Xbp1-independent Ire1 signaling is required for photoreceptor differentiation and rhabdomere morphogenesis in Drosophila

SUMMARY The Unfolded Protein Response (UPR) is composed by homeostatic signaling pathways that are activated by excessive protein misfolding in the endoplasmic reticulum (ER). Ire1 signaling is an important mediator of the UPR, leading to the activation of the transcription factor Xbp1. Here, we show that Drosophila Ire1 mutant photoreceptors have defects in the delivery of Rhodopsin-1 to the rhabdomere and in the secretion of Spacemaker/Eyes shut into the inter-rhabdomeral space. However, these defects are not observed in Xbp1 mutant photoreceptors. Ire1 mutant retinas have higher mRNA levels for targets of regulated Ire1-dependent decay (RIDD), including for the fatty acid transport protein (fatp). Importantly, downregulation of fatp by RNA interference rescues the Rhodopsin-1 delivery defects observed in Ire1 mutant photoreceptors. Our results show that the role of Ire1 during photoreceptor differentiation is independent of Xbp1 function and demonstrate the physiological relevance of the RIDD mechanism in this specific paradigm.


INTRODUCTION
The endoplasmic reticulum (ER) is the cell organelle where secretory and membrane proteins are synthesized and folded. When the folding capacity of the ER is impaired, the presence of incorrectly folded (misfolded) proteins in the ER causes ER stress and activates the Unfolded Protein Response (UPR), which helps to restore homeostasis in the ER (Ron and Walter, 2007;Walter and Ron, 2011). In higher eukaryotes, the activation of the UPR is accomplished via three signaling pathways induced by ER-resident molecular ER stress sensors: Protein kinase (PKR)-like ER kinase (PERK), Activating transcription factor 6 (ATF6) and Inositol-requiring enzyme 1 (Ire1).
Being conserved in all eukaryotes, Ire1 contains an ER luminal domain, which is involved in the recognition of misfolded proteins (Credle et al., 2005;Gardner and Walter, 2011), and cytoplasmic endoribonuclease and kinase domains, which are involved in the activation of downstream pathways. Activated Ire1 mediates the non-conventional splicing of an intron from X-box binding protein 1 (Xbp1) mRNA (or HAC1 mRNA, the yeast Xbp1 ortholog), causing a frame-shift during translation, thereby introducing a new carboxyl domain in the Xbp1 protein (Cox and Walter, 1996;Mori et al., 1996;Yoshida et al., 2001;Calfon et al., 2002;Shen et al., 2001). Xbp1 spliced is an effective transcription factor that regulates the expression of ER chaperones and other target genes (Acosta-Alvear et al., 2007).
In addition to mediating Xbp1 mRNA splicing, cell culture studies demonstrated that Ire1 promotes the degradation of mRNAs encoding ER-targeted proteins, a processed called RIDD (regulated Ire1-dependent decay), to reduce the load of ER client proteins during ER stress (Hollien and Weissman, 2006;Han et al., 2009;Hollien et al., 2009). The cytosolic domain of mammalian IRE1 binds Traf2 (tumour necrosis factor receptor-associated factor 2, Urano, 2000), an upstream activator of the c-Jun N-terminal Kinase (JNK) signaling pathway. This IRE1/Traf2 interaction is also independent of Xbp1 splicing and may lead to the activation of apoptosis after prolonged ER stress (Yoneda et al., 2001).
In the Drosophila photoreceptor cells, the rhabdomere is the light sensing organelle, a stack of photosensitive apical microvilli that is formed during the second half of pupal development (Cagan and Ready, 1989;Tepass and Harris, 2007). The rhabdomere is formed in the apical domain of each photoreceptor cell, which after a 90° rotation extends its apical domain along the proximal-distal axis of the retina. The growth of the rhabdomere requires the delivery of large amounts of membrane and proteins into this structure, imposing a considerable demand to the cellular mechanisms controlling protein folding and membrane production in the ER.
Among the proteins targeted to the developing rhabdomeres are the rhodopsins, the light sensitive proteins, and other proteins involved in the transduction of the light stimuli. Rhodopsin-1 (Rh1) is a seven transmembrane domain protein that starts to be expressed by 78% of pupal life (Kumar and Ready, 1995), and is delivered to the rhabdomeres of the outer photoreceptors (R1-R6), in a trafficking process that requires the activity of Rab11, MyosinV and dRip11 (Li et al., 2007;Satoh et al., 2005). The delivery of Rh1 to the rhabdomere is required for rhabdomere morphogenesis, since in Rh1 null mutants the rhabdomere does not form, causing degeneration of the photoreceptors (Kumar et al., 1997;Kumar and Ready, 1995).
In mammalians, the microRNA mir-708 is up-regulated by CHOP to control Rhodopsin expression levels and prevent an excessive Rhodopsin load into the ER (Behrman et al., 2011). In Drosophila, Ire1 signaling is activated in the photoreceptors upon expression of Rh1 folding mutants (Ryoo et al., 2007;Griciuc et al., 2010;Kang and Ryoo, 2009) or in ninaA mutations that cause the accumulation of misfolded Rh1 in the ER (Mendes et al., 2009). However, the role of Ire1 signaling during normal photoreceptor differentiation remains unknown. In this study, we show that Ire1 signaling is activated in the photoreceptors during pupal stages of Drosophila development. Ire1 mutant photoreceptors have defects in the delivery of Rh1 to the rhabdomere and the secretion of Spacemaker/Eyes shut (Spam/Eys) into the inter-rhabdomeral space. Surprisingly, Xbp1 null mutant photoreceptors have a milder phenotype with no defects in Rh1 delivery into the rhabdomere or spam/eys secretion. Targets of regulated Ire1-dependent decay (RIDD) are up-regulated in Ire1 mutant retinas, including the fatty acid transport protein (fatp), a known regulator of Rh1 protein levels (Dourlen et al., 2012). Finally, we show that the regulation of fatp levels by RIDD is critical for normal Rh1 delivery into the rhabdomere.

