We have taken two complementary approaches to evaluate the respective contributions of the fucosyltransferase and chaperone activities of OFUT1 to Notch signaling. First, we employed a mutant isoform of OFUT1, OFUT1^R245A. This mutation alters an invariant arginine within the putative GDP binding site, and it eliminates detectable fucosyltransferase activity, while retaining chaperone activity 10. A genomic rescue construct carrying the Ofut1^R245A mutation was created, and then introduced into flies by P element-mediated transformation. The ability of the OFUT1^R245A isoform to rescue Notch signaling phenotypes was then assayed in animals lacking wild-type OFUT1. As Ofut1 is provided maternally to embryos, this was done by making germline clones with a null allele of Ofut1, Ofut1^4R6, in animals containing the Ofut1^R245A genomic construct inserted on the same chromosome arm. In the absence of any rescue construct, animals lacking maternal and zygotic Ofut1 exhibit a strong neurogenic phenotype, in which impairment of lateral inhibition causes excess neurons to be produced at the expense of epidermal cells 9. This phenotype is characteristic of mutations in Notch pathway genes and can be visualized by staining with antibodies against a marker for neuronal cells, ELAV (Figure 1B, cf. Figure 1A). Significantly, when Ofut1^R245A is provided to Ofut1^4R6 germline clones embryos, embryonic neurogenesis, as revealed by ELAV staining, appears indistinguishable from that in wild-type embryos (Figure 1C). These embryos lacking O-FucT activity can hatch, but die as first instar larvae. This result implies that the fucosyltransferase activity of OFUT1 is not essential for Notch signaling during lateral inhibition.

As an independent test of the requirement for Notch fucosylation, we analyzed a mutant in which wild-type OFUT1 is present but unable to fucosylate Notch owing to the absence of its donor substrate, GDP-fucose. GDP-fucose is generated in the cytoplasm from GDP-mannose by enzymes including GDP-D-mannose dehydratase (GMD). Since Drosophila lack a salvage pathway for GDP-fucose synthesis 12, and GDP-fucose is the common donor substrate for all fucosyltransferases, animals lacking GMD are expected to be unable to effect all types of fucosylation. Animals that are mutant for a null allele of Gmd, Gmd¹ 10, die as larvae with no neurogenic phenotype (not shown), but the survival of Gmd zygotic mutants could have been a result of maternally provided product. Thus, requirements for Gmd during embryonic development were evaluated by making germline clones. Confirmation of the general deficit in fucosylation in these embryos was provided by staining with anti-HRP antibodies, which recognize a fucose-containing epitope on neuronal N-glycans 13. Anti-HRP staining was completely eliminated in animals lacking both maternal and zyogotic contributions of Gmd (Figure 1D). Strikingly, however, Notch signaling in these embryos, as visualized by ELAV staining, was indistinguishable from wild type (Figure 1D) and Gmd germline clone embryos are able to complete embryogenesis and hatch before dying as first instar larvae. These results establish that all fucose modifications are dispensable for lateral inhibition, as well as all other developmental processes essential for the hatching of a Drosophila embryo.

Notch signaling participates in many different processes throughout development and many of the factors that modulate Notch signaling are context specific. For example, fng is dispensable for Notch functions during embryonic neurogenesis, but plays a critical role in the developing wing. In the wing, fng is expressed by dorsal cells and its positive and negative effects on signaling by Delta and Ser, respectively, position a stripe of Notch activation along the edge of fng expression at the D-V boundary 5. The critical role of fng in positioning Notch activation is evidenced by the observation that establishing novel fng expression boundaries by creating fng mutant clones in dorsal cells results in the induction of ectopic stripes of Notch activation, which can be visualized by examining expression of targets of Notch signaling, such as Wingless (WG; see Figure 2C) 14. In contrast, Ofut1 mutant clones are associated with a complete loss of Notch activation in the wing (Figure 2D) 915, equivalent to that observed in Notch mutant clones.

