A novel connection between the fidelity of re-replication control and the fidelity of chromosome segregation

We have shown that centromeric re-replication provides a highly potent way to induce chromosomal instability and aneuploidy in cells where the mitotic segregation machinery is intrinsically intact. This suggests that the mitotic mechanisms that preserve segregation fidelity are not designed to handle the problems that arise when centromeres are re-replicated. As a consequence chromosome segregation fidelity is dependent on the fidelity of re-replication control, establishing an important connection between these very distinct processes. Moreover, as discussed later, it raises the possibility that the decreased fidelity of both processes that is seen in cancer cells may be related.

Centromeric re-replication induces missegregation

The aneuploidy arising from centromeric re-replication is generated in part through missegregation of both sister chromatids to one daughter cell. Exactly how this missegregation occurs remains to be determined, but centromeric re-replication has the potential to perturb the segregation machinery in at least three ways: (1) disruption of kinetochores; (2) disruption of centromeric cohesion; and (3) attachment of a single sister chromatid to microtubules from both spindle poles (i.e., bipolar attachment).

In budding yeast, kinetochores inherited from the previous cell cycle are disrupted by passage of the replication fork but then rapidly reassemble onto the newly replicated centromeres [34]. In principle, a similar disruption followed by reassembly could occur with re-replication forks since budding yeast kinetochores can assemble and become functional throughout the cell cycle [35]. Moreover, our observation that more than two centromeres in a re-replicating strain can be microscopically resolved in a microtubule dependent manner is consistent with the presence of functional kinetochores on many re-replicated centromeres. Nonetheless, we cannot rule out some of the missegregation we detected in our sectored colony assay arising from a failure to reassemble kinetochores on a minority of re-replicated centromeres.

The microtubule dependence of the separation of re-replicated centromeres also suggests that pericentromeric cohesion is often preserved following centromeric re-replication. Such cohesion is presumably responsible for the collapse of re-replicated and separated CEN5 spots to a single spot following the disruption of microtubules. We note, however, that in the less frequent cases when four spots were present in a cell, they often collapsed to two spots rather than one, suggesting that in some instances, particularly when more than one centromere re-replicates, pericentromeric cohesion may be compromised.

For the majority of centromeric re-replication bubbles that retain functional kinetochores and pericentromeric cohesion, attachment of re-replicated centromeres to microtubules from opposite poles will result in bipolar attachment of the re-replicated sister chromatid to the mitotic spindle (Fig. 6). Our observation of three or four resolvable centromeres moving around the nucleus in a microtubule-dependent manner is consistent with such bipolar attachments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Possible ways for centromeric re-replication to perturb chromosome segregation.

Normal sister chromatids are bilaterally symmetric and held together via cohesin to ensure their bi-orientation with respect to the spindle poles. Centromere re-replication disrupts this bilateral symmetry and can lead to abnormal bipolar attachment of a single chromatid to both spindle poles. During anaphase, this bipolar attachment could lead to a 2:0 segregation pattern. Alternatively the affected sister chromatid could break and repair in a RAD52-dependent manner to produce a 2:1 segregation pattern. Also conceivable but not shown are disruption of kinetochore function or pericentromeric cohesion by re-replication.

https://doi.org/10.1371/journal.pgen.1005039.g006

Importantly, this potential source of missegregation can, in principle, be established without triggering the two surveillance mechanisms that normally ensure faithful segregation. One of these mechanisms prevents merotelic kinetochore attachments, i.e. bipolar attachment of microtubules to individual kinetochores. Normally, because each sister chromatid has only a single kinetochore, this mechanism plays a critical role in preventing bipolar attachment of the chromatid to the mitotic spindle. However, if centromeres re-replicate, the presence of two kinetochores on a sister chromatid permits its bipolar attachment to the spindle without requiring merotelic attachment to either kinetochore. The second mechanism, the spindle assembly checkpoint, which detects the absence of tension on kinetochores, would also not be triggered because the tension generated by bipolar attachment to re-replicated kinetochores would be transmitted to their sister kinetochore(s) via sister chromatid cohesion. The inability of either surveillance mechanisms to sense and correct bipolar attachment to centromerically re-replicated sister chromatids could help explain why centromeric re-replication is such a potent inducer of missegregation and aneuploidy.

It should be noted that the re-replication bubbles that enable these bipolar attachments are transient chromosomal structure. Exactly how these bubbles eventually disappear is not clear, although they can conceivably be resolved during the normal course of a subsequent S phase. Nonetheless, because the bubbles are transient, centromeric re-replication can induce simple aneuploidy in a hit-and-run fashion. This stands in contrast to dicentric chromosomes, whose bipolar attachments inevitably lead to chromosome breakage and rearrangement [36–38].

Centromeric re-replication induces chromosome gain

In addition to inducing aneuploidy by missegregating sister chromatids, centromeric re-replication can also induce aneuploidy by the formation of extra sister chromatids, as manifested by the appearance of 2:1 segregation events. The dependence of at least half of these events on RAD52, a gene essential for most homologous recombination in budding yeast [39], implies that many of these extra chromatids are generated by chromosome breakage and recombinational repair.

The breakage is not surprising. It may arise from bipolar spindle tension being placed on centromeric re-replication bubbles by, or it might arise simply because of the susceptibility of re-replication forks to breakage [18]. What is striking, is the apparent efficiency with which the repair of these breaks can be channeled into the formation of extra sister chromatids. Again, this outcome contrasts sharply with the chromosomal rearrangements that result from dicentric chromosome breaks [37], and it reinforces the notion that chromosomes with re-replicated centromeres are not simply dicentric chromosomes in a different guise. Whether similar generation of aneuploidy by extra sister chromatid formation can occur in mammalian cells with their much larger chromosomes remains to be seen. Nonetheless, our studies have uncovered a novel way in which aneuploidy can be generated.

