Signal-Mediated Dynamic Retention of Glycosyltransferases in the Golgi

See allHide authors and affiliations

Science  18 Jul 2008:
Vol. 321, Issue 5887, pp. 404-407
DOI: 10.1126/science.1159411


Golgi-resident glycosyltransferases are a family of enzymes that sequentially modify glycoproteins in a subcompartment-specific manner. These type II integral membrane proteins are characterized by a short cytoplasmically exposed amino-terminal tail and a luminal enzymatic domain. The cytoplasmic tails play a role in the localization of glycosyltransferases, and coat protein complex I (COPI) vesicle–mediated retrograde transport is also involved in their Golgi localization. However, the tails of these enzymes lack known COPI-binding motifs. Here, we found that Vps74p bound to a pentameric motif present in the cytoplasmic tails of the majority of yeast Golgi-localized glycosyltransferases, as well as to COPI. We propose that Vps74p maintains the steady-state localization of Golgi glycosyltransferases dynamically, by promoting their incorporation into COPI-coated vesicles.

The Golgi apparatus functions both as a sorting hub for newly synthesized proteins and as the location of posttranslational modification of glycoproteins. Protein glycosylation is initiated in the endoplasmic reticulum (ER). Glycoproteins then undergo extensive sequential carbohydrate chain elongations and elaborations mediated by a group of Golgi-localized glycosyltransferases. Although they catalyze distinct glycosylation reactions and display characteristic cisternal distributions, Golgi-resident glycosyltransferases are all type II integral membrane proteins consisting of a short cytoplasmically exposed N terminus (tail), a single membrane-spanning region, and a luminally orientated stalk region and catalytic domain.

A variety of features have been identified that influence the localization of these enzymes (13): the transmembrane domain (48), the luminally oriented noncatalytic region (6, 8), interactions between catalytic domains (9, 10), and the cytoplasmic tail (1113). The steady-state distribution of these enzymes is maintained by a dynamic process that involves their retrieval from late Golgi cisterna to early cisterna, as well as between the Golgi and the ER, whereupon they transit back to the cisterna on which they function (14, 15). Although the features of glycosyltransferases that direct a particular enzyme to one cisterna or another at steady-state varies, presumably coat protein complex I (COPI)–coated vesicles are the predominant means by which the majority of these enzymes are recycled (14, 16). However, Golgi-resident glycosyltransferases lack canonical COPI-binding motifs in their cytoplasmic tails, and we sought to identify proteins that could bind the tails of these enzymes and could facilitate their incorporation into COPI-coated vesicles.

We identified Vps74p in a genetic screen for dosage suppressors of the lethality associated with deletion of the essential Golgi-resident SNARE (soluble N-ethylmaleimide–sensitive factor attachment protein receptor), Sft1p, a protein required for retrograde traffic within the Golgi (17). Vps74p is a member of a protein family that includes the mammalian proteins GPP34 (GmX33α) and GPP34R (GmX33β) (1821). Although deletion of VPS74 has no obvious effects on yeast cell growth, vps74Δ cells are known to missort carboxypeptidase Y (CPY), which suggests that Vps74p is involved in protein trafficking (22). We observed that vps74Δ cells were defective in the processing of the glycosylphosphatidylinositol (GPI)–anchored protein Gas1p. Gas1p is modified by the addition of N- and O-linked carbohydrates as it traverses the Golgi en route to the cell surface (Fig. 1A) (23). Comparisons of the fate of Gas1p in yeast strains lacking enzymes involved in N- and O-glycosylation revealed a similar processing defect in cells lacking the α1,2-mannosyltransferase, Kre2p (Fig. 1, A and B) (24). We considered the possibility that Vps74p might be required for Gas1p transport and examined the trafficking of a functional Gas1p–monomeric red fluorescent protein (Gas1p-mRFP) fusion protein (25) in cells lacking Vps74p. However, the location of Gas1p-mRFP in vps74Δ cells was indistinguishable from that in wild-type cells (Fig. 1C).

Fig. 1.