Ire1 signaling is activated in the photoreceptors during pupal stages
To detect Ire1 signaling activity in the photoreceptors during development, we used the Xbp1-EGFP reporter (Ryoo et al., 2007), in which EGFP becomes in frame with Xbp1 only upon Ire1-dependent mRNA splicing. We expressed Xbp1-EGFP in the developing eye under the control of the GMR-Gal4 driver, which is active during larval stages in the eye imaginal disc posterior to the morphogenetic furrow and in all cells of the eye during pupal and adult stages. During larval stages, endogenous Xbp1-EGFP was not observed in the photoreceptors ( Figure 1A), but it could be induced upon culture of the imaginal discs in 2 mM dithiotheitol (DTT) for 4 hours ( Figure 1B), a condition that interferes with oxidative protein folding in the ER and activates Ire1. During pupal stages, Xbp1-EGFP was not observed in the photoreceptors at 24 hours of pupal life ( Figure 1C), but it became evident during the second half of pupal development (Figures 1D,48 hours and 1E,72 hours). In the adult, Xbp1-EGFP was observed overlapping with some Homothorax positive lattice cells (Wildonger et al., 2005), but not in the photoreceptors (Elav marker). These results show that Ire1 signaling is transiently activated in the photoreceptors during mid and late stages of pupal development.

