Population Dynamics of a Lac- Strain of Escherichia Coli During Selection for Lactose Utilization

Population Dynamics of a Lac- Strain of Escherichia coli During Selection for Lactose Utilization

Patricia L. Foster

Department of Environmental Health, Boston University School of Public Health, Boston University School of Medicine, Boston, Massachusetts 021 18 Manuscript received February 25, 1994 Accepted for publication June 18, 1994

ABSTRACT During selectionfor lactose utilization,Lac' revertants of FC40, a Lac- strainof Escherichia coli, appear at a high rate. Yet, no Lac' revertants appear in the absence of lactose, or in its presenceif the cells have another, unfulfilled requirement for growth. This study investigatesmore fully the population dynamics of FC40 when incubated in the absence of a carbon source or when undergoing selection for lactose utilization. In the absence of a carbon source, the viable cell numbers do not change over 6 days. When incubated in liquid lactose medium, Lac- cells do not undergo any measurable increase innumbers or in turbidity for at least 2 days. When FC40 is plated on lactose minimum medium in the presence of scavenger cells,the upper limit to the amount of growth of Lac- cells during 5 days is one doubling, and there is no evidence for turnover (i.e., a balance between growthand death).The presence of a minority population that could form microcolonies was not detected. The implications of these results, plus the fact that the appearance of Lac' revertants during lactose selection is nearly constant with time, are discussed in reference to several models that have been postulated to account for adaptive mutations.

T is generally believed that all spontaneous mutations had another, unfulfilled growth requirement (CAIRNS I arise at random as a result of errors made during and FOSTER1991). normal DNA replication. However, an increasing body The fact that FC40 produces far more revertants after of evidence suggests that spontaneous mutations can plating than during priorgrowth allowedus to study the also arise among populations of cells subjected to nu- requirements for post-plating mutation, and also to dis- tritional deprivation (reviewed in FOSTER1993). Further- miss relatively easilycertain trivial explanations for our more, in some cases the process appears to be "adaptive" results. Such explanations fall into two main classes: (1) in the sense that the only mutations that arise are those deviations from the classical Luria-Delbriickdistribution that provide a growth advantage to the cell (CAIRNS et al. obtained from a fluctuation test can be due to depar- 1988; HALL 1990; CAIRNSand FOSTER1991; STEELEand tures from the assumptions about cell growth and mu- JINKS-ROBERTSON1992). tation thatunderlie the distribution (TESSMAN1988; In previous publications (CAIRNSand FOSTER1991; CHARLESWORTHet al. 1988; LENSKIet al. 1989; STEWART FOSTERand CAIRNS1992) we described a case ofadaptive et al. 1990; SARKAR 1991); (2) mutations that arise during mutation in a strain of Escherichia coli, FC40, that can- selection could be the result of growth of non-mutant not metabolize lactose because of a frameshift mutation cells on the selective plates (PARTRIDGEand MORGAN affecting the lac2 gene. Fluctuation tests showed that 1988; LENSKIet al. 1988; LENSKIand MITTLER1993a). many ofthe Lac' mutants that appearedearly after FC40 The first explanation is unlikely to be importantwith was plated onto minimum lactose plates had the clonal FC40. STEWARTet al. (1990) modeled several factors that distribution expected if the mutations givingrise to might produce deviations from the Luria-Delbruck dis- them occurred during the nonselective growth of the tribution toward the Poisson, namely poor growth of cultures priorto plating (LURIA andDELBR~CK 1943; LEA mutant cells during nonselective growth, poor plating and COULSON1949). However, new Lac' colonies con- efficiencyof mutants, and a generation-independent tinued to appear at a constant rate and these had the component of the mutation rate. The firstof these Poisson distribution expected if mutations were occur- proved not to be true upon testing-Lac' revertants of ring on the selective plates. From these results we con- FC40 did not differ from Lac- cells in their growth rate cluded that 90-95% of the Lac+ revertants that were on glycerol (CAIRNS and FOSTER1991). Even the most detected during a week that the cells were on lactose extreme examples of poor plating efficiency considered plates were due to mutations that occurredafter the cells by STEWARTet al. (1990) did not produce a distribution had beenplated. These revertants did not arise if lactose in which 90% of the mutants appeared to be part of a was not present, norin the presence of lactose if the cells Poisson component. In addition, these models would

