Differences in homologous recombination and maintenance of heteropolyploidy between Haloferax volcanii and Haloferax mediterranei

Abstract Polyploidy, the phenomenon of having more than one copy of the genome in an organism, is common among haloarchaea. While providing short-term benefits for DNA repair, polyploidy is generally regarded as an “evolutionary trap” that by the notion of the Muller's ratchet will inevitably conclude in the species' decline or even extinction due to a gradual reduction in fitness. In most reported cases of polyploidy in archaea, the genetic state of the organism is considered as homoploidy i.e. all copies of the genome are identical. Here we demonstrate that while this is indeed the prevalent genetic status in the halophilic archaeon Haloferax volcanii, its close relative H. mediterranei maintains a prolonged heteroploidy state in a nonselective environment once a second allele is introduced. Moreover, a strong genetic linkage was observed between two distant loci in H. mediterranei indicating a low rate of homologous recombination while almost no such linkage was shown in H. volcanii indicating a high rate of recombination in the latter species. We suggest that H. volcanii escapes Muller's ratchet by means of an effective chromosome-equalizing gene-conversion mechanism facilitated by highly active homologous recombination, whereas H. mediterranei must elude the ratchet via a different, yet to be elucidated mechanism.


Introduction
Polyploidy, having multiple copies of the same chromosome in a cell, is a well-known property of different haloarchaeal species and was demonstrated as well as characterized in Haloferax volcanii, H. mediterranei, and H. salinarum (Breuert et al. 2006). Polyploidy is also found in multiple additional members of the Euryarcheota (Ludt and Soppa 2019). The number of chromosomal copies a cell harbors depends on the growth phase of the culture and varies from 1 (Malandrin et al. 1999) and up to 55 (Spaans et al. 2015). Potential evolutionary benefits of polyploidy vary from one species to another and include the use of DNA as a phosphate storage , and an improved ability to repair DNA breaks (Hildenbrand et al. 2011;Perez-Arnaiz et al. 2020), including those generated by autoimmunity caused by the microbial immune system CRISPR-Cas (Stachler et al. 2017).
Homologous recombination is facilitated by polyploidy since it guarantees multiple wild-type templates that could be used for double strain break repair in archaea (Kish and DiRuggiero 2008;Jones and Baxter 2017) and bacteria (Zahradka et al. 2006). Polyploidy however could also be costly, especially in harsh mutagenic environments because recessive deleterious mutations can easily accumulate in the genome. Such a process can later gradually lead to unfit or even nonviable daughter cells lowering the fitness of the species. Gene conversion and equalization of the genome copies (Soppa 2011) as well as lateral gene transfer and homologous recombination (Papke et al. 2004) was suggested as possible strategies to overcome the accumulation of deleterious alleles and a way to lessen the mutation load (Kondrashov 1994) and escape genetic degradation via Muller's ratchet (Markov and Kaznacheev 2016) in those asexual nonmitotic organisms.
While polyploidy is common in many prokaryotes (archaea and some bacteria ;Soppa 2011;Markov and Kaznacheev 2016), those polyploids are generally regarded as homopolyploid, namely having genome copies that contain the same genetic information.
This contrasts with the prevalent polyploidy in plants that display autopolyploidy (a multiplication of the chromosomal set) in some species alongside allopolyploidy (chromosome doubling accompanied by hybridization of two parents, often of different species) in others (Soltis et al. 2015;Stuessy and Weiss-Schneeweiss 2019).
In the absence of mitosis, polyploidy could lead to an accumulation of deleterious mutations, and an ongoing reduction in the fitness of the population, unless there are mechanisms that can counter this process (Soppa 2011;Markov and Kaznacheev 2016). One such mechanism could be lateral gene (or in this case allele) transfer between cells and indeed H. volcanii and H. mediterranei are known in their ability to create long-lasting cytoplasmic bridges that enable the passage of DNA within the species (Rosenshine et al. 1989) and even creating a heterodiploid fused interspecies cell hybrid (Naor et al. 2012). These haloarchaea thus somewhat resemble plant allopolyploid species. These heteropolyploid hybrids are possible to maintain under selective pressure but are quickly resolved back to homopolyploidy in the absence of such selection. While it is well established that the application of selective pressure on H. volcanii can preserve a heteropolyploid state created by cell-cell mating or an artificial semi-heteropolyploidy created via plasmid transformation , those so-called merodiploid states are lost when the selective pressure is relieved and the genome returns to its homoploid state.
When considering the polyploidy of H. volcanii, the ease in which genomic deletions and mutations are introduced into this species is quite surprising (Bitan-Banin et al. 2003;Allers and Mevarech 2005). A genome-homogenizing process that agrees with the observations discussed above seems to underlie the relative ease of creating gene knockouts in H. volcanii despite its large genome copy number. Similar observations have been made in the methanogen Methanococcus maripaludis (Hildenbrand et al. 2011). In contrast, we have observed that when using the same vectors and procedures commonly applied for making gene knockouts for H. volcanii, H. mediterranei gene knockouts require laborious passages in selective media, even under low phosphate conditions where ploidy is reduced (Turgeman-Grott, personal communication). This observation suggests that despite the 86% sequence identity between H. mediterranei and H. volcanii, H. mediterranei does not homogenize its genome copies as efficiently as H. volcanii. This implies that the two related species might also demonstrate additional differences in replication and DNA repair.
The ability of a polyploid cell to equalize quickly up to 20 copies of its genome under selective pressure could be explained either by segregation and a strong heteropolyploidy fitness cost or by a robust homologous recombination and gene-conversion mechanism (Hildenbrand et al. 2011); the latter are presumably also important for additional cellular processes (Barilla 2016;White and Allers 2018). In this paper, we have studied the ability of both H. volcanii and H. mediterranei to retain a heteropolyploid state artificially created by cell-cell fusion, when selection is relieved. We also examine their respective subsequent recombination rates in the absence of selection. We show that when selection is relieved, H. volcanii equalizes its multiple genome copies efficiently while H. mediterranei does not. We also show surprising differences in recombination rates between the two species, using both cellcell mating and plasmid-chromosome recombination assays. While spontaneous homologous recombination in H. volcanii is extremely common, a strong genetic linkage is observed between two distant loci in H. mediterranei. Comparative RNA-seq displayed no obvious differences in expression levels of factors of the DNA replication pathway. A significant difference in expression levels of factors of the homologous recombination pathway was observed in stationary-phase H. mediterranei cells, with an increase in the recombination mediator radB expression.

