Chloroplast division site placement requires dimerization of the ARC11/AtMinD1 protein in Arabidopsis

Chloroplasts, photosynthetic organelles in plants, have evolved from an ancestral cyanobacterium harboring their own genomes and internal membrane systems related to the components of modern cyanobacteria (Gray, 1999). During evolution, the majority of prokaryotic genes were lost or transferred from chloroplasts to the eukaryotic host nuclei, adapting to eukaryotic gene expression systems and gaining cellular functions by targeting their gene products to chloroplasts or other subcellular compartments (Martin et al., 2002). However, not only cyanobacterial-derived but also a number of non-cyanobacterial-like proteins are thought to be targeted to chloroplasts (Martin et al., 2002), implying that modern chloroplasts are maintained by processes that require inherent prokaryotic and acquired eukaryotic systems.

Chloroplast division is an integral part of chloroplast development (Pyke, 1999), and chloroplasts divide from pre-existing plastids by binary fission (Leech and Pyke, 1988; Kuroiwa et al., 1998; Pyke, 1999; Hashimoto, 2003). It has recently been shown that chloroplast division represents a multifaceted process involving complex interdependency of both bacterial/cyanobacterial-derived and eukaryotic proteins (Hashimoto, 2003; Miyagishima et al., 2003; Gao et al., 2003).

Because of the involvement of bacterial/cyanobacterial-like proteins during chloroplast division, bacterial cell division has been used as a paradigm in dissecting the mechanism of chloroplast division. The bacterial-like FtsZ gene, AtFtsZ1-1, was the first nuclear-encoded chloroplast division component identified from Arabidopsis thaliana (Osteryoung and Vierling, 1995; Osteryoung et al., 1998), and most plants contain nuclear genes with high similarity to the bacterial cell division proteins FtsZ, MinD and MinE (Strepp et al., 1998; Osteryoung et al., 1998; Beech et al., 2000; Colletti et al., 2000; Itoh et al., 2001; Maple et al., 2002). Bacterial cell division is initiated by FtsZ, a cytoplasmic tubulin-like GTPase that assembles into a cytokinetic ring (Z-ring) at midcell in order to recruit other cell division proteins (Bi and Lutkenhaus, 1991; Erickson et al., 1996; Rothfield et al., 1999). Spatial regulation of Z-ring formation depends on the minB operon-encoded proteins, MinC, MinD and MinE, mutations of which result in asymmetrical cell division, formation of anucleate minicells and filamentation (de Boer et al., 1989). MinD is a membrane-associated ATPase and functions as an inhibitory cell division regulator by forming a complex with the division inhibitor MinC in the presence of ATP, thereby preventing FtsZ polymerization (de Boer et al., 1991; Huang and Lutkenhaus, 1996; Hu et al., 1999). MinCD activity is dynamically controlled by the topological specificity factor MinE. MinE stimulates MinD ATPase activity and subsequent membrane release, allowing MinCD oscillatory behavior, which ensures division initiation only at midcell (Raskin and de Boer, 1997; Raskin and de Boer, 1999; Hu and Lutkenhaus, 2001; Fu et al., 2001; Hale et al., 2001; Margolin, 2001; Lutkenhaus and Sundaramoorthy, 2003). It has recently been reported that MinD assembles into a dimer or polymer in the presence of ATP (Hu et al., 2002; Suefuji et al., 2003; Hu and Lutkenhaus, 2003), which is predicted to be important for its function (Lutkenhaus and Sundaramoorthy, 2003).

AtMinD1, a MinD homolog, was isolated from Arabidopsis and its role in chloroplast division was established by observations revealing that transgenic plants with reduced AtMinD1 levels show asymmetric chloroplast division, whereas elevated AtMinD1 levels show chloroplast division inhibition (Colletti et al., 2000; Kanamaru et al., 2000; Dinkins et al., 2001). More recently, the minE homolog AtMinE1 was identified in Arabidopsis (Itoh et al., 2001; Maple et al., 2002) and its role in chloroplast division site determination has been shown where AtMinE1 overexpression results in loss of topological specificity (Maple et al., 2002). Taken together, these studies, in addition to studies on plant FtsZ proteins, show that the prokaryotic part of the chloroplast division machinery is essential during the initiation stage of chloroplast division (Vitha et al., 2001; Kuroiwa et al., 2002; Miyagishima et al., 2003). In addition to Min protein homologs the bacterial/cyanobacterial origin of the division process has been further underlined by the identification of ARTEMIS (Fulgosi et al., 2002) and the Synechococcus Ftn2-like protein (Vitha et al., 2003). However, to date the underlying molecular mechanisms of chloroplast division remain unknown.

