The diatom Phaeodactylum tricornutum harbors a plastid that is surrounded by four membranes and evolved by way of secondary endosymbiosis. Like land plants, most of its plastid proteins are encoded as preproteins on the nuclear genome of the host cell and are resultantly redirected into the organelle. Because two more membranes are present in diatoms than the one pair surrounding primary plastids, the targeting situation is obviously different and more complex. In this work, we focus on preprotein transport across the second outermost plastid membrane—an issue that was experimentally inaccessible until now. We provide first indications that our hypothesis of an ERAD (ER-associated degradation)-derived preprotein transport system might be correct. Our data demonstrate that the symbiont-specific Der1 proteins, sDer1-1 and sDer1-2, form an oligomeric complex within the second outermost membrane of the complex plastid. Moreover, we present first evidence that the complex interacts with transit peptides of preproteins being transported across this membrane into the periplastidal compartment but not with transit peptides of stromal-targeted proteins. Thus, the sDer1 complex might have an additional role in discriminating preproteins that are transported across the two outermost membranes from preproteins directed across all four membranes of the complex plastid. Altogether, our studies of the symbiont-specific ERAD-like machinery of diatoms suggest that a preexisting cellular machinery was recycled to fulfill a novel function during the transition of a former free-living eukaryote into a secondary endosymbiont.
Proteins that are incorrectly folded must either be refolded into their correct structure or—if that fails—targeted for degradation. Especially in compartments lacking a protein degradation machinery, proteins designated for destruction must be targeted to a subcellular localization providing the necessary activities. This is well studied for the endoplasmic reticulum, which harbors sorting and retrotranslocation mechanisms for misfolded luminal proteins as well as for misfolded membrane proteins (Meusser et al. 2005; Nakatsukasa and Brodsky 2008). In the case of luminal proteins, a protein complex termed ERAD-L (ER-associated degradation of luminal proteins) retrotranslocates misfolded proteins into the cytosol, where they are ubiquitinated and degraded by the proteasome (Ismail and Ng 2006). Recently, we have postulated that components derived of an ERAD-L machinery were relocated and converted to fulfill duties in preprotein translocation in some plastids (Sommer et al. 2007).
Many ecologically important algae like diatoms, which are believed to be responsible for one-fifth of primary productivity on earth (Field et al. 1998) as well as highly pathogenic parasites like Plasmodium falciparum came about by intracellular enslavement of a red alga within another unicellular eukaryotic cell (Douglas et al. 2001; Stoebe and Maier 2002; Cavalier-Smith 2003). By way of this process, termed secondary endosymbiosis, organisms evolved, which harbor second hand plastids surrounded by three or four membranes (Gould et al. 2008).
Like in other phototropic eukaryotes, the plastid genome of secondarily evolved algae encodes only a small subset of proteins acting in the stroma or the thylakoids, whereas most plastid proteins are encoded as preproteins in the nucleus of the host cell. Thus, nucleus-encoded plastid proteins have to traverse two, three, four, or—if localized in the thylakoid lumen—five membranes to reach their final destination (Gould et al. 2007; Hempel et al. 2007). To ensure correct targeting in such a complex system, all nuclear-encoded plastid proteins carry a bipartite topogenic signal (BTS) at the N-terminus, which consists of a signal peptide and a transit peptide and which is recognized by specific machineries in each membrane (Bolte et al. 2009).
In case of heterokontophytes, haptophytes, cryptophytes, and apicomplexa, which originated by the enslavement of a red algal endosymbiont, the so-called complex plastid is surrounded by four membranes (Hempel et al. 2007). Analyses of the protein import mechanisms led to the impression that transport components already present in the progenitors of the symbioses were recycled. For example, in diatoms, a Sec61 complex enables protein transport across the outermost membrane, which is connected to the ER of the host cell (see schematic overview, fig. 1A) (Bolte et al. 2009). Subsequently, the signal peptide is cleaved off exposing the second part of the topogenic signal, the transit peptide, which is essential for targeting across the remaining membranes. For transport across the second outermost plastid membrane (II.), we postulate that the symbiont-specific ERAD complex was converted to mediate preprotein translocation (Sommer et al. 2007). The translocon of the following membrane, the third outermost membrane (III.), remains elusive, whereas transport across the innermost membrane (IV.) is most likely facilitated by a translocon of the inner chloroplast membrane (van Dooren et al. 2008). Finally, proteins for the thylakoids may use similar import mechanisms as found in land plants (Gould et al. 2007).
