Induction of neuron-specific glycosylation by Tollo/Toll-8, aDrosophila Toll-like receptor expressed in non-neural cells

Eukaryotic cells are enveloped within a complex coating of carbohydrate. Composed of the pendant oligosaccharide moieties of glycoproteins, glycolipids and proteoglycans, the glycocalyx constitutes the interface at which cells interact with each other and with their environment. Glycans mediate cellular recognition and adhesion, facilitate protein maturation, regulate protein activity through allostery and bioavailability, influence receptor ligation,and modulate transmembrane signaling(Varki, 1993;Hammond et al., 1994;Tsuda et al., 1999;Sharrow and Tiemeyer, 2001;Moloney et al., 2000;Vyas et al., 2002). Specific oligosaccharide structural elements participate in each of these processes,requiring spatial and temporal regulation of glycan synthesis in developing and mature organisms. As a consequence of regulated expression,oligosaccharide structures are among the most discriminating markers for cellular differentiation in complex tissues(Feizi, 1985;Dodd and Jessell, 1985;Allendoerfer et al., 1999;Matthews et al., 2002). Despite extensive descriptions of specific oligosaccharide distributions, the cellular and molecular mechanisms by which cells achieve expression of their characteristic portfolio of surface glycans are largely unknown.

Antibodies raised against the plant glycoprotein, Horseradish Peroxidase(HRP), crossreact with an N-linked oligosaccharide epitope that is distributed throughout the Drosophila melanogaster nervous system and is also expressed in a small, well characterized subset of non-neural tissues(Jan and Jan, 1982;Snow et al., 1987). Two mutations abolish expression of the HRP epitope. In the first, designatednac, the epitope is absent in the larval, pupal and adult nervous system (Katz et al., 1988). The molecular nature of the nac mutation is unknown, but affected adults exhibit sensory afferent defasciculation and behavioral phenotypes(Whitlock, 1993;Phillis et al., 1993). The second mutation that abolishes HRP-epitope expression is carried on the TM3 balancer chromosome, an extensively rearranged form of the third chromosome(Snow et al., 1987). TM3 homozygotes do not express the HRP epitope in the embryonic nervous system but do produce the glycan in the expected non-neural tissues. Therefore, it is likely that the structural genes necessary for synthesis of the HRP epitope are intact and that the TM3 mutation alters a gene that regulates tissue-specific glycosylation.

We report characterization of the TM3 locus that abolishes HRP-epitope expression. The affected gene, which we named `tollo' encodes a member of the family of cell surface receptors with homology to the Toll protein (Toll-like receptors, TLRs). Genome sequence characterization re-identified the tollo locus, resulting in its designation as`toll-8' (Tauszig et al.,2000). The founding member of the TLR family (Toll) was originally identified as a component of the signaling pathway that induces dorsal-ventral polarity in the Drosophila embryo(Anderson et al., 1985). Subsequently, TLRs have also been shown to participate in innate immune responses in Drosophila and other organisms by transducing pattern recognition signals (Hoffman et al.,1999; Medzhitov et al.,1997; Williams et al.,1997; Ip and Levine,1994). We now add the induction of tissue-specific glycosylation to the list of TLR functions.

Rabbit antiserum against HRP (anti-HRP antibody) and secondary antibodies were obtained from Jackson Laboratories (Westgrove, PA). Monoclonal antibodies 1D4, BP102 and 22C10 were provided by Corey Goodman, University of California at Berkeley. Genomic clones in P1 phage were obtained from the Artavanis-Tsakonas laboratory and grown using standard methods.

A non-complementation screen using the third chromosome deletion set(Bloomington Stock Center) was performed to determine which region of the TM3 balancer chromosome is responsible for loss of the HRP epitope. For deletion stocks maintained over the TM3 balancer, embryos were collected and stained for loss of the HRP epitope. For stocks not already over the TM3 balancer,progeny were collected from a cross of the deletion stock to a TM3 stock in which the balancer carried an embryonically expressed lacZ marker(D, ry/TM3-DZ, P{ry+t^7.2=H22.7}, Bloomington Stock Center). The lacZ activity, detectable in maxillary segments from stage 12-17, allowed the HRP-status of blue embryos to be interpreted as complementation/non-complementation of the TM3 mutation.

