Biotechnology November, 1993. Published in Feb. 1994. Volume 12: 156-163. From: Richard C. Strohman, Emeritus Professor of Molecular & Cell Biology, University of California, Berkeley, CA 94720 (510) 642-4941
The range of human phenotypes/diseases for which our burgeoning bio-molecular data base is sufficient to provide understanding, diagnosis and therapy is small. Only 2% of our disease load is related to monogenic causality, and even here the final phenotype is modulated by many factors. Monogenic logic cannot moreover, be applied to the 98% of our most important sources of premature disability and death. This article provides an analysis of the limits of genetic thinking in biotechnology and describes the outline for another approach to understanding complex cellular/physiological systems. In this outline, rules governing physiological regulation and cellular and higher levels of organization are located not in the genome, but in interactive epigenetic networks which themselves organize genomic response to environmental signaling.
Biomedical sciences are based mostly on a genetic paradigm committed to the idea that major diseases will be diagnosed and treated through gene technology (1). While this approach has been eminently satisfactory for true monogenic diseases constituting only about 2% of our total disease load (2), and while much effort has been made to convince us of the value of gene diagnosis for our more common but complex multifactorial diseases (3), it is becoming clear that genetic analysis in itself will not serve to predict, diagnose, or treat diseases like polygenic cancer, or hypertension, or other complex human phenotypes (4,5,6). As suggested below, the failure of genetic analysis to predict outcome for complex behaviors in which many genes are involved is accounted for by the newly discovered examples of epigenetic regulation in cells and in interactive physiological systems. These discoveries embrace, on the one hand, the fact of redundant genes (7,8) and, on the other hand, the fact that cellular networks accomplish similar endpoints through multiple overlapping pathways (see below). Both of these phenomena tend to confound predictions based on linear genetic logic.
Epigenetic control is an old idea in biology (9), and reflects the certainty that, when many genes are involved, simple linear analysis is insufficient to provide for reliable prediction. In multigenic cases the system is dominated by a nonlinear logic in which environmental and developmental signaling play a transcendent role in determining phenotype. Here the outcome is one in which genes may be necessary but not sufficient to provide for prediction and diagnosis. While epigenetic control is well known, the logic of its operation is only now beginning to be studied (see below).
What follows in this article is an analysis of the current biomedical paradigm and a documentation of areas in which progress is predicted to be less than adequate and therefore where the paradigm might be flawed. As Thomas Kuhn (10) has pointed out, it takes many years of work within a paradigm before ingrained flaws and new perspectives become apparent. In modern biotechnology, where large investments in material and human resources are so easily tied to promising but ultimately flawed concepts, it is essential that we examine our paradigm closely and often.
The major assumption of modern biomedical research is that unique genes have unique effects. This assumption is essential in the following areas:
Medical genetics, which seeks isomorphic mapping of human diseases to Mendelian genes (1,3).
Molecular biology, which seeks to identify unique genetically based mechanisms driving cellular processes (1).
Developmental biology, which presupposes (i) the presence of genetic programs, (ii) additivity of gene effects, and (iii) the ability to map complex developmental stages to additive programmatic sequences in DNA (8).
These assumptions and presuppositions, now experiencing major problems (4,5,11), are also the major features of the human genome project or HGP. The HGP has become the centerpiece of the biomedical paradigm and has distilled a simplistic guide for future research and application. This guide is summarized as follows:
a.
All major non infectious diseases are caused by defective genes.
b.
Diagnosis and therapy is available through genetic analysis alone.
c.
Aging and other complex human behavior is genetic and all may be mapped to Mendelian factors.
As Brenner (11) and others have pointed out, however, the uniqueness assumption of genetic determinism:
-----Unique Genes ----------> Unique effects
is undermined by an emerging body of evidence showing functional informational redundancy in cell regulation. Here the focus is on redundant genes ... that more than one gene may specify any given function (7). In this case the reductionistic plan to associate genetic causality with complex phenotype is brought into question since the major research approach, saturation mutagenesis, depends completely on the uniqueness equation. This approach to understanding disease will generate a map or network of factors which interact to provide a useful background for a complex phenotype. But, as argued below, ultimate behavior is not encoded in DNA but rather in the environmentally interactive cellular epigenetic network which includes the genome.
What most biologists have assumed for years, but have never really formalized, is that every cell contains not one, but two, informational systems; the first is genetic and the second epigenetic. The familiar genetic system of: DNA ----> RNA -----> Protein -----> Phenotype is applicable to a small range of human phenotypes. In biomedicine it is restricted to monogenic diseases like Duchenne muscular dystrophy, hemophilia, and a host of other diseases (12). However, these diseases remain a small percentage of our disease load and account for less than 2% of the total (2) (see below).
An epigenetic system may be said to be chaotic in that, while it is impossible to predict which alternative pathway will be used, it will be possible to determine potential for adaptive change under precisely defined initial conditions (15,16). As Skinner points out, chaos theory holds that patterns of behavior linked to vast numbers of interacting elements, that appear to be random and indeterminate, are actually repeatable patterns obeying simple rules. The system is thus a determinative chaotic system open to new approaches that combine linear genetic with non-linear complex system (epigenetic) analysis (16). Looked at in this way we may predict a new opportunity in biotechnology; viz., the definition of complex system parameters and specific environmental perturbations that elicit unique disease/health outcome ( see reference 17 for example).
Examples of redundancy and epigenetic regulation will also be found among the recent spate of knockout experiments in which the cell or organism develops a normal phenotype in spite of the deletion of both copies of an "important" gene (see below, VIB3a). Other examples are found in cytokine research where linkages of cells to local matrix provides context for intracellular modulation of a variety of activities (18).
We now look at specific instances where mainstream genetic determinism has found itself in conflict with new (and old) findings from basic research in genetics, and in other areas of molecular and cell biology. Most of these conflicts may be resolved by recasting them in epigenetic terms.
