On solid plates for leucine starvation, a reverse mutation formed only one leu⁺ revertant papilla on the flat colony. Therefore a papilla can be regard as one mutation event. The mutations accumulated in a time-dependent manner (Fig. 1). Leu⁺ revertant papillae did not appear until 2 days after plating due to the phenotype lag. The papillae observed in the first 2–3 days probably existed pre-plated. After that, new mutations continuously occurred and more papillae formed. These were regarded as mutations occurred under leucine starving pressure.

The results in Fig. 1 also indicated that reversion rates were not constant during the leucine starvation process. In the early period, fewer reversions took place per day; while in the later period, more reversions occurred per day. For example, during day 4–9, the average reversions per day were 5.14, but during day 14–18, the average reversions per day were 26.44. On the other hand, the number of viable cells decreased only about 17-fold during the same period. Thus the mutation rates were more than 100-fold higher during day 14–18 than that during day 4–9. This result was further supported by the data in Table 1 and Table 2. In Table 1, mutation rates of other types in liquid cultures also increased during starvation compared to those in rich medium-LB. In Table 2, especially on the micro-dropped plates, the number of revertant papillae per micro-drop increased with time of starvation.

Death rate of cells on the spread selective plates was about 13.67% higher than that of on the dropped plates (Fig 2). But on both types of plate, the number of viable cells was similar regardless of the presence or absence of Ampicillin. It was previously regarded as the evidence that in the starvation process, cells did not divide, so adaptive mutations occurrence under selective conditions was regarded as cell division-independent 11, i.e., growth-independent. However the data in Table 2,3 were not in line with it.

The data in Table 2 and Table 3 indicated that the occurrence of leu⁺ reverse mutations was growth-dependent under leucine starvation conditions. Although the death rates of cells on the spread MM+P plates and on the dropped MM+P plates was similar with or without the addition of Ampicillin (Fig. 2), and the reverse mutations under leucine starvation increased with the time prolonged (Fig. 1). These two results were similar to the previous investigations reported by Hall 9. They were previously regarded as the evidence that in selective conditions, cells did not divide, and adaptive mutation was growth independent 11. Our explanation for this discrepancy is that: under selective conditions, only small a part of cells divided, and these divided cells were so few compared to the mass of the viable but non-dividing cells that they could not be monitored by the penicillin method.

Although the cells in divisions may only account for a small proportion of the total viable cells under leucine starving conditions, they might become the important source for the appearance of leu⁺ revertants. This was consistent with the results from Table 2, which indicated that the occurrence of mutation was cell density and growth dependent, and was further supported by the results in Fig. 4 of the leucine concentration determination and by the reversions occurred in the media where scavenger bacteria were pre-cultured (Table 3). Besides, extracellular DNA and nucleotides could also be consumed as the source of carbon and energy supporting the bacteria cells to grow under nutrient deprived conditions 20. This also suggested that a small proportion of the population cells could grow in nutrient deprived conditions by absorbing the amino acids and peptides or absorbing DNA and nucleotides, which may leak from the other viable or dead cells.

It has already been proved that mutations under selection require trace nutrient which the cell is defect in the function to synthesize or metabolize. Hall used minimal medium containing 5 μM tryptophan to select trp⁺ revertants 9, Fani 7 reported the number of his⁺ revertants was always greater in plates containing trace of histidine (0.25 ug/ml) than those without histidine in E. coli FB184 hisA. All these results also suggested that occurrence of mutations under selection were growth-dependent.

Fluctuation analysis of mutations in liquid culture indicated that mutations occurred spontaneously during leucine starvation. It is well known that mutations occurred spontaneously under nonselective conditions in bacteria. From the results 2.3, we could see that under starving condition, there was still trace amount of leucine leaked from other cells in the medium, the reverse mutations could originated from the minority divided cells by the support of this amount of leucine, and then would have the Luria-Delbrück distribution patterns.

The spectra of leuB gene of leu⁺ revertants, isolated from leucine starvation conditions, did not show specificity. This was in agreement with our conclusion that mutations under starvation were growth dependent and occurred spontaneously.

Our results in fig. 1 indicated reversion rates under leucine starvation condition were not constant. In the first few days (2–3 days, in Fig. 1), the leu⁺ reversions accumulated quickly, and they were regarded as existing pre-plated and formed papillae at different time intervals because of the phenotype lag. In the days followed, new reversions accumulated very slowly, and the mutation rate during this period (day 4–10) was low. Later on, reversions accumulated more quickly, and the mutation rate was high. This stage could sustain for some time (day 12–21) and decreased at last.

