Genetic Diagnosis of Endocrine Disorders

PhD                     

CHAPTER 1

INTRODUCTION

Mutation is a sudden, inheritable change in the genome [1]. The change needs to be sudden, that is, the change must be present in one cellular generation but not in the preceding generation. Mutations were originally defined as necessarily involving the gametes, but somatic mutations are now recognized as an important process. The phenomenon of recombination will often couple two sequences to give a molecule with such novel properties that it appears to have been a mutation, but geneticists have not usually classified such events as mutations. As will be seen in the discussion of chromosome aberrations (below) this separation between recombination and mutation is not clear-cut. The change also needs to be transmitted. An alteration in the DNA structure that prevents its replication is an inactivation, but not a mutation. On the other hand there are reversible changes possible in the structure, but not the coding properties of the nucleotides, for example, cytosine or adenine methylation. These changes can be inherited, but since the pairing properties of the nucleotide are not altered they are considered as epigenetic change, notwithstanding the large effect such modifications can have on function. Mutations may involve single nucleotides, in which case we speak of point mutations. A few or many nucleotides may be added or deleted. Whole genes or groups of genes may be deleted, duplicated or moved to other chromosomes, or whole chromosomes may be added or lost as a result of errors in cell division. All such changes may have drastic effects on the life of the organism. On the other hand, many changes in the DNA have no discernible effect on function. As long as the interest of geneticists concentrated on Mendelian traits (or phenotypes), attributable to the action of a single gene, it was relatively simple to distinguish functionally significant mutations. The discovery that humans may differ at approximately one nucleotide in every thousand or, given a genome size of about three billion, at three million possible sites, the single nucleotide polymorphisms or SNPs complicate the attempt to determine which changes are significant [2]. An additional complexity comes from the discoveries that differences in sequence statistically associated with different diseases are often found to be located in gene deserts, regions in which there are no known genes [3].

Although only about 25,000 protein coding genes have been recognized, these account for less than 5% of the total DNA. However, it appears that almost half (50%) of the total DNA is transcribed. At least some of these transcripts have a function. Up to one thousand are processed into small (20–23 nucleotide) sequences called microRNA (miRNA) with key roles in regulating a variety of cell functions. The key players in this regulation are nucleotides 2 to 8 of the miRNA, the so-called seed region [4, 5].

Insofar as the sequences coding these miRNAs have a physiological effect on the organism, they need to be considered as genes in the same sense as protein coding genes. The DNA giving rise to these miRNAs is subject to mutation by the same mechanisms for base change discussed in the chapter. In addition, RNA editing changes the pairing specificity of sequences by conversion of particular adenosines (A) to inosine (I) by deamination. It is conceivable that alterations in the editing system could result in major changes in the function of particular miRNAs. Recently two point mutations in the seed region of one of the miRNA genes, miR-96, have been shown to segregate in a Mendelian way and to be responsible for a form of deafness [6]. It is likely that single base changes in other miRNA genes will be identified with other cases of inherited disease.

The total amount of DNA involved in coding for micro-RNAs is small, even though the initial transcripts before processing are several kilobases long. Even 1000 several kilobase genes would only constitute about 0.1% of a total genome of 3.3 × 10⁹ bases. However, we may yet find out that the 45% or so of transcribed DNA not accounted for, but mutable as a result of the processes described above, has some function. We are still at the beginning of our understanding of the structure of the human genome.

