Polyamines and cancer: old molecules, new understanding

The amino-acid-derived polyamines have long been associated with cell growth and cancer, and specific oncogenes and tumour-suppressor genes regulate polyamine metabolism. Inhibition of polyamine synthesis has proven to be generally ineffective as an anticancer strategy in clinical trials, but it is a potent cancer chemoprevention strategy in preclinical studies. Clinical trials, with well-defined goals, are now underway to evaluate the chemopreventive efficacy of inhibitors of polyamine synthesis in a range of tissues.

Polyamines are organic cations that are derived from amino acids and are found in all organisms. Polyamine research dates back several centuries (TIMELINE 1). In 1678, Van Leewenheuk identified crystals in semen that were later identified as the tetraamine spermine; the diamine putrescine was first identified in microbes in the late 1800s, and the triamine spermidine was identified in the early twentieth century 1 . Putrescine, spermidine and spermine are the main polyamines found in prokaryotes and eukaryotes, but other amines have been identified (for example, in extreme thermophiles) 1 .
In some bacteria, putrescine derives from arginine decarboxylation through a two-step mechanism 1 . In mammals, the diamine putrescine is synthesized in the cytoplasm as a consequence of the decarboxylation of ornithine, an amino acid that is not found in proteins and that is produced as part of the UREA CYCLE, which involves both cytosolic and mitochondrial enzymes (FIG. 1). Putrescine is the precursor used in spermidine and spermine biosynthesis. All urea-cycle enzymes are expressed in the liver and, to a lesser degree, in the intestinal mucosa 2 . In addition, certain urea-cycle enzymes, including those involved in polyamine metabolism, are widely expressed in all tissues. So, although urea-cycle enzymes are expressed primarily in the liver and intestine, polyamines are made in all tissues. Polyamines are also obtained from the diet (foods that are high in polyamines include cheese and red meat 3 ) and from other sources, such as intestinal bacteria 1,4 . These externally derived polyamines are transported into cells from extracellular spaces.
Genetic studies in the 1980s by Herbert and Celia Tabor and colleagues demonstrated that polyamines are essential for optimal growth and viability in bacteria 5 and yeast 6 . John Cleveland and colleagues have recently extended this paradigm to mammals, with the observation that the gene that encodes ornithine decarboxylase (ODC) -the enzyme required for the first stage in polyamine synthesis (FIG. 1) -is essential in mice. The mechanism of this requirement seems to involve the suppression of apoptosis by ODC in the developing mouse embryo 7 . Polyamines have been widely implicated in the growth and development of a range of mammalian tissues and in remodelling processes associated with tissue repair (BOX 1). Interestingly, polyamine synthesis is downregulated as cells become senescent 8 in many tissues in adults 9 .
The association of increased polyamine synthesis with cell growth and cancer was first reported in the late 1960s. Russell and Snyder reported high levels of ODC activity in regenerating rat liver and in several human cancers 10 . Subsequently, Anderson and Heby observed similar changes in tumour-cell polyamine levels 11 . Russell and others went on to show that polyamine levels were increased in the urine of patients with cancer, although studies so far have not proven that urine polyamine levels ODC 15 . ODC is also regulated by androgens in the prostate gland 16 , and the gene encoding ODC is markedly induced in human prostate cancer 17 . In addition to these links between ODC activity and cancer risk factors, the use of pharmacological inhibitors of polyamine metabolism has implicated polyamines in a range of cancers.
Difluoromethylornithine (DFMO), which irreversibly inactivates ODC, is the most widely studied example of a polyamine-metabolism inhibitor that suppresses cancer development in animal models 18 and has been evaluated in clinical trials. Agents that target other polyamine metabolic enzymes (FIG. 1) are also potent inhibitors of cancer growth in experimental model systems 19,20 , and some of these agents are either under evaluation or have been evaluated in human clinical cancer therapeutic trials 21,22 .
Despite the fact that the link between polyamines and cancer has been known for more than 30 years, until recently, knowledge of the specific mechanisms by which polyamine metabolism is altered during carcinogenesis has been lacking, as has evidence to indicate that polyamines have a causative rather than associative role in cancer. New advances in understanding the roles of polyamines in cancer are discussed here, as are the uses of inhibitors of polyamine metabolism for cancer treatment and prevention.

