Arabidopsis ATPase family gene 1-like protein 1 is a calmodulin-binding AAA+-ATPase with a dual localization in chloroplasts and mitochondria


  • Johanna Bussemer,

    1.  Department of Biology I, LMU Munich, Planegg-Martinsried, Germany
    Search for more papers by this author
  • Fatima Chigri,

    1.  Department of Biology I, LMU Munich, Planegg-Martinsried, Germany
    2.  Center for Integrated Protein Science (Munich) at the Department of Biology of the LMU Munich, Planegg-Martinsried, Germany
    Search for more papers by this author
  • Ute C. Vothknecht

    1.  Department of Biology I, LMU Munich, Planegg-Martinsried, Germany
    2.  Center for Integrated Protein Science (Munich) at the Department of Biology of the LMU Munich, Planegg-Martinsried, Germany
    Search for more papers by this author

U. Vothknecht, Department of Biology 1, Großhaderner Strasse 2-4, D-82152 Planegg-Martinsried, Germany
Fax: +49 89 2180 74661
Tel: +49 89 2180 74660


Members of the AAA+-ATPase superfamily (ATPases associated with various cellular activities) are found in all kingdoms of life and they are involved in very diverse cellular processes, including protein degradation, membrane fusion or cell division. The Arabidopsis genome encodes approximately 140 different proteins that are putative members of this superfamily, although the exact function of most of these proteins remains unknown. Using affinity chromatography on calmodulin-agarose with chloroplast proteins, we purified a 50 kDa protein encoded by AT4G30490 with similarity to the ATPase family gene 1 protein from yeast. Structural analysis showed that the protein possesses a single AAA-domain characteristic for members of the AAA+-ATPase superfamily and that this contains all features specific to proteins of the ATPase family gene 1-like subfamily. In vitro pull-down as well as cross-linking assays corroborate calcium-dependent binding of the protein to calmodulin. The calmodulin binding domain could be located to a region of 20 amino acids within the AAA-domain in close proximity to the Walker A motif. Our analysis further showed that the protein is localized in both mitochondria and chloroplasts, further supporting the incorporation of both endosymbiotic organelles into the calcium-signaling network of the cell. Localization of the same calmodulin-binding protein into mitochondria and chloroplasts could be a means to provide a coordinated regulation of processes in both organelles by calcium signals.

Structured digital abstract


ATPases associated with various cellular activities


ATPase family gene 1


Arabidopsis AFG1-like protein 1


calmodulin-binding domain


C-terminal domain


1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride


mature form of AFG1L1


N-terminal domain


network protein sequence analysis


N-ethylmaleimide sensitive fusion protein


precursor protein of AFG1L1


small subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase, Tic, translocon on the inner membrane of chloroplasts


voltage-dependent anion channel


AAA+-ATPases (ATPases associated with various cellular activities) are involved in a wide range of cellular processes [1,2]. These include functions that are evolutionary conserved in all kingdoms (i.e. protein degradation or DNA replication) as well as more specialized tasks such as membrane fusion or movement along the cytoskeleton [3]. AAA+-ATPases constitute a large and functionally diverse subfamily of ATPases containing a P-loop NTPase domain and are defined by a 200–250 amino acid long ATPase domain [1,2]. This so-called AAA-domain contains the Walker A and B motifs as well as more specific features for ATP binding and hydrolysis that distinguish these proteins from classical P-loop NTPases [4]. Another hallmark of AAA+-ATPases are the conformational changes that they undergo upon binding and/or hydrolysis of ATP, which are utilized to remodel their respective target substrates [5]. Moreover, almost all AAA+-ATPases assemble into catalytically active oligomers, often hexamers. Although they most likely share a common catalytic mechanism, they also contain a large number of additional domains (i.e. protease domains) besides the AAA-domain that account for the various activities of AAA+-ATPases [1].

AAA+-ATPases were initially defined by a set of proteins containing two AAA-domains but the family has subsequently been extended to include proteins that possess only a single AAA-domain. Proteins of this extended superfamily of AAA+-ATPases are conserved in prokaryotes and eukaryotes [2,6]. Most eukaryotic genomes contain approximately 50–80 potential AAA+-ATPases but the Arabidopsis genome encodes approximately 140 different proteins that are putative members of this superfamily [1,5]. This is the highest diversity found so far in any organism. Consequently, AAA+-ATPases have been identified in both chloroplasts and mitochondria and they are usually derived from the bacterial ancestor of these organelles [7–9]. Although some AAA+-ATPases are well characterized, the exact function of the majority of putative AAA+-ATPases remains unknown.

