The ParA’s function is realized by two separate proteins in the partitioning system of Myxococcus plasmid pMF1

The par operon in the sole myxobacterial plasmid pMF1 includes a function-unknown parC gene in front of the classical parA and parB genes. Removal of parC severely reduced plasmid stability, but ex-situ compensations of parC did not restore the par system function. Individual expression of parA formed insoluble proteins, while co-expression of parC before parA produced a soluble ParC-ParA heterodimer. ParA alone had no ATPase activity and no polymerization, while ParC addition aided ParA to restore the activities. Fusing ParC and ParA in different ways all produced soluble proteins and some restored ATPase activity or increased plasmid stability. Protein interaction model analysis and experiments revealed that ParC structurally mimics the N-terminal of Ia-type SopA (ParA), endowing the Myxococcus ParA protein to play functions by shifting of ParC between two sites on ParA surface. The present results highlight that ParC functions as a part of ParA to support its soluble expression and function, and the separation of ParC and ParA into two proteins in structure enables the ParC ‘fragment’ to shift in a larger range around ParA to function during partitioning. Author summary Our work on ParC here provides a new example for the evolution of multi-domain protein. ParC and ParA are two proteins, but their expression and function act as a whole, which proposes a new regulatory model for bacterial par system, and also provides research ideas and materials for the study of functional coordination and evolution of ParA domains in the future.


Introduction 25
Partitioning (par) system exists widely in bacteria and archaea, participating in the 26 isolation and allocation of chromosome and plasmid into daughter cells during the cell 27 division [1-3]. A par system typically consists of three components: an NTPase (usually 28 named ParA), a DNA-binding protein ParB, and one or more cis-acting sequences parS. 29 ParB binds specifically onto the parS sequences and assembles into a high-order 30 partitioning complex, on which the ParA proteins are further added by binding to the 31 ParB proteins, and act as an energy machinery by hydrolyzing NTP to move the ParB-32 parS complex apart. The two par genes are usually found in the same operon, with the 33 parS elements located within or adjacent to this operon [4][5][6][7]. 34 According to the ParA sequence similarity, bacterial par systems are divided into 35 three main types [7,8]. Type I par system, which has a deviant Walker-box ATPase, is 36 found in most bacterial and archaeal chromosomes and plasmids, and is probably the 37 most ubiquitous type of active partitioning system in nature. In addition to ATPase 38 activity, Type I ParA has the capability of auto-aggregating into dimers and then more 39 complex polymers [9][10][11]. Type I par system can be further divided into two 40 subfamilies, Ia and Ib [12][13][14]. Type Ia ParA, such as the ParA proteins encoded in P1 41 and F plasmids, contains a Walker-box region and an N-terminal winged helix-turn-42 helix (HTH) domain, which plays an auto-regulation role for the transcription of parAB 43 operon [11][12][13][14]. Type Ib ParA is some small proteins (192-308aa), containing only the The pMF1 plasmid was discovered in Myxococcus fulvus 124B02, and is the first 49 and as yet the sole endogenous plasmid that is able to replicate autonomously in 50 myxobacterial cells [17]. The plasmid is 18,634 bp in size, containing 23 genes 51 (pMF1.1-pMF1.23). Many of the plasmid genes have their homologs in different 52 myxobacterial genomes, while the others have not yet found their homologs in the 53 GenBank database (S1 Table), which suggest that pMF1 has a long standing co-54 adaption within myxobacteria [18,19]. We previously determined that the plasmid  [17,20,21]. pMF1 employed at least two 58 strategies for its stable inheritance in Myxococcus; one is the partitioning system 59 encoded by the pMF1. , while the second is a post-segregational 60 killing system of nuclease toxin and immune protein encoded by the pMF1.20 and 61 pMF1.19 gene pair [23]. In the partitioning system, pMF1.22 is predicted a parA gene, 62 while pMF1.23 encodes a DNA-binding protein (parB), which is able to specifically 63 bind to the parS iteron DNA sequences [22]. Based on the ParA protein sequence 64 similarity and the autogenously repressing regulation of ParB on the expression of the 65 par genes, the pMF1 partitioning system was suggested to belong to the Ib system [22]. 66 Specifically, the pMF1 par system has a third function-unknown gene, pMF1.21 67 (named parC), in front of parA and parB. The three genes are in the same transcriptional 68 unit, and parC is able to enhance the DNA-binding activity of ParB and regulate gene 69 expression of the par loci [18,22,23].

