Deficient Notch signaling associated with neurogenic pecanex is compensated for by the unfolded protein response in Drosophila

Cell-cell signaling mediated by the Notch (N) receptor is implicated in a wide variety of developmental processes in multicellular organisms, across phyla (Artavanis-Tsakonas et al., 1999; Kopan and Ilagan, 2009; Cau and Blader, 2009). In humans, N-signaling abnormalities cause diseases that include leukemia, other cancers, and pulmonary arterial hypertension (Ellisen et al., 1991; Nicolas et al., 2003; Li et al., 2009). Drosophila N encodes a transmembrane receptor with 36 epidermal growth factor (EGF)-like repeats in its extracellular domain (Wharton et al., 1985). During maturation of N, its extracellular domain is cleaved by Furin protease (S1 cleavage) in the Golgi (Logeat et al., 1998; Kidd and Lieber, 2002; Lake et al., 2009). After reaching the cell surface, the binding of N to its transmembrane ligand, Delta or Serrate, leads to a second cleavage in the extracellular domain of N by Kuzbanian (Kuz)/ADAM10 or ADAM17 (S2 cleavage). This cleavage removes most of the N extracellular domain and produces a membrane-tethered form of the N intracellular domain (NEXT) (Kopan and Goate, 2000). Subsequently, NEXT is cleaved within its transmembrane domain by γ-secretase (S3 cleavage), which liberates the intracellular domain, termed NICD (Mumm and Kopan, 2000). NICD then translocates to the nucleus and regulates the transcription of downstream genes (Struhl et al., 1993; Lecourtois and Schweisguth, 1995).

N requires various post-translational modifications to its extracellular domain to be activated. For example, O-glycosylation of the N extracellular domain by O-fucosyltransferase 1 (O-fut1) and Fringe regulates the binding between N and its ligands (Bruckner et al., 2000). O-fut1 is also known to act as an N-specific chaperone in Drosophila (Okajima et al., 2005). In addition, analysis of a Drosophila thiol oxidase, endoplasmic reticulum (ER) oxidoreductin 1-like (Ero1L), showed that disulfide-bond formation in the extracellular domain of N is indispensable for the activation of the N signal (Tien et al., 2008).

Many roles played by N signaling in Drosophila development are crucial and have been studied extensively. Its best-known role during the early development of the central nervous system, is to prevent cells that neighbor a neuroblast from choosing the neuroblast fate, a phenomenon called ‘lateral inhibition’ (Simpson, 1990). This is achieved when the neuroblast-fated cell activates N signaling in its neighbors; these cells become epidermoblasts. Thus, disruption of N signaling in Drosophila embryos results in the failure of lateral inhibition and the consequent hyperplasia of neuroblasts at the expense of epidermoblasts (Cau and Blader, 2009), which is referred to as the ‘neurogenic’ phenotype (Simpson, 1990). Because most of the genes that encode N-signaling components are essential for lateral inhibition, these genes were first identified by the neurogenic phenotype resulting from their disruption (Lehmann et al., 1983).

pecanex (pcx) was originally identified as a mutant showing recessive female sterility (Perrimon et al., 1984). Thus, pcx homozygous or hemizygous embryos obtained from pcx heterozygous females survive until adulthood. However, embryos obtained from pcx homozygous females mated with pcx hemizygous males, which are fertile, show neuronal hyperplasia, i.e. the neurogenic phenotype, suggesting that the maternally supplied pcx function rescues this phenotype (LaBonne and Mahowald, 1985). Therefore, pcx is considered to be a maternal neurogenic gene. pcx encodes a multi-pass transmembrane protein consisting of 3433 amino acids that is highly conserved from Drosophila to humans (LaBonne et al., 1989). A rat homolog of pcx, pecanex1, is expressed in spermatocytes and probably functions in the testes (Geisinger et al., 2005). However, no molecular function of the Pcx protein has been identified in any species. Here, we established that pcx is an N-signaling component in Drosophila. We also provide evidence that Pcx might be involved in ER functioning.