Once a region of interest was identified (by the fz^M21interval 70D2/3-71E4/5), additional deletions were obtained to refine the location of the affected TM3 locus. Deletion lines and their breakpoints from sources other than the Bloomington Stock Center were as follows:D^5rv14 (70D1/2-71C1/2; A. Carpenter, University of Cambridge), Brd¹⁵ (71A1/2-71C1/2; J. Posakony, University of California at San Diego) and fz^D21 (70D2-70E8,Umeå Stock Center). Other deletions were obtained from the Bloomington Stock Center: Df(3L)BK10 (71C3 - 71E5), D^5rv5 (70C3/4 -70F5/71A1), fz^GS1a (70D1 - 70E7) andfz^GF3b (70C1/2 - 70D4/5). The Gal4 driver stocks,rhomboid-Gal4, ELAV-Gal4 and hsp70-Gal4 were obtained from Tian Xu and a second chromosome UAS-lacZ line was obtained from Haig Keshishian (both at Yale University).

Embryo collections were dechorionated, fixed, devitellinized, stained with antibodies and staged according to standard methods(Patel, 1994;Campos-Ortega and Hartenstein,1985). Primary antibody dilutions were 1:500 for rabbit anti-HRP,1:3 for BP102, 1:5 for 22C10 and 1:10 for 1D4. lacZ activity was detected in embryos as previously described(Klämbt et al., 1991). The distribution of tollo mRNA in embryos was visualized by in situ hybridization using single-stranded DNA probes labeled with digoxigenin(Tautz and Pfeifle, 1989). For co-localization of tollo mRNA and mAb 22C10 staining, in situ hybridization was performed first, using digoxigenin-11-UTP-labeled RNA probes prepared by in vitro transcription(Kopczynski et al., 1996). For RNA and DNA probes, a 2.4 kb EcoR1 genomic fragment (tollo nucleotide 2796 to 5230) served as template. Embryos were routinely examined with sense and antisense probes prepared from plasmid templates bearing insert in opposite orientations.

Poly-A⁺ RNA was isolated from OreR orBrd¹⁵/TM3 embryonic total RNA by hybridization to biotinylated poly-dT oligonucleotide followed by capture and release from streptavidin magnetic beads as recommended by the manufacturer(Boehringer/Roche). To generate probe for Reverse Northern analysis,poly-A⁺ mRNA was dephosphorylated with calf intestinal phosphatase and then end-labeled with γ-³²P-ATP by T4 Polynucleotide Kinase (Sambrook et al.,1989). Drosophila genomic DNA in P1 phage was digested with restriction enzymes, electrophoresed, transferred to nylon and probed with end-labeled mRNA prepared from OreR or Brd¹⁵/TM3 embryos.

A 2.4 kb EcoRI fragment that demonstrated differential hybridization was isolated from P1 phage DS06206, subcloned into pBluescript(Stratagene) and sequenced in both directions by multiple rounds of overlapping sequence acquisition with the dideoxy chain termination method (W. M. Keck, Biotechnology Resource Center at Yale University). The open-reading frame (ORF) identified in the EcoRI fragment was extended in the 5′ direction by overlapping sequencing reactions that used DNA isolated from P1 phage DS06206 as template. The resulting genomic DNA sequence,containing a 4038 nucleotide ORF, 17 nucleotides of 5′-, and 1189 nucleotides of 3′-UTR, was submitted to GenBank (Accession Number,AF204158).

Sequence was extended in the 3′ direction by 3′-RACE (Clontech Marathon) with poly-A⁺ RNA isolated from OreR orBrd¹⁵/TM3 embryos and nested primer sets(Frohman, 1993). Distinct bands of 1.6 kb from OreR and 1.2 kb from Brd¹⁵/TM3 were obtained. Primer design incorporated restriction sites (5′ PstI and 3′ XhoI) for ease of subcloning. The final 3′-RACE reactions were digested with PstI and XhoI, yielding fragments (1.1 kb from OreR and 0.9 kb from Brd¹⁵/TM3)that were subcloned into pBluescript and sequenced. Promoter components were identified by the NNPP2.1 prediction tool (Berkeley Drosophila Genome Project). Other sequence analysis used BLAST programs available through the National Center for Biotechnology Information(Altschul et al., 1997). Northern analysis used poly-A⁺ RNA isolated from overnight embryo collections and genomic Southern analysis used genomic DNA prepared from adults. Both blotting procedures were performed by standard methods using random-primed ³²P-labeled DNA probes(Sambrook et al., 1989).

