Yeast strains and Plasmids

The list of yeast strains and plasmids utilized in this study is given in S1 and S2 Tables, respectively. The specific figures and table depicting the experiments in which they were employed are also indicated. The presence or absence of the native 2-micron plasmid in a given strain is denoted as [Cir⁺] or [Cir⁰], respectively. This designation does not include ARS-based or 2-micron-derived plasmid constructs. The genotype of a strain containing such engineered plasmids, but not the native plasmid, is still referred to as [Cir⁰] with the resident plasmid spelled out.

Construction of yeast plasmids

The yeast plasmids (S1 Fig) used for genetic assays were constructed by a strategy analogous to that described previously [41]. The rationale is to generate two requisite linear DNA fragments in vitro by PCR amplification, and allow them to self-assemble the desired circular plasmid in vivo by homologous recombination/repair. Recombination is facilitated by overlapping sequences that these fragments carry at their ends. A suitable marker (ADE2) contained in one of the fragments permits the selection of plasmid-containing cells. We used a constant DNA fragment corresponding to the A-form of the plasmid [42] that included, in sequential order, the 2-micron plasmid RAF1 gene, ADE2 inserted into the plasmid HpaI site, STB, ORI, a copy of the inverted repeat, and the REP2 gene. The other variable fragment included the REP1 gene, the second copy of the inverted repeat, and the FLP gene with the incorporated modifications. Sequences adjoining REP1 and RAF1 provided homology at one end. Homology at the other end came from sequences adjacent to FLP and REP2. An equimolar mixture of the constant fragment with one of the variable fragments was used to transform an ade2 [Cir⁰] yeast strain to adenine prototrophy. DNA samples isolated from a subset of the transformants were analyzed by PCR to identify those that contained the correct plasmid. Critical regions of the plasmid, including the modified FLP locus and the recombination junction regions, were further verified by DNA sequencing. Once a parental strain harboring the correct plasmid was established, subsequent transfer of the plasmid to other recipient strains was performed by transformation using isolated total DNA.

Engineering yeast chromosomes

Integration of exogenous DNA cassettes into a specific chromosome locale was accomplished by one of three methods based on homology-dependent double strand break repair: (1) using a linearized integrative plasmid cut within the region of homology, (2) using PCR-amplified DNA fragments with flanking homology, or (3) using the CRISPR (Cas9-sgRNA) technology. The first two methods required selection of a marker included in the incoming/editing DNA; no selection was required for the third method. All constructs were authenticated by DNA sequencing.

Purification of Flp and Flp-SUMO

Native Flp, Flp-HA-His8 and Flp-SUMO-HA-His8, as well as mutant derivatives of the tagged proteins, were overexpressed in E. coli cells using the pBAD system (Invitrogen). Purification of untagged Flp was carried out using previously described procedures [43–45]. Purification of the tagged proteins included an additional first step of nickel chromatography, followed by dialysis to remove the imidazole present in the elution buffer. The final preparations were ≥ 85% pure, as judged by SDS-PAGE and densitometric scanning of the Coomassie Blue stained bands. Protein concentrations in the final preparations were determined using the Bradford assay.

Analyses of colony morphology and plasmid loss

Overnight cultures were prepared from purified single transformant colonies containing individual 2-micron circle-derived plasmids by growing them selectively (in medium lacking adenine) at 30°C. These cultures were diluted in YEPD medium (n = 0; 10⁴ cells/ml) and grown for 10 generations (n = 10) at 30°C. Aliquots were plated out from n = 0 and n = 10 cultures on YEPD medium. Nibbled and mini-colonies were counted after incubating the plates for 5 days at 26°C. A founder cell that had lost the plasmid, and the plasmid-borne ADE2 marker, gave rise to a fully red (non-sectored) smooth colony. Such colonies were excluded from the total population in calculating the fraction of nibbled colonies. Plasmid loss rate per generation ‘I’ (for instability) was estimated from plates incubated at 30°C based on fully red and total colony counts. I = (1/10) x [ln (f₀/f₁₀)] [46], where f₀ and f₁₀ are the fractions of plasmid containing cells (yielding colonies other than the fully red ones) at n = 0 and n = 10, respectively. The sample size for the individual estimates of plasmid loss rate and the fraction of nibbled or mini-colonies was a minimum of 800 colonies. The assays were performed as at least three repetitions.

Real-time PCR for plasmid copy number determination

The PCR protocols followed those described by Chen et al. [31], and utilized the same plasmid and reference chromosomal amplicons as well as the primer pairs described by them. The amplification reactions were carried out with ABI PRISM 7900HT SDS using the SYBR Green Master Mix (Applied Biosystems). The number of cycles required to reach the C_T number (preset threshold) for each DNA sample was calculated from six separate experiments. The relative change in the copy number of a plasmid between two strains, normalized to the chromosomal reference sequence, was calculated by the 2^-ΔΔCT analysis [47].

Protein stability assays

The expression cassette for Flp-SUMO controlled by the GAL1 promoter was inserted at the TRP1 locus (thus disrupting it) on chromosome IV in [Cir⁰] wild type, siz1Δ siz2Δ, slx5Δ and slx8Δ strains. Aliquots of raffinose-grown overnight cultures were inoculated into raffinose medium, grown to mid-log phase at 30°C, and induction was performed by transferring them to 2% galactose with continued incubation at 30°C. The control (uninduced) cells were transferred from raffinose to glucose medium and incubated at 30°C. At 2hr, cells were spun down, washed, and suspended in TE buffer before adding cycloheximide (100 μg/ml) to arrest protein synthesis. Cells removed at intervals over a 60 min time course were treated with lysis buffer (50 mM HEPES, pH 7.0; 75 mM KCl., 1 mM MgCl₂, 1 mM EGTA, 0.5% Triton X-100, I mM DTT, 1 mM PMSF and one protease inhibitor tablet from Roche/50 ml). Cell extracts prepared by bead beating (5 min; 4°C) were fractionated by 12% SDS-PAGE, and analyzed by quantitative western blotting. Flp-SUMO bands were detected by anti-HA antibody (BioLegend) at 1:1000 dilution, and normalized against actin bands visualized using anti-β-actin antibody (Gene Tex) at 1:1000 dilution.

Protein stability upon proteasome inhibition by MG-132

Proteasome function was inhibited with MG-132 according to published procedures [48]. The following modifications were made to the standard protocols for measuring protein turnover. Overnight raffinose cultures were grown in synthetic medium without ammonium sulfate, and supplemented with 0.1% proline as well as other appropriate amino acids. In addition, the re-inoculation medium for obtaining mid-log phase cells for galactose induction included 0.003% SDS. The induction period was 2 hr, with 75 μm MG-132 (Biomol, Plymouth Meeting, PA) being added at 90 min. Control cells received an equivalent volume of DMSO, the solvent for MG-132. The rest of the procedure—cycloheximide treatment, preparation and fractionation of cell extracts, and western blotting—was performed as described under the previous section on protein stability assays.

Rad52 foci assay for double strand DNA breaks

The experimental strains were derived from [Cir⁰] wild type and siz1Δ siz2Δ strains expressing the GAL1 promoter driven Flp-SUMO (see the section above on protein stability assays) or from an analogous set of strains in which Flp-SUMO was replaced by Flp. The plasmid pADE2-Flp (S1 Fig) was introduced into these strains, and maintained by adenine selection. A CEN-TRP1-plasmid expressing Rad52-YFP from the native RAD52 promoter was also maintained in them by selection. The conditions for Flp or Flp SUMO induction were the same as those described for protein stability estimates in the absence of MG-132. Cells induced for 2 hr in galactose and the corresponding uninduced control cells (2 hr in glucose) were collected, washed and fixed in formaldehyde for scoring fluorescent foci. Each set of assays was repeated three times.

