Perhaps the most important consequence, historically, of a failure to appreciate the implications of evolutionary biology for the biomedical sciences, lies with the importance that evolutionary biologists place on variation, both within and between evolving populations of organisms. As noted by Burggrem and Bemis:
While comparative physiologists have made an art of avoiding the study of variation, such heritable variation nonetheless is the source of evolutionary changes in physiology as well as for all other types of characters [:201].
Ignoring interspecific differences and intraspecific variation, there has been a historic trend for comparative physiologists to revert to pre-evolutionary typological thinking involving a focus on paradigm "model" species. Again, as observed by Burggren and Bemis:
Yet the use of "cockroach as insect," "frog as amphibian," or "the turtle as reptile" persists, in spite of clear evidence of the dangers of this approach. Not surprisingly, this type of comparative physiology has neither contributed much to evolutionary theories nor drawn upon them to formulate and test hypotheses in evolutionary physiology [:206].
These problems can also be illustrated by a consideration of the importance of interspecific variation, intraspecific variation and gene regulation in the context of pharmacology and toxicology – where the focus is all too often on "mouse or rat as mammal" – and, in particular, as "human being."
In the United States, 14 to 16 million animals are used in biomedical research each year. The vast majority of mammals (85 to 90 %) employed in such research, aimed at benefiting humans, are rodents . Primate species are not a significant part of the total partly because they are difficult and expensive to house, and partly because, in the case of such species as gorillas, chimpanzees, and orangutans, they are close to extinction.
How then, is evolutionary biology relevant to a discussion of the use of animals in biomedical research aimed at benefiting humans? In terms of the pattern of evolutionary relationships, the line leading to modern humans seems to have diverged from the lineage leading to modern rodents about 70 million years ago, thus representing an accrual of some140 million years of independent evolution. The lineage leading to modern mice seems to have diverged from that leading to modern rats some 17 million years ago. It is quite easy to conclude that rats and mice are more closely related to each other than either is to humans.
From a genetic point of view, the human genome project has revealed that the human genome consists of some 30,000 genes. The mouse genome is about the same size as the human genome . Moreover, reflecting common ancestry, counterparts (or "orthologs") of many human genes have been identified in both mice and rats (notwithstanding differences in chromosomal arrangement). From the standpoint of genetic "base-pair similarity," humans, rats, and mice are remarkably similar. But the devil of genetic differences between individuals of a species, or the genetic differences between members of different species, lies in the details.
Mammals are diploid organisms, which means they have two sets of chromosomes, one from each parent. Such chromosomes in a diploid individual are said to be homologous because that they have the same pattern of genes along the chromosome. The location of a given gene on a chromosome is known as its locus. For a given locus, different versions of a gene – as the alleles – may exist in an individual (limited to two versions) and/or in a population (two or more versions). Such allelic variation generates variation with respect to the genotypes found in a population, and is thus a major source of genetic polymorphisms.
Though each individual has two alleles at a given locus (one from each parent), a large population of such individuals may exhibit several (more than two) alleles for a given gene. The various relative frequencies of alleles may be computer for any population. Different alleles typically have different biological properties. When these properties influence the reproductive success of the organisms bearing them, with the effect that different organisms in the population leave behind different numbers of offspring, then evolution occurs – over successive generations, the frequencies with which given alleles are found in the population changes. Certainly, allele frequencies can change for other reasons too, but this need not concern us here.
The main implication of evolutionary biology for our inquiries is the uncontroversial observation that in natural populations (whether of mice or humans), there is typically variation with respect to the alleles that are present. But typical laboratory populations of (say) mice are represented by highly inbred strains or varieties. The value of an inbred strain is supposed to lie in its relative genetic homogeneity. The hope is that individuals belonging to such strains should respond similarly when similarly stimulated (perhaps with drugs or toxins). The use of highly inbred individuals is a way to control for the real genetic variation in natural populations which can confound the results and conclusions of laboratory experiments. Thus, the problem of interspecies extrapolation from rodents to humans (where there are genetic similarities, but not genetic identities) is exacerbated by the fact that human populations will often not only contain alleles very different from those in rodent populations (where similar genes can be identified), but will also typically exhibit allelic variation that is absent in the (homogeneous) laboratory rodent populations used to model them. The "model" is confounded both by lack of identical (or even similar) biological properties of alleles and by lack of overall genetic variation.