Many metabolic processes essential for plant viability take place in mitochondria and chloroplasts and thus their biogenesis and function has to be carefully balanced in accordance with the developmental stage and metabolic requirements of the cell. Therefore, both organelles are tightly integrated into the diverse regulatory networks of the cell. Calcium has long been acknowledged as one of the most important signaling components in plants and is crucial for the activation of environmental stress responses as well as for the regulation of developmental processes [10,11]. Many abiotic and biotic signals are transduced into a cellular response by temporal and spatial changes in calcium concentration [12,13]. One of the most important transducers of calcium signals is calmodulin, which comprises a relatively small and acidic protein that is ubiquitously found in eukaryotes [14]. Calmodulin possesses no catalytic activity of its own but contains a very flexible central α-helix that enables binding to a high variety of target proteins. Upon binding of calcium, calmodulin undergoes a conformational change that alters its affinity to downstream targets, thereby permitting a signal specific response. Experimental evidence indicates the existence of calcium/calmodulin regulation in chloroplasts and mitochondria [15–22]. Moreover, several proteins with potential calmodulin-binding sites have been identified in both organelles [9,16,20,23]. Nevertheless, the impact of calcium and calmodulin on mitochondria and chloroplast function is not well understood.

In the present study, we describe the identification of the Arabidopsis thaliana ATPase family gene 1 (AFG1)-like protein 1 (AFG1L1), which belongs to the extended superfamily of AAA+-ATPases. The protein binds calmodulin in a calcium-dependent fashion via a calmodulin-binding site within its catalytic AAA-domain. Moreover, AFG1L1 is dual-localized in chloroplasts and mitochondria, thereby incorporating both endosymbiotic organelles into the calcium-signaling network of the eukaryotic cell.


Identification of AFG1L1 as a new calmodulin-binding protein of chloroplasts

To identify new components of the chloroplast calcium regulation network, we performed affinity chromatography on calmodulin-agarose using proteins isolated from different Arabidopsis chloroplast subfractions. By using this approach, a 32 kDa subunit of the translocon on the inner membrane of chloroplasts (Tic) was identified previously as the predominant calmodulin-binding protein of the chloroplast inner envelope [23]. The application of a fraction containing extrinsic thylakoid membrane proteins resulted in the specific binding of a small number of proteins to calmodulin-agarose (Fig. 1A). The two most prominent proteins had a molecular mass of approximately 40 and 50 kDa, respectively, and these proteins were analyzed by MS. Several peptide masses obtained for the 50 kDa protein could be matched to the predicted protein sequence of AT4G30490 from Arabidopsis (Fig. 1B).

Figure 1.

 Purification of At4g30490 by affinity chromatography on calmodulin-agarose. (A) Silver staining of a protein fraction obtained by affinity chromatography of chloroplast proteins on calmodulin-agarose. The part of the gel that was used for peptide mapping is indicated. (B) Deduced amino acid sequence of At4g30490. Gray bars below the sequence indicate the peptides found by tandem MS that matched to this protein. Bold characters depict the part of the protein that belongs to the predicted AAA-domain. Typical motifs of AAA+-ATPases are indicated by black bars above the sequence. Some important residues are marked by asterisks. The triangles indicate potential cleavage sites for the transit peptide as predicted using different software (Table 1). (C) Graphical display of the secondary structure of At4g30490. The sequence of α-helices and β-strands comprising the AAA-domain are depicted in black. TP, transit peptide.

At4g30490 is annotated as a potential member of the Arabidopsis AAA+-ATPase superfamily (Fig. 1B, C). The protein shows strong similarity to the yeast AFG1 protein and thus most likely belongs to the AFG1-like subfamily [2]. To confirm this annotation, we performed a secondary structure prediction analysis using network protein sequence analysis (NPS@). Combining the results of the NPS@ prediction with the alignment provided by Iyer et al. [2] for the classification of AAA+-ATPases, we identified specific features of At4g30490. From approximately amino acid residues 110–390, our analysis revealed a sequence of α-helices and β-strands that is typical for the AAA-domain of this protein family (Fig. 1C). Furthermore, the sequence contains several motifs specific for members of the AAA+-ATPase superfamily, including the Walker A and B motifs, sensor 1 and 2, and the arginine finger (Fig. 1B). All in all, our analysis strongly supports At4g30490 as being a member of the AAA+-ATPase superfamily. Because of its similarity to the yeast AFG1-protein, we named the protein Arabidopsis AFG1-like protein 1 (AFG1L1).