Vps74p is required for Kre2p retention in the Golgi. (A) The location and sequential order of action of glycosyltransferases involved in Gas1p processing (29). (B) vps74Δ cells have a flaw similar to the Gas1p-processing defect of cells lacking Kre2p. The relative positions of fully processed and glycosylated Gas1p (mGas1p) and the ER-resident precursor form of Gas1p (pGas1p) are indicated. (C) Transport of Gas1p-mRFP to the cell surface is unaffected in vps74Δ cells. Scale bar, 5 μm. (D) Vps74p binds to membranes. Whole-cell extracts (WCEs) were subjected to sequential rounds of centrifugation at 13,000g (13K) and 100,000g (100K), and the pellet (P) and supernatants (S) were subjected to SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and immunostaining. (E) Vps74p is a peripheral membrane protein. Yeast WCEs were extracted with lysis buffer alone (LB) or with 1% Triton X100 (T×100), 0.1 M Na2CO3 pH 11.5 (pH), or 1 M NaCl (NaCl) in lysis buffer. (F) Kre2CT-mRFP but not GFP-Rer1p is mislocalized to the vacuole lumen in vps74Δ cells. For a detailed description of reagents and methods used, see (24). DIC, differential interference contrast.

Like GPP34 (18), Vps74p is a peripheral membrane protein (Fig. 1, D and E) and, as such, would only be capable of direct physical interaction with the cytoplasmically orientated N-terminal 11 amino acids of Kre2p. We constructed a protein hybrid in which the cytoplasmic and transmembrane regions of Kre2p were fused to the N-terminal end of mRFP (Kre2CT-mRFP). In wild-type cells, Kre2CT-mRFP was localized to numerous puncta throughout the cytoplasm, which partially overlapped that of a green fluorescent protein (GFP) fusion to the Golgi-resident protein, Rer1p (GFP-Rer1p) (Fig. 1F). Furthermore, as do Rer1p (26) and the Mnn9p glycosyltransferase complex (16), Kre2CT-GFP cycled between the Golgi and the ER via a COPI- and SNARE–dependent mechanism (fig. S1). In vps74Δ cells, however, Kre2CT-mRFP was localized in the lumen of the vacuole (Fig. 1F). This effect was not due to a general deficiency in the retention of Golgi-resident membrane proteins, because the localization of GFP-Rer1p to the Golgi was unaffected. These findings are consistent with a role for Vps74p in the dynamic retention of Kre2p in the Golgi and identify the N-terminal 11 amino acids of the protein as sufficient to mediate retention.

Yeast cells in which particular genes involved in glycosylation have been deleted show reduced viability in the presence of calcofluor white (CW). Cells lacking VPS74 were more sensitive to CW than kre2Δ cells (fig. S2), so we reasoned that Vps74p might mediate the Golgi localization of additional glycosyltransferases. When VPS74 was deleted from strains expressing Mnn9-GFP, Mnn2-GFP, or Ktr6CT-GFP, these proteins were indeed mislocalized—either to a vesicle-like haze (Mnn9-GFP), or to the lumen of the vacuole (Mnn2-GFP and Ktr6CT-GFP) (Fig. 2A). Alignment of the N-terminal cytoplasmic tails of members of the yeast glycosyltransferase family revealed the presence of a semiconserved motif (F/L)-(L/V)-(S/T) (27) in the cytoplasmic tails of several members of the family (Table 1). We substituted Ala for Phe4-Leu5-Ser6 (FLS) in Kre2p and assessed the fate of these substitutions on the localization of kre2 (AAA6)CT-GFP, as well as on the capacity of the mutant full-length protein, kre2(AAA6), to process Gas1p in kre2Δ cells. kre2(AAA6)CT-GFP was mislocalized to the interior of the vacuole in wild-type cells (Fig. 2B), and cells expressing kre2(AAA6) as their sole source of the enzyme were indistinguishable from cells lacking the KRE2 gene in their inability to fully process Gas1p (Fig. 2C). We suggest that the defect in Gas1p processing was related to the instability of Kre2p in cells lacking Vps74p. Indeed, Kre2p was significantly less abundant in vps74Δ cells, and the abundance of kre2(AAA6) in wild-type cells was very similar to that of Kre2p in vps74Δ cells (Fig. 2D). In addition, Gas1p processing in kre2Δ cells expressing N-terminal Kre2p hybrid proteins generated from glycosyltransferases containing FLS-like motifs [with the exception of Anp1 (Table 1)] was indistinguishable from that in wild-type cells (Fig. 3A). Further evidence that the FLS-like motif was important for recognition by Vps74p was obtained from a yeast two-hybrid assay in which the wild-type or FLS Ala-substituted cytoplasmic tail of Kre2p [kre2C3(AAA6)] was used to assess binding by Vps74p. Although Vps74p was capable of binding to the wild-type cytoplasmic tail sequence of Kre2p, no interaction was observed between Vps74p and kre2C3(AAA6) (Fig. 3B). Vps74p binds directly to the cytoplasmic tail of Kre2p, and substitutions to the FLS motif reduce binding (Fig. 3C and fig. S3).

