Research Article

Ubiquitin-dependent chloroplast-associated protein degradation in plants

See allHide authors and affiliations

Science  22 Feb 2019:
Vol. 363, Issue 6429, eaav4467
DOI: 10.1126/science.aav4467

Chloroplast-associated protein degradation

Protein degradation is vital for cellular functions, and it operates selectively with distinct mechanisms in different organelles. Some organellar proteins are targeted by the ubiquitin-proteasome system (UPS)—a major proteolytic network in the eukaryotic cytosol. In such cases, the organelle membrane presents a substantial barrier to protein degradation. Working in the model plant Arabidopsis, Ling et al. identified mechanisms underlying the UPS-mediated degradation of proteins in the outer membrane of chloroplasts (the organelles responsible for photosynthesis). They identified an Omp85-type β-barrel outer membrane channel and a cytosolic AAA+ chaperone that fulfill conductance and motor functions in the retrotranslocation of target proteins from chloroplasts. This process thus enabled outer membrane protein processing by the cytosolic proteasome. Such chloroplast-associated protein degradation was initiated by ubiquitination of the targets by the chloroplast-localized E3 ubiquitin ligase SP1.

Science, this issue p. eaav4467

Structured Abstract


Chloroplasts are plant organelles responsible for the bulk of terrestrial photosynthetic primary production. They evolved via endosymbiosis from a cyanobacterial organism more than a billion years ago. The biogenesis and operation of chloroplasts depends on the assembly and homeostasis of thousands of nucleus-encoded proteins, which together constitute a large part of the organellar proteome. These proteins are imported by multiprotein translocases in each of the chloroplast envelope membranes after translation in the cytosol. Chloroplast proteins are subject to proteolytic regulation, which plays vital roles in maintaining normal organellar functions and in delivering responses to developmental and environmental cues. Turnover of internal chloroplast proteins is controlled by proteases inherited from the organelle’s prokaryotic ancestor, but mechanisms underlying the degradation of chloroplast outer envelope membrane (OEM) proteins are poorly defined.


We previously showed that components of the protein import translocases in the OEM (so-called TOC proteins) are ubiquitinated by the OEM-localized ubiquitin E3 ligase SP1 and subsequently degraded by the cytosolic 26S proteasome. Inherent in this process is a need to extricate the target proteins from the chloroplast membrane (they are integral membrane proteins), and to achieve this there must exist mechanisms to overcome the physical and thermodynamic barriers to extraction. This implies that additional factors are involved in OEM protein degradation, and we sought to identify these by applying forward genetics and proteomic analysis in the plant Arabidopsis.


We identified two factors required for the degradation of chloroplast OEM proteins: SP2 and CDC48. The former is an Omp85-type β-barrel channel of prokaryotic origin located in the OEM, and the latter is a conserved eukaryotic AAA+ chaperone located in the cytosol. We observed that inactivation of either component triggers the selective overaccumulation of target proteins, specifically at the chloroplast envelope. We used genetic analyses to demonstrate that SP2 and CDC48 act together in the same proteolytic pathway as the SP1 E3 ligase and physical interaction studies to show that the three components can form a complex at the surface of the chloroplast. Furthermore, by applying complementary in vivo and in vitro assays, we demonstrated that the SP2 and CDC48 proteins cooperate to bring about the extraction (“retrotranslocation”) of ubiquitinated proteins from the OEM. Overall, the data are consistent with a model (see the figure) in which SP2 and CDC48 fulfil conductance and motor functions, respectively, in the retrotranslocation of OEM proteins ubiquitinated by SP1 to enable their proteasomal degradation in the cytosol. These results extend the range of known functions of Omp85 superfamily proteins (which heretofore included bacterial protein secretion, membrane protein biogenesis, and organelle protein import) and of CDC48 (which has a well-characterized role in endoplasmic reticulum–associated protein degradation). The broader importance of this proteolytic mechanism was demonstrated by physiological analyses of plants with altered SP2 activity, which revealed defects in organellar functions, plant development, and viability.


Collectively, our results describe a multicomponent system for chloroplast envelope protein removal, dependent on the cytosolic ubiquitin-proteasome system, which is critically important for plant growth. A key part of the system is a protein retrotranslocation mechanism of chimeric prokaryotic-eukaryotic ancestry that operates at the surface of the organelle. We refer to this proteolytic system as chloroplast-associated protein degradation, or CHLORAD.

Chloroplast-associated protein degradation.

CHLORAD is a proteolytic system that selectively removes chloroplast OEM proteins, including TOC components of the chloroplast protein import machinery. The SP1 E3 ligase directs the ubiquitination (Ub) of targets; it has a RING finger (RNF) domain for ubiquitin-conjugating enzyme (E2) recruitment and an intermembrane space (IMS) domain that binds to its targets. The SP2 and CDC48 proteins mediate target retrotranslocation to the cytosol, respectively providing a conduit and driving force for the process. Upon release to the cytosol, targets are degraded by the 26S proteasome (26SP). Additional, as yet unknown factors are shown in gray.


Chloroplasts contain thousands of nucleus-encoded proteins that are imported from the cytosol by translocases in the chloroplast envelope membranes. Proteolytic regulation of the translocases is critically important, but little is known about the underlying mechanisms. We applied forward genetics and proteomics in Arabidopsis to identify factors required for chloroplast outer envelope membrane (OEM) protein degradation. We identified SP2, an Omp85-type β-barrel channel of the OEM, and CDC48, a cytosolic AAA+ (ATPase associated with diverse cellular activities) chaperone. Both proteins acted in the same pathway as the ubiquitin E3 ligase SP1, which regulates OEM translocase components. SP2 and CDC48 cooperated to bring about retrotranslocation of ubiquitinated substrates from the OEM (fulfilling conductance and motor functions, respectively), enabling degradation of the substrates by the 26S proteasome in the cytosol. Such chloroplast-associated protein degradation (CHLORAD) is vital for organellar functions and plant development.

Chloroplasts are plant organelles responsible for the bulk of terrestrial photosynthetic primary production, and they evolved via endosymbiosis from a cyanobacterial organism more than 1 billion years ago (1). The modern chloroplast proteome comprises ~3000 proteins, most of which are nucleus encoded and imported posttranslationally by multiprotein translocases located in the organelle’s outer and inner envelope membranes; these translocases are termed TOC and TIC (translocons at the outer and inner envelope membranes of chloroplasts), respectively (25). The biogenesis and operation of chloroplasts requires not only the assembly of the constituent organellar proteins but also their coordinated homeostasis. Turnover of internal chloroplast proteins is governed by several prokaryotic-type proteases inherited from the endosymbiont (6). By contrast, outer envelope membrane (OEM) proteins are degraded by the cytosolic ubiquitin-proteasome system (UPS) via poorly understood mechanisms (1).

The RING (really interesting new gene)-type ubiquitin E3 ligase SP1 is located in the chloroplast OEM, where it mediates the ubiquitination of OEM components of the chloroplast protein import machinery (so-called TOC proteins), thereby promoting their degradation by the cytosolic 26S proteasome (7). The TOC components affected by SP1 include the receptors Toc159 and Toc33 and the channel protein Toc75. Such SP1-mediated regulation of the TOC apparatus changes the organellar proteome, which in turn influences the developmental fate and functions of the organelle (e.g., enabling plant adaptation to abiotic stress) (7, 8). Although the role of SP1 in marking proteins for degradation is clear, other aspects of this chloroplast protein degradation system have remained obscure. Because TOC proteins are integral membrane components, additional factors are most likely required to overcome the physical and energetic barriers to their extraction from the membrane before degradation in the cytosol, as is the case in other membrane-associated proteolytic systems (911).

