There seems to be some solid evidence that transcription promotes mutation because the untranscribed strand is able to form secondary structures which expose bases to chemical mutagenesis.
Here is a recent paper about transcription-associated mutagenesis:
Kim H et al.(2010) Transcription-associated mutagenesis increases protein sequence diversity more effectively than does random mutagenesis in Escherichia coli. PLoS One 5(5):e10567. doi: 10.1371/journal.pone.0010567.
From the abstract:
During transcription, the nontranscribed DNA strand becomes single-stranded DNA (ssDNA), which can form secondary structures. Unpaired bases in the ssDNA are less protected from mutagens and hence experience more mutations than do paired bases. These mutations are called transcription-associated mutations. Transcription-associated mutagenesis is increased under stress and depends on the DNA sequence. Therefore, selection might significantly influence protein-coding sequences in terms of the transcription-associated mutability per transcription event under stress to improve the survival of Escherichia coli.
The authors cite a number of papers in their introduction which document the phenomenon that you could follow up. Just in case the focus on a bacterial system puts you off, the Kim et al. paper has in turn been cited in:
Wright et al. (2011) The roles of transcription and genotoxins underlying p53 mutagenesis in vivo. CARCINOGENESIS 32:1559-1567
Abstract in full:
Transcription drives supercoiling which forms and stabilizes single-stranded (ss) DNA secondary structures with loops exposing G and C bases that are intrinsically mutable and vulnerable to non-enzymatic hydrolytic reactions. Since many studies in prokaryotes have shown direct correlations between the frequencies of transcription and mutation, we conducted in silico analyses using the computer program, mfg, which simulates transcription and predicts the location of known mutable bases in loops of high-stability secondary structures. Mfg analyses of the p53 tumor suppressor gene predicted the location of mutable bases and mutation frequencies correlated with the extent to which these mutable bases were exposed in secondary structures. In vitro analyses have now confirmed that the 12 most mutable bases in p53 are in fact located in predicted ssDNA loops of these structures. Data show that genotoxins have two independent effects on mutagenesis and the incidence of cancer: Firstly, they activate p53 transcription, which increases the number of exposed mutable bases and also increases mutation frequency. Secondly, genotoxins increase the frequency of G-to-T transversions resulting in a decrease in G-to-A and C mutations. This precise compensatory shift in the 'fate' of G mutations has no impact on mutation frequency. Moreover, it is consistent with our proposed mechanism of mutagenesis in which the frequency of G exposure in ssDNA via transcription is rate limiting for mutation frequency in vivo.