Late Golgi Research Articles

Activation and Functional Characterization of the Mosaic Receptor SorLA/LR11

We previously isolated and sequenced the approximately 250-kDa type 1 receptor sorLA/LR11, a mosaic protein with elements characterizing the Vps10p domain receptor family as well as the low density lipoprotein receptor family. The N terminus of the Vps10p domain comprises a consensus sequence for cleavage by furin ((50)RRKR(53)) that precedes a truncation found in sorLA isolated from human brain. Here we show that sorLA, like sortilin-1/neurotensin receptor-3, whose lumenal domain consists of a Vps10p domain only, is synthesized as a proreceptor that is cleaved by furin in late Golgi compartments. We show that the truncation conditions the Vps10p domain for propeptide inhibitable binding of neuropeptides and the receptor-associated protein. We further demonstrate that avid binding of the receptor-associated protein, apolipoprotein E, and lipoprotein lipase not inhibited by propeptide occurs to sites located in other lumenal domains. In transfected cells, about 10% of full-length sorLA were expressed on the cell surface capable of mediating endocytosis. However, the major pool of receptors was found in late Golgi compartments, suggesting possible interaction with newly synthesized ligands. The results show that sorLA, following activation by truncation, binds multiple ligands and may mediate both endocytosis and sorting.

Influence of KDEL on the fate of trimeric or assembly-defective phaseolin: selective use of an alternative route to vacuoles.

The tetrapeptide KDEL is commonly found at the C terminus of soluble proteins of the endoplasmic reticulum (ER), and it contributes to their localization by interacting with a receptor that recycles between the Golgi complex and the ER. We investigated the effects of the addition of KDEL to phaseolin, a protein normally delivered from the ER to storage vacuoles via the Golgi complex. We show that KDEL prevents acquisition of trans-Golgi-specific glycan modifications and causes interactions with the chaperone BiP that are distinct from the ones between BiP and defective proteins. KDEL markedly increases the stability of phaseolin, but a small proportion of phaseolin-KDEL slowly reaches the vacuole without undergoing Golgi-mediated glycan modifications, in a process that can be inhibited by brefeldin A but not monensin. Our results indicate that KDEL can operate with high efficiency before proteins can reach the late Golgi cisternae but allows or promotes delivery to vacuoles via an alternative mechanism. However, addition of KDEL does not alter the destiny of an assembly-defective form of phaseolin, suggesting that the plant ER quality control mechanism is dominant over KDEL effects.

A role for actin, Cdc1p, and Myo2p in the inheritance of late Golgi elements in Saccharomyces cerevisiae.

In Saccharomyces cerevisiae, Golgi elements are present in the bud very early in the cell cycle. We have analyzed this Golgi inheritance process using fluorescence microscopy and genetics. In rapidly growing cells, late Golgi elements show an actin-dependent concentration at sites of polarized growth. Late Golgi elements are apparently transported into the bud along actin cables and are also retained in the bud by a mechanism that may involve actin. A visual screen for mutants defective in the inheritance of late Golgi elements yielded multiple alleles of CDC1. Mutations in CDC1 severely depolarize the actin cytoskeleton, and these mutations prevent late Golgi elements from being retained in the bud. The efficient localization of late Golgi elements to the bud requires the type V myosin Myo2p, further suggesting that actin plays a role in Golgi inheritance. Surprisingly, early and late Golgi elements are inherited by different pathways, with early Golgi elements localizing to the bud in a Cdc1p- and Myo2p-independent manner. We propose that early Golgi elements arise from ER membranes that are present in the bud. These two pathways of Golgi inheritance in S. cerevisiae resemble Golgi inheritance pathways in vertebrate cells.

The sodium/proton exchanger Nhx1p is required for endosomal protein trafficking in the yeast Saccharomyces cerevisiae.

We show that the vacuolar protein sorting gene VPS44 is identical to NHX1, a gene that encodes a sodium/proton exchanger. The Saccharomyces cerevisiae protein Nhx1p shows high homology to mammalian sodium/proton exchangers of the NHE family. Nhx1p is thought to transport sodium ions into the prevacuole compartment in exchange for protons. Pulse-chase experiments show that approximately 35% of the newly synthesized soluble vacuolar protein carboxypeptidase Y is missorted in nhx1 delta cells, and is secreted from the cell. nhx1 delta cells accumulate late Golgi, prevacuole, and lysosome markers in an aberrant structure next to the vacuole, and late Golgi proteins are proteolytically cleaved more rapidly than in wild-type cells. Our results show that efficient transport out of the prevacuolar compartment requires Nhx1p, and that nhx1 delta cells exhibit phenotypes characteristic of the "class E" group of vps mutants. In addition, we show that Nhx1p is required for protein trafficking even in the absence of the vacuolar ATPase. Our analysis of Nhx1p provides the first evidence that a sodium/proton exchange protein is important for correct protein sorting, and that intraorganellar ion balance may be important for endosomal function in yeast.