Plasmid bearing a full-length tollo insert was constructed with DNA from two overlapping Drosophila genomic clones in P1 phage. The 2.4 kb EcoRI fragment isolated from P1 clone DS06206, containing 3′ tollo sequence, was ligated into the EcoRI site of pBluescript I (KS)+. The resulting plasmid was digested with SalI(cuts in the 5′ polylinker) and MluI (cuts in tollo3′-UTR) to accept a 7.5 kb SalI/MluI fragment prepared from P1 clone DS05329. In addition to a complete tollo ORF, the resulting plasmid (designated pGtollo) contains 2.5 kb of 5′and 1.2 kb of 3′ genomic sequence.

To generate a rescue construct in which tollo coding sequence was placed under control of UAS elements, a NotI site was introduced into the full-length tollo construct (pGtollo) by PCR such that the enzyme cut 11 nucleotides upstream from the first ATG codon. NotI digest of the resulting construct, designated pGtolloN_-11,released a 5.2 kb fragment that was ligated into NotI-linearized pUAST transformation vector to produce pUASTtollo(Brand et al., 1994).

To generate a transformation construct in which Gal4 expression was controlled by tollo 5′ genomic sequence, a NotI site was introduced into cloned genomic DNA by PCR (pGtollo template) such that the enzyme cut seven nucleotides upstream from the first ATG codon. The amplified fragment was subcloned into pCR2.1 and recovered as a 2.5 kb fragment with a 5′ blunted SalI end and a 3′NotI end. After ligation into the pCaSpeR3 transformation vector(previously digested with StuI/NotI), Gal4 coding sequence was added as a NotI fragment, prepared from pGaTN, to yield a construct designated ptolloGal4(Brand and Perrimon,1993).

Transformant lines were produced by injection of rescue constructs intow¹¹¹⁸; Sb,Δ2-3/TM6b flies by standard procedures. Transformation with the pUASTtollo or ptolloGal4 vector yielded insertions on both the X and second chromosomes that were then crossed into the Brd¹⁵/TM3-lacZ background.

Embryos homozygous for the TM3 balancer chromosome fail to express the HRP epitope in the ventral nerve cord and peripheral nervous system, although expression is maintained in three non-neural tissues(Fig. 1A,B). TM3 rearrangement breakpoints provide likely candidates for regions of the third chromosome that affect HRP-epitope expression. To assess the relevance of these breakpoints and to ensure that a gene located between TM3 breakpoints was not missed,overlapping deletion stocks that together cover approximately 80% of the third chromosome were screened for their ability to complement loss of the HRP epitope in a TM3 background (Fig. 2). The smallest non-complementing deletion(Fig. 1C, breakpoints 71A1/2-71C1/2) is carried in a Bearded stock designatedBrd¹⁵ (Leviten and Posakony, 1996). The combination Brd¹⁵/TM3 orBrd¹⁵/TM3-lacZ produces viable and fertile adults.

Of the relevant genotypes, only Brd¹⁵ homozygotes display gross morphologic aberrations (Fig. 1D-G). Brd¹⁵/Brd¹⁵ embryos develop normally until stage 14 when defects in the formation of anterior terminal structures become apparent. In particular, retraction of the clypeolabrum is stalled in Brd¹⁵ homozygotes, causing the supraesophageal ganglia (embryonic brain lobes) to appear exteriorized. The head involution defect provides an unambiguous, reliable diagnostic for theBrd¹⁵ homozygous genotype(Fig. 1G, arrowhead). Examination of neural tissue integrity by monoclonal antibody staining demonstrates that longitudinal and commissural bundles are present in the central nervous system (mAb BP102), appropriate cell numbers and approximate cellular relations are preserved in the peripheral nervous system (mAb 22C10),and efferent motor pathways develop normally (mAb 1D4) inBrd¹⁵ homozygotes. Thus, neural differentiation and axon extension, to the extent that they are revealed by these mAb markers, are unaffected by Brd¹⁵.