The tension between population genetics and medical genetics has been described repeatedly by Lewontin (5) and analyzed most recently by Wahlsten (6). In brief, the argument is that the major statistical tool, analysis of variance, or ANOVA, used to assign quantitative weight to genetic causality, is insensitive to heredity-environment interaction. This insensitivity is minimized in agricultural breeding experiments for which ANOVA was designed because large sample size is normally the rule. In medical genetic studies (extended families) or in behavior genetics (twin studies), the sample sizes are relatively small so that error is large in detecting lack of interaction between heredity and environment. As Wahlsten points out, a newer statistical approach, multiple regression, is replacing ANOVA, but for the kinds of studies we are discussing the two procedures are essentially equivalent. Experts in agricultural genetics generally detect significant interaction between genes and environment and are extremely cautious in applying heritability coefficients or in assigning any significant numerical value to genetic cause when dealing with complex traits. Their position is that if gene effects are interactive (not additive) with environmental effects, it is incorrect to use ANOVA for assessing genetic contribution to a particular phenotype across a range of environments. Medical geneticists, however, using the same ANOVA but with significantly smaller sample size, not surprisingly do not detect interaction and therefore assume that heredity and environment are additive. They then assign great significance to heritability coefficients and are confident these numbers describe quantitatively the contribution of separate heredity and environment to any particular phenotype. We have a medical literature, then, that asserts with great confidence, but with serious theoretical reservations from sectors of population genetics, that this or that complex disease, while having an environmental component, also has a separate genetic component that can be discovered and utilized in pursuit of some hypothetical treatment strategy. It is beyond the scope of this review to enter this controversy fully; it is enough to state the minimum conclusion that medical genetics, with a linear view of gene-disease causality, finds itself in serious debate with a significant segment of its parent science of population genetics, which sees complex traits, including disease, as highly interactive, epigenetic, and impossible to reduce to genetic elements alone.
Genetic determinism in medicine may be traced to early successes, notably that of Garrod in 1908 who mapped a metabolic disease, alkaptonuria, to a Mendelian inheritance pattern. Since that time a large number of such monogenic diseases have been discovered and there is a general misconception that all diseases are available to monogenic logic and to solution through gene therapy of some kind. In fact, the total percentage of monogenic diseases has remained constant at less than 2%. While rare monogenic diseases are legitimate targets of the new technology most of the rest of the 98% of human diseases, including cancer and heart diseases, are not. The latter are polygenic, multifactorial diseases for which genes may be necessary but not sufficient (2).
As Table 1 shows, diseases may be distributed according to whether they are determined before or after fertilization. Those (2%) determined before fertilization are, of course, genetic, and are mostly not preventable. Of those determined (98%) after fertilization, there may be multiple causality, including early developmental effects, but in theory at least these are all preventable. Patterns of disease distribution having to do with secular trends, and environmental changes for large (migrating) populations also point to environmental rather than genetic determinism for modern diseases like cancer and heart diseases (19).
Restriction fragment length polymorphisms (RFLP) are being used to generate maps of genes and gene products which interact to produce a disease phenotype. The general idea here is that unique DNA sequences (mutations) can be linked to inheritance of phenotype and then mapped to specific chromosomes. Ultimately this analysis may lead to mutated genes of known function, and, theoretically, to gene or gene product replacement therapy. While this approach is applicable to single gene diseases it is highly suspect when applied to polygenic, multifactorial diseases (5,20,21).
The starting point for much of RFLP work was the analysis by Lander and Bottstein (22) applied to the hypertensive rat (23). This work revealed linkage of hypertension to a mutation at the ACE (angiotension converting enzyme) locus, a gene responsible for converting rennin to angiotension, a protein crucial to blood pressure regulation. Subsequent work, however, showed that ACE mutation was not linked with hypertension in humans (24). More recent studies provide a strong suggestion that, even for the hypertensive rat, early developmental changes will neutralize ACE mutation and provide for near normal phenotype. Thus, if young rats are taken from genetic mothers bearing the ACE mutation and nursed by normal mothers of a related strain, the pups show decreased levels of hypertension (25).
Myocardial infarction in humans has also been linked to ACE mutation (24). However, in this study many individuals were identified with the identical mutation who had no heart disease. Clearly, other factors are involved. How many other genes or other factors might there be? In studies like this the question is rarely asked. But the physiology of heart function clearly reveals that ACE related diseases will most likely be multifactorial, polygenic entities. If so, then one expects that each of the many genes will have a small effect, redundancy will be present, and any one gene or even several functionally related genes may be necessary but not sufficient to precipitate a heart disease (20,21). In other words, one anticipates that in this situation genetic diagnosis will not be a robust predictor of phenotype. The environment and individual natural history will be major determining factors. In the case of angiotension-related function it is clear that redundant epigenetic regulation will dominate a single genetic defect. Why? It is well known that in the normal or diseased human ventricle ACE is a minor source of conversion of rennin to angiotension II. There are many other (gene coded) serine proteases that provide for 90% of ventricular angiotension II levels26.
We conclude that ACE mutation will predict neither hypertension or myocardial infarction in humans. While an ACE mutation might have some effect, at the wider physiological-nervous system level there will be further interactional complexity and phenotypic adaptation including central nervous system override of rennin production. These and other elements of the hypertensive control network will confound simple genetic determinism. Examples include complex cortical and medullary regulation of heart and blood pressure rhythms that are exquisitely sensitive to environmental input and personal experience (27). Nevertheless, the biomedical community persists in calling for the use of an ACE gene screen to predict tendency for heart diseases (28). Why?
Molecular biologists are compelled to find as much detail as possible in gene based networks like the one for hypertension, and RFLP approaches do provide the appropriate tool. In time, a gene map of extremely high density for this network will become available. But such maps, for each multifactorial trait or disease, will include perhaps hundreds of genes and interactive gene products all with input from environment. The complexity of such a system can only be described as chaotic; there will be little of predictive value in the individual bits of genetic information defining this system. Rather, hypertension or other disease phenotype will be defined by the system as a whole and the responses the system makes to the appropriate internal and external signaling pattern. The search for a screen that depends on a single genetic variant is an understandable oversimplified approach born of an optimism that all controls reside in genetic elements. But, as has been pointed out above (5,20,21), such optimism, realistic when applied to the rare monogenic diseases, is misplaced when applied to complex multifactorial systems like hypertension.
Two major diseases, schizophrenia and bipolar disease (manic depression) offer interesting alternatives to a strict genetic focus. In both cases we have had a long history of multi-million dollar funding for genetic causality with nothing to show in the way of understanding or therapy. In schizophrenia, the non-concordance for monozygotic twins (MZ) is 50% so clearly genes are not the only answer. Nevertheless, great emphasis is placed on DNA analysis to detect gene mutations for this disease. But the most recent studies have provided a null result. Thus, the mapping of schizophrenia to a locus on chromosome 5 in a study of English and Icelandic families (29) was annulled by a larger study using Swedish families (30). The net result of these studies was that probably several genes were involved. Precisely the right question. The next question, rarely asked, is how many other genes may be involved. It might be five or ten or 100; nobody knows. But clearly, with more than 3-4 genes involved, epigenetic regulation will dominate (31).