Culture conditions
Nonselective medium refers to Hv-YPC medium (Allers et al. 2004) supplemented with 50 μg mL −1 thymidine. Selective media were based on Hv-Ca medium (Allers et al. 2004) supplemented with 50 μg mL −1 of uracil, tryptophan, or thymidine, as needed. All growth steps were carried out at 45°C. Liquid cultures were grown in 2.5 mL of medium, with agitation at 180 rpm.

H. volcanii and H. mediterranei strain construction
H. volcanii strains used in this study were derived from the parental strain H133 (Allers et al. 2004) that contains the deletions ΔtrpA (HVO_0789, tryptophan auxotrophy) ΔhdrB (HVO_2919, thymidine auxotrophy; Tables 1 and 2). H133 also contains deletions in the leuB and pyrE2 genes that were not used in this work. Furthermore, deletion of the crtI (HVO_2528) gene, which is involved in biosynthesis of the carotenoid pigment bacterioruberin, and is not essential nor has an effect on fitness (Turkowyd et al. 2020) resulted in a white phenotype. Strain WR780 [ΔtrpA, ΔhdrB, 1585::trpA (red)] was constructed by replacing the nonessential ASC-like gene HVO_1585 in the ΔcrtI strain with the trpA gene. Strain WR807 [ΔtrpA, ΔhdrB, ΔcrtI 1585::hdrB (white)] was constructed by replacing HVO_1585 with hdrB. Essentially the same process was repeated with H. mediterranei, but for practical reasons the pyrE gene (HFX_0318, uracil auxotrophy) was used as a selective marker instead of folA to generate strain UG482, along with trpA (HFX_0748) to generate strain UG480 This gene has been used before as a selectable marker in H. mediterranei (Naor et al. 2012)