Activation of Xbp1-EGFP in the photoreceptors requires Ire1
To test if the Xbp1-EGFP reporter expression observed in the photoreceptors corresponds to bona fide Ire1 activity, we did experiments where we tested Xbp1-EGFP expression in mosaic eyes containing Ire1 mutant clones. The homozygous lethal line PBac{WH}Ire1 f02170 contains a piggyBac transposon insertion in the open reading frame of Ire1 ( Figure 2A). We recombined PBac{WH}Ire1 f02170 into FRT82B to make mosaic clones with the Flipase/FRT technique (Golic, 1991). Mosaic eyes were induced with eyeless-Flipase and Xbp1-EGFP was expressed under GMR-Gal4 control in order to be analyzed both in the 3 rd larval instar (Figures 2B-2E) and pupal stages ( Figures 2F-2I). Xbp1-EGFP was not observed in clones homozygous for PBac{WH}Ire1 f02170 either in larval eye discs treated with 2 mM DTT ( Figure 2B) or in the pupa ( Figure 2F). Xbp1-EGFP expression was observed in clones of a precise excision of PBac{WH}Ire1 f02170 that reverts the Ire1 genomic region to wild-type ( Figures 2C and 2G). Expression of UAS-Ire1 under the control of GMR-Gal4 ( Figures 2D and 2H) or a genomic construct covering the Ire1 region (Pacman CH322-07H05, Figures 2E and 2I) rescued Xbp1-EGFP expression in the background of PBac{WH}Ire1 f02170 clones. Ire1 kinase (Ire1 K576A ) or RNase (Ire1 H890A ) mutants failed to rescue Xbp1-EGFP expression ( Figure S1). These results demonstrate that the PBac{WH}Ire1 f02170 insertion is a strong mutant allele of Ire1 and that Xbp1-EGFP activation in the photoreceptors requires Ire1 activity.

Ire1 is not required for photoreceptor specification or for the establishment and maintenance of apical/basal polarity in the photoreceptors
The expression of Xbp1-EGFP in the photoreceptors during the second half of pupal life suggests that Ire1 signaling might be required for photoreceptor differentiation during these stages. However, it is also possible that the Xbp1-EGFP reporter is not sensitive enough to detect low levels of Ire1 activity that could have a role during earlier stages of photoreceptor development. To test this possibility, we analyzed the expression of photoreceptor specification markers in the 3 rd larval instar stage in Ire1 mutant clones. Expression of Elav (pan-neural marker) and Runt (marker of R7 and R8, Kaminker et al., 2002) was normal in Ire1 homozygous mutant clones ( Figure 3A). We also analyzed markers of apical/basal polarity, since rhabdomere morphogenesis in the pupa requires proper establishment and maintenance of apical/basal polarity in the photoreceptors (Pellikka et al., 2002;Izaddoost et al., 2002;Knust, 2007;Walther and Pichaud, 2010). In the 40h-50h pupa, Ire1 mutant clones exhibited normal localization of Armadillo ( Figure 3B), DaPKC ( Figure 3B), Crumbs ( Figure 3C) and Cadherin ( Figure S2) in the photoreceptors, indicating no defect in apical/ basal polarity at these stages.

Ire1 is required for Spacemaker/Eyes shut secretion and formation of the interrhabdomeral space
The inter-rhabdomeral space (IRS) is an extracellular lumen between the rhabdomeres enclosed by the 4 cone cells that surround each ommatidial cluster. The formation of the IRS requires the secretion by the photoreceptors of the agrin/perclan-related protein Spacemaker/ Eyes shut (Spam/Eys) (Husain et al., 2006;Zelhof et al., 2006). By 62 h pupation, Spam/Eys was observed in IRS of wild-type ommatidia. In contrast, Ire1 mutant ommatidia had low levels of Spam/Eys in the IRS, with Spam/Eys retained in the photoreceptor cell body ( Figure 3D). As consequence of this defect in Spam/Eys secretion, Ire1 mutant ommatidia showed an IRS with reduced size, which we quantified by measuring the IRS area of adjacent wild-type and Ire1 mutant ommadia ( Figure 3H). This defect in Spam/Eys secretion and IRS formation was not observed in clones of a precise excision of PBac{WH}Ire1 f02170 ( Figure 3E) and could be rescued by GMR-Gal4>UAS-Ire1 ( Figure 3F) or the CH322-07H05 genomic construct covering the Ire1 region ( Figure 3G).

