Gene amplification during chemostat growth of yeast in sulfate-limiting medium

A chemostat is a culture vessel with ports for the continuous, controlled addition of media (and air) and an overflow valve to collect spent medium and a representative sample of the cells from the chemostat. By limiting a single nutrient, cells can be grown continuously under stressful conditions for hundreds of generations [13]. The rate of medium inflow sets the growth rate of the culture, but individual cells that have acquired a beneficial mutation, specific to the imposed selection, will eventually sweep the population as it gives their descendants a survival advantage [1]. By sampling cells from the overflow valve, we can follow the progression of different amplification events over time.

Among the many sulfur-limiting chemostats we have run with a prototrophic haploid yeast strain (FY4), we invariably recover cells that have amplified the sulfur transporter gene SUL1 [1,2,4,14,15]. An example of these analyses is shown in Fig 1A–1D. We detect changes in karyotype by CHEF gel electrophoresis, a form of electrophoresis that separates intact yeast chromosomes. In the ethidium bromide-stained gel (Fig 1A), there is an obvious change in the size of chr II on day 13, followed by its slow disappearance over the course of the remaining days. Hybridizing the Southern blot with a CEN2 probe (Fig 1B) confirms the identity of the changed chromosome and explains its apparent disappearance—with time, a second, larger version of chr II that co-migrates with chr XIII and XVI outcompetes cells with the initial chr II variant. By probing the same blot with SUL1 (Fig 1C) and comparing the hybridization intensity of SUL1 relative to CEN2 on the same chromosome, we estimate that the number of copies of SUL1 present on chr II progress from 1 to 3–5, without passing through copy numbers of 2 and 4.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Detection of gene amplification during sulfate-limited chemostat growth.

To follow karyotypic changes during ~200 generations (33-34 days) of growth in chemostats limiting for sulfate we ran Contour-clamped Homogenous Electric Field (CHEF) gels and performed array Comparative Genome Hybridization (aCGH). Shown are examples of changes occurring in wild type (A-D) and ars228Δ (E-H) strains. A) Ethidium bromide-stained CHEF gel of the wild type population sampled at intervals across 33 days of growth. The expected position of chr II is indicated by the red arrow; for identities and sizes of other chromosomes see Fig 3G. Hybridizations of the CHEF gel with probes for CEN2 (B) and SUL1 (C) reveal step-wise changes in the size of chr II. The comparison of hybridization intensity of SUL1 relative to CEN2 indicates discrete increases in the number of copies of SUL1 from 1 to 3 to 5 over time. D) ArrayCGH of clone 1490 (blue arrow in panel C) from day 33 reveals the region of chr II that is included in the amplification event (~779 to 802 kb) and the number of copies (~5; log2 ratios of 0, 1, 2, 3, 4 equate to copy numbers of 1, 2, 4, 8, 16.) This pattern of amplification is consistent with tandem repetition of the amplicon that alternates in orientation, leaving the distal chromosomal sequences intact. E) The ethidium bromide-stained CHEF gel of the ars228Δ mutant appears similar to that of the wild type strain, but hybridizations with the CEN2 (F) and SUL1 (G) probes indicate the persistence of multiple forms of chr II in the population. H) ArrayCGH of clone 1475 derived from day 33 (blue arrow in panel G) shows amplification of a SUL1 fragment that extends through the telomere and is present in five copies. I) Nanopore sequencing of clone 1475 reveals a single junction that joins the amplified fragment to the right telomere of chr II. Individual long reads confirm multiple tandem repeats.

https://doi.org/10.1371/journal.pgen.1012214.g001

To determine the extent of amplified flanking sequences we performed array comparative genome hybridization (aCGH, Fig 1D). In this example, we isolated a single clone (1490) from the homogeneous final population and compared genome-wide copy number of the clone to the same strain on day 0. The amplified sequences include a ~ 25 kb fragment that contains the SUL1 locus that is present at ~5 copies (Fig 1D). These additional ~100 kb are consistent with the size of chr II measured on the last day in the CHEF gel. These five copies form a tandem array of repeating units that alternate in their orientation and create junctions of head-to-head and tail-to-tail configurations similar to those found by Araya et al. [16]. In a recent study of 31 independent sulfate-limited chemostats [2] we confirmed that inverted amplicons are the predominant form of SUL1 amplification in wild type haploid laboratory yeast strains. Among the short-read sequences from the 31 populations analyzed on the final day (day 33–35), we observed 50 unique junctions mapping between CEN2 and SUL1 that were consistent with a genomic inversion at short, closely spaced inverted repeats and only a small number of other types of junctions.

