Following the Dynamics of Structural Variants in Experimentally Evolved Populations

Structural variants (SVs) are alterations of the genomic sequence, generally affecting 50 bp or more. The four categories of described SVs are large insertions, large deletions, inversions, and duplications. Until recently, more attention has been devoted to single-nucleotide variants (SNVs) than to structural variants, in terms of their phenotypic effects and their role as genetic determinants of disease, or their contribution to adaptation. This is probably because it is easier to both detect SNVs and predict their phenotypic effects. However, short- and long-read deep sequencing technologies have strongly improved the detection of SVs, at least in single individual or clonal genomes¹. In parallel, their phenotypic effects have been better characterized, and many examples of their implication as genetic determinants of human disease^2,3 or adaptation to a new environment⁴ have been documented.

Deletions and insertions, often due to mobile genetic element (MGE) insertions, are much more disruptive than single nucleotide polymorphisms (SNPs) and lead to frameshift mutations and protein structure modifications. Deletions and MGE insertions within genes almost always result in gene inactivation, and insertions into non-coding regions can lead to repression or constitutive expression of adjacent genes when insertion sequences (ISs) contain promoter or termination sequences⁵. While the knockout of essential genes leads to clear detrimental effects on bacterial fitness, the loss of non-essential genes is beneficial in some cases. Despite their inherent costs, duplications can also be advantageous, and participate in adaptation as they lead to a change in gene dosage; an increase in the activity of a specific protein can be advantageous depending on the conditions⁶.

Microbial experimental evolution populations are usually started with clones. This initial absence of genetic diversity, combined with the "closed environment" characteristic of test tubes, leads to a very limited potential of evolution by gene gain through horizontal gene transfer and recombination. In these specific conditions, the contribution to adaptation of deletions, duplications, and intragenomic MGE insertion is particularly important; bacteria often adapt through loss-of-function mutations (mainly due to deletions or MGE insertions), affecting genes that are not useful in stable, often nutrient-rich, monoculture artificial environments⁷. In the longest running E. coli evolution experiment, IS150 insertions are particularly frequent amongst populations evolved after 50,000 generations, with IS elements representing 35% of mutations that reach high frequency in populations that retain their ancestral point mutation rate⁸.

Evolve and resequence studies couple experimental evolution and next-generation sequencing (NGS) technologies to investigate how bacteria adapt, at the phenotypic and genomic levels, to different environmental conditions and stresses, such as different carbon and energy sources, antibiotics, and osmotic stress^9,10,11. These studies typically obtain genomic information on the evolved populations or clones solely at the experimental end point, and in some cases, at a number of intermediate time points^12,13,14. These data provide insight into genes and pathways involved in the adaptation to a given environment, but rarely allow researchers to follow the dynamics of de novo emerging and sweeping alleles over time.

One approach to follow these dynamics is to choose a limited number of segregating alleles of interest (because of the function of the genes they affect, because they sweep in parallel in independent populations, etc.) and use amplicon sequencing to quantify the allele proportion, pooling many time points in the same sequencing run¹⁵. This method has been successfully used to follow the dynamics of small size variants (SNPs or 1 bp indels) in experimental¹⁶ and natural¹⁷ populations of microbes. However, in the case of larger indels or MGE insertions, the size difference of the amplicons induces PCR efficiency differences, which distort the relationship between read and allele proportions. In certain cases, the size difference between the two alleles is superior to the classical length of the amplicon. Here, we coupled a triplet PCR technique with automated parallel capillary electrophoresis to quantify the relative frequency of an insertion allele based on size discrimination. This approach allows the exploitation of underused experimental time points to determine the dynamics of an emerging mutant allele and to follow its frequency to fixation or loss, in a cost-effective manner. We applied this method to track emerging mutS^- alleles, mutated through an IS10 insertion, providing the mutated genotype with a hypermutator phenotype.