Plasmid and primers
Plasmids used to create H. volcanii strains were constructed by joining ∼500 bp of each flanking sequence of the gene to be deleted into the suicide vector pTA131 (Allers et al. 2004) using specific restriction enzymes constructed on primers (Table 3). When trpA or hdrB was used to replace the deleted gene, the replacement gene under the ferredoxin promoter (P fdx ) was excised from pTA106 (trpA) or pTA187 (hdrB) (Altman-Price and Mevarech 2009) with restriction enzymes EcoRI and XbaI (NEB) and inserted by restriction and ligation between the flanking regions into pTA131 cut with the appropriate enzymes.
Plasmids used to create H. mediterranei strains were constructed using restriction enzymes or the Gibson-assembly method as described in Table 4. Briefly, the sequences to be joined were amplified by PCR with ∼25 bp long overlapping regions on either side (Table 3). ∼30 ng of each purified fragment were mixed and incubated with 10 μL Gibson-Assembly Master Mix (NEB) with pTA131 that was amplified with suitable primers, for 1 h, in a final volume of 20 μL. Ligated plasmids were transformed into E. coli (DH5α) using electroporation. Transformed bacteria were grown on LB-Agar plates supplemented with 100 μg mL −1 ampicillin. All plasmids were sequenced to ensure they contained the desired inserts.

Haloarchaeal strain construction
All transformations of archaea were performed using the PEG600 protocol described in the HaloHandbook (https://haloarchaea. com/halohandbook/). Deletion strains were constructed by transforming cells with pTA131 plasmid containing 500 bp-long flanking regions of the targeted gene, and further selection steps as previously described (Bitan-Banin et al. 2003). Gene replacements were done by transforming archaea with approximately 2 μg of linear, purified PCR products composed of 500 bp-long flanking regions of the unwanted gene surrounding the gene to be inserted. Transformed colonies were then selected by plating on Hv-Ca media supplemented with the appropriate metabolites. All strains were verified using PCR and selective plating.

Mating protocol
Cultures were grown overnight and adjusted to an OD 600 of 1. 2 mL of each strain were mixed and filtered through a 0.2 μm filter using a syringe to dispose of excess media. The filter was placed on a nonselective plate and incubated overnight. Next, the cells were suspended in selective media and plated on selective plates. The

Heteropolyploids growth assay
Cultures were grown overnight and diluted to an OD 600 of 1. A volume of 100 µL of each culture were used to inoculate 2 mL of rich medium. The culture was left to shake at 45°C and samples were taken as described in the text. For the 1-week assays, 100 µL of the culture was used to inoculate a fresh 2 mL medium every 1-2 days.

Pop-in assay
Plasmid-chromosome recombination assays, otherwise referred to here as "pop-in" assays, were performed as described in Lestini et al. (2010). Transformations were plated on Hv-Ca and Hv-YPC plates to analyze the transformation efficiency and viable count, respectively. The number of colonies on both plates was counted and the recombination rate was calculated as a fraction of transformants on Hv-Ca plates compared to Hv-YPC.

The heteropolyploid state is stable under selection
To investigate differences in genome-homogenization and recombination between H. mediterranei and H. volcanii, we generated two pairs of strains, an H. volcanii pair and an H. mediterranei pair, to be employed in mating assays (either volcanii-volcanii or mediterraneimediterranei mating). The H. volcanii paternal strain (Allers et al. 2004) contained deletions in the trpA and hdrB genes (tryptophan and thymidine auxotrophy respectively) and the two strains were constructed such that the trpA or hdrB cassettes were inserted back into the same ectopic location, instead of the ASC-like gene HVO1585, which was previously demonstrated to be a nonessential open reading frame (Moshe Mevarech, personal communication). This genetic background allowed for the selection of mated cells on growth medium lacking tryptophan and thymidine. Typically, such mating assays result in a homopolyploid cell containing a genome harboring both cassettes, which would be the outcome of recombination events. The unique construction of the current mated pair however prevents such recombination events from happening as the cassettes were inserted into the same genomic location, hence the mated cells are forced to contain two different types of genomes and are essentially held under selection in a heteropolyploid state. In addition to the metabolic selectable marker, a second marker was introduced to one of the pair strains-a deletion in the crtI (HVO_2528) gene, which is involved in the biosynthesis of the carotenoid pigment bacterioruberin and is not essential nor has an effect on fitness (Turkowyd et al. 2020). This deletion resulted in a white phenotype (WR807) in contrast with the other strain that remained naturally pigmented (red) (WR780). Essentially the same process was repeated with H. mediterranei but for practical reasons the pyrE gene (HFX_0318, uracil auxotrophy) was used as a selective marker instead of folA (hdrB homologue) to generate strain UG482 (white), along with trpA (HFX_0748) to generate strain UG480 (pink-red, see "Materials and methods"). In total, two pairs of strains were constructed, each of which contained two different markers: one for with UG482, respectively. The heteropolyploid state of the mating products was confirmed by PCR amplification of the loci that were under selection, using the primers T36 + T37 and T28 + T29 (Table 3). Supplementary Fig. 1 provides a schematic drawing describing the assay and results.

