New Methods for the Comprehensive Assessment of Changes in Gene Expression During the Life Course

The expression of a large number of genes changes during the life course. These changes are thought to underlie the mechanism of tissue differentiation during development, the modulation of physiology during pubescence, the maintenance of physiological homeostasis in the adult, the reaction of tissues to disease states, and, perhaps, some fundamental processes of aging. Proteins, rather than the ribonucleic acid messages (messenger RNA or mRNA) that relay the information in the DNA genetic code to the protein synthesizing machinery of cells, are the proximal agents of these modulations. It is difficult, however, to comprehensively assess changes in proteins over the life course. Only a small proportion of the tens of thousands of proteins in a cell can be specifically identified by available antibodies. Many critically important proteins, such as transcription factors, which regulate the expression of many other genes, are present in only trace amounts. Methods such as two-dimensional electrophoresis lack sufficient sensitivity for their identification. Highly specific sensitive-quantitative methods have been developed, however, for the identification of the levels of a wide range of mRNA in cells and tissues. These offer the best hope for the development of a battery of molecular biomarkers of aging, although the protein approach, proteomics, is being vigorously pursued by a number of laboratories.

The physical vehicles for the large-scale screening of mRNAs are sometimes referred to as massive microarrays or gene chips because the thousands of stretches of the DNA sequences that represent specific genes that allow one to identify levels of expression of mRNAs are systematically arrayed as minute samples in rows and columns on various physical substrates, such as glass slides, membranes, or composite materials. The DNA sequences are of two types. The approach pioneered by the Affymetrix company (Table 6-1) uses short strings of the nucleotide coding units of a single strand of DNA (oligonucleotides) from several exons of interest. Exons are the multiple, physically separated segments of the DNA sequence of a gene that are spliced together to provide the mRNA; the intervening sequences, which may have regulatory information, are called introns. These oligonucleotides hybridize with mRNA that is purified from various cells or tissues (e.g., from cells or tissues from a young individual) and labeled with radioactive or fluorescent tags. For the latter, one can label the material from one specimen with a fluorescent chemical that, when optically excited by a laser, emits a signal in the green region of the spectrum of light. A different set of mRNAs (for example, from the homologous cell type or tissue from an older subject) can be labeled with another fluorescent compound, such as one that emits a red signal. These two sets of labeled mRNAs can be mixed and allowed to jointly hybridize to the oligonucleotide array of human cognate genes to determine the intensity of hybridization. If the expression of a given gene does not change with age, one would observe a yellow signal (an equal mixture of the red and green signals). A dominance of a red or green signal would indicate either an increase or decrease in the expression of that gene in the cells or tissues of the older individual.

TABLE 6-1

Some Commercial Resources for Assay of Human Gene Expression Using cDNA Microarrays.

A second approach uses cDNAs instead of DNA oligonucleotides to make the microarrays. The symbol “cDNA” refers to a DNA copy made from the cognate mRNA, using an enzyme called reverse transcriptase. These cDNAs are denatured so that single strands are available to hybridize with single-stranded molecules of mRNA, forming duplex molecules. The cDNAs are typically arrayed on glass slides or membranes. Directions for making such arrays can be found in the web site of the laboratory of Dr. Pat Brown of Stanford University (http://cmgm.stanford.edu/pbrown/). Commercially available sources of such microarrays can now be obtained, and can include access to software for the analysis of the results. Table 6-1 summarizes some current major sources of such materials. The reader will note, however, that these are very expensive reagents. The associated informatics software (e.g., Rosetta Inpharmatics, Inc., http://www.rii.com/about/overview.htm) required for the analysis of the vast amounts of data that emerge from such studies, including algorithms for cluster analysis, can also be very expensive. Moreover, there is still a lack of rigorous statistical analysis of many of the results. The need for replicate independent experiments for such statistical analysis will put such an approach out of the range of most laboratories and certainly could not reasonably be applied for the study of large populations.