Now apply the use of our "model" to a consideration of the biomedical study of drug and toxin metabolism. The enzyme system that plays an important role in xenobiotic (drug and toxin) metabolism is the cytochrome P450 system. Some 500 different P450 enzymes have been characterized by description of their DNA sequences, and members of a given species may carry 40–50 of these different enzymes . For ease of reference, we will use "CYPs" as an abbreviation for the Cytochrome P450 enzyme group in the following discussion.
First, some terminology should be introduced. The CYPs represent a superfamily of genes and each gene (and its product enzyme) is named according to the similarity of its DNA sequence to other genes in the superfamily. The following example will help. Consider CYP 1A2. The first number designates the family the gene belongs to, which is determined on the basis of at least 40% sequence similarity. The letter following then designates a subfamily, determined on the basis of at least 59% sequence similarity. The last number identifies the specific gene (or enzyme). CYP 1A2 and CYP 3A4, for example, belong to different families within the CYP superfamily. By contrast, CYP 2C9 and CYP 2D6 belong to same family, but different subfamilies. We know also that each gene is composed of alleles, which may differ, so specific alleles are denoted by an asterisk and additional number. CYP 2D6*10 refers to a specific allelic variant (*10) of the CYP2D6 gene and so on.
Human intraspecific variation
Within biomedical sciences, we all too readily speak of mice and humans as if all mice and all humans were the same. For many reasons, this is an error from an evolutionary perspective. In our current example, human CYP polymorphisms can manifest themselves in the form of intraspecific (i.e., individual) differences in drug metabolism. Two genes, CYP 2D6 and CYP 2C19 are particularly important since they affect how people metabolize approximately 25% of the drugs on the market .
Sipes and Gandolfi  observed that with respect to the antihypertensive agent debrisoquine, some 3 to 10 percent of Caucasians are poor metabolizers because they are homozygous for 2 nonfunctional alleles for CYP 2D6, the gene source of debrisoquine 4-hydroxylase enzyme. There appear to be more than 75 allelic variants of CYP 2D6 circulating in human populations .
Among these 75 variants, frequencies of the distribution of alleles vary among different ethnic populations: for example, individuals homozygous for the *10 allele have low CYP 2D6 gene activity and are found in nearly 20% of the Japanese population – a figure that differs from both Caucasian and Chinese populations . Studies in molecular genetics indicate that actual cause of reduced activity of the CYP 2D6 gene is variable and complex. Causal factors range from single nucleotide polymorphisms in the coding sequences, to effective deletions of the gene itself, to polymorphisms that affect the splicing of CYP 2D6 . But on the other side of the coin, there are rapid metabolizers with high CYP 2D6 gene activity, related to the fact they possess duplicates of the gene (some with as many as thirteen copies). High metabolizers with high CYP 2D6 gene activity require more than the standard doses of drugs to achieve therapeutic responses. It should be obvious that these important human differences could never have been revealed by nonhuman animal studies.
Consider the metabolism of a specific drug, such as the antiepileptic drug mephenytoin. More than 20% of the Japanese population are poor metabolizers (compared to about 3% of the Caucasian population ). Enzymes in the CYP 2C subfamily have been shown to be responsible for mephenytoin metabolism, with CYP 2C19 responsible for the main enzyme, (S)-mephenytoin 4'-hydroxlase . Poor metabolizers appear to make a stable, but defective protein . The presence of CYP 2C19*2 and *3 alleles account for 99% of poor metabolizers within oriental populations and 87% of Caucasian poor metabolizers.
These examples represent only a minute sample of what is known about polymorphisms with respect to the specific enzymes and substrates (drugs, in this case) mentioned. But they highlight the importance of paying attention to intraspecific variation when considering metabolic activity. Partly for these reasons, Collins has recently pointed out:
In the field of metabolism, as well as some segments of toxicity and efficacy, there has been a major shift from animal-derived data to human-based data. Except for comparative studies to assess interspecies differences, animal studies have declined in importance. Part of this shift is driven be an appreciation for the uncertainty in cross-species metabolic pathways. From the practical side, the well-organized, readily available supply of human tissue has fueled this shift [:238].
