Plastid division represents a fundamental biological process essential for plant development; however, the molecular basis of symmetric plastid division is unclear. AtMinE1 plays a pivotal role in selection of the plastid division site in concert with AtMinD1. AtMinE1 localises to discrete foci in chloroplasts and interacts with AtMinD1, which shows a similar localisation pattern. Here, we investigate the importance of Min protein complex formation during the chloroplast division process. Dissection of the assembly of the Min protein complex and determination of the interdependency of complex assembly and localisation in planta allow us to present a model of the molecular basis of selection of the division site in plastids. Moreover, functional analysis of AtMinE1 in bacteria demonstrates the level of functional conservation and divergence of the plastidic MinE proteins.
The mechanisms used by cells or organelles to ensure that topologically defined differentiation events, such as division, occur at the correct place, are poorly understood. It is well-established that chloroplasts, in the same way as Escherichia coli, divide symmetrically by binary fission, to produce two approximately equal-sized daughter chloroplasts (Boasson et al., 1972); however, our understanding of the molecular basis of this is unclear.
Symmetric division of cells and organelles requires that the midpoint is identified and that the division machinery is subsequently assembled at this site. This process is well characterised in E. coli where the Min system, which consists of three proteins: MinC, MinD and MinE, works in concert with nucleoid occlusion to ensure symmetric cell division (Fig. 1) (de Boer et al., 1989). MinC is an inhibitor of cell division and is activated by MinD. However, the MinCD division inhibition complex lacks site specificity: that is, septation is prevented at all potential division sites unless MinE is also present. MinE suppresses the action of MinCD at midcell points but not at the cell poles. To achieve this, MinE harbours two separable domains: first, an anti-MinCD domain (AMD) that is necessary and sufficient for counteracting MinCD division inhibitor activity, and second, a topological specificity domain (TSD) that confers the ability of MinE to suppress the activity of MinCD specifically at the cell centre (Fig. 1) (Zhao et al., 1995; Pichoff et al., 1995; King et al., 1999).
Arabidopsis harbours homologues of the E. coli MinE and MinD proteins, although no MinC protein has been identified. Perturbation of the levels of the Arabidopsis AtMinE1 and AtMinD1 proteins leads to asymmetric chloroplast division, strongly suggesting the involvement of a plant Min-like system in selection of the division site (Colletti et al., 2000; Maple et al., 2002; Fujiwara et al., 2004). Furthermore, like the bacterial system, initial evidence indicates that AtMinE1 function is mediated through protein-protein interactions. AtMinE1 has been demonstrated to dimerise and to interact with AtMinD1 and ARC3 (Maple et al., 2005; Maple et al., 2007). The interaction of AtMinE1 and AtMinD1 is also required for stimulation of the ATPase activity of AtMinD1 (Aldridge and Møller, 2005). However, despite some functional similarities between the two systems, the absence of a MinC homologue in Arabidopsis, differences in the cation dependence of MinD on AtMinD1 ATPase activity (de Boer et al., 1991; Aldridge and Møller, 2005) and the evolution of plant-specific plastid division components (cf. Maple and Møller, 2007) all suggest that the plant Min proteins might have, in part, evolved novel functions and mechanisms.
MinE proteins are found in many different organisms, including bacteria, moss and higher plants. The availability of protein sequences from a diverse range of organisms provides a rich resource for the analysis of the functional conservation and divergence of AtMinE1. Because protein function is ultimately defined by amino acid composition, comparisons were made between MinE proteins from bacteria and plants to identify regions that are potentially functionally important. Although plant MinE proteins are, on average, only ∼20% conserved in comparison with the E. coli MinE protein, alignments reveal two areas of high homology, allowing the identification of the putative AtMinE1 AMD and TSD (supplementary material Fig. S1). The TSD of E. coli possesses a novel fold that comprises a long α-helix and a large antiparallel β-hairpin (King et al., 2000) and, strikingly, analysis of the secondary structure of AtMinE1 reveals that, despite the divergence in primary sequence, these structural elements are predicted to be conserved. Plant MinE proteins harbour N-terminal extensions that have been shown to be genuine plastid targeting signals (Itoh et al., 2001). Furthermore, plant MinE proteins harbour a C-terminal extension (CTE) unique to plastidic MinE proteins (supplementary material Fig. S1). The CTEs vary in length from 29 (Arabidopsis) to 60 (moss) amino acids in length.
To gain further insight into how the Min proteins in Arabidopsis act together during plastid division, we have analysed the function of the different domains of AtMinE1. Our detailed analysis of the AtMinE1-AtMinD1 protein complex enables us to propose a model of Min complex formation in plastids and to assess the level of functional conservation in comparison with E. coli Min proteins.