It is possible that centromeric re-replication can also lead to other chromosomal consequences that we did not observe either because they are lethal or because we did not score them. For example, the large rise in RAD52-dependent 2:1 segregation events in a dnl4 mutant background suggests that there may be other competing fates for chromosome breakage events that involve nonhomologous end-joining. In our primary colony screen, we focused on colonies that were significantly induced by centromeric re-replication and that we anticipated would be most straightforward to interpret, namely red-white and red-pink sectored colonies. Analysis of other colonies with different or more complex shapes and color patterns may uncover other types of chromosomal loss, gain, or rearrangements induced by centromeric re-replication.

A potential new source of aneuploidy in cancer

The connection we have established between decreased fidelity of re-replication control and decreased fidelity of chromosome segregation may be relevant to cancer as compromised fidelity for both processes have been observed in cancer cells. The decreased fidelity of chromosome segregation is well established in cancer [40]. Approximately 90% of solid tumors and 50% of hematopoietic cancers are aneuploid, with many exhibiting chromosomal instability [41,42]. Moreover, there are increasing hints that aneuploidy may contribute to tumorigenesis by promoting genomic instability [43,44].

How chromosomal instability arises in cancers is still an open question. Increasing attention has been placed on non-mitotic perturbations that can disrupt chromosome segregation because mitotic genes directly involved in kinetochore function, spindle function, cohesion, or the spindle assembly checkpoint are rarely mutated in sequenced cancer genomes [40,45]. One such perturbation is the accumulation of excess centrosomes, the microtubule organizing centers at the poles of spindles, as this can lead to incorrect merotelic attachment of sister chromatids to more than one spindle pole [46]. However, despite being present in many cancers, excess centrosomes are not observed in all cancers displaying chromosomal instability [47], raising the question of what other non-mitotic perturbations contribute to this instability.

Our observation that centromeric re-replication is a potent inducer of aneuploidy offers one such perturbation. This possibility is encouraged by accumulating evidence that re-replication may occur in cancer and contribute to oncogenesis [3,48,49]. Specifically, moderately elevated levels of the replication initiation proteins Cdc6 and Cdt1, which in high amounts induces detectable re-replication in cell culture and model organisms [27,50–55], has been observed in multiple types of primary tumors [56–61]. In addition, moderate overexpression of Cdt1 has been shown to potentiate carcinogenesis in mouse models [58,62,63]. Re-replication has yet to be directly confirmed in any of these settings [50,58,62,63], but this is likely due to the fact that levels of re-replication currently detectable by conventional replication assays cause extensive DNA damage and cell lethality [5,25–28,50,51,64–66]. Hence, only lower levels of re-replication can be compatible with cancer cell viability, and detecting such cryptic re-replication will require the development of more sensitive replication assays [67].

Finally, re-replication that does occur in human cells may well involve centromeric DNA, as this DNA was found to be preferentially re-replicated when overt re-replication was induced in Schizosaccharomyces pombe or S2 cells of Drosophila melanogaster [68,69]. Indeed, cryptic centromeric re-replication could account for the chromosomal instability that was induced in the absence of detectable re-replication by moderate overexpression of Cdt1 in normal human fibroblasts [70].

In summary, cryptic re-replication of centromeres could be occurring in cancer cells with deregulated initiation proteins, and our work shows that such re-replication is highly potent at inducing chromosomal instability and aneuploidy. Further investigation will be needed to determine whether this hypothesis applies to specific cancers, but our work offers a potential source of chromosomal instability and aneuploidy that has not been considered before. When combined with our previous observation that re-replication is a potent inducer of segmental amplifications and duplications [6], this work also displays the versatility with which re-replication can induce genomic instability.

Oligonucleotides

Oligonucleotides used for disrupting RAD52, DNL4, or HMRa, inserting the URA3 marker, or inserting the targeting marker TRP1 for the tet operator (tetO) array insertion are listed in S11 Table.

Plasmids

The plasmids pSH006 and pSH005 were constructed for integrating the copy number reporter ade3–2p [11] with or without ARS317, respectively, at ChrV_160kb near CEN5. The plasmid pSH006 contains the kanMX-ade3–2p-ARS317 cassette excised from pBJL2890 [6] with XbaI and StuI, and pSH005 contains the kanMX-ade3–2p cassette excised from pBJL2889 [6] with XbaI and StuI. In both plasmids the cassettes are flanked by homology sequences from ChrV_160 as listed in S10 Table. To make pSH008, pSH006 was cut with BsaBI and XbaI, filled in, and ligated to remove a secondary NotI site.

The plasmids pSH020 and pSH019 were constructed for integrating ade3–2w with or without ARS317, respectively, at ChrV_548kb near the right end of Chromosome V. The plasmid pSH020 contains a natMX-ade3–2w-ARS317 cassette based on the cassette in pSH008 but with natMX replacing kanMX and ade3–2w replacing ade3–2p. The allele ade3–2w is a derivative of ade3–2p where a frameshift mutation has been engineered at the beginning (between +45 and 46) of the open reading frame by inserting a single nucleotide. The cassette in pSH020 is flanked by homology sequences from ChrV_548 that are listed in S10 Table. The plasmid pSH019 is equivalent to pSH005 but has no ARS317 in the cassette.

The plasmid pSH013 was constructed to introduce bar1∆ by two-step gene replacement. It contains genomic sequences upstream and downstream of the BAR1 open reading frame (separated by a SpeI restriction site) cloned into pRS306 [71].