Characterization of CD1e, a Third Type of CD1 Molecule Expressed in Dendritic Cells

Dendritic cells express several alternatively spliced CD1e mRNAs. These molecules encode proteins characterized by the presence of either one, two, or three alpha domains and either a 51- or 63-amino acid cytoplasmic domain. Moreover, mRNAs encoding isoforms lacking the transmembrane domain are observed. Several of these CD1e isoforms were expressed in transfected cells, and two of them, with three alpha domains, displayed a particular processing pathway. These latter isoforms slowly leave the endoplasmic reticulum due to the presence of atypical dilysine motifs in the cytoplasmic tail. These molecules are associated with the beta(2)-microglobulin and accumulate in late Golgi and late endosomal compartments. In the latter compartments, they are cleaved into soluble forms that appear to be stable. In dendritic cells, these isoforms are mainly located in the Golgi apparatus, and upon maturation they are redistributed to late endosomal compartments. This work demonstrates the existence of CD1e molecules. As compared with other CD1 molecules, CD1e displays fundamentally different properties and therefore may represent a third type of CD1 molecules.

Evidence supporting a late Golgi location for lactosylceramide to ganglioside GM3 conversion.

Ganglioside GM2 synthase and other enzymes required for complex ganglioside synthesis were localized recently to the trans Golgi network (TGN). However, there are conflicting reports as to the location of GM3 synthase; originally this enzyme was detected in the early Golgi of rat liver but a recent report localized it to the late Golgi. We have used chimeric forms of ganglioside GM2 synthase to determine if the location of lactosylceramide (LacCer) to GM3 conversion in Chinese hamster ovary (CHO) cells was the early or late Golgi. Our approach tested whether GM3 could be utilized as a substrate by GM2 synthase chimeras which were targeted to compartments earlier than the trans Golgi, i.e., GM3 produced in the cis Golgi should be utilized by GM2 synthase located anywhere in the Golgi whereas GM3 produced in the trans Golgi should only be used by GM2 synthase located in the trans Golgi or TGN. Comparison of cell lines stably expressing these chimeras revealed that the in vivo functional activity of GM2 synthase decreased progressively as the enzyme was targeted to earlier compartments; specifically, the percentage of GM3 converted to GM2 was 83-86% for wild type enzyme, 70% for the medial Golgi targeted enzyme, 13% for the ER and cis Golgi targeted enzyme, and only 1.7% for the ER targeted enzyme. Thus, these data are consistent with a late Golgi location for LacCer to GM3 conversion in these cells.

The Low pH in Trans-Golgi Triggers Autocatalytic Cleavage of Pre-α-inhibitor Heavy Chain Precursor

Pre-alpha-inhibitor is a plasma protein whose physiological function is still unknown, but in vitro studies suggest that it might be involved in inflammatory reactions. Pre-alpha-inhibitor consists of a 25- and a 75-kDa polypeptide: bikunin and heavy chain 3 (H3), respectively. H3 is synthesized with a 30-kDa C-terminal extension, which is released in the Golgi complex through cleavage between an Asp and a Pro residue. We now provide evidence that this cleavage is triggered by the low pH in the late Golgi and occurs through an intramolecular process. First, incubation in vitro of the H3 precursor (proH3) at pH 6.0 or lower results in rapid cleavage of the protein. Second, the rate of the cleavage reaction does not depend on the concentration of proH3 and is not affected by the presence of various protease inhibitors. Third, raising the pH in organelles of cells producing proH3 abolishes cleavage during secretion. The amino acid residues near the cleavage site of proH3 differ from those of previously described self-cleaving proteins, indicating that the mechanisms of scission are different.

Ric1p and Rgp1p form a complex that catalyses nucleotide exchange on Ypt6p.