In this work, we concentrate on preprotein transport across the second outermost plastid membrane of the diatom Phaeodactylum tricornutum and provide the first experimental support that our hypothesis of plastid preprotein transport via an ERAD-derived system might be correct. This symbiont-specific ERAD-derived transport machinery will be referred to as SELMA (symbiont-specific ERAD-like machinery) from this point forward to clearly separate it from the genuine ERAD system of the host. Initially, the basis for our assumption came from the finding that in cryptophytes, heterokontophytes, and apicomplexa components of the ERAD system were detected in two independent versions, one set of ERAD-L components for the host and a second set of homologous versions for the symbiont (Sommer et al. 2007). In cryptophytes, most of the symbiont-specific ERAD components are encoded on the nucleomorph, the highly reduced nucleus of the red algal endosymbiont. Further analyses revealed that in heterokontophytes and apicomplexa, which are lacking a nucleomorph genome, these symbiont-specific ERAD factors are encoded as preproteins on the host nucleus and are equipped with N-terminal extensions (Sommer et al. 2007). Such N-terminal BTSs have been shown to direct GFP fusion proteins into the space between the second and third outermost plastid membrane, the periplastidal compartment (PPC), that is, the former cytosol of the red algal symbiont (Sommer et al. 2007). The fact that a symbiont-specific ERAD-like machinery (SELMA) was retained in all these organisms, although the ER within the highly reduced PPC was lost, is surprising and highlights the important degradation-independent role of SELMA. Because ERAD provides a complex retrotranslocation machinery for protein transport from the ER to the cytosol, the system seems predestined to fulfill the postulated function.
One central component of the ERAD-L system is the membrane protein Der1 (degradation at the ER) (Knop et al. 1996), which is a favored candidate for forming the retrotranslocation channel and was detected in two host-specific versions (hDer1-1, hDer1-2) and two symbiont-specific versions (sDer1-1 and sDer1-2) not only in P. tricornutum but also in other relatives (Sommer et al. 2007). In order to specify SELMA of diatoms more precisely, we performed localization studies with both sDer1 of P. tricornutum. Our results show that the membrane domains of both sDer1 proteins are embedded in the second outermost plastid membrane as we previously postulated. Interaction studies with the sDer1 proteins indicated that sDer1-1 and sDer1-2 form homo-oligomers and hetero-oligomers, a finding which may be important to define the pore-forming unit of SELMA. Finally, we present first evidence that SELMA is indeed involved in preprotein targeting across the second outermost plastid membrane and may additionally represent a point of discrimination for proteins transported either across two or four membranes of complex plastids.
Materials and Methods
Polyclonal antibodies against sDer1-1 and sDer1-2 of P. tricornutum were raised in rabbit and affinity purified. Antibodies are specific for peptides within the C-terminal region of the sDer1 proteins. The GFP antibody used for coimmunoprecipitation is a polyclonal antibody directed against the full-length protein (Biomol). The BiP antibody used as a control for carbonate extraction is commercially available from Abcam.
All P. tricornutum sequences were amplified using standard polymerase chain reaction (PCR) conditions. All sequences can be retrieved from P. tricornutum database (PhatrDBv2.0). For standard GFP localization studies, sDer1-1 (ID 31697), sDer1-2 (ID 35965), and hDer1-2 (ID 37614) sequences were cloned full length in front of eGFP into pPhaT1 as described previously (Sommer et al. 2007). For self-assembling GFP and split-GFP analyses, a modified pPhaT1 vector (for original pPhaT1, see Zaslavskaia et al. 2000) was used, which allows dual protein expression under inducible conditions. The pPha-DUAL[2xNR] vector contains two multiple cloning sites both under the control of the endogenous nitrate reductase promoter, which can be regulated by a switch from ammonium- to nitrate-containing medium. All topogenic signals used for self-assembling GFP and split-GFP studies were described and analyzed previously (Kilian and Kroth 2005; Gould, Sommer, Kroth, et al. 2006; Sommer et al. 2007). Mutations within the topogenic signals of Hsp70 and AtpC were introduced by PCR changing the first codon of the Hsp70 transit peptide from CAT to TTT. In case of AtpC, the first amino acid of the transit peptide was changed by replacing TTC by GCC. C-terminally truncated versions of sDer1-1 and sDer1-2 were created by PCR defining the cut off for sDer1-1 after amino acid 223 and after amino acid 250 in case of sDer1-2, respectively.
