Membrane topology and sequence requirements for oil body targeting of oleosin


For correspondence (fax +1 403 220 0704; e-mail

Present address: School of Biological Sciences, University of Manchester, 2.205 Stopford Building, Oxford Road, Manchester M13 9PT, UK.


Oleosin protein is targeted to oil bodies via the endoplasmic reticulum (ER) and consists of a lipid-submerged hydrophobic (H) domain that is flanked by cytosolic hydrophilic domains. We investigated the relationship between oleosin ER topology and its subsequent ability to target to oil bodies. Oleosin variants were created to yield differing ER membrane topologies and tagged with a reporter enzyme. Localisation was assessed by fractionation after transient expression in embryonic cells. Membrane-straddled topologies with N-terminal sequence in the ER lumen and C-terminal sequence in the cytosol were unable to target to oil bodies efficiently. Similarly, a translocated topology with only ER membrane and lumenal sequence was unable to target to oil bodies efficiently. Both topology variants accumulated proportionately higher in ER microsomal fractions, demonstrating a block in transferring from ER to oil bodies. The residual oil body accumulation for the inverted topology was shown to be because of partial adoption of native ER membrane topology, using a reporter variant, which becomes inactivated by ER-mediated glycosylation. In addition, the importance of H domain sequence for oil body targeting was assessed using variants that maintain native ER topology. The central proline knot motif (PKM) has previously been shown to be critical for oil body targeting, but here the arms of the H domain flanking this motif were shown to be interchangeable with only a moderate reduction in oil body targeting. We conclude that oil body targeting of oleosin depends on a specific ER membrane topology but does not require a specific sequence in the H domain flanking arms.


Protein topology on membranes plays a critical role in their function or activity. For instance, the prion protein (PrP) is able to cause neurodegenerative disease when present as a variant form possessing non-native topology on the endoplasmic reticulum (ER) membrane (Hegde et al., 1998). The ductin protein adopts two alternative orientations, both of which perform different functions (Hegde and Lingappa, 1997); one is a subunit of vacuolar H+-ATPase, while the other is a component of a connexon channel located in gap junctions. Oleosin is a protein that adopts a unique ER topology prior to its final destination on the surface of plant oil bodies (Abell et al., 1997, 2002), and we therefore conducted an investigation of the role its ER topology plays in its subsequent targeting to oil bodies.

Oleosin proteins associate with oil bodies in seeds and anthers (reviewed by Frandsen et al., 2001; Huang, 1996; Napier et al., 1996) and possess a tripartite structure comprising a hydrophobic (H) core of approximately 70 residues, flanked by two hydrophilic domains at the N- and C-termini (N and C domains, respectively; Figure 1). A universally conserved proline knot motif (PKM) lies at the centre of the H domain, consisting of three prolines within a 12-residue sequence. Oil bodies consist of a TAG core enclosed by a phospholipid monolayer, with a diameter of 0.5–2.0 µm. The H domain is embedded within this structure.

Figure 1.

Arabidopsis oleosin domain structure and mutations.

The native protein sequence is divided into domains of hydrophilic N-terminus (N domain), hydrophobic central domain (H domain), and hydrophilic C-terminus (C domain). The H domain is further divided into three subdomains: N-terminal H(N), PKM, and the C-terminal H(C). The residues changes made in the T1 mutation are indicated in the C domain.

Oleosin–β-glucuronidase (GUS) fusions were shown to require the H domain, but not the N or C domain, for oil body targeting in planta (van Rooijen and Moloney, 1995b). It has also been shown that the PKM is necessary for the ER to oil body transition (Abell et al., 1997). Therefore, the data suggest that the H domain is the key element within the oleosin for determination of oil body targeting, yet it is difficult to perform structural analysis on lipid-submerged domains. Original data obtained in solvent led to predictions of antiparallel β-strand structure throughout the H domain (Huang, 1992). In contrast, more recent experimental analysis performed on oleosin in a native state suggests that the H domain consists mostly of α-helix (Lacey et al., 1998).

It is well established that oleosins are synthesised at the ER membrane both in vitro (Hills et al., 1993; Loer and Herman, 1993) and in planta, prior to their localisation to oil bodies (Abell et al., 1997; Qu et al., 1986; Wanner et al., 1981). Indeed, if oleosins are expressed in tissues lacking oil bodies, they accumulate in the ER membranes (Beaudoin and Napier, 2000). It is also found that oleosins are targeted to ER membranes in a signal recognition particle (SRP)-dependent manner in both yeast (Beaudoin et al., 2000) and mammalian membranes (Abell et al., 2002), indicating that SRP is also likely to be required in planta. Consistent with the involvement of SRP, multiple independent ER-targeting sequences have been identified in the H domain (Abell et al., 2002). Furthermore, recent studies of ER localisation suggest that oleosin may be synthesised on specific TAG-synthesising subdomains of ER (Lacey and Hills, 1996). Taken together, these findings suggest a model in which oleosins bud from the ER membrane with TAG at specific sites to form nascent oil bodies. The role of the essential PKM may be to cause a strain in the structure of the H domain, which can then be relieved upon transition from the ER phospholipid bilayer to the TAG core of a nascent oil body. Oleosin can target efficiently to lipid bodies in both yeast (Beaudoin et al., 2000; Ting et al., 1997) and mammalian cells (Hope et al., 2002), suggesting that the targeting mechanisms are conserved across a wide range of organisms.

