16.3.1 The convenient nomenclature of A and a alleles hides a vast diversity of DNA sequence changes

Over 750 different cystic fibrosis mutant alleles have been described, and a similar number of different mutations in the β-globin gene. There is no reason why these should all fit into a few tidy categories. In principle however, mutation of a gene might cause a phenotypic change in either of two ways:

• the product may have reduced or no function (loss of function mutation - an amorph or hypomorph in the terminology of Box 16.2);
• the product may do something positively abnormal (gain of function mutation - a hypermorph or neomorph).

Loss of function mutations most often produce recessive phenotypes. For most gene products the precise quantity is not crucial, and we can get by on half the normal amount. Thus most inborn errors of metabolism are recessive. For some gene products, however, 50% of the normal level is not sufficient for normal function, and haploinsufficiency produces an abnormal phenotype, which is therefore inherited in a dominant manner (see Section 16.4.3). Sometimes also a nonfunctional mutant polypeptide interferes with the function of the normal allele in a heterozygous person, giving a dominant negative effect (an antimorph in the terminology of Box 16.2 - see Section 16.4.4).

Gain of function mutations usually cause dominant phenotypes, because the presence of a normal allele does not prevent the mutant allele from behaving abnormally. Often this involves a control or signaling system behaving inappropriately - signaling when it should not, or failing to switch a process off when it should. Sometimes the gain of function involves the product doing something novel - a protein containing an expanded polyglutamine repeat forming abnormal aggregates, for example.

Inevitably some mutations cannot easily be classified as either loss or gain of function. Has a permanently open ion channel lost the function of closing or gained the function of inappropriate opening? A dominant negative mutant allele has lost its function but also does something positively abnormal. Nevertheless, the distinction between loss of function and gain of function is a useful first tool for thinking about molecular pathology.

16.3.2 Loss of function is likely when point mutations in a gene produce the same pathological change as deletions

Purely genetic evidence, without biochemical studies, can often suggest whether a phenotype is caused by loss or gain of function. When a clinical phenotype results from loss of function of a gene, we would expect any change that inactivates the gene product to produce the same clinical result. We should be able to find point mutations which have the same effect as mutations that delete or disrupt the gene. Waardenburg syndrome Type 1 (MIM 193500) provides an example. As Figure 16.1 shows, causative mutations in the PAX3 gene include amino acid substitutions, frameshifts, splicing mutations, and in some patients complete deletion of the PAX3 sequence. Since all these events produce the same clinical result, its cause must be loss of function of PAX3. Similarly, among diseases caused by unstable trinucleotide repeats (see Box 16.7), Fragile-X and Friedreich ataxia are occasionally caused by other types of mutation in their respective genes, pointing to loss of function, whereas Huntington disease is never seen with any other type of mutation, suggesting a gain of function.

Figure 16.1

Loss of function mutations in the PAX3 gene. The 10 exons of the gene are shown as boxes, with the connecting introns not to scale. The shaded areas show the sequences encoding the two DNA-binding domains of the PAX3 protein. Note that mutations that (more...)

Box 16.7

Unstable expanding repeats - a novel cause of disease. Unstable expanding trinucleotide repeats were an entirely novel and unprecedented disease mechanism when first discovered in 1991, and they raise two major questions: What is the mechanism of the (more...)

16.3.3 Gain of function is likely when only a specific mutation in a gene produces a given pathology

Gain of function is likely to require a much more specific change than loss of function. The mutational spectrum in gain-of-function conditions should be correspondingly more restricted, and the same condition should not be produced by deletion or disruption of the gene. Likely examples include Huntington disease (see Box 16.7), and achondroplasia (MIM 100800: short-limbed dwarfism). Virtually all achondroplastics have one of two mutations in the fibroblast growth factor receptor gene FGFR3, both of which cause the same amino acid change, G380R (Bellus et al., 1995). Other mutations in the same gene produce other syndromes (Section 16.7.3). For unknown reasons, the mutation rate for G380R is extraordinarily high, so that achondroplasia is one of the commoner genetic abnormalities, despite requiring a very specific DNA sequence change.

Mutational homogeneity is a good first indicator of a gain of function, but there are other reasons why a single mutation may account for all or most cases of a disease:

• diseases where what one observes is directly related to the gene product itself, rather than a more remote consequence of the genetic change, may be defined in terms of a particular variant product, as in sickle cell disease (see Box 16.5);
• a molecular mechanism may make a certain sequence change in a gene much more likely than any other change - e.g. the CGG expansion in Fragile-X syndrome (see Box 16.8);
• there may be a founder effect - for example, certain disease mutations are common among Ashkenazi Jews, presumably reflecting mutations present in a fairly small number of founders of the present Ashkenazi population (Motulsky, 1995);
• selection favoring heterozygotes (Section 3.3.2) enhances founder effects and often results in one or a few specific mutations being common in a population.

Box 16.5

Hemoglobinopathies. Hemoglobinopathies occupy a special place in clinical genetics for many reasons. They are by far the most common serious mendelian diseases on a worldwide scale. Globins illuminate important aspects of evolution of the genome (see (more...)