Identification and phenotypic and molecular analysis of SP2

To improve understanding of the SP1-dependent proteolytic pathway, we revisited the forward-genetics screen (7) that originally identified SP1; this was a screen for extragenic suppressors of the Arabidopsis Toc33 mutation plastid protein import1 (ppi1), which causes chlorosis because of defective chloroplast protein import (12). In addition to sp1 mutants, mutants with lesions at a second, unlinked locus were identified in the screen; these were termed suppressor of ppi1 locus2 (sp2) mutants. Double mutant sp2 ppi1 plants were larger and greener than the ppi1 progenitor (Fig. 1, A to C) and exhibited substantial improvements in chloroplast development and chloroplast protein import capacity (Fig. 1, D to F); in all of these respects, the sp2 mutants were phenotypically very similar to sp1 mutants (7).

Fig. 1 The sp2 mutation suppresses the phenotype of the Toc33 knockout mutation, ppi1.

(A) Visible phenotypes of 30-day-old sp2-1 ppi1 suppressor mutant and control plants grown on soil. (B) Chlorophyll contents of 10-day-old sp2-1 ppi1 suppressor mutant and control seedlings grown in vitro. (C to E) Transmission electron microscopy analysis of the ultrastructure of cotyledon chloroplasts in 10-day-old sp2-1 ppi1 suppressor mutant and control plants grown in vitro. Typical organelles are shown (C). Scale bar, 2 μm. These and other micrographs were used to estimate chloroplast cross-sectional area (D) and thylakoid development (E). (F) Analysis of protein import into chloroplasts isolated from sp2-1 ppi1 suppressor mutant and control plants by using 35S-labeled Rubisco small subunit precursor protein as the import substrate, and corresponding quantification of the maturation (mat) of the radiolabeled precursor protein (pre). A representative phosphor screen image is shown (top); times (t) indicate minutes after the start of each import reaction. Together with similar images from two additional experiments, this was used to conduct the quantitative analysis shown (bottom); the amount of imported, mature protein in each sample was expressed as a percentage of that present at the final time point for the WT. (G) Domain map of the SP2 protein. Gray box, β-barrel domain; black boxes, predicted transmembrane spans. The sites of amino acid substitutions in two sp2 mutant alleles are indicated with gray triangles. Numbers indicate amino acid positions. R, arginine; H, histidine; G, glycine. (H and I) Independent mutant alleles of sp2 suppress ppi1 in similar fashion to sp2-1. The sp2-2 ppi1 and sp2-3 ppi1 mutants were identified in the EMS mutagenesis screen, whereas the sp2-4 T-DNA insertion mutant was obtained from a stock center and crossed to the ppi1 line. The plants were grown in vitro for 10 days before photography (H) and chlorophyll content analysis (I). The chlorophyll values for sp2-2 and sp2-3, which cause missense mutations, were statistically significantly different from that for sp2-1 (Student’s t test, P ≤ 0.01), suggesting that they are weak alleles. All values are means ± SEM (n ≥ 3 experiments or samples).

The SP2 locus (At3g44160) (Fig. 1G) was identified by using a combination of genetic mapping and whole-genome sequencing of the three independent mutant alleles identified in the screen (fig. S1). The original alleles were phenotypically similar to one another and to a transferred-DNA (T-DNA) insertion mutant affecting the same gene (Fig. 1, H and I, and fig. S1C); unless specifically stated otherwise, the latter (sp2-4, a null allele) was used in all subsequent analyses. The encoded protein is a member of the Omp85 superfamily of β-barrels, which are involved in protein biogenesis and transport and are widely distributed in the outer membranes of bacteria, mitochondria, and chloroplasts (13, 14). The SP2 protein is of unknown function, but it is broadly conserved in the angiosperms (flowering plants) and is closely related to the chloroplast outer membrane protein OEP80 (a protein that has also been termed Toc75-V) (fig. S2) (15). The function of OEP80 is also uncertain (15), although it has been proposed to mediate outer membrane protein biogenesis (16, 17) by analogy with well-characterized homologs in bacteria (BamA and TamA) and mitochondria (Sam50/Tob55) (13, 14).

The SP2 protein is located in the chloroplast OEM (figs. S3 and S4) (15), and it has been shown to have the capacity to form a membrane channel, like other members of the Omp85 superfamily (18). Unlike OEP80, SP2 lacks an N-terminal POTRA (polypeptide transport–associated) domain (such domains typically mediate protein-protein interactions), suggesting that the two proteins have functionally diverged (Fig. 1G and fig. S2B), a notion that is also supported by phylogenetic analyses of the proteins (fig. S2A) (17). Accordingly, OEP80 and SP2 have diametrically opposing effects on TOC protein abundance, as is evident upon comparing published results [showing that OEP80 knockdown depletes TOC proteins (16)] with those discussed below.

Analysis of the effects of SP2 on TOC proteins and plant development

To shed light on the role of SP2, further genetic analyses were conducted. In addition to ppi1, two other TOC mutations (hypomorphic alleles of the genes encoding Toc159 and Toc75) (16, 19) were phenotypically suppressed by sp2 (fig. S5). By contrast, mutations that cause chlorosis because of lesions affecting the TIC apparatus of the inner envelope membrane were not suppressed by sp2 (fig. S6). Together, these data implied a close functional relationship between SP2 and the TOC apparatus, and this notion was supported by the restored accumulation of Toc75 protein in sp2 toc double mutant plants (Fig. 2, A to D, and fig. S7). In all of these respects, the sp2 mutants were phenotypically very similar to sp1 mutants (7).

Fig. 2 Functional analysis of SP2 reveals roles in chloroplast proteostasis and development.

(A to F) Immunoblot analyses of total protein extracts (two loading amounts per sample) from the indicated genotypes, including sp2 ppi1 suppressor mutants [(A) and (B)], sp2 and toc75-III-3 single and double mutants [(C) and (D)], and two different SP2-OX lines [(E) and (F)]. Slp1, mitochondrial stomatin-like protein 1 (a nonchloroplastic membrane protein that served as a loading control); H3, nuclear histone H3. (G) Leaf senescence analysis of the indicated genotypes using mature rosette leaves induced to senesce by covering with aluminum foil. Typical control (uncovered) and senescent (covered) leaves are shown (left). The maximum photochemical efficiency of photosystem II (Fv/Fm) was measured to estimate the extent of senescence (right); the covered values for sp2-4 and SP2-OX were statistically significantly different from that for the WT (Student’s t test, P ≤ 0.0003). (H) Abiotic stress tolerance analysis of the indicated genotypes using 14-day-old plants grown in vitro on NaCl medium. Typical plants (left) and chlorophyll contents (right) are shown. The chlorophyll values for the mutant and overexpressor plants were statistically significantly different from that for the WT (Student’s t test, P < 0.004). All values are means ± SEM (n ≥ 3 experiments or samples).

The close functional connection between SP2 and the TOC apparatus was further emphasized by the observation that overexpression of SP2 triggers the specific depletion of TOC proteins (Fig. 2, E and F), resembling closely the effect of SP1 overexpression (7). Moreover, SP2 [like SP1 previously (7)] was shown to interact physically with TOC components in coimmunoprecipitation (co-IP) experiments (fig. S4), suggesting that its effect on TOC accumulation is mediated through direct physical interaction with TOC proteins. The unlikely possibility that SP2 also influences peroxisomal protein import (20, 21) could be ruled out because of the absence of an effect on the abundance of peroxisomal protein import machinery components (fig. S8).

Further similarities between SP2 and SP1 were observed when the expression profiles of the two genes were compared (fig. S9) and when sp2 mutant and SP2 overexpressor (SP2-OX) plants were analyzed physiologically in relation to leaf senescence (Fig. 2G) and salt stress tolerance (Fig. 2H and fig. S10). The activity of SP1 promotes both leaf senescence and abiotic stress tolerance (which it does by reconfiguring the chloroplast protein import machinery to produce the necessary organellar proteome changes) (7, 8), and a very similar pattern of phenotypes was observed for SP2 (Fig. 2, G and H).

