N-linked glycosylation is a prevalent protein modification in eukaryotic cells. Although glycosylation plays an important role in cell signaling during development, a role for N-linked glycosylation in embryonic patterning has not previously been described. In a screen for maternal factors involved in embryo patterning, we isolated mutations in Drosophila ALG5, a UDP-glucose:dolichyl-phosphate glucosyltransferase. Based on the embryonic cuticle phenotype, we designated the ALG5 locus wollknäuel(wol). Mutations in wol result in posterior segmentation phenotypes, reduced Dpp signaling, as well as impaired mesoderm invagination and germband elongation at gastrulation. The segmentation phenotype can be attributed to a post-transcriptional effect on expression of the transcription factor Caudal, whereas wol acts upstream of Dpp signalin by regulating dpp expression. The wol/ALG5 cDNA was able to partially complement the hypoglycosylation phenotype of alg5mutant S. cerevisiae, whereas the two wol mutant alleles failed to complement. We show that reduced glycosylation in wolmutant embryos triggers endoplasmic reticulum stress and the unfolded protein response (UPR). As a result, phosphorylation of the translation factor eIF2α is increased. We propose a model in which translation of a few maternal mRNAs, including caudal, are particularly sensitive to increased eIF2α phosphorylation. According to this view, inappropriate UPR activation can cause specific patterning defects during embryo development.
A common protein modification in eukaryotic cells is N-linked glycosylation, which occurs on the majority of proteins synthesized in the endoplasmic reticulum (ER), where a pre-assembled oligosaccharide chain is transferred to the nascent polypeptide(Helenius and Aebi, 2004). Protein glycosylation has several purposes: it is needed for proper folding and quality control in the ER, it targets some proteins to different cellular compartments, and it can affect protein function. Accumulation of unfolded proteins within the ER triggers the unfolded protein response (UPR)(Zhang and Kaufman, 2004). This response increases the folding capacity of the ER and decreases the folding demand via three ER transmembrane proteins, IRE1 (inositol-requiring 1), ATF6 (activating transcription factor 6) and PERK (PKR-like endoplasmic reticulum kinase). Activation of IRE1 and ATF6 causes transcriptional activation of genes needed for folding in the ER, whereas PERK activation results in a general decrease in translation initiation and selective translation of specific mRNAs through phosphorylation of eIF2α.
Transfer of Met-tRNAi to the 40S ribosomal subunit is accomplished by GTP-bound eIF2 (Proud,2005). Following recognition of the AUG start codon, GTP is hydrolyzed and the eIF2-GDP complex released from the ribosome. Exchange of GDP for GTP is mediated by eIF2B, and is regulated by phosphorylation of the alpha subunit of eIF2 at a conserved serine residue, which generates a competitive inhibitor of eIF2B.
In this work, we show that mutations in wollknäuel, a UDP-glucose:dolichyl-phosphate glucosyltransferase involved in N-linked protein glycosylation, disrupts Drosophila embryo development by affecting the expression of a few key gene regulators. Reduced glycosylation efficiency in wol mutant embryos triggers the UPR. As a result,phosphorylation of eIF2α is increased. We propose that some mRNAs are more sensitive to eIF2α phosphorylation than others, and that this causes specific patterning defects.
MATERIALS AND METHODS
Fly stocks, P-element transformation and germline clones
Oregon-R or w1118 were used as wild-type controls. The wol alleles 2L-284 (wol1) and 2L-267 (wol2), as well as the cad allele 2L-264,were generated on an FRT2L-40A-containing chromosome in a germline clone EMS screen performed in Tübingen(Luschnig et al., 2004).
A P-element plasmid containing a 2.8 kb wol/ALG5 genomic region was constructed by PCR amplification from genomic DNA, cloned into pCaSpeR4(Thummel and Pirrotta, 1992),and injected into w1118 embryos according to standard procedures. Two insertions on the X-chromosome were used in rescue experiments.
Germline clones were produced as described(Qi et al., 2008). Males containing a transgene misexpressing dpp in the Krüppel(Kr) domain (gift from Hilary Ashe, University of Manchester, UK)were crossed to wol1 germline clone females. Expression of dpp was activated by FLPing out transcriptional stop signals downstream of the Kr promoter(Struhl et al., 1993).
Cuticle preparation, in situ hybridization and immunofluorescence
Cuticles were prepared as described by Wieschaus and Nüsslein-Volhard(Wieschaus and Nüsslein-Volhard,1998) and examined using dark-field microscopy. Whole-mount RNA in situ hybridization using digoxigenin-labeled probes was performed as described previously (Jiang et al., 1991; Tautz and Pfeifle, 1989). Double labeling was performed as described(Kosman and Small, 1997).