Twelve recessive accumulation and replication of chloroplasts (arc) mutants, showing altered number and morphology of chloroplasts in mesophyll cells, have been identified from Arabidopsis (Pyke, 1997; Pyke, 1999; Marrison et al., 1999). Among them, arc11 is unique in that it produces a reduced and heterogeneous population of chloroplasts in mesophyll cells (Marrison et al., 1999). arc11 was originally isolated from Ler transposon insertion lines and was mapped to Arabidopsis chromosome V, but was shown not to be associated with the transposon (Marrison et al., 1999).

In this study we begin to unravel the underlying molecular mechanism of chloroplast division site placement in Arabidopsis. We show that the arc11 mutant phenotype is derived from a single point mutation in the AtMinD1 gene and that arc11 chloroplasts show remarkable defects in division site placement, an observation previously not reported. We further reveal that AtMinD1 can form homodimers but that dimerization is abolished by the single point mutation found in AtMinD1. Moreover, the mutated ARC11 protein shows aberrant intraplastidic localization patterns, suggesting that AtMinD1 dimerization and appropriate localization is vital for correct chloroplast division in Arabidopsis.

Seeds of A. thaliana ecotypes Landsberg erecta (Ler) and Columbia (Col) and the arc11 mutant (Marrison et al., 1999) (Ler background, Ohio State University, USA) were surface-sterilized with 70% (v/v) ethanol and 1% (w/v) sodium hypochlorite, 0.2% (v/v) Tween-20, and sown on 0.8% (w/v) agar-containing Murashige-Skoog (MS) medium (Wako Jun-yaku, Japan) supplemented with Gamborg's B5 vitamins (Wako Jun-yaku) and 1% (w/v) sucrose. Seeds of Nicotiana tabacum cultivar bright yellow 4 were also surface-sterilized and sown on MS agar medium supplemented with B5 vitamins and 3% (w/v) sucrose. Plants were grown in plant growth incubators at 22°C for Arabidopsis and 28°C for tobacco under continuous white light illumination (100 μE/m²second).

Whole plant seedlings or organs were mounted under glass coverslips, and observed using confocal laser scanning microscopy (CLSM) (TCS-NT, Leica Microsystems, Germany) (Fujiwara and Yoshida, 2001) or epifluorescence microscopy (TE2000, Nikon, Japan, equipped with a Hamamatsu ORCA-ER; IX70, Olympus, Japan, equipped with an Olympus DP50-C). Digital images were imported into the RGB channels of Adobe Photoshop ver. 6.0 (Adobe Systems, USA).

Total DNA from Col, Ler and arc11 was extracted using the DNeasy Plant Mini Kit (Qiagen, Germany). The genomic copy of AtMinD1 was PCR amplified using PfuTurbo DNA polymerase (Stratagene, USA) and the oligonucleotide primers, MDF 5′-ACGCTCAGAAACATTTCTGTC-3′ and MDR 5′-CGTTCGGTTCGGTTCGATC-3′. The PCR product (1.1 kb) was subjected to direct DNA sequencing. The Monsanto Arabidopsis Landsberg Sequence (http://www.arabidopsis.org/) was utilized to compare AtMinD1 sequences between ecotypes.

Wild-type AtMinD1 was expressed, fused to the N-terminus of a double influenza hemagglutinin (dHA)-epitope tag, under the control of the AtMinD1 upstream genomic promoter region. Two complementary oligonucleotide primers, for a dHA epitope (YPYDVPDYAGYPYDVPDYAG), were annealed and introduced into the BamHI site of pBluescript II SK+ (Stratagene, USA) to yield pSK-HA. A 1.9 kb Col genomic DNA fragment, amplified by PCR with MD-HindIII 5′-GGAAGCTTTGGATATCTTGATC-3′ (restriction site underlined) and MD3 5′-ATGGATCCGCCAAAGAAAGAGAAGAAGC-3′, was digested with BamHI and HindIII and ligated into pSK-HA to yield pMD-HA. A 1.9 kb HindIII-SacI fragment of pMD-HA, comprising a 0.9 kb upstream genomic region of AtMinD1, the AtMinD1 ORF and a dHA-coding sequence, was introduced into the pBI-Hyg/35S-NosT vector (Yamaguchi et al., 1999) by simultaneously removing the CaMV35S promoter to yield pBIH-MD-HA. pBIH-MD-HA was employed for Agrobacterium-mediated Arabidopsis transformation by the floral dip method (Clough and Bent, 1998; Itoh et al., 2001). A total of 55 transformed (T₁) seedlings were selected on MS plates containing hygromycin (25 μg/l, Boehringer Mannheim, Germany), and T₂ to T₄ progenies were used for microscopic characterizations. Stable T₄ seedlings, grown on antibiotic-free MS plates were subjected to quantitative RT-PCR analysis.