The existence of intraspecific variation is but a foretaste of the biological problems confronting those who seek to use animals to model human biomedical phenomena. As Darwin observed in the Origin of Species:
As each species tends by its geometrical rate of reproduction to increase inordinately in number; and as the modified descendants of each species will be enabled to increase by as much as they become more diversified in habits and structure, so as to be able to seize on many and widely different places in the economy of nature, there will be a constant tendency of natural selection to preserve the most divergent offspring of any one species.
Hence, during a long continued course of modification, the slight differences characteristic of varieties of the same species, tend to be augmented into the greater differences characteristic of species of the same genus [:108].
In other words, one effect of evolutionary processes in the formation of new species, is essentially to amplify the differences that existed in the varieties belonging to the common ancestor from which the new species descend in the course of evolutionary time. Thus, further bad news lies in the fact that interspecific variation is likely to be even more of a problem for the animal modeler than the already confounding intraspecific variation we have just discussed.
Extrapolation between rodent species
As noted above, rats and mice are more closely related to each other than either is to humans. While intraspecific variation is important within rat and mouse (and human) populations – marked differences exist between different strains of mice and different strains of rats with respect to drug metabolism and susceptibility to diseases such as cancer. Interspecific extrapolation between rats and mice has proved to be no simple matter – rats are not particularly good models for mice! Thus as Hoffman has observed:
Correspondence between mouse and rat, the two most commonly used species in carcinogenicity tests, is not especially high. For 73 compounds evaluated by Tennant et al., the concordance between mouse and rat was 67%. Moreover, in an evaluative study by Griesemer and Cueto, only 44 of 98 agents that were carcinogenic in either rats or mice were carcinogenic in both species [:216].
Extrapolation from rodents to humans
There is an enormous literature on the problems associated with extrapolation from rodents to humans. We will briefly examine three examples to highlight the difficulties encountered in any such enterprise.
As aptly described in one of the leading textbooks in cell biology:
The mouse is the most widely used model organism for the study of cancer, yet the spectrum of cancers seen in mice differs dramatically from that seen in humans. The great majority of mouse cancers are sarcomas and leukemias, whereas more than 80% of human cancers are carcinomas – cancers of epithelia where rapid cell turnover occurs. Many therapies have been found to cure cancers in mice; but when the same treatments are tried in humans they usually fail [:1347].
There are 26 known human carcinogens (the list of probable carcinogens is somewhat longer). Of these 26 carcinogens, humans are exposed to seven by inhalation. Do carcinogenicity assays involving rodents convey information about human risk? Two decades ago, Salsburg observed of the 26 known human carcinogens:
Most of these compounds have been shown to cause cancer in some animal model. However, many of the successful animal models involve the production of injection site sarcomas or the use of species other than mice or rats. If we restrict attention to long-term feeding studies with mice or rats, only 7 of the 19 human non-inhalation carcinogens (36.8%) have been shown to cause cancer. If we consider long-term feeding or inhalation studies and examine all 26, only 12 (46.2%) have been shown to cause cancer in rats or mice after chronic exposure by feeding or inhalation. ... On the basis of probability theory we would have been better off to toss a coin [:64].
Should we be alarmed if a substance induces cancer in rats or mice? Probably not, especially in view of the fact that rodents have exhibited carcinogenic responses to 19 out of 20 substances suspected of being non-carcinogenic in humans . Thus, the data available today do not support the assumption that these particular animal "models" actually are models for human carcinogenicity studies.
Differences with respect to gene regulation may be illustrated by the following example. It has been shown that xenobiotics induce transcription of certain families of CYPs by activating nuclear receptors. CYP 3As, for example, are regulated by the pregnane X receptor (PXR). Studies have been performed on human and mouse orthologs of PXR. Moore et al. commented upon the results of these studies as follows:
However, comparison of PXR from four different species shows that this receptor has diverged considerably in the course of evolution. The human, rabbit and rodent PXR are all roughly equally divergent and share only ~70% amino acid identity. This divergence in PXR is an important component of cross-species differences in the regulation of CYP3A expression by xenobiotics [:15126].