As a test of our proposed model for how these inverted amplicons occur (ODIRA, [3]) we had previously deleted the origin of replication at the 3’ end of SUL1 (ARS228) and grown the ars228Δ strain in seven independent sulfate-limited chemostats [4]. Four of the seven cultures showed SUL1 amplification patterns consistent with ODIRA, but there were three exceptions that we had not characterized in detail. In one of those three ars228Δ chemostats we observed chr II variation (Fig 1E–1G) with several copy number versions (2, 5, and 10 copies) persisting through the last day. The single clone (1475) from day 33 of the ars228Δ chemostat that we examined by aCGH contains a similarly sized amplicon to the wild-type clone (1490), but in this case, the amplicon extends through the telomere (Fig 1H). To validate this presumed amplicon structure and to identify the precise location of the amplicon boundary, we performed long read (Oxford Nanopore) sequencing of this clone’s genome. We found a single split-read at the site that corresponds with the jump in copy number in the aCGH profile. It is a junction between the telomere of chr II and a unique sequence centromere-proximal to SUL1 (Fig 1I) and is the same type of junction we found among four of the short-read sequences from the 31 chemostats of wild type haploids in which ODIRA events had predominated (S1 Fig; [2]). Several observations are consistent with the terminal, tandem repeat structure for this clone proposed in Fig 1I: the only chromosome with an altered size on the CHEF gel was chr II; sequencing returned only a single junction sequence; and individual long reads spanned multiple, tandem, direct repeats of the SUL1 region. Therefore, deleting the origin adjacent to SUL1 results in changes in the subtelomeric region that cannot be explained by ODIRA.

Because the non-ODIRA amplicons had junctions between a chromosomal telomere and internal telomeric sequences, we asked if a mutation affecting telomere maintenance would alter the cellular response to growth in sulfate limited conditions. Among the many proteins that bind yeast telomeric sequences is the Ku70/80 complex, which is important for telomere maintenance and genome stability [17]. Ku70/80 binds directly to chromosome ends, helping to protect them from degradation by exonucleases and from end-to-end fusion events. Ku70/80 also binds to telomeric DNA through its interactions with Rap1, Sir4 and the RNA component of telomerase (Tlc1) and influences telomere maintenance [18,19]. Cells lacking either component of the Ku complex have shorter telomeric G_1-3T tracts and show increased levels of genome instability [20–23].

After long-term growth of a yku70Δ strain in two sulfate-limited chemostats, we analyzed DNA from the populations by CHEF gel electrophoresis, Southern blotting and aCGH (yku70Δ-A and-B; Fig 2). The ethidium bromide-stained CHEF gels (Fig 2A and 2D) and their Southern analyses (Fig 2B and 2E) suggest that multiple amplicons with variable numbers of copies of SUL1 (numbers with dashed red lines) had persisted through the last day of chemostat growth. (The identity of “Chr?” is likely either chr XIII or chr XVI since that doublet decreases in intensity in the second half of the time course.) aCGH of the populations (Fig 2C and 2F) confirm the existence of multiple unique amplicon junctions—a minimum of four in yku70Δ-A and two in yku70Δ-B. From these analyses it is unclear how the different amplicons are distributed across the different versions of chr II found in the CHEF gels.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Deletion of yku70Δ results exclusively in terminal amplification of SUL1.

Chemostats yku70Δ-A and yku70Δ-B contained independent evolutions of the yku70Δ mutant. The ethidium bromide-stained CHEF gels (A and D) and their Southern analyses using CEN2 and SUL1 as probes (B and E) identify chr II and measure the extent of SUL1 amplification, respectively. aCGH of the populations (C and F) indicate average copy numbers of ~8 and ~7 for the yku70Δ-A and -B populations, respectively, with multiple junctions.

https://doi.org/10.1371/journal.pgen.1012214.g002

The aCGH data were noteworthy in that the amplicons not only extended through the right telomere of chr II but together they represented a minimum of six different junctions. In chemostats with wild type cells it is rare to find more than two unique junctions [2]. The multiple amplicons in chemostat yku70Δ-A appeared to arise simultaneously each with five or more copies of the SUL1 region, and one amplicon appeared to be located on a chromosome other than chr II (Fig 2B). Because chr II is the primary chromosome to show a significant change in size on the CHEF gel, we concluded that the majority of these different chr II fragments must be appended to or incorporated into chr II (Fig 2C). One explanation for the arrangement of the multiple amplicons is that each of the fragments is found in a different subset of cells in the population. A second possibility is that there is a single event that contains a mixture of different-sized repeats. The results from chemostat yku70Δ-B (Fig 2D–2F) were similar to those of yku70Δ-A.