70
In this study, we revealed that pMF1.21 is functionally essential for the partitioning 71 system in Myxococcus cells. We determined that parC located in front of parA is critical ParC is a small acidic protein with 87 amino acid residues and the parC gene 82 locates in front of parA in the pMF1 partitioning operon. Quantitative PCR 83 amplification indicated that, in the pMF1-harboring M. fulvus 124B02 strain, as well as 84 the par system-containing shuttle plasmid pZJY4111, parC transcribed slightly higher 85 than parA, but lower than parB, which were comparatively at similar transcriptional 86 levels as the replication genes (S1B Fig). To determine whether parC is required for 87 the partitioning function, we made an in-frame deletion of the parC gene in pZJY4111 88 (Fig 1A), which contains the ori region and par system of pMF1 [22]. Compared to 89 approximately 60% retention of pZJY4111 in DZ1 strain after 144 h of incubation in 90 the absence of antibiotic selection, the retention of pZJY4111ΔparC was dramatically 91 declined to 20% after 48 h, and to almost zero after 96 h of incubation (Fig 1B). The 92 above results indicated that the parC gene plays a crucial role in the partitioning system 93 for the plasmid maintenance in Myxococcus cells. 94 We experimentally compensated the deletion by ex-situ inserting the parC gene 95 with its own promoter in pZJY4111∆parC, producing pZJY4111∆parC::parC (plasmid 96 I in Fig 1C). We performed the same insertion of the parC gene with its own promoter 6 97 in the pZJY4111 plasmid, producing pZJY4111::parC (plasmid II). There was a 203-98 bp interval space sequence between the deficient par operon and the parC 99 compensation gene. However, the compensation did not restore the plasmid stability 100 phenotype, and plasmid II exhibited similar inheritance ability as pZJY4111 in M. 101 xanthus DZ1 (Fig 1D). To verify the location effects of parC on plasmid stability, we 102 placed parC at different places in the deficient par operon of pZJY4111ΔparC, either 103 after parA, forming the parA-parC-parB cascade (plasmid III); after parB, forming the 104 parA-parB-parC cascade (plasmid IV); or in front of parA and behind the parC 105 residuals (plasmid V) (Fig 1C). The plasmid stability assays indicated that plasmid III 106 or IV did not restore the plasmid maintenance; whereas plasmid V completely restored 107 the plasmid stability with almost the same curve as the wild type plasmid pZJY4111 108 (Fig 1D). 109 Notably, in the original parCAB operon, the coding sequences of parC and parA 110 overlaps four bases (ATGA), and the inserted parC gene in plasmid V also overlapped 111 the four bases with the parA gene. To investigate whether the overlap had an effect on 112 partitioning, we added the parC gene in front of parA by replacing ATGA with 113 TGAATG (plasmid VI; Fig 1C). Stability of plasmid VI reduced to some extent as 114 compared with pZJY4111or in-situ compensating plasmid V, but significantly higher 115 than those plasmids with ectopically inserted parC (Fig 1D). The transcriptions of parC 116 in M. xanthus DZ1 were confirmed by quantitative PCR amplification (Fig 1E). The  ParA heterodimer to play functions 123 To investigate the interactions between parA and parC in expression, we 124 constructed the two genes into the pET15b expression vector in the same arrangement 125 as they were in the compensation plasmids, and expressed them in E. coli, respectively.

126
When parC and parA overlapped four bases (as in plasmid V), an obvious soluble band 127 with approximately 35 kDa in size was induced (Fig 1F). For the adjacent parC and 128 parA genes without overlapped bases (as in plasmid VI), the same soluble protein band 129 was observed, but with a small amount of insoluble 24 kDa proteins (in ParA size).

130
However, when parC was placed behind parA (as in plasmid III) or the two genes were 131 transcribed separately (as in plasmid I), no obviously soluble desired band appeared.

132
The results indicated that when parC was in front of parA, the expressed ParC and ParA 133 proteins appeared to combine immediately into soluble ParC-ParA heterodimer; 134 otherwise, they were in insoluble forms, probably as monomers or oligomers.

