Research Article

Chaperone Activity of Protein O-Fucosyltransferase 1 Promotes Notch Receptor Folding

See allHide authors and affiliations

Science  11 Mar 2005:
Vol. 307, Issue 5715, pp. 1599-1603
DOI: 10.1126/science.1108995


Notch proteins are receptors for a conserved signaling pathway that affects numerous cell fate decisions. We found that in Drosophila, Protein O-fucosyltransferase 1 (OFUT1), an enzyme that glycosylates epidermal growth factor–like domains of Notch, also has a distinct Notch chaperone activity. OFUT1 is an endoplasmic reticulum protein, and its localization was essential for function in vivo. OFUT1 could bind to Notch, was required for the trafficking of wild-type Notch out of the endoplasmic reticulum, and could partially rescue defects in secretion and ligand binding associated with Notch point mutations. This ability of OFUT1 to facilitate folding of Notch did not require its fucosyltransferase activity. Thus, a glycosyltransferase can bind its substrate in the endoplasmic reticulum to facilitate normal folding.

The extracellular domain of Notch proteins consists largely of tandemly repeated epidermal growth factor–like (EGF) domains. Notch folding must be tightly controlled, because disruption of a single EGF domain can dominantly perturb Notch, as in the human congenital syndrome CADASIL (cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy). Notch receptors are modified within their EGF domains by the addition of fucose to Ser or Thr residues (O-fucose) (1, 2). Drosophila Notch contains 36 EGF domains, 23 of which can potentially be O-fucosylated. O-fucosylation is catalyzed by Protein O-fucosyltransferase 1 (O-FucT-1), which is encoded by Ofut1 in Drosophila and by Pofut1 in mammals (3). Ofut1 and Pofut1 are essential for Notch signaling in vivo (47).

Notch accumulates in the ER in OFUT1-depleted cells. We expressed an Ofut1 RNA interference (RNAi) construct, iOfut1, in a stripe of cells in Drosophila wing discs and confirmed that OFUT1 protein levels were decreased (Fig. 1A). This resulted in elevated Notch protein staining, as also seen in clones of Ofut1 mutant cells (4) (Fig. 1B) (fig. S1). Notch transcription was not affected by Ofut1 RNAi in wing discs (Fig. 1C). Activation of Notch by its ligands induces Notch processing (8), which suggests that the accumulation of Notch might derive from its inability to interact with ligands. However, clones of cells mutant for genes required for ligand-dependent cleavages of Notch (9) or doubly mutant for the two Notch ligands Delta and Serrate (Fig. 1D) did not result in elevated Notch protein. Thus, OFUT1 has an influence on Notch that is distinct from any requirement for Notch-ligand binding (5, 7).

Fig. 1.

OFUT1 depletion results in accumulation of Notch in the ER. Wing discs are shown with dorsal up and anterior to the left. Panels (A) to (C) and (E) and (F) show expression of iOfut1 under ptc-Gal4 control (UAS-iOfut1 ptc-Gal4). In this and later figures, panels marked with primes show separate channels of the same disc. Additional controls are in fig. S1. (A) Disc stained with antiserum to OFUT1 (brown) reveals depletion effected by iOfut1; inset shows magnification of the boxed area. (B) Disc stained with anti-Notch (magenta) reveals elevated levels in OFUT1-depleted cells. Presence of a UAS-GFP transgene allows GFP (green fluorescent protein) to mark ptc-Gal4–expressing cells. Vertical section is shown at bottom. The boxed areas labeled “E,F” and “G” show approximate locations of images in (E) and (F) and in fig. S1, e to h (G), respectively. (C) In situ hybridization to Notch mRNA. (D) Delta Serrate mutant clones (green, white arrow) do not exhibit elevated Notch staining. Vertical section is shown at bottom. (E) Localization of Notch, but not Crumbs, is altered in OFUT1-depleted cells; note the paucity of white staining on the left side, reflective of the absence of Notch at the apical membrane. (F) Notch overlaps with an ER marker, anti-KDEL, in OFUT1-depleted cells. The dashed line in (E) and (F) marks the edge of the ptc expression stripe. (E) presents a more apical section than (F); mislocalization of Notch is most evident in (E), whereas elevated levels are most evident in (F).