Cells lacking the GTPase Ypt6p have defects in intracellular traffic and are temperature sensitive. Their growth is severely impaired by additional mutation of IMH1, which encodes a non-essential Golgi-associated coiled-coil protein. A screen for mutants that, like ypt6, specifically impair the growth of imh1 cells led to the identification of RIC1. Ric1p forms a tight complex with a previously uncharacterized protein, Rgp1p. The Ric1p-Rgp1p complex binds Ypt6p in a nucleotide-dependent manner, and purified Ric1p-Rgp1 stimulates guanine nucleotide exchange on Ypt6p in vitro. Deletion of RIC1 or RGP1, like that of YPT6, blocks the recycling of the exocytic SNARE Snc1p from early endosomes to the Golgi and causes temperature-sensitive growth, but this defect can be relieved by overexpression of YPT6. Ric1p largely colocalizes with the late Golgi marker Sec7p. Ypt6p shows a similar distribution, but this is altered when RIC1 or RGP1 is mutated. We infer that the Ric1p-Rgp1p complex serves to activate Ypt6p on Golgi membranes by nucleotide exchange, and that this is required for efficient fusion of endosome-derived vesicles with the Golgi.

Pep3p/Pep5p Complex: A Putative Docking Factor at Multiple Steps of Vesicular Transport to the Vacuole of Saccharomyces cerevisiae

Pep3p and Pep5p are known to be necessary for trafficking of hydrolase precursors to the vacuole and for vacuolar biogenesis. These proteins are present in a hetero-oligomeric complex that mediates transport at the vacuolar membrane. PEP5 interacts genetically with VPS8, implicating Pep5p in the earlier Golgi to endosome step and/or in recycling from the endosome to the Golgi. To understand further the cellular roles of Pep3p and Pep5p, we isolated and characterized a set of pep3 conditional mutants. Characterization of mutants revealed that pep3(ts) mutants are defective in the endosomal and nonendosomal Golgi to vacuole transport pathways, in the cytoplasm to vacuole targeting pathway, in recycling from the endosome back to the late Golgi, and in endocytosis. PEP3 interacts genetically with two members of the endosomal SNARE complex, PEP12 (t-SNARE) and PEP7 (homologue of mammalian EEA1); Pep3p and Pep5p associate physically with Pep7p as revealed by two-hybrid analysis. Our results suggest that a core Pep3p/Pep5p complex promotes vesicular docking/fusion reactions in conjunction with SNARE proteins at multiple steps in transport routes to the vacuole. We propose that this complex may be responsible for tethering transport vesicles on target membranes.

Endocytotic uptake and retrograde transport of a virally encoded killer toxin in yeast.

We demonstrate that a virally encoded yeast 'killer' toxin is entering its eukaryotic target cell by endocytosis, subsequently travelling the yeast secretory pathway in reverse to exhibit its lethal effect. The K28 killer toxin is a secreted alpha/beta heterodimer that kills sensitive yeasts in a receptor-mediated fashion by blocking DNA synthesis in the nucleus. In vivo processing of the toxin precursor results in a protein whose beta-C-terminus carries the endoplasmic reticulum (ER) retention signal HDEL, which, as we show here, is essential for retrograde toxin transport. Yeast end3/4 mutants as well as cells lacking the HDEL receptor (Deltaerd2) or mutants defective in Golgi-to-ER protein recycling (erd1) are toxin resistant because the toxin can no longer enter and/or retrograde pass the cell. Site-directed mutagenesis further indicated that the toxin's beta-HDEL motif ensures retrograde transport, although in a toxin-secreting yeast the beta-C-terminus is initially masked by an R residue (beta-HDELR) until Kex1p cleavage uncovers the toxin's targeting signal in a late Golgi compartment. Prevention of Kex1p processing results in high-level secretion of a biologically inactive protein incapable of re-entering the secretory pathway. Finally, we present evidence that ER-to-cytosol toxin export is mediated by the Sec61p translocon and requires functional copies of the lumenal ER chaperones Kar2p and Cne1p.

Effect of the truncation of the C-terminal region of Kex2 endoprotease on processing of the recombinant Rhizopus oryzae lipase precursor in the co-expression system in yeast

The Saccharomyces cerevisiae Kex2p is a membrane-bound endoprotease and cleaves at a C-terminal site of Lys–Arg of the α-factor precursor in late Golgi compartment. The complete cleavage of a secreted recombinant Rhizopus oryzae lipase (ROL) precursor (rProROL) having an internal Lys–Arg has been examined by the co-expression of the KEX2 gene. The cleavage of rProROL was not sufficient. The co-expression system of the truncated gene encoding the soluble form of Kex2p (sKex2p) lacking C-terminal 201 amino acids containing the transmembrane domain (TMD) was constructed. sKex2p co-expressed was detected in the culture medium and the carboxyl site of Lys(-30)–Arg(-29) in the prosequence of rProROL was completely cleaved. These results revealed that on the process of secretion, including in the Golgi apparatus and secretion vesicles, attack of sKex2p to the produced protein occurred more efficiently than that of Kex2p. The cells having the co-expression system of the gene encoding sKex2p will be useful for production of foreign proteins designed to cleave the internal Lys–Arg site for their activation or for the application of the prepro-α-factor leader sequence.