Formation of the Min complex in plastids
AtMinE1 and AtMinD1 have been shown to interact in the yeast two-hybrid system and in living chloroplasts, indicating that the two proteins function as part of a Min protein complex during chloroplast division (Maple et al., 2005). To investigate the assembly of the Min complex, we utilised the yeast two-hybrid system to determine the domains of AtMinE1 that mediate the interactions within the Min complex. The plastid division proteins were expressed as fusions to the Gal4 activation domain (AD) or the Gal4 DNA-binding domain (BD) from pGADT7 and pGBKT7, respectively (Clontech). Initially, a series of N- and C-terminal truncations of AtMinE1 were fused to the AD to generate constructs AD-AtMinE11-197, AD-AtMinE11-169, AD-AtMinE135-141, AD-AtMinE135-229 and AD-AtMinE1142-229. Yeast HF7c cells were co-transformed with each of the AD constructs and either pGBKT7 or BD-AtMinE1 or BD-AtMinD1 (Maple et al., 2005). The relative strength of the interactions was assayed by quantifying the growth of the yeast on synthetic drop-out media lacking tryptophan and leucine and synthetic drop-out media lacking tryptophan, leucine and histidine after four days (Maple et al., 2005). The ability of the full-length AtMinE1 and AtMinD1 proteins to interact was used as a positive control and AD-AtMinE1 was able to interact with both BD-AtMinE1 and BD-AtMinD1 (Maple et al., 2005) (Fig. 2A; supplementary material Fig. S2). Furthermore, this interaction was found not to be dependent on the first exon of AtMinE1 (amino acids 1-34) as AD-AtMinE135-229 was able to interact with both BD-AtMinE1 and BD-AtMinD1 (Fig. 2A; supplementary material Fig. S2).
Analysis of the interaction of BD-AtMinE1 with the N- and C-terminal truncations of AtMinE1 revealed that AD-AtMinE1142-229 but not AD-AtMinE135-141 was able to interact with the full-length AtMinE1 fusion. Furthermore AD-AtMinE11-197, which lacks the plant-specific CTE, and AD-AtMinE11-169 also interact with BD-AtMinE1, indicating that the minimal region required for the dimerisation of AtMinE1 is amino acids 142-169, within the predicted TSD. However, dimerisation of AtMinE1 in yeast is stronger between the full-length AtMinE1 fusion proteins, indicating that amino acids 170-229, or elements within this region, contribute to the stability of AtMinE1 dimerisation (Fig. 2A; supplementary material Fig. S2).
In contrast to the dimerisation of AtMinE1, BD-AtMinD1 was unable to interact with AD-AtMinE1142-229, but an interaction was detected with AD-AtMinE135-141. Consistent with this finding, BD-AtMinD1 also interacts with AD-AtMinE11-197, AD-AtMinE11-169 and AD-AtMinE135-229 (Fig. 2A; supplementary material Fig. S2). To verify that the observed interactions were not due to the GAL4 fragment, the interaction studies were repeated in the reverse orientations, with the truncations of AtMinE1 fused to the BD and coexpressed in yeast HF7c cells with AD, AD-E or AD-D. The same pattern of results was obtained (supplementary material Fig. S2). Furthermore, to confirm these interactions in planta, bimolecular fluorescence complementation (BiFC) assays were performed in living chloroplasts. The N-terminal domain of AtMinE1 (AtMinE11-141) and the C-terminal domain fused to the AtABC1 transit peptide (TP) (TP.AtMinE1142-229) were fused to YFP1-154 (NY) in the vector pWEN-NY (Maple et al., 2005). The cDNA fusion cassettes were then transferred to the binary vector pBA002 (Kost et al., 1998) to generate AtMinE11-141-NY and TP.AtMinE1142-229-NY. The resulting constructs were used for co-infiltration into tobacco leaf cells. The interaction between full-length AtMinE1 and AtMinD1 has been demonstrated in BiFC assays, using biolistic transformation of the proteins fused to NY and CY (YFP155-238) (Maple et al., 2005). Consistent with this result, infiltration of AtMinE1-NY or AtMinD1-NY with either AtMinE1-CY or AtMinD1-CY resulted in clear reconstitution of YFP indistinguishable from that observed when AtMinE1 or AtMinD1 is fused to full-length YFP (Fig. 2B). Coexpression of TP.AtMinE1142-229-NY and AtMinE1-CY resulted in the reconstitution of YFP, which was observed as fluorescent spots in chloroplasts (Fig. 2B). Similarly coexpression of AtMinE11-141-NY and AtMinD1-CY resulted in one or more spots of fluorescence, always in close proximity to the chloroplast envelope (Fig. 2B). Coexpression of TP.AtMinE1142-229-NY with AtMinD1-CY, or of AtMinE11-141-NY with AtMinE1-CY, generated no signal (Fig. 2B). These data establish that the N-terminal domain of AtMinE1 (amino acids 35-141) is required for its interaction with AtMinD1 and that the C-terminal domain of AtMinE1 (amino acids 142-229) mediates AtMinE1 dimerisation or oligomerisation. Negative controls confirmed that these interactions were specific, and coexpression of each cDNA fusion pair in the negative controls was confirmed by RT-PCR (Material and Methods; supplementary material Fig. S3).
