Bacterial strains and growth conditions
All bacterial strains are derivatives of Salmonella enterica ssp. enterica serovar Typhimurium strain LT2 (except for three Escherichia coli strains that were used as PCR templates for amplification of selection cassettes used in λ Red recombineering). For transferring chromosomal markers between strains, generalized transduction with phage P22 HT 105/1 int-20121 or a defective prophage derivative of it (see below; “Construction and use of an artificial Gene Transfer Agent”) was used. Lysogeny broth (LB; 10 g/L Tryptone [Sigma], 5 g/L yeast extract [Sigma], 10 g/L NaCl) was used as rich medium, and was supplemented with 15 g/L Bacto agar to make LB agar (LA) plates. Salt-free LB (LB prepared without NaCl) was used for preparing cells for transformation by electroporation. SOC medium22 was used for recovery after transformations. As minimal medium, M9 medium23 was prepared with twice the amounts of M9 salts, glucose, CaCl2 and MgCl2 in order to allow growth to a higher cell density. Minimal medium was supplemented with tryptophan (W; 0.1 mM [1x] or 5 µM [0.05x, for tryptophan-limited medium]), histidine (H; 0.1 mM), and guanosine (Guo; 0.3 mM) when needed. For simplicity, the base medium is referred to as 2× M9, with any additional additions indicated as + H (histidine) + W (tryptophan) and + Guo (guanosine). To make solid M9 plates, 15 g/L of Bacto agar was added. All growth of bacteria was done at 37°C (except when preparing cells for λ Red recombineering). Antibiotics (trimethoprim [tmp; 10 mg/L; Sigma], tetracyclin [tet; 7.5 mg/L; Sigma], chloramphenicol [cam; 12.5 mg/L; Sigma]) were added to the medium only when needed for selecting recombinants after transformations and transductions.
PCR and Sanger sequencing
DNA for λ Red recombineering and Sanger sequencing was prepared by PCR using Phusion DNA polymerase (Thermo Fisher). All PCR reactions were made with bacterial cell suspensions as templates; either part of a single fresh colony from an agar plate or a ~ 1 µl sample from a frozen (-80°C) glycerol- or DMSO stock culture was re‑suspended in 100 µl ultrapure water (Sigma). Of this, 1 µl was used per 20 µl of PCR reaction. PCR cycling parameters have been described elsewhere24–26. For de-salting, removal of oligonucleotides and primer-dimers, and concentration of the DNA, PCR products were precipitated with polyethylene glycol (PEG). Briefly, PCR reactions were mixed 1:1 with a 2× PEG precipitation solution (24% [w/v] PEG 8000 [Sigma], 20 mM MgCl2 in ultra-pure water [Sigma]) and incubated for 10 min at room temperature. DNA was pelleted by centrifugation at ≥ 14,000 × g for 10 min, washed once in ≥ 1 volume of 70% ethanol and re-centrifuged at ≥ 14,000 × g for 10 min, after which it was air-dried. For local sequencing, DNA was dissolved in ultra-pure water, mixed with a sequencing primer (Table S4), and sent to Eurofins Genomics (Ebersberg, Germany) for Sanger sequencing. DNA for recombineering was dissolved in 2–10 µl of ultra-pure water.
λ Red Recombineering
Mutations, genetic markers, and reporter genes were constructed and inserted using different λ Red recombineering strategies as described elsewhere25–27. Cultures of bacterial strains carrying the λ Red helper plasmid pSIM5-Tet28 were grown in salt-free LB plus 0.2% (w/v) glucose at 30°C overnight, after which the cultures were diluted 100-fold into 25 ml of the same medium and allowed to grow at 30°C for one hour. To induce expression of the λ recombination genes, the cultures were transferred to a 42°C shaking water bath and incubated for another 15 min. After cooling in an ice-water bath for several minutes the cultures were pelleted by centrifugation and washed once in 1 ml ice-cold de-ionized water or 10% glycerol. The cells were resuspended in 200 µl of ice-cold de-ionized water or 10% glycerol. For each transformation, DNA (0.2–2 µl of de-salted and concentrated PCR product) was mixed with 20 µl cell-suspension in an electroporation cuvette (Bio-Rad, 1 mm gap) and electroporated (2.5 kV, 200 Ω, 25 µF in a Gene Pulser Xcell; BioRad). Immediately after electroporation the cell suspensions were re-suspended in 200 µl pre-warmed (42°C) SOC, transferred to a 10- or 50 ml plastic tube and incubated at 42°C for 15 minutes before plating on selective medium.