Luv1p/Rki1p/Tcs3p/Vps54p, a yeast protein that localizes to the late Golgi and early endosome, is required for normal vacuolar morphology.

We have characterized LUV1/RKI1/TCS3/VPS54, a novel yeast gene required to maintain normal vacuolar morphology. The luv1 mutant was identified in a genetic screen for mutants requiring the phosphatase calcineurin for vegetative growth. luv1 mutants lack a morphologically intact vacuole and instead accumulate small vesicles that are acidified and contain the vacuolar proteins alkaline phosphatase and carboxypeptidase Y and the vacuolar membrane H(+)-ATPase. Endocytosis appears qualitatively normal in luv1 mutants, but some portion (28%) of carboxypeptidase Y is secreted. luv1 mutants are sensitive to several ions (Zn(2+), Mn(2+), and Cd(2+)) and to pH extremes. These mutants are also sensitive to hygromycin B, caffeine, and FK506, a specific inhibitor of calcineurin. Some vacuolar protein-sorting mutants display similar drug and ion sensitivities, including sensitivity to FK506. Luv1p sediments at 100,000 x g and can be solubilized by salt or carbonate, indicating that it is a peripheral membrane protein. A Green Fluorescent Protein-Luv1 fusion protein colocalizes with the dye FM 4-64 at the endosome, and hemagglutinin-tagged Luv1p colocalizes with the trans-Golgi network/endosomal protease Kex2p. Computer analysis predicts a short coiled-coil domain in Luv1p. We propose that this protein maintains traffic through or the integrity of the early endosome and that this function is required for proper vacuolar morphology.

Uncoupling of Myelin Assembly and Schwann Cell Differentiation by Transgenic Overexpression of Peripheral Myelin Protein 22

We have generated previously transgenic rats that overexpress peripheral myelin protein 22 (PMP22) in Schwann cells. In the nerves of these animals, Schwann cells have segregated with axons to the normal 1:1 ratio but remain arrested at the promyelinating stage, apparently unable to elaborate myelin sheaths. We have examined gene expression of these dysmyelinating Schwann cells using semiquantitative reverse transcription-PCR and immunofluorescence analysis. Unexpectedly, Schwann cell differentiation appears to proceed normally at the molecular level when monitored by the expression of mRNAs encoding major structural proteins of myelin. Furthermore, an aberrant coexpression of early and late Schwann cell markers was observed. PMP22 itself acquires complex glycosylation, suggesting that trafficking of the myelin protein through the endoplasmic reticulum is not significantly impaired. We suggest that PMP22, when overexpressed, accumulates in a late Golgi-cell membrane compartment and uncouples myelin assembly from the underlying program of Schwann cell differentiation.

Transport and rearrangement of the intra-acrosomal protein acrin1 (MN7) during spermiogenesis in the guinea pig testis.

We have recently shown that a 90-kDa glycoprotein, acrin1 (MN7), is exclusively localized in the dorsal region of the acrosomal apical segment of mature guinea pig sperm, and that its location changes during epididymal maturation. The present study examined the process of transport and organization of this protein in the acrosome during spermatogenesis in the guinea pig testis. Immunoperoxidase electron microscopy showed stage-specific localization of acrin1 within the developing acrosome, as follows: acrin1 first appeared in the proacrosomic vesicles of the early Golgi phase spermatids, and it was then localized in the electron-lucent matrix region of the acrosomic vesicles of the late Golgi phase spermatids. During the cap phase, acrin1 was abundant in the electron-lucent matrix of the acrosomal apical segment and in the head-cap region (principal segment). acrin1 became more restricted to the peripheral region of the electron-lucent matrix of acrosome phase spermatids and it was localized in the electron-lucent dorsal matrix region of maturation phase spermatids. In the final step of spermiogenesis, acrin1 disappeared from the equatorial and principal segments, and it was finally confined to the dorsal matrix region of the acrosomal apical segment. In addition, Western blot analysis showed that acrin1 of testes and epididymal sperm was of the identical size, indicating that acrin1 is not proteolytically modified during epididymal sperm maturation. These results indicate that acrosome morphogenesis is closely associated with the rearrangement of acrosomal proteins.

Cloning and characterization of ADAM28: evidence for autocatalytic pro-domain removal and for cell surface localization of mature ADAM28

The metalloprotease disintegrins are a family of membrane-anchored glycoproteins with diverse functions in fertilization, myoblast fusion, neurogenesis and protein ectodomain shedding. Here we report a cDNA sequence, encoding a metalloprotease disintegrin, termed ADAM28 (‘a disintegrin and metalloprotease 28’), which was cloned from mouse lung. From protein sequence comparisons, ADAM28 is more closely related to snake venom metalloproteases (SVMPs) than to other ADAMs, and hence may cleave similar substrates to SVMPs, perhaps including components of the extracellular matrix. Northern blot analysis of selected mouse tissues revealed that ADAM28 is expressed highly and in alternatively spliced forms in the epididymis, suggesting a possible role in sperm maturation, and at lower levels in lung. The intracellular maturation of ADAM28 expressed in COS-7 cells resembles that of other ADAMs, in that ADAM28 is made as a precursor and processed to a mature form in a late Golgi compartment of the secretory pathway. Most or all of the mature, and thus presumably catalytically active, form of ADAM28 in COS-7 cells is accessible to cell surface trypsinization, suggesting that ADAM28 functions mainly on the cell surface. A mutation converting the catalytic-site glutamate residue into alanine abolishes pro-domain removal, even though this mutant form of ADAM28 can be transported to the cell surface in a manner similar to the wild-type protein. This suggests that pro-domain removal and maturation of ADAM28 may be, at least in part, autocatalytic. This is in contrast with several other ADAMs, for which furin-like proprotein convertases are involved in pro-domain removal, and in which a glutamate-to-alanine mutation in the catalytic site does not alter pro-domain removal.

Cloning and characterization of ADAM28: evidence for autocatalytic pro-domain removal and for cell surface localization of mature ADAM28

The metalloprotease disintegrins are a family of membrane-anchored glycoproteins with diverse functions in fertilization, myoblast fusion, neurogenesis and protein ectodomain shedding. Here we report a cDNA sequence, encoding a metalloprotease disintegrin, termed ADAM28 ('a disintegrin and metalloprotease 28'), which was cloned from mouse lung. From protein sequence comparisons, ADAM28 is more closely related to snake venom metalloproteases (SVMPs) than to other ADAMs, and hence may cleave similar substrates to SVMPs, perhaps including components of the extracellular matrix. Northern blot analysis of selected mouse tissues revealed that ADAM28 is expressed highly and in alternatively spliced forms in the epididymis, suggesting a possible role in sperm maturation, and at lower levels in lung. The intracellular maturation of ADAM28 expressed in COS-7 cells resembles that of other ADAMs, in that ADAM28 is made as a precursor and processed to a mature form in a late Golgi compartment of the secretory pathway. Most or all of the mature, and thus presumably catalytically active, form of ADAM28 in COS-7 cells is accessible to cell surface trypsinization, suggesting that ADAM28 functions mainly on the cell surface. A mutation converting the catalytic-site glutamate residue into alanine abolishes pro-domain removal, even though this mutant form of ADAM28 can be transported to the cell surface in a manner similar to the wild-type protein. This suggests that pro-domain removal and maturation of ADAM28 may be, at least in part, autocatalytic. This is in contrast with several other ADAMs, for which furin-like proprotein convertases are involved in pro-domain removal, and in which a glutamate-to-alanine mutation in the catalytic site does not alter pro-domain removal.

Formation of Insoluble Oligomers Correlates with ST6Gal I Stable Localization in the Golgi

The ST6Gal I is a sialyltransferase that functions in the late Golgi to modify the N-linked oligosaccharides of glycoproteins. The ST6Gal I is expressed as two isoforms with a single amino acid difference in their catalytic domains. The STcys isoform is stably retained in the cell and is predominantly found in the Golgi, whereas the STtyr isoform is only transiently localized in the Golgi and is cleaved and secreted from a post-Golgi compartment. These two ST6Gal I isoforms were used to explore the role of the bilayer thickness mechanism and oligomerization in Golgi localization. Analysis of STcys and STtyr proteins with longer transmembrane regions suggested that the bilayer thickness mechanism is not the predominant mechanism used for ST6Gal I Golgi localization. In contrast, the formation and quantity of Triton X-100-insoluble oligomers was correlated with the stable or transient localization of the ST6Gal I isoforms in the Golgi. Nearly 100% of the STcys and only 13% of the STtyr were found as Triton-insoluble oligomers when Golgi membranes of COS-1 cells expressing these proteins were solubilized at pH 6.3, the pH of the late Golgi. In contrast, both proteins were found in the soluble fraction when these membranes were solubilized at pH 8.0. Analysis of other mutants suggested that a conformational change in the catalytic domain rather than increased disulfide bond-based cross-linking is the basis for the increased ability of STcys protein to form oligomers and the stable localization of STcys protein in the Golgi.

Flipping for coats in late-Golgi membranes?

Genetic screens override investigators’ predilections and permit the internal logic of cells to point the way. Following this strategy, Chen et al.1 uncovered a possible role for lipid asymmetry in clathrin function in the late Golgi. Their starting point was the small GTP-binding protein ADP-ribosylation factor (ARF), which mediates assembly of COPI coats in the cis-Golgi and clathrin coats in the trans-Golgi.

Molecular identification, tissue distribution and subcellular localization of mST3GalV/GM3 synthase

A molecular screen for a mouse homologue of a Drosophila carbohydrate binding protein, called Gliolectin, yielded a cDNA encoding mST3GalV/GM3 synthase (CMP-NeuAc: lactosylceramide alpha2, 3-sialyltransferase). By in situ hybridization and immunohistochemistry, mST3GalV exhibits differential expression in neural and non-neural tissues. Although expressed by all neurons in the central nervous system, neuronal populations that contribute their axons to myelinated efferent projections, such as cerebellar Purkinje cells and spinal motorneurons, demonstrate the highest ST3GalV expression. When stained with anti-mST3GalV antiserum (designated CS2), subpopulations of neurons display an elaborate Golgi apparatus, frequently extending into one or more dendritic processes. The extended spatial distribution of the neuronal Golgi apparatus, particularly in spinal motorneurons, allowed the confocal immunohistochemical colocalization of mST3GalV with markers for medial/trans-Golgi but not the cis-Golgi or trans-Golgi network, consistent with previous observations suggesting that ganglioside glycosyltransferases are enriched in late Golgi compartments. Among non-neural tissues, liver and testes demonstrate cell-type specific CS2 staining. In liver, endothelial cells lining a ring of sinusoids, concentric with the central vein, express mST3GalV. Kupffer cells are also stained with CS2 antiserum but hepatocyte expression is undetectable. In the seminiferous tubules of the testes, ST3GalV is found in somatic (Leydig, Sertoli) and early germline cells (spermatogonia and primary spermatocytes); the epididymal epithelium exhibits intense ST3GalV expression. Since GM3 is a precursor for the synthesis of a- and b-series gangliosides, the range of mST3GalV/GM3 synthase expression among various cell populations indicates that certain cell types possess greater reliance on ganglioside function than others.

Medial Golgi but Not Late Golgi Glycosyltransferases Exist as High Molecular Weight Complexes: ROLE OF LUMINAL DOMAIN IN COMPLEX FORMATION AND LOCALIZATION

To investigate the organization of Golgi glycosyltransferases and their mechanism of localization, we have compared the properties of a number of medial and late acting Golgi enzymes. The medial Golgi enzymes, N-acetylglucosaminyltransferase I and II (GnTI and GnTII) required high salt for solubilization and migrated as high molecular weight complexes on sucrose density gradients. In contrast, the late acting Golgi enzymes, beta1,4-galactosyltransferase and alpha1, 2-fucosyltransferase, were readily solubilized in low salt and migrated as monomers/dimers by sucrose density gradient centrifugation. Analysis of membrane-bound GnTI chimeras indicates that the formation of high molecular weight complexes does not require the transmembrane domain and cytoplasmic tail sequences of GnTI. Furthermore, a soluble form of GnTI, containing the stem region and catalytic domain, accumulated in the Golgi prior to secretion, in contrast to beta1,4-galactosyltransferase. Soluble GnTI, which also associated with high molecular weight complexes, was comparable with membrane-bound GnTI in its ability to glycosylate newly synthesized glycoproteins in vivo. Mutation of charged residues within the stem region of GnTI, known to be important for "kin recognition", had no effect on the efficiency of Golgi localization, the inclusion into high molecular weight complexes, nor functional activity in vivo. The differences in behavior between the medial and late acting Golgi enzymes may contribute to their differential localization and their ability to glycosylate efficiently in the correct Golgi subcompartment.