Box 16.8

Laboratory diagnosis of fragile X. Cytogenetic testing - the fragile site is seen only when cells are grown under conditions of folate or thymidine depletion. For unknown reasons, it is never present in more than 50% of cells, and the frequency is often (more...)

16.3.4 Deciding whether a DNA sequence change in a gene is pathogenic can be difficult

Not every sequence variant seen in an affected person is necessarily pathogenic. If the genome-wide average heterozygosity of 0.0032 is applied to coding sequences, then screening a panel of 100 patients for mutations in a 3-kb coding sequence would reveal about 500 sequence changes. Even allowing for the much higher conservation of coding sequences, screening on such a scale will almost certainly reveal some rare nonpathogenic sequence variants, as well as pathogenic changes. If each variant occurs in one person in 10 000 in the population, it will not show up in any panel of normal controls of realistic size. How does one decide whether a sequence variant is pathogenic?

If the pathogenic mechanism is gain of function, then as explained above (Section 16.3.3), the mutation is likely to be very specific. Any sequence change different from the standard mutation is probably not pathogenic, at least for the disease in question. Loss of function mutations are usually much more heterogeneous. Only a functional test, either in vitro (Chapter 20) or in vivo (Chapter 21), can definitively show whether a DNA sequence change in a gene affects the function, but useful clues can be obtained by considering the nature of the sequence change (Box 16.4).

Box 16.4

Guidelines for deciding whether a DNA sequence change is pathogenic. Deletions of the whole gene, nonsense mutations and frameshifts are almost certain to destroy the gene function. Mutations that change the conserved GT…AG nucleotides flanking (more...)

16.4.1 Many different changes to a gene can cause loss of function

Not surprisingly, there are many ways of reducing or abolishing the function of a gene product (Table 16.1 and Figure 16.2). Some of these have been discussed in Section 9.4. The hemoglobinopathies (Box 16.5) exemplify many of these mechanisms especially well. In fact, globin mutations can be found to illustrate virtually every process described in this book. Readers with a particular interest in these diseases are recommended to consult one of the excellent reviews of this topic (e.g. Weatherall et al., 1995; see further reading).

Table 16.1

Eleven ways to reduce or abolish the function of a gene product (see Table 9.5 for a classification of mutations by their nature and location in the gene).

Figure 16.2

Deletions of α-globin genes in α-thalassemia. Normal copies of chromosome 16 carry two active α-globin genes and an inactive pseudogene arranged in tandem. Repeat blocks (labeled X and Z) may misalign, allowing unequal crossover. (more...)

When considering the likely result of a mutation on the gene product, some points to bear in mind are as follows:

• Small deletions and insertions have a much more drastic effect on the gene product if they introduce a frameshift (that is, if they add or remove a number of nucleotides that is not an exact multiple of 3). Deletions in the dystrophin (DMD) gene provide striking examples (Figure 16.3). Almost regardless of the size of the deletion, frameshifting deletions produce the lethal Duchenne muscular dystrophy, in which no dystrophin is produced, whereas nonframeshifting mutations cause the milder Becker form, in which dystrophin is present but abnormal.
• Nonsense mutations often trigger mRNA instability (see Section 9.4.6 and Hentze and Kulozik, 1999) rather than cause production of a truncated protein.
• Base substitutions in coding sequences may be pathogenic because of an effect on splicing or because they destroy an embedded signal (a nuclear localization signal, for example), rather than because of their effect on the amino acids encoded. Activation of a cryptic splice site is particularly hard to predict - see Section 9.4.5 and Berget (1995).

Figure 16.3

Deletions in the central part of the dystrophin gene associated with Becker and Duchenne muscular dystrophy. Numbered boxes represent exons 43–55. Deletions that generate frameshifts cause the lethal DMD, while frame-neutral deletions cause the (more...)

16.4.2 Epigenetic modification can abolish gene function even without a DNA sequence change

Heritable changes that do not depend on changes in a DNA sequence are called epigenetic (see Section 8.1). They may affect expression of a gene or the properties of its product. A set of mini-reviews in the 1 May 1998 issue of Cell (Vol 93, pp 301–337) discuss many of the diverse facets of epigenetics. Important epigenetic mechanisms include:

• DNA methylation. Silencing of an intact gene by methylation of adjacent control sequences is a normal part of development, differentiation and X-inactivation - but methylation can occasionally cause pathogenic loss of function. In many tumors, for example, function of the CDKN2A tumor suppressor gene is abrogated by methylation of the promoter rather than by mutating its DNA sequence (Chapter 18). In Fragile-X syndrome, the FMR1 gene is silenced by methylation, although in this case the methylation is triggered by a local DNA sequence change, expansion of a trinucleotide repeat (Box 16.8).
• Changes in chromatin configuration as a result of chromosomal rearrangements can also up-regulate or silence expression of an intact gene - for example the MYC oncogene is over-expressed when a translocation places it in the transcriptionally active immunoglobulin region (Figure 18.7).
• Imprinting (Sections 3.2.4, 8 and 8.5.4). Imprinted genes are a particularly intriguing example of epigenetic modification. Their expression is controlled by patterns of methylation that differ according to the parental origin of the gene. When either the imprinting mechanism malfunctions or the parental origin is not as expected, pathogenic loss of function or inappropriate expression can occur in intact genes. Several human diseases involve imprinted genes, the best known being Prader-Willi and Angelman syndromes (Box 16.6).
• Changes in protein conformation. Sometimes it appears that a conformational change can propagate through a population of protein molecules, converting them from a stable native conformation into a new form with different properties. The process may be analogous to crystallization. The behavior of prion proteins (Prusiner et al., 1998) is the most striking example, but natural processes of protein aggregation into subcellular structures might also be seen in this light.

