Mechanisms by which transcription can regulate somatic hypermutation

DNA/RNA hybrid substrates modulate the catalytic activity of purified AID

Activation-induced cytidine deaminase (AID) converts cytidine to uridine at Immunoglobulin (Ig) loci, initiating somatic hypermutation and class switching of antibodies. In vitro, AID acts on single stranded DNA (ssDNA), but neither double-stranded DNA (dsDNA) oligonucleotides nor RNA, and it is believed that transcription is the in vivo generator of ssDNA targeted by AID. It is also known that the Ig loci, particularly the switch (S) regions targeted by AID are rich in transcription-generated DNA/RNA hybrids. Here, we examined the binding and catalytic behavior of purified AID on DNA/RNA hybrid substrates bearing either random sequences or GC-rich sequences simulating Ig S regions. If substrates were made up of a random sequence, AID preferred substrates composed entirely of DNA over DNA/RNA hybrids. In contrast, if substrates were composed of S region sequences, AID preferred to mutate DNA/RNA hybrids over substrates composed entirely of DNA. Accordingly, AID exhibited a significantly higher affinity for binding DNA/RNA hybrid substrates composed specifically of S region sequences, than any other substrates composed of DNA. Thus, in the absence of any other cellular processes or factors, AID itself favors binding and mutating DNA/RNA hybrids composed of S region sequences. AID:DNA/RNA complex formation and supporting mutational analyses suggest that recognition of DNA/RNA hybrids is an inherent structural property of AID.

ADAR deaminase A-to-I editing of DNA and RNA moieties of RNA:DNA hybrids has implications for the mechanism of Ig somatic hypermutation

The implications are discussed of recently published biochemical studies on ADAR-mediated A-to-I DNA and RNA deamination at RNA:DNA hybrids. The significance of these data are related to previous work on strand-biased and codon-context mutation signatures in B lymphocytes and cancer genomes. Those studies have established that there are two significant strand biases at A:T and G:C base pairs, A-site mutations exceed T-site mutations (A>>T) by 2.9 fold and G-site mutations exceed C-site mutations (G>>C) by 1.7 fold. Both these strand biases are inconsistent with alternative "DNA Deamination" mechanisms, yet are expected consequences of the RNA/RT-based "Reverse Transcriptase" mechanism of immunoglobulin (Ig) somatic hypermutation (SHM). The A-to-I DNA editing component at RNA:DNA hybrids that is likely to occur in Transcription Bubbles, while important, is of far lower A-to-I editing efficiency than in dsRNA substrates. The RNA moiety of RNA:DNA hybrids is also edited at similar lower frequencies relative to the editing rate at dsRNA substrates. Further, if the A-to-I DNA editing at RNA:DNA hybrids were the sole cause of A-to-I (read as A-to-G) mutation events for Ig SHM in vivo then the exact opposite strand biases at A:T base pairs (T>>A) of what is actually observed (A>>T) would be predicted. It is concluded that the strand-biased somatic mutation patterns at both A:T and G:C base pairs in vivo are best interpreted by the sequential steps of the RNA/RT-based mechanism. Further, the direct DNA A-to-I deamination at Transcription Bubbles is expected to contribute to the T-to-C component of the strand-biased Ig SHM spectrum.

Gene Conversion-Like Events in the Diversification of Human Rearranged IGHV3-23*01 Gene Sequences

Gene conversion (GCV), a mechanism mediated by activation-induced cytidine deaminase (AID) is well established as a mechanism of immunoglobulin diversification in a few species. However, definitive evidence of GCV-like events in human immunoglobulin genes is scarce. The lack of evidence of GCV in human rearranged immunoglobulin gene sequences is puzzling given the presence of highly similar germline donors and the presence of all the enzymatic machinery required for GCV. In this study, we undertook a computational analysis of rearranged IGHV3-23(*)01 gene sequences from common variable immunodeficiency (CVID) patients, AID-deficient patients, and healthy individuals to survey "GCV-like" activities. We analyzed rearranged IGHV3-23(*)01 gene sequences obtained from total PBMC RNA and single-cell polymerase chain reaction of individual B cell lysates. Our search identified strong evidence of GCV-like activity. We observed that GCV-like tracts are flanked by AID hotspot motifs. Structural modeling of IGHV3-23(*)01 gene sequence revealed that hypermutable bases flanking GCV-like tracts are in the single stranded DNA (ssDNA) of stable stem-loop structures (SLSs). ssDNA is inherently fragile and also an optimal target for AID. We speculate that GCV could have been initiated by the targeting of hypermutable bases in ssDNA state in stable SLSs, plausibly by AID. We have observed that the frequency of GCV-like events is significantly higher in rearranged IGHV3-23-(*)01 sequences from healthy individuals compared to that of CVID patients. We did not observe GCV-like events in rearranged IGHV3-23-(*)01 sequences from AID-deficient patients. GCV, unlike somatic hypermutation (SHM), can result in multiple base substitutions that can alter many amino acids. The extensive changes in antibody affinity by GCV-like events would be instrumental in protecting humans against pathogens that diversify their genome by antigenic shift.

