the origin of mutants under selection: how natural

The Origin of Mutants Under Selection:How Natural Selection Mimics Mutagenesis(Adaptive Mutation)

Sophie Maisnier-Patin and John R. Roth

Department of Microbiology and Molecular Genetic, University of California, Davis, California 95616

Correspondence: [email protected]

Selection detects mutants but does not cause mutations. Contrary to this dictum, Cairns andFoster plated a leaky lac mutant of Escherichia coli on lactose medium and saw revertant(Lacþ) colonies accumulate with time above a nongrowing lawn. This result suggested thatbacteria might mutagenize their own genome when growth is blocked. However, this con-clusion is suspect in the light of recent evidence that revertant colonies are initiated bypreexisting cells with multiple copies the conjugative F0lac plasmid, which carries the lacmutation. Some plated cells have multiple copies of the simple F0lac plasmid. This providessufficient LacZ activity to support plasmid replication but not cell division. In nongrowingcells, repeated plasmid replication increases the likelihood of a reversion event. Reversion tolacþ triggers exponential cell growth leading to a stable Lacþ revertant colony. In 10% ofthese plated cells, the high-copy plasmid includes an internal tandem lac duplication, whichprovides even more LacZ activity—sufficient to support slow growth and formation of anunstable Lacþ colony. Cells with multiple copies of the F0lac plasmid have an increasedmutation rate, because the plasmid encodes the error-prone (mutagenic) DNA polymerase,DinB. Without DinB, unstable and stable Lacþ revertant types form in equal numbers andboth types arise with no mutagenesis. Amplification and selection are central to behavior ofthe Cairns–Foster system, whereas mutagenesis is a system-specific side effect or artifactcaused by coamplification of dinB with lac. Study of this system has revealed several broadlyapplicable principles. In all populations, gene duplications are frequent stable genetic poly-morphisms, common near-neutral mutant alleles can gain a positive phenotype when am-plified under selection, and natural selection can operate without cell division when vari-ability is generated by overreplication of local genome subregions.

AN INITIAL POINT TO CONSIDER

In considering the possibility of regulated mu-tagenesis, it is useful to consider a basic, but

often forgotten truism. Most new mutations are

near-neutral or deleterious types that are eas-ily missed or ignored. Similarly, copy-numberchanges of any particular gene are extremelycommon (1025/cell/division in Salmonella)but hard to detect. In contrast, rare beneficial

Editor: Howard Ochman

Additional Perspectives on Microbial Evolution available at www.cshperspectives.org

Copyright # 2015 Cold Spring Harbor Laboratory Press; all rights reserved; doi: 10.1101/cshperspect.a018176

Cite this article as Cold Spring Harb Perspect Biol 2015;7:a018176

1

Laboratory Press at CALIFORNIA-DAVIS on July 23, 2015 - Published by Cold Spring Harborhttp://cshperspectives.cshlp.org/Downloaded from

mailto:[email protected]

mailto:[email protected]

http://www.cshperspectives.org



http://cshperspectives.cshlp.org/

mutations with big effects are obvious becauseselection allows them to increase in frequency.Genetic analysis has misdirected our attentiontoward these rare large-effect beneficial chang-es. It is easy to forget the “silent majority”—the near-neutral and deleterious changes—but it is important to remember that naturalselection sees almost everything and is alwayswatching.

THE MAJOR QUESTION UNDERCONSIDERATION

Bacterial genetics uses stringent selection con-ditions to detect rare mutants. Genetic analysisof bacteria was made possible by the demon-stration that strong selection could reliably de-tect and enumerate preexisting mutants with-out contributing to their formation. Classicalexperiments by Luria and Delbruck (1943), Le-derberg and Lederberg (1952), and Newcombe(1949) used lethal selections to show that mu-tants arise before selection and do not requiregrowth limitation for their formation. Theseselections could only detect preexisting mutantsand, therefore, could not show whether or notsome adaptive mutations form at a higher ratein response to selection. However, the classicexperiments were interpreted as evidence thatselection does not affect mutation rates and ledto the ultimate conclusion that all mutants ariseas replication errors that are random with re-spect to functional or selective consequences.

Later experiments seemed to contradict thisconclusion. In several systems, microbial growthwas blocked by nonlethal conditions and mu-tant colonies accumulated over time above thelawn of plated cells (Shapiro 1984; Cairns et al.1988; Hall 1988; Steele and Jinks-Robertson1992; Maenhaut-Michel and Shapiro 1994;Sung and Yasbin 2002). The behavior of thesesystems was interpreted as evidence that non-growing cells increased their own mutation ratesto help them “get off the dime.” The cost ofmutagenesis (deleterious mutations) was notevaluated and the possibility that cells grow un-der selection before mutation was not eliminat-ed. If cells grow under selection, any observedmutants could result from unappreciated parent

cell divisions or from proliferation of preexistingsmall-effect mutants rather than an increasedmutation rate. When selection is imposed on agrowing population, the effect of mutagenesison mutant frequency is hard to quantify in theface of the enormous exponential effects of pos-itive selection.

To avoid interference from selection, muta-tion rates are classically measured by focusingattention on mutant types that can grow wellunder some permissive growth conditions andcan be detected later by stringent selection (re-viewed by Rosche and Foster 2006). Genome-wide mutation rates have also been estimated byexamining the whole genome sequence of pop-ulations that have been passed through repeatedsmall bottlenecks to minimize accumulationof fitter mutants and loss of mutants that im-pair growth (Lee et al. 2012). These devices solvethe rate measurement problem, but avoid thereal issue—“do adaptive (beneficial) mutationsoccur at a higher rate under selection”?

The Cairns–Foster system (Cairns et al.1988; Cairns and Foster 1991) was designed tosolve these problems. The goal was to assess theformation rate of adaptive or beneficial muta-tions under selective conditions that preventgrowth, but are not lethal. In interpreting theCairns–Foster experiment, it is assumed thatif cells are not dividing and chromosomes arenot replicating then any increase in mutant fre-quency must be caused by mutagenesis andnot selection. To eliminate growth, conditionswere carefully adjusted (using scavenger cells, asdescribed below). Selection was deliberately setat a stringency that is just enough to preventgrowth (Cairns and Foster 1991). The parentalcells do not grow, but are poised such that veryminor improvements may allow growth to start.The system has several problems. There is noeasy way to compare mutation rates in growingand nongrowing populations. During growth,one measures mutations/cell per generation.Without growth, one measures mutations/cellper unit time. It is hard to completely eliminatecryptic growth of a subset of the plated popu-lation or within slowly developing, partially re-vertant colonies. Selection never sleeps. We willsuggest later that the problem is even worse—

S. Maisnier-Patin and J.R. Roth

2 Cite this article as Cold Spring Harb Perspect Biol 2015;7:a018176



selection may even operate even when cell divi-sion is blocked.

PHENOMENOLOGY OF THE CAIRNS–FOSTER SYSTEM—HOW THE EXPERIMENTIS DONE

In the basic Escherichia coli tester strain, the lacoperon is located on a low-copy F0lac plasmid(see Fig. 1) (Cairns and Foster 1991). The lacZgene (b-galactosidase) is fused to the lacI (re-pressor) gene and transcribed from a constitu-tive (iQ) promoter. Expression of LacZ activity isprevented by a leaky þ1 frameshift mutationwithin the lacI portion of the hybrid gene, whichreduces the level of b-galactosidase (LacZ ac-tivity) to 2% of that seen in revertants (Foster1994). Cells are pregrown in minimal glycerolmedium, washed, and plated (108) on lactosemedium, in which any residual growth is pre-vented by a 10-fold excess of scavenger cells witha lac deletion mutation. Scavengers cannot uselactose or revert, but can consume any residualnutrients in the agar or compounds excretedfrom revertant colonies (notably galactose).The number of scavenger cells is set to just barelyprevent growth of the tester, which is poised onthe brink of growth (Cairns and Foster 1991).

The course of an experiment is described inFigures 2 and 3. Tester cells (108) are grown non-

selectively, washed, and plated at day 0. The Lacþ

revertant colonies start appearing by day 2 andaccumulate linearly for �5–6 days thereafter.During this time, the plated lawn shows littlegrowth and this stasis is not a balance of deathand growth (Foster 1994). The colonies appear-ing first—before day 2—are initiated by preex-isting full lacþ revertants (with a compensating21 frameshift mutation). Colonies that accu-mulate later, above a nongrowing lawn, are thecritical revertants to explain. The yield of theselate revertants depends on residual LacZ expres-sion from the leaky mutant gene. This residualexpression could, in principle, support very slowlawngrowth orcouldsimplysupplyenergyneed-ed for the reversion process, a point that willbecome critical. There are two types of late rever-tant colonies. Ninety percent of colonies appear-ing on day 5 contain cells with a stable lacþmu-tation (a compensating 21 frameshift). Theremaining colonies (10%) contain cells with atandem amplification (10–100 copies) of theoriginal mutant lac allele. Most of these are stan-dard direct-order repeats with short (10 bp)junction sequences (or short junction [SJ]types). About 20% of unstables have tandem re-peats arranged in alternating orientation (tan-dem inversion duplication [TID]). Stable typesaccumulate linearly, unstable types exponential-ly (Fig. 3).