Genetics 138: 253-261 (October, 1994) 254 P. L. Foster not account for the fact thatLac’ revertants continued For the experiment presented in Figure 3, the amount of to appear at a constant rate,and that the late appearing time required for starving cells to achievetwo generations was mutants had a Poisson distribution. The last hypothe- determined. A saturated culture of FC333 in minimum glycerol medium was incubated with shaking for a week. Every day sis, thatthere is a timedependentbut generation- an aliquot was diluted 1:lO into fresh minimum glycerolme- independent component of mutation, is indistinguish- dium, and growth was monitored with a Klett colorimeter. The able from ours. lag period increased by about 15 min each day, but the sub- We had also considered it improbable that the Lac’ sequent growth rates were the same. revertants of FC40 that arose after platingwere due to In the experiment presented in Figure 4, plates were respread by adding 50 pl of saline to the surface and spreading growth of Lac- cells on the plates. We would have de- with a glass spreader held stationary while the plates were spin- tected the 50-100-fold increase in the population that ning on a turntable. This technique was sufficient to disrupt would be necessary to produce the Lac’ mutants at preexisting Lac’ colonies of as few as four cells, which could the generation-dependent mutation rate (CAIRNS and be detected as colonies that subsequently appeared in a spiral pattern. 1991; FOSTER1993). However, it has been argued FOSTER The allelicstatus of rpoS was determined by the H,O, that the true amountof cell growth might be obscured “bubbling assay” described previously (ZAMBRANO et al. by the death of some of thecells, and that the surviving 1993). Briefly, a drop of H20, was placed on patches of cells cells, or a subpopulation of them, might have a high grown on LB plates and the degree of bubbling (indicating mutation rate (LENSKIand MITTLER1993a,b). Cell death catalase activity) compared to cells carrying rpoS+, rpoS819 and rpoS::kan alleles that were patched onto the same plates could also account for the failure to recover Lac’ mu- (all strains obtained from R. KOLTER). tants in the absence of selection for them (PARTRIDGE Statisticalmethods: Results are presented as means ? and MORGAN1988; LENSIUand MITTLER1993a,b). In this the standard error of the mean (SEM) , which was calculated as paper these and similar concerns are addressedby de- ( s2/n)where s2 is the variance. The 95% confidence intervals can be derived fromthe SEM by multiplying by 2 (large samples) tailing the population dynamics FC40of in the presence or by t with n - 1 d.f. (small samples) (ZUWAXIF 1970). To and absence of lactose. make more precise comparisons between means,the sampling data in Tables 1-3 and from CAIRNSand FOSTER(1991) were assumed to be normallydistributed with approximately equal MATERIALSAND METHODS variances, and used in one- or two-factor analysis of variance Bacterial strains and plasmids: All the strains used have a (ANOVAR). The error meansquare (= the residual vari- common parent, P9OC (COULONDREand MILLER1977). FC40 ance or the error variance) was then used as the variance in and the scavenger strain, FC29, havebeen previously described one- or two-tailed t tests to compare means. Thus, the 95% (CAIRNS and FOSTER1991). Briefly, FC40 isrifampicin-resistant confidence limits for the difference between two means is t X (Rif)and has an F’ carrying a lad-lacZ fusion with a + 1 base [s2/( n, + n,)] where s2 = the error mean square, n1 and n, pair frameshiftmutation, lacI33, in the lad coding sequence are for the means being compared, and t is taken with the (&OS and MILLER1981). FC29 is rifampicin-sensitive (Rip) degrees of freedom of the error mean square. and has anF’ carrying a deletion allele of 1acZ. FC281 is FC40 but rcfll7::Tn10; FC333is as FC40 but Rif and strepto- RESULTS mycin-resistant (Strep‘) and was derived from a spontaneous Strep‘ mutant of P9OC. All genetic manipulations used stand- FC40 cells neither grow nor die during lactose selec- ard techniques (MILLER1972). tion: The first question is can FC40 grow with lactose as Media: Bacteriawere cultivated as previously described a carbon source, i.e., how “leaky” is the lac allele? The (CAIRNS andFOSTER 1991). Liquid minimum medium or minimum plates were M9 (MILLER1972) with 20 pg/ml thiamine amount of @galactosidase produced by the mutant is and 0.1% lactose(Difco; filter-sterilized), 0.1 5% glycerol low, about 2.5 Miller units, compared to200-300 Miller (Sigma; filter-sterilized),or no carbon source. 0.001% gelatin units produced by a full Lac’ revertant. However, 2.5 was added to liquid minimum media.LB was as in MILLER (1972). units is not 0 (as measured, for example, in FC29, that Lactose MacConkey (Difco) plates were prepared according to has a deletion in ZacZ) . In addition, agar supplies some the manufacturer’s instructions. Antibiotic concentrations were: nutrients and both FC40 and FC29 will produce small 100 pg/ml rifampicin, 100 pg/ml streptomycin, 50 pg/ml car- benicillin, 10 pg/ml tetracycline in minimum medium and 20 colonies if plated at low density on minimum lactose pg/ml tetracycline in rich medium. 5Brom&hloro-3-indolyl- plates. To scavenge contaminants, 109FC29cells are usu- Pu-galactopyranoside (Xgal) was added at 40 pg/ml. ally plated with FC40. We previously published an ex- Scavenged minimum lactose mediumwas prepared by grow- periment inwhich the growthof FC40 was monitored on ing a culture of FC29 to saturation in minimum glycerol me- such plates by taking plugs of cells off the plates and dium, centrifuging the cells, resuspending them in the same volume of minimum lactose medium,and shaking the cultureat titeringfor viability (Figure 2 in CAIRNSand FOSTER 37” for 3-5 h. The cells were then removed by centrifugation, 10 1991). There was little change in cell numbers during pg/ml thiamine added (in case the FC29 cells had utilized some the 5 days of the experiment. The maximum difference of the thiamine), and the supernatant filtersterilized twice. occurred onday 4, when the number of cells was meas- Experimental techniques: Independent cultures were gen- ured to be 72% higher than on day 0 (measured im- erated by growing cells tosaturation in minimum glycerol medium, diluting this culture by into fresh medium, dis- mediately after thecells wereplated). Using the residual pensing I-ml aliquots into the appropriate number of tubes, variance and degreesof freedom calculated fromall of and allowing the cells to again reach saturation. the data (see MATERIALS AND METHODS),the upper limit Population Dynamics of Lac- E. coli 255

TABLE 1 0.20 The mean ho0and viable cell numbers of cultures of FC40 incubated in liquid minimum lactose medium and in minimum medium without a carbon source 0.15 '460,

No carbon source in medium Lactose medium 0 0 0.1 0 Day Mean Day n SEM Mean n SEM <= 0 0.0122 10 0.0011 0.0125 10 0.0007 0.7 0.01230.0011 10 0.0115 10 0.0006 1 0.0118 10 0.0008 0.0114 10 0.0007 0.05 1.7 0.0108 10 0.0007 0.0125 8 0.0011 2 0.0109 2 0.0008 10 0.0128 6 0.0014 2.7 0.0097 10 0.0009 0.0070 1 3 0.0094 3 10 0.0011 0.0100 1