Mechanisms of polyamine upregulation in cancer
The mechanistic basis of the association between cancer risk factors and polyamine metabolism has become clearer during the past decade, particularly through studies of colon cancer. The relationships between two tumour-suppressor genes and two oncogenes implicated in colon carcinogenesis and genes regulating polyamine levels in the colonic mucosa are shown in FIG. 2. ODC is upregulated in the intestinal mucosa of individuals with familial adenomatous polyposis (FAP) 23 , a heritable form of colon cancer. The tumour suppressor adenomatous polyposis coli (APC) is mutated or lost in the germline of individuals with FAP 24,25 and in somatic intestinal epithelial cells that develop into colonic polyps in individuals who are not genetically predisposed to colon cancer 26 . In colon tumour cells, loss of APC function leads to increased expression of the MYC 27 oncogene, aberrant expression are an adequate prognostic marker of cancer 12 . The role of urinary polyamines as a predictive marker for response to therapy remains to be established.
Early work in the lab of Roswell Boutwell showed that chemical promoters of skin carcinogenesis induced ODC 13 , and ODC activity is now known to be induced in normal tissues by a range of environmental and genetic cancer risk factors. For example, ultraviolet light induces Odc gene expression in rodent models of squamous-cell skin cancer, and inhibitors of ODC suppress this process 14 . In addition, asbestos, the causative agent of the deadly lung cancer mesothelioma, is a potent inducer of Summary • Polyamines are naturally occurring organic cations found in plants, animals and microbes. They are formed by the enzymatic decarboxylation of the amino acids ornithine or arginine.
• Ornithine decarboxylase (ODC) is the first enzyme in the polyamine synthesis pathway in mammals and is the target for difluoromethylornithine (DFMO), a substrate analogue and specific inhibitor that irreversibly inactivates ODC when it binds to the active site of the enzyme.
• ODC and several other polyamine metabolic proteins are essential for normal cell and tissue functions, including growth, development and tissue repair. ODC and polyamine content are increased in many cancers arising from epithelial tissues, such as the skin and colon.
• Polyamines exert their effects in eukaryotic cells in part by regulating specific gene expression.
• In murine and human colonic mucosal tissue, ODC is negatively regulated by the adenomatous polyposis coli (APC) tumour-suppressor gene. APC is mutated or deleted in the germline of people with familial adenomatous polyposis (FAP), a genetic syndrome associated with a high risk of colon cancer. APC is also mutated or deleted in somatic colon epithelial cells in most sporadic, or non-genetic, forms of colon cancer.  34 . ODC has been shown to cooperate with RAS in skin carcinogenesis in mice 35 .

• Loss of APC function causes an increase in ODC activity and polyamine biosynthesis, and tumour formation in
ODC is not the only polyamine metabolic gene that is regulated by oncogenes and tumoursuppressor genes. The ODC regulator antizyme (OAZ), which targets ODC for PROTEASOMAL DEGRADATION, is also regulated by APC in the Apc Min/+ mouse model. When wild-type APC is lost in intestinal epithelia of these mice, decreased OAZ activity contributes to increased ODC levels 32 (FIG. 2). In addition, expression of spermidine/spermine N 1 -acetyltransferase (SSAT) is negatively regulated by the KRAS oncogene, which is commonly mutated -and, as a result, aberrantly activated -in human colon cancer and other gastrointestinal of which is associated with the development of Burkitt's lymphoma and several epithelial cancers, in addition to colorectal cancer 28 . MYC encodes a transcription factor that is required for the proliferation of some normal cells, but when overexpressed leads to uncontrolled growth and cancer 29 Figure 1 | Polyamine metabolism in mammals. The metabolism of arginine, which is produced in the urea cycle, through the action of arginase, results in the production of ornithine (other details of the urea cycle are not shown). Ornithine decarboxylase (ODC) is required for the first step in polyamine synthesis, in which ornithine is decarboxylated to produce putrescine. Decarboxylation of S-adenosylmethionine (SAM), by S-adenosylmethionine decarboxylase (AMD) yields decarboxylated SAM (dcSAM), which donates its propyl amine moiety (not shown) for the formation of spermidine and spermine by spermidine synthase (SRM) and spermine synthase (SMS), respectively. The spermidine/spermine N 1 -acetyltransferase (SSAT) is a propylamine acetyltransferase that monoacetylates spermidine and can either mono-or di-acetylate spermine. These acetylated polyamines have at least two potential fates. Diamines and acetylated polyamines are substrates for export by the putative transporter DAX (diamine exporter) 116 , and are then eliminated in urine 1 . Acetylated spermidine and spermine are also substrates for a flavin-dependent polyamine oxidase (PAO), which catalyses their conversion back to putrescine. A spermine oxidase (SMO), which can oxidize non-acetylated spermine, has recently been characterized 117 , although its physiological role has not been established. Putrescine, spermidine and spermine can also be imported from extracellular compartments through a transport mechanism that is not well defined.

