Introduction

Bacterial cells have a fundamental need to divide by binary fission through accurate spatial and temporal regulation of septum formation, producing two daughter cells of equal size1. In the Gram-negative bacterium E. coli, a tubulin like protein (FtsZ) polymerizes at mid-cell to form a ring structure (Z-ring) to ensure cell division, which is often spatially regulated by the Min system, including MinC, MinD and MinE, through a pole-to-pole oscillation2, 3. MinC, as an inhibitor of the Z-ring, is recruited to the membrane by MinD and prevents FtsZ polymerization. MinE has been shown as an anti-MinCD component. The binding of a MinE dimer to MinCD has been reported to trigger the release of MinC followed by stimulating the ATPase activity of MinD and its dissociation from the membrane, thereby protecting the central division site to avoid inhibition by MinCD4. MinC contains two separate domains: an N-terminal domain, which directly interacts with FtsZ and prevents its polymerization; and a C-terminal domain, which forms a constitutive dimer and interacts with MinD as well as with FtsZ5. Ghosal et al. found that, in vitro, MinC and MinD together form a new class of nucleotide-dependent and alternating copolymeric filaments6. These copolymers were shown not to be physiologically relevant7, but are still useful in studying MinD-MinC interactions in a biochemical context. During the oscillation, MinD, along with MinC and MinE, is present in a polar zone flanked near mid-cell by the MinE ring4, 8, 9. In this mechanism of cell division positioning, each of the three proteins (MinC, MinD and MinE) plays indispensable functional roles and must be present. In contrast, in the Gram-positive bacterium B. subtilis, the MinCD complex does not appear to oscillate because there exists no MinE. Due to interaction with the pole-anchored DivIVA, the MinCD complex remains concentrated in two polar regions, thereby preventing division from taking place near the poles10, 11. In this scenario, MinJ is needed to bridge DivIVA and MinD12, 13, resulting in a different cell division pattern from min oscillation in the Gram-negative bacteria.

Archaea are a third domain of life in addition to Bacteria and Eukarya14. Archaea were initially viewed as extremophiles living in harsh environments, such as hot springs and salt lakes15. The mechanisms of cell division in Archaea differ from those in bacteria in many aspects, including DNA replication and membrane organization16. Almost all members of the Euryarchaeota, one of the five archaeal phyla, encode FtsZ and, thus, are thought to possess a bacterial-type division mechanism17, 18. However, the exact mechanism of cell division of Archaea has not been investigated and thus remains completely unknown. Further, the evolutionary relationship with regard to cell division between Archaea and Eukaryotes is also unclear. A. aeolicu is one of a handful of species in the Aquificae phylum, an unusual group of thermophilic bacteria that is thought to be the oldest species of bacteria19. Similarly, T. maritima resides in hot springs as well as hydrothermal vents20. The ideal environment for the organism is a water temperature of 80 °C (176 °F). Although T. maritima and A. aeolicu have been determined as members of Gram-negative bacteria, they reside in extreme environments similar to members of the domain Archaea21. Therefore, study of T. maritima and A. aeolicu will help understand the interaction and evolutionary relationship among Bacteria and Archaea.

In this work, we determined the crystal structure of the dimeric C-terminal domain of MinC from E. coli (EcMinCCTD). EcMinCCTD forms a dimer between the two β-sheets in each subunit, as observed in the TmMinCCTD structure22. However, both EcMinCCTD and TmMinCCTD lack an α-helix (helix3) at their C-terminal tail compared to AaMinCCTD which forms another interaction interface with MinD6. By fusing helix3 to the C-terminus of EcMinC, we studied its effect on cell morphology and cell growth, revealing that Aahelix3 impaired normal cell division in E. coli. Results of a co-pelleting assay and binding free energy calculation further led to our conclusion that the Aahelix3 enhances the interaction of AaMinCD complex, which may hinder the role of MinE in the dissociation of MinCD complex. Sequence analysis shows that there is no minE gene and MinC with the extra helix3 in non-oscillation cell division systems. Taken together, we propose an evolutionary relationship to explain the different mechanisms in cell divisions in Bacteria and Archaea.

