Response to the Commentaries on "Does Caloric Restriction Induce Hormesis?"

Angelo Turturro Ph.D., D.A.B.T. *

Bruce S. Hass, Ph.D.

Ronald W. Hart, Ph.D.

* Corresponding Author: Division of Biometry and Risk Assessment, National Center for Toxicological Research (NCTR),

Food and Drug Administration, Jefferson, AR 72079

Tel: 870-543-7340, Fax: 870-543-7332

Email: aturturro@nctr.fda.gov

Division of Genetic and Reproductive Toxicology

Tel: 870-543-7365, Fax: 870-543-7065

Email: bhass@nctr.fda.gov

Office of the Director

Tel: 870-543-7000, Fax: 870-543-7332

Email: rhart@nctr.fda.gov



First, we wish to thank all the respondents for their comments. They have added immeasurably to our understanding of the issues.

We thank George Gray for his gracious comments. We agree with his elaboration of some of the critical issues in the relationship of agent-induced body weight (BW) depression and caloric restriction (CR). We also concur with his call for more work to clarify these issues.

Sprott repeats the argument, championed by Keenan and his co-workers (Keenan and Soper, 1995), that ad libitum (AL) feeding for rodents is producing a sick animal, much as infectious diseases once ravaged animal colonies. We agree with the general theme of his discussion, and have in fact proposed a cost-effective technique of dietary control to address this (Turturro et al., 1995). However, we have also shown that the effects of BW are on a continuum, with a non-linear dose-response curve (Turturro et al., 1993-1998a, 1998b). Thus CR does not simply avoid "obese" AL-fed animals, but illustrates that the growth rate of the animal is an important factor influencing aging and toxicity studies.

Wilson's analysis, profound as always, has clearly identified one of the major issues in the debate about chronic tests, i.e., the need to minimize unwanted "noise" by the control of BW. The questions he raises about the need to define malnutrition are good ones. We would operationally define malnutrition as a deficiency in a food element that leads to the compromise of some essential bodily subsystem (e.g., decrease in immune function) or an increase in disease. The effective nutritive of the animal could be evaluated through monitoring of internal nutritional state (e.g., blood levels of minerals or water-soluble vitamins), although this does not guarantee the adequacy of every trace element co-factor. Fortunately, there is some guidance in this area (NRC, 1995).

Keenan's discussion illustrates a number of points about CR in detail. We agree with most of his conclusions, and it was for reasons such as these that we elected to use a BW marker.
As to whether some CR paradigms may not influence the incidences of certain tumors in a toxicity experiment, we feel that without dietary control there is no definitive way to compare different experiments (Turturro et al., 1995). For instance, for mammary tumors:

1) The inhibitory effects of CR on mammary tumors has been repeatedly demonstrated in mice and rats (e.g., Weindruch and Walford, 1988; Sheldon et al., 1996);

2) We, of course, have demonstrated that the incidence these tumors is a function of BW at a certain stage in development (Turturro et al., 1998a).

3) The mechanism of this effect may be through the inhibition of prolactin secretion (Engleman et al., 1993).

Whether the prolactin inhibition that occurs at a given BW is sufficient to produce an observable change in tumor occurrence is dependent on a number of factors. These include: the BW at which the study is started (i.e., initial prolactin levels), the method used to evaluate tumor incidence, the level of spontaneous tumor occurrence, the duration of the study, etc. For example, if the initial levels of prolactin are high enough to be in the range where there is little change in the tumor response with changes in dose (i.e., similar to saturation), halving the prolactin

level may not produce a noticeable effect on tumor incidence. In trying to identify the impact CR has on any given effect, the paradigm should be viewed in the context of a dose-response curve. Appropriate use of a marker, while seldom done, can help accomplish this.

We are grateful to the efforts of Keenan and his colleagues in providing details of their model. We also appreciate the usefulness of their data in developing a standard model for the genotype they use.

We are not as concerned about general inconsistencies in the definitions of the components of hormesis as Boxenbaum appears to be. We feel there are operational definitions of most of the terms used in the definitions, e.g., growth, proliferation, so that we can make some meaningful general statements about the process. However, Boxenbaum is right in asking for precision in this area, which is why we discussed at length the various definitions of hormesis.

Stevenson and Sielken first addressed the issue of the relevance of CR results in rodents and other species to humans based on the proposal cited in the paper that the calories/unit weight was constant in a lifespan. This has been shown not to be true for CR (Yu et al., 1985; Duffy et al., 1989; McCarter and McGee, 1989; McCarter and Palmer, 1992). These studies clearly showed that CR doesn't effect metabolic rate in the long term and that, since the animals live longer, total caloric use is higher.