POINT MUTATIONS

We know more about the mechanisms by which point mutations occur and of their effects than any of the other changes (Table 1.1). The original association of a single point mutation with a particular disease was the discovery that sickle cell anemia was, in Pauling’s words, a molecular disease since all of its signs can be traced to a substitution of a glutamic acid by a valine in the hemoglobin molecule. This change was later shown to be due to a change from GAG (glutamic acid) to GTG (valine) [7]. The association of genetic change with a particular disease dates back at least a century to Garrod’s discovery of the Mendelian inheritance of alkaptonuria, but Garrod had no idea of the nature of the genetic change. The second reason is methodological. Before the age of PCR and of DNA sequencing, mutations were most easily studied in the bacteria and the bacterial viruses. Ernst Freese and Seymour Benzer first systematized the possible single base changes. These investigators and their co-workers coined the terms transition to denote the change from one purine to another or of one pyrimidine to another. The four possible transitions are cytosine (C) to thymine (T) and its reverse, and adenine (A) to guanine (G) and its reverse. Freese and Benzer defined transversions as changes from a purine to a pyrimidine or the reverse. So a change from an A or a G to a C or a T, and the reverse C or T to A or G were defined as transversions. Shortly afterwards, Brenner pointed out that some of the putative transversions were actually additions or deletions of one or two nucleotides [8]. These were later called frameshifts, because of the discovery that the genetic code was formed of triplets, groups of three nucleotides read together to specify a particular amino acid. The addition (or deletion) of any number of nucleotides not divisible by three would result in a change in the reading frame of three nucleotides at a time, thereby changing the amino acid composition of all amino acids downstream of the coding change. The details of the genetic code, as elucidated in the 1960s, also indicated that such frameshifts might not only result in major changes in the amino acid composition of a protein but might also produce unexpected termination codons as a result of the shift. Point mutations that resulted in protein terminations were at first termed nonsense mutations as opposed to missense mutations that resulted in the substitution of one amino acid for another. The nonsense mutations did not make sense, i.e. did not specify any amino acid. There are three such codons, now called termination (ter) codons: UAA, UAG, and UGA. Since messenger RNA is the molecule that is actually read by the protein synthesizing machinery, the code is an RNA code with U(racil) substituted for T(hymine). One of the stop codons, UGA, is read as tryptophan in mitochondria and the mitochondrial code includes a few other variations: AGG and AGA are mitochondrial stop codons instead of coding for arginine, and AUA codes for methionine instead of isoleucine. There are 64 possible codons but only 20 (natural) amino acids and the code is degenerate, or redundant, in that several codons can specify the same amino acid. Point mutations within genes that do not change the meaning (amino acid coded) of the codon are termed synonymous or silent as opposed to non-synonymous changes. Although there was some initial confusion about the necessity of punctuation between the triplet codons, it was realized that if reading of the code began at a fixed site, and if the reading frame was designed to read three nucleotides at a time, the correct sequence of amino acids would automatically be produced.

This terminology was developed before it was realized that there were large amounts of non-coding DNA. Even though it was (is) used, frameshift has no meaning for mutations within such non-coding regions. A more recent term for small insertions or deletions, regardless of their physiological effect is indel. Notwithstanding, or because of the universal nature of the code, there still remain some mysteries. The codon AUG specifies methionine and it also signals start, but not always. UGA specifies stop but some UGA codons specify the 21st amino acid, selenocysteine. The recoding of UGA is determined by the surrounding sequence, a characteristic stem loop located in the 3'-end of mammalian mRNAs.

MUTAGENIC AGENTS

It was first assumed that the fidelity of normal replication stemmed from the stability of the A:T and G:C base pairs resulting from hydrogen bonding. The success of the early workers on the molecular nature of mutations came from their ability to account for the specificity of a variety of mutagenic base analogs and other mutagenic agents. They were able to draw acceptable alternative base pairings resulting from the incorporation of these compounds into DNA, or by their distortion of the replication machinery. For a number of years research on mutational mechanisms consisted largely of formulations of how changes in DNA structure resulting from treatment with, or incorporation of, mutagenic chemicals could change the base pairing properties of the replicating DNA so that mistakes were made. The supposition by Bruce Ames that Carcinogens are Mutagens [9] (which did not, it turned out, mean that all carcinogens are mutagens) prompted chemists to study the changes in DNA structure produced by reaction with chemical and physical mutagens. Alkylating agents such as methyl nitrosourea and the chemotherapeutically active mustard gas derivatives were shown to react with individual nucleotides to produce multiple changes. Production of O6 methyl guanine by agents such as methyl nitrosourea or methyl nitronitrosoguanidine was shown to promote mistaken base pairing, making understandable the highly mutagenic characteristics of such compounds. A major development was the discovery that metabolic systems in the host activated ingested compounds making it possible for them to react with DNA. Carcinogenic polycyclic hydrocarbons and aflatoxins are converted to epoxide derivatives with the participation of the cytochrome p450 system. These epoxides react directly with DNA producing mutagenic adducts. The Ames assay, a bacterial test for mutagenic activity, was modified to account for such activation by the incorporation of a liver extract to media on which presumptive mutagens were tested [9].