AFG1L1 only contains a single AAA-domain (Fig. 1C). The C-terminus of the protein contains further α-helical and β-strand structures reminiscent of a degenerated second AAA-domain, although major sequence features are missing. No similarity to other domains in the database could be detected for this region of the protein, and thus we labeled it the C-terminal domain (CD). The protein furthermore contains an N-terminal domain (ND) prior to the α-0 helix of the AAA-domain. Such NDs are often found in AAA+-ATPases and are believed to be important for substrate interaction [6]. In the case of AFG1L1, the ND is preceded by approximately 60 amino acids whose sequence displays no similarity to other proteins in the database. Because the protein was isolated from a chloroplast protein subfraction, this part of the protein most likely represents the transit peptide.

AFG1L1 binds to calmodulin in a calcium-dependent manner

To verify the interaction between AFG1L1 and calmodulin, we performed affinity chromatography on calmodulin-agarose with recombinant AFG1L1. A construct was obtained comprising amino acids 60–497 of the deduced protein sequence of AT4G30490 fused to a C-terminal His6-tag (Fig. 2A). The protein lacks the first 60 amino acids of the full-length AFG1L1, which is approximately the length of the transit peptide as deduced by in vitro import experiments. AFG1L1 was heterologously expressed in Escherichia coli and purified by affinity chromatography on Ni2+-nitrilotriacetic acid agarose. The purified protein was used for pull-down assays with calmodulin-agarose. Under conditions that included calcium in all buffers, AFG1L1 bound to the column (Fig. 2B, upper panel) and could be eluted with an excess of calmodulin. This behavior is identical to that of Tic32 (Fig. 2B, third panel), a protein known to interact with calmodulin [23]. By contrast, the control protein SSU (small subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase) was recovered in the flow-through of the column (Fig. 2B, bottom panel). No binding of AFG1L1 to calmodulin-agarose was observed when calcium was substituted by EDTA/EGTA (Fig. 2B, second panel), indicating that the interaction of AFG1L1 with calmodulin is calcium-dependent.

Figure 2.

 Calcium-dependent interaction between AFG1L1 and calmodulin. (A) At4g30490 without the first 60 amino acids (AFG1L1) was expressed in E. coli. Therefore, the protein lacks the potential transit peptide (TP) as predicted from the import studies. A His6-tag was added for purification purposes. (B) Purified, recombinant AFG1L1, Tic32 and SSU were incubated with calmodulin-agarose in the presence of CaCl2 or EDTA/EGTA. Bound protein was eluted from the column matrix by an excess of calmodulin and the fractions were analyzed by Coomassie staining. L, load; FT, flow through; W, wash; E, eluate. (C) Recombinant AFG1L1 and exogenous calmodulin (CaM) were treated with the zero-Å cross-linker EDC in the presence of either CaCl2 or EDTA/EGTA. Cross-linking reactions were subsequently analyzed by immunodecoration with α-AFG1L1.

To further elucidate the interaction between AFG1L1 and calmodulin, we performed cross-linking experiments using the zero-Å cross-linker 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC) with recombinant AFG1L1 and commercially available calmodulin. The cross-linking reaction was performed in the presence and absence of calcium and subsequently analyzed by SDS/PAGE and western blotting using an antiserum raised against recombinant AFG1L1 (Fig. 2C). The antiserum clearly recognizes the recombinant AFG1L1 protein as indicated by the immunoreactive band at around 50 kDa that only occurs in the presence of AFG1L1. In the presence of calcium, a new immunodetectable band appeared when both cross-linker and calmodulin were present in the reaction (Fig. 2C, upper panel). The size of the cross-linking product is approximately 65 kDa and thus fits well with the binding of AFG1L1 to calmodulin, which has a molecular mass of 16 kDa. No cross-linking product occurred when calcium was substituted by EDTA/EGTA (Fig. 2C, lower panel) or in the absence of calmodulin (Fig. 2C, upper panel).

Binding of calmodulin to AFG1L1 occurs within the AAA-domain

To determine the calmodulin-binding domain (CaMBD) of AFG1L1, we created various constructs lacking different parts of the protein, which were subsequently tested in a pull-down assay on calmodulin-agarose as described above. When either the CD (AFG1L60–380) or ND (AFG1L103–497) of AFG1L1 was missing (Fig. 3A, panel 2 and 3), binding to calmodulin could still be observed, indicating that neither domain is responsible for the interaction. We then tested AFG1L1 constructs with further truncations from the N-terminus. Although a construct lacking the first 133 amino acids (AFG1L133–497) was still able to bind to the calmodulin matrix (Fig. 3A, panel 4), the construct starting at amino acid 159 (AFG1L159–497) showed no binding (Fig. 3A, panel 5). Taken together, these results imply that the CaMBD is located somewhere in the proximity of amino acids 133–159.