The topology of oleosin on microsomal membranes has been determined by protease protection assays, finding that the H domain is mostly protected within the membrane, while the N and C domains are exposed to the cytosolic surface (Abell et al., 1997, 2002). This topology is consistent with the ER-budding model described above, obviating the need for the hydrophilic domains to cross the hydrophobic lipid bilayer. To understand the pathway and mechanisms of oil body biogenesis, it is crucial that we understand the role of oleosin topology in the ER membrane.

The unique ER topology of oleosin is largely governed by the length of its H domain (Abell et al., 2002). The absence of targeting information in the N domain causes it to remain in the cytosol, while the H domain prevents translocation of the C domain by an unknown mechanism, which depends on its length and/or hydrophobicity. It is possible to cause translocation of the N domain by addition of an N-terminal signal sequence. However, this disruption of topology does not affect the localisation of the C domain.

In this study, the requirements of the H domain for oil body targeting were investigated by two approaches. First, the role of oleosin topology was examined by manipulating the topology of oleosin–GUS and GUS–oleosin fusions. The ability of these topology variants to target to oil bodies was then tested in planta. Second, the sequence requirements of the H domain for oil body targeting were investigated. By using H domain variants with native ER membrane topology, it was possible to examine sequence requirements of the H domain independently of topology.


A membrane-straddled topology of oleosin targets inefficiently to oil bodies

The ER topology of oleosin was initially manipulated to yield a membrane-straddling variant; C/BWTOLG comprises the native oleosin with an N-terminal signal peptide (B), C-terminal GUS (G), and transcriptionally regulated by the tandem 35S promoter (C/). The signal peptide was taken from the N-terminus of the berberine bridge enzyme (BBE; Facchini et al., 1996), and it has been shown to cause translocation of the oleosin N domain in vitro (Abell et al., 2002). Indeed, when a version without the GUS extension was analysed by protease protection, the results were consistent with a straddled topology. Therefore, BWTOLG should occupy the same topology with the C-terminal GUS located in the cytosol. All in planta expression was conducted with both flax embryos and canola suspension cells, both of which synthesise oil bodies (Weselake et al., 1998). In all cases, the trends of the data were identical between these two cell types and the data showing least statistical variation are presented. Figure 2 shows that when BWTOLG is expressed transiently in planta, it is more poorly accumulated compared to WTOLG, the same construct without a signal sequence (10%), and BWTOLG also targets to oil bodies more poorly (59% compared to 88%).

Figure 2.

Membrane-straddled oleosin–GUS targets inefficiently to oil bodies (OB).

Constructs encoding WTOLG, BWTOLG or BT1OLG were transiently expressed in planta with duplicate samples. Each sample was fractionated into OB, microsomes (MS), and supernatant, and then assayed for GUS activity. Total GUS activity is displayed as a percentage of WTOLG, and microsomal and oil body activities are calculated as a percentage of total activity for each construct. Standard error bars are indicated above each bar. Comparison of means shows that both BWTOLG and BT1OLG are expressed at significantly lower levels (Tukey–Kramer test; P < 0.05) than WTOLG, and that the proportion associated with OB is significantly different (Tukey–Kramer test; P < 0.05) for WTOLG, BWTOLG and BT1OLG. The expected ER topology of the constructs is depicted underneath.

It appears that BWTOLG accumulates in the microsomal fraction (34% compared to 7%), but fails to make the transition into oil bodies efficiently. The low levels of oil body accumulation show that some oil body targeting is still possible. However, it is possible that incomplete recognition of the N-terminal signal sequence in BWTOLG allows for some inconsistency of topology. If the N-terminal signal sequence evades detection, BWTOLG could insert with the same topology as that followed by native oleosin. This mechanism could produce a fraction of molecules with normal oil body-targeting characteristics.

A variant of BWTOLG was constructed in which the positive residues flanking the C-terminal end of the H domain were replaced with negative residues, named BT1OLG. This variant is expected to occupy the same topology as that occupied by BWTOLG, according to results from the corresponding forms lacking GUS extension in in vitro translation experiments (Abell et al., 2002). In agreement with its topology expectations, BT1OLG accumulation and targeting characteristics closely resemble those of BWTOLG (Figure 2). The negative charges cause a small decrease in oil body targeting accompanied by a corresponding increase in microsomal targeting. This may be because of maintenance of a tighter association with the ER membrane (see Abell et al., 2002).

Translocated oleosin displaying no cytoplasmic domains targets inefficiently to oil bodies

The straddled form of oleosin is a partial alteration of the oleosin topology in which a domain is still exposed to the cytoplasm. Because of the constrained nature of the H domain and the following sequence, it is clear that the C domain could not be inverted without significant disruption of the H domain and may be impossible to achieve experimentally (Abell et al., 2002). Therefore, we attempted to design an oleosin construct, which would facilitate a complete inversion of topology. To achieve this, the C domain was deleted and any necessary sequences (signals or reporters) were added to the N-terminus. This meant using the GUS reporter as an N-terminal fusion, the functionality of which had not been previously tested. The GUS coding sequence used in these experiments was that described by Farrell and Beachy (1990) in which the aspartic acid residue at position (asp)-358 was replaced by a serine residue. This mutation eliminates the possibility of N-glycosylation and subsequent loss of GUS enzymatic activity. Two constructs were assembled: GWTOL:C- comprises GUS (G) fused to the N-terminus of a C-terminal-deleted oleosin (WTOL:C-), while BGWTOL:C- has a further N-terminal addition of the BBE signal sequence. Both constructs were transcriptionally regulated by the tandem 35S promoter (C/) for expression in planta, or lacking the tandem 35S promoter for in vitro transcription and translation.

