The parABSm system is involved in megaplasmid partitioning and genome integrity maintenance in Thermus thermophilus

Abstract The characteristics of the parABS system in polyploid bacteria are barely understood. We initially analyzed the physiological functions and mechanisms of the megaplasmid parABSm system in the thermophilic polyploid bacterium Thermus thermophilus. Deletion of parABm was possible only when a plasmid-born copy of parABm was provided, indicating that these genes are conditionally essential. The cell morphology of the parABm deletion mutant (ΔparABm) was changed to some extent, and in certain extra-large or twisted cells, the nucleoids were dispersed and damaged. Compared with that of the wild type, the frequency of anucleate cells was significantly increased. Genome content analyses showed that ΔparABm had lost ∼160 kb of megaplasmid and ∼23 kb of chromosomal sequences, respectively. Genome fluorescent tagging and PFGE experiments demonstrated that the truncated megaplasmid was frequently interlinked and could not be segregated correctly; thus, certain daughter cells eventually lost the entire megaplasmid and became twisted or enlarged with damaged nucleoids. Further, we found that when the megaplasmid was lost in these cells, the toxins encoded by the megaplasmid toxin–antitoxin (TA) systems (VapBC64_65 and VapBC142_143) would exert detrimental effects, such as to fragment DNA. Thus, parABSm might ensure the existence of these TA systems, thereby preventing genomic degradation. Together, our results suggested that in T. thermophilus, the megaplasmid-encoded parABS system plays an essential role in the megaplasmid partitioning process; also it is an important determination factor for the genome integrity maintenance.


Introduction
Plasmids are ubiquitous in prokaryotic bacteria and play essential roles in cell metabolism, pathogenesis, and species evolution. Naturally occurring plasmids come in different sizes, namely, plasmids (small, usually 10 3 -10 5 bp, including multicopy nonconjugative plasmids and low-copy conjugative plasmids), megaplasmids (the thresholds for minimum megaplasmid size are difficult to be defined, but they are normally >5% genome size), and "chromids" (different from megaplasmid and frequently encode essential genes) (Hall et al. 2022). How plasmids can be stably maintained with a constant copy number in host cells is of great interest for microbiologists. It is suggested that plasmid positioning and segregation mode depend on their copy numbers (Salje 2010;Million-Weaver and Camps 2014;Baxter and Funnell 2014). Plasmids with high copy number tend to segregate randomly in a passive way (Nordström and Gerdes 2003;Million-Weaver and Camps 2014;Reyes-Lamothe et al. 2014). By contrast, low-copy-number plasmids (normally 1-2 copies of plasmid per cell, e.g. F and P1 plasmids in Escherichia coli) have evolved active partitioning systems to ensure their faithful inheritance to daughter host cells (Million-Weaver and Camps 2014; Baxter and Funnell 2014;Bouet and Funnell 2019). These plasmid partitioning systems are usually consisted of three elements: one or several copies of cis-acting site (centromere-like sequence area), one centromere-binding protein (adaptor protein), and one proteinencoding ATPase or GTPase (motor protein) (Million-Weaver and Camps 2014; Bouet and Funnell 2019). The adaptor and motor protein-encoding genes are arranged in one operon and autoregulated; the centromeric sequence is often positioned near the operon and is composed of tandem repeat (direct or inverted) sequences (Million-Weaver and Camps 2014; Hu et al. 2017;Bouet and Funnell 2019). For low-copy-number plasmids, disruption of either one or three of these partitioning factors would cause segregation defects, thus yielding plasmid loss at high rates (Ebersbach and Gerdes 2001;Li et al. 2004).
Based on the enzyme activity type of the motor protein, the plasmid partitioning systems have been classified into three types (Bouet and Funnell 2019). The most prevalent one is the type I systems which feature a Walker-type ATPase, for example, the wellstudied parABS system of the E. coli P1 plasmid (Gerdes et al. 2000;Bouet and Funnell 2019). In parABS, parS is the centromere site, ParA is the ATPase, and ParB is the parS-binding protein. When the parABS system starts to partition plasmids, ParB binds parS and extends to neighboring DNA forming a large nucleoprotein complex, while ParA proteins are polymerized to filaments. The ParA filaments can contact the ParB-parS complex, once upon ATP hydrolysis, they will rapidly disassemble, thereby pulling plasmids to the quarter-cell position prior to cell division (Gerdes et al. 2000;Baxter and Funnell 2014;Hu et al. 2017;Bouet and Funnell 2019). Type II and type III plasmid partitioning systems employ actin-like ATPase and tubulin-like GTPase as the motor proteins, respectively, and also work via a dynamic motor protein polymerization mechanism (Møller-Jensen et al. 2002;Larsen et al. 2007;Bouet and Funnell 2019). Besides being present in the low-copy-number plasmids, homologs of the type I system (i.e. the parABS system) were later found in two-thirds of bacterial chromosomes that have been sequenced (Gerdes et al. 2000;Jalal and Le 2020). However, in contrast to the plasmid counterpart, the chromosomally encoded parABS system has pleiotropic functions (Murray and Errington 2008). For example, they were shown to be involved in chromosome replication and segregation in Bacillus subtilis (Lee and Grossman 2006); initiation of cytokinesis and regulation of DNA replication initiation in Caulobacter crescentus (Mohl et al. 2001;Murray and Errington 2008); cell growth, chromosome segregation, and cell motility in Pseudomonas aeruginosa (Bartosik et al. 2004;Lasocki et al. 2007); and cell morphology maintenance in Pseudomonas putida (Lewis et al. 2002). Further, the participation of the chromosomal parABS orthologs in the chromosome segregation process is not as significant as that of their plasmid equivalents (Badrinarayanan et al. 2015). Although deletion of the chromosomal parAB genes in some bacterial species (e.g. B. subtilis and P. aeruginosa) could elevate number of anucleate cells, most cells still exhibited normal chromosome segregation (Lee and Grossman 2006;Lasocki et al. 2007). In some cases, mutation of these genes would not cause cell growth defect or increasement of anucleate cells (e.g. in Vibrio cholerae and Thermus thermophilus) (Fogel and Waldor 2006;Li et al. 2015). Therefore, it is currently explained that the chromosomal parABS systems mainly function in the segregation of origin-proximal regions but not of bulk chromosomes (Badrinarayanan et al. 2015).
Besides low-copy-number plasmids and bacterial chromosomes, certain bacterial megaplasmids also encode the parABS orthologs (e.g. Deinococcus radiodurans and T. thermophilus); however, their physiological functions and mechanisms were nearly unknown (Li et al. 2015;Maurya et al. 2019). In recent decade, a series of bacteria were determined to be polyploid, which means one single cell contains multiple copies (>3) of chromosome (Soppa 2014). How these multiple genome copies can be partitioned to the daughter cells (i.e., via a stringent or a random pattern) remains unknown. Moreover, whether the parABS orthologs in these genomes play roles in their vertical transmission also remains to be mysterious. T. thermophilus is able to grow from 50°C to 80°C, and due to its various merits, it has been established as a model organism for studying thermophilic bacteria. The genome of T. thermophilus HB27 (NCBI: GCA_000008125.1) is constituted by a chromosome (1.89 Mb) and a megaplasmid (0.27 Mb), which are both shown to be present with 4-5 copies per cell under slow growth condition (Ohtani et al. 2010;Li 2019a). Both the chromosome and megaplasmid encode the parABS orthologs (termed as parABS c and parABS m , respectively). The megaplasmid also encodes two sets of toxinantitoxin (TA) loci (vapBC belonging to type II TA), which normally function to prevent the survival of plasmid-free cells by a postsegregational killing mechanism or play roles in response to cellular stresses (Kamruzzaman et al. 2021). The TA systems are composed of two genes encoding toxin and antitoxin, respectively; the antitoxins act as antidotes for the toxins. When both genes were disrupted, the concentrations of antitoxins in the cells would be rapidly decreased due to their instability property (Kamruzzaman et al. 2021). The stable and unopposed toxins would then exert their toxic effects leading to plasmid loss or cell death (Bardaji et al. 2019;Kamruzzaman et al. 2021). The toxins can kill the cells by targeting a variety of important cellular processes, including cytoskeleton formation and cell-wall synthesis, membrane integrity, DNA replication, transcription, and translation (Fraikin et al. 2020). VapBC comprises the largest TA family, and its toxin VapC normally functions as a nuclease which may have different target specificities in different organisms (Arcus et al. 2011;Fan et al. 2017).
Our previous studies showed that the chromosomal-encoded parABS c system was not required for the chromosome segregation (Li et al. 2015). The megaplasmid parAB m genes were undeletable; through partially knocking down the genes, we found that the parAB m knockdown mutant demonstrated growth and nucleoid defects and deletion of a portion of the megaplasmid sequences (Li et al. 2015). Therefore, we hypothesized that if the parAB m genes were completely deleted, there might be more detrimental effects for the cells, such as deletion of larger portion of the genome (including the megaplasmid and chromosome) or loss of the entire megaplasmid, nucleoid fragmentation, cell morphology variation, cell death, and severe megaplasmid segregation defect. Also, it is interesting to reveal the mechanism about how the parABS m system functions to maintain the megaplasmid. On the basis of our previous research, in this study, we detailedly addressed the physiological role and working mechanism of the megaplasmid-encoded parABS m system. We found that complete deletion of parAB m genes could cause severe cell growth defect and yield deletion of megaplasmid and chromosome sequences. A fraction of ΔparAB m mutant cells became enlarged or twisted and with aberrant and fragmented nucleoids, in which the truncated megaplasmid was frequently mislocalized. TUNEL (TdT-mediated dUTP nick-end labeling) assay showed that this kind of cells was mostly dead cells. We further proved that deletion of parAB m could eventually lead to loss of the entire megaplasmid in these aberrant cells, thereby triggering TA system-mediated nucleoid degradation and cell death. Together, it was concluded that in T. thermophilus, the parABS m system is not only essential for the megaplasmid partitioning but also critical for the genomic sequence and structure integrity maintenance.