Species differences are not just associated with evolution of the structures of CYP enzymes, they are also associated with evolution in the molecules that regulate the expressions (on, off, or actual amount) of the genes coding for those enzymes as well. The regulatory role of PXR is indeed medically significant. The following provides an excellent example of this significance.
The CYP 3A family is particularly important in the context of xenobiotic metabolism because, as Jones et al., have noted:
The CYP 3A gene products are among the most abundant of the monooxygenases in mammalian liver and intestine. In humans, CYP 3A4 is involved in the metabolism of more than 50% of all drugs as well as a variety of other xenobiotics and endogenous substances, including steroids [:27].
One drug that is of interest in this regard is troglitazone (marketed as Rezulin®) and used in the treatment of type-II diabetes. Troglitazone was removed from the market in the U.S. in March 2000. Despite the fact that it had been shown to be safe and effective in rodent studies , more than 65 people died (two-thirds were women), and many other required liver transplants as a result of Rezulin® toxicity. In clinical trials involving a total of 2500 human subjects, about 2% showed alanine aminotransferase (ALT) levels more than 3 times the upper limit of normal. ALT levels this high are an indicator of active liver disease (see [:114–119] for details of how Rezulin® came to market on the FDA "fast track").
The class of drugs to which troglitazone belongs was developed using rodent models of insulin resistance, but without prior knowledge of the cellular target . It is now known that troglitazone achieves its therapeutic effects by binding to the PPARγ nuclear receptor. But the concentrations required to activate PPARγ also activate the PXR nuclear receptor in humans – something it did not do in rats and mice . Thus, one immediate consequence of this interspecific difference is that human patients taking troglitazone experienced increased CYP 3A4 activity. Jones et al., comment:
Our data showing that troglitazone activates human PXR at concentrations similar to those required to activate PPARγ provide an explanation for its interactions with other drugs, including oral contraceptives. Interestingly, the relative lack of activity of troglitazone on the mouse or rat PXR may explain why these effects were not reported in animal toxicology studies. Additional studies will be required to determine whether PXR also plays a role in the hepatotoxicity observed with troglitazone. In this regard it is interesting that that the PXR ligand rifampicin has also been associated with hepatotoxicity in humans [:36].
Recently, it has been argued that the increased CYP 3A4 activity associated with troglitazone activation of human PXR results in the metabolism of troglitazone to a reactive quinone which has been hypothesized as the cause of hepatotoxicity . Examples like this could be multiplied for our discussion here. However, the point is made that at the molecular level of life there are medically significant differences between species. These differences may arise from evolved differences in catalytic activity of enzymes, from evolved differences in the regulation of gene expression, or even as by-products of interactions created by the introduction of xenobiotics, never "seen" by nature or evolution.
The consequences of the belief that humans and rodents are the same molecular animal dressed up differently can be (and have been) catastrophic. As Goldstein recently put it in an editorial in the New England Journal of Medicine:
One of the most striking features of modern medicines is how often they fail to work. Even when they do work, they are often associated with serious adverse reactions. Indeed adverse reactions to drugs rank as one of the leading causes of death and illness in the developed world [:553].
Endocrinology and Public Policy
Observations of species differences in the context of comparative endocrinology have led at least some observers to give serious consideration to evolution's consequences. Thus Hart commented:
It has proved heuristically useful in studies on estrogens. . . to adopt the unifying concept that species differences in estrogen toxicity mirror species differences. . . in estrogen endocrinology. The poor predictiveness of animal studies for humans thus becomes comprehensible in terms of interspecies variations in endocrinology [:213].
This matter is very urgent because it has become clear that a large number of substances in the environment have impact on estrogenic, androgenic and thyroid hormone activity. The US Environmental Protection Agency's (EPA) endocrine disruptor study program will be employed to examine these issues with a view to human safety and well-being.
