Structural analyses of the bacterial primosomal protein DnaB reveal that it is a tetramer and forms a complex with a primosomal re-initiation protein

The DnaB primosomal protein from Gram-positive bacteria plays a key role in DNA replication and restart as a loader protein for the recruitment of replisome cascade proteins. Previous investigations have established that DnaB is composed of an N-terminal domain, a middle domain, and a C-terminal domain. However, structural evidence for how DnaB functions at the atomic level is lacking. Here, we report the crystal structure of DnaB, encompassing the N-terminal and middle domains (residues 1–300), from Geobacillus stearothermophilus (GstDnaB1–300) at 2.8 Å resolution. Our structure revealed that GstDnaB1–300 forms a tetramer with two basket-like architectures, a finding consistent with those from solution studies using analytical ultracentrifugation. Furthermore, our results from both GST pulldown assays and analytical ultracentrifugation show that GstDnaB1–300 is sufficient to form a complex with PriA, the primosomal reinitiation protein. Moreover, with the aid of small angle X-ray scattering experiments, we also determined the structural envelope of full-length DnaB (GstDnaBFL) in solution. These small angle X-ray scattering studies indicated that GstDnaBFL has an elongated conformation and that the protruding density envelopes originating from GstDnaB1–300 could completely accommodate the GstDnaB C-terminal domain (residues 301–461). Taken together with biochemical assays, our results suggest that GstDnaB uses different domains to distinguish the PriA interaction and single-stranded DNA binding. These findings can further extend our understanding of primosomal assembly in replication restart.

Accurate and timely DNA replication is essential for cell proliferation in all organisms, which relies on a number of interprotein interactions (1)(2)(3). Sequential assembly of the primosome complex is one the most important processes that drives DNA replication in all bacterial species. In bacteria, DNA replication is initiated when DnaA recognizes and binds to the origin of replication (oriC) (4,5) where all the DNA unwinding machinery is assembled. Subsequently, the hexameric helicaseloader complex is recruited followed by joining of the primase DnaG to complete primosome and replisome assembly (6 -8).
During the elongation phase of DNA replication, replication forks are frequently challenged by mutations, nicks, base modifications, and other factors that could result in fork arrest or collapse (3,9,10). Once the DNA replication process is stalled, it undergoes a reinitiation process, and the primosome re-initiation protein PriA is recruited at the respective sites (11). The proteins DnaA, PriA, and the replicative helicase are well-conserved in both Gram-negative bacteria (such as Escherichia coli) and Gram-positive bacteria (such as Bacillus subtilis) (12)(13)(14). However, DnaA and PriA in E. coli and B. subtilis differ in their recruitment of subsequent loader proteins before recruitment of the replicative helicase. During the normal replication process in Gram-negative E. coli, DnaA recognizes oriC sites and recruits DnaB helicase with the help of the helicase loader protein DnaC and promotes the association of the primase DnaG (2,(15)(16)(17)(18). In instances of abnormal DNA repair and restart, PriA recognizes the stalled forks and sequentially recruits three other primosomal reinitiation proteins (PriB, PriC, and DnaT) and a loader protein DnaC to trigger re-loading of hexameric replicative helicase as a helicase-loader complex DnaB-DnaC to restart the replication process (13). These two distinct primosome reinitiation processes share the same helicase (DnaB) and primase (DnaG) in the replisome system. However, in Gram-positive bacteria homologues of PriB, PriC, and DnaT are absent. Instead, only the primosomal loader proteins DnaB and DnaD are found, which are mediated by both DnaA and PriA to initiate and restart the replication process at oriC and stalled forks, respectively (12,19,20). Assembly of DnaB and DnaD preloader proteins at replication sites promotes loading of the helicase DnaC (the replicative helicase homologous to E. coli DnaB) in complex with the loader protein DnaI (the helicase loader homologous to E. coli DnaC) in an ATP-dependent manner (6,21) to unwind the parental DNA template.
Previous studies suggested that Gram-positive bacterial loader protein DnaB is a multifunctional protein. Its interaction with DnaD on single-stranded DNA-binding protein-coated DNA has been demonstrated (22), and its cooperative role with DnaI in acting as a "co-loader" for the recruitment of DnaC helicase has also been shown (23). Furthermore, the cell mem-brane attachment properties of DnaB have been described (24 -27), and these have been shown to regulate the recruitment of DnaD to the membrane-attached oriC region of bacterial DNA in a supercoiled state (28,29). This process is thought to be essential for regulating DNA replication and the re-initiation process (30). Although previous low resolution structural studies using atomic force microscopy (AFM) 2 and electron microscopy (EM) showed that DnaB from B. subtilis (BsuDnaB) assembles as a non-globular tetramer (in a square or spiral shape) (7,29,31), the functional significance of adopting this conformational state and its structural details remain unclear. Due to the lack of full-length structures of DnaB or its homologues, our understanding of the mechanics involved in effective DNA processing during the replication initiation and restart processes are limited. This shortcoming warrants further structural and functional studies on DnaB at higher resolution to address its functional involvement in the association of proteins such as PriA and DnaD and to aid our understanding of the regulation of bacterial replisome priming and the process of DNA replication and restart.
In this study we determined the crystal structure of the truncated form (residues 1-300) of DnaB from Geobacillus stearothermophilus (GstDnaB 1-300 ) at 2.8 Å resolution. The GstDnaB 1-300 crystal structure describes a unique tetrameric organization comprising two dimers arranged in a domainswap conformation. The results from our GST pulldown assays and analytical ultracentrifugation (AUC) suggest that GstDnaB physically interacts with GstPriA. Furthermore, small angle X-ray scattering analyses (SAXS) of both GstDnaB FL and GstDnaB 1-300 corroborate the consistency of the structural conformation in solution. Our results enhance understanding of the domain arrangement of DnaB FL and its biological role in association with the replication re-initiation protein PriA.