In the Stevenson and Sielken discussion of the definitions of hormesis, we suggest that the definitions used in our paper are in general easier to evaluate operationally. We feel that terms such as "generally considered to be detrimental" or "cannot be anticipated" are subjective. They depend on the imagination and expertise of the person evaluating the data. On the other hand, a non-monotonic dose-response curve is operationally fairly well defined.

Finally, we do not understand the reason for the exclusion of "essential agents", especially when mortality is the endpoint of concern. At zero dose whether all the animals in a cohort die (an essential), or only more than the number at some low dose, simply alters the height of one arm of the non-monotonic dose-response curve.

The discussion concerning mortality curves emphasizes the Gompertz-Markham plots for humans. We do not know how to interpret the significance of changes such as these. We have already reported the effects of CR using a Gompertz analysis (Turturro and Hart, 1992) and found no consistent effect with DR. We did mention that this inconsistency is not easy to notice if only animals within the same strain/diet combination are compared. Also, it is not clear how stable the results of this type of analysis are. For instance, Figure 1 gives the survival curve of male B6D2F1 mice (Turturro et al., in press).

Figure 1. Mortality of AL and CR Fed Male B6D2F1 Mice. Survival in % and age in days. For details see Turturro et al., 1999.

Figure 2 is a Gompertzian analysis of these data (using Sacher, 1977, with the addition of an arbitrary constant 12 added to the abscissa to avoid negative numbers).

Figure 2. Compertzian Analysis Using AL Animals to 868 Days of Age With AL and CR. 12 is added to abscissa to avoid negative numbers. Age in days. Dotted lines are extrapolations outside the data range.

The data from AL animals living longer than 868 days of age are excluded (The dotted lines are extrapolations from the observed range). The slopes of the two lines are different, but the intercept is close to the same. If the data from animals up to 1148 days of age are included, we derive Figure 3, which indicates that the slopes are the same, but the intercepts are different.

Figure 3. Gompertzian Analysis Using Al Animals to 1148 Days of Age With Al and CR. 12 is added to abscissa to avoid negative numbers. Age in days. Dotted lines are extrapolations outside the data range.

This occurs because there is a slowing of the death rate near the end of the lifespan in the AL fed animals, coincident with the emergence of a subset of long-lived low BW animals. As this subset is included, it alters the slope and intercept of the curve. There does not appear to be an equivalent subset in the CR animals, presumably because the variability in BW of the CR animals is kept to a minimum (Turturro et al., 1998b). We favor some of the improvements in this form of analysis that Finch has recommended (Finch, 1994).

One speculation that arises out of his analysis is whether CR truly slows the aging process or whether it simply reduces the mean and standard deviation of the population to that of the small minority that exhibit maximum achievable lifespan. This does not settle the questions of why the small subset lives longer than normal and the significance of the adverse effects the paradigm has on reproduction.

We do appreciate the efforts of modeling detrimental and beneficial aspects of agents performed by Sielken and Stevenson, and feel that models such as these may eventually be helpful in "tallying up" the ultimate result of the application of a complex stimulus, such as CR, to an organism.

Many of the points discussed for Stevenson and Sielken are also relevant to the commentary of Neafsey. We appreciate her work on hormesis and CR, but feel that her conclusions are bounded by her CR dataset. Additionally, the analyses for the chemicals that are presented are puzzling. Before the first death in a chronic study (and in regions where the death rate is zero) a Gompertzian function has no meaning, since the logartihm of zero is undefined. The Gompertzian value at zero time is a theoretical construct, i.e., the intercept with the abscissa assuming a straight line. Most analyses are limited to the region of the mortality curve where the change in the death rate with time is approximately exponential. They almost always exclude the region near the origin, even in large-scale human studies, as was done by Sacher (1977) who excluded information before ten years of age (for humans). The death rates that occur in that region are usually decreasing from an initial high perinatal level (not increasing exponentially). In our figures, we are able to present data closer to the origin than usual since: 1) the starting time for data collection of mortality is post-weaning; and 2) the mortality data represent approximately 6100 male mice, almost equally distributed between AL and DR. Thus, we are past the perinatal age at the start of the experiment, and we have data from early timepoints because of the large number of animals on the study.

Since the death rates in the relatively small number of animals used in a chronic study are zero in the early stages of the study, we would not expect to explain the behavior of a curve in a region where it is undefined. Neafsey's curve-fitting approach uses assumptions to fill these regions. We have learned from our experiences in attempting to model carcinogenesis results that extrapolation into non-observable regions should be carried out with an appreciation of how the assumptions used drive the process (Turturro and Hart, 1987). Interestingly, in regions that are defined, the plots in her Fig. 2 look like our Fig. 3, and the plots in her Fig. 3 look like our Fig. 2. The plots in her Fig. 4 look similar to the plots of our data if all the AL animals are included (not shown).