Despite the low overall sequence conservation of the MinE proteins, amino acids 120-140 of AtMinE1 are >60% similar to the analogous region (amino acids 11-31) of the E. coli MinE protein (Fig. 3A). As this region mediates the interaction between the E. coli MinE and MinD proteins (Ma et al., 2003), the possibility that the interaction mechanism is conserved in plastids was investigated. The interaction between BD-AtMinE135-141 and AD-AtMinD1 was found to be strong (supplementary material Fig. S4) and was used to investigate the importance of the conserved region of the AMD. Initially, to test the hypothesis, two truncations of AtMinE1 were generated as N-terminal fusions to the BD: first, the highly conserved region was removed from AtMinE135-141 to generate BD-AtMinE135-117 and, second, the highly conserved region was added to AtMinE1142-229, to generate BD-AtMinE1117-229 (supplementary material Fig. S4). Yeast HF7c cells were co-transformed with each of these constructs and with AD-AtMinD1 or AD. Consistent with the idea that amino acids 118-141 of AtMinE1 are required for the interaction with AtMinD1, a positive interaction was detected between AD-AtMinD1 and BD-AtMinE1117-229, but no interaction was detected between AD-AtMinD1 and BD-AtMinE135-117. This result was further confirmed by BiFC assays in living chloroplasts: coexpression of AtMinD1-CY and TP.AtMinE1117-229-NY resulted in clear reconstitution of YFP in spots in the chloroplast stroma (Fig. 2B).
To test further the hypothesis that AtMinE135-141 is required for the interaction with AtMinD1, a series of point-mutation constructs was generated within the region using site-directed mutagenesis to substitute small nonpolar residues for large, polar residues and vice versa. Initially, a series of 11 constructs was generated in which amino acids on both faces of the putative α-helix, including the nine amino acids that are conserved from E. coli to Arabidopsis, were substituted (A124, I126, A127, K128, Q129, R130, L131, K132, I134, L135 and R139; Fig. 3A). Quantitative analysis revealed that mutations in L131, I134 and L135, all of which lie on one face of the α-helix (face 1; Fig. 3), abolished the interaction of AtMinE135-141 with AtMinD1, whereas substitution of A124, I126, A127, K128, Q129, R130, K132 and R139 had no effect (Fig. 3B; Table 1; supplementary material Fig. S5).
To corroborate these results, it was predicted that, if the correct region of AtMinE1 that interacts with AtMinD1 had been identified, substitution of another amino acid on face one of the AtMinE1120-140 α-helix would also eliminate the heterodimerisation with AtMinD1. Primers were designed to mutate aspartic acid (D) 138 to isoleucine (I) to generate AtMinE135-141,D138I. It was found that BD-AtMinE135-141,D138I is unable to interact with AD-AtMinD1 (Fig. 3B; Table 1; supplementary material Fig. S5). The possibility that the lack of interaction was due to unstable fusion proteins was ruled out as the point mutants generated could interact with ARC3, a stromal plastid division component that ineracts with AtMinE1 (Maple et al., 2007; data not shown). This result is consistent with the finding that the interaction between AtMinE1 and AtMinD1 is mediated by one face of the predicted AtMinE1120-140 α-helix and a complementary site within AtMinD1.
Dependency of AtMinE1 localisation on Min protein complex formation
AtMinE1 localises to discrete loci, close to one end of chloroplasts in both Arabidopsis and tobacco (Maple et al., 2002; Maple et al., 2005). To analyse the roles of the AtMinE1 domains on localisation, and additionally the dependency of the correct localisation on the formation of the Min complex, the C-terminal truncations of AtMinE1 (AtMinE11-197, AtMinE11-169 and AtMinE11-144) and the N-terminal truncations fused to the AtABC1 transit peptide (TP) (TP.AtMinE1117-229 and TP.AtMinE1142-229) were fused to YFP or CFP in the pWEN18 or pWEN15 vectors (Kost et al., 1998), respectively. The resulting fusions, AtMinE11-197-YFP, AtMinE11-169-YFP, AtMinE11-141-CFP, TP.AtMinE1117-229-YFP and TP.AtMinE1142-229-YFP, were each transiently expressed in tobacco and Arabidopsis by biolistic transformation.
Consistent with previous reports, AtMinE1-YFP (Maple et al., 2005) localises to discrete spots in both Arabidopsis and tobacco (Fig. 4). In striking contrast, AtMinE11-197-YFP in tobacco formed short, curved filaments in association with the chloroplast membrane (Fig. 4). Detailed analysis in Arabidopsis revealed that AtMinE11-197-YFP frequently formed long filaments encircling nearly the entire circumference of chloroplasts (Fig. 4), although, unlike the FtsZ Z-rings (Vitha et al., 2001; McAndrew et al., 2001), AtMinE11-197-YFP filaments never formed a complete ring-like structure. The ability of AtMinE1 to form filaments was abolished when the two predicted β-sheets of the TSD were removed as AtMinE11-169-YFP was observed as multiple spots (Fig. 4).
AtMinE11-141-CFP showed altered localisation and occurred throughout the stroma, often in association with the chloroplast inner membrane (Fig. 4). TP.AtMinE1142-229-YFP was also unable to localise correctly and instead localised throughout the stroma, never in association with the membrane. TP.AtMinE1142-229-YFP was also observed to localise to diffuse patches, forming large spots or stripes within the stroma (Fig. 4). These data indicate that either the interaction of AtMinE1 with AtMinD1 or the dimerisation of AtMinE1 alone is not sufficient for correct localisation of AtMinE1. To investigate the possibility that the correct localisation of AtMinE1 is dependent on the formation of the Min protein complex, the localisation pattern of TP.AtMinE1117-229-YFP, which harbours the determinants required for the interaction of AtMinE1 with both AtMinE1 and AtMinD1, was analysed. Strikingly, TP.AtMinE1117-229-YFP localises indistinguishably from full-length AtMinE1-YFP, forming one or two discrete spots in close association with the chloroplast envelope in both tobacco and Arabidopsis (Fig. 4).