The proximal breakpoint of the Brd¹⁵ deletion (71C1/2)overlaps a TM3 rearrangement breakpoint at 71C(Lindsley and Zimm, 1992). Therefore, P1 phage clones that map to the 71C1/2 interval were obtained and probed with ³²P-end-labeled mRNA prepared from embryo collections from OreR or Brd¹⁵/TM3 stocks. The P1 phage clone designated DS06206 contains a 2.4 kb EcoRI fragment and a 1.1 kbBamHI/XbaI fragment that are both transcribed in OreR but not detected in Brd¹⁵/TM3 embryos(Fig. 3A-C). Subsequent sequence analysis placed the 1.1 kb fragment within the 2.4 kb fragment. Probe prepared from the 2.4 kb EcoRI fragment was used to probe northern blots of poly-A⁺ RNA isolated from OreR orBrd¹⁵/TM3 embryos. A 6.5-7.0 kb band was identified in the OreR preparation that was not detected in Brd¹⁵/TM3 poly-A⁺ RNA (Fig. 3D). Genomic Southern analysis demonstrates multiple restriction fragment length polymorphisms in the Brd¹⁵/TM3 genotype. Probe prepared from the 2.4 kb EcoRI fragment hybridizes to a 1.5 kbHindIII fragment in OreR that is shifted to approximately 6 kb inBrd¹⁵/TM3. In turn, probe prepared from the 1.5 kbHindIII fragment identifies the same polymorphism as well as 876 bpHindIII/Xba1 and 280 bp PstI fragments that also differentiate the two genotypes (Fig. 3E-G).

The sequenced 2.4 kb EcoRI genomic fragment yielded an open reading frame of 1250 bp which was extended to 4038 bp in length by further genomic sequencing (Fig. 4). Motifs in the predicted protein define it as a member of the Toll-like Receptor family (Hashimoto et al.,1988). Leucine-rich repeats and cysteine-rich domains are found in the extracellular portion of the molecule and a Toll homology (TH) domain is present in the C-terminal region. Based on the TH similarity and on the inability of Brd¹⁵/TM3 adults to reliably extricate themselves from their food, we named the gene `tollo', a Finnish word roughly translated as `stupid'.

A total of 6735 nucleotides were sequenced, extending from 17 bp upstream of the ORF to 2677 bp beyond the first in-frame stop codon, and found to be co-linear with genomic sequence in GenBank Accession Number AE003531 and withDrosophila cDNA sequence LD 33590(Adams et al., 2000;Rubin et al., 2000). The sequence predicts that the HindIII, XbaI and PstI polymorphisms observed in Brd¹⁵/TM3 lie in the 3′UTR of the gene (Fig. 3E-G). To more precisely define the polymorphism, 3′-RACE was performed on poly-A⁺ RNA isolated from OreR and Brd¹⁵/TM3 embryos. The fragment amplified from Brd¹⁵/TM3 embryos yielded 969 nucleotides of sequence of which the first 237 matched previously sequenced genomic DNA. However, TM3 sequence diverged from wild-type at a position corresponding to nucleotide 5635, 1.6 kb downstream from the first in-frame stop codon of tollo (nucleotide 4039) and within the 3′ UTR predicted by mRNA size (Fig. 3D, Fig. 4). Sequence obtained for the first 732 bases of divergence matches aDrosophila transposable element designated `412', GenBank Accession Number X04132 (Yuki et al.,1986). It was not determined whether the divergent sequence reflects the insertion of an intact transposable element or identifies the site of the TM3 rearrangement breakpoint previously mapped to 71C.

