Hormesis, the biological and toxicological concept that small quantities have opposite effects from large quantities, is reviewed with emphasis on its relevance to nutrition.
Hormetic and other dose–response relationships are categorized, depicted, and discussed. Evidence for nutritional hormesis is presented for essential vitamin and mineral nutrients, dietary restriction, alcohol (ethanol), natural dietary and some synthetic pesticides, some herbicides, and acrylamide. Some of the different hormetic mechanisms that have been proposed are reviewed.
The credence and relevance of hormesis to nutrition are considered to be established. The roles of hormesis in nutritional research and in formulating nutritional guidelines are discussed.
The New York City Department of Health and Mental Hygiene.
Moderate levels of exercise promote good health, whereas excessive levels are debilitating (Melzer et al., 2004). In molecular pharmacology, many chemicals are known to have opposite effects as a function of dosage (e.g., the antibiotics penicillin, erythromycin and streptomycin promote bacterial growth at low doses, contrary to effects at higher doses). It has long been recognized that mild forms of stress can promote mental and physical function whereas extreme stress is more likely to cause mental anguish and physical ailments, the Yerkes-Dodson Law in experimental psychology (Yerkes and Dodson, 1908). In stating that only the dose makes a thing not a poison, Paracelsus (the supposed model for Goethe's Dr Faustus) recognized some six centuries ago that in medicine the efficacy of toxic chemicals depends on dosage. These examples of small quantities having effects opposite from that of large quantities are commonly termed hormesis. The hormesis concept claims that as the dose of an agent being studied is reduced, the response of the end point being measured does not simply get smaller and smaller, drifting into background noise, but may actually reverse course and change to an opposite response. Hormesis proponents believe it to be commonly manifested in both biology and toxicology, and highly generalizable according to biological model tested, end point measured, and chemical class/physical agent employed. This review will discuss evidence for hormesis in human nutrition, the possible relevance of hormesis to nutritional research, and the formulation of hormetic nutritional guidelines.
Hormesis – definition and dose–response forms
Figure 1 (adapted from Davis and Svendsgaard, 1990) provides some representative dose–response forms. Dose–response relationships have usually been characterized by either the threshold model (Figure 1a) or the linear non-threshold (LNT) model (Figure 1b). Since the consolidation of toxicological and pharmacological conceptual thinking, the most accepted dose response model in these disciplines has been the threshold model, which assumes that dose has no effect until a threshold is reached, at which point response increases linearly with dose. According to the LNT model, response is directly proportional to dose without any threshold, so that some level of response is always present, even at the lowest possible dose level. The LNT model has become the standard model for assessing the health risk of chemical carcinogens and radiation by regulatory agencies in many countries; while for non-carcinogens, the same regulatory agencies assume that there is a threshold dose, below which there is no health risk.
Hormesis is characterized by dose–response relationships having two distinct phases (i.e., biphasic, non-monotonic) whose forms depend on end points measured. Examples include inverted U-shaped curves (Figure 1c) showing low-dose enhancement when reduction is expected (normal function end points displaying this type curve include growth, fecundity, longevity and cognitive function); J-shaped curves (Figure 1d), which are allied to U-shaped curves, showing low-dose reduction when enhancement is expected (adverse dysfunction end points displaying this type curve include carcinogenesis, mutagenesis and disease incidence). Note that hormesis not only challenges the threshold and LNT models but also more importantly suggests that as dose decreases there are not only quantitative changes in measured responses but also qualitative changes vis-à-vis both control (background) and high-dose levels.
Figure 2 (adapted from Eaton and Klaassen, 2001) is a depiction of hypothetical dose–response relationships contributing to a postulated hormetic effect, in this case a U-shaped dose–response curve. Low doses produce a response demarked Protective in Figure 2b. Increased doses produce an oppositely postulated response demarked Adverse in Figure 2a. The combined effect curve (Figure 2c) shows a hormetic U-shaped dose–response.
The definition of hormesis adopted here is that of adaptive nonlinear biphasic dose–response relationships characterized by small quantities having opposite effects from large quantities, that is, small doses elicit opposite responses to that seen at high doses. This definition purposefully sidesteps the potentially vexing issue of beneficial/harmful effects, which should be deferred to subsequent evaluation of the biological and ecological content of the response.
Hormesis – current background
There is a great deal of current interest in the topics of hormesis and hormetic-like responses. The most effective current proponents of hormesis are Edward Calabrese's group at the University of Massachusetts who have systematically developed methodologies for evaluating dose–response relationships (Calabrese, 2005a). They have reported hormetic-like biphasic dose/concentration responses for numerous endogenous agonists, including NO, adenosine, opioids, adrenergic agents, prostaglandins, estrogens, androgens, 5-hydroxytryptamine and dopamine (Calabrese and Baldwin, 2001a); inorganic agents including numerous toxic substances such as arsenic, cadmium, lead, mercury, etc., and chemotherapeutic agents, peptides and ethanol (Calabrese and Baldwin, 2003a); as well as processes such as apoptosis (Calabrese, 2001). The hormetic chemotherapeutic agents studied included antibacterial, antiviral, antitumor and antiangiogenesis agents, with their quantitative and temporal dose responses being similar to that reported for both chemicals and radiation (Calabrese and Baldwin, 2003b). Hormetic-like biphasic dose–responses have been reported in over 130 human tumor cell lines from a wide range of agents, including antineoplastics, toxic substances (i.e., environmental pollutants), non-neoplastic drugs, endogenous agonists and phyto-compounds (Calabrese, 2005b). Extensive hormetic-like biphasic dose–response relationships have been reported in immunology, with over 90 different immune response-related end points induced by over 70 endogenous agonists, over 200 drugs, over 40 environmental contaminants, and over 30 animal models spanning over a dozen mammalian and human cell lines (Calabrese, 2005c). The clinical implications for hormesis in humans are presently essentially unknown, although numerous investigators have raised some important concerns. For example, Brandes (2005) reports that certain antidepressant drugs in addition to producing biphasic growth responses in in vitro cancer cells and stimulating cancer growth in rodents also correlates with an increased risk of breast and other cancers in some, but not all, patients on these drugs.
