Rhomboid proteins in the chloroplast envelope affect the level of allene oxide synthase in Arabidopsis thaliana


  • Ronit Rimon Knopf,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Hebrew University of Jerusalem, Rehovot 76100, Israel
    Search for more papers by this author
  • Ari Feder,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Hebrew University of Jerusalem, Rehovot 76100, Israel
    Search for more papers by this author
    • Present address: Newe Yaar Research Center, Agricultural Research Organization (ARO), PO Box 1021, Ramat Yishay 30095, Israel.

  • Kristin Mayer,

    1. Institute of Plant Physiology and Biotechnology, University of Hohenheim, 70593 Stuttgart, Germany
    Search for more papers by this author
  • Albina Lin,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Hebrew University of Jerusalem, Rehovot 76100, Israel
    Search for more papers by this author
  • Mor Rozenberg,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Hebrew University of Jerusalem, Rehovot 76100, Israel
    Search for more papers by this author
  • Andreas Schaller,

    1. Institute of Plant Physiology and Biotechnology, University of Hohenheim, 70593 Stuttgart, Germany
    Search for more papers by this author
  • Zach Adam

    Corresponding author
    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Hebrew University of Jerusalem, Rehovot 76100, Israel
      (e-mail zach@agri.huji.ac.il).
    Search for more papers by this author

(e-mail zach@agri.huji.ac.il).


Rhomboids are intra-membrane serine proteases whose sequences are found in nearly all organisms. They are involved in a variety of biological functions in both eukaryotes and prokaryotes. Localization assays revealed that two Arabidopsis thaliana rhomboid-like proteases (AtRBL), AtRBL8 and AtRBL9, are targeted to the chloroplast. Using transgenic plants expressing epitope-tagged AtRBL9, we localized AtRBL9 to the chloroplast inner envelope membrane, with both its N- and C-termini facing the stroma. Mass spectrometry analyses confirmed this localization, and suggested that this is also the case for AtRBL8. Both are proteins of very low abundance. The results of size-exclusion chromatography implied that AtRBL9 forms homo-oligomers. In search of a putative function, a comparative proteomic analysis was performed on wild-type and double-knockout plants, lacking both AtRBL8 and AtRBL9, using the iTRAQ method. Of 180 envelope proteins, the level of only a few was either increased or decreased in the mutant line. One of the latter, allene oxide synthase, is involved in jasmonic acid biosynthesis. This observation provides an explanation for the recently reported aberration in flower morphology that is associated with the loss of AtRBL8.


The chloroplast is not only the photosynthetic organelle, but is also involved in the synthesis of lipids, amino acids, secondary metabolites and plant hormones. In addition to the plethora of enzymes participating in these processes, chloroplasts contain an elaborate proteolytic machinery that is involved in diverse functions such as protein import and targeting, general protein quality control, and responses to changes in environmental conditions. The majority of these proteases are either entirely soluble, or are membrane-anchored with their proteolytic domain facing a soluble compartment. Only a few are intra-membrane proteases (Adam et al., 2006; Sakamoto, 2006; Kato and Sakamoto, 2010; Olinares et al., 2011).

Unlike the great majority of proteases, the active sites of intra-membrane proteases are embedded in the lipid bilayer where they hydrolyze a peptide bond within a transmembrane helix of their substrates, thereby releasing them from the membrane (Urban and Freeman, 2002; Weihofen and Martoglio, 2003; Wolfe and Kopan, 2004; Wolfe, 2009a; Lal and Caplan, 2011). Although membrane-embedded, these proteases employ the same basic hydrolytic mechanisms as their soluble counterparts (Wolfe, 2009b). One of the best-characterized intra-membrane proteases is the rhomboid protease, which belongs to the family of serine proteases (Urban, 2010). The first identified rhomboid protease, Rho-1, was found to be responsible for cleavage of growth factor Spitz in the Golgi apparatus of Drosophila melanogaster, thereby releasing it from the membrane, allowing its subsequent secretion from the cell and activation of the epidermal growth factor receptor (EGFR) pathway (Lee et al., 2001; Urban et al., 2001). Structural studies revealed that rhomboids are Ser–His dyad proteases composed of six core transmembrane helices forming a V-shaped cavity within the membrane. Lateral movement of a helix and a loop opens this cavity to the aqueous phase, allowing entry of the substrate and water molecules, thus providing a hydrophilic environment where the cleavage reaction takes place (Wang et al., 2006; Wu et al., 2006; Ben-Shem et al., 2007; Lemieux et al., 2007; Vinothkumar, 2011).

Sequence analysis of various plant species revealed the presence of multiple rhomboid-like sequences in each (Koonin et al., 2003; Garcia-Lorenzo et al., 2006; Tripathi and Sowdhamini, 2006; Kmiec-Wisniewska et al., 2008). However, only some of these, 13 in the case of Arabidopsis (Lemberg and Freeman, 2007), are predicted to encode active proteases. AtRBL1 and AtRBL2 (for nomenclature, see Lemberg and Freeman, 2007; and Table S1) localize to the Golgi apparatus and are expressed in all plant tissues (Kanaoka et al., 2005). Disruption of each of these genes did not lead to any visible phenotype, and only AtRBL2 is capable of cleaving the Drosophila substrates Spitz and Keren (Kanaoka et al., 2005). AtRBL9 and AtPARL were found to localize to chloroplasts and mitochondria, respectively (Kmiec-Wisniewska et al., 2008), but no further information on the function or intra-organelle localization of AtRBL9 is available. A recent study demonstrated that AtRBL8 localizes to both chloroplasts and root plastids, and correlated its function mainly with root growth, floral development and fertility (Thompson et al., 2012). Furthermore, these authors speculated that the floral aberrations of rbl8 knockout mutants resulted from alteration in jasmonate synthesis.