To disentangle the different amplicon sizes found in the yku70Δ chemostats, and to see if they were changing over time, we isolated clones from day 17 and day 33 from yku70Δ-A (Fig 3A) and days 13, 17 and 33 from yku70Δ-B (S2A Fig). We found 33 clones with unique karyotypes, and all but one had additional copies of SUL1. By comparing the ethidium-stained gel with the hybridization results for CEN2 (Figs 3B and S2B) and SUL1 (Figs 3C and S2C) we observed that 1) all clones had only a single copy of chr II with some clones retaining the parental-sized chr II; 2) clones with the parental-sized chr II had amplified versions of SUL1 on some other chromosome; and 3) many chromosomes had increased in length but the increase was not due to SUL1 amplification. We conclude that the multiple amplicons seen in Fig 2 reflect a mixed population of cells, with different subsets having different amplicons. The explanation for the wide-spread changes in chromosome size was revealed by hybridization with a Y’ probe (Figs 3D, S2D and S3): many chromosomes had amplified copies of subtelomeric Y’ sequences and on average, by 33 days cells had doubled their Y’ content (from ~16 copies to ~30 copies; S3 Fig). All unique bands on the CHEF gel can be accounted for by either SUL1 or Y’ amplification or both. That these changes are selected for by growth in the sulfate-limited chemostat is illustrated by parallel analysis of independent clones from the day 0 sample (Fig 3E–3G). Among these 18 clones there is only a single chromosome with a modest increase in Y’ hybridization (Fig 3G; red star).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Clones from the yku70Δ-A chemostat have highly variable karyotypes.

To determine the structure of the four terminal amplicons in the yku70Δ-A chemostat and yku70∆-B (S2 Fig) we isolated clones from different days and analyzed them by CHEF gels and Southern hybridization. We compared these clones with 18 yku70Δ clones isolated from day 0. The red arrows in each panel indicate the expected position of chr II and the hash marks mark the position of other chromosomes, identified in (G). The ethidium bromide-stained CHEF gels (A and E) were analyzed by Southern blotting using probes from CEN2 (B), SUL1 (C and F) and the Y’ from chr V-R (D and G). The CEN2 probe identifies which chromosomes are chr II. In ten of the evolved clones (B; boxed in red) chr II is significantly larger and each has more than 1 copy of SUL1 (C). The SUL1 probe also identified other chromosomes that contain the SUL1 region (C; bands boxed in cyan are not chr II); however, there are other chromosomes with altered mobility that are not accounted for by SUL1 addition (e.g., band marked with * in A). Hybridization with the Y’ probe reveals multiple additions in the evolved clones (D) but only a single Y‘ addition in the day zero clones (G; red star). All other day 0 clones have the expected distribution of Y’ elements (chr I, III, and XI, in gray type, do not have a Y’ on either end.).

https://doi.org/10.1371/journal.pgen.1012214.g003

Long read sequencing confirms SUL1 amplicon structures in the yeast yku70Δ mutant

To identify the precise sequence of the amplicon junctions generated in the yku70Δ mutant strain, we performed Oxford Nanopore sequencing on the population sample from day 34 of yku70Δ-A. We found split reads (Fig 4A) at the same coordinates that flanked the discontinuities in copy number in the aCGH data (Fig 2C) along with four additional low-frequency junctions on chr II-R (Fig 4B). At seven of these genomic locations in the reference genome (Saccharomyces Genome Database, SGD) are short stretches of telomeric repeats (G_1-3T/C_1-3A) and at the eighth junction is a sequence homologous to a region just 45 bp in from the right telomere of chr II (Fig 4B). From the long reads we determined the structure of the amplicons associated with each junction: most long reads revealed tandem repeats of the telomeric SUL1 fragment appended directly to the right telomere of chr II (Fig 4C and 4D). Lower frequency events involved the addition of a Y’ to the SUL1 amplicon (Fig 4E) or translocation of the terminal SUL1 fragment to a different chromosomal telomere (Fig 4F). For the sequences summarized in Fig 4D, the preexisting ITSs had undergone an expansion from just a few base pairs to up to nearly 220 bp (Fig 4B) and the first sixteen base pairs of the expanded ITS were identical to the first sixteen base pairs of the chr II-R telomere (5’-GGGTGTGTGGGTGTGG-3’). This sequence identity suggests that the native chr II-R telomere was providing the additional G_1-3T sequence at the SUL1 amplicon junctions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Nanopore sequencing of the yku70Δ-A chemostat population.