Real-time RT-PCR was employed to quantify transcript levels of AtMinD1 in Ler, arc11 and transgenic plants (Shimada et al., 2003). Total RNA was extracted from 15-day-old seedlings using the guanidine-hydrochloride method, and complementary DNAs (cDNAs) were synthesized from DNase I-treated RNA with random primers using the Super Script First-strand Synthesis System (Invitrogen Life Technologies, USA) (Shimada et al., 2003). Quantitative RT-PCR was performed with a model 7700 sequence detector and a TaqMan Universal PCR Master Mix (Applied Biosystems, USA). Total AtMinD1 transcripts were monitored using AtMinD1-specific PCR primers, MD-FOR 5′-AGAGAGAGACCGACATTTGCG-3′ and MD-REV 5′-CGCGTATCGTCGTTATCACCT-3′, and a TaqMan probe MD-TAQ 5′-FAM (6-carboxyfluorescein)-TCGTCTTTCCAACACCGCCTTTTCCTAMRA-3′ (6-carboxytetramethylrhodamine), and endogenous AtMinD1 transcripts were monitored using AtMinD1 mRNA 3′-untranslated region (UTR)-specific primers MD-FOR2 5′-TCCTAGGACAATGTGGAATTCTACTG-3′ and MD-REV2 5′-CAGAAATCAAGAACCTCAAGAACAAA-3′, and a TaqMan probe MD-TAQ2 5′-FAMTGTTGCTGAGTTCAAGCTCTGATTCTTATGCCTAMRA-3′. The 18S ribosomal RNA was employed as an internal control.

The entire coding sequence of AtMinD1 was amplified by PCR from arc11 genomic DNA with MD-XbaI 5′-ATTTCTAGATCTGTGGAGACAGCTGAAG-3′ and MD-SacI 5′-TCCGAGCTCCATTTAGCCGCCAAAGAA-3′. A 1.0 kb PCR product was digested with XbaI and SacI and introduced into pBI-Hyg/35S-NosT under the control of the cauliflower mosaic virus 35S (CaMV35S) promoter to yield pBIH-arc11. pBIH-arc11 was employed for Agrobacterium-mediated Arabidopsis transformation. A total of 20 T₁ seedlings were selected on MS plates containing kanamycin (50 μg/l, Nacalai Tesque, Japan) or hygromycin (25 μg/l, Boehringer). Hypocotyls and primary leaves of T₂ and T₃ seedlings were used for microscopic characterizations. Stable T₃ seedlings were subjected to RT-PCR analysis.

Full-length AtMinD1 and AtMinD1(A296G) proteins were transiently expressed as green fluorescent protein (GFP) or yellow fluorescent protein (YFP) fusions in planta. pGFP2/AtMinD1 was used as an original vector (Maple et al., 2002). To construct pGFP2/AtMinD1(A296G), the entire coding region of AtMinD1(A296G) was amplified with MIN/1 and MIN/5 (Maple et al., 2002), digested with XhoI and KpnI, and ligated into pGFP2 (Kost et al., 1998). Expression vectors for YFP fusion proteins were made from the GFP expression vectors, by replacing the GFP coding sequence with that of YFP (EYFP, Clontech). To verify the N-terminal extensions of AtMinD1 for a chloroplast targeting transit peptide, the 64 amino acid-coding sequence of AtMinD1 was amplified by PCR using MD-SalI (Kanamaru et al., 2000) and MD-NcoI3 5′-TTCCCATGGTGATAACGACGATACG-3′. The PCR product was digested with SalI and NcoI and ligated into the CaMV35S-sGFP(S65T)-NOS (Niwa et al., 1999). The expression vectors were introduced into young tobacco leaf cells by particle bombardment. Over 30 cells were analyzed by epifluorescence microscopy using a Nikon TE2000.

Saccharomyces cerevisiae strain AH109 (MATa, trp1-901, leu2-3, 112, ura3-52, his3-200, gal4Δ, gal80Δ, LYS2::GAL1_UAS-GAL1_TATA-HIS3, GAL2_UAS-GAL2_TATA-ADE2, URA3::MEL1_UAS-MEL1_TATA-lacZ, MEL1) and plasmids pGADT7 and pGBKT7 encoding the Gal4 activation domain and the Gal4 DNA binding domain, respectively, were derived from MATCHMAKER Two-Hybrid System ver. 3 (Clontech Laboratories, USA). Full-length coding sequences of AtMinD1 and AtMinD1(A296G) were PCR amplified using MIND/5 5′-TACATATGGCGTCTCTGAGATTGTTC-3′ and MIND/7 5′-ATGGATCCTTAGCCGCCAAAGAAAGAGAAGAAGCC-3′, digested with NdeI and BamHI, and ligated into pGADT7 or pGBKT7. Yeast AH109 was cotransformed with pGADT7- and pGBKT7-derived constructs by electroporation. pGADT7-Rec and SV40 Large-T antigen PCR fragment and pGBKT7-53 (Clontech), and pGADT7 and pGBKT7 empty vectors, were also cotransformed as a positive and a negative control, respectively. A pGADT7-derived vector containing an Arabidopsis cDNA, which was isolated as a candidate for an AtMinD1 interacting protein in the two-hybrid screen (M.T.F. and S.G.M., unpublished), was cotransformed with pGBKT7-AtMinD1 as a weak positive control. Transformants were selected on yeast dropout (SD) media plates lacking leucine (Leu) and tryptophan (Trp) (Clontech), and fresh colonies were then streaked on SD plates lacking Leu, Trp and histidine (His). Cells were cultured on plates for 4 to 9 days at 26-30°C. Yeast growth was classified into four classes based on histidine auxotrophy from three independent experiments: +++, positive growth distinguished 2 days after culture; ++, 3 days; +, 5-6 days; –, normal background growth.

