Factors that influence bidirectional long-tract homozygosis due to double-strand break repair in Candida albicans

Abstract

Genomic rearrangements have been associated with the acquisition of adaptive phenotypes, allowing organisms to efficiently generate new favorable genetic combinations. The diploid genome of Candida albicans is highly plastic, displaying numerous genomic rearrangements that are often the by-product of the repair of DNA breaks. For example, DNA double-strand breaks (DSB) repair using homologous-recombination pathways are a major source of loss-of-heterozygosity (LOH), observed ubiquitously in both clinical and laboratory strains of C. albicans. Mechanisms such as break-induced replication (BIR) or mitotic crossover (MCO) can result in long tracts of LOH, spanning hundreds of kilobases until the telomere. Analysis of I-SceI-induced BIR/MCO tracts in C. albicans revealed that the homozygosis tracts can ascend several kilobases toward the centromere, displaying homozygosis from the break site toward the centromere. We sought to investigate the molecular mechanisms that could contribute to this phenotype by characterizing a series of C. albicans DNA repair mutants, including pol32^-/-, msh2^-/-, mph1^-/-, and mus81^-/-. The impact of deleting these genes on genome stability revealed functional differences between Saccharomyces cerevisiae (a model DNA repair organism) and C. albicans. In addition, we demonstrated that ascending LOH tracts toward the centromere are associated with intrinsic features of BIR and potentially involve the mismatch repair pathway which acts upon natural heterozygous positions. Overall, this mechanistic approach to study LOH deepens our limited characterization of DNA repair pathways in C. albicans and brings forth the notion that centromere proximal alleles from DNA break sites are not guarded from undergoing LOH.

Introduction

Genomic instability is defined by the increased frequency of mutations or genetic rearrangements. It facilitates rapid adaptation to changing environments and is often considered a hallmark of pathogenic species and cancer (Maslov and Vijg 2009; Darmon and Leach 2014; Covo 2020). Better understanding of the biological implications of genomic instability has often relied on fundamental mechanistic studies conducted in the baker’s yeast, Saccharomyces cerevisiae, and aimed at evaluating the effect of perturbations in molecular mechanisms involved in DNA maintenance with regards to genomic rearrangements. The opportunistic fungal pathogen Candida albicans is an emerging powerful model organism to study the effect of genome dynamics on fungal pathogenesis (Legrand et al. 2019), notably due to its diploid highly heterozygous and plastic genome (Hirakawa et al. 2015; Ropars et al. 2018; Wang et al. 2018). The genome of C. albicans is highly tolerant to genomic rearrangements (Fischer et al. 2006), as copy number variations and loss-of-heterozygosity (LOH) are frequently acquired upon exposure to stress conditions (Selmecki et al. 2005; Forche et al. 2011; Harrison et al. 2014; Ene et al. 2018). Indeed, LOH of various sizes are ubiquitous to both laboratory and clinical isolates, and can span across several nucleotides or even affect entire chromosomes (Ene et al. 2018; Ropars et al. 2018; Wang et al. 2018). In recent years, several gross chromosomal rearrangements (GCRs) have been associated with adaptive phenotypes, including antifungal resistance (Selmecki et al. 2008; Todd et al. 2019), or niche specialization by augmentation or diminution of virulence in vivo (Tso et al. 2018; Forche et al. 2019; Liang et al. 2019). Thus, understanding the molecular mechanisms leading to genomic rearrangements is key to better understand C. albicans biology and pathogenesis.

Despite the beneficial outcomes of genomic instability, genomic maintenance is essential to ensure the transmission of a complete and faithful DNA sequence to the subsequent generations. All cells recurrently suffer DNA damages caused by intrinsic and extrinsic factors leading to DNA single- or double-strand breaks (DSB), threatening genomic integrity. DNA-DSB repair can be conducted through numerous molecular mechanisms, which can be classified into two major categories: nonhomologous end joining (NHEJ) and homologous-recombination (HR)-mediated repair. NHEJ is described as a “cut and paste” DNA repair mechanism because it involves little (if any) modifications of the broken DNA ends, with the DNA break being identified and ligated together. In contrast, several different pathways of HR-mediated repair exist, all requiring extensive modifications of the broken DNA ends and an intact repair template, such as the homologous chromosome or the sister chromatid, in order to complete the repair. Indeed, as described in S. cerevisiae, HR-mediated repair is a lengthier process requiring DNA damage detection, 5’-end resection leading to single-strand DNA (ssDNA), homology search, ssDNA invasion and DNA synthesis, and therefore necessitates the coordination of several cellular processes. In addition, DNA-DSB repair can be associated with the generation of genomic rearrangements, i.e., LOH, translocations, or copy number variations (Malkova et al. 2000; Ramakrishnan et al. 2018; Piazza et al. 2019). For instance, different repair pathways will result in different LOH sizes, ranging from a single locus to several kilobases. Indeed, mitotic crossover (MCO) or BIR between homologs can cause long-tract LOH, which can span over hundreds of thousands of bases, stretching from the breakpoint to the telomere. BIR comes into play to repair the break when only one end of the DSB shares homology with a template and is characterized as highly inaccurate, due to frequent template switching and slippage of polymerase δ resulting in 1,000-fold increase in indels (Smith et al. 2007; Deem et al. 2011). The mutagenic property of BIR has also been illustrated in C. albicans, as a higher frequency of de novo mutations has been reported in large LOH tracts by Ene and collaborators (Ene et al. 2018). In addition, BIR in S. cerevisiae requires extensive 5’-end resection leading to the exposure of long ssDNA which is more sensitive to DNA-damaging agents and may lead to the formation of point mutation clusters (Sakofsky et al. 2014, 2019; Elango et al. 2019).

In C. albicans, only a limited number of studies focused on the effects of DNA repair pathway disruption on overall genomic stability, mainly through the characterization of deletion mutants of genes involved in excision repair (Base Excision Repair—BER, Nucleotide Excision Repair—NER), and DNA Mismatch Repair (MMR) or DSB repair. More recently, two studies investigated the outcomes of I-SceI- and Cas9 nuclease-mediated DNA-DSBs in C. albicans, by characterizing the repair mechanisms at stake (Feri et al. 2016; Vyas et al. 2018). These studies demonstrated that DNA-DSBs are predominantly associated with HR repair in C. albicans, as illustrated by the frequent occurrence of LOH upon DNA-DSB repair (Feri et al. 2016; Vyas et al. 2018). These LOH, that are frequently observed in the C. albicans population, are counter-intuitive in regards to the several studies that have highlighted the importance of maintaining heterozygosity as it has been shown to correlate with strain fitness (Hickman et al. 2013; Hirakawa et al. 2015). There appears to be an equilibrium between genomic stability, which ensures longevity of the lineage and maintenance of genomic integrity, and genomic instability which promotes the acquisition of new allele combinations potentially favoring adaptation. This equilibrium may be influenced by extrinsic factors since the nature of molecular mechanisms leading to LOH in C. albicans varies based on growth conditions, e.g., exposure to H₂O₂ predominantly results in short-tract LOH while fluconazole and growth at 39°C favor the appearance of long-tract LOH (Forche et al. 2011). In the context of a human pathogen, environmental stimuli that would trigger genome instability could activate specific DNA repair pathways, consequently dictating the LOH size and therefore the translation of these homozygosis tracts into phenotypic diversity. Fidelity of repair mechanisms can also play an important role in the latter described equilibrium though, this has not been extensively investigated in C. albicans. The reliability of DNA damage repair remains to be thoroughly investigated as it plays an important role in diploid heterozygous strains where maximum heterozygosity should ideally be maintained as it has been shown to be beneficial in the long-term.

