Biology Reference
In-Depth Information
with this explanation, the relative base substitution rates observed in the context
of human genetic disease (Cooper and Krawczak, 1993; Krawczak and Cooper,
1996a) were found to be remarkably similar to those derived by Hess
et al
. (1994)
in an extensive analysis of human gene/pseudogene alignments. This pattern of
similarity was still apparent after allowance had been made for the DNA sequence
environment at the site of mutation by the use of nearest neighbor-dependent
mutation rates (Krawczak and Cooper, 1996a); the only notable difference was a
slight under-representation among pseudogene mutations of C
A
substitutions. This was held to be explicable in terms of either a lower level of
germline methylation or, perhaps more likely, a deficiency of the 5mC-containing
CG dinucleotides ('CG suppression'; Bird, 1980) in non-coding DNA sequences.
The claim that relative mutation rates exhibit a strong positive correlation
between pathological mutations in coding sequences and evolutionarily fixed
mutations in non-coding sequences (Krawczak and Cooper, 1996a) might appear
surprising in the light of other experimental results. For example, a study of
murine
aprt
gene sequences has claimed that silent mutations accumulate more
slowly in transcribed sequences, possibly due to preferential DNA repair (Boulikas,
1992; Turker
et al
., 1993). Further, Hanawalt (1990) has shown that
in vitro
-induced
pyrimidine dimers and interstrand DNA crosslinks are repaired with a substan-
tially higher efficiency in active genes than in noncoding regions. Although this
type of lesion is specific to the action of particular exogenous chemical mutagens
and irradiation, the idea of a system which is generally more effective at removing
endogenous mutations from coding DNA as opposed to noncoding DNA is appeal-
ing. This is because efficient DNA repair should only have conferred a substantial
selective advantage in coding regions. By contrast, the results of Krawczak and
Cooper (1996a) suggest that the relative contribution (via variable efficiency) of dif-
ferent DNA repair pathways to the generation of mutations is unlikely to differ
substantially between intragenic and intergenic sequences.
To what extent is the likelihood of generation of a mutation related to its phe-
notypic consequences? When codon substitutions causing genetic disease were
categorised according to whether they were neutral or whether they changed the
hydrophobicity or polarity of the encoded amino acid residue, it emerged that
neutral changes were characterized by larger likelihoods of generation via muta-
tion than nonneutral substitutions (Krawczak and Cooper, 1996a). This disparity
suggested that selection has operated on the cellular DNA repair machinery in
such a way as to optimize the removal of the latter type of mutation. If nonneutral
changes were more likely to result in a disadvantageous (disease) phenotype than
neutral substitutions, then any repair bias operating against these changes at the
DNA level would have had a selective advantage.
The hypothesis of a mutational repair bias was further supported by the finding
that the likelihood of generation of an amino acid substitution in humans is neg-
atively correlated with its likelihood of coming to clinical attention (Krawczak
and Cooper, 1996a). The extent of the correlation was, however, found to differ
dramatically between different types of substitution. Thus, a significant decrease
in mutation generation likelihood was only associated with an increased likeli-
hood of clinical observation for substitutions which affected hydrophilic and
polar residues. Various explanations may be considered to account for these
T and G