Figure 3.

 Identification of the CaMBD of AFG1L1. (A) Various recombinant AFG1L1 constructs were purified and subsequently incubated in equal amounts with calmodulin-agarose in the presence of calcium. After extensive washing, bound protein was eluted from the column matrix by an excess of calmodulin and fractions were analyzed by Coomassie staining. Index numbers indicate the amino acid residues in correlation to the AFG1L1 full-length protein. A corresponding schematic representation of the domain structure of full-length AFG1L1 and the various deletion constructs is given on the right. ND and CD are shown as white boxes, whereas the AAA-domain is indicated by a gray box. Calmodulin-binding as determined above is indicated. L, load; FT, flow through; W, wash; E, eluate (B) Graphical display of the secondary structure of AFG1L1. The red box depicts the localization of the CaMBD within the AAA-domain N-proximal to the Walker A motif. TP, transit peptide.

To identify the calmodulin-binding domain of AFG1L1 more precisely, we analyzed this region bioinformatically (Fig. 4). From amino acids 144–161, we could model an amphiphilic α-helix with a positive net charge of 3 (Fig. 4B), which is characteristic of many calmodulin binding proteins [24]. Furthermore, we were able to determine two established motifs for the conserved positions of hydrophobic amino acids within putative CaMBDs, named 1–10 and 1–14, respectively (Fig. 4A). Therefore, we tested a protein construct lacking amino acids residues 141–161 (AFG1LΔ141–161) in the calmodulin-agarose pull-down assay. We could detect no binding to the calmodulin matrix (Fig. 3A, panel 6), indicating that the calmodulin binding domain is indeed located within this region of the AFG1L1. Interestingly, these results place the CaMBD directly within the AAA-domain, just N-proximal to the Walker A motif (Fig. 3B).

Figure 4.

 Conservation of the CaMBD in AFG1-like proteins from higher plants. (A) Amino acid sequence of the deduced CaMBD of AFG1L1 from positions 144–161. Hydrophobic amino acid residues within the CaMBD sequence are indicated in bold. Two potential motifs (1–10 and 1–14) for calmodulin-binding are indicated below the sequence. (B) Helical wheel representation of the CaMBD of AFG1L1. The CaMBD forms an amphiphilic α-helix. Hydrophobic amino acid residues are indicated by gray circles. The line separates the hydrophilic and hydrophobic side of the helix. (C) Alignment of the amino acid sequence of AFG1L1 from A. thaliana (At3g30490) comprising the putative CaMBD and the adjacent Walker A motif with corresponding sequences of homologs from other plant organisms, such as P. trichocarpa (ABK95418), V. vinifera (CAN65875) and A. thaliana (AFG1L2, At4g28070) as well as nonplant organisms, such as Saccharomyces cerevisiae (AFG1, NP_010862), H. sapiens (EAW48377) and D. melanogaster (NP_610780). Amino acid residues identical to the AFG1L1 sequence are boxed in black, whereas conserved residues are boxed in gray.

Homology searches had revealed the presence of AFG1-like proteins in different organisms such as yeast, human and other plants, including a second homolog in Arabidopsis, which we called AFG1L2. By comparison of the calmodulin-binding domain of AFG1L1 with homologs from other plant and nonplant species, we could observe a high conservation of the CaMBD among the plant species such as Vitis vinifera and Populus trichocarpa (Fig. 4C). By contrast, no conservation was found within this region in case of nonplant species such as Homo sapiens and Drospohila melanogaster. Therefore, it is likely that the calmodulin-binding domain of AFG1L1 is a plant-specific trait that was added to the AFG1 family later in evolution. A partial conservation of the CaMBD could be observed for AFG1L2, although further studies are required to ensure whether this protein is capable of interacting with calmodulin.

AFG1L1 is dual-localized in mitochondria and chloroplasts

Because we originally purified AFG1L1 from chloroplasts, we further analyzed its protein sequence using bioinformatic tools that predict a possible subcellular localization from the primary amino acid sequence (Table 1). Interestingly, the majority of the prediction software available identifies a potential N-terminal targeting sequence but favors a localization of AFG1L1 in the mitochondrion rather than the chloroplast.