Our previously developed I-SceI inducible DNA-DSB system coupled to a fluorescence-based LOH reporter system, permits the detection, quantification, and sorting of C. albicans cells that had undergone a long-tract LOH, as a result of HR-mediated DNA-DSB repair (Feri et al. 2016). By using this setup, we sought out to investigate the fidelity of BIR/MCO in C. albicans, as they were proposed to be the main molecular mechanism giving rise to the long-tract LOH often observed in laboratory and clinical isolates (Hirakawa et al. 2015; Ropars et al. 2018; Wang et al. 2018). Analysis of whole-genome sequencing data of several isolates which have undergone I-SceI-dependent BIR/MCO revealed that long-tract LOH are quite often associated with unexpected events of homozygosis from the I-SceI break site toward the centromere. In order to explore the factors that could influence these unexpected events, we investigated the impact of deleting four genes whose S. cerevisiae orthologs are known to be important for completion of the HR process. POL32 encodes a nonessential subunit of the DNA polymerase δ complex. Although not needed for normal DNA replication, Pol32 is required for BIR (Lydeard et al. 2007). During BIR, the 3’ end invades its template DNA and heteroduplexes can be formed between the annealed sequences in the displacement loop or D-loop. If the region contains heterozygous SNPs, components of the MMR machinery, such as Msh2 (Reenan and Kolodner 1992), could intervene to repair the mismatches at the level of the heteroduplexes, transforming a heterozygous region into a homozygous region. Finally, particular proteins will regulate the disassembly of recombination intermediates D-loops into noncrossover (NCO) or crossover (CO) outcomes of repair. Among them are Mph1, a helicase that allows the displacement or dissolution of the D-loop and favors NCO outcomes (Prakash et al. 2009), and Mus81, a resolvase important for resolution of Holliday junctions and for promoting CO events (Ho et al. 2010). Although extensively characterized in the budding yeast S. cerevisiae, the role of these genes in C. albicans overall genome integrity has not been addressed except for C. albicans msh2^-/- mutants that display an increase in repetitive-tract instability and in which drug resistance was shown to arise more rapidly (Legrand et al. 2007).

In this study, we describe for the first time, that long-tract homozygosis start sites do not perfectly coincide with DNA break sites in C. albicans and the extent of ascending homozygosis tracts. By characterizing pol32^-/-, msh2^-/-, mph1^-/-, or mus81^-/- deletion mutants, we propose that (i) extensive 5’-end resection occurs over several kilobases and (i) the MMR machinery potentially contributes to those ascending LOH tracts toward the centromere upon BIR/MCO-mediated DNA-DSB repair in C. albicans. Overall, we gain a better understanding of the factors that influence the extent of BIR/MCO-mediated genomic changes.

Material and Methods

Strains and culturing conditions

C. albicans strains described in the study are derived from the reference strains SC5314, parental strain CEC4088, and CEC4012(Leu^-) (Feri et al. 2016) possessing the BFP/GFP LOH reporter system (positions 471,021–481,176) and the I-SceI target-sequence (positions 776,698–778,104) on the left arm of Chr4. Yeast cells were cultured on/in rich YPD medium (1% yeast extract, 2% peptone, and 2% dextrose). Synthetic defined (SD) (0.67% yeast nitrogen base without amino acids, 2% dextrose) and Synthetic complete (SC) (0.67% yeast nitrogen base without amino acids, 2% dextrose, 0.08% drop-out mix with all the essential amino-acids) media were used for selection. A yeast carbon base medium with bovine serum albumin (YCB-BSA) and supplemented with amino-acids was used to recycle LEU2 auxotrophic marker during strain construction. Solid media were obtained by adding 2% agar.

All C. albicans strains are listed in Supplementary Table S3.

Induction of Tet-On system

The I-SceI gene is found under the control of the Tet-On promoter. Activation of I-SceI protein production and induction of DNA-DSB at targeted site, is conducted by inoculating YPD + anhydrotetracycline (ATc) (3 µg/mL) (Thermofisher ACROS Organics™) with a preculture (liquid SC-His-Arg medium overnight at 30°C). Induction is performed for 8 hours at 30 °C followed by an overnight recovery in YPD. The same induced and noninduced cultures were used for both the 5-FOA selection assay and the quantification of LOH frequency using flow cytometry.

Cell sorting

Large debris and filamentous cells that could obstruct the tubing system of the cytometer were filtered using BD Falcon™ Cell strainers from induced and noninduced cultures. The MoFlo^® Astrios™ flow cytometer was used to analyze and sort the cells of interest. For each sorted gate, 1,000 cells were recovered in 400 μL of liquid YPD medium, plated immediately after cell sorting on four YPD Petri plates and incubated at 30°C for 48 h before collection of results.

Characterization of molecular mechanism used leading to LOH

Single colonies recovered from either cell sorting or 5-FOA plates were isolated and culture in YPD at 30°C overnight. Characterization of the molecular mechanism leading to LOH was conducted as described below for every single colony isolated. Functionality of auxotrophic markers was evaluated by drop tests on SC medium with appropriate drop-out amino acid, depending on tested marker. Overnight saturated cultures in YPD of selected strains were spotted on YPD (control), SC-His (test presence of BFP-HIS1 cassette), SC-Arg (test presence of GFP-ARG4 cassette), and SC-Ura (test presence of URA3-TS cassette) and placed at 30°C for 24 h to observe presence and absence of growth. The fluorescence status of everyone was also assessed using flow cytometry [MACSQuant^® Analyzer (Miltenyi Biotec)], where 10,000 cells were analyzed per culture and identified as mono-BFP, mono-GFP, or biofluorescent (BFP and GFP). An aliquot of each culture was used to extract DNA permitting genotyping of Chr4 and the accurate identification of each LOH size. The gDNA was extracted using the EPICENTER MasterPure™ Yeast DNA Purification Kit followed by DNA elution in 100 μL of sterile water. Allele composition at two SNP positions, SNP156 (left arm of Chr4) and SNP95 (right arm of Chr4) was examined using the SNP-RFLP technique as described in Forche et al. (Forche et al. 2009), with restriction enzyme TaqI and AluI, respectively. Theoretically, upon I-SceI mediated DNA-DSB the URA3 marker associated with the I-SceI target sequence is lost. Thus, we screen for presence of absence of URA3 by conducting PCR reactions using primers from Supplementary Table S4. By taking in account all results obtained throughout the phenotypic and genotypic profiling, we were able to confirm the presence of LOH and identify the size of each LOH tract, consequently the molecular mechanism resulting in LOH.

KO strains

Construction

To investigate which molecular mechanism(s) is (are) responsible for extending the homozygosis from the DNA-DSB locus toward the centromere, we investigated the effects of knocking-out 4 different genes (POL32, MSH2, MPH1, and MUS81), key players in different DNA repair pathways. These KO strains were constructed in SC5314 derived, CEC4088 (Leu^-) and CEC4012 (Leu^-) possessing the I-SceI TS on HapA and HapB, respectively. With the transient CRISPR-Cas9 system (Min et al. 2016), homozygous deletion mutants were obtained in one round of transformation using a recyclable leucine flipper cassettes. We constructed a FLP-LEU2 plasmid by switching out the SAT1 gene in the pSFS1A, pFLP-SAT1 plasmid (Reuss et al. 2004), with the LEU2 gene. We amplified the LEU2 gene from the pFA-CmLEU2 plasmid (Schaub et al. 2006) using primers containing PvuII and NsiI restriction sites and ligated into the backbone of pSFS1, where removal of SAT1 was conducted by double digestion with Blp1 and PstI followed by Klenow treatment. By recycling of the LEU2 auxotrophic marker, each strain was then complemented with a full-length copy of GPI16, as Chr4L harbors a recessive lethal allele on HapB between the I-SceI TS and the BFP/GFP LOH reporter system, rescuing individuals which become homozygous for Chr4B (Feri et al. 2016).