A) A read-depth plot of the split reads from Oxford Nanopore sequencing of the yku70Δ-A day 34 population was consistent with the positions of copy number change that were identified in the aCGH profile (Fig 2C). B) We identified eight unique split reads from this region of chr II and found that seven of the eight junctions occurred between telomeric G_1-3T sequences and an internal sequence centromere proximal of SUL1. At the seven internal sites were unique, short regions of G_1-3T homology (an internal telomeric sequence, ITS). The eighth junction occurred between an internal sequence that had homology to a sequence 45 bp into the subterminal region of chr II-right. The reference genome coordinate of the junction sequence, the number of supporting reads, the length of the expanded ITS and the numbers of repeats on the longest reads are given. C) Illustration of the telomere structure of four of the chromosomes in the yku70Δ parent strain. The colors of the Y’s are from the genome-wide analysis shown in S4A Fig. Long reads across the eight junctions provided information on the identity of the chromosome that had acquired the SUL1 telomeric fragment, the number of repeats added, and the chr II coordinate of the amplicon junction (numbers in kb to the right of each cartoon). They fell into three classes: D) terminal addition of the SUL1 fragment directly to the telomere or sub-telomere of chr II-R; E) terminal addition of the SUL1 fragment with interspersed Y’s from other chromosome locations; and F) translocations of the terminal SUL1 fragment to the telomeres of other chromosomes. The expanded ITSs in D and E began with 5’-GGGTGTGTGGGTGTGG-3’, the same sequence found at the junction between chr II’s right telomere and the adjacent subtelomeric DNA.

https://doi.org/10.1371/journal.pgen.1012214.g004

Structure of Y’ amplicons in the yku70Δ mutant

Since all clones from days 17 and 33 of the yku70Δ-A evolution had one or more chromosomes with additional copies of Y’ (Fig 3D), we searched Nanopore split reads near telomeres and found many instances of tandem arrays of Y’s. We wanted to address three questions: 1) is there a specific Y’ that is amplified; 2) within an array, how homogeneous are tandem Y’s and expanded ITSs; and 3) how are amplification of Y’ and SUL1 related? To classify Y’s and assign them accurately to chromosomes, we defined a section of the Y’ region (bounded by specific sequence tags common to all Y’s curated in SGD; “trimmed Y’s”) and determined alignments and phylogenetic relationships of the Y’s in the reference genome (sacCer3) with those in FY4 and the yku70Δ parent (S4 Fig). These comparisons defined 13 Y’ classes that we could assign to specific chromosomes (S4 Fig). We then used the set of trimmed Y’s from FY4 to compare individual Y’s from tandem amplicons using multiple alignments and phylogenetic trees.

Examples of four Nanopore reads that contain tandem Y’ repeats are diagrammed in Fig 5. The first three reads (Fig 5A–5C) are composed of homogeneous Y’s that match the Y’ that is resident on that specific chromosome. The fourth read (Fig 5D) is a repeating trimer of two distinct Y’s. Because this read (as well as read 3) did not contain unique sequences, we could not definitively map it to a particular chromosome. However, based on the phylogenetic trees (S5 Fig) we generated for the Y’s from each of these four reads we can conclude that the sources of the repeating Y’s are from four different telomeres.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Tandem arrays of Y’s and expanded ITSs are homogeneous and are derived from the resident telomere.

In addition to junctions involved in SUL1 amplification, another major form of junction we identified in yku70Δ-A involved tandem Y’ elements separated by expanded ITSs of homogeneous lengths. A-D) Individual Nanopore reads 1 (37672 bp), 2 (111251 bp), 3 (50690 bp) and 4 (59060 bp) are illustrated along with the Y’s from the parental chromosomes for comparison. Square brackets in B and D: sequences that were repeated n times as tandem blocks. Multiple alignments of the expanded ITSs in Reads 1-3 (S6A–S6D Fig) and Read 4 (E) indicate they are unique to each amplicon and homogeneous across the Y’ array. The identities of Y’s (bottom right) were derived by multiple alignments and phylogenetic trees (see S5A–S5D Fig).

https://doi.org/10.1371/journal.pgen.1012214.g005

Not only are the contiguous Y’s homogeneous, the expanded ITSs also appear to be related. The similarity is reflected in the lengths of the expanded ITS (Fig 5A–5D) but is also confirmed by sequence alignments (Figs 5E and S6). There is striking sequence similarity within the groups of tandem, expanded ITSs (S6A–S6C Figs) but each forms clear clades in the phylogenetic analysis (S6D Fig). From these analyses, we conclude that there is no obvious preference for which Y’ is amplified; however, the resident Y’ and its adjacent telomere appear to act as the seed for generating homogeneity in the tandem array of Y’ and expanded ITSs.