Box 1 | Polyamines in normal growth, development and tissue repair
Polyamines have long been associated with cell proliferation, hypertrophy (increase in cell size) and tissue growth, and are known to be involved in the development of several tissue types. Substantial evidence for the functional involvement of polyamines in the normal development of the intestinal tract has accumulated. Early studies indicated that polyamine-synthesis inhibitors disrupted intestinal development in mice 91,92 . Subsequently, it was found that the repair of gastric and duodenal injury was also dependent on polyamine metabolism 93,94 . The role of polyamines in tissue repair might be to facilitate tissue remodelling, as has been reported for certain types of lung damage 95 . Inhibition of polyamine synthesis suppresses wound healing in rodents 96 . Polyamines have also been implicated in the development and function of both male 97 and female reproductive organs 98,99 . Aberrant expression of spermidine/spermine N 1 -acetyltransferase influences specific gene expression controlling reproductive-tract tissue growth and function 100 .
In addition, increases in polyamine metabolic-enzyme activities and tissue polyamine levels have also been associated with the normal growth and hypertrophy of several other tissues -including skin, breast, kidney and heart -in rodents 1 . However, pharmacological or genetic suppression of these enzymatic activities does not prevent either cardiac or renal hypertrophy responses 101,102 . So, polyamines are functionally involved in growth responses of some tissues, but are only associated with growth responses in others.
suppressor genes. These results indicate how polyamine levels become increased in colorectal and other gastrointestinal cancers.
Polyamine levels are also increased in other epithelial cancers, including skin and prostate cancers. HRAS cooperates with ODC in skin carcinogenesis, but there is as yet no evidence to implicate HRAS in the regulation of polyamine metabolic enzymes in this tissue 35 . ODC transcription is regulated by androgens in mammalian and human prostate cancer cells 38 . Inappropriate MYC expression due to chromosomal translocation is the cancers 36 . The KRAS-dependent signalling pathway regulates SSAT transcription through a mechanism involving peroxisome-proliferator-activated receptor-γ (PPARγ), a putative tumour suppressor. PPARγ positively regulates the transcription of SSAT through a PPAR response element (PPRE) in the SSAT promoter 37 . KRAS suppresses SSAT transcription by inhibiting PPARγ expression and subsequent binding to the SSAT promoter 36 . So, polyamine synthesis and catabolism are both regulated by signalling pathways that are influenced by oncogenes and tumour- Cell-culture studies indicate that wild-type APC suppresses transcription of MYC, which is an activator of ornithine decarboxylase (ODC) transcription. Wild-type APC also acts to regulate ODC antizyme (OAZ), a protein that regulates ODC activity by targeting it for degradation. In normal cells and tissues, most of the KRAS protein is inactive as a signalling molecule. The tumour-suppressor peroxisome-proliferatoractivated receptor-γ (PPARγ), which activates spermidine/spermine N 1 -acetyltransferase (SSAT) transcription, is repressed by active KRAS. Therefore, in normal cells and tissues, wild-type APC and KRAS lead to reduced proliferation, increased apoptosis and reduced neoplasia. b | When APC is mutated or deleted, as occurs in individuals with familial adenomatous polyposis, ODC expression is increased. Loss of APC function results in a decrease in OAZ, which contributes, along with transcriptional activation as discussed above, to increased ODC activity. Oncogenic mutations in KRAS, which prevent hydrolysis of bound GTP and activate KRAS signalling activity, suppress polyamine catabolism. Consequently, mutant APC and KRAS act to promote neoplasia by increasing polyamine biosynthesis and suppressing polyamine catabolism. Increased levels of polyamines are associated with increased cell growth, decreased apoptosis and expression of genes, including some involved in tissue remodelling. These processes contribute to normal development, tissue repair and, when deregulated, neoplasia.
that extremely high polyamine content can cause apoptosis and consequent ulceration [42][43][44] . However, these cases of polyamine-induced apoptosis occur primarily when intracellular polyamine levels are abnormally high, and are a consequence of loss of regulation of polyamine homeostasis. Polyamines are necessary for blood-vessel development (angiogenesis) occurring in response to damage to normal tissues or tumour growth. Inhibition of polyamine synthesis blocks angiogenesis in models of gastric ulceration 42 and in tumour models 45,46 . Polyamine metabolism also contributes to argininedependent effects on colon tumour cell growth 45,46 . Arginine is metabolized to either ornithine, requiring the enzyme arginase (FIG. 1), or nitric oxide. ODC can be inactivated by nitric oxide through nitrosylation 47. Arginine and its catabolite NG-hydroxy-L-arginine can induce cell growth arrest by inhibiting arginase in a manner that is rescued by exogenous polyamines 48 . Presumably, inhibition of arginase prevents polyamine production both by suppressing ornithine production and by favouring nitric oxide production, which inhibits ODC.