Table 1.   Evaluation of the intracellular localization of AFG1L1 using different prediction programs. The protein sequence of At4g30490 was submitted to different software for prediction of intracellular localization. Only subsequent output predicting an organellar localization is shown.
ProgramsPrediction of localizationProbability
(cleavage site)
  1. a Cut-off ≥ 6.21; b reliability class, from 1 to 5, where 1 indicates the strongest prediction; c probability; numbers in parenthesis indicate the predicted cleavage site.

ipsort X7.20a[40]
targetp XRC 3b (25)[41,42]
mitoprot ii X0.9337c (73)[43]
predotar X0.81c[44]
multilocX 0.39c[45]

In light of the subcellular targeting prediction for AFG1L1, we performed in vitro import experiments to verify the localization of AFG1L1 in either chloroplasts or mitochondria. Radioactively-labeled protein obtained by in vitro translation from full-length cDNA of AFG1L1 was used for import studies into isolated organelles from pea (Fig. 5). In vitro translation resulted in the formation of a 55 kDa protein, as expected for AFG1L1 with its N-terminal transit peptide (i.e. precursor protein of AFG1L1; pAFG1L1). After incubation of the translation product with isolated chloroplasts, a radioactive-labeled protein of approximately 50 kDa appeared, which is neither present in the translation reaction, nor in the 0 min control (Fig. 5A). This protein was furthermore protected from proteolysis, indicating that it represents the mature form of AFG1L1 (mAFG1L1) after import and removal of the transit peptide. mAFG1L1 first appeared after approximately 5 min and its amount increased in intensity over time (Fig. 5A). These results indicate that AFG1L1 can be imported into chloroplasts in a time-dependent manner. We also performed import assays using isolated pea leaf mitochondria (Fig. 5B). As before, the incubation of mitochondria with pAFG1L1 resulted in the appearance of a smaller 50 kDa protein in a time-dependent fashion. This protein was protected from protease treatment, indicating that pAFG1L1 is imported into mitochondria and cleaved into a smaller mature protein after translocation (Fig. 5B).

Figure 5.

 Dual localization of AFG1L1 in chloroplasts and mitochondria. 35S-labeled precursor protein of AFG1L1 (pAFG1L1) was imported into isolated chloroplasts (A) or mitochondria (B) from pea. Import reactions were carried out at 25 °C and the reaction was stopped after different times by re-isolation of the organelles. Chloroplasts and mitochondria corresponding to one half of each import reaction were treated with thermolysin, whereas the other half remained untreated. Samples were subsequently analyzed by SDS/PAGE and phosphoimaging. TL, one-tenth of the translation product used in the assay; mAFG1L1, mature form of AFG1L1; Th, thermolysin. (C) 35S-labeled pAFG1L1 was incubated with isolated chloroplasts (chloro) and mitochondria (mito) in a dual-import reaction for 20 min at 25 °C. Organelles were re-isolated after termination of the import reaction and analyzed separately. (D) Immunoblot analysis of isolated chloroplasts (chloro) and mitochondria (mito) from P. sativum and A. thaliana. Total organellar proteins were analyzed by immunodecoration with α-AFG1L1, as well as antisera against mitochondrial (VDAC) and chloroplast (Tic62) proteins. Tic62, 62 kDa subunit of the translocon on the inner membrane of chloroplasts; VDAC, voltage-dependent anion channel.

It has been observed that translocation into plant organelles is not always faithful in vitro if only a single organelle is present in the assay. This obstacle can be overcome by simultaneous import into isolated mitochondria and chloroplasts, termed a dual-import system [25]. We thus performed dual-import experiments using mitochondria and chloroplast in the same import reaction (Fig. 5C). After 20 min of import, both organelles were separated and analyzed independently. As before, we could observe the formation of protease-protected mAFG1L1 in chloroplasts as well as in mitochondria (Fig. 5C), indicating that the protein translocated into both organelles simultaneously. In mitochondria, a second protein band of a slightly smaller size than mAFG1L1 is visible (Fig. 5C, asterisk). This most likely presents a degradation product of mAFG1L1 that is formed after longer periods of time, as necessary, for the isolation of the mitochondria after the dual-import reaction.

To verify the dual localization of AFG1L1 in both, chloroplasts and mitochondria, we raised an antibody against AFG1L1. When used for immunodecoration of different cellular protein fractions from pea as well as Arabidopsis, the antibody clearly recognized a protein of the appropriate size in both mitochondria and chloroplasts (Fig. 5D, upper panels). Tests with antisera against chloroplast and mitochondrial proteins revealed no cross-contamination of chloroplasts with mitochondria and vice versa (Fig. 5D, lower panels). Thus, the immunoblot results strongly support the localization of AFG1L1 in both organelles.