Model for formation of telomeric tandem arrays

While unequal sister chromatid exchange is an obvious mechanism for producing tandem sequence arrays, the short telomeric tracts at the sites of both SUL1 and Y’ amplicons in the ars228Δ and yku70Δ-A and -B mutants suggest a mechanism that involves microhomology such as microhomology-mediated break induced replication (mmBIR). As the name suggests, a break is the initiating event. Telomeres are not usually seen as DNA breaks as they are protected by protein complexes that shield them from exonucleases and end-recognizing machinery that would initiate DNA repair. However, in the yku70Δ mutant, telomeres have lost a major protective protein and fail to efficiently recruit telomerase. As a result, the G_1-3T telomeric tracts undergo drastic shortening (from a mean of 539 bp in the wild-type to 204 bp in the yku70Δ parent; our Oxford Nanopore data). It is the short, transiently unprotected 3’ overhangs that we propose are being recognized as “breaks” and are initiating an mmBIR process.

In Fig 6A we outline an mmBIR event initiated by an unprotected telomere. It is the proposed mechanism for generating type-I survivors in telomerase null mutants (reviewed in [11,12]); however, it does not provide an explanation for why the sites of the tandem junctions or their intervening, expanded ITSs would be related by descent. With these and other concerns in mind, we propose a version of mmBIR that occurs “in cis” (Fig 6B). In summarizing the possible structural changes that could arise from mmBIR, Hastings et al. [24] describe rolling circle replication as a consequence of template switching in cis after replication fork collapse. Here, we are suggesting that the right telomere of chr II invades an ITS lying centromere proximal of SUL1 to initiate conservative replication. When the migrating bubble reaches the initial site of telomere invasion, it now encounters the extended ITS created by that first invasion and continues through the junction sequence into the newly appended repeat. There is technically no end in sight as the BIR bubble continues to re-replicate the sequences just added. One way in which replication could terminate is for the migrating bubble to encounter a single stranded nick on the C_1-3A strand. This model explains why the SUL1 amplicons are preferentially found on chr II, why the expanded ITS sequences in tandem repeats are identical (or nearly so), why amplicons suddenly appear with five or more repeats (as in yku70Δ-A, Fig 2A and 2B) and why tandem Y’s are often identical and related to the resident Y’ instead of reflecting the natural Y’ variation seen across the genome.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Models that account for tandem telomeric addition.

The short stretches of G_1-3T sequences (ITSs) at the sites of both SUL1 and Y’ amplicons in the yku70Δ mutant suggest a form of break induced replication that is initiated by microhomology (microhomology mediated break induced replication—mmBIR). A) mmBIR in trans: When an unprotected telomere (red arrow) invades a sister chromatid, homologue or a non-homologous chromosome at an ITS (red-cyan duplex) the BIR machinery initiates conservative replication that proceeds to the end of the chromosome through a migrating replication bubble. This mechanism can explain the production of a single tandem copy if the invading telomere is from the sister chromatid, or can explain translocations of a Y’ or a SUL1 fragment when a chromosome other than chr II-R provides the invading telomere. B) mmBIR in cis: When the unprotected telomere invades an ITS on the same chromosome arm it also creates the second copy of SUL1 and adjacent sequences. However, when the bubble reaches the site of the old telomere, it now encounters the junction created by the invasion and simply copies it and the appended sequences a second, third, or more times. While there is technically no circular DNA involved in this mechanism, it is similar to rolling-circle replication identified in some bacterial plasmids and the yeast 2-micron plasmid during amplification [34]. Hence, we refer to this form of mmBIR as “pseudo rolling-circle mmBIR”.

https://doi.org/10.1371/journal.pgen.1012214.g006

Deletion of genes involved in DNA metabolism affects the mode of SUL1 amplification

The results from ars228Δ and yku70Δ chemostats suggested that the various pathways of SUL1 amplification are under genetic control and that mutations in specific genes reduce the incidence of ODIRA events. We expanded our sulfate-limited chemostat analysis to include mutant strains with deletions in genes involved in DNA metabolism (results for nine of the mutants are shown in Figs 7 and S7). We ran two (or four) independent chemostats for each of the mutants, examined karyotypic changes occurring during chemostat growth by CHEF gels combined with hybridization to a SUL1 probe (S7 Fig) and performed aCGH on the population from the final day of growth (day 33–35; Fig 7). One of the mutants (exo1Δ) failed to amplify SUL1 or any other unique genomic region. Most populations appeared to have a single amplicon present on the last day. To confirm the form of the amplicons, we had the populations sequenced using Oxford Nanopore long reads and discovered that, in addition to inverted amplicons, there were direct amplicons, ones where the amplified region extended through the right telomere, two translocation events and one extrachromosomal, inverted linear fragment where the junction was composed of sequences from the endogenous 2-micron plasmid (Fig 7, cyan, purple, red, orange, green bars, respectively; S1 Table). Only the two chemostats of the rad5Δ mutant (Fig 7) produced exclusively inverted repeats. The telomeric amplicons in the chemostats of msh2Δ, sae2Δ, sgf73Δ and pol32Δ had junctions with expanded ITSs (S1 Table) formed by invasion of a telomere at pre-existing, short ITSs.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Characterization of chemostat populations of nine gene knockouts by aCGH.