This situation is likely to change, however, for several reasons. First, improvements in the technology are likely to decrease the variance. Second, the enormous interest and demand for such approaches will certainly lead to substantial decreases in costs. Third, more modest arrays designed for the investigations of genes of special interest to specific studies are likely to be developed by individual laboratories or by specialized centers within major institutions at reduced costs. Fourth, and most importantly, once a few particularly robust biomarkers of specific processes of aging have been discovered and replicated using microarrays, one can employ other, more rigorously quantitative and more rapid methods based upon the polymerase chain reaction (PCR).

PCR greatly amplifies small amounts of DNA, either directly from genomic DNA or from mRNA that has been converted to DNA using reverse transcriptase enzymes. The authors have successfully employed one such method in studies of groups of subjects with autopsy-documented dementias of the Alzheimer type (DAT) and compared the results with materials from other neurological degenerative disorders and neurologically normal age-matched controls. We found changes in the expression of a specific gene in relatively unaffected portions of the brain of DAT patients that we believe could represent a relatively early stage in the evolution of the disease process (Hu et al., 2000). The method is called quantitative reverse transcription-competitive PCR (RT-cPCR). The assay is extremely sensitive and very specific. For example, we could measure, in the same reaction, two alternatively spliced forms of the same gene (called isoforms) that differed in only six nucleotide base pairs. This should be considered in the context of the fact that each of the two sets of human genes in an individual cell has about 3.1 billion base pairs. The isoform with the extra base pairs, which results in a protein with an additional two amino acids, turned out to be an exclusive marker of neuronal cells (Hu et al., 1999). It would be feasible to perform approximately 1,000 such assays, which would score for both isoforms, in no more than 50 days. We estimate the cost per assay of both isoforms to be no more than $7, including the salary of a technician and the costs of all reagents. The application of robotic methods should greatly decrease the cost per specimen for very large-scale studies.

Once suitable conditions have been worked out, the methodology could be applied, in principle, to any species of mRNA from any source of mRNA, including cells from peripheral blood samples; from small biopsies of subcutaneous fat, skin, or skeletal muscle; or from cultured cells derived from such biopsies. There are already interesting clues as to what general classes of mRNA are deserving of further analysis in larger populations of aging animals and aging human subjects. A microarray analysis of the skeletal muscles of aging cohorts of mice revealed alterations of gene regulation involving loci included in several functionally related disease clusters (Lee et al., 1999). These included genes involved in energy metabolism, protein metabolism, DNA repair, and in the responses of cells to a variety of stress factors, such as oxidative and thermal stress. A large proportion of these changes could be reversed by partial caloric restriction, a well-established method in rodents for the enhancement of life span and the postponement of several late-life diseases. An interesting example of a specific gene whose expression was markedly diminished in aged muscles was one coding for a mitochondrial protease that is involved in the biogenesis of mitochondria, an organelle that is in abundance in skeletal muscles and that is the powerhouse of our cells. The use of expression arrays for the study of aging in human subjects has so far been confined to cultured skin fibroblast-like cells from only a few subjects (Shelton et al., 1999; Ly et al., 2000), but there is much more research in progress.

Other Uses of Microarray Technologies

These microarray technologies could also be adapted to determine the prevalence of a wide variety of inborn alterations in gene dosage, either deletions or DNA amplifications (Pinkel et al., 1998; Trent et al., 1997). To do this, one hybridizes fragments of genomic DNA, rather than messenger RNA, against an array of cognate DNA sequences. This approach has only just begun to be applied to human tissues and has so far been used only to demonstrate that one can get linear signals of gene dosage for genes on the X chromosome over the range of one to five copies of X chromosomes. Given a number of technical improvements, we may soon be in the position of being able to determine the contribution of inborn changes in gene dosage to late-life disorders. Malignant neoplasms are cogent examples. We know that individuals born with only one copy of a tumor suppressor gene are more likely to develop cancer or to progress towards cancer as a result of a lesion developing in the remaining good copy of that gene sometime during the individual's life span. There are likely to be hundreds of such genes. It would be of great interest to compare the relative prevalence of deletions in such genes in various populations and in different epochs. There is epidemiological evidence of secular trends towards the postponement of reproduction in many developed countries (Armitage and Babb, 1996; Guyer et al., 1998; Speroff, 1994; Ford and Nault, 1996; Kaufmann et al., 1998; Hoshi et al., 1999). It is known that advanced paternal age is associated with increased risk of various mutations (Risch et al., 1987; Vogel and Ralhenberg, 1975). James Crow, a leading scholar in this field, has concluded that “the age of the father is the main factor determining the human spontaneous mutation rate, and probably the total mutation rate” (1993). An example of the importance of these issues to late-life diseases is the recent evidence that advanced paternal age is associated with increased risk of prostate cancer in offspring (Zhang et al., 1999). Could this be due, in part, to an increasing burden of deletions in tumor suppressor genes?