Overexpression of AtMinE1 truncations
Overexpression of AtMinE1 leads to misplacement of the chloroplast division site (Maple et al., 2002). In order to investigate the effects of disrupting the interactions and localisation of AtMinE1 on the plastid division process, truncations of AtMinE1 were placed under the control of the cauliflower mosaic virus 35S promoter (35S) in pBA002 (Kost et al., 1998). The resulting transgenes, 35S-AtMinE1, 35S-AtMinE11-197, 35S-AtMinE11-169, 35S-AtMinE11-141, 35S-TP.AtMinE1117-229 and 35S-TP.AtMinE1142-229 (amino acid number indicated), were each transformed into Arabidopsis and primary transformants analysed by grid confocal microscopy. Analysis of the wild-type Arabidopsis lines revealed a population of equally sized chloroplasts in mesophyll cells (Fig. 5A). The division profiles of chloroplasts are visualised more readily in petiole cells, and all chloroplast division sites in these cells were symmetric (Fig. 5A, white arrowhead). Subsequently, rosette leaves of 26 lines expressing the 35S-AtMinE1 transgene were examined, where 13 lines showed heterogeneity in chloroplast size and number (Table 2). Analysis of the division profiles of chloroplasts in petiole cells of these lines revealed that this arose from the misplacement of the division site (Fig. 5A, white arrowhead; 35S-AtMinE1#10). Five lines expressing 35S-AtMinE1 (Fig. 5A; 35S-AtMinE1#23) contained greatly enlarged chloroplasts. Division profiles were examined in petioles showing that lines with greatly enlarged chloroplasts exhibited few or no division events, suggesting a complete arrest of chloroplast division.
Subsequently, the effects of expressing truncations of AtMinE1 were examined (Table 2). It was found that, in a proportion of lines expressing 35S-AtMinE11-197, 35S-AtMinE11-169, 35S-AtMinE11-141, 35S-TP.AtMinE1117-229 or 35S-TP.AtMinE1142-229, the chloroplast population in mesophyll cells was enlarged and heterogeneous in shape (Fig. 5A; Table 2). In each of these lines, the sites of chloroplast division were frequently observed towards the poles of the chloroplasts or multiple division sites along the length of the chloroplasts in petiole cells (Fig. 5A, white arrowhead). Moreover, in a number of lines expressing each transgene, each mesophyll cell contained just one to two greatly enlarged chloroplasts, and this again correlated with an absence of division sites in the petiole cells (Fig. 5A; Table 2).
Semi-quantitative RT-PCR revealed that, for all transgenes, a 2-6-fold increase in AtMinE1 transcript levels over wild-type levels was sufficient to disrupt the placement of the division site, resulting in misplacement of the division site and a population of heterogeneously sized chloroplasts (Fig. 5B). Furthermore, severe disruption of the division process was correlated with a more severe increase in AtMinE1 transcript levels (10-36-fold over wild-type; Fig. 5B). These data indicate that disruption of the plastid division machinery by overexpression of AtMinE1 can occur through interaction with either the endogenous AtMinE1 or AtMinD1, regardless of the localisation patterns displayed by the truncations (Fig. 4).
Conservation of the function of MinE
Analysis of the primary structures of plant MinE proteins indicates that there has been divergence from the bacterial Min proteins (supplementary material Fig. S1). However, the discovery that the mode of Min protein complex formation appears to be conserved raises the question of how conserved the function of AtMinE1 actually is. Overexpression of AtMinE1 in E. coli mimics the effects of overexpression of the endogenous MinE protein (de Boer et al., 1989; Maple et al., 2002), demonstrating that E. coli is a suitable model system to study the plant MinE proteins. As an initial step in assessing the degree to which AtMinE1 has retained its ancestral functions, we sought to determine whether AtMinE1 can complement an E. coli minE– strain.
The E. coli strain PB114(λDB173) [ΔminCDE(Plac::minCD)] lacks the chromosomal minB operon but is lysogenic for a λ derivative expressing the minC and minD genes under control of the lac promoter (de Boer et al., 1989). AtMinE1 was placed under the control of the constitutive PaadA promoter in the low-copy-number plasmid pGB2 (Churchwood et al., 1984). PB114(λDB173) was transformed with pGB2-AtMinE1, alongside the empty pGB2 vector as a control. Induction of MinCD with 0.05 mM IPTG in the host strain containing the empty pGB2 vector resulted in a complete arrest of division, as shown by the formation of nonseptate filamentous cells (Fig. 6A; supplementary material Table S1). By contrast, induction of MinCD in the presence of the PaadA-AtMinE1 plasmid released the inhibition of cell division, as shown by the decrease in average cell length and the presence of division septa (Fig. 6A; supplementary material Table S1). AtMinE1 did not do this in a topologically specific manner, as shown by the incidence of minicells (Fig. 6A, white arrows) and both symmetric and non-symmetric division events (Fig. 6A, black arrowheads and black arrows, respectively). To ascertain whether the `minicelling' phenotype is due to an imbalance in the levels of MinC, MinD and AtMinE1, the levels of MinC and MinD were adjusted by varying the levels of IPTG. At the lowest levels of IPTG (0.01-0.03 mM), minicells were still evident and the levels of MinCD were insufficient to block cell division and cause filamentation of the control strain (supplementary material Table S1). At higher levels of IPTG (0.2-4.0 mM), AtMinE1 still counteracted the MinCD-induced division block, but the degree to which this occurred became less dramatic, and the continued presence of minicells (Fig. 6A, white arrows) indicated that the release of the MinCD-induced division block still did not occur in a topologically specific manner (Fig. 6A; supplementary material Table S1).