First, GWTOL:C- and BGWTOL:C- were tested by translation in vitro for microsomal targeting and topology. Figure 3(a) shows that both GUS–oleosin constructs target efficiently to microsomes, with most of the protein existing in the pelleted microsome fractions. The membrane-associated polypeptide is also resistant to carbonate washes, confirming true membrane association (data not shown). Treatment with proteinase K results in complete loss of full-length GWTOL:C-, while BGWTOL:C- is mostly resistant to protease degradation. As GWTOL:C- incorporates into membranes less efficiently than BGWTOL:C- does, a direct comparison of protease-protected membrane products can be made from a 40% loading of protease-protected BGWTOL:C-. The protected fragment can still be clearly observed, and it is also detected in a 20% loading. Therefore, BGWTOL:C- is largely protected whereas GWTOL:C- is not. Further addition of Triton X-100 to the protease degradation allows the protease to digest full-length BGWTOL:C-. We concluded that the N-terminal GUS is located, as expected, outside the microsome for GWTOL:C- and inside the microsomal lumen for BGWTOL:C-. Preprolactin (pPL) possesses a long signal sequence that is cleaved upon translocation, resulting in an observable increase in gel mobility. Full-length pPL is found predominantly in the supernatant fraction, while the cleaved form is found almost exclusively in the pellet, thereby confirming the efficiency of membrane isolation. Protease specifically degrades the non-translocated full-length form, leaving the majority of the cleaved prolactin intact. The more efficient protection of prolactin compared to membrane-associated BGWTOL:C- is likely to reflect a portion of BGWTOL:C- that is membrane integrated but not translocated, i.e. the N-terminal signal sequence is bypassed and BGWTOL:C- adopts the topology of GWTOL:C-. This is unsurprising for an in vitro assay given that targeting by signal sequences rarely approaches 100%. The partial cleavage of pPL in Figure 3(a) is a typical example.

Figure 3.

Inverted oleosin–GUS targets inefficiently to oil bodies (OB).

(a) In vitro translation of GWTOL:C-, BGWTOL:C- and pPL in the presence of canine pancreatic microsomes (MS). The products were fractionated into supernatant (S) and pellet (P), and samples were treated with proteinase K (PK) and/or Triton X-100 (TX) where indicated. Polypeptide fragments were analysed by SDS–PAGE using 12% polyacrylamide. Extra lanes were included for BGWTOL:C- in which 40 and 20% volumes of PK-treated pellet fractions were loaded. The full-length pPL and signal-sequence-cleaved Prolactin (PL) forms of pPL are indicated. The deduced ER topology of the constructs is depicted underneath.

(b) Constructs encoding WTOLG, GWTOL:C- or BGWTOL:C- were transiently expressed in planta with four independent samples. Each sample was fractionated into OB, MS and supernatant, and then assayed for GUS activity. Total GUS activity is displayed as a percentage of WTOLG, and microsomal and oil body activities are calculated as a percentage of total activity for each construct. Standard error bars are indicated above each bar. Comparison of means shows that both WTOLG and GWTOL:C- are expressed at significantly higher levels, accumulate significantly more in OB and significantly less in MS (Tukey–Kramer test; P < 0.05) than BGWTOL:C-. The expected ER topology of the constructs is depicted underneath.

(c) Constructs encoding WTOLG and BGWTOL were transiently expressed in planta with triplicate samples. Each sample was fractionated into OB, MS and supernatant, and then assayed for GUS activity. Total GUS activity is displayed as a percentage of WTOLG, and microsomal and oil body activities are calculated as a percentage of total activity for each construct. Standard error bars are indicated above each bar. Comparison of means shows that WTOLG and BGWTOL are expressed, and targeted to both OB and MS at levels significantly different from one another (Tukey–Kramer test; P < 0.05). The expected ER topology of the constructs is depicted underneath.

Figure 3(b) shows that GWTOL:C- (GUS–oleosin) is capable of targeting at a comparable efficiency to WTOLG (oleosin–GUS). However, the GUS–oleosin with lumenal orientation, BGWTOL:C-, accumulates poorly compared to GWTOL:C- (3 and 98%, respectively) when expressed in planta. Although overall accumulation is poor, BGWTOL:C- does target to both microsomal (45%) and oil body (54%) fractions. Therefore, BGWTOL:C- accumulates in microsomes, but is impaired in its ability to make the transition to oil bodies. As in the experiments with BWTOLG, it is not possible to determine the process by which residual BGWTOL:C- becomes associated with oil bodies. Any product in which the N-terminal signal sequence is not recognised will assume a topology identical to GWTOL:C-, and thereby targets to oil bodies by a conventional route.