Localization Analyses Using Self-assembling GFP
The self-assembling GFP system was established earlier (Cabantous et al. 2005) and used previously to specify protein localization (van Dooren et al. 2008). To check on sDer1-1 and sDer1-2 localization and topology, we fused the small GFP fragment (GFP-11) to the C-terminus of the full-length sDer1-1/sDer1-2 protein and the C-terminally truncated versions of sDer1-1 and sDer1-2, respectively. The large GFP fragment (GFP1–10) was fused to the C-terminus of marker proteins for different compartments of the cell. As an ER marker, we used the endogenous protein disulfide-isomerase in full length (ID 44937), whereas the BTS of sHsp70 was used to direct the large GFP fragment to the PPC. As a negative control, the large GFP fragment was fused to the ER marker and cotransfected with the small GFP fragment fused to the BTS of stromally targeted AtpC. As a positive control, both fragments were directed to the PPC using the topogenic signals of sHsp70 and sUbc4, respectively. All combinations of sDer1 proteins and marker proteins mentioned in the text were cloned into pPha-DUAL[2xNR]. Phaeodactylum tricornutum cells were transformed as described previously (Apt et al. 1996), with the exception that transformation was carried out under noninduced conditions in medium containing 1.5 mM NH4 as solely nitrogen source. Transformants were checked by colony PCR for genomic integration of both constructs, and positive colonies were grown in liquid medium containing 1.5 mM NH4. For microscopic analysis, cells were harvested, washed once in nitrogen free medium, and transferred to NO3 (0.9 mM) containing medium to induce expression of self-assembling GFP constructs. After 6 h of induction, transformants for all construct types were analyzed with a confocal laser scanning microscope Leica TCS SP2 using a HCX PL APO 63x/1.32–0.6 oil Ph3 CS objective. GFP and chlorophyll fluorescence was excited at 488 nm, filtered with beam splitter TD 488/543/633, and detected by two different photomultiplier tubes with a bandwidth of 500–520 and 625–720 nm for GFP and chlorophyll fluorescence, respectively.
In vivo Protein-protein Interaction Analysis Using the Spilt-GFP System
The principle of split-GFP was described previously (Ghosh et al. 2000; Kerppola 2006). For protein–protein interaction studies, we expressed the proteins of interest with the C-terminus fused to the split eGFP fragments N-GFP (aa 1–155) and C-GFP (aa 156–245), respectively. Split-GFP constructs were cloned into pPha-DUAL[2x NR]. For interaction studies, we used the following combinations. Homo-oligomerization of sDer1-1 and sDer1-2 was checked by fusing the proteins to both split-GFP fragments. Hetero-oligomerization was checked by fusing sDer1-1 to N-GFP and sDer1-2 to C-GFP. Interaction of topogenic signals with both sDer1 proteins was analyzed by fusing the topogenic signals (BTS of AtpC, FcpD, sHsp70, and sCdc48) to N-GFP and the sDer1 proteins to C-GFP. As negative control, both GFP fragments were directed to the PPC by fusing them with the topogenic signals of sDer1-1 and sDer1-2. Transformation was carried out under noninduced conditions (see Methods on Localization Analyses Using Self-assembling GFP), and transformants were checked by colony PCR for integration of both constructs. Positive colonies were cultured on plates containing 0.9 mM NO3 and were analyzed with a confocal laser scanning microscope.
Phaeodactylum tricornutum cells were grown in liquid culture as described previously (Apt et al. 1996). For coimmunoprecipitation, 3 × 109P. tricornutum cells were harvested by centrifugation (5 min, 1,000 × g), washed once in phosphate buffered saline, and resuspended in 5 ml IP buffer (50 mM Tris HCl ph 7.5, 200 mM KOAc, 1 mM EDTA, 10% glycerol, 10 mM NEM, 25 μg/ml e64d, and protease inhibitor cocktail). Cells were disrupted using a French press (1,000 psi, five repeats), and intact cells were removed by centrifugation (5 min, 1,000 × g). Membranes were collected by centrifugation (30 min, 120,000 × g) and resuspended in 1.5 ml IP buffer. For membrane solubilization, 2% Digitonin (Calbiochem) was added and the extract was incubated for 4 h at 4 °C with agitation. After centrifugation (10 min, 10,000 × g), the supernatant was loaded on 100 μl of a 50% suspension of protein-A agarose beads (Thermo Scientific) covalently cross-linked to GFP antibody/sDer1-2 antibody, respectively. Beads were incubated with agitation over night at 4 °C. After washing the beads twice with IP buffer containing 0.5% Digitonin, proteins were eluted with 50 μl 1× SDS-buffer at 95 °C for 5 min. Aliquots of 25 μl were used for SDS-PAGE and Western blot analysis.