Stabilised DNA secondary structures with increasing transcription localise hypermutable bases for somatic hypermutation in IGHV3-23

Somatic hypermutation (SHM) mediated by activation-induced cytidine deaminase (AID) is a transcription-coupled mechanism most responsible for generating high affinity antibodies. An issue remaining enigmatic in SHM is how AID is preferentially targeted during transcription to hypermutable bases in its substrates (WRC motifs) on both DNA strands. AID targets only single stranded DNA. By modelling the dynamical behaviour of IGHV3-23 DNA, a commonly used human variable gene segment, we observed that hypermutable bases on the non-transcribed strand are paired whereas those on transcribed strand are mostly unpaired. Hypermutable bases (both paired and unpaired) are made accessible to AID in stabilised secondary structures formed with increasing transcription levels. This observation provides a rationale for the hypermutable bases on both the strands of DNA being targeted to a similar extent despite having differences in unpairedness. We propose that increasing transcription and RNAP II stalling resulting in the formation and stabilisation of stem-loop structures with AID hotspots in negatively supercoiled region can localise the hypermutable bases of both strands of DNA, to AID-mediated SHM.

Transcriptional de-repression and Mfd are mutagenic in stressed Bacillus subtilis cells

In recent years, it has been proposed that conflicts between transcription and active chromosomal replication engender genome instability events. Furthermore, transcription elongation factors have been reported to prevent conflicts between transcription and replication and avoid genome instability. Here, we examined transcriptional de-repression as a genetic diversity-producing agent and showed, through the use of physiological and genetic means, that transcriptional de-represssion of a leuC defective allele leads to accumulation of Leu(+) mutations. We also showed, by using riboswitches that activate transcription in conditions of tyrosine or methionine starvation, that the effect of transcriptional de-repression of the leuC construct on the accumulation of Leu(+) mutations was independent of selection. We examined the role of Mfd, a transcription elongation factor involved in DNA repair, in this process and showed that proficiency of this factor promotes mutagenic events. These results are in stark contrast to previous reports in Escherichia coli, which showed that Mfd prevents replication fork collapses. Because our assays place cells under non-growing conditions, by starving them for two amino acids, we surmised that the Mfd mutagenic process associated with transcriptional de-repression does not result from conflicts with chromosomal replication. These results raise the interesting concept that transcription elongation factors may serve two functions in cells. In growing conditions, these factors prevent the generation of mutations, while in stress or non-growing conditions they mediate the production of genetic diversity.

Genome-wide analysis of DNA turnover and gene expression in stationary-phase Saccharomyces cerevisiae

Exponential-phase yeast cells readily enter stationary phase when transferred to fresh, carbon-deficient medium, and can remain fully viable for up to several months. It is known that stationary-phase prokaryotic cells may still synthesize substantial amounts of DNA. Although the basis of this phenomenon remains unclear, this DNA synthesis may be the result of DNA maintenance and repair, recombination, and stress-induced transposition of mobile elements, which may occur in the absence of DNA replication. To the best of our knowledge, the existence of DNA turnover in stationary-phase unicellular eukaryotes remains largely unstudied. By performing cDNA-spotted (i.e. ORF) microarray analysis of stationary cultures of a haploid Saccharomyces cerevisiae strain, we demonstrated on a genomic scale the localization of a DNA-turnover marker [5-bromo-2'-deoxyuridine (BrdU); an analogue of thymidine], indicative of DNA synthesis in discrete, multiple sites across the genome. Exponential-phase cells on the other hand, exhibited a uniform, total genomic DNA synthesis pattern, possibly the result of DNA replication. Interestingly, BrdU-labelled sites exhibited a significant overlap with highly expressed features. We also found that the distribution among chromosomes of BrdU-labelled and expressed features deviates from random distribution; this was also observed for the overlapping set. Ty1 retrotransposon genes were also found to be labelled with BrdU, evidence for transposition during stationary phase; however, they were not significantly expressed. We discuss the relevance and possible connection of these results to DNA repair, mutation and related phenomena in higher eukaryotes.

