tGolgin-1 (golgin-245, trans golgi p230) and golgin-97 are members of a family of peripheral membrane proteins of unknown function that localize to the trans Golgi network (TGN) through a conserved C-terminal GRIP domain. We have probed for GRIP protein function by assessing the consequences of overexpressing isolated GRIP domains. By semi-quantitative immunofluorescence microscopy we found that high level expression of epitope-tagged, GRIP domain-containing fragments of tGolgin-1 or golgin-97 specifically altered the characteristic pericentriolar distribution of TGN integral membrane and coat components. Concomitantly, vesicular transport from the TGN to the plasma membrane and furin-dependent cleavage of substrate proteins in the TGN were inhibited. Mutagenesis of a conserved tyrosine in the tGolgin-1 GRIP domain abolished these effects. GRIP domain overexpression had little effect on the distribution of most Golgi stack resident proteins and no effect on markers of other organelles. Electron microscopy analyses of GRIP domain-overexpressing cells revealed distended perinuclear vacuoles and a proliferation of multivesicular late endosomes to which the TGN resident protein TGN46 was largely mislocalized. These studies, the first to address the function of GRIP domain-containing proteins in higher eukaryotes, suggest that some or all of these proteins and/or their ligands function in maintaining the integrity of the TGN by regulating resident protein localization.
The trans Golgi network (TGN) is a series of interconnected tubules and vesicles at the trans face of the Golgi stack that functions in the processing and sorting of glycoproteins and glycolipids at the interface of the biosynthetic and endosomal pathways (Griffiths and Simons, 1986; Traub and Kornfeld, 1997). TGN structure is dynamic, subject to constant influx and efflux of membrane from and to both secretory and endosomal compartments. Such dynamics require efficient membrane recycling to maintain a constant steady state composition of lipids and proteins. Hence, TGN resident integral membrane proteins, including glycosyl modifying enzymes (Geuze and Morre, 1991), proprotein processing enzymes (Seidah and Chretien, 1997; Varlamov and Fricker, 1998), SNARE proteins (Lewis et al., 2000; Mallard et al., 2002; Siniossoglou and Pelham, 2001), and putative cargo binding proteins such as TGN38/TGN46 (Banting and Ponnambalam, 1997), maintain a steady state accumulation within the TGN by active retention and recycling (Bryant and Stevens, 1997; Ghosh et al., 1998; Mallet and Maxfield, 1999; Ponnambalam et al., 1994). The best characterized recycling pathways involve retrieval of resident proteins and glycolipids from endosomes (Molloy et al., 1999; Rohn et al., 2000). The molecular mechanisms that regulate both the efflux of membrane to the cell surface and the retrieval from endosomes are only beginning to be understood, and include sorting signals on cargo proteins and specific cytoplasmic components to effect cargo movement (Mallard et al., 2002; Rohn et al., 2000). Among these components there are likely to exist as yet unidentified tethering factors (Lowe et al., 1998; Pfeffer, 1999).
In an effort to elucidate effectors regulating TGN biogenesis, we have focused on a group of large peripheral membrane proteins of unknown function characterized by an extensive predicted coiled-coil structure and a conserved C-terminal GRIP (Golgin-97, RanBP2α, Imh1p and trans golgi p230) domain (Barr, 1999; Kjer-Nielsen et al., 1999a; Kjer-Nielsen et al., 1999b; Munro and Nichols, 1999). The GRIP domain confers localization to the TGN and associated vesicles for all four mammalian GRIP domain-containing proteins (GRIP proteins): tGolgin-1 [(Cowan et al., 2002) also known as trans golgi p230 (Erlich et al., 1996), golgin-245 (Fritzler et al., 1995) and 256 kDa golgin (H. P. Seelig, GenBank accession no. X82834)], golgin-97, GCC88 and GCC185 (Brown et al., 2001; Gleeson et al., 1996; Luke et al., 2003). The single yeast GRIP protein, Imh1p, similarly localizes to the late Golgi (Panic et al., 2003; Setty et al., 2003). No function has yet been assigned to any GRIP protein, but indirect evidence supports a role in vesicular traffic at the TGN. Several GRIP domains bind to the small Arf-like GTPase, Arl1 (Lu et al., 2001; Panic et al., 2003; Setty et al., 2003; Van Valkenburgh et al., 2001), which itself may regulate Golgi and/or TGN structure (Lu et al., 2001; Van Valkenburgh et al., 2001; Panic et al., 2003). tGolgin-1 is associated with a class of vesicles released from Golgi membrane fractions in an in vitro budding assay (Brown et al., 2001; Gleeson et al., 1996). Finally, IMH1 displays both multicopy suppression and synthetic lethality with mutations in genes encoding the yeast Rab6 homologue, YPT6, and its nucleotide exchange factor, RIC1 (Li and Warner, 1996; Siniossoglou et al., 2000; Tsukada and Gallwitz, 1996; Tsukada et al., 1999), which themselves are implicated in endosome to TGN recycling (Bensen et al., 2001; Siniossoglou and Pelham, 2001; Tsukada et al., 1999). Given these data, and their structural similarities to known tethering proteins (Pfeffer, 1999), we hypothesized that mammalian GRIP proteins function in regulating protein recycling from endosomes to the TGN.