An epigenetic alternative to a genetic cause is that schizophrenia, together with many other complex human "mental" phenotypes, is a developmental disease. It is clear, for example, that minute changes in wiring of the nervous system are due to non genetic changes (32). Environmental input during early development could be crucial so that unequal sharing of uterine resources by monozygotic twins (33) could result in changes in embryonic wiring diagrams that would in turn provide an adequate basis for phenotypic outcome later on. Indeed, when one looks for anatomical differences between MZ twins, one see that a related micro anatomical change is present in the affected but not in the normal member of the pair (34). Thus MZ twins of identical genotype will display different behaviors related to epigenetic change; in this case perhaps to minor differences in the way the human cortex is wired during development. It is too early to decide the merits of the wiring hypothesis. It is enough for us to note that such developmental studies have been much less "attractive" to the research community than are approaches designed to identify specific genes.
Bipolar disease (mental depression) offers a similar conclusion. Early studies suggested a strong linkage to specific loci (35) but these studies were contradicted by later ones using larger families (36) so that now there appears to be agreement that manic depression is related to many genes (37,38). How many we do not know. But it appears that the search for single genes leads, in this case, to the conclusion of complex epigenetic regulation in which genes together with dominant environmental signals (perhaps early developmental signals) determine the outcome. Once again, genes may be necessary but not sufficient. Nevertheless, there is now a renewed search for a unitary molecular mechanism for this disease (39). Such searches disregard epigenetic theory which recognizes that traits are most likely highly interactive and, while many proximal molecular and genetic factors will be discovered, a distal controlling set of environmental factors will most likely be found to be crucial. How might environments regulate gene expression?
There are several examples of environments providing strong constraints on genetic determinism. The first may be found in a series of studies (see reference 40 for review) in which rat populations are exposed to strong selection for ability to solve maze problems. Within seven generations, in standard rearing environments, two separate populations were selected with very significant abilities to solve mazes; a bright population and a dull one. These populations bred true for these phenotypes. However, when bright rats (pups) are given a "dull" developmental environment they perform as dull rats in later life. The reverse is true when dull rats are raised in "bright" environments. As Gilbert Gottlieb has pointed out, we inherit not only genes but an entire developmental manifold which provides overriding controls on the way gene expression manifests as overt phenotype (40). Another possibility, as discussed above, is the case of hypertensive rats pups raised by surrogate mothers where the adopted pups display a less intense phenotype even though they carry the ACE mutation. Other examples are found in work on MZ twins who show extremely different phenotypes although they have, of course, identical genes (33). Aerobic stress and muscle training can combine to change levels of gene product concentration and even specific gene repression and activation as discussed in section III (13,14). Finally, we tend to forget that molecular biology itself was built on the operon theory in which genes are turned on and off by environmental signals31. In the operon case procaryotic gene control of enzyme expression may be understood in terms of a linear mechanism. In the case of epigenetic controls, however, it will be a much more complex pathway (Figure 1); one in which even strong gene effects will be redirected by the entire genetic background in conjunction with different environmental settings.
Cancer, in its multiple forms, has often been described as one of our most multifactorial and enigmatic diseases. While strong evidence exists for a genetic background, for many cancers much of this evidence is potentially confounded by congenital and familial effects; we forget that many things are inherited in addition to genes (33). In addition, we know that many forms of cancer have strong environmental determinants. For example, epidemiological studies clearly reveal that migration of human populations results in new patterns of cancer in which the group takes on diseases reflective of their new environment, and abandons diseases common to their relatives who remain at home and with whom there is a shared genetic background (19). Current research emphasis is on mutation in tumor supressor genes which, while they will play some role in cancer, may also prove to be constrained by other factors. Below we analyze several cases and conclude that an epigenetic basis for cancer is an attractive but missing research component.
The most recent trend has been to associate unique cancers with mutation in growth control or tumor suppression genes such as p53 or retinoblastoma (rb) (41,42). These genes code for DNA-binding proteins that are presumed to delay or inhibit cell replication. Mutation in both alleles would then produce defective regulation of growth and tumor formation. If one of these genes is defective at birth then one inherits a tendency for cancer; the disease itself is then predicted to occur when, through somatic mutation, the second allele is also defective (12). But it is now clear that some form of redundancy for both of these genes is present in cells making it difficult or impossible to use mutational analysis alone for predicting cancer. For example, a mouse has been constructed with both p53 alleles absent (43). In this case, it was expected that growth control in all cells would be defective with dire effects for all affected individuals. However, the affected animals were all normal at birth, and early development and growth was normal. It was only after adulthood was reached that some, but not all, of these individuals developed tumors in excess of that found in control populations. Clearly, the early phases of development that depend on stringent growth controls remain independent of p53 input or have redundant pathways around p53. The same must be said of the p53 mutant adult individuals that did not show any tumor formation. Finally, when a normal p53 gene is inserted into cancer cells it may or may not restore normal growth regulation (44).
More recent evidence shows that p53 protein may form heterodimers with many other cellular proteins (45) including replication protein A which is involved in the initial stage of DNA replication (46). Thus, p53 regulation is a prime candidate for epigenetic control in which the final effect is modulated by a complex interaction of many bits of genetic and environmental information. Much is learned about DNA replication in p53 studies, but the emerging picture shows us, not single gene control of cancer, but complex interactive regulation. Epigenetic interactions of p53 protein with other gene products forms a basis for explaining the varied effects observed when p53 is mutated in different genetic backgrounds (43) or when wild type p53 fails to restore normal growth regulation to p53 defective cells (44).
A similar story may be told for the retinoblastoma (rb) mutation (see reference 47 for review) which arises either spontaneously or via heredity associated with a deletion in or absence of chromosome 13 in 20-30% of affected cases (48,49). But 20-30% is not 100%, so clearly other factors are involved. Many individual rb tumors do not show mutated rb genes (50). In addition, while expression of wild type rb in some rb-defective cells will restore normal growth (51), such transfection and expression fails to produce normal growth when these cells are transplanted to the eye of nude mice (52). Finally homozygous rb knockout in the mouse is lethal but only late in development after lineage determination is complete and after millions of cell replications have been completed (42). This gene, therefore, while it plays an important role in cell replication, is not essential during early development. We must assume that it displays epigenetic interaction with redundancy, at least, at the genetic level.
Human breast cancer work is heavily invested in genetic causality even though a large population study tells us that less than 2.5% of breast cancer is associated with genetic determinism (53). Many mutations are found in later stages of a variety of tumors, but it remains uncertain whether these mutations are the cause or the effect of earlier non-genetic lesions which, if reversed soon enough, would have deflected the tumors altogether (54,55).