The plasmid pSR14 was obtained from the Dave Morgan Lab (UCSF) with permission from its original source, the Susan Gasser Lab (Friedrich Miescher Institute). This plasmid contains 128 tandem copies of tetO, the LEU2 marker, and target sequence for its integration into the genome [72].

The plasmid pCUP1-TetR-tdTomato-ADE2 was obtained from Dan Liu in the Dave Morgan Lab (UCSF). It expressed a Tet repressor (TetR) linked to the fluorescent protein tdTomato under the control of the pCUP1 promoter. The fusion construct was integrated into the middle of ADE2, and the entire fragment inserted into pRS406. The fusion protein lacks a nuclear localization sequence.

The plasmid pBJL2667 expresses GFP-Tub1 fusion protein under the control of the pHIS3 promoter. This plasmid contains pHIS3-GFP-TUB1 fusion construct from the multiple cloning site of pRS306 in the plasmid pAFS91 (Aaron Straight, Stanford) [73] subcloned into multiple cloning site of pRS304.

Sequences files for all plasmids are available upon request.

Strain construction

All haploid and diploid strains used in our experiments (S9 Table) are in a genetic background that can conditionally induce reinitiation of DNA replication, most prominently from the origin ARS317. They also have one homolog of Chromosome V marked with the ade3–2p copy number reporter to the right its centromere CEN5 (ChrV_160kb). This ade3–2p marked homolog also contains ARS317 close to the centromere CEN5 (ChrV_160kb), ARS317 near the right end of the chromosome (ChrV_548kb), or no ARS317 at all. The basic strategy for generating these diploids was to mate a MATa haploid strain containing the ade3–2p marked Chromosome V with a congenic MATα strain containing a URA3 marked Chromosome V.

All haploid strains used in these matings were derived from the haploid strain YL3155 [6] (MATa ORC2-(NotI, SgrAI) orc6(S116A) leu2 ura3–52 trp1–289 ade2 ade3 MCM7–2NLS bar1::LEU2 CAN1 HMRa). YJL3155 is primed to reinitiate DNA replication because the MCM7–2NLS allele makes the Mcm2–7 complex constitutively nuclear [74], and the orc6(S116A) partially prevents CDK inhibition of the origin recognition complex (ORC) by mutating one of the 11 CDK consensus phosphorylation sites in ORC [75]. To generate the haploid strains used in the matings, various combinations of changes to the following loci were made:

MATα—MATα haploids were obtained by switching from MATa using pGAL-HO in pSB283 [76].

hmra∆::hphMX—for both haploids, the endogenous ARS317 was removed by deleting the entire HMRa locus. This was done by integrating an hmra∆::hphMX disruption fragment that was generated by two-step PCR amplification from pAG26 [77] using primers described in S11 Table.

bar1∆—for both haploids, bar1::LEU2 was converted to bar1∆ by loop-in/loop-out gene replacement using BsrG1 linearized pSH013.

ura3∆::{ACT1term-pGAL1/10-delntCDC6,cdk2A-CDC6term}—for both haploids, a galactose-inducible, stabilized version of CDC6 was integrated in place of the uras3–52 allele by loop-in/loop-out gene replacement with SmaI linearized pKJF019 [18].

ChrV_160::{kanMX, ade3–2p, ±ARS317}—for the MATa haploids, CEN5 was marked with the copy number reporter ade3–2p by integration of a cassette containing kanMX and ade3–2p at ChrV_160kb (right of the centromere). When ARS317 was to be positioned near CEN5, we used a {kanMX, ade3–2p, ARS317} cassette excised from pSH008 with SacI and NotI. When ARS317 was not to be on the chromosome or to be located on the right arm (at ChrV_548kb), we used a {kanMX, ade3–2p} cassette excised from pSH005 with SacI and NotI.

ChrV_160::URA3—for the MATα haploids URA3 was integrated near CEN5 by one step gene insertion with a URA3 integration fragment generated by two step PCR amplification from pRS316 using the primers described in S11 Table.

ChrV_548::{natMX, ade3–2w, ±ARS317}—for the MATa haploids, a cassette containing natMX and ade3–2w (the inactive color reporter described above) was integrated near the right end at ChrV_568kb. When ARS317 was to be positioned at this location we used a {natMX, ade3–2w, ARS317} cassette excised from pSH020 excised with SacI and NotI. When ARS317 was not to be on the chromosome or to be located near CEN5, we used a {natMX, ade3–2w} cassette excised from pSH019 with SacI and NotI.

dnl4∆—to make diploid strains homozygous for the dnl4 deletion, DNL4 in both MATa and MATα haploids was deleted by integrating a dnl4∆::LEU2 disruption fragment that was generated by two step PCR amplification from pRS315 [71] using the primers described in S11 Table.

rad52∆—to make diploid strains homozygous for the rad52 deletion but wild-type for DNL4, RAD52 in both MATa and MATα haploids was deleted by integrating a rad52∆::LEU2 disruption fragment that was generated by two step PCR amplification from pRS315 using the primers described in S11 Table. To make diploid stains homozygous for both rad52 and dnl4, MATa and MATα haploids that already contained the dnl4∆::LEU2 allele had their RAD52 genes deleted using a different marker. A rad52∆::URA3 disruption fragment that was generated by two step PCR amplification from pAG36 [77] was used for the MATa haploids, and a rad52∆::natMX disruption fragment that was generated by two step PCR amplification from pAG36 was used for the MATα haploids. Primers used for the PCR are listed in S11 Table.

trp1–289::{GFP-TUB1, TRP1}—the plasmid pBJL2667 was linearized with Bsu36I and integrated via loop-in at the endogenous trp1–289 locus.

ade2::{pCUP1-TetR-tdTomato, URA3, ADE2}—the plasmid pCUP1-TetR-tdTomato-ADE2 was linearized with BglII and looped-in at the endogenous ade2 locus. Due to the inability of a functional ADE2 to confer complete adenine prototrophy without a fully functional ADE3, cells were selected for the URA3 marker in the backbone of the plasmid. However, the presence of a fully functional ADE2 prevents accumulation of the red pigment produced in an ade3–2p ade2 background, meaning colonies were now without color (white).