Bacterial strains and growth conditions
E. coli DH5a was the host strain for constructing all the plasmids and was grown in LB medium under 37°C. The wild-type (WT) T. thermophilus strain (HB27, DSM7039) and its derivative strains were grown in TB medium at 70°C or 60°C. TB medium was composed of 4 g/l yeast extract, 8 g/l trypticase peptone (BD Biosciences), and 3 g/l NaCl, and the pH value was adjusted to 7.5. The growth media were supplemented with kanamycin (50 μg/ml and 20 μg/ml for E. coli and T. thermophilus, respectively), bleomycin (Sigma-Aldrich, 15 μg/ml for E. coli and 3 μg/ml for T. thermophilus), or nalidixic acid (Nal) (100 μg/ml) when necessary.

Plasmid constructions
All plasmids and strains used in this study are summarized in Table 1, and the oligonucleotides used for PCR amplification are listed in Supplementary Table 1. The allele exchange vectors pUC-ΔparAB m ::blm, pUC::ΔparB c ::kat-parB c _sgfp, pUC::Δfdh::kat-parB m _ sgfp, pUC-ΔvapBC64_65::kat, and pUC-ΔvapBC142_143::blm were derived from the fundamental vector pUC18. For pUC-ΔparAB m ::blm, pUC-ΔvapBC64_65::kat, and pUC-ΔvapBC142_143::blm, the thermostable antibiotic genes kat and blm were, respectively, PCR amplified from pMK18 and pMB18 vectors (E. coli/T. thermophilus shuttle vectors) used in former studies (de Grado et al. 1999;Brouns et al. 2005;Li et al. 2015). Approximately 1 kb upstream and downstream flanking regions of the parAB m , vapBC64_65, or vapBC142_143 locus were PCR amplified from T. thermophilus HB27 (GCA_000008125.1), respectively. The primers used in these reactions would confer the PCR fragments 20-bp sequences overlapping pUC18 and kat/blm; thereby, the recombinant vectors could be constructed by fragment in-fusion (i.e. Gibson assembly) method (Gibson et al. 2009) (reagents were from New England Biolabs). For constructing pUC::ΔparB c :: kat-parB c _sgfp and pUC::Δfdh::kat-parB m _sgfp, pMK-parB c _sgfp and pMK-parB m _sgfp were constructed first (the schematic designs of the constructs are shown in Supplementary Fig. 1), for which parB c and parB m were, respectively, PCR amplified and translationally fused with the sgfp gene of the pMK-sgfp vector used in a previous study (Li et al. 2015). The kat-parB c _sgfp and kat-parB m _sgfp fragments were then amplified from pMK-parB c _sgfp and pMK-parB m _sgfp and cloned into the middle of the two flanking homology regions (around 1 kb) of parB c and fdh (TTP0138, encoding formate dehydrogenase) under the pUC18 background, resulting pUC::ΔparB c ::kat-parB c _sgfp and pUC::Δfdh::kat-parB m _sgfp, respectively. In these two vectors, parB c _sgfp and parB m _sgfp were both transcribed under the slp promoter (i.e. promoter of the kat gene). The complementation vector pMK-parAB m was derived from pMK18. In detail, parAB m was PCR amplified and transcriptionally fused to the kat cassette of pMK18 using the same Gibson assembly method (Gibson et al. 2009).