We used a transient CRISPR-Cas9 system (Min et al. 2016), which does not necessitate the genomic integration of either Cas9 or sgRNAs. The construction of sgRNAs and amplification of Cas9 cassettes from the pV1093 plasmid were conducted as described in Min et al. Four unique 20 bp sgRNAs (Supplementary Table S4) were designed using CHOPCHOP (Montague et al. 2014), guiding Cas9 cleavage activity to each locus of each target genes (POL32, MSH2, MPH1, and MUS81). Repair templates are constructed by PCR amplification using 120 bp primers (Supplementary Table S4), each composed of 20 bp complementary to the FLP-LEU2 cassette and 100 bp tails possessing the complementary sequences of the KO gene locus. Each primer pair was used to amplify the FLP-LEU2 cassette from plasmid pFLP-LEU2. Each cassette was amplified in a total PCR volume of 500 μL, precipitated in 100% ethanol and re-suspended in 100 μL of distilled sterile water. For each transformation, competent cells were transformed with approximately 5 μg of appropriate DNA cassette. CEC4088 and CEC4012 cells were co-transformed with 5 μg of the appropriate FLP-LEU2 cassette, 1 μg of the Cas9 cassette and 1 μg of sgRNA using the Lithium Acetate/PEG transformation protocol. Transformants were then selected on SC-Leu medium and, junction PCRs and internal target gene PCRs were performed in order identify homozygous KO transformants and to ensure proper integration of FLP-LEU2 cassette at the targeted locus (using primers from Supplementary Table S4). Confirmed homozygous KO transformants were cultured overnight at 30°C in liquid YCB-BSA medium supplemented with 100 X amino acids (Arg10g/L His10g/L Uri 2 g/L Leu 6 g/L) and platted on SD-agar plates containing various concentrations of leucine (0.02 g/L, 0.012 g/L, 0.006 g/L, 0.0012 g/L, and 0.0006 g/L) in order to select for colonies which have flipped-out the LEU2 marker, resulting in LEU2 loss. The latter strains underwent a second transformation round to integrate StuI linearized CIp-P_TDH3-LEU2-GPI16 plasmid (Feri et al. 2016) at the RPS1 locus on Chr1, ensuring GPI16 complementation. Transformants were then selected on SD medium and junction PCRs were conducted to ensure proper integration at RPS1 locus, using primers in Supplementary Table S4. All 10 constructed strains can be found in Supplementary Table S3.

Characterization

The characterization of the 10 KO strains (Supplementary Table S3) consisted of morphology, doubling time, filamentation, and genotoxic stress sensitivity assays.

The colony morphology of all strains was also assessed on solid YPD medium, at 30°C for 3 days. Images were taken with a Leica M80 Stereomicroscope at a x25 magnification. Images were captured with a DMC 2900 camera, using the Leica Application Suite (LAS) imaging software. Cell phenotypes were observed using overnight culture in YPD at 30°C with an Olympus IX83 microscope, 20x objective. Filamentation phenotypes were observed after 3–4 hours in YPD, YPD + 10% serum and Spider medium broths and imaged with the Olympus IX83 microscope, 20x objective.

The doubling times were evaluated in YPD media at 30°C by measuring the optical density with TECAN Infinite over 24 hours.

Sensitivity of strains to various genotoxic stressors was evaluated using drop test assay. Cultures grown at 30°C overnight were spotted YPD, YPD 4 mM H₂O₂, YPD 0.1 mM menadione, YPD 2 mM tBHP, YPD 100 µM camptothecin, YPD 0.03% EMS and YPD 0.01% MMS. Images were taken after 2–3 days of incubation at 30°C with a PhenoBooth Colony Counter (Singer Instruments).

Quantification of LOH frequency using flow cytometry

All flow cytometry analyses were conducted on the MACSQuant^® Analyzer (Miltenyi Biotec). Data for a maximum of 10⁶ cells per sample were analyzed using the FlowJo V10.1 software. The gates to determine the LOH frequencies were arbitrarily selected but conserved throughout sample analysis.

5-Fluoroorotic acid (5-FOA) assay

Efficiency of I-SceI-cleavage upon induction was evaluated by plating three different cell dilutions of cultures (20,000, 2,000, and 200 cells) grown in presence (induced) or absence (noninduced, control) of ATc on 5-FOA-containing plates in triplicates. Dilutions were verified by plating a volume corresponding to 100 cells on YPD plates. Plates were incubated at 30°C for 3 days before analysis. CFUs were counted in order to calculate 5-FOA^R frequency in both induced and noninduced conditions. In addition, a subset of CFUs were selected and characterized by fluorescence, auxotrophy and SNP-RFLP typing.

DNA extraction and whole-genome sequencing

Wildtype heterozygous (I-SceI TargetA or TargetB) strains along with 4–5 isolates having undergone I-SceI-dependent BIR/MCO, based on flow cytometry and SNP-typing, for each deletion background (WT, pol32^-/-, msh2^-/-, mph1^-/-, and mus81^-/-) were cultured in 5 mL of liquid SD medium overnight at 30°C and DNA was extracted following the manufacturer’s protocol using the QIAGEN QIAamp DNA Mini Kit. The DNA was eluted in a total volume of 100 µL. The genomes were sent for whole-genome sequencing at Novogene, Illumina sequencing technology. Libraries were constructed using the yeast resequencing kit: NEB Next^® Ultra™ DNA Library Prep Kit. The NovoSeq6000 platforms were used to generate 150 bp paired-ends reads. Refer to “Data and reagent availability” for information regarding data accessibility.

Analysis of DNA-DSB site

Sequences and genomic variations were analyzed as previously described in Ropars et al. (2018) and Sitterle et al. (2019). Each set of paired-end reads was mapped against the C. albicans reference genome, SC5314 haplotype A (version A22-s07-m01-r57), using Minimap2 version (Li 2018). SAMtools version 1.9 and Picard tools version 2.8.1 (http://broadinstitute.github.io/picard) were then used to filter, sort, and convert SAM files. SNPs were called using Genome Analysis Toolkit version 3.6, according to the GATK Best Practices. SNPs were filtered using the following parameters: VariantFiltration, QD < 2.0, LowQD, ReadPosRankSum<-8.0, LowRankSum, FS > 60.0, HightFS, MQRankSum < −12.5, MQRankSum, MQ < 40.0, LowMQ, HaplotypeScore > 13.0. Coverages were also calculated using the Genome Analysis Toolkit. The GATK variant filtration walker (VariantAnnotator) was used to add allele balance information to VCF files. The value of allele balance at heterozygous sites (ABHet) is a number that varies between 0 and 1. ABHet is calculated as the number of references reads from individuals with heterozygous genotypes divided by the total number of reads from such individuals.

In order to princely analyze I-SceI break sites, heterozygous positions between 650,000 and 1,050,000 for Chr4 were extracted from the GATK SNPs calling files (filtered with best practice VCF, mentioned above) from sequencing data of heterozygous WT I-SceI TargetA and I-SceI TargetB strains. This generated two heterozygous positions reference lists, subsequently used for comparison with each strain. For each sequenced isolate we also extracted all information from the GATK SNPs calling files (VCF) for genomic region between positions 650,000 and 1,050,000 on Chr4 and using a Python custom script (available upon request) we investigated the status of each position of interest within each sequenced isolate. Because we aligned read data onto haplotype A of SC5314 our interpretation for each strain was: for a given position, if the VCF file indicated (i) a defined heterozygous SNP we assigned the position as heterozygous, (ii) a homozygous SNP, we considered that positions to be homozygous for haplotype B, or (iii) if no information for this position was found in the VCF file it was assigned as a homozygous haplotype A SNP. The final tables which summarize the comparison results between (appropriate) WT heterozygous strain and each sequenced isolate were then used to precisely analyze the region of interest and evaluate the ascending LOH size by identifying the start of homozygosis tracts.

Results

Unexpected long-tract homozygosis centromere-proximal to the I-SceI-induced double-strand break

We previously described the use of a LOH reporter system coupled with a DNA-DSB inducing system taking advantage of the S. cerevisiae I-SceI meganuclease to investigate the mechanisms involved in DNA-DSB repair in C. albicans (Loll-Krippleber et al. 2015; Feri et al. 2016). The strains generated by Feri et al. possess (i) the BFP/GFP LOH reporter system (position 471,021–481,176) and (ii) the I-SceI target sequence (position 776,698–778,104) (“I-SceI TargetA” on Hap A or “I-SceI TargetB” on Hap B) on the left arm of Chr4 (Chr4L) and (iii) a full-length copy of GPI16 at the RPS1 locus in order to compensate for the recessive lethal allele on Chr4B in the SC5314 background, as well as (iv) the I-SceI gene under the control of the inducible promoter P_TET (Feri et al. 2016) (Supplementary Figure S1). These strains allow the induction of a DNA-DSB on Chr4L, the recovery of cells having undergone long-tract LOH upon DNA-DSB repair and the precise study of the resulting homozygosis tracts. Here, we aimed to assess the fidelity of homology-directed repair mechanisms resulting in long-tract LOH events in C. albicans. We selected five clones derived from both parental strains (i.e., carrying the I-SceI target sequence either on Chr4A or Chr4B) and having undergone long-tract LOH based on marker gene loss and SNP-RFLP analysis (Supplementary Figure S2). These strains were whole-genome sequenced. Allele balance at heterozygous positions (ABHet) and sequencing depth confirmed that all ten sequenced isolates had undergone BIR/MCO (Figure 1A, Supplementary Table S1).