All experiments were performed at 25°C on standard Drosophila culture medium. Canton-S was used as wild type. The mutants used were: pcx³, a loss-of-function mutant (Mohler, 1977; Mohler and Carroll, 1984); Df(1)ED6574 and Df(1)ED409, deletions uncovering the pcx locus (Yan et al., 2009); N^55E11, a null mutant (Kidd et al., 1983); and Presenilin^C1 (Psn^C1), a null mutant (Lukinova et al., 1999). The Gal4 lines used were: wingless-Gal4 (wg-Gal4) (Pfeiffer et al., 2000), armadillo-Gal4 (arm-Gal4) (Sanson et al., 1996), Aloxg-Gal4 (Taniguchi et al., 2011), matα-Gal4 (Bossing et al., 2002) and MS1096 (Capdevila and Guerrero, 1994). The UAS lines used were: UAS-NICD (Go et al., 1998), UAS-pcxGFP (see below), UAS-Hsc70-3{wt} (Elefant and Palter, 1999), UAS-Hsc70-3K97S (Elefant and Palter, 1999), UAS-Xbp1-RB (Ryoo et al., 2007), UAS-endoplasmic reticulum-Cyan fluorescent protein (UAS-ER-CFP), a GFP variant with an ER-retention signal (KDEL) (BD Biosciences), and UAS-O-fut1 (Sasamura et al., 2003). hs-ΔECN expresses ΔECN under the control of a heat-shock promoter (Rebay et al., 1993). The heat-shock conditions were as described in Rebay et al. (Rebay et al., 1993). The Enhancer of split [E(spl)] m8-lacZ line carries a lacZ reporter controlled by the E(spl) m8 enhancer (Lecourtois and Schweisguth, 1995). P{Crey}1b overexpresses cre recombinase (cre) in the female germ line (Siegal and Hartl, 1996). tubP-Gal80^ts overexpresses temperature-sensitive Gal80 (Hewes et al., 2006).

The pcx homo/hemizygous embryos shown in Fig. 1B,F, Fig. 2B,D,F,H,I, Fig. 3C,E,G-M, Fig. 5B,D,G,H and Fig. 6A-H,K-M′,O-O″ were obtained by the following genetic crosses. Fig. 1B: pcx³/pcx³ × pcx³/Y; Fig. 1F: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-pcxGFP/UAS-pcxGFP; Fig. 2B: pcx³/pcx³; E(spl) m8-lacZ/E(spl) m8-lacZ × pcx³/Y; E(spl) m8-lacZ/E(spl) m8-lacZ; Fig. 2D: pcx³/pcx³ × pcx/Y; Fig. 2F: pcx³/pcx³ × pcx³/Y; Fig. 2H: pcx³/pcx³ × pcx³/Y; Fig. 2I: pcx³, cre/pcx³, cre; UAS-pcxGFP/UAS-pcxGFP × pcx³/Y; Aloxg-Gal4/Aloxg-Gal4; Fig. 3C: pcx³/pcx³; hs-ΔECN/hs-ΔECN × pcx³/Y; hs-ΔECN/hs-ΔECN; Fig. 3E: pcx³/pcx³; matα-Gal4/matα-Gal4 × pcx³/Y; UAS-NICD/UAS-NICD; Fig. 3F,G: pcx³/pcx³; hs-ΔECN/hs-ΔECN × pcx³/Y; hs-ΔECN/hs-ΔECN; Fig. 3H-J: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-NICD/UAS-NICD; Fig. 5B,D: pcx³/pcx³ × pcx³/Y; Fig. 5G: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-pcxGFP/UAS-pcxGFP; Fig. 5H: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-NICD/UAS-NICD; Fig. 6A-C: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-Xbp1-RB/UAS-Xbp1-RB; Fig. 6D,E: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-Hsc70-3{wt}/UAS-Hsc70-3{wt}; Fig. 6F-H: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-Hsc70-3K97S/UAS-Hsc70-3K97S; Fig. 6K: pcx³/pcx³; matα-Gal4/matα-Gal4 x pcx³/Y; UAS-Hsc70-3K97S/UAS-Hsc70-3K97S; Fig. 6L,L′: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-Xbp1-RB/UAS-Xbp1-RB; Fig. 6M,M′: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-Hsc70-3K97S/UAS-Hsc70-3K97S; Fig. 6O-O″: pcx³/pcx³; arm-Gal4/arm-Gal4 × pcx³/Y; UAS-O-fut1/UAS-O-fut1. pcx³ and other pcx³ derivatives carrying transgenes in other chromosomes were balanced with FM6, Bar (B). Thus, pcx³/pcx³ females and pcx³/Y males were identified based on the absence of the B phenotype.