Mutant generations
For generation of T. thermophilus ΔparAB m mutant, the common method for deletion of bacterial essential genes was utilized (a schematic description of the mutant generation procedure is shown in Supplementary Fig. 2) (Kruse et al. 2005;Yamaichi et al. 2007). Initially, the parAB m gene deletion vector (pUC-ΔparAB m :: blm, linearized by HindIII before using) and the exogenous plasmid encoding ParAB m (pMK-parAB m ) were co-transformed to the WT T. thermophilus cells (natural competent). After addition of DNA, the cells were exposed to appropriate conditions (70°C, 180 rpm/min) for 3 h to grow, also to absorb and recombine DNA. During this period, at first, the megaplasmid parAB m (including parS m on parB m ) was still present in the cells, the megaplasmid/ plasmid-expressed ParB m would bind the parS m site, and ParA m would also provide force for pushing/or pulling the ParB m -parS m complex. Since the copy number of pMK-parAB m could reach 4-10 copies/cell, in addition, it contained a strong promoter (slp) in front of the parAB m genes (de Grado et al. 1999), the ParAB m proteins from which could be overexpressed. Constant overexpression of ParAB m from the exogenous plasmid would automatically render the megaplasmid-encoded ParAB m proteins unessential. Consequently, the antibiotic gene blm from the gene deletion vector pUC-ΔparAB m ::blm would easily replace the megaplasmid parAB m genes. The transformation reaction was then spread on TB plate supplemented with kanamycin for the selection of cells containing pMK-parAB m . The colonies of the parAB m deletion mutant were distinguishable from other transformants which only contained pMK-parAB m and with the megaplasmid parAB m present, as they were apparently small and white (upon loss of the megaplasmid parAB m , a large portion of the genome sequences including the blm gene was sequentially lost, resulting this aberrant phenotypes, see results). Finally, to remove the pMK-parAB m vector from the parAB m deletion mutant cells, three white and small transformants were randomly picked and, respectively, inoculated in liquid TB medium without any selection and grown (70°C, 180 rpm/min) for 72 h, and then the cultures were probably diluted and plated on TB plate without any selection. Approximately 100 colonies of each transformant from these TB plates were picked and streaked on three types of plates, respectively, i.e. TB + kanamycin, TB + bleomycin, and TB. And streaks grown only on TB plate but not on the other two types This study of plates were defined as ΔparAB m mutants that have lost the pMK-parAB m vector. Three ΔparAB m streaks were then randomly selected for genotype confirmation by PCR (using two sets of primers binding the control gene and parA m or parB m genes, respectively) and/or genome sequencing. To confirm that it was the exogenously expressed ParAB m proteins but not the shuttle vector (i.e. pMK18) itself that would recover the complete parAB m deletion mutant, we also co-transformed pUC-ΔparAB m ::blm and pMK18 to the WT T. thermophilus cells followed by mutant selection using the similar method (a schematic illustration is shown in Supplementary Fig. 3). For generation of ΔvapBC64_65::kat, the pUC-ΔvapBC64_65::kat vector was linearized by HindIII and transformed to the WT T. thermophilus cells. The transformants were selected on TB plate supplemented with kanamycin, and the genotype was confirmed by PCR using primers flanking the deletion region. For generation of the double-gene deletion mutant ΔvapBC64_65/142_143, the pUC-ΔvapBC142_143::blm vector (linearized by HindIII) was transformed to the ΔvapBC64_65::kat mutant cells, followed by selecting on TB + kanamycin + bleomycin plate. The mutant genotype was also confirmed by PCR. The generation procedures of

Analyses of DNA content
PCR for determining the megaplasmid and chromosome sequence loss was performed based on standard protocol, except that two pairs of primers were used in one reaction, one pair of which was amplifying the target locus, and the other was amplifying the control gene region [i.e. TTC0825 (734 bp) or TTC1220 (890 bp) for detecting megaplasmid deletion and TTC0439 (857 bp) for detecting chromosome deletion]. β-Glucosidase (Bgl) activity assay was performed according to the method described in Li et al. (2015), for which three independently grown cell cultures were used. The quantitative PCR (qPCR) experiment for assessing the relative megaplasmid copy number was essentially performed as described previously (Breuert et al. 2006). Two megaplasmid loci (TTP0108 and TTP0161) were selected as the qPCR amplification targets, and one chromosomal locus near the oriC (TTC1609) was chosen as the internal reference amplification region (the primers used are listed in Supplementary Table 1). The amplicons were between 100 and 200 bp; three biological and three technical repeats were carried out for each strain. The relative quantification 2 −ΔΔCt method was used to calculate the relative (to WT) megaplasmid copy number of the ΔparAB m or ΔparAB m /ParAB m strain. For genome sequencing, genomic DNA of the ΔparAB m or ΔvapBC64_65/142_143 mutant strain was respectively isolated by Monarch Genomic DNA Purification Kit (New England Biolabs) based on the protocol provided by the manufacturer, and the purified gDNA (suitable for Next Generation Sequencing) was sequenced by BGI Technology (Shenzhen, China). The following genome sequence alignment method for determining genome deletion in the mutants was essentially based on previous studies (Bardaji et al. 2019;Ramijan et al. 2020). Specifically, the CLC Genomics Workbench software (8.5.1) was then used to analyze the Illumina reads. Raw Illumina reads of the ΔparAB m or ΔvapBC64_65/142_143 mutant were imported and mapped to the reference genome of T. thermophilus HB27 (NCBI reference sequence: GCA_000008125.1) via the "Map reads to reference" function in the NGS core tools. Pulsed-field gel electrophoresis (PFGE) was essentially performed as described by Herschleb et al. (2007). The WT and ΔparAB m strains were grown to exponentially growing phase, and the cell pellets were collected by centrifugation at 4,000 rpm/min for 20 min, respectively. The cell-holding agarose plugs were prepared using SeaKem Gold agarose (Lonza). The genomic DNA in the plugs was directly exposed for electrophoresis without digestion. One percent of SeaKem Gold agarose prepared in 0.5 × TBE was used for gel casting. The electrophoresis was carried out in a PFGE CHEF-DR III variable angle system (Bio-Rad), and the samples were run in the following conditions for 24 h: 120 degree included angle, 6 V/cm, 8-50 s switch time ramp, and 14°C.