Figure 1.

Analysis of the DNA-DSB site in cells having undergone BIR/MCO in wild type C. albicans genetic background display homozygosis from the DNA break site toward the centromere. (A) Plots showing the allele balance at heterozygous positions across the eight chromosomes. Sequencing confirms the presence of homozygosis, a long tract LOH, on the left arm of Chr4 (either haplotype A or B according to the strain) in mono-fluorescent isolates having undergone BIR/MCO. Additional LOH events have previously been described in parental strains of “I-SceI TargetA” and “I-SceI TargetB” [LOH on Chr2 (Loll-Krippleber et al. 2015)] and in SC5314 [LOH on Chr3 and Chr7 (Abbey et al. 2011)]. The orange vertical line on Chr4 represents the I-SceI-break site. (B) Each diagram represents an individual sequenced isolate having undergone BIR/MCO, as determined by flow cytometry and SNP typing, following I-SceI induction targeting DNA-DSB on haplotype A (I-SceI TargetA) or haplotype B (I-SceI TargetB). Representation of a 175 kb region surrounding the I-SceI-break site (orange) was built by monitoring the homozygous/heterozygous status using 1 kb bin sizes. Several unexpected homozygosis regions located between the I-SceI-break site and the centromere are identifiable within these diagrams, length (bp) indicated on the right side of each diagram. Identified in red writing are isolates having undergone spontaneous LOH on Chr4L and those with an asterisk (*) display interspersed regions of heterozygosis within long homozygosis tracts. The location of the repeat sequence RB2-4a is identified in pink.

In order to assess the fidelity of BIR/MCO in C. albicans, we conducted a detailed characterization of the Chr4 genomic region between positions 650,000 and 1,050,000, surrounding the I-SceI break site. All heterozygous SNPs found in both parental strains were identified within this region of interest. Allele status at these positions was investigated for each of the ten C. albicans clones having undergone BIR/MCO, to precisely identify the start site of the homozygosis tract. As illustrated in Figure 1B, our data revealed a great variability in terms of anchoring position for the selected LOH, ranging from 898 to 118,465 bp upstream the I-SceI target site in 7 out of 10 clones. Clone 5 displayed homozygosis with interspersed heterozygous regions which extended almost exactly from the I-SceI site toward the telomere, likely to represent template switching (Figure 1B). For 2 out of 10 clones (6 and 7), the homozygosis tract detected at the BFP/GFP locus did not go up all the way to the I-SceI site, despite the actual cut by I-SceI as illustrated by the loss of the I-SceI target site-associated auxotrophic marker URA3. In these two clones, the DNA-DSB had likely been repaired by gene conversion (GC) and had been accompanied by an I-SceI-independent BIR/MCO event distal to the I-SceI site starting at positions 628,652 and 523,190, respectively. It is worth noting that all isolates displayed identical patterns of transitions between HapA and HapB homozygosis (as illustrated by the blue or green lines at the very left end of the 175 kb window). These are likely to be due to errors in haplotyping of the reference strain used to align reads and are not representative of systematic template switching.

Phenotypic characterization of deletion mutants of genes involved in various aspects of DNA repair and recombination

Although these extended regions of homozygosis toward the centromere may result from I-SceI-independent BIR, spontaneous BIR events are still rare and are unlikely to have occurred here with such a high frequency, suggesting that extended homozygosis regions are an intrinsic feature of the molecular mechanisms involved in the repair of I-SceI-induced DNA-DSB in C. albicans. In order to identify the molecular mechanism(s) at stake, we sought out to evaluate the impact of deleting genes known to be involved in various aspects of DNA repair and recombination on the homozygosis events occurring at the I-SceI site upon HR-mediated DNA repair through BIR/MCO. Homozygous deletion mutants were constructed in both strains carrying the I-SceI target sequence on Chr4A or Chr4B for the following genes: (i) the POL32 gene, encoding a component of the Pol δ complex involved in DNA synthesis during BIR, (ii) the MSH2 gene, involved in identification of MMR substrates, (iii) the MPH1 gene, involved in D-loop displacement and favoring NCO outcomes of HR-repair or (iv) the MUS81 gene encoding a resolvase involved in CO events. Because of a heterozygous recessive lethal mutation in the GPI16 gene on Chr4B, these mutant strains were also complemented with WT GPI16 allele to ensure viability of Chr4B homozygous cells (Feri et al. 2016). Deletion mutants were obtained for all genes using the CRISPR-Cas9 approach and confirmed by PCR.

Growth rates of the deletion mutants were evaluated in standard laboratory conditions (YPD 30 °C). The mus81^-/- mutants displayed a drastic growth defect in 24 hours growth kinetic experiments (Supplementary Figure S4). The other mutant strains, pol32^-/-, msh2^-/-, and mph1^-/- displayed doubling times similar to WT strains though pol32^-/- mutant strains appeared to have a slight growth defect (Supplementary Figure S4B). Upon observation of colonies on agar media, we noticed that mus81^-/- mutants formed much smaller colonies with wrinkle edges while the other mutant strains formed smooth round colonies comparable to WT strains (Supplementary Figure S4A). In liquid cultures (YPD 30°C), similarly to WT, msh2^-/- and mph1^-/- mutant strains grow in yeast form. However, pseudohyphae were observable in cultures of pol32^-/- mutants and even more frequent with mus81^-/- mutants, reminiscent of genotoxic stress-induced filamentation (Figure 2A). In contrast, no obvious defects in filamentation were observed in any mutant strains in liquid filamentation-inducing conditions (YPD + 10% serum or SPIDER media at 37°C for 3 h), as compared to WT strains (Figure S5).

Figure 2.

Characterization of DNA repair KO mutants. (A) Phenotypic characterization of the cell morphologies of wildtype and KO mutant strains. YPD liquid culture were imaged after an overnight incubation at 30°C. (B) Characterization of genotoxic sensitivity of wild-type and KO mutant strains. Spot assay conducted on YPD medium containing either; 4 mM H₂O₂, 2 mM tBHP, 0.01% MMS, 0.03% EMS or 100 µM camptothecin. Images were taken after 2 days at 30°C. (C) Characterization of spontaneous genomic instability of KO mutants as compared to WT strains. Histogram representing the fold changes (±SD) of mono-fluorescent cells (mono-BFP or mono-GFP) between WT and each KO strains (n = 4). Statistical differences in overall mono-fluorescent cells (sum of mono-BFP and mono-GFP) were assessed using nonparametric Kruskal-Wallis (P > 0.0001) and post hoc Dunn’s multiple comparison tests (**P < 0.01 and ***P < 0.001). Example of FACS profile in Figure S3.

We also conducted drop dilution assays to evaluate the sensitivity of the deletion mutants to various DNA-damaging agents, namely oxidative agents (4 mM H₂O₂ and 2 mM tBHP), DNA alkylating agents (0.01% MMS and 0.03% EMS), and 100 µM camptothecin. Growth, as compared to WT, was observed after 48 hours at 30 °C. While the pol32^-/- and mph1^-/- mutants are sensitive only to 0.01% MMS, the mus81^-/- deletion mutants demonstrate sensitivity upon exposure to all five DNA-damaging agents. (Figure 2B). While growth of the two msh2^-/- strains did not seem to be affected by H₂O₂, tBHP, EMS and camptothecin, the two strains exhibited contrasting sensitivity profiles on 0.01% MMS, with the strain carrying the I-SceI target sequence on HapA being more sensitive to MMS than the strain carrying the I-SceI target sequence on HapB (Figure 2B). Of interest, Legrand and collaborators have reported that independent msh2^-/- null mutants behave similarly to our msh2^-/- mutant carrying the I-SceI target sequence on the Chr4B homolog, displaying no increased sensitivity to MMS (Legrand et al. 2007). Discrepancies between msh2^-/- deletion mutants may be associated with the accumulation of point mutations, as disruption of the MMR pathway is associated with a mutator phenotype (Drotschmann et al. 1999; Healey et al. 2016; Boyce et al. 2017). Although the pol32^-/-, msh2^-/-, and mph1^-/- mutants did not display a growth defect in the presence of tBHP, we noticed that these mutants displayed smaller colonies as compared to the WT strain (Figure 2B).