Tester strain+1

lacl

lacldinB+

lacF′lac128

oriT

lacZ

lacZ

Deletion

lacY

lacY

lacA

lacA

Parent strain

Figure 1. The strains used in the Cairns–Foster system. The Escherichia coli tester strain carries an F0lac plasmidwith a mutant lac allele; the chromosomal lac region has been deleted. On the F0lac plasmid, the lacI and lacZgenes have been fused so they encode a single protein with b-galactosidase (LacZ) activity. This hybrid gene istranscribed from a constitutively expressed lacIQ promoter. The ability of the tester to grow on lactose is blockedby a þ1 frameshift mutation within the lacI portion of the hybrid gene. This mutation is leaky and allowsproduction of about 2% of the revertant enzyme activity (Cairns and Foster 1991). The F0lac plasmid also carriesthe dinB gene, which encodes an SOS-inducible error-prone (mutagenic) DNA repair polymerase. The transfer(tra) functions of plasmid F0lac128 that allow conjugation are constitutively expressed. Conjugal replication andtransfer processes are initiated at a transfer origin (oriT) by the single-strand endonuclease TraI.

Origin of Mutations Under Selection

Cite this article as Cold Spring Harb Perspect Biol 2015;7:a018176 3



NUMEROLOGY OF THE CAIRNS–FOSTERSYSTEM—THE MUTATION PROBLEMTO BE SOLVED

The curious thing about the Cairns–Foster sys-tem is the abundance of revertant colonies and

the paucity of evidence for any general muta-genesis that might cause their appearance. Thelac ! lacþ reversion rate is 1028/cell/divisionduring nonselective growth in liquid medium(Rosche and Foster 2006). A lawn of 108 platedcells produces 100 colonies over 5 to 6 days

Plate 108Added revertant colonies accumulate

under selection

Day 5Day 4Day 2

Initiated by preexistingrevertant cells

NB X-gal

Stable Lac+ Unstable Lac+

10%90%

Days of incubation on lactose

Lac– mutant(glycerol)

cells

Lactose

xx

Figure 2. The Cairns–Foster reversion experiment. Cells of the strain described above are pregrown on glycerol,washed, and plated on lactose medium. The few revertant colonies that appear on lactose within 2 days areinitiated by fully lacþ revertant cells that arose during prior nonselective pregrowth. More Lacþ coloniesaccumulate on the plate over the next 4 days. On day 5, 90% of new late colonies are made up of stable lacþ

cells that have acquired a compensating frameshift mutation and, therefore, a functional lacþallele. The other10% of revertant colonies are made up of cells with an amplified array of the original mutant lac allele.Amplification-bearing cells are unstable and frequently lose their Lacþ phenotype. When these revertants arestreaked on rich, nonselective medium containing the chromogenic LacZ substrate X-gal, they form sectored(blue/white) colonies that reveal their frequent loss of ability to use lactose (see Fig. 3).

100

10

Unstable

Lac+

colo

ny n

umbe

r

2 4 6Days under selection Inverse

(TID)

Direct(SJ)

Unstable Lac+ revertant types n = 10–100

n

Stable Lac+ revertant

LacZ– parentfs

lacI-lacZ

+

Stable( I X

( I X l )

I )

( I I )

( I X l )( I X l )( I X l )

n( I X l )( I X l ) ( I X l )( I X l )

Figure 3. The time course of accumulating stable and unstable revertants in the Cairns–Foster experiment. Afterthe tester strain is plated on lactose medium (on day 0), revertant colonies accumulate over several days. Stablelacþ colonies accumulate linearly with time, whereas unstable Lacþ colonies accumulate exponentially withtime. The genotype of the parent tester strain is diagrammed at the top right. Stable Lacþ colonies have acquired acompensating (21) frameshift mutation and thereby a lacþ allele. Cells in unstable Lacþ revertant colonies havemultiple copies (n) of the original leaky mutant lac allele, arranged either as tandem direct-order repeats (shortjunction [SJ]) or expanded tandem inversion duplications (TIDs) with copies in an alternating orientation.Cells with sufficient copies of the partially functional mutant lac allele can grow on lactose and form a sectoredblue/white colony on rich X-gal medium, owing to loss of lac copies.





(about 100-fold more than expected) and doesso with no apparent lawn growth and very littlegeneral mutagenesis. If every plated cell dividedonce, the number of revertants could be ex-plained by a 100-fold increase in general muta-tion rate. However, the lawn of parent cells showsno increase in the frequency of unselected mu-tations (Torkelson et al. 1997; Rosche and Foster1999; Slechta et al. 2002). The Lacþ revertantcolonies showa 20-fold increase in the frequencyof unselected associated mutations, which is un-evenly distributed. Ninety percent of Lacþ rever-tants show no evidence of mutagenesis, whereas10% have experienced about a 100-fold increasein genome-wide mutation frequency (Roscheand Foster 1999). It seems clear that some muta-genesis occurs but is insufficient to explain therevertants, and is distributed in interesting ways.

THREE SUGGESTIONS TO EXPLAIN THEUNEVEN DISTRIBUTION OF MUTAGENESIS

Several ideas have been proposed to explain howLacþ revertant number can increase 100-foldwith very little general genome-wide mutagen-esis. The first two ideas below have been used,individually or together, as parts of models thatexplain the Lacþmutants by stress-induced mu-tagenesis. The third idea (our favorite) suggeststhat stress does not induce mutagenesis at all.It attributes the modest observed increases inmutation rate to a nonessential artifactual sideeffect of growth under selection in this system.

Directed Mutagenesis—Stress InducesPredominantly Beneficial Mutations in AllCells of the Population

This model proposes a regulatory mechanismthat senses the physiological problem and directsmutagenesis to sites that improve growth (Fosterand Cairns 1992; Foster 1993). Direction to use-ful sites explains how the number of Lacþ rever-tants might increase more than the number ofmutants at large. That is, the mechanism makespredominantly valuable changes. Although amechanism to direct mutagenesis may seem dif-ficult to imagine, clever ways of achieving thishave been suggested (Stahl 1988). Those include

an increase in transcription errors, or a failure inthe postreplicative mismatch correction system.These possibilities were later rejected (Foster andCairns 1992; Stahl 1992). One important earlyobservation was that the recombinase RecA andsome DNA synthesis are required for appearanceof Lacþ revertants under selection (Cairns andFoster 1991; Foster and Cairns 1992). The selec-tive-amplification model described below relieson RecA and has the effect of directing mutagen-esis to the precise positions that limit growth(Roth et al. 1996).

Later experiments showed that mutagenesisis not directed in the strictest sense. Starvationof lac mutants on lactose stimulates selectivereversion of mutations in other genes on the F0

plasmid (þ1 frameshifts in tetA), not just in lacgenes (Foster 1997). During starvation on lac-tose, tetracycline-resistant mutants (TetAþ) ac-cumulated at the same linear rate as did Lacþ

revertants. This was evidence that limitation ofgrowth on lactose does not direct mutationsspecifically to the lac operon, but also affectsother sites close to lac on F0lac. The formationrate of both apparently directed mutation typescan be enhanced by DinB if the dinBþ gene isalso present on the plasmid that carries tetA andlac (see model below).

Hypermutable States—Stress InducesGenome-Wide Mutagenesis in a Subset(1023) of the Plated Population

This model was suggested by Barry Hall (1990)and was supported by the discovery that Lacþ

revertants have an increased probability of car-rying associated unselected mutations (Torkel-son et al. 1997; Rosche and Foster 1999; Slechtaet al. 2002). The Lacþ revertant cells with asso-ciated mutations do not arise in cells with aheritable mutator mutation (Rosche and Foster1999). The frequency of unselected mutants inthe starved population is low, and the small frac-tion of those cells that were mutagenized hasbeen estimated at about 1 cell in a 1000 (Roscheand Foster 1999; Bull et al. 2000). If the accu-mulated 100 Lacþ revertants were initiated bycells that arose in the mutagenized subset ofcells, they would have to arise from 105 instead





108 plated cells. Generating 100 revertants from105 cells (a rate of 1023) would require a 105-foldincrease in mutation rate over that measuredin unselected cells. When inflicted on nongrow-ing cells, this intense mutagenesis is higherthan can be achieved by any chemical mutagenand is expected to add about eight null muta-tions to each Lacþ genome (Roth et al. 2003).Thus, general mutagenesis alone cannot explainthe revertants appearing in the Cairns–Fostersystem, but might contribute if combined withthe directed mutagenesis described above.