Analyses of nucleoid morphology and segregation
For analyzing cell and nucleoid morphology, exponentially growing cells were collected, and the cell pellets were resuspended in 1 × PBS. DAPI (4′,6-diamidino-2-phenylindole-dihydrochloride) and Neuro-DiO were then used to stain the nucleoid and cell membrane, with final concentrations of 1 and 10 μg/ml, respectively. After incubating at RT for 30 min, the residual dyes were washed by 1 × PBS for three times, and the cells were resuspended in the same buffer. An appropriate amount (10 μl) of the cells was then mounted on glass slides followed by microscopic analysis. For visualizing the chromosome and megaplasmid segregations of the WT and ΔparAB m cells, the ParB-parS fluorescent tagging system was used (Hu et al. 2007;Badrinarayanan et al. 2015). Specifically, the allele exchange vector pUC::ΔparB c :: kat-parB c _sgfp or pUC::Δfdh::kat-parB m _sgfp was individually transformed into WT or ΔparAB m , respectively. The transformants were selected by kanamycin, and the correct recombinants were confirmed by PCR (with primers flanking the homology regions). In the correct recombinants (i.e. ΔparB c ::kat-parB c _sgfp, Δfdh:: kat-parB m _sgfp, ΔparAB m ΔparB c ::kat-parB c _sgfp, or ΔparAB m Δfdh:: kat-parB m _sgfp), the chromosomal parB c or the megaplasmid fdh gene was, respectively, replaced by kat-parB c _sgfp or kat-parB m _sgfp fusion genes ( Supplementary Fig. 1). ParB c -sGFP and ParB m -sGFP were able to bind their cognate parS c and parS m sites, which were, respectively, residing in the chromosomal replication origin-proximal region and the megaplasmid parB m gene itself. Thus, the subcellular localization and segregation pattern of the ParB c -sGFP and ParB m -sGFP fluorescent signals reflected those of the chromosome and megaplasmid, respectively. The four strains expressing ParB c -sGFP or ParB m -sGFP were then grown to exponential phase, and the cells were collected for fluorescence microscopy. All the fluorescence microscopic experiments were performed with a Nikon Ni-U microscope. The cell length, width, the sizes of the DAPI-stained nucleoids, and the ParB c -sGFP/ParB m -sGFP foci numbers were all analyzed by using ImageJ software (NIH, USA). The nucleoid amount and nucleoid morphology of the cells were reflected by the dimension and pattern of the DAPI-stained area, respectively (Schneider et al. 2007). The significance of differences of these parameters between two strains was tested by Student's t-test (P < 0.01 indicates significant difference, P > 0.05 indicates no difference).

TUNEL assay
Although TUNEL was initially used to analyze DNA fragmentation in eukaryotic cells, it was later also shown to be effective for detecting nucleoid damage and cell death in prokaryotic cells (Rohwer and Azam 2000). TUNEL is based on the principle that when DNAs are damaged, their free 3′-OH ends will be exposed, which can be enzymatically labeled with dUTP-fluorescein isothiocyanate by using terminal deoxynucleotidyl transferase (TdT). The nucleoids in the cells will thus be fluorescently labeled which can be easily detected using fluorescence microscopy or flow cytometry (Rohwer and Azam 2000). In our study, to assess DNA degradation and cell death of the ΔparAB m , ΔvapBC64_65/142_143, and Nal-treated cells, TUNEL assay was performed using the In Situ Cell Death Detection Kit, Fluorescein (Roche Applied Science), according to the protocol provided by the manufacturer. The fluorescein-dUTP labeled broken DNA of these cells was observed by fluorescence microscopy.

Generation of parAB m gene deletion mutant
The chromosomally encoded parABS locus normally resides in the very vicinity of the replication origin region of the chromosome. Likewise, the T. thermophilus megaplasmid parABS m is also positioned near the megaplasmid replication origin region. parABS m is composed of parA m (TTP0084) and parB m (TTP0083) genes transcribed in one operon and a parS m sequence which is residing in the parB m gene itself (5′-AAGGACGCGTCCTT-3′) (Li et al. 2015). T. thermophilus is polyploid; when a gene is essential, the deletion mutant will remain heterozygous which contains both the WT and the mutant allele at the same gene locus (Ohtani et al. 2010;Li et al. 2015;Li and Gao 2021). Using a standard gene exchange method (replaced by a thermostable bleomycin resistance gene cassette blm), we found that the parAB m complete deletion mutant was impossible to be obtained (i.e. remained heterozygous) (Li et al. 2015), unless an exogenous plasmid (pMK18-parAB m ) encoding the WT copy of ParAB m was co-transformed with the gene deletion vector (pUC-ΔparAB m ::blm) into the T. thermophilus cells (see Materials and methods). This result implied that both of the parA m and parB m genes are conditionally essential. Without antibiotic (kanamycin) selection, pMK18-parAB m could be easily lost from the host cells (the loss of the plasmid was confirmed by growth test of the mutant in TB medium supplemented with kanamycin), thereby resulting pure ΔparAB m mutant (see Materials and methods). PCR and genome sequencing results confirmed that the parAB m operon has been completely deleted in the ΔparAB m mutant ( Fig. 1; Fig. 3c and d). It is worthy to note that, in the above experiment, it was the exogenously expressed ParAB m proteins but not the shuttle vector (i.e. pMK18) itself that worked to recover the complete parAB m deletion mutant. As when we cotransformed pUC-ΔparAB m ::blm and pMK18 to the WT T. thermophilus cells followed by the similar mutant selection procedure ( Supplementary Fig. 3), no white transformant colonies were observed, which indicated that no large megaplasmid deletion had occurred (comparing with the phenotypes of the complete parAB m deletion mutant; see the following results). Further, PCR analysis confirmed that all the resulting parAB m deletion mutants were heterozygous (containing both WT-parAB m and ΔparAB m ::blm alleles) but not complete deletion ( Supplementary Fig. 4).