Due to the ubiquity of rhomboid proteases and their importance in various organisms, including bacteria and mitochondria, we aimed to identify and characterize chloroplast rhomboids. Here we show that at least two rhomboid proteases, AtRBL8 and AtRBL9, reside in the chloroplast envelope, with their N- and C-termini oriented toward the stroma. The absence of both leads to alterations in the level of several envelope proteins. Among these is the enzyme allene oxide synthase (AOS), which is involved in the biosynthesis of jasmonic acid, providing a functional link between chloroplast rhomboids and the recently reported reduced fertility in the absence of AtRBL8 (Thompson et al., 2012).


Prediction of chloroplast-targeted rhomboid-like proteases

In order to identify and characterize chloroplast rhomboid-like proteases, we searched for putative Arabidopsis thaliana rhomboid-like (AtRBL) proteins using protein blast in the National Center for Biotechnology Information server. Our final list of putative AtRBLs was composed strictly of proteins containing the conserved residues asparagine, serine and histidine in the appropriate positions, with the serine located within the conserved GASG motif (Lee et al., 2001; Figure S1), on the assumption that such proteins will be active (Table 1). Once the set of putatively active AtRBLs was established, we searched for predicted chloroplast-targeted proteins using subcellular localization prediction software. There have been cases of ambiguity in the targeting signals, thus a mitochondrial-targeting prediction was also considered as a possible chloroplast-targeting prediction. The strongest chloroplast-targeting prediction consensus was found for AtPARL, AtRBL1, AtRBL8 and AtRBL9, but a weaker or no consensus was found for AtRBL2–6 (Table 1).

Table 1. Prediction of subcellular localization of the various rhomboid-like proteins
Gene nameaLocusProtein IDSubcellular localization prediction
  1. P, plastid; M, mitochondria; O, other.

  2. The putatively active AtRBLs are listed.

  3. Loci and protein IDs correspond to data retrieved from the National Center for Biotechnology Information database.

  4. aNomenclature according to Lemberg and Freeman (2007).

AtPARLAt1g18600 NP_564058 PMMM
AtRBL1At2g29050 NP_180469 POOP
AtRBL2At1g63120 NP_176500 OOOP
AtRBL3At5g07250 NP_196342 OOOP
AtRBL4At3g53780 NP_850698 OOOO
AtRBL5At1g52580 NP_175667 OOOO
AtRBL6At1g12750 NP_172735 OMOO
AtRBL8At1g25290 NP_173900 OPPO
AtRBL9At5g25752 NP_680221 PMMM

AtRBL8 and AtRBL9 are localized to chloroplast membranes

At the onset of our study, experimental data regarding rhomboid localization in Arabidopsis was available only for AtRBL1 and AtRBL2, which were both found to localize to the Golgi apparatus (Kanaoka et al., 2005). Thus, we determined the localization of AtPARL, AtRBL3, AtRBL8 and AtRBL9. To this end, we fused the cDNA sequences encoding the N-termini of these proteins (containing their transit peptides) to that of GFP, transiently expressed the chimeric genes in Arabidopsis protoplasts, and monitored the accumulation of GFP using confocal microscopy (Figure 1). AtPARL–GFP co-localized with the mitochondrial reporter Mitotracker, similarly to the pattern for the mitochondrial control ATPsyn–GFP (a fusion of GFP to the ATP synthase β-subunit transit peptide), suggesting mitochondrial localization of AtPARL (in agreement with the results of Kmiec-Wisniewska et al., 2008). AtRBL8–GFP and AtRBL9–GFP co-localized with chlorophyll autofluorescence, similar to the chloroplast control SSU–GFP (a fusion of GFP to the Rubisco small subunit transit peptide), suggesting their chloroplast localization (in agreement with the results of Kmiec-Wisniewska et al., 2008 and the recently published study by Thompson et al., 2012). AtRBL3–GFP co-localized neither with Mitotracker nor with chlorophyll autofluorescence. However, it showed an expression pattern that was distinct from that of GFP alone, suggesting it is localized elsewhere than the cystosol (GFP alone), mitochondria or chloroplasts. The chloroplast localization of AtRBL8 and 9 was further confirmed using chloroplast import assays of full-length in vitro synthesized proteins into isolated pea chloroplasts (Figure S2). Although import efficiency is low (an expected phenomenon given their hydrophobicity; Bölter and Soll, 2011), it is apparent that both AtRBL8 and 9 were imported into chloroplasts, underwent maturation, and partitioned with the membrane fraction.

Figure 1.

 Subcellular localization of GFP fusion proteins in Arabidopsis protoplasts.
The proteins whose targeting peptides were used are indicated on the left. Chlorophyll autofluorescence, GFP and Mitotracker fluorescence are shown. For chloroplast localization, GFP and chlorophyll images were merged, and for mitochondria localization, GFP and Mitotracker images were merged. GFP was used as a control to show the targeting pattern of GFP alone, without a transit peptide, SSU–GFP is a fusion of GFP to the Rubisco small subunit transit peptide (chloroplast control), and ATPsyn–GFP is a fusion of GFP to the ATP synthase β-subunit transit peptide (mitochondrial control).

AtRBL9 is localized to the chloroplast envelope

To examine the sub-chloroplast membrane localization of AtRBL8 and AtRBL9 in vivo, we generated transgenic Arabidopsis lines expressing C-terminally HA-tagged proteins (AtRBL–HA); only the AtRBL9–HA transgenic lines accumulated detectable levels of the protein. As expected, AtRBL9–HA was detected in total protein extracts and isolated chloroplasts of both AtRBL9–HA1 and AtRBL9–HA5 lines, and was not detected in wild-type (WT) plants (Figure 2a). To determine its sub-chloroplast localization, intact chloroplasts were isolated, lysed and fractionated into thylakoid and envelope membranes. Samples were immunoblotted using marker antibodies for thylakoids (OEE33) and envelopes (CAC3 and Tic40). Consistent with their known localization, OEE33 was detected only in thylakoids whereas CAC3 and Tic40 were found exclusively in the envelope (Figure 2b). The AtRBL9–HA distribution pattern follows that of the envelope proteins.

