs cause secondary cancers. In comparison with tumor promoters, anticancer drugs are more cytotoxic, because they are aimed at cancer cells and used at maximally tolerated doses. Therefore, normal cells do not easily acquire resistance to anticancer drugs. In other words, they rarely become malignant following chemotherapy, but if they do, such chemotherapy-caused cancers are especially resistant to further therapy.88, 89, 90, 91 Secondary leukemia is often inhibited apoptotic response.92 Chemotherapy-related leukemias account for 10–20% of all acute myeloid leukemia (AML) cases and are resistant.88 Secondary tumors include brain tumors, leukemia, breast cancer, thyroid cancer, bone and soft-tissue sarcoma.93 Radiation, alkylating agents and topoisomerase II inhibitors are most common causes for secondary tumors.89, 90, 91, 92, 93, 94 Leukemias induced by alkylating agents usually follow a long period of latency and present as myelodysplasia with unbalanced chromosomal aberrations such as deletions of chromosome 13, 5 or 7. DNA topoisomerase II inhibitors cause leukemias with translocations of the MLL gene at chromosome band 11q23 or, less often, t(8;21), t(3;21), inv(16), t(8;16), t(15;17) or t(9;22). Between 2 and 12% of patients who receive topoisomerase inhibitors have developed t-AML.28, 95

Carcinogens as anticancer drugs

Arsenic and phorbol ester (TPA) exemplify carcinogens turned into anticancer drugs. Still this is considered as an exception. The prevailing notion is that anticancer drugs and carcinogens are opposites. Likewise, there is no search for new anticancer drugs among carcinogens. This may explain a reluctance to introduce the classic tumor promoter TPA to the clinic. In clinical trials, TPA has demonstrated effectiveness in therapy of drug-resistant leukemia.24

It is a misconception that an agent that causes cancer cannot be used as anticancer drug. In reality, alkylating and other DNA-damaging drugs, radiation, hormones and antihormones are all carcinogenic. There are startling examples. Alcohol is one of the causes of hepatocarcinomas. On the other hand, percutaneous ethanol injections are used for therapy of hepatocarcinomas.96 Furthermore, low ethanol concentrations (1 mmol) inhibit cell proliferation and increase apoptosis more strongly in hepatocarcinoma cells than in normal hepatocytes.97 Ethanol injection can lengthen survival in patients with hepatocellular carcinoma.98

Carcinogens are cytotoxic to certain cancer cells. Cancer cells are resistant to agents that cause this particular cancer. More than 65 years ago, it was noted that tumors that are caused by alkylating carcinogens are resistant to alkylating drugs.41 Yet, alkylating agents may be effective against some cancers such as testicular cancer and childhood leukemia. DNA-damaging drugs, which are effective in testicular cancer and leukemias, are not necessarily effective in secondary cancers. Similarly, cancer that has been promoted by TPA will be probably resistant to therapy with TPA. But leukemia cells that are resistant to DNA-damaging agents may retain sensitivity to TPA.24 Also, secondary cancers caused by chemotherapy are expected to be resistant to DNA-damaging drugs. But they may be sensitive to Gleevec.99

As shown in Figure 4, if agent A causes particular cancer, then agent A cannot ‘cure’ the same cancer. Thus, a futile therapy can be avoided. Cancer cells are still sensitive (or hypersensitive) to different classes of cytotoxic agents (agent B). In contrast, if agent B had caused cancer, then agent A (Figure 4b) may be used for therapy. A carcinogen turns into an anticancer drug and vice versa (Figure 4).