The embryo shown in supplementary material Fig. S1A was obtained by the genetic cross pcx³/pcx³ × +/Y. Female embryos were selected by immunostaining with an anti-Sex lethal antibody (Bopp et al., 1991).

Mosaics were generated in embryos of pcx³P{Crey}1b/pcx³; Aloxg-Gal4/UAS-pcxGFP or pcx³P{Crey}1b/Y; Aloxg-Gal4/UAS-pcxGFP (Taniguchi et al., 2011; Siegal and Hartl, 1996). To obtain these embryos, we crossed pcx³P{Crey}1b/pcx³P{Crey}1b; UAS-pcxGFP females with pcx³/Y; Aloxg-Gal4 males.

To obtain full-length pcx cDNA, we performed rapid amplification of cDNA ends (RACE) with an already-known partial pcx cDNA sequence (SD01552). EGFP cDNA (Clontech) was combined in-frame with the 5′ end of the full-length pcx cDNA using PCR. The resulting fragment was inserted into the NotI site of the pUAST vector (Brand and Perrimon, 1993). pUAS-pcxGFP was introduced into the Drosophila genome using P element-mediated transformation (Brand and Perrimon, 1993).

To produce MS1096/UAS-pcxGFP; Gal80ts, we crossed MS1096 females with UAS-pcxGFP, Gal80ts males at 25°C. The second instar larvae of MS1096/UAS-pcxGFP, Gal80ts were displaced to 30°C and raised to the third instar stage. These third instar larvae were dissected and immunostained.

The antibody staining of embryos (Rhyu et al., 1994), wing imaginal discs (Matsuno et al., 2002) and S2 cells (Trammell et al., 2008) was performed as previously described. Confocal microscopy images were collected on an LSM 510 META (Zeiss) and analyzed on an LSM Image Browser. The following primary antibodies were used: rat anti-Elav (7E8A10, 1:20) (O’Neill et al., 1994), mouse anti-Elav (9F8A9, 1:20) (O’Neill et al., 1994), rat anti-GFP (GF090R, 1:200; Nacalai Tesque), mouse anti-Sex-lethal (M18, 1:20) (Bopp et al., 1991), mouse anti-β-galactosidase (Z378B, 1:100; Promega), mouse anti-Engrailed (1:25) (Braid et al., 2010), rabbit anti-active MAPK (1:200, Promega), mouse anti-RFP (1:500, MBL), mouse anti-Protein disulfide isomerase (Pdi) (1:100, Stressgen) (Vaux et al., 1990), rabbit anti-GM130 (1:50, Abcam), rabbit anti-COPII (1:100, ABR), guinea pig anti-Hrs (1:400) (Lloyd et al., 2002), rabbit anti-Rab7 (1:5000) (Tanaka and Nakamura, 2008), rabbit anti-Rab11 (1:10,000) (Tanaka and Nakamura, 2008), mouse anti-Notch intracellular domain (C17. 9C6, 1:200) (Fehon et al., 1990) and rat anti-Drosophila E-Cadherin (DE-Cad) (DCAD2, 1:10) (Oda et al., 1994). The fluorescent secondary antibodies, Cy3-conjugated goat anti-mouse (Jackson ImmunoResearch), Alexa488-conjugated goat anti-rat (Molecular Probes), Cy3-conjugated goat anti-rabbit (Jackson ImmunoResearch) and Cy3-conjugated donkey anti-guinea pig (Jackson ImmunoResearch) were used at 1:500.

Each protein sample was prepared from five embryos of wild-type or UAS-pcxGFP/arm-Gal4, incubated at 68°C for 10 minutes, and subjected to 5% SDS-PAGE and western blotting as described previously (Crevel et al., 2001). To detect PcxGFP and β-tubulin, anti-GFP (1:1000) and anti-β-tubulin antibodies (E7, 1:2000) (Wong et al., 2010) were used, respectively.

The pcx and single-minded (sim) RNA probes were prepared, and the in situ hybridization of embryos was performed as described previously (Takashima and Murakami, 2001).

For electron microscopy, specimens were prepared as described previously (Tepass and Hartenstein, 1994). These specimens were observed by electron microscopy using standard techniques, as described previously (Suzuki and Hirosawa, 1994).

Primers and total RNA of embryos were prepared as described previously (Haecker et al., 2008).