Figure 2.

 Intra-organellar localization of AtRBL9.
(a) Total protein extracts (T) and intact chloroplasts (C), isolated from WT and HA1 and HA5 transgenic lines expressing AtRBL9–HA were separated by SDS–PAGE and immunoblotted with antibodies as indicated. Each sample contained 2 μg chlorophyll.
(b) Intact chloroplasts were isolated from the transgenic plants, hypotonically lysed and fractionated on a sucrose gradient into envelope (E) and thylakoid (T) membranes. Samples were separated by SDS–PAGE and immunoblotted with antibodies as indicated. Each sample contained 5 μg protein.
(c) Envelope membranes were isolated from chloroplasts of various plant lines. Aliquots of protein (15 μg) were loaded in each lane, separated on SDS–PAGE, and immunoblotted using an AtRBL9-specific antibody. For detection with the HA antibody, only 1.5 μg protein (1:10) were loaded.
For all panels, the molecular weight of marker proteins in kDa is indicated on the right.

When envelope preparations of various plant lines were probed with an AtRBL9-specific antibody (Figure 2c), a band was detected at the expected size in WT, the AtRBL8-2 knockout mutant and the AtRBL9–HA1 transgenic line, but not in an AtRBL9 knockout line (see description of knockout lines below). The band detected by the AtRBL9 antibody in the AtRBL9–HA1 transgenic line was the same size as the one detected by the HA antibody, further confirming that AtRBL9 localizes to the envelope. Interestingly, these results are consistent with the results of a proteomic analysis of chloroplast envelopes (Ferro et al., 2010). As the vast majority of outer envelope proteins are not subjected to maturation (Li and Chiu, 2010), whereas both AtRBL8 and AtRBL9 underwent maturation during protein import (Figure S1), these results suggest that RBL9 localizes to the inner envelope membrane.

To determine the topology of AtRBL9 in the inner envelope membrane, we performed protease protection assays in intact chloroplasts. This method is based on the differential permeability of the outer envelope membrane (OEM) to trypsin and thermolysin; thermolysin cannot penetrate the OEM, and therefore degrades only proteins on the outer surface of the chloroplast, whereas trypsin penetrates the OEM and degrades proteins located both on the outer surface of the chloroplast and in the inter-membrane space (McAndrew et al., 2001). Intact chloroplasts isolated from AtRBL9–HA plants were treated with both proteases, and the degradation pattern of AtRBL9–HA was monitored (Figure 3a). PDV2, an outer envelope protein facing the cytosol (Glynn et al., 2008), was efficiently degraded in the presence of either thermolysin or trypsin, whereas OEP37, an outer envelope protein (Schleiff et al., 2003), was resistant to thermolysin but degraded in the presence of trypsin. A similar pattern was observed for APG1, an inner envelope protein whose soluble domain faces the inter-membrane space (Viana et al., 2010). The OEP37 and APG1 degradation patterns confirmed that trypsin did indeed penetrate the OEM, and was also active in the inter-membrane space. FtsH1, a thylakoid membrane protein facing the stroma (Lindahl et al., 1996), was mostly resistant to both proteases, suggesting that trypsin, at the concentration used, did not substantially penetrate the inner envelope membrane. In the presence of 2% Triton X-100, which disrupts membrane compartmentalization, FtsH1 was fully degraded by trypsin, but not by thermolysin.

Figure 3.

 Membrane topology of AtRBL9–HA.
(a) Intact chloroplasts isolated from AtRBL9–HA transgenic plants were treated with thermolysin or trypsin with or without 2% Triton X-100. Samples were subjected to SDS–PAGE and immunoblotted using antibodies as indicated. For HA detection, 1.5 μg chlorophyll equivalents were loaded onto the gel, and, for the rest of the antibodies, 5–10 μg chlorophyll equivalents were loaded. The protease concentrations used for the treatment were 100 μg ml−1 (+), 200 μg ml−1 (2+) and 400 μg ml−1 (4+).
(b) Intact and hypotonically lysed chloroplasts isolated from AtRBL9–HA transgenic plants were treated with trypsin. Samples were subjected to SDS–PAGE and immunoblotted. For HA detection, 1.5 μg chlorophyll equivalents were used. For Tic40 detection, 5 μg chlorophyll equivalents were used. Protease concentrations used for the treatment were 0 μg ml−1 (0), 50 μg ml−1 (1), 100 μg ml−1 (2), 250 μg ml−1 (3) and 500 μg ml−1 (4). In treatment 2*, trypsin inhibitor was added to the reaction mixture together with the protease.
For all panels, the molecular weight of marker proteins in kDa is indicated on the right.