Selective Improvement of PreexistingSmall-Effect Mutants—Stress Is NotMutagenic—Only an Agent of Selection

This model proposes that cells with multiplecopies of the leaky mutant lac allele arise beforeselection and initiate the colonies that appearlater on lactose plates. Because of the residualactivity of the mutant lac allele, multiple copiesof the mutant gene might provide sufficient en-ergy to support growth or at least allow repeatedplasmid replication under selection. Repeatedreplication of multiple lac target sites might en-hance the likelihood of an improving mutationusing only the basal mutation rate. This lac am-plification may occur in slowly growing cells orin nondividing cells that continue to replicatetheir F0lac plasmid under selection.

The enhanced yield of Lacþ revertants underselection and the associated mutagenesis seenin 10% of Lac revertants is attributed in partto activity of the error-prone DinB polymerase.This polymerase is encoded by a gene locatednear lac on the F0lac plasmid, and, therefore, issubject to increased expression in strains withmultiple F0lac copies or tandem duplicationslarge enough to include both the lac and dinBgenes.

THE PHILOSOPHICAL CONUNDRUMPOSED BY THE CAIRNS–FOSTER SYSTEM

Work on the Cairns–Foster system has generat-ed a body of high-quality data. Parties to thedebate accept most of these results. A few re-maining conflicts will be discussed below. De-

spite their general acceptance, these results havebeen interpreted in diametrically opposite ways.One side concludes that cells have mechanismsto create mutations whenever growth is blocked(“adaptive mutation,” “stress-induced muta-genesis,” “directed mutation,” “stationary phasemutagenesis,” or “hypermutable states”). Theother side (that is, us) interprets the same resultsas evidence that growth limitation does notchange the mutation rate, but serves as an agentof selection that favors growth and improve-ment of preexisting partial revertants. The fail-ure to resolve this issue may result from a ten-dency of both sides to test predictions thatverify their point of view rather than doing ex-periments that decide between the two generalpossibilities. Experiments discussed below mayprove decisive.

FACTORS THAT AFFECT THE YIELDOF STABLE AND UNSTABLE REVERTANTS—THE MAJOR FACTS

The accumulated body of information has beenreviewed repeatedly, both from the viewpoint ofstress-induced mutagenesis (Foster 1993, 1999,2005, 2007; Rosenberg 2001; Hersh et al. 2004;Galhardo et al. 2007; Rosenberg et al. 2012) andfrom the position of selection alone (Roth andAndersson 2004; Roth et al. 2006; Anderssonet al. 2011). Below, we describe the most strikingfindings. These results are agreed on but inter-preted differently. Later, we will combine theseresults, resolve some data conflicts, and offer acomprehensive view of reversion in the Cairns–Foster system that has broader implications forevolutionary processes.

F0 Plasmid Transfer Replication

The yield of Lacþ revertants is essentially elimi-nated by mutations that inactivate the ability ofthe F0lac plasmid to transfer conjugatively to arecipient cell (Foster and Trimarchi 1995a; Ga-litski and Roth 1995; Radicella et al. 1995; Peterset al. 1996). This is most striking for strains lack-ing TraI—a plasmid protein with DNA nickaseand helicase activities, both of which contributeto transfer replication and to reversion (Traxler





and Minkley 1988; Foster and Trimarchi 1995a).During the reversion process, a Tn10 inserted onthe F0 plasmid is subject to frequent loss, which isknown be stimulated by single-strandedness(Syvanen et al. 1986). Because conjugal replica-tion origin produces a single-strand product,frequent Tn10 loss suggested that reversion in-volves plasmid transfer or transfer-associatedreplication from the oriT transfer origin (Godoyand Fox 2000). However, transfer of the wholeplasmid is seldom associated with reversion(Foster and Trimarchi 1995a,b; S Maisnier-Patinand JR Roth, unpubl.). These results do noteliminate the possibility that TraI initiates inter-nal F0lac replication from the transfer origin(oriT) or contributes to transfer within a singlecell or among daughter cells. The F-plasmidconjugation functions are needed for both stableand unstable Lacþ revertants (Ponder et al.2005). Consistent with the above points, the ef-fect of growth limitation on lac reversion de-pends on the leaky lac allele being located onthe F0 plasmid (Slechta et al. 2003).

Recombination Proficiency

Lack of the RecA (DNA strand exchange) pro-tein or the RecBC/RuvABC proteins (double-strand break repair) reduces revertant numberseverely (Cairns and Foster 1991; Harris et al.1994). Other recombination defects cause agenerally smaller reduction (He et al. 2006).The several schools of thought suggest that re-combination either plays a role in initiating mu-tagenic DNA replication or contributes to theprocess of gene amplification. Ability to recom-bine, like ability to transfer, is required for bothstable and unstable Lacþ revertant colonies.

The DinB Error-Prone DNA RepairPolymerase

This polymerase (also called Pol IV) belongs tothe Y-family of translesion DNA polymerasesand is part of the SOS response to DNA damage(Friedberg et al. 2005). DinB can efficiently rep-licate a template containing damaged bases thatblock normal replication, especially bulky le-sions formed at the N2-position of deoxygua-

nosine (Jarosz et al. 2007; Walsh et al. 2011).DinB can also elongate misaligned primersand incorporate modified nucleotides (Nohmi2006). Overexpression of DinB leads to sponta-neous mutagenesis of undamaged DNA. In par-ticular, a high level of DinB makes 21 mutationsby template slippage in repetitive sequences(Kim et al. 2001; Nohmi 2006)—the type of se-quence found at the site of the lac mutation inthe Cairns system.

In the absence of DinB, the number of stableLacþ revertants drops about 10-fold but theyield of unstable revertants is unaffected (Mc-Kenzie et al. 2001). Without DinB, stable andunstable revertants appear in equal numbers (a1:1 instead of a 9:1 ratio) and total revertantyield is reduced about fivefold (see Fig. 4).Thus, DinB contributes substantially, but isnot absolutely required for the accumulationof revertants under selection.

Duplication and Amplification of the lacOperon

Gene duplications form at a high rate and cometo a high steady-state frequency during nonse-lective growth (Reams et al. 2010). This steadystate is maintained by the balance between highformation rates on the one hand and high lossrates plus fitness cost on the other. Cells withhigher amplification are expected to come tolower steady-state frequencies. Cells whose F0laccarries a large duplication (100 kb) of the plas-mid lac region are carried at a high steady-statefrequency of 1/500. These common lac duplica-tions form on F0lac between flanking copies ofIS3 (1.2 kb). Unstable Lacþ revertant coloniesconsist of cells with an amplification of a smallerlac region (20–30 kb). The duplication under-lying these amplifications often form among re-petitive extragenic palindromic (REP) elements(Bachellier et al. 1999; Kofoid et al. 2003; Slacket al. 2006; Kugelberg et al. 2010). The cells inunstable Lacþ revertant colonies grow under se-lection because they have 10–100 copies of asmall region (lower fitness cost) that includesthe leaky mutant lac allele.

The amplifications found in unstable Lacþ-

revertants are of two types. Some have tandem





direct-order repeats with a short junction (SJ)sequence and appear to arise by an exchangeamong very short repeats (10 bp). Others havean array of tandem copies in alternating ori-entation (head-to-head and tail-to-tail) (Ku-gelberg et al. 2006, 2010; Slack et al. 2006).These are known as tandem inversion dupli-cations (TIDs) (see Fig. 3). Both duplicationtypes allow slow growth on lactose and expandinto higher amplifications by unequal recom-bination among the directly repeated sequences(Kugelberg et al. 2006, 2010). Copy-numberexpansion is expected whenever the growthimprovement provided by additional lac cop-ies exceeds the fitness cost of the amplifi-cation.

TWO DIAMETRICALLY CONTRADICTORYMODELS—AND HOW THEY GREW

The two general models described below use thesame basic facts outlined above, but explain re-version in the Cairns–Foster system in differentways. In the first, colonies are initiated on theselection plate when an evolved stress-inducedcellular mechanism causes a discontinuousevent in a nongrowing cell. This event eithercreates a new large-effect frameshift mutation inthe lacI-lacZ gene or a sudden amplification ofmutant lac copy number. In the second model,colonies are initiated by preexisting mutantswith a lac duplication. These cells grow slowlyat first and improve under selection. We will

100

50

10

2 4 6Days under selection

Lac+

colo

ny n

umbe

rNo

DinB

WithDinB

TotalLac+

revertants

100

50

10


Lac+

colo

ny n

umbe

r

No Tral

TotalLac+

revertants

100

50

10


Lac+

colo

ny n

umbe

r

No RecA

TotalLac+

revertants

100

50

10


Lac+

colo

ny n

umbe

r

Unstable

With DinB

Stable

100

50

10


Lac+

colo

ny n

umbe

r

Unstable

Stable

Without DinB

Figure 4. Major contributors to revertant yield. A considerable body of data has been accumulated to test theeffect of various chromosome- and plasmid-encoded functions on revertant yield. Top left shows dependence ofrevertant yield on the error-prone repair polymerase DinB (4- to 5-fold). Top center shows that, in a DinBþ

strain, stable revertants are about 90% of total and unstable revertants are 10%. Top right shows that, withoutDinB, the number of stable revertants is reduced about 10-fold and becomes roughly equal to the number ofunstable revertants (Foster 2000; McKenzie et al. 2001). That is, even without the mutagenic DinB polymerase,the number of stable revertants increases with time under selection. Two of the most important functions forreversion are TraI (bottom left), which nicks and unwinds plasmid DNA during conjugative transfer, and RecA, astrand-exchange protein that is central to homologous recombination. Mutants in recBC and ruvABC genes havesimilarly strong effects on revertant yield.





first describe one simple version of these twoopposite models and then we will considersome data conflicts and present a unified modelthat we think accounts for all of the availableinformation.