Box 16.6

Molecular pathology of Prader-Willi and Angelman syndromes. Prader-Willi syndrome (PWS) and Angelman syndrome (AS) are both caused by problems with differentially imprinted genes at 15q11-q13. They exemplify the complicated molecular pathology associated (more...)

16.4.3 Haploinsufficiency describes the case where a 50% reduction in the level of gene function causes an abnormal phenotype

Loss of function mutations tend to be recessive because heterozygotes often function perfectly normally. Sometimes this is because feedback loops compensate for the reduced dosage by increasing transcription or the activity of the gene product, but in many cases the cell and organism are able to function normally with only a 50% level of gene action. Only relatively few genes show haploinsufficiency; Table 16.2 lists some examples.

Table 16.2

Phenotypes probably caused by haploinsufficiency (see text for details).

One might reasonably ask why there should be dosage sensitivity for any gene product. Why has natural selection not managed things better? If a gene is expressed so that two copies make a barely sufficient amount of product, selection for variants with higher levels of expression should lead to the evolution of a more robust organism, with no obvious price to be paid. The answer is that in most cases this has indeed happened, which is why relatively few genes are dosage-sensitive. Sometimes perhaps, if the gene product is needed in large quantities, the total synthetic capacity of the cell, even at maximum transcription levels, may be insufficient if only one copy of the gene is present. An example may be elastin. In people heterozygous for a deletion or loss of function mutation of elastin, for the most part the elastic tissues (skin, lung, blood vessels) work normally, but often the aorta, a highly elastic tissue, shows some degree of narrowing just above the heart (supravalvular aortic stenosis), which may require surgery (see Section 16.8.1).

Certain gene functions, however, are inherently dosage-sensitive (Fisher and Scambler, 1994). These include:

• gene products that are part of a quantitative signaling system whose function depends on partial or variable occupancy of a receptor, DNA-binding site, etc.;
• gene products that compete with each other to determine a developmental or metabolic switch;
• gene products that co-operate with each other in interactions with fixed stoichiometry (such as the α and β globins or many structural proteins).

In each case the gene product is titrated against something else in the cell. What matters is not the correct absolute level of product, but the correct relative levels of interacting products. Genes whose products act essentially alone, such as many soluble enzymes of metabolism, seldom show dosage effects. Pathological effects caused by gene dosage depend on interactions, and so are subject to modification by changes elsewhere in the genome. Thus these dominant conditions often show highly variable expression (see Section 16.6.3).

16.4.4 Mutations in proteins that work as dimers or multimers sometimes produce dominant negative effects

A dominant negative effect occurs when a mutant polypeptide not only loses its own function, but also interferes with the product of the normal allele in a heterozygote. Dominant negative mutations cause more severe effects than deletion or nonsense mutations in the same gene. Some sort of physical association of the normal and mutant products is required for a dominant negative effect. Structural proteins that contribute to multimeric structures are vulnerable to dominant negative effects. Collagens provide a classic example.

Fibrillar collagens, the major structural proteins of connective tissue, are built of triple helices of polypeptide chains, sometimes homotrimers, sometimes heterotrimers, that are assembled into close-packed crosslinked arrays to form rigid fibrils. In newly synthesized polypeptide chains (preprocollagen), N- and C-terminal propeptides flank a regular repeating sequence (Gly- X-Y)_n, where either X or Y is usually proline, and the other is any amino acid. Three preprocollagen chains associate and wind into a triple helix under control of the C-terminal propeptide. After formation of the triple helix, the N- and C-terminal propeptides are cleaved off (Figure 16.4). A polypeptide that complexes with normal chains, but then wrecks the triple helix can reduce the yield of functional collagen to well below 50%. The molecular pathology of collagen mutations is very rich, and is discussed below (see Section 16.6.1, Table 16.6B).

Figure 16.4

Dominant negative effects of collagen gene mutations. Collagen fibrils are built of arrays of triple-helical procollagen units. The type I procollagen comprises two chains encoded by the COL1A1 gene and one encoded by COL1A2. Null mutations in either (more...)

Table 16.6B

Molecular classification of the connective tissue diseases.

Nonstructural proteins that dimerize or oligomerize also show dominant negative effects. For example, transcription factors of the b-HLH-Zip family (see Figure 8.8) bind DNA as dimers. Mutants that cannot dimerize often cause recessive phenotypes, but mutants that are able to sequester functioning molecules into inactive dimers give dominant phenotypes (Hemesath et al., 1994). The ion channels in cell membranes provide another example of multimeric structures that are subject to dominant negative effects (Section 16.6.1).