In vitro FRT-binding assays with Flp or Flp-SUMO

The binding assays were performed in 30 μl individual mixtures incubated on ice for 20 min using the buffer conditions described by Prasad et al. [49]. The substrate DNA fragment (0.05 pmol per binding reaction) was 262 bp long, and contained one FRT site. Aliquots were fractionated by electrophoresis in 5% polyacrylamide gels (29:1 crosslinking) at 4°C in 1x TBE duffer. The bound complexes and the unbound substrate were visualized by autoradiography or phosphor imaging.

Recombination, strand cleavage and strand joining assays

The conditions for in vitro recombination were similar to those described previously [50,51]. Each 30 μl reaction mixture contained 0.2 pmol plasmid substrate (with two FRT sites oriented head-to-tail) and 1 pmol of purified Flp or Flp-SUMO. At the end of the 30°C incubation period (from 0.5 to 30 min), the reactions were stopped by treatment with 0.2% SDS (final concentration) followed by proteinase K treatment (50 μg/reaction sample). DNA purified by chloroform-phenol extraction and ethanol precipitation was digested with NdeI and EcoRV. The digestion products were separated by 1% agarose gel electrophoresis, and DNA bands were visualized by ethidium bromide staining.

Strand cleavage and strand joining reactions were carried out in the recombination buffer with 0.05 pmol of the respective ³²P-labeled half-site substrates per reaction and Flp or Flp-SUMO ranging from 0.2 pmol to 2 pmol. At the end of 30 min incubation at 30°C, reactions were stopped by adding 0.2% SDS, and processed without proteinase K treatment. The cleavage and joining reactions were analyzed by electrophoresis in 12% SDS-polyacrylamide (29:1 crosslinking) and 12% polyacrylamide-urea (19:1 crosslinking) gels, followed by phosphorimaging or autoradiography.

Strand cleavage by a complementing pair of R191A and Y343F mutants; strand joining by a Y343F mutant in the presence of an R191A mutant

First, tubes were set up in pairs on ice with one tube within a pair containing the ³²P-labeled half-site plus the R191A mutant, and the other containing the same labeled half-site plus the Y343F mutant. The amounts of half-site and protein in each tube were 0.05 pmol and 0.5 pmol, respectively, in 15 μl of 1.5x recombination buffer. Following 10 min on ice to allow full occupancy of the half-site by protein, 7.5 μl each of the binding mixture were withdrawn from each set of paired tubes, and transferred simultaneously to fresh tubes (maintained at 30°C) containing 2.5 pmol of an unlabeled DNA fragment with one FRT site in 15 μl 0.5x recombination buffer. The contents were gently mixed in each tube and incubated for 30 min. Except for the difference in the substrate half-sites, the strand cleavage and joining reactions were similar in other respects. The reactions were analyzed by SDS-polyacrylamide gel (12%) electrophoresis (for cleavage) and by polyacrylamide-urea gel (12%) electrophoresis (for joining).

Detection and quantitation of DNA and protein bands

Radioactively labeled DNA bands, captured on a phosphor storage screen (Bio-Rad), were scanned using a Typhoon Trio Phosphorimager (GE-Healthcare). Unlabeled DNA bands were visualized in agarose gels by ethidium bromide staining. Protein bands were detected in western blot analyses using Pierce^TM ECL protocol (ThermoFisher Scientific). Image analysis and quantitation of band intensities were performed using the software Quantity One (Bio-Rad; version 4.5.1). For recombination, strand cleavage and strand joining assays, the extent of reaction was estimated as the ratio of the intensity of product band(s) to the sum of the intensities of substrate and product bands. For DNA binding, the ratios of the bound C-I and C-II complexes to the sum of C-I, C-II and unbound DNA were determined. Protein bands were quantitated against actin as the internal control. Multiple exposures were used to compensate for large intensity differences between individual bands. Appropriate correction factors were applied, based on the linear ranges of intensity variation.

Substitution of Flp-SUMO fusion for Flp suppresses Flp-induced cell lethality in a siz1Δ siz2Δ strain

Impairment in the regulation of Flp-mediated amplification leads to high 2-micron plasmid copy number [31,35,52], which induces characteristic nibbling at colony edges. This phenotype, which is more conspicuous at 20°C than at 30°C, is due to differences in plasmid copy number in individual cell lineages, resulting in variable growth inhibition and cell mortality among them. Loss of plasmid restores normal growth and smooth edges. Over time, plasmid-free [Cir⁰] cells tend to rise in the population. Thus, colony morphology and plasmid loss rates are reliable reporters of the mean plasmid load carried by cells, and indirectly of the Flp level/activity in them.

The steady-state level of sumoylated Flp in a wild type strain is ~10% of total Flp [31]. The predominant site of SUMO conjugation, mediated by Siz1 and Siz2, is Lys-375 [31] located < 50 amino acids upstream of the carboxyl-terminus (Ile-423). Replacement of Lys-375 by arginine partially recapitulates the effects of siz1Δ siz2Δ in a wild type background, yielding ~4-fold increase in Flp, ~2-fold higher plasmid copy number, and consistently more abnormal colonies on plates incubated at 20°C [31]. Given the relative proximity of Lys-375 and Ile-423, we suspected that Flp containing the SUMO moiety (amino acids 1–96) as a carboxyl-terminal extension is likely to functionally mimic Flp(K375-SUMO). If so, Flp-SUMO may justifiably be utilized as a surrogate for Flp(K375-SUMO) in addressing potential differences between native and modified Flp in their relative stability in vivo as well as their DNA recognition and catalytic properties in vivo and/or in vitro. It is nearly impossible to study exclusively the naturally sumoylated Flp in a cell, as it would be diluted out by the excess unmodified version. In addition, the extent of the modification (~10%) makes it technically quite challenging to obtain sufficient quantities of Flp(K375-SUMO) for in vitro analyses.

In order to test whether Flp-SUMO can redress the effects of siz1Δ siz2Δ on colony morphology, we transformed [Cir⁰] strains with 2-micron plasmid derivatives engineered to express Flp or Flp-SUMO under the control of the native FLP-promoter (S1 Fig). Except for manipulations of the FLP locus and an insertion of the ADE2 marker, the reporter plasmids retained the overall organization of the native 2-micron plasmid genome. Note that the expressed Flp, Flp-SUMO and their variants contained the HA-His8 epitope tag at their carboxyl-termini. For simplicity, these proteins as well as the plasmids expressing them are referred to without mentioning the tag.

The large fraction of nibbled colonies (> 75% at 26°C) in the siz1Δ siz2Δ strain containing pADE2-Flp was substantially reduced (~1%) when the strain harbored pADE2-Flp-SUMO (Fig 2A and 2B; S3 Table). The enlarged images of colonies shown above S3 Table highlight the difference between smooth and nibbled edges. The frequency of mini-colonies in the population, signifying highly retarded cell growth or extensive cell death, also showed a corresponding reduction (from ~18% to ~3%) (S3 Table). The nibbled- and mini-colony phenotypes were consistent with a ~5-fold increase in the mean copy number of pADE2-Flp in siz1Δ siz2Δ compared to the wild type (S2 Fig). For comparison, the increase in the native 2-micron plasmid copy number in the mutant strain was ~10-fold (S2 Fig). As Flp-SUMO is recombination-competent (S3 Fig), the lack of nibbling in siz1Δ siz2Δ harboring pADE2-Flp-SUMO was not due to Flp-SUMO being inactive in generating FRT-nicks, which are intermediates in the recombination reaction (and potential initiators of BIR).