Members of the superfamily of AAA+-ATPases are found in all kingdoms of life [2,6]. They are involved in very diverse cellular processes, including protein degradation, membrane fusion or cell division, but share a common mode of action via protein remodeling, often in a regulatory role. In the present study, we describe the identification of AFG1L1 from chloroplasts, a protein putatively annotated as a member of the AFG1-like subfamily of AAA+-ATPases [2]. Our secondary structure analysis of the deduced amino acid sequence clearly supports this annotation. Furthermore, the protein contains all sequence motifs and features that are characteristic of this family and shares significant sequence homology to AFG1 of yeast. AFG1 and AFG1L1 are members of the extended superfamily of AAA+-ATPases with distinct features not found in the classical clade of AAA-ATPases, most prominently the presence of only a single AAA-domain. Although AAA+-ATPases are common in both eukaryotic as well as prokaryotic organisms, all known chloroplast and mitochondrial members of this family are derived from the bacterial endosymbiont [2,6]. These proteins accomplish functions in the organelles that are identical or closely related to their original function in the bacterium. Phylogenetic analysis by Iyer et al. [2] suggests that the AFG1-like subfamily goes as far back as the origin of bacteria and that the AFG1-like proteins found in eukaryotes are derived from the α-proteobacterial ancestor of the mitochondrion. This would imply that AFG1L1 originally had a sole localization within this organelle. This is supported by the fact that BLAST searches reveal close homologs of AFG1L1 not only in other plants and Chlamydomonas, but also in human, fungi and other nonplant eukaryotes. More importantly, close homologs of AFG1L1 exist in many α-proteobacteria but not in cyanobacteria, making it unlikely that the protein was derived from the phylogenetic ancestor of the chloroplast. We thus believe that AFG1L1 was originally a mitochondrial protein and that the dual localization of AFG1L1 in chloroplasts and mitochondria is a newly-acquired secondary trait. Because no homolog exists in cyanobacteria, it is furthermore likely that the dual-targeted protein has obtained a new function required in both organelles. This is supported by the fact that the Arabidopsis genome contains a second homolog to AFG1 from yeast encoded by AT4G28070, which, as demonstrated by in vitro import experiments, appears to be localized solely in the mitochondrion (data not shown). This protein would be sufficient to provide the original function of AFG1 in plant mitochondria.

It appears that targeting of AFG1L1 into chloroplasts and mitochondria is mediated by the same, ambiguous transit peptide. This is demonstrated by import experiments as well as immunodecoration, both of which indicate that the mature AFG1L1 protein has an identical size in both organelles. The ambiguous nature of the transit peptide severely hampers a clear in silico determination of organellar localization. Most, but not all, programs favor a mitochondrial localization, providing further evidence that the mitochondrion was the original target for the transit peptide and chloroplast localization has been achieved by its alteration and adaptation. No evidence for further subcellular targeting within either organelle (i.e. by a bipartite presequence or internal sorting signal) could be obtained from the sequence analysis. Nevertheless, although AFG1L1 does not appear to be a membrane protein, a membrane association is likely. In the case of chloroplasts, this is supported by the fact that the protein was originally purified from an extrinsic thylakoid protein fraction, indicating that the protein can associate with this membrane on the stromal side. In the case of mitochondria, an association with the inner membrane from the matrix side is in accordance with the localization of AFG1 from yeast [26].

The Arabidopsis genome has revealed the largest set of AAA+-ATPases to date, indicating a strong requirement for the often regulatory role of these proteins [1]. The functional diversity of AAA+-ATPases is determined by small insertions into the AAA-domain or by additional N- or C-terminal extensions. For example, small N-terminal extensions have been shown to control substrate specificity and oligomeric assembly of AAA+-ATPases [6]. In the case of AFG1L1, we have identified the protein by affinity chromatography on calmodulin-agarose indicating the presence of an additional calmodulin-interacting domain. We could clearly demonstrate that AFG1L1 binds calmodulin in a calcium-dependent fashion, a feature that might regulate substrate interaction or enzymatic activity of AFG1L1. The calmodulin-binding domain required for this interaction is located within the catalytic AAA-domain of AFG1L1, just N-proximal to the Walker A motif. Therefore, we believe that the interaction of AFG1L1 with calmodulin might trigger the catalytic activity of this AAA+-ATPase. Interestingly, the calcium-binding domain is encoded in the beginning of exon 4 together with the Walker A motif, which is necessary for the catalytic activity of most AAA+-ATPases.

The acquisition of a calmodulin-binding domain to AFG1L1 appears to be a more recent feature that was added after the endosymbiotic creation of mitochondria and chloroplasts. Because no calmodulin binding to any AFG1-like protein has so far been described, AFG1L1 might be a specific addition to the members of the AFG1-like protein family. This acquisition appears to be a plant-specific trait because, in contrast to other plant AFG1 homologs, no calmodulin-binding domain could be modeled for homologs of nonplant organisms such as yeast or humans. This is in accordance with the fact that plants contain a number of unique calmodulin targets that are not present in other eukaryotes [27].