Post-translational regulation of activation-induced cytidine deaminase

The assembled immunoglobulin genes in the B cells of mice and humans are altered by distinct processes known as class switch recombination (CSR) and somatic hypermutation, leading to diversification of the antibody repertoire. These two DNA modification processes are initiated by the B cell-specific protein factor activation-induced cytidine deaminase (AID). AID is post-translationally modified by phosphorylation at multiple sites, although functional significance during CSR has been implicated only for phosphorylation at serine-38 (S38). Although multiple laboratories have demonstrated that AID function is regulated via phosphorylation at S38, the precise biological role of S38 phosphorylation has been a topic of debate. Here, we discuss our interpretation of the significance of AID regulation via phosphorylation and also discuss how this form of AID regulation may have evolved in higher organisms.

Mechanism of somatic hypermutation: critical analysis of strand biased mutation signatures at A:T and G:C base pairs

The DNA sequence data of the somatic hypermutation (SHM) field published since 1984 has been critically reviewed. The analysis has revealed three strand biased mutation signatures. The first concerns the mutations generated at G:C base pairs in mice genetically deficient in uracil-DNA glycosylase and MSH2-MSH6-mediated mismatch repair. Such mice display the AID deaminase footprint and here C mutations exceed G mutations at least 1.5-fold. This supports earlier and more recent studies claiming that dC-to-dU deaminations occur preferentially in the single stranded DNA regions of the displaced nontranscribed strand (NTS) during transcription. The second concerns the signature generated in immunised mice where G mutations exceed C mutations by at least 1.7-fold. This is a newly identified strand bias which has previously gone undetected. It is consistent with the polynucleotide polymerisation signature of RNA polymerase II copying the template DNA strand carrying AID-mediated lesions generated at C bases, viz. uracils and abasic sites. A reverse transcription step would then need to intervene to fix the mutation pattern in DNA. The third concerns the long recognised strand biased signature generated in normal aged or actively immunised mice whereby A mutations exceed T mutations by two- to three-fold. It is argued that this pattern is best understood as a combination of adenosine-to-inosine (A-to-I) RNA editing followed by a reverse transcription step fixing the A-to-G, as well as A-to-T and A-to-C, as strand biased mutation signatures in DNA. The reasons why the AID-linked RNA polymerase II mutation signature had previously gone undetected are discussed with regard to limitations of standard PCR-based SHM assay techniques. It is concluded that the most economical SHM mechanism involves both DNA and RNA deaminations coupled to a reverse transcription process, most likely involving DNA polymerase eta acting in its reverse transcriptase mode. Experimental approaches to differentiate this RNA-based model from the standard DNA deamination model are discussed.

I. VH gene transcription creates stabilized secondary structures for coordinated mutagenesis during somatic hypermutation

During the adaptive immune response, antigen challenge triggers a million-fold increase in mutation rates in the variable-region antibody genes. The frequency of mutation is causally and directly linked to transcription, which provides ssDNA and drives supercoiling that stabilizes secondary structures containing unpaired, intrinsically mutable bases. Simulation analysis of transcription in VH5 reveals a dominant 65nt secondary structure in the non-transcribed strand containing six sites of mutable ssDNA that have also been identified independently in human B cell lines and in primary mouse B cells. This dominant structure inter-converts briefly with less stable structures and is formed repeatedly during transcription, due to periodic pauses and backtracking. In effect, this creates a stable yet dynamic "mutability platform" consisting of ever-changing patterns of unpaired bases that are simultaneously exposed and therefore able to coordinate mutagenesis. Such a complex of secondary structures may be the source of ssDNA for enzyme-based diversification, which ultimately results in high affinity antibodies.

