The divergent domains of the NEFA and nucleobindin proteins are derived from an EF-hand ancestor

The human protein NEFA (DNA binding, EF-hand, Acidic region) has previously been isolated from a KM3 cell line and immunolocalized on the plasma membrane, in the cytoplasma, and in the culture medium. Sequence analysis of a cDNA clone encoding NEFA identified a hydrophilic domain, two EF-hands, and a leucine zipper at the C- terminus. These characters are shared with nucleobindin (Nuc). In this paper we have further characterized NEFA and probed its evolutionary origins. Circular dichroism (CD) spectra of recombinant NEFA indicated a helical content of 51% and showed that the EF-hands are capable of binding Ca2+. Experiments with recombinant NEFA and synthesized peptides revealed that the leucine zipper cannot form a homodimer. The leucine zipper may allow heterodimer formation of NEFA and an unknown protein. Phylogenetic analyses suggest that this protein is derived from a four-domain EF-hand ancestor with subsequent duplications and fusions. The leucine zipper and putative DNA-binding domains of NEFA have evolved secondarily from existing EF-hand sequences. These analyses provide insights into how complex proteins may originate and trace the precursor of NEFA to the common ancestor of eukaryotes.      

CALNUC (nucleobindin) is localized in the Golgi apparatus in insect cells

A mouse monoclonal antibody 12B1 was raised against Golgi fractions from Sf21 insect cells and selected as Golgi-specific by immunostaining of the cells. The antigen was purified from the cells by immunoaffinity chromatography with the monoclonal antibody, and its N-terminal and internal amino acid sequences were determined. Based on the partial amino acid sequences, cDNA encoding the antigen protein was cloned and sequenced. The amino acid sequence deduced from the cDNA nucleotide sequence showed a homology to those of CALNUC family proteins, CALNUC (or nucleobindin, a calcium-binding Golgi protein with DNA-binding activity) and protein NEFA (a cell surface protein with DNA-binding, EF-hand, and acidic domains). The insect protein had two EF-hand loops at the same sites as the mammalian CALNUC family proteins, but had no leucine zipper which the mammalian homologues commonly have. An electron microscopic immunoperoxidase study demonstrated that the insect protein was localized in the cis-Golgi cisternae and cis-Golgi networks. Since this localization is identical to that of mammalian CALNUC, the insect protein was considered to be a homologue of CALNUC rather than that of NEFA. Assays involving proteinase K digestion, sodium carbonate extraction and Triton X-114 extraction revealed that the insect CALNUC-like protein was a soluble protein tightly associated with the luminal surface of Golgi membranes as reported for mammalian CALNUC. The insect protein was also shown to have calcium-binding activity as does mammalian CALNUC. These data verify that the insect protein is CALNUC. The existence of CALNUC in insect cells suggests that CALNUC is an essential calcium-binding Golgi protein in a wide range of the animal kingdom. A phylogenetic tree analysis, however, suggested that NEFA was derived from CALNUC long after the segregation of a mammalian ancestor from an insect ancestor.     

Heterologous overexpression of human NEFA and studies on the two EF-hand calcium-binding sites.

Human NEFA is an EF-hand, leucine zipper protein containing a signal sequence. To confirm the calcium binding capacity of NEFA, recombinant NEFA analogous to the mature protein and mutants with deletions in the EF-hand domain were expressed in Pichia pastoris and secreted into the culture medium at high yield. The calcium binding activity of each purified protein was measured by a modified equilibrium dialysis using the fluorescent Ca2+ indicator FURA-2 and atomic absorption spectroscopy. A stoichiometry of 2 mol Ca2+/mol NEFA was determined. The Ca2+ binding constants were resolved by intrinsic fluorescence spectroscopy. Fluorescence titration exhibited two classes of Ca2+ binding sites with Kd values of 0.08 microM and 0.2 microM. Circular dichroism (CD) spectroscopy showed an increase from 30 to 43% in the amount of alpha-helix in NEFA after addition of calcium ions. Limited proteolytic digestion indicated a Ca2+ dependent conformational change accompanied by an altered accessibility to the enzyme.     