Results

Structural characterization of EcMinCCTD

Since full-length EcMinC structure has not been determined, we initially aimed to crystallize the full-length protein. After 3~6 weeks, crystals grew to the maximum size (Fig. S1). SDS-PAGE analysis of the harvested crystals showed that these crystals resulted from a degradation fragment of the MinC protein at a molecular weight of ~13 kD. Furthermore, western blotting results demonstrated that this fragment of MinC had the C-terminal 6 × His tag (data not shown), which suggested that our crystals were from EcMinC’s C-terminal domain (MinCCTD). This is likely caused by degradation of the flexible linker between the N- and C-terminal domains of MinC, which is easily susceptible to non-specific proteolysis. Subsequently, we created the C-terminal domain (pET22b-MinCCTD) construct for crystallization. However, MinCCTD protein was very unstable and prone to precipitation. As a result, we went back to the initial method of using full-length MinC to obtain MinCCTD crystals. It is inferred that MinCNTD, although cleaved off over time, initially helps to promote the correct folding and solubility of MinCCTD.

The crystal structure of EcMinCCTD was solved at 3.0 Å resolution using the molecular replacement method. The search model was the C-terminal domain of TmMinC (PDB: 1HF2). EcMinCCTD forms a dimer within the asymmetric unit via domain swapping (Fig. 1A). The tight dimer formed through the MinCCTD beta helical domain is similar to that of TmMinCCTD (Fig. 1B, left). The two MinCCTD structures superimposed with a root-mean-square deviation (RMSD) of 1.4 Å for the 110 Cα atoms (residues 122–231).

Figure 1
figure 1

Overall structure of the dimeric MinC C-terminal domain from E. coli (EcMinCCTD). (A) The ribbon diagrams of two monomers are in green and magenta. (B) Left: structural comparison of EcMinCCTD (yellow) and the TmMinC C-terminal domain (TmMinCCTD, red). Right: structural comparison of EcMinCCTD (yellow) and the AaMinC C-terminal domain (AaMinCCTD, orange). Superimposition was performed using the program PyMOL.

Previously, an A. aeolicus MinCCTD-MinD complex (AaMinCD, PDB: 4V02) and an E. coli MinD (EcMinD) dimer crystal structure had been solved (PDB: 3Q9L). Compared to AaMinCCTD, EcMinCCTD and TmMinCCTD both lack a helix3 at its C-terminus (Fig. 1B). We superimposed our EcMinCCTD structure and the EcMinD structure (PDB: 3Q9L) onto the AaMinCD co-crystal structure (PDB: 4V02), resulting in an assembled EcMinCD complex (Fig. 2A). In this EcMinCD model structure, the RSGQ motif of MinC and MinD’s helix 7 make up an interaction interface as expected. In earlier studies the highly-conserved residues D154 in EcMinD and RSGQ motif in EcMinC were identified to be very important for MinC-MinD interaction23, 24 (Fig. 2A). The helix3 of AaMinCCTD forms interaction interface 2 with helix 8 of MinD, which is absent in both EcMinC and TmMinC (Fig. 2B). We superimposed the assembled EcMinCD model complex (Fig. 2C) on EcMinDE complex (PDB: 3R9J), which shows that the interaction interface of the EcMinDE complex is more compact than that of EcMinCD. More importantly, the interface surfaces clearly overlap, consistent with the previous findings25, 26. Thus the MinE contact helix would compete with MinC for binding to MinD, which explains why MinE could stimulate the release of MinC.

Figure 2
figure 2

Structure of assembled EcMinCD complex and its superimposition with AaMinCD complex and EcMinDE complex. (A) Assembled EcMinCD complex using EcMinCCTD and EcMinD superimposed with AaMinCD (green: EcMinD, yellow: EcMinCCTD). The highly-conserved residues D154 in EcMinD and RSGQ motif in EcMinC are known to be critical for MinC-MinD interaction. (B) Superimposition of EcMinCD with AaMinCD (green: EcMinD, yellow: EcMinCCTD, blue: AaMinD, orange: AaMinCCTD). Helix3 of AaMinCCTD forms interface 2 with helix 8 of MinD, which is absent in EcMinCD complex. (C) Superimposition of EcMinCD with EcMinDE (green: EcMinD, yellow: EcMinCCTD, magenta: EcMinE). We observe that the MinE contact helix competes with MinC for binding to MinD and the interface surfaces clearly overlap25, 26, which provides explanation of why MinE can stimulate the release of MinC.