The latest focus in breast cancer has been on a familial study where linkage has been established for "cancer tendency" to a locus on chromosome 17q21 (56). The lod score for linkage was 5.98, well in the range to insure high probability of association between the cancer and the genetic anomaly. While the technology of this linkage study may be assumed to be state of the art, we must also be aware of its problems. For example, we do not know what the frequency of this mutation might be in the general population nor do we know the extent to which other mutations might be present in the suspect or other chromosomes of the affected women. Neither do we know the pleiotropic and epistatic effects that other genes might have in altering the penetrance of the suspected mutation. These are all questions of fundamental importance in elementary genetics (31). We also have no knowledge of what environmental influences might be required in order for tumor formation to occur in the presence of the mutation. The mutation may be necessary but not sufficient; it may require specific environments or specific other genetic background.
Finally, as mentioned at the outset of this discussion, we also know that while there may be some relationship of family history to breast cancer only 2.5% of breast tumors are genetic in origin (53).
Even for monogenic diseases like CF the case is extremely complex. Over 350 different mutations have been found for the CFTR (cystic fibrosis transmembrane regulator) (57) and a wide variation in phenotype is associated with these genotypic changes (58). Present molecular biology reveals a bewildering array of symptoms in humans and animals with major CFTR mutations leaving open the question of the degree to which mutated CFTR function itself may be compensated for by epigenetic regulation (59). For example, 67% of cystic fibrosis cases carry a mutation at codon F508 of the CFTR gene . Those homozygous for this mutation show symptoms at early age, including pancreatic insufficiency, but also display a wide phenotypic variance for pulmonary manifestation (58). An additional mutation at codon r117h, when found as a compound heterozygote with F508, shows many of the symptoms of the F508 homozygote but with much later age at diagnosis (about 10 vs. 2 years). However, the r117h mutation turns out to be highly prevalent ( about 10% of all patients screened) and is apparently involved in diseases other than CF58. It also now appears that a channel other than CFTR may be involved since the alternative channel may show enhanced activity with UTP and partially compensate for defective CFTR function (60). Thus, even this classical Mendelian trait is now seen as one modulated by many other genes and open to epigenetic regulation. Under these circumstances any genetic screening for CF will need to cope with multiple mutations, variation in expression as a function of other genes and environments, and the persistent possibility that, for many mutants, the phenotypic outcome will remain in the realm of uncertainty.
Computed tomography and magnetic resonance imaging have become widespread and extremely expensive additions to diagnosis. The problems inherent in imaging techniques have been recently analyzed (61) and are discussed briefly below as prologue to similar problems turning up with molecular measurements which, while extremely sensitive, are also without proven meaning when applied to disease manifestation.
Imaging techniques, because they are so sensitive, often measure not disease itself but early changes in tissue which are taken as evidence that disease will develop. Early changes may however be extremely misleading since they often reflect reversible processes or those with extremely long lag time to any clinical manifestation. As our machines are able to detect the most incipient stages we experience several problems (61). First, as exemplified by thyroid cancer, is the problem of defining diseases as cellular changes that always progress to serious morbidity. In this case clinical cancer (tumor size > 2 cm) is only 0.1% in adults between age 50-70 years. However, autopsy using increasingly thin sections of the gland could reveal at least one papillary carcinoma in 36% of adults. It was calculated that as sections became thinner autopsy would show verifiable papillary cancer in 100% of cases (62). These "tumors" discovered at earliest stage represent an enormous reservoir of detectable but subclinical disease. Under these circumstances and for a variety of diseases the patient may never experience clinical symptoms but, under aggressive medical management, may become involved in unnecessary and expensive medical procedures that are predicted to have little positive effect.
The second major problem that arises has to do with the effect on reported disease frequency where frequency increases as the degree of measurement sensitivity increases. However, without any manifestation, early stage diagnosis makes it appear as if we are experiencing large increases in the disease itself. The third problem is the effect of statistical evaluations of various therapies for a disease. As the time between diagnosis and manifestation increases it is made to appear as if various therapies are working even when nothing in the way of treatment need be involved in the statistical analysis (61).
Nowhere in medical technology have we greater sensitivity of measurement than in antigen and nucleic acid chemistry. The possibility exists however that these measurements are often without predictive value for the diseases for which their measurement was designed.
Increased levels of scrutiny can, for example, explain recent reported increased prevalence in breast, prostate, and thyroid cancer61. Testing for carcinoembryonic antigen (CEA) to predict colon cancer is the latest example of an ineffective and wasteful procedure (63). Prostate antigen testing together with other evaluation may prove useful. However these tests, used alone, can provide for an enormous increase in reported prevalence, in increased apparent time of survival, and, unless carefully applied, could lead to unnecessary treatment.
Clinical diagnosis is on solid ground when measuring clearly defined end points (extremes of anatomy, or physiology, or high titers of biological entities such as virus or bacteria) for which there is good correlation to disease. However, when clinical tests rely on so-called surrogate markers the ground becomes soft. Measuring actual hepatitis B virus titers is an example of a useful (hard) endpoint (64). Measuring the concentration of a liver enzyme as a sign of virus-induced hepatitis C liver damage (65) is a dubious (soft) indicator since enzyme levels represent epigenetic endpoints with many pathway outcomes in which is imbedded much redundancy.
Non-A, non-B hepatitis (NANBH) remains one of our most poorly defined diseases. Nevertheless, hepatitis C virus (HCV) is assumed to be the prime cause (66). Whether or not there is an entity corresponding to HCV is actually not clear since the virus itself has never been isolated. Rather, the molecular basis for HCV rests on detection of an RNA in NANBH patients that did not hybridize with host genomic DNA (67), and subsequent detection of an antibody present in blood of chimps and patients with liver disease (68). This RNA has been cloned and is positive-stranded with regard to the encoded NANBH antigen. In addition, by indirect methods a number of proteins have been identified that are encoded by the putative HCV genome (69). All together, this very impressive molecular biology has clearly shown much similarity between the reconstructed "agent" or HCV and flavi- or pestivirus. But causality between the presence of a putative virus and liver disease remains completely unproved, and a long term study has failed to measure any increase in mortality from all causes after transfusion-associated NANBH (70).