In wild-type wing cells, Notch is localized near the adherens junctions, similar to the apical protein Crumbs (10, 11) (Fig. 1E). In OFUT1-depleted cells, Notch accumulated throughout the cytoplasm both basally and apically (Fig. 1, B and E). By comparison with Crumbs, Notch levels at the apical membrane were decreased relative to wild-type cells (Fig. 1E). Crumbs contains 30 extracellular EGF domains, many of which could potentially be O-fucosylated (2, 11), and the localization of the other known substrates of OFUT1, Delta and Serrate (12), was not changed in OFUT1-depleted cells (fig. S1). Thus, OFUT1 is specifically required for cell surface localization of Notch.

To identify where Notch trafficking becomes impaired, we compared the localization of Notch to markers of subcellular compartments. Notch staining in cells lacking OFUT1 overlapped markers of the endoplasmic reticulum (ER) (anti-KDEL and Boca) (Fig. 1F) (fig. S1), but did not overlap markers for Golgi (anti-Golgi), early endosomes (Hrs, GFP:rab5), or late endosomes (GFP:rab7) (fig. S1). Cultured Drosophila S2 cells express OFUT1 but lack a functional Notch gene. A fraction of transfected Notch in S2 cells appeared on the cell surface, whereas the rest remained in the ER (Fig.2A)(fig.S2). When OFUT1 was depleted from S2 cells by RNAi, cell surface staining of Notch was eliminated and only ER staining remained (Fig. 2B). We also examined a ∼120-kD Notch proteolytic fragment, which is thought to be generated by processing steps subsequent to secretion out of the ER (13). In OFUT1-depleted cells, this processed form of Notch was undetectable (Fig. 2F). We also assayed the secretion into the medium of a Notch–alkaline phosphatase (N:AP) fusion protein consisting of the 36 EGF domains of Notch fused to AP. A strong decrease in N:AP secretion was observed in OFUT1-depleted cells, whereas secretion of the analogous Delta and Serrate fusion proteins DL:AP and SER:AP was much less affected (Fig. 2G). The deficit in Notch secretion reflects a specific requirement for an O-FucT-1, because it can be reversed by cotransfection with mouse Pofut1 (Fig. 2). Thus, OFUT1 is required for normal trafficking of Notch out of the ER.

Fig. 2.

OFUT1 is essential for secretion of Notch in S2 cells. (A to E) S2 cells transfected to express Notch, stained before detergent treatment for cell surface Notch (ECN, red) and after detergent for intracellular Notch (ICN, green). In (B) to (E), cells were also treated with double-stranded RNA (dsRNA) corresponding to Ofut1; in (C) to (E), cells were also cotransfected with Pofut1 (C), Pofut1R245A (D), or Boca (E). (F) Immunoblot of S2 cell lysates probed with antibodies to the Notch intracellular domain and to OFUT1 and tubulin (as controls). Where indicated, lysates are from cells transfected with Notch, Boca, Pofut1, or Pofut1R245A and treated with Ofut1 dsRNA. Upper arrow points to full-length Notch; lower arrow points to a 120-kD fragment whose appearance requires OFUT1. (G) Secretion of AP-tagged Notch, Delta, or Serrate extracellular domains. Relative AP activity secreted into conditioned media is shown from transfected S2 cells, cells treated with dsRNA corresponding to Ofut1, and, where indicated, cells cotransfected with Boca, Pofut1, or Pofut1R245A. AP activity was normalized for transfection efficiency by luciferase activity in lysates (AP/luc).

OFUT1 is an ER protein. The influence of OFUT1 on Notch in the ER was unexpected, because transporter activity for guanosine diphosphate (GDP)–fucose, the donor substrate for fucosyltransferases, has been detected only on Golgi membranes (14), and all other fucosyltransferases examined to date are Golgi proteins (15). However, mammalian O-FucT-1 is modified with high mannose–type N-glycans, which could be consistent with localization in the ER (3). Indeed, in S2 cells, OFUT1 colocalized with ER markers (GFP:KDEL and anti-KDEL) (Fig. 3A) and did not overlap with Golgi markers (Fig. 3B). Similarly, in the wing disc, OFUT1 partially colocalized with an ER marker (anti-KDEL, Fig. 3C).

Fig. 3.