Stress-Induced Mutagenesis

In this model, none of the plated tester cells growunder selection but all remain alive and respondto starvation by inducing a set of genes that leadto DinB-dependent mutagenesis. Increased ex-pression of DinB is mediated by LexA, repressorof the SOS DNA repair system, and by RpoS,activator of genes during slow growth (Laytonand Foster 2003; Lombardo et al. 2004; Gal-hardo et al. 2009). Although the tester cells arenot dividing, the leakiness of the lac allele isthought to give enough energy for occasionalfiring of the vegetative plasmid origin on F0

and DNA synthesis. If a replication fork encoun-ters a nick produced at the conjugational repli-cation origin oriT by the endonuclease TraI, thefork collapses creating a double-strand break.Normally, such ends are processed in an error-free manner by RecA/BC-dependent recombi-nation. However, during growth limitation, themodel proposes that repair becomes mutagenicbecause RpoS and DinB proteins are up-regu-lated. Both of these proteins are crucial to theformation of the point mutations (21 frame-shifts) that produce a lacþ allele and a stablerevertant colony. In other tester cells, starvationcauses a sudden lac amplification, and the mul-tiple mutant lac alleles allow growth and forma-tion of an unstable revertant colony. In thismodel, both stable and unstable revertant typesare independently initiated by discrete eventsthat are induced under selection by a stress-in-duced mechanism. These events all occur innongrowing cells.

Amplification Under Selection

In the initial form of this model, the reversionprocess starts with a duplication-bearing cellformed during nonselective pregrowth beforeplating. Such cells are extremely common (1/500) and grow very slowly under selection. Theirgrowth rate increases progressively by expansion

of the tandem array of lac alleles. Ultimately, themutant lac allele is replicated a sufficient num-ber of times that one copy acquires a reversionevent to lacþ. This can occur at the standardunenhanced mutation rate. As soon as a frame-shift mutation generates a revertant lacþ allele,growth accelerates and a stable revertant colonyforms. If a deletion removes a junction of theinitial duplication, the repeat size is reduced.This lowers the fitness cost of amplification, al-lowing higher amplification and formation ofan unstable revertant colony.

This model requires no mutagenesis, butdoes benefit from multiple copies of the dinBþ

gene. Mutation rates do increase in colonieswhose lac amplification includes the nearbydinB gene, which happens to lie only 16 kbaway from lac on the F0lac plasmid. This modelwas developed based on experiments with theSalmonella version of this system in which thelac mutation supports higher lawn growth. Themodel seems to explain the behavior of Salmo-nella, but does not account for behavior of theE. coli system in which much less growth occursunder selection.

FOUR DATA CONFLICTS THAT HAVEDELAYED A FINAL DECISION

Below are four questions whose answers shouldhelp decide between “stress-induced mutagen-esis” and “amplification under selection.” Wethink that these answers are near.

Are Revertants Initiated before or afterSelection?

A decisive question regarding these models iswhether revertant colonies are initiated by cellsthat arise on the selection plate, possibly in-duced by stress. Alternatively, colonies could beinitiated by weak Lacþ cells that form beforeplating as proposed by the selection models.One of the strongest arguments in favor of“stress-induced mutagenesis” is the outcomeof a Luria–Delbruck fluctuation test applied toreversion in the Cairns–Foster system (Cairnsand Foster 1991). Multiple parallel cultureswere plated independently on selective medium,





and day 5 revertant yields were found to show aPoisson distribution. The absence of “jackpots”characteristic of a Luria–Delbruck distributionwas interpreted as evidence that Lacþ coloniescannot be initiated before selection but mustarise on the selection plate. This result is op-posite to that of classical experiments showingthat mutants detected by laboratory selectionsarise before selection (Luria and Delbruck 1943;Newcombe 1949; Lederberg and Lederberg1952). The results were taken as evidence thatthe Lacþ colonies arising in the Cairns–Fostersystem were initiated on the plate by mutationsformed in response to selection or “stress.”

Since the time those fluctuation tests weredone, it has become clear that duplication mu-tations are immune to Luria–Delbruck fluctu-ation tests (Reams et al. 2010). The frequency ofduplications and higher amplifications in unse-lected cultures comes to a steady-state frequencythat is maintained by a balance between the highrate of gene duplication formation (1025/cell/generation) on the one hand, and the even high-er loss rate (typically 1022/cell/generation) plusthe fitness cost of the duplication on the otherhand. The forces that drive the frequency ofcells with a gene duplication (or amplification)toward steady state act on any aberrant frequen-cy caused by timing of formation events. Thisminimizes the occurrence of jackpots. Becausethese steady states obscure fluctuation, a differ-ent test was required to determine whether cellswith preexisting duplications initiate revertantcolonies that appear under selection.

Recent evidence suggests that revertant col-onies are initiated by cells that form before se-lection and thus cannot be stress induced (Sanoet al. 2014). This was determined by placing thetetA gene near lac on the F0lac plasmid. Multiplecopies of tetA are toxic when induced by theanalogue anhydrotetracycline (AnTc). If cellswith multiple copies of lac initiate revertants,many of those cells should also have multiplecopies of the nearby tetA gene and be sensi-tive to inhibition by AnTc. Reducing the fre-quency of high-copy variant cells (loweringtheir steady-state frequency) should reduce thenumber of revertant colonies seen later underselection. When cultures of the tester strain with

tetA on the F0lac plasmid were pregrown in thepresence of AnTc, the number of Lacþ rever-tants was reduced sharply. The revertant num-ber is restored by a few additional generationsof growth in the absence of AnTc. Unexpectedly,the reduction in revertant yield was seen re-gardless of the position of tetA on the F0lac plas-mid. This suggested that the critical cells havemultiple copies of the whole F0lac plasmid, rath-er than an amplification of a small lac region.This finding is parallel with the “amplificationunder selection” model, except that amplifica-tion occurs at the level of plasmid copy numberinstead of by tandem repeats of only the lacregion. Cells with multiple copies of the F0lacplasmid initiate both the stable and unstableLacþ revertants. If all revertants are initiatedbefore selection, we conclude they cannot be“stress-induced.”

Must dinB Be Located on the F0lac Plasmid?

In the original “amplification under selection”model, expression of DinB is increased by selec-tive coamplification of the dinB and lac genes.This requires that the two genes be located closetogether, as they happen to be in the originaltester strain. In contrast, “mutagenesis” mod-els” propose that stress increases the level ofDinB protein by transcriptional gene regulatorymechanisms (e.g., RpoS, LexA) and gene posi-tion should be irrelevant. We found a clear effectof dinB gene position on revertant yield—theLacþ revertant number was severely decreasedby removing the dinBþ allele from its positioncis to lac on the F0lac plasmid (Slechta et al.2003). Functionality of the chromosomal dinBallele was not required. The conflict arose whenthe same experiment (performed in anotherlaboratory) showed no effect of position(McKenzie et al. 2001). We have reconstructedall of the critical strains and still get our originalresult—a full yield of stable revertants requiresa functional dinBþ allele located on the sameplasmid as lac. Resolving this “he said, she said”stand-off will require more testing, but we areconvinced that dinB gene position is critical andobserved mutagenesis requires coamplificationof dinB with lac.





Is There Growth before Mutation UnderSelection?

Mutagenesis models depend on mutants arisingin nongrowing cells. In contrast, our originalamplification–selection model relies on growthunder selection. Most other systems used tosupport stress-induced mutagenesis have noteliminated growth under selection, making ithard to decide what is happening (Mittelman2013). In the Cairns–Foster system, supportersof mutagenesis have taken pains to show thatcells are not growing (see below). Supportersof selective amplification argue that these testshave missed important growth within develop-ing colonies, which is systematically missed byassays of the plate population. The question ofgrowth could be decisive.

A considerable body of evidence supportsthe belief that mutations arise in nongrowingcells. This evidence must be accounted for inany final model. Notice that the preponderanceof evidence below argues against growth:

1. The total number of parental cells in the lawn(outside of colonies) does not increase for 4days. The constant lawn population is not abalance between growth and death (Foster1994). However, a large population (108) isbeing tested so one doubling would be hardto detect and yet would provide 1028 oppor-tunities for mutation. In addition, these testsdo not assess growth within the developingcolony.

2. If growth occurs within colonies, as suggest-ed by the original amplification model, thenthe number of visible colonies is expected toincrease exponentially because an exponen-tially increasing number of precursor cells ispresent in each colony. The graphs describ-ing revertant accumulation show an upwardflex in some experiments but, in general, thisaccumulation is linear with time, consistentwith very little general population growthover 4 days. This argues against growth with-in the developing colony before the reversionevent.