The present antibody test for HCV is used nationally to monitor the blood supply at an estimated cost of millions of dollars annually. But the antibody is constructed from an RNA sequence, presumed to be viral but, of unknown origin. While the antibody test has been given the credit for lowering the rate of post-transfusion NANBH it has also been noted that this credit may be due more to the intensive concomitant screening for HIV and the subsequent elimination of many risky donors (66). Here we may have a case of a meaningless test based upon a high tech ability to measure, with exquisite sensitivity (cDNA/polymerase chain reaction or PCR), extremely small levels of biological molecules. The initial measurements of putative HCV RNA required enormous amplification using (PCR) but our actual experience establishing clinical meaning for such low levels of a marker like this is nil. In one study PCR results could only correlate with serological tests in 40% of the cases (66). In addition, while the literature dealing with HCV reports viremia as a characteristic of HCV infection, viremia is always in terms of PCR measurement (71). As Kary Mullis, the inventor of PCR, himself states, "The vice of the PCR is that it can find the biochemical equivalent of the needle in the haystack. Viral fragments that are present only in minute quantities can be amplified and identified, but this tells us nothing about whether replicating virus is present in sufficient quantities to do harm" (72).
PCR has also proved to be crucial for diagnosis of HIV-related AIDS where it is used to amplify vanishingly small amounts of HIV sequences.
The HIV-AIDS hypothesis remains plagued by the fact that most AIDS patients, until end stage disease, rarely show HIV viremia (classically defined by actual virus replication), and diagnosis continues to rely on PCR and antibody measurements. In HCV and in HIV we may have abandoned, at least at great scientific cost , traditional rules (Koch's Postulates) for establishing disease causality (73).
The number of genetic marker tests possible is increasing every month as more and more RFLPs are catalogued, so that it becomes inevitable that many associations will be found between markers and disease. For reasons already given results of these associations will be extremely difficult to interrupt. In addition there are economic, legal, and ethical complications to screening for genetic markers.
Markers are now advanced for disease tendency for complex polygenic diseases like heart and cancer problems for which many genes must be involved; but can we afford it? As discussed above, such tests will be confounded by informational redundancy at both the genomic and at the epigenomic levels, and will be accompanied by unknown levels of false positives and false negatives. Many tests will have to be done. Consider the following: there are an estimated 100,000 human genes. If only one percent is active in some chain of causality for a complex disease like hypertension or schizophrenia then, with a minimum million patients annually and a minimal cost of $10 per test, our annual gene testing bill is $10 billion. With polygenic diseases in which 100 genes are involved, RFLP measurements will actually measure some relationship of the disease with all 10022. But with redundant genes, and epigenetic adaptation the measurements may signify nothing.
Legal and ethical issues are related to studies involving population groups. If gene linkage studies are used in a population with subgroups tending to mate among themselves, and if we find that they score high (or low) for a mental trait and at the same time show association with particular markers, we could be learning the obvious: that they interbreed and therefore show a frequency for particular sequences different from other subgroups or from the population as a whole. But this difference need have nothing to do with the mental trait in question (74) This particular criticism is a major problem for behavioral genetics, in which there is a continuing attempt to discover the genetic basis for human intelligence differences among people using this RFLP technology (75).
A very large economic/ethical problem comes when high sensitivity molecular testing meets clinical practice and "defensive medicine" — in which not doing a test poses a threat of malpractice. There is great profit to be made by biomedical firms convinced they are doing the right thing in offering state-of-the-art molecular testing. The doctor on the front line will drive profit forward (without sharing in it unless the doctor has a financial interest in the technology) even when it is clear that these tests have dubious value. Once a technology is released, it will be extremely difficult to take it away. Given the number of test possibilities that can be done in the name of "disease tendency," the nation's health bill could easily be inflated by billions of dollars annually. This kind of thing is already happening. The projected costs for a cholesterol screen in the U.S., for those age 65+ only, is 2-16 billion dollars depending upon the treatment used (76), and with no predicted improvement in either morbidity or life expectancy. When one adds the testing cost for younger people (25 - 64 years of age) alarmed at the cholesterol scare touted everywhere, the cost may be closer to $30 billion per year (77).
Clearly, the use of gene and/or gene product testing without adequate analysis of predictive power or value needs close monitoring.
Finally, with the announcement in December,1993 of mutations in "mutator genes" as a cause of colon cancer (78) there was an immediate call from scientists and the leadership of the HGP for a genetic screen using a test priced at $1000 per individual (NY Times Dec.3,1993). But the test does not suggest a therapy other than what we already know from disease natural history related to diet and surgery.
A major assumption of modern biomedical thinking is that genetic inheritance is the only inheritance. But biologists have always known this to be incomplete and we are now rediscovering the nature of our over-simplified paradigm (79). For example, a recent report makes clear that even sex determination is influenced by epigenetic factors and that uterine exposure to hormones has a profound effect in some organisms (80). In modern developmental biology the second major assumption of biomedicine ... that genetic programs are a script for phenotype... is being abandoned. There is no isomorphic mapping of complex phenotype to Mendelian factors32, and the mechanism by which the organism elicits phenotypic variability from isogenic or near isogenic situations remains a profound mystery. Thus, the work on sibling species reveals that organisms may remain constant in morphology over millions of years even while they are enormously divergent at the level of DNA (81). Humans and chimps are shown to be nearly identical in genetic terms revealing that the organism is able to draw vastly different phenotypes from highly similar genotypes (82). Thus, profound questions are raised concerning the assumption of gene programming. First, that there appears to be less of a relationship between genetic and morphological complexity than we have thought. Second, if the program is not in the genes, and organisms clearly are programmed, then where is the program? These and newer variations of complex, non-linear themes tend to be suppressed by our near monolithic commitment to molecular genetic mechanisms. This review has suggested that the molecular reductionist program to explain life has serious limits and that new epigenetic approaches to genetic regulation will be crucial. What might these new approaches be?
John Maddox, the editor of Nature, has written that modern biology, in concentrating on mechanism, has neglected theoretical approaches that might provide structure to the enormous data base accumulated by strictly molecular inquiry (83), and has suggested that such a conceptual structure might include a quantitative approach to dynamical cellular properties such as concentration fluxes of molecules which would control gene expression (84). Numerical characterization of these properties might then provide a basis for theory construction concerning regulation at levels higher than the gene. Theoretical physicist-cum-biologist, Walter Elsasser, has in fact laid out a basic description of a holistic theoretical biology in which dynamical properties play the role of higher order regulation (85). It is apparent that new research opportunities need to be created that will encourage work on these dynamical systems, and the theoretical structure hinted at by Maddox and Elsasser may lie, at least partially, in theory of complex systems (15,16).