Expression of tollo mRNA is first detected in the cellular blastoderm, initially as prominent bands at both ends of the embryo(Fig. 5A). Very rapidly,tollo mRNA appears in dorsoventral bands repeated along the entire length of the embryo (Fig. 5B). As germband retraction begins, late in stage 12, the bands of tollomRNA span the entire width of the germband(Fig. 5C,D), placing Tollo expression within ectodermal domains from which ventral nerve cord precursor cells differentiate and delaminate. By the time germband retraction is complete (stage 13), tollo mRNA expression disappears from the ventral ectoderm that underlies the delaminated, discrete nerve cord. Expression of the HRP epitope in the ventral nerve cord is first detected reproducibly at stage 14, shortly after tollo mRNA decreases. Thus,Tollo expression in ventral ectoderm coincides with a period of maximal contact with differentiating neurons and disappears once neurons segregate from the ectoderm to form a consolidated ventral nerve cord.

In the lateral ectoderm, tollo mRNA is found in distinct,segmentally repeated domains at stage 13. Together, these repeated domains form continuous anteroposterior stripes of ectodermal expression(Fig. 5E,F). Within each domain, expression is not uniform. Cells at the segment boundaries express higher levels of tollo mRNA, forming ectodermal pockets that are partly lined with Tollo-expressing cells. By early stage 15, tollomRNA is greatly reduced in the lateral ectoderm and expressing domains are attenuated to a few cells immediately adjacent to segment boundaries. Expression of the HRP epitope in the peripheral nervous system is first detected reproducibly at stage 15, shortly after tollo mRNA expression has decreased in the lateral ectoderm.

Consistent with the determination that the TM3 chromosome has not lost the entire tollo gene (Figs3,4), hybridization signal was detected in TM3/TM3 embryos (Fig. 5G). Thus, RNA that contains tollo sequence is produced in TM3 embryos despite being undetectable in blotted poly-A⁺ mRNA preparations (Fig. 3D). Hybridization to Brd¹⁵/Brd¹⁵ embryos was not detected at any stage, indicating that the anti-sense probe is specific for tollo and does not cross-hybridize to other embryonically expressed Toll-like receptors(Fig. 5H).

The position of Tollo-expressing domains along the dorsoventral axis of the lateral ectoderm closely approximates the site of proneural cluster formation(Blochinger et al., 1990). Therefore, tollo mRNA expression was localized relative to the position of differentiating neurons in the peripheral nervous system(Fig. 6A,B). Within the lateral domains of Tollo expression found in each segment, maturing neurons occupy patches that display reduced or undetectable tollo mRNA(Fig. 6A). At stage 14, all but the earliest neurons to differentiate (which have actively begun to migrate away from their birthplace towards their final embryonic positions) are found in close association with ectodermal cells that express tollo mRNA(Fig. 6B). Thus, in the peripheral nervous system, as in the ventral nerve cord, Tollo expression coincides temporally with periods of neural differentiation that are characterized by maximal contact between the ectoderm and neural precursor cells.

To determine whether tollo is sufficient to rescue expression of the HRP epitope in the Brd¹⁵ homozygote, a transformation construct was generated (pUASTtollo) that placedtollo-coding sequence under the control of UAS elements(Brand et al., 1994). A second transformation construct was prepared (ptolloGal4) that placed Gal4 expression under control of 2.5 kb of Drosophila genomic DNA found immediately upstream of the tollo initiation codon(Brand and Perrimon, 1993).Tollo-Gal4 transformant lines were crossed to a UAS-lacZreporter line and embryo collections were stained for β-galactosidase activity. Both in the germband extended embryo at stage 12(Fig. 7A,B) and in the lateral ectoderm of the stage 13 embryo (Fig. 7C,D), lacZ activity matched the distribution oftollo mRNA detected by in situ hybridization. Thus, the 2.5 kb of genomic DNA incorporated into the ptolloGal4 transformation vector contains control sequences sufficient to recapitulate normal Tollo expression.