Expression vectors for AtMinD1-CFP (cyan fluorescent protein) and AtMinD1-YFP were constructed from pGFP2/AtMinD1, by replacing the GFP coding sequence with those of CFP (ECFP, Clontech) and YFP (EYFP, Clontech), respectively. For single and dual expression of AtMinD1-CFP and/or AtMinD1-YFP, plasmid DNA was introduced into tobacco leaf cells by particle bombardment as described above. FRET assays between fluorophores, CFP (donor) and YFP (acceptor), were performed with a Nikon TE2000, a Hamamatsu ORCA-ER cooled CCD camera and the OpenLab (Improvision, Coventry, UK) system software using 60× N.A. 1.4 objective lens (Nikon) and filters for CFP (exciter S436/10, emitter S470/30) and YFP (exciter HQ500/20, emitter S535/30) (Chroma Technologies, USA). For the acceptor photobleaching experiment, YFP was bleached by continuous illumination of excitation at the maximal strength for 20 seconds.

Chloroplasts undergo active membrane expansion and division during development from undifferentiated proplastids in meristems (Pyke et al., 1999). In higher plants chloroplast division is not synchronized, which is evident from the variety of chloroplast division states visible in various tissues and cells (Fig. 1A). To efficiently characterize division states of arc11 chloroplasts, we made use of hypocotyls and primary leaf petioles at an early stage of seedling development. Chlorophyll autofluorescence imaging by CLSM showed that a wide range of green tissues in arc11 contain a decreased number of mostly elongated and expanded chloroplasts with irregular shapes as compared with wild-type (WT, Ler) tissues, which contain round to ellipsoidal or dumbbell-shaped chloroplasts (Fig. 1A). More detailed single chloroplast imaging using higher magnifications revealed that the morphological abnormalities of arc11 chloroplasts are clearly linked to defects in division, as revealed by the existence of multiple constrictions (Fig. 1B; see also supplementary figure). Normal chloroplast division follows a coordinated pathway of events: slight envelope expansion, initial central constriction at a single site, further membrane constriction, and membrane fission as the final stage (see Fig. 1B, inset). By contrast, arc11 chloroplasts did not display a single central constriction but showed a varying number of constrictions from one to several, the most severe phenotype showing six constriction sites (Fig. 1B; see supplementary figure). Moreover, these constriction sites were placed randomly, but in parallel, along the long axis of chloroplasts (Fig. 1B). The arrayed chloroplast morphology indicates that membrane constriction in arc11 initiates and proceeds regardless of division initiation at other sites. The asymmetrical placement of constriction sites and the continuing membrane expansion results in a heterogeneous population of chloroplasts in terms of size and shape, including spherical mini-chloroplasts less than 2 μm in diameter (Fig. 1B, white arrowhead). Despite this, the envelope membrane still appears to grow with polarity during the division process. In more mature cells such as mesophyll cells, multiple arrayed chloroplasts could not be detected, whereas elongated and filamentous chloroplasts could be observed (data not shown) as described previously (Marrison et al., 1999).

It is clear that arc11 chloroplasts mislocalize the chloroplast division apparatus during division, a phenotype previously not reported. Furthermore, the observed phenotypes of dividing arc11 chloroplasts show significant resemblance to the min mutants (Bi and Lutkenhaus, 1993), in addition to the observed phenotype of AtMinE1 overexpressing plants (Maple et al., 2002).

It was previously reported that the asymmetric division of arc11 mesophyll chloroplasts showed similarity to the chloroplast morphology observed in Arabidopsis AtMinD1 antisense plants (Colletti et al., 2000). Furthermore, the arc11 locus was mapped close to the SSLP marker nga139 on Arabidopsis chromosome V (Marrison et al., 1999), which is in close proximity to the AtMinD1 locus (Fig. 2A). To examine our hypothesis that arc11 could be allelic to AtMinD1, we investigated the expression profile and the DNA sequence of AtMinD1 in arc11. Quantification of AtMinD1 transcripts using AtMinD1/18S rRNA ratio calculations indicated that the transcript abundance in Ler and arc11 seedlings is almost identical, with relative amounts of 1.40±0.14 (L er=1) (Fig. 3). However, AtMinD1 sequence analysis in Col, Ler and arc11 revealed that a single nucleotide substitution is present in AtMinD1 in arc11 (Fig. 2B). This cytosine to guanine nucleotide substitution results in a missense A296G mutation within α-helix 11 at the C-terminus of AtMinD1 (Fig. 2). Interestingly, alignment analysis of MinD amino acid sequences from bacteria to plants revealed that Ala 296 shares limited conservation to green plants. Moreover, the recent crystal structure and functional analyses of bacterial MinD proteins (Cordell and Löwe, 2001; Hayashi et al., 2001; Sakai et al., 2001) has revealed that Ala 296 is located distantly from crucial motifs and amino acid residues important for direct nucleotide binding and interaction with MinC. To date, no functional implications have been reported for α-helix 11 in any species.