Phaeodactylum tricornutum cells were harvested by centrifugation, resuspended in membrane solubilization buffer (50 mM NaCl, 50 mM imidazole/HCl ph 7, 2 mM 6-aminohexanoic acid, 1 mM EDTA, 8.5% sucrose), and disrupted with a French press (1,000 psi, five repeats). Intact cells were removed by centrifugation (5 min, 1,000 × g), and membranes within the supernatant were collected by centrifugation for 30 min at 120,000 × g. The membrane pellet was solubilized in carbonate buffer pH 11.5 and incubated on ice for 30 min. Membranes were again collected by centrifugation and proteins of membrane and supernatant fractions were precipitated using trichloroacetic acid. Equal portions of all fractions were loaded on SDS-PAGE and analyzed by Western blot using specific antibodies against sDer1-1 and sDer1-2.
For electron microscopic analyses, wild-type P. tricornutum cells were harvested by centrifugation (5 min, 1,000 × g) and fixed for 4 h with 0.02% glutaraldehyde and 4% paraformaldehyde in f/2 medium. Cells were washed in Sorensen's buffer, dehydrated in a graded ethanol series, and embedded in lowicryl resin. Sections were cut with a diamond knife and mounted on pioloform-coated grids. Primary antibodies against sDer1-1 and sDer1-2 were detected with a secondary antibody coupled to 30 nm gold particles (Biocell). Samples were poststained with uranyl acetate and lead citrate under standard conditions.
Localization of the symbiont-specific ERAD-like machinery (SELMA)
To characterize SELMA, we concentrated on the genetically accessible diatom P. tricornutum (Bozarth et al. 2008). Recently, we showed that P. tricornutum encodes in addition to the host-specific ERAD-L system, a second ERAD-like system (SELMA), which is localized within the symbiont (Sommer et al. 2007) and which we postulate to be involved in plastid preprotein targeting. One of the first components identified in a symbiont-specific version was the protein Der1, which is a core component of the genuine ERAD-L pathway and as a membrane protein a candidate for the translocation channel. Der1 is present in two host-specific versions (hDer1-1, hDer1-2) and in silico analyses revealed two additional copies (sDer1-1, sDer1-2), which are equipped with a targeting signal specific for the remnant cytoplasm of the eukaryotic symbiont (Sommer et al. 2007). In order to localize these membrane proteins within the cell, we first expressed proteins, in which the complete sDer1 copies are C-terminally fused to GFP. As shown in figure 2A, a host Der1 fusion shows GFP fluorescence in the ER (c), whereas the sDer1-1 and sDer1-2 fusion proteins localize in a blob-like structure (a + b), which was shown earlier to represent the PPC (Gould, Sommer, Hadfi, et al. 2006; Gould, Sommer, Kroth, et al. 2006). In order to demonstrate that the sDer1 proteins are indeed membrane-integrated proteins as it is known for the Der1 proteins from yeast and the mammal system (Hitt and Wolf 2004; Lilley and Ploegh 2004), we used carbonate extraction to separate integral membrane proteins from membrane-associated proteins and the soluble fraction. As indicated in figure 2B, both proteins, sDer1-1 and sDer1-2, were detected in the membrane fraction exclusively. Further evidence for membrane localization came from electron microscopic studies with immunogold labeling. Polyclonal antibodies directed against sDer1-1 and sDer1-2 adorn one of the plastid-surrounding membranes (fig. 2C).