Viable Lac+ cells per ml X 10" 0 1 2 3 4 No carbon source Lactose Days Day Mean Day n SEM Mean Day n SEM FIGURE1.-The A,, of 10 cultures of FC40 incubated in liq- 3 0.98 10 0.07 1.7, 2, 2.7 7 0.99 0.06 uid minimum lactose medium (see legend to Table 1). Aliquots of 10 p1 (containing about lo' cells) from 10 independent cultures of FC40 were each inoculated into 1 ml of minimum medium A,, of all the cultures on day 0, and the upper limit without a carbon source and into 1 ml of scavenged minimum lactose medium (see MATERIALS AND METHODS) in disposable cuvettes. These (95% confidence)of this difference is +21%. Similarly, were incubated at 37" and the A,,, taken twice a day against blanks the difference between the mean A,,, of the lactose cul- of each medium likewise incubated. When an increase in the absor- tures and themean A,,, of the cultures without a carbon bance of a lactose culture was first detected, it was titered on lactose MacConkey plates; all increases in absorbance correlated with the source on day 2 was +18%, and the upper limit (95% appearance of Lac' cells (red on lactose MacConkey plates). The confidence) of this difference is +39%. number of Lac- cells (white on lactose MacConkey plates) were also counted, if possible. (In three cases the initial increase in turbidity A separate series of experiments showed that Lac' occurred during the night and the culture was overwhelmed by revertants arise normally in liquid lactose medium. Lac'/cells by the next day.) At the end of the experiment, all of the About 10, cells in 0.1 ml of minimum lactose medium cultures without a carbon source were also titered on lactose MacCo- were inoculated into microtiter wells and the turbidity nkey plates. SEM = standard error of the mean. monitored daily. After day2, the proportion of nontur- (95% confidence)of this difference is +96%. Thus,this bid wells turning turbid each day was constant, and by experiment showed that FC40 cellscan, atmost, double day 4 anaverage of 0.4mutation perwell had occurred. during 5 days ofincubation on minimum lactose plates. Since, even in liquid medium, it takes about 2 days for Turbidity is a more sensitive measure of cell growth most Lac' mutants to become detectable (Figure l), and is little affected by cell death with a nonlysing strain the mutations that give rise to Lac' mutants observable of bacteria like FC40. In the experiment presented in on days 3-4 must have already occurred by day 2 (Figure Table 1 and Figure 1, FC40 cells wereincubated in liquid 2). These results will befurther considered in the lactose minimum medium and liquid minimum me- DISCUSSION. dium without a carbon source. The absorbance at 600 As mentioned above, it has been suggested that the nm (A6,,,,)was measured twice a day until all the lactose cell numbers on lactose plates give the illusion of sta- cultures had been overtaken by the growth of Lac' re- bility because some cells are able to utilize the lactose or vertants. During this time, the numbers of viable Lac- its contaminants, but othercells cannot and die(LENSKI cells did not change (Table1). It is clear in Figure 1 that and MITTLER1993a). If so, then death should become there was no detectable increase in A,, of any lactose apparent when cells are held without a carbon source. culture until the growth of Lac' cells became obvious No loss of viability was detected in the experiment in (P = 0.68). In fact, until that moment, there were no liquid medium described above (Table 1).Viability was significant differences in A,,, among thelactose cultures also monitored on plates without a carbon source (Table and the cultureswithout a carbon source (Table 1; P = 2) and, here too, there was no detectable decline in 0.70). These results strongly suggest that the cells re- viable cell number for 6 days (P = 0.82). It might be ceived no growth benefit from the lactose. However, argued thateven this apparent stability is a result of turn- even if the cells can grow on lactose but the methods over; some cells may growby utilizing nutrients released used were not sensitive enough to detect it, these data by their dying neighbors. To test this possibility, a few allow an upper limit to be placed on the amount of Strep' cells were plated with FC40 and scavengers on growth possible. By day 2, the 6 remaining lactose cul- minimum plates without a carbon source. On day 0 and tures had amean A,, that was 4% higher than themean 4, these plates were overlaid with top agar containing 256 Foster P. L.

TABLE 2

0.9 Viable counts of FC40 on minimum plates without a carbon source

~ ~- ~ ~~ 0.a Viable cells X lo-' on day:

0.7 5 4 0 3 1 2 6

0.6 Mean 1.28 1.34 1.50 1.39 1.46 1.36 1.43 n 10 10 10 10 10 10 10 0.5 SEM 0.10 0.12 0.13 0.08 0.08 0.12 0.14 0.4 Aliquots containing 1.3 X 10' FC40 cells from 10 independent cultures were spread with about lo9 scavenger cells on 70 minimum 0.3 plates without acarbon source. Five plugs were taken from one plate of each set each day and titered on LErifampicin plates, as described 0.2 (CAIRNS and FOSTER1991). Day 0 sampleswere taken immediately after the cells were spread. 0.1 TABLE 3 0.0 0 1 2 3 4 The survivai of a population of Strep' ce& on minimum lactose Days plates and minimum plates without a carbon source