Although the ability of Arabidopsis AtMinE1 to partially complement the MinE deficiency in E. coli is remarkable, the issue of why AtMinE1 cannot fully complement this strain suggests some evolutionary divergence in function. As a first step to investigate this, we tested whether the Arabidopsis and E. coli Min proteins could interact in the yeast two-hybrid system. The E. coli MinE and MinD cDNAs were cloned into the pGADT7 and pGBKT7 vectors to generate BD-MinD, BD-MinE, AD-MinD and AD-MinE. Yeast HF7c cells were cotransformed with each of these vectors and with the AtMinE1, AtMinE135-141, AtMinE1142-229 and AtMinD1 constructs. As a positive control, we analysed the interactions that mediate the formation of the Min protein complex in E. coli and, consistent with published results, found that the E. coli MinE protein can dimerise and can interact with MinD and that MinD can also dimerise (Table 3) (Huang et al., 1996). Next, the abilities of the E. coli and Arabidopsis proteins to interact were analysed. Surprisingly, no interactions were detected between the full-length AtMinE1 or AtMinD1 proteins with MinE or MinD (Table 3). However, a strong interaction was detected between the N-terminal domain of AtMinE135-141 and MinD and between AtMinE1142-229 and MinE. These data suggest that, although the determinants in AtMinE1 that mediate the interactions within the Min protein complex are conserved, there might be changes in the stability of these interactions (Table 3).
In E. coli, AtMinE1-GFP localises to the cell poles and at different points along the cells in an analogous way to that of the endogenous MinE protein (Maple et al., 2002). Complementary to the studies in plastids, the localisation of the AtMinE1 domain fusion proteins was analysed in E. coli. The AtMinE1 cDNA fusions to YFP or CFP were placed under the control of the T7 promoter in the pRSET-A vector (Invitrogen). E. coli BL21(DE3)pLysS cells were transformed with each vector alongside the empty pRSET-A vector or pRSET-A–YFP as controls and the fusion proteins induced using 0.05 mM IPTG. Cells containing the empty pRSET-A vector showed no fluorescence, and cells expressing YFP alone showed uniform fluorescence throughout the cell (Fig. 6B). As expected, the full-length AtMinE1-YFP and AtMinE1117-229-YFP fusion proteins localised to one or two spots at the poles of the cells (Fig. 6B). AtMinE11-197-YFP and AtMinE11-169-YFP were also able to localise in a similar way to that of the full-length AtMinE1-YFP fusion protein, forming discrete spots at the poles of the bacterial cells, although both were also observed to form multiple spots along the length of the cell (Fig. 6B). Similar to the results obtained in plastids, AtMinE11-141 localises throughout E. coli cells, although the resolution of the images is not sufficient to determine whether the fusion protein is membrane localised (Fig. 6B). Unlike the pattern observed in plants, the AtMinE1142-229-YFP fusion protein always localised to one or two spots, indicating that correct localisation of AtMinE1 in E. coli might not require interaction with the endogenous MinD protein (Fig. 6B).
Interestingly, it was observed that, when expression of AtMinE1-YFP was induced using 2 mM IPTG, analogous to overexpression of AtMinE1 and MinE, AtMinE1-YFP induced a minicelling phenotype in E. coli (de Boer et al., 1989; Maple et al., 2002) (Fig. 5C). In E. coli, overexpression of the MinE AMD or TSD also results in a minicelling phenotype due to the ability of the domain to counteract MinCD-mediated cell division inhibition or to interfere with selection of the division site, respectively (Zhang et al., 1998). Overexpression of AtMinE11-141-CFP resulted in a minicelling phenotype, with the appearance of small minicells (white arrows; Fig. 6C). By contrast, overexpression of AtMinE1142-229-YFP did not result in the obvious formation of minicells, and symmetric division sites were observed (black arrowhead; Fig. 6C). Overexpression of all other AtMinE1 truncations also resulted in a minicelling phenotype (Fig. 6C), suggesting that disruption of the E. coli division machinery by these truncations occurs though the interaction with MinD.
How chloroplasts place their division site to ensure symmetric chloroplast division is important in understanding how the chloroplast division process takes place. The misplacement of the division site in Arabidopsis plants with altered levels of AtMinE1 (Maple et al., 2002) or AtMinD1 (Colletti et al., 2000; Fujiwara et al., 2004) strongly implicates these two proteins in the process, but to date little is known about how they function to achieve symmetric chloroplast division. To begin to understand this process, we analysed the detailed function of AtMinE1.
