β amyloid protein precursor-like (Appl) is a Ras1/MAPK-regulated gene required for axonal targeting in Drosophila photoreceptor neurons

Differentiation of photoreceptors in the Drosophila compound eye requires activation of Ras1 and mitogen-activated protein kinase (MAPK) downstream of a receptor tyrosine kinase (RTK). In most photoreceptor precursors, Ras1/MAPK is activated by the RTK Drosophila Epidermal Growth Factor Receptor (DER) (Freeman, 1996). In the R7 photoreceptor precursor, however, the same Ras1/MAPK cassette is activated by an additional RTK, the Sevenless RTK (Sev) (Banerjee et al., 1987b; Hafen et al., 1987). Both RTKs are required for R7 determination, since significantly higher levels of Ras1/MAPK activation are necessary for R7 to overcome specific repressive mechanisms that would otherwise transform it into a non-neural cone cell. Sev is expressed in several photoreceptor precursors and in the four cone cells (Banerjee et al., 1987a; Tomlinson et al., 1987). However, only the presumptive R7 will activate the Sev receptor. Although the complex regulatory network involved in R7 specification has been extensively investigated, little is known about the genetic program that responds to these RTK signals.

Here, we explore the transcriptional profile downstream of Ras/MAPK-mediated Sev signaling and show that β amyloid protein precursor-like (Appl), the fly ortholog of the human β amyloid precursor protein (APP) (Hardy and Selkoe, 2002; Rosen et al., 1989) is a direct target of Ras1/MAPK required for R7 function. Appl belongs to the APP family, which is conserved across species from Caenorhabditis elegans to mammals (Daigle and Li, 1993; Rosen et al., 1989). In humans, APP is the precursor of the Aβ peptide, which is involved in the development of Alzheimer's disease. In addition to mutations in the coding region that favor cleavage to produce the Aβ-peptide, aberrant expression of APP has also been associated with Alzheimer's disease (AD) (Koo et al., 1990; Palmert et al., 1988). Remarkably, different ligands of membrane receptors with intrinsic tyrosine kinase activity modulate APP expression in mammalian cell cultures (Cosgaya et al., 1996; Lahiri and Nall, 1995; Ohyagi and Tabira, 1993; Ruiz-León and Pascual, 2001). Moreover, some APP regulatory sequences that respond to Ras/MAPK have been characterized in PC12 cells (Villa et al., 2001).

Despite some evidences in cultured cells, little is known about APP regulation in tissues and organs in vivo. Therefore, we decided to use Drosophila advanced genetic and genomic tools to analyze Ras/MAPK regulation of Appl expression. In flies, Appl is specifically expressed in postmitotic neurons at all stages of development (Martin-Morris and White, 1990) and has been implicated in axonal transport (Gunawardena and Goldstein, 2001; Torroja et al., 1999a), neuronal development (Li et al., 2004; Merdes et al., 2004) synaptic bouton formation (Ashley et al., 2005; Torroja et al., 1999b) response to traumatic brain injury (Leyssen et al., 2005) and protection against neurodegeneration of processed Appl products (Wentzell et al., 2012). Furthermore, human APP rescues the behavioral phenotype of the Drosophila Appl loss-of-function mutant, indicating an evolutionarily conserved role (Luo et al., 1992). In this work we studied the function of Appl in the R7 and found that it is required for R7 targeting and light discrimination. In addition, it is not yet clear whether Appl as well as mammalian APP transcriptional activation responds to Ras1 or to neural differentiation. We took advantage of the Drosophila genetic tools to study Ras1-mediated activation of Appl at single cell resolution in the intact developing retina. Our study provides a description of the Ras1 transcriptional regulation of Appl and offers useful insights into the mechanisms by which RTKs and the Ras/MAPK pathway regulate the differentiation of a neural cell type. Furthermore, our findings may be of relevance to understanding the pathophysiology of human disorders such as Alzheimer's disease.

To investigate the transcriptome controlled by Sev in the R7 photoreceptor, we analyzed microarrays using eye imaginal discs from third instar larvae in the sev mutants sev^S11 and sev^d2 compared with wild-type controls. The gain-of-function (GOF) allele sev^S11 encodes a ligand-independent activated Sev receptor and generates flies with extra R7 photoreceptors, whereas the loss-of-function (LOF) allele sev^d2 impairs the formation of R7 (Basler et al., 1991). The entire set of microarrays was normalized following the same protocol (see Materials and Methods). This kind of standardization allowed us to include a third computational comparison: sev^s11 against sev^d2 (GOF/LOF). The number of genes with significantly modified expression (1.5-fold change, and false-discovery-rate-corrected P<0.05) is shown in supplementary material Table S1 .

As Ras1 acts through the MAPK cascade to activate transcription (Plotnikov et al., 2011), we restricted the list of genes to those that were transcriptionally activated. We defined this population as target genes, which correspond to genes with increased expression in GOF/wild type or GOF/LOF, and those with decreased expression in LOF/wild type. The total number of genes obtained was 233. The functional annotation of target genes using Gene Ontology (GO) is shown in Fig. 1.

MAPK controls neural development through phosphorylation of the transcriptional repressor Yan and the transcriptional activator Pointed-P2 (PntP2) (Brunner et al., 1994). Both proteins are able to recognize ETS binding sites, DNA consensus sequences that are well characterized and conserved (Oikawa and Yamada, 2003). To restrict the list of target genes to those candidates directly regulated by the MAPK cascade, we performed a computational search for ETS binding sites in the upstream promoters (1000 bp upstream from the transcription start site) and introns of the target genes using the genomes of 12 Drosophilidae to strengthen the predictions (Clark et al., 2007). Thirty-eight genes with ETS binding sites were identified as ETS target genes (supplementary material Table S2). The GO annotation of these genes contained categories consistent with the molecular program activated by Ras1 in the R7 cell (Fig. 1B). Appl was classified within the over-represented category ‘neurogenesis’ containing putative neural genes that respond to Ras1 in R7 neural precursors (supplementary material Table S2). Hereafter, we decided to focus on Appl regulation because of the biomedical implications.

Appl mRNA was found in photoreceptors as they are recruited into developing clusters (Fig. 2A). It first appears in the precursor of R8, followed by R2/5, R3/4, R1/6, and finally R7 (Fig. 2A,B). It was undetectable in unspecified cells and cone cells (Fig. 2A,B), in agreement with previous studies (Luo et al., 1990; Torroja et al., 1996). Antibodies directed against the N-terminal ectodomain, show a punctuate distribution at the membrane of all photoreceptors and developing axons (Torroja et al., 1996). We re-examined the expression of Appl protein in the eye and observed differences in Appl distribution between photoreceptors. Confocal sagittal and horizontal sections of eye discs showed Appl to be predominantly expressed in R7 and R8 photoreceptors (Fig. 2B). This prevailing localization was more evident in the posterior part of the disc, where all cells are already specified. High levels of Appl in R7 correlate with the increased expression of Appl detected in sev GOF arrays. Appl upregulation by activated Sev was further confirmed by in situ hybridization and immunostaining of sev^S11 eye discs (Fig. 2C,D). As shown in Fig. 2C, greater Appl mRNA accumulation was observed in sev^S11 eye discs compared with wild-type discs, and higher levels of Appl protein were found in cells transformed to an R7 fate (compare Fig. 2B and 2D).