To allow a closer comparison between the straddled and ‘inverted’ forms of oleosin reporters, we constructed a form of C/BGWTOL:C- with a full-length oleosin, named C/BGWTOL. This will essentially occupy the same topology as that occupied by inverted BGWTOL:C-, but with the oleosin C domain extending into the cytosol. When expressed transiently, Figure 3(c) shows that BGWTOL accumulates very poorly compared to WTOLG (3 and 100%, respectively). It also targets to oil bodies more poorly (41% compared to 78%), and targets at a similar level in microsomal fractions (13 and 10%). These data show that this alternative straddled form behaves in a manner similar to both the BWTOLG straddled form and the translocated form of oleosin lacking the C-terminal (BGWTOL:C-). Clearly, the overall accumulation and oil body targeting is similar for this membrane-straddled construct compared to the other membrane-straddled forms and to those lacking a cytoplasmic C-terminal. Surprisingly, the microsomal targeting is not significantly enhanced as in the other topology variants. It is possible that it is more susceptible to degradation.

Oleosin targets to oil bodies by a single topological pathway

The ability of straddled and inverted forms of oleosin to target to oil bodies suggests that there may be more than one topological pathway to the oil body. However, it is also possible that a small proportion of the constructs possessing an N-terminal signal sequence may evade translocation. This possibility was examined by creating a form of inverted BGWTOL:C-, in which the GUS reporter has an asparagine residue at position 358 within the GUS sequence (replacing the serine residue in other constructs), named BG(N)WTOL:C-. It has been shown that N-linked glycosylation of GUS at this asparagine causes a more than 100-fold reduction in GUS activity (Alcocher et al., 2003; Iturriaga et al., 1989). Therefore, it was possible to determine whether the GUS portion of the fusion protein became exposed to the ER lumen during its movement to oil bodies.

Figure 4 shows that upon expression in planta, the GUS activity accumulated on oil bodies is similar for both C/BGWTOL:C- and C/BG(N)WTOL:C-. However, the GUS activity accumulated in the microsomal fraction is significantly lower (approximately twofold reduction) for C/BG(N)WTOL:C-. This reduction in microsomal activity demonstrates that the majority of GUS–oleosin is oriented such that GUS is lumenally displayed, and then becomes inactivated by glycosylation (Alcocher et al., 2003; Iturriaga et al., 1989). If this form of GUS–oleosin were able to target to oil bodies, we would expect to observe at least some reduction in the GUS activity associated with oil bodies. The fact that the oil body GUS activity remains at similar levels indicates that the GUS–oleosin reaching the oil body is likely to be targeted because of a failure of N-terminal signal peptide recognition, and therefore indicates that GUS–oleosin is not able to traffic to oil bodies via a lumenally exposed form.

Figure 4.

Inverted oleosin targets to oil bodies (OB) by a conventional pathway.

Constructs encoding BGWTOL:C- (labelled SER to indicate serine at position 358 of GUS) and BG(N)WTOL:C- (labelled ASN to indicate glycosylatable asparagine at position 358 of GUS) were transiently expressed in planta with triplicate samples. Each sample was fractionated into OB, microsomes (MS) and supernatant, and then assayed for GUS activity. Total GUS activity is normalised to the mean oil body accumulation for BGWTOL:C-, whereas the percentage targeting figures are means of GUS distribution for that construct. Standard error bars are indicated above each bar. Comparison of means shows that BGWTOL:C- and BG(N)WTOL:C- accumulate to significantly different levels in microsomes (Tukey–Kramer test; P < 0.05). The expected ER topology of the construct is depicted underneath, depending on whether the signal sequence causes translocation. GUS can be glycosylated (forked symbol) in the ER lumen.

The oleosin H domain can be reconstructed to retain oil body targeting

The oleosin H domain has been shown to be necessary for both ER targeting, ER topology (Abell et al., 2002), and for oil body targeting (van Rooijen and Moloney, 1995b). Furthermore, it has been shown that the PKM is essential for oil body targeting without affecting ER accumulation (Abell et al., 1997). Therefore, we were interested in determining which features of the H domain, outside the conserved PKM, are important for oil body targeting.

Two oleosin variants with rearranged H domains provided the most useful material to initiate dissection of the H domain. In T2OLEO, the C-terminal arm of the H domain (H(C)) was replaced with a second copy of the N-terminal arm (H(N)), thereby leaving the overall length of the H domain and the PKM unaffected. Similarly, in T3OLEO, the H(N) arm was replaced with a second copy of H(C). Both of these variants were combined with the T1 mutation described for BWTOLG, placing negatively charged residues immediately C-terminal to the H domain. Importantly, it has been shown that both T2OLEO and T3OLEO are able to target efficiently to microsomes, and adopt the same topology as that displayed by WTOLEO (Abell et al., 2002). This characteristic allowed the investigation of sequence requirements for oil body targeting, independently of ER targeting and topology considerations, assuming that these constructs behave in planta as they do in vitro. Both T2OLEO and T3OLEO were tagged with a C-terminal GUS, and their expression were driven by a 35S promoter (C/) to yield C/T2OLG and C/T3OLG.