Ire1 is required for Rh1 delivery into the rhabdomere
By 72h pupa, we observed that Ire1 mutant photoreceptors have defective delivery of Rh1 into the rhabdomeres, with retention of Rh1 in the photoreceptors cell body ( Figure 4A). We observed this phenotype in later pupal stages and in the adult, with concomitant degeneration of the rhabdomeres ( Figure S3). This defect in Rh1 delivery into the rhabdomere was not observed in clones of a precise excision of PBac{WH}Ire1 f02170 ( Figure  4B) and could be rescued by GMR-Gal4>UAS-Ire1 ( Figure 4C) and CH322-07H05 genomic construct ( Figure 4D). Ire1 kinase (Ire1 K576A ) or RNase (Ire1 H890A ) mutants failed to rescue Rh1 localization defect ( Figure S1). We also observed that Ire1 mutant photoreceptors have reduced levels of ER markers, such as the ER chaperone BiP/GRP78 (Figures 4E, S4A and S4B) and the Rh1 specific chaperone ninaA ( Figure 4F and S4A). The levels of the ER chaperone Calnexin (Cnx), however, were unchanged between wild-type and Ire1 mutant cells ( Figure 4G).
The cytoplasmic retention of Rh1 and the lower levels of the ER chaperones BiP/GRP78 and ninaA, suggested that Rh1 could be accumulating as misfolded protein in Ire1 mutant photoreceptors. In ninaA mutations, Rh1 accumulates as an immature, glycosylated form, while the levels of mature Rh1 are reduced (Colley et al., 1991;Ondek et al., 1992;Baker et al., 1994). We performed Western Blot on protein extracts from Ire1 mutant eyes and although Rh1 mature form levels were reduced, no immature Rh1 form was observed ( Figure 4H). This result indicates that, although ninaA levels are reduced in Ire1 mutant photoreceptors, residual ninaA activity could be sufficient to process Rh1 to the mature form.

Xbp1 is not required for Spacemaker/Eyes shut secretion or Rh1 delivery into the rhabdomere
Next, we investigated if Xbp1 mutants also exhibit the same defects in Spam/Eys secretion and Rh1 delivery into the rhabdomere as observed in Ire1 mutant photoreceptors. We previously used a lethal P element insertion upstream of Xbp1 open reading frame, P{lacW}Xbp1 K13803 , which accelerates retinal degeneration induced by ninaE dominant mutations (Ryoo et al., 2007). Clones of P{lacW}Xbp1 K13803 did not present the defects in Spam/Eys secretion and Rh1 delivery into the rhabdomere observed in the Ire1 mutation (data not shown). However, P{lacW}Xbp1 K13803 still has residual expression of Xbp1 mRNA (Ryoo et al., 2007), that could be sufficient to rescue potential defects in the photoreceptors. For this reason, we decided to generate Xbp1 protein null deletion mutants with no residual function of the Xbp1 protein. Using Δ23 transposase, we induced imprecise excisions of 2 P elements (P[PTT-GB]Xbp1 CB02061 and P[SUPor-P]CG9418 KG05183 ) neighboring Xbp1, which resulted in the identification of 3 different deletion mutations (Excisions 79, 81 and 250, Figure 5A). We recombined each deletion mutant into FRT42D chromosome to generate mosaic clones in the eye. We observed identical phenotypes with these 3 Xbp1 deletion mutants.
In contrast to Ire1 mutants, Xbp1 mutant ommatidia presented normal levels of Spam/Eys secretion into the inter-rhabdomeral space with no retention of Spam/Eys in the cell body of the photoreceptors ( Figure 5B). Likewise, Rh1 was delivered to the rhabdomere ( Figures 5C  and 5D), although at 72 h. pupa we observed lower levels of Rh1 in the rhabdomeres of Xbp1 mutant photoreceptors than in wild type (quantification in Figure 5E). In contrast, at 82 h. pupa, Xbp1 mutant rhabdomeres presented higher levels of Rh1 in the rhabdomere than wild-type ( Figures 5D and 5E). Furthermore, at this stage, while wild-type showed the typical crescent-shaped gradient of concentration with higher levels at the rhabdomere base, Xbp1 mutant photoreceptors showed an even distribution of Rh1 in the rhabdomere (magnifications in Figure 5D). Our interpretation of these results is that in Xbp1 mutant ommatidia, Rh1 is delivered into the rhabdomere, although there is a developmental delay in this process. As in Ire1 mutant photoreceptors, Xbp1 mutants had reduced levels of the ER markers BiP/GRP78 (Figures 5F, S4A and S4B) and Protein disulphide isomerase (Pdi, Figure 5G). Cnx was unchanged between Xbp1 mutant and wild-type cells ( Figure 5H).
The branches of the UPR are interconnected and loss of one pathway may lead to increased activation of parallel pathways during ER stress. Upregulation of ATF4 is induced by activation of the PERK pathway and we observed upregulation of ATF4 in Ire1 mutant clones but not in Xbp1 mutants (Figures S4C and S4D). Overall, these results demonstrate that Ire1 mutants have a more severe rhabdomere morphogenesis phenotype than Xbp1 mutants, indicating that Ire1 function in this context should be mediated by Xbp1independent mechanisms.