However, although polyamines seem to be associated with numerous cellular processes, a key criticism of polyamine research has been that the specific mechanisms underlying their modes of action -including those in cancer cells -have not been defined. Over the past few years, polyamines have been shown to affect specific gene expression through both transcriptional and post-transcriptional processes (BOX 2). cDNA microarray technologies have been used to identify polyamine-regulated genes in cancer cells. In one example, cells derived from human colon tumours were treated with DFMO. This ODC inhibitor decreased cellular putrescine and spermidine levels and also blocked the ability of these cells to form tumours in severe combined immunodeficient mice when added to drinking water 41 . DFMO suppressed the expression of several genes involved in two types of cell-cell interactions, TIGHT JUNCTIONS and GAP JUNCTIONS. Tight-junction proteins have been thought to act as tumour suppressors 49 , whereas gap-junction proteins have been implicated in cell communication involved in carcinogenesis 50 . Experiments in cell culture showed that the effects of the ODC inhibitor could be rescued by exogenous putrescine, supporting the interpretation that these genes are regulated by polyamine-dependent mechanisms. Other groups have observed similar effects of DFMO on cell-junction proteins 51,52 , although alterations seem to be cell-type specific.
Polyamines contribute to the regulation of other cancer-related functions, including apoptosis and proliferation. Termination of fetal development is associated with increased apoptosis in embryonic cells lacking ODC. As discussed earlier, Odc-knockout mice die early during embryogenesis 7 . Increased levels of polyamine have long been associated with proliferation, and recent experiments with genetically altered rodents have confirmed the important roles of these molecules in both normal and neoplastic growth 53 . basis for neoplasia in Burkitt's lymphoma 29 . Inappropriate expression of MYC in models of this disease results in inappropriate ODC expression and tumorigenic phenotypes, including increased proliferation and decreased apoptosis 30,31 . In addition, ODC is regulated by the putative oncogene eukaryotic translation initiation factor 4E (eIF4E) in a range of cancers 39 .

Evidence for a causative role in cancer
A recent study of genetic variability affecting ODC expression has provided evidence that increased polyamine synthesis and retention in the cell has a causative role in human cancer colon cancer, rather than being a purely associative effect 40 . The relationship of a SINGLE-NUCLEOTIDE POLYMORPHISM (SNP) affecting ODC promoter activity to risk of COLON-POLYP recurrence in a human colon-polyp prevention study was evaluated. The ODC SNP (G315A) occurs 315 nucleotides downstream of the ODC transcriptional start site and is located between two consensus E-boxes in the promoter region. The transcriptional activator MYC and the transcriptional repressor MAD1 bind to these elements. The frequency of this SNP has been measured in several groups of people, including participants in a colon cancer prevention trial. Fifty-five percent of participants in this trial, all of whom previously had a colon polyp, were homozygous G at this locus; thirty-five percent were heterozygous G/A and ten percent were homozygous A. The SNP was shown to have functional consequences for ODC expression. The transcriptional repressor MAD1, a MYC antagonist, selectively repressed ODC transcription in an A-allelespecific manner. Finally, it was found that the A-allele was associated with a statistically significant reduction in risk of colon-polyp recurrence and that this risk was even further reduced in participants who reported taking aspirin regularly. Aspirin use does not affect allelespecific transcription of ODC, but does affect polyamine catabolism by inducing the transcription of SSAT (see later).
The ODC A-allele therefore seems to be a potential prognostic factor for polyp risk in humans, and allele number might be a predictor of human response to aspirin as a colon cancer-preventive agent. The ODC A-allele favours binding of the transcriptional repressor MAD1. So, the ODC A-allele might work in concert with aspirin to lower tissue polyamine levels and, therefore, the risk of colon-polyp recurrence.