135
The 35 kDa band in Figure 1F was cut, digested with trypsin and identified by mass  spectrometry as the polymer of ParC and ParA and their ratio is 1:1, appeared and 172 increased (Fig 1F). Thus, the presence of ParC did promote the aggregation of ParA 173 protein into ParC-ParA polymers. Further ATPase experiments found that ParA 174 proteins alone had almost no ATPase activity, but retrieved ATPase activity when 175 mixed with ParC (Fig 1I). The ATPase activity reached the highest at the ParC-ParA 176 ratio of 1:1, and more ParC had no increase to the activity.

178
To investigate the potential interaction patterns between ParC and ParA, we 179 constructed their structure models using the i-tasser (Iterative thread assembly 180 optimization) method [27]. The Myxococcus ParA protein was structurally composed 181 of five beta folds in center and outside around with eight alpha helixes (Fig 2A). forming an active nucleotide "sandwich" dimer with ATP in the middle [24][25][26]. Region 196 II is on the outside of H7/8, partly overlapped with the binding interface 2 of ParF. 197 Region I locates in the H5-β5-H6 region, covering an irregular sequence in front of H5. 198 Region I has the highest binding energy, far exceeding the other two sites, leading to 199 the preference of binding upon region I to form ParA homodimer (ParA-(I)-ParA).

200
However, homo-dimerization on region I reduces the binding ability of the other two 201 aggregation sites, especially in region III, the binding energy of which turns to -9.13 202 kJ/mol in the homo-dimer (Fig 2C), hindering further active dimerization at region III, 36.34 to -7.13 at C1 site, and increases significantly from -14.91 to -63.87 at C2 site, 220 respectively (Fig 2E). As a result, ParC on ParA dimer in the ParC-A-A-C tetramer 221 moves from C1 to C2 site, and the site of the tetramer for further aggregation is changed 222 from region II to region I (Fig 2F). ParCA exhibited ATPase activity, which, however, was much lower than that of ParC-237 ParA heterodimer (Fig 3A). Because the binding energies of the N-terminal binding  We mutated the C2 region by substituting D157V158 to A157A158. As 258 predicted, the binding ability of the ParA mutant to ParC at C2 will decrease, and a 259 third ParC binding site will appear at the H11 region. With the dimerization of ParA 260 with C2 site mutation, ParC did not bind to C2 site, but to H11 helix (S6 Fig). 261 Experimental assays showed that the point mutations in C2 region did not affect the 262 soluble co-expression of ParC / ParA protein, but significantly reduced the retention of 263 plasmid X (Fig 3C), which further confirmed that the binding of ParC and ParA at C2 264 region is important for the plasmid partitioning. shows that when ParA was dimerized, whether ParA bound ATP or ADP, ParC is only 296 able to bind to C2 site. When ParA was in the monomer form, ParA-ADP binds to ParC 297 at C1 site, but ParA-ATP binds to ParC at both C1 or C2 sites and preferentially at C2 ParC moves back to C1 position (Fig 4). 307 Gene fusion/fission is a major contributor to evolution of multi-domain proteins 308 [42]. Here, ParC and ParA function as a whole, but separate into two proteins, which 309 provides more interaction plasticity and thus is more efficient than one fusion protein 310 in plasmid partitioning. Their interfaces can be changed in a larger range, which 311 contributes to better protein-protein interactions and increases the protein activity and 312 precision adjustability. This is a novel regulation method of bacterial par system, 313 different from any known Ia or Ib ParA proteins and will be a good model for studying   The Escherichia coli strains were cultivated in LB medium at 37C. When required, 40 321 μg/ml of kanamycin and 100 µg/ml ampicillin were added into the medium.  Protein-protein docking analysis 378 We further predicted the interaction pattern between the ParA and ParC proteins.

379
The modeled structures were submitted as targets to PRISM protein-protein docking                          , Three genes, each having 594 many homologs in different genomes only of myxobacteria. , Three genes, each 595 having a single homolog in many different genomes, but the genes with the highest 596 similarity come from myxobacteria. , Ten genes, having no homology in the survey.

597
Details are referred to S1