Figure 4
figure 4

The role reversal of anticancer drugs and carcinogens. (a) Cytotoxic agent A is a carcinogen because it inhibits proliferation of normal cells, selecting for cells with mutations (m). Cancer cells are still sensitive to agent B. In this scenario, mutations that are selected by agent A render cells resistant to agent A and sensitive to agent B. (b) Agent B is a carcinogen because it inhibits proliferation of normal cells, selecting cells with distinct types of mutations (m). Accumulation of mutations results in cancer that is resistant to agent B. Cytotoxic agent A, which at low doses causes cancer, inhibits proliferation of cancer cells that are caused by agent B. Cancer cells are sensitive to agent A and resistant to agent B

Full size image

It is important to emphasize that overlap between carcinogens and anticancer drugs is not accidental. Any agent that is cytotoxic can cause cancer and can be used for cancer therapy, precisely because it is cytotoxic. In theory, DMBA, okadaic acid and aflatoxin, for instance, may be developed as anticancer drugs. Okadaic acid, an inhibitor of serine/threonine protein phosphatases 1 and 2A, is a tumor promoter. Okadaic acid also induces apoptosis in most tumor cell lines and could be considered as an anticancer drug. Furthermore, cells with mutations in Ras are more sensitive to okadaic acid100, 101 implying its potentially chemopreventive properties. As another example, many effects of the carcinogen benzo(a)pyrene dihydrodiol epoxide (BPDE) and anticancer drug camptothecin (CPT) are similar. Low doses (50–100 nM) of BPDE activate Chk1 causing a reversible S-phase arrest, while higher doses cause permanent S-phase arrest and cell death.102 The cytotoxicity of CPT is also S-phase specific and CHK1-regulated S-phase checkpoint response protected cells from CPT-induced killing.103 Can BPDE be used therapeutically as an S-phase-specific drug that induces a reversible Chk-1-dependent arrest? And, if ethanol can be used for the therapy of liver cancer, why not to consider a one step further: cytotoxic agents from the tobacco smoke may restrain growth of certain types of lung cancer cells in some nonsmokers. This example is merely hypothetical.

Empirical chemoprevention: a double-edge sword

It is often emphasized that chemopreventive agents can inhibit oncogenic pathways, inhibit growth and induce apoptosis in cancer cells. Then, what is the difference between anticancer and chemopreventive agents? Furthermore, if agents are cytotoxic to cancer cells, then, like chemotherapy, they must cause severe side effects. Finally, cancer cells are relatively resistant to apoptosis. Therefore, if an agent can induce apoptosis even in cancer cells, then it can kill normal cells too. If so, side effects would be inevitable. Conventional chemopreventive agents are expected to be administered long-term to healthy individuals. As emphasized, even minor adverse side effects would be unacceptable.104 Actually, to cause apoptosis in cancer cells in vitro, potential chemopreventive agents are used at intolerably high concentrations. For example, chemopreventive agents from the green tea induce apoptosis and growth arrest in vitro at doses that are not achievable in humans via tea consumption.105 Perhaps, dietary chemopreventive agents, at physiological concentrations, do not cause apoptosis in cancer cells.

Second problem is that chemopreventive agents and carcinogens overlap (Figure 1). As any cytotoxic agents, chemopreventive agents may select for oncogenic resistance. If some premalignant cells are resistant to chemopreventive agent (due to additional oncogenic mutations), then this chemopreventive agent will select oncogenic cells, thus driving tumor progression. Moreover, the same agent may be chemopreventive at early stages, but may promote cancer later in tumor progression. For example, TGF-β and retinoic acid inhibit the formation of benign tumors, but enhance progression to carcinomas.32, 66

Selective chemoprevention: exploiting oncogene-addiction

Unlike anticancer drugs, chemopreventive agents should not necessarily affect advanced cancer cells. Chemopreventive agents should be selectively cytotoxic to premalignant cells. A chemopreventive agent must provide selective disadvantage (by causing apoptosis, growth arrest or differentiation) to cells with a single oncogenic alteration (Figure 5). But is this possible? Carcinogens select for cells with an oncogenic mutation, precisely because such a mutation renders cells resistant. How may chemopreventive agents selectively suppress such resistant cells? In other words, why would carcinogen-resistant cells be hypersensitive to other agents? First, while rendering cells resistant to one condition, an oncogenic mutation can cause a collateral sensitivity to another condition. c-myc renders cells resistant to growth arrest but sensitive to apoptosis, whereas Ras can prevent apoptosis but favors cell senescence.3, 48, 106, 107

