Single-cell copy number variant detection reveals the dynamics and diversity of adaptation

Abstract

Copy number variants (CNVs) are a pervasive source of genetic variation and evolutionary potential, but the dynamics and diversity of CNVs within evolving populations remain unclear. Long-term evolution experiments in chemostats provide an ideal system for studying the molecular processes underlying CNV formation and the temporal dynamics with which they are generated, selected, and maintained. Here, we developed a fluorescent CNV reporter to detect de novo gene amplifications and deletions in individual cells. We used the CNV reporter in Saccharomyces cerevisiae to study CNV formation at the GAP1 locus, which encodes the general amino acid permease, in different nutrient-limited chemostat conditions. We find that under strong selection, GAP1 CNVs are repeatedly generated and selected during the early stages of adaptive evolution, resulting in predictable dynamics. Molecular characterization of CNV-containing lineages shows that the CNV reporter detects different classes of CNVs, including aneuploidies, nonreciprocal translocations, tandem duplications, and complex CNVs. Despite GAP1's proximity to repeat sequences that facilitate intrachromosomal recombination, breakpoint analysis revealed that short inverted repeat sequences mediate formation of at least 50% of GAP1 CNVs. Inverted repeat sequences are also found at breakpoints at the DUR3 locus, where CNVs are selected in urea-limited chemostats. Analysis of 28 CNV breakpoints indicates that inverted repeats are typically 8 nucleotides in length and separated by 40 bases. The features of these CNVs are consistent with origin-dependent inverted-repeat amplification (ODIRA), suggesting that replication-based mechanisms of CNV formation may be a common source of gene amplification. We combined the CNV reporter with barcode lineage tracking and found that 102-104 independent CNV-containing lineages initially compete within populations, resulting in extreme clonal interference. However, only a small number (18-21) of CNV lineages ever constitute more than 1% of the CNV subpopulation, and as selection progresses, the diversity of CNV lineages declines. Our study introduces a novel means of studying CNVs in heterogeneous cell populations and provides insight into their dynamics, diversity, and formation mechanisms in the context of adaptive evolution.

Fig 1. Fluorescent protein signal is proportional to gene copy number.

(A) Protein fluorescence increases with increasing copies of the mCitrine gene. We determined the fluorescence of haploid and diploid cells containing variable numbers of a constitutively expressed mCitrine gene integrated at either the HO locus and/or the dubious ORF, YLR123C. The two-copy diploid is heterozygous at both loci. Each distribution was estimated using 100,000 single-cell measurements normalized by forward scatter. (B) Schematic representation of how the fluorescent reporter enables CNV detection in heterogeneous evolving populations through quantitative changes in protein fluorescence. Data and computer code used to generate this figure can be accessed in OSF: https://osf.io/fxhze/. a.u., arbitrary units; CNV, copy number variant.

Fig 2. Dynamics of GAP1 CNVs in evolving populations.

(A) Normalized distributions of single-cell fluorescence over time for a representative GAP1 CNV reporter strain and one- and two-copy control strains evolving in glutamine-limited chemostats. Single-cell fluorescence is normalized by the forward scatter measurement of the cell. (B) Normalized median fluorescence for each population evolving in glutamine- (n = 9), urea- (n = 9), and glucose-limited (n = 8) chemostats. The fluorescence of the one- and two-copy control strains is plotted for reference (gray dotted lines). (C) Estimates of the proportion of cells with GAP1 amplifications over time for nine glutamine-limited populations containing the GAP1 CNV reporter. Data and computer code used to generate this figure can be accessed in OSF: https://osf.io/fxhze/. a.u., arbitrary units; CNV, copy number variant.

Fig 3. Diversity and fitness effects of GAP1 CNVs.

(A) Representative sequence read depth plot from a glutamine-limited clone (gln_01_c4). The nucleotide coordinates of GAP1 in our CNV reporter strain are Chromosome XI: 518438–520246 (blue line). Estimated breakpoint boundaries are shown in red. Read depth was normalized to the average read depth on Chromosome XI. Reads at each nucleotide position were randomly downsampled for presentation purposes. (B) Read depth–based estimates of GAP1 copy number are positively correlated with median fluorescence of glutamine-limited clones, indicating that fluorescence is informative about the copy number of de novo CNVs. (C) Schematic representation of CNVs identified in clones isolated from glutamine-limited populations. The relative fitness of each clone is also indicated. Copy number and CNV boundaries were estimated using read depth. This schematic is simplified for presentation purposes: the reported copy number refers specifically to the GAP1 coding sequence and does not necessarily reflect copy number throughout the entire CNV, which may vary. For read depth measurements across the entirety of Chromosome XI, see S2 Text. Data and computer code used to generate this figure can be accessed in OSF: https://osf.io/fxhze/. CNV, copy number variant; LTR, long terminal repeat; N/A, not applicable; g150, generation 150; g250, generation 250.

Fig 4. Inverted repeats mediate CNV formation.

Nucleotide (“nt”) resolution of CNV breakpoints for (A) GAP1 and (B) DUR3 CNVs were identified using a combination of discordant and split reads. To characterize novel sequence, we identified all supporting split reads, performed de novo assembly, and aligned the resulting sequence against the reference genome. Sequences in the reference genome (blue) are inversely oriented in the assembled contig, suggesting an inverted structure within CNVs. (C) Schematic representation of replication-based CNV formation. After fork stalling, fork regression results in the newly replicated inverted repeat sequence annealing to the complementary sequence and ligating to the lagging strand. (D–E) Distribution of sequence features across 28 breakpoints at the GAP1 and DUR3 loci that contain inverted repeats. Data and computer code used to generate this figure can be accessed in OSF: https://osf.io/fxhze/. CNV, copy number variant.

Fig 5. Lineage tracking reveals extensive clonal interference among CNV-containing lineages.

(A) We used FACS to fractionate cells containing GAP1 CNVs from two populations at four time points (dashed black lines) and performed barcode sequencing. (B) Using a sample- and time point–specific false positive correction, we identified 7,067, 973, 131, and 76 barcodes in one population (bc01; left) and 5,305, 5,351, 583, and 28 barcodes in another population (bc02; right), at generations 70, 90, 150, and 270, respectively. Each barcode found at >1% frequency in at least one time point is represented by a unique color in the plot, for a total of 21 barcodes in bc01 and 18 barcodes in bc02. All other lineages that are never detected at >1% frequency are shown in gray. Lineages denoted by a * are found at >1% frequency in both populations. Data and computer code used to generate this figure can be accessed in OSF: https://osf.io/fxhze/. CNV, copy number variant; FACS, fluorescence-activated cell sorting.