TABLE 1.1 Definitions

We live in an environment that is essentially 55.6 Molar water. Given the law of mass action, hydrolytic reactions are inevitable. It has been estimated that we lose about 18,000 bases in every 24-hour period as a result of hydrolysis of the glycosidic bond [1]. The abasic sites so created are intrinsically mutagenic and organisms have devised a set of enzymes to survey and repair the damage. However, the most reactive mutagen in our environment is undoubtedly oxygen. Breathing, however unavoidable, is inherently dangerous! The electron transport chain by which ATP is generated results in the generation of reactive oxygen species (ROS) that produce the hydroxyl radical OH. When formed in proximity to DNA this species produces a variety of oxidation products, of which a guanine with a saturated imidazole ring (8-oxoguanine) is the most important. It is so important that enzymes utilizing different tactics are produced to overcome the effects [10]. The first, OGG1, is a glycosylase that removes the damaged base and sets off the base excision repair sequence. 8-oxoguanine in a DNA template strand has an increased probability of pairing with adenine resulting in G:C→T:A transversions. The human enzyme MYH (bacterial mutY homolog) removes As inserted opposite such damaged Gs. Base excision repair then occurs. Most of the time, the correct C will be inserted, but if not, the cycle can be repeated until an apoptic reaction ensues.

Some mutagenic changes, particularly those close to repeated sequences, seem to defy the specificity rules. Studies of the influence of the surrounding sequence on mutation permitted the deduction, supported by evidence with in vitro models, that what at first appeared to be base substitution mutagenesis might actually be the result of slippage errors in DNA replication [11]. Frameshifts (indels) are also likely to occur at regions of repeated sequence and the models account for such changes as a result of misalignments of either template or newly synthesized strand as a result of slippage during replication. Dissociation and reassociation of DNA strands occurs repeatedly and when there are repeated sequences, the reassociation may occur so that some of the repeated sequences in the newly synthesized strand loop out of the structure making addition of extra bases possible.

SPECIFICITY RULES

Most (early) considerations of mutational specificity and frequency focused on considerations of hydrogen bonding and on the base pairs suggested by Watson and Crick. Occasionally, an alternative pairing, the Hoogsteen base pair, was suggested to account for particular cases of specificity [12]. Notwithstanding the ability of such models to account for the data, it is not at all clear that hydrogen bonding is critical for successful DNA synthesis [13]. A non-polar thymine analog, (2,4 difluorotoluene) unable to form hydrogen-bonded base pairs can be incorporated selectively opposite adenine or an equivalent non-polar analog. The results obtained with such non-polar analogs depend on the polymerase used for the study. For some enzymes shape of the substrate is the determining factor; for others hydrogen bonding remains an important feature. Such studies with different polymerases indicate that the chemical nature of the substrate nucleotide is not the sole determinant of mutagenic specificity.

A favorite subject for study has been an abasic site, a point in the sequence at which a base has been removed without breaking the backbone DNA chain. In the absence of a base to serve as template, one might suppose that if replication were to proceed it would do so with bases added at random. In fact, this is not how it works [14]. Although the abasic site is a barrier to normal synthesis, some polymerases are able to insert bases opposite such sites and to extend DNA chains past the lesion. Bacterial replicative enzymes tend to insert As opposite abasic sites and there are structural studies to indicate why this should be the preferred result. Yeast will insert a C or an A. Yeast, it turns out, contains a special protein (rev1), essentially a nucleotidyl transferase, that inserts Cs at the end of growing chains and will insert a C opposite an abasic site. Mammalian cells are likely to insert Gs or As. The mutagenic specificity in different organisms turns out to be due to the specificities of their polymerases. The surprising finding, particularly in view of the emphasis on base pairing in the 1980s, is that both the frequency and specificity of the errors are determined by the particular polymerase used for replication [15, 16].

There are some rules that apply to most of the polymerases. Particular mutagens do react with DNA to produce products with unique pairing properties that predict the type of mutation to be obtained. The important guanine modifications have an increased propensity to pair with adenine, leading to G:C→T:A transversions. Guanine also reacts with many of the metabolically activated polycyclic hydrocarbon derivatives implicated as carcinogens and the resulting products when replicated also appear to give primarily G→T transversions followed by G→C transversions. In contrast, most spontaneous mutations at sites of G are transitions, explained as the conversion by deamination of methylC in G(methyl)C pairs to G:T which replicates to give a G(methyl)C→A:T transition. Survey of genes for the types of mutation observed, so-called mutational spectra, can give clues as to the origin of the mutations observed. So, for example, the prevalence of G→T mutations in lung cancers is taken as evidence for the role of tobacco related polyaromatic hydrocarbons and nitrosoamines as mutagens [17]. Unfortunately, such mutagenic spectra cannot identify the precise lesion or agent involved. In the case just cited either oxidation of guanine or the addition of a polyaromatic hydrocarbon group lead to the same mutational change.