Identification of CDC48 as a mediator of OEM protein degradation

The identification of a channel-forming component (SP2) that putatively cooperates with SP1 in TOC protein degradation raised a parallel with endoplasmic reticulum (ER)–associated protein degradation (ERAD) (9, 10, 22), where polytopic membrane proteins form a retrotranslocon to enable substrate extraction from the membrane (2325). In that system, the conserved eukaryotic, multifunctional AAA+ (ATPase associated with diverse cellular activities) chaperone CDC48 (p97) (9, 10, 26) forms a cytosolic ATP-powered motor to drive such retrotranslocation. We sought similar regulators of the TOC apparatus by using co-IP followed by mass spectrometry and identified CDC48 as a minor TOC-associated component (fig. S11). The major Arabidopsis isoform, CDC48A (At3g09840) (27) (Fig. 3A), is ubiquitous in the cytosol, but its association with chloroplasts was evident upon analyzing lysed, cytosol-free cells containing a CDC48 mutant with stabilized substrate binding (fig. S12).

Fig. 3 Functional analysis of CDC48 reveals roles in chloroplast proteostasis and development.

(A) Domain map of the CDC48A protein. Gray boxes, ATPase domains (D1 and D2). The sites of the DN and Trap mutations are indicated with gray triangles. Numbers indicate amino acid positions. K, lysine; A, alanine; E, glutamic acid; Q, glutamine. (B) Visible phenotypes of 9-day-old plants expressing the CDC48-DN protein or an equivalent nonmutated control protein (CDC48-WT), both induced with estradiol for 2 days. Two independent transgenic lines are shown for each construct. (C to E) Immunoblot analyses of total leaf [(C) and (D)] and protoplast (E) protein extracts from the CDC48-WT and CDC48-DN transgenic plants after induction with estradiol or mock control treatment lacking the inducer. For (C) and (D), plants were simultaneously treated with 200 mM mannitol for 2 days before protein extraction to trigger stress-dependent TOC protein degradation (8); this was essential to observe the effects of CDC48-DN in this assay. (F and G) BiFC analysis of the CDC48-TOC interaction. Reconstitution of YFP fluorescence was assessed after transient coexpression of the indicated pairs of fusion proteins, which carry complementary N- or C-terminal YFP fragments (nY and cY, respectively). OEP7 and CDKA1 acted as OEM and cytosolic controls, respectively. Representative images are shown (F). Scale bar, 10 μm. In addition, the frequency of protoplasts showing a BiFC signal was quantified (G). All values are means ± SEM (n = 3 experiments).

Because CDC48 is an essential component in Arabidopsis, we used plants that inducibly overexpressed a dominant-negative (DN) form of the protein (CDC48-DN) (Fig. 3A) (27) to assess its function. The expression of CDC48-DN caused chlorosis (Fig. 3B), indicating defective chloroplast biogenesis, as well as a buildup of reactive oxygen species (ROS), paralleling an effect of the sp1 mutation (8) (fig. S13). In the case of sp1, such ROS overaccumulation was attributed to a failure to properly regulate the TOC apparatus and chloroplast protein import, leading to a deregulation of photosynthetic activity (8). Similarly, the effects of CDC48-DN observed in this study were linked to the specific overaccumulation of TOC proteins (Fig. 3, C and D), revealing phenotypic similarity to sp2 mutants (Fig. 2, A to D) in addition to sp1 mutants (7). The SP1 protein, which is subject to UPS-dependent autoregulation (fig. S14) (7), also accumulated in response to CDC48-DN expression (Fig. 3E), indicating that SP1 is degraded via the same processes as TOC proteins.

Because CDC48 is an abundant cellular constituent distributed throughout the nucleocytosolic compartment, we wished to localize its functional links to the chloroplast protein import machinery specifically to the OEM in intact cells. By using bimolecular fluorescence complementation (BiFC), a method that uses the reconstitution of yellow fluorescent protein (YFP) fluorescence to report on protein-protein interactions of interest, we demonstrated the physical interaction of CDC48 with Toc159 specifically at the chloroplast envelope (Fig. 3, F and G). Separate experiments further revealed that the CDC48-DN–triggered accumulation of OEM proteins (shown in Fig. 3, C and D) occurred specifically at the envelope in vivo (fig. S15). Together, these data supported a direct role for the CDC48 chaperone in OEM protein degradation, at the surface of the organelle.

Analysis of the functional and physical relationships among SP1, SP2, and CDC48

Having identified two new components that apparently mediate OEM protein degradation, like SP1, we next addressed whether the three components act together in a common pathway. We began by considering the relationship between SP1 and SP2. The absence of phenotypic additivity in sp1 sp2 double mutants (in the ppi1 background) in relation to plant greening and Toc75 protein accumulation (Fig. 4, A to D) supported the notion that SP1 and SP2 function together. Accordingly, SP2 was essential for SP1 action, because the sp2 mutation abrogated the effect of SP1 overexpression (fig. S16).

Fig. 4 SP2 acts in the same pathway of TOC protein degradation as SP1.

(A to D) Analysis of an sp1 sp2 ppi1 triple mutant. Triple mutant plants were compared with both sp1 ppi1 mutants and sp2 ppi1 mutants in relation to the extent of suppression of ppi1. No phenotypic additivity in the triple mutants was apparent upon analysis of visible phenotypes (A), chlorophyll contents (B), or the abundance of Toc75 protein as determined by immunoblotting [(C) and (D)]; two loading amounts per sample were analyzed. (E to G) Analysis of plants simultaneously overexpressing both SP1 and SP2. Double-OX plants were compared with both single SP-OX genotypes, revealing a synergistic interaction in relation to the visible phenotype (E) (left), chlorophyll content (E) (right), and TOC protein depletion as analyzed by immunoblotting [(F) and (G)]. Values are means ± SEM (n = 4 to 5 experiments).

Moreover, whereas the overexpression of neither SP1 nor SP2 individually affected plant greening in the wild-type (WT) background under standard conditions (fig. S10), the simultaneous overexpression of both genes caused strong chlorosis linked to severe depletion of TOC proteins (Fig. 4, E to G), indicating functional interdependency between SP1 and SP2. Such interactions are often observed where the components interact physically, and may arise through mutual stabilization (28, 29). SP1 and SP2 comigrated (with each other and with the TOC apparatus) on native gels, suggesting their coexistence in high–molecular weight complexes that either include or exclude the TOC apparatus (fig. S17A). In accordance with this interpretation, SP1 and SP2 interacted specifically in in vitro pull-down experiments (fig. S17B).

Next, the functional relationships between CDC48 and the two SP components were assessed. This was done by determining the dependence of SP overexpression effects upon CDC48 function. We observed that the ability of the overexpression of either SP1 or SP2 to trigger TOC protein depletion was blocked when CDC48-DN was coexpressed (Fig. 5). These data indicated that the functions of the SP components in OEM protein degradation require CDC48 and, thus, that all three factors act in the same proteolytic pathway. In agreement with this conclusion, reciprocal co-IP assays demonstrated physical associations among CDC48, SP2, SP1, and TOC proteins (Fig. 6 and fig. S18).

Fig. 5 CDC48 acts in the same pathway of TOC protein degradation as SP1 and SP2.

Analysis of the effects of CDC48-DN on the ability of SP1 or SP2 overexpression to trigger TOC protein depletion. Transgenic plants carrying the following constructs were analyzed by immunoblotting: CDC48-WT, CDC48-DN, SP1-OX, and the latter two in combination (A and B) and CDC48-WT, CDC48-DN, SP2-OX, and the latter two in combination (C and D). In each case, the analysis was done after treatment with [or without (mock)] estradiol (E2) to induce the expression of the CDC48 constructs. All values are means ± SEM (n = 3 to 4 experiments).

Fig. 6 CDC48 and SP2 interact physically with each other and with OEM proteins.