Fig. 2.

Deletion of VPS74 results in the mislocalization and degradation of a subset of Golgi glycosyltransferases. (A) Mnn2-GFP, Mnn9-GFP, and Ktr6CT-GFP are mislocalized in vps74Δ cells. (B) Amino acid substitutions to the FLS6 sequence in the cytoplasmic tail of Kre2 [kre2(AAA6)CT-GFP] results in mislocalization of the protein. (A and B) Scale bar, 5 μm. (C) Cells expressing kre2(AAA6) as their sole source of Kre2p are defective in the processing of Gas1p. (D) kre2(AAA6) is unstable in vps74Δ cells. (+) or (–) indicates endoglycosidase H treatment. The glycosylated (gly) and deglycosylated (de-gly) forms of Kre2p, see arrows. For a detailed description of reagents and methods used, see (24). WT, wild type.

Fig. 3.

Vps74p binds to the cytoplasmic tails of Golgi glycosyltransferases and to coatomer. (A) Glycosyltransferase-Kre2p chimeras bearing FLS-like motifs restore Gas1p processing in kre2Δ cells. (B) The FLS6 sequence in the cytoplasmic tail of Kre2p is required for interaction with Vps74p in a yeast two-hybrid assay. AD and BD refer to the activation and binding domains of GAL4, respectively. (C) Vps74p binds directly to the cytoplasmic tail of Kre2p. (Top) Immunoblot of 5% copurified proteins; (middle and bottom) 20% of input proteins. (Graph) Quantification of the results (means ± SD). A fusion protein of kre2C3(AAA6) with glutathione S-transferase [kre2C3(AAA6)-GST] (26.4 ± 18.6%) and GST alone (0.1 ± 0.3%). (D) Overexpression of VPS74 redistributes Kre2CT-GFP from the Golgi to the ER. The arrows highlight the ER localization of Kre2CT-GFP. Scale bar, 5 μm. (E) GST-Vps74p binds to coatomer from yeast WCEs. The binding of GST-Vps74p to coatomer was 7.5 ± 0.4 (means ± SD) times as great as that of GST alone. (F) Vps74p binds to full-length GST-Ret2p and to the trunk domain (amino acids 1 to 372) of GST-Sec26p. (Top, left) Proteins purified on glutathione beads after incubation with histidine (His)6-tagged Vps74p; (top, right) 1× input. The arrows indicate the positions of the recombinant proteins. (Bottom) Immunostaining of Vps74p bound to GST fusion proteins. The relative amounts of Vps74p input proteins mixed with GST fusion proteins are shown at 1× and 10×. (G) GPP34R suppresses the instability of Kre2p in vps74Δ cells. For a detailed description of reagents and methods used, see (24).

Table 1.