The Mammalian Calcium-binding Protein, Nucleobindin (CALNUC), Is a Golgi Resident Protein

We have identified CALNUC, an EF-hand, Ca²⁺-binding protein, as a Golgi resident protein. CALNUC corresponds to a previously identified EF-hand/calcium-binding protein known as nucleobindin. CALNUC interacts with Gα_i3 subunits in the yeast two-hybrid system and in GST-CALNUC pull-down assays. Analysis of deletion mutants demonstrated that the EF-hand and intervening acidic regions are the site of CALNUC's interaction with Gα_i3. CALNUC is found in both cytosolic and membrane fractions. The membrane pool is tightly associated with the luminal surface of Golgi membranes. CALNUC is widely expressed, as it is detected by immunofluorescence in the Golgi region of all tissues and cell lines examined. By immunoelectron microscopy, CALNUC is localized to cis-Golgi cisternae and the cis-Golgi network (CGN). CALNUC is the major Ca²⁺-binding protein detected by ⁴⁵Ca²⁺-binding assay on Golgi fractions. The properties of CALNUC and its high homology to calreticulin suggest that it may play a key role in calcium homeostasis in the CGN and cis-Golgi cisternae.     

Molecular cloning of nucleobindin, a novel DNA-binding protein that contains both a signal peptide and a leucine zipper structure.

We have previously reported that KML1-7 cells cloned from a lupus-prone MRL/l mouse produced a soluble factor that preferentially expanded anti-DNA antibody production across the H-2 barrier. We purified this factor, a 55 kD protein that we termed nucleobindin (Nuc), and obtained its cDNA clone. Although the gene for Nuc encodes a signal peptide and, in fact, Nuc was identified as a secreted protein, Nuc had a DNA-binding property. The putative polypeptide predicted from the cDNA sequence featured a signal peptide, a leucine zipper structure and a basic amino acid-rich region. The DNA-binding property of Nuc was destroyed by deletion of either the leucine zipper structure or the basic amino acid-rich region. The amino acid sequences of Nuc are highly conserved between mouse and human. We discuss the possible role of Nuc in autoimmunity.     

A putative leucine zipper within the herpes simplex virus type 1 UL6 protein is required for portal ring formation

The herpes simplex virus type 1 UL6 protein forms a 12-subunit ring structure at a unique capsid vertex which functions as a conduit for encapsidation of the viral genome. To characterize UL6 protein domains that are involved in intersubunit interactions and interactions with other capsid proteins, we engineered a set of deletion mutants spanning the entire gene. Three deletion constructs, D-5 (Delta 198-295), D-6 (Delta 322-416), and D-LZ (Delta 409-473, in which a putative leucine zipper was removed), were introduced into the viral genome. All three mutant viruses produced only B capsids, indicating a defect in encapsidation. Western blot analysis showed that the UL6 protein was present in the capsids isolated from two mutants, D-6 and D-LZ. The protein encoded by D-5, on the other hand, was not associated with capsids and was instead localized in the cytoplasm of the infected cells, indicating that this deletion affected the nuclear transport of the portal protein. The UL6 protein from the KOS strain (wild type) and the D-6 mutant were purified from insect cells infected with recombinant baculoviruses and shown to form ring structures as assessed by sucrose gradient centrifugation and electron microscopy. In contrast, the D-LZ mutant protein formed aggregates that sedimented throughout the sucrose gradient as a heterogeneous mixture and did not yield stable ring structures. A mutant (L429E L436E) in which two of the heptad leucines of the putative zipper were replaced with glutamate residues also failed to form stable rings. Our results suggest that the integrity of the leucine zipper region is important for oligomer interactions and stable ring formation, which in turn are required for genome encapsidation.     

Origin of the NEFA and Nuc Signal Sequences

The human protein NEFA binds calcium, contains a leucine zipper repeat that does not form a homodimer, and is proposed (along with the homologous Nuc protein) to have a common evolutionary history with an EF-hand ancestor. We have isolated and characterized the N-terminal domain of NEFA that contains a signal sequence inferred from both endoproteinase Asp-N (Asp-N) and tryptic digests. Analysis of this N-terminal sequence shows significant similarity to the conserved multiple domains of the mitochondrial carrier family (MCF) proteins. The leader sequence of Nuc is, however, most similar to the signal sequences of membrane and/or secreted proteins (e.g., mouse insulin-like growth factor receptor). We suggest that the divergent NEFA and Nuc N-terminal sequences may have independent origins and that the common high hydrophobicity governs their targeting to the ER. These results provide insights into signal sequence evolution and the multiple origins of protein targeting. Received: 20 February 1997 / Accepted: 28 July 1997     