To evaluate requirements for Notch O-fucosylation during wing development, we again took advantage of the ability to rescue the chaperone, but not the fucosyltransferase, activities of OFUT1 by expression of Ofut1^R245A. Ofut1^R245A was expressed in Ofut1^4R6 null mutant clones using two different methods. In one set of experiments, we expressed Ofut1^R245A from a genomic rescue construct, such that Ofut1^R245A was expressed under its own promoter. In an alternative approach, Ofut1^R245A was over-expressed in Ofut1^4R6 mutant clones using the mosaic analysis with a repressible cell marker (MARCM) technique 16, which employs the UAS-Gal4 system to drive expression under the control of a heterologous promoter. Both methods yielded similar results: when Ofut1^R245A expression is similar to endogenous Ofut1 levels, the rescued clones phenocopy fng mutant clones (Figures 2E and 2F, cf. Figure 2C). Thus, where the clones intersected the normal D-V boundary, normal WG expression is lost, whereas the borders of clones within the dorsal compartment can be associated with ectopic WG expression. This striking phenotype has two important implications. First, because WG expression can be induced within mutant cells expressing only Ofut1^R245A, it indicates that, as during embryonic neurogenesis, modification of Notch with O-fucose is not required for it to function as a receptor. Second, it indicates that the absence of O-fucose is functionally equivalent to the absence of elongated (i.e. FNG-modified) O-fucose in the developing wing.

We also attempted to extend these observations by examining Gmd mutants. Animals that are mutant for a null allele, Gmd¹, exhibit decreased growth and loss of WG expression in the wing imaginal disks, consistent with a deficit in Notch signaling 10. To investigate whether this phenotype reflects a specific requirement for Gmd in FNG-dependent Notch signaling, or a more general requirement for Gmd in Notch signaling, we created clones of cells homozygous for Gmd¹. However, in most cases these did not show obvious Notch-loss-of-function phenotypes (not shown). Only when the Minute technique was used to generate disks in which all, or almost all, of the wing was composed of mutant tissue was a loss of WG expression observed, and in all cases the loss of WG expression was non-autonomous (Figure 2G). A non-autonomous phenotype in clones has been reported previously for a mutation in UDP-glucose dehydrogenase, which is required for the synthesis of heparan sulfate 17. We suggest that a general non-autonomy of mutations in genes that participate in nucleotide sugar biosynthesis could be explained if nucleotide sugars can diffuse through cells via gap junctions. Owing this non-autonomy, Gmd clones could not be directly compared with fng or Ofut1 clones.

Loss of Ofut1 impairs Notch signaling, but over-expression of Ofut1 can also impair Notch signaling 7. This observation led to the suggestion that high levels of Notch O-fucosylation might suppress Notch signaling. However, the determination that it is the chaperone activity, rather than the fucosyltransferase activity, that is universally required for Notch signaling prompted us to re-examine the basis for the inhibition of Notch signaling associated with OFUT1 over-expression. Towards that end, Ofut1^R245A was expressed at high levels in the developing notum under the control of ap-Gal4. Expression of wild-type Ofut1 (UAS-Ofut1 [8.2]) under ap-Gal4 control interferes with Notch-mediated lateral inhibition, and so results in the production of extra bristles (Figure 3B, cf. Figure 3A) 7. The effect on bristles was only evident in microchaetes, but not in macrochaetes. Over-expression of Ofut1^R245A (UAS-Ofut1^R245A[28.3]) under ap-Gal4 control results in a similar phenotype (Figure 3C). This observation, together with another recent study 18, indicate that the over-expression phenotype is independent of OFUT1's fucosyltransferase activity. Interestingly, when much stronger Ofut1 expression (relative to Figure 3B) was induced in a thin stripe of cells along the anterior-posterior (A-P) compartment boundary using the ptc-Gal4 driver with UAS-Ofut1 [11.1], WG expression was inhibited non-autonomously (Figure 3E, cf. Figure 3D). As extracellular concentrations of nucleotide sugars are generally thought to be too low to support glycosylation, this non-autonomous affect is also consistent with a fucosylation-independent activity.

The observation that the fucosyltransferase activity of OFUT1 is not required for all Notch signaling events implies that the previously described Notch chaperone activity of OFUT1 is its most critical function. Recently, however, our identification of this chaperone activity has been questioned and two additional activities for OFUT1 have been proposed in its place: the first, independent of its fucosyltransferase activity, promotes endocytosis of Notch from the plasma membrane to the early endosome 18; the second, dependent upon fucosyltransferase activity, promotes trafficking of Notch from the plasma membrane to the sub-apical complex and adherens junctions 18. The two most critical experiments distinguishing between the chaperone model and these trafficking models for OFUT1 function relate to the localization of Notch in cells lacking OFUT1.