Download:

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

Fig 2. Colony morphology in the presence of plasmids expressing Flp, Flp-SUMO or Flp(H305L) and plasmid stability.

The 2-micron plasmid derivatives pADE2-Flp (A), pADE2-Flp-SUMO (B) and pADE2-Flp(H305L) (C) are schematically illustrated in S1 Fig, and their relevant features described in the legend. The start of non-selective growth in YEPD liquid medium, following overnight growth in medium lacking adenine, is indicated by n = 0. Additional growth for ten generations is indicated by n = 10. The colony morphologies shown here (A-C) (see also S3 Table) are from YEPD plates incubated at 26°C for 5 days. Only white colonies (composed of plasmid containing cells), or fully white patches in colonies with red sectors, were scored for nibbling. Red colonies resulting from plasmid loss were omitted. Magnified views of representative normal and aberrant colonies are shown above S3 Table The plasmid loss rate per generation (D) was calculated from colonies incubated for 3–4 days at 30°C. An asterisk indicates p < 0.05 by the two-tailed test for significance. ‘WT’ denotes the wild type strain (in other figures as well).

https://doi.org/10.1371/journal.pgen.1008193.g002

The catalytic variant Flp(H305L) expressed by pADE2-Flp(H305L) (S1 Fig) is strongly defective in strand joining in vitro and in vivo in yeast [53–55], and is expected to cause an accumulation of the FRT-nicked intermediate. The variant is inactive in recombination. The presence of pADE2-Flp(H305L), contrary to expectation, produced few nibbled (1–2%) or mini-colonies (~3%) either in the wild type or in the siz1Δ siz2Δ strain at 26°C (Fig 2C; S3 Table). Presumably, sumoylation of Flp(H305L) at Lys-375 in the wild type strain was sufficient to suppress excessive FRT-nicking. The lack of nibbling even in the siz1Δ siz2Δ strain was likely due to high pADE2-Flp(H305L) missegregation, signified by the relative abundance of red (ade2) and red-sectored colonies (Fig 2C) (see also plasmid loss rates in Fig 2D). As a result, mother cultures (n = 0 in Fig 2) would be enriched in cells with low copy numbers of pADE2-Flp-SUMO as well as plasmid-free cells capable of resuming growth when provided with adenine. Such cells would be further enriched during non-selective growth (from n = 0 to n = 10) due to their fitness advantage. In fact, overexpression of Flp(H305L) in [Cir⁺] cells is a convenient method for rapidly curing them of the endogenous 2-micron plasmid [56].

In sum, the aberrant cell growth typical of under-sumoylation of Flp is alleviated by the expression of Flp-SUMO from the native FLP-promoter. The lack of anticipated nibbling in the siz1Δ siz2Δ strain from Flp(H305L) expression is the result of the high rate of plasmid loss induced by this Flp variant.

Flp or Flp(H305L), but not Flp-SUMO, increases plasmid loss in a siz1Δ siz2Δ strain

As equal segregation of 2-micron plasmid molecules occurs in physical association with chromosomes [8–11], the high molecular weight hyper-amplified plasmid concatemers in a siz1Δ siz2Δ strain are likely to interfere with this process. Furthermore, the deficiency in sumoylation of Rep1 and Rep2 partitioning proteins might have an additional effect on segregation [32]. Plasmid-free cells, having higher fitness, will tend to outgrow their plasmid-containing counterparts. ‘Apparent’ plasmid stability during non-selective growth provides a reasonable test of the potential salutary effect of Flp-SUMO under conditions that proscribe normal sumoylation of Flp.

The pADE2-Flp plasmid showed a significantly higher loss rate in the siz1Δ siz2Δ strain compared to the wild type (Fig 2D; left pair of histograms). By contrast, there was a modest improvement in the stability of pADE2-Flp-SUMO in the mutant compared to the wild type (Fig 2D; middle pair of histograms). The lower basal stability of pADE2-Flp-SUMO than pADE2-Flp in the wild type strain (Fig 2D; green histograms of the left and middle pairs) might result from the larger size of pADE2-Flp-SUMO, the particular modification of the FLP locus, or from potential additional sumoylation at Lys-375 (which would be ameliorated by siz1Δ siz2Δ). Consistent with the expected increase in steady state DNA damage at FRT and the attendant plasmid hyper-amplification, pADE2-Flp(H305L) was more unstable than pADE2-Flp in the wild type strain (Fig 2D; green histograms of the left and right pairs). The instability of pADE2-Flp(H305L) was worsened by siz1Δ siz2Δ (Fig 2D; right pair of histograms), suggesting that Flp(H305L), analogous to Flp, is also downregulated via sumoylation.

The comparable loss rates of pADE2-Flp(H305L) and pADE2-Flp in the wild type and the siz1Δ siz2Δ strains, respectively, (Fig 2D; orange histogram of the left pair and green histogram of the right pair) might appear to suggest that normal sumoylation of Flp(H305L) and strongly reduced sumoylation of Flp are more or less equivalent with respect to the amount of strand nicks that the two proteins produce at FRT. However, unlike Flp, Flp(H305L) cannot resolve an amplified plasmid concatemer by recombination, nor can it counter plasmid missegregation by recombination-mediated amplification (Fig 1A and 1B). These factors may aggravate the instability of pADE2-Flp(H305L) in the siz1Δ siz2Δ strain.

The plasmid stability results demonstrate that the increased formation and expansion of plasmid-free cells triggered by FRT-nicks in a siz1Δ siz2Δ host is strongly suppressed when this strain expresses Flp-SUMO instead of Flp or Flp(H305L).

Cell lethality caused by Flp(H305L) in conjunction with mus81Δ or yen1Δ is efficiently reversed by replacement of Flp(H305L) with Flp(H305L)-SUMO

The BIR pathway promotes the repair of one-ended double strand breaks—such as those resulting from the arrival of a replication fork at a Flp-nicked FRT site [31,33,39,57,58]. Consistent with the mechanism diagrammed in Fig 1C, the formation of amplified high molecular weight plasmid DNA is dependent on strand cleavage by Flp, and requires Pol32 as well as Rad proteins involved in BIR [33]. The initial D-loop intermediate of BIR formed by strand invasion of homologous DNA may be processed into a replication fork. Alternatively, it may mature into a Holliday junction by convergence with a replication fork from the opposite direction (Fig 3A; left). The coalescence of two D-loops expanding in opposite directions would generate a double Holliday junction (Fig 3A; right). The organization of the two FRT sites and the location of the bi-directional replication origin within the 2-micron plasmid provide opportunities for a Flp-mediated BIR D-loop to meet a replication fork, or a second such D-loop, approaching it head-on (Fig 3A). As the Flp-induced single strand nicks may occur at one or both of the plasmid FRT sites, and on either DNA strand within a site, head-to-tail configuration of two D-loops or a D-loop and a replication fork is also possible. Crystal structures and in vitro experiments rule out double strand breaks at FRT by Flp [50,59–61], minimizing the probability of two-ended double strand break repair in FRT DNA damage.

Download:

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

Fig 3. Flp(H305L)-SUMO, in contrast to Flp(H305L), mollifies the requirement for Mus81 and Yen1 in cells containing chromosomal plus extra-chromosomal FRT sites.