H. volcanii and H. mediterranei lose heterozygosity at different rates under nonselective conditions
The heteropolyploid state was stable as long as the mating products were kept under selection but started decreasing at different rates when selection was relieved as described next. To explore heterozygosity maintenance in the absence of selective pressure, single heteropolyploid H. volcanii or H. mediterranei colonies were picked into nonselective YPC liquid media and samples were taken 48 h (5 biological replicates) and 1 week (5 additional biological replicates) after inoculation (fresh growth medium was added every other day) and were plated on nonselective media. 50 colonies were then taken from each plate and streaked on two selective media types each with selection to only one of the metabolic markers. The number of colonies that grew on both types of media was enumerated and used to calculate the percentage of heterozygous colonies in the population. To corroborate the results of these viable counts, the genotype of about 20% of the colonies was also determined by PCR and was found to match the growth phenotype ( Supplementary Fig. 2). We observed nearcomplete loss of heterozygosity in H. volcanii, but not in H. mediterranei after 1 week of growth. After 48 h in nonselective media, both H. volcanii and H. mediterranei mating products showed a decrease of the heteropolyploid state to an average of approximately 30% of screened colonies. The replicates showed a wide variation of the heteropolyploid state ranging from 4% to 56% in H. mediterranei and from 8% to 48% in H. volcanii. However, a near-complete loss of heterozygosity in H. volcanii, but not H. mediterranei was observed after 1 week, at which point H. mediterranei was still showing an approximately 30% heteropolyploidy (ranging from 6% to 88%) while 4 of the 6 H. volcanii replicates showed no heteropolyploidy and the other 2 replicates showed 4% and 10% heteropolyploidy (Fig. 1).

Recombination rate is much higher in H. volcanii than in H. mediterranei
Since each chromosome of the heteropolyploid cells described above contained one color marker and one selective marker, there are four possible genotype combinations that can, in theory, exist in selected cells. Notably, since only two of these four combinations are present in any of the original strains of each species, emergence of the two other combinations can only result from recombination events in cells that contain both types of the original chromosomes.
Immediately after mating, mated cells are heteropolyploid, yet once selection is relaxed cells can segregate to a homopolyploid state (as above). We next examined the recombination rate of segregated cells in the two species. Since each strain has a unique combination of markers, it was possible to determine whether segregated (homopolyploid) cells were similar to one of the original strains or contained a new combination of color and selective marker. The colonies obtained after 48 h and inferred to be homopolyploid in the previous analysis were thus analyzed (Fig. 2). H. volcanii colonies demonstrated a high rate of recombination where about 50% of the colonies retained the original (parental) marker combination while the other 50% showed a recombinant genome. In contrast, about 80% of H. mediterranei cells retained the original (parental) markers combination. These results suggest that in H. volcanii the recombination rate is higher than in H. mediterranei, especially taking into account that the genomic distance between the markers used is double in H. volcanii compared to H. mediterranei (∼0.5 M bp in H. mediterranei and ∼1 M bp in H. volcanii). Colony color-based results were validated by PCR (conducted on a sample of 5% of the colonies) and confirmed in 100% of cases.
To directly compare the recombination rates of these species, we then performed plasmid-chromosome recombination (integration or "pop-in") experiments in which H. mediterranei and H. volcanii cells were transformed with a nonreplicating plasmid bearing a pyrE2 selectable marker. In such experiments, only cells that integrate the plasmid into their chromosome via homologous recombination can grow. Four plasmids for H. mediterranei (pTA2684, pTA2688, pTA2692, pTA2696) and four plasmids for H. volcanii (pTA2686, pTA2690, pTA2694, pTA2698) were constructed for this purpose. Each plasmid was constructed in a pTA131 background with a DNA fragment of approximately 1,000 bp that corresponds to a portion of either the H. mediterranei  or H. volcanii genome. Orthologous genes were chosen for H. mediterranei and H. volcanii since they occupy similar positions in their respective genomes and share high gene similarity between species. All plasmids carry the pyrE2 selectable marker (pyrE2 from H. volcanii) that is suitable for both organisms (Table 3). Recombination rates were calculated as the percentage of the cells that integrated the plasmids compared to the viable count. Transformation efficiency was calculated using a replicating plasmid (pTA230) to control for the effect of transformation efficiency. We observed that for all DNA fragments, the plasmid integration rates were generally over 5-fold higher in H. volcanii than in H. mediterranei (Fig. 3). Taken together with the segregation experiments described above, these findings indicate significantly higher recombination rates in H. volcanii compared to H. mediterranei.