This method requires two target alleles with a ≥5% difference in size. First, primer triplets are designed to produce similarly sized fragments, which share a common primer. Second, PCR conditions are optimized, and a calibration curve is produced using mixes of wild-type (WT) and mutant gDNA. Lastly, samples are amplified by PCR, and the relative frequency of each allele is quantified by parallel quantitative capillary electrophoresis.

Setting up this protocol requires precise knowledge of the insertion, deletion, inversion, or duplication point within the ancestral sequence. This information is usually obtained by whole-genome sequencing (WGS) of the end or intermediate point samples. In the following protocol, the general principle for the case of an insertion mutation is given for each step, alongside a representative case where the frequency of an IS10 insertion in the mutS gene in an experimental evolution population of E. coli is followed. In this population, WGS of the endpoint population identified the insertion of a 1,329 bp IS10 between positions 2,463 and 2,471, resulting in the duplication of this insertion site. This method is applicable to the three other SV types, and the specificities of each case are given in the discussion.

1. Design of triplet primers

1. Use classic primer design practices (18-24 bp, 40%-60% GC content, start/end with G/C pairs where possible, T_m difference < 5 °C) to produce primers FW1 and RV1. Design primers to amplify a short amplicon on the WT allele around the mutant allele insertion site (Figure 1).
   NOTE: The amplicon size can range from 100 bp to up to 3,000 bp, in line with the DNA size ladder used in the capillary electrophoresis. In this example, a 155 bp amplicon was amplified. The small fragment size chosen here prevents off-target amplification of the whole IS10 insertion sequence (see section 2).
2. Design a second forward primer FW2 within the insertion sequence, to produce a second amplicon that is approximately 5% larger or smaller than the WT amplicon (Figure 1). This 5% size difference is the minimum size difference that the parallel capillary electrophoresis device could reliably distinguish. Therefore, design the primers such that the two amplicons have a difference in size, which is above but as close as possible to the relative threshold.
   NOTE: Be sure to minimize the T_m difference and primer dimer formation. In this example, a second forward primer was designed to produce a 226 bp amplicon, 71 bp larger than the WT amplicon. In this representative example, the primer sequences are as follows:
   FW1:AAAGCATTTCGCCGAACGCC
   RV1: GCGATAAATCCACTCCAGCGCC
   FW2: AGTTCGCTTAGGCATGGAAG

Figure 1: Schema of triplet primer design on mutS WT gene and mutant mutS IS10 insertion. The black triangle represents the IS10 insertion site in the mutS gene. The WT gene is in blue, and the IS10 is in orange. Primers FW1 and RV1 flag the IS10 insertion site and produce a 155 bp WT amplicon. The RV1 primer and the intra-IS10 primer FW2 produce a second 226 bp amplicon. Please click here to view a larger version of this figure.

2. Optimization of PCR conditions

1. Grow an overnight culture of fixed WT and mutant allele clones.
2. Extract the DNA using any kit.
3. Quantify the DNA.
4. Prepare the sample of DNA by diluting the extraction of WT and mutant DNA to 5 ng/µL. Mix the two DNA samples in a 50/50 ratio.
5. Amplify 10 ng of the three DNA samples (WT, mutant, 50/50 mix) using a 2x ready-to-use PCR master mix, 0.5 µM FW1 primer, 1 µM RV1 primer, and 0.5 µM FW2 primer in a 20 µL reaction volume. Use a 2% agarose gel to migrate the PCR product by classical electrophoresis, and determine the optimal PCR conditions.
   NOTE: Elongation times should be minimized to prevent formation of the FW1 and RV1 amplicon on the mutant allele. The annealing temperature should be adjusted to minimize biased amplification of the alleles and non-specific amplification.
   1. To follow the program in this example, use the following settings: 98 °C for 10 s, followed by 25 cycles of 98 °C for 1 s, 58 °C for 15 s, 72 °C for 8 s, and a final elongation step at 72 °C for 1 min.
      ​NOTE: The elongation time was reduced to 8 s to prevent amplification of the >1,000 bp product from the forward primer and the RV1 primer on the mutant allele (mutS with IS10 insertion).