A. The single FRT site inserted into chromosome VI between replication origins ARS606 and ARS607 [53] is indicated at the top left. In the [Cir⁰] strain, this is the only target for strand cleavage by Flp(H305L) or Flp(H305L)-SUMO. Additional FRT targets are presented by multiple copies of the 2-micron plasmid in the [Cir⁺] strain (top right). The schematic diagrams below depict the consequences of BIR triggered by strand nicks at FRT coupled to DNA replication. The merger of an expanding D-loop with an oncoming replication fork (left) will result in a repair intermediate with a DNA branch at one end. The encounter between two expanding BIR-induced D-loops (right) will generate an intermediate with DNA branches at both ends. Each of these branches can mature into a Holliday junction via strand ligation. Resolution of these intermediates may be performed by Yen1 or Mus81-Mms4 or both. B-D. Serial dilutions of [Cir⁰] or [Cir⁺] cells, uninduced (glucose) or induced (galactose) for the expression of Flp(H305L) or Flp(H305L)-SUMO, were spotted on YEPD plates, and growth was examined after 3-day incubation at 30°C. The inoculum at the left-most spots was ~5 x 10⁴ cells, with the spots to their right representing sequential 10-fold dilutions.

https://doi.org/10.1371/journal.pgen.1008193.g003

The resolution of specialized DNA structures such as Holliday junctions and D-loops in yeast require Yen1 and/or Mus81-Mms4 activities [53]. Induction of Flp(H305L) in a strain containing an FRT site inserted between two strong replication origins in a chromosome results in poor viability in the yen1Δ mus81Δ background [53]. We utilized isogenic [Cir⁰] and [Cir⁺] strains harboring an identical chromosomal insertion of FRT to test whether the presence of additional copies of plasmid-borne FRT sites would aggravate the Flp(H305L)—yen1Δ or the Flp(H305L)-mus81Δ effect, and whether cell survival can be improved by replacing Flp(H305L) by Flp(H305L)-SUMO.

In the [Cir⁰] strain, mus81Δ caused a decrease in colony forming units upon Flp(H305L) induction, with yen1Δ and yen1Δ mus81Δ displaying a stronger effect (Fig 3B). The loss of viable colonies from each single mutation as well as the double mutation was magnified in the [Cir⁺] strain (Fig 3C). Expression of Flp(H305L)-SUMO in place of Flp(H305L) restored cell survival in the single mutants to nearly the same level as in the wild type (Fig 3D). The palliative response to Flp(H305L)-SUMO, though not as strong, was evident in the double mutant as well (Fig 3D).

In the absence of Mus81 or Yen1 or both, the branched intermediates of Flp(H305L)-induced BIR (three or four-way DNA junctions) appear to accumulate, to the detriment of the cell. The apparent reduction of these intermediates in the presence of Flp(H305L)-SUMO is consistent with an abatement in the formation of unsealed strand nicks at FRT that precedes the BIR events. Furthermore, these results corroborate the previous inference that the DNA damage induced by Flp(H305L) at plasmid FRT sites may be masked by accelerated plasmid loss from cells.

The pADE2-Flp-SUMO plasmid causes cell lethality, and is destabilized, in the absence of Slx5 or Slx8

The Slx5-Slx8 STUbL (sumo targeted ubiquitin ligase) regulates a wide range of cellular functions in yeast that include gene expression, quality control of nuclear proteins, DNA damage repair, and chromosome stability [62–66]. The coupling to ubiquitin-proteasome systems via Slx5-Slx8-mediated recognition of conjugated SUMO, or native surface features that mimic SUMO, may bring about not only the degradation of particular target proteins but also the functional re-localization of multi-subunit protein machines. Examples include proteolysis of the Mot1 transcription factor [67] or the Matα2 repressor [68], and the relocation of double strand DNA breaks in G1 cells to repair centers stationed at the nuclear periphery [69]. Ubiquitin-dependent proteolysis of sumoylated Flp appears to be one important mechanism for the post-translational regulation of Flp. The levels of Flp, including its SUMO-conjugated form, are increased in slx5Δ and slx8Δ strains [33]. We therefore tested whether the beneficial effects of Flp-SUMO observed in the siz1Δ siz2Δ strain would be reversed in an slx5Δ or slx8Δ mutant.

The slx5Δ and slx8Δ strains containing pADE2-Flp or pADE2-Flp-SUMO (Fig 4A and 4B; S3 Table) formed nibbled colonies in contrast to the wild type strain containing either plasmid and the siz1Δ siz2Δ strain containing pADE2-Flp-SUMO (Fig 2A and 2B; S3 Table). The fraction of mini-colonies caused by pADE2-Flp was also elevated in the slx8Δ mutant compared to the wild type (S3 Table). Mini-colonies induced by pADE2-Flp-SUMO in the slx5Δ and slx8Δ strains were only slightly more than those in the wild type or the siz1Δ siz2Δ mutant (S3 Table).

Download:

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

Fig 4. The pADE2-Flp-SUMO plasmid fails to prevent nibbled colonies and increased frequency of plasmid-free cells in silx5Δ and slx8Δ strains.

The protocols for scoring colony phenotypes (A, B) and estimating plasmid loss rates (C) were similar to those outlined under Fig 2. The cells seeded on plates at n = 0 and n = 10 were grown at 26°C for 5 days for scoring nibbled and mini-colonies (see S3 Table). Values for plasmid loss were derived from colonies on plates incubated at 30°C, and are directly comparable to those in Fig 2D. An asterisk denotes p < 0.05.

https://doi.org/10.1371/journal.pgen.1008193.g004

The loss of pADE2-Flp-SUMO was more rapid in the slx5Δ and slx8Δ strains than in the wild type (Fig 4C) or the siz1Δ siz2Δ (Fig 2D) strains. The slx5Δ or slx8Δ mutation caused instability of pADE2-Flp as well (Fig 4C), the plasmid loss rates being comparable to that in the siz1Δ siz2Δ strain (Fig 2D).

A key point underscored by these findings is that both slx5Δ and slx8Δ, but not siz1Δ siz2Δ, display the phenotypes of aberrant pADE2-Flp-SUMO amplification (Figs 2D and 4C; S3 Table). Furthermore, slx8Δ phenocopies siz1Δ siz2Δ closely when they contain pADE2-Flp (Figs 2D and 4C; S3 Table). The trend is similar for slx5Δ albeit less striking than slx8Δ. Some of the quantitative differences among individual mutants may be reconciled by the involvement of the ubiquitin and SUMO pathways in the pleiotropic regulation of a variety of cellular processes.

The phenotypic responses of the mutants to pADE2-Flp or pADE2-Flp-SUMO suggest that normal regulation of Flp requires sumoylated Flp to be ubiquitinated and channeled to the proteasome for degradation. Furthermore, SUMO conjugated physiologically by Siz1-Siz2 (principally to Lys-375) within Flp and SUMO attached artificially to the Flp carboxyl-terminus are functionally interchangeable in recognition by the Slx5-Slx8 STUbL. As a result, Flp becomes deleterious in siz1Δ siz2Δ, slx5Δ and slx8Δ; Flp-SUMO is harmful only in the latter two. Corroborating these interpretations, the half-life of Flp-SUMO was higher in slx5Δ compared to that in the wild type or the siz1Δ siz2Δ mutant, while Flp-SUMO degradation was almost completely blocked by slx8Δ (Fig 5A and 5B). The turnover of Flp-SUMO in the wild type and siz1Δ siz2Δ was suppressed by the proteasome inhibitor MG-132 (Fig 5C and 5D).

Download:

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

Fig 5. Turnover of Flp-SUMO is slowed down by siz1Δ siz2Δ, slx5Δ, slx8Δ or by proteasome inhibition.