3. Newcomb spreading experiments did notshow evidence that lacþ revertants arise

from cells growing under selection beforemutation (Newcombe 1949). This classicexperiment is a contemporary of the Luria–Delbruck fluctuation test and the Leder-bergs’ replica-printing experiment, whichoriginally showed that mutations are notcaused by the selection. Newcombe platedbacteria cells on nonselective medium andallowed time for a few divisions. He thenspread the population of the plate beforeimposing a selection (phage resistance). Ifresistant mutants were present in the platedpopulation and grew before selection (whichthey did), the act of spreading multiple cellswithin a developing colony should seedmultiple colonies that could be detected afterselection was imposed. Spreading should notincrease selected mutant number if muta-tions are induced by exposure to selection.

If colonies are initiated by rare preexistingcells that grow under selection before reversion,then respreading the lawn shortly after platingshould increase the number of revertants de-tected after several days under selection. Thishappened in Newcomb’s experiment, indicatingthat the mutant clones arose and grew beforeselection. In the Cairns–Foster experiment, thisexperiment would test whether preexistingslow-growing cells grow before acquiring a fulllacþ reversion mutation, as suggested by theamplification model. This experiment is diffi-cult because spreading must be performed aftersome growth has occurred but before any devel-oping clone has acquired a full lacþ allele. It isalso crucial to avoid preexisting Lacþ cells,which will certainly grow and produce manycolonies after spreading. The experiment wastried by Foster (1994) and by us (S Quinones-Soto and Y Toofan, unpubl.). Both attemptsfailed to provide evidence of preexisting cellsthat grow before reversion.

Are Amplifications Remodeled UnderSelection?

The original “amplification–selection” modelproposed that revertants were initiated by cellswith a duplication or amplification of the large





100-kb lac region among IS3 copies. These cellswere suggested to grow and amplify a bit underselection and then acquire either a point muta-tion to lacþ (aided by their multiple copies ofDinB) or a join point deletion. The join pointdeletion reduces the size of the repeated unit andthereby the fitness cost so as to allow high am-plification under selection. Without growth,this remodeling process seems unlikely.

Since then, we have found evidence that thecommon large IS3-mediated duplications makeno contribution to reversion and the short-er duplications (see Fig. 3) that underlie theamplifications in unstable revertants actuallyarise during nonselective pregrowth (Kugelberget al. 2010). These short duplications are suffi-ciently frequent in the plated population to ex-plain all of the unstable Lacþ revertants (Reamset al. 2012). If SJ and TID duplications arise byremodeling of a preexisting duplication, thismust all occur during growth before selection.Because unstable revertants are initiated by cellsthat arise before plating, they cannot be “stress-induced” and are essentially already remodeledbefore selection. We will propose below that theprecursors of unstable revertants grow immedi-ately after plating and improve under selectionby further expansion of their amplified array. Asexpected for exponentially growing clones thatbecome visible with time, the number of unsta-ble revertants increases exponentially over a se-ries of days (Fig. 4). It seems clear that growth iscentral to development of these colonies andthey are initiated by cells that arise before selec-tion. The open question of growth involves pri-marily the early history of stable Lacþ revertants.

THE DEEPER MEANING OF GROWTHAND MUTAGENESIS

The question of growth under selection is crit-ical because growth and mutagenesis are essen-tially trade-offs in trying to account for the 100revertants that arise in this system. If growth istruly eliminated by selection, as seems likely,mutagenesis of some kind seems essential toexplain the observed mutations. If growth oc-curs, then the exponential expansion may pro-vide sufficient acts of replication to produce the

revertants using the standard unenhanced basalmutation rate. We have resisted the idea of mu-tagenesis because the idea of an evolved mech-anism to increase the genome-wide mutationrate temporarily seemed numerically unsani-tary. Mathematical modeling of the “hypermu-table state” model suggested that temporary up-regulation of genome-wide mutagation rate in asubset of the population is unlikely to be usefullong term because of the high cost of deleteriousmutations and the persistence of deleteriousmutations after any episode of hypermutability(Roth et al. 2003). This modeling does not elim-inate the mechanistically problematic possibil-ity of “directed mutation.”

Our persistent confidence that growth mustoccur but somehow escape the tests listed abovewas based on the finding that stable Lacþ rever-tant colonies include rare cells that are unstableLacþ (Hendrickson et al. 2002). The growth andamplification model predicts that stable rever-tant colonies should contain a few unstable pre-cursor cells with an amplified mutant lac allele.Such ancestral cells were identified by their for-mation of sectored (blue/white) colonies whenstreaked on rich X-gal medium and were takenas evidence for amplification and growth with-in the developing colony before reversion. Theseunstable Lacþ cells represented about one cellin 1000 stable Lacþ cells (Hendrickson et al.2002). These sectored colonies are visually in-distinguishable from the extensively character-ized cells in unstable Lacþ colonies, which carryan extensive lac amplification. However, Has-tings et al. (2004) reported that the rare unsta-ble cells within otherwise stable Lacþ coloniesshowed an unstable Lacþ phenotype that is notheritable. We have confirmed their observationand found that the rare unstable cells do not havelac amplifications (which are heritable) butseem to have multiple copies of the F0lac plas-mid, some lacþ and some lac. This is an extreme-ly unstable (poorly heritable) situation attribut-ed to the mechanisms that control plasmid copynumber and distribution. This result suggestedthat lacþ revertant alleles might not arise in cellsgrowing with amplification, but rather in cellswith multiple copies of the whole F0lac plasmid.This is consistent with the evidence above that





revertants are initiated by cells with multipleplasmids copies. These results reopened thequestion of growth, because cells with multipleplasmid copies seem too unstable to form exten-sive clones under selection but might persistwhen division is blocked.

Without growth it is difficult to explain therevertant number without some mutagenesis.

A NEW MODEL RESOLVES THECONFLICTS

This model is consistent with the entire bodyof information accumulated by both sides ofthe Cairns debate and outlined above. In thismodel, reversion starts before selection (witha plasmid copy-number increase) and the crit-ical frameshift mutations occur under selectionwithout prior growth and with no programmedincrease in mutation rate:

1. The major feature of this model is that bothstable and unstable Lacþ revertants are ini-tiated by cells with an increased F0lac plas-mid copy number. These cells arise beforeselection so the revertant colonies they ini-tiate cannot be stress induced.

2. Although the initiator cells with multiplecopies of F0lac arise before selection, theirprecise structure is unclear. These cells seemto have evaded standard plasmid copy-number controls. These plasmids may havetransfer replication forks that have switchedto rolling circle replication and producedouble-stranded linear products that maybe degraded as they are made. Alternatively,actual transfer might be occurring as sug-gested in earlier studies (Galitski and Roth1995; Radicella et al. 1995; Peters et al. 1996;Godoy and Fox 2000). Transfer could injecta single-strand end that becomes doublestranded, recombines, and initiates a rollingcircle replication fork using the recipientplasmid as template. Additional replicationof lac in these nongrowing cells may resultfrom nicks introduced at oriT, which be-come double-strand breaks when hit by areplication fork, as suggested previously(Foster et al. 1996; Harris et al. 1996). Re-

gardless of the nature of these cells, they rep-licate lac repeatedly over the course of sever-al days using energy provided by the leakylac allele. Each act of lac replication providesan opportunity for reversion. The DNAsingle strands exposed during this processinduce an SOS response. The absence ofcell growth induces the RpoS regulon. Thesetwo features have been suggested previously(Foster 2007; Galhardo et al. 2007).

3. After plating, these initiator cells do notgrow exponentially, because they lack suffi-cient energy to fire their chromosomal rep-lication forks. They use the energy providedby multiple lac copies to repeatedly repli-cate and repair their plasmids. The plasmidreplication (from an unregulated origin)exploits available energy and allows contin-ued copying of the plasmid in nondividingcells. The failure of cells to divide leavesthem unable to easily reduce their plasmidcopy number until growth resumes.

4. Formation of stable revertants (21 frame-shifts) is enhanced about 10-fold by thepresence of a dinBþ gene anywhere on theF0lac plasmid. An increased level of DinBprotein results from multiple dinB geneson the several plasmid copies. Mutationsto lacþ form during overreplication of lacon the multiple plasmid copies. This repli-cation alone is sufficient for some rever-sion, but the revertant yield is enhancedabout 10-fold because DinB is amplifiedas part of the plasmid. Because cells arenot dividing, the main chromosome isnot replicating and is not subject to DinBmutagenesis.

5. The events described above do not reflect aprocess that is stress induced in a standardsense. Selective conditions prevent cell di-vision but allow repeated local overreplica-tion of the F0lac plasmid. This overreplica-tion is excessive only because cell divisionand chromosome replication are blocked.Extra plasmid copies do not segregate.