The use of massive oligonucleotide arrays can also be used for the detection of a wide range of polymorphisms (Halushka et al., 1999; Sapolsky et al., 1999; Chee et al., 1996). In contrast to mutations, which are rare events (∼10^–6), polymorphisms represent common genetic variants (alleles). By definition, their allelic frequencies are at least 1-2 percent of all alleles at the particular genetic locus. Polymorphisms are being used for a very exciting emerging technology involving the use of ordered arrays of human genes cloned in what are called bacterial artificial chromosomes (BAC clones). These can be used for mapping disease genes via hybridization of genomic DNA using the principles of genomic mismatch repair and linkage disequilibrium (Cheung et al., 1998). (For further details, see the internet site of the laboratory of Vivian G. Cheung, http://genomics.med.upenn.edu/vcheung/.) The method takes advantage of the fact that we humans have single nucleotide polymorphisms (SNPs) at an average of at least one for each one thousand of our base pairs. Chromosomal domains indicative of identity by descent for such SNPs permit one to map and clone disease genes of interest in the context of a vast amount of information on other regions of the genome. The term “identity by descent” refers to the sharing of an allelic form of a gene in two individuals because they have both been inherited from the same common ancestor. By contrast, the term “identity by state” refers to the situation in which such sharing is coincidental. This technology may well be the method of choice for the genomic characterization of large populations.

The polymerase chain reaction (PCR) has also been adapted to the analysis of tissue sections so that levels of gene expression (Barlati et al., 1999; Lee et al., 1996), the detection of genomic deletions (Chiang et al., 1999), and the detection of infectious agents (Fredericks and Relman, 1996; Montone and Litzky, 1995) can be related to specific cell types. The development of methods for the quantitative analysis of gene expression in single cells is of exceptional importance to gerontology because as people age, there is a shift in the population heterogeneity of complex tissues. Methods such as preparative cell sorting can provide one solution to this problem (Darzynkiewicz et al., 1994). It is often used, for example, to analyze specific subsets of peripheral blood lymphocytes.

Genetic Linkage Studies: The Identification of Pedigrees and Their Use for the Discovery of Major Genes that Modulate Life Span and Senescent Phenotypes

Although not systematic, large-scale screenings of populations with specific phenotypes of interest to gerontologists have led to the identification of unusual pedigrees with patterns of segregation consistent with the hypothesis that mutation at a single major genetic locus is responsible for the phenotype. These studies take advantage of the availability of a large number of genetic markers and the principle that a marker that is in close physical proximity to a disease gene is rarely separated from that disease gene during meiosis, thus “traveling” together from generation to generation. This is the basis of genetic linkage analysis, which permits the mapping of a disease gene to a specific segment of a specific chromosome. Given such mapping, it is possible to eventually clone the responsible mutant gene. The discoveries of autosomal dominant mutations at the beta amyloid precursor protein locus and at the two presenilin loci are good examples (Blacker and Tanzi, 1998). These mutations cause early-onset DAT. Mutation at the presenilin 1 locus is particularly virulent, leading to dementia as early as age 26 in a pedigree with the diagnosis confirmed by autopsy (Martin et al., 1991). These mutations affect virtually all carriers if they live long enough. The dementias are usually interpreted as resulting from dominant gain-of-function mutations, meaning that the mutation produced a protein having a new, deleterious function. The evidence is far from definitive on this point, however.