Because the genes targeted for deletion are key players of DNA metabolism, we also assessed genome stability in the mutant strains. We took advantage of the BFP/GFP LOH reporter system, localized on Chr4L, to evaluate genome stability in the deletion mutants under standard laboratory conditions (YPD, 30°C). While deletion of POL32 resulted in only a very small augmentation in LOH frequency at the BFP/GFP locus as compared to WT, the highest increases were observed for mph1^-/- and mus81^-/- mutants with a fold change of 8x and 58x, respectively, as compared to WT (Figure 2C, Supplementary Figure S3). Significant differences in spontaneous LOH frequency were observed between strains (Kruskal-Wallis test, P < 0.0001) (Figure 2C). It is worth noting that we saw the same response in terms of LOH frequency in the pol32^-/-, mph1^-/-, and mus81^-/- mutants, regardless of the haplotype targeted by I-SceI. In contrast, the extent of LOH frequency increase varied between the two msh2^-/- mutants depending on the haplotype targeted by I-SceI, with a fold change of 2x and 7x, respectively for HapB and HapA (Figure 2C, Supplementary Figure S3). Subtle phenotypic differences between the two msh2^-/- mutants will be further discussed below.

In addition, we assessed the functionality of the inducible DNA-DSB system in the deletion mutant strains. Because the I-SceI target sequence is associated with the URA3 auxotrophic marker, the occurrence of a DNA-DSB at the I-SceI target site, followed by efficient repair, can be monitored by counter-selection on 5-FOA-containing medium. The 5-FOA^R colonies arise from cells that have likely undergone and successfully repaired an I-SceI-mediated DNA-DSB. Therefore, fold changes between the frequencies of 5-FOA^R colonies in noninduced vs. induced conditions can be used as indicators of cleavage efficacy. The inducible I-SceI cleavage system appeared to be functional in all four deletion mutants as illustrated by average fold changes ranging from 25.5 to 2,047 (Figure 3A). Noticeably, the pol32^-/- and mus81^-/- mutants displayed fold changes that were in the lower range, below 100, and that could be explained by an increased mortality rate of the cells undergoing a DNA-DSB in these genetic backgrounds.

Figure 3.

Augmentations of LOH events in WT and KO strains upon I-SceI induction. (A) Average fold changes in the appearance of 5-FOA^R colonies (±SD) between I-SceI induced and noninduced conditions (n = 3) are illustrated in the histogram. (B) Average fold changes of mono-fluorescent cells (±SD) between I-SceI induced and noninduced conditions (n = 4) are represented in the histogram. Example of FACS profile in Supplementary Figure S3.

Functionality of the inducible I-SceI cleavage system was also assessed by flow cytometry (Supplementary Figure S3), focusing this time on DNA-DSB repair resulting in long-tract LOH. In Figure 3B, we compared the fold changes in terms of increase in mono-fluorescent cells in all strains, and observed an augmentation in mono-fluorescent cells upon I-SceI induction in all strains, ranging from 1.5- to 75.2-fold. We observe that deletion of the gene of interest reduces the fold changes of mono-fluorescent cells in comparison to the WT. Appearance of long-tract LOH upon I-SceI induction was decreased in pol32^-/-, mus81^-/-, mph1^-/,^- and msh2^-/- strains, in decreasing order of severity as compared to WT (Figure 3B).

To test for a potential shift in the molecular mechanisms used to mend the induced DNA-DSB, we conducted a characterization of repair mechanisms in induced conditions, where DNA-DSB are predominantly I-SceI-dependent. We took advantage of the auxotrophic markers associated with each component of the BFP/GFP LOH reporter system and with the I-SceI target sequence, combined with SNP-RFLP to evaluate the size of LOH events, as summarized in Supplementary Figure S2. The size of I-SceI-dependent LOH events in each mutant was assessed by analyzing a total of 58 to 62 5-FOA^R colonies (2 strains × 26 to 32 colonies each), using cytometry and spotting on supplemented media to check fluorescence status and auxotrophy status, respectively. As illustrated in Figure 4A, the majority of colonies (≥80%), which underwent I-SceI-dependent DNA-DSB, are characterized by short-tract LOH, likely resulting from GC. Thus, similarly to WT strain, I-SceI-dependent DNA-DSB in the deletion mutants are mainly repaired through GC, a molecular mechanism which limits the LOH to the DNA-DSB locus. We then focused on characterizing the molecular mechanisms associated with I-SceI-dependent long-tract LOH events. To do so, we defined two populations of mono-fluorescent cells for characterization after I-SceI-dependent DNA cleavage: the trans population which includes cells that have lost the fluorescence marker carried by the Chr4 homolog targeted by I-SceI and the cis population which includes cells which have become mono-fluorescent but retain the fluorescence marker carried by the Chr4 homolog targeted by I-SceI; each population being representative of distinct molecular mechanisms at stake in I-SceI-dependent long-tract LOH. To specifically assess the impact of each gene of interest on the molecular mechanisms leading to I-SceI-dependent long-tract LOH, we performed fluorescence-activated cell sorting on both trans and cis mono-fluorescent populations of WT and deletion mutants. We recovered 1,000 colonies per population using cell sorting, of which 70–80 colonies (35–40 colonies/mono-fluorescent population, for two strains) were selected for further characterization by flow cytometry and SNP-typing (Supplementary Figure S2). We found that LOH events giving rise to the cis population are all the result of GC with CO in both the WT and mutant strains (Figure 4B). For the trans populations, we observed three profiles: (i) the WT strains display 80% of the long-tract LOH resulting from BIR/MCO/CT and 20% from chromosome loss, (ii) the mph1^-/- and msh2^-/- mutants displayed a 2.5- to 5-fold decrease respectively in the population proportion of long-tract LOH resulting from chromosome loss and (iii) the pol32^-/- and mus81^-/- mutants displayed a 2- to 3.5-fold increase in chromosome loss (Figure 4C). Overall, the four candidate genes altered the distribution of the investigated molecular mechanisms upon I-SceI-mediated DNA-DSB. It is also worth noting that despite the subtle phenotypic differences observed between the two msh2^-/- mutants mentioned above, the pair (I-SceI TargetA and TargetB) displayed similar proportions of repair mechanisms identified. Based on our observations, further analysis was conducted by genetic background, with no distinction between the two mutants of a pair.

Figure 4.

Analysis of molecular mechanisms used in I-SceI induced conditions. (A) Percentage of molecular mechanisms leading to LOH among 5-FOA^R colonies in I-SceI induced condition (n = 58–62). Long-tract LOH span from the I-SceI target site until the BFP/GFP LOH reporter system while short-tract LOH is restricted to the I-SceI break site. (B and C) Characterization of mono-fluorescent-sorted colonies within cis (B) and trans (C) populations in I-SceI induced condition (n = 70–80). A subset of sorted cells was characterized by flow cytometry, auxotrophy spotting, and SNP-typing to profile the homozygosis tract on the left arm of Chr4 and determine the molecular mechanism used for I-SceI-induced DNA-DSB. Abbreviations of molecular mechanisms are as follows; gene conversion (GC), break-induced replication (BIR), MCO, gene conversion with crossover (GC with CO). Colonies were regrouped in four categories: BIR/MCO/CT (I-SceI-dependent LOH spanning from the I-SceI break site until the Chr4L telomere), CL (I-SceI-dependent LOH spanning across entire Chr4), GC w CO (I-SceI-dependent LOH spanning from the I-SceI break site until the Chr4L telomere) and I-SceI indep. (I-SceI-independent event leading to mono-fluorescence). For detailed analysis of molecular mechanisms please refer to Supplementary Figure S2.