6. When any one of the many copies of themutant lac allele acquires a lacþ reversion





event (21 frameshift), energy is immedi-ately supplied, allowing resumption of celldivision and chromosome replication. Allplasmids with nonrevertant lac copies arelost by segregation as the cell divides andcopy-number controls come into play. Therevertant F0lacþ copy is selectively main-tained in cells with a normalized copy num-ber. This stable lacþ colony contains no cellswith a tandem lacþ amplification, but mayinclude a few cells that have not yet dilutedout all of their unreverted F0lac copies.These explain the rare cells in a stable Lacþ

colony that show nonheritable unstableLacþ phenotypes.

7. The unstable Lacþ revertants are also initi-ated by preexisting cells with multiple cop-ies of an F0lac plasmid, but the plasmidin these cells carry an internal lac duplica-tion (SJ or TID). Cells with multiple copiesof a duplication-bearing F0lac plasmid haveenough energy to fire their replication ori-gins and divide. Their plasmid copy num-ber can be maintained in part by continuedreplication (requiring Tra functions), butdrops progressively by unequal plasmid par-titioning during growth. As plasmid copynumber returns toward normal, the lac am-plification expands and progressively im-proves growth. These cells ultimately pro-duce a colony of unstable Lacþ colonies.

8. In principle, this model is identical to theinitial selected tandem amplification mod-el described above except that: (1) Plasmidcopy number increases before selection andis maintained without growth. Stable ele-vated plasmids copy numbers in nongrow-ing cells take the place of tandem copyamplification during growth under selec-tion. (2) Local overreplication of the wholeplasmid in nongrowing cells replaces cellgrowth to provide opportunities for rever-sion.

9. Preexisting cells with multiple copies of anF0lac plasmid may be equivalent to the rare“hypermutable” cells in mutagenesis mod-els. Under selection, cells in this state repli-cate their F plasmid and lac region in the

presence of high levels of DinB, but do notreplicate their chromosome, which is notgenerally mutagenized.

10. Mutagenesis in this model is “directed” intwo senses. Only the F0lac replicates in thepresence of elevated DinB levels. Followinga rare act of reversion, the population ofunreverted lac alleles that allowed this eventis lost and only the rare lacþ allele is retainedselectively. This has the effect of directingmutagenesis to the precise base pairs thatlimit ability to grow on lactose (Roth et al.1996). This point will be discussed morebelow.

RELEVANCE OF THE CAIRNS–FOSTERSYSTEM TO EVOLUTION UNDERSELECTION

Since the advent of the Cairns–Foster system, ithas become clear that copy-number variationsare the most common genetic changes in natu-ral populations and in populations of metazo-an somatic cells (reviewed by Reams and Roth2014). Selective geneamplification underlies pro-gression of many malignancies (Albertson 2006).Our original selection–amplification model forthe Cairns system involved growth and im-provement under selection. This first modelseems to explain the Salmonella version of theCairns–Foster system and has been experimen-tally verified for a variety of other biologicalsituations, including evolution of bacterial anti-biotic resistance (Nilsson et al. 2006; Sandegrenand Andersson 2009; Sun et al. 2009; Paulanderet al. 2010; Pranting and Andersson 2011)and adaptation of poxvirus to host defenses(Elde et al. 2012). Most generally, the selectiveamplification model offers an explanation forevolution of novel genes under selection (Berg-thorsson et al. 2007) and has been experi-mentally shown to produce a new gene within3000 generations of growth under selection(Nasvall et al. 2012). Thus, the Cairns–Fostersystem has suggested a broadly relevant biolog-ical process.

Despite these successes, it now appears thatour original model does not quite explain the





E. coli version of the Cairns–Foster system inwhich very little growth occurs. Resolution ofseveral conflicts led to the new model describedabove, which does explain the E. coli system.This new model may also provide answers toquestions with broad evolutionary relevance.

How Does Selected Amplification MimicMutagenesis?

Despite a considerable fitness cost, duplicationsand amplifications are held as stable polymor-phisms in unselected populations (Reams et al.2010). These steady-state frequencies are highbecause duplications and amplifications format a high rate. Steady-state frequencies increasewhen selection favors cells with multiple copiesof some near-neutral allele. Amplification ofsuch an allele can provide a selective value thatexceeds the cost of the duplication itself. Thisallows lineages with amplifications to expandand add copies of a growth-limiting allele tothe genome. These cells grow faster and replicatemore copies of the rate-limiting allele at eachdivision. Each act of replication provides an op-portunity for further mutations that improvethe functionality of the limiting allele. In the

Cairns–Foster system, the mutant lac allele hasa very low ability to support growth on lactoseand is, in a sense, a near-neutral allele that canprovide a beneficial phenotype when amplified.Thus, selection gives the appearance of muta-genesis that actually results from replicatingmore copies of the target allele (see Fig. 5).

How Can Selection Appear to Direct Mutationto Critical Sites?

Direction of mutation to selectively valuablesites is a side effect of amplification under se-lection. As soon as an improving mutation oc-curs in any one of the multiple lac copies, selec-tion holds the improved allele and allows lossof other copies of the same gene. Deleteriousalleles of the gene in question are selectivelylost from the lineage. The apparent directednessis caused by the dynamic control of gene copynumber—loss and reamplification occurs con-tinuously and affects genes carried in eithertandem arrays or on multicopy plasmids. Inessence, the multiple gene copies in an ampli-fied array (or multiple plasmid copies in a cell)constitute a population that is under selectionto improve. The probability that a final rever-

Growthwithout

selection

Growth with selected gene amplification

Extra gene copies added to genome

Extra nonmutant gene copies lostwhen selection is relaxed by new allele

Mutation made more likelyby selective amplification

Beneficial new allele selectively retained

Figure 5. How lac amplification enhances revertant yield without mutagenesis. Each act of lac replicationprovides an opportunity for a reversion event (frameshift). A cell with only one lac copy has one opportunityto revert for each cell division or plasmid replication (see left column). As the lac allele amplifies, each cell gainsadditional chances for a reversion event with no increase in mutation rate (right column). If the amplified arrayimproves growth on lactose, then the lineage on the right expands faster than that on the left (without anamplification) and this growth also adds to the likelihood of a reversion event. In effect, amplification directsmutations to the exact base pairs that limit growth, because once a revertant allele forms, selection holds onlythat allele, whereas the nonrevertants alleles are no longer selectively retained. The likelihood of an unselectedmutation near lac is the same with and without selection, because only the revertant allele is kept in the genome.





tant plasmid has an associated unselected mu-tation is the same as the probability of that as-sociated mutation arising in any single copy ofthe gene with no selection or amplification.Thus, amplification enhances the likelihoodthat some lacþ mutation will occur, but doesnot enhance the likelihood of an associated un-selected mutation, because selection fixes onlythe revertant allele. Mutations are, in effect, di-rected to the exact base pairs that provide theselected phenotype and are not made more like-ly at other sites.

Natural Selection without Reproduction

It is generally assumed that natural selection re-quires cell division because improvement (orpurification) is attributed to differential repro-ductive success of cells with different genotypes.In the Cairns–Foster system, the plated cellpopulation is subjected to selection that pre-vents growth ( just barely). The growth-limitinggene (on the F0lac plasmid) is repeatedly copieduntil an improving mutation occurs. A rever-tant allele triggers exponential growth and isselectively held, whereas nonrevertant allelesare lost by segregation as soon as cell divisionstarts and the population expands.

The increase in lac copy number occurs be-fore selection, by some loss of plasmid copy-number control. On lactose medium, cells can-not divide and have no way to reduce their copynumber. They can, however, replicate their plas-mids and use DNA repair mechanisms to healplasmid breaks introduced by the TraI proteinand/or recombine DNA fragments with recipi-ent plasmids during redundant conjugation.Thus, the plasmid is a subset of the genomethat is overreplicated in nondividing cells. Cellsare trapped in this state until a reversion eventoccurs.

Mutagenesis by overreplication may be rel-evant to progression of some malignancies (Al-bertson 2006). Somatic cells that grow very littlemay repeatedly repair regions of their genomesurrounding fragile sites, which are subject tofrequent breakage. In this situation, nongrow-ing cells are under strong selection to grow andmay repeatedly replicate specific regions of their

genome during break repair. This may be mim-icked in the Cairns–Foster system in whichbreaks induced at oriT stimulate mutagenesisof the F0lac plasmid. We submit that the phe-nomenon of break-induced mutagenesis (Sheeet al. 2011; Rosenberg et al. 2012) may reflectrepeated repair and local overreplication innongrowing cells and involve no increase in mu-tation rate.

CONCLUDING REMARKS

At first glance, the Cairns–Foster system ap-pears simple and seems to show stress-inducedmutagenesis of nongrowing cells. This conflictswith the classical conclusion that selection andmutation are independent. Deeper inspectionreveals that the behavior of this system doesnot require mutagenesis but relies on severalaspects of selection. The apparent direction ofmutations to valuable sites results from selectiveamplification of the growth-limiting gene andselective retention of the improved allele (a formof selection). The modest increase in generalmutation rate caused by DinB is an artifact ow-ing to the chance proximity of the dinB and lacgenes, which enables selective coamplification.Cells in a hypermutable state are those withmultiple copies of the F0lac plasmid that in-cludes lac and dinB. These cells arise before se-lection and are not stress induced.