Comparative RNA-seq analysis of the DNA replication and homologous recombination pathways in H. mediterranei and H. volcanii
To determine whether a difference in RNA expression levels could explain the variations described above, a comparative analysis of DNA replication and homologous recombination factors in both Haloferax species was carried out (a full RNA data are provided as a Supplementary File. The raw data are available). Most genes involved in DNA replication and replication-restart showed no statistically significant differences in expression levels between the two species (Supplementary File 1, Tables 1-4). In terms of the homologous recombination pathway, although radA, mre11, rad50, and genes encoding DNA ligases had similar expression levels in both species, an increase in radB (Wardell et al. 2017) expression was noted in H. mediterranei stationary-phase cells. This difference could result in divergent processing of DNA breaks and explain the decreased homologous recombination rate in H. mediterranei.

Discussion
H. volcanii and H. mediterranei are two closely related halophilic archaea species. Naor et al. (2012) reported an average nucleotide sequence identity of 86.6% between H. volcanii strain DS2 and H. mediterranei R4 strain. They also showed that these strains are able to perform intra-species cell-cell fusion followed by an exchange of genomic fragments via recombination and have a similar genome architecture. However, these two species have a strikingly different gene content in their respective megaplasmids (3 additional replicons in each species; Hartman et al. 2010;Han et al. 2012). Despite their similarities, the two species demonstrate a variety of genetic differences, one of which is discussed in this work. We show that while H. volcanii seems to equalize its genome copies quickly and effectively, H. mediterranei can maintain a state of semi-heterozygosity/heteropolyploidy for many generations; this corresponds to a high frequency of HR events in H. volcanii, and much lower rates in H. mediterranei.
Two possible mechanisms could underlie these differences in heterozygosity; modifications in chromosome segregation or variations in gene-conversion and recombination mechanisms. The knowledge of archaeal chromosome segregation is limited and sparse. While it was suggested that the same SegAB-dependent mechanism may apply in monoploid or diploid archaea of both Euryarchaeota and Crenarchaeota (Barilla 2016), little is known about the mechanism of chromosome segregation in polyploid archaea. Importantly, most Euryarchaeota and halophilic archaea in particular are highly polyploid (Chant and Dennis 1986;Breuert et al. 2006;Liu et al. 2013;Zerulla and Soppa 2014;Spaans et al. 2015;Barilla 2016;Ludt and Soppa 2019), further complicating segregation issues.
While not much is known about differences in chromosome segregation between the two species, evidence about differences in the way that recombination is connected to replication does exist. An H. volcanii strain lacking all four origins of replication was strictly dependent on the RadA recombinase (Hawkins et al. 2013), hence it was proposed that recombination alone can support DNA replication in H volcanii. This phenomenon was also later suggested for the hyperthermophilic archaeon T. kodakarensis, which is also polyploid (Gehring et al. 2017). In contrast, an H. mediterranei strain that had its three major chromosomal origins deleted was viable but slow growing and was shown to be dependent on a dormant origin (oriC4) that became essential in the absence of the other three (Yang et al. 2015). Furthermore, unlike the origin-deleted mutant of H. volcanii, that H. mediterranei mutant showed higher radA expression but was viable in a radA suppressed background, indicating that it was dependent on the cryptic origin rather than on HR for replication. These past studies indicate that in the total absence of origins, H. volcanii can depend on recombination-initiated replication whereas H. mediterranei cannot. This could be attributed either to differences in genome replication or a failure of H. mediterranei HR machinery to support DNA replication.