The interaction of AFG1L1 with calmodulin would interlink the regulatory function of AAA+-ATPases with the calcium signaling network of the cell. Calmodulin is a ubiquitous transducer of calcium signals in eukaryotes and is especially important in plants [10,11]. Moreover, calcium-regulation is a eukaryotic trait and has never been shown for prokaryotes. Nevertheless, calmodulin-mediated calcium regulation has been shown for both mitochondria and chloroplasts [15–22]. Several proteins with calmodulin-binding sites have been identified in both organelles [9,16,20,23] and indications for the presence of calmodulin in chloroplasts and mitochondria have been obtained [15,28,29]. Compared to yeast or humans, the Arabidopsis genome contains over 50 genes coding for putative calmodulins and calmodulin-like proteins and, although it is not yet clear which of these are targeted into either organelle, several calmodulin-like proteins contain potential signal peptides for chloroplast and mitochondrial targeting [21].

As a result of its dual localization, AFG1L1 provides a means to simultaneously regulate organellar functions in a calcium-dependent manner. Interestingly, AFG1L1 is not the only organellar calmodulin-binding AAA+-ATPase. Previous screens for calmodulin-binding proteins have identified CIP111 (a chloroplast-localized homolog of CDC48) as a calmodulin-binding protein [9]. Thus, it appears that calcium signaling is clearly incorporated into the chloroplast’s tool kit of regulation. In the case of AFG1L1, this feature has been extended into the second endosymbiotic organelle, the mitochondrion.

So far, the function of most members of the AFG1-like subfamily remains elusive. AAA+-ATPases have been shown to be involved in a wide range of cellular processes, including protein degradation, DNA replication, movement along the cytoskeleton or membrane fusion [1,2]. AFG1 from yeast was originally described as a homolog of the AAA-ATPase N-ethylmaleimide sensitive fusion protein (NSF), an important component involved in the regulated disassembly of SNARE-complexes during membrane fusion [30]. By contrast to AFG1L1, NSF belongs to the classical clade of AAA-ATPases, containing two AAA-domains. Nevertheless, AFG1L1 contains an ND with features reminiscent of a second, degraded AAA-domain, indicating that it might have evolved from an ancestor belonging to the classical clade. Thus, AFG1L1 may play a role in chloroplast vesicle transport [18]. Alternatively, a role in the degradation of misfolded or unassembled cytochrome c oxidase subunits was suggested for AFG1 from yeast [26]. In plants, this function would likely be fulfilled by the gene product of AFG1L2, the solely mitochondrial-localized homolog of AFG1L1. Dual localization of AFG1L1 suggests a functional role in both organelles and further studies are required to determine whether the protein is rather related to NSF or involved in one of the many other processes catalyzed by AAA+-ATPases.

Experimental procedures


Bovine brain calmodulin was obtained from Alexis (Grünberg, Germany). Calmodulin-agarose was obtained from Sigma-Aldrich (Hamburg, Germany) and Ni2+-nitrilotriacetic acid agarose was obtained from Qiagen (Hilden, Germany). The cross-linking reagents EDC and sulfo-N-hydroxysulfosuccinimide were obtained from Pierce (Bonn, Germany) and the reticulocyte lysate translation system was obtained from Promega-Deutschland (Mannheim, Germany).

Affinity chromatography on calmodulin-agarose

Intact chloroplasts were obtained from leaves of 20–25-day-old Arabidopsis plants (Arabidopsis thaliana L., cultivar Columbia) as described previously [31]. Isolated chloroplasts were lysed for 30 min in 50 mm Hepes/KOH (pH 7.6) on ice and thylakoid membranes were isolated by subsequent centrifugation at 3000 g for 10 min. Extrinsic thylakoid proteins were obtained by high salt wash using 500 mm NaCl in 25 mm Hepes/KOH (pH 7.6). Membranes were incubated for 60 min on ice followed by centrifugation at 30 000 g for 15 min. The supernatant was diluted in 25 mm Hepes/KOH (pH 7.6) to 150 mm NaCl and CaCl2 was added to a final concentration of 0.1 mm before incubation with 100 μl of calmodulin-agarose slurry for 16 h at 4 °C. The agarose beads were collected by a brief low-spin centrifugation and washed three times with ten column volumes of binding buffer (25 mm Hepes/KOH, pH 7.6, 150 mm NaCl, 0.1 mm CaCl2). Bound proteins were eluted with 2% SDS. Elution fractions were analyzed by SDS/PAGE followed by silver staining.