Given the limitations of the Gompertzian analysis, especially the sensitivity of the approach to changes in subpopulations in the mortality curve, we feel that attaching significance to any differences in the undefined region of the curve is extremely speculative.

We appreciate that Neafsey wished to address the problem of uncontrolled food consumption in bioassays. However, we feel the technique of pair-feeding for chronic tests is not practical because of high cost and logistic problems. In addition, pair-feeding may not address the issues of differences in BW, which we have found to be the key to understanding much of the variability in chronic tests. An animal exposed to a toxicant, such as amphetamine may eat the same amount as pair-fed animal but may weigh less, and the tumor incidences found in the controls in the study reflected the BW, not the food consumption (Turturro et al., 1998b). We prefer the use of dietary control (Turturro et al., 1995).

Masoro addresses the question of the relationship of stress to hormesis. We have some difficulty accepting his arguments about stress since it is not clear to us what exactly he means. Does he mean to include toxic stress (i.e., second-hand smoke [to the lung]; hydrazines in mushrooms [to the liver]), disease stress (viral, bacterial, protozoid, etc.), thermal stress, etc.? How do these stressors combine: synergistically, additively, antagonis-tically, or variably depending on circumstances? As Stevenson and Sielken and we have all pointed out, the time of the application of CR is important. Does the age at which the stressor act determine whether the level of stress is low enough to be in the hormetic range or how it combines with other stresses?

Based on our analyses of CR, exercise, and thermal stress discussed in the paper, it appears that each agent has positive and negative aspects, for different organs at different intensities and at different ages. Combining them in a way to maximize the beneficial effects and minimize the negative ones is not simple. One has to be careful to not over extrapolate from our experience with one of the overall most benign stressful paradigms, CR, to ignore the potent detrimental effects associated with many of the other stressors in daily life. CR may be unique as a stressor in not elevating levels of macromolecular damage in the tissue (such as increases in the products of oxidative stress) as we have discussed previously (e.g., Hart et al., 1992).

Kirkwood and Shanley have discussed the evolutionary aspects of CR. We think the analysis, though thoughtful, has missed a major point. As discussed in Hart and Turturro (1998) and Turturro et al. (1999), based on our observations of the activity patterns of CR animals we have noted the burst of food acquisition and combative activity near feeding time in the CR animal. We feel that this burst of activity is an important component in the organism's response to CR that probably does not confer benefits for the long-term survival of the animal but is instead a mechanism to maximize the chance in the short-term of obtaining food during a shortage. In this approach, CR is pleiotropic, with the necessity of obtaining food for immediate survival outweighing the long-term negative effects of a daily increase in activity. Thus the positive effects of CR are mitigated by some of the negative consequences of this requirement, with the overall balance effecting the endpoint observed as we discussed in our paper. We feel this approach is more consistent with the complex nature of CR and its myriad effects on the organism.

The discussion in Parsons is a subtle application of the principles of natural selection to make the case that we have evolved an optimal phenotype and deviations from the requirements of that phenotype leads to hormesis. In essence, all agents in the environment are in some sense positive since, in their absence, the animal would not be behaving optimally. Although seemingly compelling, we think this approach underestimates the capacity of an organism to maintain homeostasis, avoid unpleasant stimuli, and change behavior. There are many ways to address the problems associated with a stressor in the environment, the simplest being avoidance. For food shortage, one of the simplest approaches is food acquisition behavior (as discussed above). Finally, although Parsons discusses fitness in terms of the animal being best adapted for survival in a particular niche, we feel that natural selection does not maximize survival, either individual or group. We have suggested that improved survival is simply an accidental result of improving the genomic protection and increased redundancy (Hart and Turturro, 1998). What is selected for is reproductive fitness. For example, in species that die after they breed such as Pacific salmon (Robertson, 1961), we believe animals will continue to be selected for factors increasing the ability to reproduce, e.g., to swim upstream, not for the capacity to delay breeding.

REFERENCES

Duffy, P., Feuers, R., Leakey, J., Nakamura, K., Turturro, A., and Hart, R. Effect of chronic caloric restriction on physiological variables that modulate energy metabolism in the male Fischer-344 rat. Mech. Age. Devel., 48:117-133, 1989.

Engleman, R.W., Day, N.K., Good, R.A. Calories, parity and prolactin influence mammary epithelial kinetics, and differentiation and alter other mouse mammary tumor risk. Cancer Res., 53:1188-1194, 1993.

Finch, C. Longevity, Senescene and the Genome. U. Chicago Press; Chicago, IL, 1994.

Hart, R., Chou, M., Feuers, R., Leakey, J., Duffy, P., Lyn-Cook, B., Turturro, A., and Allaben, W. Caloric restriction and chemical toxicity/carcinogenesis. Quality Assurance: Good Practice, Regulation, and Law, 1:120-131, 1992.