Growth rate, cell, and nucleoid morphology of the parAB m mutant
The ΔparAB m mutant exhibited small and white colonies which were strikingly different from the WT colonies (i.e. bigger and showing orange-yellow color) (Fig. 2a). The mutant cells also demonstrated severe growth defect when incubated in liquid medium (Fig. 2b), indicating that lack of parAB m affected the WT cell growth. Fluorescent microscopic analyses showed that the ΔparAB m cells were significantly (P < 0.01, 100 cells were counted for each strain) shorter and wider than the WT cells growing at a same growth phase ( Fig. 2c; Table 2). Some of the mutant cells even became twisted or extraordinarily enlarged (on average 1.38 ± 0.29 times wider than the WT cells, 30 enlarged cells were counted for ΔparAB m , also 30 cells were counted for WT) (Fig. 2c), indicating that deletion of parAB m has directly or indirectly rendered cell morphology change. According to the dimension and pattern of the DAPI-stained areas (Schneider et al. 2007), the nucleoid amount and nucleoid morphology of the cells were analyzed. We found that although with swelled cell shape, the nucleoids in the mutant cells were smaller compared with those in the WT cells (100 cells were measured for each strain, P < 0.01). Further, most nucleoids in the mutant cells remained unsegregated in which the replicated nucleoids (represented by two DAPI-stained regions) remained together (Fig. 2c, examples are pointed by triangles), and the frequency of anucleate cells was increased to 8.5% (only 0.4% was detected in the WT cells, Table 2). Moreover, in the extra-large or irregularly shaped cells, the nucleoids were dispersed and seemed like being damaged (Fig. 2c, pointed with arrows). Further complementation experiment showed that the above cell growth and nucleic defects in the mutant cells were irreversible, inasmuch as the aberrant phenotypes could not be complemented by exogenously expressed ParAB m (i.e. ΔparAB m /ParAB m , a strain in which parAB m was Fig. 1. Generation of the T. thermophilus ΔparAB m mutant. The mutant was confirmed by PCR using two sets of primers binding the control gene and parA m or parB m genes, respectively. The control gene used in these PCR reactions was TTC0213 residing in the chromosome. The primer pairs used for amplifying parA m , parB m , and TTC0213 were dparA m -F and dparA m -R, dparB m -F and dparB m -R, and TTC0213-F and TTC0213-R, respectively (listed in Supplementary  Table 1). 1, 2, and 3 represent three individual mutants. The 250 bp DNA ladder (Takara, Japan) was used for the agarose gel electrophoresis.
introduced in trans after the recovery of ΔparAB m ) (Fig. 2a-c). In fact, the ΔparAB m /ParAB m cells exhibited even more defective phenotypes with respect to growth and nucleoid content and morphology ( Fig. 2b and c; Table 2). Together, the above results suggested that in T. thermophilus, parAB m is important for the cell growth and nucleoid morphology maintenance.  The average cell length and width were calculated from 100 cells of each strain, respectively; for calculation of the frequency of anucleate cells, 500 WT cells, 200 ΔparAB m cells, and 250 ΔparAB m /ParAB m cells were counted, respectively. For qPCR, the locus near the chromosomal oriC (TTC1609) was chosen as an internal reference amplification region, and the relative quantification 2 −ΔΔCt method was used to calculate the relative megaplasmid copy number. The significance of difference of the cell length, cell width, and frequency of anucleate cells between ΔparAB m (or ΔparAB m /ParAB m ) and WT, and the significance of difference of the relative megaplasmid copy number between ΔparAB m and ΔparAB m /ParAB m were all tested by Student's t-test.

Genome content analyses of ΔparAB m
The wild-type T. thermophilus cells are orange-yellow due to the ability of producing carotenoid, and the enzyme catalyzing the last step of carotenoid synthesis is encoded by locus TTP0057 on the megaplasmid (Takano et al. 2011). Therefore, the white color of the ΔparAB m colony was an indicative of carotenoid synthesis deficiency (Fig. 2a), probably caused by loss of the TTP0057 locus.
To verify this speculation and further uncover the role of parAB m in genome maintenance, we measured genome content at multiple genomic sites of the mutant via various methods. Enzyme activity assay result showed that ΔparAB m had nearly no Bgl (encoded by TTP0042 on the megaplasmid) activity (Fig. 3a). The enzyme activity was even lower than the Δbgl mutant (Fig. 3a), indicating that there was not only a complete absence of TTP0042 but also of TTP0222 (encodes β-galactosidase, producing unspecific cleavage of the substrate during enzyme activity test) in the ΔparAB m mutant. Further, PCR and genome sequencing data (the sequencing data of ΔparAB m was deposited in the NCBI SRA database under BioProject accession code PRJNA909084) confirmed that in ΔparAB m , a large portion (∼160 kb) of the megaplasmid and a small portion (∼23 kb) of the chromosome were lost, respectively ( Fig. 3b and c). Notably, the same PCR result was observed in all of the three ΔparAB m mutants ( Fig. 3b;   Fig. 3. Analyses of the genome content of the ΔparAB m mutant from various aspects. a) Bgl assay results of WT, ΔparAB m , and Δbgl. Mean and SD from three biological repeats are shown. b) PCR confirming the megaplasmid and chromosomal sequence loss in ΔparAB m . In each reaction, two sets of primers were used, which could bind the control gene [TTC0825 (734 bp) or TTC1220 (890 bp) for detecting megaplasmid deletion and TTC0439 (857 bp) for detecting chromosome deletion] and the target amplification region, respectively. The genomic positions of the target amplicons were indicated in d). c) Whole genome sequencing confirming the megaplasmid and chromosomal sequence loss in ΔparAB m . Shown is the alignment of the Illumina reads to the reference genome of T. thermophilus HB27 (NCBI reference sequence: GCA_000008125.1). d) Schematic maps showing the residual megaplasmid and chromosomal regions (inner arcs) in ΔparAB m . In the megaplasmid map, boxes (i.e. surE, vapBC64_65, and "1-9") indicate the 11 genetic loci that have been tested by PCR. In the chromosome map, short lines (i.e. "chr1-5") indicate the five genetic loci that have been tested by PCR. The origin and terminus regions of the megaplasmid or of the chromosome are shown as oriC and ter, respectively. Supplementary Fig. 5), implying that the observed genome deletion was probably related to the parAB m deletion. The coordinates of the eliminated regions were roughly mapped as 1-91,449 and 165,789-232,605 on the megaplasmid and 564,199-587,420 on the chromosome (Fig. 3d). Further investigation of the deletion sites revealed that two 1031-bp direct repeating sequences were, respectively, found at the ends (i.

Chromosome and megaplasmid segregation in ΔparAB m
qPCR results showed that the average copy number of the truncated megaplasmid in the parAB m mutant was lower than that of the WT strain (Table 2). This result implied that the residual megaplasmid in the ΔparAB m cells was frequently lost during cell division. To visualize the subcellular localization and segregation patterns of the chromosome and residual megaplasmid, the ParB-parS fluorescent tagging system was utilized (see Materials and methods). ParB c -sGFP (from the chromosomally encoded parABS c system) and ParB m -sGFP would specifically bind their cognate cis-acting parS sequence forming fluorescent foci; thus, their localization pattern reflects that of the chromosome and megaplasmid, respectively. Nearly 95 and 54% WT cells contained detectable ParB c -sGFP/parS c and ParB m -sGFP/parS m foci, respectively ( Fig. 4a and b), and the average foci numbers were 6.84 ± 2.15 and 4.50 ± 1.64, respectively (100 cells from different microscopic fields were counted). Ninety-three percent of ΔparAB m cells contained ParB c -sGFP/parS c foci (the average number was 5.31 ± 1.53, 100 cells from different microscopic fields were counted), and the localization pattern of these foci was similar to that of the WT foci (Fig. 4a). However, extremely few ΔparAB m cells (lower than 10%) were found possessing the ParB m -sGFP/parS m signal (Fig. 4b), suggesting that a fraction of the mutant cells had lost the residual megaplasmid, which was in high agreement with the qPCR result. Further, the ParB m -sGFP/parS m fluorescent foci in the WT cells were regularly spaced and orderly arranged; on the contrary, in the ΔparAB m cells (especially irregularly shaped cells), they were aggregated and randomly distributed throughout the cell (Fig. 4b). Moreover, PFGE analysis also showed that doubled or multiplied megaplasmid sizes could be detected in ΔparAB m (Fig. 4c). Together, these results indicated that in the parAB m deletion cells, the chromosomes could be normally segregated; however, the replicated residual megaplasmid copies could not be properly separated or localized to the correct cellular position, thus forming dimer or multimer.

