Proliferation and Morphogenesis of the Endoplasmic Reticulum Driven by the Membrane Domain of 3-Hydroxy-3-Methylglutaryl Coenzyme A Reductase in Plant Cells^,

The membrane domain of 3-hydroxy-3-methylglutaryl CoA reductase contributes to the morphogenesis of the endoplasmic reticulum.

Abstract

The enzyme 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGR) has a key regulatory role in the mevalonate pathway for isoprenoid biosynthesis and is composed of an endoplasmic reticulum (ER)-anchoring membrane domain with low sequence similarity among eukaryotic kingdoms and a conserved cytosolic catalytic domain. Organized smooth endoplasmic reticulum (OSER) structures are common formations of hypertrophied tightly packed ER membranes devoted to specific biosynthetic and secretory functions, the biogenesis of which remains largely unexplored. We show that the membrane domain of plant HMGR suffices to trigger ER proliferation and OSER biogenesis. The proliferating membranes become highly enriched in HMGR protein, but they do not accumulate sterols, indicating a morphogenetic rather than a metabolic role for HMGR. The N-terminal MDVRRRPP motif present in most plant HMGR isoforms is not required for retention in the ER, which was previously proposed, but functions as an ER morphogenic signal. Plant OSER structures are morphologically similar to those of animal cells, emerge from tripartite ER junctions, and mainly build up beside the nuclear envelope, indicating conserved OSER biogenesis in high eukaryotes. Factors other than the OSER-inducing HMGR construct mediate the tight apposition of the proliferating membranes, implying separate ER proliferation and membrane association steps. Overexpression of the membrane domain of Arabidopsis (Arabidopsis thaliana) HMGR leads to ER hypertrophy in every tested cell type and plant species, whereas the knockout of the HMG1 gene from Arabidopsis, encoding its major HMGR isoform, causes ER aggregation at the nuclear envelope. Our results show that the membrane domain of HMGR contributes to ER morphogenesis in plant cells.

The endoplasmic reticulum (ER) consists of a cytosolic dynamic weft of tubules and cisternae, with varying general shape and extension, continuously adapted to the changing demands of its functional commitment (Staehelin, 1997; Federovitch et al., 2005; Stefano et al., 2014). A high increase in the number and extension of ER membranes (ER hypertrophy) is a common feature of eukaryotic cells specialized in biosynthesis and secretion of metabolites and proteins. This is, for instance, the case of plant gland cells involved in the production of diverse isoprenoid products (Schnepf, 1974) and mammalian adrenocortical cells involved in steroid hormone biosynthesis (Fawcett, 1981). Such processes require elevated levels of ER resident enzymes like 3-hydroxy-3-methylglutaryl CoA reductase (HMGR), cytochrome b5, and cytochrome P450 (Boutanaev et al., 2015; Storbeck et al., 2014). Interestingly, the overexpression of these enzymes in yeast (Saccharomyces cerevisiae) or mammalian cells induces proliferation and accumulation of ER membranes that, at the electron microscope, are visualized as stacked membranes (lamellae or karmellae), concentric membrane associations (whorled ER), or ordered arrays of membranes with hexagonal or cubic symmetry (crystalloid ER; Jingami et al., 1987; Wright et al., 1988). These structures have been collectively named organized smooth endoplasmic reticulum (OSER; Snapp et al., 2003). OSER structures are natural formations broadly distributed in animal and plant cells (Snapp, 2005; Almsherqi et al., 2009) and can be induced in yeast and mammalian cells by overexpression of about a dozen endogenous ER membrane proteins (Sandig et al., 1999). Overexpression studies led to the proposal that OSER structures are caused by membrane zippering because of oligomerization of the expressed protein (Takei et al., 1994; Gong et al., 1996; Yamamoto et al., 1996; Fukuda et al., 2001; Li et al., 2003; Snapp et al., 2003; Lingwood et al., 2009; Costantini et al., 2012), but other biochemical processes are likely involved in OSER biogenesis. The huge membrane proliferation observed upon expression of OSER-inducing proteins, such as yeast or human HMGR (Jingami et al., 1987; Wright et al., 1988) or diverse cytochrome P450 variants (Sandig et al., 1999), requires very active lipid and protein biosynthesis and cannot be explained by mere oligomerization. Despite long-lasting research, the mechanisms and signals triggering ER proliferation and OSER biogenesis remain elusive (Wright, 1993; Federovitch et al., 2005). Particular attention has been paid to identification of sequence or structural features in the overexpressed protein critical for its OSER-inducing effect (Jingami et al., 1987; Vergères et al., 1993; Parrish et al., 1995; Yamamoto et al., 1996; Profant et al., 1999, 2000). However, the work to date has not allowed the identification of broadly conserved peptide or structural motifs that function as a signal to trigger ER proliferation and morphogenesis.

Eukaryotic HMGR has a key regulatory role in the mevalonate pathway for isoprenoid biosynthesis (Burg and Espenshade, 2011; Rodríguez-Concepción et al., 2013). Isoprenoid products derived from the pathway are involved in diverse essential functions, including membrane biogenesis (sterols), control of growth and development (steroid hormones and cytokinins), protein prenylation (farnesyl and geranyl groups), protein glycosylation (dolichols), and respiration (ubiquinones). In plants, the mevalonate pathway also provides a wide variety of compounds required for defense or positive interactions with other organisms (Chappell, 1995). HMGR has a bipartite structure formed by an N-terminal membrane domain, with low sequence similarity among eukaryotic kingdoms, and a highly conserved catalytic domain, and the two are connected by a flexible nonconserved linker region (Liscum et al., 1985; Basson et al., 1988; Campos and Boronat, 1995). HMGR is tetrameric in animals and plants because of oligomerization of the catalytic domain (Bach et al., 1986; Wititsuwannakul et al., 1990; Istvan et al., 2000). The membrane domain of plant HMGR, which is composed of two conserved hydrophobic segments, a short lumenal loop and a cytosolic N-terminal region (Campos and Boronat, 1995), is fairly small compared with those of yeast and human HMGR, which are predicted to have eight transmembrane sequences (Olender and Simon, 1992; Profant et al., 2000). The N-terminal region is diverged among plant HMGR isoforms but contains a highly conserved N-terminal Arg motif that fits the consensus of ER retention signal described in animal type II membrane proteins (Campos and Boronat, 1995). Despite their structural differences, all eukaryotic HMGRs coincide in basic functional features. The membrane domains of yeast and human HMGR determine the targeting, turnover, and capacity to induce ER proliferation, therefore controlling the catalytic activity of the appended domain (Gil et al., 1985; Jingami et al., 1987; Skalnik et al., 1988; Olender and Simon, 1992; Parrish et al., 1995; Hampton et al., 1996). Conversely, a properly folded cytosolic domain is required to trigger karmellae biogenesis in yeast or slow down regulated HMGR degradation in human cells, indicating that the tertiary structure and function of the membrane domain are influenced by the oligomerization state of the catalytic domain (Cheng et al., 1999; Profant et al., 1999). Thus, in yeast and animal cells, membrane and catalytic HMGR domains show interdependent function.