Co-IP of SP1, TOC components, and/or CDC48 with HA-tagged WT CDC48 (A) or Myc-tagged SP2 (B) from protoplast extracts. Cells were transfected with the following constructs: CDC48-WT-HA or YFP-HA plus SP1-Myc (A) or SP2-Myc plus CDC48-WT-HA, SP1-HA, or YFP-HA (B). In both cases, YFP-HA acted as a negative control. TL, total lysate; poly-Ub, poly-ubiquitinated form; hc, heavy chain; α-HA, anti-HA; α-Myc, anti-Myc. The asterisk indicates a nonspecific cross-reacting band.

Investigating the roles of SP2 and CDC48 in OEM protein retrotranslocation

In ERAD, CDC48 drives the retrotranslocation of substrate proteins from the ER to the cytosol. To address the hypothesis that CDC48 is similarly involved in the extraction of substrates from the chloroplast OEM, we conducted retrotranslocation assays by using two complementary methodologies (23, 30, 31).

First, we used an in vivo assay that evaluates the distribution of polyubiquitinated substrate between chloroplasts and the cytosol in intact, proteasome-inhibited cells. We observed that the extraction of polyubiquitinated Toc33 to the cytosol was inhibited in cells expressing CDC48-DN, providing strong evidence that CDC48 acts in OEM protein retrotranslocation (Fig. 7, A and B). Replication of this assay with sp2 mutant and SP2-OX cells revealed a similar requirement for SP2 in the extraction of polyubiquitinated Toc33 (Fig. 7, C and D). Moreover, when both of these assays were repeated with SP1 as a second, model substrate [selected because of its high turnover rate (7)], essentially identical results were obtained (Fig. 8, A to D, and fig. S19).

Fig. 7 CDC48 and SP2 are required for the retrotranslocation of polyubiquitinated Toc33.

(A and B) Analysis of the role of CDC48 in the retrotranslocation of Toc33-HA substrate by an in vivo retrotranslocation assay. Protoplasts isolated from CDC48-WT and CDC48-DN transgenic plants that were transiently expressing Toc33-HA were treated, after estradiol induction, with 5 μM bortezomib proteasome inhibitor and then separated into chloroplast and cytosol fractions. In this assay, retrotranslocation occurred in intact cells, and the retrotranslocated Toc33-HA was protected by bortezomib inhibition, which initiated the experiment. After fractionation, Toc33-HA was immunoprecipitated from both fractions and detected by immunoblotting with anti-HA (α-HA) and anti-ubiquitin (α-Ub) (A). Retrotranslocation efficiency was assessed by quantifying the relative amounts of ubiquitinated Toc33 in the chloroplasts and cytosol (B). (C and D) Analysis of the role of SP2 in the retrotranslocation of Toc33-HA substrate by an in vivo retrotranslocation assay performed as described for (A) and (B) by using protoplasts from WT, sp2 mutant, and SP2-OX plants (and without the need for estradiol induction). Typical immunoblotting results are shown (C), along with quantification (D). All values are means ± SEM (n = 3 experiments).

Fig. 8 CDC48 and SP2 are required for the retrotranslocation of polyubiquitinated SP1.

(A to D) Analysis of the roles of CDC48 and SP2 in the retrotranslocation of SP1-HA substrate by an in vivo retrotranslocation assay. Protoplasts isolated from CDC48-WT and CDC48-DN transgenic plants [(A) and (B)] or from WT, sp2 mutant, and SP2-OX plants [(C) and (D)] that were transiently expressing SP1-HA were treated [after estradiol induction in the case of (A) and (B)] with 5 μM bortezomib proteasome inhibitor and then separated into chloroplast and cytosol fractions. In this assay, retrotranslocation occurred in intact cells (as in Fig. 7), and the retrotranslocated SP1-HA was protected by bortezomib inhibition, which initiated the experiment. After fractionation, SP1-HA was immunoprecipitated from both fractions and detected by immunoblotting with anti-HA (α-HA) and anti-ubiquitin (α-Ub) [(A) and (C)]. Retrotranslocation efficiency was assessed by quantifying the relative amounts of ubiquitinated SP1 in the chloroplasts and cytosol [(B) and (D)]. (E and F) Analysis of the role of CDC48 in the retrotranslocation of SP1-HA substrate by an in vitro retrotranslocation assay. This assay [in contrast with the in vivo assay depicted in (A) to (D)] used a cell-free reaction. Chloroplasts were prepared from estradiol-induced CDC48-DN protoplasts transiently expressing SP1-HA and FLAG-tagged ubiquitin. Cytosol was prepared from induced CDC48-WT or CDC48-DN protoplasts (that were not expressing SP1-HA or FLAG-tagged ubiquitin). Reaction mixtures composed of chloroplasts and either cytosol fraction (as indicated above the gel images) were incubated for 1 hour and then refractionated into chloroplast (left) and cytosol (right) samples; any SP1-HA detected in the cytosol at the end of the reaction must have arisen from the chloroplasts. SP1-HA was enriched from 90% of the protein extract of each fraction by IP and detected by immunoblotting with anti-HA (to detect unmodified SP1-HA) and anti-FLAG (to detect ubiquitinated SP1-HA) (E), and the results were quantified (F). All values are means ± SEM (n = 3 to 4 experiments).

To provide corroboration of the in vivo assay data, we used an in vitro retrotranslocation assay. This experiment tested the ability of CDC48-DN in a cytosol extract to interfere with the removal of polyubiquitinated SP1 from isolated chloroplasts (Fig. 8, E and F). The results closely paralleled those from the in vivo assays, because the presence of CDC48-DN was seen to inhibit OEM substrate extraction to the cytosol (in a manner that the control CDC48-WT protein did not). Thus, we concluded that both CDC48 and SP2 act in OEM protein retrotranslocation.


This work identified SP2 and CDC48 as functional partners of the SP1 E3 ligase and, in so doing, defined a multicomponent system for chloroplast protein degradation involving integral OEM and cytosolic factors. Both components act in substrate protein retrotranslocation, and, on the basis of their structural characteristics, we conclude that they provide a retrotranslocon for substrate conductance during extraction and a molecular motor needed to meet the energetic threshold of the process (fig. S20). We designate this system chloroplast-associated protein degradation (CHLORAD), in recognition of similarities with ERAD (including the involvement of the UPS, CDC48, and retrotranslocation), but emphasize that the central participation of a β-barrel channel of prokaryotic ancestry and a number of other key differences set it apart as a mechanistically and evolutionarily distinct process (9, 25).

A major function of the analogous ERAD system is the elimination of a broad range of misfolded ER proteins (9, 10, 22), which it performs alongside another role in the removal of undamaged proteins for regulatory reasons (32). The available evidence suggests that CHLORAD has a different focus, with its primary purpose being regulatory and centered on the TOC apparatus of the chloroplast protein import machinery. Moreover, whereas ERAD processes lumenal proteins in addition to ER membrane proteins, there is presently no evidence to suggest that CHLORAD similarly acts on internal chloroplast proteins, a fact which may be linked to the presence of diverse proteases of prokaryotic origin in the chloroplast interior (6). An early event in CHLORAD is substrate protein ubiquitination, mediated by SP1, but how substrates are identified and transferred to SP1 is currently unclear. The SP1 intermembrane space domain is involved in substrate binding (7), and the triggering factor appears to be related to developmental cues or environmental stress (7, 8); the stability of SP1 may also be relevant, because the protein is autoubiquitinated (7) and is itself degraded by CHLORAD. Whether CHLORAD also acts in the clearing of damaged proteins, in parallel with its regulatory functions, remains to be determined.