One might begin the merger of genetic reductionism and epigenetic complexity with those areas where multigenic systems are know to be coordinated by higher order cellular responses to environmental conditions. Nobel laureate, Barbara McClintock, who described mobile genetic elements long before they were discovered by molecular biology, had always been preoccupied with mechanisms that rapidly reorganize the genome. In one of her last reviews she wrote of the significance of responses of the genome to challenge. She ended that article as follows: "We know about the components of genomes .... We know nothing, however, about the how the cell senses danger and initiates responses to it that often are truly remarkable" (86).
At the cell level an interesting epigenetic approach to complex analysis of heart disease with multigenetic causality linked to interactive environments is the work of Sing and his group (17). At levels above the cell ... for complex physiological systems ... chaos theory builds on epigenetic thinking and already is providing new ways to think about complex systems. This is particularly true for cardiac function where sinus arrhythmia, long thought to be low level noise, or random fluctuation in heart rate, is now seen as high order chaos (16). Coupling of heart rate to brain function and thus to experience has long been appreciated as an observable patterned occurrence, but was mostly inexplicable through standard physiological experiment (87). Chaos theory is able to provide a method of revealing generic pattern in what was thought to be random variation. Recognition of these patterns allows new insights into brain-heart physiology and may even allow prediction of sudden cardiac death among patients at risk (88,89). A recent symposium on chaos in brain function makes clear what new possibilities may lie beyond a reductionistic paradigm (90).
It is here, at this interface between cell/organism and external world, that new research effort might be focused. Initial cellular responses are epigenetic in nature and involve selection of adaptive response from a bewildering array of molecular possibilities. At cellular and higher levels we expect that evolution has worked to select not just single genes but integrated behavior or generic patterns of response at all levels of biological organization (9). These patterns can not be seen by linear analysis. It is at this level that theory of complex systems might prove useful. Generic patterns, with some ultimate basis in genomic reorganization, changes in gene expression, etc., would perhaps be open to theoretical structuring. Explanation and prediction of behavior leading to cancer or other cellular pathology, and to disease of heart and other complex organs would then not need to depend entirely on an apparently endless reductionistic analysis but could rely more on understanding rules of higher level organization; rules which, themselves, have been selected, and which control downstream mechanistic elements.
1. Hood,L. 1992. In: The Code of Codes. D. J. Kevles and L. Hood. (Eds.) Harvard University Press, Cambridge, Massachusetts.
2. Mckeown, T. 1979. The Role of Medicine: Dream, Mirage or Nemesis? Princeton University Press.Princeton, New Jersey.
3. Casky, T. 1992. In: The Code of Codes. D. J. Kevles and L. Hood. (Eds.) Harvard University Press, Cambridge, Massachusetts.
4. Strohman, R.C. 1993. Ancient Genomes, Wise Bodies, Unhealthy People: Limits of Genetic Thinking in Biology and Medicine. Perspectives in Biology & Medicine 37(1):112-145.
5. Lewontin,R.C. 1992. Biology As Ideology. Harper Perennial, New York
6. Wahlsten, D. 1990. Insensitivity of the analysis of variance to heredity- environment interaction. Behav. and Brain Sci. 13:109-161.
7. Tautz, D. 1992. Redundancies, development and the flow of information. BioEssays 14:263-266
8. Wilkins, A.S. 1993. Genetic Analysis of Animal Development, 2nd Ed., p473. Wiley-Liss Press, New York.
9. Wright, S. 1941. The physiology of the gene. Physiological Reviews. 21:487-527.
10. Kuhn, T. 1962. The Structure of Scientific Revolutions. The University of Chicago Press, Chicago, Illinois.
11. Brenner, S., Dove, W., Hewrskowitz,I., and Thomas, R. 1990. Genes and development: molecular and logical themes. Genetics 126: 479-486.
12. Weatherall, D. J. 1982. The new genetics and clinical practice. Nuffield Provincial Hospitals Trust, London.
13. Coggan, A. R., Spina, R. J., R ogers, M. A., King, D. S., Brown, M., Nemeth, P. M., and Holloszy, J. O. 1990. Histochemical and enzymatic characteristics of skeletal muscle in master athletes. J. Applied Physiology 68:1896-1901.
14. Pette, D. and Dusterhoft, S. 1992. Altered gene expression in fast-twitch muscle induced by chronic low-frequency stimulation. Am. J. Physiol. 262:333-338.
15. Kaufmann, S. 1993. The Origins of Order. Oxford University Press, New York.
16. Skinner, J.E., Molnar,M., Vybiral,T., and Mitra,M. 1992. Application of chaos theory to biology and medicine. Integ. Physiol. & Behav. Sci. 27:39-53
17. Sing, C. F. and Reilley, S.L. 1993. Genetics of common diseases that aggregate but do not segregate in families. pp140-161 In: Sing, C.F., Hanis,C.L., (Eds.) Genetics of cellular, individual, family and population variability. Oxford Univ. Press, New York.
18 Nathan, C., and Sporn, M. 1991. Cytokines in context. Jour Cell Biol. 113:981-986
19. McKeown, T. 1988. The Origins of Human Disease. Basil Blackwell, Inc., New York.
20. Feldman, M. W. and Lewontin, R. C. 1975. The heritability hang-up. Science 190:1163-1168.
21. Gillespie, J. H. and Turelli, M. 1989. Genotype-environment interactions and the maintenance of polygenic variation. Genetics . 121:129-138.
22. Lander, E.S., and Botstein, D. 1986. Strategies for studying heterogeneous genetic traits in humans by using a linkage map of restriction fragment length polymorphisms. Proc. Natl. Acad. Sci. USA 83:735-737.
23. Jacob, H. J., Linkpaintner, K., Lincoln, S. E., Kusumi, K., Bunker, R. K., Mao, Y. P., Ganten, D., Dzau, V. J., and Lander, E. S. 1991. Genetic mapping of a gene causing hypertension in the stroke-prone spontaneously hypertensive rat. Cell 67:213-224.
24. Cambien, F., Poirier, O., Lecref, L., Evans, A., Cambou, J. P., Arveiler, D., Luc, G., Bard, J. M., Bara, L., Ricard, S., Tiret, L., Amouyel, P., Alhenc-Gerlas, F., and Soubrier, F. 1992. Deletion polymorphism in the gene for angiotensin-converting enzyme is a potent risk factor for mycardial infarction. Nature 359:641-644.