UAS-tollo and tollo-Gal4 transformant lines were separately prepared in the Brd¹⁵/TM3 background. HRP-epitope expression is absent from embryos collected from lines bearing either construct alone. However, when UAS-tollo andtollo-Gal4 lines are crossed to each other, HRP-epitope expression is rescued in embryos that lack (Brd¹⁵/TM3 and TM3/TM3 genotypes) and in embryos that possess(Brd¹⁵/Brd¹⁵) the head involution defect associated with the Brd¹⁵ deletion(Fig. 8A,D,G, see arrowhead). Thus, the head involution defect is independent of HRP-epitope expression.Tollo-Gal4/UAS-tollo rescues HRP-epitope expression in the ventral nerve cord (Fig. 8B,E,H) and in the peripheral nervous system(Fig. 8C,F,I).

Other Gal4 driver lines were screened for their ability to rescue the HRP epitope in UAS-tollo transformants. Neither a pan-neural driver(ELAV-Gal4) nor a mesectodermal/midline glial driver(rhomboid-Gal4) rescued oligosaccharide expression when crossed to UAS-tollo, despite their ability to drive expression in cells that make extensive contact with neuronal surfaces. Therefore, simple juxtaposition of Tollo protein and a neuron is insufficient; induction of the neuronal HRP epitope requires Tollo expression in appropriate non-neural ectodermal cells.

Heat-shock driven expression of Tollo in all cells(hsp70-Gal4/UAS-tollo) generates early embryonic lethality that precludes assessment of HRP-epitope rescue. However, in the course of these experiments, hsp70-Gal4/UAS-tollo embryos not subjected to heat shock were also collected and stained with anti-HRP antibody. Unexpectedly, unshocked embryos older than stage 15 express the HRP epitope in the salivary gland and in sensory neurons most proximal to the gland (Fig. 8J,K). Other neuronal populations were not stained, whether in the CNS or in more posterior segments of the PNS. Thus, leaky Gal4 expression in the salivary gland(verified by UAS-lacZ reporter) is sufficient to induce the HRP epitope in a tissue that does not normally express the glycan and is able to rescue the epitope in nearby sensory neurons.

Nearly a half-century of molecular biochemistry has documented tissue-specific, cell-specific, stage-specific, and disease-specific oligosaccharide presentation (Varki,1993; Dennis et al.,1999; Lowe, 2001;Paulson and Colley, 1989;Amado et al., 1999). Despite this wealth of information, few descriptions exist of molecular mechanisms that control the specificity of glycan expression(Turkington et al., 1968;Rajput et al., 1996;Qian et al., 2001). Glycan synthesis has been modulated by manipulating various transmembrane signaling pathways, indicating that receptor-mediated events at the cell surface can influence oligosaccharide profiles (Carlow et al., 2001; Wagers and Kansas, 2000; Chen et al.,1998). Although the surface receptors that transmit these signals have not been identified, our results demonstrate that the Tollo/Toll-8 transmembrane protein influences cell-specific glycan expression in theDrosophila embryo.

The localization of Tollo expression to non-neural ectodermal cells and the rescue of neural-specific glycosylation by transgenic tollo both demonstrate that non-homologous cells modulate glycan expression in adjacent tissues. The close proximity of Tollo-expressing ectodermal cells to differentiating neurons is consistent with a molecular mechanism in which neural glycosylation is influenced by the activity of a neuronal surface receptor that directly binds ectodermal Tollo. Alternatively, the molecular activity of tollo may reside entirely within the ectodermal cell,exerting an indirect influence on neural glycosylation by propagating or attenuating instructive signals subsequently interpreted by local neurons. At present, our data cannot unambiguously distinguish whether the direct or indirect mechanism applies. However, the results of HRP-epitope rescue and Tollo misexpression studies indicate requirements that both models must satisfy.

Neuronal synthesis of the HRP-glycan is rescued inBrd¹⁵/Brd¹⁵ embryos when Tollo is expressed in its wild-type ectodermal pattern(tollo-Gal4/UAS-tollo). However,ELAV-Gal4/UAS-tollo and rho-Gal4/UAS-tolloembryos fail to rescue the HRP epitope, despite driving misexpression in neurons and glia that would present Tollo to neuronal surfaces at developmental stages coincident with the normal Tollo expression pattern. Therefore, if tollo acts directly to alter neural glycosylation,ectodermal presentation of Tollo must be unique in comparison with expression in other cell types that also share contact with neurons; either the Tollo protein requires an ectodermal-specific post-translational modification for activity or an ectodermal co-factor is necessary for appropriate presentation to neurons. If tollo indirectly affects neural glycosylation by generating or influencing paracrine signals sensed by differentiating neurons,then the cellular context in which Tollo is expressed determines induction of the HRP epitope; either the relevant paracrine influence is specifically of ectodermal origin or a required tollo intracellular signaling pathway is absent from neurons and glia.