II. Correlations between secondary structure stability and mutation frequency during somatic hypermutation

The role of secondary structures and base mutability at different levels of transcription and supercoiling is analyzed in variable region antibody genes VH5, VH94 and VH186.2. The data are consistent with a model of somatic hypermutation in which increasing levels of transcription and secondary structure stability correlate with the initial formation of successive mutable sites. Encoded differences exist in stem length and the number of GC pairs at low versus high levels of transcription in CDRs. These circumstances simplify the complexities of coordinating mutagenesis by confining this process to each mutable site successively, as they form in response to increasing levels of transcription during affinity maturation.

The biochemistry of somatic hypermutation

Affinity maturation of the humoral response is mediated by somatic hypermutation of the immunoglobulin (Ig) genes and selection of higher-affinity B cell clones. Activation-induced cytidine deaminase (AID) is the first of a complex series of proteins that introduce these point mutations into variable regions of the Ig genes. AID deaminates deoxycytidine residues in single-stranded DNA to deoxyuridines, which are then processed by DNA replication, base excision repair (BER), or mismatch repair (MMR). In germinal center B cells, MMR, BER, and other factors are diverted from their normal roles in preserving genomic integrity to increase diversity within the Ig locus. Both AID and these components of an emerging error-prone mutasome are regulated on many levels by complex mechanisms that are only beginning to be elucidated.

G4-forming sequences in the non-transcribed DNA strand pose blocks to T7 RNA polymerase and mammalian RNA polymerase II

DNA sequences rich in runs of guanine have the potential to form G4 DNA, a four-stranded non-canonical DNA structure stabilized by formation and stacking of G quartets, planar arrays of four hydrogen-bonded guanines. It was reported recently that G4 DNA can be generated in Escherichia coli during transcription of plasmids containing G-rich sequences in the non-transcribed strand. In addition, a stable RNA/DNA hybrid is formed with the transcribed strand. These novel structures, termed G loops, are suppressed in recQ(+) strains, suggesting that their persistence may generate genomic instability and that the RecQ helicase may be involved in their dissolution. However, little is known about how such non-canonical DNA structures are processed when encountered by an elongating polymerase. To assess whether G4-forming sequences interfere with transcription, we studied their effect on transcription elongation by T7 RNA polymerase and mammalian RNA polymerase II. We used a reconstituted transcription system in vitro with purified polymerase and initiation factors and with substrates containing G-rich sequences in either the transcribed or non-transcribed strand downstream of the T7 promoter or the adenovirus major late promoter. We report that G-rich sequences located in the transcribed strand do not affect transcription by either polymerase, but when the sequences are located in the non-transcribed strand, they partially arrest both polymerases. The efficiency of arrest increases with negative supercoiling and also with multiple rounds of transcription compared with single events.

Ectopic expression of AID in a non-B cell line triggers A:T and G:C point mutations in non-replicating episomal vectors

Somatic hypermutation (SHM) of immunoglobulin genes is currently viewed as a two step process initiated by the deamination of deoxycytidine (C) to deoxyuridine (U), catalysed by the activation induced deaminase (AID). Phase 1 mutations arise from DNA replication across the uracil residue or the abasic site, generated by the uracil-DNA glycosylase, yielding transitions or transversions at G:C pairs. Phase 2 mutations result from the recognition of the U∶G mismatch by the Msh2/Msh6 complex (MutS Homologue), followed by the excision of the mismatched nucleotide and the repair, by the low fidelity DNA polymerase η, of the gap generated by the exonuclease I. These mutations are mainly focused at A∶T pairs. Whereas in activated B cells both G:C and A∶T pairs are equally targeted, ectopic expression of AID was shown to trigger only G:C mutations on a stably integrated reporter gene. Here we show that when using non-replicative episomal vectors containing a GFP gene, inactivated by the introduction of stop codons at various positions, a high level of EGFP positive cells was obtained after transient expression in Jurkat cells constitutively expressing AID. We show that mutations at G:C and A∶T pairs are produced. EGFP positive cells are obtained in the absence of vector replication demonstrating that the mutations are dependent only on the mismatch repair (MMR) pathway. This implies that the generation of phase 1 mutations is not a prerequisite for the expression of phase 2 mutations.