Immunohistochemistry and RNA/protein double-staining protocols are modified from Manoukian and Krause (Manoukian and Krause, 1992). Rabbit anti-Eve (1:1000 dilution) was provided by Mike Levine (University of California, Berkeley, CA), guinea pig anti-Cad(1:800) and rat anti-Hb (1:400) antibodies are described by Kosman et al.(Kosman et al., 1998) and were provided by John Reinitz and Steve Small.
Complementation tests with deficiencies were used to map the wolalleles, which failed to complement the deficiency Df(2L)TE29Aa-11. We developed SNP markers from the intergenic regions in the 260 kb interval of this deficiency. PCR products that could be distinguished by restriction fragment length polymorphisms were identified, and used in high-resolution recombination mapping with P-element strains flanking the deficiency. The wol1 allele in a white- background was crossed to the white+ P-elements l(2)k16919, located proximal to the deficiency, or l(2)k14902, positioned distal to the deficiency. We recovered 25 white-eyed recombinants out of ∼20,000 flies from the cross with the proximal P-element, and four recombinants out of∼5000 flies from the distal P-element. Genomic DNA was prepared from the recombinants, PCR amplified with SNP marker primers, and analyzed after restriction digest. One recombination event with the l(2)k16919 chromosome occurred distal to a SNP located 162 kb into the deficiency. Two out of the four recombinants with l(2)k14902 had recombined proximal to a SNP located 210 kb into the deficiency.
Genomic DNA was isolated from wol1 and wol2 homozygous mutant larvae. Exonic sequences from the genes in this interval were amplified by PCR, sequenced and compared with an FRT2L-40A chromosome from another mutant. We found mutations in the CG7870/ALG5 gene, and confirmed that lethality maps to the ALG5 locus by complementation tests with a PiggyBac insertion (PBac RBe04276, Fig. 3)that became available during the course of this work.
Saccharomyces cerevisiae strains used were derivates of YG91(Matα ade2-101 ura3-52 his3-200Δ alg5::HIS3) and YG355 (Mat ade2-101 ura3-52 his3-200Δ alg5::HIS3 wbp1-2) (Burda et al., 1996). Standard yeast media and genetic techniques were used (Guthrie and Fink,1991).
The Drosophila ALG5 cDNA was RT-PCR amplified from embryo mRNA,TA-cloned and sequenced. The wol1 and wol2 mutations were introduced by site-directed mutagenesis (Quick-change Kit, Stratagene). Blunt-ended cDNAs were cloned into the yeast expression vector pCFZ41-GPD(Mumberg et al., 1995).
Analysis of carboxypeptidase Y (CPY) in ALG5-deficient cells(YG91) was detected by western blot analysis as described(Burda et al., 1996).
Western blot and RT-PCR
Protein extracts from 2×106 untreated, 10 mM DTT-treated or eIF2α RNAitreated S2 cells, or from 30 μl 2- to 4-hour w1118 or wol1 germline clone embryos were prepared as described (Lilja et al.,2007). Proteins (7 μg) were separated by SDS-PAGE, transferred to PVDF membrane (GE Healthcare), and incubated with a rabbit phospho-eIF2α (Ser51) antibody (1:1000, Cell Signaling Technology). The membrane was re-probed with a rabbit anti-human eIF2α (residues 50-150)antibody (1:200, Abcam). HRP-coupled secondary antibodies were visualized by ECL (GE Healthcare).
Total RNA was isolated from 0- to 3-hour w1118 and wol1 germline clone embryos using TRIzol reagent(Invitrogen). PolyA-RNA was extracted from total embryo RNA as well as from mock-treated, DTT-treated and RNAi-treated cells using Dynabeads mRNA DIRECT Microkit (DYNAL), followed by reverse transcription with Superscript II(Invitrogen). Primers flanking the unconventional splice site in xbp1mRNA (Xbp1_F, 5′-CGCCAGCGCAGGCGCTGAGG-3′ and Xbp1_R,5′-CTGCTCCGCC-AGCAGACGCGC-3′) were used in 25 PCR cycles. RNAi against eIF2α was performed as described(Qi et al., 2008).
RESULTS AND DISCUSSION
Segmentation defects in wollknäuel mutant embryos
From a screen for maternal genes that are required for Drosophilaembryo development (Luschnig et al.,2004), we selected mutants that affect segmentation gene expression. We identified two mutants with defects in posterior embryo patterning, 2L-284 and 2L-267, that failed to complement each other and are thus allelic. Cuticle preparations of embryos derived from mothers harboring 2L-284 or 2L-267 germline clones revealed segmentation defects in the posterior half of the embryo, and a curled-up phenotype resulting from defects in germband elongation and retraction (Fig. 1B,C). The cuticle phenotype resembles a ball of wool,Wollknäuel in German, and we therefore named our alleles wollknäuel 1 and 2 (wol1 and wol2).