A, B. Flp-SUMO was expressed from a GAL promoter controlled expression cassette inserted into chromosome IV at the TRP1 locus. Following galactose induction for 2 hr, protein synthesis was inhibited by cycloheximide (0 min), and the levels of Flp-SUMO were followed over a 60 min time course by western blot analysis. The lanes marked ‘C’ refer to uninduced cells, incubated in glucose medium for 2 hr. Band intensities were plotted after normalizing them against those for actin as a control. The relative intensity at 0 min was assigned a value of 1.0. Flp-SUMO was visualized in the western blots using an antibody to the HA-tag present at the carboxyl-terminus. Actin was detected using an antibody to the native protein. C, D. The analyses were similar to those in A and B, except that the culture conditions were slightly modified (Materials and Methods), and the proteasome inhibitor MG-132 was added to the experimental samples at 90 min into induction (30 min prior to time 0).

https://doi.org/10.1371/journal.pgen.1008193.g005

Flp-SUMO has weaker recombinase activity than Flp in vitro

In principle, sumoylation may not only control the steady state levels of Flp in vivo but may also modulate its recognition of the FRT site and/or its active site functions. DNA damage and resultant BIR in the 2-micron plasmid may be curtailed by preferential exclusion of sumoylated Flp from FRT sites, reduced strand cleavage and/or enhanced strand joining by the modified protein, or by combinations of these attributes. In order to probe the effects of sumoylation on Flp activities, we utilized in vitro assays using purified Flp and Flp-SUMO. Based on their very similar in vivo responses to multiple mutations in the SUMO/ubiquitin pathways, Flp-SUMO is presumed to be an authentic substitute for Flp(K375-SUMO) in vitro as well. This premise is also supported by structural considerations. In the Flp recombination/synaptic complexes [59,70], the distances from Lys-375 or the carboxyl-terminus (Ile-423) to key active site residues are nearly the same, the maximum difference being ~4 Å (S4 Table). As such, potential structure-function perturbations from the SUMO moiety are not expected to be different between Flp sumoylated at Lys-375 and Flp-SUMO.

In a time-course deletion reaction between two head-to-tail FRT sites, Flp-SUMO was approximately 60% as active as Flp (Fig 6A and 6B). To be consistent with in vivo experiments, the in vitro analyses were performed using proteins containing the HA-His8 tag at the carboxyl-terminus. Native Flp and the HA-HIS8-tagged variant were nearly identical in recombination activities (S4 Fig).

Download:

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

Fig 6. Flp-SUMO is less active than Flp in recombination between head-to-tail FRT sites placed within a circular plasmid.

A, B. The recombination reaction is schematically outlined at the top. The DNA fragments generated by EcoRV plus NdeI digestion of the substrate plasmid (labeled as ‘S’) and the recombinant products (labeled as ‘P’) were distinguished by agarose gel electrophoresis. The assays, performed at 30°C, utilized 0.2 pmol of the substrate plasmid (0.8 pmol binding site for monomeric Flp) and 1.0 pmol of Flp or Flp-SUMO (containing carboxyl-terminal HA-His8 tag). As shown in S4 Fig, there was no measurable difference between native Flp and tagged Flp in recombination activity.

https://doi.org/10.1371/journal.pgen.1008193.g006

The contra-effect of the SUMO moiety on recombination may occur during Flp-FRT association and assembly of the recombination complex, or within a pre-assembled complex. Allosteric interactions and collaborative assembly of the strand cleavage active site between neighboring FRT-bound Flp monomers are key features of the recombination reaction, which is carried out within a synaptic structure containing four Flp monomers bound to two partner FRT sites [50,59]. Thus, sumoylation of just one Flp monomer in vivo, out of two bound to an FRT site or four bound to a pair of synapsed FRT sites, may protect against DNA damage at FRT by attenuating Flp activity. The role of monomer-monomer interactions in Flp catalysis is dealt with in more detail in the context of the strand cleavage and strand joining activities of Flp (see below).

Flp has higher FRT binding efficiency and binding cooperativity than Flp-SUMO

The weaker recombinase activity of Flp-SUMO than Flp could, at last in part, be due to its impaired recognition of the FRT site. Flp binds as a monomer to each of the two 13 bp head-to-head binding elements flanking the 8 bp strand exchange region of a minimal FRT site, and cooperative interactions between the bound monomers stabilize the FRT-Flp dimer complex. We tested Flp and Flp-SUMO for differences in the formation of singly or doubly bound complexes with a DNA fragment containing one copy of the FRT site. These two complexes were resolved by their electrophoretic mobility in native polyacrylamide gels.

At identical molar ratios of protein per binding element, the combined amount of monomeric (C-I) and dimeric (C-II) complexes formed were lower for Flp-SUMO than Flp (Fig 7A and 7B). It took approximately twice as much Flp-SUMO as Flp to convert the same amount of FRT-DNA into the bound form. Furthermore, there was a striking difference in binding cooperativity between Flp-SUMO and Flp. At 1:1 molar ratio of protein to binding element (lane 2 in Fig 7A), C-II formation by Flp was favored over C-I by a factor of ~3. By contrast, Flp-SUMO formed ~2-fold more C-I than C-II under the same condition (lane 2, Fig 7B).

Download:

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

Fig 7. FRT-binding is weakened in Flp-SUMO.

A-C. Stepwise association of individual Flp monomers to the two binding elements (arrows oriented head-to-head) of an FRT site is diagrammed schematically at the top. The binding reactions contained 0.05 pmol of the DNA containing one FRT site (labeled as ‘S’), equivalent to 0.1 pmol Flp-binding element, and the indicated amounts of Flp or Flp-SUMO. The binding mixtures were incubated on ice for 20 min, and the unbound and bound DNA were fractionated by electrophoresis in native 5% polyacrylamide gels. The DNA-protein complexes corresponding to the occupancy of one binding element and both binding elements are called C-I and C-II, respectively. The subscripts ‘F’ and ‘FS’ refer to Flp and Flp-SUMO, respectively.

https://doi.org/10.1371/journal.pgen.1008193.g007

The disparity in FRT binding and cooperativity between Flp and Flp-SUMO was also highlighted by binding reactions containing a mixture of the two proteins (Fig 7C). Even at 1:3 molar ratio, Flp surpassed Flp-SUMO in FRT occupancy, as indicated by the strong reduction of the C-II complex formed by Flp-SUMO and the appearance of the Flp C-I and C-II complexes (lane 3, Fig 7C). At this ratio, the C-I complexes formed by the two proteins were roughly equal. The plot of C-I ratios underestimates the advantage of Flp over Flp-SUMO in FRT binding as it does not capture the much stronger cooperativity of Flp in C-I → C-II conversion (Fig 7A and 7B). The band corresponding to the heterodimeric C-II complex, expected to migrate just above the Flp C-II complex, was not well resolved from it. At equimolar or higher Flp to Flp-SUMO ratios (lanes 4–6, Fig 7C), the C-I and C-II complexes were predominantly the Flp-bound forms.

Extrapolation of the in vitro DNA binding results to the in vivo situation suggests that sumoylation of Flp would diminish its binding to the 2-micron plasmid FRT sites. In addition, or conversely, sumoylation of FRT-bound Flp might accelerate the dissociation of the modified protein from DNA. Either or both of these mechanisms would provide a safeguard against potentially detrimental strand cleavage events at FRT.

Flp-SUMO is attenuated in strand cleavage

The strongly conserved catalytic hexad cluster of the tyrosine site-specific recombinase family (to which Flp belongs), including the invariant tyrosine nucleophile, is represented in Flp by Arg-191, Lys-223, His-305, Arg-308, Trp-330 and Tyr-343. In the shared active site for strand cleavage, Tyr-343 from one Flp monomer is donated to the pro-active site of the second neighboring Flp monomer, which provides the other five catalytic residues [50,59]. Strand joining in the cleaved intermediate, harboring a 3’-O-phosphotyrosyl bond, is promoted by one Flp monomer utilizing the 5’-hydroxyl group from DNA as the nucleophile [50,71]. In order to appraise whether sumoylation affects the chemical competence of the Flp active site, we assayed Flp and Flp-SUMO in strand cleavage and strand joining reactions in vitro.