Leucine zipper-mediated homodimerization of the p21-activated kinase-interacting factor, beta Pix. Implication for a role in cytoskeletal reorganization

Pix, a p21-activated kinase-interacting exchange factor, is known to be involved in the regulation of Cdc42/Rac GTPases. The 85-kDa betaPix-a protein contains an Src homology 3 domain, the tandem Dbl homology and Pleckstrin homology domains, a proline-rich region, and a GIT1-binding domain. In addition to those domains, betaPix-a also contains a putative leucine zipper domain at the C-terminal end. In this study, we demonstrate that the previously identified putative leucine zipper domain mediates the formation of betaPix-a homodimers. Using in vitro and in vivo methodologies, we show that deletion of the leucine zipper domain is sufficient to abolish betaPix-a homodimerization. In NIH3T3 fibroblast cells, expression of wild type betaPix-a induces the formation of membrane ruffles. However, cells expressing the leucine zipper domain deletion mutant could not form membrane ruffle structures. Moreover, platelet-derived growth factor-mediated cytoskeletal changes were completely blocked by the leucine zipper domain deletion mutant. The results suggest that the leucine zipper domain enables betaPix-a to homodimerize, and homodimerization is essential for betaPix-a signaling functions leading to the cytoskeletal reorganization.     

Intramembrane proteolysis and endoplasmic reticulum retention of hepatitis C virus core protein

Hepatitis C virus (HCV) core protein is suggested to localize to the endoplasmic reticulum (ER) through a C-terminal hydrophobic region that acts as a membrane anchor for core protein and as a signal sequence for E1 protein. The signal sequence of core protein is further processed by signal peptide peptidase (SPP). We examined the regions of core protein responsible for ER retention and processing by SPP. Analysis of the intracellular localization of deletion mutants of HCV core protein revealed that not only the C-terminal signal-anchor sequence but also an upstream hydrophobic region from amino acid 128 to 151 is required for ER retention of core protein. Precise mutation analyses indicated that replacement of Leu(139), Val(140), and Leu(144) of core protein by Ala inhibited processing by SPP, but cleavage at the core-E1 junction by signal peptidase was maintained. Additionally, the processed E1 protein was translocated into the ER and glycosylated with high-mannose oligosaccharides. Core protein derived from the mutants was translocated into the nucleus in spite of the presence of the unprocessed C-terminal signal-anchor sequence. Although the direct association of core protein with a wild-type SPP was not observed, expression of a loss-of-function SPP mutant inhibited cleavage of the signal sequence by SPP and coimmunoprecipitation with unprocessed core protein. These results indicate that Leu(139), Val(140), and Leu(144) in core protein play crucial roles in the ER retention and SPP cleavage of HCV core protein.     

Human cytomegalovirus UL84 oligomerization and heterodimerization domains act as transdominant inhibitors of oriLyt-dependent DNA replication: evidence that IE2-UL84 and UL84-UL84 interactions are required for lytic DNA replication

Human cytomegalovirus (HCMV) UL84 encodes a 75-kDa protein required for oriLyt-dependent DNA replication and interacts with IE2 in infected and transfected cells. UL84 localizes to the nucleus of transfected and infected cells and is found in viral replication compartments. In transient assays it was shown that UL84 can interfere with the IE2-mediated transactivation of the UL112/113 promoter of HCMV. To determine whether UL84 protein-protein interactions are necessary for lytic DNA synthesis, we purified UL84 and used this protein to generate a monoclonal antibody. Using this antibody, we now show that UL84 forms a stable interaction with itself in vivo. The point of self-interaction maps to a region of the protein between amino acids 151 and 200, a domain that contains a series of highly charged amino acid residues. Coimmunoprecipitation assays determined that UL84 interacts with a protein domain present within the first 215 amino acids of IE2. We also show that an intact leucine zipper domain of UL84 is required for a stable interaction with IE2 and UL84 leucine zipper mutants fail to complement oriLyt-dependent DNA replication. UL84 leucine zipper mutants no longer interfere with IE2-mediated transactivation of the UL112/113 promoter, confirming that the leucine zipper is essential for a functional interaction with IE2. In addition, we demonstrate that both the leucine zipper and oligomerization domains of UL84 can act as transdominant-negative inhibitors of lytic replication in the transient assay, strongly suggesting that both an IE2-UL84 and a UL84-UL84 interaction are required for DNA synthesis.     