Methodologies developed by Calabrese's group have been used to generate two separate intensive and extensive hormetic databases. The more recent and complete database, The Relational Hormesis Database (a.k.a. The Hormesis Database), culled some 1450 publications to report >6000 cases of hormetic dose–response relationships for some 900 broadly diversified chemical and physical agents (Calabrese and Blain, 2004). The hormetic dose–responses satisfied rigorous a prior evaluative criteria (including to varying degrees: strength of study design, magnitude of stimulation, statistical significance and reproducibility of findings). With stimulation used in the context of Calabrese and Baldwin's (2002) definition, that is, a response opposite to that observed at higher (i.e., greater than threshold) doses, the maximum stimulatory responses were typically only approximately 30–60% greater than the concurrent control, with nearly 80% of the maximum responses being less than twice the control value. In general, the widths of the stimulatory responses were also modest, typically extending over a dose range of 20-fold (i.e., 1/20) or less immediately below the NOAEL (No Observed Adverse Effect Level – the highest dose not differing in a statistically significant manner from the control group, which serves as a toxicological quasi-threshold), and often less than 10-fold (i.e., 1/10). To reiterate, hormetic responses in the vast majority of cases within The Relational Hormesis Database were typically quite modest in both magnitude and width.
The analyses of Calabrese's group revealed the uncovered biphasic dose–response relationships to be quite common and broadly generalizable; that is such responses do not appear to be restricted to biological model, end point measured, or stressor agent, and appear to represent a basic feature of biological responsiveness to chemical and physical stressors. Furthermore, the quantitative features of the hormetic-like dose responses are remarkably similar with respect of response amplitude and width, and the relationship of the maximum response to the ‘quasi-threshold’ NOEL. Such similarities across models, end points and stressor agents suggest a similar type of adaptive strategy possibly related to resource allocation. In addition, the biphasic dose–response was also found to be more common than other competing models, even the long-pre-eminent toxicological threshold model (Calabrese and Baldwin, 2001b, 2003a).
Examples of hormetic nutrition
The following subsections will review manifestations of nutrition hormesis in the following areas: essential nutrients (vitamins and minerals), dietary restriction, alcohol (ethanol), natural dietary and some synthetic pesticides, some herbicides and acrylamide.
Hormetic essential vitamin and mineral nutrients
Vitamins and minerals are hormetic essential nutrients that are necessary to maintain human health. Small daily levels of these substances are both required and beneficial, while excessive dietary levels can lead to hypervitaminosis, tissue mineralization, or electrolyte imbalance. For example, the Seventeenth Edition of The Merck Manual of Diagnosis and Therapy published in 1999 discusses deficiency, dependence and toxicity in vitamins A, D, E, K, B6; deficiency and toxicity in the macrominerals sodium, potassium, calcium, magnesium, phosphate and chloride, and in the microminerals iron, iodine, fluorine, selenium and copper. All trace elements (i.e., microminerals) are toxic at high levels, and some (e.g., arsenic, nickel, and chromium) have been implicated in carcinogenesis.
Eaton and Klaassen (2001) in their ‘Principles of Toxicology’ chapter of the major toxicological reference Casarett and Doull's Toxicology: The Basic Science of Poisons depict the individual dose–response relationship for an essential substance required for normal physiologic function and survival such as a vitamin or trace element as being U-shaped (Figure 3). At very low doses there is a high adverse effect, which decreases with increasing dose. This is the deficiency region of the dose–response relationship for essential nutrients. As the dose is increased to a point where the deficiency no longer exists, no adverse response is detected; the organism is then in a state of homeostasis, defined as the maintenance of constant internal state for an organism's efficient function and performance to ensure that a stable physiological milieu is maintained in the face of perturbations. (Homeostasis is operationally defined as the dose range that results in neither deficiency nor toxicity.) However, as the dose is increased to higher levels, a toxicity adverse response (usually qualitatively different from that observed at deficient doses) appears and increases in magnitude with increasing dose, just as with other toxic substances. Thus, the biphasic nature of the U-shaped response can, practically speaking, be subdivided into low- and high-dose regions where the toxicity response differentially occurs (the arms of the U), plus a region of no toxic effect (the trough of the U). As specifically cited by Eaton and Klaassen, Vitamin A deficiency can have harmful vision effects while excessive Vitamin A can damage the liver or cause birth defects; and high doses of selenium can affect the brain while high doses of estrogens may increase the risk of breast cancer. Nevertheless, all three of these substances are essential for life and their deficiencies will cause harm. It is important to note that while Eaton and Klaassen have not explicitly claimed Figure 3 to be an example of hormesis, they have come very close to doing so in noting it to be conceptually similar to the hypothetical hormetic dose–response relationship that they had depicted in Figure 2. In addition, Mertz (1981) has cited many examples of essential trace elements exerting a U-shaped dose–response on physiological functioning, ranging from impairment at levels of deficient intake, to optimal function at intermediate levels, to toxicity at excessive intake levels.