Viability of the ΔparAB m cells that have lost megaplasmid
The nucleoids of certain ΔparAB m cells especially abnormal cells were dispersed or broken (Fig. 2c), implying nucleoid fragmentation and cell death of these cells. The PFGE result also indicated that the nucleoids were severely degraded in the mutant (this phenomenon was reproducible in all of the three generated parAB m mutants) (Fig. 4c). The occurrence of the DNA broken event in the ΔparAB m cells was further analyzed by TUNEL assay. For TUNEL, Nal-(be able to cause nucleoid fragmentation) treated WT cells were used as positive controls. Around 67% of irregularly shaped ΔparAB m cells with nucleoids dispersed (100 aberrant cells from different microscopic fields were counted) showed positive TUNEL fluorescence signal similar to that of the Nal-treated cells (Fig. 5), indicating that DNA degradation had occurred in these cells, which could trigger cell death. The T. thermophilus megaplasmid encodes two sets of TA modules belonging to the VapBC family, i.e. TTP0064/TTP0065 and TTP0142/TTP0143. In WT cells, the cognate antitoxins can probably counteract the activity of toxins; however, when both TA encoding genes were lost, toxins would exert damage effect, since toxins are more stable than antitoxins. In ΔparAB m , the TTP0064/TTP0065 TA locus was lost (as shown in Fig. 3d), and in some cells, the residual megaplasmid was completely cured (Fig. 4b); thus, the second TA locus (TTP0142/TTP0143) was also lost in these cells. It is possible that the observed DNA fragmentation and cell death phenomenon in ΔparAB m were caused by the activity of the two toxins. To test this, we deleted both TA loci and performed TUNEL assay of the ΔvapBC64_65/142_143 mutant ( Supplementary Fig. 6). Although the average cell length of the vapBC mutant was much longer than that of ΔparAB m , the nucleoid morphology of the majority of the ΔvapBC64_65/142_143 cells was resembling to that of the ΔparAB m cells, which appeared as broken punctates (Fig. 5). Further, around 18% of the ΔvapBC64_65/142_143 cells (100 cells from different microscopic fields were counted) had similar TUNEL fluorescence as the aberrant ΔparAB m cells and Nal-treated cells. Overall, the above results suggested that deletion of parAB m could eventually trigger loss of the whole megaplasmid in certain cells, and thus loss of the megaplasmid-encoded TA systems, which was threatening to the viability of these T. thermophilus cells.

Discussion
Compared with those of the chromosomal and low-copy-number plasmid-encoded parABS systems, the characteristics of the megaplasmid-encoded parABS ortholog were barely understood. Further, whether the parABS system can function to partition genomes in polyploid bacteria was also unknown. T. thermophilus contains multiple copies of megaplasmid which encodes parABS; therefore, it is an ideal organism for solving the above queries.