FIGURE2.-The accumulation of Lac' revertants of FC40 in Strep' Cells per plate on day: liquid lactose medium. Eight independent cultures of FC40 Strep' cells were diluted 1/100 into scavenged minimum lactose medium added 0 2 4 (see MATERIALS AND METHODS) and 1OO-pl aliquots (containing about 3 X lo6cells) were dispensed into microtiter wells(total Plates without a carbon source of 726 wells). The microtiter plate lids were sealed around the 0 Mean 1.7 1.3 2.0 0.9 0.9SEM 0.9 0.6 at edges with Parafilm and the plates incubated37". Each day 25 Mean 25.0 31.3 26.0 the wells were examined with a dissecting microscope and 1.2 1.5SEM 3.6 scored as positive at the first sign of turbidity. The leftovercell 100 Mean 107.0 120.3 108.7 suspensions, consistingof <1 ml, were incubated separately in 5.8 5.0SEM 6.5 disposable centrifuge tubes and titered every day on lactose 200 Mean 227.7 242.3 235.0 MacConkey plates until these cultures were overtakenby Lac' SEM 7.0 9.7 8.4 cells. The mutations per well was calculated as m = -ln(Po), Lactose plates where Po was the proportionof wells remaining nonturbid on 0 Mean 1.3 2.0 1.3 0.3 0.3 0.2 0.3SEM 0.3 each day. (The 45 wells that became turbidby day 2 were elimi- 75 Mean 80.3 73.7 78.3 nated from this calculation because these probably contained 5.5 3.7SEM 2.7 preexisting Lac' mutants.) The figure shows the mean num- 150 Mean 149.7 148.3 131.3 ber of mutations perwell and the mean numberof viable Lac- 4.7 4.8SEM 5.6 cells perml (in the centrifuge tubes)2 SEM with n = 8. Because 300 Mean 261.0 268.3 260.3 a Lac' mutant takes about 2 days to become detectable as a 5.8 11.9SEM 3.0 change the turbidity, the curve showing mutations per well hasVarious dilutions of three independent cultures of FC333 (as FC40 been displacedtwo days tothe left to indicate the timeby which but Strep' Rifs)were added toapproximately 10' FC40 and lo9 the mutations would have already occurred. scavengers and plated in top agar on minimum plateswithout a carbon source (first experiment) or minimum lactose plates (second glycerol and streptomycin. If the populationwas turning experiment). All plates were immediately overlaid with another 2.5 ml top agar. Then on days 0, 2 and 4, one plate from each set was over, the total number of Strep' cells on theplate might overlaid with 2.5 ml top agar containing 1% glycerol and 6 mg strep remain stable, but individual cells woulddie and thusfail tomycin. Lac+ colonies were counted and marked on the lactose to give riseto progeny. As a result, the numberof Strep' plates each day.Strep' colonies were counted 2 and 3 daysafter overlay. n = 3 in every case. colonies would decline. As shown in Table 3, every Strep' cell plated on day 0 was able to give riseto a Strep' colony 4 days later. Overall, the numberof Strep' colony form- of FC40 did not accumulate during lactose selection if ing unitswas 3% higher onday 4 than on day 0, and the the cells could not benefit from the mutation because upper limit (95% confidence) for a decline in number they lacked another requirement for growth (CAIRNS is - 1%. A similar result was obtained on minimum lac- and FOSTER1991). The experiment presented inFigure tose plates, where there were 4% fewer Strep' colony 3 asks if Lac- cells plated on lactose, or plated without forming unitson day 4 than onday 0, and the upperlimit a carbon source, accumulate mutations that give an- (95% confidence) for a decline in number is -6%. other, nonselected phenotype. Resistance to rifampicin Nonselected mutationsdo not accumulate during lac- (Rip)was chosen as the second phenotype for several tose selection: It has also been argued that thecells un- reasons: (1) during nonselected growth the mutation dergo a nonspecific increase in mutation rateunder the rate toRip is about 5-fold greater than thereversion rate conditions of selection (LENSKIand MITTLER1993a). of the Lac' allele (data not shown); (2) because rifam- However, we have previously shown that our selective picin is lethal to wild type cells, the selection is "clean," conditions are notgenerally mutagenic-Lac+ revertants no residual growth occurs, no new Lac' mutants arise, Dynam ics of Lac-Population of Dynamics E. coli 257