3. Calibration curve

1. Mix the two DNA samples, WT and mutant, in 10/90, 25/75, 40/60, 50/50, 60/40, 75/25, and 90/10 ratios.
   NOTE: Biological replicates are prepared from independent overnight bacterial cultures.
2. Amplify using optimized PCR conditions (see section 2).
3. Quantify the amplicon product.
4. Dilute the PCR products to 0.1 ng/µL.
5. Prepare the parallel capillary electrophoresis instrument.
   1. Mix fresh gel and dye (NGS quantitative analysis kit (22, 33, or 55); HS NGS fragment 1-6,000 bp for this example).
      NOTE: See the parallel capillary electrophoresis HS NGS fragment guide for detailed instructions (see the Table of Materials).
6. Replace the capillary storage solution and the inlet buffer, and place the rinse buffer plate in the correct drawer locations of the parallel capillary electrophoresis instrument.
7. Add 2 µL of HS diluent marker to 22 µL of each diluted sample in a 96-well plate.
8. Add a size ladder (DNA size ladder; range 1-6,000 bp) from a HS NGS quantitative analysis kit to one well of the 96-well plate.
9. Place the 96-well plate in the correct drawer of the parallel capillary electrophoresis instrument, and select Run on the parallel capillary electrophoresis instrument software.
10. Analyze the results using the data analysis software, which detects and identifies each peak of the size ladder, assigning each peak of the samples to their known actual size.
    1. When using a quantitative kit, use the software to determine the DNA quantity of each fragment by integrating the area under the peak, as in chromatography data analysis. Again, compare samples with known amounts in standards to quantify each peak from a sample, and calculate the ratios between the different peaks detected in the samples.
    2. Construct a calibration curve (Figure 2), linking the known proportion of the mutant allele (DNA mix) to the one measured using the parallel capillary electrophoresis instrument. This calibration curve allows the reliability of the method to be evaluated and corrected for minor amplification bias.

4. Sample preparation

1. Grow time point samples overnight in standard conditions.
2. Extract the DNA.
3. Quantify the DNA.
4. Amplify the samples using optimized PCR conditions (see section 2).
5. Run the samples in the parallel capillary electrophoresis instrument (see steps 3.5-3.10).

5. Allele quantification

1. Extract mutant allele quantities from the parallel capillary electrophoresis instrument data with the software, and calculate the actual proportions by plotting these values on the calibration curve.

Using DNA extracted from an ancestral clone and a hypermutator clone isolated from the S2.11 population at generation 1,000, we established the calibration curve shown in Figure 2. The actual mutant proportions from laboratory-prepared DNA mixes and measured by the parallel capillary electrophoresis instrument were linked by a linear relationship of slope 1.0706, with an R² of 0.9705. Additionally, there was a good agreement between biological replicates; the standard deviation was between 0.61 and 17.74 across the nine points of the standard curve.

Figure 2: Calibration standard curve. Observed versus expected proportion of WT/mutant DNA mix. Error bars represent the standard error between biological replicates. Please click here to view a larger version of this figure.

DNA was extracted from 19 time points ranging from generation 356 to generation 990, amplified and quantified by parallel capillary electrophoresis. Prosize software was used to identify and quantify each amplicon in each sample. These results were converted to allele proportion using the calibration curve.

The results shown in Figure 3 reveal a nonmonotonic trajectory of the hypermutator-inducing mutS allele. Generation 680 (day 102) was the first time point at which the mutant allele was in sufficient quantity to be detected by parallel capillary electrophoresis. The mutant allele then increased rapidly in frequency and reached 66.7% at generation 713 (33 generations later). The mutant allele frequency then stagnated and was 76% by generation 766, despite having previously increased rapidly in frequency. Unexpectedly, the mutant allele frequency then reduced from 76% to 49% over the course of 13 generations (2 days), after which it increased to fixation.