Regulation of Fatp by RIDD is critical for rhabdomere morphogenesis
Studies using cell culture approaches showed that, during ER stress, Ire1 promotes the degradation of mRNAs encoding proteins that are translated in the ER (Hollien and Weissman, 2006;Han et al., 2009;Hollien et al., 2009). This Xbp1-independent function of Ire1 was denominated as Regulated Ire1-Dependent Decay (RIDD). To investigate if RIDD occurs in the Drosophila eye, we determined by quantitative RT-PCR the mRNA levels of specific RIDD targets in wild-type and Ire1 mutant eyes ( Figure 6A). The mRNAs of genes such as Indy, Sparc and Fatp were 1,5 to 2,5 fold up-regulated in Ire1 mutant eyes, in comparison with controls. To confirm the quantitative RT-PCR results, we performed immuno-fluorescent staining in eyes with Ire1 mutant clones, using antibodies against Fatp and Sparc. We found that Fatp ( Figure 6B) and Sparc ( Figure S5A) protein levels were upregulated in Ire1 mutant clones. Furthermore, purified human Ire1 cleaved Drosophila Xbp1 and Fatp mRNAs in vitro ( Figure S5B). We found two Ire1 cleavage sites in the mRNA of Fatp and mutagenesis of these sites abrogated cleavage of Fatp mRNA by human Ire1 ( Figure S5B).
Fatp is a regulator of Rh1 protein levels, since in Fatp loss-of-function mutants Rh1 levels are up-regulated, causing photoreceptor degeneration (Dourlen et al., 2012). Furthermore, fatty acids are precursors in the biosynthesis of phosphatidic acid and an increase in Fatp levels could lead to higher levels of phosphatidic acid, due to an increased import of fatty acids into the cell. High levels of phosphatidic acid cause downregulation of Rh1 protein levels and rhabdomere defects (Raghu et al., 2009), similar to the phenotypes observed in Ire1 mutant photoreceptors. Thus, we investigated if Fatp over-expression could lead to phenotypes similar to the ones observed in Ire1 mutant photoreceptors. In fact, photoreceptors where UAS-Fatp was expressed under the control of GMR-Gal4 showed defective delivery of Rh1 to the rhabdomere ( Figure 6D). Furthermore, downregulation of Fatp levels by expression of Fatp RNAi in Ire1 mutant photoreceptors rescued the defective delivery of Rh1 to the rhabdomere ( Figure 6E). Finally, we detected higher levels of phosphatidic acid in Ire1 mutant retinas than in controls ( Figure S5C) and the defect of Rh1 delivery to the rhabdomere in Ire1 mutant photoreceptors was also rescued by overexpression of lipid phosphate phosphohydrolase (UAS-LPP (lazaro), Figure 6F), an enzyme that converts phosphatidic acid to diacylglycerol and is able to rescue the rhabdomere defects caused by excessive levels of phosphatidic acid (Raghu et al., 2009). We conclude that RIDD regulation of Fatp levels is an important part of Ire1 role during photoreceptor differentiation in the Drosophila eye.