To specify the target membrane of sDer1-1 and sDer1-2 more accurately and to determine protein topology, we made use of an established strategy involving engineered self-assembling GFP fragments (Cabantous et al. 2005). For such an approach, the C-terminal beta strand of GFP (GFP-11) is separated from the rest of the protein (GFP1–10). Only when both of these truncated GFP components are localized within the same cellular compartment will GFP self-assemble and become detectable by its fluorescence, as has recently been shown in targeting experiments in Toxoplasma gondii (van Dooren et al. 2008). By applying this approach to the diatom, we noticed that successful adaptation in P. tricornutum critically depends on the expression time of the fusion proteins (see Materials and Methods/Discussions). We C-terminally fused both sDer1-copies with the GFP-11 fragment, directing each fusion protein to its target membrane. In parallel, we targeted the second GFP fragment, GFP1–10, either to the ER or to the PPC using topogenic signals specifying intracellular targeting in the diatom (Gould, Sommer, Kroth, et al. 2006). As a result, we obtained fluorescence with both sDer1 versions only when GFP1–10 was directed to the PPC indicating that the C-terminus of both proteins is indeed localized within the PPC as predicted. There is no fluorescence in combination with the ER marker (fig. 2D, a + d and b + e). According to in silico predictions and as shown before in yeast, Der1 proteins have four transmembrane domains with the N- and C-terminus located in the cytoplasm (Hitt and Wolf 2004). If this were also to be true for the sDer1 proteins of P. tricornutum, we would predict that by deleting the C-terminal membrane helix of sDer1, the C-terminus of the protein should then point to the ER lumen. Hence, we next used such truncated versions of both sDer1s and fused these constructs to GFP-11. Fluorescence now appeared in combination with the GFP1–10 ER marker (fig. 2D, c + f), indicating that the sDer1 proteins are indeed localized within the second outermost membrane with the C-terminus in the PPC and when C-terminally truncated on the ER side, respectively. Oddly enough, we still observed fluorescence combining the truncated versions with the PPC marker (data not shown), but as the sDer1 proteins do interact with targeting signals for the PPC (results presented below), it is very likely that GFP fragments are close enough for self-assembly during transport of the PPC marker protein. Control experiments were conducted either by directing one self-assembling GFP fragment to the ER and the other to the plastid stroma (negative control) or by directing both to the PPC (positive control), which resulted in a lack of fluorescence and periplastidal GFP fluorescence, respectively (supplementary fig. S1, Supplementary Material online).
sDer1 Proteins Interact Homo- and Heterotypically
Localization studies showed that the proteins sDer1-1 and sDer1-2 are both localized within the second outermost membrane of the complex plastid. The Der1 protein in yeast and the mammal versions Derlin1 and Derlin2 are favored candidates for the retrotranslocation channel in the ERAD-L pathway as they are part of an oligomeric complex and have shown to interact with different ERAD substrates (Lilley and Ploegh 2004; Ye et al. 2004; Lilley et al. 2006; Bernardi et al. 2008). In order to inspect whether sDer1-1 and sDer1-2 do interact with each other, we first studied protein interactions by coimmunoprecipitation. Using protein extract of a culture expressing a sDer1-2 GFP fusion protein, we immunoprecipitated with an antibody against GFP. Thereby, we were able to pull down not only the sDer1-2 fusion protein but additionally the endogenous sDer1-2 protein and the sDer1-1 protein (fig. 3A, upper panel). These results indicate that the sDer1-2 protein is able to form homo-oligomers as well as hetero-oligomers together with the sDer1-1 protein. As a negative control, we used protein extract of a culture expressing GFP in the cytosol. By pulling down GFP, neither sDer1-1 nor sDer1-2 was coimmunoprecipitated (fig. 3A, lower panel). Heterotypic interaction of sDer1-2 and sDer1-1 was additionally verified by using an antibody against the endogenous sDer1-2 protein for coimmunoprecipitation. In the reverse experiment, using protein extract of a culture expressing the sDer1-1 protein as a GFP fusion, we again observed heterotypic interaction between the sDer1-1 GFP fusion and the endogenous sDer1-2 protein (supplementary fig. S2, Supplementary Material online). These in vitro results were next examined in an in vivo interaction assay using a split-GFP approach. In such an assay, GFP is split into an N-terminal and a C-terminal fragment (N-GFP and C-GFP). For interaction