Whereas ERAD uses E3 ligases with multiple transmembrane spans that may simultaneously serve as channels (9), CHLORAD requires a separate channel-forming component. This is presumably because SP1 does not have a sufficient number of transmembrane spans to enable channel formation in isolation. Our data indicate that CHLORAD is mediated by an SP1-SP2 complex in which SP2 provides a channel to enable a close link between ubiquitination and retrotranslocation. It is possible that the operation of the SP2 channel involves lateral opening to enable substrate entry from the membrane (24, 33). Unlike any of the proposed ERAD channels [which are all based on α-helical transmembrane spans (25)], SP2 belongs to the Omp85 superfamily of β-barrel proteins originating from prokaryotic cells (13, 14). This indicates that CHLORAD recruited a channel of prokaryotic origin and that this component underwent neofunctionalization during the evolution that followed endosymbiosis. Consistent with this notion, SP2 is unusual in that it does not possess a partner-protein–interacting POTRA domain, which is a typical feature of Omp85-type proteins in other systems. The data imply that SP2 instead collaborates with SP1 to provide this recruitment function (7), and we hypothesize that the loss of the POTRA domain was a key step in the protein’s acquisition of a new function. As has been proposed for ERAD (10), it is possible that more than one channel operates in CHLORAD (e.g., under different circumstances); this may account for the absence of a complete block in retrotranslocation in sp2 mutants.

Once CHORAD has been initiated, the conserved eukaryotic chaperone CDC48 is recruited to the chloroplast from the cytosol. This step may involve the binding of CDC48 to the polyubiquitin chain on a substrate protein with the assistance of unknown cofactors or with a putative chloroplast OEM tethering factor (9, 10, 22). Either way, the chaperone then drives substrate extraction to the cytosol for proteasomal degradation, which it does by overcoming the energetic barrier to the removal of proteins from the membrane (fig. S20). The functional cooperation of CDC48 with SP2 in this retrotranslocation process is notable, because it indicates that CHLORAD has a chimeric prokaryotic-eukaryotic origin, much like the chloroplast protein import machinery, which also has an Omp85 superfamily protein (Toc75) at its core (25). Quite likely, other factors are involved in the CHLORAD process, and these may exist within the SP1-SP2 core complex or act peripherally in the broader proteolytic pathway.

Previously published data concerning SP1 function (7, 8), together with results reported here concerning SP2, show clearly that the CHLORAD system is critically important for plant growth and development, with roles linked to the reconfiguration of the organellar proteome and functions and to organelle developmental fate. These observations provide a clear indication of the physiological importance of CHLORAD and suggest potential strategies to improve crop performance (e.g., under environmental stress).

Materials and methods

Plant material and growth conditions

All Arabidopsis thaliana plants were of the Columbia-0 (Col-0) ecotype, except the ppi1 line used for the genetic mapping of sp2, which was introgressed into Landsberg erecta (Ler) through seven outcrosses. The sp1-1, sp1-3, ppi1, tic40-4, hsp93-V-1, ppi2-3 (fts1), and toc75-III-3 (mar1) mutants, as well as the 35S promoter–driven SP1 overexpressor (SP1-OX) transgenic line, have all been described previously (7, 12, 16, 19, 34, 35). The sp2-4 (SALK_137135) mutant was obtained from the Salk Institute Genomic Analysis Laboratory and confirmed by polymerase chain reaction (PCR) and reverse transcription PCR (RT-PCR) analyses, as described previously (36); this mutant was phenotypically similar to the three mutants with a chemically induced sp2 allele identified in this study and unlike a previously described T-DNA mutant (37). For making double and triple mutants, the sp1-3 and sp2-4 alleles were used (because their mutations are easier to detect by PCR); double mutants were selected and verified by phenotype analysis and by PCR-based genotyping (table S1). Unless specifically stated otherwise, the sp2-4 allele was used in the experiments. For consistency with previous work (7, 8), the sp1-1 allele was used in physiological experiments.

For in vitro growth, seeds were surface sterilized, sown on Murashige-Skoog (MS) agar medium in petri plates, cold-treated at 4°C, and thereafter kept in a growth chamber, as described previously (38). All plants were grown under a long-day cycle (16 hours of light and 8 hours of darkness). For induction of CDC48-WT or CDC48-DN expression in the corresponding transgenic lines, 7-day-old plants were transferred onto MS agar medium supplemented with 4 μM estradiol (Sigma).

Physiological studies

Chlorophyll measurements were performed by using a Konica-Minolta SPAD-502 meter (39) for analysis of rosette-stage plants or by photometric quantification after extraction in N,N′-dimethylformamide (DMF) as described previously (40) for analysis of seedlings.

Dark treatments for the induction of senescence were conducted as previously described (7, 41). Developmentally equivalent leaves of 28-day-old plants were wrapped in aluminum foil while still attached to the plant and then left under standard growth conditions for 5 days. The maximum photochemical efficiency of photosystem II (Fv/Fm) was determined by measuring chlorophyll fluorescence with a CF Imager (Technologica, UK) as described previously (42). Five experiments were performed, and approximately five leaves (each one from a different plant) were analyzed per genotype in each experiment.

Salt stress experiments were conducted as described previously, with minor modifications (8). All seeds of the different genotypes used in this work were harvested at the same time. Seeds were germinated directly on MS agar medium (supplemented with 1% sucrose) containing 150 to 170 mM NaCl. Stress tolerance was assessed by measuring chlorophyll accumulation after 14 days. Three experiments were performed, and ~25 seedlings per genotype were analyzed in each experiment.

Hydrogen peroxide was detected by staining with 3,3′-diaminobenzidine (DAB) (Sigma) as previously described (43). Seven-day-old plants grown on MS agar medium were transferred onto similar medium containing 4 μM estradiol (for induction of CDC48-WT and CDC48-DN expression). The plants were left to grow for a further 2 days before the initiation of DAB staining. Each experiment used approximately five seedlings per genotype. Three experiments were performed with the same result, and typical images are presented. The area of staining was quantified by using ImageJ as described previously (8).

Identification of the sp2 mutants and genetic mapping

The original sp2 mutants (sp2-1 [sp2-310], sp2-2 [sp2-416], and sp2-3 [sp2-555]) were identified by screening the M2 progeny of 7000 M1 ppi1 seeds that had been treated with 100 mM ethyl methanesulfonate (EMS) for 3 hours by using a published procedure (7, 44). Allelism test crosses confirmed that all of the mutations are allelic. All three sp2 ppi1 mutants were backcrossed to ppi1 mutants three times before phenotypic analysis. Initial mapping of sp2 was conducted by analyzing the greenest plants in F2 populations from crosses between sp2-1 ppi1 (Col-0) and ppi1 mutants introgressed into the Ler ecotype by using PCR markers that detect Col-0/Ler polymorphisms. In a mapping population of 190 such F2 plants, six were heterozygous for the marker F21A17 at position 12285000 on the upper arm of chromosome 3 but homozygous for Col-0 downstream of that, suggesting that the suppressor mutation was in the downstream Col-0 region; F3 seedlings from these six plants were grown and verified visually to be nonsegregating, as expected for a homozygous sp2-1 ppi1 double mutant. In a second mapping population of 192 plants, the sp2 mutation was further mapped to the south of a more southerly marker, MJI6-2 at position 12597802 on the upper arm of chromosome 3. However, it was not possible to determine the position of the sp2 locus precisely because of the persistence of an “island” of Col-0 DNA in the Ler-introgressed ppi1 line, near the sp2 locus (around the chromosome 3 centromere). Thus, final identification of the gene was achieved by whole-genome sequencing.

Whole-genome sequencing and assembly

Approximately 100 mg of plant inflorescence tissue from each of the original sp2 alleles (sp2-1 ppi1, sp2-2 ppi1, and sp2-3 ppi1) and from ppi1-1 was harvested and flash-frozen in liquid nitrogen. Total genomic DNA was then extracted by using an E.Z.N.A. plant DNA kit (Omega Bio-tek) according to the manufacturer’s guidelines. The DNA samples were quantified by comparison with standards.