Sequence alignment of DnaB from different species
The Gram-positive bacterial helicase loader protein DnaB from G. stearothermophilus (GstDnaB) comprises 461 residues. Based on proteolysis assays using proteinase K and trypsin digestion studies of B. subtilis, BsuDnaB is structurally organized into three domains: residues 1-184 (N-terminal), 204 -296 (middle), and 297-472 (C-terminal) (31). The corresponding domains in GstDnaB comprise residues 1-144 (N-terminal), 201-290 (middle), and 304 -461 (Cterminal) (described in further detail in the following sections). To identify the sequence variation between species, we aligned the GstDnaB sequence with that of BsuDnaB. As shown in supplemental Fig. S1, GstDnaB exhibits 47.2% overall sequence identity with BsuDnaB. The lowest sequence similarity was observed between residues 150 and 204 and in the extreme C-terminal region (residues 430 and 460) of DnaB.

GstDnaB loader protein interacts physically with the primosome restart protein GstPriA
The DNA replication restart process and the underlying mechanics relating to the association of co-proteins in Grampositive bacteria needs further clarification. As mentioned earlier, the primosomal loader protein DnaB is unique among Gram-positive bacterial species and plays a key role in both DnaA-and PriA-mediated pathways to initiate and restart the DNA replication process at oriC and stalled forks, respectively (7, 18 -20, 22, 32, 33). In addition to localization at oriC (34), in B. subtilis the association of DnaB with loader proteins such as DnaD and DnaI at origin-independent loci is also considered to be essential for the replication restart process (19,22). In this regard it is not clear whether the DnaB loader protein associates directly or indirectly with primosomal protein PriA and other loader proteins. Hence, we aimed to test for direct association of GstDnaB with GstPriA. To this end, we performed glutathione S-transferase (GST) pulldown assays using purified recombinant full-length GstDnaB (GstDnaB FL ) and truncated forms (GstDnaB 1-300 or GstDnaB 301-461 ) as fusion proteins containing an N-terminal GST moiety (GST-GstDnaB FL , GST-GstDnaB 1-300 , or GST-GstDnaB 301-461 ). As shown in Fig. 1 A, GstPriA specifically interacted with GST-GstDnaB FL and GST-GstDnaB 1-300 , whereas interaction was completely abolished for the C-terminal region (GstDnaB 301-461 ). This suggests that the N-terminal and middle domains of GstDnaB are essential for interactions with GstPriA.
To corroborate this interaction between GstDnaB and GstPriA, we performed sedimentation-velocity experiments through AUC to provide direct evidence of complex formation in solution. Five different fragments of GstDnaB (GstDnaB FL , GstDnaB 1-300 , GstDnaB 301-461 , GstDnaB  , and GstDnaB 202-300 ) were individually mixed with an equal molar ratio of GstPriA (1:1) and then subjected to AUC. The sedimentation coefficient of GstPriA alone was 5.1 S, which corresponds to the relative molecular mass of the monomer ϳ89.5 kDa (green dashed lines in Fig. 1, B-F). The sedimentation coefficient of GstDnaB FL alone was 7.2 S, which corresponds to the relative molecular mass of the tetramer ϳ212 kDa (red dashed lines in Fig. 1B). With respect to GstPriA interaction with full-length GstDnaB, we noticed the formation of two distinct major peaks, denoted as A and B in the GstDnaB FL /GstPriA mixture. Peak A has a sedimentation coefficient of ϳ5.2 S, which is similar to the 5.1 S for GstPriA alone. However, the s value of peak B was 8.3 S, which is much greater than that for the GstDnaB FL tetramer alone, clearly indicating that this peak corresponds to a GstDnaB FL -GstPriA complex. Consistently, similar results were observed when GstDnaB 1-300 was substituted for GstDnaB FL in the above experiment (Fig. 1C); the sedimentation coefficient of peak A (5.2 S) corresponded closely to Gst-PriA alone, whereas the s value of peak B was also much higher than for the GstDnaB 1-300 tetramer alone (7.1 S), indicating formation of a complex between GstDnaB 1-300 and GstPriA. However, consistent with our earlier GST pulldown experiments, we did not find an interaction between GstDnaB 301-461 and GstPriA as a shift in the protein peaks of the mixture (blue line, Fig. 1D) was not observed, and instead, the peaks were 2 The abbreviations used are: AFM, atomic force microscopy; AUC, analytical ultracentrifugation; SAXS, small angle X-ray scattering; ITC, isothermal titration calorimetry; AFM, atomic force microscopy; SAD, single-wavelength anomalous diffraction; r.m.s.d., root mean square deviation; Se-Met, selenomethionine; NSRRC, National Synchrotron Radiation Research Center; R g , radius of gyration; ssDNA, single-stranded DNA.

Crystal structure of DnaB protein
much closer to those of both GstDnaB 301-461 and GstPriA alone (green and red dashed lines, Fig. 1D).
In addition, the s value of GstPriA-GstDnaB 1-300 complex (peak B of Fig. 1C) was 7.1 S, which is similar to the s value of GstDnaB FL tetramer. This would give an estimated molecular weight of the GstPriA-GstDnaB 1-300 complex at 7.1 S was ϳ212 kDa. Therefore, this indicated that the stoichiometry between GstPriA and GstDnaB complex most likely is 1-4. To further corroborate the stability for the complex between GstDnaB and GstPriA in solution, we also performed isothermal titration calorimetry (ITC) experiments. As shown in the supplemental Fig.  S2, the dissociation constant (K d ) between GstDnaB 1-300 and GstPriA was 47.2 M, and the C-value in ITC experiment is 1.4.
The low C-value range from 1 to 5 often leads to a large error in fitting, a result of stoichiometry (N). Fortunately, the decoupled of K d with the error in N at the low C-value permits the determination of K d when the titration proceeds to a near saturate level (35). Due to the low C value (1.4) in our original setting, we also switched the order of syringe and cell component in the experiment by titrating high concentration of GstPriA (780 M) to GstDnaB 1-300 (40 M). The best fittings of K d , N, and C values were 72.9, 0.215 and 0.12, respectively. Although the stoichiometry was quite matched to our expectations, the standard error was large due to the low C value. On the other hand, the K d had a reasonable standard error and was in a similar range compare with our other ITC data with a wild range of