16.5.1 Acquisition of a novel function is rare in inherited disease but common in cancer

Making random changes in a gene is quite likely to stop it working, but very unlikely to give it a novel function. The only mechanism that commonly generates novel functional genes is when a chromosomal rearrangement joins functional modules of two different genes (Table 16.3). Such exon-shuffling was no doubt important in evolution; for molecular pathology, it is most often noticed when it leads to cancer. Many acquired tumor-specific chromosomal rearrangements produce chimeric genes with novel activities that lead to uncontrolled cell proliferation (see Table 18.3). A rare case of an inherited point mutation conferring a novel function on a protein is the Pittsburg allele at the PI locus (MIM 107400; Figure 16.5).

Table 16.3

Mechanisms of gain of function mutations.

Figure 16.5

An inherited mutation causing a protein to gain a novel function. The α₁-antitrypsin molecule inhibits elastase. Methionine 358 in the reactive center acts as a ‘bait’ for elastase; when the peptide link between Met358 and Ser359 (more...)

16.5.2 Over-expression may be pathogenic

Gross over-expression of certain genes is common in cancer cells. The mechanisms by which somatic genetic changes produce over-expression include massive reduplication of the gene or transposition of a gene normally expressed at low level into a highly active chromatin environment. These are discussed more fully in Chapter 18.

Inherited diseases are not often caused by constitutional over-expression of a single gene. Duplication of the DSS gene on Xp21.3 causes male to female sex reversal, probably as a direct result of the doubled dosage (Bardoni et al., 1994). For the PMP22 peripheral myelin protein gene, an increase in gene dosage from two to three copies is enough to produce Charcot-Marie-Tooth disease (see Figure 16.7). Such modest increases in gene expression are probably seldom pathogenic, although a similar degree of dosage sensitivity of unidentified genes must explain many features of chromosomal trisomies (Section 16.8.2). Overactivity of an abnormal gene product, with normal transcription and translation of the gene, can produce similar effects.

Figure 16.7

Gene dosage effects with the PMP22 gene. Most patients with Charcot-Marie-Tooth disease are heterozygous for a 1.5-Mb duplication at 17p11.2, including the gene for peripheral myelin protein, PMP22 (black square). A patient homozygous for the duplication (more...)

16.5.3 Qualitative changes in a gene product can cause gain of function

Although gains of truly novel functions are very rare in inherited disease, activating mutations that modify cellular signaling responses quite often produce dominant phenotypes. The G-protein coupled hormone receptors provide good examples. Many hormones exert their effects on target cells by binding to the extracellular domains of transmembrane receptors. Binding of ligand causes the cytoplasmic tail of the receptor to catalyse conversion of an inactive (GDP-bound) G-protein into an active (GTP-bound) form, and this relays the signal further by stimulating adenylyl cyclase. Some mutations cause receptors to activate adenylyl cyclase even in the absence of ligand.

• Familial male precocious puberty (MIM 176410: onset of puberty by age 4 in affected boys) is found with a constitutively active luteinizing hormone receptor.
• Autosomal dominant thyroid hyperplasia can be caused by an activating mutation in the thyroid stimulating hormone receptor (see MIM 275200).
• Jansen's metaphyseal chondrodystrophy (MIM 156400: a disorder of bone growth) can be caused by a constitutionally active parathyroid hormone receptor.
• A constitutionally active G_sα protein (part of the G-protein) causes McCune-Albright syndrome or polyostotic fibrous dysplasia (PFD, MIM 174800). PFD is known only as a somatic condition in mosaics - probably constitutional mutations would be lethal. Depending on the tissues carrying the mutant cell line, the result is polyostotic fibrous dysplasia, café- au-lait spots, sexual precocity and other hyperfunctional endocrinopathies. Loss of function mutations of the same gene often underlie a different disease, Albright's hereditary osteodystrophy (see Table 16.5).

Table 16.5

Examples of genes responsible for more than one disease.

16.6.1 For loss of function mutations the phenotypic effect depends on the residual level of gene product

The DNA sequence changes described in Table 16.1 can cause varying degrees of loss of function. Many amino acid substitutions have little or no effect, while some mutations will totally abolish the function. A mutation may be present in one or both copies of a gene. When both homologues are affected, they may be affected unequally - people with autosomal recessive conditions are often compound heterozygotes, with two different mutations. If both mutations cause loss of function, but to differing degrees, the least severe allele will dictate the level of residual function.

Figure 16.6 shows four possible relations between the level of residual gene function and the clinical phenotype.

Figure 16.6

Four possible relationships between loss of function and clinical phenotype. See text for discussion.