Previous studies proposed that pcx encodes an N-signaling component, based on its mutant phenotypes (Perrimon et al., 1984; LaBonne et al., 1989). However, although pcx homo/hemizygotes lacking maternal pcx show the neurogenic phenotype, the contribution of pcx to N signaling has not been examined directly (Perrimon et al., 1984). A similar neurogenic phenotype is observed in embryos homozygous for N, as described previously (Fig. 1C) (Poulson, 1937; Perrimon et al., 1984). Consistent with Perrimon’s report, we confirmed that embryos homo/hemizygous for pcx³ and lacking a maternal contribution of pcx showed the neurogenic phenotype in all cases examined (n>50) (Fig. 1B) (Perrimon et al., 1984). In the rest of this paper, these embryos are referred to as, ‘pcx^m/z embryos’. As reported previously, the maternal neurogenic phenotype of pcx was paternally rescued in 80% of the embryos (n=57) (supplementary material Fig. S1A) (LaBonne et al., 1989). Df(1)ED6574 and Df(1)ED409 are deletions lacking the pcx locus (Yan et al., 2009). We also found that Df(1)ED6574/pcx³ or Df(1)ED409/pcx³ females mated with pcx³/Y males produced embryos with the neurogenic phenotype in all cases examined (n=22, supplementary material Fig. S1B; n=24, supplementary material Fig. S1C). The extent of the neurogenic phenotype in these embryos was equivalent to that of pcx^m/z embryos, suggesting that pcx³ is a null allele.

Next, we determined the molecular lesion of the pcx³ mutant by sequencing its pcx locus. A nonsense mutation was found in the genomic DNA sequence of the pcx locus corresponding to the 2030th amino acid of the Pcx protein, which resulted in the production of a truncated Pcx protein (Fig. 1D). This mutant protein lacks the C-terminal half, which contains an evolutionarily conserved Pecanex C domain (Fig. 1D) (Gilbert et al., 1992). To confirm that the disruption of pcx functions is fully responsible for the neurogenic phenotype, we examined whether pcx overexpression in pcx^m/z embryos could rescue this phenotype. We overexpressed GFP-tagged Pcx protein (PcxGFP), and detected this protein in the extracts of UAS-pcxGFP/arm-Gal4, but not wild-type embryos, by western blotting (Fig. 1E). The neuronal hyperplasia in pcx^m/z embryos was effectively suppressed by the overexpression of pcxGFP, at 73% frequency (n=26) (Fig. 1F). This result demonstrated that the maternal neurogenic phenotype of pcx³ was caused by the mutation of the pcx gene. These results also showed that pcxGFP retains pcx’s wild-type function.

We then examined the expression pattern of pcx in embryos and imaginal discs by in situ hybridization. pcx expression was strong in the early embryos, from stage 1 to 4, and then diminished from stage 5 (Fig. 1G-J). These results are consistent with the previous finding that pcx is a maternal neurogenic gene (Perrimon et al., 1984). By contrast, no pcx expression was detected in imaginal discs (data not shown). This might explain why adult pcx homozygotes do not show obvious defects besides female sterility (Perrimon et al., 1984), although other explanations are also possible.

The maternal neurogenic phenotype associated with the pcx mutant supported the idea that pcx encodes an essential component of N signaling. However, neuronal hyperplasia can also be induced by the mutation of genes that do not contribute to N signaling directly, such as shaggy (Simpson et al., 1988). Therefore, to confirm the involvement of pcx in N signaling, we examined the expression of two N-signal target genes, m8 and single-minded (sim), in pcx^m/z embryos.

m8 is a member of the Enhancer of split [E(spl)] complex and encodes a basic helix-loop-helix protein (Delidakis and Artavanis-Tsakonas, 1992; Knust et al., 1992). The transcription of m8 is a direct target of N signaling (Bailey and Posakony, 1995).We detected m8 expression using [E(spl)] m8-lacZ, which carries a lacZ reporter controlled by the E(spl) m8 enhancer (Lecourtois and Schweisguth, 1995). As shown in Fig. 2A, m8 expression was detected in the central and peripheral nervous systems of wild-type embryos. By contrast, the m8 expression was drastically reduced in pcx^m/z embryos, in all cases examined (Fig. 2B).

sim is expressed in mesectoderm cells and is required for the specification of the midline cells that arise from them (Crews et al., 1988). Furthermore, sim expression depends on the activation of N signaling in these cells (Hong et al., 2008). In wild-type cells, sim expression was detected by in situ hybridization in a single row of mesectoderm cells in the lateral half of each embryo, as reported previously (Fig. 2C) (Morel and Schweisguth, 2000). We found that the row of cells expressing sim was severely interrupted in pcx^m/z embryos, in all cases examined (n=22), indicating that sim expression was reduced in these embryos (Fig. 2D). These results indicate that pcx has an essential role in N signaling.