It is quite interesting that there is as yet no convincing evidence for autosomal recessive loci resulting in unusual susceptibility to DAT. Such mutations represent a loss of function, typically of an enzyme. One certainly has reason to suspect that such mutations might be found if one were to examine very large numbers of pedigrees in different parts of the world. One would especially want to search within populations having high degrees of consanguineous marriages, as two rare recessive mutations are more likely to show up in the offspring of such matings. Examples would include populations of Hindus in the Southern part of India, where first cousin and uncle-niece marriages are still prevalent. Muslims throughout India and other parts of the world, such as the Middle East, also sanction first cousin marriages. The advantage of a focus on India, of course, is that its population continues to grow at a rapid rate and is now approaching one billion.

The rationale for suspecting that an enzyme deficiency can cause a form of DAT comes from current views concerning its pathogenesis. These views center upon the processing of the beta amyloid precursor protein (Selkoe, 2000). This processing is carried out by enzymes that catalyze the cutting of the protein at various positions in the sequence of its amino acids. A “bad” way to process the protein is with enzymes known as beta and gamma secretases. The products of such processing are small peptides, called beta amyloids; these form insoluble aggregates that deposit in the brain, including its blood vessels. Beta amyloid is thought by many to be a major cause of the loss of synaptic connections between neurons and of the neurons themselves. There is evidence that a product of the presenilin 1 gene is in fact a gamma secretase or a protein closely associated with the relevant gamma secretase (Li et al., 2000). The “good” way to process the beta amyloid precursor protein is with an enzyme known as alpha secretase. It cuts roughly in the middle of the piece of protein destined to make the beta amyloid peptide, thus preventing the accumulation of this substance. It is therefore reasonable to believe that a deficiency of alpha secretase would accelerate the development of DAT.

Genetic diseases caused by single Mendelian genes, whether autosomal dominant, autosomal recessive, or X-linked recessive, although quite rare, are relatively easy to diagnose. Nowadays, once diagnosed, it is relatively straightforward to map and clone the responsible gene, especially if multiple pedigrees with the same phenotype or large affected kindreds can be identified. This is the reason why screening very large populations or focusing on screening a particular geographic area or ethnic group is important. An example of the importance of the latter strategy is the success in collecting a group of related pedigrees that permitted the mapping and cloning of the presenilin 2 gene mutation. It turned out that the original set of affected pedigrees were all ethnic Volga Germans (Bird et al., 1988). Their ancestors migrated from the Hesse area of Germany to one of two small villages on the west bank of the Volga River in about 1760. They did this at the behest of Catherine the Great, the Czarina of Russia, who was a fellow German. This “genetic founder effect” gave assurance that one was dealing with a common cause of familial DAT. One could therefore map the responsible mutations even with a relatively small number of pedigrees, since they were basically part of one large kindred.

All three autosomal dominant DAT mutations, those at the presenilin 1 and 2 loci and those at the beta amyloid precursor protein locus, result in relatively early-onset dementias. There is considerable variation in the age of onset for the case of the presenilin 2 mutations, however, including rare apparent “escapees” who may live into the ninth decade without becoming affected.