Overexpression of isolated GRIP domains by transfection competitively displaces endogenous GRIP proteins from Golgi membranes (Kjer-Nielsen et al., 1999a; Kjer-Nielsen et al., 1999b), presumably by competition for a limiting GRIP domain binding site. Such displacement would be expected to interfere with the function of endogenous GRIP proteins. We show here that in cells in which GRIP domain-containing fragments from tGolgin-1 or golgin-97 are overexpressed, the structure, resident protein localization and function of the TGN are largely disrupted. The results suggest that GRIP proteins or their ligands function in the maintenance of TGN integrity, probably through regulating TGN membrane protein localization.
Materials and Methods
cDNA clones encoding mouse tGolgin-1 and plasmids encoding N-terminally HA11 (influenza hemagglutinin 11) epitope-tagged truncated forms of tGolgin-1 encoding the C-terminal 312, 186, 81 or 50 amino acids (C312, C186, C81, C50), and a point mutant of C312, C312(Y2187A) (in which the codon for the critical GRIP domain tyrosine residue was altered to that for alanine) in the mammalian expression vector pCDM8.1 (Bonifacino et al., 1990) are described elsewhere (Cowan et al., 2002). T7 epitope-tagged C312 in pCDM8.1 was prepared by two-step polymerase chain reaction (Higuchi et al., 1988) using primers encoding the epitope tag and a Kozak consensus start site. A fragment encoding the C-terminal 179 amino acids of golgin-97 (G97-C179) was amplified by reverse transcriptase coupled PCR (RT-PCR) from RNA isolated from a human melanoma, MNT-1, and subcloned into pCDM8.1-HA11. Sequences of all PCR-derived inserts and of junctions of subcloned fragments were verified by automated dideoxy sequencing. Details of the sequence of PCR primers used for plasmid construction are available upon request. The following constructs have been described previously as indicated: Tac in pCDM8.1 (Leonard et al., 1984); Tac chimeric proteins with the cytoplasmic domains of TGN38 [TGG (Humphrey et al., 1993)], furin [TTF (Voorhees et al., 1995)] or Lamp1 [TTL1 (Marks et al., 1996)], or the di-leucine-based sorting signal of CD3γ [Tac-DKQTLL (Letourneur and Klausner, 1992; Marks et al., 1996)] in pCDM8.1; Pmel17 in pCI (Berson et al., 2001); furin with a C-terminal HA11 epitope tag in pXS (Bosshart et al., 1994); and the ts045 variant of the vesicular stomatitis virus glycoprotein conjugated to enhanced green fluorescent protein [VSV-G-EGFP (Presley et al., 1997)].
Cell culture and transfections
HeLa cells were maintained in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum. Cells were transfected using calcium phosphate precipitation as described previously (Marks et al., 1996). For most experiments, cells grown on coverslips in six-well dishes were transfected with 7 μg of total DNA; in cases where moderate expression levels were desired, 50-100 ng of the desired expression construct were used. Except where noted, cells were analyzed 2-3 days following transfection. Stable transfectants of the CHO cell variant TRVb-1 expressing Tac chimeric proteins TGG and TTF (Ghosh et al., 1998; Mallet and Maxfield, 1999), provided by Dr F. R. Maxfield (Weil Medical College of Cornell University, New York, NY), were transiently transfected using Superfect (Qiagen, Valencia, CA) or GenePORTER (Gene Therapy Systems, San Diego, CA) according to the manufacturers' instructions.
Antibodies were to the following molecules (sources are given in parentheses). HA11 epitope (mAb 16D12, Covance Research Products, Richmond, CA; mAbs 12CA5 and 3F10, Roche Molecular Biochemicals, Indianapolis, IN); TGN46 [rabbit anti-TGN46 (Prescott et al., 1997); sheep anti-TGN46, Serotec, Oxford, UK)]; cation-independent mannose 6-phosphate receptor (mAb αMPR300, Affinity BioReagents, Golden, CO); Tac (Santini et al., 1998); golgin-97 (rb αgolgin-97 and mAb αgolgin-97, E. K. L. Chan, Scripps Res. Inst., La Jolla, CA); giantin (Linstedt and Hauri, 1993); Rab6 (C-19 rb αRab6, Santa Cruz Biotechnology, Santa Cruz, CA); ERGIC-53 (Schweizer et al., 1988); mannosidase II (Moremen and Touster, 1985); galactosyltransferase (Berger and Hesford, 1985); MG160 (Gonatas et al., 1989); AP-1 (Ahle et al., 1988); Lamp1 (Carlsson et al., 1988; Mane et al., 1989); Lamp2 (Mane et al., 1989); CD63 (mAb αCD63, Beckman Coulter, Fullerton, CA); transferrin receptor (mAb B3/25, Roche Molecular Biochemicals); T7 epitope (mAb from Novagen, Madison, WI). mAbs to EEA1, human tGolgin-1 (p230), GM130, p115 and syntaxin 6 were from Becton Dickenson/Transduction Laboratories (San Diego, CA). Secondary antibodies to mouse, rabbit, sheep or rat IgG conjugated to rhodamine red X (RRX), fluorescein isothiocyanate (FITC), aminomethylcoumarin acetate (AMCA), or gold (6 nm and 12 nm) were from Jackson ImmunoResearch (West Grove, PA). FITC- and Texas Red-conjugated secondary antibodies to mouse IgG isotypes were from Southern Biotechnology (Birmingham, AL).