before rifampicin was added. Theoutgrowth period was chosen to be sufficient to allow starving cells two gen- 70 r---" erations of growth (measure in liquid medium, see MA- 6o T 1 t TERIALS AND METHODS). This period was chosen because at thecell density plated (lo'), few newmutations giving the Rif phenotype would be likely to occur. That the outgrowth period was sufficient to allow a detectable number of recently arisen Rif mutants to be expressed was confirmed in two ways: (1) a fluctuation test incor- porating this amount of outgrowth on plates before rifampicin was added resulted in a distribution of Rip mutants with no indication of phenotypic lag (data not shown); (2) when ads- ogf cells in stationary phase were similarlygiven the same amount of outgrowth, the number of Rif mutants increased 5-fold (data not 0 1 2 3 Days shown); these cells, whichare missing all alkyltransferase activity, accumulate base substitution mutations in station- FIGURE3.-The number of Rip mutations of FC333 occurring during lactose selection. Aliquots (100 pl) from 5 inde- ary phase (REBECK and SAMSON 1991; FOSTERand Gums pendent cultures ofFC333 (as FC40 but Rif" Strep') were 1992; MACKAY et al. 1994). mixed with about 6 X lo9 scavengers and plated in top agar As shown in Figure 3, when FC333, a Rip equivalent on minimum lactose plates and minimum plates without a of FC40, was plated on minimum lactose plates, there carbon source. All plates were immediately overlaidwith another 2 ml top agar. On days 0,2 and 3, two lactose plates and was no increase in Rif mutants during thetime that the two plates withouta carbon source from each set were overlaid mutations giving rise to the 50 Lac' mutants appearing with 2 ml top agar containing 1% glycerol and 6 mg strepto- by day 5 must have occurred. Thus, there appears notto mycin (to kill the scavenger cells). One of each type of plate be a general increase in mutation rates under the selec- from each set was then immediately overlaidwith 5 ml top agar containing 3 mg rifampicin; the other plates were incubated tive conditions, although it is possible that some muta- at 37" to allow Rif mutants to be expressed, and then overlaid genic process occurs that affects only particular muta- with rifampicin. The time of outgrowth was 4 hr on day 0,4.5 tions or only certain regions of the chromosome hr on day 2, and 4.75 hr on day 3; these times were chosen to ( DANCHIN1988; GRAFXN 1988; LENSKI and MITTLER allow equivalent amounts of growth (see MATER~ALSAND METH- 1993a). But even with these caveats, the experiment, at ODS). Lac+ colonies were counted and marked on the lactose plates each day. Rif colonies werecounted 2 and 3 days after the least, demonstrates again that theLac' mutants that overlay. No symbol, cumulativeLac' colonies on lactose plates; arise are not the result of simple population changes, circles, Rif colonies after outgrowth; squares, Rip colonies because such changes would be expectedto also give rise without outgrowth; closed symbols, Rif colonies on lactose to Rif mutants. plates; open symbols, Rif colonies on plates withouta carbon source. The Lac' curve hasbeen displaced two days to the left Lac+ mutants are unlikely to have arisenfrom a grow- to indicate the time at which the mutations giving riseto either ing subpopulation of cells: While the results presented Lac' or Rif mutants would have alreadyoccurred. The figure above showthat theoverall population is stable, they do shows the means 2 the SEM with n = 5. not exclude the possibility that all the Lac' revertants and plates can be incubated as long as necessary to en- arise in a subpopulation of cells that has a growth ad- sure that all mutants have appeared; (3) the mutational vantage on lactose and a high mutation rate toLac'. As events that give rise to Rif are simple base-substitutions originally hypothesized by LENSKIet al. (1988, 1989), a in the /3 subunit of RNA polymerase (OVCHINNIKOVet al. mutant population that can weakly utilize lactose ap 1983;J1~ andGROSS 1989) and areresponsive to avariety pears during nonselective growth of the cultures; after mutagenic treatments (MILLER1972); (4) base substitu- plating on lactose, these mutants grow into impercep tions have been shown to give riseto late-appearing mu- tible microcolonies and sustain mutations that give the tants in several systems (CAIRNS et d. 1988; HALL1990, final full Lac' phenotype. These intermediates can be 1991;FOSTER and CAIRNS 1992; PR~VALand CEBULA1992). thought of as cells with an imperfectly reverted allele The difficulty with the Rif phenotype is that, like (LENSKIand MITTLER 1993a), or cells carrying an extra- other drug resistances, it may not be immediately ex- genic suppressor ( MAURICE FOX,personal communica- pressed. Although the majority of Rif alleles are domi- tion), orcells that have duplicated or further amplified nant when tested in merodiploids (OVCHINNIKOVet al. the unreverted allele (JOHN ROTHand FRANKLIN STAHL, 1983; JIN and GROSS1989), a cell with a newly arisen personal communication). mutation may require time to synthesize enough of the The intermediate model has at least three mutant protein togive the Rip phenotype. For this rea- parameters-the initial number of intermediates (which son, the cells were givena period of growth on glycerol is determined by the mutation rate to the intermediate), 258 P. L. Foster the growth rate of the intermediate,and thesecond mutation rate to the final genotype (LENSKIet al. 1989). Obviously, the initial number of intermediates at the time of plating cannot beless than the final number of fully Lac' mutants that appear (forFC40, the frequency of intermediates would have to be greater than lo"). The growth rate of theintermediates must be fast enough to start yielding revertants within a few hours of plating, but the combination of the number of intermediates and theirgrowth must not be so large that they produce a detectable increase in turbidity or viable count. If the intermediateswere the result of mutations that occurred during prior growth, as proposed by LENSKI et al. (1989), then the intermediateswould havea clonal distribution, and the final distribution of Lac' mutants would reflect this. However, if the intermediatesarise by 012345678 some other mechanism, such as unstable amplification Days during the last stages of prior growth (JOHN ROTH and FIGURE4.