OFUT1 is an ER protein. (A) S2 cells transfected with an ER marker (GFP:KDEL, green) and stained for OFUT1 (red). (B) S2 cells stained for Golgi (anti-Golgi, green) and OFUT1. (C) Horizontal (top) and vertical (bottom) sections of a ptc-Gal4 UAS-iOfut1 wing disc stained for ER (anti-KDEL, green) and OFUT1. OFUT1 was depleted to the left of the dashed lines by RNAi. (D) Alignment of C-terminal sequences of various O-FucT-1s, with proposed KDEL signal underlined. (E) Immunoblot with antibody to S-tag epitope, showing localization of S-tagged OFUT1 and two OFUT1 mutants: deletion of the C terminus (ΔHEEL) and recreation of the OFUT1SH allele. The localization of both mutants is shifted from cell lysate to culture medium. (F) Assays on a Factor VII substrate of OFUT1 activity in conditioned media from cells expressing wild-type or mutant forms of OFUT1.

Soluble ER proteins are retained by a retrograde transport system that recognizes the C-terminal tetrapeptide KDEL and related variants. The presence of KDEL-like sequences is conserved in OFUT1 orthologs (Fig. 3D), which suggests that OFUT1 is a soluble ER protein. When OFUT1 was overexpressed, a small amount was secreted, consistent with observations that the KDEL receptor can be saturated (16); also, the mobility of OFUT1 on SDS–polyacrylamide gel electrophoresis (PAGE) gels in the culture media was indistinguishable from that in cell lysates, as expected if the N terminus is a signal peptide (fig. S3). Moreover, the N terminus of OFUT1 could be substituted with a signal peptide from another protein, Bip, without affecting its localization. Available antisera are directed against the C terminus of OFUT1 (5), but we were able to insert an S-epitope tag into Bip:OFUT1 without affecting its localization or activity; hence, we used this chimera to examine the requirement for the C-terminal HEEL (17) sequence. Deletion of these four amino acids substantially increased secretion of OFUT1 into the medium (Fig. 3, E and F).

A mutation in Ofut1, Ofut1SH, causes a severe loss-of-function phenotype in flies and is associated with replacement of the seven C-terminal amino acids of OFUT1 by four other amino acids (5). Because the C terminus includes an ER retention signal, this mutation might derive from improper localization rather than loss of enzymatic activity. To test this idea, we created a variant of Bip:OFUT1 carrying the C-terminal Ofut1SH mutation. This protein was secreted into the medium as efficiently as was Bip:OFUT1 lacking the HEEL sequence (Fig. 3E). Moreover, it displayed substantial enzymatic activity (Fig. 3F). Thus, Ofut1SH is a mislocalization mutant, and retention in the ER is required for normal OFUT1 function in vivo.

Fucosyltransferase activity of OFUT1 is not required for Notch secretion. Because the ER lacks detectable GDP-fucose transporter activity, we thought that OFUT1 might facilitate Notch folding independently of its fucosyltransferase activity. A comparative study of sequences from α1,2-fucosyltransferases and α1,6-fucosyltransferases identified a GXHXR(R/H) motif that is implicated in donor substrate binding (18, 19); O-fucosyltransferases include a related motif (20). The first R in this motif is invariant among all three fucosyltransferase families (Fig. 4A) and is essential for the activity of α1,6- and α1,2-fucosyltransferases in vitro and in vivo (18, 19, 21). We mutated this R in OFUT1, making OFUT1R245K and OFUT1R245A. Secreted versions of these mutants, including a C-terminal V5-hexahistidine tag, were expressed in S2 cells. The mutant chimeras were secreted as well as was the wild-type chimera, which suggests that they folded normally (fig. S3), but lacked fucosyltransferase activity (Fig. 4E) (fig. S3).

Fig. 4.