Polyamine functions in cancer
Polyamines affect numerous processes in carcinogenesis. Increased polyamine levels are associated with increased cell proliferation, decreased apoptosis and increased expression of genes affecting tumour invasion and metastasis. Conversely, suppression of polyamine levels is associated with decreased cell growth, increased apoptosis and decreased expression of genes affecting tumour invasion and metastasis 37,41 . These generalizations need to be placed in context. Several reports, including our own, have documented inhibitor that has proven to be effective in the treatment of certain hyperproliferative and infectious diseases, including removal of excess facial hair in women and in the treatment of African sleeping sickness 57 . The Merrell group also developed reagents that inhibited AMD 58 and the flavin-dependent polyamine oxidase (PAO), which is involved in polyamine catabolism 59 . PAO catalyses a reaction to produce the shorter-chain amines putrescine and spermidine, as an alternative to synthesis from amino acids. Some evidence exists to support the hypothesis that combinations of polyamine-synthesis inhibitors and PAO inhibitors might be more potent antiproliferative strategies than either type of inhibitor alone 60 . In addition to the development of drugs that target specific enzymes in the polyamine pathway, structural analogues of the polyamines themselves have been synthesized and evaluated as potential anticancer drugs. The rationale for this latter approach includes the fact that polyamines participate in both the negative regulation of polyamine biosynthetic enzymes and the positive regulation of polyamine catabolic enzymes. As listed in TABLE 1, Phase I (toxicity assessment) trials have recently been reported for the AMD inhibitor SAM4861 (REF. 58) and several polyamine analogues that do not specifically target a particular enzyme, but rather have features similar to the polyamines themselves. One Phase II (efficacy assessment) trial found the polyamine analogue DENSPM to be well tolerated, but without evidence of therapeutic benefit 22 . Other Phase II trials are ongoing, but results have not yet been reported.
The development of these selective inhibitors of polyamine metabolism enabled the closer examination of effects of polyamine depletion in cell culture -depletion of intracellular polyamines suppressed growth, but was not generally toxic to cells 61 . The inhibitors were also evaluated in combination with cytotoxic anticancer agents (drugs and ionizing radiation) and were found to have modest interactive effects in animal models 62 . From the late 1970s and throughout the 1980s, DFMO was actively evaluated as an anticancer drug, either alone or in combination with other agents, in clinical trials. DFMO was chosen over other ODC inhibitors in these early clinical trials for several reasons, including activity in animal models and pharmacokinetic properties in humans that allowed favourable physiologically significant concentrations of the drug in serum. DFMO, either alone or in combination with other chemotherapies or radiotherapy, was generally found to have little antitumour activity in Phase I trials. Phase II trials also showed DFMO to be generally ineffective as a treatment for patients with leukaemias and brain tumours 63 . The reason for the lack of cancer therapeutic efficacy of DFMO might be a consequence of the finding that this agent does not generally kill cells. As successful therapies are generally cytotoxic, either through direct or indirect action, lack of cancer therapeutic efficacy of DFMO is not that surprising. However, DFMO does suppress the expression of genes that are involved in cell proliferation, tissue remodelling and/or tumour invasion 41,64 . These

Use of polyamine metabolic inhibitors
Interest in targeting polyamine metabolism as a potential strategy for cancer chemotherapy was stimulated in the early 1970s by a study from Williams-Ashman and Schenone 54 . This work indicated that methylglyoxal (bis) guanylhydrazone (MGBG), a drug used in the treatment of leukaemia, inhibited S-adenosylmethionine decarboxylase (AMD) -a key enzyme in polyamine synthesis (FIG. 1; TABLE 1) -and the formation of spermidine and spermine. However, MGBG applications in cancer therapy were limited because of drug-induced toxicity, especially to self-renewing normal tissues, including the bone marrow and intestinal tract.
In the mid-1970s, the Centre de Recherche Merrell International (later affiliated with The Dow Chemical Company), a pharmaceutical firm in Strasbourg, France, began a research programme to develop other, less toxic, inhibitors of polyamine metabolism and to test their efficacy in cancer therapy. Investigators at this institute demonstrated that inhibitors of ODC suppressed mammalian tumour cell growth 55 . Those findings spurred the development of a wide range of very selective ODC inhibitors, including DFMO 56 . DFMO is a specific ODC RIBOSOME Particles composed of RNA and protein that are sites of protein synthesis.

Box 2 | Mechanisms of polyamine-dependent gene expression
Ornithine decarboxylase (ODC) and its diamine product putrescine participate in a range of cellular processes affecting cell behaviours. Several specific mechanisms influencing gene expression have been described that require unique polyamines or putrescine , which is their diamine precursor. In the early 1980s, Myung Park and collaborators demonstrated that the polyamine spermidine was the essential substrate for the posttranslational modification of the putative eukaryotic translation inhibition factor 5A (eIF5A) 103 . Subsequent studies have shown that the Saccharomyces cerevisiae equivalent of eIF5A is an essential gene in this yeast 104 , but is not essential for general protein synthesis 105,106 . Rather, eIF5A seems to be involved in the processing of specific RNAs 107,108 .