Crystal structure of DnaB protein
concentration. Therefore, we only discussed the K d from ITC but did not take N into consideration. In addition, our ITC result indicated that the interaction between GstDnaB and Gst-PriA is dynamic. This dynamic interaction may be crucial for the formation, rearrangement and dissociation of the primosomal complexes. To identify the GstPriA-binding region in the GstDnaB 1-300 protein segment, we prepared two truncated forms of the protein (GstDnaB 1-145 and GstDnaB 160 -300 ). However, neither truncated form was soluble. Hence, we then generated two other soluble truncated forms (GstDnaB  and GstDnaB 202-300 ) and subjected them to sedimentation-velocity AUC. As shown in Fig. 1E, the mixture of GstDnaB  and GstPriA produced a broader peak (blue line) covering tetrameric GstDnaB 1-200 alone and monomeric GstPriA alone (red and green dashed lines), but a shift in the peak or a change to the sedimentation coefficient was not observed. This result indicates that GstDnaB  is not sufficient to interact with GstPriA. We did not observe complex formation between GstDnaB 202-300 and GstPriA; protein peaks of this mixture (blue line, Fig. 1F) did not shift noticeably, and its sedimentation coefficient was similar to that of GstDnaB 202-300 and Gst-PriA alone (green and red dashed lines, Fig. 1F). In brief, our sedimentation-velocity AUC results further support the GST pulldown assays in demonstrating DnaB interaction with PriA. Both the N-terminal and middle domain regions are required to establish this interaction, and the connecting flexible linker might have some additional role in PriA interaction. We found that the C-terminal domain of GstDnaB is not involved in PriA interaction, but it may have another role in DNA binding or in the recruitment of other primosome machinery for further DNA unwinding, and this topic warrants further study.

Overall structure of GstDnaB 1-300 monomer
Previous structural studies of DnaB have been limited to a low resolution using atomic force microscopy (AFM) and EM (29,31), which lack detailed structural information at the atomic level. To unravel the structural details of GstDnaB to better understand its biological role in primosome assembly and replication restart, we first attempted to crystallize fulllength GstDnaB (GstDnaB FL ). However, we could not obtain crystals, perhaps due to the flexibility of the linker region connecting the middle and C-terminal domains. Hence, we generated a truncated version of the protein encompassing the N-terminal and middle domains (residues 1-300: GstDnaB 1-300 ), which we previously showed to be functionally important for associations with PriA. A previous study also showed that BsuDnaB is truncated at its C terminus in a growth-phase-dependent manner (34). These observations suggested to us that structural study of this portion of the protein (i.e. DnaB 1-300 ) would have biological implications for its functional importance in primosome assembly.
The crystal structure of GstDnaB 1-300 was determined at 2.8 Å resolution via single-wavelength anomalous diffraction (SAD) phasing ( Table 1). As expected, the GstDnaB 1-300 monomer consisted of the N-terminal and middle domains connected by a long linker ( Fig. 2A). The N-terminal domain (residues 3-144) was composed of three ␤-strands and eight ␣-helices, whereas the middle domain (residues 161-288) was composed of two ␤-strands and seven ␣-helices (Fig. 2, A and B). The three ␤-strands in the N-terminal domain were organized into a three-stranded ␤-sheet, in which ␤1 and ␤3 are parallel but ␤2 and ␤3 are antiparallel. In the middle domain, composed of seven helices (␣9-␣15) and two ␤-strands (␤5 and ␤6), four ␣-helices (␣11, ␣13, ␣14, and ␣15) were juxtaposed, and the two ␤-strands (␤5 and ␤6) formed an antiparallel ␤-sheet. These two domains were connected by a long linker region (residues 145-160) containing one ␤-strand (␤4), which forms one ␤-sheet with ␤1-␤3 of the neighboring monomer (supplemental Fig. S3). Although a well-defined electron density was observed for most of the GstDnaB 1-300 structure, we did not see a continuous density between residues 172 and 189 in the middle domain (dashed line in Fig. 2B). Surprisingly, residues between 190 and 202 formed a long random structure that connected to ␣10 ( Fig. 2A; supplemental Fig. S4). Residues Asp-190 -Ile-192 form three main-chain hydrogen bonds with residues Phe-144 -Arg-146 of the linker region to further stabilize this random structure. We speculate that this region may form a large loop with the unobserved region (residues 172-189), further highlighting the dynamic nature of the protein.

Structural organization and assembly of GstDnaB as a tetramer
Although previous AFM and EM structural studies at low resolution had shown that BsuDnaB is a tetramer with a squarelike architecture (29), atomic details of the arrangement of the functional tetramer are lacking. In agreement with previous observations, we observed a D2 tetrameric conformation (dimer of dimers) for the GstDnaB 1-300 in our crystal structure Interestingly, the GstDnaB 1-300 dimers (Chains A and B or Chains C and D) were assembled in a domain-swap conformation, and the overall dimer was basket-like (Fig. 3A). Surprisingly, the overall tetrameric conformation of GstDnaB 1-300 in our crystal structure differs significantly from those of previously observed low-resolution structures of BsuDnaB reported in AFM and EM studies (29,31). AFM results revealed a square-like architecture of BsuDnaB, measuring 21.6 nm and with a distinct central space formed between the four monomers of the tetramer. In contrast, a 24 Å resolution EM reconstruction map demonstrated that the BsuDnaB tetramer adopted a spiral conformation with a distinct and large central space. Although both of these studies showed some similarity in the BsuDnaB tetramer conformation, differences between the low-resolution structures are suggestive of a highly flexible protein. Our GstDnaB 1-300 crystal structure did not reveal a central space or hole between the four monomers of the tetramer. Instead, our overall structure had a completely new conformation in which two dimers of GstDnaB 1-300 are aligned back-to-back and form a large interface at the center of the tetramer.
To resolve the structural differences and domain arrangement, we first calculated our GstDnaB 1-300 X-ray structure to a 24 Å density map and compared it with the EM structure of BsuDnaB (the accession code EMD-1225 in the Electron Microscopy Data Bank) at the same resolution. As shown in supplemental Fig. S5, A-C, GstDnaB 1-300 tetrameric structure (colored in pink) did not show square or spiral arrangement with distinct central holes as observed in the EM structure of BsuDnaB (colored in green). Instead, the GstDnaB tetrameric structure was arranged in a letter X-like manner and lacked any similarity in comparison with the structural arrangement with BsuDnaB. Interestingly, although the GstDnaB 1-300 tetrameric structure lacks C-terminal domain, its overall volume appear to be comparable with that of the BsuDnaB EM structure at 24 Å resolution (supplemental Fig. S5, A-C). Meanwhile, as the AFM structural data of BsuDnaB (29) is not available in data bank and it has been shown that both EM and AFM low-resolution structures BsuDnaB share similarities (31), we did not further attempt to compare with our structure.