The parABS m system works to guard the genome sequences
To delete the parAB m operon, we co-transformed a parAB m gene deletion vector (pUC-ΔparAB m ::blm) and an exogenous plasmid encoding ParAB m (pMK-parAB m ) to the WT T. thermophilus cells. During the transformation, the cells were grown for 3 h after addition of the plasmid DNAs. In the beginning, the megaplasmid parAB m (including parS m on parB m ) was present in the cells, and the megaplasmid/plasmid-expressed ParB m would bind the parS m site; thus, the whole ParA m -ParB m -parS m system would work properly. However, the ParA m and ParB m proteins were constantly overexpressed from pMK-parAB m , thus after a period of time, they would eventually render the megaplasmid-encoded ParAB m unessential. Sequentially, the blm antibiotic marker from pUC-ΔparAB m ::blm would easily replace the parAB m operon (as illustrated in Supplementary Fig. 2), thereby resulting the complete parAB m deletion.
Through deletion of the parAB m operon followed by phenotype observation, we found that the T. thermophilus parABS m system was not only important for the megaplasmid segregation but surprisingly also played a role in genome integrity maintenance. PCR, genome sequencing, and PFGE experiments showed that the ΔparAB m mutant had lost ∼160 kb (including the oriC region and the megaplasmid replication initiator gene repA) of the megaplasmid and ∼23 kb of the chromosome sequences, respectively ( Fig. 3b-d; Fig. 4c). Notably, there are two long direct repeats (1031 bp) at the two ends of the lost megaplasmid region and four ISs adjacent to the chromosomal deletion start. Therefore, it is highly possible that the megaplasmid deletion was caused by homologous recombination between the two repeats, and the chromosomal deletion might be mediated by transposition of these ISs. Regarding how the residual megaplasmid replicated after loss of the oriC site, we hypothesized that it would either integrate into the chromosome or maintain auto-replicating by itself. According to the PFGE data (Fig. 4c), we found that the residual megaplasmid was not integrated into the chromosome (if it was, there should be no megaplasmid bands upon PFGE analysis). Thus, it is possible that the megaplasmid would choose another site as the oriC site, and the replication initiation protein might be from the chromosome. Indeed, unlike that of the monoploid bacteria, the GC skew (software for detecting bacterial replication origin and terminus regions) (Grigoriev 1998) result of the T. thermophilus megaplasmid is asymmetrical and contains many high-G/high-C shift points (Supplementary Fig. 7). A similar GC skew pattern was also observed in the genomes of other polyploid bacteria, such as Synechococcus elongatus PCC 7942, Synechocystis sp. PCC 6803, Gloeobacter violaceus PCC 7421, and D. radiodurans (Watanabe 2020). This indicated that the megaplasmid of T. thermophilus contains multiple oriC-like regions, which may support autonomous replication after the most qualified oriC was deleted. A same result has been uncovered in Anabaena sp. PCC 7120 (Watanabe 2020). Further, as demonstrated by Fig. 2, the phenotypic defects (e.g. the cell growth, morphological and nucleic The chromosomal and megaplasmid signals were represented by ParB c -sGFP and ParB m -sGFP fluorescent foci, respectively. Arrows are pointing the apparent fluorescent foci. Bars, 2 μm. c) PFGE analysis of the undigested genomic DNA isolated from the WT or ΔparAB m cells. The arrowheads at the top of the PFGE gel image indicate the chromosome bands, and the other arrowheads indicate the megaplasmid bands. 1, 2, and 3 indicate the PFGE result of the three ΔparAB m mutants. In mutant 2, the genomic DNA was completely degraded, and although with genome degradation, mutants 1 and 3 showed faint ladder-like megaplasmid bands. Note: in both WT and mutant, the circular megaplasmids ran slower than the corresponding linear marker molecule during PFGE. defects) of the ΔparAB m mutant could not be complemented by the exogenously expressed ParAB m . This result was as expected. Deletion of the megaplasmid parAB m in ΔparAB m meant loss of the megaplasmid parS m site (residing in parB m ); therefore, there was no cis-reacting site of the ParB m protein during the complementation experiment. Additionally, the mutant cell has already lost a large portion of the genome sequence. Nevertheless, it is worthy to note that this result does not contradict the principle that we used to generate the ΔparAB m mutant, due to the fact that during the mutant generation process, parS m was present at first and was lost afterward (see Materials and methods).
Moreover, we found that when parAB m was deleted, the replicated megaplasmid could not be transmitted to the correct cellular position or segregated properly (Fig. 4b and c), thus forming intertwined structure in which the repeating regions might directly contact each other, whereby recombinases might easily recognize the complex and execute sequence excision. This interpretation is reasonable, as although plasmids (especially multicopy number plasmids) can provide certain benefits for bacteria, they are meanwhile metabolic burdens to the cell (Million-Weaver and Camps 2014). To increase their stability, plasmids usually carry dedicated genetic determinants, namely, plasmid partitioning determinants (e.g. the parABS system) (Bouet and Funnell 2019;Baxter et al. 2020), post-segregational killing systems (e.g. the TA system) (Hernández-Arriaga et al. 2014), and multimer resolution systems (some recombinases, e.g. XerCD) (Colloms 2013;Cameranesi et al. 2018). In addition to these three plasmid stability elements, ISs are also suggested to be important for the architecture of plasmids (or even chromosomes) (Bardaji et al. 2011). When bacteria were subjected to growth stress or when the plasmid stability determinants were missing, genetic changes of the plasmids could possibly occur (Bardaji et al. 2019;Ramijan et al. 2020). Loss or rearrangement of plasmid sequences was also observed in some other bacteria. For example, genomes of actinobacteria were shown to readily undergo rearrangements or deletions, and the size of the deletion genome could be more than 1 Mb (Hoff et al. 2018). Ramijan et al. (2020) provided evidence showing that protoplast formation and regeneration (i.e. a stressful process) in the filamentous actinobacterium Kitasatospora viridifaciens could trigger severe megaplasmid sequence loss and genome rearrangement. In Pseudomonas syringae, inactivation of the TA systems from its virulence plasmid pPsv48C could also lead to high-frequent deletions of the plasmid sequence (Bardaji et al. 2019). In the case of K. viridifaciens, the genome variation was suggested to be the consequence of transpositions of IS upon environmental stress (Ramijan et al. 2020). In P. syringae, that was triggered by recombinations between two copies of miniature inverted-repeat transposable elements (MITEs) or one-ended transposition of IS801 once the TA system was lost (Bardaji et al. 2019). The similar event has also been revealed in Shigella sonnei, in which the absence of plasmid TA systems (i.e. ccdAB and gmvAT) was found to contribute to pINV loss, and organization of ISs on pINV could determine plasmid plasticity (Martyn et al. 2022).
In our situation, it seemed that the large genome sequence deletion in ΔparAB m was fundamentally caused by the deletion of the parAB m genes; since the genome deletion at the same region was not a random event, it was found in all of the three ΔparAB m mutants ( Fig. 3b; Supplementary Fig. 5). Unlike the situations in the bacteria exemplified above, it was not caused by the loss of the TA system, on account of that genome deletion at the same region as that of ΔparAB m was not found in the ΔvapBC64_65/142_143 mutant (the sequencing data was deposited in the NCBI SRA database under accession code PRJNA909084) ( Supplementary Figs. 8 and 9). Further, it seemed that the large genome deletion did also not happen before the parAB m deletion, since there were no important or potential genes in the deleted genomic region of ΔparAB m that had to be lost to allow for the survival of the parAB m mutant (Supplementary Table 2). In fact, without the initial generation of the ΔparAB m ::blm genotype, a spontaneous recombination of the two repeating sequences (1031 bp) of the megaplasmid would not possibly happen. The two repeats are genomically and also spatially separated (Henne et al. 2004), and in the numerous previous studies from our and other groups (Ohtani et al. 2010(Ohtani et al. , 2016Angelov et al. 2013;Li et al. 2015;Li 2019b), no spontaneous recombination event of the two repeating elements could be observed in the WT strain or in any other megaplasmid gene deletion mutants (typically, also not observed in the vapBC deletion mutant in this study). Apparently, when parAB m was deleted, the megaplasmid could not be segregated properly ( Fig. 4a and b), thus became entangled (Fig. 4b and c); at this moment, the repeating elements might directly interact each other, and thereby the recombinases would recognize the complex and perform the excision. Taken together, it seems logical that the large megaplasmid sequence (including the later inserted blm marker) in ΔparAB m was lost by homologous recombination, after the parAB m genes were deleted.
Regarding the mechanisms for how loss of TA could lead to deletion of genome sequences, it has been previously shown that in Shigella spp., the VapBC family TA system could stabilize local sequences by preventing IS-mediated deletions (Pilla et al. 2017). Therefore, we hypothesized that the megaplasmid parAB m might work to guard the genome sequence in a similar mechanism. Intriguingly, to our knowledge, before this study, none of the plasmid or chromosome deletion event was found to be directly attributed to parAB disruption, although it is theoretically highly possible. The reason could be that for the low-copy-number plasmid, the parABS system is primarily important, disruption of which would cause partitioning defect, thus loss of the whole plasmid but not a portion of sequence (Million-Weaver and Camps 2014; Baxter and Funnell 2014;Bouet and Funnell 2019). For the chromosomes, the parABS system is not significant for its maintenance, as redundant mechanisms (e.g. SMC, TopoIV, and Ftsk-XerCD systems) can act to segregate the chromosome (Badrinarayanan et al. 2015). Numerous studies have shown that deletion of parABS had negligible effect for the chromosome segregation (Lewis et al. 2002;Bartosik et al. 2004;Fogel and Waldor 2006;Lasocki et al. 2007;Li et al. 2015). To some extent, the T. thermophilus megaplasmid can be defined as a second chromosome, as it carries certain essential genes for cell viability (e.g. TTP0161 and TTP0162 encoding the α and β subunit of class I ribonucleotide reductase which is a key enzyme for deoxyribonucleotide synthesis) (Henne et al. 2004;Ohtani et al. 2016). However, on the other hand, it is still a metabolic burden for the whole cell due to its big genome size. Therefore, it is conceivable that the parABS m system might have been evolved to protect the essential sequences but not the whole megaplasmid.