We stained embryos derived from wol germline clones (hereafter called wol embryos) with an engrailed (en) RNA probe (Fig. 1D-F). In the posterior, every second en stripe was missing. Double staining for Even skipped (Eve) protein and en RNA showed that the missing stripes correspond to odd-numbered stripes (Fig. 1G,H). Expression of en regulators was altered in wol embryos (see Fig. S1 in the supplementary material). We therefore examined gap gene expression, as gap proteins control expression of the en regulators. The posterior giant (gt) expression domain was severely reduced (Fig. 1K), although it recovered at later embryo stages (not shown), and the posterior knirps (kni) stripe was shifted posteriorly in wol embryos (Fig. 1, compare M with N).
A major activator of gt and kni expression is Caudal(Cad), which forms a posterior-anterior protein gradient, being translationally repressed by Bicoid in the anterior part of the embryo(Rivera-Pomar and Jackle,1996). Whereas the maternal cad RNA was present in normal amounts in wol embryos (Fig. 1,compare O with P), we found much less Cad protein in wolembryos than in the wild type (Fig. 1Q,R). This result suggests that either Cad protein stability or the efficiency of Cad translation is affected by the wol mutations. Segmentation gene expression in wol embryos was very similar to that found in embryos derived from cad germline clones(Fig. 1I,L). There was no failure in either the terminal system or in the posterior system, both of which control gap gene expression (see Fig. S2 in the supplementary material). We therefore favor the notion that reduced Caudal levels cause the segmentation phenotype in wol embryos.
wol is required for dorsal-ventral patterning and gastrulation movements
We investigated dorsal-ventral patterning in wol embryos by examining rhomboid (rho) expression. The rho gene is expressed in two ventrolateral bands in response to the protein Dorsal(Stathopoulos and Levine,2002). In addition, rho is activated in dorsal cells by signaling from the TGF-β protein Decapentaplegic (Dpp)(Fig. 2A, arrow). In wol mutant embryos, the Dpp-dependent rho expression pattern was selectively affected (Fig. 2B). We overexpressed Dpp from a Kr enhancer in a central domain of transgenic embryos and monitored Dpp activity through another downstream target gene, Race (also known as Ance)(Fig. 2C)(Rusch and Levine, 1997). As expected, Race mRNA was absent from wol mutant embryos(Fig. 2D). However, in wol embryos expressing ectopic Dpp (purple in Fig. 2E), Raceexpression was restored (brown in Fig. 2E). From this result, we conclude that wol activity is not needed for transduction of the Dpp signal, but is acting upstream of Dpp signaling. We found that expression of dpp itself is reduced in wol embryos as compared with wild type(Fig. 2F,G), which probably explains the failure to activate Dpp target genes. It appears that an unknown maternal regulator of dpp expression or, alternatively, mRNA stability is dependent on wild-type wol activity.
We also noted problems with cell movements during gastrulation in wol mutant embryos. As a result, germband elongation does not proceed normally (Fig. 1 and data not shown), and mesoderm invagination is disturbed (see Fig. S3 in the supplementary material).
wol encodes a UDP-glucose:dolichyl-phosphate glucosyltransferase
Using SNP markers, we mapped the wol locus to a 48 kb interval on chromosome 2. We sequenced the exons from the nine genes in this interval from wol homozygous larvae. In the wol1 allele we found an A-to-T transversion, and in the wol2 allele a G-to-A transition in the gene annotated as CG7870(Fig. 3A). CG7870 is predicted to encode a 326 amino acid protein - the UDP-glucose:dolichyl-phosphate glycosyltransferase ALG5 (Heesen et al.,1994). In wol1, there is a R209W substitution within the glycosyltransferase domain, whereas W316 is changed to a stop in wol2. The lethality of wol mutants could be rescued by a transgene containing the ALG5 genomic locus. Whole-mount in situ hybridization showed that wol RNA is maternally contributed and expressed zygotically in the salivary glands and proventriculus(Fig. 3B-D), tissues where a lot of protein secretion takes place.