-The accumulation of Lac+ revertants of FC40 FRANKLINSTAHL, personal communication), then theini- after respreading. Aliquots of approximately 3 x lo6 cells of FC40 from 5 independent cultureswere spread with approxi- tial population of intermediates might very well have a mately 3 X lo9scavenger cells on 60 minimum lactose plates. Poisson distribution. Thus, the intermediatehypothesis Lac' colonies were counted each day, and5,15 and 18 plates cannot be dismissed simplyon the basis of the distribu- with no visible Lac' colonies at 30X magnification were re- tion of the numbers of Lac' revertants in a fluctuation spread on days 1, 2 and 3. Of these, several actually had one or undetected Lac' colonies, which were recognized by test. However, the fact that Lac' revertants of FC40 ap- two the appearanceof many Lac' coloniesin a spiral pattern2 days pear at nearly a constant rate during lactose selection after respreading. Most of these plates were subsequentlydis- (CAIRNS and FOSTER1991) puts severe restraints on any carded, but the number of Lac' colonies estimated to have hypothesis that supposes growing intermediates. If mu- been present on the day of respreading was included in the tants arise from a population thatis increasingwith time, totals. The figure shows the mean cumulative numberLac+ of colonies per plate. The final numberof plates contributingto whether this population is large or small, then the rate the curves was 5, 12 and 13 respread on days 1, 2 and 3, re- at which mutants appear should increase with time un- spectively; and 60, 55, 40 and 22 non-repread on days 0, 1, 2 less the mutation rate is, for some unknown reason, in- and 3-8, respectively. Curves from the respread plates appear versely proportional to the population size. displaced to the right because ofmost these plates had no Lac' The best data on the rate at which Lac' revertants colonies on the day of respreading. arise during lactose selection are from a large fluctuation test that we previously published (Figure 3 and examined each day at 30X magnification against a grid, Table 1 in CAIRNS and FOSTER1991). One hundred and and, over the course of 5 days, the position of every small twenty independent cultures of FC40 were plated with colony recorded. Every colony observed became a full- a 10-fold excess of scavenger cells on minimum lactose sized Lac+ colony visibleto the unaided eye by the next plates, and Lac' colonies were counted every day for 7 day. This result indicated either that every microcolony days. The colonies that appeared on day 2 were due to rapidly gives rise to a fully Lac' mutant, or that all mi- events that occurred during priorgrowth ofthe cultures, crocolonies are restricted to less than lo4cells. In either and the numbers of colonies appearing on day 3 were case, the second mutation rate Lac'to would haveto be low because about 30% of the revertants take somewhat high. To test these possibilities, lo6 FC40cells were longer than2 days to produce visible colonies. However, spread with scavengerson minimum lactose plates, and from days 4 to 7, the rate at which revertants appeared on days 1, 2 and 3, plates with no visible Lac' colonies was constant to within 20%, and the limit (95% confi- were respread. As shown inFigure 4, respreading didnot dence level) for either an increase or a decrease in the result in an increase in the rate of appearance of Lac' rate is 25%. colonies. An increase would be expectedif the mutation The existence of microcolonies was also looked for rate was sufficiently high that a large proportion of mi- directly by microscopically examining populations of crocolonies would normally sustain more than one mu- FC40 on minimum lactose plates. In a preliminary ex- tation, or if growth of the microcolonies was limited for periment, itwas determined that at30X magnification, some density-dependent reason (for example, diffusion a colony of lo4 cells was readily observable. FC40 cells of nutrients into the colony). ( los) from 9 independent cultureswere mixed with lo9 The respreadingexperiment alsoallowed a more scavengers and spread on lactose plates. The plates were detailed examination of the population on minimum Population Dynamics of Lac- E. coli 259 lactose plates. The plates were respread with a continu- plified copies of the lac- allele that confer a full Lac' ous, circular motion so that any small colony of Lac+ phenotype; theamplified array would be resolved asthe revertants that had not been detected on the day of re- Lac' colony develops (JOHN ROTHand FRANKLINSTAHL, spreading would be recognizable two days later as an arc personal communication). Dependingon the mutation of regrown colonies. Were Lac+ mutants arising from rate within the amplified array, such a populationwould microcolonies, then each Lac' colony appearing after not necessarily have been detected by the respreading respreading shouldlikewise be surroundedby an arcof experiment. To test this possibility, newly arisen Lac+ regrowing microcolonies. Each dayduring therespread- colonies of FC281 werestreaked on minimum glycerol- ing experiment the area surroundingall ofthe approxi- Tet plates containing X-gal (Lac+ cells are blue on this mately 130 Lac+ colonies that arose after respreading medium), andsimultaneously on minimum lactose-Tet was carefully examined at 30X magnification, and ap plates. If Lac+ mutants arise from an amplified array, proximately half of these areas were also examined at "young" Lac+ colonies should give mixtures of blue and loox magnification. In only two cases were potential white colonies and/or blue colonies that segregate white microcolonies observed. On oneplate, 8 small colonies cells on the nonselective X-gal medium. In all, 82 colo- appeared 3 days after respreading. All of these subse- nies isolated over 4 days weretested, and 15 of these gave quently became fully Lac', and were probably due to a a mixture of blue and white colonies. However, the same colony of slowgrowing revertants that had been present colonies that gave mixed streaks on X-gal alsogave on the day of respreading. On anotherplate which had mixed colony sizes on lactose. On retesting from the been respread on day 3, 13 small colonies appeared on lactose plates, all of the colonies that were small on lac- day 6, but never developed into large Lac' colonies. tose were whiteon X-gal, and all ofthe colonies that were These may have been contaminants, or, possibly, mu- large on lactose were blue on X-gal. Because amplifylng tants that can scavenge stationary phase cells (see be- cells would havebeen bothstabilized and selected for on low). But, these two isolated instances, which together the lactose plates, this result indicates that the white accounted for less than 20% of the Lac+ colonies that colonies in the original streaks were most likely con- appeared after respreading, provide little support for taminating nonreverted cells. No sectoring colonies the hypothesis that Lac' mutants arise from a micro- were found. Thus, although these results are not de- colony population. finitive, they provide little evidencethat amplification Amplification of the lac- allele does not appear to of the lacregiongives rise to a detectable microcolony allow cells to multiply on lactose: One explanation for population. the appearance of Lac+ colonies after selection is that Lac' revertants of FC40 do not have mutationsin the the Lac- cells can become phenotypically Lac' by am- stationary-phasesigma factor: As E. coli enters into sta- plifying the unrevertedallele. Although the respreading tionary phase, a number of genes are induced whose experiment would probably have detected this, Lac+ expression depends on the stationary-phasesigma factor colonies were also explicitly tested for amplification of encoded by rpoS (reviewed inKOLTER et al. 1993).Upon the lac allele. As shown by TLSTYet al. (1984), amplifi- prolonged incubation in stationary phase, cells can ap- cation is unstable during nonselective