Impact of Aahelix3 on cell morphology and cell growth

Next, we set to find out whether the helix 3 of AaMinC is crucial for MinC’s function. We fused the Aahelix3 to the C-terminus of EcMinC, resulting in a hybrid MinC (EcMinC-Aahelix3) to be used in the subsequent functional experiments. SDS-PAGE and western results showed that this hybrid protein expressed well (Fig. S2) and CD analysis revealed that it was properly folded compared to the original MinC (Fig. S3). In addition, MALLS and SEC results (Fig. S4) of MBP-EcMinC and MBP-EcMinC-Aahelix3 showed that both of these two proteins formed stable dimers, which is consistent with the fact that MinC forms dimer in the AaMinCD complex structure (Fig. 2B).

We anticipated that the addition of a 10-aa peptide at the C-terminal tail would have a functional consequence and alter EcMinC’s characteristics in cell division regulation. To test whether the additional Aahelix3 affects the cell morphology of E. coli cells in vivo, we carried out SEM experiments using strains containing EcMinC-Aahelix3 (Figs 3 and S5). As shown in the SEM images (Fig. 3A–D), the cell length of BL21(DE3) (wild-type or WT) strains was about 2.6 μm. MinCCTD overexpression strains were a bit longer compared to WT strains (~3.5 μm, Fig. 3C), suggesting that without MinCNTD, overproduction of MinCCTD could not inhibit the FtsZ filament formation at mid-cell. However, MinCCTD-helix3 strains exhibited much longer cell morphology than WT (>10 μm vs 2.6 μm), strongly suggesting that helix3 influenced the normal cell division in cells. To exclude the effect of endogenous MinC in cells, we carried out complementation experiment using MinC-KO strains and BW25113 (WT strain). As shown in Fig. 3E–H, MinC-KO cells were longer than BW25113 cells (~4 μm vs 1.5 μm). Consistent with our prediction, complementation of EcMinC to MinC-KO strains restored normal cell length similar to BW25113 cells. Surprisingly, complementation of EcMinC-Aahelix3 to MinC-KO strains could not help cells regain the ability of cell division regulation, in which the cells exhibited cell length of ~6 μm. Furthermore, we monitored the cell morphology of live cells of these corresponding strains through standard light micrographs and results were generally consistent to the SEM images (Fig. S6).

Figure 3
figure 3

Scanning electron micrographs of derivatives of MinC expression in E. coli BL21(DE3) or MinC-KO. (A) E. coli BL21(DE3) (WT); (B) EcMinC-Aahelix3-BL21(DE3); (C) EcMinCCTD-BL21(DE3); (D) EcMinCCTD-Aahelix3-BL21(DE3); (E) BW25113; (F) MinC-KO; (G) EcMinC-MinC-KO; (H) EcMinC-Aahelix3-MinC-KO. Cells of EcMinCCTD-Aahelix3 exhibited much longer cell morphology than WT (A~D) and complementation of MinC-helix3 to MinC-KO strains could not help cells regain the ability of cell division regulation (E~H), verifying that the Aahelix3 influenced the normal cell division in cells. Scale bars represent 5 μm.

To assess the influence of EcMinC-Aahelix3 on the activity of EcMinC, we compared growth curves of WT strain and strains overexpressing EcMinC or EcMinC-Aahelix3. Growth of the EcMinC overexpression strain was slower than WT (Fig. 4), consistent with the observation that overexpression of MinC inhibits cell division27. Nevertheless, growth of the EcMinC-Aahelix3 overexpression strain was slower than the EcMinC overexpression strain, suggesting that helix3 increased the inhibitory activity of EcMinC on cell division.