1. A simple recessive condition. People heterozygous for a mutation that totally abolishes gene function are phenotypically normal, provided their remaining allele is not significantly defective.
2. A dominant condition caused by haploinsufficiency. In reality, this simple situation is rare. If a 50% reduction in gene product causes symptoms, a more severe reduction will probably have more severe effects.
3. A recessive condition with graded severity. Among many examples are:

   • mutants in the X-linked hypoxanthine guanine phosphoribosyl transferase (HPRT) gene. The extent of residual enzyme activity in mutants correlates well with the clinical phenotype of affected males (Table 16.4).
   • Reduced copy numbers of α-globin genes produce successively more severe effects. As shown in Figure 16.2, most people have four copies of the α-globin gene (αα/αα). People with three copies (αα/α-) are healthy; those with two (whether the phase is α-/α- or αα/--) suffer mild α-thalassemia; those with only one gene (α-/--) have severe disease, while lack of all four α genes (--/--) causes lethal hydrops fetalis.
4. Closely related to the situation in (c), decreasing residual function of a gene may extend the phenotype, perhaps causing a condition with a different clinical label. Depending on the position of the thresholds, several different situations can arise:

   • Several related recessive conditions may be caused by successive reductions in gene function at a single locus. For example, extracellular matrix is rich in sulfated proteoglycans like heparan sulfate and chondroitin sulfate, and defects in sulfate transport interfere with skeletal development. Loss of function mutations in the DTDST sulfate transporter cause three related autosomal recessive skeletal dysplasias, diastrophic dysplasia (MIM 226600), atelosteogenesis II (MIM 256050) and achrondrogenesis Type 1B (MIM 600972), depending on the extent of loss (Hastbacka et al., 1996).
   • Moderate reduction in function, caused by either haploinsufficiency or dominant negative effects, may produce a dominant condition, while very severe reduction in homozygotes may produce a recessive condition. The dominantly inherited Romano-Ward syndrome (MIM 192500: cardiac arrhythmia) is caused by dominant negative mutations in the KVLQT1 K⁺ channel; in transfected Xenopus oocytes the heterozygote ion channels have about 20% of normal activity. A simple loss of function mutation in the same gene has no clinical effect in heterozygotes (with 50% function) but causes the recessive Jervell and Lange-Nielsen syndrome (MIM 220400: heart problems and hearing loss) in homozygotes. In the Xenopus assay, ion channels from JLN patients totally lack function (Wollnik et al., 1997).
   • Mutations in the same gene can produce two or more dominant conditions, the milder one by simple haploinsufficiency, and more severe forms through dominant negative effects. This happens in the COL1A1 or COL1A2 genes that encode type I collagen (Figure 16.4). Mutations in these genes usually produce osteogenesis imperfecta (OI; brittle bone disease). Frameshifts and nonsense mutations produce type 1 OI, the mildest form, while amino acid substitutions in the Gly-X-Y repeated units are seen in the more severe types II, III and IV OI. The genotype- phenotype relationship is quite subtle. Substitution of glycine by a bulkier amino acid in the Gly-X-Y unit has a dominant negative effect by disrupting the close packing of the collagen triple helix. The helix is assembled starting at the C-terminal end, and substitution of glycines close to that end has a more severe effect than substitutions nearer the N-terminal end. Skipping of exon 6 (of COL1A1 or COL1A2) has a quite different effect. The site for cleavage of the N-terminal propeptide is lost and abnormal collagen is produced that causes Ehlers Danlos syndrome Type VII (MIM 130060; laxity of skin and joints). A different function has been lost, and a different phenotype results.

Table 16.4

Consequences of decreasing function of hypoxanthine guanine phosphoribosyl transferase.

16.6.2 Loss of function and gain of function mutations in the same gene will cause different diseases

We have seen that loss of function mutations in the PAX3 gene cause the developmental abnormality Type 1 Waardenburg syndrome (Figure 16.1). A totally different phenotype is seen when an acquired chromosomal translocation creates a novel chimeric gene by fusing PAX3 to another transcription factor gene, FKHR in a somatic cell. The gain of function of this hybrid transcription factor causes the development of the childhood tumor, alveolar rhabdomyosarcoma (see Table 18.3).

A striking example concerns the RET gene. RET encodes a receptor that straddles the cell membrane. When its ligand (GDNF) binds to the extracellular domain it induces dimerization of the receptors, which then transmit the signal into the cell via tyrosine kinase modules in their cytoplasmic domain. A variety of loss of function mutations - frameshifts, nonsense mutations and amino acid substitutions that interfere with the post-translational maturation of the RET protein - are one cause of Hirschsprung's disease (MIM 142623; intractable constipation caused by absence of enteric ganglia in the bowel - see Section 19.5.2). Certain very specific missense mutations in the RET gene are seen in a totally different set of diseases, familial medullary thyroid carcinoma and the related but more extensive multiple endocrine neoplasia type 2. These are gain of function mutations, producing receptor that reacts excessively to ligand or is constitutively active and dimerizes even in the absence of ligand. Curiously, some people with missense mutations affecting cysteines 618 or 620 suffer from both thyroid cancer and Hirschsprung disease - simultaneous loss and gain of function. This reminds us that loss of function and gain of function are not always simple scalar quantities; mutations may have different effects in the different cell types in which a gene is expressed.