25. Myers, M.M., Brunelli, S.A., Squire,J.M., Shindeldecker, R.D., and Hofer, M.A. 1989. Maternal behavior of SHR rats and its relationship to offspring blood pressures. Develop. Psychobiol. 22(1):29-53.
26. Urata, H., Healy, B., Stewart, R. W., Bumpus, F. M., and Husain, A. Angiotensin. 1990. II-forming pathways in normal and failing human hearts. Circulation Res. 66:883-890.
27. Peterson, L.H. 1972. in Neural and Psychological Mechanisms in Cardiovascular Disease. Milano, Italy: Casa Editrice.
28. Kurtz, T.W. 1992. The ACE of hearts. Nature . 359:588-589
29. Sherringtion, R., Brynjolfsson, J., Petursson, H., Potter, M., Dudleston, K., Barraclough, B., Wasmuth, J., Dobbs, M., and Gurling, H. 1988. Localization of a susceptibility locus for schizophrenia on chromosome 5. Nature 336:164-167.
30. Kennedy, J. L., Giuffra, L. A., Moises, H. W., Cavalli-Sforza, L. L., Pakstis, A. J., Kidd, J. R., Catiglione, C. M., Sjogren, B., Wetterberg, L., and Kidd, K. K. 1988. Evidence against linkage of schizophrenia to markers on chromosome 5 in a northern Swedish pedigree. Nature. 336:167-170.
31. Ayala, F. J. and Kiger, J. A. Jr. 1984 Modern Genetics, 2nd ed. Benjamin/Cummings Pub. Co., Menlo Park, California.
32. Stent, G. 1981. Strength and weakness of the genetic approach to the development of the nervous system. Ann. Rev. Neurosci. 4:163-194.
33. Phillips, D. I. W. 1993. Twin studies in medical research: can they tell us whether diseases are genetically determined? The Lancet. 341:1008-1009
34. Bracha, H. S., Torrey, E. F., Bigelow, L. B., Lohr, J. B., and Linigton, B. B. 1991. Subtle signs of prenatal maldevelopment of the hand ectoderm in schizophrenia: a preliminary monozygotic twin study. Biol. Psychiatry. 30:719-725.
35. Egeland, J. A., Gerhard ,D.S., Pauls, D.L., Sussex, J.N., Kidd, K.K., Allen,C.R., Hostetter, A.M., & Housman, D.E. 1987. Biopolar affective disorders linked to DNA markers on chromosome 11. Nature 325, 783-787.
36. Kelsoe, J. R., Ginns, E. I., Egeland, J. A., Gerhard, D. S., Goldstein, A. M., Bale, S. J., Pauls, D. L., Long, R. T., Kidd, K. K., Conte, G., Housman, D. E., and Paul, S. M. 1989. Re-evaluation of the linkage relationship between chromosome 11p loci and the gene for bipolar affective disorder in the Old Order Amish. Nature 342:238-243.
37. Hodgkinson, S., Sherrington, R., Gurling, H., Marchbanks, R., Reeders, S., Mallet, J., Mcinnis, M., Peturrson, H., and Brynjolfsson, J. 1987. Molecular genetic evidence for heterogenity in manic depression. Nature. 325:805-806.
38. Baron, M. 1991. X-linkage and manic-depressive illness: a reassessment. Social Biol. 38:179-88.
39. New hunt on for bipolar genes. 1993 Science 262:651
40. Gottlieb, G. 1992. Individual Development and Evolution In:The Genesis of Novel Behavior. Oxford University Press, Oxford.
41. Levine, A. J., Momand, J., and Finlay, C. A. 1991. The p53 tumour suppressor gene. Nature 351:453-456.
42. Jacks, T., Faneli, A., Schmitt, E. M., Bronson, R. T. Goodell, M. A. and Weinberg, R. A. 1992. Effects of an Rb mutation in the mouse. Nature 359:295-300.
43 Donehower, L. A., Harvey, M., Slagle, B. L., Mcarthur, M. J., Montgomery Jr., C. A., Butel, J. S., and Bradley, A. 1992. Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumors. Nature 356:215-221.
44. Baker, S. J., Markowitz, S., Fearon, E. R., Willson, J. K., and Vogelstein, B. 1990. Suppression of human colorectal carcinoma cell growth by wild-type p53. Science 249:912-5.
45. Pietenpol, J. A. and Vogelstein, B. 1993. No room at the p53 inn. Nature 356:17-18.
46. Dutta, A., Ruppert, J. M., Aster, J. C., Winchester, E. 1993. Inhibition of DNA replication factor RPA by p53. Nature 365:79-82.
47. Duesberg, P. H. and Schwartz, J. R. 1992. Latent viruses and mutated oncogenes: no evidence for pathogenicity. Progress in Nucleic Acid Research and Molecular Biology. 43:135-204.
48. Knudson, A. 1985. Hereditary cancer, oncogenes, and antioncogenes. Cancer Res. 45:1437-43.
49. Benedict, W. F., Banetjee, A., Mark, C., and Murphee, A. 1983. Nonrandom chromosomal changes in untreated retinoblastoma. Cancer Genetics and Cyotgenetics . 10:311-333.
50. Gardner, H. A., Gallie, B. L., Knight, L. A., and Phillips, R. A. 1982. Multiple karyotypic changes in retinoblastoma tumor cells: presence of normal chromosome No. 13 in most tumors. Cancer Genetics and Cyotgenetics 6:201-211.
51. Bookstein, R., Shew, J. Y., Chen, P. L., Scully, P., and Lee, W. H. 1990. Suppression of tumorigenicity of human prostate carcinoma cells by replacing a mutated RB gene. Science 247:712-5.
52. Xu, H. J., Sumegi, J., Hu, S. X., Banerjee, A., Uzvolgyi, E., Klein, G., Benedict, W. F. 1991. Intraocular tumor function of RB reconstituted retinoblastoma cells. Cancer Research 51:4481-5.
53. Colditz, G. A., Willett, W. C., Hunter, D. J., Stampfer,M. J., Manson, J. E., Hennekens, C. H., Rosner, B. A., and Speizer, F. E. 1993. Family history, age, and risk of breast cancer. JAMA. 270(3):338-343.
54. Rubin, H. 1985. Cancer as a dynamic developmental disease. Cancer Research 45:2935-2942.
55. Farber, E. and Rubin,H. 1991. Cellular adaptation in the origin and development of cancer. Cancer Research 51:2751-2761.
56. Hall, J. M., Lee, M. K., Newman, B., Morrow, J. E., Anderson, L. A., Huey, B. and King, M. C.1990. Linkage of early-onset familial breast cancer to chromosome 17q21. Science 250:1684-1689.