In Arabidopsis (Arabidopsis thaliana), two genes encode three HMGR isoforms: HMGR1S and HMGR1L are encoded by HMG1, and HMGR2 is encoded by HMG2 (Enjuto et al., 1994; Lumbreras et al., 1995). Isozymes HMGR1S and HMGR1L are identical in sequence, but HMGR1L has an N-terminal extension of 50-amino acid residues (1L extra region). HMGR1S transcript is expressed in the whole plant, whereas those encoding HMGR2 or HMGR1L are detected at a much lower level and only in meristems, seedlings, roots, and inflorescences (Enjuto et al., 1994, 1995; Lumbreras et al., 1995). Chimeras formed by the membrane domain of plant HMGR fused to GFP localize in the ER network but accumulate in highly fluorescent cytosolic bodies (Leivar et al., 2005; Merret et al., 2007). In this work, we show that these bodies are OSER structures. Expression of the membrane domain of HMGR1S or HMGR2 fused to GFP leads to reversible formation of OSER structures in every cell type of diverse plant species. The conserved N-terminal Arg motif is not an ER retention signal but is required to induce ER proliferation and morphogenesis. In addition, the knock of the HMG1 gene from Arabidopsis causes severe defects in ER morphology around the nucleus. These results indicate that the membrane domain of plant HMGR is naturally involved in the morphogenesis of the ER.

RESULTS

1S:GFP and 2:GFP Induce Biogenesis of ER Aggregates, But 1L:GFP Is Retained in the ER Network

Plant HMGR is primarily targeted to the ER by interaction of the highly conserved hydrophobic sequences of its membrane domain with the signal recognition particle (Campos and Boronat, 1995). To find out whether the less conserved regions flanking these sequences might influence the final subcellular location, we designed chimeras of the three Arabidopsis HMGR isoforms carrying the fluorescent marker GFP instead of the catalytic domain (Fig. 1A). Chimeras 1S:GFP, 1L:GFP, and 2:GFP contain the membrane domains of Arabidopsis HMGR1S, HMGR1L, and HMGR2, respectively. The chimeras were transiently expressed by biolistic transfection or agroinfection in different cell types from Arabidopsis, Nicotiana benthamiana, onion (Allium cepa), and leek (Allium porrum). In epidermal, vascular, mesophyll, trichome, and culture cells from Arabidopsis and epidermal cells from onion and leek, 1S:GFP and 2:GFP are detected in the ER network but mainly accumulate in highly fluorescent vesicular structures, which remain close to the ER (Fig. 1, B–E; Supplemental Figs. S1, A–D, S2, A–C, and S3, A–F). In all assayed systems, these structures emerge as small spherical bodies in three-branch junctions of the cortical ER or beside the nuclear envelope (Fig. 1, B–D; Supplemental Figs. S2, A–C and S3, A–F) and remain physically connected to the ER network, which was confirmed by three-dimensional reconstruction of confocal microscopy images (Supplemental Movie S1). They can become large fluorescent bodies usually located near the nucleus (Fig. 1D). In transfected N. benthamiana leaves, 1S:GFP was expressed at high levels in epidermal cells and mostly accumulated in large, densely packed structures around the nucleus (Fig. 1G). In parallel transfections of Arabidopsis, N. benthamiana, and onion, the ER lumen markers GFP-Lys-Asp-Glu-Leu (GFP-KDEL) and GFP-His-Asp-Glu-Leu (GFP-HDEL) outlined an ER network characteristic of the targeted cell type (Fig. 1I; Supplemental Figs. S1E, S2E, and S3I), confirming that the aggregates were formed de novo upon 1S:GFP or 2:GFP expression. In contrast to 1S:GFP and 2:GFP, the 1L:GFP chimera was exclusively found in the ER network in all transfected plant systems, which included epidermal cells from Arabidopsis, N. benthamiana, and onion and an undifferentiated cell line from Arabidopsis (Fig. 1, F and H; Supplemental Figs. S2D and S3, G and H). Because 1S:GFP and 1L:GFP behave differently, formation of the ER-derived aggregates is likely prevented by the 1L extra region.

Figure 1.

ER morphology after expression of 1S:GFP, 2:GFP, and 1L:GFP in Arabidopsis and N. benthamiana cells. A, Schematic representation in the ER membrane of HMGR:GFP chimeras used in this study. Chimeras 1S:GFP, 1L:GFP, and 2:GFP have the membrane domains of Arabidopsis HMGR1S, HMGR1L, and HMGR2, respectively, fused to C-terminal GFP. Construct 1S is in the membrane domain of HMGR1S alone. B to L, Arabidopsis (B–F) and N. benthamiana (G–L) leaves were transfected with constructs encoding the HMGR chimera (1S:GFP, 2:GFP, or 1L:GFP) or a control ER lumen marker (GFP-HDEL or T3RE) and visualized by confocal laser microscopy. J to L, Cotransfection with 1S:GFP and T3RE was visualized by confocal laser microscopy as indicated. The images are Z projections of 10 (B–F), 50 (G–I), or 20 (J–L) sections from transfected Arabidopsis epidermal (B, C, and F), Arabidopsis vascular (D and E), or N. benthamiana epidermal (G–L) cells. N, Nucleus. Bars = 10 μm.

The physical connection of the 1S:GFP-containing aggregates to the ER network suggested but did not prove that they are part of the ER. To examine their identity, we cotransfected 1S:GFP with T3RE, which is a Discosoma spp. red fluorescent protein derivative targeted to the ER lumen (Forés et al., 2006). In N. benthamiana, Arabidopsis, and onion epidermal cells, there was a perfect match between 1S:GFP and T3RE fluorescence in the ER network, nuclear envelope, and spherical structures (Fig. 1, J–L; Supplemental Fig. S4). Because T3RE carries a functional ER retention motif (C-terminal KDEL), the physical coincidence implies that the 1S:GFP-containing aggregates belong to the ER. The 1S:GFP chimera can be used as a general marker of the altered ER: it labels all ER domains, including the induced ER membrane aggregates.

The Biogenesis of ER Aggregates Induced by 1S:GFP Is Reversible

The 1S:GFP chimera was assayed also by stable transformation of Arabidopsis plants. Six independent transgenic lines expressing high levels of 1S:GFP in the whole plant (Fig. 2A) were brought to homozygosity to prevent segregation in progeny. Alteration of the ER morphology was general in 7-d-old 1S:GFP plants. ER-derived vesicular structures were observed, for instance, in cotyledon epidermal cells, root hairs, and stomata guard cells (Fig. 2, B–D). Although the transgenic lines had a slightly delayed growth with respect to the wild type, they were phenotypically normal. The observed disturbance of the ER structure did not block ER function and plant development. As in the transient expression assays, 1S:GFP aggregates developed from ER network junctions (Fig. 2E) and attained the largest size at the nuclear envelope (Fig. 2, B–D). Interestingly, 1S:GFP expression vanished between weeks 3 and 4 of plant development (Fig. 2F) but reproducibly reappeared upon seed germination, at least until generation five. The ER was altered in cells still expressing 1S:GFP (Fig. 2G) but had normal network structure in cells without the chimera, which was indicated by transient expression of lumenal ER markers (Fig. 2, H and I). Biolistic transfection with the ER marker was conducted shortly after disappearance of 1S:GFP, ruling out that the observed network could belong to a newly differentiated cell. The transfected leaves had the expected anatomy, with no signs of plant tissue degeneration or regeneration (Fig. 2F). We conclude that the loss of 1S:GFP allowed progressive recovery of normal ER architecture, which was altered again in the next generation, because of reappearance of the chimera. ER disturbance driven by 1S:GFP does not irreversibly affect the organelle and is compatible with cell integrity and function.

Figure 2.

Reversible formation of ER aggregates in transgenic 1S:GFP Arabidopsis plants. Transgenic plants were analyzed by stereomicroscopy (A and F) and confocal laser microscopy (B–E and G–I) to detect the fluorescence of 1S:GFP (green; A–G), endogenous chlorophyll (red; F), or ER lumen markers (red or green; H and I). A, Ten-day-old transgenic seedlings have widespread 1S:GFP expression. B to D, The chimera induces ER-derived aggregates in all cell types, including leaf epidermal cells (B and E), root hairs (C), and stomata guard cells (D). F, After 3 weeks of growth, 1S:GFP vanishes, allowing detection of red fluorescence from chlorophyll. G to I, ER-derived aggregates are present in leaf sectors still expressing 1S:GFP (G) but not in areas without the chimera, which is shown by biolistic transfection of ER lumen markers (H and I). Confocal microscope images are Z projections of 12 (B, D, E, H, and I) and 21 (G) sections from the transfected cells, except in C, which corresponds to a single section. N, Nucleus. Bars = 3 mm (A and F) and 10 μm (B to E and G to I).

The 1S:GFP-Induced ER Aggregates Are OSER Structures That Result from Membrane Proliferation

To further characterize the ER-derived vesicular structures, we analyzed N. benthamiana and Arabidopsis 1S:GFP-expressing organs by transmission electron microscopy (EM). In N. benthamiana epidermal cells, 1S:GFP expression correlates with the appearance of densely packed ER membrane aggregates (Fig. 3, A and B), which were never seen in nontransfected organs. The hypertrophied ER had ordered membranes with crystalloid, lamellar, or whorled patterns (Fig. 3B), such as the OSER structures previously described in animal cells (Snapp et al., 2003). Also coincident with them, 1S:GFP-induced OSER structures have a narrow cytosolic space of about 10 nm between membranes, whereas the broader areas correspond to the ER lumen (Fig. 3, D and E). Ribosomes are excluded from the aggregates, which only contain smooth ER (Fig. 3, B, D, and E). Similar smooth ER patterns with narrow cytosolic space were observed in all organs and cell types of transgenic 1S:GFP Arabidopsis plants (Fig. 3, F–I). ER regions with crystalloid, whorled, or lamellar membranes were interconnected in the same aggregate, indicating that the pattern may depend on the orientation of the section (Fig. 3, B, D, and F–I). The crystalloid structures of both Arabidopsis and N. benthamiana cells frequently appear beside the nuclear envelope (Fig. 3, C and F), which is in agreement with the observations at the confocal microscope. Ultrastructural and confocal microscope analyses indicate that expression of 1S:GFP or 2:GFP leads to a high increase of the amount of ER membranes. The 1S:GFP and 2:GFP chimeras are very effective inducers of membrane proliferation and OSER biogenesis in plant cells.

Figure 3.

Ultrastructural analysis of 1S:GFP-induced ER aggregates. Agroinfected N. benthamiana leaves (A–E) and transgenic Arabidopsis seedlings (F–I) expressing 1S:GFP were imaged by transmission EM. A to E, In N. benthamiana leaf epidermal cells, the aggregates are composed of interconnected crystalloid, lamellar, and whorled smooth ER (A and B) frequently located beside the nucleus (C). Whorled structures contain alternating lumenal (aligned dots) and more electron-dense cytosolic spaces that exclude ribosomes (white arrowheads; D). The narrow (10 nm) electron-dense space between membranes (arrowhead) shows continuity with the cytosol (E). F to I, In all cell types of transgenic 1S:GFP Arabidopsis seedlings, the ER membrane aggregates also accumulate around the nucleus and have diverse interconnected patterns. Cr, Crystalloid; La, lamellar; N, nucleus; n, nucleolus; W, whorled. Bars = 2 µm (A), 0.5 µm (B), 1 µm (C and F), 200 nm (D), 100 nm (E), and 0.5 µm (G–I).

The OSER Structures Have High Levels of 1S:GFP and HMGR But Low Sterol Content

To confirm the identity of 1S:GFP-derived OSER structures and determine whether they also contain endogenous HMGR, we performed double-immunocytochemical analyses on transgenic 1S:GFP Arabidopsis plants, with polyclonal antibodies raised against GFP or the catalytic domain of HMGR1. For a panoramic view of single cells, we first ran whole-mount immunocytochemistry on 6-d-old seedlings. In cotyledon parenchymal cells, signals corresponding to 1S:GFP and endogenous HMGR (fluorescing in green and red, respectively) physically coincided in an irregular mass of ER around the blue-labeled nucleus (Fig. 4, A–C). The ER was concentrated surrounding this organelle, with poor cortical ER (Fig. 4A).

Figure 4.

1S:GFP-induced OSER structures have high levels of 1S:GFP and HMGR but low sterol content. A to C, Whole-mount immunocytochemical analysis of transgenic 1S:GFP Arabidopsis seedlings. Immunodetection of 1S:GFP (green) and HMGR (red) in cotyledon parenchymal cells visualized by confocal microscopy. The two proteins colocalize in ER aggregates around the DAPI-stained (blue) nucleus. D to H, Immunocytochemical EM analysis of transgenic 1S:GFP Arabidopsis seedlings. 1S:GFP (18-nm particle) and endogenous HMGR (12-nm particle) were immunodetected in leaf epidermal cells. The two proteins colocalize on membranes of crystalloid aggregates (D, E, and G), whorled aggregates (F), and nuclear envelope (arrowhead in G). No signal is observed in the control without primary antibodies (H). I, Quantification of HMG1 and HMG2 transcripts and HMGR protein in transgenic 1S:GFP seedlings. Transcript and protein levels were measured by qRT-PCR or immunoblot and are represented relative to the HMG1 transcript or HMGR1S protein of wild-type seedlings, respectively, as the average ± sem of three independent experiments. Electrophoretic migration of HMGR1L (69 kD) and HMGR1S (63 kD) is indicated. Coomassie-stained Rubisco was used as a loading control and is shown below the immunoblot. J to K, Sterol and 1S:GFP distribution in root cortical cells in transgenic 1S:GFP seedlings. Samples were stained with filipin III and visualized by confocal microscopy. Free sterols show poor colocalization with 1S:GFP. Confocal microscope images are Z projections of 10 (A–C) or 15 (J–l) sections. Cr, Crystalloid; N, nucleus; W, whorled; wt, wild type. Bars = 10 µm (A–C), 1 µm (D), 200 nm (E and F), 0.5 µm (G and H), and 30 µm (J–L).

The subcellular locations of 1S:GFP and endogenous HMGR were further analyzed by immunocytochemical ultrastructural studies. In leaf epidermal cells of the transgenic Arabidopsis seedlings, 1S:GFP and endogenous HMGR were detected at high densities on membranes of ER-derived aggregates (Fig. 4, D and E). Mixed dual labeling was observed in crystalloid, whorled, and lamellar ER aggregates, the interconnecting ER network, and the nuclear envelope (Fig. 4, E–G). No signal was detected in the control without primary antibody (Fig. 4H) or controls with primary and secondary antibodies in other cell compartments (Supplemental Fig. S5). The observation of normal morphologies for the Golgi, plasmalemma, mitochondria, and chloroplasts indicates that membrane proliferation induced by 1S:GFP is exclusive of the ER (Supplemental Fig. S5). Our results indicate that 1S:GFP and endogenous HMGR colocalize in all regions of the ER but accumulate mainly in the ER-derived aggregates, which we refer to as ER-HMGR domains. The immunodetection of GFP shows that these domains correspond to the highly fluorescent bodies identified in the confocal microscope. There is full coincidence in subcellular location between 1S:GFP and endogenous HMGR in both confocal microscopy and EM, indicating that the elimination of the catalytic domain does not alter the subcellular distribution of the HMGR protein, which is entirely determined by its membrane domain.

The high abundance of endogenous HMGR in 1S:GFP-induced OSER structures suggested that HMGR accumulates at higher levels in transgenic plants than in wild-type plants. To examine this hypothesis, crude extracts from Arabidopsis seedlings were submitted to immunoblot with an antibody specific for the HMGR catalytic domain. As shown in Figure 4I, the HMGR protein level was about 3 times higher in transgenic 1S:GFP plants than in the wild type. To ascertain whether this increase was a consequence of elevated transcript levels, we ran quantitative reverse transcription and real-time PCR with separate primer sets specific for HMG1 and HMG2 (Fig. 4I). We found that the total HMGR transcript did not increase but consistently decreased in the transgenic 1S:GFP plants. Thus, the increase in HMGR protein was not caused by higher transcript levels. The 1S:GFP chimera and endogenous HMGR accumulate in OSER structures induced by the membrane domain of HMGR1S, which has a self-stabilization effect. To find out whether the higher HMGR level caused sterol accumulation, we submitted roots of the transgenic plants to sterol-specific staining with filipin III (Boutté et al., 2011). Although the filipin signal was evident in the plasmalemma and some intracellular structures, it was nearly undetectable in 1S:GFP-induced OSER structures (Fig. 4, J–L), indicating very low levels of sterols with the unreacted 3′-β-OH group. The main role of plant HMGR in OSER biogenesis does not seem to be metabolic but morphogenetic.

The Membrane Domain of HMGR1S Suffices to Trigger OSER Biogenesis

It was previously shown that expression of dimerizing GFP fused to the cytoplasmic domain of otherwise passive ER resident proteins causes ER membrane stacking and OSER biogenesis and that equivalent constructs with a monomeric GFP variant do not have this morphogenic effect (Snapp et al., 2003; Costantini et al., 2012). To find out whether dimerization of GFP was contributing to the OSER-inducing capacity of 1S:GFP in plant cells, we designed the 1S:GFP monomeric (1S:GFPm) chimera containing a nondimerizing GFP. Of the three single-amino acid changes known to prevent GFP dimerization (Zacharias et al., 2002), we chose the A206K substitution, because it has the lowest dimerization capacity. Its dissociation constant is about 650 times higher than that of the original GFP, making the corresponding dimer essentially undetectable (Zacharias et al., 2002). The morphogenic capacity of 1S:GFPm was first analyzed by stable transformation of the Arabidopsis SALK_061790 knockout mutant (Myouga et al., 2010), which does not express HMGR1S or HMGR1L. In SALK_061790 parenchymal cells, 1S:GFPm expression converted the ER network into perinuclear and cortical ER aggregates containing the monomeric chimera, which was shown by whole-mount immunocytochemical analysis (Fig. 5A). No endogenous HMGR could be detected under the same conditions (Fig. 5B), confirming that the observed alterations were not caused by a sharp increase in HMG2 expression. In leaf epidermal cells from wild-type Arabidopsis, transient expression of 1S:GFPm led to the emergence of ER-derived vesicles from ER junctions (Fig. 5C), similar to those induced by transient or stable expression of 1S:GFP (Figs. 1B and 2E). The 1S:GFPm aggregates became as large as those induced by 1S:GFP (compare Fig. 5D with Fig. 1D). The 1S:GFPm chimera accumulated as well in thick ER strands that were connected to the thin ER network (Fig. 5E). In N. benthamiana epidermal cells, 1S:GFPm led to formation of big perinuclear ER aggregates with radially disposed thick ER strands (Fig. 5F). Thus, the appearance of thick ER strands is reiterative in cells from different plants expressing 1S:GFPm.

Figure 5.

OSER structures induced by 1S:GFPm and 1S fragment. A and B, Whole-mount immunocytochemical analysis of the SALK_061790 Arabidopsis mutant stably transformed with 1S:GFPm. Immunodetection of 1S:GFPm (A) and HMGR (B) in cotyledon parenchymal cells of 1S:GFPm/SALK_061790 seedlings visualized by confocal microscopy with DAPI-stained nuclei. C to F, Transient expression of 1S:GFPm in epidermal cells from wild-type Arabidopsis (C–E) and N. benthamiana (F) leaves visualized by confocal microscopy. G and H, Transient coexpression of 1S fragment and T3RE ER marker (G) in epidermal cells from N. benthamiana leaves using expression of T3RE alone as a control (H). I to L, Transmission EM of epidermal cells from N. benthamiana leaf after transient expression of 1S:GFPm. Confocal microscope images are Z projections of 10 (A–E), 33 (F), 15 (G), and 19 (H) sections. Cr, Crystalloid; N, nucleus; W, whorled. Bars = 10 µm (A–H), 1 µm (I), 0.5 µm (J and K), and 200 nm (L).

To confirm that the membrane domain of HMGR1S is self-sufficient to induce OSER structures, we removed GFP from the 1S:GFP chimera. This was achieved by incorporation of two consecutive stop codons at an appropriate site of the 1S:GFP coding sequence without altering other sequences in the expression plasmid. The resulting construct encoded a 20-kD HMGR1S fragment, named 1S, containing the N-terminal region, the two membrane-spanning segments, and the linker region. In N. benthamiana leaf epidermal cells, the nonfluorescent 1S fragment led to formation of ER aggregates, which were visualized by cotransfected T3RE, whereas when expressed alone, the ER lumen marker localized in the typical ER network (Fig. 5, G and H). This indicates that the membrane domain of HMGR1S devoid of an appended cytosolic domain can disturb ER morphology and induce OSER structures. ER proliferation and OSER biogenesis were not as pronounced as those induced by 1S:GFP or 1S:GFPm chimeras, probably because of lower expression or stability of the 1S fragment.

To characterize OSER structures produced in the absence of dimerizing GFP, we obtained electron micrographs of agroinfected N. benthamiana leaves. Ultrastructural analysis of the epidermal cells uncovered that expression of 1S:GFPm led to very active proliferation of the smooth ER that was converted to interconnected crystalloid, lamellar, and whorled aggregates, equivalent to those generated by 1S:GFP (Fig. 5, I–L). Particular ultrastructural features, such as exclusion of ribosomes from the aggregate and the narrow and regular cytosolic spaces between membranes (Fig. 5, K and L), also coincided with those of 1S:GFP-induced OSER formations. The apposition of the cytosolic side of the ER membranes is remarkable, because monomeric GFP cannot contribute to the close interaction. Most likely, other ER resident proteins accomplished this task. We conclude that expression of 1S:GFPm in plant cells leads to ER proliferation and association of the accumulated membranes rendering typical OSER structures.

The Conserved N-Terminal Motif of Plant HMGR Is an OSER-Inducing Signal

It was previously suggested that the Arg motif conserved in the N-terminal region of plant HMGR might be an ER retention signal (Campos and Boronat, 1995; Merret et al., 2007). As a first step to examine its function, we searched for all available N-terminal plant HMGR-encoding sequences in the National Center for Biotechnology Information database. To avoid overestimating variability, only full-length complementary DNAs and ESTs confirmed by at least two independent entries were considered. We found 305 nonredundant HMGR sequences from 158 species and 117 genera of monotyledonae and dicotyledonae, which broadly represent the flowering plant kingdom (Supplemental Data Set S1). In these sequences, the three-Arg tag is only part of a longer motif consisting of the N-terminal Met, one or two acidic residues, a hydrophobic residue, the three arginines, and two prolines (Fig. 6A); 273 of 305 plant HMGR isoforms (90%) contain this canonical sequence (Supplemental Data Set S1). From them, 194 sequences keep all three arginines, whereas 79 have introduced one- or two-amino acid substitutions. The most repeated variations of the three-Arg tag (written in one-letter code) are HRR (17 sequences), RQR (9 sequences), SRR (8 sequences), RRK (7 sequences), RRL (4 sequences), RRQ (4 sequences), HRK (3 sequences), and RRG (3 sequences). These variants keep at least two arginines or three basic residues, indicating sequence constraints for a conserved function.

Figure 6.

Role of the N-terminal motif of plant HMGR in the morphogenesis of the ER. A, N-terminal motif of plant HMGR deduced from nonredundant complementary DNA and EST entries (Supplemental Data Set S1) with MEME software (Bailey et al., 2009). Letter size is proportional to amino acid residue frequency in the referred position. The amount of motif-defining information corresponding to each amino acid position is represented in the ordinate, in bits. B to I, The original 1S:GFP (N-terminal sequence MDLRRRPPKPP) and its mutated variants ΔDLRRR-1S:GFP (sequence MGPPKPP) and ΔRRR-1S:GFP (sequence MDLAAAPPKPP) were transiently expressed by biolistic transfection in Arabidopsis leaf (B–E) or by agroinfection in Arabidopsis cell culture (F and G) and N. benthamiana leaf (H and I). Vascular (B and C), epidermal (D, E, H, and I), and in vitro culture cells (F and G) were visualized by confocal microscopy. Images are Z projections of 10 (B–G) or 50 (H and I) sections. N, Nucleus. Bars = 10 µm.

To examine the role of the conserved N-terminal motif, we obtained new variants of 1S:GFP, in which the DLRRR sequence was replaced by a single Gly residue (construct ΔDLRRR-1S:GFP) or the three arginines were replaced by three alanines (construct ΔRRR-1S:GFP; Fig. 6). The chimeras were transiently expressed in leaf and cell culture from Arabidopsis or N. benthamiana. In epidermal, vascular, or in vitro cultured cells, both mutated variants localized in the cortical ER network (Fig. 6, B, D, F, and H). No OSER structures were induced by ΔDLRRR-1S:GFP or ΔRRR-1S:GFP, despite the long-lasting expression, which depending of the system, was prolonged for 18 to 96 h. Under the same conditions, the 1S:GFP chimera induced typical OSER structures (Fig. 6, C, E, G, and I). We conclude that the conserved Arg motif is required for ER proliferation and OSER biogenesis triggered by 1S:GFP but not for retention in the ER, which should be mediated by other elements of the N-terminal domain. In plant HMGR, the conserved N-terminal motif signals ER proliferation and morphogenesis.

In Arabidopsis, the Disruption of the HMG1 Gene Causes ER Aggregation around the Nucleus

The requirement of a highly conserved HMGR motif for OSER biogenesis in plant cells and the OSER-inducing capacity of this protein in the three eukaryotic kingdoms strongly suggested that HMGR has a natural role in the morphogenesis of the ER. To further examine this hypothesis, we determined ER morphology in the SALK_061790 line from Arabidopsis, which carries a transfer DNA (T-DNA) insertion in the HMG1 gene and is the knockout for HMGR1S and HMGR1L. We transformed this line with GFP-HDEL to trace the ER lumen. Transformation was performed in heterozygous SALK_061790 plants, and the morphology of the ER was studied in generation T2, which had a segregating T-DNA insertion genotype. This allowed comparison of the three HMG1 allelic combinations on the same genetic background. The ER was characterized in cells from the root cortex of 7- to 10-d-old plants. At this age, the homozygous mutant was at the two-cotyledon stage and could be easily distinguished from homozygous wild-type and heterozygous plants, which already had true leaves (Fig. 7H). We examined the nuclear and cortical ER in root cells from 36 mutant and 55 wild-type plants. In total, 406 Z-projection images were obtained (201 from the mutant and 205 from the wild type) with reproducible ER morphologies. In the cortex from homozygous wild-type and heterozygous plants, the GFP-HDEL labeling was prominent in ER bodies and weaker in the cortical ER network and the nuclear envelope (Fig. 7, A–C). The ER bodies are natural fusiform enlargements of the ER and very abundant in Arabidopsis seedlings (Yamada et al., 2011). In the wild type, some ER bodies were around the nucleus but only occasionally in contact with the nuclear envelope (Fig. 7A). However, in homozygous mutant cells, ER bodies aggregated at the nuclear envelope (Fig. 7, D–F). Under bright field, the nucleus and surrounding ER bodies formed a clearly visible rough mass (Fig. 7F), which could be identified as well by 4′,6-diamidino-2-phenylindole (DAPI) staining (Fig. 7J). The differences between the wild type and the mutant are more evident in Z-projection images of single nuclei (Fig. 7, G and I) and videos generated from the corresponding individual sections (Supplemental Movies S2 and S3). A cross optical section of the root confirmed that the nuclear aggregates of ER bodies occurred in the cortex (just below the epidermal layer; Fig. 7, K and L). The disruption of the HMG1 gene caused concentration of ER bodies at the nuclear envelope but did not affect the cortical ER network and the ER bodies connected to it, which was observed in optical section series (compare Supplemental Movie S2 with Supplemental Movie S3). We conclude that the knockout of HMG1 gene from Arabidopsis selectively affects the organization of the ER around the nucleus, causing aggregation of ER bodies.

Figure 7.

Effect of the disruption of the HMG1 gene from Arabidopsis on the morphology of the ER. Arabidopsis plants of the SALK_061790 line heterozygous for the disrupted HMG1 gene were transformed with GFP-HDEL. At the segregating generation T2, the differentiation zone of roots from 7- to 10-d-old plants was examined by confocal laser microscopy. Images were captured after fluorescent excitation of GFP or propidium iodide (A, D, G, and I–L) under bright field (C and F), or with the merge of both channels (B and E). A to C, Morphology of the ER in the root of the heterozygous plant. The GFP-HDEL label is detected in the cortical ER, the nuclear envelope, and more intensively, ER bodies. The nuclei (indicated by arrows) appear as fusiform structures. D to F, Morphology of the ER in the root of the homozygous SALK_061790 mutant plant. The GFP-HDEL label mainly distributes in ER bodies around the nucleus. The nuclei (seen from the front; arrows) have a circular appearance with an irregular outline. G, Nuclear region in cortex cells from the SALK_061790 hmg1 mutant. The image corresponds to a Z projection of 15 sections from the GFP channel taken every 0.85 μm. The ER bodies accumulate around the nucleus (yellow N). The individual sections are shown on bright field in Supplemental Movie S2. H, Phenotype of SALK_061790 homozygous hmg1 mutant (left) and wild-type (right) plants 10 d after growth at 22°C. The mutant plant has a delayed development in both aerial and root organs and is still at the two cotyledon stage. I, Nuclear region in cortex cells from the wild type. The center of the nucleus is indicated with a yellow N. The image corresponds to a Z projection of 15 sections from the GFP channel taken every 0.52 μm. The individual sections are shown on bright field in Supplemental Movie S3. J, DAPI staining and GFP detection of homozygous mutant root. The image corresponds to a Z projection of seven sections. K and L, Propidium iodide staining and GFP detection of homozygous mutant root. L corresponds to the XZ scanning section indicated by the dashed line in K. The nuclear ER body aggregates localize in cortex cells just below the epidermal layer. A to F, K, and L correspond to single scanning sections. G, I, and J are Z projections. Bars = 20 µm (A–F, K, and L), 10 µm (G, I, and J), and 3 mm (H).

DISCUSSION

We have shown that the membrane domain of Arabidopsis HMGR is involved in the morphogenesis of the ER. On the one hand, the knockout of the HMG1 gene from Arabidopsis encoding its major HMGR isoform altered the ER morphology around the nucleus. On the other hand, the membrane domain of HMGR1S or HMGR2 fused to GFP-induced massive ER proliferation and OSER biogenesis through a specific well-conserved sequence. The OSER structures emerged from ER junctions and mostly accumulated around the nucleus. Similar alterations were induced by 1S:GFP and 2:GFP in all tested cell types from diverse plant species, indicating a conserved mechanism. The nondimerizing 1S:GFPm chimera was as effective as 1S:GFP in the induction of ER proliferation and membrane stacking, indicating that an appended dimerizing cytosolic domain is not required for the morphogenic effect. Expression of the membrane domain of HMGR1S alone also disturbed the ER network and generated OSER structures, confirming its sufficiency in the process. In agreement with the dispensability of the catalytic domain of HMGR, the induced OSER structures did not accumulate sterols. Our finding that sterols are barely detected in OSER structures is consistent with their expected low level in ER membranes (Korn, 1969) and indicates that sterol accumulation is not required for OSER biogenesis, which is in contrast to a previous proposal (Lingwood et al., 2009). Our results indicate a morphogenetic rather than a catalytic role for HMGR in OSER biogenesis, but the proliferated membranes could subsequently accommodate active enzyme machineries. It was previously proposed that the synthesis of different classes of isoprenoids could be mediated by enzymatic arrays or metabolons containing specific HMGR isozymes in particular ER domains (Chappell, 1995). Plant and animal cells engaged in high production of definite isoprenoid products typically contain OSER structures (Schnepf, 1974; Fawcett, 1981) that may contain the predicted metabolons.

We found that the three-Arg motif conserved in the N-terminal region of plant HMGR is an absolute requirement to induce ER proliferation and morphogenesis. In animal cells, Arg residues located between positions 2 and 5 of type II membrane proteins constitute an ER retention signal (Schutze et al., 1994), and a motif of Arg residues close to the N terminus was shown to accomplish an ER retention role as well in plant cells (Boulaflous et al., 2009). In contrast, the Arg tag of plant HMGR is not required for retention in the ER but represents an ER morphogenic signal. Nevertheless, the three basic residues are only a part of the N-terminal motif conserved in plant HMGR, and the two hydrophobic segments and the short lumenal region of the membrane domain are also highly conserved (Campos and Boronat, 1995). These elements may also contribute to HMGR-driven ER proliferation and morphogenesis. The requirement of the three-Arg sequence for ER morphogenesis indicates that the same motif identified in 194 of 305 analyzed plant HMGRs (64% of the total) should accomplish an equivalent role. We have experimentally confirmed Arabidopsis HMGR1S and HMGR2 as members of this group. Other HMGR variants carrying two Arg residues of three Arg positions or with conservative changes in the basic motif (an additional 26%) will likely perform the same or a very similar function. However, those HMGRs with no N-terminal Arg motif (the remaining 10%) represented by Arabidopsis HMGR1L may be unable to induce ER proliferation and OSER biogenesis. Thus, ER morphology may be regulated by expression of different HMGR isoforms in plant cells. The fact that OSER biogenesis is driven by plant HMGR through conserved sequences, its occurrence in all tested plants and cell types, its reversibility, and its compatibility with cell function and plant development reinforce that ER morphogenesis is naturally controlled by HMGR. This was confirmed experimentally in Arabidopsis, where the disruption of the HMG1 gene severely affected the morphology of the nuclear ER.

In the three eukaryotic kingdoms, HMGR exerts key regulation in sterol biosynthesis through its catalytic domain and triggers ER proliferation and OSER biogenesis through its membrane domain. This coincidence has been preserved along evolution, underlining its importance for eukaryotic life. Remarkably, the membrane domain of HMGR has retained its morphogenic capacity, despite low sequence conservation. OSER biogenesis driven by HMGR was shown here in plants and observed previously in mammalian cells and yeast in three cases using homologous expression systems (Jingami et al., 1987; Wright et al., 1988). In addition, OSER structures were induced by expression of yeast HMGR in mammalian cells or mammalian HMGR in yeast (Wright et al., 1990), indicating common mechanisms in this morphogenic phenomenon. Also conserved is the subcellular location for ER proliferation. OSER biogenesis driven by HMGR has been shown to initiate at the nuclear envelope in mammalian cells (Pathak et al., 1986), and HMGR-induced lamellae (thus named karmellae) also appear around the nucleus in yeast (Wright et al., 1988). In plant cells, OSER structures emerge from ER network junctions and around the nucleus, where they mostly accumulate. The morphogenic action of HMGR at the nuclear envelope was confirmed by disruption of the HMG1 gene from Arabidopsis, which caused ER aggregation at this site. The fact that plant HMGR requires a highly conserved motif to control morphogenesis of the ER suggests that other eukaryotic HMGRs may have alternative sequences with equivalent function.

OSER structures induced by 1S:GFP or 1S:GFPm in plant cells are similar to those observed in animals. Higher eukaryote OSER structures are formed by tightly packed interconnected regions of crystalloid, lamellar, and whorled smooth ER membranes and morphologically more complex than the lamellar OSER formations observed in yeast. Also, in a close similarity, animal and plant cell OSER structures contain membrane stacks with two distinct alternating spaces: the electron-translucent lumen, which shows a general ordered pattern but can be distended and disordered in some regions, and the more electron-dense narrow cytosolic area that is delimited by adjacent cisternae, which has a very constant width. The electron-dense aspect and even spacing of the cytosolic area indicate accumulation of regularly disposed material, likely proteins. It was suggested that the width of the cytosolic space depends on the size of the overexpressed OSER-inducing protein (Snapp et al., 2003), but this is not the case in OSER structures induced by plant HMGR. Because 1S:GFPm does not dimerize, other ER resident or cytosolic ER-interacting proteins should be involved in the close association of membranes. Interestingly, freeze-fracture EM of transfected human cells uncovered a high density of intramembrane particles in the cytosolic face of the OSER membranes and a much lower number of similar structures in the lumenal face (Anderson et al., 1983). It is tempting to speculate that the observed protoplasmic particles may correspond to protein complexes mediating membrane apposition.

It was previously noticed that the formation of OSER structures in mammalian cells correlated with the expression level of the OSER-inducing protein. When cytochrome b5 was highly expressed, it induced OSER structures, but at lower expression, the ER network was not disturbed (Snapp et al., 2003). Thus, it was argued that OSER biogenesis merely results from overexpression of ER membrane proteins (Yamamoto et al., 1996; Li et al., 2003; Snapp et al., 2003). Certainly, OSER biogenesis induced by 1S:GFP is associated to a high accumulation of the chimera. However, we found that OSER structures also accumulate endogenous HMGR protein, with concomitantly reduced HMGR transcript levels. Thus, the membrane availability induced by 1S:GFP somehow stabilizes the chimera and also, the endogenous HMGR, which was previously found in several plants to be submitted to ongoing degrading processes (Korth et al., 2000; Leivar et al., 2011; Pollier et al., 2013). Other endogenous and heterologous ER membrane proteins are also stabilized in OSER structures (Kochevar and Anderson, 1987; McLaughlin et al., 1999). Chimeras 1L:GFP, ΔDLRRR1S:GFP, and ΔRRR1S:GFP are nearly identical to 1S:GFP, with the only difference of the naturally extended N-terminal region or the elimination of the conserved N-terminal motif. In agreement with the stabilization of HMGR in OSER structures, the observed lower expression levels of 1L:GFP, ΔDLRRR1S:GFP, and ΔRRR1S:GFP might not be the cause but a consequence of their inability to induce ER proliferation and morphogenesis.

Our data indicate that ER proliferation and OSER biogenesis can be induced by ER resident proteins, such as HMGR, carrying conserved short-peptide motifs that function as ER morphogenetic signals. Previous work showed that overexpression of oligomerizing ER proteins leads to conversion of preexisting ER into OSER structures without significant membrane proliferation (Snapp et al., 2003). On this basis, we propose a two-step model for OSER biogenesis in eukaryotic cells: (1) ER proliferation induced by definite sequences and structural elements of particular ER resident proteins, such as HMGR, and (2) membrane stacking promoted by oligomerization of ER resident proteins. This model would allow induction of OSER structures by overexpression of oligomerizing ER proteins, skipping the first of two proposed steps. Abundant ER resident proteins (not necessarily the ones inducing ER proliferation) could mediate the stacking of membranes. In support of this, OSER biogenesis was triggered by expression of 1S:GFPm and 1S constructs, which do not have a cytosolic oligomerizing domain. The 1S-GFPm OSER structures had tightly stacked ER membranes with narrow cytosolic space that are indistinguishable from those induced by 1S:GFP in plant cells (this work) or other oligomerizing proteins in animal cells (Snapp et al., 2003). The stabilization of ER resident proteins in OSER structures detected here for endogenous HMGR could be favored by the availability of membrane space or the isolation from external proteases. The stabilized proteins would then contribute to OSER biogenesis by promoting membrane association. The membrane stacking step might be energy dependent, because ATP deprivation severely disrupted the ordered pattern of proliferated ER membranes (Lingwood et al., 2009). Sustained expression of the OSER-inducing protein may be needed to maintain OSER structures, because we found that disappearance of 1S:GFP in plant cells allows reversion to normal ER network. Thus, our observations suggest that OSER formation and disassembly may be controlled in a simple way by determining the expression level of morphogenic HMGR isoforms.

MATERIALS AND METHODS

Plant Material

Arabidopsis (Arabidopsis thaliana) seeds were surface sterilized by washing three times with 70% (v/v) ethanol and three times with 100% (v/v) ethanol, sowed in petri dishes with Murashige and Skoog medium (Murashige and Skoog, 1962), and vernalized at 4°C for 3 d. Seedlings were grown under long-day conditions (16-h-light/8-h-dark cycle) at 22°C to 24°C. Arabidopsis plants were grown on a perlite:vermiculite:sphagnum (1:1:11) mixture in the same light regime. The undifferentiated T87 cell line from Arabidopsis was cultured as indicated (Axelos et al., 1992). The SALK_061790 mutant from Arabidopsis (Myouga et al., 2010) carrying a T-DNA insertion in the first exon of gene HMG1 (AT1G76490) was obtained from the European Arabidopsis Stock Centre. Nicotiana benthamiana was grown on perlite-vermiculite-sphagnum in a greenhouse under natural day conditions. Fresh onions (Allium cepa) and leeks (Allium porrum) were purchased in the market and used a few hours later for biolistic transfection.

Constructs and Plant Transfection

For biolistic transfection, constructs were cloned in plasmid pGFPau (Leivar et al., 2005). For agroinfection or generation of Arabidopsis transgenic plants, constructs were cloned in plasmid pPCV002 (Ferrando et al., 2001) provided by Alejandro Ferrando (Instituto de Biología Molecular y Celular de Plantas, Valencia, Spain). In both plasmids, transcription is under control of the cauliflower mosaic virus 35S promoter, which confers high constitutive expression. Constructs were prepared by conventional PCR and cloning techniques as indicated in Supplemental Table S1 using the oligonucleotides listed in Supplemental Table S2. The pPCV002 derivatives were transformed into Agrobacterium tumefaciens strain GV3101 pMP90RK (Koncz and Schell, 1986). Transgenic Arabidopsis plants were generated by the floral dip method (Clough and Bent, 1998) using the transformed A. tumefaciens cells. Transient expression in Arabidopsis leaves, onion scale leaves, and leek scale leaves was achieved by biolistic transfection with pGFPau derivatives. Transient expression in the Arabidopsis T87 cell line (Ferrando et al., 2001) and N. benthamiana leaves (Van der Hoorn et al., 2000) was achieved by infection with A. tumefaciens (agroinfection) carrying the appropriate pPCV002 derivatives. Biolistic transfection was performed with a Biolistic PDS-1000/He System (BioRad) with a 900-pounds-force per square inch rupture disc (BioRad) as described (Lois et al., 2000). Samples were observed 18 to 48 h after biolistic transfection and 72 to 96 h after agroinfection.

Determination of Transcript and Protein Levels

HMG1 and HMG2 transcripts were quantified by quantitative reverse transcription and real-time PCR as described (Leivar et al., 2011) using the At4g26410 transcript as a normalization reference. The primer sets and the amplicon sizes for the transcripts were as follows: At4g26410, primers 264F and 264R (81 bp); HMG1, primers 764F and 764R (85 bp); and HMG2, primers HMG2F and HMG2R (92 bp). Primers are listed in Supplemental Table S2. HMGR protein levels were determined by immunoblot with anti-CD1-i antibody as described (Leivar et al., 2011).

Determination of ER Morphology in the SALK_061790 Mutant

To determine the morphology of the ER, the SALK_061790 line from Arabidopsis carrying a T-DNA insertion in the HMG1 gene was transformed with the GFP-HDEL marker of the ER lumen. Because the homozygous disruption of the Arabidopsis HMG1 gene severely affects development and fertility, transformation was performed on heterozygous plants. A. tumefaciens strain GV3101 pMP90 pVK-Her6 encoding GFP-HDEL (Batoko et al., 2000) was provided by Ian Moore (University of Oxford, UK). Transformants were selected in generation T1 by resistance to hygromycin and the appearance of GFP fluorescence in the whole plant. In general, the fluorescence was more intense in the root. Generation T1 was self-crossed to obtain T2 seeds, which were germinated in one-half-strength Murashige and Skoog medium. After 7 d of growth, homozygous mutants were still at the two-cotyledon stage, whereas homozygous wild-type and heterozygous plants had developed true leaves. The root was sectioned between days 7 and 10 for immediate observation with a confocal microscope, and the rest of the plant was transferred to a fresh plate for additional growth before confirmation of the genotype by PCR. The wild-type allele was identified with primers H1.2F and H1.3R (547-bp amplification product), and the mutant allele was identified with primers LB6 and H1.3R (358-bp product).

Stereomicroscopy and Confocal Microscopy

Arabidopsis seedlings were observed with the stereomicroscope Olympus SZX16 (with excitation at 460 to 495 nm and emission above 510 nm) to detect GFP and chlorophyll. Arabidopsis seedlings were stained with filipin III (Boutté et al., 2011), DAPI (Kapuscinski, 1995), or propidium iodide (González-García et al., 2011) and observed with the confocal microscope as reported. Confocal laser microscopy was performed with spectral microscope Olympus FV1000 (objectives UPLSAPO 60× O, numerical aperture: 1.35 and UPLSAPO 60× W, numerical aperture: 1.20) or Leica TCS SP5 (objectives HCX PL APO CS 40.0 × 1.25 oil UV, HCX PL APO 63.0 × 1.20 W corrected for UV, and HCX PL APO CS 63.0 × 1.20 water UV) at room temperature. Fluorophores were detected with the following excitation and emission wavelengths: GFP (excitation = 488 nm, emission = 500–545 nm), T3RE (excitation = 543 nm, emission = 555–700 nm), Filipin-sterol complexes (excitation = 364 nm, emission = 400–484 nm), DAPI (excitation = 405 nm, emission = 425–490 nm), propidium iodide (excitation = 559 nm, emission = 570–670 nm), Alexa Fluor-488 (excitation = 488 nm, emission = 500–545 nm), and Alexa Fluor-594 (excitation = 559 nm, emission = 570–670 nm). Images were acquired with the software FV10-ASW 4.1 (Olympus) and LAS AF 2.7.3.9723 (Leica) and processed with the software ImageJ 1.48v (http://imagej.nih.gov/ij).

Antibodies

HMGR was detected with immunosubtracted rabbit serum recognizing its catalytic domain (anti-CD1-i; Leivar et al., 2011) as the primary antibody at 1:1,000 for immunoblot, 1:500 for whole mount, and 1:1,000 for transmission EM. Anti-rabbit IgG secondary antibodies for HMGR detection were code NA934V (GE Healthcare) coupled to horseradish peroxidase at 1:5,000 for immunoblot, code Ab150068 (Abcam) coupled to Alexa Fluor-594 at 1:1,000 for whole mount, and code 111-205-144 (Jackson Immunoresearch) coupled to a 12-nm gold particle at 1:30 for transmission EM. GFP was detected with Ab-5450 (Abcam) as the primary antibody at 1:1,000 for whole mount and transmission EM. Anti-goat IgG secondary antibodies for GFP detection were code Ab150133 (Abcam) coupled to Alexa Fluor-488 at 1:1,000 for whole mount and code 705-215-147 (Jackson Immunoresearch) coupled with an 18-nm gold particle at 1:15 for transmission EM.

Immunolocalization in Whole Mount

Whole-mount in situ immunolocalization was done as described (Sauer et al., 2006) with modifications. After fixation in 4% (w/v) paraformaldehyde, seedlings were incubated in methanol to remove chlorophylls. Five cycles of freezing and thawing of the seedlings on glass slides were performed to permeate tissue followed by incubation with 2% (w/v) Driselase (D8037; Sigma) to allow for penetration of antibodies through the plant cell wall. After blocking with 3% (w/v) bovine serum albumin (BSA) solution, incubation with one or two primary antibodies simultaneously (anti-CD1-i at 1:500 and/or Ab-5450 at 1:1,000) was performed directly on the slides. Samples were then washed with phosphate-buffered saline (PBS) and incubated with one or two secondary antibodies. Secondary antibody for HMGR detection (Abcam code Ab150068; Alexa Fluor-594) was used at 1:1,000. Secondary antibody for GFP detection (Abcam code Ab150133; Alexa Fluor-488) was used at 1:1,000. After washing with PBS, samples were sealed for observation at the confocal microscope.

Transmission EM Ultrastructural Studies

Explants from N. benthamiana leaves or Arabidopsis seedlings were removed under a stereomicroscope and transferred to glass vials filled with 1.5% (v/v) paraformaldehyde and 1.5% (v/v) glutaraldehyde in phosphate buffer (pH 7.4) containing 2 mm CaCl₂. The vials were incubated at 4°C for 48 h. After washing with the phosphate-CaCl₂ buffer, samples were postfixed for 3 h at 4°C with 1% (w/v) osmium tetroxide and 0.8% (w/v) potassium ferricyanide in the same buffer. Samples were subsequently dehydrated in acetone, infiltrated with Spurr resin for 4 d, embedded in the same resin, and polymerized at 60°C for 48 h. Cell integrity was confirmed at the light microscope. Ultrathin sections were obtained using an Ultracut UC6 Ultramicrotome (Leica) and mounted on Formvar-coated copper grids. Samples were stained with 2% (w/v) uranyl acetate in water and lead citrate. Samples were observed in a JEM-1010 Electron Microscope (JEOL) equipped with a CCD Camera SIS Megaview III and the AnalySIS software.

Transmission EM Immunocytochemical Studies

Explants from Arabidopsis seedlings were cryoimmobilized by high-pressure freezing using an EM Pact (Leica) with yeast (Saccharomyces cerevisiae) paste as the filler. Freeze substitution of frozen samples was performed in an Automatic Freeze Substitution System EM AFS (Leica) with methanol containing 0.5% (w/v) uranyl acetate at −90°C for 3 d. On the fourth day, the temperature was increased slowly by 5°C per hour to −50°C. At this temperature, samples were rinsed in acetone, infiltrated, and flat embedded in Lowicryl HM20 for 10 d. Ultrathin sections were picked up on Formvar-coated nickel grids. Sample-containing grids were incubated on drops of PBS with 5% (w/v) BSA for 20 min at room temperature. After removal of the washing solution, drops of PBS with the polyclonal primary antibody (anti-CD1-i or Ab-5450 at 1:1,000) and 1% (w/v) BSA were added and incubated for 2 h. Grids were washed three times for 30 min with a drop of PBS with 0.25% (v/v) Tween 20 and incubated for 1 h in drops of PBS with the secondary antibody and 1% (w/v) BSA. Secondary antibody for HMGR detection (code 111-205-144; Jackson Immunoresearch; 12-nm gold particle) was used at 1:30. Secondary antibody for GFP detection (code 705-215-147; Jackson Immunoresearch; 18-nm gold particle) was used at 1:15. Grids were washed three times with a drop of PBS for 5 min and two times with distilled water and air dried. In control assays for nonspecific binding of the gold-conjugated antibody, the primary antibody was omitted. Sections were stained with 2% (w/v) uranyl acetate in water and lead citrate and observed in a JEM-1010 Electron Microscope (JEOL) with an SIS Mega View III CCD Camera.

Bioinformatics

The N-terminal motif of plant HMGR was obtained with MEME software version 4.9.1 (Bailey et al., 2009), which is accessible at http://meme-suite.org/ using the defect parameters.

Sequence data from this article can be found in the Arabidopsis Genome Initiative and GenBank databases under accession numbers HMGR1L (At1g76490 and AAR83122), HMGR1S (At1g76490 and AAF16652), and HMGR2 (At2g17370 and AAA67317).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. ER morphology after expression of 1S:GFP and 2:GFP in Arabidopsis cells.

• Supplemental Figure S2. ER morphology after expression of 1S:GFP, 2:GFP, and 1L:GFP in Arabidopsis cell culture.

• Supplemental Figure S3. ER morphology after expression of 1S:GFP, 2:GFP, and 1L:GFP in onion or leek epidermal cells.

• Supplemental Figure S4. The 1S:GFP-containing aggregates belong to the ER.

• Supplemental Figure S5. Controls of immunodetection and organelle morphology.

• Supplemental Table S1. Preparation of DNA constructs.

• Supplemental Table S2. Oligonucleotides.

• Supplemental Data Set S1. N-terminal sequence of plant HMGR.

• Supplemental Movie S1. Physical connection between the ER network and ER-derived aggregates.

• Supplemental Movie S2. Morphology of the ER around the nucleus in the Arabidopsis hmg1 mutant.

• Supplemental Movie S3. Morphology of the ER around the nucleus in wild-type Arabidopsis.

Glossary

BSA

bovine serum albumin

DAPI

4′,6-diamino-phenylindole

EM

electron microscopy

ER

endoplasmic reticulum

OSER

organized smooth endoplasmic reticulum

PBS

phosphate-buffered saline

qRT

quantitative reverse transcription

T-DNA

transfer DNA