Cells were fixed with 2% formaldehyde and stained as described previously (Marks et al., 1995). Where indicated, cells were treated with 1 mg/ml leupeptin or 10-50 μg/ml cycloheximide (CHX) prior to fixation. For transferrin uptake, cells on coverslips were incubated with FITC-transferrin (from Sigma; 20 μg/ml) in serum-free medium containing 0.5% bovine serum albumin for 15-20 minutes, then rinsed in warm PBS prior to fixation. Cells were analyzed on a Leica (Bannockburn, IL) DM IRBE microscope using a Hamamatsu (Hamamatsu, Japan) Orca digital camera and Improvision (Lexington, MA) OpenLab software. Semiquantitative analyses were done with the Measurements module using raw images taken at constant exposure times pre-determined to be sub-saturating for the brightest samples. Outlines were drawn around individual cells, and total fluorescence from each cell was determined by multiplying the total number of pixels within the outline by the average value per pixel (after subtracting a background value per pixel, taken from an area within the photographed field in which there were no cells). Images shown were obtained using the Volume Deconvolution module from a series of raw images in different z-axis planes.
Metabolic pulse/chase and immunoprecipitation
Cells transfected with C312, C312(Y2187A), or control vector together with pCI-Pmel17 were metabolically labeled for 30 minutes with [35S]methionine/cysteine and chased for various periods of time. Cell lysates in Triton X-100 were immunoprecipitated with anti-Pmel17, immunoprecipitates were fractionated by SDS-PAGE and gels were analyzed by phosphorimaging as described previously (Berson et al., 2001).
Electron microscopy analyses
HeLa cells cotransfected with C312 or C312(Y2187A) (42 μg/10 cm dish) and pCDM8.1-Tac (4.2 μg) (Marks et al., 1995) were harvested in PBS/5mM EDTA, washed and stained with anti-Tac antibodies conjugated to phycoerythrin (Beckman Coulter, Fullerton, CA). Fluorescently labeled cells were harvested using a FACStar Plus cell sorter. Sorted cells were fixed as described previously (McCaffery and Farquhar, 1995) after a 15 or 30 minutes incubation at 37°C with 5 mg/ml horseradish peroxidase (Sigma). For conventional electron microscopy (EM), cells were fixed in 100 mM cacodylate buffer, pH 7.4, containing 3% formaldehyde, 1.5% glutaraldehyde, and 2.5% sucrose for 1 hour, washed, and osmicated at 4°C in Palade's fixative containing 1% OsO4. Cells were then washed, treated with tannic acid, stained with uranyl acetate, dehydrated through a graded series of ethanol, and embedded in epon. 80 nm sections were cut on a LEICA UCT ultramicrotome and analyzed on a Philips 420 TEM at 80 kV. For immunogold labeling of ultrathin cryosections, cells were fixed in PBS containing 4% paraformaldehyde for 1 hour, washed and harvested. Cell pellets were cryo-protected in 2.3 M sucrose containing 20% polyvinyl pyrollidone, mounted on aluminum cryopins, and frozen in liquid nitrogen. Ultrathin cryosections were then cut on a Reichert UCT ultramicrotome equipped with an FCS cryostage, and sections were collected onto 300 mesh, formvar/carbon-coated nickel grids. Grids were washed, blocked in 10% FCS, and incubated overnight with primary antibodies (10 μg/ml). After washing, grids were incubated with gold-conjugated secondary antibodies for 2 hours, washed, and embedded in a mixture containing 3.2% polyvinyl alcohol (10× 103 mol. mass), 0.2% methyl cellulose (400 centiposes), and 0.2% uranyl acetate. Sections were analyzed on a Philips EM 410 transmission electron microscope.