growth. Cells pear thatcarry certain mutantsalleles ofrpoS that confer (lo8) fromthree independent cultures ofFC281, a the ability to scavenge the wild-type population. It has tetracycline-resistant (Tet')derivative of FC40, were been suggested that the growth these rpoS mutants plated with scavengers on minimum lactose plates. On could account for the occurrenceof other mutations day 7, 300 Lac+ colonies were picked at random onto during selection (ZAMBRANOet al. 1993). To test this minimum lactose-Tet plates. After growth, these were possibility, 40-50 Lac+ revertantsof FC40arising each replicated onto the same medium (to select for more day from days 3 to 9 were checked for their RpoS amplification and purify them away from the scaven- phenotype. All were apparently rpoS by the criterion gers), then ontorich LB-Tet plates (to allow for deam- that they produced 0, from H,O, at thewild-type rate plification), and then ontolactose MacConkey plates (to ( ZAMBRANOet al. 1993). non-selectively score for Lac' mutants, which are redon these plates). Of the 300, only22 were not red onlactose DISCUSSION MacConkey plates, but of these, 18 had also failed to The experiments described above address several of grow after replication onto the second set of lactose the hypotheses that have been proposed to explain why plates. Thus, only 4/282 Lac+ colonies, less than 2%, Lac+ revertants of FC40 arise when lactose is the sole showed anyevidence of instability.If amplification per se source of carbon, but do not arise when the cells are can render a cell Lac+, such Lac+ cells are only a small starving in the absence of lactose (CAIRNS and FOSTER minority of the mutant population. 1991). The simplest of these is that FC40 can grow on It is also possible that cells amplifylng the lac region lactose plates, either because the Lac- allele is leaky, or might grow a little on lactose plates, develop into mi- because othernutrients are present (PARTRIDGEand crocolonies, and thensustain mutations among the am- MORGAN1988; LENSKI and MITTLER 1993a,b). Further- 260 P. L. Foster more, the amount of growth might be masked by cell Then, if 3 X 10' cells of FC40are plated on lactose plates, death or turnover (LENSKIand MITKER1993a,b). The each of the 3 X lo5 cells of the hypothetical subpopu- results presented above showthat this hypothesis cannot lation could not have undergone more than 9 genera- be true-during the period when Lac' mutants are ac- tions of growth by day 2. Yet, according to the model, cumulating, the Lac- cells do notdie or turn over, and these cells must account for theapproximately 50 Lac+ can achieveless thana doubling on lactose plates colonies that appearby day 4, implyinga mutation rate (Tables 1-3 and Figure 2) (Figure 2 in CAIRNS and FOSTER greater than lo" per cell division. 1991). In addition to these theoretical considerations, mi- In the presence of lactose, FC40 accumulates Lac' croscopic observations also limit the size of any micro- mutants at a constantrate. Although there is some varia- colony population and its mutation rate. Because they tion among experiments and among laboratories, in a could not be seen at 30X magnification, microcolonies typical experiment with 3 X 10' cells on a minimum either cannot exceed lo4 cells at any point during an lactose plate, 50 to 200 new Lac' mutants have appeared experiment, or the cells must have such a high mutation after 5-7 days (CAIRNS and FOSTER 1991;HARRIS et al. rate that each microcolony gives rise to a Lac' mutant. 1994). We can now provide upper limits on the popu- However, werethe mutation rate that high,or were the lation changes that could be taking place during this growth of microcolonies limited for some density- period and still remain undetected by our methods. At dependent reason, the microcolony population would the 95% confidence level, the upper limit for the in- have been detected by respreading (Figure 4). crease in viable count is less than one doubling and the All of these results lead to the conclusion that, during upper limit for the amount of cell death or turnover is lactose selection, the Lac' mutants that arise cannot do less than 10%. Applying these results to a typical ex- so at the normal generationdependent mutation rate periment, it follows that during the period when 50 either from the population atlarge, or from a subpopu- Lac+ mutants are arising, the maximum number of lation. Certainly, with enough imagination, some com- cell divisions that could be taking placeis 3 X 10'. At bination of changing mutation rates and growth param- the normal generation-dependent mutation rate,this eters might be found that wouldexactly mimic the number of cell divisions could account foronly about resultswith FC40. For example, the constant appearance 1 mutation. of Lac' revertants could be modeled by allowing a grow- It has been argued that the mutation rate during ex- ing subpopulation to have a high mutation rate initially, ponential growth is irrelevant. Starvation could be mu- but making that mutation ratefall (for some reason) in tagenic, and the failure of Lac' mutants to accumulate inverse proportion to the population size. Or, the mu- on plates without a carbon source could be due to cell tation rate per generation could be constant, but the death (LENSKIand MITTLER1993a). The results pre- subpopulation growing on lactose might have the un- sented above also place limits on models of this kind. usual property of producing, at each division, onlyone When incubated without a carbonsource, the cell num- live daughter cell. But, such models tend to become bio- ber did not decline, and there was no evidence of turn- logically implausible. over (Tables 1-3). FC40 also did not appear to have a We are left, then, with a rather limited number of generally heightened mutation rate during starvation possibilities. Because the revertants of FC40 that arise (Figure 3). during selection are dependent on Red (CAIRNS and Several models have been proposed thathypothesize FOSTER1991), and amplification probably requires Re- a subpopulation of cells that can grow on lactose and cA's recombination function (TLSTYet al. 1984),we pre- then give riseto thefull Lac+ mutants (LENSKIet al. 1988, viously suggested that mutants arise from a subpopula- 1989; LENSKIand MITTLER1993b; JOHN ROTH and tion of cells that is ampllfylng the lac region (FOSTERand FRANKLIN STAHL,personal communication; MAURICE FOX, CAIRNS 1992; FOSTER1992). This model provides for personal communication). However, the limits on cell adaptive mutations because, within an amplifylng cell, if growth also apply to this hypothetical subpopulation. no useful mutation occurred, the amplification eventu- Because the subpopulation must be better able to grow ally would be resolved. If a mutation occurred that al- on lactose than the majority, they might have been de- lowed the cell to grow, the amplification would also be tected as an increase in turbidity when FC40 was incu- resolved but the useful mutation would be retained. bated in minimum lactose medium. However, no sig- However, this hypothesis differs from another, similar nificant increase in turbidity was