We also investigated whether other signaling pathways were affected by the absence of pcx function. To examine the activity of Wnt signaling, we detected the expression of engrailed (en), a target gene of Wnt signaling, by anti-En antibody staining (White et al., 1998). The En expression in pcx^m/z embryos was not significantly different from that observed in wild type (Fig. 2E,F). We also examined the activity of receptor tyrosine kinase signaling pathways, which can be detected by anti-phosphorylated MAPK antibody staining (Peri et al., 1999). The intensity of the anti-phosphorylated MAPK antibody staining was almost the same in the wild-type embryos and the pcx^m/z embryos (Fig. 2G,H). The results above indicate that the function of pcx might be specifically required for the activation of N signaling.

To understand the role of Pcx in N signaling, we examined whether N activation depended on Pcx activity in the signal-receiving or the signal-sending cells, by determining whether the pcx function was cell-autonomous. We recently developed a modified Cre/loxP system to efficiently induce somatic mosaic clones in Drosophila embryos (Taniguchi et al., 2011). The Aloxg-Gal4 line drives Gal4 expression as a consequence of Cre-mediated cis-recombination between its two loxP sites, leading to the overexpression of UAS-pcxGFP in clonal cells in pcx^m/z embryos. The cells comprising the mosaic clones expressing pcxGFP did not assume a neuronal fate, whereas the pcx^m/z embryonic cells surrounding them cells became neurons (100%, n=11) (Fig. 2I-L). This result suggests that the pcxGFP-expressing cells, in which the level of N signaling was higher than in the surrounding pcx^m/z cells, preferentially differentiated into epithelial cells. We also noted that the clones overexpressing pcxGFP frequently formed circular clusters, which indicated a global change in tissue architecture, probably because the cell-adhesion property was different between the neurons and epithelial cells. Thus, although these results need to be interpreted cautiously, this potential cell-autonomous behavior of the pcx gene might support the hypothesis that pcx is required in the signal-receiving cells.

To elucidate how Pcx contributes to N signaling, we examined whether various forms of N, including the membrane-tethered form of activated N (ΔECN) and the nuclear form of activated N (NICD) (Fig. 3A), could activate N signaling in embryos lacking pcx function. In wild-type embryos, ubiquitous expression of ΔECN (matα-Gal4/UAS-ΔECN) (82%, n=17) (Fig. 3B) or NICD (matα-Gal4/UAS-NICD) (91%, n=22) (Fig. 3D) resulted in the ectopic expression of sim in a few cells neighboring the row of mesectoderm cells that expressed sim endogenously, as reported previously (Morel and Schweisguth, 2000).

Although sim expression was severely reduced in pcx^m/z embryos (Fig. 2D), its expression was increased by the overexpression of ΔECN (91%, n=21) (Fig. 3C) or NICD (94%, n=16) (Fig. 3E) in pcx^m/z embryos. Thus, we speculated that NICD and ΔECN are epistatic to pcx. However, we also noted that NICD’s ability to induce ectopic sim expression was reduced in pcx^m/z embryos, compared with wild type (Fig. 3D,E). Therefore, it is possible that Pcx plays some role(s) downstream of NICD. However, a similar reduction in NICD’s ability to induce ectopic sim expression was observed in N homozygotes (supplementary material Fig. S1D), suggesting that this reduction might not be a specific effect in pcx^m/z embryos. Based on these results, we speculated that Pcx functions upstream of ΔECN and NICD, although we could not exclude the possibility that Pcx also functions downstream of NICD.