57. Tsui, L. 1992. The spectrum of cystic fibrosis mutations. TIG. 8(11):392-398
58. The Cystic Fibrosis Genotype-Phenotype Consortium, 1993. Correlation between genotype and phenotype in patients with cystic fibrosis. N. Engl. J. Med. 329(18):1308-1313.
59. Collins, F. S. 1992. Cystic Fibrosis: Molecular biology and therapeutic implications. Science 256:774-777 .
60. Miller, S. S., Jiang, C., Finkbeiner, W. E., Widdicombe, J. H., McCray Jr., P. B. Altered fluid transport across airway epithelium in cystic fibrosis. Science 262:424-427
61. Black, W.C. and Welch, H.G. 1993 Advances in diagnostic imaging and overestimation of disease prevalence and the benefits of therapy. NEJM. 328(17):1237-1243.
62. Harach, H. R., Franssila, K. O., Wasenius, V. M. 1985. Occult papillary carcinoma of the thyroid: a "normal" finding in Finland: a systematic autopsy study. Cancer 56:531-538
63. Moertel, C. G., Fleming, T. R., Macdonald, J. S., et al. 1993. An evaluation of the carcinoembryonic antigen (CEA) test for monitoring patients with resected colon cancer. Jama. 270(8):943-947.
64. Fortuin, M., Chotard, J., Jack, A. D., Maine, N. P. Mendy, M., Hall, A. J., Inskip, H. M., George, M. O., Whittle, H. C. 1993. Efficacy of hepatitis B vaccine in the Gambian expanded programme on immunisation. Lancet 341:1129-1131.
65. Tabor, E., Gerety, R. J., Drucker, J. A. 1978. Transmission of non-A, non-B hepatitis from man to chimpanzee. Lancet 1:463-465
66. Reyes, G. R., Baroudy, B. M. 1991. Molecular biology of non-A, non-B, hepatitis agents: Hepatitis C and Hapatitis E viruses. Advances in Virus Research 40:57-102.
67. Choo, Q. L., Kuo, G., Weiner, A. J., Overby, L. R., Bradley, D. W., and Houghton,M. 1989. Isolation of a cDNA clone derived from a blood-born non-A, non-B viral hepatatitis genome. Science. 244:359-362.
68. Weiner, A. J., Kuo, G., Bradley, D. W.,Bonino, F., Saracco, G., Lee, C., Rosenblatt, J., Choo, Q. L., and Houghton, M. 1990. Detection of hepatitis C viral sequences in non-A, non-B hepatitis. Lancet . 335:1-3.
69. Selby, M. J., Choo, Q., Berger, K., Kuo, G., Glazer, E., Eckart, M., Lee, C., Chien, D., Kuo, C., Houghton, M. 1993. Expression, identification and subcellular localization of the proteins encoded by the hepatitis C viral genome. Journal of General Virology 74:1103-1113
70. Seeff, L. B., Buskell-Bales, Z., Wright, E. C., Durako, S. J., Alter, H. J., Iber, F. L., Hollinger, F. B., Gitnick, G., Knodell, R. G>, Perrillo, R. P., Stevens, C. E>, Hollingsworth, C. G. 1992. Long-term mortality after transfusion-associated non-A, non-B hepatitis. NEJM . 327:1906-1911
71 Czaja, A.J. 1992. Chronic hepatitis C virus infection - A disease in waiting? NEJM 327(27):1949-1950
72. Thomas, C., Mullis, K. B., Ellison, B. J. and Johnson,P. 1993. Why there is still an HIV controversy. Nature. submitted in November 1993.
73. Duesberg, P.H. 1993. AIDS acquired by drug consumption and other noncontagious risk factors. Pharmac. Ther. 55:201-277
74. Aldhous, P. 1992 The promise and pitfalls of molecular genetics. Science 257:164-165.
75. Plomin, R., DeFries, J. C., and McClearn, G. E. (1990) Behavioral Genetics, 2nd ed. W. H. Freeman, New York.
76. Garber, A. M., Littenberg, B., Sox, H. C., Wagner, J. L., and Gluck, M. 1991. Costs and health consequences of cholesterol screening for asymptomatic older Americans. Arch. Internal Med.. 151(6):1089-1095.
77. Moore, T. J. 1989. Heart Failure: A Critical Inquiry Into American Medicine and The Revolution in Heart Care. Random House, Inc., New York.
78. Fishel, R.,Lescoe, M.K., Rao, M.R.S., Copeland, N.G., Jenkins, N.A., Garber, J., & Kolodner, R.D. 1993. The human mutator gene homolouge MSH2 and its association with hereditary non-polyposis colon cancer. Cell 75:1027-1038.
79. Sapp, J. 1987) Beyond the Gene: Cytoplasmic Inheritance and the Struggle for Authority in Genetics. Oxford University Press, Oxford.
80. Clark, M. M., Karpluk, P. and Galef, B. G. 1993. Hormonally mediated inheritance of acquired characteristics in Mongolian gerbils. Nature . 364:712
81. Nanny, D.L. 1982. Genes and phenes in Tetrahymena. BioScience. 32(10):783- 788.
82. Wilson, A. C., Carlson, S. S., and White, T. J. 1977. Biochemical Evolution. Ann. Rev. Biochem. 46:573-639.
83. Maddox,J. 1992. Is molecular biology yet a science? Nature 355: 201
84. Maddox,J. 1992. Finding wood among the trees. Nature 333:11
85. Elsasser, W 1987. Reflections on the Theory of Organisms, Orbis Publishing, Quebec.
86. McClintock,B. 1984. The significance of responses of the genome to challenge. Science 226: 792-801.
87. Bond,W.C., Bohs,C., Ebey,J., & Wolf,S. 1973. Rhythmic heart rate variability (sinus arrhythmia) related to stages of sleep. Conditional Reflex 8(2):98-107.
88. Skinner,J.E., Goldberger, Mayer-Kress,G.,& Ideker,R,E. 1990. Chaos in the heart: implications for clinical cardiology. Biotechnology 8:1018-1033.
89. Cripps,T.R., Malik,M., Farrell,T.G., & Camm,A.J. 1991. Prognostic value of reduced heart rate variability after myocardial infarction:clinical evaluation of a new analysis method. Br. Heart J. 65:14-19.
90. Basar,E. 1993. Chaotic dynamics and resonance phenomena in brain function: progress, perspectives, & thoughts. In Chaos in Brain Function ed. E.Basar. Springer-Verlag. Berlin