DNA repair mechanisms involved in events of homozygosis from the I-SceI break site toward the centromere

As mentioned above, we reported the occurrence of long-tract homozygosis centromere-proximal to the I-SceI-induced double-strand break in the WT. We further investigated whether the absence of the candidate genes altered this phenomenon. To do so, we sequenced the genomes of 9–10 isolates having undergone I-SceI-dependent BIR/MCO, for each deletion background and the WT (5 isolates of each haplotype). Whole-genome sequencing data allowed us to confirm the event of BIR/MCO on Chr4L using ABHet plots (Supplementary Figure S6). We first took advantage of the data to investigate the occurrence of GCR events occurring within the 49 C. albicans sequenced strains. For the purpose of this analysis, we decided to exclude the long-tract LOH events on Chr4L (I-SceI-induced) and GCRs which are shared by all sequenced isolates for a given genetic background (i.e., LOH on Chr5 in I-SceI TargetA msh2^-/- and mph1^-/- strains), as the latter most likely arose during the strain engineering process (Supplementary Table S2). The heatmap in Figure 5 shows the distribution of GCRs (LOH and aneuploidy events) amongst the 8 chromosomes and the different genetic backgrounds. Overall, the pol32^-/-- and mus81^-/--derived isolates possessed the most genomic rearrangements with 5 and 16 GCRs within the ten pol32^-/- and the nine mus81^-/- sequenced strains, respectively (Figure 5). In contrast, isolates with WT and msh2^-/- genetic backgrounds displayed no additional GCRs other than those induced by I-SceI on Chr4L or pre-existing ones (Figure 5). Interestingly, from a chromosome perspective, Chr5 seems to be the most impacted by GCRs in this dataset, exhibiting LOH and aneuploidy (3n or 4n) events in equal proportion (Supplementary Table S2). Also, in the mus81^-/- genetic background, Chr4 appears to be highly impacted by aneuploidies (3n or 4n) (Figure 5, Supplementary Table S2).

Figure 5.

Distribution of genome-wide GRS in wild-type and deletion strains. Heatmap of GCRs identified within the 49 C. albicans sequenced isolates, showing the density of GCRs per genetic background (n = 9–10 strains). Summary of GCRs identified within sequenced strains can be found in Supplementary Table S2. All the long-tract LOH events on Chr4L (I-SceI-induced) and GCRs which are shared by all sequenced isolates for a given genetic background (i.e., LOH on Chr5 in I-SceI TargetA msh2^-/- and mph1^-/- strains) were excluded for the heatmap analysis, as those most likely arose during the strain engineering process.

We then conducted an in-depth analysis of the 400 kb region surrounding the I-SceI-break site as described above for the WT strains. By investigating the status of 2,845 naturally heterozygous positions on Chr4L between positions 650,000 and 1,050,000, we precisely identified which positions have become homozygous upon I-SceI-cleavage in each sequenced isolate. As illustrated in Figure 6, 36% (14/39) of the mutant sequenced isolates having undergone an I-SceI-dependent BIR/MCO upon cleavage of Chr4 (HapA or HapB) displayed homozygosis which extended almost exactly from the I-SceI- break site toward the telomere (from position 778,104 to the telomere). While the remaining 64% (25/39) presented an extension of the homozygous region toward the centromere (Figure 6). Of the latter isolates, two (I-SceI TargetB mus81^-/- isolates 7 and 8) possessed homozygosis tracts that extend until position 1,044,280, beyond CEN4 (located at position 992,473–996,110), and another (I-SceI TargetA mus81^-/- isolate 2) displayed homozygosis until the RB2-4a repeat sequence (pink) (Figure 6). Indeed, homozygosis tracts of these 3 isolates probably resulted from spontaneous LOH events. The other 22 events of homozygosis ascending toward the centromere in mutant isolates ranged up to 201 kb away from the I-SceI-break site (Figure 6). In addition, within the analyzed region of interest, we observed an alternating pattern between homozygosis of the HapA (green) and HapB (blue) shared by all strains (Figure 6). The latter is most likely due to haplotyping errors found within the SC5314 reference genome and not associated with intrinsic features of BIR/MCO molecular mechanisms. However, 10/49 strains displayed homozygosis tracts with interspersed regions of heterozygosis (identified by an asterisk in Figure 6), suggesting events of template switching during BIR, independently of genetic background. Regions of heterozygosis within homozygosis tracts are also observed at repeat regions such at RB2-4a locus (highlighted in pink on Figure 6), this heterozygosis is most certainly artefactual and associated with the misassembly of repeat regions in the reference genome used to align sequencing reads. Repeat regions are often difficult to accurately assemble due to their repetitive nature. In addition, I-SceI + TargetB msh2^-/- isolate 6 displayed initially homozygosis of HapB (blue) followed by HapA (green) homozygosis until the telomere, which can potentially be explained by two independent events or the occurrence of a spontaneous GC with CO event (Figure 6).

Figure 6.

Results of sequencing data analysis at DNA-DSB site in cells having undergone BIR/MCO. Each diagram represents an individual sequenced isolate (n = 4–5) having undergone BIR/MCO, as determined by flow cytometry and SNP typing, following I-SceI induction targeting DNA-DSB on haplotype A (I-SceI TargetA) or haplotype B (I-SceI TargetB) in the corresponding genetic backgrounds. Representation of a 175 kb region surrounding the I-SceI-break site (shown in orange) was built by monitoring the homozygous/heterozygous status using 1 kb bin sizes. Numerous unexpected homozygosis regions located between the I-SceI-break site and the centromere are identifiable within these diagrams, with the length of the homozygosis centromere-proximal to the DNA-DSB indicated on the right side of each diagram in bp. Light blue writings identify isolates where homozygosis starts in the RB2-4a repeat sequences (shown in pink), while red writings identify isolates having undergone non-RB2-4a associated spontaneous LOH. Finally, an asterisk (*) show isolates that display interspersed regions of heterozygosis within long homozygosis tracts.

We then proceeded to quantify and compare the length of homozygosis tracts observed between the I-SceI-break site and the centromere between the different genetic backgrounds. We precisely identified the start of the long-tract LOH on Chr4L as the position of the most centromere proximal heterozygous SNP which became homozygous, followed by two other newly homozygous positions within a > 300 bp genomic region (Figure 7A). Thus, the difference between the position of the homozygosis tract start and the position of the I-SceI-break site on Chr4L gives the length of the region of homozygosis ascending toward the centromere (indicated in Figure 6), which can consequently be compared between genetic backgrounds. In order to accurately compare the effect of various deletion backgrounds on homozygosis tracts observed between the I-SceI-break site and the centromere, we chose to remove all I-SceI-independent LOH events on Chr4L (7 strains, indicated in blue and red writings in Figure 6). Among the remaining 42 isolates, the majority of isolates, 88% (37/42), possess ascending LOH tracts ranging from 0 to 17.4 kb while 12% of isolates (5/42) display >50 kb long ascending homozygosis toward the centromere (Figure 7B). Despite the low-sampling size, when comparing the median ascending LOH size across the five genetic backgrounds a reduction in median length is observed within the pol32^-/- (0 kb), msh2^-/- (0.4 kb), and mph1^-/- (0.7 kb) deletion strains as compared to WT (8.3 kb) (Figure 7B). Deletion of the MUS81 gene does not appear to have a negative impact on median length of homozygosis tracts between the I-SceI-break site and the centromere (Figure 7B). The same trends can be observed upon individual analysis of each targeted haplotype (I-SceI TargetA or I-SceI TargetB) (Figure 7C).

Figure 7.

Comparison of ascending LOH-tracts toward CEN4 across genetic backgrounds. (A) Illustration of how ascending LOH size was determined for each of the 59 sequenced C. albicans isolates of this study. Using sequencing data from the WT heterozygous strains, heterozygous positions between positions 650,000 and 1,050,000 of the Chr4 were extracted. For each sequenced isolate, having undergone I-SceI-dependent BIR/MCO, the status of each of the latter heterozygous positions (black asterisk) was analyzed in order to determine if it (i) remained heterozygous (black asterisk) or (ii) became homozygous (blue asterisk) for either HapA or HapB. The start site of the long-tract LOH, here defined by three consecutive homozygous SNPs (gray box), was determined by identifying the most centromere proximal position which became homozygous followed by two other consecutive positions which also became homozygous. In addition, these three homozygous positions (blue asterisks) must be spread across >300 bp, while ascending LOH size for each sequenced isolate having undergone I-SceI-dependent BIR/MCO was tabulated as the distance between the I-SceI-break site (orange) and the most centromere proximal homozygous position of the LOH break site. (B) Plotting of the length of homozygosis observed between the I-SceI-break site and the centromere (CEN4), also referred to as ascending LOH, in the context of I-SceI-dependent BIR/MCO events (blue: DNA-DSB on haplotype A, green: DNA-DSB on haplotype B) in the WT and mutant strains. The red line indicates the median ascending LOH size for each genetic background. (C) Plot of median ascending LOH sizes for each genetic background and both I-SceI-targeted haplotypes (haplotype A—TargetA or haplotype B—TargetB).