Library preparation and sequencing were conducted at the Earlham Institute (Norwich, UK). Approximately 1 to 5 μg of genomic DNA per sample at a minimum concentration of 20 ng/μl was used in sequencing library preparation. Individual barcoded Illumina TruSeq DNA libraries were generated for each genotype. The four samples were then sequenced on one lane of Illumina HiSeq 2000, which generated between 32.2 million and 42.3 million 100–base pair (bp) paired-end reads for each sample. The first five bases of the 5′ ends of the reads were removed by using fastx_trimmer v.0.0.14 (45), and any bases with a Phred quality score below 15 were removed from the 3′ end by using cutadapt (v.1.3) (46). Illumina TruSeq adaptors where also removed by using cutadapt, where a minimum overlap of 10 bases with the adaptors, a maximum error rate of 0.1, and a minimum final read length of 50 bases were set. Lastly, fastq_quality_filter v.0.0.14 (45) was used to remove sequences with a Phred score below 20 in more than 5% of the bases. Read pairs were identified by using (47).

The reads from the ppi1 single mutant were mapped to the TAIR10 A. thaliana reference genome (48) of the Phythozome v.9.0 release (49) by using clc_mapper v. (50). A consensus sequence in FASTA format was then generated by using clc_find_variations v. (50). Also, the transcript sequences from the TAIR10 release were aligned to the reference genome to facilitate manual examination of identified mutations and to visualize whether a particular mutation occurs in an exon or intron, and so on.

The three datasets from the individual sp2 double mutants were independently aligned to the ppi1 reference genome by using clc_mapper v. The ppi1 reads were also aligned to the same reference in order to identify any variable sites resulting from allelic variation in the ppi1 line.

In silico identification of mutations

The mutagen EMS used to generate the three sp2 mutants reacts with guanine in the DNA molecule and is likely to (i) cause point mutations that change guanine to adenine (or cytosine to thymine on the reverse strand). The respective mutations affecting the three sp2 mutants were furthermore expected to (ii) occur in the same gene (or corresponding promotor region) in (iii) all three mutants but (iv) not necessarily in the exact same positions. These four search criteria were implemented in the program “” (51), which takes as input a gff3 file with gene coordinates and the single-nucleotide polymorphism variant output from clc_mapper and outputs a list of names of mutated genes from each dataset. Genes found to be mutated in all the sp2 datasets were then manually examined by visualizing the alignment data with the genome viewer IGV (v.2.3) (52).

This analysis showed that each sp2 allele contains a G-to-A point mutation within the At3g44160 gene, just to the south of the chromosome 3 centromere. In sp2-1, a mutation was detected at the splice junction preceding the final exon; this was later shown to cause mis-splicing, frameshifts, and premature termination, implying that sp2-1 is a knockout allele. In sp2-2 and sp2-3, the detected mutations were both predicted to cause an amino acid substitution. Further details are provided in the fig. S1 legend. The transmembrane β-strands and the three-dimensional structure of the SP2 protein were predicted with Phyre2 by using the Intensive modeling mode, which produced a model based on five structures for bacterial TamA and BamA proteins (c4c00a, c5ekqA, c4k3bA, c4n75A, and c4k3cA) to increase confidence (53).

Phylogenetic analysis

Sequences were obtained by BLAST searches of the Phytozome 12 database (54) (table S2). Sequences were aligned by multiple alignment using fast Fourier transform (MAFFT) (55), and manual alignment adjustments were made by using Mesquite 1.12 (Tangient). Phylogeny was inferred by using MrBayes 3.2 software (56). Two runs were performed in parallel, with each using eight Markov chain Monte Carlo chains for 8 million generations and the temperature set to 0.2. The standard deviation of split frequencies (StdDev) was 0.001228 at the end of the analysis, and it was therefore assumed to have converged. Trees were sampled every 1000 generations, reaching a total of 8000 trees. Burn-in was set to 25%, and so the first 2000 trees were discarded. The resulting phylogeny was a minimum 50% consensus of the remaining 6000 sampled trees. Parameters not mentioned were retained at the default setting.

Gene identifiers

Gene sequences for the following proteins from A. thaliana were used experimentally in this study: SP2 (At3g44160), AtCDC48A (At3g09840), Toc33 (At1g02280), CDKA1 (At3g48750), Toc159 (At4g02510), OEP7 (At3g52420), OEP80 (At5g19620), and Rubisco small subunit (SSU) (At1g67090).

Plasmid constructs

All primers used are listed in table S1. The SP1-HA, YFP-HA, FLAG-tagged ubiquitin, SP1-YFP, GST-SP1flex, and YFP-Toc33 constructs have all been described previously (7, 21). The coding sequence (CDS) for the CDC48-DN mutant (AtCDC48AK254A,K527A) was amplified from a pre-existing plasmid (H6T7-DN-B, which has an ethanol-inducible promoter) (27) by using primers that add a C-terminal FLAG tag. All other Arabidopsis CDSs (including that for WT AtCDC48A, amplified with and without a FLAG tag) were PCR amplified from Col-0 cDNA, and the CDS encoding the CDC48 “Trap” mutant (AtCDC48AE581Q) was generated by overlap-extension PCR (57). The Gateway cloning system (Invitrogen) was used to make most of the constructs, and all entry clones were verified by DNA sequencing. To generate C-terminal 6×Myc tag fusion proteins, the SP1 and SP2 CDSs were cloned into the pE3c vector (58) and then subcloned into the p2GW7 35S-driven expression vector (59) for protoplast transfection (generating the SP1-Myc and SP2-Myca constructs). The SP2 CDS, with and without the Myc tag, was cloned into the pB2GW7 binary 35S-driven overexpression vector (59) for stable plant transformation (generating the SP2-OX and SP2-Myc constructs). The C-terminally FLAG-tagged WT and DN mutant CDC48 CDSs were cloned into the pMDC7 binary vector (60) for stable plant transformation, enabling estradiol-inducible transgene expression [providing more stable expression than the original ethanol-inducible system (27)] (generating the CDC48-WT and CDC48-DN constructs). Untagged WT and Trap mutant CDC48 CDSs were cloned into both a modified p2GW7 plant expression vector providing a C-terminal hemagglutinin (HA) tag (generating the CDC48-WT-HA and CDC48-Trap-HA constructs) and the p2GWC7 (for WT) or p2GWY7 (for Trap) plant expression vector (59), which provides a C-terminal cyan fluorescent protein (CFP) or YFP tag, respectively (generating the CDC48-WT-CFP and CDC48-Trap-YFP constructs). To generate C-terminally HA-tagged Toc33, the corresponding CDS was similarly cloned into the modified p2GW7 vector (generating the Toc33-HA construct). To generate N-terminally TAP (tandem affinity purification)–tagged Toc33, the corresponding CDS was cloned into the NTAPi (N-terminal TAP tag) binary vector (61) (generating the TAP-Toc33 construct). To generate BiFC constructs, selected gene sequences were cloned into pBlueScript II SK− by using a single SmaI restriction site, sequenced, and then subcloned into 5′ KpnI and 3′ XmaI sites of the pSAT4A-cEYFP-N1 vector (62) for CDC48-cYFP; 5′ XhoI and 3′ EcoRI sites of the pSAT4A-cEYFP-N1 vector (62) for CDKA1-cYFP; 5′ EcoRI and 3′ SalI sites of the pSAT4-nEYFP-C1 vector (62) for nYFP-Toc159; and 5′ EcoRI and 3′ SalI sites of the pSAT4A-nEYFP-N1 vector (62) for OEP7-nYFP.

Transient assays and stable plant transformation

Protoplast isolation and transient assays were carried out as described previously (7, 63). When required, MG132 (Sigma), epoxomicin (Merck), or bortezomib (Selleckchem) (all three chemicals prepared as a 10 mM stock solution in dimethyl sulfoxide) or E-64 (Melford) (prepared as a 10 mM stock solution in water) was added to the protoplast culture medium at 15 hours after transfection, to a final concentration of 1 to 10, 1 to 10, 5, or 10 μM, respectively; subsequently, the culture was incubated for a further 2 to 3 hours before analysis. When protoplasts isolated from the CDC48-WT and CDC48-DN transgenic lines were used, 10 μM estradiol (prepared as a 10 mM stock solution in ethanol) was included in the culture medium throughout the incubation of protoplasts (either for 15 hours in the case of transfected protoplasts or for 2 days when transfection was not needed). For XFP fluorescence and IP assays, protoplast aliquots of 0.1 ml (105 protoplasts) or 1 ml (106 protoplasts) were transfected with 5 or 100 μg of DNA, respectively, and the samples were analyzed after 15 to 18 hours.