Disruption of TGN protein localization by overexpression of GRIP domains
In earlier work, HA-epitope-tagged, truncated forms of mouse tGolgin-1 containing the C-terminal 1247, 312, 186 or 81 amino acids, encompassing the GRIP domain, were efficiently localized to the Golgi when expressed at low levels by transfection in HeLa cells or other cell lines (Cowan et al., 2002). To probe GRIP protein function, we used immunofluorescence microscopy (IFM) to analyze whether high level expression of GRIP domain-containing fragments in transiently transfected HeLa cells affected the distribution of TGN resident proteins. HA-tagged C312 (corresponding to the C-terminal 312 amino acids of mouse tGolgin-1) was used in most experiments because of its enhanced expression and stability relative to the other truncated products, but similar results were observed using all GRIP domain-containing tGolgin-1 fragments. As the expression levels of C312 increased, pericentriolar anti-HA staining gave way to a diffuse staining pattern throughout the cytoplasm (Fig. 1b,c). We defined three qualitative fluorescence patterns: level I, a narrow ribbon-like structure consistent with Golgi localization; level II, a more diffuse paranuclear/pericentriolar structure with additional cytoplasmic staining; and level III, an intense cytoplasmic staining that may mask underlying structure. The C312 distribution in these cells was often reticular, perhaps reflecting association with ER membranes or cytoskeletal elements. Semiquantitative analysis of expression levels showed a correlation between total fluorescence intensity and the staining pattern observed (Fig. 1d). Consistent with previous reports (Kjer-Nielsen et al., 1999a; Kjer-Nielsen et al., 1999b), cells with level II or III staining patterns showed reduced or eliminated pericentriolar staining for endogenous tGolgin-1 and golgin-97 (see Fig. 3), likely reflecting competition for a limited GRIP domain binding site at the TGN (Barr, 1999; Kjer-Nielsen et al., 1999b).
Surprisingly, the localization of several TGN resident integral membrane proteins was disrupted in HeLa cells with high (level III) C312 expression (Fig. 2). TGN46, which localizes to a tight pericentriolar structure in untransfected cells or cells expressing low levels (level I) of C312 (Fig. 2, stars), was present in diffuse, vacuolated structures in cells with level III C312 expression (Fig. 2a-b′); in some cells expressing extremely high levels, pericentriolar staining was completely absent (not shown). Similarly, HA-tagged furin, which localizes to the TGN in transfected HeLa cells (Bosshart et al., 1994), was mislocalized to a diffuse and `expanded' pericentriolar structure in cells with level III expression of T7-epitope tagged C312 (Fig. 2e,e′). These effects were independent of cell-type and a function of TGN localization rather than of unrelated functions of the TGN46 and furin lumenal domains. This was evident in transiently transfected CHO cell variants by the C312-induced redistribution of the chimeric proteins TGG and TTF, which bear the cytoplasmic domains of TGN38 or furin, respectively, and the lumenal and transmembrane domains of an irrelevant protein, Tac (Fig. 2g-h′,j-k′). Furthermore, AP-1, which associates with the TGN as a peripheral membrane protein, also failed to accumulate in the pericentriolar region in C312-overexpressing cells (Fig. 2m-n′). Since TGN46/TGG and furin/TTF localize to the TGN via distinct cytoplasmic targeting signals (Humphrey et al., 1993; Voorhees et al., 1995) and recycling pathways (Ghosh et al., 1998; Mallet and Maxfield, 1999), and AP-1 localization is mediated by independent mechanisms (Page and Robinson, 1995; Seaman et al., 1996), these data show that high level expression of C312 disrupted TGN localization mediated by multiple pathways. Similar results were obtained in COS, MOP8, and NRK cells (unpublished data). Note that the distribution of syntaxin 6 and the cation-independent mannose 6-phosphate receptor did not overlap significantly with the TGN in our untransfected HeLa cells, and thus no change in their distribution could be observed in cells overexpressing C312 (unpublished data).
Three approaches were used to show that the displacement of TGN proteins was due to overexpression of an intact GRIP domain. First, cells were transfected with HA-tagged C312 in which a tyrosine conserved in all GRIP domains was replaced by alanine [C312(Y2187A)]; this fragment failed to localize to the Golgi in transfected cells at any expression level (Cowan et al., 2002) (see also Barr, 1999; Kjer-Nielsen et al., 1999a; Munro and Nichols, 1999). Expression of C312(Y2187A) at levels comparable to those sufficient for the level III phenotype of intact C312 (assessed by semiquantitative analysis of anti-HA fluorescence intensity) failed to disrupt the localization of any analyzed TGN resident protein (Fig. 2c,f,i,l,o), indicating that the GRIP domain must be intact to effect TGN disruption. Second, TGN46 mislocalization was observed upon overexpression of a C-terminal fragment of golgin-97, g97.C179, containing an independent GRIP domain (Fig. 3m,n). Finally, using semiquantitative IFM analyses in triple stained HeLa cells, we found that comparable levels of C312 expression induced both the mislocalization of TGN46 and the displacement of the endogenous GRIP proteins, tGolgin-1 and golgin-97, from the Golgi (Fig. 3a-l). Taken together, these data suggest that saturation of GRIP domain binding sites on TGN membranes or on a GRIP effector molecule results in disruption of the steady state distribution of TGN resident proteins.