MUTAGENIC POLYMERASES

The conclusion that organisms evolve their own mutation rate and, by providing variation, their own rate of evolution comes from the discovery of mutations in bacteria and in yeast that make the organisms mutagen stable. For example, treatment with ultraviolet light will kill such mutant organisms, but the mutation frequency will not increase among the survivors. An interpretation of the finding is that mutation requires that an altered DNA be replicated and that during replication past the lesion produced by the mutagen, an error occurs. If the DNA lesion induced by the treatment ordinarily blocks replication, then what is termed trans lesion synthesis (TLS) is required in order for mutation to occur. The finding of mutagen stable mutants indicated that proteins different from those used for replication were required for TLS. In the 1990s it was discovered that Escherichia coli, yeast, humans and other organisms code for a set of DNA polymerases distinct from the replicative polymerases and with specificities inherent in their structure. One polymerase, polymerase eta (the polymerases are assigned Greek letter names) is adept at bypassing pyrimidine dimers produced by ultraviolet radiation. Such dimers are major products of exposure to sunlight and polymerase eta synthesizes past the T^T lesion without error, inserting two As. A deficiency in this enzyme results, in humans, in the variant form of xeroderma pigmentosum, a disease characterized by the induction of numerous skin tumors. In the absence of polymerase eta, a second enzyme, probably polymerase iota takes over, but this enzyme does make errors (mutations) when synthesizing past pyrimidine dimers. There are about 16 different human DNA polymerases known (Table 1.2) and these enzymes have been classified into different groups based on structural homologies [15, 16]. Polymerase eta is a member of the Y family of polymerases which includes Pol ι (iota), Pol κ (kappa), and Rev1 (terminal deoxycytidyl transferase). Polymerase zeta is a member of the B family of polymerases that also includes the major replicative polymerases. The Y family of polymerases is characterized by a relaxed specificity. Structural studies show that the pockets, which in replicative polymerases fit tightly around the incoming nucleotides, are relaxed permitting wrong bases to be incorporated. The enzyme REV-1 pairs an incoming C with an arginine in the protein sequence rather than with the ostensible templating base [18]! On occasion, the incorporation of incorrect bases is apparently preferred by mammalian cells as a survival mechanism, since blocked replication forks lead to lethal double strand breaks in DNA.

One characteristic of replicative DNA polymerases in vivo is their processive nature when part of the replicative complex. That is, the polymerase remains attached to the DNA growing point after incorporation of a nucleotide, poised for the addition of the next base. Damaged nucleotides in the template block this progression. In order for the Y family polymerases to promote TLS, they must in some way displace the replicative polymerase from the growing point. This is accomplished via the sliding clamp, PCNA (proliferating cell nuclear antigen) which encircles the DNA at the replication fork [19]. Addition of ubiquitin to the clamp serves as a signal for combination of the Y family polymerases which can then rotate into the active site as the replicative polymerase falls off [20]. These error-prone polymerases are not processive and after addition of a few nucleotides they fall off. The details of polymerase action at growing points containing damaged nucleotides are an object of active research at the present. Its practical significance lies in the possibility of interrupting the action of such auxiliary polymerases, thereby diminishing the overall mutation rate.

MUTATION MODIFIERS

In spite of the explanations given in many elementary texts, the free energy differences between correct and incorrect base pairs are very small, at most 0.4 kcal/mol. This means that in a water solution, in which there is much competitive hydrogen bonding, a correct base pair is only about twice as likely to form as is a mismatch. The major contribution to specificity is clearly the replicative polymerases: the structural nature of the pockets into which incoming nucleotides fit and the kinetic interactions between elongation of the chain and reversal of the reaction accounting for this specificity. The free energy differences between correct and mismatched bases for a reaction catalyzed by a replicative polymerase (Drosophila melanogaster polymerase alpha) indicated a difference of 4.9 kcal/mol, equivalent to a discrimination factor of about 1 in 3000 [21]. The in vitro measured error frequency of the different polymerases (Table 1.2) varies from a low of about 1 in 100,000 for the different B family replicative polymerases to more than 3.5 per 100 for human polymerase eta [15]. However, the mutation rate in organisms is likely to be between 10−8 and 10−9 per nucleotide per generation [22, 23], a value four to five orders of magnitude lower. Organisms attain this high level of fidelity by two major processes contributing to specificity consequent to replication: proofreading and mismatch repair (MMR).