The indirect mechanism is consistent with the function of other TLRs(Belvin and Anderson, 1996;Cantera et al., 1999;Halfon and Keshishian, 1998)and is supported by two additional observations. First, the discontinuous distribution of tollo mRNA in the neurogenic ectoderm of the ventral nerve cord indicates that tollo expression is limited to ectodermal cells contacting only a subset of the total differentiating neuron pool. Therefore, global CNS expression of the HRP epitope requires a signal unrestricted by the need for cell-cell contact. Second, HRP-epitope expression is rescued in PNS sensory neurons located near salivary glands that ectopically express Tollo, consistent with the generation of a locally active signal. Expression of the HRP epitope is not rescued in the CNS nor in more remote parts of the PNS by hsp70-Gal4/UAS-tollo, implying that temporal and physical barriers can limit Tollo activity.

The unexpected, ectopic expression of the HRP epitope in the secretory epithelium of the salivary gland indicates that some developing tissues are only one signal away from assuming an altered glycosylation phenotype. At least within the salivary gland, this result also indicates that Tollo can act directly or can generate an autocrine signal that autonomously modulates glycosylation. For neural tissue, though, elaboration of the HRP epitope is a non-autonomous neuronal behavior that requires ectodermal Tollo expression. By analogy to Toll, soluble protein ligands (like Spätzle) are prime candidates for the Tollo activator, but the full diversity of TLR ligands has yet to be characterized in any organism(Hoffman et al., 1999;Yang et al., 1998;Belvin and Anderson, 1996).

In plants and in the Drosophila adult, HRP-epitope structure has been demonstrated to contain an extensively trimmed high-mannose core carrying an α3-linked Fuc residue on the internal GlcNAc of the chitobiose.(Fabini et al., 2001;Kurosaka et al., 1991). To generate the described Drosophila HRP epitopes, high-mannose oligosaccharides must first be trimmed to a Man₃GlcNAc₂or Man₂GlcNAc₂ core. The core structure is then di-fucosylated (α3 and α6), requiring the activity of two distinct fucosyltransferases. Addition of Fuc α3 to the core requires previous and transient addition of GlcNAc to a terminal Man, yielding a di-fucosylated Man_2/3GlcNAc₂ oligosaccharide(Fabini et al., 2001;Altmann et al., 1993;Altmann et al., 1995). Therefore, trimming mannosidases, two fucosyltransferases, an N-acetylglucosaminyltransferase and a hexosaminidase constitute the minimal set of processing activities required to generate an HRP epitope. Of these activities, addition of the α3 Fuc imparts antibody recognition to the oligosaccharide.

A Drosophila fucosyltransferase that adds Fuc in α3 linkage to core GlcNAc has been characterized(Fabini et al., 2001). Designated `FucTA', the enzyme exhibits in vitro acceptor specificity appropriate for synthesis of the HRP epitope and the gene maps to 71B2, 87 kb distal to tollo. Although this lies within theBrd¹⁵ deletion, combining tollo-Gal4 with UAS-tollo results in rescue of HRP epitope expression inBrd¹⁵ homozygotes. Therefore, glycan expression is rescued by Tollo/Toll-8 in a FucTA null background. The relevance of FucTA activity to HRP-epitope expression in the embryonic nervous system remains to be determined, but Drosophila requires α3 fucosyltransferase activity and the resulting capacity to synthesize the HRP epitope in multiple contexts. Mutants that lack the HRP epitope in larval and adult stages express the oligosaccharide embryonically and epitope expression in embryonic non-neural tissue is maintained in mutants that lack the embryonic neural oligosaccharide (Snow et al.,1987; Katz et al.,1988). Thus, multiple pathways, under independent control and active in different tissues and developmental stages, lead to synthesis of the HRP epitope.