Figure 4
figure 4

Effect of Aahelix3 on cell growth. Growth curves of WT E. coli BL21(DE3) (blue), EcMinC strains (red) and EcMinC-Aahelix3 strains (green) under aerobic conditions at 37 °C. All experiments were completed in triplicate and performed twice. The standard error of the mean was used to calculate the error bars. For those points where experimental variations are too small, their error bars are not visible.

The helix3 increases the interaction between EcMinC and EcMinD

It has been reported that MinD forms copolymers with MinC in the presence of ATP6, 28. To further understand the protein interaction that occurs between MinC-helix3 and MinD, we studied complex formation in vitro using a co-pelleting assay. MinC and MinD were incubated in the absence of ATP, and then mixtures were fractionated by centrifugation. We observed that both EcMinDΔC10 and MBP-EcMinC, alone or together, were soluble and predominantly localized in the supernatant (Fig. 5). In addition, MBP-EcMinC-Aahelix3 alone was also mainly found in the supernatant. However, when MBP-EcMinC-Aahelix3 and EcMinDΔC10 were incubated together, the distribution of both proteins shifted to the pellet (Fig. 5). These results demonstrate that MinC-helix3 and MinD form large oligomers in vitro, implying that the helix3 increases the interaction between EcMinC and EcMinD. Furthermore, MBP-EcMinC and MBP-EcMinC-Aahelix3 were respectively used to interfere with the interaction between MinD and MinE using a competition experiment. Our results show that MBP-EcMinC is unable to replace MinE from the MinDE complex but MBP-EcMinC-Aahelix3 successfully competes with MinE (Fig. S7) because there was almost no MinE in the elution, again suggesting that the helix3 increases the EcMinC-EcMinD interaction.

Figure 5
figure 5

Formation of large complexes containing MinD and MinC. Mixtures containing combinations of EcMinDΔC10 (12 μM) and MBP-EcMinC (6 μM) or MBP-EcMinC-Aahelix3 (6 μM), where incubated at 25 °C for 15 min and then fractionated by centrifugation. Supernatants and pellets were analyzed by SDS–PAGE and Coomassie Blue staining. This figure is a cropped gel and the full-length gel is shown in Fig. S11.

Binding free energy calculation of protein complex

We next carried out binding free energy calculations to find out the relative interacting strength of AaMinCD, EcMinCD and EcMinDE. The 6 models we used are detailed in Methods. Results (Table 1) show that without helix3, the interaction between AaMinC-Δhelix3 and AaMinD was much weaker compared to AaMinCD (−45.86 kcal·mol−1 vs −85.54 kcal·mol−1). Moreover, interaction between EcMinC-Aahelix3 and EcMinD was stronger than EcMinCD (−41.25 kcal·mol−1 vs −31.94 kcal·mol−1), indicating that the helix3 strengthened the binding of MinCD. Intriguingly, the binding of EcMinC with EcMinD was weaker than EcMinD with EcMinE (−79.96 kcal·mol−1 vs −89.21 kcal·mol−1), which is consistent with the fact that EcMinE is able to compete with EcMinC from the EcMinCD complex and stimulate the release of EcMinC6.

Table 1 Binding free energy calculation of protein complex.