DISCUSSION
Studies in mammalian systems revealed that the Ire1/Xbp1 signaling pathway is important during development for the differentiation of secretory cells. For example, Xbp1 "knockout" mice have defects in the differentiation of antibody-secreting plasma cells (Reimold et al., 2001) and secretory cells of the exocrine glands of the pancreas (Lee et al., 2005). Presumably in these cases, activation of Ire1/Xbp1 signaling is required to increase the capacity of the ER to fold and process the high load of secreted proteins.
Our present results demonstrate that Ire1 signaling is required for photoreceptor differentiation and rhabdomere morphogenesis, a process that also imposes a high demand to the capacity of the ER to fold proteins such as Spam/Eys and Rh1. As shown, Ire1 mutant photoreceptors have defects in the secretion of Spam/Eys to the IRS and in the delivery of Rh1 to the rhabdomere. However, we see activation of the Xbp1-EGFP reporter starting at 48h of pupal development, well before when the Spam/Eys secretion and Rh1 delivery defects are observed. Presumably, the folding of other unidentified proteins during these earlier stages might also require Ire1 signaling. It is noteworthy though, that in mutant B lymphocytes modified to lack antibody production, Ire1 is still activated (and Xbp1 spliced) upon lymphocyte differentiation to plasma cells (Hu et al., 2009). Activation of Ire1/Xbp1 signaling in this context seems to be part of the process of plasma cell differentiation, independently of the accumulation of misfolded proteins in the ER lumen (van Anken et al., 2003;Hu et al., 2009). Ire1 function is also required for the regulation of the membrane lipids. In mammalians, Ire1/Xbp1 signaling regulates the biosynthesis of phospholipids and other lipids (Sriburi et al., 2007;Lee et al., 2008). A study in yeast demonstrated that Ire1 is activated by "membrane aberrancy", a condition of stress caused by the experimental depletion of inositol (Promlek et al., 2011). Activation of Ire1 in this case occurs by a mechanism that is distinct from the one involving the recognition of misfolded proteins by the luminal domain of Ire1 (Promlek et al., 2011). Furthermore, Ire1 can be activated by direct binding of flavonoids, such as quercetin, to a pocket present in the cytoplasmic domain of Ire1, in a mechanism that is also independent of the binding of misfolded proteins to Ire1 (Wiseman et al., 2010). Our present results do not clarify if Ire1 activation in the photoreceptors during pupal stages results from the accumulation of misfolded proteins in the ER lumen or an imbalance in the membrane lipids.
Our results demonstrate that Ire1 signaling is required for photoreceptor differentiation and rhabdomere morphogenesis in an Xbp1-independent manner. Studies using cell culture paradigms demonstrated that, in addition to mediating Xbp1 mRNA splicing, Ire1 also promotes RIDD (Regulated Ire1-Dependent Decay), the degradation of mRNAs encoding ER-targeted proteins (Hollien and Weissman, 2006); (Han et al., 2009) (Hollien et al., 2009), but the physiological significance of the RIDD mechanism is unknown. Our quantitative RT-PCR results show that RIDD targets are up-regulated in Ire1 mutant eyes, including fatty acid transport protein (fatp), a regulator of Rh1 protein levels (Dourlen et al., 2012). Our results show that regulation of fatp mRNA by RIDD is critical for rhabdomere morphogenesis, since the experimental downregulation of fatp mRNA by RNAi rescues the Rh1 rhabdomere delivery defect observed in Ire1 mutants.
Rh1 protein levels and Rh1 delivery to the rhabdomere is very sensitive to the levels of sphingolipids (Yonamine et al., 2011) and phosphatidic acid (Raghu et al., 2009). Increased Fatp levels may lead to an increase in the levels of fatty acids and subsequently, phosphatidic acid, which is known to downregulate Rh1 protein levels and cause rhabdomere morphogenesis defects (Raghu et al., 2009). High levels of phosphatidic acid disrupt the Arf1-dependent transport of membrane to the developing rhabdomere (Raghu et al., 2009). Our results show that phosphatidic acid levels are elevated in Ire1 mutant retinas and lowering phosphatidic acid levels by expression of LPP rescues the defects observed in Ire1 mutants, demonstrating that Ire1/Fatp dependent regulation of fatty and phosphatidic acids levels is important for rhabdomere morphogenesis in Drosophila. In addition, it is possible that the increase in phosphatidic acid levels in Ire1 mutant photoreceptors is also caused by the activation of PERK, since we observed upregulation of the PERK pathway mediator ATF4 in Ire1 mutant photoreceptors ( Figure S4C) and in cell culture models it was shown that PERK is able to phosphorylate diacylglycerol and generate phosphatidic acid (Bobrovnikova-Marjon et al., 2012).
In conclusion, our results, using well-characterized genetic tools (Ire1 and Xbp1 null mutations) and a developmental paradigm (photoreceptor differentiation in the Drosophila pupa), demonstrate the physiological relevance of Xbp1-independent mechanisms downstream of Ire1 signaling.

Drosophila stocks and molecular biology
All Drosophila stocks were raised with standard cornmeal medium, at 25°C, in a 12 hour light/dark cycle. Clones of mutant eye tissue were generated by the Flp/FRT technique (Golic, 1991), with Flipase expression under the control of the eyeless promoter (Newsome et al., 2000). The open reading frame of Ire1 was amplified from larval cDNA with the forward primer 5'ggaagatctatgggcagtcttaagaagttacc and reverse primer 5' cggggtacctcaatcctgcgttgaaggtgg and cloned into pUAST using BglII and Acc65I restriction sites. The injection of DNA into embryos to establish Drosophila transgenic lines was performed by BestGene. The CH322-07H05 (Pacman) genomic construct is a 22Kb fragment covering Ire1 and the neighbouring genes CG4662, CG11447, CG4572 and CG4686. The CH322-07H05 transgenic line was generated by using φC31 integrasemediated transgenesis (BestGene Strain #9752, attP acceptor site in 22A3). The Xbp1 deletion mutants were created by crossing flies with the P elements P[PTT-GB]Xbp1 CB02061 or P{SUPor-P}CG9418 KG05183 with Δ23 transposase flies and were characterized by PCR and sequencing. The precise excision of PBac{WH}Ire1 f02170 was generated by crossing to PBac transposase and confirmed by sequencing. UAS-Fatp and UAS-Fatp RNAi were a gift from B. Mollereau and were described in (Dourlen et al., 2012). UAS-Lpp was a gift from P. Raghu and was described in (Raghu et al., 2009).

Quantitative RT-PCR
For quantitative reverse transcription-PCR (RT-PCR) analysis, total RNA was extracted from adult heads with Trizol and Direct-zol RNA mini-prep (Zymo Research). 0,5 μg of total RNA was retro-transcribed using RevertAid First Strand (Thermo/Fermentas). Each PCR reaction was performed on 1/40 of the cDNA obtained using SSoFast EvaGreen Supermix (Bio-Rad) according to the manufacturer's instructions and using a Bio-Rad Cfx-96 detection system. All samples were analyzed in triplicates and from 3 independent RNA extractions. For each sample, the levels of mRNAs were normalized using rp49 as a loading control. PCR primers were: sparc (F: AGCTCCATCTGCATGTAGTGCTCG, R:

In vitro Ire1 mRNA cleavage assay
The in vitro Ire1 mRNA cleavage assay was performed as in (Cross et al., 2012). dXbp1 mRNA was produced in vitro with T7 RNA polymerase (Promega Ribomax Express RNA) from pOT2-dXbp1 after linearization with NotI. dFatp mRNA was produced from a PCR product obtained with F:TAATACGACTCACTATAGGGAGACCGG and R: CGGTGTGGTACAAAGGCAAGG. Fatp Mut1 was done by changing GTGGCCAATGTGctgcagGCTCAGGGCTAC into GTGGCCAATGTGctccacGCTCAGGGCTAC and Fatp Mut. 2 by changing TCCCTCCCTGctgcacAGCATCAC into TCCCTCCCTGctccaaAGCATCAC. The mRNAs (1μM) were incubated at 29°C with or without purified human Ire 1 (5 μM) and the Ire1 inhibitor 4μ8C (10 μM), kind gifts from Heather Harding. Cleavage products were resolved by electrophoresis on a 1% denaturing agarose gel stained with ethidium bromide.

Analysis of retinal phosphatidic acid
The analysis of retinal phosphatidic acid by LC-MS was performed following the protocol described in (Raghu et al., 2009). For each genotype 150 retinas were separated from the head after being frozen in liquid nitrogen and dehydrated in acetone. The retinas were homogenized using a glass piston in 1 mL methanol containing 500 ng of a 12:0/12:0 PA standard (Avanti Polar Lipids). The lipids were extracted with addition of 2 mL of chloroform and 1ml of 0.88% (w/v) KCl to split the phases. The lower organic phase was dried in a speedvac during 30 min and stored at −20°C until further analysis.
Phospholipids extracts were profiled using the LC-MS in order to identify and quantify the PA species present. The analysis was performed using a Surveyor HPLC system (Thermo Scientific) interfaced to an externally calibrated LTQ-Ion trap mass spectrometer (Thermo Scientific) with an electrospray ionization (ESI) source in negative mode. The capillary was operated at 5kV and source temperature was set to 300°C, the sheath gas and auxiliary gas setting was 20 (arbitrary units). For normal-phase chromatographic analysis, a 10 μL aliquot of phospholipid extract dissolved in 30 μL of chloroform was injected into a Luna, Phenomenex silica column (150 mm × 1.0 mm, 3um particle size). PA phospholipids species originating from fly extracts were separated using chloroform/methanol/water (90:9.5:0.5) containing 7.5 mM ethylamine changing to acetonitrile/chloroform/methanol/ water (30:30:35:5) containing 10 mM ethylamine at flow rate of 50 μL/min. The column temperature was maintained at 25°C. The relative amount of each PA species was determined dividing the area of the corresponding m/z peak by that of the internal standard.
For electron microscopy, adult eyes were fixed in 2,5% glutheraldehyde in 0,1 M cacodylate buffer (pH 7.4) and post-fixed with 1% osmium tetroxide. After dehydratation in ethanol and propylene oxide, the eyes were infiltrated with resin and mounted in silicone molds. Ultrathin sections (60nm) were cut and examined with electron transmission microscope (Jeol JEM).

Supplementary Material
Refer to Web version on PubMed Central for supplementary material. CEDOC/FCM-UNL and IGC for access to the imaging and fly facilities. We thank Hermann Steller for all the support during the initial stages of this work. This work was supported by Fundação para a Ciência e a Tecnologia through grants PEst-OE/EQB/LA0004/2011, PTDC/BIA-BCM/105217/2008, PTDC/SAU-OBD/104399/2008 and PTDC/BEX-BCM/1217/2012, by a Marie Curie International Reintegration grant (PIRG03- GA-2008-230935), and by the National Institutes of Health grants R01EY020866 and R01DK047119. D.R. is a Principal Research Fellow of the Wellcome Trust. D.S.C was supported by a doctoral fellowship from the Gulbenkian Foundation.

HIGHLIGHTS
• Ire1 is activated in the photoreceptors during Drosophila pupal development.
• The role of Ire1 in rhabdomere morphogenesis is independent of Xbp1 function.