Figure 5
figure 5

Chemoprevention versus tumor promotion. Cytotoxic agent can select either for or against mutant cells. A mutation can confirm resistance to one agent (carcinogen) and sensitivity to another (chemopreventive agent). In theory, cells bearing oncogenic mutations may be vulnerable to certain agents (chemopreventive agents)

Full size image

Similarly, mutations can render cells resistant to one agent (carcinogen) but sensitive to another (chemopreventive) agent. As the simplest example, cells with Bcr-Abl are resistant to DNA-damaging drugs but are very sensitive to inhibitors of Bcr-Abl.108 Also, p53 disruption sensitizes certain cells to DNA-damaging agents, but renders these cells resistant to the antimetabolite 5-fluorouracil.109 As another example, activation of S-phase checkpoint and induction of E2F1 render cancer cells hypersensitive to lapachone.110

Anticancer drugs that selectively suppress cells with a single oncoprotein are chemopreventive by definition. For example, Gleevec (imatinib), an inhibitor of Bcr-Abl, is selectively toxic to cells that express Bcr-Abl.111, 112 Since Gleevec selectively kills cells with Bcr-Abl, Gleevec is actually a mechanism-based chemopreventive agent for chronic myeloid leukemia (CML), which is caused by Bcr-Abl. Furthermore, there is no need for continuous administration of mechanism-based chemopreventive agents. In brief courses, Gleevec will eliminate rare cells that acquire Bcr-Abl. (Of course, to use Gleevec as chemopreventive agent, it will be necessary to identify people at high risk for CML.) BCR-ABL fusion transcripts can be detected by RT-PCR in the leukocytes of some healthy individuals.113 For CML, an early therapy is actually a mechanism-based chemoprevention. This suggests therapy with Gleevec in newly diagnosed CML before additional mutations occur. Similarly, inhibitors of the EGF receptor and ErbB2 will be more effective in premalignant cells than in advanced cancer with multiple mutations. In mice, inhibitors of the EGF receptor prevented breast cancer in ErbB2 transgenic mice114 and polyps.115

Especially at early stages of carcinogenesis, when premalignant cells depend on one sole oncogene, inactivation of such an oncogene can result in apoptosis (oncogene-addiction).116, 117 Oncogene-addiction is the Achilles heel of oncogenic resistance. Yet, further tumor progression renders cells oncogene-independent due to acquiring additional mutations. For example, c-myc-induced tumorigenesis is initially c-myc-dependent but can progress to c-myc-independent phase.118, 119, 120

Protective chemoprevention: induction of nononcogenic resistance

Carcinogenesis is selection for oncogenic resistance to carcinogens. In contrast, acquiring nononcogenic resistance may stop further tumor progression. In theory, by acquiring nononcogenic resistance, a cell avoids selective pressure without becoming oncogenic. For example, nononcogenic resistance might prevent the development of cancer (therefore, primary cancer cells are usually Pgp-negative). In colorectal carcinogenesis, Pgp is typically expressed in the adenoma stage, protecting the epithelium against further environmentally induced damage.121 This may explain why only 5% of adenomas actually progress to cancer121 and why untreated cancers are usually Pgp-negative. As another example, the DNA repair protein MGMT prevents carcinogen-induced conversion of benign tumors into malignant tumors.122 In other words, by conferring nononcogenic resistance, MGMT abolishes selective pressure. MGMT is often silenced in tumors and therefore they are sensitive to alkylating drugs.9 Reactivation of MGMT renders cancer cells resistant to chemotherapy.123

Similarly, abrogation of cytotoxic effects of aflatoxin prevents carcinogenesis in rat liver.124 Even oncogenes can prevent cancer in certain circumstances.125 Bcl-2 retards ultraviolet- and chemical-induced tumorigenesis.73 Perhaps, by causing resistance to apoptosis, Bcl-2 (in certain conditions) prevents selection for additional oncogenic mutations. (Of course, in most conditions, Bcl-2 accelerates carcinogenesis.) Pharmacological agents also can render cells nononcogenically resistant, thus preventing carcinogenesis.126, 127