Secondary structures as predictors of mutation potential in the lacZ gene of Escherichia coli

Four independent nonsense mutations were engineered into the Escherichia coli chromosomal lacZ gene, and reversion rates back to LacZ(+) phenotypes were determined. The mutation potential of bases within putative DNA secondary structures formed during transcription was predicted by a sliding-window analysis that simulates successive folding of the ssDNA creating these structures. The relative base mutabilities predicted by the mfg computer program correlated with experimentally determined reversion rates in three of the four mutants analysed. The nucleotide changes in revertants at one nonsense codon site consisted of a triple mutation, presumed to occur by a templated repair mechanism. Additionally, the effect of supercoiling on mutation was investigated and, in general, reversion rates increased with higher levels of negative supercoiling. Evidence indicates that predicted secondary structures are in fact formed in vivo and that directed mutation in response to starvation stress is dependent upon the exposure of particular bases, the stability of the structures in which these bases are unpaired and the level of DNA supercoiling within the cell.

The effect of promoter strength, supercoiling and secondary structure on mutation rates in Escherichia coli

Four mutations resulting in opal stop codons were individually engineered into a plasmid-borne chloramphenicol-resistance (cat) gene driven by the lac promoter. These four mutations were located at different sites in secondary structures. The mutations were analysed with the computer program mfg, which predicted their relative reversion frequencies. Reversion frequencies determined experimentally correlated with the mutability of the bases as predicted by mfg. To examine the effect of increased transcription on reversion frequencies, the lac promoter was replaced with the stronger tac promoter, which resulted in 12- to 30-fold increases in reversion rates. The effect of increased and decreased supercoiling was also investigated. The cat mutants had higher reversion rates in a topA mutant strain with increased negative supercoiling compared with wild-type levels, and the cat reversion rates were lower in a topA gyrB mutant strain with decreased negative supercoiling, as predicted.

Ancient Phylogenetic Beginnings of Immunoglobulin Hypermutation

Many structures and molecules closely related to those involved in the specific process of immunoglobulin (Ig) hypermutation existed before the appearance of primordial Ig genes. Consequently, these structures can be found even in animals and organisms distinct from vertebrates; likewise, homologues of hypermutation enzymes are present in a broad range of species, from bacteria to mammals. Our analysis, based predominantly on primary structure, demonstrates the existence of molecules similar to Ig domains, variable Ig domains (IGv), and antigen receptors (AR) in unicellular organisms, nonvertebrate metazoans, and nonvertebrate Coelomata, respectively. In addition, we deal here with some important structural properties of CDR1-like segments of the selected sponge adhesion molecule GCSAMS exhibiting chimerical Ig domain similarities, and demonstrate the occurrence of conserved regions corresponding to Ohno's modern intact primordial building block in the C-terminal part of IGv-related segments of nonvertebrate origin. The results of our analysis are also discussed with respect to the possible phylogeny of molecules preceding the hypothetical common antigen receptor ancestor.

Secondary structure formations of conotoxin genes: A possible role in mediating variability

Small venomous peptides called conotoxins produced by the predatory marine snail (genus Conus) present an interesting case for mutational studies. They have a high degree of amino acid variability among them yet they possess highly conserved structural elements that are defined by cysteine residues forming disulfide bridges along the length of the mature peptide. It has been observed that codons specifying these cysteines are also highly conserved. It is unknown how such codon conservation is maintained within the mature conotoxin gene since this entire region undergoes an accelerated rate of mutation. There is evidence suggesting that nucleic acids wield some influence in mechanisms that dictate the region and frequency where mutations occur in DNA. Nucleic acids exert this effect primarily through secondary structures that bring about local peaks and troughs in the energy relief of these transient formations. Secondary structure predictions of several conotoxin genes were analyzed to see if there was any correspondence between the highly variable regions of the conotoxin. Regions of the DNA encompassing the conserved Cys codons (and several other conserved amino acid codons) have been found to correspond to predicted secondary structures of higher stabilities. In stark contrast the regions of the conotoxin that have a higher degree of variation correlate to regions of lower stability. This striking co-relation allows for a simple model of inaccessibility of a mutator to these highly conserved regions of the conotoxin gene allowing them a relative degree of resistance towards change.