Although other researchers have reported the mutation rates increased under selective conditions, they did not mention or emphasis that the mutation rates were not constant. For examples, Hall 910 used E. coli FCY2B trpB^- mutant, E. coli FCY6B trpAB^- mutant, on tryptophan limited plates; Shapiro14 used E. coli MCS2 ara-lacZ, on Ara-Lac selection plates with different concentrations of glucose; Bridges21 used E. coli trp A23 mut Y under on minimal plates; Kasak22 used P. Putica PaW85 phe^- on phenol-minimal plates. All of curves (the mutants appeared in a day versus culturing time) involved several parts which with different curvatures respectively. They could not be simply regarded linear as Hall did 9.

In fact, during the selection process, the viable cells did not increase but decreased slowly. This is an important reason for elevating the mutation rate. But, the magnitude of viable cells decreased was less than that of increased mutation rate, which suggested the selection played a role on elevating the mutation rates.

Why were mutation rates inconstant during selections, and were higher under selections than that of under non-selections? Here are some possible explanations.

Firstly, there was misapprehension for defining and calculating mutation rate. Generally, mutation rate (μ) was defined as mutations per cell per division. For a certain gene in a bacterium such as E. coli, regardless of the hot spots, suppose that the generation time is about constant during the exponential phase and all the cells are synchronized, μ is proportional to the mutation rate μ_b (mutations per base pair per genome replication), according to Drake's paper 23. For bacteria growing in rich medium (such as LB) in the exponential phase, the number of cells is approximate to the number of dividing cells. So, in the normally growing populations in exponential phase, almost each originating of DNA replication can result in a complete genome and accompany a cell division, μ_b is thus almost equals to the number of damages or mistakes per base pair occurs during one genome replication, and μ_b is low.

Under selections, the divided cells are so few that it is difficult to determine the number of divided cells. Then the mutation rate is usually defined as mutations per viable cell per day79101214, marked as μ_d in this paper. If the generation time is constant during the selective process, μ_d is also proportional to the mutation rate μ and to μ_b .

Under selections, the growth of E. coli cells was severely restricted. The generation time is much longer than that under non-selective conditions. During the prolonged period of cell division, although most cells cannot replicate complete genomes, they might have initiated genome replications and produced incomplete genomes. This assumption was based on the result that 25% of the genome turned over per day during amino acid starvation as estimated by Bridges from the mutation rate of a mutT mutant 24, and based on that 20% of genome DNA was turned over in 24 hours in the buffer determined by Tang 25. Of the whole population, where small portion of cells accomplish complete genome replication, most of cells would undertake incomplete replication. Thus the accumulated spontaneous DNA damages in the uncompleted cycles, which would cause mutations when DNA replication resumes, may be fixed in the completed genome, and become true mutations. Different pathways may be used to fix DNA damages and an important one is homogenous recombination. The long generation time provides opportunities for cells to fix these DNA damages. So the mutations per base per genome replication not only include the damages in one cycle of complete genome DNA replication, but also include any accumulated DNA damages in other incomplete genome DNA replications. Thus, under selective conditions, mutation rate per base pair per genome replication becomes a collective result and would be higher than that under non-selection. Due to the same reason, a bacteria population at different growth stages during the selection progress may have different generation times and would result in different mutation rates respectively.

Secondly, changes of metabolic control system under selective conditions may also elevate the background mutation rates, which can also affect advantageous adaptive mutation rates. It is well known that in the long process of evolution of microbes, diverse mechanisms had been developed to maintain their genetic stability (μ_g) and decrease their mutation rates (μ_b) during their genome expanding23. These mechanisms may include metabolic controls over concentrations of endogenous and exogenous mutagens, pre-replication DNA repair systems, the insertion accuracy of polymerases, 3'-exonucleolytic proofreading, and several post-replication systems for repairing mismatches 22627282930313233343536. Different bacteria apply different sets of these mechanisms, and the efficiency of a particular mechanism varies among microbes.

Leucine starvation and other selective condition can be regarded as an environmental signal. Populations that have experienced a severe challenge from the selective environment might therefore be expected to change their physiological state extensively, such as stringent response, global regulation in gene expressions, and other inducible responses such as SOS. These changes are to keep cells surviving and adapting to the selective environment. During this process, the elaborate enzymatic machinery, which can keep the fidelity of DNA replication and control the spontaneous mutation rate, are destroyed; some new proteins must be synthesized using exogenous and endogenous nutrition and energy sources. In the end, populated cells could get into a relative stable states, in which, cell division becomes more difficult, generation time becomes longer; the fidelity of DNA replication decreases, the spontaneous mutation rates (adaptive mutation rates) increases.