Hormetic dietary restriction
Owing to its ability to intervene in both biological aging and life-shortening pathogenesis, dietary modification in the form of dietary restriction (DR) is the most effective and reproducible laboratory intervention for extending lifetime survival (both mean and maximum lifespan) in a variety of diverse organisms. The two DR protocols most commonly used in animal studies are: (1) every-day caloric restriction (CR) with the restricted animals being typically being given 20–40% fewer calories than the ad libitum (AL) fed control animals, and (2) intermittent fasting (IF) with the restricted animals typically going a whole day without food and then permitted to eat AL on the following day.
Both the CR and IF laboratory DR regimens provide essential nutrients and vitamins, thereby avoiding confounding effects of malnutrition. In fact, CR and IF animals are almost always healthier, sleeker, and more active than their AL counterparts, who tend to develop obesity in mid-life. While the CR laboratory regimen administered throughout life is generally more protective than adult-onset CR, both prevent adult-onset obesity, significantly extend lifespan and suppress tumorgenesis. Laboratory caloric restricted diets, which as noted are designed to provide adequate micronutrient coverage, have been found to produce reduced oxidative DNA damage (Haley-Zitlin and Richardson, 1993). Contrast this result with laboratory reports that micronutrient deficiencies produce DNA damage both through oxidative lesions and single- and double-strand breaks (Ames, 1998).
DR by either the CR or IF regimen can significantly extend the life span of rodents by some 30–40% (Mattson et al., 2002). Life-extending CR effects have been observed in, amongst others, protozoa, yeast, nematode, several insect species including fruitfly, mouse, rat, hamster, guinea-pig, dog, cow and in several non-human primate species. CR is a potent and broadly acting cancer-prevention regimen in laboratory experimental carcinogenic models, being effective in several species, for a variety of tumor types, and for spontaneous and chemically induced neoplasias. For example, CR inhibits spontaneous neoplasias in several knockout and transgenic mouse models, it suppresses the carcinogenic action of several classes of chemicals in rodents, and also inhibits several forms of radiation-induced cancer. CR in animal models also reduces age-associated neurodegenerative disorders, prevents age-assisted declines in psychomotor and spatial memory tasks, improves the brain's plasticity and ability for self-repair, as well as retarding age-associated physiological deterioration and delaying and, in some cases, preventing age-assisted diseases. These salutary effects have been found to arise from either CR (i.e., energy) or meal frequency rather than from restriction of specific toxic dietary contaminants or specific macro- or micronutrients (Yu, 1994). They occur without a reduction of mass-specific metabolic rate or level of physical activity (Hursting et al., 2003).
Observational studies suggest that dietary restriction has beneficial effects on human morbidity and mortality. Participants in the Harvard Alumni Health Study and the Nurse's Health Study with body mass indexes (a possible biological marker of CR) some 15–20% below the national average (a hallmark of ‘chronic but mild undernutrition’) had reduced mortality. Their peripheral blood lymphocytes had higher DNA repair capacity and exhibited no appreciable age-dependent decline in DNA-repair potential vis-à-vis normal subjects (Raji et al., 1998). Moreover, physiological changes analogous to those observed in CR monkeys, including high-density lipoprotein (HDL) cholesterol increases, are reported in Muslims who fast during the holy month of Ramzan. Residents of Okinawa, Japan, who are known to consume fewer calories than residents of the main Japanese islands, display lower death rates from cerebrovascular disease, cancer, and heart disease. Food shortages in some North European countries during World War II led to a sharp fall in mortality from coronary artery disease followed by sharp rise in mortality with the war's end, although these observations are difficult to interpret because of confounding factors such as malnutrition (Strom and Jensen, 1951). There are some other more controlled human demonstrations, such as the Biosphere 2 studies (Verdery and Walford, 1998), which also suggest positive effects but which unfortunately suffer from a dearth of subjects and non-optimum controls. On a more encouraging note, emerging evidence from human population studies (as well as laboratory experimental data) indicate that low calorie intake can reduce risk of Alzheimer's disease, Parkinson's disease and stroke, three of the most devastating neurodegenerative conditions in the elderly (Roth et al., 1999; Mattson et al., 2002). In addition, some as of yet small-scale clinical studies have reported that long-term CR ameliorates diastolic function and reduces atherosclerosis (Fontana et al., 2004; Meyer et al., 2006).
Anti-aging and life-prolonging effects of laboratory CR studies are accompanied by stimulation of various maintenance mechanisms, including increased efficiency of DNA repair, increased fidelity of genetic information transfer, more efficient protein synthesis, more efficient protein degradation, more efficient cell replacement and regeneration, improved cellular responsiveness, as well as fortification of the immune system. Among the specific mechanisms proposed to account for the antiaging and life-prolonging actions of CR are oxidative damage attenuation, alteration of the glucose–insulin system and alteration of the IGF-1, which ubiquitously acts on tissues to regulate growth, cell death, and development (Hursting et al., 2003; Masoro, 2003). As an additional and more general explanation, Masoro (1998) has proposed hormesis as the beneficial action(s) resulting from the response of an organism to low-intensity stress. In this context, ‘stress’ is operationally defined as a signal generated by a physical or chemical factor, the stressor, which in a living system initiates a series of events in order to counteract, adapt and survive. Some moderate intensity long-term stressors have been reported to delay aging and prolong life and include temperature shock (heat and cold), irradiation (UV-, gamma- and X-rays), heavy metals, prooxidants, alcohol, exercise, as well as CR (Rattan, 2004). Stress exposure induces various regulatory stress response proteins (e.g., glucocorticoids glucose-regulated stress proteins and/or hsps heat shock proteins) that protect cellular components as well as allow a better degradation of damaged proteins during stress. The heat shock protein system is known to protect cells against the damaging action of a broad spectrum of physiological stresses in addition to heat and CR, for example, cold, amino acid analogues, heavy metals, free radicals, exercise activity, etc. (Lindquist, 1986). These stress response proteins repair exposure-caused cellular damages and potentially even the damages present before the stress, so that organisms may be in better conditions after the stress and live longer. Other stress response mechanisms that have been proposed include induction of detoxification enzymes, cell proliferation/apoptosis, DNA repair, protein turnover and antioxidative response (Yu and Chung, 2001; Klaunig, 2005).