DNA synthesis can be considered as a series of steps in which the growing chain is elongated and then the newly inserted base is checked, or proofread, to determine if it meets built in pairing specifications (Fig. 1.1). Terminal bases that do not fit are removed by exonucleolytic action. The requisite 3'→5' nuclease activity is either built into the structure of B class polymerases or exists as a separate but closely associated protein(s). Y family polymerase members are devoid of exonuclease activity. Exonucleolytic proofreading results in about a hundred-fold increase in the fidelity of replication. Mutations in either the exonuclease or exonuclease domains have mutator properties, producing additional mutations in every round of replication. Organisms can fine tune their proofreading, and mutants (of bacterial viruses) have been isolated in which the rate of spontaneous mutation is lowered because of an increase in the efficiency of proofreading. Such an increase comes at a cost in energy since the ATPs required to provide the pyrophosphates required for polymerization are wasted. Even replication events with normal bases involve proofreading. About 6 to 13% of the polymerization events result in an excised (proofread) base [24]. The replication process can be depicted as a competition between proofreading and further elongation [14] since once the chain has been elongated five or six nucleotides beyond a mismatch it appears immune to proofreading. The elongation step is distinct from the initial addition opposite any particular base (Fig. 1.1). Some of the enzymes of the Y series are relatively efficient in the addition of a nucleotide opposite a non-pairing template but are unable to elongate the resulting product. It has been suggested that polymerase zeta, a B family polymerase, has as a function the elongation of mismatched bases inserted by iota and other error prone polymerases [16, 25]. The events in trans lesion synthesis can be described as follows: the replicative complex recognizes a mismatch or altered base. This triggers ubiquitination of the PCNA clamp which in turn results in dissociation of the replicative complex from the DNA and allows access of a Y family polymerase to the growing point [20]. A deoxynucleotide is added, and before the replicative complex and its associated proofreading activity can access the mismatch, polymerase zeta replaces the Y family polymerase and elongates for a few nucleotides. Polymerase zeta then falls off and is replaced by the normal replication complex. Ubiquitination of PCNA plays a critical regulatory role in the process [20]. A variation of this scenario suggests that replication proceeds asynchronously until the next initiation sequence, leaving a gap which is subsequently filled in by a process similar to that described. Superimposed on these events must be the availability of the different deoxynucleotides used for synthesis. Alterations in the pool size of the different DNA constituents can affect the selection of bases and altering relative pool sizes can be mutagenic [26]. Whatever the sequence of events, the result of any particular elongation attempt will be determined by the various competitions for access to the nucleotide at the growing point. The result of any single specific replication event cannot be predicted.

TABLE 1.2 Human DNA polymerases

FIGURE 1.1 Schematic outline of the competition between extension and proofreading in DNA synthesis.

MISMATCH REPAIR

Newly synthesized DNA in humans and other organisms is subject to yet another inspection by the set of proteins constituting the mismatch repair (MMR) system. These proteins detect mismatches in the DNA: both base pair mismatches and mismatches due to small additions or deletions. In bacteria, in which the process was studied first, the detection is carried out by a single protein acting as a homodimer, the mutS protein, which when bound to the mismatch recruits a second protein dimer, mutL. In the enteric bacteria this ATP dependent complex activates the endonuclease activity of a third protein, mutH, which makes a single stranded break in the error-containing strand. The nicked strand is unwound by a helicase encoded by the UvrD gene and the displaced strand is degraded by an exonuclease. The resulting single stranded gap is then filled by the replicative polymerase. The key to the successful operation of this scheme is making sure that the newly synthesized strand including the error is the one removed. In Escherichia coli this trick is accomplished by a special methylation mechanism. Adenines at GATC sites are methylated on both strands but the methylation of the newly inserted adenine is accomplished only after replication. Immediately after replication the newly synthesized strand is unmethylated. It is this hemi-methylated DNA which is the substrate for mismatch repair and it is the unmethylated, i.e. newly synthesized, strand which is removed [1].

Eukaryotic cells have a more complex, although clearly similar, mismatch repair mechanism [27]. Instead of a single mutS protein, eukaryotes have five, three of which (MSH2 [MutS homolog], MSH3 and MSH6) form dimers with slightly different specificities. There are four MutL homologues (MLH1, MLH2, PMS1 [post meiotic segregation protein]), and PMS2 which also function as heterodimers. The MSH2:MSH6 heterodimer recognizes base–base mismatches and small insertions or deletions, the MSH2:MSH3 complex specializes in recognition of larger insertions and deletions. As the names indicate, certain of these proteins also play an important role in meiosis. There is no MutH analog. The adenine methylation recognition mechanism appears confined to enteric bacteria. In vitro reconstructions of the eukaryotic mismatch repair system use a free 3'OH end (i.e. a nick in the DNA) to identify the newly synthesized strand and it appears likely that in vivo it is the growing point of the DNA (or an unligated Okazaki fragment) that provides the MMR signal. Eukaryotic MMR is more closely tied to replication as compared to the enteric bacteria. The MSH proteins have been shown to bind to PCNA which locates them at the site of the DNA growing fork [19].