To test whether the arc11 phenotype is indeed due to the A296G mutation in AtMinD1, we introduced a wild-type AtMinD1 transgene into the nuclear genome of arc11 plants. AtMinD1 was expressed as a fusion with a 21-residue double hemagglutinin (dHA) epitope tag at its C-terminus under the control of the AtMinD1 upstream genomic promoter sequence. In all, 55 T₁ plants were obtained, and chloroplast morphology and AtMinD1 expression levels of these plants were examined (Figs 3, 4). In many plants partial or full complementation phenotypes could be observed at the T₁ generation, which was associated with T-DNA insertions; however, several T₂ and T₃ progenies displayed different chloroplast phenotypes between individuals (Fig. 4F,G; data not shown). By observing hygromycin resistance (T-DNA marker) in segregated plants this observation was most probably due to changes in genomic T-DNA copy number resulting in disequilibration of the highly sensitive relationship between AtMinD1 expression levels and chloroplast division (Colletti et al., 2000; Dinkins et al., 2001). Further microscopic characterizations established three independent transgenic lines showing stable complementation phenotypes at T₃ and T₄ generations (Fig. 4C), whereas more than ten lines showed an apparent division inhibition phenotype (Fig. 4D). Using immunoblot analysis and the anti-HA antibody we confirmed that the transgene in these transgenic lines produced the expected sized protein (data not shown). Quantitative RT-PCR analyses of total and endogenous AtMinD1 transcript levels confirmed that transgenic plants showing successful complementation (lines 11HA38 and 11HA42; Fig. 4C) displayed similar (0.96 to 2.82 times to that of WT AtMinD1) AtMinD1 transgene expression levels to WT (Fig. 3). By contrast, the chloroplast division inhibition phenotype (lines 11HA2 and 11HA7; Fig. 4D) was associated with highly elevated transgene expression levels, consistent with previous observations (Fig. 3) (Colletti et al., 2000; Dinkins et al., 2001). Together, these findings show that arc11 is a loss-of-function mutant of AtMinD1 and that the observed phenotype is due to a single A296G mutation.

Interestingly, the observed chloroplast phenotype in partially complemented plants implied that slightly elevated expression levels of AtMinD1 have an effect on chloroplast envelope morphology (Fig. 4G,H). Several segregating individuals from original T₁ plant (line 11HA38) showed slight division inhibition and normal-to-larger chloroplasts, 6-9 μm in diameter, within cells (Fig. 4G). These chloroplasts displayed low heterogeneity in terms of size; however, the chloroplasts showed abnormalities in terms of envelope morphology, exhibiting distorted or surface-rugged outlines (Fig. 4G). Quantitative RT-PCR analysis for stable transgenic plants (line 11HA44; Fig. 4H) revealed that this phenotype is attributed to moderately enhanced AtMinD1 transgene expression (Fig. 3). This indicates that normal expression of AtMinD1 may be important not only for appropriate chloroplast division but also for correct envelope morphology.

MinD is an inhibitory division protein and its overexpression prevents division in Escherichia coli and chloroplasts (de Boer et al., 1989; Colletti et al., 2000; Kanamaru et al., 2000; Dinkins et al., 2001). If the A296G mutation perturbs AtMinD1 function, we would expect that AtMinD1(A296G) overproduction would not affect chloroplast division. To test this, we expressed AtMinD1(A296G) under the control of the CaMV35S promoter in WT plant (Fig. 5). A total of 20 CaMV35S-AtMinD1(A296G) transgenic plants were obtained and analyzed by microscopy and RT-PCR as described previously. From this analysis we found that the transgenic plants displayed division inhibition, producing a few but enlarged chloroplasts inside cells (Fig. 5B). In some cases the large chloroplasts seemed to be vacuolated, as observed in arc11 and arc11 transgenic plants (see Fig. 4D). Quantification of AtMinD1 transcript levels showed that endogenous AtMinD1 levels were not affected in the AtMinD1(A296G) transgenic plants, proving that the observed division inhibition is associated with AtMinD1(A296G) transgene expression. These results show that the AtMinD1(A296G) mutant protein has retained its division inhibition activity but has lost its ability to control appropriate placement of the division apparatus.