detected (Figure 1 and hypothesis proposing that amplification of the mutant Table 1).Even allowing for the imprecision of the mea- lac allele confers a growth advantage on lactose (JOHN surements, the amount of growth that the subpopula- ROTHand FRANKLINSTAHL, personal communication). tion could have achieved by day 2 could not have been The results with FC40 suggest, instead, that atany given as much as to give a 50% increase in total cell mass. It time a constant proportionof the population under lac- has been proposed that the subpopulation is initially tose selection may be amplifylng the lac region. This 0.1% of the total population (LENSKIet al. 1989;JOHN hypothesis leaves open the possibilities that the muta- ROTHand FRANKLINSTAHL, personal communication). tion rate within the amplified array may be a function of Population Dynamics of Lac- E. coli 261 the numberof copies, that the numberof copies within HALL, B. G., 1991 Spectrum of mutations that occur under selective and non-selective conditions in E. coli. Genetica 73-76. a cell may increase with time, and even that the process 84: HARRIS, R. S., S. LONGERICHand S. M. ROSENBERG,1994 Recombina- may be stimulated by the presence of lactose. But, to be tion in adaptive mutation. Science 264: 258-260. consistent with our finding that the reversion rate of JIN,D. J., and C. A. GROSS, 1989 Characterization of the pleiotropic FC40 is nearly constant duringselection, the numberof phenotypes of rifampin-resistantrpoBmutants ofEscherichia coli. J. Bacteriol. 171: 5229-5231. cells that are amplifjmg(or perhaps their extentof am- KOLTER,R., D. A. SIEGELEand A. TORMO,1993 The stationary phase plification) would have to be constant with time unless, of the bacterial life cycle. Annu. Rev. Microbiol. 47: 855-874. of course, the reversion rate per gene copy declines LEA, D.E., and C. A. COWISON,1949 The distribution of the numbers of mutants in bacterial populations. J. Genet. 49: 264-265. with time. LENSKI,R. E., and J. E. MIITLER,1993a The directed mutation con- However, there is, as yet, no evidence that amplifica- troversy and neo-Darwinism. Science 259: 188-194. tion of any sort is occurring. The only clues we have to LENSKI, R. E., and J. E. MI-ITLER,1993b Directed mutation. Science the mechanism by which adaptive mutations arise in 260: 1222-1223. LENSKI,R. E., M. SLATIUNand F. J. Am,1988 Another alternative to FC40 is that they are dependent on some function or directed mutation. Nature 337: 123-124. functions of the RecABC pathway (CAIRNSand FOSTER LENSKI, R. E., M. SLATKINand F. J. AYALA,1989 Mutation and selection 1991; FOSTER1993; HARRIS et al. 1994), and that they in bacterialpopulations: Alternatives tothe hypothesis ofdirected mutation. Proc. Natl. Acad. Sci. USA 86: 2775-2778. consist mainly of simple 1-base deletions (FOSTERand LURIA,S. E., and M. DELBRUCK,1943 Mutations of bacteria from virus TRIMARCHI1994; ROSENBERCet al. 1994). Thus, alter- sensitivity to virus resistance. Genetics 28: 491-511. native hypotheses,such as transitory, error-prone MACKEY, W. J., S. HANAND L. D. SAMSON,1994 DNA alkylation repair limits spontaneous base substitution mutations in Escherichia DNA synthesis (FOSTER1992; STAHL1992; FOSTERand coil. J. Bacteriol. 176: 3224-3230. TRIMARCHI1994), are still possible. It also remains to MILLER,J. H., 1972 Experiments in Molecular Genetics. Cold Spring be seen if the results with FC40 are generalizable to Harbor Laboratory, Cold Spring Harbor, N.Y. other strains and organisms. OVCHINNIKOV,Y. A., G. S. MONASWKAYA,S. 0. GURIEV,N. F. KALININA, E. D. SVERDLOVet al., 1983 RNA polymeraserifampicin resistance mutations in Escherichia coli:sequence changes and domi- I thank R. KOLTERand M. VOLKERTfor strains and plasmids,J. CAIRNS, nance. Mol. Gen. Genet. 190: 344-348. J. DRAKE,E. EISENSTADT,M. Fox and F. STAHLfor useful discussions,and PARTRIDGE,and L., M. J. MORGAN,1988 Is bacterial evolution random M. Fox, R. LENSKI,J. MITTLER,S. ROSENBERG,J. ROTH and F. STAHLfor or selective? Nature 336: 22. communicating unpublished results and/or hypotheses.J. PISANOand PJUVAL,M. J., and T. A. CEBULA,1992 Sequence analysis of mutations J. TRIMARCHIprovided excellent technical assistance. This work was arising during prolonged starvation of Salmonella typhirnun'um. supported by grant MCB-9213137 from the U.S. National Science Genetics 132 303-310. Foundation. REBECK, G. W., and L. SAMSON,1991 Increased spontaneous mutation and alkylation sensitivityofEscherichia coli strains lacking the ogt LITERATURE CITED 0'-methylguanine DNA repair methyltransferase. J. Bacteriol. 173: CAIRNS,J., and P. L. FOSTER,1991 Adaptive reversion of a frameshift 2068-2076. mutation in Escherichia coli. Genetics 128: 695-701. ROSENBERG,S. M., S. LONGERICH,P. GEE and R. S. HARRIS, 1994 Adap CAIRNS, J., J. OVERBAUCHand S. MILLER, 1988 The origin of mutants. tive mutation by deletions insmall mononucleotide repeats. Nature 335: 142-145. Science 265: 405-407. &os, M. P., and J. H. MILLER,1981 Genetic and sequence analysis SARKAR,S., 1991 Lamarck contreDanuin, reduction uersusstatistics: of frameshift mutations induced by ICR-191. J. Mol. Biol. 153: Conceptual issues in the controversy over directed mutagenesis 39 - 66. in bacteria, pp. 235-271 in Organismand the Origzns of Self; CHARLESWORTH,D., B. CHARLESWORTHand J. J. BULL,1988 Origin of edited by A. I. TAUBER.Kluwer Academic Publishers, The mutants disputed. Nature 336: 525. Netherlands. COULONDRE,C., and J. H. MILLER,1977 Genetic studies of the lac STAHL,F. W., 1992 Unicorns revisited. Genetics 132: 865-867. repressor. 111. Additional correlation of mutational sites with spe- STEELE,D. F., and S. JINKS-ROBERTSON,1992 An examination of adap cific amino acid residues. J. Mol. Biol. 117: 525-575. tive reversion in Saccharomyces cerevisiae. Genetics 132: 9-21. DANCHIN,A., 1988 Origin of mutants disputed. Nature 336: 527. STEWART,F.M., D. M. Go~oo~and B. R. LEVIN,1990 Fluctuation FOSTER,P. L., 1992 Directed mutation: between unicorns and goats. analysis: the probability distribution of the number of mutants J. Bacteriol. 174: 1711-1716. under different conditions. Genetics 124: 175-185. FOSTER,P. L., 1993 Adaptive mutation: the uses of adversity. Annu. TESSMAN,I., 1988 Origin of mutants disputed. Nature 336: 527. Rev. Microbiol. 47: 467-504. TLSTY,D. T., A. M. ALBERTINI and J. H. MILLER, 1984 Gene amplifi- FOSTER,P. L., and J. CAIRNS, 1992 Mechanisms ofdirected mutation. cation in the lac region of E. cole'. Cell 37: 217-224. Genetics 131: 783-789. &BRANO, M. M., D.A. SIEGELE,A. MARTA,A. TORMOand R. FOSTER,P. L., and J. M. TRIMARCHI,1994 Adaptivereversion of a KOLTER,1993 Microbial competition: Escherichia coli mu- frameshift mutation in E. coli by simple base deletions at he tants that takeover stationary phase cultures. Science 259: mopolymeric runs. Science 265: 407-409. 1757-1760, GWEN,A., 1988 Origin of mutants disputed. Nature 336: 525-526. ZLJWAYLIF,F. H., 1970 GeneralApplied Statistics. Addison-Wesley, WL,B. G., 1990 Spontaneous point mutations that occur more Reading, Mass, often when they are advantageous than when they are neutral. Genetics 126: 5-16. Communicating editor: N. R. DFUNKWATER