Structural interactions between monomers and dimers of GstDnaB
The dimeric interface of two monomers is located between chains A and B or chains C and D. The interface between the two dimers that forms the stable tetramer is formed by chain A/B and chain C/D (Fig. 3A). To determine the most likely interaction interface that could represent a major structural element in the formation of the tetramer, we used the PISA server (36) to measure the surface areas of these interfaces. The buried surfaces within the chains A and B dimer and the chains C and D dimer were 3542.1 Å 2 and 3431.5 Å 2 , respectively (Interface-1 and Interface-2 of Fig. 3, A-C, respectively). The buried surfaces of the interfaces of chains A/C and chains B/D were 1548.4 Å 2 and 1490.5 Å 2 , respectively (Interface-3 of Fig. 3,  A and D). These results suggest that the dimeric interfaces between chains A and B and chains C and D are the most likely to play a major role in the formation of GstDnaB dimer, whereas interactions between chains A and C coupled with chains B and D contribute to assembly and stabilization of the overall tetramer.
Because GstDnaB 1-300 dimer (chain A/B or chain C/D) is assembled in a domain-swap conformation, the DnaB dimeric interface is stabilized by Interface-1 and Interface-2 from two adjacent N-terminal and middle domains of DnaB (Fig. 3, B and C), with Interface-1 being formed between the two adjacent N-terminal domains of DnaB (Fig. 3B). This interaction is stabilized by reciprocal hydrogen bonds between residues Gln-23 and Glu-24 of chains A and B, and reciprocal salt bridges between residues Asp-26 and Arg-194 of chains A and B. In addition, evidence for several van der Waals interactions Crystal structure of DnaB protein between both chains was found. Interestingly, the major interaction between the DnaB monomers in the dimer occurs in Interface-2 (see Fig. 3C), which might serve as a key factor in establishing the "domain-swap" to stabilize the dimer. We found that the N-terminal domain of chain A builds an interaction interface (Interface-2) with the linker region and middle domain of adjacent monomer chain B (Fig. 3, A and C). These interactions involve residues from ␣13 and ␣15 in chain A and residues from ␤2, ␣5, ␣6, and ␣7 in chain B. Residues Lys-235, Arg-267, and Glu-282 of chain A form salt bridges with Asp-121, Glu-122, and Arg-129 of chain B, respectively. In addition, hydrogen bonding between residues Gln-156, Arg-159, Glu-196, and Tyr-199 of chain A with Asp-91, Tyr-78, Glu-85, Gln-117, and Asp-121 of chain B further strengthens the dimerization interface, which could contribute to stabilization of the DnaB dimer. The nature of this dimerization interface and the corresponding residues involved in the interactions establishing the dimer are also consistent in the other dimer of chains C and D.
As described earlier, the overall tetrameric conformation of DnaB is stabilized by the interaction interface between the two dimers (AB and CD) through Interface-3 (Fig. 3, A and D). The intermolecular interactions stabilizing this interface also involve several salt bridges, hydrogen bonds, and hydrophobic interactions. Residue Glu-54 of chain A/D forms a salt bridge with Lys-28 of chain B/C at one end of Interface-3. At the other end, hydrogen bonds are formed between residues Thr-48 of chain A/D and residue Gln-35 of chain B/C. The central part of Interface-3 possesses a distinctive hydrophobic core that is created by the hydrophobic residues Leu-55 and Leu-69 of one chain and Leu-32, Leu-136, and Phe-143 of the neighboring chain. We postulate that these hydrophobic interactions might play an additional role in the establishment or stabilization of the GstDnaB tetramer by maintaining the interdimer conformation. Surprisingly, there are 36 residues (9 residues of each monomer) involved in interface-3 interaction, and sequence alignment between GstDnaB and BsuDnaB shows that these residues involved in the GstDnaB tetramer interfaces share 44% sequence identity (56% sequence similarity). This result indicated that DnaB co-loader protein from different species may use these residues to form a stable tetramer structure. These structural interactions and the observed interfaces between DnaB monomers in the tetramer suggest that the N-terminal and middle domains of DnaB dictate the association of monomers to effect DnaB's role as a tetramer.

The N-domain and linker regions are required for tetramerization of GstDnaB
Our earlier results suggested that residues 1-300 of GstDnaB play an important role in tetramerization and functional stability. We next prepared the three truncated forms GstDnaB  (N-terminal domain alone), GstDnaB  (N-terminal domain with linker region), and GstDnaB 202-300 (middle domain alone) for AUC studies to determine which region guides formation of the functional tetramer. Unfortunately, GstDnaB 1-144 showed impaired solubility, and we could not pursue further studies.

Crystal structure of DnaB protein
AUC results for GstDnaB  and GstDnaB 202-300 showed sedimentation coefficients of 4.8 S and 1.2 S, respectively (red lines in Fig. 1, E and F). The c(s) distribution analysis corresponded to the molecular mass of ϳ97.3 kDa for DnaB  , which closely matches the tetrameric conformation. The molecular mass of DnaB 202-300 was ϳ12.5 kDa, which closely matches the monomeric conformation. Thus, our AUC analysis with different truncated forms of GstDnaB indicates that the N-terminal domain plus linker region contributes to dimer and/or tetramer formation. The significance of DnaB tetramerization has also been addressed in B. subtilis through the formation of a bead-like nucleoprotein structure with linear dsDNA (29).