The HA signal was resistant to thermolysin treatment; even in the presence of 2% Triton X-100, only slight processing was observed (Figure 3a). The signal also resisted trypsin at the high concentration of 400 μg ml−1, but was degraded when detergent was added. Although it is not clear why both FtsH1 and the HA signal resisted thermolysin treatment in the presence of detergent, the HA degradation pattern in the presence of either protease best resembles that of FtsH1, suggesting that the HA tag faces the stroma. To rule out the possibility that the HA signal was resistant to trypsin due to embedding within the membrane, intact and hypotonically lysed chloroplasts were treated with increasing trypsin concentrations (Figure 3b). Chloroplast lysis reduces the protective effect of membrane compartmentalization, but does not affect membrane embedding of the proteins. Comparing the protection pattern of Tic40 (an inner envelope protein whose soluble domain faces the stroma, Chou et al., 2003) and the HA signal suggested that their sensitivity to trypsin increases following chloroplast lysis (compare lanes 3 and 4). Both Tic40 and the HA signal appeared to slightly degrade when intact chloroplasts were treated (Figure 3a); however, when hypotonically lysed chloroplasts were treated, degradation was far more efficient (Figure 3b). These results confirm that the HA tag is protected from trypsin due to membrane compartmentalization and not membrane embedding, and that, like Tic40, it faces the stroma. As AtRBL9 is predicted to contain six transmembrane helices (Lemberg and Freeman, 2007) and was tagged at its C-terminus, it may be concluded that AtRBL9 is located in the inner envelope membrane with both its N- and C-termini facing the stroma.

AtRBL9 forms a homo-oligomer

We have often noted that the HA antibody recognizes more than one band (e.g. Figure 3). This is reminiscent of similar observations in bacterial rhomboids (e.g. Lemberg et al., 2005). Most characteristic was the additional band at approximately 50 kDa, and higher-molecular-weight bands were also observed in some cases. The predicted size of the mature protein expressed from the AtRBL9–HA construct is 27 kDa, suggesting that the higher-molecular-weight bands may represents oligomers that are too hydrophobic to dissociate even under the high reducing conditions of SDS–PAGE, similar to the case of the D1 and D2 proteins of the photosystem II complex (Kapri-Pardes et al., 2007). To assess the size of the potential oligomer, intact chloroplasts were isolated from transgenic plants, solubilized using the mild detergent n-dodecyl-β-d-maltoside (DDM) and fractionated on a Superose 6 column. The relevant fractions (in which the AtRBL–HA protein was identified) were pooled, concentrated and fractionated on a Sephacryl 200 column (Figure 4a). Analysis of the obtained fractions using the HA antibody (Figure 4b) revealed that the peak of AtRBL9–HA eluted at a volume corresponding to a size of approximately 130 kDa (Figure 4a). As the mean size of DDM micelles is 50 kDa, the AtRBL9–HA oligomer has a size of 80 kDa.

Figure 4.

 AtRBL9–HA forms oligomers.
Intact chloroplasts were solubilized using 1% DDM and loaded onto a Superose 6 column. Fractions containing the AtRBL9–HA protein were pooled and concentrated, and loaded onto a Sephacryl S-200 column.
(a) The protein concentration during fractionation was monitored on the basis of the absorbance at 280 nm (OD280). Size markers and the HA signal are indicated by arrowheads. Retention volumes are the means ± SD of two independent replicates.
(b) Samples from the Sephacryl S-200 eluted fractions were subjected to SDS–PAGE and immunoblotted using an HA antibody. Fraction numbers (Fra.) and the corresponding retention volumes (Ret.) are indicated. The molecular weight in kDa of marker proteins is indicated on the left. The peak fraction of the HA signal is indicated by an arrow. Two blots are presented, separated by a white line.

In immunoprecipitation experiments performed on intact chloroplasts solubilized by DDM, followed by SDS–PAGE, tryptic digestion and MS peptide fingerprint analysis, most identified peptides belonged to peptides of AtRBL9–HA, suggesting that it does not form a stable complex with any other protein (Figure S3a,b). AtRBL9 peptides were recovered not only from the gel piece corresponding to the monomer size, but also from pieces corresponding to the high-molecular-weight bands in the immunoblots. The sequence coverage map of the recovered peptides also suggested that processing of AtRBL9 occurs at or closely downstream of the site predicted by the ChloroP program (Figure S3c). Taken together, the monomer size of 27 kDa and the observed 80 kDa complex imply that AtRBL9–HA may form a homo-oligomer of three to four subunits. However, we do not know at this stage whether oligomerization of AtRBL9 has any functional significance.

The envelope proteome: WT versus double knockout

To determine the physiological role of AtRBL8 and AtRBL9, corresponding T-DNA insertion lines were obtained, backcrossed to WT, and crossed between themselves to produce double knockouts (see Figure S4). Under all growth conditions tested (standard short day, standard long day, long day at 27°C, long day under outdoor light intensities varying between 70 and 2500 μmol photons m−2 sec−1 at 16, 22, 28 and 34°C. Temperatures apply to outdoor light intensities), these single and double knockout plants presented a WT phenotype (e.g. Figure S4d). Additional characteristics of the knockout lines such as germination rates, total protein content, total chlorophyll content and their photosystem II activity, measured by chlorophyll fluorescence (Fv/Fm), were also the same as in WT (Figure S5). We therefore used a comparative proteomic approach, and searched for envelope proteins whose level is affected by loss of AtRBL8 and AtRBL9 (line 877). A total of 812 peptides were identified and quantified, allowing for the quantitative comparison of 180 proteins (Table S2). Localization analysis of the identified proteins using the At_Chloro database (Ferro et al., 2010), showed that only 19 of the 180 identified proteins are not envelope residents, suggesting relatively low cross-contamination of the analyzed envelope fractions by other compartments (Figure 5).

Figure 5.

 Localization of the identified proteins.
The localization of identified proteins was determined based on the At_Chloro database (Ferro et al., 2010). Env, envelope; OM, outer envelope membrane; IM, inner envelope membrane; Str, stroma; Thy, thylakoids; Mit, mitochondria; NP, non-plastid. Black bars represent envelope proteins, white bars represent thylakoid and stromal proteins, and gray bars represent non-chloroplast proteins.