But problems have been uncovered concerning the rodent strains selected to evaluate the human risk. As Spearow and Barkley have commented on the results of recent research:
. . . studies have revealed a tremendous amount of genetic variation in susceptibility to endocrine disruption by oestrogenic agents between strains of rats and mice. These studies have shown that the highly prolific, large litter size selected CD-1 mice and Sprague-Dawley rats most commonly used for product-safety testing are much more resistant to oestrogenic agents than other strains examined [:1027].
CD-1 mice are at least 16-fold more resistant than other strains of mice (including B6) to compounds that cause inhibition of testis weight, a measurement used as an indicator of androgen activity. CD-1 mice are 126-fold more resistant than B6 mice to inhibition of sperm maturation by estradiol ; the authors add:
The most favored EPA rodent model for endocrine disruptor testing, the Sprague-Dawley rat, is also more resistant than other strains to inhibition of testis weight by (DES) diesthylstiboestrol. Furthermore Sprague-Dawley rats are highly resistant while Fisher-344 rats are highly sensitive to oestradiol, DES or BisPhenol A induced hyperprolactinaemia, uterine and vaginal hypertrophy, hyperplasia, mucous secretion and c-Fos induction [:1027].
Since evolution tends to amplify differences between populations after the cessation of gene flow, short of highly fortuitous convergent evolution (nowhere demonstrated), it is unlikely that interspecific differences will be less than those observed among different strains of the same species of rodent. If not highly unlikely, it would, at the very least, be imprudent to assume in such important studies.
A good illustration of the problem here lies in the fact that the rodents selected for the study of endocrine disruption in humans have been selected for the pragmatic virtue of large litter sizes. But, as researchers have noted:
We should realize that an animal that has been selected for high fecundity regulates reproduction quite differently than unselected individuals. Furthermore, these highly prolific strains tend to be quite precocious, with many "immature" CD-1 females showing vaginal opening and elevated uterine weights even in response to the 0 dose control treatment. Such precocious sexual development and the resulting elevation of ovarian oestrogen production complicates, if not limits, the use of strains previously selected for high prolificacy for detecting osestrogenic activities in intact uterotropic assays [:1028].
Lying at the heart of an evolutionary view of animal populations is variation. Variation exists both within and between populations. With the cessation of gene flow between populations, initial variation between ancestral populations of a given species can be expected to be amplified in successive generations. Alleles rare in one such population may become common in another, and so on.
What then is to be done in the light of observations such as these? Constructive suggestions for the future course of medicine already exist that are harmonious with evolutionary theory, and emanate from such branches of biomedical science as pharmacogenetics and pharmacogenomics (a branch of pharmacology using genome-wide techniques to study inherited differences with respect to drug response). The long-term goal of pharmacogenomics is that of therapy tailored to an individual patient – therapy that reflects the uniqueness of the individual as a member of an evolving population. As has been observed by Evans, et al.,
The potential is enormous for pharmacogenomics to yield a powerful set of molecular diagnostic methods that will become routine tools with which clinicians will select medications and drug doses for individual patients. . . Genotyping methods are improving so rapidly that it will soon be simple to test for thousands of single nucleotide polymorphisms in one assay [:546–547].
What are we to do in the meantime? Population studies with respect to drug metabolism are already providing clinically relevant insights. Again, an observation by Evans, et al.,
. . . a specific genotype may be important in determining the effects of a medication for one population. . . but not for another; therefore, pharmacogenomic relations must be validated for each therapeutic indication and in different racial and ethnic groups. Remaining cognizant of these caveats will help ensure accurate elucidation of genetic determinants of drug response and facilitate the translation of pharmacogenomics into widespread clinical practice [:547].
While the specter of "race-based" medicine is sure to raise hackles (see relevant discussions [32,36-38]), we already know of many statistical associations of certain, metabolically significant, allelic variants with certain racial and ethnic groups. That is to say, two populations may differ with respect to the relative statistical frequencies of certain allelic variants. Many of these associations are simply results of long past events, such as natural or human-created barriers that separated populations. Until such a time as individualized therapy is possible, matters of ethnicity ought to be one of the factors taken into account in a rational discussion of the course of drug therapy, as these matters currently are recognized and used as a factor in genetic counseling. The pre-evolutionary, typological "one therapy fits all," possibly rooted in "Caucasian (male) as human" model, requires serious critical scrutiny.