Discussion

By taking advantage of the previously developed inducible I-SceI-mediated DNA-DSB system coupled to a BFP/GFP LOH reporter system (Feri et al. 2016), we assessed the fidelity of homology-directed repair mechanisms in C. albicans leading to long-tract LOH events, focusing on the BIR/MCO mechanisms. To do so, we performed whole-genome sequencing to precisely analyze the region surrounding the I-SceI-break site on Chr4L in cells which displayed I-SceI-dependent BIR/MCO profiles, as determined by flow cytometry and SNP typing (Supplementary Figure S2). This analysis revealed that BIR/MCO is associated with homozygosis tracts that often extend beyond the I-SceI target site toward the CEN4 centromere and extend as expected on the other side toward the left arm telomere of Chr4. We then demonstrated that deletion of POL32, MPH1, or MSH2 genes, reduced the median ascending LOH size towards the centromere, suggesting that the overflow of homozygosis toward the centromere could constitutes intrinsic features of the BIR/MCO mechanisms and also involves the MMR machinery in C. albicans upon directed I-SceI-mediated DNA-DSB.

In the human pathogenic yeast C. albicans, LOH are ubiquitously observed throughout the genomes of both laboratory and clinical isolates (Abbey et al. 2011; Ropars et al. 2018; Wang et al. 2018) and are frequently detected upon exposure to various stress conditions (Forche et al. 2011; Ciudad et al. 2016). Studying the extent of LOH occurrence can be seen as highly relevant in C. albicans, a diploid organism whose sexual reproduction potential relies on homozygosis of the mating-type locus (MTL on Chr5) in a population that is predominantly heterozygous MTLa/α. An increased tendency of Chr5 to undergo LOH could allow a cell subpopulation to become mating competent and regain heterozygosity in the population. A major source of LOH are DNA-DSBs repaired by HR-mediated molecular mechanisms leading to various LOH sizes (Malkova et al. 2000). Most of our knowledge of these mechanisms is based on fundamental studies conducted in S. cerevisiae. Although LOH is an important aspect of C. albicans biology, as illustrated by their association with the fixation of beneficial alleles leading to adaptive phenotypes (Coste et al. 2007; Tso et al. 2018; Liang and Bennett 2019), the precise characterization of genes and pathways involved in DNA repair in C. albicans remains superficial and “phenotype”-oriented. Indeed, the deciphering of underlying molecular mechanisms is crucial to further understand the biological relevance of LOH in C. albicans.

In order to investigate the fidelity of homology-directed repair, notably BIR/MCO, we characterized the effect of deletion mutation for four genes (POL32, MSH2, MPH1, and MUS81) on DNA-DSB repair, of which three (POL32, MPH1, and MUS81) had not been previously investigated in C. albicans. Our characterization of mutant strains includes (i) a general phenotypic characterization of the mutants (in terms of growth rate, cell/colony morphology, filamentation, and sensitivity to genotoxic stresses), (ii) the evaluation of spontaneous genomic instability, and (iii) the characterization of molecular mechanisms at stake upon I-SceI-mediated DNA-DSB, in order to conclude on any striking genetic rewiring between S. cerevisiae and C. albicans.

The POL32 gene is a nonessential accessory component of the polymerase δ protein complex required for BIR initiation (Huang et al. 2002; Makarova and Burgers 2015). Phenotypic characterization of homozygous deletion of POL32 in C. albicans revealed a cell morphology phenotype, reminiscent of genotoxic-stress induced filamentation (Shi et al. 2007), an increased sensitivity to MMS and only a slight increase in LOH frequency (Figure 2, B and C). Similarly, the pol32^-/- deletion mutants exhibited a reduced frequency of 5-FOA^R colonies and of cells displaying long-tract LOH (assessed by flow-cytometry) upon I-SceI-dependent DNA-DSB (Figures 3A and 4C), suggesting that the disruption of this accessory component of the Pol δ complex disrupts the efficient repair of DNA-DSB in C. albicans. Although POL32 has been shown to be dispensable for GC in S. cerevisiae (Lydeard et al. 2007), it might not be the case in C. albicans. Nevertheless, short-tract LOH is predominant upon I-SceI-dependent DNA-DSB in all strains including pol32^-/- strains (Figure 4A). Also, the characterization of long-tract LOH revealed an increased abundance of chromosome loss (Figure 4C), suggesting that POL32 is more specifically implicated in BIR in C. albicans, similar to observations made in S. cerevisiae (Lydeard et al. 2007). Interestingly, in the study by Lydeard et al., the authors reported POL32 as being essential for RAD51-dependent BIR (Lydeard et al. 2007). Indeed, Malkova et al. had previously reported RAD51-dependent and RAD51-independent BIR processes in S. cerevisiae (Malkova et al. 2001, 2005). The two processes differ in terms of site of initiation with the RAD51-independent BIR that initiates at a site that can be far from the DSB site (30 kb closer to the centromere), predominantly making use of a facilitator of BIR (FBI) sequence located in the vicinity of the origin site (Malkova et al. 2001), while the RAD51-dependent BIR does not depend on a FBI sequence to initiate repair and therefore begins within a few kb of the DSB (Malkova et al. 2005). Hence, one would expect the RAD51-independent BIR process to be the only mechanism functional in pol32^-/- deletion mutants, with long-tract LOH events that should rely on FBIs and initiate far from the I-SceI-induced DSB. However, we observed the opposite with an enrichment for long-tract LOH that initiate within a few kb of the I-SceI-induced DSB. These observations suggest differences between S. cerevisiae and C. albicans in terms of RAD51/POL32 interactions or properties of the RAD51-independent BIR.

In S. cerevisiae, the MUS81 gene is best known for its nuclease activities important for resolution of replication and recombination intermediates, notably resolution of Holliday junctions and for promoting CO events (Ho et al. 2010). In C. albicans, the mus81^-/- strains displayed the most drastic phenotypes, as illustrated by their reduced growth, their shriveled colony morphology that correlates with the presence of numerous pseudohyphal cells and a high sensitivity to several genotoxic compounds (Supplementary Figure S4, Figure 2, A and B). As mentioned above, the formation of (pseudo)hyphal cells has been reported in C. albicans as a consequence of cell cycle disruption upon genotoxic stress (Shi et al. 2007). More recently, S. cerevisiae MUS81 was shown to participate in cell cycle regulation (G2 to M transition) (Pfander and Matos 2017). Our observations suggest that, similarly to S. cerevisiae, MUS81 could play a role in cell cycle progression in C. albicans. In addition, whole-genome sequencing of C. albicans mus81^-/- isolates highlighted numerous GCRs, notably aneuploidy events (Figure 5, Supplementary Table S2). Increased rates of aneuploidy have also been observed in mouse and human mus81^-/- cells (McPherson et al. 2004; Pamidi et al. 2007). MUS81 is likely to be crucial to remove recombination and replication intermediates that link sister chromatids. In cells lacking MUS81, these intermediates would perdure, preventing proper chromosome disjunction and resulting in aneuploidy. The importance of MUS81 in HR-mediated DNA repair is illustrated by the increase in the overall spontaneous genome instability in strains lacking MUS81 (Figure 2C, Supplementary Figure S3). Also, similarly to the pol32^-/- mutants, we observed a reduced frequency of 5-FOA^R colonies and cells displaying long-tract LOH (assessed by flow-cytometry) upon I-SceI-dependent DNA-DSB (Figures 3A and 4C) in the mus81^-/- mutants, suggesting that even the GC process is impaired. In Schizosaccharomyces pombe, mus81 mutants have normal frequencies of GC but reduced frequencies of CO (Smith et al. 2003), thus GC and CO can be genetically separated. This might not be the case in C. albicans as both GC and CO seem to be reduced in mus81^-/- mutants as illustrated by the decrease in GC with CO events (Figure 4B), although we cannot refute that the high frequency of CL observed within this genetic background is skewing our appreciation of the frequency of GC with CO.