The response of populations to selection canbe rapid when it exploits two of the most com-mon mutation types known. Duplications andamplifications form frequently and are held inpopulations as stable genetic polymorphisms.Near-neutral beneficial alleles are extremelycommon and gain significant value when am-plified. Selection thus gives the appearance ofmutagenesis by driving rapid progressive genet-ic adaptation with no programmed change inmutation rate.

ACKNOWLEDGMENTS

We thank our laboratory mates Emiko Sano,John Paul Aboubechara, Semarhy Quinones-Soto, Andrew Reams, Eric Kofoid, and Douglas





Huseby for ideas and critique. We thank IvyRoush for allowing us to cite her unpublishedwork. We thank colleagues Wolf Heyer, NeilHunter, and Stephen Kowalczykowski for sug-gestions. We thank collaborators Dan Anders-son and Diarmaid Hughes of Uppsala Universi-ty for constant encouragement. Most of all, wethank John Cairns and our loyal opposition—Susan Rosenberg and Pat Foster—for their per-sistent and incisive defense of stress-inducedmutagenesis. Their models have provided con-stant stimulation even though we have remaineddubious that they are correct. The model de-scribed here for the Cairns system involves noprogrammed or induced mutagenesis. Selec-tion alone mimics all of the previously suggestedmechanisms—directed mutagenesis, hypermu-table states, and amplification during growthunder selection.

REFERENCES

Albertson DG. 2006. Gene amplification in cancer. TrendsGenet 22: 447–455.

Andersson DI, Hughes D, Roth JR. 2011. The origin of mu-tants under selection: Interactions of mutation, growth,and selection. In EcoSal-Escherichia coli and Salmonella:Cellular and molecular biology (ed. Slauch JM, et al.).ASM, Washington, DC.

Bachellier S, Clement JM, Hofnung M. 1999. Short palin-dromic repetitive DNA elements in enterobacteria: A sur-vey. Res Microbiol 150: 627–639.

Bergthorsson U, Andersson DI, Roth JR. 2007. Ohno’s di-lemma: Evolution of new genes under continuous selec-tion. Proc Natl Acad Sci 104: 17004–17009.

Bull HJ, McKenzie GJ, Hastings PJ, Rosenberg SM. 2000.Evidence that stationary-phase hypermutation in the Es-cherichia coli chromosome is promoted by recombina-tion. Genetics 154: 1427–1437.

Cairns J, Foster PL. 1991. Adaptive reversion of a frameshiftmutation in Escherichia coli. Genetics 128: 695–701.

Cairns J, Overbaugh J, Miller S. 1988. The origin of mutants.Nature 335: 142–145.

Elde NC, Child SJ, Eickbush MT, Kitzman JO, Rogers KS,Shendure J, Geballe AP, Malik HS. 2012. Poxviruses de-ploy genomic accordions to adapt rapidly against hostantiviral defenses. Cell 150: 831–841.

Foster PL. 1993. Adaptive mutation: The uses of adversity.Annu Rev Microbiol 47: 467–504.

Foster PL. 1994. Population dynamics of a Lac2 strain ofEscherichia coli during selection for lactose utilization.Genetics 138: 253–261.

Foster PL. 1997. Nonadaptive mutations occur on the F0

episome during adaptive mutation conditions in Escher-ichia coli. J Bacteriol 179: 1550–1554.

Foster PL. 1999. Mechanisms of stationary phase muta-tion: A decade of adaptive mutation. Annu Rev Genet33: 57–88.

Foster PL. 2000. Adaptive mutation in Escherichia coli. ColdSpring Harb Symp Quant Biol 65: 21–29.

Foster PL. 2005. Stress responses and genetic variation inbacteria. Mutat Res 569: 3–11.

Foster PL. 2007. Stress-induced mutagenesis in bacteria.Crit Rev Biochem Mol Biol 42: 373–397.

Foster PL, Cairns J. 1992. Mechanisms of directed mutation.Genetics 131: 783–789.

Foster PL, Trimarchi JM. 1995a. Adaptive reversion of anepisomal frameshift mutation in Escherichia coli requiresconjugal functions but not actual conjugation. Proc NatlAcad Sci 92: 5487–5490.

Foster PL, Trimarchi JM. 1995b. Conjugation is not requiredfor adaptive reversion of an episomal frameshift muta-tion in Escherichia coli. J Bacteriol 177: 6670–6671.

Foster PL, Trimarchi JM, Maurer RA. 1996. Two enzymes,both of which process recombination intermediates, haveopposite effects on adaptive mutation in Escherichia coli.Genetics 142: 25–37.

Friedberg EC, Walker GC, Siede W, Wood RD, Schultz A.2005. DNA repair and mutagenesis. ASM, Washington,DC.

Galhardo RS, Hastings PJ, Rosenberg SM. 2007. Mutation asa stress response and the regulation of evolvability. CritRev Biochem Mol Biol 42: 399–435.

Galhardo RS, Do R, Yamada M, Friedberg EC, Hastings PJ,Nohmi T, Rosenberg SM. 2009. DinB upregulation is thesole role of the SOS response in stress-induced mutagen-esis in Escherichia coli. Genetics 182: 55–68.

Galitski T, Roth JR. 1995. Evidence that F plasmid transferreplication underlies apparent adaptive mutation. Science268: 421–423.

Godoy VG, Fox MS. 2000. Transposon stability and a role forconjugational transfer in adaptive mutability. Proc NatlAcad Sci 97: 7393–7398.

Hall BG. 1988. Adaptive evolution that requires multiplespontaneous mutations. I: Mutations involving an inser-tion sequence. Genetics 120: 887–897.

Hall BG. 1990. Spontaneous point mutations that occurmore often when advantageous than when neutral. Ge-netics 126: 5–16.

Harris RS, Longerich S, Rosenberg SM. 1994. Recombina-tion in adaptive mutation. Science 264: 258–260.

Harris RS, Ross KJ, Rosenberg SM. 1996. Opposing roles ofthe Holliday junction processing systems of Escherichiacoli in recombination-dependent adaptive mutation. Ge-netics 142: 681–691.

Hastings PJ, Slack A, Petrosino JF, Rosenberg SM. 2004.Adaptive amplification and point mutation are indepen-dent mechanisms: Evidence for various stress-induciblemutation mechanisms. PLoS Biol 2: e399.

He AS, Rohatgi PR, Hersh MN, Rosenberg SM. 2006. Rolesof E. coli double-strand-break-repair proteins in stress-induced mutation. DNA Repair (Amst) 5: 258–273.

Hendrickson H, Slechta ES, Bergthorsson U, Andersson DI,Roth JR. 2002. Amplification-mutagenesis: Evidence that“directed” adaptive mutation and general hypermutabil-





ity result from growth with a selected gene amplification.Proc Natl Acad Sci 99: 2164–2169.

Hersh MN, Ponder RG, Hastings PJ, Rosenberg SM. 2004.Adaptive mutation and amplification in Escherichia coli:Two pathways of genome adaptation under stress. ResMicrobiol 155: 352–359.

Jarosz DF, Beuning PJ, Cohen SE, Walker GC. 2007. Y-familyDNA polymerases in Escherichia coli. Trends Microbiol15: 70–77.

Kim SR, Matsui K, Yamada M, Gruz P, Nohmi T. 2001. Rolesof chromosomal and episomal dinB genes encoding DNApol IV in targeted and untargeted mutagenesis in Escher-ichia coli. Mol Genet Genomics 266: 207–215.

Kofoid E, Bergthorsson U, Slechta ES, Roth JR. 2003. Forma-tion of an F0 plasmid by recombination between imper-fectly repeated chromosomal Rep sequences: Acloser lookat an old friend (F0128 pro lac). J Bacteriol 185: 660–663.

Kugelberg E, Kofoid E, Reams AB, Andersson DI, Roth JR.2006. Multiple pathways of selected gene amplificationduring adaptive mutation. Proc Natl Acad Sci 103:17319–17324.

Kugelberg E, Kofoid E, Andersson DI, Lu Y, Mellor J, RothFP, Roth JR. 2010. The tandem inversion duplication inSalmonella enterica: Selection drives unstable precursorsto final mutation types. Genetics 185: 65–80.

Layton JC, Foster PL. 2003. Error-prone DNA polymeraseIV is controlled by the stress-response sigma factor, RpoS,in Escherichia coli. Mol Microbiol 50: 549–561.

Lederberg J, Lederberg EM. 1952. Replica plating and indi-rect selection of bacterial mutants. J Bacteriol 63: 399–406.

Lee H, Popodi E, Tang H, Foster PL. 2012. Rate and molec-ular spectrum of spontaneous mutations in the bacte-rium Escherichia coli as determined by whole-genomesequencing. Proc Natl Acad Sci 109: E2774–E2783.

Lombardo MJ, Aponyi I, Rosenberg SM. 2004. Generalstress response regulator RpoS in adaptive mutation andamplification in Escherichia coli. Genetics 166: 669–680.