Protein targets for rhomboid cleavage are expected to be more abundant in line 877 compared to WT. This was observed for 11 proteins, with a reduction of >50% compared with the WT. These proteins are thus candidate substrates of chloroplast rhomboid proteases (Table 2). For six of these proteins, not all detected peptides were affected uniformly. In three of them, the levels of peptides derived from the N-terminus were lower in the WT than in the mutant, and for three proteins, the levels of C-terminal peptides were reduced. Apparently, the cleavage products that were generated from the target protein by rhomboid cleavage were not equally detectable. This may be explained by differences in the stability of the cleavage products, or differential release of either the N- or the C-terminal cleavage product from the envelope. Probable direct targets for limited proteolysis by chloroplast rhomboids include Tic21, FAD7, CAC3, PORC, APG1 and an amino oxidase family protein (Table 2). For the remaining five proteins, peptides from the entire length of the protein were similarly affected, i.e. were more abundant in the mutant than in the WT. These proteins may also be direct targets of rhomboid proteases, and thus more abundant in the rhomboid loss-of-function mutant. The most likely candidates in this group, proteins whose level is ca. 2-fold higher in the mutant compared to WT, are PORB, FtsH11 and an aldo/keto reductase family protein (Table 2). Alternatively, an unknown regulatory process may have been affected in the mutant resulting in accumulation of these proteins.

Table 2. Summary of the decreased proteins (WT < mutant)Thumbnail image of

For 13 proteins, the opposite effect was observed, i.e. they were more abundant in the WT and showed a reduction of >50% in mutant line 877 (Table 3). For six of these proteins, all detected peptides showed a similar decrease in the mutant. These proteins are not likely to be substrates of rhomboids but may be indirectly affected by rhomboid activity via a regulatory protein that is subject to rhomboid cleavage in the mutant. They include a MATH domain protein, OEP6 and 16-1, AOS, RCA, and an unknown protein (Table 3). For seven of the 13 proteins, only peptides derived from either the N- or the C-terminus were reduced in the mutant, rather than the entire length of the protein. As these proteins cannot be direct targets of rhomboids proteases, because of the lower level of some peptides in the mutant, the differential accumulation of different parts of the protein remains unexplained.

Table 3. Summary of the increased proteins (WT > mutant)Thumbnail image of

AOS level is affected by AtRBL8 and/or AtRBL9

To follow up the iTRAQ results, we performed an immunoblot analysis on selected proteins from the two representative groups (877 > WT and 877 < WT). APG1, which contains a single transmembrane domain, similar to known rhomboid substrates, was not affected by the loss of AtRBL8 and/or AtRBL9 (Figure S6). Similarly, the level and size of Tic40, which has been implied to be processed by the At1g74130 gene product (Karakasis et al., 2007), were not affected (Figure S6). However, the AOS protein displayed a pattern similar to the one observed in the iTRAQ experiments, i.e. WT plants accumulated 60% more AOS compared to the double knockout plants (Figure 6a,b), confirming that AOS is indeed indirectly affected by the loss of AtRBL8 and/or AtRBL9. Interestingly, this effect on AOS abundance is not at the transcript level, as the AOS transcript abundance in WT and the double knockout is similar (Figure 6c). These results establish a link between the presence of chloroplast rhomboid proteases in the envelope and accumulation of AOS, an enzyme in the jasmonic acid biosynthesis pathway that resides in the same membrane.

Figure 6.

 AOS accumulates to higher levels in the WT envelope compared with the double knockout mutant.
(a) Envelope membranes were prepared from WT and double knockout line 877. Various sample amounts were subjected to SDS-PAGE in Tris/Tricine and immunoblotted using the indicated antibodies. The molecular weight of marker proteins in kDa is indicated on the right.
(b) Quantification of the results shown in (a). AOS levels were normalized to the envelope protein OEP37 for quantification. Values are means ± SE of five replicates. The asterisk indicates a statistically significant difference compared with line 877 by t test (P < 0.05).
(c) The AOS transcript level was compared between WT and double knockout line 877 using quantitative PCR. Values are means ± SE of four replicates. The difference between WT and double knockout line 877 was not significant.


AtRBL8 and AtRBL9 conservation and functional redundancy

The ubiquity of rhomboid proteases within organisms of all kingdoms suggested their possible existence in chloroplasts as well. During a search for chloroplast rhomboid proteases, we identified AtRBL8 and AtRBL9 as chloroplast-targeted rhomboids, consistent with the results of Kmiec-Wisniewska et al. (2008) and Thompson et al. (2012). AtRBL8 and AtRBL9 appear to be conserved not only in higher plants but also in mosses. Phylogenetic analysis of plant rhomboid proteases in Arabidopsis thaliana, Oryza sativa, Populus trichocarpa and Physcomitrella patens showed that both AtRBL8 and AtRBL9 form distinct branches, and both have a homolog in each of these species (Kmiec-Wisniewska et al., 2008). Currently, AtRBL8 and AtRBL9 are the only rhomboid-related proteins that have been identified in the chloroplast using proteomic tools (Ferro et al., 2010). It is noteworthy that extremely low levels of AtRBL8 and AtRBL9 were measured in the above study, which may explain why we did not detect them in our proteomic analysis. This also suggests that other rhomboids may exist in the chloroplast in very low levels. However, since no additional putatively active rhomboid proteases are predicted to reside in the chloroplast (this study and Kmiec-Wisniewska et al., 2008), the existing data suggest that AtRBL8 and AtRBL9 are the only chloroplast rhomboid proteases, and that both are conserved among plants.