Hart, R.W., and Turturro, A. Evolution and dietary restriction. Exp. Geront., 33:53-60, 1998.

Keenan, K., and Soper, K. The effects of overfeeding and moderate dietary restriction on Sprague-Dawley rat survival, spontaneous carcinogenesis, chronic disease and the toxicologic response to pharmaceuticals. In: Dietary Restriction; Implications for the Design and Interpretation of Toxicity and Carcinogenicity Studies (R Hart, D. Neuman, R. Robertson, Eds.). Wash., DC: ILSI Press, 99-126, 1995.

McCarter, R., and McGee, J. Transient reduction of metabolic rate by food restriction. Am. J. Physiol., 257:E174-179, 1989.

McCarter, R., and Palmer, J. Energy metabolism and aging: a lifelong study of Fischer-344 rats. Am. J. Physiol., 263:E448-452, 1992.

NRC(National Research Council). Nutrient requirements of Laboratory Animals. 4th Rev. Ed. Wash., DC: National Academy Press, 1995.

Robertson, O. Prolongation of the life span of kokanee salmon (Oncorhynchus nerka kennerlyi) by castration before the beginning of gonad development. Proc. Natl. Acad. Sci., 47:609-621, 1961.

Sacher, G. Life table modification and life prolongation. In: Handbook of the Biology of aging (C. Finch and L. Hayflick, editors). Van Nostrand, New York, pp. 582-638, 1977.

Sheldon, W., Bucci, T., Blackwell, B., and Turturro, A. Effect of ad libitum feeding and 40% feed restriction on body weight, longevity, and neoplasms in B6C3F1, C57Bl6, and B6D2F1 mice. In: Pathobiology of the Aging Mouse, Vol. 1 (U. Mohr, D. Dungworth, C. Capen, W. Carlton, J. Sundberg, and J. Ward, Editors). ILSI Press, Wash. DC, pp. 21-26, 1996.

Turturro, A., and Hart, R. Quantifying risk and accuracy in risk assessment: The process and its role in risk management problem solving. Med. Oncol. and Tumor Parmacother., 4:125-132, 1987.

Turturro, A., and Hart, R. Dietary alteration in the rate of cancer and aging. Exper. Gerontol., 27:583-592, 1992.

Turturro, A., Duffy, P., and Hart, R. Modulation of toxicity by diet and dietary macronutrient restriction. Mut. Res., 295:151-164, 1993.

Turturro, A., Blank, K., Murasko, D., and Hart, R. Mechanisms of caloric restriction effecting aging and disease. Annals N.Y. Acad. Sci., 719:159-170, 1994.

Turturro, A., Duffy, P., and Hart, R. The effect of caloric modulation on toxicity studies. In: Dietary Restriction: Implications for the Design and Interpretation of Toxicity and Carcinogenicity Studies (R. Hart, D. Neuman, and R. Robertson, Editors) ILSI Press, Wash. DC, pp. 79-98, 1995.

Turturro, A., Duffy, P., Hart, R., and Allaben, W. Rationale for the use of dietary control in toxicity studies ­ B6C3F1 mouse. Toxicol. Pathol., 24:769-775, 1996.

Turturro, A., Leakey, J., Allaben, W., and Hart, R. Letter to the Editor, Response to Michael Festing's "Fat Rats". Nature, 389:326, 1997.

Turturro, A., Hass, B., Hart, R., and Allaben, W. Body weight impact on spontaneous and agent-induced diseases in chronic bioassays. Inter. J. Toxicol., 17:79-100, 1998.

Turturro, A., Hass, B., and Hart, R. Hormesis ­ Implications for risk assessment caloric intake (body weight) as an exemplar. Human Ecol. Risk. Assmnt., 17:454-499, 1998a.

Turturro, A., Duffy, P., and Hart, R. Antioxidation and evolution: Dietary restriction and alterations in molecular processes. In: Antioxidants in Human Health and Disease (T. Basu, N. Temple, and N. Garg, Editors). CAB International, Oxford, Eng., pp. 83-94, 1999.

Turturro, A., Witt, W., Lewis, S., Hass, B.S., Lipman, R.D., and Hart, R. Growth curves and survival characteristics of the animals used in the biomarkers of aging program. J. Gerontol. 54A:B492-B501, 1999.

Weindruch, R., and Walford, R. The Retardation of Aging and Disease by Dietary Restriction. Springfield, IL: C.C. Thomas, 1988.

Yu, B.P., Masoro, E., McMahan, A. Nutritional influences on aging of Fischer-344 rats. I. Physical, metabolic, and longevity characteristics. J. Geront., 40:657-670, 1985.