Pedigrees are now being collected by T.T. Perls and others for the purpose of using linkage analysis to identify genes that contribute to unusual longevities. One rationale for such a study was a finding consistent with the conclusion that siblings of centenarians tended to have relatively long life spans (Perls et al., 1998). It is too early to evaluate the success or failure of these studies, but there are reasons for pessimism. First of all, life span is subject to stochastic events (Finch and Kirkwood, 1999). Cancer, for example, can result in chance “hits” at particularly vulnerable parts of the genome (e.g., tumor suppressor genes). Two individuals might have the same basic rate of somatic mutations, but one may be lucky and never sustain “hits” in a tumor suppressor gene, thus escaping premature death from a cancer. Second, there is also concern that, at least for the case of human subjects, there are likely to be a very large number of genes impacting upon late-life disorders and, hence, upon life span (Martin, 1978). A proportion of such genes may be of exceptional importance, however. One such example may be the gene coding for a gene that helps to unwind double stranded DNA; when one inherits two doses of a severe mutation in that gene, it causes the Werner syndrome (Yu et al., 1996). People with that disorder display such features as premature arteriosclerosis, a variety of cancers, cataracts, osteoporosis, type 2 diabetes, gonadal atrophy, atrophy of skin and subcutaneous tissues, and premature graying and thinning of hair (Epstein et al., 1966). Twin studies indicate that perhaps a quarter of the variance of life span in humans has a genetic basis (Herskind et al., 1996). A search for these genes by large scale genomic screening of pedigrees with members exhibiting unusual longevities may therefore reveal several important genetic loci, perhaps including the Werner locus. The ascertainment of such families will require large-scale screening of populations and documentation of birth dates.

Genetic Association Studies and the Difficulties Inherent in Case-Control Designs

Most cases of DAT are not associated with single autosomal dominant genes and occur well after the age of 65 years, typically after the age of 85 years, when perhaps 50 percent of the population can be affected (Katzman and Kawas, 1998). Such a high “background” of affected individuals makes the search for genetic susceptibility factors very difficult. There is also the difficulty of matching cases and controls in genetic association studies. Typically, one sees the statement, in publications, that only Caucasian subjects were studied. But what is a Caucasian? That word obscures the enormous genetic heterogeneity in the U.S. population. Thus, differences in the prevalence of a particular polymorphism may be biased by an enrichment, in either controls or in diseased subjects, of a particular ethnic subtype having a different distribution frequency of the polymorphism of interest. This leads to both false positives and false negatives. These uncertainties have led to the development of various pedigree-based genetic association studies, including the use of various sib-pair strategies. The failure to replicate an apparently robust conclusion, however, may simply be due to the fact that the polymorphism of interest may not be relevant in a particular population in a particular environment. Despite these difficulties, however, there is evidence that polymorphic variations at a number of different genetic loci each contributes to differential susceptibility to DAT. Most of these are likely to make only small contributions to susceptibility. The best example of a major modulator of susceptibility to the common late-onset forms of DAT is the apolipoprotein E (APOE) polymorphism (Corder et al., 1993). There are likely to be other major modulators, however; one of these loci may even have a much greater impact than APOE, at least in some populations (Daw et al., 2000).

There have been several confirmations (Castro et al., 1999) of the report (Schachter et al., 1994) that long-lived individuals such as centenarians are statistically more likely to carry the APOE e2 allele and less likely to carry the e4 allele, but with interesting regional variations (Panza et al., 1999). This is consistent with a deleterious role of APOE e4 on cardiovascular disease (Lehtinen et al., 1995) and on DAT (Corder et al., 1993). The robustness of the Alzheimer effect, however, varies among different ethnic groups, the impact being less for American black and Hispanic populations (Tang et al., 1996). These association studies are also consistent with the reported protective effect of the e2 allele against DAT (Corder et al., 1993). Preliminary results have indicated enrichments of other polymorphic alleles in certain populations (e.g., De Benedictis et al., 1999).

The Paucity of Genetic Investigations of Differential Rates of Change of Specific Physiological Functions in Middle Age and the Many Advantages of Such an Approach