Figure 5 shows that when expressed in planta, T2OLG is able to accumulate in ER in similar proportions, although at significantly reduced total levels compared to WTOLG (35% compared to 100%), and with lower efficiency of oil body targeting (50% compared to 66%). Therefore, the H(C) element is clearly not essential for oil body targeting, although the native combination of H(N) and H(C) is more efficient for oil body targeting than the repeated H(N) segment. T3OLG is able to accumulate in both ER and oil bodies at similar proportions compared with WTOLG, but at reduced levels compared to WTOLG (20% compared to 100%). This suggests that these modifications result in a much more labile oleosin–GUS polypeptide. However, it is not clear whether this instability is intrinsic to the modification or the result of inefficient mobilisation onto oil bodies as previously shown by van Rooijen and Moloney (1995b). Small decreases in the proportion targeting to oil bodies are accompanied by small increases in the proportion targeting to microsomes, indicating that both H domain rearrangements cause a partial inhibition of ER to oil body movement. This implies that the H(N) element is not an absolute requirement for oil body targeting, even though the native combination of H(N) and H(C) is more efficient for oil body targeting than the repeated H(C) segment. Overall, it appears that H(N) and H(C) bear sufficient similarity that they are able to substitute for each other in the oleosin H domain. The targeting inefficiencies of the T2OLG and T3OLG may be as a result of specialisation of targeting function between H(N) and H(C). Both T2OLG and T3OLG accumulate a higher proportion of GUS activity in the microsomes but a lower proportion of activity in the oil bodies compared to WTOLG. These results suggest that both H domain variants are less able than the native oleosin to make the transition from ER to oil bodies, although there is also an impairment of ER accumulation in absolute terms.

Figure 5.

Reconstructed oleosin H domains retain oil body targeting.

Constructs encoding WTOLG, T2OLG or T3OLG were transiently expressed in planta with triplicate samples, or two sets of triplicate samples for T3OLG. Each sample was fractionated into oil bodies (OB), microsomes (MS) and supernatant, and then assayed for GUS activity. Total GUS activity is displayed as a percentage of WTOLG, whereas the percentage targeting figures are means of GUS distribution for that construct between MS and OB. Standard error bars are indicated above each bar. Comparison of means shows that both T2OLG and T3OLG are expressed at significantly lower levels (Tukey–Kramer test; P < 0.05) than WTOLG, and that the proportion associated with OB is significantly lower (Tukey–Kramer test; P < 0.05) for T2OLG than WTOLG or T3OLG. The expected ER topology of the constructs is depicted underneath.

It remains a possibility that the reduced level of accumulation and/or targeting resulted from topology alterations not detected in microsomes in vitro. Therefore, a variant of C/T2OLG was constructed with the addition of the BBE signal sequence, termed C/BT2OLG. This would only be expected to affect the targeting of T2OLG if T2OLG adopts a native topology the same as that adopted by WTOLG. Figure 6 shows that BT2OLG accumulates at lower levels than does T2OLG, comparable to the difference between BWTOLG and WTOLG (see Figure 3c). Therefore, the in planta topology of T2OLG is the same as that of WTOLG.

Figure 6.

T2OLG assumes the same topology in planta as in vitro.

Constructs encoding T2OLG or BT2OLG were transiently expressed in planta with duplicate samples. Each sample was fractionated into oil bodies (OB), microsomes (MS) and supernatant, and then assayed for GUS activity. Total GUS activity is displayed as a percentage of GUS, and microsomal and oil body activities are calculated as a percentage of total activity for each construct. Comparison of means shows that both T2OLG and BT2OLG are expressed at significantly different levels (Tukey–Kramer test; P < 0.05). This experiment is representative of at least two similar experiments for each construct. The expected ER topology of the constructs is depicted underneath.


Trafficking of oleosin to oil bodies is dependent upon its ER topology

We have assessed the role that ER topology of oleosin plays in its trafficking to oil bodies. To achieve this we analysed the in planta behaviour of ER topology variants. Although the H domain will not allow translocation of the C domain to the ER lumen (Abell et al., 2002), we were able to create topology variants possessing a lumenal N domain by the addition of an N-terminal signal peptide. When such a membrane-straddled oleosin–GUS was expressed in planta, it accumulated very poorly. The reason for poor accumulation was an inhibition of the ER to oil body transition as accumulation on the ER in vitro was not inhibited. Similar results were obtained for a membrane-straddled GUS–oleosin, and more significantly for a translocated GUS–oleosin that did not display protein on the cytosolic side. This GUS–oleosin displaying GUS lumenally is the most extreme variation of oleosin topology that can be achieved. These data provide compelling evidence that oleosin ER topology is critical for successful oil body targeting, i.e. only a cytosolic disposition will permit efficient oil body targeting.

Although the reduction in oil body targeting was accompanied by an increase in the proportion accumulated in the ER fraction, there was no large-scale accumulation in the ER fraction. This is not surprising, because a large foreign protein in the ER is likely to be identified for degradation by the ubiquitin-proteasome pathway (Baumeister et al., 1998; Mayer et al., 1998; Yeung et al., 1996). Indeed, the proline knot variant of oleosin–GUS has previously been shown to accumulate to similar levels in ER in planta compared to the native form, despite its inability to exit the ER (Abell et al., 1997). Moreover, when a proline knot mutant of oleosin was expressed in mammalian cells, its poor accumulation could be over come by treatment with a proteasome inhibitor (Hope et al., 2002).