Interactions mediating the formation of a Min protein complex in Arabidopsis
AtMinE1 and AtMinD1 form a complex in plastids, and disruption of this complex has been suggested to affect plastid division in Arabidopsis (Fujiwara et al., 2004). Initial investigation of the domains within AtMinE1 that mediate the formation of the Min complex revealed that the determinants required for dimerisation (AtMinE1142-229) and for the interaction with AtMinD1 (AtMinE135-141) are in two separable domains, harbouring the predicted TSD and AMD, respectively (Pichoff et al., 1995; Zhao et al., 1995; King et al., 1999).
In keeping with the bacterial model (Pichoff et al., 1995; Zhao et al., 1995; King et al., 1999), we show that AtMinE1142-229 is sufficient for AtMinE1 dimerisation (Fig. 2). The minimal region required for this interaction is within amino acids 141-169, but the strength of the interaction is stronger when the complete TSD and CTE are present. This could represent differences in the AtMinE1 dimer stability if determinants within amino acids 170-229 are involved in the interaction. In support of this, residues in the E. coli TSD that contribute to the hydrophobic core of the MinE TSD dimer (King et al., 2000) are conserved within AtMinE1 amino acids 156-195 (supplementary material Fig. S1).
In contrast to full-length AtMinE1, which interacts very weakly with AtMinD1 (Fig. 2) (Maple et al., 2005), AtMinE11-141 interacts strongly with AtMinD1 (supplementary material Fig. S4). This is similar to the interactions observed in E. coli (Huang et al., 1996; Ma et al., 2003) and could suggest that a conformational change in the full-length AtMinE1 protein is required in chloroplasts to unmask the AtMinD1-binding site. The interaction between AtMinE1 and AtMinD1 is dependent on a region of highly conserved amino acids in AtMinE1 (Ma et al., 2003) (amino acids 120-140; Fig. 3), and analysis of the effect of point mutations within AtMinE1124-139 on the interaction with AtMinD1 revealed that mutations in four amino acids (L131, I134, L135 and D138) abolished the interaction. The region of AtMinE1 required for the interaction with AtMinD1 or MinD is less extensive than the region identified in the E. coli MinE protein (Ma et al., 2003), which probably represents an evolutionary adaptation.
Because AtMinE1 and MinE interact with AtMinD1 through one face of an α-helix, we speculated that this site interacts with a corresponding surface on an α-helix on AtMinD1. The MinE-binding site on the MinD surface is located close to the nucleotide-binding site, and MinE competes directly for interaction with D152 with K11, one of two highly conserved lysines in the P-loop (a glycine-rich extended loop that forms part of the Walker A motif) (Zhou et al., 2005). Zhou and colleagues (Zhou et al., 2005) found that substitution of D152 in the highly conserved α-helix seven of MinD eliminated the interaction of MinD with MinE. Investigation of the effects of substituting the corresponding conserved residue in AtMinE1 (D213) had no affect on the interaction between AtMinD1 and AtMinE1 (data not shown), suggesting that the specific amino acids involved in the interaction of AtMinD1 with AtMinE1 might have changed. However, mutations of highly conserved lysine residues in the walker A motifs of AtMinD1 (K72) (Aldridge and Møller, 2005), E. coli MinD (K16) (de Boer et al., 1989) and Neisseria gonorrhea MinD (K16) (Ramirez-Arcos et al., 2002) result in loss of ATPase activity and, in the case of N. gonorrhea and Arabidopsis, this has also been demonstrated to be associated with disruption of the interaction between MinD and MinE (Ramirez-Arcos et al., 2002; Aldridge and Møller, 2005). The determination of the AtMinE1 binding site on AtMinD1 will be important for understanding the regulation of AtMinD1 ATPase activity.
Dependency of Min complex formation on AtMinE1 localisation
AtMinE1 localises to one or two discrete spots in close proximity to the chloroplast envelope (Maple et al., 2002; Maple et al., 2005). Analysis of the localisation of two separable domains of AtMinE1, sufficient for the interaction with AtMinD1 and AtMinE1 (AtMinE11-144 and TP.AtMinE1142-229, respectively), revealed that neither was able to localise in the same way as the full-length protein. AtMinE11-144 localised in close association with the chloroplast inner membrane (Figs 4, 7). Similarly, the E. coli AMD domain fused to GFP fails to accumulate in a ring structure and is evenly distributed along the periphery of cells (Raskin and de Boer, 1997). Interestingly, while the membrane association of the full-length E. coli MinE protein is dependent on the presence of functional MinD protein (Fu et al., 2001; Raskin and de Boer, 1997), MinE1-33-GFP expressed in the absence of MinD still associates with the periphery of the cells (Ma et al., 2003), raising the possibility that both MinE and AtMinE1 harbour a membrane-binding site that normally functions after MinD has delivered MinE to the membrane. Interestingly, AtMinE11-97-GFP (Itoh et al., 2001) localises throughout the chloroplast stroma, indicating that the AtMinE1 membrane association is dependent on determinants within amino acids 98-141 (Figs 4, 7). The explanation for this localisation pattern is unclear, but it is possible that, in the absence of the interaction with AtMinD1, the truncated AtMinE1 is able to interact with a currently unidentified protein required within the Min system.