Sequence analysis of MinC and MinD

The difference in the C-terminal tail of AaMinC, TmMinC and EcMinC led us to speculate that there may be some evolutionary relationships between MinCs from different groups of organisms. Therefore, we performed sequence analysis of MinCs (Fig. 6A, 240–291 aa) from various organisms. Results show that the C-terminal tail of MinCs of several typical Gram-negative bacteria are almost identical and lack helix3. Moreover, MinCs from Thermotogae family are conserved, which indicates that they all have a flexible loop similar to TmMinC (Fig. 6A, 275–282 aa). Furthermore, there is a helix3 at AaMinC’s C-terminal tail. Based on the predicted secondary structure of MinC of Gram-positive B. subtilis (Fig. S8), we propose that there is also a helix at BsMinC’s C-terminal tail. The alignment of several MinCs from Gram-positive bacteria suggests that they all likely possess a helix at their C-terminus (Fig. 6A). Using the Neighbor-Joining method29, the evolutionary history is inferred and results show that the evolutionary relationships of MinC and MinD from different organisms are consistent with their sequence analysis (Figs S9, S10). To examine a possible connection amongst cell division patterns among different organisms, we have summarized the effects on cell morphology exhibited by several relevant proteins (Tables 2, Table S3). As seen in Table 2, in E. coli which belongs to typical Gram-negative bacteria, the cell division pattern is basically dependent on spatially regulation via min oscillation. T. maritima of Thermotogae and A. aeolicus of Aquificae are both members of Thermophile, whose living conditions are similar to each other at the high temperature of 95 °C30. However, in A. aeolicu, MinE is not present and its MinC has an extra helix3. Although there is also no MinE in B. subtilis, DivIVA plays a role in the inhibition of MinC, leading to cells undergoing a different division mechanism instead of min oscillation, which is a pole-anchored pattern31. With the help of DivIVA and MinJ, the MinCD complex remains concentrated in two polar regions, thereby preventing division from taking place near the poles. Nevertheless, neither MinC nor MinE is present in Archaea cells and the cell division pattern still remains unknown.

Figure 6
figure 6

Sequence alignment of MinC from different organisms (240–291 aa) and overall structures of superimposed dimeric AaMinCCTD, TmMinCCTD and EcMinCCTD. (A) The C-terminal tails of MinC of the typical Gram-negative bacteria are highly similar. However, The C-terminal tails of MinC from several bacteria belonging to Thermophile (such as A. aerolicu and T. maritima) and several typical Gram-positive bacteria (such as B. subtilis) are longer than the Gram-negative bacteria. Alignment was performed using the program ClustalX 2.1. (B) As seen from the structures, TmMinCCTD has a flexible loop at its C-terminal tail and AaMinCCTD has a helix3 at its C-terminal tail (blue: AaMinCCTD, green: TmMinCCTD, red: EcMinCCTD).

Table 2 Preliminary summary of cell division components from different organisms.

Discussion

The structure of full-length EcMinC has not been solved yet and only the N-terminal domain of EcMinC has been determined32. To solve the EcMinC structure, we attempted to crystallize full-length EcMinC. However in the crystal structure only the CTD of EcMinC was observed, consistent with the SDS-PAGE result. However, without MinCNTD which possibly promotes protein folding and solubility, MinCCTD construct alone was very unstable and readily precipitated (data not shown). In the TmMinC structure33, only the C-terminal domain, but not the N-terminal domain was involved in dimer formation. In comparison, both domains of EcMinC are engaged in dimerization32 (this work), which hints that the dimer interface of TmMinC is weaker than that in EcMinC.

Ghosal et al. determined the complex structure of AaMinCD and discovered that MinC and MinD together form a new class of nucleotide-dependent, alternating copolymeric filaments6. A subsequent publication from the Lutkenhaus group casts reasonable doubt on the physiological relevance of these MinCD copolymers7. As seen from the structure of the AaMinCD complex, helix3 at the C-terminal tail of AaMinC and helix 8 of AaMinD form interface 2, one of the two distinct interaction interfaces between the two molecules. However, this helix3 is absent in both EcMinCCTD and TmMinCCTD. A. aeolicus is an unusual group of thermophilic bacteria, which can be found near underwater volcanoes or hot springs30. As seen in the secondary structure of MinC in different organisms (Fig. 6A), most typical Gram-negative bacteria do not have helix3 at the C-terminal tail of MinC. However, there is a helix3 at the C-terminus of AaMinC, which belongs to the thermophilic bacteria. Furthermore, there is a coil at the C-terminal tail of the MinC in several other thermophilic bacteria including T. maritima (Fig. 6). The distinct structural features in different species indicate a role of helix3 (or lack of) in different cell division styles and evolutionary adaptation. The structure of AaMinCD complex shows that the helix3 of AaMinC forms an extra interacting interface with MinD. We speculate that in addition to its implication in cell division, this extra interaction may help stabilize the protein structure and enable A. aerolicu to withstand extreme environmental conditions such as high temperature.

