Generation of mutation hotspots in ageing bacterial colonies

How do ageing bacterial colonies generate adaptive mutants? Over a period of two months, we isolated on ageing colonies outgrowing mutants able to use a new carbon source, and sequenced their genomes. This allowed us to uncover exquisite details on the molecular mechanism behind their adaptation: most mutations were located in just a few hotspots in the genome and over time, mutations increasingly originated from 8-oxo-guanosine, formed exclusively on the transcribed strand.


Introduction
Bacteria constitute a precious biological model system for studying the molecular details of ageing and evolution. Bacterial cells defective in the MutY enzyme, responsible for removing adenine nucleotides pairing with the 8-oxo oxidized variant of guanosine (8-oxo-G), exhibit a dramatic increase in the number of adaptive mutants and Bridges proposed a model to explain this phenomenon that was later termed retromutagenesis 1,2 . In this model, the process of transcription opens up the DNA double helix, enhancing the probability of mutation in the transcribed strand, but only mutations on the transcribed strand are transferred to mRNA and translate into mutants proteins that explore novel activities. Subsequently, the activity of the mutant protein may enable the cell to replicate and thereby fixes the initial adaptive mutation on both DNA strands. In agreement with the retromutagenesis model, lacZ amber mutations on the transcribed strand were recently isolated in approximately 10fold excess over mutations on the non-transcribed strand upon treatment with the mutagen nitrous acid 3 . However, the prevalence of this molecular mechanism has not not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/041525 doi: bioRxiv preprint first posted online Feb. 26, 2016; been studied in a non-mutator background and has not been validated in a whole genome context. To gain new, in-depth knowledge of the mechanisms behind adaptive mutation we wanted to study the genetic changes in a background as undisturbed as possible. For this purpose, we designed an E. coli strain incapable of fermenting maltose, plated it on rich medium with maltose, and subsequently collected all mutants starting to outgrow on colonies over the course of two months ( Figure 1).

Aging colonies give rise to mutants in a non-mutator-based experimental system
Our starter strain is a derivative of the model bacterium Escherichia coli K12 MG1655 with its cyaA gene deleted, precluding synthesis of the signalling molecule cyclic AMP. As a consequence, a great many genes involved in carbon source catabolism cannot be expressed because they belong to operons requiring the cAMP Receptor Protein (CRP) complex to be activated 4 . These cells can grow for some time on rich media, but, after having used the accessible sources present in the available medium (MacConkey medium), the cells cannot grow further but remain as small colonies caught in a quiescent state.
Around four days after plating, mutants capable of fermenting maltose started to appear as red papillae-like structures on the white colonies, and these mutants continued to appear over the next two months as some cells adapted to their environment. Mutant papillae outgrew on approximately one in 200 colonies, and progressively invaded the surface of the plate (Figure 2a) making it necessary to start not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. with a large number of plates to be able to sample mutants at the late time points. All mutant strains were purified and displayed a variety of phenotypes on different carbon sources (Supplementary Table 1).

Adaptive mutations are located in specific hotspot genes
We sequenced the genomes of 96 mutants, spanning the whole two-month period, and From the way the experiment was constructed, it could be expected that mutations in the crp gene would account for growth on maltose in a ΔcyaA background. However, the difficulty of obtaining cAMP-independent (crp*) mutants witnessed by Jon Beckwith's laboratory suggested that under exponential growth such mutations were very rare (of the order of 1 in 10 9 cells 5 ). In light of this, the ease with which we obtained such mutations in resting colonies was utterly unexpected.
It has been repeatedly observed that in E. coli laboratory strains, the rpoS gene is often inactivated 6,7 . This is also what we observed at the rpoS hotspot (point mutations in general, but also frameshifts and a deletion of the region, see Supplementary Table 2). Inactivation of the RpoS protein may have triggered an increase in age-related mutagenesis, as this transcription factor is involved in not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/041525 doi: bioRxiv preprint first posted online Feb. 26, 2016; oxidative stress response during stationary phase 8 . However, before we understand the molecular mechanism, we can only speculate why the remaining hotspot genes are targeted. The majority of adaptive mutations take place on the transcribed strand Remarkably, the mutated base is highly dominantly present on the transcribed strand within gene coding sequences (84% for G transversions and 93% for C deaminations, Figure 3a). Even more extremely, in a total of 594 crp variants we sequenced from different papillae, 99% of the G transversion and C deamination events had taken place on the transcribed strand ( Figure 3b). These observations are consistent with an increased mutation rate in transcribed regions (see Supplementary Table 2) and strongly support the retromutagenesis model (Figure 3 c-d) 1,3,12 -here in the absence of a mutagen. Importantly, the extreme strand bias considerably reduces the relevance of potential residual replication and growth in the aging colonies -a heavily debated, but hard to test, part of the controversy of adaptive mutations 13 . not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/041525 doi: bioRxiv preprint first posted online Feb. 26, 2016; In the retromutagenesis model, mutations generated on the transcribed strand are selected for. In contrast, a so-called hitchhiking mutation is per definition not selected. Are the other hotspots selected for or hitchhikers? It is interesting to compare the specific nucleotide changes in crp that occur alone (singles) with those found only in