Trafficking of the Ca2+-activated K+ channel,hIK1, is dependent upon a C-terminal leucine zipper

We demonstrate that the C-terminal truncation of hIK1 results in a loss of functional channels. This could be caused by either (i) a failure of the channel to traffic to the plasma membrane or (ii) the expression of non-functional channels. To delineate among these possibilities, a hemagglutinin epitope was inserted into the extracellular loop between transmembrane domains S3 and S4. Surface expression and channel function were measured by immunofluorescence, cell surface immunoprecipitation, and whole-cell patch clamp techniques. Although deletion of the last 14 amino acids of hIK1 (L414STOP) had no effect on plasma membrane expression and function, deletion of the last 26 amino acids (K402STOP) resulted in a complete loss of membrane expression. Mutation of the leucine heptad repeat ending at Leu(406) (L399A/L406A) completely abrogated membrane localization. Additional mutations within the heptad repeat (L385A/L392A, L392A/L406A) or of the a positions (I396A/L403A) resulted in a near-complete loss of membrane-localized channel. In contrast, mutating individual leucines did not compromise channel trafficking or function. Both membrane localization and function of L399A/L406A could be partially restored by incubation at 27 degrees C. Co-immunoprecipitation studies demonstrated that leucine zipper mutations do not compromise multimer formation. In contrast, we demonstrated that the leucine zipper region of hIK1 is capable of co-assembly and that this is dependent upon an intact leucine zipper. Finally, this leucine zipper is conserved in another member of the gene family, SK3. However, mutation of the leucine zipper in SK3 had no effect on plasma membrane localization or function. In conclusion, we demonstrate that the C-terminal leucine zipper is critical to facilitate correct folding and plasma membrane trafficking of hIK1, whereas this function is not conserved in other gene family members.     

The calcium-binding protein p54/NEFA is a novel luminal resident of medial Golgi cisternae that traffics independently of mannosidase II

A new Golgi resident, p54, has been demonstrated in several eukaryotic species and in multiple organs. Based on Triton X-114 partition, carbonate extraction and trypsin protection assays, p54 behaved as an extrinsic membrane protein, facing the luminal compartment. p54 was purified by two-dimensional electrophoresis and identified by matrix-assisted laser desorption ionization/time-of-flight (MALDI-TOF) mass spectrometry as NEFA, a calcium-binding protein (Barnikol-Watanabe et al., 1994, Biol. Chem. Hoppe Seyler, 375, 497-512). By immunofluorescence, p54/NEFA essentially colocalized with the medial Golgi marker mannosidase II, and did not overlap with the cis-Golgi marker p58, nor with the trans-Golgi network (TGN) marker TGN38. By immuno-electron microscopy, p54/NEFA localized in the medial cisternae and in Golgi-associated vesicles. p54/NEFA remained associated with mannosidase II despite Golgi disruption by nocodazole, caffeine, or, to some extent, potassium depletion (a new procedure to induce Golgi disassembly), but the two markers rapidly dissociated upon brefeldin A treatment and at metaphase, and reassociated upon drug removal and at the end of anaphase. Since p54/NEFA is a peripheral luminal membrane constituent, its distinct trafficking from the transmembrane marker mannosidase II suggests a novel Golgi retention mechanism, by strong association of this soluble protein with another integral transmembrane resident.     

Leucine zipper motif of chicken histone acetyltransferase-1 is essential for in vivo and in vitro interactions with the p48 subunit of chicken chromatin assembly factor-1

We cloned cDNA encoding chicken cytoplasmic histone acetyltransferase-1, chHAT-1, comprising 408 amino acids including a putative initiation Met. It exhibits 80.4% identity to the human homolog and possesses a typical leucine zipper motif. The glutathione S-transferase (GST) pull-down assay, involving truncated and missense mutants of the chicken chromatin assembly factor-1 (chCAF-1)p48, revealed not only that a region (comprising amino acids 376–405 of chCAF-1p48 and containing the seventh WD dipeptide motif) binds to chHAT-1 in vitro, but also that mutation of the motif has no influence on the in vitro interaction. The GST pull-down assay, involving truncated and missense chHAT-1 mutants, established that a region, comprising amino acids 380–408 of chHAT-1 and containing the leucine zipper motif, is required for its in vitro interaction with chCAF-1p48. In addition, mutation of each of four Leu residues in the leucine zipper motif prevents the in vitro interaction. The yeast two-hybrid assay revealed that all four Leu residues within the leucine zipper motif of chHAT-1 are necessary for its in vivo interaction with chCAF-1p48. These results indicate not only that the proper leucine zipper motif of chHAT-1 is essential for its interaction with chCAF-1p48, but also that the propeller structure of chCAF-1p48 expected to act as a platform for protein–protein interactions may not be necessary for this interaction of chHAT-1.     

Conserved leucines in N-terminal heptad repeat HR1 of envelope fusion protein F of group II nucleopolyhedroviruses are important for correct processing and essential for fusogenicity

The heptad repeat (HR), a conserved structural motif of class I viral fusion proteins, is responsible for the formation of a six-helix bundle structure during the envelope fusion process. The insect baculovirus F protein is a newly found budded virus envelope fusion protein which possesses common features to class I fusion proteins, such as proteolytic cleavage and the presence of an N-terminal open fusion peptide and multiple HR domains on the transmembrane subunit F(1). Similar to many vertebrate viral fusion proteins, a conserved leucine zipper motif is predicted in this HR region proximal to the fusion peptide in baculovirus F proteins. To facilitate our understanding of the functional role of this leucine zipper-like HR1 domain in baculovirus F protein synthesis, processing, and viral infectivity, key leucine residues (Leu209, Leu216, and Leu223) were replaced by alanine (A) or arginine (R), respectively. By using Autographa californica multicapsid nucleopolyhedrovirus (AcMNPV) as a pseudotype expression system, we demonstrated that all mutant F proteins incorporated into budded virus, indicating that leucine substitutions did not affect intercellular trafficking of F. Furin-like protease cleavage was not affected by any of the leucine substitutions; however, the disulfide bridging and N-linked glycosylation patterns were partly altered. Single substitutions in HR1 showed that the three leucine residues were critical for F fusogenicity and the rescue of AcMNPV infectivity. Our results support the view that the leucine zipper-like HR1 domain is important to safeguard the proper folding, glycosylation, and fusogenicity of baculovirus F proteins.     

Leucine zipper-mediated homodimerization of the adaptor protein c-Cbl. A role in c-Cbl's tyrosine phosphorylation and its association with epidermal growth factor receptor.

The 120-kDa proto-oncogenic protein c-Cbl is a multidomain adaptor protein that is phosphorylated in response to the stimulation of a broad range of cell surface receptors and participates in the assembly of signaling complexes that are formed as a result of the activation of various signal transduction pathways. Several structural features of c-Cbl, including the phosphotyrosine-binding domain, proline-rich domain, and motifs containing phosphotyrosine and phosphoserine residues, mediate the association of c-Cbl with other components of these complexes. In addition to those domains that have been demonstrated to play a role in the binding of c-Cbl to other signaling molecules, c-Cbl also contains a RING finger motif and a putative leucine zipper. In this study, we demonstrate that the previously identified putative leucine zipper mediates the formation of Cbl homodimers. Using the yeast two-hybrid system, we show that deletion of the leucine zipper domain is sufficient to abolish Cbl homodimerization, while Cbl mutants carrying extensive N-terminal truncations retain the ability to dimerize with the full-length Cbl. The requirement of the leucine zipper for the homodimerization of Cbl was confirmed by in vitro binding assays, using deletion variants of the C-terminal half of Cbl with and without the leucine zipper domain, and in cells using Myc and green fluorescent protein (GFP) N-terminal-tagged Cbl variants. In cells, the deletion of the leucine zipper caused a decrease in both the tyrosine phosphorylation of Cbl and its association with the epidermal growth factor receptor following stimulation with epidermal growth factor, thus demonstrating a role for the leucine zipper in c-Cbl's signaling functions. Thus, the leucine zipper domain enables c-Cbl to homodimerize, and homodimerization influences Cbl's signaling function, modulating the activity of Cbl itself and/or affecting Cbl's associations with other signaling proteins in the cell.     