The parABS m system can determine cell viability
In addition to causing genome sequence deletion, the TA systems can also determine cell viability (Kamruzzaman et al. 2021). The cells with TA loci lost might be aberrantly shaped, undivided, or have fragmented nucleoid. For example, when the megaplasmidderived chromosome (ChrII) of V. cholerae was eliminated upon parAB2 deletion, the 13 TA loci on which were also eliminated, such that the toxins started to cause detrimental cytologic changes. Consequently, the cells became hypertrophic, undivided, and had condensed nucleoid, which would die eventually (Yamaichi et al. 2007). A number of TA deletion mutants have been shown to have distinct or deleterious phenotypes (Arcus et al. 2011). For example, ΔfitAB mutant of Neisseria gonorrhoeae showed a faster traversing rate through epithelial cells than the wild type (Wilbur et al. 2005); a hipBA deletion mutant of E. coli showed a reduction in antibiotic tolerance (Schumacher et al. 2009); and deletion of vapBC3 and vapBC4 could impair the infection ability of Mycobacterium tuberculosis in animal models (Agarwal et al. 2018). It has been suggested that besides its ability of degrading RNA or glutamyl-tRNA synthetase (Christensen and Gerdes 2003;Germain et al. 2013), the TA systems (e.g. VapBC) can also regulate the intracellular metabolisms, such as to regulate the levels of branched-chain amino acids which are proposed to play essential roles in monitoring the nutritional supply and physiological state of the cell (Frampton et al. 2012). For instance, deletions of the total three TA loci (VapBC, MazEF, and Phd/Doc, ΔTAtriple) in Mycobacterium smegmatis could result in severe survival defect. Further experiments showed that there was a significant difference in the levels of branched-chain amino acids between the wild-type and the ΔTAtriple mutant, which would cause DNA damage and cell death. We thus hypothesized that the megaplasmid-encoded TA loci of T. thermophilus may also play roles in the intracellular metabolism regulations. Under this circumstance, when both the toxin and antitoxin were absent, the ongoing phenotypes of the ΔvapBC64_65/142_143 mutant should be possibly observed (Fig. 5). In the future work, it is worthy to test the metabolic changes of this mutant. Nevertheless, in our study, deletion of parAB m led to the excision of ∼160 kb megaplasmid sequences, which included one TA locus (vapBC64_65). The smaller megaplasmid would not be segregated correctly; thereby, certain daughter cells eventually lost the whole megaplasmid. These cells became swell or twisted and had dispersed nucleoids (Fig. 2c). TUNEL assay showed that they were actually dead cells (Fig. 5). The nucleoid morphology of these cells was similar to that of the ΔvapBC64_65/142_143 mutant (Fig. 5). Therefore, we suggested that the combined actions of the megaplasmid-encoded toxins likely to some aspects contributed to the observed phenotype of the dead ΔparAB m cells. This indicated that through guarding the genome sequences (including the TA loci), the T. thermophilus parABS m system can also determine cell viability.

The parABS m system mediates the megaplasmid segregation
Besides its role in genome integrity maintenance, parABS m was also involved in megaplasmid localization and segregation. Fluorescent tracking of the ΔparAB m genomes demonstrated that the truncated megaplasmid was frequently mislocalized or even completely disappeared (Fig. 4b). Also, the PFGE analysis showed that the smaller megaplasmid was likely to form multimers (Fig. 4c). These results indicated that although the megaplasmid present with multicopy in one T. thermophilus cell, it still requires active segregation machineries. High-copy-number plasmids (>20) normally lack active partition systems and are believed to be segregated randomly during cell division, since the frequency of daughter cells receiving no plasmid is practically low (Million-Weaver and Camps 2014). Our data suggested that in T. thermophilus, the copy number of the megaplasmid was not sufficient for random segregation event to occur. Additionally, the megaplasmid copies were regularly arranged and equally spaced in the WT cells (Fig. 4b), which also implied that they could be stringently segregated. Further, it seemed that except for parABS m , there was no redundancy of the mechanisms that account for the megaplasmid segregation in T. thermophilus. This was in a striking contrast to the segregation of chromosomes but analogous to that of the low-copy-number plasmids. For most of the bacterial chromosomes, deletions of the parAB genes, their segregations would proceed normally (Lewis et al. 2002;Bartosik et al. 2004;Fogel and Waldor 2006;Lasocki et al. 2007;Li et al. 2015). For instance, deletion of the parA1 gene in V. cholerae did not cause evident chromosome segregation defect (Fogel and Waldor 2006), and in the B. subtilis parB (spo0J) deletion mutant, most cells still exhibited regular chromosome segregations (Lee and Grossman 2006). In fact, it is now widely accepted that the chromosomal parABS system mainly functions to motivate the chromosomal origin region but not bulk DNA segregation (Jalal and Le 2020). The segregation of the bulk chromosomal DNA is accomplished by SMC complex (structural maintenance of chromosomes) and in the meanwhile disentangled by topoisomerase (e.g. TopoIV) (Badrinarayanan et al. 2015). Moreover, the T. thermophilus megaplasmid parABS m system is phylogenetically more close to the plasmid-encoded parABS (Li et al. 2015). Together, our results indicated that unlike the chromosomal parABS, the megaplasmid parABS system was important for mediating the megaplasmid segregation, even under the context that the replicon was present with multicopy in one cell.

Conclusions
Through providing a plasmid-born copy of parAB m , we generated the T. thermophilus parAB m gene deletion mutant. The genome content analysis results showed that the ΔparAB m mutant had lost ∼160 kb of the megaplasmid and ∼23 kb of the chromosomal sequences, respectively. The truncated megaplasmid would not be segregated correctly; thus, certain daughter cells eventually lost the entire megaplasmid and became twisted or enlarged containing dispersed nucleoids. We further found that when the megaplasmid was lost, the megaplasmid-encoded TA systems (VapBC64_65 and VapBC142_143) were also eliminated; thereby, the toxins would exert detrimental effects, such as to fragment DNA. Since the genome sequence deletion at the same region was not detected in the ΔvapBC64_65/142_143 mutant, we concluded that the megaplasmid and chromosome excision event in ΔparAB m was related to the deletion of parAB m . Thus, it seems that ParAB m can act to prevent recombinase or transposasemediated genome instability. To sum up, our results suggested that in T. thermophilus, the megaplasmid parABS system plays essential roles in both megaplasmid segregation and genome integrity maintenance. We initially characterized the functions of a megaplasmid-encoded parABS system in a polyploid thermophilic bacterium and revealed a new factor (i.e. parABS) for determining bacterial genome integrity.

Data availability
Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and supplemental material. Supplementary file contains all supplemental tables and figures. The sequencing data was deposited in the NCBI SRA database under accession code PRJNA909084.
Supplemental material available at G3 online.