GstDnaB binds ssDNA substrate through its C-terminal domain
As studies have shown that DnaB plays a prominent role in the DNA replication initiation process at both oriC and stalled forks, we next tested the DNA-binding ability of GstDnaB FL and its truncated forms. We wondered if the C-terminal region plays an important role in DNA-binding activity of Geobacillus species. Hence, we tested purified full-length and two truncated forms of GstDnaB (GstDnaB 1-300 and GstDnaB 301-461 ) for ssDNA (dT40)-binding activity by electrophoresis mobility shift assays (EMSA). As shown in Fig. 4A, EMSA results revealed that increasing concentrations of GstDnaB FL exhibited increased ssDNA-binding activity. Surprisingly, we did not observe ssDNA-binding activity for GstDnaB 1-300 even with a protein concentration of 240 M (Fig. 4B). This observation conflicts with a previous study of ssDNA substrate binding in the Bacillus species (34). However, ssDNA-binding activity was observed for GstDnaB 301-461 with increasing protein concen-trations (Fig. 4C). This observation is in accordance with an earlier study (34). To calculate the dissociation constants of DnaB and ssDNA binding, we quantified the florescent signal in each DNA band. As shown in Fig. 4D, DnaB FL had the highest ssDNA-binding affinity, with an apparent K d of 34.9 M followed by GstDnaB 301-461 (K d ϭ 115.4 M) and with no binding activity for GstDnaB 1-300 . These results indicate that DNA binding by GstDnaB is predominantly mediated by the C-terminal domain. Overall, these results highlight the functional difference between two truncated forms of GstDnaB representing the N-terminal plus middle domain and the C-terminal domain, respectively, in terms of PriA interaction and ssDNA binding.

Comparisons with other protein structures
To further understand the structure and function of GstDnaB, we first performed a structural homology search for the N-terminal domain of GstDnaB (GstDnaB-NTD) (residues 1-144) using the Dali server (37). As shown in Fig. 5A, the proteins RepA (38) from Staphylococcus aureus (SaRepA) (PDB ID code 4PT7) and TFE (39) from Sulfolobus solfataricus (SsTFE) (PDB ID code 1Q1H) showed the highest structural homology with GstDnaB-NTD despite having lower sequence identities with it of 18 and 10%, respectively. We next superimposed GstDnaB-NTD with SaRepA-NTD and with SsTFE-NTD and found that in both cases they were well-aligned, with r.m.s.d. of 2.6 Å and 2.1 Å, respectively (Fig. 5A). Both SaRepA and SsTFE lack the first ␤-strand (␤1) observed in GstDnaB-NTD, and their ␤2 and ␤3 strands are folded into a characteristic winged helix-turn-helix fold that is structurally well-conserved in both proteins. In RepA, the winged helix-turn-helix folded region is proposed to have DNA-binding activity and

Crystal structure of DnaB protein
plays a key role in origin recognition by plasmids during replication (38). In contrast, bacterial SsTFE-NTD lacks DNA-binding activity (39), similar to GstDnaB-NTD. We calculated the surface potentials of these three proteins. As shown in Fig. 5B, SaRepA-NTD has a much more positively charged surface potential than GstDnaB-NTD and SsTFE-NTD. This finding may explain why we did not observe ssDNA-binding for GstDnaB 1-300 . Although the N-terminal domains of GstDnaB and SaRepA exhibit structural similarities, their functional roles and their associations with other partner proteins differ.
We then performed the Dali homology search for the middle domain of GstDnaB (GstDnaB-MD) (residues 161-290). Our results showed that the N-terminal domain of DNA polymerase ␣-primase B-subunit (p68N) (40) was the most structurally homologous to GstDnaB-MD. Despite the similarity in their architectures (Fig. 5C), an amino acid sequence alignment between the middle domain of GstDnaB and p68N revealed low sequence similarity, and the chemical nature of their surfaces differed considerably (Fig. 5D). The middle domain of GstDnaB is characterized by an electrostatic surface with both basic and acidic charges, whereas the N-terminal domain of p68N displays a predominantly acidic surface with a few small hydrophobic patches. As discussed above, the middle domain of GstDnaB plays a role in mediating interactions with PriA. However, the N-terminal domain of p68N physically interacts with the T antigen helicase domain. Despite the differences between the surfaces of these two domains of GstDnaB and p68N, they have a similar function in mediating proteinprotein interactions.