In support for the hormesis hypothesis, Masoro (2003) has cited findings that single-gene mutations that extend the life of invertebrate species also increase the ability of these organisms to cope with damaging agents. Recent neurodegenerative laboratory studies lend additional strong support to hormetic stress mechanisms (Mattson et al., 2002; Anson et al., 2003). It was found that stress induced by every-other-day feeding (IF, with the laboratory animals going a whole day without food and then eating AL on the next day) sometimes produced even more positive results than CR feeding every day. The observed reductions in neurodegenerative disorders were attributed to activation at the cellular and molecular levels of hormetic responses in which neurons increased production of neurotropic factors and stress proteins (specifically hsps heat shock proteins and glucose-regulated stress proteins). The intriguing possibility that the beneficial effects of IF can at least in part be dissociated from caloric intake is further supported by the finding that targeted deletion of the insulin receptor in adipose cells results in increased longevity without a reduction in caloric intake (Bluher et al., 2003).
Hormetic alcohol (ethanol)
The review of Pohorecky (1977) provides strong psychological, behavioral and biochemical evidence for alcohol (ethanol) affecting the experimental behavior of animals and humans in a biphasic dose–dependent manner, with low-doses being excitory and high-doses depressant. For non-human mammalian biological end points (including, amongst others, the liver, kidney, heart, spleen, endocrine, embryonic development and birth weight), Calabrese and Baldwin's (2003c) assessment of the toxicological and pharmacological literature demonstrated ethanol-induced biphasic dose responses over a wide range. In most cases, the low-dose hormetic responses were judged to be statistically significant and reliable. They observed striking similarities across studies in both hormetic dose-ranges and hormetic dose–response magnitudes, with the maximum hormetic response approaching two-fold but was usually 25–40% greater than the control value. Furthermore, the maximum hormetic response was generally within a factor of 2–4 of the estimated NOAEL. They argued against a common explanatory mechanism, and for a diverse plethora of underlying mechanisms. They also stated that regardless of actual response mechanisms, the similarity of dose–response characteristics likely reflected an important overall regulatory strategy built into the framework of a homeostatic control system. Calabrese and Baldwin (2003c) further noted that these collective findings closely resembled the biphasic hormetic responses to a wide range of toxic chemical agents that they had previously reported (Calabrese and Baldwin, 1997).
While there are a considerable number of published investigations suggesting that chronic alcoholics have an increased susceptibility to infectious diseases (in particular pneumonia), Hallengren and Forsgren (1978) have presented evidence that at low to moderate blood alcohol concentrations there is an increased adherence, phagocytosis and chemotaxis of human polypmorphonuclear leucocytes. The magnitude of the detected response is consistent with that seen for hormetic responses, with the hormetic concentration range being about four-fold. It would appear that these results have important ramifications for biomedical researchers and clinicians.
Human health benefits of regular light-to-moderate alcohol consumption vis-à-vis abstinence include decreased risks of dementia, diabetes, osteoporosis, and cardiovascular disease and death (Standridge et al., 2004). Both case–control and cohort epidemiological studies have consistently shown that light-to-moderate drinkers are at lower risk of cardiovascular disease and death than nondrinkers, although heavier alcohol consumption can negatively affect the neurologic, gastrointestinal, hematologic, immune, psychiatric and musculoskeletal organ systems. The hormetic biphasic effects of alcohol consumption on total human mortality is a well studied and documented phenomenon with the basic J-shape of the dose–response curve confirmed in the vast majority of studies and being remarkably stable and independent of assessment measure (Gaziano and Buring, 1998; Rehm, 2000). For example, Gronbaek (2004) cited 19 separate epidemiological studies conducted over the last three decades that describe the impact of alcohol on total mortality as J-shaped, with a beneficial effect of regular light-to-moderate intake (10–40 g per day, or one to three alcoholic beverages), and a detrimental effect of both lower and higher intake. These studies have been reported by a large number of different research teams, with the effects observed in populations of different genders, races and nationalities. The total mortality risk reduction at light-to-moderate levels is likely due to lower risk of fatal cardiovascular disease without dramatic increases in other causes of death. Rehm (2000) notes that the J-shape results mainly from the beneficial effect of moderate alcohol consumption on ischemic cardiovascular conditions coupled with the detrimental effects of drinking on other health conditions. While the risk curves of detrimental effects may vary between linear, exponential or threshold effects, they always seem to be monotonic: the more consumption, the higher the disease-specific mortality risk. Rehm further notes that this means that the J-shaped curve should not be expected and cannot be found in populations with no or few coronary heart disease (CHD) deaths, for example, in some developing countries or in younger age groups.