A second example of a specific mechanism of polyamine-mediated gene expression is the formation of ODC antizyme (OAZ), a protein that targets ODC for proteasomal degradation 109  The activity of DFMO as an inhibitor of carcinogenesis in experimental models indicated that it might be a potent chemopreventive agent in humans. DFMO inhibits intestinal polyp formation in Apc Min/+ mice 32 . APC mutations are found in almost all sporadic (nonhereditary) forms of colon cancer, and ODC levels are increased in colon polyps in humans 18 . Our group has therefore carried out chemoprevention trials with DFMO in patients with FAP and sporadic colon polyps.
The aim of these trials was to adopt the simplest possible strategy for drug delivery to maximize compliance in these individuals, who had not yet developed cancer. Serum levels of DFMO and tissue levels of polyamine were measured (BOX 3). Although the serum half-life was known to be in the order of hours, single daily oral doses of the drug were able to suppress certain tissue polyamine content 66 .
ODC is also upregulated in other intraepithelial neoplasias (IENs) -non-invasive precursors of epithelial cancers. Development of DFMO as a potential chemopreventive agent for use in human diseases has been most systematically developed for colorectal cancer, non-melanoma cutaneous cancer, bladder cancer and cervical cancer, with a somewhat less aggressive approach in Barrett's oesophagus and breast cancer. A summary of the current status of definitive clinical trials and evidence supporting their undertaking is provided in TABLE 2. Identification of DFMO toxicities. For a candidate compound to be adopted in a preventive setting, it is essential that the drug causes minimal or no toxicity 67 . This requirement is dictated by the fact that cancer prevention involves treatment of basically healthy individuals who are at risk of, cancer, but have not developed the disease. Therefore, identifying the limiting dose for DFMO -that which produces no evident side effects, but results in the desired biochemical effect in the tissue/organ under study (in this case the colon or skin) -was crucial before implementation of definitive or pivotal trials, which could support New Drug Applications to the Food and Drug Administration (FDA).
features of DFMO might contribute to the potency of this drug as a cancer-preventive agent, at least in experimental models.

Use of DFMO for cancer prevention
In contrast to the rather moderate effects of DFMO on models in which cancer is already established, DFMO is a potent inhibitor of carcinogenesis. Chemical and physical carcinogenesis proceed by a series of events, including an initiation phase involving DNA damage, leading to gene mutations, followed by a promotion phase characterized by proliferation of initiated cells. Elegant studies by Weeks et al. 65 showed that DFMO suppressed skin carcinogenesis by blocking the promotion phase of this process. The mechanism of this effect is thought to involve inhibition of the increased cell proliferation that is associated with the promotion phase. Subsequent studies indicated that DFMO was a potent inhibitor of epithelial carcinogenesis in a range of models, including those for skin, breast and colon 18 .

Box 3 | Measuring polyamine variables in clinical studies
Early in our clinical chemoprevention studies with difluoromethylornithine (DFMO), we decided that we needed to measure tissue variables that would indicate the efficacy of the treatment. We conducted a clinical evaluation of variability in a range of factors affecting tissue polyamine levels 115 . These variables included measures of ornithine decarboxylase (ODC) and spermidine/spermine N 1 -acetyltransferase RNA levels and enzyme activities as factors affecting polyamine levels and direct measures of polyamine levels. This study identified several sources of error in measurements of these variables in human tissues and indicated that direct assessment of polyamine levels in target tissues using highperformance liquid chromatography methods was the most reliable measure of polyamine metabolism for predicting the effects of intervention using agents like DFMO. Assessment of the suppression of ODC enzyme activity in skin biopsy samples might also be a reliable marker of the effects of DFMO in patients undergoing treatment for skin cancer prevention 14,73 . Attempts to use an easily accessible tissue such as shed buccal mucosal cells as a surrogate for other internal organs were unsuccessful, largely because of contamination of the buccal mucosal cells by endogenous bacteria and artificial increases of polyamine levels 75 . Other important variables that have been identified include age, as both putrescine levels and the ratio of spermidine to spermine, and changes in these variables as a function of DFMO treatment, decrease as a function of donor age 66 . mucosa, used DFMO alone 9,71-77 . Subsequently, studies in experimental model systems demonstrated that several signalling pathways were de-regulated in colon carcinogenesis, and prevention strategies using combinations of inhibitors were more effective than singleagent strategies 40 . Specifically, combinations of DFMO and non-steroidal anti-inflammatory drugs (NSAIDs) were shown to be potent combinations in experimental models. NSAID use has been consistently associated with a reduced risk of colon cancer 79 . There are two aspects to the rationale for combining ODC inhibitors such as DFMO with NSAIDs (FIG. 3). First, cyclooxygenases (COXs) -which are inhibited by NSAIDs -are potent modifiers of APC-dependent intestinal carcinogenesis in mouse models. Genetic or pharmacological suppression of COX2 markedly reduces numbers of intestinal tumours 80 , and the COX2 inhibitor celecoxib suppressed polyp size in a clinical trial of patients with FAP 81 . COX2 acts on arachidonic acid to produce prostaglandins. The specific mechanisms involved in COX2-dependent tumorigenesis are not yet fully defined, but might involve activation of gene expression mediated by cell-surface or intracellular prostaglandin receptors 82 . Some, or all, of these mechanisms might be totally independent of polyamines. Therefore, COX2 and ODC seem to act as modifiers downstream of the APC tumour-suppressor gene in both animal models and in humans. Cell-culture studies indicate that DFMO and the NSAID sulindac act additively to suppress colon tumour cell viability 64 , indicating that these agents are acting through independent mechanisms. It is unknown whether NSAIDs can suppress tumour formation by mechanisms that are totally independent of COXs and polyamines.