Figure 3: Dynamics of ancestral (WT) and derived (mutS::IS10) alleles over the course of a 1,000-generation experimental evolution. The ancestral allele frequency is in gray, and the derived allele frequency is in blue. Please click here to view a larger version of this figure.

Figure 4: Primer design and amplicon position for quantification of the four types of structural variants. Amplified fragments are depicted in black and white. Arrows represent primers at their hybridization site; dotted arrows represent primers hybridizing but not participating in any fragment amplification, either because there is no paired primer (duplication and inversion) or the elongation time has been reduced to avoid the amplification of large amplicons (insertion and deletion). Please click here to view a larger version of this figure.

Here, we have proposed a cost-effective method that allows the dynamics of emerging adaptive SV alleles in experimental evolution populations to be followed. This method couples classic PCR techniques and automated parallel capillary electrophoresis, allowing for the relative quantities of two alleles to be determined. Once set up, it permits the quantification of allele proportions in many samples in parallel, and is much less expensive than WGS. This method can be seen as an equivalent to amplicon sequencing for non-SNP mutations and as a solution to the technical limitations of amplicon sequencing of large SVs. We have shown that the method accurately quantifies the proportion of non-SNP mutants, using a calibration curve that overcomes PCR bias (Figure 2).

Characterizing SV allele dynamics can be useful to identify linked mutations through their parallel temporal dynamics, or to calculate selection coefficients and identify their variations across time. In the example of application we have presented here, we identified, by end-point WGS, that hypermutators were segregating or fixed in various experimental evolution E. coli populations¹⁸. These populations were considered hypermutator as they had disruptive mutations (IS insertion) in one of the methyl mismatch repair (MMR) genes and had accumulated many more mutations over the 1,000 generations of experimental evolution than the ones not carrying mutations in the MMR genes (24.38 ± 1.75 vs. 126.7 ± 19.57). Mutations in MMR genes are known to lead to an increased mutation rate, allowing bacteria to sample many more mutations than their baseline mutation rates permit^12,13. However, a large increase in mutation rate also translates to a faster accumulation of deleterious mutations and a greater mutational load^14,15. Mutators have been predicted to be transiently advantageous in populations adapting to new environments, as the proportion of beneficial mutations is larger in this situation. The beneficial genetic backgrounds generated under high mutations rate allows mutators to propagate through a population by hitchhiking on the beneficial mutations they generate¹⁹. Temporal dynamics of hypermutator propagation have been investigated by simulations, and with the method we propose here, it is possible to follow the dynamics of IS-mediated mutator alleles in experimental populations¹⁹.

In our representative results, the SV is an IS insertion, but the method described here can be easily extended to the other types of SV (see Figure 4). The triplet primer design follows exactly the same logic for deletion quantification as for insertion quantification: two primers sit around the deletion point and the third primer sits within the deleted sequence. For inversions and duplications, two of the primers sit around one of the breakpoints and the third one sits inside the inverted/duplicated fragment, close to the other breakpoint and directed toward the outside of the fragment. For insertions, deletions, and inversions, each amplicon is specific to one of the alleles, ancestral or derived. For duplications, one of the amplicons is common to the two alleles, and the other one is specific to the derived allele. This has to be considered when calculating the proportion of alleles from the calibration curve.

Finally, we have developed this method to follow an SV in experimental evolution populations of bacteria, for which "metagenomic" sequencing at the population level is common practice. However, this method can also be applied to experimental evolution populations of multicellular organism in a pool-seq approach, or to intra-host viral or microbial pathogen sampling¹⁷.

This method has some limitations. First, the alleles being compared must differ in size by at least 5%. This prevents us from following the dynamics of alleles produced by very small insertions or deletions. Fortunately, these allele dynamics may be followed by amplicon sequencing. Second, the sequence of the derived allele must be known to allow for the design of the primers around the insertion or deletion or breakpoints. Third, primer design and optimization are necessary for each site. Thus, this method is appropriate for following a limited number of target allele pairs of interest. If many alleles are to be followed in the same population, we suggest NGS of multiple time points.