studies, each potential interaction partner is fused to one of the GFP fragments, whereby only interaction between the proteins of interest can allow N-GFP and C-GFP to assemble and subsequently show fluorescence (Kerppola 2006). Using the split-GFP approach, we detected homotypic interactions for sDer1-2 as well as for sDer1-1. Additionally, we observed the formation of a hetero-oligomeric complex between sDer1-1 and sDer1-2. In all cases, we obtained a blob-like structure, indicating the formation of oligomeric interactions of the sDer1 versions in the PPC (fig. 3B, a–c). As a negative control, we fused only the topogenic signals of the sDer1 proteins to the GFP fragments, directing both fragments to the PPC. The negative control showed no GFP fluorescence (fig. 3B, d)
sDer1 Proteins Interact with the Transit Peptide of PPC Proteins but Not with Stromal Proteins
We have shown that both sDer1 proteins are located within the second outermost membrane of the complex plastid and are able to interact homo- and heterotypically. As we postulate that SELMA is involved in plastid preprotein targeting, we next wanted to investigate the interaction of sDer1 proteins with a potential target during transport, the transit peptide of nucleus-encoded proteins. For this, we again used the split-GFP approach, in which fluorescence is—according to our experimental strategy—only recovered when the sDer1 proteins are interacting with either the transit peptide of a PPC-located Hsp70 or of a stroma-specific AtpC protein. The topogenic signals of both proteins, Hsp70 and AtpC, were previously shown to have targeted GFP into the PPC or the stroma, respectively (Kilian and Kroth 2005; Gould, Sommer, Kroth, et al. 2006). For the split-GFP assay, we fused the sDer1 proteins C-terminally with the C-GFP fragment and the topogenic signals of either Hsp70 or AtpC with the second fragment, N-GFP. Interestingly, in the split-GFP assay sDer1-1 as well as sDer1-2 show interaction with the transit peptide of the PPC-located Hsp70 but not with the transit peptide of the stromal protein AtpC (fig. 4, a–d). The same results were obtained when testing another stroma- (FcpD) and PPC-specific (sCdc48) topogenic signal for interaction with the sDer1 complex (supplementary fig. S3, Supplementary Material online). We recently showed that the first amino acid of the transit peptide is an important signal for discriminating PPC-targeted proteins from stroma-targeted proteins (Gould, Sommer, Hadfi, et al. 2006). While stroma-targeted proteins contain an aromatic amino acid (or in rare cases a leucin) at the first position of the transit peptide, the same position is strictly nonaromatic in case of periplasmatic proteins (Kilian and Kroth 2005; Gould, Sommer, Hadfi, et al. 2006; Gruber et al. 2007). Furthermore, we have previously demonstrated that by mutating a nonaromatic to an aromatic amino acid at this position, a PPC-specific protein can be directed across all four membranes into the plastid stroma (Gould, Sommer, Hadfi, et al. 2006). Hence, we became increasingly interested in whether an aromatic amino acid at the first position of the transit peptide may contribute to the differential binding of transit peptides to the sDer1 proteins. We therefore changed the phenylalanine (F), located at the first position of the transit peptide of AtpC into an alanine (A), and checked again on sDer1 interaction. Interestingly, this modification led to the recovery of GFP fluorescence in the split-GFP system applied (fig. 4, g + h). In a reverse experiment, we introduced a phenylalanine at the first position of the symbiont-specific, PPC-located Hsp70 transit peptide. This construct showed no GFP fluorescence anymore, indicating that modification of the first amino acid of the transit peptide suppresses interaction with the sDer1 proteins (fig. 4, e + f).
Protein transport into plastids surrounded by four membranes was long a subject of much speculation. Recently, we have identified a symbiont-specific ERAD-like machinery (SELMA) in secondarily evolved algae, which harbor a complex plastid of red algal origin (Sommer et al. 2007). As SELMA was detected in cryptophytes, heterokontophytes, and apicomplexa (Sommer et al. 2007) and because one of the central functions of ERAD is retrotranslocation of proteins across a biomembrane (Tsai et al. 2002; Ismail and Ng 2006; Vembar and Brodsky 2008), we had postulated that the symbiontic ERAD-like system provides the translocation machinery for transport across the second outermost membrane of plastids surrounded by four membranes (Sommer et al. 2007). Due to the occurrence of two versions of sDer1 proteins, the supposition was made that SELMA may not only form a pore in the second but also in the third outermost membrane (Gould et al. 2008).