Population samples from the final day of chemostats with nine different gene knockouts were examined by array CGH. With the exception of exo1Δ, each population generated one or more SUL1 amplicons (SUL1 is marked by the vertical yellow bar). The amplified regions were mostly restricted to the terminal 50 kb (exceptions msh2Δ-A and rad54Δ-A) and most cultures had resolved to a single predominant amplicon by the end of the experiment (except for chemostats msh2Δ-A, sae2Δ-B, pol32Δ-A, and rad52Δ-B). All junctions were precisely identified by Oxford Nanopore sequencing (see S1 Table) and are indicated by chr II coordinates (in kb). Colored bars indicate amplification through the telomere (red), inverted (ODIRA) amplicons (cyan), direct repeats (purple), translocations (orange) and free linear fragments (green).

https://doi.org/10.1371/journal.pgen.1012214.g007

To quantify the effects of different mutations on the ODIRA pathway of gene amplification, we moved four of the mutations (yku70Δ, sae2Δ, rad5Δ, and rad54Δ) into our test strain (s2-1) that harbors a split-URA3 cassette (S8A Fig; [2]). The position of the two overlapping partial ura3 fragments generates a specific chromosomal rearrangement when a URA3 gene is produced by an ODIRA-like event (S8B Fig). We chose these four mutants because they spanned the range of only ODIRA events (rad5Δ), only terminal events (yku70Δ), and a mixture of four forms of amplification (sae2Δ and rad54Δ). In wild type cells, the frequency of ODIRA events is 22% (11/50 ura+ clones; [2]). ODIRA events were significantly reduced in three of the mutants (rad5Δ, yku70Δ, rad54Δ) and increased slightly in the sae2Δ mutant (S8C Fig).

We cannot say definitively, without further work, how the different mutant genes we tested might be participating in the ODIRA or mmBIR pathways. The failure to recover amplicons in the exo1Δ chemostats may be coincidence or it could mean that Exo1 is essential for gene amplification. In addition, we cannot say whether Yku70 or Sgf73 (whose deletions produced only telomeric SUL1 amplicons, Figs 2 and 7) are required for the ODIRA mechanism or their absence makes telomeric invasion more likely, or perhaps both. It would require measuring absolute rates of these two mechanisms in the chemostats—something that we are currently unable to do experimentally.

Test of the cis-mmBIR mechanism for tandem addition of subtelomeric sequences

The genetic requirements for mmBIR are well established (reviewed in [11,12]). In particular, mmBIR is dependent on the DNA polymerase that contains a subunit encoded by POL32. Since the two pol32Δ chemostats (Fig 7) both produced amplicons of SUL1 that were terminally duplicated, it would appear on the surface that our model cannot be correct; however, these results are nevertheless compatible with our model, as terminal duplications could also be created by non-homologous end joining or by unequal sister chromatid exchange. To determine to what extent Pol32 was involved in terminal duplications we tested a pol32Δ yku70Δ double mutant. To simplify the analysis and avoid the multiple sulfate-limited chemostats that would be required, we employed a batch culture protocol to ask whether simply passaging strains would produce the same unstable Y’ phenotype. We performed three technical replicates for each of four strains (wild type, yku70Δ, pol32Δ, and pol32Δ yku70Δ), growing them for 22 consecutive days (~130 generations) in the same sulfate-limiting medium that we had used in the chemostats. Because batch culture cannot reproduce the same selective conditions for SUL1 amplification that a chemostat can, we did not expect to see SUL1 amplification (and we recovered none) but we were able to reproduce the Y’ instability.

At the end of the 22 days of culture, we isolated 12 clones from each of the 12 cultures and performed CHEF gels and Y’ hybridizations of the Southern blots (Figs 8A–8D and S9). It is clear by inspecting the Y’ autoradiograms of the yku70Δ Southern blots that there are no parental karyotypes remaining and that many of the chromosomes with Y’ additions have a very significant increase in the amount of Y’ hybridization on most of the altered chromosomes relative to the hybridization levels on chromosomes in the other three strains (Figs 8A–8D and S9). We identified a total of 7 of 34 clones with Y’ additions in the wild type strain and 34 of 34 in the yku70Δ strain (S9 Fig). This result clearly indicated that continuous passaging of the yku70Δ strain was sufficient to result in Y’ amplification. In contrast, among the 35 unique clones of the pol32Δ yku70Δ strain only eight had an altered chromosome size, a frequency that was comparable to the number of clones we found in the pol32Δ strain (S9 Fig). Thus, deleting POL32 strongly reduces the Y’ amplifications caused by deletion of YKU70, underscoring the important role that Pol32 plays, specifically, in the pathway that occurs when telomeres are destabilized by the absence of the Ku heterodimer. Nevertheless, the fact that there were non-zero numbers of terminal Y’ amplification events in the absence of Pol32 confirms that there is not a single pathway to generate terminal duplications.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Batch culture evolution of yku70Δ confirms that Y’ amplification events are decreased when combined with a pol32Δ mutation.