As the only rhomboid proteases in the chloroplast, their function cannot be substituted for by other rhomboid proteases; nevertheless, the possibility of functional redundancy with other chloroplast proteases cannot be excluded. For example, in the case of Drosophila Rho-1, residual levels of Spitz cleavage occur independently of Rho-1. This residual cleavage is sensitive to the protease inhibitor batimastat, suggesting partial functional overlap with metalloproteases (Urban et al., 2001). Examples of metalloproteases present in the same membrane compartment as AtRBL8 and AtRBL9 are AraSP (Bölter et al., 2006) and FtsH11 (this study and Ferro et al., 2010). This implies that rhomboid function is of such importance that alternative pathways are employed to bypass their deficiency, or, alternatively, that their substrates may be residually cleaved by other proteases independently of the rhomboids’ physiological role.

Possible function of chloroplast rhomboids

In this study, AtRBL9 was shown to reside in the inner envelope membrane, with both its termini facing the stroma. AtRBL8 was found to localize to the chloroplast envelope using a GFP targeting assay (Thompson et al., 2012) and in a proteomic analysis (Ferro et al., 2010). Additionally, AtRBL8 underwent maturation during import (Figure S2). Therefore, we suggest that AtRBL8 localizes to the inner envelope as AtRBL9 does. Due to similarity in the distribution of positive charges in the hydrophilic loops of both proteins, we predict that they have the same topology (both are localized with their N-termini facing the stroma). Although the active site of rhomboids is buried within the membrane bilayer, it is not placed in the middle, but instead is located at the outer leaflet of the membrane, closer to the membrane–water interphase (Urban et al., 2001; Wang et al., 2006, 2007; Ben-Shem et al., 2007). These topological findings and structural constraints imply a scenario in which, following cleavage of a substrate by AtRBL8 or AtRBL9, its soluble domain is released to the inter-membrane space (Figure 7).

Figure 7.

 Proposed model for chloroplast rhomboid topology and cleavage directionality.
The chloroplast rhomboid protease in the inner envelope membrane is represented schematically. The catalytic Ser and His residues are shown. The position of the active site within the membrane results in cleavage in the upper part of the substrate transmembrane helix, and release of the cleavage product to the inter-membrane space. IEM, inner envelope membrane; IMS, inter-membrane space.

The scenario outlined above fits in with a few possible functions. One is in chloroplast retrograde signaling. Although retrograde signaling research has advanced greatly, the actual signals have yet to be identified (Pfannschmidt, 2010; Inaba et al., 2011). Release of a proteinaceous factor into the inter-membrane space may be a step towards release of a signaling molecule, similar to Drosophila rhomboid Rho-1, whose cleavage substrate Spitz is secreted from the cell to activate the EGFR pathway (Lee et al., 2001; Urban et al., 2001). Moreover, there are chloroplast-localized proteins that are later relocated from the chloroplast to the nucleus (Krause and Krupinska, 2009), supporting the idea of a proteinaceous factor taking part in retrograde signaling. Such examples are the transcription factors pBrp and Tsip1, which are attached to the chloroplast envelope, and, under specific conditions, are released from the envelope and relocate to the nucleus (Lagrange et al., 2003; Ham et al., 2006). The suggested mechanisms for their release include, among others, proteolytic cleavage of membrane-associated proteins (Krause and Krupinska, 2009). Accordingly, a recent study suggested that intra-membrane proteolysis, mediated by an unknown serine protease, releases the PTM transcription factor from the membrane, allowing its relocation to the nucleus and activation of the ABI4 transcription factor (Sun et al., 2011).

Another possible function is import, either to the inter-membrane space or the inner envelope membrane. Currently, Tic22 and MGD1 are the only known residents of the inter-membrane space. Although both use the TOC complex to cross the outer envelope, Tic22 does not require the action of stromal factors or ATP to reach its final destination, while MGD1 requires ATP and the SPP protease for its full translocation (Vojta et al., 2007). Nevertheless, both proteins are processed to maturation, with Tic22 suggested to be processed by an unknown inter-membrane space protease (Vojta et al., 2007). The topological characteristics of AtRBL8 and AtRBL9, as well as the fact that rhomboids do not require co-factors or ATP (Lemberg et al., 2005; Urban and Wolfe, 2005), suggest the possibility of their involvement in Tic22 processing during its import.

Protein targeting to the inner envelope is another possibility for rhomboid function. In general, inner envelope residents reach their destination either via a stop-transfer pathway or a re-insertion (post-import) pathway (Li and Chiu, 2010). A study that aimed to identify the various routes to the inner envelope membrane suggested the existence of five different pathways (Firlej-Kwoka et al., 2008). In that study, Tic40, HP29b, PIC1 and PPT were suggested to undergo multiple cleavage steps during import, not only in the stroma. This again suggests a possible role for rhomboids in import and processing of these proteins, leading to their correct insertion into the membrane. Tic40 has indeed been suggested to be processed by chloroplast rhomboids (Karakasis et al., 2007). However, the gene product tested in that study (At1g74130) does not possess the required residues in the appropriate positions for activity. Furthermore, we show here that Tic40 was not affected by loss of AtRBL8 and/or AtRBL9 (Figure S6), both at the abundance and size of the protein. However, involvement of AtRBL8 and/or AtRBL9 in the import and processing of other inner envelope residents should not be ruled out.

Two other possible substrates identified in the proteomic analysis are the inner envelope proteins FtsH11 and aldo/keto reductase. In the absence of specific antibodies, they could not be further studied. Nevertheless, their hydrophobicity plots suggest the existence of one or two transmembrane helices. Thus it would be interesting in the future to determine whether their import and processing are affected by AtRBL8 and/or AtRBL9.