Reactions were performed using half-site substrates containing a single Flp binding element and one scissile phosphate [71–74]. Interactions between two Flp monomers, each bound to a half-site, permits the assembly of the shared active site required for strand cleavage by the Tyr-343 nucleophile [50]. The half-sites are suitably configured, as described below, to avoid interference from the joining reaction during cleavage assays and vice versa.

In the strand cleavage substrate, the scissile phosphate is followed by a truncated 3 nt segment from the strand exchange region (Fig 8A). The full 8 nt complement of the exchange region in the opposite strand ends in 5’-phosphate, which prevents strand joining by a hydroxyl group at this position. Furthermore, diffusion of the unstably hydrogen bonded trinucleotide product of cleavage away from the reaction center minimizes cleavage reversal.

Download:

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

Fig 8. The strand cleavage and joining efficiencies are lower for Flp-SUMO than Flp.

A. In the schematic diagram of the strand cleavage substrate (S), the single Flp binding element is signified by the parallel lines ending in an arrowhead, with ‘P’ indicating the scissile phosphate. The 5’-³²P label on the strand containing ‘P’ is shown by the asterisk. The 5’-hydroxyl group on the complementary strand is blocked by phosphorylation. Each reaction, containing 0.05 pmol half-site plus the indicated amounts of Flp or Flp-SUMO, was incubated for 30 min at 30°C, and was analyzed by electrophoresis in 12% SDS-polyacrylamide gels. The cleavage products formed by Flp and Flp-SUMO are labeled as CP_F and CP_FS, respectively. In the plots of the product yields from three separate experiments with each protein, the results with native Flp (lacking the HA-His8 tag) are also included. B. In the schematic diagram of the strand joining substrate (5’-³²P-labeled; asterisk), the tyrosyl group attached to the scissile phosphate mimics the strand-cleaved recombination intermediate. The free 5’-hydroxyl group on the non-radioactive strand attacks the phosphotyrosyl bond in a ‘pseudo-joining’ reaction to give a hairpin product. The reactions were performed under the conditions employed for cleavage, and were analyzed by electrophoresis in 12% denaturing polyacrylamide gels. The hairpin products formed by Flp and Flp-SUMO are identical, and are labeled as ‘HP’.

https://doi.org/10.1371/journal.pgen.1008193.g008

In the strand joining substrate, the scissile phosphate is covalently linked to a tyrosyl moiety (Fig 8B) to mimic the phosphotyrosyl bridge between Flp and DNA formed during strand cleavage. The hairpin product of the joining reaction is refractory to cleavage, as it cannot support the Flp-Flp interactions required for active site assembly in trans.

Flp-SUMO was less active than Flp in strand cleavage as well as strand joining. The half-maximal reaction required a ~3-fold higher amount of Flp-SUMO for both steps (Fig 8A and 8B). Flp-SUMO did catch up with Flp in its V_max for joining, and required a ~3-fold higher protein amount. However, at a ~5-fold higher amount, Flp-SUMO was close to (but still below) saturation cleavage by Flp. Thus, the SUMO attachment reduces the catalytic activity of Flp, the adverse effect on strand cleavage being stronger than that on strand joining.

Collectively, the binding and activity assays (Figs 6–8) suggest that the differences in the catalytic efficiencies of Flp and Flp-SUMO can be accounted for primarily by reduced FRT binding affinity and cooperativity imposed by the modification, perhaps with some impairment of active site function in addition. The latter possibility was tested more directly by strand cleavage complementation and strand joining assays using cleavage-incompetent Flp mutants (described below).

Flp-SUMO is less competent than Flp in the phosphate activation step of strand cleavage, but not so in the tyrosine donation step

The two steps of single strand exchange during Flp recombination involves the assembly of the shared cleavage pocket (and trans-donation of Tyr-343) within individual FRT sites or between synapsed partner FRT sites [50,59] (Fig 9A). The shared active site makes it possible to catalytically complement a Flp monomer lacking Tyr-343 with one lacking a pentad residue within its pro-active site [50,71] (Fig 9B). For example, Flp(Y343F) bound to a half-FRT site can activate its scissile phosphate, which may then be cleaved by Tyr-343 from Flp(R191A) bound to a second half-FRT site. The individual mutants themselves are inactive in cleavage. We took advantage of catalytically complementing Flp and Flp-SUMO partners to ask whether Flp-SUMO is affected in the phosphate activation step or in the tyrosine donation step of strand cleavage.

Download:

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

Fig 9. Flp-SUMO is attenuated at the phosphate activation step of strand cleavage but not at the Tyr-343 nucleophile donation step.

A. Strand cleavage complementation by neighboring Flp monomers is schematically illustrated in the context of the recombination synapse. The four Flp monomers bound to the synapsed FRT sites assemble two active sites for strand cleavage at a time. Two monomers activate the scissile phosphates (green dots), to be cleaved by Tyr-343 nucleophiles received from the other two monomers. The scissile phosphates refractory to cleavage are shown as red dots. The Holliday junction formed by the first strand cleavage/exchange step isomerizes to switch the Tyr-343 donors to recipients and vice versa. The ensuing second strand cleavage/exchange step completes recombination. The FRT sites are oriented by marking them left (L) to right (R). B. The schematic diagram at the top shows the delivery of Tyr-343 from Flp(R191A) to cleave the half-site bound by Flp(Y343F). The indicated proteins were pre-bound separately to the 5’-³²P-labeled half-site substrate (S), and mixed together to assay cleavage. The reactions were analyzed by SDS-polyacrylamide gel electrophoresis as described under Fig 8. The cleavage product formed by Flp(R191A) and Flp(R191A)-SUMO are indicated as CP_F and CP_FS, respectively. The normalized cleavage yields from three separate assays are plotted as histograms with the output from the Flp(R191A)-Flp(Y343F) reaction as 1.0 C. The strand joining reaction in a half-site mediated by a bound Flp(Y343F) monomer is schematically diagrammed at top. The presence of a half-site bound by Flp(R191A) may potentially modulate the joining efficiency of Flp(Y343F) via monomer-monomer interactions. Flp(R191A) itself has no joining activity. The ³²P-labeled half-site was bound separately by the mutants, and then mixed. The hairpin product (HP) and the unreacted substrate (S) were separated by denaturing polyacrylamide gel electrophoresis (see Fig 8). In the histogram plots representing three independent assays, the strand joining efficiency is normalized against a value of 1.0 for the Flp(Y343F) reaction containing Flp(R191A).

https://doi.org/10.1371/journal.pgen.1008193.g009

The reactions were done by pre-binding the 5’-end labelled half-site separately with each one of a pair of complementing mutants, and mixing them in the presence of an excess unlabeled FRT to soak up unbound proteins (Fig 9B). With Flp(R191A) as the tyrosine donor, the cleavage output was higher (~1.6-fold) with Flp(Y343F) than with Flp(Y343F)-SUMO as its partner (lanes 1 and 2, Fig 9B). A similar difference in cleavage (~1.5-fold) was noted between Flp(Y343F) and Flp(Y343F)-SUMO when the tyrosine donor was Flp(R191A)-SUMO (lanes 3 and 4, Fig 9B). Flp(R191A)-SUMO was fully competent at tyrosine donation, yielding better cleavage than Flp(R191A) when partnered with Flp(Y343F) (~1.5-fold higher; lanes 1 and 3, Fig 9B) or with Flp(Y343F)-SUMO (~1.6-fold higher; lanes 2 and 4, Fig 9B). The relative cleavage efficiencies are shown as histogram plots with a value of 1.0 assigned for the Flp(R191A)-Flp(Y343F) pair.