We examined the role of a third gene involved in HR-mediated repair, MPH1, which encodes a helicase promoting D-loop displacement and NCO outcomes during HR-mediated DNA repair (Prakash et al. 2009). Our results suggest that a similar role is undertaken by MPH1 in C. albicans, as its deletion enhances spontaneous genomic instability (Figure 2C, Supplementary Figure S3) and results in an augmentation in GC with CO as shown by characterization of I-SceI-dependent long-tract LOH events (Figure 4C). In addition, similar to observations made in S. cerevisiae where mph1^-/- strains favor BIR (Mehta et al. 2017), our characterization illustrates an increase in BIR/MCO events within I-SceI-dependent long-tract LOH events (Figure 4C). The anti-CO function of MPH1 was shown to require the activity of the MMR component MutSα (Tay et al. 2010). Thus, unsurprisingly we observed an increase in GC with CO within our msh2^-/- genetic background (Figure 4B). The MSH2 knock-out phenotype is the same as previously reported in C. albicans by Legrand et al. (2007). However, we described a slightly increased genome instability and observed a contrasting sensitivity to MMS between our two strains (Figure 2B). Defects in the MMR pathway lead to mutation accumulation in cells (Drotschmann et al. 1999; Shcherbakova and Kunkel 1999; Goellner 2020). Because our msh2^-/- I-SceI TargetB strain possesses a genotoxic sensitivity profile similar to the msh2^-/- mutants previously described (Legrand et al. 2007), we suspect that the msh2^-/- I-SceI TargetA strain underwent a mutation rendering it sensitive to MMS as it did not demonstrate any other obvious discrepancies with the msh2^-/- I-SceI Target B strain, e.g., discrepancies between msh2^-/- strains was not observed in latter analysis of the proportion of I-SceI-dependent molecular mechanisms. Overall, our thorough phenotypic characterization highlighted new leads worth of further investigation to better understand the subtle differences between S. cerevisiae and C. albicans in terms of maintenance of genome integrity.

We then studied the importance of each of these four genes on ascending LOH tracts toward the centromere upon DNA-DSB repair. We initially observed that, as expected, homozygosis would span from the DNA-DSB site toward the telomere but would also ascend up to 118 kb from the DNA-DSB break site toward the centromere in WT isolates having undergone BIR/MCO as a consequence of I-SceI-dependent DNA-DSB (Figure 1B). This observation implies that the effect of LOH is not only restricted to the alleles distal to the DSB, as more commonly assumed, but could also impact alleles proximal to the DSB, and this over hundreds kilobases. We sought out to try to understand which molecular mechanisms could be involved in these unexpected homozygous regions by investigating precisely the homozygosis tracts situated around I-SceI-break sites of multiple isolates that displayed I-SceI-dependent BIR/MCO profiles (Supplementary Figure S2) from each previously described deletion background (Figure 6). Despite our small sample size, our data suggest that null mutants of either pol32^-/-, msh2^-/-, or mph1^-/-, displayed shorter median ascending LOH tracts toward the centromere as compared to the wild-type genetic background, independently of which haplotype was targeted by I-SceI nuclease (Figure 7, B and C). As repeat sequences are known for their recombination properties, our initial analysis took into consideration the potential impact of repeat sequence RB2, a sub-region of major repeat sequences (MRSs) of C. albicans, on the occurrence and length of ascending LOH. We also accounted for the presence of other repeat sequences, such as the recently described long repeat sequences (Todd et al. 2019) which can be found between the I-SceI target site and CEN4. We observed that few LOH breaks sites precisely coincide with the several long repeats located between positions 780,294 and 782,728 on Chr4, in proximity of the I-SceI target site. We cannot exclude that these sequences could be responsible for the ascending 898 bp long LOH events detected, as defined by our assay (Figures 6 and 7A). However, when we remove the BIR/MCO isolates displaying 898 bp long ascending LOH from our analysis, the median ascending LOH size remains inferior in pol32^-/-, msh2^-/-, and mph1^-/- genetic backgrounds as compared to WT suggesting that these repeat sequences do not influence our conclusions.

Deletion of POL32 in S. cerevisiae is often used to inhibit BIR as it was described to be essential for BIR and dispensable for GC (Lydeard et al. 2007). Shorter ascending LOH tracts upon POL32 deletion could imply that these overflows of LOH toward the centromere may be associated with the BIR molecular mechanism, as inhibition of the BIR process seems to reduce their appearance (Figure 7). BIR has been described in detail in S. cerevisiae, and shown to require extensive 5’ resection as the majority of BIR events required >2.5 kb of 5’ resection, of which 50% of these events showed ≥15.5 kb resection stretches and a few events even spanned more than 27.5 kb (Chung et al. 2010). Our observations suggest that extensive 5’ resection is also a feature of BIR in C. albicans and thus could explain why homozygosis tracts associated with BIR often do not start exactly in the close vicinity of the DNA-DSB and overflow toward the centromere. Indeed, extensive 5’ resection leading to a long 3’ single-strand DNA renders more efficient the homology search in heterozygous organisms. We also described a reduction in ascending LOH size within the MSH2 deletion background suggesting that the MMR machinery may also be implicated. The MMR pathway plays an important role by augmenting replication fidelity and genome stability through correction of base mismatch errors. The MMR machinery is spatially and temporally coupled to DNA replication and is active behind proofreading replicative DNA polymerases Polδ, resulting in a reduction in polymerase activity-associated mutations (Drotschmann et al. 1999; Harfe and Jinks-Robertson 2000; Healey et al. 2016; Boyce et al. 2017). The MMR machinery is also recruited along the DNA replication machinery at the site of DNA-DSB repaired by BIR. In these 5’ resection regions, MMR may “correct” natural heterozygous positions resulting in homozygosis from the DNA-DSB site toward the centromere. Lastly, the median ascending LOH size seems reduced in mph1^-/- genetic background (Figure 7B). This observation could be explained by the stabilization of the D-loop upon loss of the heteroduplex DNA dissolvase Mph1, allowing to hypothesize that either less 5’ resection could be needed to stabilize BIR, which is consistent with the finding of Mehta et al. showing that BIR is favored upon MPH1 deletion (Mehta et al. 2017), or that MPH1 deletion could favors COs or MCO. With our current system, we cannot differentiate BIR from MCO events although BIR has been described as a highly inaccurate process leading to > 1,000 times more indels than during normal replication in S. cerevisiae (Deem et al. 2011). A precise analysis of our sequencing data could allow us to quantify and compare indel levels between isolates in hopes to distinguish both molecular mechanisms. However, each one of the above hypotheses regarding reduction of median ascending LOH size cannot explain the >20 kb ascending events, thus we cannot refute that these latter events occurred spontaneously and are not associated with I-SceI-dependent DNA-DSB repair. Of interest, upon removal of the >20 kb ascending LOH events, we still observe an effect of pol32^-/-, msh2^-/-, and mph1^-/- deletion on the median LOH size. In addition, these ascending homozygosis tracts could also be influenced by 3’ single-strand DNA instability or nucleases activity during DNA-DSB repair, two processes which remain to be explored in C. albicans.

Overall, BIR/MCO in C. albicans are frequently associated with extended regions of homozygosis from the DNA-DSB locus toward the centromere. Despite our small sampling size, we also propose that the MMR pathway could act upon natural heterozygous positions during DNA-DSB repair by BIR/MCO and participate in the overflow of homozygosis toward the centromere. For the first time, we describe that long-tract homozygosis start sites do not perfectly coincide directly with DNA break sites in C. albicans. This notion is striking in this pathogenic yeast as heterozygosity levels are positively correlated with strain fitness, while here, we show that LOH through BIR is a messy process in C. albicans which potentially impacts dozens of additional alleles proximal to the centromere.

Data and reagent availability

Genome sequences of the 51 engineered C. albicans isolates described in this study have been deposited in the NCBI Sequence Read Archive under BioProject ID PRJNA674870. Supplemental Material available at figshare DOI: https://doi.org/10.25386/genetics.14054156. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables.