The endoplasmic reticulum (ER) is a network of membrane sheets and tubules connected via three-way junctions. A family of proteins, the reticulons, are responsible for shaping the tubular ER. Reticulons interact with other tubule-forming proteins (Dp1 and Yop1p) and the GTPase atlastin. The Arabidopsis homologue of Dp1/Yop1p is HVA22.
We show here that a seed-specific isoform of HVA22 labels the ER in tobacco (Nicotiana tabacum) cells but its overexpression does not alter ER morphology. The closest plant homologue of atlastin is RHD3. We show that RHD3-like 2 (RL2), the seed-specific isoform of RHD3, locates to the ER without affecting its shape or Golgi mobility. Expression of RL2-bearing mutations within its GTPase domain induces the formation of large ER strands, suggesting that a functional GTPase domain is important for the formation of three-way junctions.
Coexpression of the reticulon RTNLB13 with RL2 resulted in a dramatic alteration of the ER network. This alteration did not depend on an active GTPase domain but required a functional reticulon, as no effect on ER morphology was seen when RL2 was coexpressed with a nonfunctional RTNLB13. RL2 and its GTPase mutants coimmunoprecipitate with RTNLB13.
These results indicate that RL2 and RTNLB13 act together in modulating ER morphology.
The plant endoplasmic reticulum (ER) is the gateway to the secretory pathway and oversees a large number of biological processes, among which are the synthesis, quality control and export of secretory proteins. The ER, a very dynamic organelle, is continuous with the nuclear envelope, comprises a network of tubules and cisternae, much of which is cortical, and in plants are associated mainly with the actin cytoskeleton (Sparkes et al., 2009). A fundamental yet unsolved question is how the shape of the ER informs its biosynthetic capacity. Very recent research in both animals and plants has begun to elucidate which factors are involved in conferring and maintaining ER shape (reviewed in Sparkes et al., 2009, 2011). We have recently shown that a large family of integral membrane proteins, the reticulons (RTNs), are sufficient to induce curvature of the plant ER membrane and generate tubules (Tolley et al., 2008, 2010; Sparkes et al., 2010), in a manner similar to their mammalian and yeast counterparts (Voeltz et al., 2006; Hu et al., 2008). In plants, RTNs are represented by large gene families, with the Arabidopsis family having 21 members (Oertle et al., 2003; Nziengui et al., 2007). In yeast and mammals, DP1/Yop1, proteins with similar transmembrane topology to RTN and comparable function, work in concert with reticulons both in vivo (Voeltz et al., 2006) and in vitro (Hu et al., 2008). The closest DP1/Yop1 homologue in plants is HVA22, one isoform of which has been recently shown to localize to the ER (Chen et al., 2011), but it is not yet clear whether it can shape the ER membrane in the same way as its animal and yeast relatives.
Membrane fusion events within the secretory pathway are generally mediated by SNARE proteins. While plant ER SNAREs have been characterized, they appear to regulate vesicle fusion within the bidirectional trafficking between the ER and Golgi, and have yet to be shown to be involved in homotypic ER membrane fusion (Bubeck et al., 2008; Lerich et al., 2012). Recently, a family of large, dynamin-like GTPases, the atlastins, was proposed to mediate ER membrane fusion in a GTP-dependent manner (Hu et al., 2009; Orso et al., 2009). Atlastins are integral membrane proteins with a large N-terminal cytosolic region containing the GTPase domain, two transmembrane domains and a cytosolic C-terminus of variable length. Very recent structural analysis of the atlastin cytosolic domain (Bian et al., 2011) and mutagenesis studies (Moss et al., 2011; Pendin et al., 2011) point towards a mechanistic model where dimerization of the atlastin GTPase domains and a subsequent conformational change, possibly driven by GTP hydrolysis, bring two membranes into close contact and facilitate their fusion (Moss et al., 2011; Pendin et al., 2011). As a consequence, atlastin mutants with impaired nucleotide-binding activity are incapable of promoting liposome fusion in vitro (Bian et al., 2011) and several mutations in vivo present an unbranched ER network, where tubules form large cable-like bundles instead of anastomosing into the normally observed three-way junctions (Hu et al., 2008; Pendin et al., 2011).
Atlastins have been shown to interact with RTNs and DP1/YOP1 in animal and yeast cells (Hu et al., 2009; Orso et al., 2009). It is therefore possible that the interplay between reticulons and atlastins, that is, between membrane-shaping and membrane-fusing proteins, may underpin the shared, distinctive cortical ER network architecture among eukaryotic cells. The closest plant homologue of atlastin is RHD3 (Wang et al., 1997; Zheng et al., 2004). Three isoforms of this protein are present in the Arabidopsis genome (Hu et al., 2003) and recently the most abundant isoform of RHD3 (At3g13870), which is mostly expressed in vegetative tissues, was shown to be localized to the tubular ER and to complement the rhd3 mutation. A GTPase domain mutant was shown to phenocopy the rhd3 mutation (Chen et al., 2011) and an intact C-terminal domain was also shown to be necessary for RHD3 function (Stefano et al., 2012). While RHD3 has a role in the formation of ER junctions, an rhd3 null mutant is still capable of forming tubular fusions in the peripheral ER, indicating that this protein is important but not essential for this process (Stefano et al., 2012). The question remains as to whether RHD3 also interacts with reticulons.
Our previous work has focused on a reticulon isoform, RTNLB13, which is predicted to be expressed solely in embryos of maturing seeds (Sparkes et al., 2011). In this report, we characterize the DP1/Yop1 homologue HVA22b and one isoform of the atlastin homologue RHD3 (At5g45160), recently renamed RHD-like2 (RL2; Chen et al., 2011), which are both predicted to be present at the same developmental stages as RTNLB13 (Supporting Information, Fig. S1). RL2 was recently shown to be functionally interchangeable with the more widely expressed RHD3 (Chen et al., 2011).
We asked whether HVA22b and RL2 have ER-shaping properties and whether they have the capacity to interact with RTNLB13. Here we show, using transient expression in tobacco epidermal cells, that while HVA22b and wildtype RL2 have no apparent effect on the ER membrane and network, GTPase domain mutations in RL2 can affect cortical ER structure. RTNLB13 and RL2 proteins physically interact and, when coexpressed, can induce major alterations to the architecture of the cortical ER network. We also show that this additive effect requires a functional reticulon protein.
Materials and Methods
The coding sequences of RHD3-like2 (At5g45160) and HVA22b (At5g62490) were amplified from genomic DNA using the following primers: RL2 forward 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCCCGCCAATGGGTGAAAATGATGATGGATGCTCAACTCAAC-3′ and RL2 reverse 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCCATCTGACTAATCTCACTCTCTTGCACGTTG-3′ and HVA22b forward 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCCCGCCAATGAGTTCCGGAATCGGAAG-3′ and HVA22b reverse 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCCTAGTAGATATAGGCGTCATC-3′.
All cloned products were cloned into the Gateway entry vector pDONR207 and then into their respective destination vectors: pB7RWG2 for C-terminal mRFP and pB7WGF2 for N-terminal eGFP (Karimi et al., 2005). Binary constructs were transformed into the Agrobacterium tumefaciens strain GV1301.
Green or red fluorescent protein (GFP or RFP) fused to the ER retrieval signal HDEL (GFP-HDEL or RFP HDEL, respectively) were used as an ER marker for localization experiments. The signal anchor sequence from a rat sialyl transferase fused to GFP (ST-GFP; Boevink et al., 1998) was used as a Golgi marker.
For the construction of the Derlin 1 reporter, the cDNA encoding AtDer1 (At4g29330) was obtained from the Nottingham Arabidopsis Stock Centre (NASC). The cDNA was amplified by PCR to add ClaI and KpnI sites at its 5′ and 3′ ends, respectively, and cloned downstream of the 35S promoter in pDE109 (Denecke et al., 1990). The cDNA for yellow fluorescent protein (YFP) was cloned in frame with the 3′ end of the derlin coding sequence using KpnI and XbaI, upstream of the nos terminator cassette of pDE. The whole expression cassette was then extracted from pDE with EcoRI and HindIII and cloned into the same sites of pGREEN0029 (Hellens et al., 2000). The binary vector was introduced into A. tumefaciens C58::pSOUP (Hellens et al., 2000).
The construction of RTNLB13-YFP and RTNLB13-ΔTM4 has already been described (Tolley et al., 2008, 2010). RLD2-K53A, RL2-R171Q and RL2-S54N were generated by QuickChange mutagenesis (Kunkel, 1985) using the following primers: forward RL2-K53A 5′-CCTCAATCTTCTGGAGCGTCTACTCTTTTGAAC-3′ and reverse 5′-GTTCAAAAGAGTAGACGCTCCAGAAGATTGAGG-3′; RL2-R171Q forward 5′-CTTTTGTTTGTGATCCAAGATAAGACCAAAACT-3′; and reverse 5′-AGTTTTGGTCTTATCTTGGATCACAAACAAAAG-3′; RL2-S54N forward 5′ CAATCTTCTGGAAAGAATACTCTTTTGAACCATTTG-3′; and 5′CAAATGGTTCAAAAGAGTATTCTTTCCAGAAGATTG-3′.
Transient expression and confocal microscopy
Nicotiana tabacum L. cv Petit Havana SR1 was grown as described previously (Sparkes et al., 2006). A. tumefaciens cultures were infiltrated at the following OD600 ST-GFP 0.04, GFP-HDEL/RFP HDEL 0.04, RL2-RFP, the GTPase mutants (RL2-K53A-RFP, RL2-S54N-RFP, and RL2-R171Q-RFP) and GFP- or -HVA22b 0.05. Segments of infiltrated leaves were observed after 3–4 d with a Leica TCS SP5 confocal microscope (Leica Microsystems, Milton Keynes, UK) equipped with a ×63 (1.3NA) water immersion objective. GFP was excited at 488 nm and detected in the 495–520 nm range. YFP was excited at 514 nm and detected in the 525–550 nm range. RFP was excited at 561 nm and detected in the 571–638 nm range. Simultaneous detection of YFP and RFP was performed by combining these settings in the sequential scanning facility of the microscope, according to the manufacturer's instructions. The confocal microscope settings were kept constant throughout experiments.
Data acquisition and subsequent analysis using Volocity software version 3 (Perkin Elmer) for Golgi body tracking were carried according to (Sparkes et al., 2008) and (Avisar et al., 2009).
Nicotiana tabacum leaves were agroinfiltrated with both myc-RTNLB13 and RL2-RFP or its relative GTPase mutant constructs. Leaf sectors were homogenized in homogenization buffer (150 mM Tris–HCl pH 7.5, 150 mM NaCl, 1.5% (v/v) Triton X-100), supplemented immediately before use with ‘Complete’ protease inhibitor cocktail (Roche Diagnostics, Burgess Hill, UK) and subjected to immunoprecipitation with anti-myc monoclonal antibody 9E10 and Protein A Sepharose beads. The beads were subsequently washed three times using NET-Gel buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% (v/v) Nonidet P-40, 0.25% (w/v) gelatine, 0.02% (w/v) NaN3) and resuspended in sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer. Immunoselected polypeptides were resolved by SDS-PAGE, transferred to nitrocellulose filters and subjected to immunoblotting with either anti-myc or anti-DsRed antibody (Invitrogen).
Results and Discussion
HVA22b labels the cortical ER but does not affect its morphology
As a first step towards the characterization of HVA22b and RL2 proteins, we studied their intracellular location and effect on the ER network in the transient expression system of agroinfiltration in tobacco epidermal cells. AtHVA22 is the closest homologue of DP1/Yop1, a protein structurally similar to RTN in yeast and mammalian cells (Voeltz et al., 2006). There are five isoforms of HVA22 in Arabidopsis (Chen et al., 2002), with HVA22b (At5g62490) having a seed-limited expression pattern comparable to RTNLB13 and RL2 (Fig. S1). We generated fusions between HVA22b and GFP. When expressed in tobacco epidermal cells by agroinfiltration (Sparkes et al., 2006), GFP-HVA22b labels the ER uniformly (Fig. 1a–c) and, unlike RTN, it is also found in ER cisternae and the nuclear envelope (Fig. 1d–f). No apparent constriction of ER tubules, as normally observed with overexpressed RTNLB13 (Tolley et al., 2008), can be seen. This localization matches that reported for a different isoform, AtHVA22d (Chen et al., 2011). These results may indicate that HVA22 is not serving an ER-shaping structural role in plants. Indeed, sequence homology with Yop1 is relatively low (33.6% similarity, 18.9% identity with EMBOSS Needle tool at the European Bioinformatics Institute, www.ebi.ac.uk). More importantly, HVA22b has a predicted transmembrane topology that is different from the reticulon-like ‘w’ topology of Yop1 (Fig. S2), which may explain its lack of ability to induce membrane curvature and/or constriction of ER tubules.
RL2 labels the ER network and mutations in the GTPase domains affect ER morphology
We generated a fusion between RL2 and RFP and expressed it in tobacco epidermal leaf cells. Fig. 1(g–i) shows that, like HVA22b, RL2-RFP labels the whole of the ER network and appears to have no apparent structural effect on tubular ER. Therefore, it seems that constitutive overexpression of RL2 does not induce obvious morphological changes in tobacco epidermal cells.
Given that the GTPase activity of atlastin (Stefano et al., 2012) is key to its capacity to oligomerize and facilitate ER membrane fusion (Hu et al., 2009; Bian et al., 2011; Moss et al., 2011; Pendin et al., 2011), we tested whether RL2 localization and effect on ER shape were affected by mutations within its predicted GTPase domain. We introduced the following point mutations into RL2-RFP: K53A (corresponding to K80A in ATL1; Hu et al., 2009), S54N in the P-loop region (G1, same as Chen et al., 2011), and R171Q in the G4 guanosine binding motif. In tobacco leaf epidermal cells, all mutants still labelled the cortical ER network (Fig. 2), but its morphology appeared to be altered. A large number of cable-like structures were present in cells expressing K53A and R171Q (Fig. 2c,d,g,h), at the expense of the normal network of three-way junctions and the occasional, thicker bundles observed upon expression of wildtype RL2-RFP (Figs 1h, 2a,b). The appearance of these large ER strands is reminiscent of the ER phenotype observed in the rhd3-1 mutant (Zheng et al., 2004). This probably indicates that the GTPase mutations have a dominant negative effect, outcompeting the native protein. Surprisingly, a second mutation within the G1 motif, S54N, located to punctate structures on the ER (Fig. 2e,f). When S54N was coexpressed with the ER luminal marker GFP-HDEL, the cortical ER appeared to more cisternal in structure (Fig. S3D–F) – an effect reminiscent of the phenotype observed upon overexpression of the ER membrane marker GFP-calnexin (Irons et al., 2003) and Fig. S6(d–f).
When wildtype atlastin is overexpressed in mammalian cells, the ER assumes an abnormal, punctate appearance (Moss et al., 2011). Normal ER morphology is restored when expressing mutants impaired in membrane fusion (Moss et al., 2011). This is compatible with the predicted fusogenic function of atlastin. Ours and others' observations (Chen et al., 2011) seem to indicate a different behaviour for the plant atlastin homologues. Overexpression of wildtype RL2 does not affect ER morphology, whereas both down-regulation and expression of GTPase mutants result in ER morphology alterations. Intriguingly, the vegetative isoform of RHD3 cannot complement a mutant lacking Sey1p, the yeast atlastin homologue (Chen et al., 2011), perhaps indicating functional divergence of the protein or of essential interacting partners.
We also tested whether overexpression of wildtype or mutant RL2 had an effect on Golgi body motility, as indicated for the vegetative form of RHD3 (Chen et al., 2011). We coexpressed RL2 and its relative mutants with the Golgi marker ST-GFP (Boevink et al., 1998). We tested the effects in at least 20 cells taken from three independent experiments from cells coexpressing the Golgi and RL2 marker. Golgi bodies display a range of movements such as saltatory, stop-go, fast, and uni/birectional. In order to quantify the effects of RL2 on Golgi dynamics, velocity, displacement rate and meandering index were monitored. Displacement is the shortest straight-line distance between the beginning and end of a track, while the meandering index is calculated by dividing the displacement rate with the velocity, and therefore provides a measure of Golgi trajectory. For example, a meandering index of 1 corresponds to a Golgi body moving with a straight trajectory, whereas a lower number indicates a more random motion. All of these parameters were calculated for > 280 Golgi bodies per construct and plotted as a cumulative distribution frequency (Fig. 3; Sparkes et al., 2008). Comparisons between CDF curves were generated using the Kolmogorov-Smirnov statistical test (KS test). The velocities and displacement rates of Golgi bodies in cells expressing wild type RL2 or the S54N mutant do not appear to be significantly different to control cells only expressing the Golgi marker. However, velocity and displacement rate CDF plots for RL2 mutants K53A and R171Q mutants were significantly different to the control (P <0.06). As can be seen from the meandering index plot (D/V), Golgi bodies in cells expressing the RL2 mutant K53A display a more random motion than under control and other conditions. As it has been proposed that Golgi bodies move over or with the ER network in an actin dependent manner (DaSilva et al., 2004; Sparkes et al., 2008), such differences in movement between the expression of mutant RHD3 and the wild type could be ascribed to the reorganization of the network into a more cable-like form witnessed under mutant RLD2 expression.
Coexpression of RL2 and RTNLB13 affects the morphology of the cortical ER network
Our results indicate that RL2 labels the cortical ER network but mutations that affect its nucleotide-binding capacity lead to an apparent increase in the number of large ER strands and a reduction of ER regions presenting normal three-way junctions (Fig. 2). This hints at a role of RL2 in organizing the classic geometrical network of the cortical ER possibly by mediating homotypic fusion of ER tubules. In mammalian and yeast cells, atlastins and reticulons have been shown to interact (Hu et al., 2009). We therefore tested whether the overexpression of both RL2-RFP and RTNLB13-YFP affected ER morphology. When both constructs were co-infiltrated into tobacco leaves, at low magnification cells coexpressing both proteins presented very bright fluorescent spots, not present when RL2-RFP is coexpressed with GFP-HDEL (Fig. S4; compare A,C with B,D). At higher magnification, these bright areas appeared to be part of a severely altered cortical ER network (Fig. 4a–c, compare with adjacent cell expressing RTNLB13 only, asterisk). The number of three way junctions was drastically reduced and large membrane cables and aggregates became apparent (Fig. 4d–f). It is possible that this effect is a consequence of the overexpression of two ER membrane proteins and bears no relation to the function of either RL2 or RTNLB13. Therefore we tested whether the same phenotype could be caused by coexpressing RL2 and a ‘neutral’ ER membrane marker. Given the current paucity of polytopic ER membrane proteins available as markers in plants, we produced a new marker by generating a YFP fusion to the Arabidopsis homologue of derlin 1. Human derlin proteins 1-3 (Lilley & Ploegh, 2004; Oda et al., 2006) are homologues of Saccharomyces cerevisiae Der1p (Knop et al., 1996). All derlins function in the ER membrane as components of a larger complex where they may facilitate the recognition and retro-translocation of certain misfolded proteins from the ER to the cytosol in a protein quality control pathway. In Arabidopsis, there are three derlin homologues: one (At4g29330) shows homology to mammalian Derlin-1 and is annotated as Derlin 1; two other genes (At4g21810 and At4g04860) have closer homology to Der2 and are named Derlin 2.1 and Derlin 2.2, respectively (Fig. S5A). Arabidopsis Derlin1 is closer to its human counterpart than to Derlin 2.1 and 2.2 (Fig. S5A). When expressed in tobacco protoplasts alongside a plant ERAD substrate, ricin A chain (RTA; Di Cola et al., 2001), Derlin 1 does not seem to affect the degradation kinetics of RTA (Fig. S5B). More importantly for the present work, when expressed in tobacco leaves together with the luminal marker RFP-HDEL, derlin 1-YFP labelled the ER network without apparently affecting its shape, unlike more popular markers such as GFP-calnexin, which induces ER cisternae (Irons et al., 2003; Runions et al., 2006), and various reticulon isoforms, which induce constrictions in ER tubules (Sparkes et al., 2010; Fig. S6). Therefore, while derlin 1 may not function in the ERAD pathway its apparent lack of effect on ER morphology makes it an attractive potential polytopic (Fig. S5C) ER membrane marker.
When Derlin1-YFP was coexpressed with RL2, the ER network appeared normal (Fig. 4g–i). It is therefore unlikely that the drastic remodelling observed upon RL2 and RTNLB13 coexpression is a nonspecific effect arising from the simultaneous overexpression of two ER membrane proteins.
The disruption of the ER was also observed when RTNLB13 was coexpressed with the RL2 GTPase mutants (Fig. 5a–i). In the case of the most severe mutant, S54N, loss of network architecture was further aggravated by the membrane fragmentation phenotype observed with S54N alone (Fig. 5g–i). This indicates that the effect of coexpressing RTNLB13 and RL2 may be additive.
Given that the severe morphological alteration of the ER persists when GTPase mutants of RL2 are expressed, we asked whether this additive effect depended on the reticulon protein being functional. We therefore coexpressed RHD3 with a mutant of RTNLB13 (ΔTM4; Tolley et al., 2010), in which each transmembrane segment was shortened to match the ‘standard’ length (17 residues) of an ER membrane protein (Brandizzi et al., 2002). We have previously shown that ΔTM4, while still residing in the ER membrane, is no longer capable of inducing tubule constrictions, or converting ER sheets into tubules (Tolley et al., 2010). Indeed, when ΔTM4 was coexpressed with RL2, the ER network appeared unperturbed (Fig. 5j–l). Given that the ΔTM4 mutant protein appears to have comparable stability to wildtype RTNLB13, this indicates that the major disruption of tubular ER specifically arises from the combined action of RL2 and RTNLB13, and that such additive effect requires a functional reticulon.
In mammalian cells, interaction between ATL1 atlastin and the reticulon isoforms Rtn3c or Rtn4a appears to be mediated by their respective transmembrane domains; accordingly, interaction persists in ATL1 mutants with altered nucleotide binding activity (Hu et al., 2009). Our results seem to corroborate this observation, because the additive effect of RL2 and RTNLB13 is maintained in the GTPase mutants, while shortening of the TMD of RTNLB13 prevents the additive effect seen when coexpressed with RL2.
RL2 interacts with RTNLB13
Given the spectacular ER disruption phenotype observed by the simultaneous expression of RL2 and RTNLB13, we hypotheszsed that these proteins must interact physically. To test this, we performed coimmunoprecipitation experiments on infiltrated leaf sectors expressing a myc-tagged version of RTNLB13 (Tolley et al., 2008) and the RFP-tagged RL2. Fig. 6 shows that RL2-RFP can be coimmunoprecipitated with RTNLB13-myc. Coimmunoprecipitation with RTNLB13- myc is also observed for each of the GTPase mutants (Fig. 6). This further confirms the hypothesis that interaction between reticulon and atlastin homologues does not require a functional GTPase domain.
In conclusion, the data presented here confirm that the plant homologues of atlastin the RHD3 family are involved in the geometrical organization of the ER network, perhaps through regulation and formation of three-way junctions between tubules. This role may be regulated by an interaction with the membrane-curving reticulon proteins, although such an interaction is not dependent on the GTPase activity of RL2. The exact nature of this interaction and the potential role of RL2 in ER tubule fusion remain to be ascertained.
H.L. was supported by a BBSRC studentship and I.S. by an Oxford Brookes Nigel Groome Fellowship. We are grateful to Christopher Snowden for the cloning of Arabidopsis Derlin1.