SAXS analysis reveals an elongated shape for GstDnaB FL
Because crystallization of GstDnaB FL proved challenging due to its high flexibility, we performed solution scattering experiments to resolve the overall tetrameric structure of the fulllength protein. SAXS analysis was conducted for both GstDnaB FL and GstDnaB 1-300 . The radius of gyration (R g ), maximum dimension (D max ), Guinier plots, and pair distance distribution plots for both GstDnaB FL and GstDnaB 1-300 were calculated using the SAXS profile (Fig. 6, A-D; supplemental Table S1). As expected, R g values (derived from the Guinier plots using the program PRIMUS) for GstDnaB FL and GstDnaB 1-300 were significantly different; 57.45 Ϯ 0.199 Å and 35.42 Ϯ 0.036 Å, respectively (supplemental Table S1). This result clearly suggests that the overall dimensions of the SAXS shells and the structural architecture of GstDnaB FL and GstDnaB 1-300 are different. Indirect Fourier transformation of the curves calculated using the GNOM program and the experimental curves fitted with SAXS data for both GstDnaB FL and GstDnaB 1-300 are shown in Fig. 6, A and B. The observed linearity of the Guinier plots in the lower q-region indicates that the scattered intensities follow the Guinier law, suggesting that both protein samples were monodispersed and is indicative of good quality data (inner panel of Fig. 6, A and B). By using P(r) functions, we next determined D max for both proteins and measured values of 200 Å for GstDnaB FL and 112 Å for GstDnaB 1-300 (Fig. 6, C and D). The significant difference in the Dmax values suggests that GstDnaB FL possesses an extended or elongated conformation compared with that of GstDnaB 1-300 .
As shown in supplemental Figs. S6 and S7, we generated eight molecular envelopes for both GstDnaB FL and GstDnaB 1-300 from the scattering data using the ab initio modeling feature of the GASBOR program. These eight molecular envelopes were averaged by the program DAMAVER (41). The resulting "overall" envelope of GstDnaB 1-300 is elliptical, with dimensions relatively consistent with those of our crystal structure in the tetrameric conformation. To compare our crystal structure of GstDnaB 1-300 with the overall ab initio envelope derived from solution scattering, we performed docking using the SUPCOMB program (42). As expected, our crystal structure fitted well with the GstDnaB 1-300 SAXS envelope (Fig. 6E). Furthermore, the observed experimental SAXS curve aligned well with the theoretical curve calculated from the crystal structure of GstDnaB 1-300 using CRYSOL, with an acceptable Chi () value of 3.46 (Fig. 6G). This suggests consistency of the  Figure 6. Small angle X-ray scattering characterization of GstDnaB 1-300 and GstDnaB FL . A and B, experimental X-ray scattering curves (black lines) and theoretical fitting curves (red lines) of DnaB 1-300 and DnaB FL were generated with GNOM. The insets show the Guinier plots. C and D, the distance distributions of DnaB 1-300 and DnaB FL . E and F, ab initio models of DnaB 1-300 and DnaB FL . The models were obtained from GASBOR and superimposed with the crystal structure of DnaB 1-300 using the SUPCOMB program. The color scheme is the same as that for Fig. 2A. G, experimental SAXS data (black line) and calculated scattering curves for the crystal structural of DnaB 1-300 (red line) fitted by CRYSOL.

Crystal structure of DnaB protein
tetrameric conformation of GstDnaB 1-300 in both crystal form and in solution.
The overall GstDnaB FL envelope from the SAXS curve was elongated in shape, with an additional protruding density corresponding to the C-terminal region (residues 301-461). The protruding density was also found to be well-connected to the middle domain, so that the overall envelope resembled an "X-like" structure. To validate that the additional density could accommodate the C-terminal region of GstDnaB, we fitted the tetrameric conformation of the GstDnaB 1-300 crystal structure into the central region of the GstDnaB FL envelope using the SUPCOMB program. The docked crystal structure fit well into the SAXS envelope of GstDnaB FL , leaving additional density at the four corners that could adequately accommodate the C-terminal domain of GstDnaB (Fig. 6F).
Taken together, the results from our generated overall SAXS envelope and the corresponding docking studies support the unique structural conformation of the GstDnaB FL tetramer and the flexible nature of the C-terminal domain. This flexible C-terminal domain protruding from the stable tetramer points to the functional dynamics of GstDnaB in recognition and coiling of DNA substrate.

Discussion
Gram-positive bacterial DnaB is an essential protein required for both initiation and restart of parental DNA replication (19). Bacterial DnaB also associates with both DnaA and PriA loader proteins to initiate the cascade of DNA replication at oriC and stalled replication forks, respectively. Our structural and biochemical studies provide insights into the functional mechanics of this loader protein in primosome assembly. As shown in Fig. 3A, GstDnaB 1-300 forms a unique architecture of two basket-like structures in a domain-swap conformation. Our SAXS-derived envelope studies provide evidence that GstDnaB FL has an X-shaped conformation in solution. GstDnaB FL is an elongation of the crystal structure of GstDnaB 1-300 with a protruding C-terminal domain. Although previous studies concluded that BsuDnaB is a tetramer (consistent with our studies), the conformational architecture of GstDnaB differed from it greatly. We speculate that the previous low resolution EM tetrameric structure (29, 31) may be incorrect due to bias caused by choosing an inappropriate ␥-complex clamp loader as an initial model, resulting in a different tetrameric arrangement. However, we cannot rule out the possibility that the different conformations might be due to the low sequence similarity of residues 148 -205 (supplemental Fig. S1) from different bacterial species. As observed in our crystal structure, residues in this variable region (residues 148 -205) also play some role in the formation of an interface between the monomers in the tetramer (Fig. 3).
Based on our crystal structure and SAXS data analysis, the X-shaped architecture of tetrameric GstDnaB FL has biological implications. The protruding GstDnaB C-terminal domains on both sides of the tetrameric conformation might be essential for its functional role in DNA binding, as these domains possess clusters of highly positively charged residues. Previous functional studies have shown that the BsuDnaB C-terminal region alone is sufficient to interact with both ssDNA and dsDNA substrate (34). The GstDnaB C-terminal region alone possesses 35 positively charged residues, the majority of which are clustered between residues 365 and 400 (see supplemental Fig. S1). Such a large number of highly positive residues clustering at one locus was not found for GstDnaB  . Although the overall tetrameric surface charge of GstDnaB 1-300 shows some exposed positively charged residues (supplemental Fig. S8), we did not detect noticeable DNA-binding activity even at protein concentrations of 240 M. Instead, GstDnaB 1-300 exhibited interaction with GstPriA, and this association was relatively stable under ultracentrifugation. Association of DnaB with PriA is one of the key steps in the assembly of the primosome machinery during replication restart. BsuDnaB was not found to associate with PriA (22), which conflicts with our GstDnaB functional studies, but the non-association of DnaB with PriA in Bacillus species might be due to weak interactions under the in vitro conditions or low concentrations (10 M) used in that study (22).
In addition, the lack of PriA-binding activity for GstDnaB 1-200 (see Fig. 1) suggests the importance of the middle domain in the PriA association. Our crystal structure and solution-scattering studies of GstDnaB 1-300 together with our in vitro functional studies of complex formation with PriA demonstrate the functional and structural roles of the N-terminal and middle domains of GstDnaB. Our ssDNA-binding studies also show that GstDnaB 301-461 (but not GstDnaB 1-300 ) is involved in DNA binding.
In summary, we determined the crystal structure of a twodomain construct of DnaB from the Gram-positive G. stearothermophilus. The structure suggests a dimer-dimer tetrameric arrangement for DnaB and provides insights into the interdomain communication between the N-terminal and middle domains. Our biochemical assays also distinguished the PriAbinding and ssDNA-binding domains of GstDnaB, which extends our understanding of primosomal assembly and its mechanics.