The transmembrane domain of murine alpha-mannosidase IB is a major determinant of Golgi localization

Murine alpha1,2-mannosidase IB is a type II transmembrane protein localized to the Golgi apparatus where it is involved in the biogenesis of complex and hybrid N-glycans. This enzyme consists of a cytoplasmic tail, a transmembrane domain followed by a "stem" region and a large C-terminal catalytic domain. To analyze the determinants of targeting, we constructed various deletion mutants of murine alpha1,2-mannosidase IB as well as alpha1,2-mannosidase IB/yeast alpha1,2-mannosidase and alpha1,2-mannosidase IB/GFP chimeras and localized these proteins by fluorescence microscopy, when expressed transiently in COS7 cells. Replacing the catalytic domain of alpha1,2-mannosidase IB with that of the homologous yeast alpha1,2-mannosidase and deleting the "stem" region in this chimera had no effect on Golgi targeting, but caused increased cell surface localization. The N-terminal tagged protein lacking a catalytic domain was also localized to the Golgi. In the latter case, when the stem region was partially or completely removed, the protein was found in both the ER and the Golgi. A chimera consisting of the alpha1,2-mannosidase IB N-terminal region (cytoplasmic and transmembrane domains plus 10 amino acids of the "stem" region) and GFP was localized mainly to the Golgi. Deletion of 30 out of 35 amino acids in the cytoplasmic tail had no effect on Golgi localization. A GFP chimera lacking the entire cytoplasmic tail was found in both the ER and the Golgi. These results indicate that the transmembrane domain of alpha1,2-mannosidase IB is a major determinant of Golgi localization.     

Expression and function of the leucine zipper protein Par-4 in apoptosis.

The prostate apoptosis response-4 (par-4) gene was identified by differential screening for genes that are upregulated when prostate cancer cells are induced to undergo apoptosis. The par-4 gene is induced by apoptotic signals but not by growth-arresting, necrotic, or growth-stimulatory signals. The deduced amino acid sequence of par-4 predicts a protein with a leucine zipper domain at its carboxy terminus. We have recently shown that the Par-4 protein binds, via its leucine zipper domain, to the zinc finger domain of Wilms' tumor protein WT1 (R. W. Johnstone et al., Mol. Cell. Biol. 16:6945-6956, 1996). In experiments aimed at determining the functional role of par-4 in apoptosis, an antisense par-4 oligomer abrogated par-4 expression and activator-driven apoptosis in rat prostate cancer cell line AT-3, suggesting that par-4 is required for apoptosis in these cells. Consistent with a functional role for par-4 in apoptosis, ectopic overexpression of par-4 in prostate cancer cell line PC-3 and melanoma cell line A375-C6 conferred supersensitivity to apoptotic stimuli. Transfection studies with deletion mutants of Par-4 revealed that full-length Par-4, but not mutants that lacked the leucine zipper domain of Par-4, conferred enhanced sensitivity to apoptotic stimuli. Most importantly, ectopic coexpression of the leucine zipper domain of Par-4 inhibited the ability of Par-4 to enhance apoptosis. Finally, ectopic expression of WT1 attenuated apoptosis, and coexpression of Par-4 but not a leucine zipperless mutant of Par-4 rescued the cells from the antiapoptotic effect of WT1. These findings suggest that the leucine zipper domain is required for the Par-4 protein to function in apoptosis.     

NUCB1, the Drosophila melanogaster homolog of the mammalian EF-hand proteins NEFA and nucleobindin.

Mammalian NEFA and nucleobindin are calcium-binding proteins containing a signal peptide, two EF-hand motifs, acidic and basic regions and a leucine-zipper motif. Although they have been discussed to be involved in autoimmunity, apoptosis and calcium homeostasis in the Golgi apparatus and bone matrix, their exact role remains unknown. Here we report the cloning of their Drosophila homolog, nucb1, as well as the analysis of its expression pattern during embryogenesis and the subcellular localization of the NUCB1 protein. The nucb1 mRNA and the NUCB1 protein were found to be expressed maternally and zygotically, and they accumulate ubiquitously at low levels during all embryonic stages due to a maternal component. From stage 11 onward, high levels of zygotic expression can be detected specifically in the salivary glands and their placodes. In contrast to the known mammalian family members, the NUCB1 protein localizes in a subpattern of cytoplasmic substructures, probably the Golgi apparatus.