From the point of view of gerontology, it can be argued that medical scientists have been preoccupied with deleterious phenotypes. One would very much like to know the genetic contributions to unusually robust retention of specific physiological functions during aging. I have argued that one should begin such studies with middle-aged cohorts. First of all, population geneticists have determined that the force of natural selection has essentially disappeared for phenotypes that do not emerge until after around age 40 or 45 (reviewed by Martin et al., 1996). Thus, we are in a position to detect differential rates of change of relevant phenotypes beginning in middle age. Secondly, assays for specific physiological functions are less likely to be compromised by various co-morbidities that would be the case were one to investigate differential rates of decline of specific functions in much older individuals. Thirdly, a genetic analysis would benefit from the potential availability of DNA samples for at least three generations when the index cases are middle-aged subjects. Fourth, there are much larger populations of middle-aged subjects, most of whom would be fully cooperative and capable of undergoing various stress tests without undue danger to their health. Fifth, middle-aged sibs of such index cases should be readily available, thus permitting sib-pair genetic studies. One such approach would be to screen a very large population of such subjects for individuals who score in the upper one percentile for a specific physiological function. One could then identify sib pairs exhibiting extreme concordance or extreme concordance for the trait of interest; such methods can greatly increase the statistical power of the genetic analysis (Zhang and Risch, 1996; Risch and Zhang 1995, 1996). Such an experimental design could lead to the definition of alleles that underlie what one might call “elite” aging, as opposed to mere “successful” aging. These alleles could serve as entry points to mechanistic studies that could lead to interventions of benefit to much larger segments of our population.

What physiological functions could be so investigated? The answer is that virtually all physiological functions could be studied, given the availability of sufficiently sensitive assays. Ideally, such assays should also be relatively inexpensive, relatively noninvasive, and amenable to longitudinal study. A possible example would be the delayed paragraph recall test for hippocampal function (Golomb et al., 1994). The tests can be made more difficult in order to detect unusual degrees of preservation of short-term memory.

Relevant Data from Anatomic Pathology

Surgical Pathology

In the United States, essentially all tissues removed at the time of surgery are examined by certified pathologists and their trainees. In the vast majority of cases, small samples are immersed in neutral formalin solutions within a few hours of their removal and fixed for periods of less than 24 hours, following which they are embedded in paraffin, sectioned to produce slices of around 6-8 microns in thickness, dehydrated in a series of ethanol solutions (with the unfortunate consequence of elimination of most lipids), stained with hematoxylin (for nucleic acids) and eosin (for proteins), and mounted on slides with coverslips for examination with the light microscope. Both the microscopic slides and the paraffin blocks are typically archived for many decades. This procedure is among the most venerable and informative in the entire field of medical practice and medical research. It is still the basis for diagnosis throughout the world. It provides a vast storehouse of definitive diagnostic historical information for large populations. It also provides the potential to uncover new definitive information with the use of an expanding toolkit of monoclonal antibodies to specific protein epitopes that have not been obscured by the formalin fixation. Nucleic acid probes, including those that provide amplification of genomic DNA, the DNA of infectious agents, and, to some extent, messenger RNA species, may also be applied to the archived materials. One can often carry out successful genotyping of loci of relevance to epidemiologists using such materials from long-deceased individuals. A good example is the tri-allelic polymorphism for the APOE gene, which greatly influences susceptibility to cardiovascular disease and dementias of the Alzheimer type, as noted above. It is in fact now standard practice to include APOE genotyping of all individuals involved in epidemiological studies of dementias.

At a few medical centers, segments of some surgically removed tissue are preserved cryobiologically. Cryopreserved surgical tissues can be utilized for a much greater variety of morphological, biochemical, and immunohistochemical studies. For optimal long-term preservation of tissues, preservation in liquid nitrogen is required. This expensive methodology, which requires careful monitoring and regular replenishments of nitrogen, might eventually be replaced, however, by long-term storage in improved mechanical freezers. A systematic program of cryopreservation of subsets of surgically removed tissues would provide vastly increased opportunities for population-based research on human health and disease. From such living tissues, a variety of cell types could be established in culture, providing unlimited amounts of materials for large-scale genotyping and investigations of variations of gene expression. The recently introduced massive microarray (chip) technologies, discussed above under the section on Genetics, could provide vast amounts of information given the availability of such tissues. Peripheral blood samples can of course provide suitable sources of DNA for genotyping, but they are much more limited as regards investigations of variations in gene expression, since one is dealing with only those cell types in circulating blood.