Plasmid construction
The coding regions of full-length DnaB (DnaB FL ) and truncated forms (DnaB 1-300 , DnaB 301-461 , DnaB  , and DnaB 202-300 ) were individually generated by PCR amplification of genomic DNA isolated from G. stearothermophilus using Pfu DNA polymerase (Stratagene). NdeI and XhoI restriction sites were incorporated into forward and reverse primers, respectively, to permit insertion of amplified PCR products into the pET21b vector (Novagen) with a C-terminal His 6 tag for protein expression in E. coli. For N-terminal GST-tagged DnaB, the PCR-amplified coding region of DnaB FL or its truncated forms (DnaB 1-300 and DnaB 301-461 ) were cloned into a pGEX-2TK vector using the BamHI and EcoRI sites.
PCR amplification of the gene encoding full-length PriA (residues 1-801) was performed in a similar manner as that for DnaB, except for incorporation of NdeI and HindIII restriction sites. The resulting amplified fragment was cloned into pET21b (Novagen) with a C-terminal His 6 tag.

Crystal structure of DnaB protein Protein expression and purification
Expression and purification procedures for all His 6 -tagged proteins (DnaB FL, DnaB 1-300 , DnaB 301-461 , DnaB 1-200 , DnaB 201-300 , and PriA) were similar. Briefly, the individual plasmid was transformed and expressed in E. coli BL21 (DE3) cells. Transformed E. coli cells were cultured in LB medium containing 100 g/ml ampicillin and grown at 37°C until optical density (A 600 ) reached 0.6. Overexpression of target protein was induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside and further incubated for 12 h at 20°C. All purification procedures were performed at 4°C. Cells were harvested by centrifugation and then suspended in an optimized buffer containing 20 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 5 mM imidazole (buffer A). The cell suspension was lysed with an M-110L Microfluidizer apparatus (Microfluidics) and then centrifuged at 30,000 ϫ g for 30 min. The soluble cell extract was loaded onto a HisTrap HP (Ni 2ϩ -chelating) column equilibrated with buffer A. The column was washed with 10 column volumes of buffer A plus 50 mM imidazole to remove impurities. His 6 -tagged proteins were eluted with buffer A plus 200 mM imidazole. Eluted protein fractions were further purified by Superdex 200 gel filtration chromatography equilibrated with 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 5 mM ␤-mercaptoethanol (buffer B).
For the production of selenomethionine-substituted DnaB 1-300 protein (SeMet-DnaB 1-300 ), cells were grown and expressed in Overnight Express Autoinduction System 2 medium (Novagen), and then cell suspensions were purified in a similar manner as that for DnaB 1-300 . All purified proteins were subjected to SDS-PAGE analysis, and purities of Ͼ95% were confirmed by peptide mass spectroscopy.
Recombinant GST-tagged proteins (GST-DnaB FL , GST-DnaB 1-300 , and GST-DnaB 301-461 ) were expressed as described above, and harvested cells were resuspended in 1ϫ phosphatebuffered saline (PBS) and then lysed. The lysate was clarified by centrifugation, and the resulting supernatant was loaded onto a GSTrap HP column equilibrated in buffer containing 1ϫ PBS buffer. The column was washed with 10 column volumes of 1ϫ PBS buffer. Bound GST-tagged protein was then eluted with an elution buffer containing 50 mM Tris-HCl (pH 7.4), 200 mM NaCl, and 10 mM glutathione. Eluted protein was further purified by gel filtration chromatography on a Superdex 200 column in 25 mM HEPES (pH 7.4), 500 mM NaCl, and 1 mM DTT before being subjected to SDS-PAGE analysis to confirm protein purity.

Crystallization and X-ray data collection
Crystallization trials were set up by using 15 mg/ml GstDnaB 1-300 in a buffer containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 5 mM ␤-mercaptoethanol. Initial screening was performed by using a Phoenix robot platform (Rigaku) with commercially available screening reagents (Hampton Research, Molecular Dimension, and Qiagen Ltd.) at 25°C. Native crystals of GstDnaB 1-300 (0.15 mm ϫ 0.15 mm ϫ 0.02 mm) were obtained by the hanging-drop vapor diffusion method against a buffer containing 0.1 M Tris (pH 8.5) and 1.0 M ammonium sulfate. To solve the crystallographic phase problem, SeMet-GstDnaB 1-300 crystals were obtained against buffer containing 25% v/v ethylene glycol under hanging-drop vapor diffusion conditions and diffracted to 2.8 Å resolution. Both crystals belong to the space group P2 1 2 1 2 1 , with cell dimensions of a ϭ 110.04 Å, b ϭ 117.12 Å, and c ϭ 159.13 Å. Single-wavelength anomalous diffraction data were generated using synchrotron X-ray radiation at Beamline 15A1 of the National Synchrotron Radiation Research Center, Taiwan, and collected with a Rayonix MX300HE CCD Area Detector at 100 K. Selenium single-wavelength anomalous dispersion (Se-SAD) data acquired at a peak wavelength of 0.9792 Å was used to phase the SeMet-GstDnaB 1-300 crystal. The dataset was processed using the HKL2000 program suite (43), and statistics are shown in Table 1.