Epidemiological studies indicate that all types of alcoholic beverages are associated with increased cancer risk at higher drinking levels, suggesting that ethanol itself causes the detrimental effect rather than any particular type of beverage. A great deal of recent research has focused on the possible benefits of specific alcoholic beverages. Some have postulated that wine's polyphenolic antioxidants including resveratrol and proanthocyanidin reduce risk of cardiovascular disease. Resveratrol has been shown to exert its effects on cells by inducing a stress response mediated by the sirtuin protein family (Tissenbaum and Guarente, 2001; Howitz et al., 2003). At this point in time the totality of evidence suggests that the major beneficial component of alcoholic beverages on cardiovascular mortality is in fact ethanol per se rather than some other component (Rimm et al., 1996). Gronbaek (2004) states that it is likely that any apparent additional beneficial effect of wine on health in addition to the effect of ethanol itself is a consequence of confounding.
Several plausible mechanisms for the cardioprotective effect of light-to-moderate alcohol intake have been suggested (Korthuis, 2004; Standridge et al., 2004). One is its role in increasing plasma HDL cholesterol concentrations and in reducing low-density lipoprotein (LDL) cholesterol concentrations. Such changes have been suggested to account for approximately half of alcohol's protective effect against CHD. (Commonly referred to as the good cholesterol, HDL binds with cholesterol and brings it back to the liver for elimination or reprocessing, thereby lowering total cholesterol levels in body tissues and its build-up on arterial walls, and in a sense reversing the atherosclerotic process.) Free radical scavenging and inhibition of LDL oxidation by free radicals have also been proposed as mechanisms; as well as reduction/inhibition of platelet adhesion and aggregation, clotting factor concentrations, vascular smooth muscle cell proliferation and migration, and inflammation involving monocytes and T lymphocytes. Of particular interest in light of previous reference to hsps heat shock proteins are animal studies showing that oxidative stress from low dose alcohol consumption induces several heat shock and antioxidant proteins, which may serve as cardioprotective proteins in hearts subsequently exposed to ischemia/reperfusion (I/R) injury (Sato et al., 2004). Also of interest is the surprising finding that alcoholics had significantly less DNA damage than controls and demonstrated higher capacity for DNA repair, possibly because excessive ethanol rather than damaging DNA (it is not genotoxic) induces antioxidant and repair enzymes (Pool-Zobel et al., 2004).
Hormetic synthetic and natural dietary pesticides
Pesticides are substances that prevent, destroy or repel pests. While most intended pesticides are synthetic chemicals, some are plant derivatives, organic dusts or biological agents/products such as bacteria or their toxins. There is evidence that some synthetic chemical pesticides that enhance tumor formation at high doses may affect a reduction in tumor incidence at lower doses. For example, initial studies of DDT (1,1-bis(p-chlorophenyl)-2,2,2-trichloroethane) with the rat model used by the United States National Toxicology Program suggested a hormetic response at low doses and carcinogenicity at higher doses, ‘showing a U-shaped curve at very low doses’ (Sukata et al., 2002). A later follow-up study reported that a low dose of DDT can inhibit the formation of both preneoplastic and neoplastic lesions in the rat liver in an initiation-promotion model, with the dose–response curve showing a J-shape hormetic phenomenon (Kushida et al., 2005). DDT and similar organochloride insecticides have been of concern because of their unusual lipophilicity, persistence because of their chlorine substituents and concentration in adipose tissue. However, many thousands of chlorinated chemicals are produced in nature and natural pesticides can also bioconcentrate if they are fat-soluble. Furthermore, there is no convincing epidemiological evidence of a carcinogenic hazard of DDT to humans (Key and Reeves, 1994). Specifically, ‘the literature does not contain even one peer-reviewed independently replicated study linking DDT exposure to any adverse health outcome’ (Attaran and Maharaj, 2000). In support of laboratory studies reporting hormetic effects is the confirmatory speculation of Morse (1998) that low doses of pesticides may be responsible for a hormetically induced resurgence of insects and mites in secondary pest outbreaks following pesticide applications on agricultural commodities. A major impact of this hormesis effect would be its leading to the need for additional pesticide treatments resulting in a spiraling increase in pesticide usage, termed the ‘pesticide syndrome’ in the entomological literature.
As opposed to synthetic chemical pesticides, Ames and Gold (2000) and Gold et al. (2003) have reported the ubiquitous presence of natural pesticides in foods. About 99.9% of all chemicals to which humans are exposed (mainly through food) are of natural origin. The amounts of synthetic chemical pesticide residues in plant foods, for example, are extremely low compared to the amounts of natural ‘pesticides’ produced by plants themselves. Of all dietary pesticides that humans eat, 99.99% are natural: these are chemicals produced by plants to defend themselves against fungi, insects and other animal predators. Each plant produces a different array of such chemicals. On average, the Western diet includes roughly 5000–10 000 different natural pesticides and their breakdown products. Americans eat about 1500 mg of natural pesticides per person per day, which is about 10 000 times more than they consume of synthetic pesticide residues. Even though only a small proportion of natural pesticides have been tested for carcinogenicity, circa year 2003 half of those tested (38/72) have been found to be rodent carcinogens. For example, more than 1000 naturally occurring chemicals have been described in coffee: 30 have been tested and 21 have been found to be rodent carcinogens in high-dose tests. In a single cup of coffee, the natural chemicals that are rodent carcinogens are about equal in weight to an entire year's worth of synthetic pesticide residues that are rodent carcinogens, even though only 3% of the natural chemicals in roasted coffee have been adequately tested for carcinogenicity. Cooking of foods produces burnt materials – about 2000 mg per person per day – that also contain many rodent carcinogens. In contrast, the residues of 200 synthetic chemicals measured by the United States Federal Drug Administration, including the synthetic pesticides thought to be of greatest importance, average only about 0.09 mg per person per day. Naturally occurring pesticides that are rodent carcinogens are ubiquitous in fruits, vegetables, herbs, and spices. In summation, it is probable that almost every fruit and vegetable in the supermarket contains natural pesticides that are rodent carcinogens, that no diet can be free of chemicals identified as carcinogens in high-dose rodent tests, and that exposure to synthetic rodent carcinogens is small compared to the natural background of rodent carcinogens.