All DNA oligonucleotides are listed in Table S4. For re-construction of mutations in hisA and trpA, the target genes containing the mutations were PCR amplified from evolved populations or isolated clones from evolved populations. In some cases, the gene was amplified as two overlapping fragments using a mutant and a wildtype as template in order to avoid one mutation present in a mutant containing several mutations. The PCR products were transformed into strains carrying the selectable and counter-selectable cassette Atox126 (GenBank accession: MN207489; containing dhfr [trimethoprim resistance], PrhaB-orph11 [DNAse toxin orphan-11 from E. coli EC869 under the control of the rhamnose inducible rhaB promoter], and amilCP [blue chromoprotein from Acropora millepora]) in hisA or trpA, selecting for loss of the cassette on M9 + 0.3% (w/v) rhamnose + histidine + tryptophan plates. Deletions, some single nucleotide substitutions, and small insertions were constructed using DIRex25,26, using “half-cassettes” that were generated by amplifying overlapping parts of Atox1 and Atox2 (GenBank accession: MN207490) as described previously25,26.
Evolution experiments
Experiment 1 (populations 1–1 to 1–8) was started from separate cultures of DA52864 (ΔtrpF ΔPgal ΔgalE::Tlux). In addition to the deletion of the trpF coding sequence (corresponding to the C-terminal domain of the fused TrpCF protein), it carries a deletion of the gal operon promoter and has the galE gene replaced by the lux transcriptional terminator24. These mutations make the strain conditionally resistant to phage P22 infection (resistant in the absence of galactose but sensitive in the presence of galactose), and were used in order to avoid contamination by phage P22 and avoid selection of P22 resistant mutants, which we have occasionally seen in previous experiments. After a first overnight growth in medium containing both histidine and tryptophan, samples from each population were passaged 1:100 into 5 mL tryptophan-limited medium (2× M9 + 0.05× W) in 50 mL conical bottom screwcap tubes, and were continually passaged into the same medium once every seven days for 20 cycles, after which they were passaged 1:1000 into the same medium every 3–4 days (twice a week) for an additional 30 cycles, totaling approximately 430 (20*6.64 + 30*10) generations. As each cycle started with fresh tryptophan-containing medium, growth for the first few generations occurred without any selection for TrpF activity. Once tryptophan was depleted, only mutants able to synthesize tryptophan (due to mutations that generated TrpF activity in another enzyme) could grow, making selection for TrpF activity intermittent. Samples from the populations were frozen at -80°C in 20% DMSO after cycle 10, 15, 20, 25, 40 and 50.
Experiment 2 (lineages 2mut + 1–2mut + 24 and 2mutS1–2mutS24) was done similarly to experiment 1, but was started from separate cultures of DA62207 and DA64168 (ΔtrpF and ΔtrpF mutS; see supplementary Table S5 for the detailed genotypes of these strains). In addition to the trpF deletion and the mutS mutation, these strains contain the arabinose-inducible defective P22 prophage GTA22 (below), which makes them resistant to lytic growth of phage P22 and makes them convenient both as donors and recipients in transductions. These 48 populations were passaged with 100× dilutions into 5 ml of fresh tryptophan-limited medium (2× M9 + 0.05× W) once every seven days for 13 cycles, at which point all but three mutS populations grew visibly denser. Samples from the populations were frozen at -80°C in 20% glycerol after cycle 4, 5, 6, 7, 8, 10, 12 and 13.