Table 16.5 lists a number of cases where mutations in a single gene can result in more than one disease. Usually the gain of function mutant produces a qualitatively abnormal protein. Occasionally a simple dosage effect can be pathogenic. The peripheral myelin protein gene PMP22 is an example. Unequal crossovers between repeat sequences on chromosome 17p11 create duplications or deletions of a 1.5 Mb region that contains the PMP22 gene (Figure 16.7). Heterozygous carriers of the deletion or duplication have one copy or three copies, respectively, of this gene. People who have only a single copy suffer from hereditary neuropathy with pressure palsies or tomaculous neuropathy (MIM 162500), while as mentioned above, people with three copies have a clinically different neuropathy, Charcot-Marie-Tooth disease 1A (CMT1A; MIM 118220).

16.6.3 Variability within families is evidence of modifier genes or chance effects

Many mendelian conditions are clinically variable even between affected members of the same family who carry exactly the same mutation. Intrafamilial variability must be caused by some combination of the effects of other unlinked genes (modifier genes) and environmental effects (including chance events). Phenotypes depending on haploinsufficiency are especially sensitive to the effects of modifiers, as discussed above (Section 16.4.3). Waardenburg syndrome is a typical example: Figure 16.1 shows the evidence that this dominant condition is caused by haploinsufficiency, and Figure 3.5C shows typical intrafamilial variation.

An example of how a modifier might work comes from an interesting family with apparent digenic inheritance of ocular albinism (Morell et al., 1997). Tyrosinase is a key enzyme of melanocytes; deficiency leads to oculocutaneous albinism (MIM 203100). A common variant of the tyrosinase gene, R402Q, encodes an enzyme with reduced activity, but the residual activity is sufficiently high for even homozygotes to be phenotypically normal. However, in the family reported by Morell et al., people carrying one or two copies of R402Q showed ocular albinism (a mild form of oculocutaneous albinism) when they also carried a mutation in MITF, a gene involved in differentiation of melanocytes. Mutations in MITF alone do not cause ocular albinism.

Intrafamilial variability is a big problem in genetic counseling because families contemplating childbearing want to know how severely affected a child would be. Thus there is a clinical as well as a scientific motivation to identify modifier genes. The work of Easton et al. (1993) on neurofibromatosis type 1 shows how statistical analysis of clinical phenotypes within large families can provide evidence for modifier genes. These might then be sought by a whole genome search, comparing clinically concordant and discordant relatives - but very large samples would probably be needed for success. The role of pure chance should also not be ignored, especially in conditions with a patchy phenotype. Examples include patchy depigmentation in Waardenburg syndrome, and the variable numbers of neurofibromata or polyps in neurofibromatosis type 1 (MIM 162200) and adenomatous polyposis coli (MIM 175100), respectively.

Candidate modifier genes may be suggested by knowledge of the biochemical interactions of the primary gene product, or from studies in mice, where the necessary genetic analysis is feasible. An example of the latter is the identification of phospholipase A as a major modifier of the number of tumors seen in a mouse model of polyposis coli. The human counterpart (PLA2G2A, MIM 172411) has been investigated as a modifier of the severity of adenomatous polyposis coli. Results to date are promising but not conclusive (lod score of 2). Whatever the final verdict, this work provides a model of how clinically important modifiers might be identified.

16.6.4 In mitochondrial diseases heteroplasmy and instability complicate the relationship between genotype and phenotype

Mitochondrial diseases (Section 3.1.5) can be caused by point mutations, deletions or duplications that abolish the function of genes in the densely packed mitochondrial genome (Figure 7.2). Cells typically contain thousands of mtDNA molecules. A major complication is that cells can be homoplasmic (every mtDNA molecule carries the causative mutation) or heteroplasmic (cells have a mixed population of normal and mutant mitochondrial DNA). Additionally, both the mutations and the heteroplasmy often seem to evolve with time within an individual. The same individual can carry both deletions and duplications, and the proportion can change with time (Poulton et al., 1993).

Phenotype-genotype correlations are particularly hard to establish with mitochondrial diseases. The same sequence change is frequently seen in people with different syndromes, and attempts to explain severity of symptoms by differing degrees of heteroplasmy have not been convincing - for example, 60–70% of people with Leber's hereditary optic atrophy (MIM 535000: sudden irreversible loss of vision) have a substitution at nt 11778 of the mitochondrial genome, but some patients appear to be homoplasmic while others, no less severely affected, are heteroplasmic. Possible reasons for the difficulties include:

• heteroplasmy can be tissue-specific, and the tissue that is examined (typically blood or muscle) may not be the critical tissue in the pathogenesis;
• mtDNA is much more variable than nuclear DNA, and some syndromes may depend on the combination of the reported mutation with other unidentified variants;
• some mitochondrial diseases seem to be of a quantitative nature: small mutational changes accumulate that reduce the energy-generating capacity of the mitochondrion, and at some threshold deficit clinical symptoms appear;
• many mitochondrial functions are encoded by nuclear genes (see Box 7.1), so that nuclear variation can be an important cause or modifier of mitochondrial phenotype.

The MITOMAP database of mitochondrial mutations (http://infinity.gen.emory.edu/mitomap.html) summarizes all the information on phenotypes and genotypes, and shows just how great is the challenge of relating them.