Thus, SUMO attachment to Flp diminishes its ability to activate an adjacent scissile phosphate, and consequently causes a decrease in the probability of its cleavage. However, Flp-SUMO is more active in Tyr-343 donation for cleavage of an activated scissile phosphate. The two opposing effects would more or less cancel out. Taken together, FRT binding, FRT cleavage and cleavage complementation data suggest that weakening the cleavage potential per se of Flp by sumoylation is unlikely to play a significant role in protecting FRT against strand nicks in vivo. More prominent is the reduced FRT binding affinity and cooperativity resulting from the modification. Note that these effects would have been minimized in the complementation assays, as they utilized pre-bound half-sites.

Flp and Flp-SUMO are equally efficient in the strand joining step

The strand joining reaction, executed by a single Flp monomer, does not require the active site Tyr-343 [72,75]. Flp and Flp(Y343F) are equally efficient in this reaction. We tested strand joining in half-site substrates pre-bound by Flp(Y343F) or Flp(Y343F)-SUMO, thereby avoiding the obfuscating effects on DNA binding.

A half-site bound by the Y343F mutant was mixed with one bound by the R191A mutant (Fig 9C), exactly as in cleavage complementation (Fig 9B). As the mutant pair may engage in the interactions responsible for the assembly of the shared active site, the native conditions under which joining occurs are recapitulated. Furthermore, this experimental design makes it possible to test whether the joining activity of a given Y343F mutant protein is differentially affected by whether the partner R191A mutant contains SUMO or not.

Flp(Y343F) and Flp(Y343F)-SUMO gave equal amounts of the hairpin product when the reaction also contained the Flp(R191A)-bound half-site (lanes 1 and 2, Fig 9C) or the half-site bound by Flp(R191A)-SUMO (lanes 3 and 4, Fig 9C). The histogram plots show the relative joining efficiencies for each protein pair normalized against a value of 1.0 for Flp(Y343F) in the presence of Flp(R191A).

Comparable joining activities of native Flp and Flp-SUMO would suggest that enhanced strand joining (or cleavage reversal) by SUMO-modified Flp is probably not a contributing factor in lowering the steady state levels of FRT nicks in vivo.

Fewer strand nicks at plasmid FRT sites by Flp-SUMO compared to Flp in siz1Δ siz2 Δ cells are consistent with lower Flp-SUMO occupancy of these sites

The 2-micron plasmid molecules are organized in vivo into nucleosome-beaded mini-chromatin circles [76–78]. There is concern that the in vitro behavior of Flp-SUMO towards FRT on naked DNA may not be strictly relevant in vivo. DNase I sensitivity assays [79] suggest that FRT sites in the 2-micron plasmid are relatively nucleosome-free. In order to verify that the in vitro differences between Flp and Flp-SUMO are valid in vivo, we estimated the amount of strand nicks at FRT elicited by a relatively brief but sustained burst of induction of each protein in a cell biological assay. This assay, though indirect, avoids potential problems in quantitating FRT nicks by alkaline gel electrophoresis posed by variabilities in the extraction of DNA covalently linked to Flp.

The experimental [Cir⁰] strains, expressing RAD52-YFP, contained the pADE2-Flp plasmid, which has two FRT sites per molecule (S1 Fig) and a copy number similar to that of the 2-micron plasmid (S2 Fig). Flp or Flp-SUMO was expressed in this strain from the same chromosome locale (TRP1; Chromosome IV) under GAL promoter control. The small background level of Flp expressed by its native promoter from pADE2-Flp does not interfere with the assay, as it would be swamped out by the galactose-induced levels of Flp or Flp-SUMO. Strand nicks formed at FRT were visualized indirectly as Rad52-YFP foci associated with the double strand breaks that such nicks give rise to upon DNA replication [80].

In the wild type strain, the percentage of cells containing Rad52-YFP foci were low (~6% or lower; Fig 10A and 10B) with or without induction of Flp or Flp-SUMO. The foci in uninduced (glucose- grown) cells represent background DNA damage plus any damage at plasmid FRT sites due to Flp expressed from pADE2-Flp. The percentage of foci-containing cells was higher in the glucose-grown siz1Δ siz2Δ strain, (~12 to ~18%) (Fig 10A and 10B). Presumably, this increase signifies general DNA repair deficiency caused by the mutations combined with increased FRT strand nicks due to deficient sumoylation of the plasmid-expressed Flp. Induction of Flp by galactose raised the fraction of foci-containing cells from ~18% to ~29% (Fig 10A), signifying a further increase in the frequency of strand nicks at FRT. By contrast, there was no such increase upon similar induction of Flp-SUMO (Fig 10B). Our findings are consistent with a previous demonstration of increased Rad52 foci formation due to siz1Δ siz2Δ in a [Cir⁺] strain compared to the [Cir⁰] control [33]. In a strain lacking galactose inducible Flp or Flp-SUMO, but containing the pADE2-Flp plasmid, the fractions of Rad52 foci containing cells were not significantly different in glucose- versus galactose-grown siz1Δ siz2Δ (Fig 10C). This was also the case for the wild type, as expected.

Download:

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

Fig 10. Flp-SUMO mediates fewer strand nicks at plasmid FRT sites in vivo than Flp.

Representative cells containing one, two or more than two Rad52-YFP foci per nucleus (DAPI-stained chromosomes) are shown at the top. Most foci-containing cells displayed a single focus (75–80%). The remainder contained predominantly two foci, and only very few contained three or more. A, B. Rad52-YFP foci were counted in log-phase cells incubated in medium containing glucose (uninduced) or galactose (induced for Flp or Flp-SUMO expression) for 2 hr at 30°C. Cells were fixed in formaldehyde immediately after the incubation period, and examined by fluorescence microscopy. The plotted data are derived from >250 cells for each pair of histograms representing an experiment. The asterisk indicates p < 0.05. C. The assays were done as in A and B, except that the strains did not contain the GAL promoter controlled expression cassettes for Flp or Flp-SUMO. D. The levels of the induced proteins were assayed by western blotting, with actin as the control for normalization, as described under Fig 5. Both Flp and Flp-SUMO were detected using an antibody to the HA-tag present in them.

https://doi.org/10.1371/journal.pgen.1008193.g010

The striking difference between Flp and Flp-SUMO in foci formation in siz1Δ siz2Δ is consistent with the decreased occupancy of FRT sites by Flp-SUMO, as suggested by the in vitro results. More rapid turnover of Flp-SUMO in siz1Δ siz2Δ compared to Flp cannot explain the difference, as these foci were scored immediately following an optimal 2 hr galactose induction of both proteins. In fact, the induced level of Flp-SUMO at this time point was higher than that of Flp in both the wild type and mutant strains (Fig 10D).