While seldom explored, cryopreserved surgically excised tissues can also be used for experimental investigations of organ or organoid cultures (Thesleff and Sahlberg, 1999; de Boer et al., 1996; Kruk and Auersperg, 1992; Kopf-Maier and Kolon, 1992). Unlike conventional cell cultures, these methods provide a more normal simulation of what obtains in the living organisms, including physiological cell densities and various interactions among cells and their matrix components. To the best of our knowledge, this powerful methodology has never been employed for any large-scale studies of materials from human populations.

Given the increasing economic constraints of medical services, preservation of even a subset of routinely obtained surgical tissues is very unlikely to be initiated without some form of subsidization by private and/or governmental agencies. This would of course require a prospective, peer-reviewed research proposal. Let us give a single possible example. Periodic flexible endoscopic examinations of the sigmoid colon and rectum have become standard medical procedures for subjects over the age of 50 years in order to detect and remove potentially pre-malignant hyperplastic and adenomatous polyps (Khullar and DiSario, 1997). A substantial fraction of older subjects are found to have such lesions, which are routinely fixed in formalin and examined microscopically. A small portion of such lesions and adjacent normal tissues, together with small biopsies from the normal tissues of subjects who are examined but who do not have polyps, could be routinely cryobiologically preserved or simply frozen as materials for a variety of research questions, given suitable informed consent. One such set of questions would relate to the recent surprising observations that many thousands of somatic mutations can be found in such lesions (Stoler et al., 1999) (Box 6-1). The methodology was the relatively simple yet informative inter-simple sequence repeat polymerase chain reaction, which can sample mutations over a very large portion of the genome. The use of frozen or cryopreserved mucosal/submucosal fragments for the detection of chemical and viral mutagenic agents and the response of these tissues to challenge by such agents would have the potential to elucidate underlying mechanisms. A viral agent known as the JC virus should be an interesting candidate as an environmental etiologic agent (Laghi et al., 1999; Neel, 1999).

BOX 6-1

Calculation of the Total Numbers of Large-Scale Mutations (insertions, deletions and translocations) in a Typical Colon Cancer Cell. Source: From D.C. Stoler et al., 1999, with modifications by George M. Martin that give a greater number of mutations (more...)

Somatic mutations, by definition, are changes in the sequence of nucleotide base pairs of DNA. Epigenetic changes also occur during aging. In contrast to mutations, these involve chemical changes that are superimposed upon a normal sequence of DNA. These chemical changes are of two principal types: the addition or removal of methyl groups from cytosines; and the addition or removal of acetyl groups from histones, which are basic proteins closely associated with DNA that are thought to participate in the modulation of the expression of large blocks of genes. These epigenetic changes may be of basic importance to our understanding of the pathobiology of aging (Holliday, 1993; Sinclair et al., 1998; Young and Smith, 2000). In the context of the present discussion, a particularly interesting epigenetic change has been documented to occur in aging epithelial cells of the human colonic mucosa. It has been found that there is a steady increase, with aging, in the levels of methylation of a region of the promoter of a gene coding for an estrogen receptor (Issa et al., 1994). Promoters are DNA sequences that are essential for the transcription of genes into mRNA. The methylation of promoters is generally associated with gene silencing. Estrogen receptors participate in the regulation of the proliferation of colonic mucosal cells. When the receptor is active, the binding with estrogen (forms of which circulate in both males and females) inhibits proliferation. Methylation of the estrogen receptor thus leads to a dysregulation of the proliferative homeostasis of colonic epithelial cells. Not surprisingly, regions of the colon in which this occurs are exactly those that are highly susceptible to colonic cancer (Issa et al., 1994).

Population scientists should participate in such research, as the prevalence of colon cancer varies substantially in different regions of the country (Devesa et al., 1999; some of these data are available as county-specific maps on the National Cancer Institute website: www.nci.nih.gov/atlas/). Is this related to regional differences in diet or environmental mutagens such as the JC virus? Are there secular changes in the prevalence of the molecular alterations described above? Is the epigenetic component a robust marker of the aging colonic mucosa in all populations?