Analysis of AtMinE11-141-CFP and TP.AtMinE1142-229-YFP localisation suggests that correct localisation of AtMinE1 is dependent on the interaction with itself and AtMinD1. TP.AtMinE1117-229-YFP, which harbours the determinants required for both AtMinE1 and AtMinD1 interactions (Fig. 2), localises indistinguishably from full-length AtMinE1. This decisively demonstrates that dimerisation of AtMinE1 and interaction with AtMinD1 are necessary and sufficient for the correct localisation of AtMinE1 and emphasises the importance of the formation of the Min protein complex during plastid division (Fig. 7).
Given that TP.AtMinE1142-229-YFP, which can interact with both the full-length AtMinE1 and AtMinD1 protein, exhibits a wild-type localisation pattern, the same could have been expected of AtMinE11-169-YFP and AtMinE11-197-YFP, which contain the determinants sufficient for the interaction with both AtMinE1 and AtMinD1 (Fig. 2). However, the AtMinE11-169-YFP fusion protein forms multiple speckles in chloroplasts and in E. coli, suggesting that the localisation pattern is influenced by the stability of the fusion protein. Strikingly, AtMinE11-197-YFP, which lacks the CTE, was found to form filaments and incomplete ring-like structures in chloroplasts (Fig. 4). MinD forms filaments in the presence of ATP and phospholipids in vitro (Hu et al., 2002; Suefuji et al., 2002), and, in E. coli, MinE and MinD are organised into extended spirals that wind around the cell (Shih et al., 2003). Our data suggest that AtMinE1 has the potential to form continuous filaments, and, if this is the case, it is expected that AtMinD1 will form similar structures. Recently, it has been shown that AtMinE1 can interact with ARC3 (Maple et al., 2007), and it is exciting to speculate that, as part of AtMinE1 function, AtMinE1 and ARC3 act together in a ring-like structure during chloroplast division. Whether they do function this way will merit further study.
Conservation and divergence of AtMinE1
The mode of function of AtMinE1 is highly conserved with that of the bacterial MinE proteins, as not only overexpression of AtMinE1 induces minicelling in E. coli but AtMinE1 also partially complements MinE deficiency (de Boer et al., 1989). These data raise the outstanding question as to whether the mechanism of Z-ring positioning by the Min proteins is identical in bacteria and plastids or whether differences have evolved.
The determinants required for the interactions within the Min complex are highly conserved in AtMinE1 from the E. coli mechanism of complex assembly (Huang et al., 1996; King et al., 1999; King et al., 2000; Ma et al., 2003) (Figs 2, 3). This is further supported by the finding that interactions were detected between AtMinE135-141 and MinD, and between AtMinE1142-229 and MinE (Table 3). The lack of interaction between the full-length Arabidopsis and E. coli Min proteins might reflect an evolutionary change in the stability of the interactions or the overall 3D topology of the complex and could in part explain why AtMinE1 cannot fully complement the E. coli minE mutant. MinE and AtMinE1 both localise to the poles of E. coli cells and show dynamic localisation patterns (Fig. 6B) (Hale et al., 2001; Maple et al., 2002). Removal of the CTE makes AtMinE11-197 more like prokaryotic MinE at the secondary-structure level and it is therefore not surprising that AtMinE11-197-YFP localises in a similar way to AtMinE1-YFP or MinE-GFP in E. coli (Fig. 6C).
The results of the complementation assay suggest that AtMinE1 has retained its ability to counteract the MinCD cell-division inhibitor complex but not in a topologically defined manner. Overexpression of the AMD or TSD in wild-type E. coli leads to minicelling as a result of the ability of the domains to interact with MinD and interfere with selection of the division site, respectively (Zhao et al., 1995). Analysis of the ability of the truncations of AtMinE1 to affect cell division revealed that only AtMinE1142-229 did not result in asymmetric cell division. This suggests that the Arabidopsis TSD is not fully functional in E. coli, possibly because two amino acid residues (D45 and V49) crucial for the topological specificity function of MinE are poorly conserved in the plant MinE proteins (supplementary material Fig. S1) (King et al., 2000).
Our results demonstrate that the determinants required for the formation of the Min protein complex and the interdependency on localisation are highly conserved. However, the differences in domain structure and divergence in amino acid sequence could reflect the adaptation of the Min complex to function within the plastid environment and the replacement of MinC with novel protein(s), potentially including ARC3, with which AtMinE1 and AtMinD1 both interact (Maple et al., 2007).
Conclusions and future studies
AtMinE1 and AtMinD1 are involved in selection of the division site, and the functional analysis of Min complex assembly and localisation allows us to present a molecular model of selection of the division site in plastids. AtMinD1 associates with the chloroplast envelope through a C-terminal α-helix (Fujiwara et al., 2004). AtMinE1 dimerises through the TSD, and these dimers are recruited to AtMinD1 through a direct interaction of a conserved α-helix in the AMD with a complementary binding site on AtMinD1. The Min complex is visualised as discrete foci close to the chloroplast envelope, and the interaction of AtMinE1 stimulates AtMinD1 ATPase activity (Aldridge and Møller, 2005), possibly leading to the detachment and relocation of the Min complex. The next challenge will be to establish the relationship of ARC3 to the Min complex, the molecular basis of Min protein relocation and the possibility that the complex moves along a coiled scaffold within the plastids.
Materials and Methods
Arabidopsis cv. Colombia and Nicotiana tabacum cv. Samsun were used for all experiments and propagated as described previously (Maple et al., 2002).