Structural determination and refinement
Identification of selenium positions and generation of the initial SAD phase of SeMet-GstDnaB 1-300 crystals at 2.8 Å resolution were performed using PHENIX (44,45). After identifying 21 selenium positions, the initial phases were further refined using the maximum likelihood density modification algorithm in PHENIX. In the crystal there are four molecules in an asymmetric unit. Due to the limited resolution, automatic model building using PHENIX was not successful, so we used Coot (46) for model building. The final SeMet-GstDnaB 1-300 structure was obtained through iterative cycles of rebuilding and refinement using Coot (46) and PHENIX (44,45), respectively. A total of 85 water molecules were found in the crystal structure by the general water molecule selection tool in PHENIX. The final structure was refined to an R factor of 22.8% and a R free value of 26.3% for the 1978 reflections (3.8%) of randomly chosen reflections. Due to the absence of an electron density for the region corresponding to residues 172-189, we could not trace and validate the secondary structure. The residues in the individual chains were identified as follows: chain A (residues 3-171 and 190 -288), chain B (residues 7-171 and 190 -288), chain C (residues 7-171 and 190 -288), and chain D (residues 6 -171 and 190 -288). A PHENIX-generated Ramachandran plot and theangles for GstDnaB 1-300 show that 96.9% of residues are in the most favored regions, and 3.1% are in allowed regions. However, due to the unclear or discontinuous electron density map located in several loop regions, there are still some residues from different chains with poor fit to the density. Refinement statistics are summarized in Table 1. The figures were produced using PyMOL (53). Solvent-accessible and interface areas were calculated by PISA (36). The atomic coordinate and the structure factor for GstDnaB  have been deposited in the Protein Data Bank under accession code 5WTN.

GST pulldown assays
Purified GST (200 g) or three GST-tagged DnaB proteins (200 g; GST-DnaB FL , GST-DnaB 1-300 , GST-DnaB 301-461 ) were individually mixed with 10 l of glutathione agarose beads at room temperature for 30 min. The beads were pelleted by centrifugation and washed with 1 ml of a binding buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, and 0.05% Tween 20. The resulting immobilized indi-Crystal structure of DnaB protein vidual proteins were then incubated with 700 g of PriA monomer at room temperature for 30 min. Beads were pelleted and washed 5 times with 1 ml of a wash buffer containing 50 mM HEPES (pH 7.5), 100 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, and 0.05% Tween 20. The bound proteins were eluted from the beads by boiling in SDS sample buffer before being subjected to SDS-PAGE. Western blotting was then performed to detect PriA interactions with GST alone or with the three GST-tagged DnaB proteins.

Isothermal titration calorimetry
Experiments were carried out using the MicroCal iTC200 system (GE Healthcare). For all isotherms, the sample cell was filled with 0.1 mM GstPriA in 20 mM Tris-HCl (pH 8.0) and 150 mM NaCl, and the syringe was loaded with 1 mM GstDnaB 1-300 in the same buffer. The binding isotherm was obtained from 20 injections of GstDnaB 1-300 into GstPriA at 25°C. For the first titration, an injection volume of 1 l was used; for subsequent titrations, 2 l of GstDnaB 1-300 was injected with an interval of 180 s. The stirring speed and reference power were 1000 rpm and 5 cal/s, respectively. Binding isotherms were integrated and analyzed using Origin v7.0 software (MicroCal).

Small angle X-ray scattering analyses
Protein samples for SAXS analysis were collected by gel filtration. The SAXS experiments for DnaB FL and DnaB 1-300 were performed at Beamline 23A at the National Synchrotron Radiation Research Center (NSRRC) in Taiwan, equipped with a MAR 165 CCD detector at a sample-to-detector distance of 2.5 m. The X-ray wavelength was 0.8285 Å, with an energy of 15.0 keV, and the collection time was 300 s. Protein sample concentrations were 14 mg/ml and 20 mg/ml for DnaB FL and DnaB 1-300 , respectively. The gel filtration buffer (20 mM Tris (pH 8.0), 300 mM NaCl, and 5 mM ␤-mercaptoethanol) was used as the solvent blank. No sample aggregation was observed during the measurements. Background scattering from the buffer alone was subtracted from the sample, and data were scaled using PRIMUS (48). Primary data reduction was done with an NSRRC 23A SXAS data reduction program and further analyzed using the ATSAS package (49). The R g was calculated from a Guinier plot using PRIMUS. The pair distance distribution function P(r) and the maximum distance D max were calculated using GNOM (50). The low-resolution shapes of protein samples were determined as ab initio models from the scattering data by GASBOR (51). The eight ab initio models from the eight individual GASBOR runs had a similar overall architecture. The eight ab initio models were aligned, averaged, and scored with a normalized structural difference (NSD) using DAMEVER (41). The eight ab initio models of DnaB FL and DnaB 1-300 agreed well, yielding 1.071 Ϯ 0.336 and 1.231 Ϯ 0.088 (NSD Ϯ S.D.), respectively. The SAXS data were fitted with the crystal structure of DnaB 1-300 using CRYSOL (52). The SAXS envelopes and crystal structure of DnaB 1-300 were superimposed using SUPCOMB (42). Data collection and other SAXS measurements together with the programs used are summarized in supplemental Table S1.

Electrophoresis mobility shift assays
To determine ssDNA-binding ability, interactions of purified recombinant proteins (GstDnaB FL , GstDnaB 1-300 , and GstDnaB 301-461 ) with ssDNA (dT40) substrate were resolved on 8% polyacrylamide gels. All purified proteins were analysis by SDS-PAGE to confirm the purity (supplemental Fig. S9). Binding assays were carried out in a final volume of 20 l of DNA-protein reaction mixture in buffer B for 30 min at 20°C, with the proteins at the indicated concentrations against 20 nM dT40 ssDNA substrate labeled at the 5Ј end with Cy5 (MDBio, Inc.). Post-reaction, the samples were mixed with 6ϫ gel loading buffer (60 mM Tris-HCl (pH 7.6) and 30% glycerol) and the protein-ssDNA complexes were electrophoresed on 8% native polyacrylamide gels in 0.5ϫ TBE buffer (45 mM Tris borate, 1 mM EDTA (pH 8.0)) at 4°C for 160 min. The gels were immediately scanned for the florescence signals using the Cy5 channel with Typhoon FLA 9000 (GE Health) for visualization of DNA bands. The fluorescent signals in each DNA bands were quantified by using ImageJ software. Results of three independent repeats for ssDNA-protein affinity assays were measured. The