Organisms deficient in their ability to make one of the mismatch repair proteins have increased mutation rates. The medical interest in MMR dates from the discovery that an inherited colon carcinoma syndrome can be traced to a deficiency in the MMR proteins [28, 29]. The most frequent culprits are the hMLH1 (human mutL homolog) and hMSH2 genes, followed by hPMS2 and hMSH6. Analyses of tumor tissue show that the promoters of these MMR genes are frequent targets of epigenetic inactivation by methylation [30]. The absence of a functional MMR system is often signaled by an increase in microsatellite instability. The microsatellites are regions of mono-, or di-nucleotide repeats (e.g. CACACACACA) that are polymorphic, i.e. in which the actual number of repeat units at a particular location differs among individuals. The number of repeat units at each locus is inherited and is the basis of much DNA fingerprinting. Instability is observed as a detectable increase in the number of such repeats, easily demonstrated by gel electrophoresis. Individuals deficient in MMR may have thousands of microsatellite instabilities throughout the genome but a panel of five selected loci is generally used for testing. Instability at two loci serves as a positive signal [31]. Bound MMR proteins may also serve as a signal for apoptosis. Organisms deficient in the O6-methylguanine methyltransferase protein are exquisitely sensitive to killing by methylating agents. The cells become much less sensitive when made MMR defective, possibly because of a loss of a signal from the MMR proteins combined at the O6methylG:T mismatch. The mechanism is important because MMR deficient cells with mutagenic lesions that should be signals for apoptosis, survive, reproduce and propagate mutations [32].

One of the unexpected discoveries of the 1990s was the finding that about 30 mostly neurodegenerative diseases, including Fragile X syndrome, Huntington chorea, myotonic dystrophy and Friedrich’s ataxia are due to expansions of simple repeats in the genome [33]. For example, a CAG sequence which occurs from 6 to about 35 times in normal individuals expands to up to 100 repeats in individuals affected with Huntington’s chorea. In Friedrich’s ataxia, a normal GAA sequence occurring from 7 to 22 times in an intron may expand to 200 to 1700 units. The origin of these mutations is mysterious. Model systems have been developed in which the influence of slippage during replication has been studied. It seems that there is a role for the mismatch repair system [34] but exactly what sets off these changes is a mystery.

MUTATION OUTSIDE THE REPLICATION CYCLE

Although it is simplest to consider mutation as a consequence of DNA replication, DNA turnover is not limited to the S phase. Insofar as DNA repair processes involve excision and reconstitution of the excised region, there is a chance for mutation during the repair process, particularly if an error prone polymerase is recruited to fill in the gap. Such an error will create a mismatch, but correction of such mismatches during interphase might occur using either strand as a template and resulting in fixation of such mutations. This sequence seems to account for the phenomenon of somatic hypermutation during the immune response: perhaps the only instance in which mutations are an essential part of a normal biological process [35]. The first step, which depends on transcription of the immunoglobin gene, involves activation of a single strand specific cytidine deaminase (activation-induced cytidine deaminase, AID). This enzyme converts cytidine to uracil creating a U:G mismatch. The mismatch is recognized either by the MMR pathway or by uracil glycosylase. During the ensuing repair processes, involving polymerase eta and the rev1 cytidine terminal transferase protein, mutations are generated. It has not been determined whether somatic hypermutation occurs during the S phase, but there is no reason why it should be limited to this period. Somatic hypermutation is limited to portions of the immunoglobulin gene, in particular cells at only limited stages of their differentiation. It is not clear what gives the process its specificity. Its interest lies not only in the importance of the immune phenomenon itself, but in the possibility that under particular circumstances something similar might occur, say, in tumorigenesis. This possibility is highlighted by the discovery of somatic hypermutations in the BCL6 and CD95 genes [36].