Fucosyltransferase activity is not required for Notch secretion. (A) Alignment of α1,2-fucosyltransferases, α1,6-fucosyltransferases, and O-fucosyltransferases identifies an invariant Arg residue (arrow) in a region implicated in GDP-fucose binding (18). (B) Clone of cells mutant for Ofut14R6 and marked by the absence of OFUT1 staining (green). Notch is mislocalized in mutant cells (outlined by dashes); elevated levels are only visible basally. Asterisks identify regions where Notch is out of the plane of focus. (C) Clone mutant for Ofut14R6 and simultaneously expressing OFUT1R245A under tubulin-Gal4 control. Notch secretion to the cell surface is restored, and accumulation inside cells is reduced. Antiserum to OFUT1 recognizes OFUT1R245A; the clone is marked by elevated staining (green). (D) Notch protein localization (red) in Ofut1SH overlaps an ER marker (Boca, green) and is complementary to an apical membrane marker (E-cadherin, cyan). (E) Assays on purified N:AP or control (Fc:AP) substrates of fucosyltransferase activity, using affinity-purified OFUT1:V5, OFUT1R245A:V5, or a mock purification (-Control) from untransfected cells as enzyme sources. (F) Schematic of the Gmd locus; exons are shown as bars and the open reading frame (nucleotides 61 to 1245) is in pink. In Gmd1 DNA, a deletion begins at nucleotide 65 and ends in an intron; the predicted peptide encoded by this allele is only two amino acids. (G to I) Wing imaginal discs stained for expression of WG (yellow) from wild-type (G), Ofut1SH (H), and Gmd1 (I). WG at the D-V boundary (arrowhead) is regulated by Notch and is missing from Ofut1 and Gmd mutants. WG in the hinge (arrows) is regulated separately but illustrates reduced growth. (J) Notch protein localization in Gmd1 is largely complementary to Boca and partially overlaps E-cadherin. Arrows highlight Notch at the apical membrane.

We then investigated the ability of OFUT1R245A to promote Notch secretion. In developing wing discs, we created clones of cells that were simultaneously mutant for a null allele of the endogenous Ofut1 gene and expressed OFUT1R245A under the control of a heterologous promoter. Expression of this enzymatically inactive form of OFUT1 restored Notch localization to the cell surface and partially suppressed intracellular staining of Notch (Fig. 4, B and C). To conduct rescue experiments in S2 cells, we engineered the R245A mutation into murine Pofut1 (Pofut1R245A) (fig. S3). Expression of Pofut1R245A in OFUT1-depleted S2 cells partially restored the trafficking of Notch to the cell surface, as assayed by immunostaining, production of the 120-kD cleavage product, and secretion of N:AP (Fig. 2). A control ER protein, the low-density lipoprotein (LDL) receptor chaperone Boca, could not restore Notch secretion (Fig. 2). Thus, OFUT1 can facilitate the secretion of Notch independently of its fucosyltransferase activity.

As an independent test of the lack of requirement for O-fucose in Notch secretion, we created and analyzed mutations in the predicted Drosophila GDP-mannose 4,6-dehydratase (Gmd) (Fig. 4F), which is required for GDP-fucose biosynthesis and hence for all forms of fucosylation in Drosophila. A null mutant, Gmd1, could survive to third instar. These animals exhibited severely decreased growth of the wing disc and lacked expression of Notch target genes at the dorsal-ventral boundary [e.g., Wingless (WG)] (Fig. 4I). These phenotypes are consistent with the expected consequences of absence of O-fucose and were similar to those observed in Ofut1SH (Fig. 4H). However, in Ofut1SH mutants Notch accumulated in the ER (Fig. 4D), but in Gmd1 mutants, Notch did not accumulate in the ER and could be detected on the cell surface (Fig. 4J).

OFUT1 binds Notch. The failure of Notch to be secreted from the ER in the absence of OFUT1 could reflect a chaperone activity analogous to the action of Boca, which facilitates the folding of LDL receptors by transiently associating with a specific protein domain (16, 22, 23). In the absence of OFUT1, misfolded Notch would then be retained in the ER by quality control mechanisms (24, 25). This hypothesis predicts that it should be possible to detect an association between Notch and OFUT1. Mammalian O-FucT-1 has affinity for EGF domains, as it was partially purified by EGF domain affinity chromatography (3). To detect binding between OFUT1 and Drosophila Notch, we coexpressed OFUT1:V5:His and N:AP in S2 cells and then used coimmunoprecipitation to assay the association between Notch and OFUT1. Specific binding of OFUT1 or OFUT1R245A to N:AP was reproducibly detected (Fig. 5A), as was weak binding to Delta (fig. S4).

Fig. 5.