Flp-SUMO is an authentic substitute for Flp(K375-SUMO)

An important advantage of Flp-SUMO is that it permits direct comparison with unmodified Flp in the DNA binding and cleavage properties of the two purified proteins. We have not utilized in vitro SUMO conjugation [82,83] to Flp because of uncertainties in the extent of modification and the potential for conjugation at semi-consensus or non-consensus lysine residues. Structural considerations suggest that the folded SUMO domain linked to Lys-375 or to Ile-423 would be spaced approximately equally from the body of Flp (S4 Table). Furthermore, Flp-SUMO mirrors Flp(Lys-375-SUMO) in vivo in that both require Slx5-Slx8 for their normal regulation, while Siz1-Siz2 can be bypassed in the case of Flp-SUMO but not Flp. Equally compelling is the concordance between the reduced in vitro FRT binding/cleavage by Flp-SUMO (Figs 7 and 8) and the inferred extents of in vivo FRT DNA damage effected by unmodified Flp versus natively modified Flp or Flp-SUMO (Flp >> Flp-SUMO in siz1Δ siz2Δ and Flp(K375-SUMO) = Flp-SUMO in wild type) (Fig 10). Extrapolating the interpretations from Flp-SUMO to Flp(K375-SUMO) is therefore justified on multiple grounds.

Flp-SUMO mimics Flp(K375-SUMO) in blocking BIR-induced hyper-amplification of the 2-micron plasmid

The adverse physiological effects of gross undersumoylation of Flp (siz1Δ siz2Δ)—retarded growth, premature cell death and increased plasmid loss—are consistent with BIR-induced hyper-amplification of FRT-containing plasmids. These phenotypes are strongly ameliorated by switching from Flp to Flp-SUMO expression. According to prior genetic and biochemical work, BIR is triggered by Flp-mediated strand nicks at FRT in conjunction with plasmid replication [31,33]. Mutations in BIR pathways suppress unregulated plasmid amplification [33]. High rates of plasmid loss may occasionally mask the typical BIR phenotypes—as was observed with the hyper-cleaving variant Flp(H305L)—because of the growth advantage of plasmid-cured cells. However, induction of Flp(H305L) leads to cell death when resolution of the branched DNA intermediates of BIR is impaired by yen1Δ, mus81Δ or yen1Δ mus81Δ. Cell killing by Flp(H305L), which correlates with FRT copy number, is strongly curbed when these mutants express Flp(H305L)-SUMO instead. The salvaging effects of SUMO fusion to Flp are dependent on the Slx5-Slx8 STUbL, suggesting that SUMO modification of Flp is followed by ubiquitination and proteasome-mediated degradation (Fig 11B). The half-lives of Flp-SUMO in the wild type and mutant strains, and in the presence of the proteasome inhibitor MG-132, are consistent with this scenario. Natively sumoylated Flp has also been shown to be regulated by Slx5-Slx8 [33]. The cumulative results suggest that the mechanisms responsible for SUMO recognition in Flp(K375-SUMO) and downstream processing are shared by Flp-SUMO as well.

Normal SUMO conjugation and artificial SUMO fusion exert similar biological effects despite the stoichiometric difference in modification

There are precedents for SUMO and ubiquitin fusion proteins satisfying the functional roles of their naturally modified counterparts. Examples in yeast include the recombination protein Rad52 as well as transcriptional regulatory proteins [84–87]. The properties of Flp-SUMO revealed in our studies vis a vis the siz1Δ siz2Δ and slx5Δ or slx8Δ host strains are in general conformity with the behavior of other biologically active SUMO fusion proteins.

While sumoylation may signal protein degradation in certain instances, it may act as a safeguard against degradation in others. The opposing functions may be reconciled if the SUMO-acceptor lysine is also the target for ubiquitination by an STUbL [88]. In one case, the addition of ubiquitin may occur in concert with the removal of SUMO by Ulp1. In the other, the pre-existing SUMO may sterically block the action of a ubiquitin ligase. The apparent deregulation of Flp by interfering with either SUMO conjugation (siz1Δ siz2Δ) [31,33] or deconjugation (ulp1) [34] would fit into the first model. It is not known whether Lys-375 of Flp is the site for both sumoylation and ubiquitination under native conditions. If it is, the functional similarity between Flp and Flp-SUMO suggests that SUMO can activate Slx5-Slx8 in cis or in trans during ubiquitination.

In spite of the large difference in stoichiometry between SUMO conjugation at Lys-375 (~10%) [31] and SUMO fusion at Ile-423 (100%), the modifications are nearly indistinguishable in their biological roles. Dynamic modification of Lys-375 by deconjugation and reutilization of the SUMO moiety, targeted modification of FRT-bound Flp or selective exclusion of modified Flp from FRT may overcome the limitations of substoichiometric modification. Deregulation of Flp by ulp1 [34] would be consistent with the dynamic nature of SUMO conjugated to Lys-375. These general principles—the catalytic nature of the modification, its compartmentalization, and/or its dominance in dimeric or oligomeric protein assemblies—may explain similar biological effects produced by substoichiometric modification of native proteins and stoichiometric modification of the corresponding engineered fusion proteins.

SUMO conjugation lowers FRT-association and FRT-cleavage by Flp

As revealed by in vitro analyses, Flp-SUMO has weaker affinity for FRT than Flp, and is less cooperative in FRT-binding. Furthermore, the underrepresentation of FRT-bound Flp-SUMO in binding reactions with mixtures of Flp and Flp-SUMO suggests negative cooperativity between the two. As strand cleavage requires the collaboration of two Flp monomers [50,59], the in vivo implications of negative cooperativity are significant in the context of a single FRT site or a pair of synapsed FRT sites. The reduced association or enhanced dissociation of a sumoylated Flp monomer would protect FRT from strand cleavage, and the potential initiation of BIR, even if it were stably bound by an unmodified Flp monomer (Fig 11C). Similar dissociation within four recombinase monomers bound to a pair of synapsed FRT sites would cause these sites to disengage from each other. Normal plasmid amplification by Flp-mediated reconfiguration of replication forks can thus be regulated as well (Fig 11D).

There are several examples for the modulation of the DNA binding affinity of proteins by SUMO or ubiquitin conjugation, in particular proteins associated with DNA damage repair [65,89]. The modification promotes protein dissociation from DNA in most instances [90–95], although the reverse trend has also been observed [96–98].The properties of Flp-SUMO follow the general theme of fine-tuning DNA-protein interactions through post-translational modification. However, this is the first time that this phenomenon has been demonstrated in the regulation of a site-specific DNA recombinase.

Self- and host-imposed controls regulate the population of selfish genomes

As an extrachromosomal element, the 2-micron plasmid derives its evolutionary fitness from its high transmission fidelity during cell division, and its ability to restore copy number in case of a glitch during a segregation event [2]. The steady state plasmid copy number of 40–60 molecules in a haploid nucleus is an optimized maximum value. Higher copy numbers reduce the fitness of the host, and indirectly harm the plasmid. Central to copy number maintenance is the regulation of Flp through modulation of gene expression by plasmid-coded proteins (Fig 11A) [24,25,27] and through post-translational modification by the host’s SUMO and ubiquitin conjugation machineries (Fig 11B–11D) [31,33]. The post-translational control is two-pronged. As suggested by previous work [31,33], and confirmed by the present study utilizing Flp-SUMO, proteasome-mediated turnover of Flp, sequentially modified by SUMO and ubiquitin conjugation, lessens the probability of unsealed strand nicks at FRT and unwarranted increase in plasmid copy number via BIR (Fig 11B). At another level, as suggested by the in vitro and in vivo properties of Flp-SUMO, the lower propensity of SUMO-conjugated Flp to stay associated with FRT protects it from excessive strand nicks (Fig 11C). By controlling the level and activity of Flp (Fig 11A–11D), SUMO conjugation would also prevent hyper-amplification of the plasmid by FRT x FRT recombination coupled to plasmid replication (Fig 11D) [21–23].

Thus, safeguards against plasmid overpopulation through moderation of selfishness exercised by the plasmid itself and through preventive measures implemented by the host guarantee the mutual compatibility between a dependent genome and its guardian genome over evolutionary time.