The strain used in this study was purchased from ATCC.

Escherichia coli ATCC 31343 F- hsdS20 ((r- B m- B)) ara-14 leuB6 proA2 lacY1 galK2 rpsI20 xyl-5 mtl-1 supE44 lambda-.

Escherichia coli ATCC 35577 F- recA1 uvrA6 phr-1 thi-1 thr-1 leuB6 lacY1 galK2 ara14 xyl15 mtl1 proA2 argE3 rpsL31 tsx-33 supE44 gyrA98 lambda-.

The rich complete medium was Luria broth (LB), consisting of 10 g of peptone, 5 g of yeast extract powder, and 10 g NaCl per liter. pH7.2. Minimal medium (MM) consisting of 0.5 g of sodium citrate, 7 g of K₂HPO₄·2H₂O, 3 g of KH₂PO₄·2H₂O, 1 g of (NH₄)₂SO₄, 0.1 g of MgSO₄·7H₂O, 2 g of glucose, and 1 mg of thiamine per liter. Selective medium containing, in addition to the minimal medium, 30 mg leucine per liter for selecting pro⁺ mutants (MM+L), or 30 mg proline for selecting leu⁺ mutants (MM+P). Complete synthesized medium (CM) containing, in addition to minimal medium, 30 mg leucine and 30 mg proline per liter. CM-Lac medium was used to select lac⁺ revertants, it was same to CM except that glucose was replaced by lactose.

Cells were grown at 37°C for 12–16 hours in LB medium. Then the cells were centrifuged, washed, and resuspended to about 1 × 10⁹ cells/ml. The 10 μl aliquots were distributed on the MM+P selective plates and on CM plates. All plates were incubated in a humidified chamber at 37°C. On dropped plates, the suspension formed "flat colony" that was normal in appearance except for a slight dimpled. The number of leu⁺ revertants papillae on the flat colonies was counted each day; and the number of viable cells in the flat colony was determined by transferring them onto LB plates.

On solid plates, only one papilla formed after a mutation occurred. So, the number of mutations is approximately equal to the number of revertant papillae. The mutation rates is determined according to the expression μ = M/N, where M is number of reverse mutations, is about equal to the number of revertant papillae N_m. N is the number of viable cells in the flat colonies with no papillae, these colonies were present on the plates 2 days earlier than the papillae counted. The mutation frequency is determined according to the expression f = N_m/N.

Cells were grown at 37°C for 12–16 hours in LB medium . Then the cells were centrifuged, washed, and resuspended to about 1 × 10⁹ cells/ml. The 2-ml aliquots were transferred into 18 ml MM+P, CM and LB medium, respectively. A serial of 0.5 ml aliquots from each well mixed suspension were dropped into 48 wells of a microtilter dish and incubated at 37°C. The number of leu⁺ revertant were determined by plating the cultures onto MM+P plates, the number of viable cells were determined by diluting and plating the cultures onto LB plates.

The mutation rates were calculated by the P₀ methods of Luria and Delbrück 16. According to the equation of μ = -lnP₀/N, where P₀ is the proportion of cultures with no revertants, and N is the mean number of cells per culture. The mutation frequency is determined according to the equation of f = ∑N_m/nN. Where, ∑N_m is the total number of revertants in all cultures, N is the mean number of cells per culture, and n is the number of cultures.

The PCR primers were based upon the nucleotides sequence from the GenBank (accession no.D17631) The primer 1 is 5'-gctaccgatattgtcgagtc-3', 110 bp upstream the initiate code 'atg'; the primer 2 is 5'-cgtcttagccatgattacac-3', 14 bp downstream the terminate code 'taa'. The PCR product is 1216 bp, containing the full length of leuB gene (1092 bp).

All PCR reactions and PCR products purification have done according to the manufacturer's instructions of Qiagen PCR kit. The purified PCR products were separated and sequenced by cycle sequencing according to the manufacturer's instructions on an ABI Model 3700 automated sequencer by United Gene Holdings, Ltd.(Shanghai, China).

Cells were grown at 37°C for 12 hours in LB medium. Then the cells were centrifuged, washed, and resuspended. The aliquots were distributed in 0.85%NaCl, MM, MM+P and CM media respectively, and continuously incubated at 37°C. After every period, part of the cultures were picked off to determination the number of viable cells by transferring appropriate culture onto LB plates. The leucine concentrations in the cultures (discarding the cells) and in the supernatants of the cells lysates were determined by using HITACHI 835 high speed amino acids analyzer. See additional file 1, 2 for the information about this analyzer and the determination conditions (Additional File 1 and Additional File 2).