Heterologous expression of AFG1L1

Full-length AFG1L1 lacking the N-terminal 60 amino acids as well as various truncated AFG1L1 constructs were obtained by PCR on cDNA from A. thaliana using primers containing the restriction sites for NcoI and XhoI. The PCR products were cloned into pET21d (Novagen, Schwalbach, Germany) in frame with a C-terminal His6-tag. All constructs were heterologously expressed in E. coli BL21(DE3) cells. Purification of expressed proteins was performed on Ni2+-nitrilotriacetic acid agarose (Qiagen) under denaturing conditions in accordance with the manufacturer’s instructions.

Pull-down assay on calmodulin-agarose

Twenty micrograms of either recombinant full-length AFG1L1 as well as truncated AFG1L1 constructs, SSU or Tic32 were incubated with 20 μl of calmodulin-agarose slurry for 90 min at 4 °C in a buffer containing 10 mm Tris/HCl (pH 8.0), 50 mm NaH2PO4, 4 m urea and 0.1 mm CaCl2 as described previously [32]. For control experiments, calcium was substituted by 5 mm EDTA/EGTA. Agarose beads were collected by a brief low-spin centrifugation and washed three times with four column volumes of the respective binding buffer. Bound proteins were eluted competitively with 10 μm calmodulin and further analyzed by SDS/PAGE and Coomassie blue staining.

Cross-linking of AFG1L1 with calmodulin

Twenty micrograms of recombinant AFG1L1 was incubated in the presence or absence of 2 μm calmodulin in 50 mm Hepes/KOH (pH 7.6), 0.1 mm CaCl2 for 2 h at 4 °C. For control experiments, calcium was substituted with 5 mm EDTA/EGTA. Cross-linking was performed for 30 min at 21 °C in the presence of 2 mm EDC and 5 mm sulfo-N-hydroxysulfosuccinimide as described previously [33]. The cross-linking reactions were analyzed by SDS/PAGE and immunodecoration with α-AFG1L1.

Localization of AFG1L1 by western blot analysis

Purified recombinant AFG1L1 protein was used to raise antibodies in rabbit (Biogenes, Berlin, Germany). Intact chloroplasts and mitochondria were isolated from pea and Arabidopsis as previously described [31,34–36]. Western blot analysis was performed using α-AFG1L1 as well as α-Tic62, and α-voltage-dependent anion channel (VDAC). Immunoreactive proteins were detected by secondary antibodies coupled to either horseradish peroxidase utilizing an ECL detection kit (GE Healthcare Europe GmbH, Freiburg, Germany) or coupled to alkaline phosphatase with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as substrates.

Protein import into chloroplasts and mitochondria

Full-length AFG1L1 was amplified by PCR and subsequently cloned into pSP65. mRNA was obtained using SP6 RNA-polymerase and 35S-labeled protein was subsequently produced with the reticulocyte lysate system from Promega-Deutschland (Mannheim, Germany) for 60 min at 30 °C in the presence of 35S-labeled methionine (1175 Ci·mmol−1; PerkinElmer LAS Germany GmbH, Rodgau, Germany). The translation mixture was centrifuged for 10 min at 256 000 g to remove aggregated proteins and the supernatant was used for in vitro import assays.

Import into either isolated chloroplasts or mitochondria from leaves of 10-12-day-old peas (Pisum sativum, cultivar Arvika) and dual-targeting experiments were performed as described previously [25,35,37]. In brief, chloroplasts and mitochondria from P. sativum were isolated separately as described above. Import reactions were carried out for 20 min at 25 °C, followed by protease treatment with thermolysin. For dual-import experiments, mitochondria and chloroplasts were separated after import by centrifugation through a 4% (v/v) percoll cushion and fractions containing the organelles were collected individually, separated on SDS/PAGE and analysed by phospho-imaging using a Fuji FLA-3000 (Fujifilm Europe GmbH, Düsseldorf, Germany).

Structure analysis

Secondary structure prediction was performed using NPS@ (PBIL, Lyon, France) [38]. Analysis of the calmodulin-binding domain was performed using the Calmodulin Target Database [39].


We would like to thank J. Soll (LMU Munich) for kindly providing expression clones of SSU and Tic32 as well as the antibody against Tic62. The antiserum against VDAC was a kind gift of J. Whelan (Perth University). The technical assistance of Claudia Sippel is gratefully acknowledged. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (SFB-TR1) to U.C.V.