The strict dependence on the ER topology for oil body targeting supports a model in which oleosin incorporates into developing oil bodies at the ER membrane (Figure 7). In such a scheme, oleosins would need to make a transition from the phospholipid bilayer to a region containing TAG. The subsequent budding of the oil body into the cytosol would be more favourable if the embedded oleosins already presented their hydrophilic domains to the cytosol. The native topology is consistent with such a requirement whereas a membrane-straddled or inverted form might present difficulties for extracting the hydrophilic domains from the ER lumen.

Figure 7.

Model of oleosin transition from ER to oil bodies (OB).

Native oleosins are proposed to move laterally within the ER membrane until they flow into a region of TAG enclosed between the ER membrane leaflets. The drive for this movement may be provided by the thermodynamic benefit of moving the H domain from a phospholipid bilayer into the TAG droplet, where interactions within the H domain might occur. This is denoted as a transition from a constrained to a relaxed state. The transition from membrane to oil droplet is possible for oleosin variants with rearranged H domains (T2 and T3 mutations), because of the remaining interaction between the two arms of the H domain. The nascent oil body blebs into the cytosol with its associated oleosin proteins. Straddled or inverted oleosins are able to accumulate in ER membranes, but are unable to move into OB, presumably because of the necessity to position both hydrophilic domains on the cytosolic surface.

The strict dependence on topology is consistent with the differences between oleosins and mammalian apolipoprotein B (apoB; compared by Murphy and Vance, 1999). Whereas oleosin is present on the surface of cytosolic oil bodies, apoB forms a coat around lipoproteins, which initially form in the ER lumen (Innerarity et al., 1996; Yao and Mcleod, 1994). The complete translocation of apoB and absence of transmembrane spans allow it to associate with the lumenal face of the ER membrane, thereby facilitating its combination with phospholipid and TAG to form lumenal lipoproteins. Therefore, it appears that membrane topology is a common critical factor in the association of proteins with lipid bodies.

The H domain hydrophobic arms do not have strict sequence constraints

The oleosin H domain has previously been shown to possess essential targeting and insertion properties (Abell et al., 2002; van Rooijen and Moloney, 1995b). Furthermore, the PKM was shown to be essential for oil body targeting, even though it is targeted to the ER membrane in planta, and with the correct topology (Abell et al., 1997). Here, we investigated the role played by the hydrophobic arms, H(N) and H(C), that flank the PKM. Although the rearranged H domains with duplicated H(N) (T2OLG) and duplicated H(C) (T3OLG) do not target to oil bodies as efficiently as WTOLG does, it is clear that the H(N) and H(C) sequences are not strict requirements for oil body targeting. The rearrangements in T2OLG and T3OLG are substantial, yet do not abolish oil body targeting. The implication is that the H domain sequences outside the PKM are not highly specific but do play some role in facilitating the transition to oil bodies. The increased proportion of GUS activity in microsomes at the expense of oil bodies in the case of T2OLG and T3OLG, compared to WTOLG, suggests that both H(N) and H(C) play a role in the transition from ER to oil bodies.

The native combination of H(N) and H(C) may promote a structural interaction; for example, the α-helical structure demonstrated by Lacey et al. (1999) may coil around itself. This interhelical interaction could be stabilised by intercalation of the polar residues spaced regularly throughout the hydrophobic arms, thereby shielding them from the hydrophobic environment in an oil body. In support of this model, the structure of bacteriorhodopsin shows that the more hydrophilic faces tend to pack against one another (von Heijne, 1994). It is also possible that the native combination of H(N) and H(C) is preferred because they specialise in facilitating different steps in the transition. For example, H(N) might act as a more efficient signal sequence for ER targeting whereas the sequence of H(C) might be optimised for the transition from phospholipid to TAG. A final explanation for the preferred pairing of H(N) and H(C) is that this combination may represent the optimal hydrophobicity of the H domain. H(C) contains a higher proportion of hydrophobic residues than H(N) does.

Model of oleosin trafficking to oil bodies

We have demonstrated that the oil body targeting of oleosin is dependent on the adoption of a native topology in the ER membrane, but that considerable variation in the sequence of the H domain hydrophobic arms is permissible. However, the optimal targeting of the native H domain suggests more general structural functions of H(N) and H(C). The importance of both topology and hydrophobic domain structure is consistent with the idea that the transition of an oleosin polypeptide from the ER to the oil body is driven thermodynamically, i.e. by folding constraints, which are relieved, and intramolecular interactions, which are permitted when the oleosin reaches a nascent oil body and can undergo relaxation. These findings are also relevant to the use of oleosins as carriers for recombinant proteins, which has been shown to be advantageous for expression of a variety of therapeutic proteins in seeds (van Rooijen and Moloney, 1995a). The rules described herein may be used to assist in the design of such oleosin–recombinant protein translational fusions.

Experimental procedures

Plasmid construction

Plasmid construction was carried out using standard procedures (Sambrook et al., 1989). Site directed mutagenesis was performed by PCR, using Pwo polymerase (Roche, Mannheim, Germany) according to manufacturer's instructions. Oligonucleotides were synthesised using a Cyclone Plus DNA Synthesiser (Millipore, Billerica, MA, USA) or an Oligo 1000 m DNA synthesiser (Beckman, Mississauga, ON, Canada). All oleosin variants generated by PCR were confirmed by automated dye-terminator sequencing (DNA Sequencing Laboratory, University of Calgary, Alberta, Canada). Constructs in which the GUS gene was amplified by PCR were verified by testing at least two independent clones for in planta GUS activity; similar levels of GUS activity between independent clones was taken to imply the successful amplification of the GUS gene. Oleosin constructs were based on the cDNA sequence from the plasmid YAP230T7 (obtained from the Arabidopsis Biological Resource Centre). N-terminal signal peptides were obtained from the BBE, present in bbe1 (Facchini et al., 1996). pPL cDNA was present in pSP64 (Siegel and Walter, 1988).

C/WTOLG was constructed from a version of pGNOS (van Rooijen and Moloney, 1995b) with a tandem 35S promoter (Ca2) upstream of the GUS gene, named pC/ntGUS, and then inserting a copy of an Arabidopsis oleosin cDNA between the promoter and GUS. pC/ntGUS was cut with NcoI and a PCR fragment containing oleosin (amplified with Bamoleo (5′-CGCGGATCCATGGCGGATACAGCTAGA-3′; BamHI and NcoI sites underlined) and GVR01 (5′-AATCCCATGGATCCTCGTGGAACGAGAGTAGTGTGCTGGCCACCACGAGTACGGTCACGGTC-3′; NcoI and BamHI sites underlined) from pWTOLEO (Abell et al., 1997) and cut with NcoI) was inserted. pC/BWTOLG was constructed by the ligation of Ca2 (pCa2 cut with ScaI and XbaI) and B82WTOLEO (pB82WTOLEO (Abell et al. 2002) cut with SpeI and ScaI) into pC/WTOLG (cut with ScaI). pC/BT1OLG and pC/BT2OLG were constructed as pC/BWTOLG but using pB82T1OLEO and pB82T2OLEO, respectively, in place of pB82WTOLEO. pC/T2OLG and pC/T3OLG were constructed using the same strategy, but by cutting the oleosin segment with BamHI and ScaI from pB82T2OLEO and pB82T3OLEO, respectively. pC/GWTOL:C- was constructed by ligating GUS (PCR fragment using T7 and BAMXSEQ2 (5′-CGCGGATCCTCTTCCTTCGATTTGTTTGCCTCCCTGC-3′; BamHI site underlined) on pGNOS template, cut with BamHI), WTOLEO:C- (pWTOLEO:C- (Abell et al., 1997) cut with BamHI and PstI), and NOS terminator (pGNOS cut with PstI and KpnI), ligated into pCa2 cut with BamHI and KpnI. pC/BGWTOL:C- was constructed from pC/GWTOL:C- by deleting the NcoI–NcoI GUS fragment and replacing it with B25 signal sequence (PCR fragment using BSP1 (5′-CGTTCCATGGCAATGTGCAGAAGCTTAA-3′; NcoI site underlined) and SPD-25 (5′-ATTATCACTAGTATCACCACCTCGTACGCATGT-3′; SpeI site underlined) on BBE template, cut with NcoI and SpeI) and GUS (PCR fragment using BGU1 (5′-GGTGATACTAGTGATAATATGGTCCGTCCTGTAGAA-3′; SpeI site underlined) and BGU2 (5′-GAAACCATGGATCCTCGTGGAACGAGTTGTTTGCCTCCCTGCTG-3′; NcoI and BamHI sites underlined) on pGNOS template, cut with SpeI and NcoI). In pC/BG(N)WTOL:C-, the GUS in pC/GBWTOL:C- was modified by exchange of a fragment between internal SpeI and BstBI sites, with the corresponding fragment from pPVOLG (Abell et al., 1997). pC/BGWTOL was constructed by deletion of the AgeI–KpnI segment and replacing it with a 3′ segment of oleosin (pWTOLEO cut with AgeI and PstI) and NOS (pGNOS cut with PstI and KpnI).

In vitro translation

Constructs were linearised after the termination codon, then transcribed using T7 RNA polymerase with CAP analogue, according to manufacturer's instructions (New England Biolabs, Beverly, MA, USA). Transcript was then used to program a rabbit reticulocyte translation lysate according to manufacturer's instructions (Promega, Madison, WI, USA); 10 µl reactions contained 0.1 µg transcript, and products were radiolabelled with 35S-methionine (New England Nuclear, Boston, MA, USA). Canine pancreatic microsomes were supplemented at 0.5 Eq per 10 µl. After incubation at 30°C for 60 min, the reactions were layered onto 100-µl HSC buffer (250 mm sucrose, 500 mm KOAc, 5 mm Mg(OAc)2, 50 mm Hepes–KOH, pH 7.9) and centrifuged at 100 000 g for 10 min in a TLA100 rotor (Beckman, Mississauga, ON, Canada). The supernatant was collected and the pellet was resuspended in LSC buffer (250 mm sucrose, 100 mm KOAc, 5 mm (MgOAc)2, 50 mm Hepes–KOH, pH 7.9) for direct analysis or further treatment as follows. One portion was treated with proteinase K at 100 µg ml−1 for 30 min at 0°C. The protease was inactivated by addition of 5 mm phenylmethyl sulfonyl fluoride. A third sample was treated with 1% Triton X-100 in combination with protease. Samples were heated at 70°C for 10 min with an equal volume of loading buffer, and products were separated by SDS–PAGE (12% polyacrylamide). Gels were fixed and dried before being monitored using phosphorimager plates and a Fujifilm BAS-1800 phosphorimager. Quantification was performed using aida software.