We previously reported that an AtMinD1-GFP fusion protein localizes to discrete speckle(s), which tend to localize as single spots at polar regions in ellipsoidal chloroplasts (Maple et al., 2002). To investigate the effect of the A296G mutation on intraplastidic behavior of AtMinD1, we expressed an AtMinD1(A296G)-YFP fusion protein in live plant cells (Fig. 6). As observed previously in transgenic plants, expression of a WT AtMinD1-GFP or AtMinD1-YFP fusion protein in leaf epidermal cells revealed discrete fluorescence signals to single stromal spots/speckles (Fig. 6; Table 1). By contrast, expression of an AtMinD1(A296G)-GFP or AtMinD1(A296G)-YFP fusion protein in leaf epidermal cells resulted in chloroplasts showing either large and distorted fluorescent aggregates and/or multiple speckles (Fig. 6; Table 1). These data show that the A296G mutation in ARC11/AtMinD1 results in abnormal AtMinD1 intraplastidic localization patterns and implies an important role for α-helix 11 in terms of correct AtMinD1 localization as part of the chloroplast division pathway.

It has been shown that bacterial MinD forms dimers or polymers in the presence of ATP on the membrane (Hu et al., 2002; Suefuji et al., 2003), and it is possible that the intraplastidic mislocalization of AtMinD1(A296G) inside chloroplasts is either due to loss of direct interaction with the envelope region itself or due to loss of protein-protein interaction capabilities. To gain further insights into this we examined protein-protein interactions of AtMinD1 using the yeast two-hybrid system (Fig. 7). Full-length WT AtMinD1 proteins were fused to the C-terminus of the Gal4 activation domain (AD-AtMinD1) and to the Gal4 DNA binding domain (BD-AtMinD1), respectively, and expressed in yeast AH109 cells. As a marker for protein-protein interactions we made use of the ability of AH109 to only grow in the absence of His on positive protein-protein interactions. We found that His auxotrophy was only restored in yeast cells cotransformed with both AD-AtMinD1 and BD-AtMinD1, showing that AtMinD1 can form homodimers (Fig. 7). The AtMinD1 interaction appears to be relatively weak and/or transient in yeast cells, which was further verified by the observation that individual yeast transformants exhibited slightly different growth rates on His selection plates (data not shown). Although the AtMinD1 interactions observed are relatively weak, repeated experiments have conclusively confirmed that AtMinD1 is capable of forming homodimers in vivo. In sharp contrast, His auxotrophy was not restored in yeast cells cotransformed with both AD-AtMinD1 and BD-AtMinD1(A296G), suggesting that the A296G mutation in ARC11 abolishes its ability to form homodimers (Fig. 7). This was further confirmed by the lack of His auxotrophy restoration in yeast cells cotransformed with both AD-AtMinD1(A296G) and BD-AtMinD1(A296G). Taken together, our results suggest that AtMinD1 is capable of forming homodimers in vivo and that the loss of homodimerization of AtMinD1 in arc11 results in intraplastidic mislocalization of AtMinD1, ultimately culminating in misplacement of the chloroplast division apparatus.

To corroborate our yeast two-hybrid results we analyzed AtMinD1 protein-protein interactions using Fluorescence Resonance Energy Transfer (FRET) technology (Gadella et al., 1999; Miyawaki and Tsien, 2000). FRET is an energy transfer process from a donor fluorophore to an acceptor fluorophore when donor and acceptor fluorophore are in close proximity (∼100 Å); it has recently been utilized to analyze live cell events such as calcium ion dynamics and protein-protein interactions in planta (Gadella et al., 1997; Allen et al., 1999; Mas et al., 2000). On the basis of their spectroscopic properties cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) represent an ideal FRET fluorophore pair in vivo: energy flow from the donor (CFP) to the acceptor (YFP) enables capture of YFP emission upon CFP excitation. We constructed two vectors expressing AtMinD1 as a fusion to CFP and YFP and transiently expressed these in tobacco leaf cells by particle bombardment. As expected, both fusion proteins, in their respective CFP and YFP fluorescent channels, showed a discrete bright speckle(s) within chloroplasts (Fig. 8A), as observed with AtMinD1-YFP (Fig. 6). In cells where AtMinD1-CFP and AtMinD1-YFP were co-expressed, a significant increase in fluorescence intensity was observed upon CFP excitation in the FRET channel (excitation filter 436/10; emission filter 535/30) as compared with the signal (bleedthrough) when using single fluorescent protein fusions (Fig. 8A), suggesting that FRET occurs from the donor (AtMinD1-CFP) to the acceptor (AtMinD1-YFP) in chloroplasts. To verify these findings we performed acceptor photobleaching experiments (Miyawaki and Tsien, 1999) by applying high-intensity YFP excitation for a short time period to minimize chloroplast damage (Fig. 8B). After photobleaching a clear increase of the donor fluorescence (CFP) was observed in the fluorescent regions within chloroplasts (Fig. 8B), showing a reduction in energy transfer, which is diagnostic of FRET. We detected no increase in AtMinD1-CFP fluorescence by this treatment in single fluorophore control experiments, but rather a fluorescence signal decrease as expected (data not shown). Thus, our data show that FRET occurred between AtMinD1-CFP and AtMinD1-YFP and that AtMinD1 proteins physically interact inside chloroplasts.