The AOS level is affected by AtRBL8 and/or AtRBL9

We used iTRAQ analysis to search for differences at the protein level within the envelope membranes of WT and rhomboid double knockout plants. This analysis highlighted AOS, an enzyme of the jasmonate biosynthesis pathway (Schaller and Stintzi, 2009), as an enzyme whose level is affected by these rhomboids. In their absence, AOS accumulated to 50–60% of WT levels (Figure 6 and Table 3). AOS is a cytochrome P450 of the CYP74 family (Song et al., 1993), which catalyzes the first specific reaction in the formation of 12-oxophytodienoic acid (OPDA), a signaling compound with multiple functions that is also the immediate precursor of jasmonic acid (Laudert et al., 1996). The AOS enzyme is targeted to the chloroplast envelope in a pathway that requires ATP and proteins (Froehlich et al., 2001), and is bound to the membrane through a large non-polar patch at the enzyme surface, and not by a transmembrane helix or a lipid anchor (Schaller and Stintzi, 2009). The Arabidopsis genome contains a single intronless gene encoding AOS. It is expressed in rosette leaves and is induced by wounding, among other signals, both at the transcript and protein levels (Laudert and Weiler, 1998; Kubigsteltig et al., 1999). Plants in which the AOS gene is knocked out are male-sterile and display lower jasmonic acid synthesis levels following wounding (von Malek et al., 2002; Park et al., 2002). Thompson et al. (2012) showed that AtRBL8 knockout plants present diffuse floral abnormalities and some fertility defects in early inflorescences, and implied the involvement of AtRBL8 in jasmonic acid biosynthesis. Our results provide a functional link between these defects and rhomboid function, with chloroplast rhomboids (most probably AtRBL8) being involved in AOS accumulation in the envelope membrane.

Experimental Procedures


Previously published putative Arabidopsis rhomboid-like protein sequences (Koonin et al., 2003) were used for a blast search of the Arabidopsis thaliana genome database using the blastp algorithm. The retrieved sequences were then aligned to known rhomboid sequences using multalin software (http://multalin.toulouse.inra.fr/multalin/multalin.html; Corpet, 1988). Sequences lacking the structurally and functionally important conserved residues were eliminated. Subcellular localization was predicted using ChloroP (http://www.cbs.dtu.dk/services/ChloroP/; Emanuelsson et al., 1999), TargetP (http://www.cbs.dtu.dk/services/TargetP/; Emanuelsson et al., 2000), Predotar (http://urgi.versailles.inra.fr/predotar/predotar.html; Small et al., 2004) and PSORT (http://psort.hgc.jp/form.html).

Plant material

Arabidopsis thaliana plants, ecotype Columbia-0, were grown on Kekkila peat (Kekkila, http://www.kekkila.com) at 22°C and 16 h illumination of 90–120 μmol photons m−2 sec−1. Transgenic lines expressing AtRBL9 fused to the HA epitope were generated by cloning relevant sequences into vector pCB302-3 (Xiang et al., 1999), and transformation into WT plants using the floral-dip method (Clough and Bent, 1998). T-DNA insertion lines for AtRBL8 (SALK lines N578681 and SALK N537037) and AtRBL9 (SAIL line N842060) were obtained from the Salk Institute (http://www.salk.edu/). Both knockout lines were backcrossed with WT three times, and then were crossed with each other to generate a double knockout.

GFP localization assay

The cDNA sequences encoding the N-termini of the described proteins were fused to GFP in the pTH2GFP expression vector (Chiu et al., 1996), and transiently expressed under the control of the CaMV 35S promoter in Arabidopsis protoplasts, as previously described (Sade et al., 2009). GFP fluorescence was visualized using confocal laser scanning microscope LSM 510 (Zeiss, http://www.zeiss.com/) in multi-channel mode. Mitochondria were detected by Mitotracker Orange CMTMRos staining (Invitrogen, http://www.invitrogen.com). Images were processed using Zeiss LSM 510 image browser software.

Quantitative PCR

Total RNA was extracted from 3-week-old Arabidopsis plants using a standard procedure. Reverse transcription was performed using RevertAidTM M-MuLV reverse transcriptase (Fermentas, http://www.fermentas.com). The quantitative PCR reaction was performed using SYBR® Green technology on the MX3000P system (Stratagene, http://www.stratagene.com). For each genotype, at least four independent biological replicates were analysed in triplicate. Relative expression of AOS was quantified using the relative standard method, with ACT2 (At3g18780) as the normalizer transcript.

Leaf extracts, chloroplast and envelope preparations, and protease protection assay

Total protein extracts were obtained by grinding leaf tissue in 0.2 m Tris/HCl pH 6.8, 5 m urea, 8% SDS, 10% glycerol, 20%β-mercaptoethanol, followed by incubation at 40°C for 5 min. The extract was cleared by centrifugation at 13 000 g for 15 min. Intact chloroplasts were prepared from rosette leaves of 8-week-old plants as previously described (Aronsson and Jarvis, 2002). Intact chloroplasts were further fractionated as described by Block et al. (2002). For the protease protection assay, intact chloroplasts (75 μg chlorophyll) were treated with thermolysin (type X; Sigma, http://www.sigmaaldrich.com) or trypsin (Boehringer Mannheim, http://www.roche-applied-science.com) at the indicated concentrations in HS buffer (200 μl, 50 mm HEPES/KOH pH 8.0, 0.3 m sorbitol) in the presence of 0.8 mm CaCl2 and, when indicated, 2% Triton X-100. The reaction was performed on ice for 10 min and quenched by addition of HS buffer containing 55 mm EDTA and one tablet of mini protease inhibitor cocktail (Roche, http://www.roche.com). In the trypsin treatment, the quenching buffer was supplemented with 7.5–75 μg trypsin inhibitor (type I-S from soybean; Sigma), based on the amount of trypsin in the reaction. Following protease treatment, intact chloroplasts were re-isolated on a 40% Percoll cushion (Sigma) and re-suspended in HS buffer.