Cancer therapy: selective drugs and selective combinations

As we discussed, selection for resistance to carcinogens and anticancer drugs are essentially identical processes. There are several consequences. First, anticancer drugs are carcinogenic and may further drive tumor progression. In order to occur, however, tumor progression must be preceded by tumor regression (remission) (Figure 3). Second, carcinogenic agents may become new anticancer drugs. Third, ‘carcinogenic history’ of a particular cancer may indicate which drugs will be ineffective. For example, cancer caused by arsenic will not be sensitive to arsenic-based therapy, and doxorubicin (DOX)-induced secondary cancer will be probably insensitive to standard chemotherapy. Thus, patients may avoid futile therapy.

In rare cancers that depend on a single oncogenic mutation (e.g., Bcr-Abl in CML and the EGF receptor with activating mutation in lung cancer), drugs that inhibit such oncogenes can selectively kill leukemia and cancer cells.128, 129, 130, 131 Technically, the therapy with an agent that targets a single oncogenic alteration and selects against cells with such mutations is ‘chemoprevention’ (as defined herein). In other words, it is chemoprevention that is used as cancer therapy. In general, however, cancer cells have multiple mutations and are intrinsically resistant to growth arrest and apoptosis. It seems easier to kill normal cells than cancer cells. Furthermore, following chemotherapy, cancer cells can acquire nononcogenic resistance (e.g., MRP). How then can anticancer drugs be selectively cytotoxic to such resistant cells? It is possible to exploit resistance using antagonistic drug combinations.132 For example, a combination of apoptosis-inducing agents and caspase inhibitors is clearly antagonistic. But these combinations can selectively kill cancer cells that are ‘resistant’ to inhibitors of caspases.133, 134, 135 This will allow one to use apoptosis-inducing agents selectively. As another example, pretreatment with DOX antagonizes cytotoxicity of paclitaxel (PTX). Low concentrations of DOX cause G2 arrest in normal cells but not in cancer cells with mutant p53 (oncogenic resistance) and with MRP (nononcogenic resistance).132 While DOX protects nonresistant normal cells, PTX kills only DOX-resistant cells. In theory, the combination DOX+PTX will eventually select for resistance. Resistance to this combination could be due to re-acquiring sensitivity to DOX. Although selection for resistance is not avoidable, it can be exploited for therapy.

Unifying model

The common feature of carcinogenesis, therapy and chemoprevention is selection for resistance due to suppression of sensitive cells. Whether normal or malignant cells (or cells at different stages of tumor progression) are more sensitive determines the ‘direction’ of the process: carcinogenesis versus therapy/prevention (Figure 5).

For example, DNA-damaging agents such as etoposide (an inhibitor of topoisomerase II) can cure testicular cancer.136 Topoisomerase II inhibitors are also very cytotoxic to normal epithelial and bone marrow cells. Also, they can induce the translocation of Bcr-Abl (t9;22) in bone marrow cells.95 Most importantly, Bcr-Abl-expressing cells are resistant to apoptosis caused by DNA-damaging drugs.5, 137 Therefore, continuous exposure to DNA-damaging agents (as well as endogenous cytotoxic agents) may cause selective expansion of Bcr-Abl-expressing cells, eventually resulting in leukemia (Figure 6a). At this point, it would be futile to treat such leukemia by DNA-damaging drugs, even though they are effective in primary leukemias and testicular cancer. Instead, the Achilles heel is the dependence on Bcr-Abl. This could be exploited for therapy (and potentially for chemoprevention). The Bcr-Abl kinase inhibitor Gleevec is selectively cytotoxic to Bcr-Abl-expressing cells, thus causing remission (Figure 6b). On the other hand, it selects for resistant leukemic clones. Once resistance occurs, opposite strategies may be used in sequence. One striking example is the reversion from blast to chronic phase of leukemia after withdrawal of Gleevec.138 Finally and remarkably, Gleevec also may cause cancer, acting as a carcinogen. Gleevec causes side effects due to inhibition of certain normal cells that are dependent on c-Kit and PDGF.139, 140 Then (like any agent that is toxic to certain normal cells), it may select for oncogenic resistance. In fact, the development of squamous cell carcinoma of the skin141 and c-Kit-expressing small-cell lung cancer142 has been reported following therapy of CML with Gleevec.