To confirm these results, we also examined whether ΔECN and NICD could rescue the neurogenic phenotype associated with the absence of pcx function. As reported previously, in wild-type embryos, the overexpression of NICD (93%, n=14) suppressed neuronal differentiation, which is called the ‘anti-neurogenic phenotype’ (supplementary material Fig. S1E) (Lieber et al., 1993). Overexpression of ΔECN (90%, n=20) (Fig. 3F,G) or NICD (92%, n=20) (Fig. 3H-J) suppressed the neurogenic phenotype of pcx^m/z embryos. This result was compatible with the idea presented above that Pcx functions upstream of ΔECN and NICD. However, these analyses did not clarify whether N, Dl or both were affected in the pcx^m/z embryos.

To gain insight into the biochemical roles of Pcx, we examined its subcellular localization. Because it has been difficult to obtain a specific antibody against Pcx, we decided to study the subcellular localization of PcxGFP, which rescued the maternal neurogenic phenotype of pcx³, as described above (Fig. 1E,F).

We expressed pcxGFP driven by arm-Gal4 in the neuroectoderm of embryos at stage 14, because pcx^m/z embryos exhibit their phenotype in this tissue and at this stage. Under these conditions, PcxGFP mostly colocalized with the Pdi-positive ER (Fig. 4A). By contrast, we did not detect the colocalization of PcxGFP with markers of the Golgi (GM130), ER-Golgi intermediate compartment termed ERGIC (COPII), or endocytic compartments, including the early endosomes (Hrs), late endosomes (Rab7) and recycling endosomes (Rab11) (Fig. 4B-F).

To reduce the possibility that the overexpression of pcxGFP led to the mislocalization of its product, we drove the pcxGFP expression weakly, using the Gal4-UAS system in combination with a temperature-sensitive Gal 80, Gal80^ts, in the wing imaginal discs of third instar larvae (Hewes et al., 2006). Under these conditions, we could control the expression level of pcxGFP so that it was just above the detection limit of the product by immunostaining (data not shown). Consistent with the results obtained in embryos, PcxGFP was mostly colocalized with an ER marker (Pdi), but not with other markers (supplementary material Fig. S2A-F) (McKay et al., 1995; Bobinnec et al., 2003). These results suggest that Pcx might be an ER-resident protein, although the localization of the endogenous protein still needs to be determined.

The specific localization of Pcx-GFP to the ER suggested that its function might be ER-related. Protein disulfide isomerase (Pdi) is located specifically in the ER (Roth and Pierce, 1987; Koivu and Myllyla, 1987). To detect possible defects in the ER, wild-type embryos and pcx^m/z embryos were stained with an anti-Pdi antibody. The ER appeared to be normal in pcx^m/z embryos at stage 5 (Fig. 5A-B′, top). However, an abnormality of the ER was observable in these embryos at stage 9, when neuroblast segregation is just starting (Fig. 5A-B′, middle) (Hartenstein and Campos-Ortega, 1984). Enlarged ER was observed predominantly in the region corresponding to the dorsal epidermis of the wild-type embryos at stage 14 (Fig. 5A-B′, bottom) (Bokor and DiNardo, 1996). These observations suggest that the sensitivity to the absence of pcx function might vary among cells. A similar enlargement was detected in pcx^m/z embryos expressing KDEL-CFP, which encodes CFP with an ER retention signal (supplementary material Fig. S3A-B′). To confirm the surplus of ER in the cells of pcx^m/z embryos, we also observed these embryos by electron microscopy. Consistent with the results above, an overabundance of ER was observed in the electron microscopy images (Fig. 5C,D). By contrast, the structures of the ER-Golgi intermediate compartment (ERGIC), the Golgi, and the endocytic compartments, including the early endosomes (Rab5), late endosomes (Rab7) and recycling endosomes (Rab11) were normal in these embryos (supplementary material Fig. S3C-L′).

The abnormal ER phenotype was not observed in embryos homozygous for N or in embryos lacking zygotic and maternal Psn, an N-signaling component (Fig. 5E-F′), although these embryos showed the neurogenic phenotype (Simpson, 1990; Ye et al., 1999). These results indicate that the enlargement of the ER was not due to the disruption of N signaling. We also found that this ER defect was rescued by the overexpression of pcx-GFP in pcx^m/z embryos, indicating that this phenotype was induced by the lack of pcx function (Fig. 5G,G′). Therefore, we speculated that the enlargement of the ER might be the reason for the severe reduction in N signaling in these embryos. However, the overexpression of NICD also rescued the ER enlargement in these embryos (Fig. 5H,H′), even though a disruption of N signaling did not cause the ER enlargement and pcx is required upstream of NICD. Therefore, we speculate that the ectopic activation of N signaling by the overexpressed NICD also influences the structure of the ER.

Based on the fact that Pcx was mostly detected in the ER and that defects in the ER structure were found in pcx^m/z embryos, we next examined whether Pcx participates in some biological process that takes place in the ER. For example, the proper folding of N is known to be essential for its activation. The formation of disulfide bonds in the extracellular domain of N in Drosophila is known to require Ero1L, which has a broad role in protein folding in yeast (Tien et al., 2008).

When unfolded or misfolded proteins appear in the ER, the organelle suffers a type of stress (ER stress) that can induce apoptosis (Lee, 1987; Gething and Sambrook, 1992; Li et al., 2006). Cells suffering from ER stress can reduce the stress level through a response known as the unfolded protein response (UPR) (Kaufman, 1999). In stressed cells, Ire-1, which is an ER-tethered endonuclease that acts as a sensor of ER stress, splices the Xbp1 mRNA, which encodes a basic helix-loop-helix (bHLH) transcription factor; this factor then activates the transcription of genes that promote protein folding (Lee et al., 2003; Yoshida et al., 2003). Therefore, the ectopic production of spliced Xbp1 (Xbp1-RB) mRNA results in the transcriptional induction of these genes (Back et al., 2006). We produced Xbp1-RB mRNA in pcx^m/z embryos and observed that the neurogenic phenotype was effectively suppressed (Fig. 6B,C,I). In 31% (n=21) of these embryos, the metameric structures of the central nervous system were restored (Fig. 6C,I).

Heat-shock cognate 70-3 (Hsc70-3) encodes Drosophila binding protein (Bip), a major chaperone that recognizes misfolded proteins in the lumen of the ER (Rubin et al., 1993). We found that the overexpression of wild-type Hsc70-3, Hsc70-3{wt}, weakly rescued the neurogenic phenotype of pcx^m/z embryos (67%, n=12) (Fig. 6D,E,I). A dominant-negative form of Hsc70-3, Hsc70-3K97S, induces the UPR through the activation of Xbp1 (Elefant and Palter, 1999). Overexpression of Hsc70-3K97S suppressed the neurogenic phenotype of pcx^m/z embryos as efficiently as Xbp1-RB (Fig. 6F-I). In 29% of these embryos (n=21), an almost normal central nervous system was observed (Fig. 6H,I). To confirm that the rescue of the neurogenic phenotype by the upregulation of the UPR was due to the recovery of N signaling, we examined the expression of sim (Hong et al., 2008). The expression of sim in mesectoderm cells was restored by the overexpression of Hsc70-3K97S in pcx^m/z embryos, at 76% frequency (n=21) (Fig. 6J,K). These results suggested that the attenuation of N signaling associated with the absence of pcx function could be restored by the upregulation of some ER functions that are under control of the UPR.

Next, we examined whether the enlargement of the ER in pcx^m/z embryos was rescued by inducing the UPR. The ER enlargement was efficiently suppressed by the overexpression of Xbp1-RB (74%, n=21) or Hsc70-3K97S (67%, n=21) in these embryos (Fig. 6L-M′). Therefore, the condition causing the enlargement of the ER in the absence of pcx function, which could be responsible for the disruption of N signaling, is restored by the upregulation of the UPR.

To understand the possible roles of pcx in the UPR, we detected the spliced form of endogenous Xbp-1 (Xbp1-RB) mRNA, which reflects the activation of the UPR in vivo, by RT-PCR (Haecker et al., 2008). As reported previously (Elephant and Palter, 1999), overexpression of UAS-Hsc70-3K97S, which encodes a dominant-negative form of Hsc70 driven by arm-Gal4, induced the production of Xbp1-RB mRNA (Fig. 6N, lane 4). In pcx^m/z embryos, the overexpression of UAS-Hsc70-3K97S also induced Xbp1-RB mRNA as efficiently as in the wild-type background (Fig. 6N, lane 5). These results suggest that pcx is not essential for induction of the UPR. We also found that overexpressing pcx did not induce an ectopic UPR in wild-type embryos (Fig. 6N, lane 3). However, we detected a weak but reproducible induction of Xbp1-RB mRNA production in pcx^m/z embryos (Fig. 6N, lane 2). Thus, an ectopic UPR might be induced in pcx^m/z embryos, although ectopic apoptosis was not detected in these embryos (data not shown).