Transgenic lines carrying the SP2-OX, SP2-Myc, CDC48-WT, and CDC48-DN constructs were generated by Agrobacterium-mediated transformation (16, 36). Transformants were selected by using MS medium containing either phosphinothricin (for the SP2 constructs) or hygromycin B (for the CDC48 constructs). At least 12 T2 lines for each transformation were analyzed, and at least two lines with a single T-DNA insertion (which showed a 3:1 segregation on selective MS medium in the T2 generation) were chosen for further analysis.


Transmission electron microscopy was performed as described previously (16). Measurements were recorded by using at least 30 different plastids per genotype and were representative of three individuals per genotype. Chloroplast cross-sectional area was estimated as described previously (16, 34) by using the equation π × 0.25 × length × width. Numbers of thylakoid lamellae per granal stack and of interconnections between granal stacks were determined as previously described (7, 16) in at least 96 resolvable grana across three individuals per genotype.

All fluorescence microscopy and BiFC experiments were conducted at least twice with the same results, and typical images are presented. For the imaging of CFP, YFP, and chlorophyll fluorescence signals, in most cases (except for fig. S14) protoplasts were examined by using a Zeiss LSM 510 META laser-scanning confocal microscope (Carl Zeiss) as described previously (8). To visualize signals associated with chloroplasts without interference from cytosolic signals, protoplasts were ruptured by gently tapping the cover glass; this enabled the release of the cytosol and of intact chloroplasts. For fig. S14, fluorescence images were captured by using a Nikon Eclipse TE-2000E inverted microscope as described previously (36).

For BiFC assays, plasmid DNA for two constructs [one nYFP (N-terminal YFP fragment) fusion and one cYFP (C-terminal YFP fragment) fusion] was cotransfected into WT protoplasts and then YFP signals were analyzed by confocal imaging. All images were captured by using the same settings to enable comparisons. The frequency of protoplasts that successfully expressed a BiFC YFP fluorescence signal was determined by counting the number of positive cells and the total number of cells per microscope field (each field typically contained ~40 protoplasts) in approximately five microscope fields per transfection. Each combination of constructs was analyzed three times to evaluate the frequency.

In vitro translation and in vitro pull-down analysis

The SP2 and OEP80 CDSs were cloned into pBlueScript II SK− by using a single SmaI restriction site and verified by DNA sequencing. The SSU precursor construct was described previously, as was the in vitro transcription-translation procedure (16, 64).

The purification of glutathione S-transferase (GST)–SP1flex and GST proteins from bacteria and the procedure used for in vitro pull-down analysis were described previously (7).

Chloroplast isolation, protein import, and topology analysis

Chloroplasts were isolated from 14-day-old in vitro–grown plants (or, when stated, from protoplasts). Isolations, protein import, and protease treatments were performed as described previously (16, 38, 6568). The presented chloroplast protein import data are representative of three independent experiments.

Immunoblotting, immunoprecipitation, and blue native PAGE

Immunoblotting was performed as previously described (34, 69) with minor modifications. Total protein samples of 10 to 20 μg, prepared from seedlings, were typically analyzed. Primary antibodies were as follows. To identify TOC and TIC proteins, we used anti–atToc75-III antibody (36), anti-atToc159 antibody (70), anti-atToc33 (G-domain) antibody (36), anti-atTic110 antibody (71, 72), and anti-atTic40 antibody (36). To identify non-TOC OEM proteins, we used anti-OEP80 antibody (73) and anti-SFR2 antibody (74). To identify chloroplast stromal proteins, we used anti-cpHsc70 (AgriSera, AS08 348) (75), anti-Hsp93 (heat shock protein, 93 kDa) antibody (16, 76), and anti-PRPL35 antibody (7). To identify proteins of other cellular compartments, we used anti-Slp1 (mitochondria) (77), anti-calreticulin (ER) (78, 79), anti-H3 histone (Abcam) (nucleus) (36), and anti-PEX13 and anti-PEX14 (the latter from Agrisera, AS08 372) (peroxisome) (20, 21). Other primary antibodies we used were anti–HA tag (Sigma), anti–c-Myc tag (Sigma), anti–green fluorescent protein (GFP) (which detects both GFP and YFP) (Sigma), and anti–FLAG tag (Sigma). As most of the proteins analyzed in this study were membrane proteins, we used Slp1 (a nonchloroplastic membrane protein) and Tic110 [an internal chloroplast membrane protein unaffected by CHLORAD (7, 8)] as loading controls.

Secondary antibodies were anti–rabbit immunoglobulin G (IgG) conjugated with horseradish peroxidase (Santa Cruz Biotechnology) or, in the case of anti–c-Myc and anti-FLAG, anti–mouse IgG conjugated with horseradish peroxidase (GE Healthcare). Chemiluminescence was detected by using ECL Plus Western blotting detection reagents (GE Healthcare) and an LAS-4000 imager (Fujifilm). Band intensities were quantified by using Aida software (Raytest). Quantification data were obtained from the results of at least three experiments all showing a similar trend. Typical images are shown in all figures.

For the IP of HA-tagged proteins, total protein (~500 mg) was extracted from protoplasts in IP buffer (25 mM tris-HCl, pH 7.5; 150 mM NaCl; 1 mM EDTA; 1% Triton X-100) containing 0.5% plant protease inhibitor cocktail (PPIC) (Sigma) and centrifuged at 20,000 × g for 10 min at 4°C. The clear lysate was then incubated with 50 μl of EZview Red anti-HA affinity gel (Sigma) for 2 hours to overnight at 4°C with slow rotation. After six washes with 500 μl of IP washing buffer (25 mM tris-HCl, pH 7.5; 150 mM NaCl; 1 mM EDTA; 0.5% Triton X-100), bound proteins were eluted by boiling in 2× SDS–polyacrylamide gel electrophoresis (PAGE) loading buffer [50 mM tris-HCl, pH 6.8; 20% glycerol; 1% sodium dodecyl sulfate (SDS); and 0.1 M dithiothreitol (DTT)] for 5 min and analyzed by SDS-PAGE and immunoblotting. A similar procedure was adopted for the IP of Myc-tagged proteins, except that 50 μl of EZview Red anti–c-Myc affinity gel (Sigma) was used instead of the anti-HA gel. For detection of ubiquitinated proteins, the IP buffer also contained 10 mM N-ethylmaleimide (NEM) (Sigma).

Two-dimensional blue native (BN)–PAGE was performed by using a procedure described previously (80).

TAP and mass spectrometry

Chloroplasts were isolated from a complemented ppi1 mutant line carrying the TAP:Toc33 construct and then used as starting material for TAP. The TAP procedure was performed as described previously (81), with the omission of the secondary affinity purification step, which was not essential for our analysis. The tobacco etch virus nuclear-inclusion-a endopeptidase eluates were concentrated 1:10 by using Vivaspin 500 ultrafiltration spin columns (Sartorius Stedim Biotech), boiled with 1 volume of 2× SDS-PAGE loading buffer, and loaded onto SDS-PAGE gels for analysis. Silver staining was used to visualize proteins and estimate their sizes and migration positions. For the identification of CDC48, the 75- to 100-kDa region of a Coomassie brilliant blue–stained SDS-PAGE gel slice was subjected to in-gel trypsin digestion and liquid chromatography–tandem mass spectrometry analysis. Scaffold (Proteome Software) and Mascot database searches were used to interpret the results.