Figure 6
figure 6

From carcinogenesis to therapy. (a) Carcinogenesis. By inducing apoptosis in normal hematopoietic cells, DNA-damaging agents select for cells with Bcr-Abl. (b) Therapy. By inducing apoptosis in Bcr-Abl-expressing cells, Gleevec selects for normal hematopoietic cells, thus causing remission. (c) Progression. By sparing Bcr-Abl-expressing cells having additional oncogenic mutations (Res-Bcr-Abl), Gleevec selects for resistant leukemia cells and disease progression

Full size image

Given that carcinogenic and therapeutic effects are determined by cellular context, the term carcinoagents can include carcinogens, physiological regulators and anticancer agents. Carcinoagents are ambivalent. They suppress one type of cells while promoting survival of another type. Such suppression could be due to apoptosis, nonapoptotic (slow) cell death, cell senescence, differentiation and reversible cycle arrest. Although we cannot discuss this in detail, apoptosis is considered a goal of cancer therapy, the preferential outcome of therapy.4, 104, 135, 143, 144, 145, 146, 147, 148, 149, 150 This suggestion is based on the observation that curable malignancies are usually apoptosis-prone. Returning to the beginning of this article, the question is how can apoptosis be a goal of cancer therapy, if apoptosis avoidance is a hallmark of cancer. First, cancer cells can be resistant to DNA damage but sensitive to kinase inhibitors and vice versa, depending on the carcinogenic history (Figures 4 and 6). Second, cancer arises from apoptosis-prone tissues. For example, cancer does not originate from arterial smooth muscle cells, which are extremely resistant to apoptosis. In contrast, inefficient hematopoiesis due to extensive apoptosis of defective cells leads to leukemia. Apoptosis-prone tissues become malignant because they are apoptosis prone and/or controlled by restrictive regulators. For example, normal breast and prostate cells are tightly controlled by growth factors and hormones and are apoptosis-prone. Like exogenous carcinoagents, physiological restrictions drive carcinogenesis. Later, by invading and outgrowing their blood supply, tumors do not need any exogenous carcinogens to drive further selection. A tumor itself creates carcinogenic conditions to select for resistance. Thus, cancer cells become resistant compared with their parental normal cells but not necessarily compared with arterial SMC and dermal fibroblasts, for instance. This will be a topic for a future review.

Future directions

As announced by Andrew von Eschenbach, ‘NCI sets goal of eliminating suffering and death due to cancer by 2015’.151 To cure and prevent cancer, therapeutic and chemopreventive agents and their combinations can be selectively cytotoxic to cancer and premalignant cells while sparing normal cells. Three potential strategies, aimed at cancer- and tissue-specific targets and selective combinations aimed at universal targets with protection of normal cells, can control cancer.152 Yet, any effective therapy is limited by selection for resistance and potential tumor promotion. Therefore, opposite therapeutic strategies should be used in sequence.152

To control cancer, it is important to realize, as discussed here, that carcinogenesis and cancer therapy are identical processes and there is little difference between agents that promote and halt cancer. Similarities between carcinogens and anticancer drugs indicate that carcinogens offer a potential source of new anticancer drugs to cancers induced by other carcinogens (specifically, the type of therapy must be selected on the understanding that the inducer cannot also be the therapy). The term carcinoagents emphasizes the dual nature of cytotoxic agents. Carcinoagents include not only carcinogens and drugs but also physiological restrictive signaling molecules such as TGF-β and immunological suppressors, for instance. All could be used to control cancer.