There are numerous epidemiological studies supporting the likelihood that nutritional preemption is a useful strategy to reduce cancer risk at multiple body sites, with the greatest protection attributable to greater fruit and vegetable consumption. For example, both the international review panel of the World Cancer Research Fund – American Institute for Cancer Research (1997), and the Chief Medical Officer's Committee on Medical Aspects of Food and Nutrition Policy (COMA), British Department of Health (1998) recommend fruit and vegetable consumption to combat cancer. Epidemiological and other studies supporting the protective role of fruits and vegetables against cancer have been reviewed by Hayes (2005), including prospective epidemiological studies of atomic-bomb survivors that show significant and convincing protection against various cancers (Sauvaget et al., 2003, 2004).
What accounts for the paradox that plant foods contain carcinogens and also protect against cancer? Hormesis has been alluded to as an explanation by Parsons (2000), Rico (2002), and Johnson and Bruunsgaard (1998). It has been suggested that the plant diet brings together many different toxic chemicals which when ingested at low doses stimulate the chemo-defense system and enhance host resistance, while the chemo-defense system is overwhelmed at higher doses. It has been further proposed that the toxic phytochemicals in fruits and vegetables that protect against insects and other predators may induce a mild hormetic stress response in cells (Mattson, 2005). Induction of mild cellular responses have been cursorily invoked to explain the hormetic actions of the phytochemicals isothiocyanates, present in broccoli, and quercetin, present in red grape seeds (Faulkner et al., 1998; Blardi et al., 1999).
As of this date little detailed explanations of the hormetic effects of natural dietary pesticides have been offered, although there is a wealth of available information regarding the efficacy of dietary nutrients as anticarcinogenic agents. Circa year 2002, more than 500 dietary compounds have been identified as potential modifiers of the cancer process, including allium compounds, carotenoids, coumarins, dietary fiber, dithiolthiones, flavonoids, folate, glucosinolates, indoles, inositol hexaphosphate, isoflavones, isothiocyanates, limonene, plant sterols, phytosterois, protease inhibitors, saponins, selenium, and Vitamins C (ascorbate) and E (tocopherols) (Potter and Steinmetz, 1996; Safe et al., 1999; Milner, 2002). Some of the specific mechanisms by which nutrients may suppress the cancer process include the following: inhibiting genetic damage caused by exogenous agents by inhibiting carcinogen uptake, retarding activation, enhancing detoxification, scavenging oxygen radicals, and preventing DNA binding; influencing the repair of structural/functional genetic defects by enhancing endogenous repair and restoring proper methylation; eliminating damaged cells or clones by inducing apoptosis, promoting differentiation and enhancing immunosurveillance; and, suppressing growth and clonal evolution by slowing or stopping proliferation, retarding angiogenesis and inhibiting invasion (Potter and Steinmetz, 1996; Steinmetz and Potter, 1996; Milner, 2002).
Hormetic herbicides: dioxin and phosfon
Seventy-five phenoxy acid compounds constitute the group labeled dioxins (polychlorinated dibenzo-p-dioxins). The most acutely toxic of the polychlorinated dioxin isomers is considered to be 2,3,7,8-tetrachlorodibenzo-p-dioxin, or TCDD. A stable compound, dioxin breaks down slowly in the environment when protected from the ultraviolet rays of the sun, but once it finds its way into the soil it persists for a long time, remaining tightly bound to soil particles. Since dioxin is highly insoluble in water, it tends not to migrate very fast in soil. Dioxins are not manufactured directly for use in any product. Rather, they are formed as unwanted byproducts in the production of herbicides such as 2,4,5-T, in wood preservatives made from chlorophenols, in the incomplete combustion of wood products and industrial and municipal wastes, and in the combustion of diesel fuel (Kociba and Schwetz, 1982).
Dioxin is highly toxic to some species of animals (while less so to others), and at high doses an animal carcinogen. The largest, longest, and most detailed laboratory dioxin experiment entailed a 2-year TCDD study of 485 white rats carried by Kociba et al. (1978). This study formed the basis of the United States Environmental Protection Agency (EPA) declaring dioxin a human carcinogen, with aspects of this study having been criticized by Keenan et al. (1991) and Paustenbach et al. (1991). But the Kociba study also showed that dioxin has an anticancer effect at low doses, reducing tumor incidences of the pituary, uterus, mammary glands, pancreas and adrenals. At these dioxin levels striking U-shaped dose–response relationships were shown when all tumors were combined; and at the lowest dose the females displayed a modest decrease in liver tumors, the critical target organ for the EPA cancer risk assessment (Cook, 1994). Complementary laboratory evidence for hormesis is provided by the report that TCDD in Sprague–Dawley rats affected cell-mediated immunity by producing a striking inverted U-shaped response curve, in that low doses enhanced and high doses suppressed this immune function and that humoral-mediated immunity might also be stimulated by TCDD (Fan et al., 1996).