Although previous studies described asymmetrical, single constriction sites during chloroplast division events in mesophyll cells of arc11 and AtMinD1 antisense plants (Marrison et al., 1999; Colletti et al., 2000), we have shown here that hypocotyls and petioles of young arc11 seedlings contain mainly elongated and asymmetrically dividing multiple-arrayed chloroplasts (Fig. 1B; see supplementary figure). The observed differences between previous studies and our studies most probably reflect the use of different tissues at different stages of development. Multiple-arrayed chloroplasts have also been observed in hypocotyls of plants with elevated AtMinE1 levels (Maple et al., 2002), and combined, these results suggest that elongating young tissues represent an excellent source for the observation of dividing chloroplasts in live cells.

The asymmetrical division and mini-chloroplast phenotype of arc11 chloroplasts resembles that of the bacterial min phenotype, rather than the division-inhibited filamentous phenotype (de Boer et al., 1989). Our results expand and reinforce previous studies (Marrison et al., 1999; Colletti et al., 2000) which suggested that AtMinD1 represents a conserved chloroplast division component involved in the correct placement of the division machinery. The elongated morphology of arc11 chloroplasts might be explained by polar membrane growth during chloroplast division and subsequent multiple division site formations. It has been indicated through physiological studies on the arc mutants that the relationship between division and envelope membrane expansion is independent and compensatory when one process is inhibited (Pyke, 1997). Division arrest or delays might therefore cause extended polar growth of envelope membranes; however, the biogenesis of multiple-arrayed chloroplasts is an issue that deserves further study.

arc11 was successfully complemented by the expression of AtMinD1-dHA at approximately equal levels to endogenous AtMinD1 transcript levels observed in WT plants. Unexpectedly, we were only able to identify three transgenic plants showing a stable complementation phenotype. As most chloroplast division genes are known to affect chloroplast division as a result of their elevated or decreased expression levels, it is highly probable that regulatory elements, the chromosomal location and/or copy number of the AtMinD1-dHA T-DNA affect spatial and temporal expression patterns resulting in aberrant chloroplast division. Interestingly, observations of transgenic lines showing partial complementation and weak division inhibition suggested that regulated expression of AtMinD1 may also be important in maintaining envelope membrane morphology (Fig. 4G,H). In partially complemented arc11 plants the chloroplasts were relatively uniform in size, but on closer examination the chloroplast envelopes appeared both distorted and rugged on the surface (Fig. 4G,H) as compared with WT (Fig. 4E) and fully complemented lines (Fig. 4F). Although most known plants defective in chloroplast division display an abnormal chloroplast shape, the rugged chloroplast envelopes, as shown in Fig. 4G and 4H, were specific to the partially complemented arc11 plants: examination of chloroplasts from AtFtsZ and AtMinE1 transgenic plants and all the other 11 arc mutants did not reveal rugged and distorted chloroplast envelopes (unpublished). It appears therefore that AtMinD1 may affect the mechanism of controlling chloroplast envelope membrane morphology, either directly or indirectly, and that it is expression-level dependent.

The mutated amino acid residue in arc11, Ala 296, is a plant-specific conserved residue present in α-helix 11 (secondary elements are based on structural analyses of Pyrococcus MinD), which is absent in prokaryotic MinD proteins (Fig. 2B). α-helix 11 is close to the extreme C-terminal amphipathic helix, which is essential for membrane association of MinD in E. coli (Szeto et al., 2002; Hu and Lutkenhaus, 2003) (Fig. 2). Although no functional significance of α-helix 11 has to date been reported in either prokaryotic or eukaryotic MinD proteins, our results show that the A296G mutation in AtMinD1 causes aberrant localization inside chloroplasts. In contrast to WT AtMinD1, which shows distinct single polar localization patterns, AtMinD1(A296G) localizes to multiple spots/speckles throughout the stroma and to large abnormal spots when expressed in WT leaf cells (Fig. 6; Table 1). This indicates that a disruption of α-helix 11 renders AtMinD1 unable to localize appropriately. The A296G substitution in AtMinD1 may either lead to an overall conformational change of the entire protein or to a more localized conformational change only affecting the C-terminal part of AtMinD1. We favor the latter, as AtMinD1(A296G) has retained chloroplast division inhibitory activity (Fig. 5) and because the extreme C-terminal part of MinD, which is conserved from eubacteria to chloroplasts, is surface exposed and responsible for membrane localization through an amphipathic helix in bacteria (Szeto et al., 2002; Hu and Lutkenhaus, 2003). It is possible that the introduction of a Gly residue in α-helix 11 results in helix disruption due to the great rotational freedom of Gly residues which, in turn, leads to a direct loss of appropriate localization inside chloroplasts.