In this study, we have investigated the cellular localization of both sDer1 proteins of P. tricornutum and studied their interactions. First, we showed by carbonate extraction that sDer1-1 and sDer1-2 are indeed integral membrane proteins (fig. 2B), which are localized within one of the plastid surrounding membranes as indicated by electron microscopic investigations (fig. 2C). Next, we expressed the full-length proteins fused to GFP, highlighting either membrane two or three as the potential target membrane, as we obtained GFP fluorescence in an intensive blob-like structure, corresponding to the lumen between the second and third outermost plastid membrane, the PPC (fig. 2A). Finally, we adapted a self-assembling GFP approach to the diatom to clarify the target membrane and topology of sDer1-1 and sDer1-2, noticing in the process that this system reacts very sensitive to the expression time of the GFP constructs. At expression times longer than 24 h, self-assembly of the GFP fragments was obtained in nearly all combinations as indicated by fluorescence. This may indicate that transport intermediates are fixed by overexpression of the GFP-11 and GFP1–10 fusion proteins, which led to assembly of GFP regardless the fusion protein combination investigated. Therefore, in order to avoid overloading the transport mechanisms, we used shorter expression times (6 h), which resulted in the finding that fluorescence is recovered only when full-length sDer1 proteins are combined with the PPC marker and when C-terminally truncated versions are transformed together with the fragment directed to the ER. Thus, our data imply that the sDer1 proteins are indeed located within the second outermost plastid membrane with the C-terminus in the PPC.
Although the ERAD-L pathway of yeast and mammals has been studied in great detail and many of the components involved in protein recognition, translocation, and degradation have been identified in the last 12 years, the pore-forming capacity of the ERAD-L protein complex has yet to be specified. The Sec61 complex and the Der1 family proteins are so far most likely candidates to form the retrotranslocation channel (Ye et al. 2004; Meusser et al. 2005; Nakatsukasa and Brodsky 2008; Willer et al. 2008). In a bioinformatic approach, we have investigated whether a second, symbiont-specific Sec61 complex could be identified in P. tricornutum in addition to the host copies. This failed not only for diatoms but also for all investigated databases of secondarily evolved organisms with red algal origin such as cryptophytes and apicomplexa, indicating that Sec61 could very well have no involvement in SELMA-dependent translocation (Sommer et al. 2007). However, it remains possible that the symbiont-specific ERAD-derived system does not consist of all components necessary in ERAD-L. In that case, the lack of a symbiont-specific Sec61 would not necessarily contribute to the discussions on retrotranslocation in ERAD-L. In any case, another complex than Sec61 provides the pore-forming components for SELMA, the symbiontic ERAD-like system in the second outermost plastid membrane of diatoms.
Besides Sec61, the Der1 proteins are favored as candidates for the translocation channel in ERAD-L, which is best studied in yeast and mammals (Lilley and Ploegh 2004; Ye et al. 2004; Lilley and Ploegh 2005). But because the proteins are only about 26 kDa in size and have only four transmembrane domains, oligomerization is definitely a prerequisite for channel formation. In order to investigate protein–protein interactions of both versions of sDer1 in P. tricornutum, we first coimmunoprecipitated specifically bound proteins of a sDer1-2 GFP fusion protein and could thereby show that sDer1-2 forms homo-oligomers and can also interact with sDer1-1 in vitro. These results were completely confirmed using the split-GFP system as an in vivo interaction test, which furthermore demonstrated that sDer1 interacts homo-typically as well. As in these experiments, the fluorescence signals were identical to that found in localization studies with sDer1-GFP fusion proteins, the interactions are at the second outermost membrane of the complex plastid of the diatom.
Thus, the interaction profile of sDer1 proteins in diatoms is similar to the mammalian system, in which two of the identified three mammalian Der1 family proteins (Derlin1 and Derlin2) interact with each other (Lilley and Ploegh 2005), but different to the yeast system, in which the two Der1 homologs (Der1p and Dfm1p) are localized in different complexes (Goder et al. 2008). The fact that both sDer1 proteins of P. tricornutum form homo- as well as hetero-oligomers gives further support to our localization data that both proteins are localized within the same membrane and therefore excludes a role of SELMA in protein transport at the third outermost plastid membrane.