Chaperone activity of OFUT1. (A) Coimmunoprecipitation experiment with V5-tagged OFUT1, OFUT1R245A, or Boca, and N:AP or Fc:AP. The indicated proteins were coexpressed in S2 cells and isolated from conditioned media. Input amounts of V5-tagged proteins are shown by immunoblot (top) and of AP-tagged proteins by AP assay. Fc:AP is expressed and secreted more robustly than N:AP, but nonetheless it does not bind OFUT1. Coimmunoprecipitation is shown by the immunoblot at bottom. (B) Relative secretion of N:AP and N:AP mutants in S2 cells, expressed as a normalized ratio of AP to luciferase activity when coexpressed with OFUT1R245A (+R245A) or Boca. (C) Binding of wild-type or mutant forms of N:AP (or, as a negative control, Fc:AP) to Delta-expressing (left of each pair of bars) or control S2 cells (right of each pair). AP fusions were transfected into wild-type (OFUT1-expressing) S2 cells; where indicated, cells were cotransfected with OFUT1R245A or Boca. Equal amounts of protein (AP activity 4000 mOD/min) were used in each case.

Chaperone activity of OFUT1 can promote the folding of wild-type and mutant Notch. To further evaluate the hypothesis that OFUT1 is a Notch chaper-one, we assayed its influence on the secretion and ligand-binding activity of Notch. These experiments were conducted with N:AP so that amounts of Notch could be quantified (5, 26). OFUT1R245A enhanced both N:AP secretion and ligand binding (Fig. 5, B and C) (fig. S4), which suggests that at endogenous levels of OFUT1, a fraction of N:AP is misfolded. We also examined the influence of OFUT1 on mutant forms of Notch. N:AP-EGF23-32f is a form of N:AP that contains single base changes in each of the EGF domains 23 through 32, such that the site of O-fucose attachment is changed from the normal Ser or Thr to Ala or Val. This mutant was secreted at less than one-tenth the rate of wild-type N:AP, and it bound Delta-expressing cells less than half as well as did wild-type N:AP (Fig. 5). OFUT1R245A promoted the secretion and ligand binding of N:AP-EGF23-32f (Fig. 5).

CADASIL is a disease characterized by strokes, dementia, and early death; it is caused by point mutations in Notch3 (27) that either add or eliminate a Cys in one of the EGF domains. It is not yet clear whether the etiology of CADASIL results from alteration of normal Notch3 function or from the accumulation of protein aggregates (28, 29). In any case, the molecular nature of CADASIL mutations implies that they influence protein folding. By sequence alignment, we modeled two CADASIL mutations, R141C and C542Y, onto Drosophila Notch, and created the equivalent mutations, S200C and C599Y. Secretion and ligand binding of N:APS200C were not significantly different from that of wild-type N:AP. However, both secretion and ligand binding of N:APC599Y were impaired, and coexpression with OFUT1R245A enhanced N:APC599Y secretion and ligand binding (Fig. 5). Thus, OFUT1R245A promoted the secretion and activity of both wild-type and mutant forms of Notch, supporting the conclusion that OFUT1 is a Notch chaperone.

Chaperone activity and glycosyltransferases. The observations that OFUT1 binds to Notch, that OFUT1 is an ER protein, and that Notch secretion requires OFUT1 (but not its fucosyltransferase activity) together suggest a model for OFUT1 action (fig. S5). OFUT1 could associate with folded EGF domains of Notch in the ER during folding of the extracellular domain. This association could be required to prevent inappropriate inter- or intramolecular associations of EGF domains, which would otherwise interfere with further Notch folding. OFUT1 would catalyze the transfer of fucose either during folding (if O-fucosylation can occur in the ER) or subsequently (if it happens in another compartment). In the absence of OFUT1, individual EGF domains could still fold, but the extracellular domain would not assemble correctly, resulting in misfolded and/or aggregated Notch. Implicit in this model is that the specificity of OFUT1 for folding of Notch lies in the need to block inappropriate associations of Notch EGF domains. By contrast, the folding reactions for other EGF domain–containing proteins, such as Serrate, Delta, and Crumbs, apparently do not present similar possibilities for inappropriate association and hence do not require OFUT1.

Other glycosyltransferases with specific protein substrates have been identified, including O-FucT-2 (30), O-glucosyltransferase (31), O-mannosyltransferase (32), and C-mannosyltransferase (33). O-and C-mannosyltransferases appear to be ER glycosyltransferases (32, 34), and the subcellular distributions of O-FucT-2 and O-glucosyltransferase and have not yet been determined. Thus, ER glycosyltransferases with substrate specificity might constitute a class of quality control proteins.

Supporting Online Material

Materials and Methods

Figs. S1 to S5


References and Notes

View Abstract

Navigate This Article