The UPR increases chaperone activities in the ER (Lee et al., 2003). Thus, it is possible that upregulation of the UPR could restore the defective folding of N in pcx^m/z embryos. To test this possibility, we overexpressed O-fut1, which has an N-specific chaperone activity, in pcx^m/z embryos. We found that the neurogenic phenotype and the ER enlargement were suppressed in 31% (n=29) of the pcx^m/z embryos overexpressing O-fut1 (Fig. 6O-O″), although the suppression was less efficient than that seen with the ectopic induction of the UPR. Although further analysis is required, this result is consistent with the idea that the disruption of N’s folding might account, in part, for the attenuation of N signaling in pcx^m/z embryos. It is conceivable that the transportation of N, and possibly Dl, was disrupted in pcx^m/z embryos. However, N and Dl were properly localized to the apical region of the epithelial cells where Drosophila E-cadherin was accumulated at stage 5, when the expression of sim had already started (supplementary material Fig. S4). These results suggest that the transportation of N was not severely disrupted in pcx^m/z embryos, even if the folding of N was disrupted.

The Pcx family proteins are evolutionarily conserved, large transmembrane proteins with multi-pass transmembrane domains (LaBonne et al., 1989). However, no motifs that might suggest Pcx’s biochemical function have been found in its amino acid sequence. Although pcx was previously suggested to be involved in N signaling, based on the neurogenic phenotype associated with its mutant in Drosophila, this possibility had not been explored. In this study, we provide evidence that Pcx is a component of the N-signaling pathway.

In pcx^m/z embryos, the ER was abnormally enlarged. Various factors regulating the architecture of the ER have been identified. In Drosophila, Atlastin, a dynamin-like GTPase, is required for fusion of the ER membrane (Orso et al., 2009). Thus, the overexpression of Atlastin induces an enlarged ER (Orso et al., 2009). In addition, the peripheral ER shows two distinct structures: tubules and sheets (Puhka et al., 2007). Several factors organizing the shape of the ER membrane into tubules or sheets have been identified (English et al., 2009). Therefore, Pcx might contribute to the regulatory machinery that accomplishes the normal organization of the ER.

In pcx^m/z embryos, the enlarged ER was observed predominantly in the region corresponding to the dorsal epidermis of wild-type embryos (Bokor and DiNardo, 1996). Therefore, sensitivity to the absence of pcx function might differ among groups of cells. This distinct behavior could reflect differences in the cell-cycle phase or level of UPR activity.

Although our results showed that the reduction of N signaling was not responsible for the enlargement of the ER in pcx^m/z embryos, the ectopic activation of N signaling by overexpression of NICD also suppressed this ER defect. We speculate that the ectopic activation of N signaling might affect the progression of the cell-cycle or the level of UPR, which could in turn affect the regulation of the ER architecture. It has been shown that N signaling directly or indirectly affects the cell cycle (Johnston and Edgar, 1998; Simon et al., 2009). However, the biological significance and mechanisms of this phenomenon remain elusive.

We found that induction of the UPR suppressed the ER enlargement in pcx^m/z embryos. The suppression of the ER enlargement by the expression of genes that induce the UPR coincided with the rescue of N signaling activity in these embryos. Therefore, the reduced N signaling in pcx^m/z embryos might be attributable to the enlargement of the ER. However, we cannot exclude the possibility that pcx is independently involved in the activation of N signaling and the regulation of the ER architecture. Nevertheless, our results suggest that some downstream events induced by the UPR compensate for the defect of N signaling associated with the absence of pcx function. We found that overexpression of O-fut1, an N-specific chaperone, partially compensated for the loss of pcx function (Okajima et al., 2005). Thus, a disruption of N signaling in the absence of pcx function might be partly due to the mis-folding of N, which is consistent with our hypothesis that pcx acts upstream of the activated forms of N and probably functions in signal-receiving cells.

The UPR induces various downstream events, including the attenuation of protein synthesis, the enhancement of misfolded ER protein degradation, and the induction of genes encoding various chaperones (Kaufman, 1999). Therefore, in future experiments, it will be important to determine the specific defects that are compensated for by the UPR in the absence of pcx function.

We thank the Bloomington Drosophila Stock Center (Indiana) and the Developmental Studies Hybridoma Bank (University of Iowa).