Four strains (wild type, yku70Δ, pol32Δ, and yku70Δ pol32Δ) were serially transferred for 22 days (~140 generations). On day 22, 11-12 colonies from each culture were examined by CHEF gel and Southern hybridization with a Y’ probe (A-D). Clones with altered Y’ hybridization are indicated by cyan circles. Two clones from yku70Δ (B; c4 and c11, indicated by the red asterisks) were sequenced with Oxford Nanopore Technology and their telomeric regions examined for amplification. The chromosomal assignments of Y’ telomeres in clones 4 and 11 are diagrammed in E and F. Structural alterations are highlighted by blue shading. The phylogenetic analysis of the Y’s from these two clones are shown in S10A and S10B Fig. Southern blots of CHEF gels from twelve clones from each of two additional technical replicates are shown in S9 Fig. The telomere maps and phylogenetic analysis of the three additional clones are given in S10C–S10E Fig.

https://doi.org/10.1371/journal.pgen.1012214.g008

To confirm the structure of the Y’ amplicons that arose in the batch culture of the yku70Δ strain we had five clones sequenced using Oxford Nanopore technology (c4, c11, c6, c12 and c10; Figs 8E, 8F, S9 and S10). Each had tandem additions of homogeneous Y’s that were related to the resident Y’ and were separated by expanded ITSs of similar size and sequence, resembling those presented in S6 Fig. Additional features we identified in the clones were the secondary transfer of amplified Y’ arrays to other chromosomes, presumably through non-allelic homologous recombination (e.g., c4 chr XIV-L; c11 chr X-L; Fig 8E and 8F) and two hybrid Y’s with the 5’ end of one chromosome and the 3’ end of another chromosome (Fig 8F chr II-R; S10B Fig chr VIII-R; S10D Fig chr XVI-R). We also found examples where a Y’ had been added to the end of a chromosome that had previously lacked a Y’ (Fig 8F chr II-R; S10D Fig chr II-R and chr XI-L), indicating that mmBIR occurs in trans at a reduced frequency.

Terminal amplification in the human genome explained by mmBIR in cis

The Telomere-to-Telomere (T-2-T) human genome assembly has added important sequence details in repeated parts of the genome, including refinements at telomeres. In a scan of the human T-2-T genome (https://genome.ucsc.edu) we focused on the recent addition to the reference genome of ~160 kb of sequences at the chromosome 18p telomere. In that region are four ~45 kb repeats that include several annotated genes (including genes for tubulin variants and for LINC RNAs) (Figs 9A and S11). Between the direct repeating units are internal stretches of the CCCTAA telomeric repeats that are strikingly similar to each other in size (~650 bp) but significantly shorter than the average terminal telomeric CCCTAA tract (Figs 9B and S11). The distribution of sizes is comparable to the distribution of telomeric lengths in yeast and the expanded ITSs in the Y’ tandem arrays we found in the yku70Δ mutant (Fig 9C). This pattern at Chr18p is similar in size and structure to terminal SUL1 amplicons we found arising in the yeast yku70Δ strain and suggests the possibility that this human amplification event may have occurred by mmBIR pseudo-rolling-circle replication during a period of telomere stress.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 9. Tandem telomeric amplification on human chromosome 18p.

Human telomeres are capped by the sequence TTAGGG/CCCTAA, and as in yeast, these repeats are found throughout the genome as ITSs. A) A simplified drawing of ~300 kb of telomere adjacent sequence of Chr18p from the T-to-T human genome sequence illustrates the telomere and several adjacent ITSs (red asterisks). Several genetic landmarks (LINC, AP and TUBB8B1) illustrate the repeated nature of this region and separating each repeat is an ITS of relative uniform size (~650 bp). B) A histogram of telomere tract length was compiled from the T-to-T data for the 23 chromosomes and compared to the ITS lengths from Chr18p, highlighted in red in the histogram. (A screen shot of this region taken from the human genome browser is shown in more detail in S11 Fig). C) A similar histogram of telomere lengths in FY4 yeast (gray) and the ITSs from the four tandem Y’s from Fig 5 (red, orange, yellow and green) suggest a common mechanism for terminal amplification events from yeast to humans.