In vivo retrotranslocation assays

The method used for in vivo retrotranslocation assays was adapted and modified from similar approaches commonly applied in ERAD studies (23). Transfected protoplasts were used for these assays to facilitate the detection of the substrate and enable the efficient and uniform application of proteasome inhibitor; in addition, protoplastation applies a stress that triggers the TOC degradation process (8). First, the substrate (Toc33-HA or SP1-HA) and ubiquitin were transiently overexpressed in 106 protoplasts for each genotype to increase the detection sensitivity for higher–molecular weight (ubiquitinated) forms of the substrate (82). The transformed protoplasts were incubated for 15 hours, and then bortezomib was applied to a final concentration of 5 μM before an additional 3 hours of incubation. Subsequent fractionation steps to produce separate chloroplast and cytosol samples were all carried out on ice or at 4°C and used previously described procedures with modifications (65, 83). Protoplasts were pelleted by centrifugation at 100 × g for 2 min and gently resuspended with protoplast-washing buffer (500 mM mannitol; 4 mM 4-morpholineethanesulfonic acid–KOH, pH 5.6). Then the protoplasts were pelleted again and resuspended by gentle agitation in 500 μl of HS buffer [50 mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES)–NaOH, pH 8.0; 0.3 M sorbitol] containing 0.5% PPIC and 5 μM bortezomib and gently forced twice through 10-μm nylon mesh to release chloroplasts. The collected flow-through was centrifuged at 1000 × g for 5 min to produce a chloroplast-containing pellet and a cytosol-containing supernatant (S1). The pellet was gently resuspended in 500 μl of HS buffer, and the chloroplasts were purified by a two-step Percoll (Fisher Scientific) gradient (38). Intact chloroplasts were washed with 500 μl of HS buffer and then pelleted by centrifugation at 1000 × g for 5 min. The S1 sample was centrifuged at 10,000 × g for 15 min. The resulting supernatant (S10) was recovered and ultracentrifuged at 100,000 × g for 1 hour, producing a further supernatant (S100) that was concentrated to 50 μl by using Vivaspin 500 ultrafiltration spin columns; this was the cytosolic fraction. The pelleted chloroplasts were lysed in 100 μl of denaturing buffer (25 mM tris-HCl, pH 7.5; 150 mM NaCl; 5 mM EDTA; 10 mM NEM; 1% SDS; 2% Sarcosyl; 5 mM DTT) containing PPIC, whereas the cytosolic fraction was mixed with 50 μl of 2× denaturing buffer containing PPIC. Lastly, the substrate protein was purified by IP by using a previously described procedure to improve the sensitivity of detection of the ubiquitinated protein (7). Experiments were repeated three times, and similar results were obtained.

In vitro retrotranslocation assays

The method for in vitro retrotranslocation assays was adapted and modified from previous reports (84, 85). Approximately 4 × 106 protoplasts from CDC48-WT or CDC48-DN plants were incubated with 10 μM estradiol for 2 days to induce abundant expression of CDC48 protein in the cytosol. After incubation, the protoplasts were washed once with 2 ml of protoplast-washing buffer, and then a cytosolic fraction was prepared from each genotype, essentially as described above except that the protoplasts were resuspended in import buffer [50 mM HEPES-NaOH, pH 8.0; 3 mM MgSO4; 0.3 M sorbitol; 5 mM Mg–adenosine triphosphate (ATP); 20 mM gluconic acid (potassium salt); 10 mM NaHCO3; 0.2% bovine serum albumin] containing 5 μM bortezomib and complete EDTA-free protease inhibitor cocktail (PIC) (Roche) (the latter according to the manufacturer’s instructions), instead of HS buffer. Cytosol was concentrated to ~10 mg/ml by using Vivaspin 500 ultrafiltration spin columns and was used immediately or aliquoted and stored at −80°C. Chloroplasts were prepared from 106 CDC48-DN protoplasts transiently overexpressing SP1-HA and FLAG-tagged ubiquitin (and induced with 10 μM estradiol) by using the procedure described above. Chloroplast pellets were gently resuspended in reaction buffer [import buffer containing 30 mM MgATP, ubiquitin (Sigma) at 0.1 μg/μl, 5 μM bortezomib, and PIC] and then immediately used in in vitro retrotranslocation reactions. Each reaction mixture contained reaction buffer (to a final reaction volume of 50 μl), 10 μl of cytosol, and 2 × 107 chloroplasts. Reaction mixtures were incubated at 25°C for 1 hour with occasional agitation to resuspend the chloroplasts, and reactions were then stopped by the addition of 1 μl of 250 mM NEM to a final concentration of 5 mM. Subsequently, at 4°C, the chloroplasts were pelleted by centrifugation at 10,000 × g for 1 min, and the resulting supernatant was further centrifuged at 20,000 × g for 10 min to produce another supernatant corresponding to the cytosolic fraction. The SP1-HA protein was enriched from the resulting chloroplast and cytosol fractions by IP, as described above. Ubiquitinated SP1 protein (either resident in the chloroplasts or retrotranslocated into the cytosol) was detected by anti-FLAG immunoblotting; use of transiently expressed FLAG-tagged ubiquitin in this way increased the detection sensitivity. Experiments were repeated three times, and similar results were obtained.

Statistical analysis

Statistical calculations (mean, SEM, and t test) were performed by using Microsoft Excel software. The statistical significance of differences between two experimental groups was assessed by using a two-tailed Student’s t test. Differences between two datasets were considered significant at P < 0.05.

Supplementary Materials

Figs. S1 to S20

Tables S1 and S2

References (8692)

References and Notes

Acknowledgments: We thank M. Rashbrooke for assistance with initial analyses and rough mapping of sp2; Y. Zeng, J. Bédard, and N. Li for technical assistance; N. Allcock and S. Hyman (University of Leicester EM Laboratory) for electron microscopy; A. R. Bottrill (University of Leicester Protein Nucleic Acid Chemistry Laboratory) for mass spectrometry; L. Dolan for comments on the manuscript; L. J. Sweetlove (Slp1), J. Denecke (calreticulin), and B. Bartel (PEX13) for antibodies; F. Wu for the Toc159 BiFC construct; S. Y. Bednarek for the H6T7-DN-B construct; G. Benvenuto (and Addgene) for the pE3c construct; J. Chory and J. Woodson for ppi2-3 (fts1) seeds; and SIGnAL and NASC for the sp2-4 allele. Funding: This work was supported by grants from the BBSRC (BB/D016541/1, BB/K018442/1, BB/R009333/1, and BB/R016984/1) to R.P.J.; the Royal Society Rosenheim Research Fellowship to R.P.J.; a Department of Plant Sciences DPhil studentship to R.P.J and W.B.; a Gatsby Sainsbury Ph.D. studentship to R.T.; and a Carl Tryggers Stiftelse för Vetenskaplig Forskning fellowship (CTS 11:479) to M.T. Author contributions: Q.L. and W.B. formulated the research plan, designed and performed experiments, and interpreted results. Q.L. carried out the functional analysis of CDC48 and contributed to the preparation of the manuscript. W.B. completed genetic, molecular, physiological, and detailed functional studies of SP2, with assistance from Q.L. R.T. performed the genetic mapping of sp2, prepared samples for whole-genome sequencing, and conducted protein import assays with assistance from T.D.S. M.T. analyzed the whole-genome sequencing data and identified a candidate sp2 locus. P.L. performed the TAP experiment. A.B. conducted the mutant screen and identified the original sp2 alleles. R.P.J. conceived of the study, supervised the work, analyzed the data, and prepared the manuscript. All authors discussed the results and commented on the manuscript. Competing interests: This work is the subject of pending UK patent applications GB 1803833.1, GB 1803834.9, and GB 1815206.6 (the inventor is R.P.J. in each case), which cover the use of CHLORAD to manipulate plastid development in crops. Data and materials availability: Sequencing data used in this study, as well as the reference genome for ppi1, are available at the National Center for Biotechnology Information under Bioproject PRJNA488548. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRoteomics IDEntifications (PRIDE) repository with the identifier PXD010954. All other data are available in the manuscript or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article