First, it was claimed that in contrast to our report that Notch is not secreted to the cell surface in the absence of OFUT1 10, Notch is secreted to the plasma membrane in Ofut1 mutant cells 18. This claim was based on an assay in which anti-Notch antibodies were added to live disks. However, this is not an effective assay for cell surface localization, because a cell surface receptor is not required for bulk endocytosis (e.g. even fluorescent dextran is efficiently endocytosed by disk cells 19), and once endocytosed, antibodies could be spread throughout the secretory pathway and then accumulate wherever there are significant epitope concentrations. A standard assay for cell surface localization is to determine whether the accessibility of epitopes requires membrane permeabilization. Thus, we performed immunostaining using antibodies directed against the extracellular domain of Notch, both in the presence and in the absence of detergent. When wing disk cells are permeabilized with detergent, wild-type cells exhibit both intracellular staining as well as apical hexagonal staining, corresponding to the normal cell surface localization of Notch near the adherens junctions (Figures 4A and 4B). Ofut1 mutant cells show increased Notch staining (Figures 4A and 4B), which we have previously shown overlaps with ER markers 10 (see Additional file 1). To demonstrate that the Notch accumulation near the apical surface in Ofut1 mutant cells is not exposed on the cell surface, we stained disks without detergent treatment. Under these conditions, no staining was observed in Ofut1 mutant cells, whereas neighboring wild-type cells exhibit normal apical surface staining (Figures 4C and 4D). These results are consistent with previous data using Drosophila S2 cells 10 and reconfirm that OFUT1 is required for secretion of Notch to the cell surface in wing disk cells.

The second critical piece of evidence presented against the chaperone model was a claim that the Notch that accumulates in Ofut1 mutant cells is not actually in the ER. This claim relies on the belief that the ER is homogeneous organelle, that is, that all ER proteins are homogeneously distributed. Both our studies 10 and those of Sasaki et al.18 reveal overlap between Notch and ER markers in Ofut1 mutant or RNAi-depleted cells. At the same time, we agree that Notch does not overlap perfectly with ER markers in all focal planes; indeed, it is for this reason that we originally thought that Notch was not in the ER in the absence of OFUT1 7. However, we now think that differences between Notch and other markers reflect ER heterogeneity. In support of this idea, we have investigated the relative distributions of five different ER markers in wing imaginal disks cells: two dedicated ER chaperones (OFUT1 and Boca), a classic ER protein marker (Calnexin), a synthetic ER protein (GFP with a KDEL ER retention signal added) and bulk ER proteins (using an anti-KDEL antibody). Each of these ER markers exhibits partial, but not perfect, overlap with other ER markers (see Figures 4E and 4F and Additional file 1) 10. This observation of ER heterogeneity emphasizes that the lack of perfect correspondence among markers cannot be taken as compelling evidence for the absence of ER localization of Notch in Ofut1 mutant cells. Together with observations that the accumulated Notch in Ofut1 cells does not overlap markers for any other vesicles or organelles 1018, that the distribution of Notch in Ofut1 mutant cells is similar to the distribution of OFUT1 itself 10, which by several criteria is an ER protein 1011, and that Notch does partially overlap with a variety of ER markers in Ofut1 mutant or depleted cells 1018, these observations are consistent with the conclusion that Notch accumulation within Ofut1 mutant cells is in the ER.

Ofut1^4R6 was obtained from K Matsuno 9. UAS-Ofut1^R245A[28.3] and Gmd¹ are described in 10. The Gal4 drivers used were ptc-Gal4 (Flybase ID number; FBti0002124), ap-Gal4 (FBti0002785). UAS-iOfut1 [16.2], UAS-iOfut1 [12.3], UAS-Ofut1 [11.1] and UAS-Ofut1 [8.2] (insertion on chromosome 2) were obtained as described previously 7. UAS-Ofut1 [11.1] provides stronger expression than UAS-Ofut1[8.2].

To obtain mutants lacking both maternal and zygotic Gmd, germline clones of Gmd¹ were made by crossing Gmd¹FRT 40A/CyO virgins to hs flp; ovo^DFRT 40A/CyO males and heat shocking the progeny at 38°C for 1 h on two successive days during first and second larval instar. Then, hs flp; Gmd¹FRT40A/ovo^DFRT 40A female progeny that contained Gmd¹ germline clones were crossed to Gmd¹/CyO; twi-Gal4 UAS-GFP males.