The second aspect of the rationale for combining ODC inhibitors and NSAIDs is that SSAT is a transcriptional target for several NSAIDs. The nonselective COX1/COX2 inhibitor sulindac activates PPARγ, which recognizes a unique DNA sequence in the SSAT promoter to induce transcription of this Testing DFMO as a treatment, described above, had already identified changes in hearing as a potentially limiting toxicity in prevention trials 63 . However, a series of clinical studies by our group and investigators at Wisconsin, primarily targeting people with colon and skin IENs, have established that the hearing changes affect PURE TONE only, are rapidly reversible following discontinuation of the drug, and that a low dose of DFMO can be used that does not produce measurable changes in hearing, but nevertheless results in substantial lowering of polyamine levels and/or ODC inhibition [68][69][70] .
These results also have implications for the use of DFMO in other organs. If polyamine levels cannot be substantially reduced in an organ at the limiting dose (about 0.5 g/m 2 /day) then use of the drug in a particular chemoprevention setting might not be realistic. For example, a limiting dose of DFMO does lower polyamine levels substantially in cervical tissue, but does not do so in oral or breast tissue, and probably the bladder 9,71-77 . In many tissues, including the colonic mucosa, DFMO suppresses putrescine and spermidine, but not spermine 66,78 . In prostate tissue, which contains 5-10 times more spermine than either putrescine or spermidine, a one-month treatment with DFMO did reduce spermine levels 76 . Therefore, the effects of DFMO in specific tissues depend both on the expression of the target enzyme and the regulation of the individual amines. The expression of ODC, the target for DFMO, might not be sufficiently increased and might therefore contribute to the growth of IENs in these tissues. In addition, DFMO did not reverse Barrett's oesophagus lesions -the failure to affect either tissue polyamines or other markers might have contributed to the fact that clinical trials have not progressed for these IENs (TABLE 2).

Use of DFMO in combination with NSAIDs.
Our initial colon cancer prevention trials, to assess the safety and efficacy of suppressing polyamine content in colorectal PURE TONE A single frequency tone measured as part of clinical audiometric evaluations. *Preclinical evidence includes effects on both cell and animal models of cancer 18 . ‡ Supportive evidence includes effects of agents on surrogate markers of cancer development (for example, suppression of tissue polyamines) at doses that cause minimal toxicity to patients. § Further information on these chemoprevention trials can be found in the online links box.  Sulindac and celecoxib are both non-steroidal anti-inflammatory agents. ¶ Cytology and measures of proliferation. DFMO, difluoromethylornithine.
Ongoing DFMO chemoprevention trials. The first goal of our cancer chemoprevention trials was to define a dose of DFMO that was both safe and effective. Our dose de-escalation studies and subsequent randomized placebo-controlled studies demonstrated that oral doses of DFMO between 0.25 and 0.5 g/m 2 /day for times as short as 1 month and as long as 1 year were effective in reducing rectal polyamine content in humans [85][86][87] . These studies also demonstrated that these doses, which can reduce rectal polyamine levels, did not produce detectable toxicities in excess of those observed in the group receiving placebo.
A recently completed clinical trial assessed the toxicity of DFMO in combination with low doses of sulindac. Final toxicity results from this trial will be available within 2 years. Although follow up in this 3-year treatment trial is not yet complete, and the study is still blinded, no detectable differences in toxicities have been identified between the two study groups (F.L.M., E.W.G. and C. McLaren, unpublished observations). Based on these results, a Phase III study has been initiated to determine if DFMO plus sulindac can reduce colon-polyp recurrence. Results for this polyp-recurrence trial, which also uses a 3-year treatment duration, will be available within 5 years, taking into account time for patient accrual to the study and treatment time.