Based on our in silico findings that heterokontophytes, cryptophytes, and apicomplexa have retained a symbiont-specific ERAD-derived machinery even though the ER within the PPC was not retained, we postulated that this machinery may meet a novel, degradation-independent role in preprotein transport across the second outermost plastid membrane of four membrane bound plastids with red algal origin (Sommer et al. 2007). However, there was hitherto no direct experimental data to support that hypothesis. The results presented in this study give a first hint that SELMA might indeed be involved in preprotein targeting as we could show that both sDer1 proteins interact with the topogenic signal of a PPC-located protein on the periplastidal side (fig. 4). Such interactions were analyzed in vivo using the split-GFP system that can also detect only transient interactions. Once the GFP fragments assemble due to the interaction of the fused proteins, the complex is trapped and GFP fluorescence remains stable (Kerppola 2006). This is exceedingly advantageous for studying the presumably transient interaction between the sDer1 complex and the PPC-specific transit peptide, which is terminated shortly thereafter upon the cleaving of the transit peptide in the PPC (Deschamps et al. 2006).
In addition, the interaction studies shown in this publication revealed that the sDer1 complex might be involved in discriminating between proteins directed to the PPC or across the two further membranes into the plastid stroma. This is because both sDer1 proteins interact with the topogenic signal of PPC-targeted proteins, while failing to do so with the topogenic signal of stroma-targeted proteins (fig. 4).
Concerning discrimination between PPC-located and stromal proteins, it was previously shown that the first amino acid of the transit peptide is critical for directing a nucleus-encoded protein either to the stroma (in case of an aromatic amino acid) or to the PPC (in case of most nonaromatic amino acids) (Kilian and Kroth 2005; Gould, Sommer, Hadfi, et al. 2006; Gruber et al. 2007). This set of results show a progression of this paradigm in that the aromatic amino acid at +1 position of the transit peptide seems to be a discriminating factor already during transport by SELMA. This was seen in the split-GFP assay, which indicated that the aromatic amino acid at +1 position suppresses binding to sDer1. Moreover, in the reverse experiment, the replacement of this aromatic amino acid in a stromal protein transit peptide was able to induce sDer1 binding (fig. 4). We can thus extend the model for protein transport across the second outermost membrane (fig. 1B): After translocation across the outermost membrane via Sec61 and excision of the signal peptide, the transit peptide of nucleus-encoded proteins is recognized by a still unknown receptor and/or directly by a SELMA component. The sDer1 complex is involved in the decision if preproteins are directed further across the third and fourth outermost membranes or not. PPC-resident proteins are possibly withheld from further transport in order to be processed, whereas stromal proteins do not bind to sDer1, thus allowing further transport across the plastid envelope in a mostly unfolded conformation (fig. 1B). Other interpretations of these data are nonetheless conceivable because different transport routes for PPC (SELMA-dependent route) and stromal proteins (SELMA-independent route) would lead to the same results.
In summary, our studies indicate that a preexisting eukaryotic machinery was relocalized from the ER of the endosymbiont to the second outermost plastid membrane and adapted to a new transport function during the transition from a free-living red alga into a secondary endosymbiont.
In this study, we present first experimental support that the symbiont-specific ERAD-like machinery (SELMA) of diatoms is localized within the second outermost plastid membrane. We could demonstrate that periplastidal preproteins interact with the SELMA complex suggesting that the ERAD-derived translocation system is indeed involved in plastid preprotein targeting across that specific membrane as we previously postulated. Furthermore, we presented first indication that the sDer1-complex might play an additional role in discriminating preproteins that are transported across two membranes into the PPC and stromal proteins which cross all four membranes. Because applicative tools for knockout or knockdown studies are presently unavailable for P. tricornutum, the final confirmation of SELMA-dependent preprotein translocation in diatoms still remains elusive. However, as the phylogenetically related group of apicomplexa also have retained a symbiont-specific ERAD-like machinery within their apicoplast and a knockdown method is available for some of these organisms, an apicomplexan system could be more suitable to investigate the effects of knockdowns of SELMA components.
We are grateful to Geoff Waldo's group (Los Alamos, USA) for providing templates for self-assembling GFP fragments GFP-11 and GFP1–10. Furthermore, we thank Marianne Johannsen and Erhard Mörschel for immunoelectron microscopic analyses. We would like to thank Andrew Bozarth, Gregor Felsner, and Maik Sommer for critical reading of the manuscript. This work was supported by the German Research Foundation (SFB 593 and Graduate School 1216).