Since dioxin is found in trace amounts in some herbicides, autoexhaust and the incineration process, people could be exposed to it just by eating meat, fish, eggs or diary products. There is strong epidemiological evidence supporting hormetic effects. The herbicide 2,4,5-T contains the potent dioxin toxin TCDD. In the Vietnam War, Agent Orange was the military code name for a mixture of the dioxin herbicide 2,4,5-T and the nondioxin herbicide 2,4-D (which itself evinces hormesis in oyster growth (Davis and Hidu, 1969)). There were some 1200 United States servicemen who as members of the ‘Ranch Hand’ team were heavily exposed to Agent Orange, mixing and loading it, spending hours in its spray mist and accumulating it on their clothing. Those servicemen exposed to both the Southeast Asia environment and dioxin were found to have significant reductions of elevated systemic cancer and total skin cancers compared to their colleagues only exposed to the Southeast Asia environment (termed Comparisons) who showed increases in systemic, melanoma and total skin cancers (Kayajanian, 2000). For Ranch Hand veterans there is a significant 50% reduction in total cancers at the highest dioxin body burden levels compared to low-level background body burden (Kayajanian, 2001). These results are consistent with the failure to detect prostate cancer increase of the Ranch Hand group vis-à-vis Comparisons in association with either monitored TCDD levels or time served in Southeast Asia (Pavuk et al., 2006). The latter study suggested that a longer service in the Southeast Asia environment and exposures other than TCDD may have increased the risk of prostate cancer in Comparison veterans.
Other epidemiological studies are also pertinent. Of especial interest are some recent case–control studies that report J-shaped soft tissue sarcoma dose–response curves for low body burdens of dioxin and three polychlorinated biphenyl compounds (Tuomisto et al., 2004, 2005). These results were reported to support similar J-shaped dose–responses in animal studies, and were cited as possible examples of hormesis. Long-term monitoring of victims of major dioxin industrial accidents at Saveso, Italy in 1976 and at Monsanto Chemical's Nitro, West Virginia plant in 1949 failed to reveal any deleterious effects other than temporary chloracne (a severe acnelike skin disorder), and short-term reversible nerve dysfunction (Cole et al., 2003). A long-term mortality study of a cohort of some 2000 male chemical production workers previously exposed to substantial dioxin levels (some substantial enough to result in chloracne) found no coherent evidence of increased cancer (Bodner et al., 2003). From analyses of three occupational cohorts, Starr (2003) reported zero additional cancer deaths from all exposures to dioxin-like compounds including those arising from dietary intake (whereas Crump et al. (2003) from analyses of essentially the same data concluded that dioxin toxic equivalent exposures within roughly three-fold of current background levels may be carcinogenic). Chemical plant worker mortality data underlying the 1990 National Institute of Occupational Safety and Health (NIOSH) Report on Dioxin has been used by Kayajanian (1999) to identify tissue/organ/system sites at which dioxin exposure caused or was associated with a reduction in mortality. Kayajanian (2002) reported that the NIOSH data showed a hormetic J-shaped dioxin dose–response curve, and also cited evidence that the Saveso cancer data showed hormetic J-shaped responses.
The synthetic plant growth inhibitor phosfon (2,4-dichlorobenzyl tributyl phosphonium chloride) exhibits a hormetic inverted U-shaped dose–response relationship on the growth of peppermint, Mentha piperita (Calabrese, 1999; Calabrese and Baldwin, 2003b; Kaiser, 2003). Growth stimulation occurred at lower phosfon doses with diminution of the stimulatory response at higher doses, leading to marked inhibitory growth depending on the specific range of doses employed. This growth-pattern hormesis was attributed to overcompensation to an initial disruption in homeostasis.
Acrylamide is an industrial chemical also found in cigarette smoke. Acrylamide also occurs as a natural product of cooking, rather than as a food contaminant, as a result of a Maillard chemical reaction between reducing sugars and specific amino acids (e.g., asparagines within foods upon exposure to high heat). Recently, relatively high levels of acrylamide were unexpectedly detected in widely consumed food items, notably French fries, potato chips and bread. Acrylamide is a known human neurotoxin. It has been classified as a group 2A carcinogen (‘probably carcinogenic to humans’) by the International Agency for Research on Cancer (although the data suggesting that acrylamide may cause cancer in humans is derived from only one strain of one animal species). This has sparked intensive investigations regarding its occurrence, chemistry and toxicology/potential health risk in the human diet.
Retrospective studies on the association of cancer incidence and dietary acrylamide in Sweden (Mucci et al., 2003a, Mucci et al., 2003b, 2004, 2005) and in Italy/Switzerland (Pelucchi et al., 2003, 2006) could not provide evidence for an association between high and low acrylamide intake and cancer incidence of various organs. A hormetic effect was reported by Mucci et al. (2003a): ‘unexpectedly, an inverse trend was found for large bowel cancer with a 40% reduced risk in the highest compared to lowest quartiles of known acrylamide intake,’ while Mucci et al. (2003b) reported decreased risk of colorectal and kidney cancers with increasing acrylamide dose. Pelucchi et al. (2003) found no evidence of an increased risk of cancer with higher fried potatoes consumption, an important source of dietary acrylamide, and also confirmed the inverse association with large bowel cancer. A large prospective study found no evidence that dietary intake of acrylamide is associated with cancers of the colon or rectum (Mucci et al., 2006). In addition, a study of some 9000 workers exposed to acrylamide between the years 1925 and 1976 found a statistically significant decrease in deaths from all causes (Collins et al., 1989). A follow-up study of these workers through year 1994 corroborated many of the initial findings (Marsh et al., 1999).