Recent studies have shown that dimerization or polymerization of MinD is strictly ATP-dependent and is vital for MinD mode of action (Hu et al., 2002; Suefuji et al., 2003; Hu and Lutkenhaus, 2003). We have shown here that AtMinD1 forms homodimers (Figs 7, 8) and that the A296G mutation results in the loss of AtMinD1 dimerization capacity (Fig. 7), which ultimately results in division site misplacement (Fig. 1). Although unlikely, due to the distant localization of A296 to the nucleotide binding site in AtMinD1, it is possible that ATP binding is affected by the A296G mutation in AtMinD1 resulting in loss of dimerization. Alternatively, and more probable, is that loss of dimerization capacity is due to a local C-terminal conformational disruption caused by the A296G substitution in arc11. This further suggests that the C-terminal region, and more specifically α-helix 11 of AtMinD1, is responsible for AtMinD1 dimerization. This is in slight contrast to E. coli MinD, which requires its nucleotide binding domain for dimerization purposes. However, E. coli MinD dimerization most probably also requires amino acid residues distant to the nucleotide binding domain. Despite this, it is possible that MinD in plants has acquired new dimerization properties from other ATPases different from that of classical prokaryotic MinD proteins during evolution generating an evolutionary MinD hybrid. For example, the NifH ATPase shows a good fit when superimposed on the E. coli MinD structure and NifH contains a C-terminal surface-exposed α-helical region, absent in E. coli MinD, involved in NifH dimerization (Lutkenhaus and Sundaramoorthy, 2003). It is possible that AtMinD1 represents an evolutionary protein hybrid containing both classical prokaryotic MinD domains in addition to domains acquired from other P-loop ATPases.

In prokaryotic systems it is not known whether MinD dimerization occurs after or before membrane association, and although we favor the hypothesis that AtMinD1 forms dimers before localization, it is possible that appropriate AtMinD1 localization occurs before dimerization. This is an area that requires further study.

We have shown that AtMinD1 displays functional dimerization (Figs 7, 8) and we suggest that AtMinD1 dimerization is important for correct intraplastidic localization patterns (Fig. 6). In E. coli it is proposed that MinD-mediated division site placement follows a coordinated sequence of events: MinD forms homodimers when bound to ATP; MinD then interacts with the FtsZ polymerization inhibitor MinC and moves to the membrane; MinE then interacts with the MinCD complex and stimulates ATP hydrolysis and membrane release (Lutkenhaus and Sundaramoorthy, 2003). Although no MinC protein has been identified in plants we propose that a MinC-like division inhibitor is probably present in Arabidopsis because overexpression of WT AtMinD1 in Arabidopsis results in chloroplast division inhibition (Colletti et al., 2000; Kanamaru et al., 2000; Dinkins et al., 2001) (Figs 3, 4), which is suggestive of excessive MinC-like protein activation and loss of FtsZ polymerization. On the basis of these data and our findings presented here, we propose a new working model for AtMinD1-mediated chloroplast division site placement in Arabidopsis (Fig. 9). In WT plants AtMinD1 forms dimers inside chloroplasts followed by recruitment and activation of a putative MinC-like protein. Following this the protein complex mediates appropriate placement of the division machinery, ensuring a single central constriction site by inhibiting FtsZ polymerization at inappropriate sites (Fig. 9A). In arc11 chloroplasts AtMinD1 is unable to form dimers (Fig. 9C) but retains its ability to interact with a putative division inhibitor protein as indicated by the loss of division site formation in arc11 overexpressing plants (Fig. 9B). This leads to mislocalization of the protein complex, ultimately resulting in division site misplacement and multiple constriction sites (Fig. 9C).

We have shown here that AtMinD1-mediated division site placement in Arabidopsis chloroplasts follows a coordinated sequence of events. Although a MinC-like protein remains to be identified in plants, this initial insight into the molecular machinery controlling division site placement in Arabidopsis chloroplasts has paved the way for new findings into the fundamental biological process of chloroplast division.

The authors thank Arabidopsis Biological Resource Center (Ohio State University, USA) and Joanne L. Marrison and Rachel M. Leech (University of York, UK) for providing the arc11 seeds, Masanobu Nakamura (University of Tokyo, Japan) and Akira Nagatani (Kyoto University, Japan) for the pBI-Hyg/35S-NosT vector, and Yasuo Niwa (University of Shizuoka, Japan) for the CaMV35S-sGFP(S65T)-nos vector. We also thank members of the Yoshida and the Møller laboratories for helpful discussions and technical assistance. This work was in part supported by Grants-in-Aid 14760068 from the Ministry of Education, Culture, Sports, Science and Technology of Japan to M.T.F., grants from the Special Postdoctoral Research Program of RIKEN to M.T.F. and R.I., and from BBSRC (91/C17189 and 91/REI18421), The Royal Society (574006.G503/23280/SM) and HEFCE (HIRF/046) to S.G.M.