Computational analyses show A-to-G mutations correlate with nascent mRNA hairpins at somatic hypermutation hotspots

Activation-induced cytidine deaminase (AID) initiates Phase I somatic hypermutation (SHM) of antibody genes by deaminating deoxy-cytosine to deoxy-uracil (C-to-U). These lesions trigger Phase II, a poorly understood process of error-prone repair targeting A-T pairs by DNA polymerase eta (Pol eta). Since Pol eta is also a reverse transcriptase, Phase II could involve copying off RNA as well as DNA templates. We explore this idea further since in an RNA-based pathway it is conceivable that adenosine-to-inosine (A-to-I) RNA editing causes A-to-G transitions since I like G pairs with C. Adenosine deaminases (ADARs) are known to preferentially edit A nucleotides that are preceded by an A or U (W) in double-stranded RNA substrates. On this assumption and using a theoretical bioinformatics approach we show that a significant and specific correlation (P<0.002) exists between the frequency of WA-to-WG mutations and the number of mRNA hairpins that could potentially form at the mutation site. This implies roles for both RNA editing and reverse transcription during SHM in vivo and suggests definitive genetic experiments targeting the appropriate ADAR1 isoform (gammaINF-ADAR1) and/or Ig pre-mRNA templates.

Hypothesis: biological role for J-C intronic matrix attachment regions in the molecular mechanism of antigen-driven somatic hypermutation

A major function of J-C intronic matrix attachment regions (MAR) during immune diversification via somatic hypermutation (SHM) at immunoglobulin loci may be to manipulate the topology of DNA within the upstream target domain. The suggestion that SHM induction requires MAR-induced torsional strain, in conjunction with DNA remodelling at the J-C intron, completes the definition of a cogent paradigm within which all extant molecular data on the issue may be interpreted. Moreover, the suggestion that a mutagenic mechanism relieves MAR-generated superhelicity could provide an indication as to the evolutionary basis of SHM.

Advances in mechanisms of genetic instability related to hereditary neurological diseases

Substantial progress has been realized in the past several years in our understanding of the molecular mechanisms responsible for the expansions and deletions (genetic instabilities) of repeating tri-, tetra- and pentanucleotide repeating sequences associated with a number of hereditary neurological diseases. These instabilities occur by replication, recombination and repair processes, probably acting in concert, due to slippage of the DNA complementary strands relative to each other. The biophysical properties of the folded-back repeating sequence strands play a critical role in these instabilities. Non-B DNA structural elements (hairpins and slipped structures, DNA unwinding elements, tetraplexes, triplexes and sticky DNA) are described. The replication mechanisms are influenced by pausing of the replication fork, orientation of the repeat strands, location of the repeat sequences relative to replication origins and the flap endonuclease. Methyl-directed mismatch repair, nucleotide excision repair, and repair of damage caused by mutagens are discussed. Genetic recombination and double-strand break repair advances in Escherichia coli, yeast and mammalian models are reviewed. Furthermore, the newly discovered capacities of certain triplet repeat sequences to cause gross chromosomal rearrangements are discussed.

On the molecular mechanism of somatic hypermutation of rearranged immunoglobulin genes

Somatic hypermutation (SHM) diversifies the genes that encode immunoglobulin variable regions in antigen-activated germinal centre B lymphocytes. Available evidence strongly suggests that DNA deamination potentiates phase I SHM and subsequently triggers phase II SHM. A concise review of this evidence is followed by a detailed critique of two possible models which suggest that polymerase-eta potentiates phase II SHM via either its DNA-dependent or its RNA-dependent DNA synthetic activity. Quantitative analysis, in the context of extant data that define the features of SHM, favours the RNA-dependent mechanism.

Stress-directed adaptive mutations and evolution

Comparative biochemistry demonstrates that the metabolites, complex biochemical networks, enzymes and regulatory mechanisms essential to all living cells are conserved in amazing detail throughout evolution. Thus, in order to evolve, an organism must overcome new adverse conditions without creating different but equally dangerous alterations in its ongoing successful metabolic relationship with its environment. Evidence suggests that stable long-term acquisitive evolution results from minor increases in mutation rates of genes related to a particular stress, with minimal disturbance to the balanced and resilient metabolism critical for responding to an unpredictable environment. Microorganisms have evolved specific biochemical feedback mechanisms that direct mutations to genes derepressed by starvation or other stressors in their environment. Transcription of the activated genes creates localized supercoiling and DNA secondary structures with unpaired bases vulnerable to mutation. The resulting mutants provide appropriate variants for selection by the stress involved, thus accelerating evolution with minimal random damage to the genome. This model has successfully predicted mutation frequencies in genes of E. coli and humans. Stressed cells observed in the laboratory over hundreds of generations accumulate mutations that also arise by this mechanism. When this occurs in repair-deficient mutator strains with high rates of random mutation, the specific stress-directed mutations are also enhanced.