16.7.1 The gene underlying a disease may not be the obvious one

16.7.2 Locus heterogeneity is the rule rather than the exception

Locus heterogeneity describes the situation where the same disease can be caused by mutations in several different genes. It is important to think about the biological role of a gene product, and the molecules with which it interacts, rather than expecting a one-to-one relationship between genes and syndromes. As we saw in Section 3.1.4, clinical syndromes often result from failure or malfunction of a developmental or physiological pathway; equally, many cellular structures and functions depend on multicomponent protein aggregates. If the correct functioning of several genes is required, then mutations in any of the genes may cause the same, or a very similar, phenotype.

Once again, the collagens (Figure 16.4; Section 16.6.1) provide good examples. We have seen that type I collagen, the major collagen of skin, bone, tendon and ligaments, is built of triple helices comprising two α(1) chains and one α(2). Mutations in either the COL1A1 or COL1A2 genes cause the same condition, dominant osteogenesis imperfecta. Type II collagen forms fibrils in cartilage and other tissues including the vitreous of the eye. It is made of homotrimeric helices of COL2A1 chains. Different mutations in the COL2A1 gene result in an overlapping spectrum of skeletal dysplasias including Stickler syndrome, spondyloepiphyseal dysplasia and Kniest dysplasia. A similar phenotype can result from mutations in the type XI collagen, which is a minor component of the type II fibril. In all these cases, which syndrome is produced depends on the overall effect on the final collagen fibrils, and not on which gene is mutated.

16.7.3 Mutations in different members of a gene family can underlie a series of related or overlapping syndromes

Mutations in the genes encoding fibroblast growth factor receptors exemplify the way phenotypes can depend on a network of interactions rather than a single linear pathway. The ten fibroblast growth factors govern important developmental processes through four cell surface receptors, FGFR1-4. Most tissues express multiple FGFRs, including splice variants of each. The FGF receptors are receptor tyrosine kinases that act in a similar manner to the RET protein described above: signal transduction requires receptor dimerization, and this can involve homodimers or heterodimers. FGFR mutants could produce an altered balance of splice forms, change the balance of homo- and heterodimers, reduce signaling by a dominant negative effect or produce constitutionally active dimers. Thus there is the potential for complex genetic effects.

Very specific mutations of the receptor genes are responsible for a series of dominant disorders of skeletal growth (Figure 16.8). Mutations in FGFR2 on 10q25 are found in Crouzon, Jackson-Weiss, Pfeiffer and Apert syndromes, while different specific mutations in FGFR3 at 4p16 produce achondroplasia, thanatophoric dysplasia types 1 and 2, hypochondroplasia, Crouzon syndrome with acanthosis nigricans and Muencke's coronal craniosynostosis. Some patients with Pfeiffer syndrome have a mutation in FGFR1. For clinical descriptions, references and an introduction to the molecular pathology of these syndromes, see OMIM and Wilkie (1997). The very specific nature of the mutations suggests a gain of function, and the achondroplasia, thanatophoric dysplasia and Crouzon mutants have been shown to produce receptors with varying degrees of constitutive (ligand independent) activation when transfected into certain types of cells (Naski et al., 1996).

Figure 16.8

FGFR mutations. Three of the four highly homologous fibroblast growth factor receptors are shown. Each receptor tyrosine kinase has three immunoglobulin-like extracellular domains (held by S-S bridges), a transmembrane domain, and paired intracellular (more...)

16.7.4 Clinical and molecular classifications are alternative tools for thinking about diseases, and each is valid in its own sphere

The connective tissue disorders caused by collagen gene mutations, which have been a recurring theme in this chapter, illustrate the difference between clinical and molecular classifications of diseases.

• All mendelian diseases can be classified on a molecular basis (Table 16.6B), first by the locus involved and second by the particular mutant allele at that locus.
• Genetic diseases can also be classified clinically according to the symptoms and the prognosis, as well as by the pathogenesis (Table 16.6A). Clinical categories defined in this way may not correspond exactly to a molecular classification, but they may be more useful for suggesting the prognosis and how the patient should be managed.

Table 16.6A

Clinical classification of the connective tissue diseases. OI, osteogenesis imperfecta (brittle bone disease); SED, spondylo-epiphyseal dysplasia; EDS, Ehlers-Danlos syndrome.

Clinical labels are not simply conventions. They evolve as knowledge of the underlying genetics advances - diseases are lumped together (Duchenne and Becker muscular dystrophy) or split (BRCA1 breast cancer from sporadic breast cancer). A molecular classification is essential for molecular diagnosis, and it may allow more accurate counseling - for example, molecular analysis shows that unaffected parents who have more than one affected child with osteogenesis imperfecta are not carriers of a recessive form of OI, but germinal mosaics (see Figure 3.8). However, a full-blown molecular classification is not always clinically useful - for example, although OMIM lists nine loci causing Usher syndrome (recessive deafness and blindness), clinically it is only useful to distinguish three types, which vary in their severity. Thus a molecular classification illuminates rather than supersedes the clinical classification.