Luria SE, Delbruck M. 1943. Mutations of bacteria fromvirus sensitivity to virus resistance. Genetics 28: 491–511.

Maenhaut-Michel G, Shapiro JA. 1994. The roles of starva-tion and selective substrates in the emergence of araB-lacZ fusion clones. EMBO J 13: 5229–5239.

McKenzie G, Lee P, Lombardo M-J, Hastings P, Rosenberg S.2001. SOS mutator DNA polymerase IV functions inadaptive mutation and not adaptive amplification. Mo-lecular Cell 7: 571–579.

Mittelman D. 2013. Stress-induced mutagenesis. Springer,New York.

Nasvall J, Sun L, Roth JR, Andersson DI. 2012. Real-timeevolution of new genes by innovation, amplification, anddivergence. Science 338: 384–387.

Newcombe HB. 1949. Origin of bacterial variants. Nature164: 150.

Nilsson AI, Zorzet A, Kanth A, Dahlstrom S, Berg OG, An-dersson DI. 2006. Reducing the fitness cost of antibioticresistance by amplification of initiator tRNA genes. ProcNatl Acad Sci 103: 6976–6981.

Nohmi T. 2006. Environmental stress and lesion-bypassDNA polymerases. Annu Rev Microbiol 60: 231–253.

Paulander W, Andersson DI, Maisnier-Patin S. 2010. Am-plification of the gene for isoleucyl-tRNA synthetasefacilitates adaptation to the fitness cost of mupirocinresistance in Salmonella enterica. Genetics 185: 305–312.

Peters JE, Bartoszyk IM, Dheer S, Benson SA. 1996. Redun-dant homosexual F transfer facilitates selection-inducedreversion of plasmid mutations. J Bacteriol 178: 3037–3043.

Ponder RG, Fonville NC, Rosenberg SM. 2005. A switchfrom high-fidelity to error-prone DNA double-strandbreak repair underlies stress-induced mutation. Mol Cell19: 791–804.

Pranting M, Andersson DI. 2011. Escape from growth re-striction in small colony variants of Salmonella typhimu-rium by gene amplification and mutation. Mol Microbiol79: 305–315.

Radicella JP, Park PU, Fox MS. 1995. Adaptive mutationin Escherichia coli: A role for conjugation. Science 268:418–420.

Reams AB, Roth JR. 2014. Mechanisms of gene duplica-tion and amplification. In Recombination mechanisms(ed. Kowalczykowski S, Hunter N, Heyer W-D). ColdSpring Harbor Laboratory Press, Cold Spring Harbor,NY.

Reams AB, Kofoid E, Savageau M, Roth JR. 2010. Duplica-tion frequency in a population of Salmonella entericarapidly approaches steady state with or without recom-bination. Genetics 184: 1077–1094.

Reams AB, Kofoid E, Kugelberg E, Roth JR. 2012. Multiplepathways of duplication formation with and without re-combination (RecA) in Salmonella enterica. Genetics 192:397–415.

Rosche WA, Foster PL. 1999. The role of transient hyper-mutators in adaptive mutation in Escherichia coli. ProcNatl Acad Sci 96: 6862–6867.

Rosche WA, Foster PA. 2006. Methods for determiningspontaneous mutation rates. Methods Enzymol 409:195–213.

Rosenberg S. 2001. Evolving responsively: Adaptive muta-tion. Nat Rev Genet 2: 504–514.

Rosenberg SM, Shee C, Frisch RL, Hastings PJ. 2012. Stress-induced mutation via DNA breaks in Escherichia coli: Amolecular mechanism with implications for evolutionand medicine. Bioessays 34: 885–892.

Roth JR, Andersson DI. 2004. Amplification-mutagenesis—How growth under selection contributes to the originof genetic diversity and explains the phenomenon ofadaptive mutation. Res Microbiol 155: 342–351.

Roth JR, Benson N, Galitski T, Haack K, Lawrence JG, MieselL. 1996. Rearrangement of the bacterial chromosome:Formation and applications. In Escherichia coli andSalmonella (ed. Neidhardt FC), pp. 2256–2276. ASM,Washington, DC.

Roth JR, Kofoid E, Roth FP, Berg OG, Seger J, Andersson DI.2003. Regulating general mutation rates: Examinationof the hypermutable state model for Cairnisian adaptivemutation. Genetics 163: 1483–1496.

Roth JR, Kugelberg E, Reams AB, Kofoid E, AnderssonDI. 2006. Origin of mutations under selection: The





adaptive mutation controversy. Annu Rev Microbiol 60:477–501.

Sandegren L, Andersson DI. 2009. Bacterial gene amplifica-tion: Implications for the evolution of antibiotic resis-tance. Nat Rev Microbiol 7: 578–588.

Sano E, Maisnier-Patin S, Aboubechara J, Quinones-Soto S,Roth JR. 2014. Plasmid copy number underlies adaptivemutability in bacteria. Genetics 198: 919–933.

Shapiro JA. 1984. Observations on the formation of clonescontaining araB-lacZ cistron fusions. Mol Gen Genet 194:79–90.

Shee C, Gibson JL, Darrow MC, Gonzalez C, Rosenberg SM.2011. Impact of a stress-inducible switch to mutagenicrepair of DNA breaks on mutation in Escherichia coli. ProcNatl Acad Sci 108: 13659–13664.

Slack A, Thornton PC, Magner DB, Rosenberg SM, HastingsPJ. 2006. On the mechanism of gene amplificationinduced under stress in Escherichia coli. PLoS Genet2: e48.

Slechta ES, Liu J, Andersson DI, Roth JR. 2002. Evidence thatselected amplification of a bacterial lac frameshift allelestimulates Lacþ reversion (adaptive mutation) with orwithout general hypermutability. Genetics 161: 945–956.

Slechta ES, Bunny KL, Kugelberg E, Kofoid E, Andersson DI,Roth JR. 2003. Adaptive mutation: General mutagenesisis not a programmed response to stress but results fromrare coamplification of dinB with lac. Proc Natl Acad Sci100: 12847–12852.

Stahl FW. 1988. News and views: A unicorn in the garden.Nature 355: 112–113.

Stahl F. 1992. Unicorns revisited. Genetics 132: 865–867.

Steele DF, Jinks-Robertson S. 1992. An examination ofadaptive reversion in Saccharomyces cerevisiae. Genetics132: 9–21.

Sun S, Berg OG, Roth JR, Andersson DI. 2009. Contributionof gene amplification to evolution of increased antibioticresistance in Salmonella typhimurium. Genetics 182:1183–1195.

Sung HM, Yasbin RE. 2002. Adaptive, or stationary-phase,mutagenesis, a component of bacterial differentiation inBacillus subtilis. J Bacteriol 184: 5641–5653.

Syvanen M, Hopkins JD, Griffin TJt, Liang TY, Ippen-IhlerK, Kolodner R. 1986. Stimulation of precise excision andrecombination by conjugal proficient F0 plasmids. MolGen Genet 203: 1–7.

Torkelson J, Harris RS, Lombardo M-J, Nagendran J, ThulinC, Rosenberg SM. 1997. Genome-wide hypermutationin a subpopulation of stationary-phase cells underliesrecombination-dependent adaptive mutation. EMBO J16: 3303–3311.

Traxler BA, Minkley EG Jr. 1988. Evidence that DNA heli-case I and oriT site-specific nicking are both functionsof the F TraI protein. J Mol Biol 204: 205–209.

Walsh JM, Hawver LA, Beuning PJ. 2011. Escherichia coli Yfamily DNA polymerases. Front Biosci (Landmark Ed) 16:3164–3182.





2015; doi: 10.1101/cshperspect.a018176Cold Spring Harb Perspect Biol Sophie Maisnier-Patin and John R. Roth Mutagenesis (Adaptive Mutation)The Origin of Mutants Under Selection: How Natural Selection Mimics

Subject Collection Microbial Evolution

Mutation)Natural Selection Mimics Mutagenesis (Adaptive The Origin of Mutants Under Selection: How

Sophie Maisnier-Patin and John R. Roth

EvolutionPaleobiological Perspectives on Early Microbial

Andrew H. Knoll

Campylobacter coli and Campylobacter jejuniThe Evolution of

Samuel K. Sheppard and Martin C.J. MaidenPreexisting GenesEvolution of New Functions De Novo and from

Joakim NäsvallDan I. Andersson, Jon Jerlström-Hultqvist and

http://cshperspectives.cshlp.org/cgi/collection/ For additional articles in this collection, see

Copyright © 2015 Cold Spring Harbor Laboratory Press; all rights reserved


http://cshperspectives.cshlp.org/cgi/collection/

http://cshperspectives.cshlp.org/cgi/collection/

http://cshperspectives.cshlp.org/cgi/adclick/?ad=42791&adclick=true&url=http%3A%2F%2Fwww.genelink.com%2Fnewsite%2Fproducts%2FmodoligosINFO.asp


the origin of mutants under selection: how natural

Documents