Our results suggest superficially that loss of the HRP epitope is of relatively little consequence. However, the component of the HRP epitope structure that imparts antibody recognition may be distinct from the functional domain of the oligosaccharide. Therefore, mutations in genes such as tollo, which affect specific carbohydrate expression, may not immediately reveal oligosaccharide function. The nac mutant, which lacks larval, pupal and adult expression of the HRP epitope, exhibits grossly normal nervous system morphology (Katz et al., 1988; Phillis et al.,1993). Highly penetrant axon defasciculation errors are present in the nac adult but only become apparent when afferent projections arising from discrete subsets of dye-labeled sensory neurons in the wing margin are visualized at their entry point into the central nervous system(Whitlock, 1993). Until techniques of similar resolution are applied to embryos that lack the HRP epitope, the functional significance of loss of this tissue-specific glycan cannot be fully evaluated.

In Drosophila, the HRP epitope is present on several neural proteins, many of which are also expressed in non-neural tissue where they lack the glycan (Desai et al.,1994; Snow et al.,1987; Sun et al., 1995; Wang et al., 1994). Thus, cells determine whether or not to construct the HRP epitope on a particular glycoprotein based on the tissue in which the protein is expressed, rather than on a signal intrinsic to the polypeptide. While tollo/toll-8 demonstrates that such tissue-specific glycan expression can be achieved through the activity of a Toll-like receptor, the correlation between Drosophila TLR expression and specific glycosylation patterns cannot be comprehensively assessed before glycan characterization in the Drosophila embryo is greatly expanded(Fabini et al., 2001;Seppo and Tiemeyer, 2000). Nonetheless, distributions of other TLRs exhibit spatial and temporal overlap with Tollo expression, raising the possibility that TLRs sculpt embryonic glycosylation patterns through combinatorial activation of glycosylation pathways in interacting domains of developing tissues(Hashimoto et al., 1988;Eldon et al., 1994;Chiang and Beachy, 1994;Stathopolous and Levine, 2002).

TLRs mediate pattern recognition (frequently glycan-based) as part of the innate, non-adaptive immune response in Drosophila and vertebrates(Hoffman et al., 1999;Medzhitov et al., 1997;Williams et al., 1997;Yang et al., 1998). However,only a subset of Drosophila TLRs induce defensive responses. TLR family members appear divided into clans that function in innate immunity or that fulfill developmental needs (Tauszig et al., 2000). The capacity to control glycosylation could unite the TLR family in support of a common cause, to produce appropriate spatial and temporal patterns of cell-specific glycosylation. Expressed by immune cell types that participate in tissue surveillance, TLRs are positioned to locally influence cellular glycosylation in response to pathogen, thereby coupling innate detection of non-self patterns with expression of protective glycans on host cells. In addition, further analysis of the distribution and function of TLRs may indicate that the constitutive maintenance of diverse tissue glycan profiles is generally an active process in which glycan expression is continually renewed or responsively modified by TLR-mediated signaling. In mature tissues and in the embryo, the expression of glycans must be orchestrated to coincide with the appearance of relevant carbohydrate binding proteins that mediate cell adhesion and recognition(Varki, 1993;Sharrow and Tiemeyer, 2001;Feinberg et al., 2001;Song and Zipser, 1995;Vyas et al., 2002). Therefore,broader mechanisms that impart specificity to cell-cell interactions are likely to be revealed with further characterization of the pathway by whichtollo/toll-8 controls oligosaccharide expression.

The authors are indebted to Marc Schwartz, Reed Kelso and Jay Goodman for deletion mapping; Corey Goodman for supplying antibodies; and Carl Hashimoto,Lynn Cooley and members of their respective laboratories for advice. Digital image acquisition was expertly facilitated by Peter Tiemeyer. Grant support from the NIH-NICD (HD33878), from the March of Dimes (Basil O'Connor Award),and from the Patrick and Catherine Weldon Donaghue Foundation is gratefully acknowledged. M. S. and P. M. received support from NIH-NIGMS (GM07223). A. S. was funded by the Finnish Academy and by a Postdoctoral Fellowship from the HFSPO.