When we initiated our clinical studies, there was great concern in the medical community regarding the potential hearing-related toxicity of DFMO. In the various human epithelial tissues studied, the dose of DFMO that was effective in lowering polyamine levels or inhibiting ODC ranged from about 0.25-1.0 g/m 2 /day 73,76,78 . As the dose of DFMO that produces noticeable hearing changes is about 1.0 g/m 2 /day, lower doses might be both efficacious and safe.
There are several biological factors that could adversely affect the potentially beneficial long-term outcomes of DFMO use in cancer prevention. These could include amplification of ODC in some tissues, as has been reported in cell-culture models 88 , or increased uptake of polyamines from bacterial flora in the intestinal lumen. Both of these mechanisms could overcome the inhibitory effects of DFMO. Additionally, unexpected long-term toxicities might emerge.

Implications and future directions
'Proof of principle' for chemoprevention of cancer in humans has been established [85][86][87] . However, toxicity has abrogated the widespread adoption of agents, such as retinoids, for the prevention of cervical and oral cancers 89 . Recent results with finasteride in the prevention of prostate cancer do indicate that the development of new tumours could be reduced significantly in an at-risk population. However, more high-grade tumours were detected in the finasteride treatment group, raising the concern that these might be a consequence of drug use 121 . Consequently, the future application of this drug in cancer prevention is clouded by this potential risk. By contrast, DFMO toxicity has been well studied in a systematic manner, gene 37 . NSAIDs that are structurally unrelated to sulindac, like aspirin, also induce SSAT transcription 40 by both COX-dependent and -independent mechanisms 37,40 . Aspirin seems to activate SSAT transcription through NF-κB and AP-1 consensus elements in the SSAT promoter (N. Babbar, R. Casero and E. W. G., unpublished observations). One mechanism of the chemopreventive and therapeutic effects of certain NSAIDs is their ability to induce apoptosis. NSAID-induced apoptosis can be reversed, in part, by exogenous polyamines 37,83 , indicating that NSAIDinduced SSAT induction and subsequent polyamine export are causally involved in NSAID-induced apoptosis. Consequently, NSAIDs act as inducers of polyamine catabolism and export, and complement inhibitors of polyamine synthesis in acting to lower tissue polyamine levels. So, ODC inhibitors and NSAIDs counter the effects of mutation of tumoursuppressor genes, such as APC andPPARγ, and oncogenes, such as MYC and KRAS, which act to increase tissue polyamines in cancer. Combinations of DFMO and NSAIDs work at least additively in several models of colon and intestinal carcinogenesis 67,84 , corroborating this model. It is hoped that these results of combination chemoprevention in experimental model systems will translate to use in patients with colorectal cancer. polyamines in normal growth, development and tissue repair, and how these processes go awry in cancer, could be manipulated for future therapeutic benefit. The concept of combination therapy for chemoprevention is important, as combinations of agents are generally more effective than single agents in animal models. Polyamine levels can increase because of synthesis and/or uptake, and their metabolism is highly regulated by a series of catabolic enzymes in mammals. Consequently, it should not be surprising that attaining the goal of reducing high polyamine levels in tissues during cancer development might require targeting two or more of these processes. The finding that the polymorphism affecting ODC-promoter activity was most profoundly associated with colon-polyp recurrence in individuals taking aspirin (an activator of polyamine catabolism), indicates that interventions at several points in polyamine metabolism might be necessary to optimally repress the development of epithelial cancers. We have focused on NSAIDs as the agents used in clinical prevention trials in combination with DFMO. However, targeting features of polyamine metabolism, such as polyamine uptake and efflux, and/or catabolism, with DFMO or other polyamine-synthesis inhibitors might also be useful strategies for the prevention of epithelial cancers. and side effects are unlikely to prevent its use in cancer prevention if efficacy is demonstrated.
Polyamines continue to be molecules that hold fascination for biologists, chemists, molecular biologists and clinical researchers. Although these molecules have been known for over 300 years, a mechanistic understanding of their roles in normal and disease processes has only been developed in the past 30 years (FIG. 3). The first definitive clinical trials for cancer prevention are still in progress. Several aspects of polyamine metabolism and function present numerous experimental clinical opportunities. Although this review has focused on the promise of targeting polyamine metabolism in cancer prevention, increasing understanding of the role of these molecules in human cancer might lead to new ways of using these agents for cancer therapy. In this regard, a recent report indicates that DFMO might improve survival in certain patients with brain tumours 90 .
Our understanding of the upstream regulation of ODC has increased markedly in the past few years and the roles of APC, MYC and related transcriptional activators and repressors offer unique opportunities for intervention. Similarly, the downstream function of polyamines and their important roles in angiogenesis and invasion have recently been identified (discussed earlier). New information relating to the role of the