SPONTANEOUS MUTATION AND TUMORIGENESIS

Mutations are clearly important in tumorigenesis and it is clear that natural selection, in the sense of selection of the clones fittest for reproduction in the host’s environment, plays a major role. A first question is whether tumors are hypermutable or whether the accumulation of mutations in tumors can be accounted for solely by selection. Hypermutability is an old suggestion put in modern terms in a series of papers by L. Loeb [37]. A parallel series of papers, summarized by W. Bodmer [38], argues that the spontaneous human mutation rate is sufficient to account for the mutations observed. Mismatch repair deficiency and its associated increased mutation rate is certainly associated with a sub-class of colon carcinomas. However, such deficiency accounts for only about 15% of colon carcinomas and the tumors have properties (e.g. a relatively stable karyotype) that distinguish them from other solid tumors. It is also (now) clear that tumor DNA is separated from normal tissue by many mutations. When analyzed by methods that sequence DNA derived from tumors and neighboring surrounding tissue, the tumors show many more mutations than the surrounding normal tissue. In one study in which 274 megabases of tumor DNA was screened for point mutations in exons of protein kinase genes the investigators detected 1007 mutations: 921 single base changes, 54 nonsense mutations and 219 silent mutations [39]. Assuming an average nucleotide size of 1500 bases per gene this finding implies a mutation frequency of about 0.5% (0.005), or about one new mutation for every 200 genes. One of the observations in this, and similar studies, is that no two tumors have precisely the same pattern of mutation – there is little overlap in the mutations observed although some mutations, e.g. in TP53, do tend to recur. Since we expect that each of us inherits about three new mutations in 50,000 genes (two sets of 25,000) this seems to show that the tumors must be highly mutable. The adherents of the selection is sufficient school argue that this need not be so. Direct measurements of the somatic mutation rate in human cells are rare but a recent measurement gives a value of 1.06 × 10−6 mutations per cell division for the PIG-A gene [22]. There are 249 amino acids in this gene coded for by 747 nucleotides, which gives a mutation rate of 1.4 × 10−9per nucleotide per generation. (For comparison, an estimate of the average mutation rate for 20 genes involved in Mendelian disease is 1.8 × 10−8per nucleotide per generation [23].) The adult human has approximately 5 × 10¹³ to 10¹⁴ cells (this number is at best a guess by pathologists). Since the number of divisions required to produce a population of N cells is N–1; this means that there are 1.41 × 10−9× 2 × 3 × 10⁹ nucleotides (the number of nucleotides in a diploid genome) × 10¹⁴ divisions or 8.5 × 10¹⁴ new mutations produced during development. Only about 5% or less of these will be in the genes, but even so this means 4 × 10¹³ mutations distributed throughout 25,000 genes. This is about equal to the number of cells, so that most cells will have at least one new mutation in their genes. Based on the Poisson distribution, many cells will have no mutations but others will have two, three, four or more. If, as seems to be indicated by recent whole genome association studies, the nongenic region of the genome is not as devoid of information as we suppose, the number of effective mutations will be even higher. Even in the absence of a mutator activity the selectionists argue that there will be sufficient genetic variation to drive tumor progression. The experimental observations are that mutations in genes affecting DNA repair, and therefore mutation rate, are, as a class, frequently found mutated in tumors. The debate has practical consequences. The common observation that cancers arise in the elderly is often interpreted as indicating that there needs to be an accumulation of individual (mutational) changes before an overt tumor develops. Many estimates suggest that five or six changes need to accumulate. Even a slight decrease in mutation rate would raise the age of incidence to greater than the average life span. If cancers have a mutator phenotype, it is argued that it may be easier to find drugs to counteract this phenotype and thus increase the age of first incidence to greater than the average life span!

The frequent occurrence of new mutations in all somatic cells makes it important to distinguish between mutations that are important in the etiology of a condition being investigated and those present by accident. For example, consider two cells spatially separated in a tissue. Both will have accumulated a different complement of neutral mutations during their separate development. Collection of the tissue and sequencing by standard methods will not reveal any of these new mutations because sequences that represent less than about 10% of the genome are dismissed as noise in standard sequencing. Suppose now that one of these cells develops a new mutation that leads to proliferation. Cells of this new clone will all contain the new mutation but in addition, they will contain all of the mutations that have occurred in that particular cell during its development. Sequencing of a tumor derived from that clone will reveal not only the mutation responsible for the selection (the driver), but also all of the others (the passengers or hitchhikers) that have accumulated previous to the transformation.

The evidence for this theoretical formulation comes from an examination of the types of mutation observed. If mutation is random and if there is no selection, then about 25 to 30% of all mutations, depending on the codon usage of the gene being studied, should be silent or synonymous, resulting in no amino acid change. Demonstrating that the proportion of silent mutations is what would be expected for random mutation is a sign that such mutations have not been selected. In the data set referred to above [17], 219 out of 1007 mutations or about 20% were silent, close to, but slightly below the expected value for pure random mutation. The investigators conclude that some of the mutations are indeed drivers and functionally significant, but it takes both sophisticated statistics and biological insight to determine which [40].