Our observations that the recombination frequency appears to be much lower in H. mediterranei than in H. volcanii suggest that homologous recombination operates at a different level in those two species. Genome similarity is consistent with protein similarity. The two species generally have highly similar protein orthologs with a median amino acid identity of 86% (calculated with http://ekhidna2.biocenter.helsinki.fi/AAI/; Medlar et al. 2018), and both have the entire set of known recombination proteins. For example, RadA and RadB (a paralog of RadA that has been shown to mediate HR in H. volcanii; Wardell et al. 2017), of H. volcanii and H. mediterranei are 94% and 95% identical. However, Rad50 and Mre11 are surprisingly divergent, sharing only 82% and 78% identity, respectively. Thus, the difference between these species could lie at the level of resection of DNA by the Mre11-Rad50 complex, which should be tested experimentally by swapping these Fig. 3. Rate of recombination of H. volcanii vs H. mediterranei, measured using a "pop-in" assay. Four nonreplicative plasmids for H. mediterranei (pTA2684, pTA2688, pTA2692, pTA2696) and four nonreplicative plasmids for H. volcanii (pTA2686, pTA2690, pTA2694, pTA2698) were transformed into the cells. Recombination rates were calculated as the percentage of the cells that integrated the plasmids compared to the viable count. Transformation efficiency was calculated using a replicating plasmid (pTA230) to control for the effect of transformation efficiency.
factors. Alternatively, a difference in RadB affinity for singlestranded DNA, or the differences in gene expression (as determined in our RNA-seq data) between the two species could explain the differences described above.
Polyploidy confers an array of immediate advantages to the organism including multiple homologous templates for mutational and strand break repair. However, in the absence of an accurate DNA segregation mechanism, the evolutionary long-term consequences of polyploidy, according to the logic of Muller's ratchet (Muller 1964) and later models (Markov and Kaznacheev 2016) would lead to a high segregation load and the accumulation of deleterious recessive alleles, even resulting in giving rise to nonviable daughter cells. Therefore, a polyploid yet nonmitotic organism would face a strong purifying selection against the existence of the species. By extension, any extant polyploid organism has either found a way to escape the ratchet (Ludt and Soppa 2019) or at least substantially slow it down.
Several mechanisms have been suggested that can result in the organisms overcoming the evolutionary trap of polyploidy. Interestingly at least half of these suggested strategies are coupled with or mechanistically connected to recombination (Markov and Kaznacheev 2016): Ploidy cycle, namely the periodic decrease in polyploidy such that the daughter cells are subjected to selection, equalization of the chromosomes by gene-conversion thus exposing said deleterious alleles to selection in their homozygote state, lateral gene transfer and recombination that were shown to have a synergistic effect of rescuing otherwise "doomed" populations by creating viable chromosomes from the fragments of existing genes, and acquired chromosomes the latter phenomenon potentially enhanced by cell-cell mating.
Here we show that while H. volcanii might escape the ratchet via efficient chromosomes equalization, H. mediterranei is much slower in doing so. According to the models, recombination is an important part of many ratchet evading mechanisms, but it shows a much higher rate in H. volcanii than in H. mediterranei. How H. mediterranei escapes Muller's ratchet while still maintaining substantial "heterozygosity" for multiple generations remains to be elucidated.

Data availability
Raw RNA-seq data were generated at University of Nottingham. Derived data supporting the findings of this study are available in a separate supplementary file. The raw sequence data for the RNA-seq experiments are available at SRA, accessions ERS13562943-ERS13562948. Strains and plasmids are available upon request.
Supplemental material available at G3 online.

Conflicts of interest
None declared.