ChrV_151::{LEU2, tetOx128}—The plasmid pSR14 was linearized with AscI and stability integrated into ChrV at 151 kb, which is to the left of CEN5 (the re-replication cassette is to the right of the centromere). To integrate this construct at the proper location, targeting homology was generated and integrated first. A fragment containing TRP1 and specific targeting sequence was generated by PCR amplification from pRS314 using primers described in S11 Table. This construct was placed at ChrV_151, which provided the targeting homology for the sequences at the end of the linearized pSR14. Cells were visually inspected for bright red spots prior to strain archiving.

Media and cell growth

Cells were grown in or on YEP medium [78] supplemented with 2% wt/vol dextrose (YEPD), 8% wt/vol dextrose (YEP8D), or 3% wt/vol raffinose + 0.05% wt/vol dextrose (to form YEPRd). The color development plates were synthetic base with 2% wt/vol dextrose (SD), with the final amino acid concentrations as follows: adenine [10 μg/mL], uracil [20 μg/mL], tryptophan [20 μg/mL], histidine [20 μg/mL], arginine [20 μg/mL], methionine [20 μg/mL], tyrosine [30 μg/mL], leucine [60 μg/mL], isoleucine [30 μg/mL], lysine [30 μg/mL], phenyalanine [50 μg/mL], glutamate [100 μg/mL], aspartate [100 μg/mL], valine [150 μg/mL], threonine [200 μg/mL], serine [200 μg/mL]. These plates contain low adenine and are referred to as SDClowA plates. All cell growth was performed at 30°C except where otherwise noted.

The color from the ade3–2p reporter was most consistent when SDClowA plates were poured in a precise manner three days prior to use. For a 2 L batch, the following was added together in a 4 L flask: 13.4g Yeast Nitrogen base without amino acids, 40g Bacto-agar, 1 stir bar, and 1850mL MQH₂O. The opening of the flask was covered with foil, and the mix was stirred for 5 min. Media was autoclaved on a liquid cycle (30 min at 121°C with slow exhaust) in a dry autoclavable plastic tray. To prevent excessive heating the flask was removed as soon as the jacket pressure allowed the door to open, and the media was stirred for 10 min to mix and cool. At this point powdered amino acid mix and 100mL 40% Dextrose were added. The media then stirred for an additional 10 min to further mix and cool. Plates were poured using a PourBoy 4 plate pouring machine to dispense 33 mL of media per plate. These SDClowA plates were stacked unwrapped to allow them to dry but shielded from light until use in the assay.

For microscopy, cells were imaged in SDC-Trp media to maintain selective pressure on the integrated GFP-Tub1 construct. Prior to imaging, cells were grown and induced in S media supplemented with 3% wt/vol raffinose + 0.05% wt/vol dextrose (to form SRd). Amino acid concentrations for both medias are as follows: adenine [40 μg/mL], uracil [40 μg/mL], tryptophan [0 μg/mL], histidine [40 μg/mL], arginine [40 μg/mL], methionine [40 μg/mL], tyrosine [60 μg/mL], leucine [120 μg/mL], isoleucine [60 μg/mL], lysine [60 μg/mL], phenyalanine [100 μg/mL], glutamate [200 μg/mL], aspartate [200 μg/mL], valine [300 μg/mL], threonine [400 μg/mL], serine [400 μg/mL].

Re-replication induction strategy

Re-replication was conditionally induced in a strain background where ARS317 preferentially reinitiates [4]. The strain background was deregulated in three ways: (1) (MCM7–2NLS)—the CDK-driven export of Mcm2–7 from the nucleus [74,79,80] was blocked by fusing a constitutive nuclear localization signal onto the endogenously expressed Mcm7; (2) (pGAL-∆ntcdc6-cdk2A)—the CDK inhibition of Cdc6, which occurs through transcriptional regulation [81], phosphorylation-directed degradation [82–84], and direct CDK binding [85], was disrupted by expressing an extra copy of Cdc6 lacking CDK phosphorylation and binding sites under a galactose-inducible promoter; and (3) (orc6(S116a))—the CDK inhibition of ORC by phosphorylation of Orc2 and Orc6 was constitutively yet minimally perturbed by eliminating one of four CDK consensus phosphorylation sites on Orc6 [6].

To induce re-replication, galactose is added to express ∆ntcdc6-cdk2A from the pGAL promoter. This increases the G2 DNA copy number of ARS317 by 50% from 2C to 3C. Despite the constitutive deregulation of Mcm7 and Orc6, no detectable re-replication could be detected by microarray comparative genomic hybridization (our most sensitive copy number assay) in the absence of ∆ntcdc6-cdk2A induction. Nonetheless, in principle cryptic and sporadic re-replication events could occur throughout the genome before the induction. To focus on the consequences of re-replication specifically induced at CEN5 from ARS317, we use congenic strains either lacking ARS317 at CEN5 or with ARS317 located on the arm of the chromosome as controls for colony sectoring frequency and aneuploid frequency.

Sectoring assay following re-replication induction

Colony color development and sectoring frequencies were most reproducible when freshly thawed cells were used and the re-replication induction and plating were performed in a precise manner. Yeast were thawed from frozen glycerol stocks onto YEPD plates and grown at 30°C. The following day, this patch was used to inoculate 25 mL of YEPD and was grown to an OD₆₀₀ between 0.4 and 0.5 over the course of 4–5 hours at 30°C in a shaking (250 rpm) waterbath. From this culture, we inoculated the experimental culture grown at 30°C in non-repressive rich media containing 3% raffinose and 0.05% dextrose (YEPRd) so that after 13–15 hours, the culture would be growing exponentially at an OD₆₀₀ between 0.4 and 0.5.

To arrest cells, nocodazole (US Biological, Salem, MA; Cat. no. N3000) was added to the culture at a final concentration of 15 μg/mL. After two hours of incubation, effective mitotic arrest was confirmed microscopically (>90% large budded cells) and 40% galactose was added to the arrested culture to a final concentration of 2.7%. Galactose induction was allowed to proceed for three hours before cells were washed and plated. A modification of this assay was used for YJL9631 and YJL9633, which have ARS317 integrated at the right end of Chromosome V. Total reinitiation from the right end of Chromosome V in these strains was higher after 3 hours of induction than the centromeric reinitiation from strains with ARS317 integrated near CEN5, possibly from the low-level endogenous re-replication occurring on the end of the arm. Thus, to compare the consequences of equivalent amounts of reinitiation at the two locations, we reduced the galatose induction time for YJL9631 and YJL9633 to two hours. This was done by delaying the galatose induction until 3 hours after addition of nocodazole so that all strains were exposed to nocodazole for the same amount of time (5 hr). Cell cycle arrest and maintenance of arrest during the induction of re-replication were confirmed by flow cytometry.

After the induction, cells were diluted into YEPD to an estimated concentration of 1000 CFU/mL based on (1) the OD₆₀₀ of the culture at the time of nocodazole addition (using a conversion factor of 2 x 10⁷/mL at an OD₆₀₀ = 1.0) and (2) the expected viability drop following the induction of re-replication (see below). From this dilution, 200μL (~200 CFU) were promptly spread onto SDClowA plates using 6–8 sterile glass beads per plate. The plates were incubated face-up at 30°C in the middle of an air incubator for 5–7 days, after which the plates were placed into a dark drawer at room temperature (18–22°C) for further color development. After 2–3 days there was optimal distinction between red and pink colors, and the plates were scored for color-sectored colonies.

After the total number of colonies was counted, plates were manually screened using a Leica Modular Stereomicroscope MZ6 and a Leica KL1500 LCD 150-watt halogen cold light source and ring-light fixture. Lamp was set to 3200 K with the aperture 80% open. To be scored, sectored colonies were required to be at least 1 mm in diameter and have a red portion that was less than 75% but more than 25% of the total colony. This range of sector sizes was chosen to accommodate differences in the time the two daughter cells might take to recover from the cell cycle arrest or possible differences in doubling times for the two daughter lineages. We were stringent about sector color, however. The pink portion of each red/pink colony had to have the same tint as the surrounding pink colonies (indicative of the color a single copy of ade3–2p produced), and the red portion had to be of an unmistakably darker tint than the pink portion. The white portion had to be completely white and without colored tint. Colonies containing more than two colors were not scored. All sectors had to originate from the center of the colony, indicating they had been formed in an early cell division immediately following centromeric re-replication. Colonies with colored sectors originating from outside the center were not scored. The results of all trials reported in this study are listed in S7 Table.

After all plates were screened and sectors tallied, colonies were picked from their original plate and struck onto a new SDClowA plate, then incubated as described above to develop color. This colony purification process allowed us to obtain clonal isolates from each sector of the colony. From the streak, individual colonies were picked (one of each color) and patched onto rich media with 2% dextrose. These patches provided the material for glycerol freezer stocks, kept in 96-well plates at—80°C.

The viability drop during the plating described above was determined in preliminary experiments by plating out an estimated 200 CFU per plate based solely on the OD₆₀₀, then measuring the actual number of colonies appearing. Measured and estimated colony numbers were the same (within 10–15%) for cells that were arrested in nocodazole for 2 hr, indicating that there was little viability drop at the T = 0 hr time point. After a 3 hr induction of re-replication (T = 3 hr), however, there was an ~4x drop in viability due in large part to the prolonged exposure to nocodazole (5 hr). Importantly, this viability drop was not dependent on the presence or absence of ARS317 near CEN5, whereas the sectoring frequency among the surviving colonies was. We thus inferred that this frequency was independent of the viability drop and was representative of the sectoring frequency that the entire population would have displayed in the absence of a viability drop. Finally, we note that the T = 3 hr viability drop was somewhat different in the recombination mutants we examined: ~6x for rad52∆ and rad52∆ dnl4∆, and ~3x for dnl4∆.

Statistical analysis of sectoring frequencies

To determine if the mean red/white-colony or red/pink-colony frequencies after centromeric re-replication was significantly higher than that of the no-ARS317 control (Fig. 2 & Fig. 3A), a one-tailed, two-sample t-test was applied using Excel. To determine whether the mean red/pink-colony frequencies after centromeric re-replication in recombination mutant backgrounds (rad52∆, dnl4∆, rad52 dnl4∆) were significantly different than in wild-type recombination backgrounds a two-tailed, two-sample t-test was applied using Excel. All sectoring frequencies used in the t-tests are listed in S7 Table. All p values, regardless of significance, are reported in the appropriate figure legends. Asterisks used in the figures indicate level of confidence, with three (***) indicating a p-value of <0.001, and two (**) indicating a p-value between 0.01 and 0.001.

Monitoring re-replication via aCGH

The remainder of the cells not used in the platings described above were harvested after the induction of re-replication and their DNA extracted as previously described (referred to as “method 2”) [18]. We have referred to this protocol as the “Clean Genomic Prep” on the GEO microarray database. This DNA was hybridized against reference DNA from YJL8590 (collected from a nocodazole-arrested culture that did not undergo re-replication) in the manner described previously [4]. Briefly, reference DNA was labeled with Cy5 fluorescent dye and DNA from the re-replicating strain was labeled with Cy3 fluorescent dye. Equal quantities were combined and competitively hybridized to an in-house printed microarray containing approximately 13,000 genic and intergenic PCR product elements representing the entire S. cerevisiae genome (GEO platform number GPL3412) at 63°C for at least 20 hours. Arrays were then scanned using an Axon Scanner 4B and analyzed as described previously [4]. All microarray data is deposited in the Gene Expression Omnibus [86] (http://www.ncbi.nlm.nih.gov/geo/) with accession number GSE55641.

Copy number determination via aCGH

Sectored colonies were chosen at random for analysis of copy number across the genome. For each such colony both purified sectors were thawed from the glycerol freezer plate stock onto YEPD plates and grown at 30°C overnight. A part of each patch was then used to inoculate 5 mL liquid YEP8D, which was grown to saturation over 2–3 days at 30°C with most cells arrested in G0 phase.

To obtain the DNA, the extraction method used in the re-replication DNA extraction was simplified to handle numerous small cultures at once. This DNA protocol is referred to as the “Small DNA Prep” on the GEO microarray database. 1 mL from a saturation culture was moved to a 2 mL screw-capped tube and spun down in a Eppendorf 5417C microfuge at 14,000 rpm for 3 min. Cell pellets were washed twice with 1 mL water and spun down at 14,000 rpm for 3 min for each wash. Cell pellets were then flash-frozen in liquid nitrogen and either processed immediately or stored at—80°C for future extraction. To the frozen cell pellets, 200 μL lysis buffer (2% Triton X-100, 1% SDS, 100 mM NaCl, 10 mM TrisCl pH 8, 20 mM EDTA) was added and tubes were allowed to rock gently at 4°C for 10 min to mix in the buffer as well as thaw the pellet. Then, 400 μL small glass beads and 200 μL phenol:chloroform:isoamyl alcohol (25:24:1) were added to the tube and vortexed immediately to mix. Tubes were vortexed on high for 10 min in a Vortex genie, after which 400 μL 1X TE (10 mM TrisCl pH 7.5, 1 mM EDTA) and 400 μL phenol:chloroform:isoamyl alcohol (25:24:1) were added. The tubes were vortexed to mix then spun down at 13300 rpm for 10 min in the microfuge on soft. 500 μL of the clear upper aqueous phase was transferred to new screw-capped tubes, containing 500 μL chloroform:isoamyl alcohol (24:1), and vortexed well. Tubes were spun at 5100 rpm for 10 min in the microfuge, and 450 μL of the top phase was moved to a new microfuge tube. Volumes were brought up to 500μL with 50μL 1x TE pH 7.5, and 400 μL isopropanol and 5 μL 5 M NaCl were added followed by vortexing. Tubes were spun at 10600 rpm for 10 min in the microfuge. The supernatant was carefully aspirated out of microfuge tubes, and the pellet was washed with 750 μL 70% ethanol by vortexing well. Tubes were spun at 10600 rpm for 5 min in the microfuge and the supernatant was carefully aspirated out. DNA pellets were dried for 10 min with gentle heat in a speedvac. Pellets were resuspended in 175 μL 1x TE pH 7.5 by heating in a 37°C waterbath with vortexing for 15 min. 1 μL RNase (Qiagen, Cat. no. 19101, DNase free, 100mg/mL) was added to each tube, which were inverted to mix and incubated in the 37°C waterbath for 1 hour. After the incubation, 100 μg Proteinase K (Roche, Cat. no. 03115852001) was added to each tube and inverted to mix, then placed in a 55°C waterbath to incubate for 30 min. Then, 400 μL chloroform:isoamyl alcohol (24:1) was added to each tube followed by vortexing and a spin at 7500 rpm for 10 min in the microfuge. The top 150 μL was removed and added to new microfuge tubes, followed by 120 μL isopropanol and 1.5 μL 5 M NaCl. Tubes were vortexed and spun at 9600rpm for 10 min in the microfuge. The supernatant was carefully aspirated out of microfuge tubes, and the pellet was washed with 750 μL 70% ethanol by vortexing well. Tubes were spun at 9600 rpm for 5 min in the microfuge and the wash was carefully aspirated out. DNA pellets were dried for 10 min with gentle heat in a speedvac. Pellets were dissolved in 50 μL 2 mM Tris pH 7.8 and placed in the 37°C waterbath for 15 min to ensure complete resuspension. DNA was then kept at—20°C until used for aCGH.

DNA from the sector isolates was hybridized against reference DNA from YJL8590 (collected from a nocodazole-arrested culture that did not undergo re-replication) in the manner described previously [4]. Briefly, reference DNA was labeled with Cy5 fluorescent dye and DNA from the color sector was labeled with Cy3 fluorescent dye. Equal quantities were combined and applied to an in-house printed microarray (GEO platform number GPL3412) and hybridized at 63°C for at least 20 hours. Arrays were then scanned using an Axon Scanner 4B and analyzed as described previously to generate copy number information across the genome [4]. All microarray data is deposited in the Gene Expression Omnibus [86] (http://www.ncbi.nlm.nih.gov/geo/) with accession number GSE55641.

To interpret the copy number of Chromosome V in the red or white sectors, we had to take into account the fact Chromosome V monosomy and trisomy are not stable in the long term. Populations of diploid cells containing these aneuploidies are eventually taken over by cells that have become disomic for Chromosome V. Hence, during the growth, freezing, and thawing of these aneuploidy cells, the population average of Chromosome V copy number gradually declines from 3C to 2C (for trisomy) or increases from 1C to 2C (for monosomy). In contrast, the populations of cells that start off euploid with a 2C copy number of Chromosome V never show any significant difference from 2.00 C during continued growth. Hence, we scored red sectors with Chromosome V copy number > 2.2C as trisomic for Chromosome V and white sectors with Chromosome V copy number < 1.8C as monosomic for Chromosome V. Both criteria had to hold in order for a red/white colony to be scored as a 2:0 segregation event. For red/pink colonies to be scored as a 2:1 segregation event, Chromosome V copy number in the red sectors had to be > 2.2C and in the pink sector at 2.00 C.

During our copy number analysis we discovered several sectored colonies, in which the Chromosome V copy number in the red and/or white sectors was significantly different than 2.00C, but not different enough to satisfy the thresholds of > 2.2C and < 1.8C respectively. Because we suspected that the stress of freeze-thawing placed selective pressure on the outgrowth of euploid cells, we isolated DNA directly from the frozen cells to perform aCGH. Cells from the frozen stock were placed directly in a screw-cap tube with 1 mL water, then processed the same as described for saturation cultures. In almost all of these cases, the new copy number analysis established that the red sectors had a much greater than 2.2C copy number for Chromosome V, while the white sectors had a much less than 1.8C copy number for Chromosome V.

Live-cell microscopy: Cell growth and re-replication induction

Cell growth and re-replication induction were performed in a similar manner as the re-replication assays described above. Yeast were thawed from frozen glycerol stocks onto SDC-Trp plates and grown at 30°C. The following day, this patch was used to inoculate 25 mL of SDC-Trp and was grown to an OD₆₀₀ between 0.2 and 0.4 over the course of 6 hours at 30°C in a shaking (250 rpm) water bath. From this culture, we inoculated the experimental culture grown at 30°C in non-repressive rich media containing 3% raffinose and 0.05% dextrose (SRd) so that after 13.5 hours, the culture would be growing exponentially at an OD₆₀₀ between 0.3 and 0.5 and be amenable to rapid galactose induction.

To monitor the exertion of metaphase spindle tension on re-replicated centromeres, we could not pre-arrest cells in metaphase with nocodazole before inducing re-replication (as outlined in Fig. 1A). Instead, we took advantage of the fact that re-replication itself induces a metaphase arrest with an active mitotic spindle because of the DNA damage response triggered by re-replication [18, 26]. Hence, we induced re-replication in exponentially growing cells by adding 40% galactose to a final concentration of 2.7%. For video microscopy (Fig. 5C; S1 and S2 Movies), galactose induction was allowed to proceed for approximately 3 hours, after which a small sample (500 μL) was removed and imaged live (see “Video microscopy” below). For time-lapse imaging (Figs. 5A and 5B), cultures were washed by vacuum filtration using 10 volumes of pre-warmed sterile water, then resuspended in 25 mL pre-warmed SDC-Trp media to limit further galactose induction of re-replication. From this new resuspension, a small sample (500 μL) was imaged continuously for approximately 20 min before nocodazole was added (see “Time-lapse imaging before nocodazole treatment”below) and then for approximately 2 hours after nocodazole was added (see “Time-lapse imaging after nocodazole treatment” below).

Live-cell microscopy: Imaging re-replicated cells

All live imaging was conducted in a temperature-controlled chamber maintained at 30°C with yeast immobilized to the bottom of a chambered coverslip (Lab-Tek/Thermo Fisher, Cat. no.12565401). Briefly, the bottom of the chamber was coated using 100 μL of 0.5 μg/mL Concanavalin A Type IV (Sigma-Aldrich, Cat. no. C2010–25MG) and allowed to air dry in the dark at 30°C for 1 hour. Roughly 3.5x10⁶ cells in 500 μL was pipetted into the chamber and allowed to sit for 15 min at 30°C, after which the chamber was gently washed twice to remove excess cells. Chambers were then imaged on the Deltavision deconvolution microscope (Applied Precision) using SoftWorx image acquisition software.

Video microscopy. Re-replicating cell spindle and spot dynamics (Fig. 5C; S1 and S2 Movies) were visualized with a 100× 1.40 UPLS Apo objective (Olympus) and a CoolSNAP_HQ / ICX285 camera. To visualize DNA, Hoechst 33342 stain was added at a final concentration of 10 μg/mL to the cells prior to imaging. A set of three images (one for each channel) was taken every 6 seconds for 10 min; the Z plane remained unchanged and was prevented from drifting using the UltimateFocus feature. The settings for each excitation wavelength are as follows: RD-TR-PE (red channel; ex: 555 nm, em: 617 nm; 32% power; 0.6 s exposure), FITC (green channel; ex: 490 nm, em: 528 nm; 50% power, 1 s exposure), and DAPI (blue channel; ex: 360 nm, em: 457 nm; 32% power; 0.2 s exposure).

Time-lapse imaging before nocodazole treatment. Spot and spindle visualization before and after nocodazole (Fig. 5A) was imaged with a 60× 1.42 NA Plan Apo objective (Olympus) and a CoolSNAP_HQ / ICX285 camera. No Hoechst stain was added to these samples. X, Y, and Z coordinates were recorded for 25 locations on the chambered coverslip, 10 for the no ARS317 strain and 15 for the ARS317 at CEN5 strain. For each of these points, a Z-stack was taken ±1 μm from the recorded Z coordinate in steps of 0.5 μm for both the red and green channels. This recorded Z coordinate was prevented from drifting using the UltimateFocus technology. Four stacks were acquired for each marked position as quickly as possible, resulting in images taken every 7 min for nearly 20 min. The settings for each excitation wavelength are as follows: RD-TR-PE (red channel; ex: 555 nm, em: 617 nm; 32% power; 0.8 s exposure), and FITC (green channel; ex: 490 nm, em: 528 nm; 50% power, 1 s exposure).

Time-lapse imaging after nocodazole treatment. Media containing nocodazole was carefully added to the chambers with a curved pasteur pipet so as not to disturb chamber position on the microscope stage. Immediately before addition, nocodazole was added to 500 μL fresh SDC-Trp media and then dripped into each chamber for a final nocodazole concentration of 75 μg/mL. Nocodazole was kept separate from the media until the last minute to minimize its precipitation out of the aqueous media. Using the same 25 points as the pre-nocodazole imaging, a Z-stack was taken ±1 μm from the recorded Z coordinate in steps of 0.5 μm for both the red and green channels. This recorded Z coordinate was prevented from drifting using the UltimateFocus technology. Four stacks were acquired for each marked position every 10 min for 2 hours. The settings for each excitation wavelength are as follows: RD-TR-PE (red channel; ex: 555 nm, em: 617 nm; 32% power; 0.8 s exposure), and FITC (green channel; ex: 490 nm, em: 528 nm; 50% power, 1 s exposure).

Live-cell microscopy: Spot analysis of re-replicated cells

Quantification of spot number before nocodazole treatment. All images were assigned with a random numerical prefix using the free demo version of Renamerox (Branox Technologies) so that they may be scored blindly. Using the free imaging software Fiji [87], a single Z-stack was opened and cells meeting specific criteria were annotated with a letter. The criteria are as follows: the cell must remain large-budded with a short (≤ 2 μm) spindle indicating a stable metaphase arrest without entry into anaphase. Using the Z- and time-dimension, the number of centromeric spots for each cell satisfying the criteria was determined and recorded. Centromere-breathing and rapid movement of spots across Z sections made it necessary to look through all sections of a Z-stack and through all four time frames during our assessment of the number of resolvable spots (see S6 Table).

It should be noted that roughly 8% of cells scored contained spots with noticeable differences in intensity by eye (our most sensitive way of detecting these differences). Such difference could conceivably arise from a number of factors, including: (1) the presence of overlapping unresolved centromeres in some spots; (2) differences in how close the Z sections cut through the center of each spot; (3) possible lag times in restoring complete binding of fluorescent proteins after their displacement by re-replication. The 8% of cells we flagged were those showing spot intensity differences that persisted through multiple pre-nocodazole time frames.

Within these cells, we were particularly interested in determining how many spots likely represented two (or more) centromeres that were never resolved in any time frame, as this would give us an indication of the extent to which centromeres were undercounted. To do this, we looked for spots whose intensity was more than 170% of a neighboring spot within the same cell. Quantification of spot intensity was performed using Fiji. Because of high signal to noise ratio, large variability in the noise for different Z sections, and movement of spots among the Z sections, each spot was quantified using the Z section and time point where the intensity of the spot was greatest. The total pixel intensity for a fixed-size circle was measured both when the circle was centered over the spot and when the circle was placed just outside the spot. The difference was then calculated to determine the spot intensity corrected for the neighboring noise. We note that the surrounding noise could be quite different for the different spots in the same cell, as it was lower for spots near the periphery of the cell body (e.g. within the bud neck) and was higher for spots closer toward the center. Based on this quantification of spot intensity, the spots that potentially represented two or more unresolved centromeres comprised only 2–3% of the total number of spots we counted. Because of this low proportion and the challenges described above of accurately quantifying spot intensity, we decided to adhere to the accepted practice in the field of counting the number of fluorescently marked chromosomal spots without adjusting for spot intensity. Nonetheless, this trial quantification provided further indication that we were not significantly undercounting the number of centromeres following their induced re-replication.

Quantification of spot number before nocodazole treatment. All images were assigned with a random numerical prefix using the free demo version of Renamerox (Branox Technologies) so that they could be scored blindly. Using the imaging software Fiji [87], single Z-stacks were opened until the file matching the pre-nocodazole treatment was found. Then, using the annotated pre-nocodazole image as a guide, each cell was revisited after 2 hours of nocodazole treatment. At this specific timepoint, the Z-planes were examined to determine how many spots were present. Several spots that were originally scored in the pre-nocodazole treatment were discarded due to one of the following reasons: the cell underwent anaphase before the nocodazole treatment had an effect; the cell’s spots separated into opposite lobes after spindle disappearance, indicating they were mid-anaphase; the spindle was not broken down; the centromeric spot was not visible at the final 2 hour timepoint, either because it lost signal or it became out of focus; or the cell died. Thus only large-budded cells in metaphase prior to nocodazole with spindle breakdown and visible spot(s) 2 hours after nocodazole were scored. For each strain in each trial, ≥ 100 cells were scored. Values were recorded and can be seen graphically in Fig. 5B or in S6 Table.

It should be noted that all images were scored with only moderate and uniform changes to brightness and contrast as needed.

Image creation for Fig. 5, S6 Fig, and S1 and S2 Movies

For images viewed in Fig. 5A and S6 Fig, the entire images were adjusted uniformly for brightness and contrast only.

For images viewed in Fig. 5C and S1 and S2 Movies, the red channel of this montage was subjected to Fiji’s bleach correction plugin using an exponential line fit. The resulting image fit a curve with R² > 0.99. To remove background, it was subtracted using Fiji’s background subtraction, using a rolling ball radius of 200 pixels. The green channel was altered only in brightness and contrast. The blue channel underwent a simple ratio (0.5) bleach correction using Fiji’s bleach correction plugin.