16.8.1 Contiguous gene and microdeletion syndromes bridge the gap between single gene and chromosomal syndromes

If our 3000 Mb genome contains 50 000–100 000 genes, a deletion of a megabase or so, which is too small to be seen under the microscope, may still involve dozens of genes. An increasing number of well characterized clinical syndromes are proving to be caused by such microdeletions, or occasionally microduplications (Table 16.7). Once the cause is recognized, further cases can easily be diagnosed by fluorescence in situ hybridization (Section 10.1.4) using a probe from the deleted region.

Table 16.7

Syndromes often caused by autosomal chromosomal microdeletions.

Autosomal microdeletion syndromes

Not all syndromes that can be associated with microdeletions are true microdeletion syndromes (reviewed by Budarf and Emanuel, 1997). For example Alagille syndrome (MIM 118450) is seen in patients with a microdeletion at 20p11, but 93% of Alagille patients have no deletion. The cause of the syndrome in all cases is haploinsufficiency for a single gene, JAG1 located at 20p11, due to deletions or point mutations. True microdeletion syndromes (sometimes called segmental aneusomy syndromes) are caused by haploinsufficiency of several genes. Langer-Giedion syndrome (LGS, MIM 150230) is an example of an autosomal contiguous gene syndrome. LGS is caused by deletion of two adjacent dosage-sensitive genes in 8q24, and possibly a third gene causing mental retardation. Mutation of the TRPS1 gene alone produces the typical face, bulbous nose and sparse hair of LGS; mutation of the adjacent EXT1 gene produces multiple exostoses. LGS combines these features with mental retardation. However, autosomal microdeletions do not in general produce clear-cut contiguous gene syndromes representing the cumulative effect of all the deleted genes. Homozygous deletions are usually lethal, and in a heterozygote only the few dosage-sensitive genes will affect the phenotype (Figure 16.9). Thus it is a challenge to work out which of the deleted genes causes which aspect of the syndrome.

Figure 16.9

X-linked and autosomal microdeletion syndromes. On the X chromosome, deletion of genes X₁ or X₅ is lethal in males. Patients 1–3 show a nested series of contiguous gene syndromes, patient 4 a nonoverlapping contiguous gene syndrome. On the autosome, (more...)

Williams syndrome (WLS; MIM 194050) provides an interesting example of these problems. People with WLS have a recognizable face, they are growth retarded, as infants they may have life-threatening hypercalcemia, and they often have supravalvular aortic stenosis (SVAS). As mentioned above (Section 16.4.3), isolated SVAS sometimes occurs as an autosomal dominant condition (MIM 185500). Dominant SVAS was mapped to 7q11.23 and shown sometimes to result from deletion or disruption of the elastin gene. This provided the clue for identifying the microdeletion in WLS. Typically 1–2 Mb of DNA is deleted, including the elastin gene. This explains the SVAS component of WLS. Potentially the facial features could also be caused by a deficiency of the connective tissue protein elastin, but this is evidently not the case because people with simple elastin mutations often have SVAS but do not have the characteristic Williams face. Several other genes have been identified that are deleted in WLS, but it is not easy to decide the role of each gene in the syndrome. This is especially true for the intriguing mental phenotype. People with WLS are usually moderately mentally retarded, to about the same extent as people with Down syndrome, but they have a very distinctive cognitive profile and personality (Tassabehji et al., 1999). They are highly sociable, often musical, and talk remarkably well, but have a specific inability at manipulating shapes (visuospatial constructive ability). Identifying the relevant genes from among all those deleted in WLS might provide an entry to identifying genetic determinants of normal human cognition and behavior. Other microdeletion syndromes are also associated with specific behaviors, and so unraveling the constituent genes of these syndromes is an area of great current research interest.

16.8.2 The major effects of chromosomal aneuploidies may be caused by dosage imbalances in a few identifiable genes

Monosomies and trisomies probably owe their characteristic phenotypes to a few major gene effects superimposed on many minor disturbances of development. Genes where a 50% increase of dosage has a major effect must be very uncommon, and so it should be possible to identify the few genes that cause the major features of, for example, Down syndrome (DS). Studies of patients with translocations show that the Down syndrome critical region is in 21q22.2; trisomy of other parts of chromosome 21 does not cause DS. At least two candidate genes for mental retardation have been identified from this region: DYRK, a gene whose Drosophila and mouse homologs (minibrain) produces dosagesensitive learning defects (Smith and Rubin, 1997), and DSCAM, a brain-specific cell adhesion molecule (Yamakawa et al., 1998).

The abnormalities of Turner syndrome might be due to haploinsufficiency at the earliest stages of development, before X inactivation takes place, but more probably stem from haploinsufficiency of genes that escape X inactivation and that have a functional Y counterpart. Only 18 such genes are known (Lahn and Page, 1997). Some people with partial deletions of Xp have a Turner or partial Turner phenotype, and their deletions can be used to try to pinpoint the genes responsible for particular features of the syndrome (see Zinn et al., 1998). Other candidates are suggested by overlapping mendelian conditions - the pseudoautosomal region at the tip of Xp (Section 2.4.3) contains a homeobox gene SHOX that is very occasionally mutated in chromosomally normal people with short stature or with Leri-Weill dyschondreostosis (MIM 127300). Thus SHOX is a candidate for the short stature of Turner syndrome.