Growth curves and growth rate determinations
Overnight cultures (in 1 mL 2× M9 + 0.4% glucose + 0.1 mM histidine + 0.1 mM tryptophan) were started either from isolated colonies of pure re-constructed clones or from thawed 50 µL samples from freezer stocks of evolved populations. Pure clones were grown as five biological replicates while populations were grown as a single sample per population. After approximately 24 hours of growth at 37°C the cultures were diluted 1:1000 into tryptophan-free (2× M9 + 0.4% glucose) and tryptophan-limited (same but with 5 µM tryptophan) medium, after which 300 µL of each diluted culture was transferred to a well in a Bioscreen honeycomb plate. To avoid evaporation of media, all empty wells were filled with 300 µL medium or water, and the plates were sealed by taping around the edges. The plates were incubated in a Bioscreen C reader (Oy Growth Curves, Turku, Finland) at 37°C, with continuous moderate shaking, and reading of optical density at 600 nm (OD600) every 4 minutes for 7 days. Despite taping the edges, the wells around the edges of the plate reached a higher final OD than other replicates of the same strain and upon inspection these cultures showed a significant loss of volume. For growth rate determinations, the OD600 data was plotted to find the exponential phase (which ended at about OD600 = 0.18 for the cultures in tryptophan-limited media) and the data for only the exponential phase was transferred to Prism 9 or 10 (GraphPad Software). The data was fitted (without subtracting any blank value) to the exponential growth equation Y(t) = b + Y0*ekt, where Y(t) is the OD600 at time t, b is the contribution of the medium and the plastic of the individual well of the Bioscreen plate to the OD600 (“blank”), Y0 is the contribution of cells to the OD at time t = 0, k is the growth rate (min− 1), and t is the time (min). Differences of < 5% in k between two strains were considered insignificant due to limitations such as the precision of OD measurements and well-to-well variation in the Bioscreen plates. The average growth rate up to 90% of the maximum OD for each culture (Fig. 3) was determined as follows: First the OD600 data was blank-subtracted using the b value from the curve fitting (above) as blank. For each culture the first data point above Y0 (Y> Y0) and the first point where OD reached above 90% of the maximum OD (Y90%ODmax) within the first three days were determined. The average growth rates between those two points were determined as kavg = [ln(Y90%ODmax)-ln(Y> Y0)]/[t90%ODmax-t> Y0] and are expressed in day− 1.
Whole Genome Re-sequencing
Genomic DNA was prepared from 1–2 ml samples from evolving populations using MasterPure (Epicentre). Libraries for MiSeq sequencing (in-house) were made using Nextera DNA library prep and indexing kits (Illumina). Some samples were sequenced by BGI (Beijing, China) using their proprietary technology. Analysis of sequence reads were done with CLC Genomics Workbench (Qiagen). After quality-based trimming to remove any ambiguities and low-quality reads, the reads were mapped against a reference genome containing all modifications present in the ancestral strains. Indels and SNPs were detected using the low frequency variant detector, and structural rearrangements (duplications or deletions) were found using visual scanning of mapped read depth in combination with the structural rearrangement tool in CLC. Amplification copy number in the single population (DA65458; 2mutS8) that contained an amplification of a relevant target gene (hisA) was estimated by dividing the mean read depth of the structural genes in the his-operon (7136 bp at 156.49× read depth) with the mean of the mean read depths of four identically-sized regions (4× 7136 bp at 19.95× read depth) just outside of the amplified region. One relevant mutation (trpA[Pro62Fs]) was reported by CLC genomics workbench to be detected in 30–60% of the reads covering that position. However, when the reads were re-mapped to a reference sequence containing the mutation, 82–100% of the reads from the relevant populations mapped perfectly to the mutation.
Isolation and sequencing of clones from evolved populations
In order to determine the order of appearance of mutations in populations with more than one mutation in the target gene (hisA or trpA) during the evolution experiments, clones were isolated from frozen stocks of earlier time-points, and the target genes were PCR amplified to generate templates for Sanger sequencing. For each frozen population, a streak was made to isolate single colonies on LA plates. Eight colonies from each streak were picked and streaked once more on LA plates, and from each of these eight clones a single colony was used as template for PCR. Some clones with known hisA or trpA alleles were frozen in the strain collection and used as templates in PCR reactions for re-construction of mutations in fresh (unevolved) genetic background.