The AtABC1 chloroplast transit peptide (amino acids 1-63) (Møller et al., 2001) was fused to AtMinE1117-229 and AtMinE1142-229by SOEing PCR. Unless otherwise stated, all vectors were generated using the primers listed in supplementary material Table S3: cDNAs were amplified with the appropriate primer pair and cloned into pPCR-Script (Stratagene) before being subcloned into the destination vector. Subscripts in the names of plasmids indicate the protein products (e.g. pGBKT7/AtMinE11-141) or the mutant allele (e.g. pGBKT7/AtMinE1A124R). All vectors used in the study, and the genotypes, are listed in supplementary material Table S2.
Yeast two-hybrid analysis
Yeast HF7c cells were co-transformed with combinations of pGADT7 and pGBKT7 vectors according to the manufacturer's instructions (Clontech), followed by quantitative protein-protein interaction analysis (Maple et al., 2005).
Image capture and analysis
Fluorescence image acquisition was performed on a Nikon TE-2000U inverted fluorescence microscope and filters for CFP (exciter S436/10, emitter S470/30), YFP (exciter HQ500/20, emitter S535/30) and chlorophyll autofluorescence (exciter HQ630/30, emitter HQ680/40) (Chroma Technologies) equipped with a Hamamatsu Orca ER 1394 cooled CCD camera. Volocity II software (Improvision) was used to capture 0.5 μm Z-sections and to generate extended-focus images.
Bimolecular fluorescence complementation assays
Fusions of full-length cDNAs to the N- and C-terminal halves of YFP (NY and CY, respectively) were constructed in pWEN-NY and pWEN-CY (Maple et al., 2005) and transferred to the binary vector pBA002. Assays were performed by Agrobacterium co-infiltration, whereby cells transformed with each construct were individually grown to an OD600, mixed in a 1:1 ratio and infiltrated into the abaxial side of young tobacco leaves (Yang et al., 2000). Infiltrated regions of tobacco leaf were analysed after 48-72 hours and assays repeated in triplicate. As a negative control, the AtMinE1-NY fusions were co-infiltrated with the inner-membrane-associated protein GC1 fused to CY (GC1-CY; supplementary material Table S2) and no detectable fluorescence was generated. Coexpression of both of the NY and CY fusion cassettes in each negative sample was confirmed by RT-PCR. Total RNA samples were isolated using GenElute™ mammalian total RNA miniprep kit and treated with DNAse I (Sigma). First-strand cDNA was synthesised using 2 μg total RNA, primer 5′-(T)17(A/G/C)N-3′ and M-MLV RT (Promega). A twelfth of the RT reactions was used for 35-cycle PCR reaction using fusion-cassette-specific primers (supplementary material Fig. S3 and supplementary material Table S3).
Complementation of E. coli
AtMinE1 was cloned into pGB2 (Churchward et al., 1984), thereby placing AtMinE1 downstream of the constitutive PaadA promoter. This construct, alongside pGB2 as a control, was transformed into PB114 (λDB173) [ΔminCDE(Plac-minCD)] (de Boer et al., 1989). Overnight cultures were diluted 1:100 into LB medium; isopropyl 3-D-thiogalactoside (IPTG) was then added to 0.01-4.0 mM, unless otherwise stated. The cultures were grown at 37°C for a further 3 hours before microscopic analysis.
AtMinE1 truncations were fused to YFP or CFP in pWEN18 or pWEN15 (Kost et al., 1998), respectively. For localisation studies in Arabidopsis and tobacco, constructs were transfected into leaves by particle bombardment (Kost et al., 1998) and analysed after 48 hours. For localisation studies in E. coli, the fusion cassettes and YFP alone were placed under the control of the T7 promoter in pRSET-A (Invitrogen). BL21(DE3)pLysS E. coli cells transformed with these vectors were grown overnight in LB medium containing 1% glucose and ampicillin and chloramphenicol. After 16 hours, the cells were subcultured in fresh LB medium lacking glucose but containing antibiotics and 0.05 mM IPTG. After 2 hours of growth at 37°C, the cells were viewed by epifluorescence microscopy.
AtMinE1 truncations were placed under the control of the CaMV35S promoter in pBA002. Transgenic Arabidopsis plants were generated using the Agrobacterium-mediated floral-dip method (Clough and Bent, 1998) and primary transformants were selected on 15 μg/ml DL-phosphinotricin (Melford Laboratories). Semi-quantitative RT-PCR of AtMinE1 and actin expression was conducted using gene-specific primers (supplementary material Table S2) as described above, although a twelfth of the RT-PCR reactions was used for 18-cycle (AtMinE1142-229) or 20-cycle PCR reaction (actin and AtMinE11-141). Quantification of expression levels was performed using ImageJ (Abramoff et al., 2004).
This work was supported by the The Leverhulme Trust (F/00 212/M), The Functional Genomics (FUGE) Program/The Research Council of Norway and The EMBO Young Investigator Programme to S.G.M. We kindly thank Lawrence Rothfield, University of Connecticut, CT, for providing PB114(λDB173) [ΔminCDE(Plac::minCD)] and for helpful discussions.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/120/19/3446/DC1
- Accepted July 4, 2007.
- © The Company of Biologists Limited 2007