It is widely accepted that MinC is a division inhibitor which is able to prevent Z ring formation and causes filamentation when overproduced even in the absence of MinD and MinE34. The C-terminal domain of MinC interacts with MinD, while its N-terminal domain directly inhibits FtsZ filamentation35. Our SEM experiments show that EcMinCCTD-helix3 strains exhibited longer cells than EcMinCCTD strains (Fig. 3). Moreover, our cell growth curves demonstrate that the Aahelix3 increased MinC’s inhibitory activity in vivo (Fig. 4). Based on the literature and our findings, we propose that the fusion of helix3 to EcMinC leads to disruption of cell division because there exists stronger interaction in MinC-helix3/MinD complex so that MinE cannot trigger the release of MinC. Consistently, the co-pelleting results show that MinC-helix3 could form large complexes with MinD in vitro (Fig. 5). SEM images (Fig. 3) reveal that complementation of EcMinC could rescue the loss of cell division inhibition in MinC-KO strain, while complementation of EcMinC-Aahelix3 could not rescue the inhibitory ability. When MinC was transformed into MinC-KO, the induced MinC helped cells to undergo normal cell division. However, when EcMinC-Aahelix3 was transformed into MinC-KO, the helix3 of MinC interfered MinE’s function of triggering the release of MinC, hence impairing cells’ normal cell division.

In E. coli, MinC and MinD undergo a rapid pole-to-pole oscillation in response to MinE, which causes a periodic block of the polar division sites36, 37. MinE is reported to bind to MinCD to trigger the release of MinC and stimulates the ATPase activity of MinD4. Afterwards, MinD is dissociated from the membrane, thus the central division site is protected from inhibition by MinCD. The superposition of EcMinDE and EcMinCD structures reveals that the MinE contact helix would compete with MinC for binding to MinD since the interface surfaces clearly overlap (Fig. 2C)6, which is consistent with previous findings25, 26. In addition, our binding free energy calculation of EcMinCD and EcMinDE (Table 1) reveals that there is a stronger interaction of EcMinDE than EcMinCD complex, leading to our speculation that the interaction of EcMinCD complex is weaker than AaMinCD due to solely one interaction interface, which is consistent with our negative results of pull-down experiments using EcMinC and EcMinD in vitro (data not shown). Only in this case is EcMinE able to replace MinC from the MinCD complex and stimulate MinC’s dissociation from the membrane. As shown in Table 2, in Gram-negative bacteria such as E. coli, there is no helix3 at the C-terminal tail of EcMinC and EcMinE is present. However, in Gram-positive bacteria, DivIVA, instead of MinE, plays a role in inhibition of MinCD38 and cell undergoes a different division mechanism. There is no MinE and DivIVA in A. aerolicu, we speculate that there also exists no min oscillation in A. aerolicu cell cycle due to the intimate interaction between AaMinCD. Moreover, in Archaea we find that both MinC and MinE are lost and cells would undergo a totally different, although yet unknown, division mechanism.

In summary, in this study we determined the crystal structure of the dimeric C-terminal domain of EcMinC. Through the structural alignment of AaMinCCTD, our EcMinCCTD and TmMinCCTD, we were intrigued by the fact that there exists an extra helix3 at AaMinCCTD’s C-terminal tail, which forms a tight interaction with MinD. SEM experiments and cell growth assays demonstrate that Aahelix3 fused at the C-terminus of EcMinC interfered normal cell division. Combined with results of binding free energy calculation and co-pelleting assay, we conclude that the interaction of AaMinCD complex is stronger than EcMinCD. Sequence analysis of MinC and MinD in different organisms and comparison of cell division pattern with B. subtilis have enabled us to propose that T. maritima, A. aerolicu and Archaea undergo different cell division cycles. The interaction between MinE and helix3 of MinC (or lack thereof) in different organisms provides a clue for explaining the different mechanisms in cell divisions among Bacteria and Archaea, although the detailed evolutionary changes in cell division pattern and mechanism remain to be investigated.