THE ROLE OF DNA STRUCTURE

Although the Human Genome Project was officially completed in 2003, a 2008 paper [41] concludes that what was sequenced may actually represent only a minor allele! Continued advances in techniques for the analysis of DNA coupled with analysis of the genomes of a wider range of individuals confirms the findings that not only are there large numbers of single nucleotide polymorphisms (SNPs) but that structural variations including insertions, deletions and inversions of the DNA sequence may involve more base pairs than are found in the SNPs. Changes in copy number have been associated with both Mendelian and complex human traits [42]. The repeated elements in our genome promote instability by at least two major mechanisms, transposition and unequal crossing over, and it may be that these mechanisms are at least as likely to be responsible for mutagenic change (as defined phenotypically) as are any of the point mutational mechanisms. One of the startling revelations of the year is the observation that a group of identical twins display mosaicism in the copy number of their genes [43]. Copy number variation was observed in twins both concordant and discordant for particular traits. A deletion associated with MLL was found in one twin previously diagnosed with the disease. The genetic identity of monozygotic twins has been a given in genetic research for years. The report suggests that recombination events can occur during mitotic development of individuals and that such variations may have medical relevance.

Aberrant recombination of repeated elements in the genome can result in duplication, deletion or inversion of large regions of the genome (Fig. 1.2). Insertional mutagenesis due to the introduction of a transposed element into a gene can result in disruption of the gene product. The most prevalent transposable elements are the SINES (short interspersed nuclear elements) and the LINES (long interspersed elements). Full sized LINE elements are about 6.1 kb in size and contain coding sequences for gene products essential for their transposition within the genome. Humans may have 200,000 to 500,000 of these elements in their genome, of which the most frequent is the L1 (LINE-1). Only a few such elements retain their capacity to catalyze movement [44]. The others have suffered a variety of sequence changes in the course of history resulting in their inactivation. The SINE elements, of which the most prevalent are the Alu sequences of about 300 bases, may be present in approximately 1,000,000 copies throughout the human genome. These elements cannot catalyze their own movement but can apparently transpose by utilizing some of the enzymatic machinery produced by LINE elements. Transposons are recognizable by the target site duplications (TSD) that are found at either end of the insertion sites. Such duplications of about 2 to 10 bases occur as a result of the insertion mechanism, which at one point involves making a staggered nick in the double helix somewhat similar to the staggered cuts made by restriction enzymes, although there seems to be no specific sequence recognition for the insertion sites.

FIGURE 1.2 Cartoon showing how repeated elements in the genome can lead to chromosomal aberrations. (a) Duplication due to unequal crossing over. During meiosis recombination of homologous chromosomes containing displaced repeated elements can result in one recombinant obtaining an extra copy of the repeated element and the second chromosome losing a copy. (b) Deletion due to two transposable (repeated) elements inserted in the same chromosome in the same orientation. Pairing of the homologous regions of the duplicated transposable elements followed by recombination results in a daughter chromosome in which the genetic material between the chromosomes has been deleted. The loop including a copy of the transposon and of the genes intervening is presumed to be lost in replication. Recombination between non-homologous chromosomes containing such repeated sequences can lead to translocations. (c) Inversion due to two transposable elements inserted into the same chromosome in opposite orientations. As shown in the diagram, recombination will result in inversion of the chromosome order.

Well over 30 pathological conditions have been associated with new insertions of Alu sequences and about 11 with insertion of L1 elements [45]. Eight of the L1 elements are reported on the X chromosome and 9 of the Alu sequence are X–linked. This unexplained excess on the X does not seem to be accounted for by the bias derived from the necessary dominance of all mutations in males. It has been estimated that there is a new Alu transposition about once in every 20 births and that the ratio of disease producing transpositions as compared to single nucleotide changes is 1 in 2000.

The actual process of retrotransposition is estimated to account for only about 15% of the events leading to structural variation, possibly because most of the elements are no longer active [44]. Most of the structural aberrations occur as a result of recombination events between repeated sequences [41]. Unequal crossing over (Fig. 1.2) can produce both addition and deletion of particular sequences. The production of deletions and inversions depends on whether the repeated sequences occur in the same or inverted orientation (Fig. 1.2b,c). Recombination between repeats in different chromosomes can result in translocations. Our current understanding of the structure of the genome therefore indicates it to include numerous sources of instability. Most such changes will be eliminated when they occur at meiosis because they result in abnormal development. Changes occurring in somatic cells may result in pathological change and we are only beginning to understand such instability. The numerous repeats and retroviral like elements in our genome act as potential