To obtain mutants lacking both maternal and zygotic Ofut1, germline clones of Ofut1^4R6 were made by crossing FRT42B [G13] Ofut1^4R6/CyO virgins to hs-flp [122]; FRT42B [G13] ovo^D/CyO males and heat shocking the progeny at 38°C for 1 h on two successive days during the second and early third larval instar. Then, hs-flp [122]; FRT42B [G13] Ofut1^4R6/FRT42B [G13] ovo^D female progeny that contained Ofut1^4R6 germline clones were crossed to FRT42B [G13] Ofut1^4R6/CyO, twi-Gal4 UAS-GFP males.

To generate larger clone for Gmd¹in disks, the Minute technique 27 was employed. M(2) Ubi-GFP FRT40A/CyO act-GFP flies were crossed to yflp; FRT40A Gmd¹/CyO flies and the progeny were heat shocked at 38°C for 1 h during first larval instar. The homozygous Gmd¹ cells produced by mitotic recombination lack the Minute mutation and, thus, divide more quickly than the surrounding cells heterozygous for Minute. The homozygous Minute clones cannot survive.

To express Ofut1^R245A in Ofut1 mutant backgrounds, the MARCM technique 2728 was used. This technique allows the generation of mosaic clones that are mutant for one gene and expressing another gene under the control of a UAS promoter. Mosaic clones were generated by crossing y w hs-Flp [122] tub-Gal4 UAS-GFP:nls; FRT42B [G13] hsp:Myc tub-Gal80 [LL2]/CyO females to FRT42B [G13] Ofut1^4R6/CyO; UAS-Ofut1^R245A [28.3]/TM6B males. The expression of Ofut1^R245A is suppressed by the ubiquitous expression of the Gal80 transcriptional repressor. However, FRT-mediated mitotic recombination generating mutant clones for Ofut1^4R6 simultaneously removes tub-Gal80, which leads to disinhibition of Gal4 and, thus, drives the expression of both the GFP marker and Ofut1^R245A.

For genomic rescue experiments, a 3.8 kb BamHI/EcoRI fragment comprising the Ofut1 genomic sequence flanked by partial sequences of the neighboring genes CG8257 and CG8309, was cloned into the corresponding sites of pCasper. The R245A mutation was introduced by site directed mutagenesis (Stratagene). The resulting construct, pCasper-Ofut1^R245A, was subjected to transposon-mediated germline transformation. To recombine with Ofut1^4R6, lines with insertions on the right arm of chromosome 2 were selected. Three lines, Ofut1^R245A[18.1] (insertion; 43C), Ofut1^R245A[33.1] (insertion; 53E) and Ofut1^R245A[23.1] (insertion; 59A-C), are obtained. Among them, the expression was confirmed in Ofut1^R245A[18.1] and Ofut1^R245A[23.1]. As the recombination between Ofut1^R245A[18.1] and FRT42B [G13] was difficult owing to the proximity of each insertion site, we used Ofut1^R245A[23.1] for further analysis.

For expression of GFP:KDEL in wing disk cells, the GFP:KDEL coding fragment was isolated by polymerase chain reaction (PCR) from pMT/Bip-GFP:KDEL 10, cloned into pUAST and transformed in to Drosophila.

Antibody staining was performed essentially as described previously 5. Antibodies used were mouse anti-WG (4D4, DSHB), mouse anti-Notch (C458.2H, DSHB), rat anti-ELAV (7E8A10, DSHB), rabbit anti-HRP (Cappel), rabbit anti-Notch (Intracellular Notch, E. Giniger), guinea pig anti-OFUT1 10, guinea pig anti-Boca antibody 29, mouse anti-KDEL antibody (Stressgen) and rabbit anti-Calnexin antibody (Stressgen).

For cell surface Notch staining of wing disks, third instar larvae were dissected and fixed for 30 min with the fixative solution containing 4% formaldehyde (Polysciences Inc. #18814), 0.1 M Pipes (pH7.2) and 50 mM NaCl. After washing three times for 5 min with phosphate buffered saline (PBS; 10 mM sodium phosphate, 2.7 mM KCl, 137 mM NaCl) followed by blocking with PBS plus 5% donkey serum (PBS-DS) for 30 min, disks were incubated with mouse anti-extracellular Notch (C458.2H; DSHB) in PBS-DS overnight. After washing with PBS (three times for 5 min) followed by blocking with PBS-DS for 30 min, disks were incubated for 3 h with Cy3-conjugated donkey anti-mouse IgG (Jacson Laboratories) in PBS-DS containing 0.05% BSA and 0.005% Triton-X 100, and then washed three times for 5 min with PBS containing 1% BSA and 0.1% Triton-X 100 (PBT).