To determine whether GRIP domain overexpression interfered with a dynamic process, we tested whether the change in TGN46 distribution was reversible. Because C312 has a short half-life (Fig. 4a; see also Fig. 7), it could be rapidly depleted by CHX treatment of transfected HeLa cells; treatment for 1-4 hours increased the fraction of transgene-positive cells with level I staining at the expense of cells with level III staining (Fig. 4b), indicating a loss of cells with high level C312 expression. Analysis of this same population of cells for TGN46 localization indicated that the fraction of cells with a wild-type pericentriolar TGN46 staining pattern increased over time of CHX treatment (Fig. 4c). Thus, the effect of GRIP domain overexpression on TGN46 distribution is reversible.
Specificity of protein localization defects in GRIP domain overexpressing cells
To determine whether the effects of GRIP domain overexpression were limited to the TGN, we analyzed the distribution of residents of other organelles in cells expressing high levels of C312. Staining patterns for Golgi stack integral (giantin and mannosidase II; ManII) and peripheral (GM130 and p115) membrane proteins were unaffected by overexpression of C312 (Fig. 5A,a-h) or of other GRIP domain-containing fragments (unpublished data). Furthermore, there were no consistent effects on the distribution of actin filaments, cell surface proteins, or markers of the endoplasmic reticulum (ER), ER/Golgi intermediate compartment, early endosomes, or late endosomes/lysosomes (Table 1). Pericentriolar staining for β1,4 galactosyltransferase (GalT), an enzyme that localizes to both the TGN and the trans Golgi cisternae, was disrupted in about half of the cells that overexpressed C312, but not in cells that overexpressed C312(Y2187A) (Fig. 5Ba-f). This effect was observed less consistently than for TGN46 or other TGN resident proteins (Fig. 2) and did not correlate with C312 expression level (unpublished data), perhaps reflecting the distribution of GalT to both the TGN and Golgi stacks (Nilsson et al., 1993; Rabouille et al., 1995) and cycling between these compartments (Cole et al., 1998; Miesenböck and Rothman, 1995). The results indicate that GRIP domain overexpression affects primarily TGN structure and/or protein localization, with minimal effects on the Golgi stacks and no discernible effects on other membranous organelles.
Rab6 has been implicated as a binding partner for the GRIP domains of tGolgin-1 and golgin-97 (Barr, 1999). Rab6 localization to the Golgi was only marginally affected in C312-overexpressing cells (Fig. 5g-l and Table 1); as for GalT, loss of tight pericentriolar Rab6 staining was observed in only a fraction (25.7%) of C312-overexpressing cells and did not correlate with C312 expression levels. This suggests that Rab6 localization to the Golgi is not dependent on binding of any particular GRIP-domain containing protein, and that excess GRIP domain cannot sequester Rab6 from membranes. Similar results were obtained with a more established GRIP domain binding partner, Arl1 (data not shown), which in yeast is known to regulate – but not to be regulated by – GRIP domain Golgi localization (Panic et al., 2003; Setty et al., 2003).
TGN functional defects induced by GRIP domain overexpression
We next tested whether the alterations in TGN protein distribution observed upon overexpression of GRIP domains underlie TGN functional defects. First, we assayed protein transport from the ER to the plasma membrane using an EGFP-tagged form of the temperature-sensitive ts045 variant of VSV-G (Bergmann et al., 1981; Presley et al., 1997). VSV-G-EGFP misfolds at the restrictive temperature of 39°C and is thus retained within the ER of transfected HeLa cells, producing a reticular pattern throughout the cytoplasm visible by fluorescence microscopy (Fig. 6a) (Presley et al., 1997). When shifted to 32°C, accumulated VSV-G-EGFP synchronously exits the ER and traverses the Golgi complex en route to the plasma membrane. VSV-G-EGFP largely accumulates at the Golgi by 30 minutes and at the plasma membrane by 60 minutes (Fig. 6b-c). Coexpression of VSV-G-EGFP with very high levels of C312(Y2187A), which lacks a functional GRIP domain, had no effect on the kinetics of VSV-G-EGFP transport (compare Fig. 6a-c to d-f). In cells expressing comparable levels of C312, VSV-G-EGFP staining was also similar to the controls at 39°C and after 30 minutes at 32°C (Fig. 6g,h), indicating normal folding kinetics and ER to Golgi transport. However, after 60-120 min, most C312-expressing cells retained VSV-G-EGFP within the Golgi region (Fig. 6i,j), indicating a block in Golgi to plasma membrane transport. By 3 hours, some VSV-G-EGFP reached the plasma membrane in most cells (Fig. 6k), indicating that the block was not absolute, but still a substantial proportion of cells displayed prominent pericentriolar fluorescence not observed in the controls. The Golgi to plasma membrane block was dependent on C312 expression level; in cells expressing lower levels of the C312, transport proceeded normally such that plasma membrane staining was apparent by 60 minutes (Fig. 6l), whereas in cells expressing extremely high levels of C312, no plasma membrane fluorescence was observed even after 3 hours (Fig. 6m).
As a second test of TGN function, we assessed the ability of proprotein convertases of the furin family to cleave a substrate protein. Furin localizes predominantly to the TGN (Shapiro et al., 1997) where it cleaves target proproteins at dibasic recognition sites (Nakayama, 1997). Pmel17, a glycoprotein found in melanosome precursors in pigment cells, is cleaved by furin or a related proprotein convertase en route to late endosomes in transfected HeLa cells (Berson et al., 2003). To determine whether furin cleavage is compromised in cells overexpressing GRIP domains, Pmel17 processing was assessed by metabolic pulse/chase and immunoprecipitation in transfected HeLa cells expressing Pmel17 without or with excess C312 or C312(Y2187A). In cells expressing Pmel17 alone, a single ∼100 kDa band (P1) present in cell lysates from pulse-labeled cells matured first to a slower migrating species (P2) by oligosaccharide processing in the Golgi, and then to two faster migrating species (Mα and Mβ) by proprotein convertase cleavage (Fig. 7b, lanes 1-4) as observed previously (Berson et al., 2001; Berson et al., 2003) and outlined schematically in Fig. 7a. By 4 hours, most of the P1 and P2 was converted to Mα and Mβ (lane 4). The kinetics of processing was virtually unchanged in cells co-expressing C312(Y2187A) (lanes 9-12). However, in cells co-expressing C312 at comparable levels (Fig. 7c), P2 accumulated and Mα and Mβ were not substantially generated (lanes 5-8). The formation of P2, which is resistant to digestion by endoglycosidase H (Berson et al., 2001; Berson et al., 2003), indicates that these cells had no, or minimal, defects in oligosaccharide processing within the Golgi stack. However, P2 accumulation coupled with Mα and Mβ depletion indicate that proprotein convertase cleavage of Pmel17 was blocked. These data indicate that overexpression of GRIP domains compromises either enzymatic functions or the interactions between enzyme and substrate within the TGN.
Despite these defects, some sorting functions normally ascribed to the TGN appeared to remain intact in C312-overexpressing cells. Transport to late endosomes was unaltered relative to controls as assessed by the steady state localization of several endogenous late endosomal and lysosomal proteins and the transport kinetics of cotransfected chimeric proteins (Table 1). These results suggest that C312 overexpression affects only certain TGN subdomains or the cycling of only certain cargo.
Ultrastructural defects and mislocalization of TGN46 in cells overexpressing GRIP domains
To determine the effects of C312 overexpression on TGN morphology in more detail, cells were analyzed by EM. Cells expressing C312 or C312(Y2187A) were enriched by cell sorting following cotransfection with a plasma membrane marker, Tac; by IFM, >90% of sorted cells expressed C312 or C312(Y2187A), the majority at high levels. Sorted cells were incubated with horseradish peroxidase (HRP) for 15-30 minutes at 37°C as a marker of fluid phase endocytosis prior to fixation, and then analyzed by EM either for morphology after embedding in epon or for immuno-EM after indirect immunogold labeling of ultrathin cryosections with antibodies to HA, HRP, and/or TGN46.
HeLa cells transfected with C312(Y2187A) displayed a relatively normal morphology, similar to untransfected cells, with intact, flattened pericentriolar Golgi cisternae (Fig. 8a). Cells transfected with C312 were more heterogeneous, but displayed general defects in Golgi morphology. Intact Golgi cisternae were difficult to find in most C312-expressing cells. They were replaced by large, pericentriolar vacuolated structures (Fig. 8b,c). The vacuoles varied in number, reflecting an apparent proliferation of Golgi/TGN rather than simply engorged cisternae. The structures often displaced a large fraction of the cytoplasm, and some of the vacuoles contained internal membrane sheets. Multivesicular endosomes (mve) were also more abundant in cells expressing C312 than those expressing C312(Y2187A) (e.g., see Fig. 8c). These data indicate that GRIP domain overexpression results in a large scale disruption of Golgi/TGN architecture and alterations in the abundance of late endosomes.
Immuno-EM analyses (Fig. 9) revealed the basis for the phenotypes observed by IFM. In cells expressing high levels of C312(Y2187A), TGN46 was localized as expected to cisternae and tubulovesicular structures at the trans side of the Golgi stack (Fig. 9A). A similar pattern was observed in cells expressing low levels of C312, in which TGN46 and C312 were colocalized in these structures (unpublished data). In cells expressing high levels of C312, however, TGN46 was instead detected primarily in two types of membrane compartments. One type comprised large vacuolated structures with few internal membranes (Fig. 9B,C), probably corresponding to those observed by conventional EM (Fig. 8). These structures were labeled abundantly on the cytoplasmic face by anti-HA antibodies detecting the C312 transgene (anti-HA labeling was also observed throughout the cytoplasm and in large, electron dense inclusions; unpublished data). The second class of compartments comprised multivesicular structures with numerous intralumenal vesicles (Fig. 9D-F). These structures were mve, since they were also labeled by HRP following internalization for 15 or 30 min (Fig. 9E,F). TGN46 labeling in mve was found both on the limiting and intralumenal membranes, suggesting targeting for degradation in lysosomes. Since TGN46 normally bypasses the mve and trafficks through recycling endosomes en route to the TGN (Ghosh et al., 1998), these data suggest that GRIP domain overexpression results in TGN46 mistargeting to both late endocytic organelles and enlarged, vacuolated Golgi-derived structures, thus interfering with the normal delivery of TGN46 from the endocytic pathway to the TGN.
Mammalian GRIP protein function has remained undefined, and a role for the yeast GRIP protein, Imh1p, in Golgi maintenance is inferred only from indirect genetic interactions. We have shown that overexpression of isolated GRIP domains in cultured mammalian cells results in specific disruption of TGN morphology, protein localization and function. The data provide evidence that at least one GRIP protein and/or GRIP ligand functions in TGN maintenance, probably by regulating recycling from endosomes.
Specificity of GRIP-domain-dependent disruption of vesicular transport
The displacement of endogenous GRIP proteins by expression of increasing doses of exogenous GRIP domains had been shown previously (Kjer-Nielsen et al., 1999a; Kjer-Nielsen et al., 1999b), but this is the first report to describe a concomitant redistribution of all TGN resident proteins. That TGN disruption required a functional GRIP domain was confirmed by (1) induction of a similar phenotype by hyper-expression of all GRIP-domain containing fragments of tGolgin-1 and golgin-97, (2) the inactivity of N-terminal tGolgin-1 fragments that lack the GRIP domain, and (3) the inactivity of a tGolgin-1 GRIP domain fragment with a mutation in the critical tyrosine residue. The innate instability of C-terminal fragments of tGolgin-1 (see Figs 4, 7) and the requirement for high level expression of GRIP domain-containing fragments to disrupt TGN localization may explain why this phenotype was not previously noted. Although TGN homeostasis was affected by overexpression of both tGolgin-1 and golgin-97 GRIP domains, subtle differences in the IFM staining pattern of the residual TGN and in the effects on some Golgi stack residents suggest some specificity in the function of individual GRIP domains.
With few exceptions, the disrupting effects of GRIP domain overexpression were limited to the TGN, with the strongest effects on TGN46 distribution. The effects did not extend to other organelles, consistent with the localization of GRIP proteins to the TGN (Brown et al., 2001; Gleeson et al., 1996; Luke et al., 2003). Effects on the Golgi stack were mixed. The distribution of most Golgi resident proteins by IFM was unaltered in GRIP overexpressing cells relative to controls, and Golgi function was largely unaffected based on the pericentriolar accumulation of ER-released VSV-G and on the largely unchanged kinetics of acquisition of Golgi modifications to Pmel17. Other effects on the Golgi may have been secondary to TGN disruption, perhaps as a consequence of cycling of stack residents through the TGN (Johnston et al., 1994). Two Golgi stack residents that also localize to the TGN (Martinez et al., 1994; Rabouille et al., 1995), GalT and Rab6, were redistributed in only a fraction of cells that expressed GRIP domain fragments at a threshold level sufficient to disrupt TGN46 distribution and not in a manner that correlated with the expression level of GRIP. The failure to observe characteristic Golgi stacks by EM in cells overexpressing GRIP domains was probably because they were obscured by the bloated TGN membranes and/or because of slight Golgi cisternal dilation proximal to the centriole. Finally, the redistribution of the early Golgi v-SNARE, GS28, in cells overexpressing the tGolgin-1-derived C312 (unpublished data) may, as for Rab6 and GalT, reflect a more general distribution of this vSNARE throughout the Golgi and TGN (Nagahama et al., 1996).