Methods

Strain materials and plasmid construction

Standard methods were used for plasmid construction. The plasmids utilized in this study are listed in Table S1. The coding sequence for MinC was amplified from E. coli K12 and inserted into pET22b expression plasmid through NdeΙ and XhoΙ to express a soluble construct containing a 6 × His at the C-terminus. For SEM and cell growth experiments, in order to obtain fusion protein of EcMinC-Aahelix3 and EcMinCCTD-Aahelix3, the sequence of helix3 from AaMinCCTD was fused to the C-terminus of EcMinC/EcMinCCTD using two-step site-directed mutagenesis. The resulting sequence was inserted into pET22b through NdeΙ and XhoΙ. For constructs used in MinCD co-pelleting assay and competition experiment, genes encoding MBP-EcMinC or MBP-EcMinC-Aahelix3 were inserted into a modified pET-28b vector (mod-pET28b). The N-terminal His-tag in the original pET-28b vector was replaced by an MBP-tag through NcoΙ and NdeΙ and EcMinC/EcMinC-Aahelix3 were inserted through NdeΙ and XhoΙ. After insertion, EcMinC or EcMinC-Aahelix3 was expressed with MBP at its N-terminus and with or without His-tag at its C-terminus. EcMinD was truncated by 10 amino acids at the C-terminus (ΔC10) to increase solubility by removing the amphipathic helix (MTS) and inserted into pET22b through NdeΙ and XhoΙ to express MinDΔC10-6 × His protein. Gene encoding EcMinE was inserted into pET22b through NdeΙ and XhoΙ and a strep-tag was fused to the C-terminus of MinE through site-directed mutagenesis. Bacterial strain TOP10 was used for general cloning and plasmid maintenance. BL21(DE3) strain was used for protein expression, SEM and cell growth experiments. Strains of E. coli BW25113 (wild-type) and MinC-KO were obtained from E. coli Genetic Resources at Yale CGSC (http://cgsc.biology.yale.edu/KeioList.php) and used for SEM experiments.

Protein expression and purification

For protein preparation and crystallization trials, BL21(DE3) E. coli cells were transformed by plasmid of pET22b-MinC and cultivated in LB medium. Cells were induced at OD600 ~ 0.8 by 0.5 mM isopropy-β-D-thiogalactoside (IPTG) and cultured at 16 °C for 20 h and harvested for disruption by sonication in buffer containing 20 mM Tris-HCl, pH 7.5, 200 mM NaCl (buffer A). After centrifugation, the supernatant was mixed with 2 ml of Ni-agarose resin. Next, beads were washed with 50 ml buffer A containing 30 mM imidazole and 300 mM imidazole was used to elute protein. Eluted samples were pooled and subjected to a size-exclusion chromatography (HiLoad superdex16/60 S200, GE Healthcare) using an AKTA avant system (GE Healthcare) equilibrated with buffer containing 20 mM Tris-HCl, pH 8.0, 100 mM NaCl. Fractions containing the desired protein were pooled and concentrated for crystallization screening. For proteins used in co-pelleting assays, MBP-EcMinC, MBP-EcMinC-Aahelix3 and MinDΔC10 were expressed and purified in a similar way except that proteins were purified in buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl (buffer B). The purified proteins were concentrated and stored with 20% glycerol at −80 °C until use. Proteins were analyzed by SDS-PAGE and visualized by Coomassie Brilliant Blue (CBB) staining. For western blotting analysis of the EcMinC-Aahelix3 construct, eluted protein from Ni-agarose resin were subjected to SDS-polyacrylamide gel electrophoresis, transferred onto PVDF membrane and blocked for 1 h with TBST containing 5% milk. A mouse monoclonal anti-His-HRP conjugated antibody (Tianjin Sungene Biotech) was used for primary antibody and goat alkaline phosphatase-conjugated anti-mouse secondary antibody was used for protein detection.