Canola cell suspension cultures

The cell suspension culture was derived from winter oilseed rape microspores (Brassica napus L. cv. Jet Neuf), provided by Dr Randall. J. Weselake (University of Lethbridge). They were maintained according to Simmonds et al. (1991), which includes subculturing every 2 weeks. Cells for transient expression were supplemented with sucrose to a final concentration of 14% at the point of subculturing, to increase the levels of TAG and oleosin synthesis (Weselake et al., 1998), and then harvested 7 days after subculturing.

Biolistic transient expression

The DNA was coated onto gold particles, according to manufacturer's instructions (Bio-Rad, Hercules, CA, USA) with minor modifications. Gold particles, 1.6 µm in diameter (approximately 3 mg), in 50 µl water were mixed with 5–10 µg of plasmid DNA, 50 µl of 2.5 m CaCl2 and 20 µl of 0.1 m spermidine (free base, tissue culture grade, Sigma Chemical Co., St Louis, MO, USA), incubated on ice for 5 min, then vortexed for 4 min. This preparation was then pelleted at 10 000 r.p.m. for 15 sec in a microfuge, and the particles were washed with absolute ethanol (HPLC grade, Fisher, Hampton, NH, USA) before resuspension in 60 µl of absolute ethanol. Portions of gold particles were analysed by agarose gel electrophoresis to verify the success and consistency of DNA coating.

For embryos, 10 flax mid-cotyledonary embryos were laid side by side in random orientation on this disc, filling a 2-cm-diameter circle at the centre of the plate. For suspension cells, 1 ml of cells at 50% v/v were spread onto a sterile filter disc (Whatman GF/A glass microfibre, 24 mm) on a vacuum filter, and allowed to dry almost to completion. The filter disc with cells was then placed on a second sterile filter paper disc (Whatman #1, 42.5 mm), which was moistened with cell suspension media supplemented with 10 µm racemic abscisic acid (ABA, Sigma Chemical Co.), 10 µg ml−1 nystatin, and 5 µg ml−1 chloramphenicol, and placed in a 60-mm Petri plate.

Biolistics was carried out using the PDS-1000/Helium particle gun (Bio-Rad), using 900 psi rupture discs, a target distance of 11 cm, a chamber vacuum of 685 mm Hg, and 15–18 µm of gold/DNA suspension per macrocarrier.

After delivery, the embryos were transferred to a standard mixture of NLN medium (Lichter, 1982), containing 10 µm ABA. Suspension cells were moistened further by the addition of 700 µl suspension cell media, supplemented with ABA, nystatin, and chloramphenicol as described above. Embryos or suspension cells were left in the dark at room temperature for 2 days before processing.

Oil body/microsomal fractionation for transient assays

Embryos or suspension cells were homogenised by mortar and pestle on ice, with the addition of 400 µl microsome buffer (MB; 50 mm Hepes–KOH, pH 7.5, 0.25 m sucrose, 10 mm KCl, 5 mm EGTA, 62.5 mm KOAc, 5 mm MgCl2, 5 mm DTT) with the addition of 2 mm PMSF and 5 µl yeast protease inhibitor cocktail (Sigma). The homogenate was transferred to a 1.5-ml Eppendorf tube containing approximately 2 mg polyvinylpolypyrrolidone (PVPP) and thoroughly mixed, then spun at 5000 g for 5 min at 4°C to pellet the cell debris. The resulting supernatant was transferred to a 0.5-ml Eppendorf tube and spun at 10 000 g for 10 min at 4°C. The cleared supernatant was carefully withdrawn from underneath the oil pad, using a needle syringe, and transferred to a 2-ml ultracentrifuge tube, then spun at 174 000 g maximum for 90 min at 4°C. The final supernatant was removed and the microsomal pellet was resuspended in GUS extraction buffer (50 mm Na2HPO4, pH 7.0, 10 mmβ-mercaptoethanol, 10 mm Na2EDTA, 0.1% Triton X-100, 0.1% sarkosyl; Jefferson, 1987). The oil pad was washed by two successive washes using 400 µl GUS extraction buffer.

GUS assays

Fractions were assayed for GUS activity by the methods described by Jefferson (1987). Samples were incubated with 4-methyl umbelliferone glucuronide (MUG) and sampled at three time points. Sodium carbonate (0.2 m) was used to stop the reaction. The hydrolysis product 4-methyl umbelliferone (MU) was measured by fluorometry on a Hitachi F-2000 fluorescence spectrophotometer (excitation at 365 nm; emission at 455 nm).


We would like to thank Dr Randall J. Weselake for assistance with cell suspension cultures and Mr David Bird for donation of the bbe1 DNA template. We gratefully acknowledge the financial support of the Natural Sciences and Engineering Research Council (Research Partnerships Program) for the award of an Industrial Research Chair to M.M.M. and for Research Grant #3490.