The ALG5 enzyme is involved in N-linked protein glycosylation by transferring glucose from UDP-glucose to dolichyl-phosphate, a lipid residing in the ER membrane (Runge et al.,1984). This glucose is added to the oligosaccharide chain that is assembled in the ER prior to transfer to the nascent polypeptide by oligosaccharyl transferase (Helenius and Aebi, 2004). Substrate recognition by oligosaccharyl transferase is greatly diminished in the absence of terminal glucoses. Mutations that disrupt ALG5 function are therefore expected to cause an accumulation of hypoglycosylated proteins.
To investigate whether the wol1 and wol2 mutations affect the function of ALG5, we performed a complementation assay in Saccharomyces cerevisiae. As previously shown, mutations in yeast alg5 lead to hypoglycosylation of secreted proteins (Heesen et al., 1994). This can be assayed by processing of the carboxypeptidase Y (CPY) protein, a vacuolar protein with four N-linked oligosaccharides. In aΔ alg5 yeast strain, CPY glycoforms lacking one or two oligosaccharide chains accumulate (Fig. 3E). Introduction of yeast or Drosophila ALG5 cDNA into the Δalg5 strain restored the glycosylation phenotype of CPY,whereas cDNAs with the wol1 or wol2mutations failed to do so (Fig. 3E). In a yeast growth assay, ALG5 cDNAs with the wolmutations only weakly rescued, or failed to rescue, the growth phenotype (see Fig. S4 in the supplementary material). These results suggest that the wol mutations either impair the catalytic activity of the ALG5 protein or lead to protein destabilization.
The unfolded protein response is triggered in wol mutant embryos
An important function of N-linked glycosylation is to aid the folding of proteins in the ER (Helenius and Aebi,2004). Accumulation of unfolded proteins in the ER induces the UPR that activates the IRE1 endoribonuclease. In Drosophila, this leads to the removal of a 23 bp intron from xbp1 mRNA in the cytoplasm(Plongthongkum et al., 2007; Ryoo et al., 2007; Souid et al., 2007), which generates a translational frameshift that gives rise to transcriptionally active Xbp1 protein (Plongthongkum et al.,2007).
We examined xbp1 splicing in wol mutant embryos by RT-PCR. Fig. 4A shows that one band with the size expected from unspliced xbp1 mRNA is obtained from wild-type embryos, whereas both spliced and unspliced products were detected in wol embryos. To confirm that this band corresponds to ER-stress-induced xbp1 splicing, we isolated mRNA from untreated,tunicamycin- or DTT-treated S2 cells. Tunicamycin inhibits N-linked glycosylation, whereas DTT prevents thioester bond formation, and both treatments generate unfolded proteins in the ER. As shown in Fig. 4A, these drug treatments resulted in more of the smaller, spliced xbp1 mRNA than was found in untreated S2 cells. We conclude that the UPR is triggered by tunicamycin and DTT, as well as by mutations in wol.
Another branch of the UPR is activation of the kinase PERK that results in eIF2α phosphorylation and attenuation of translational initiation(Kaufman, 2004). Drosophila PERK (also known as PEK) is maternally contributed to the embryo (Pomar et al., 2003). As shown by the western blot in Fig. 4B, a 1.5-fold increase in eIF2α phosphorylation was detected in wol embryos as compared with wild type, whereas the total amount of eIF2α remained unchanged. As a control, we prepared protein extracts from S2 cells treated with DTT or with eIF2αdouble-stranded RNA. eIF2α phosphorylation was increased by DTT, whereas both the eIF2α and the phospho-eIF2α bands disappeared in extracts from eIF2α RNAi-treated cells (Fig. 4B).
Taken together, these results confirm that the UPR is activated in wol mutants, and indicate that translation might be attenuated in wol embryos. We propose that this causes the observed patterning defects. According to this model, reduced maternal wol activity would lead to accumulation of unfolded proteins in the ER in early embryos, with a consequent transient increase in eIF2α phosphorylation. Translation of selected maternal mRNAs, including cad and the activator of dpp expression, would be particularly sensitive to increased eIF2α phosphorylation. Reduced amounts of these transcription factors result in disruption of posterior segmentation and of dorsal-ventral patterning. Although the UPR plays important developmental and physiological roles in C. elegans, Drosophila and mammals(Ryoo et al., 2007; Shen et al., 2001; Shen et al., 2005; Souid et al., 2007; Wu and Kaufman, 2006), this is the first report to indicate that inappropriate UPR activation may disrupt embryonic patterning.
Supplementary material for this article is available at http://dev.biologists.org/cgi/content/full/135/10/1745
We thank Hilary Ashe, John Reinitz and Steve Small for providing reagents,and Monika Björk at the WCN fly facility for embryo injections. A.H. was supported by a Wenner-Gren fellowship, and grants from the Swedish Research Council to M.M. supported this work.