How might TGN dynamics be disrupted by GRIP domain overexpression? The expansion of membrane observed by EM and the failure to sort VSV-G from the Golgi to the plasma membrane suggest a defect in TGN export, consistent with in vitro budding of tGolgin-1-bound membranes from purified Golgi stacks (Gleeson et al., 1996). However, disrupted Golgi export is probably a secondary consequence of a primary failure to properly localize TGN resident proteins for several reasons. First, defective TGN export would not explain the redistribution of TGN resident proteins to peripheral structures. Second, this redistribution occurred at GRIP expression levels similar to those required to displace tGolgin-1 from the Golgi, suggesting that TGN46 mislocalization was a primary effect of competition for GRIP domain binding sites. Third, TGN46 was largely mislocalized to endosomes, from which TGN46 is normally recycled to the TGN (Ghosh et al., 1998; Mallet and Maxfield, 1999). We thus favor the interpretation that competition for GRIP domain binding sites interferes with the recycling of TGN46 from endosomes to the TGN (Fig. 10). The secretory defect could then result secondarily from the depletion of factors from the TGN that follow a similar recycling pathway and that are required for subsequent budding of plasma membrane-bound cargo, such as cargo recruitment proteins (Rojo et al., 1997) or v-SNAREs (Gurunathan et al., 2000; Salem et al., 1998; Springer and Schekman, 1998) (Fig. 10). Displacement of TGN resident proteins that follow a different recycling pathway, such as GalT and furin, might also be secondary to changes in TGN architecture. This model would explain why we did not observe redistribution of endosomal residents (Table 1), which rely on distinct cargo recruitment proteins and SNAREs with distinct recycling pathways. Our model predicts that only the post-endocytic endosome-to-TGN recycling step would be blocked by GRIP domain overexpression, consistent with qualitative assessments that failed to detect defects in internalization, recycling, and late endocytic delivery of several internalized cargo proteins (unpublished data).
Potential mechanisms of TGN disruption by GRIP domain overexpression
What might be the molecular basis for the effects of GRIP domain overexpression? One potential explanation is competitive displacement from GRIP domain binding sites of endogenous GRIP proteins that are required for TGN maintenance. The GRIP proteins may play a direct role in maintaining TGN homeostasis, or serve as regulators of an effector (or effectors) of TGN maintenance such that overexpression of the GRIP domain would saturate binding sites and block effector function. The structural similarity of GRIP proteins to tethering factors involved in membrane fusion events at other sites within the secretory and endosomal system (reviewed by Pfeffer, 1996; Pfeffer, 1999) or to components of the `Golgi matrix' required for stacking Golgi cisternae (Warren and Shorter, 2002) would be consistent with such a function. A second potential explanation for the effects is that overexpressed GRIP domains sequester GRIP domain ligands that function in TGN maintenance. Such ligands may interact with additional effectors, and thus their sequestration by excess GRIP domains could block functions in which GRIP proteins are not directly involved. The best candidate for such a ligand is Arl1. GTP-bound Arl1 associates with GRIP domains and other effectors in vitro (Lu et al., 2001; Panic et al., 2003; Setty et al., 2003; Van Valkenburgh et al., 2001) and in yeast is required for GRIP protein recruitment to the TGN (Panic et al., 2003; Setty et al., 2003). Moreover, the vacuolization of the Golgi and TGN observed here is similar to that observed upon overexpression of a predicted GTP-locked form of Arl1 (Lu et al., 2001; Van Valkenburgh et al., 2001), and another putative Arl1 effector in yeast, the VFT complex (Panic et al., 2003), has been implicated in regulating recycling from endosomes to the TGN (Siniossoglou et al., 2001; Siniossoglou et al., 2002). Another candidate GRIP ligand Rab6 (Barr, 1999), is a known effector of endosome to TGN recycling (Mallard et al., 2002) but less likely to be responsible for the observed phenotype; a GTP-locked form of Rab6 induces TGN46 redistribution (Martinez et al., 1994), but only at extremely high levels of expression and the effect is phenotypically distinct from that induced by GRIP overexpression (unpublished data). GRIP domains might also conceivably sequester lipid ligands, such as phosphatidylinositol phosphates which are known to regulate the morphology and function of the TGN and Golgi (Godi et al., 1999; Munro, 1998). Biochemical analyses of GRIP domain binding to Golgi membranes should help to identify potential GRIP effectors and the mechanism of GRIP overexpression-induced TGN disruption.
The authors wish to thank Sean Jordan for technical support, B. Wendland, C. Machamer and C. G. Burd for critical review of the manuscript, M. I. Greene for support of D.M.G., S. Munro, M. A. Lemmon and C. G. Burd for helpful discussions, H. Pletscher for help with cell sorting, and all of the investigators who graciously sent us reagents, particularly E. Berger, J. S. Bonifacino, T. Braulke, E. K. L. Chan, M. Fritzler, M. Fukuda, N. Gonatas, H.-P. Hauri, W. Hong, L. Johannes, J. Lippincott-Schwartz, F. Maxfield, K. Moremen, S. Munro and V. Ponnambalam. This work was primarily supported by American Cancer Society Research Project Grant RPG-00-238-01-CSM to M.S.M. and by National Science Foundation grant NSF-DBI-0099706 to J.M.M. A.Y. was supported in part by grant DAMB17-01-1-0365 from the US Dept. of the Army. S.S. was partly supported by training grant 5 T32 CA 09140 from the National Cancer Institute. D.M.G. was supported by grants from the Lucille P. Markey Charitable Trust and the National Multiple Sclerosis Society.
- Accepted July 7, 2003.
- © The Company of Biologists Limited 2003