Immunoblot analysis

Samples were separated on 12–15% polyacrylamide in Tris/glycine buffer unless indicated otherwise, and electroblotted onto nitrocellulose membrane, or, for HA detection, polyvinylidene fluoride membrane. Antibodies used and their dilutions were: HA (Sigma, monoclonal), 1:1500; OEE33 (Zaltsman et al., 2005), 1:10 000; GST–FtsH1 (Zaltsman et al., 2005), 1:5000; Tic40 (Agrisera, http://www.agrisera.com) 1:1000; AtRBL9 (Agrisera), 1:750; CAC3 (Ke et al., 2000), 1:20 000; PDV2 (Glynn et al., 2008), 1:3000; OEP37 (Schleiff et al., 2003), 1:1000; APG1 (Viana et al., 2010), 1:500. All blots were developed using enhanced chemiluminescence (ECL) reagents (Sigma) for horseradish peroxidase-conjugated secondary antibodies.

iTRAQ and mass spectrometry analysis

Envelope samples were subjected to iTRAQ analysis as follows. Four protein pellets (50 μg each) were re-suspended separately in 8 m urea in 100 mm ammonium bicarbonate. The samples were reduced using 2.8 mm dithiothreitol (60°C for 30 min), modified using 8.8 mm iodoacetamide in 100 mm ammonium bicarbonate at room temperature for 30 min, and digested in 2 m urea, 25 mm ammonium bicarbonate using modified trypsin (Promega, http://www.promega.com) at a 1:50 enzyme:substrate ratio, overnight at 37°C, followed by a second digestion for 4 h. The tryptic peptides were desalted using C18 tips (Harvard Inc., www.harvardapparatus.com), dried and resuspended in 100 mm HEPES (pH 7.3). The iTRAQ™ reagent was brought to room temperature and mixed with ethanol. After vortexing and spinning, each of the reagents was transferred to a sample tube (30:70 sample:reagent ratio). The tubes were incubated at room temperature for 1 h. All four iTRAQ™ reagent-labeled samples were combined, purified on C18, and resuspended in 0.1% formic acid. The combined labeled peptides (60 μg) were separated in an online 2D chromatography protocol (MuDPiT, multidimensional protein identification technology, see http://www.proteome.soton.ac.uk/mudpit.htm). First, the peptides were loaded on a 15 mm BioX-SCX column (Thermo Scientific, http://www.thermoscientific.com) and eluted using salt steps of 0, 30, 100, 200 and 500 mm ammonium formate in 5% acetonitrile and 0.1% formic acid, pH 3. The eluted peptides were further resolved by capillary reverse-phase chromatography (75 μm internal diameter, 20 cm fused-silica capillaries, self-packed with 3 μm Reprosil-Aqua C18). The peptides were eluted using a 125 min gradient (5–40% acetonitrile containing 0.1% formic acid), followed by a wash step of 95% acetonitrile for 15 min. The flow rate was approximately 0.2 μl min−1 and the peptides were analysed using an Orbitrap mass spectrometer Thermo-Fischer, http://www.thermofisher.com). Mass spectrometry was performed in positive mode using a repetitively full MS scan followed by collision-induced dissociation and higher-energy collision dissociation of the three most dominant ions selected from the first MS scan. The mass spectrometry data were analyzed and compared using sequest 3.31 (Thermo-Fischer) searching the ‘PLANTS’ part of the NCBInr database (http://blast.ncbi.nlm.nih.gov). All these procedures were performed at the Smoler Proteomic Research Center (The Technion, Haifa, Israel).

Gel filtration

Intact chloroplasts were isolated from transgenic plants. Protein (1 mg) from the intact chloroplasts was pelleted by centrifugation (10,000 × g, 5 min at 4°C) and resuspended in 125 μl buffer A (100 mm Tris pH8.0, 50 mm NaCl), and then supplemented with 125 μl buffer B [100 mm Tris pH 8.0, 50 mm NaCl, 2% DDM Sigma), 2× protease inhibitor cocktail (Roche)]. The suspension was incubated on ice for 15 min with occasional vortexing for complete solubilization. Following solubilization, the sample was ultracentrifuged for 30 min at 29 000 g. The resulting supernatant was fractionated using a Superose 6 column (Amersham) in an AKTA explorer instrument (Amersham, http://www.gelifesciences.com). The column was first equilibrated in buffer A, then loaded with sample and eluted with buffer A supplemented with 0.1% DDM at a rate of 200 μl min−1. Fractions (300 μl) were collected, of which 45 μl were used for SDS–PAGE. Fractions containing the HA signal were pooled and concentrated to a volume of 300 μl using an Amicon Ultracel 5K ultracentrifugal filter device (Millipore, http://www.millipore.com), supplemented with protease inhibitor cocktail and fractionated again on a Sephacryl S-200 column (Amersham). Elution of protein was performed using buffer A supplemented with 0.05% DDM. The protein markers used with both columns were apoprotein (443 kDa), alcohol dehydrogenase (150 kDa), BSA (66 kDa) and carbonic anhydrase (29 kDa; all purchased from Sigma).


We thank Elzbieta Glazer (Department of Biochemistry & Biophysics, Stockholm University) for the mitochondrial ATPsyn–GFP-encoding construct, Menachem Moshelion for the assistance in the GFP transient expression assays, and Gal Dvorkin for help with statistical analysis. Antibodies were generously provided by Bettina Bölter (Department of Botany, University of Munich) (OEP37), Katherine Osteryoung, (Plant Biology Department, Michigan State University) (PDV2), Danny Schnell (Department of Biochemistry & Molecular Biology, University of Massachusetts) (APG1), Stephan Pollman (AOS) and Eve Wurtele (CAC3). This work was supported by grants from the Israel Science Foundation (903/04 and 385/08) to Z.A., and from the Hohenheim University/Hebrew University Collaborative Research Program to A.S. and Z.A.