As already noted, stress has been suggested as a hormesis-inducing agent and has been invoked to explain hormesis arising from dietary restriction, natural dietary pesticides and ethanol consumption. While specific hormetic inducing mechanisms have been advanced, prominent among them being stress protein induction; there is now less expectation that any one single molecular mechanism can explain hormetic effects. This judgment arises from the fact that hormetic effects have been observed for a broad range of species, substances and biological end points (such as growth, longevity, reproduction, immune response, and many physiological and metabolic responses). In fact, there appear to be numerous ways in which biological systems can manifest hormetic biphasic dose–response relationships. What is seen among the array of particular hormetic mechanisms is a general commonality in quantitative features of hormetic dose–responses. The usual modest amplitudes of hormetic responses suggest that hormesis operates within a framework designed to conserve resources. Such a limited, yet efficient, use of resources is consistent with the hypothesis that hormesis is a modest overcompensation to the disruption of homeostasis, a concept proposed more than a century ago by Townsend (1899) and later experimentally supported by Branham (1929). (Arguments have been advanced by Rattan (2001) to replace the term homeostasis with the term homeodynamics to reflect the dynamic nature of living processes in an ever-changing lifeline.) This overcompensation (other terms in the literature being over-shoot and rebound) is believed to be the manifestation of a broad biological strategy, with the specific mechanisms unique to each system being simply biological tactics. Linked in this way to a control system designed to restore homeostasis, only modest overcompensation would be predicted, and this has proven to be the case. This overcompensation not only negates the negative effects, but also offers protection against its later recurrence.
Stebbing had proposed that the disturbance of homeostasis is the driving force behind hormetic response, with organisms having biochemical and physiological control mechanisms that regulate internal processes and respond to external disturbances such as that caused by external stress or damage (Stebbing, 1982, 2003). After the homeostasis compensatory response neutralizes the disturbance and restores equilibrium, the fitness of the individual is optimized to respond to subsequent, more serious, challenges. However, high levels of stress or damage disrupt the organism beyond its limits of recovery, causing irreparable damage. Although the specific details of the hormetic mechanism may differ depending on how the organism is being challenged; the overall strategy is the same, the maintenance of constant internal state to ensure an organism's efficient functioning and performance (i.e., homeostasis) that in turn produces biphasic dose–response relationships.
An alternative explanation involves direct response to stress in the receptor systems that control the broad range of critical processes underlying most organismal activities. Hormetic biphasic dose–response relationships acting via the receptor-based mechanism have been explained from having two receptor subtypes with markedly different agonist affinities, with different receptor subtypes existing in different quantities, having different affinities for binding an agonist, and having different effects that lead to either a stimulatory or inhibitory pathway (Calabrese, 2004). At low concentrations with small amounts of the agonist present, a high-affinity receptor would be responsible for the dominant response. However, it is postulated that low-affinity receptors exist in greater quantities. So as the concentration increases, the number of occupied low-affinity receptors will eventually outnumber occupied high-affinity receptors and begin to dominate the response. Thus, low levels of a compound can have the opposite effect of high levels of the same compound. As a result of the combined presence of both receptor subtypes, it would then be possible to account for hormetic biphasic dose–response relationships.
This review has considered the current and historical background of hormesis, the concept that small quantities have opposite effects from large quantities. Hormetic dose–response relationships were reviewed, categorized, and depicted vis-à-vis other dose–response relationships. Evidence for biphasic dose responses was presented for essential vitamin and mineral nutrients, dietary restriction, ethanol consumption, natural dietary and some synthetic pesticides, some herbicides and acrylamide. Some of the evidence is quite extensive. For example, there are at least 19 separate reports of alcohol's impact on mortality being biphasic J-shaped; while the evidence for the paradoxical effects of fruits and vegetables are also compelling. The hormesis paradigm appears to account for and explain many paradoxical nutritional effects in a logical and consistent manner.
Those engaged in nutritional research and in formulating nutritional guidelines should be cognizant of the following facts. As discussed, hormetic effects are relatively modest. Although modest, such effects can be quite important, for example, in molecular pharmacology where concerted low-level measurements have yielded a treasure trove of hormetic effects (Calabrese, 2005c). Any consideration of hormesis should also take into account that it may be a two-edged sword. To wit, hormetic effects with reduction of dose do not necessarily induce subjectively salutary changes. On the contrary, account must be taken of the fact that hormesis, if present, could induce subjectively negative effects. For example, some chemotherapeutics used to treat tumors at high concentrations enhance tumor growth at lower concentrations. Calabrese et al. (2006) have detailed this potential problem in the field of drug development – one scenario the case where low-dosage represents an adverse/unwanted effect (e.g., tumor cell proliferation by antitumor drugs), the obverse scenario the case where low-dosage defines a beneficial/intended therapeutic zone (e.g., cognition enhancement in Alzheimer's disease treatment). Another caveat in formulating and following nutritional guidelines is the possible duality of hormetic end point effects, with different end points displaying either positive or deleterious effects at the same dose levels. For example, in laboratory studies, low doses of cadmium help prevent some cancers while also acting as an endocrine disrupter that could lead to cancer (Kaiser, 2003). Nevertheless these caveats should not detract from the fact that in toto the hormesis paradigm holds great promise in the field of nutrition and nutritionists should be aware of the possibilities of hormesis, both pro and con, both in their research and their formulation of nutritional guidelines.
Proponents of hormesis believe it to be commonly manifested in both biology and toxicology, and highly generalizable according to biological models tested, end points measured, and chemical class/physical agents employed. In view of the evidence presented here it is suggested that nutritionists and those engaged in nutritional studies render nutritional hormesis serious consideration.
I wish to acknowledge the stakhanovite efforts and services of the staff of the William Hallock Park Memorial Public Health Library of the New York City Department of Health and Mental Health Hygiene and the technical assistance of my colleague Raymond Ford. I have no conflicts of interest that are either directly or indirectly relevant to the content of this article.
About this article
Critical Reviews in Food Science and Nutrition (2017)