Crystallization and X-Ray Diffraction

Soluble MinC protein (15 mg/ml) was crystallized by hanging-drop vapor diffusion method at 20 °C mixing 2 μL protein solution with 2 μL reservoir solution containing 0.1 M Tris-HCl, pH 7.2, 0.44 M sodium/potassium L-(+)-tartrate. Over time, MinC protein was degraded into two separate domains (MinCNTD and MinCCTD) and needle-shaped crystals of MinCCTD were grown over a period of 3–6 weeks. Crystals were cryoprotected in reservoir solution with 20–30% glycerol and flash frozen in liquid nitrogen. X-Ray diffraction data were collected at the beam line 13B1 of National Synchrotron Radiation Research Center (Hsinchu, Taiwan) and processed with HKL-300039.

Structure determination, refinement and analysis

The structure was determined by molecular replacement using Phaser 40 using the C-terminal domain structure of TmMinC (PDB: 1HF2) as a search model. Model building and refinement were performed using Coot 41 and phenix.refine 42, respectively. The figures were generated using PyMOL (http://www.pymol.org/). The atomic coordinates and structure factors have been deposited in the Protein Data Bank as entry 5XDM. The data collection and refinement statistics are summarized in Table S2.

Scanning electron microscopy (SEM)

For scanning electron microscopy sample preparation, overnight cultures of strains containing MinC derivatives were diluted to OD600 0.05 in fresh LB at 37 °C. Strains were induced after 4 h with 1 mM IPTG and cultured overnight at 16 °C under 150 rpm shaking. Cells were then harvested by centrifugation for 10 min at 5000 rpm. After wash three times using PBS (pH 7.4), cells were then dehydrated through 70% acetone solution and dropped onto foil which was glued onto a metal specimen holder after the cells had dried. The SEM images of cells were obtained using a SU8010 Scanning Electron Microscope (Hitachi). The cell length was measured (over 50 per each sample) for statistical analysis using Graphpad Prism 6 (Graphpad software, Inc).

Cell growth conditions

Different MinC strains were cultured in LB overnight and diluted to the same OD600. 1 mM IPTG was added to cultures of each strain. The cultures were then incubated on 96-well plates with shaking at 37 °C for 12 h and cell growth was monitored by taking OD600 measurements hourly. During measurements, the optical path length is approximately 5 mm. Each sample was assayed in triplicate.

Co-pelleting assay

For the MinCD co-pelleting assay, MBP-EcMinC, MBP-EcMinC-Aahelix3 and MinDΔC10 proteins were used. Proteins were purified as described before. Mixtures (50 μL) in assembly buffer (50 mM MES, pH 6.5, 100 mM KCl, 10 mM MgCl2) containing MinD (12 μM) and MinC (6 μM) were incubated for 15 min at 25 °C, then centrifuged at 14 000 × g for 30 min. After that, the supernatant and pellets were analyzed by SDS-PAGE. Control experiments were carried out in a similar way using individual protein.

Binding free energy calculation

After 5 ns MD simulation, stable models of protein complex were obtained. The molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method43,44,45 implemented in the AMBER14 package was employed to calculate the binding free energy of protein complex. The binding free energy was obtained through calculating the free energy differences of ligand, receptor, and their complex as follows:

$${\rm{\Delta }}{G}_{binding}={G}_{complex}-{G}_{ligand}-{G}_{receptor}$$

The 6 models (Table 1) we used were described as bellows: model 1: AaMinCD complex (PDB: 4V02); model 2: the helix3 of AaMinC in AaMinCD complex was eliminated to constitute AaMinC-Δhelix3/AaMinD complex; model 3: assembled EcMinCD complex using EcMinCCTD (this work) and EcMinD (PDB: 3Q9L) superimposed with AaMinCD; model 4: the Aahelix3 was added to the C-terminus of EcMinC to constitute EcMinC-Aahelix3/EcMinD complex. In models 1–4, the binding free energy of one chain of MinD with the other three chains in the complex was calculated. In model 5, EcMinDE complex (PDB: 3R9J), the binding free energy of two MinD chains with two MinE chains was calculated. In model 6, assembled EcMinCD complex (same as model 3), the binding free energy of two MinD chains with two MinC chains was calculated.