https://doi.org/10.1371/journal.pgen.1012214.g009

The initial step in our proposal for mmBIR in cis shares structural similarity to the T-loops found in mammalian chromosomes [25]. These T-loops form by invasion of the single stranded G-rich strand into the upstream duplex tract of telomeric repeats and can range from 1 to 25 kb in size. They are formed and stabilized by the Shelterin complex and are thought to protect chromosome ends from degradation and/or recombination [25]. The authors speculated on the potential for the invading 3’ end to initiate replication, and it is now commonly thought that replication from this imbedded 3’ end is responsible for the lengthening of the G-rich telomeric tract in ALT type-II survivors. More recent work has shown that invasion of the 3’ end of the G-rich strand can also associate more distally with ITSs [26], but no role for replication priming has been proposed. More long read sequencing, particularly of telomere-adjacent regions, in individuals or tumors suspected of copy number disorders will potentially determine whether there is a role for telomere associated mmBIR in cis in human disorders.

Strains and culture conditions

All strains in this study were derived from FY4 (a derivative of S288C) or its close relative BY4741 (https://sites.google.com/view/yeastgenome-help/more-about-yeast/commonly-used-strains). The split-URA3 strain, s2-1 [2], was used to quantify the effect of yku70Δ, sae2Δ, rad54∆ and rad5Δ on the frequency of ODIRA amplification events. Medium and culture conditions for the sulfate-limited chemostats are as described [1,28]. Batch culture passaging was carried out in 6 well plates with 5 ml of sulfate-limited medium at 30˚C. A 100-fold dilution was made every 24 hours for 22 days and performed in triplicate for each strain. Twelve clones from each culture from day22 were isolated for further analysis. All gene deletion mutants were obtained from the KanMX yeast knock out collection [29,30] and/or created by transformation using PCR fragments and confirmed by PCR and Sanger sequencing (Eurofins). The ARS228 deletion was created by scrambling the ARS consensus sequence [4].

Molecular analysis of clones and cell populations

Periodically, during growth in chemostats or serial transfer, genomic DNA from cell populations was analyzed by CHEF gel electrophoresis [31, reviewed in 32] and Southern blotting [14] to detect karyotypic changes. The population or single clones from the final day of chemostat growth were analyzed by aCGH [2,14,31,33] to identify the segments of the genome that had been amplified. The final population of one chemostat (yku70Δ-A) was sequenced in house using Oxford Nanopore Technology. Sequences and contigs (S1 Data) of clones or populations from other chemostats, batch culture evolutions and the parental FY4 and yku70Δ parent strains were performed by Plasmidsaurus using Oxford Nanopore Technology (S2 Table). Raw reads from these data were analyzed using BBEdit (Bare Bones Software) and alignment to the reference genome to identify amplification junctions. Short read sequence data for populations of 31 independent sulfate-limited chemostats are from Martin et al., 2024 [2].

Characterization of SUL1 and Y’ amplicons

Split reads from Oxford Nanopore sequencing of the population from the yku70Δ-A chemostat were examined manually in IGV to identify SUL1 amplicon junctions. Sequence coverage of individual junctions ranged from 1 to 62. (The single split read for the junction at 786.502 kb was present twice on the same fragment as part of a SUL1 triplication.) Among the longer reads (>~40 kb) we found evidence for tandem amplicons of 3–4 repeats for five of the junctions. We were also able to identify translocations of the SUL1 fragment to two other chromosomes.

The split reads from Oxford Nanopore sequencing of yku70Δ-A also confirmed the tandem structure of Y’s with intervening, expanded ITSs. To determine the identities of the Y’s involved, we used tandem Y’s in a BLAST search and found chr XII (left or right) was always the best match, presumably because chr XII in SGD has tandem Y’s at each end. However, when the same sequence was split into individual Y’s, other telomeres provided the best BLAST match. Because some chromosomal ends are incomplete in SGD we focused on a section of the Y’ region (between 5’ATGGAAATTGAAA and 5’ATGCACTTGCGAGATC) common to all Y’s in SGD and performed multiple alignments (CLUSTALW) and determined phylogenetic relationships (FastTree full) of the Y’s in the reference genome, and compared them to Y’s in FY4 and the yku70Δ strains (S4 Fig). The phylogenetic relationships allowed us to define thirteen Y’ classes (S4A Fig Legend) and their distribution across the sixteen yeast chromosomes (S4A Fig). (Trees generated by PhyML or RAxML had identical topologies.) The Y’ identities on individual chromosome ends for the three strains mostly agree. We used this same strategy of creating multiple alignments and phylogenetic trees to identify the origin of Y’s in the tandem repeats found in our experimental samples. However, for Y’s we compared tandem repeats from single long reads to the yku70Δ parent consensus sequences. Even with the relative high frequency of sequencing errors it was still possible to identify the closest Y’ relative for the individual Y’s in the tandem repeats.