Formation of adaptive mutations shows age-dependence
not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/041525 doi: bioRxiv preprint first posted online Feb. 26, 2016; Time is an important factor in the development of mutant papillae and the sampling of adaptive mutants over two months enabled us to observe interesting age-related trends. Firstly, a minimum of four days of incubation is required for adaptive mutations to occur in our experimental setup. Secondly, the frequency of papillae with rpoS mutations increases over time, suggesting an increasingly selective advantage with age ( Figure 4a). Alternatively, the mechanism(s) leading to rpoS mutations are becoming more prominent over time. Thirdly, G transversion mutations in crp increases over time, paralleled by a decrease in C-deaminations (Figure 4b).
Our observations are not consistent with a general increase in mutagenesis, as the frequency of rifampicin resistance is of the order of four mutants per 10 8 cells as generally observed 14 . Furthermore, removal of RecA, a previously suggested key actor in adaptive mutations 15 , displayed only a marginally negative effect on the total number of papillae formed (Supplementary Figure 1).

Discussion
In the present study, we sequenced 96 adaptive mutants isolated during a two-month period, after having arisen from a minimally perturbed genetic background. To our knowledge, a similar experiment has not been performed at this time scale previously. Importantly, time was an essential factor not only for the development of adaptive mutation, but also seemingly for the mutational mechanism and we show that a highly dominating proportion of the adaptive mutations are 8-oxo-G modifications on the transcribed strand. This strongly supports the retromutagenesis model as a major mechanism behind adaptive mutations. Were this process generalised to multicellular organisms it would be a fertile contribution to the initiation of cancer that parallels ageing. not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.

Methods
Because the fnr gene is a homologue of crp, we used a starter strain previously identified as MG1655 at the ECGSC1 shown to harbour a ca. 40 kb deletion in the fnr region 16 . The cyaA deletion was introduced with lambda Red-induced recombination 17 . In pilot experiments, we had noticed that the presence of a pTrc99A plasmid (Pharmacia Biotech) carrying the E. coli tig gene (PCR amplified with the oligonucleotides CATGCCATGGTGAGGTAACAAGATGCAAGTTTC 3' and

5' CGCGGATCCAATTACGCCTGCTGGTTCATC 3' and cloned into the NcoI and
BamHI restriction sites) produced twice as many mutants that in absence of the tig plasmid, typically allowing us to recover around 200 mutants in each two months experiment, a figure quite convenient to get significant observations. Thus K12 Mg1655 (cyaA::cat delta fnr, pTrctig) is our starter strain. The recA deletion was from the Keio collections 18 that was used to prepare a P1 lysate by standard procedures.
For growth on plates the lactose-free MacConkey medium was used 19 (Difco MacConkey Agar Base) supplemented with 1% carbohydrate (maltose in most experiments) or glycerol and chloramphenicol, 5 mg/litre, ampicillin, 100 mg/litre, and 1 mM IPTG. For subsequent 3-times purification of papillae, chloramphenicol, ampicillin and IPTG were omitted from the plates. To obtain isolated colonies on the plates, early stationary-phase grown bacteria from the LB liquid culture (37°C) were diluted in sterile water containing 9 g/l sodium chloride, to the concentration of 2.5×10 4 bacteria/ml and 100 µl of the bacterial suspension was spread onto MacConkey plates. The plates were subsequently placed in plastic boxes containing beakers with water to ensure constant humidity and placed for the duration of the not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/041525 doi: bioRxiv preprint first posted online Feb. 26, 2016; experiment in an incubator at 37°C. To test for mutator phenotypes, rifampicin agar plates were prepared using the LB medium agar supplemented with 100 mg/litre rifampicin prepared in methanol. At all times the plates were wrapped in aluminium foil for protection from light.    Over time, there is a steady increase in 8-oxoG mutations whereas the C-deamination associated mutations decrease. not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.