The amino acid sequences of the cytoplasmic tails of yeast Golgi-resident glycosyltransferases (27). FLS-like motifs are highlighted in bold italic font. Glycosyltransferases bearing presumptive Vps74p-binding (F/L)(L/I/V)-X-X-(R/K) motifs are highlighted in bold font and their corresponding motifs are underlined and in italics.

View this table:

The inability of the Anp1-Kre2 hybrid protein to restore Gas1p processing to kre2Δ cells (Fig. 3A) was puzzling, so we considered the possibility that the Vps74p glycosyltransferase recognition site contained residues in addition to the FLS-like motif. A recent study suggests an important role for dibasic sequences in the recycling of Mnn9p from the Golgi to the ER (13), and our yeast two-hybrid assays, Kre2p stability studies, and live-cell imaging experiments supported a requirement for the KR8 peptide in the Vps74-dependent retention of Kre2p (fig. S4). In addition, we observed no requirement for the polar residue within the FLS-like motif of Kre2p, whereas F4 and L5 were essential for binding to Vps74p (fig. S5). Thus, the Vps74p-binding site on yeast Golgi-resident glycosyltransferases conforms to a consensus sequence of (F/L)-(L/I/V)-X-X-(R/K), a motif that is present in 16 Golgi-resident enzymes (Table 1). To test the validity of this proposal, we examined the fate of Mnn5-GFP, which contains a sequence that conforms to the consensus (LIXXK6), in vps74Δ cells. Consistent with our prediction, Mnn5-GFP was mislocalized to the vacuole lumen in these cells (fig. S6).

Vps74p could facilitate Golgi retention of glycosyltransferases either by blocking anterograde trafficking of these enzymes from the Golgi or by assisting their recycling through Golgi cisterna or from the Golgi to the ER, in a process mediated via the COPI coat (fig. S1). To distinguish between these possibilities, we examined the fate of Kre2CT-GFP in cells that constitutively overexpressed Vps74p. We reasoned that if Vps74p functioned in an ER recycling pathway, overexpression of the protein might be expected to shift the steady-state distribution of Kre2CT-GFP from the Golgi to the ER. This is indeed the case; the effect of VPS74 overexpression on the steady-state localization of Kre2CT-GFP was not due to an indirect block in ER export or to a general acceleration of retrograde transport (Fig. 3D).

A role for Vps74p in retrograde transport of Kre2CT-GFP was also supported by our genetic studies. In addition to VPS74, we also identified genes encoding components of the Golgi vesicle coat complex COPI or its substrates (fig. S7). Because Sft1p is required for retrograde transport within the Golgi (17), we surmised that overexpression of these genes compensated for this deficiency in sft1Δ cells. In addition, genetic interactions were observed between VPS74 and mutants defective in ER-Golgi trafficking (fig. S7).

Given that the Mnn9p complex (16) and Kre2CT-GFP (fig. S1) cycle between the Golgi and ER in a COPI-dependent manner, we considered the possibility that Vps74p might function together with COPI to mediate the recycling of glycosyltransferases. Consistent with this hypothesis, Sec26p (βCOP) was dependent on Vps74p for its ability to support the growth of sft1Δ cells (fig. S8A), and Vps74p bound to coatomer (Fig. 3E), as well as to the trunk region of Sec26p and Ret2p (δCOP) in in vitro mixing assays (fig. S8B and Fig. 3F).

We suggest that Vps74p functions as a component of the Golgi-ER vesicular transport machinery as a glycosyltransferase-sorting receptor for the COPI coat and, in so doing, controls the steady-state distribution of these enzymes by preventing their transport beyond the Golgi. Additional support for this model comes from a recent study in which the requirement for Vps74p in the retention of glycosyltransferases was also reported (28). Although the human homologs of Vps74p, GPP34, and GPP34R suppress the phenotypes of vps74Δ cells (Fig. 3G and fig. S9), plants and protozoa appear to lack orthologs of Vps74p. Thus, multiple mechanisms (13) likely function to ensure the correct localization of Golgi-resident glycosyltransferases in cells.

Supporting Online Material

Materials and Methods

Figs. S1 to S9

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article