Construction and use of an artificial Gene Transfer Agent
After transductions with phage P22, it is of critical importance to isolate phage-free bacterial clones to avoid the lytic spread of phage in cultures as well as to avoid the selection of phage-resistant mutants or lysogens that are impossible to use in later transductions. Additionally, in long-term evolution experiments, we have occasionally found evolving populations to be infected by P22, or to carry mutations that provide resistance to P22, which could be problematic for later genetic experiments. Therefore, an inducible artificial Gene Transfer Agent (GTA; here referred to as GTA22) was constructed (using λ Red recombineering) by three modifications to a P22 HT 105/1 sieA44 lysogen. The mutation HT 105/121 in makes DNA packaging into new P22 virions less specific, allowing a high frequency of generalized transduction. The sieA44 mutation prevents the lysogen from blocking entry of DNA from P22 (and related phages) in the periplasm, making the lysogen work as a recipient in transductions. However, P22 also encodes a phase-variable O-antigen conversion locus, gtrABC29, which when it is expressed prevents adsorption of P22 to the cell (P22 uses the O-antigen as receptor), making the cells resistant to superinfection (and prevents transduction).
To convert this prophage into an inducible gene transfer agent, both ends, including the attachment site, of the prophage were deleted (using DIRex25,26). One deletion (ΔthrW-kil) removes 6.9 kb of DNA from the “left” end of the prophage, including the left copy of the P22 attachment site (in the thrW tRNA gene), the integrase and excisionase genes (int and xis), the recombination genes (abc1, abc2 and erf) and the septation inhibitor (kil). The other deletion removes 3 kb of DNA including the right copy of the attachment site (a P22 encoded partial duplicate of the thrW gene) and the phage-encoded O-antigen conversion locus (gtrABC). Secondly, in order to make strains containing GTA22 work as donor in transductions the prophage was made inducible by placing (using λ Red recombineering) a copy of the P22 antirepressor (antP22) after the L-arabinose inducible ParaBAD promoter in the host genome. Without the attachment site and the integrase/excisionase the prophage cannot excise and circularize its genome when it is induced, and any replication forks starting at the phage origin of replication extends into the surrounding bacterial host genome30. Without the O-antigen conversion genes and a functional sieA gene the lysogen cannot prevent adsorption of P22, and will therefore accept incoming DNA from P22 virions. But, as GTA22 still contains a functional P22 repressor (c2) and other regulatory elements, the lysogen is resistant to lytic growth of any superinfecting P22 genome. As packing of the P22 genome into new virions is unidirectional and initiates at a single pac site in the phage genome30, GTA22 (being unable to excise and circularize) is unable to pack a complete copy of its genome into new virions, making it essentially avirulent (but it could theoretically form rare virulent virions if a copy of its genome is circularized through illegitimate recombination). Thus, a strain containing GTA22 is resistant to lytic infection by P22 but works efficiently as recipient in transductions. Furthermore, it never (or very rarely) forms virulent virions, which removes the need for screening to avoid phage infected clones.
In order to induce GTA22, overnight cultures were diluted 1:20 into LB containing 0.1% (w/v) L-arabinose. After incubation for at least 6 hours to allow lysis, any surviving bacteria were killed by vigorous shaking with chloroform. Lysates were cleared by centrifugation (≥ 14,000 ×g for 1–3 min) before being used in transductions. For transductions, 0.1–10 µl of cleared lysate was mixed with 100 µl LB supplemented with 0.4% glucose (to prevent expression of ParaBAD-antP22 in the recipient due to remaining arabinose in the lysate), after which 100 µl overnight culture of the recipient strain was added. After 30 min incubation at 37°C the transductions were plated on appropriate selective media, and the plates were incubated until colonies appeared (overnight on LA, 24–48 h on M9).
Generation of a mutator allele
In order to generate a reversible knock-out of mutS, an insertion of the selectable and counter selectable Acatsac1 cassette (GenBank accession: MF124798; containing the genes amilCP, cat, and sacB, conferring blue color, chloramphenicol resistance, and sucrose sensitivity, respectively) was designed as a duplication-insertion (DUP-In)24 so that it would generate a duplication with a single copy of the Acatsac1 cassette inserted between two truncated mutS copies, one containing the first ~ 1.5 kb of the 2.5 kb mutS gene, and the other containing the last ~ 1.5 kb. As none of the mutS copies are complete they are unlikely to be functional (as was verified from the large amounts of genomic mutations in the evolved mutator populations compared to the isogenic non-mutator populations; Table S2, parts 2 and 3), and due to the presence of the cassette the colonies are blue, chloramphenicol resistant and sucrose sensitive. The reason for making it as a DUP-In was to be able to revert the mutator allele (by plating on sucrose to find cells that lost the cassette by recombination between the 500 bp repeats) in clones isolated from the evolved populations, but this was not tested.