Next, we examined the relation between high Appl expression and R7 function. Adult flies normally display a phototactic response involving a preference for UV over visible light that relies on perception of UV light by the R7 photoreceptor (Harris et al., 1976). As a consequence, R7 function can be analyzed using a spectral-preference assay (see Materials and Methods). We used this assay to assess the functional role of Appl in R7 and found that flies carrying the null mutation Appl^d exhibited reduced UV preference compared to wild-type flies (Fig. 3A). Since Appl^d eyes contain all photoreceptors, including R7 (supplementary material Fig. S1A), Appl does not contribute to neural specification. We therefore assessed the possibility that Appl is involved in regulating R7 axonal targeting. Photoreceptor neurons establish synapses in the peripheral lamina and the deeper medulla of the optic lobe. The R8 and R7 neurons of each ommatidium project their axons into the same horizontal column of the medulla but synapse in different layers: M3 and M6, respectively (Ting and Lee, 2007) (Fig. 3B). R7 axons in Appl^d flies were found to be mistargeted in 2% of the R7 photoreceptors examined (Fig. 3C). This lack of a severe phenotype in Appl^d mutant eyes was not surprising, since several molecules are known to be involved in axonal targeting (Hadjieconomou et al., 2011). Thus, other transmembrane proteins might mask the Appl^d phenotype. To address this possibility, we sensitized Appl^d flies using different genetic backgrounds in heterozygosis for other alleles of membrane proteins frazzled (fra³) (Kolodziej et al., 1996), fasciclinIII (fasIII^E25) (Chiba et al., 1995), NCadherin (CadN^m19) (Iwai et al., 1997) and neurotactin (Nrt¹) (Speicher et al., 1998) and tested them alone and in combination with Appl^d in the spectral preference assay. We found that Appl^d flies carrying any of those mutant alleles had a greater frequency of incorrect response to UV light (supplementary material Fig. S1B). The strongest phenotype was observed in the combination of Appl^d with axon guidance protein neurotactin (Nrt) (Fig. 3A), which confirms their genetic interaction (Merdes et al., 2004). These data suggest that Appl appears to cooperate with multiple guidance receptors for correct R7 targeting. We therefore analyzed the targeting of R7 to the M6 layer in Appl^d/Appl^d;Nrt¹/+ flies. We recovered phenotypes with differences in severity that can be divided into three categories: (1) weak phenotype (56% of brains) that resembles that of Appl^d mutants with 3% of mistargeted R7; (2) intermediate (30%) phenotype with 12.5% of mistargeted R7; and (3) strong (14%) phenotype in which the axons are completely disorganized (n = 14; Fig. 3D). These findings further support a role for Appl in R7 axon targeting and suggest a function for Appl in R7-mediated discrimination of UV light.

To evaluate whether Ras1 signaling is required for Appl expression we generated clones of mutant alleles that interfere with or enhance Ras1 pathway activity. We first induced clones expressing a dominant-negative form of DER (DER^DN). These clones caused complete loss of Appl expression (Fig. 4A), indicating that DER is required for Appl expression. We next analyzed clones expressing the constitutively active form of Ras1, Ras^V12. We observed that expression of Ras^V12 is sufficient to cell-autonomously activate Appl in the eye (Fig. 4B). Ras1 activity can induce photoreceptor differentiation, whereas DER^DN prevents photoreceptor formation (Freeman, 1996; Spencer et al., 1998). Therefore, to clarify whether Appl expression is downstream of Ras1 signaling or alternatively an indirect consequence of photoreceptor specification, we analyzed the R8 precursor, which does not require Sev and DER for its specification (Yang and Baker, 2001) We generated clones carrying a null allele of Ras1 (Ras1^ΔC40b) near the morphogenetic furrow that precedes photoreceptor differentiation and then examined the zinc finger protein encoded by senseless, which is a specific marker of the R8 precursor. We found that Appl expression was severely reduced or absent in those clones (Fig. 4C), indicating that Ras1 controls its expression independently of photoreceptor specification.

The transcriptional output of the canonical Ras1/MAPK pathway is usually mediated by the transcription factor PntP2 (Brunner et al., 1994; O'Neill et al., 1994). To assess whether PntP2 mediates Ras1/MAPK activation of Appl, we generated twin clones of the loss-of-function allele pnt^Δ88. In these clones, we observed R8 cells with severely reduced Appl expression (Fig. 4D) that strongly resembled the expression in Ras1 loss-of-function clones. This result indicates that PntP2 mediates Ras1/MAPK activation of Appl expression and is consistent with a putative direct regulation of Appl through PntP2 binding to specific ETS regulatory sequences.

Consistent with a proposed role of ETS binding sites, we found four conserved ETS predictions distributed in two Appl introns (ETS1and ETS4; Fig. 4E; supplementary material Fig. S2). To evaluate whether PntP2 is able to induce gene expression through binding to Appl ETS sites, we generated enhancer-lacZ reporters for all four ETS sites in flies and found that Ras^v12 is sufficient to activate ETS1 and 3, when ectopically expressed in clones (supplementary material Fig. S3). None of the constructs produced endogenous β-galactosidase expression in wild-type eye discs. This may be because the binding sites for other transcriptional regulators that cooperate with the endogenous expression are missing in this construct. Nevertheless, to validate PntP2 binding, we performed chromatin immunoprecipitation (ChIP) assays with eye imaginal discs. Owing to the lack of antibodies against PntP2, we generated a transgenic fly carrying a HA-tagged PntP2 under the UAS inducible promoter (supplementary material Fig. S4A). The UAS-PntP2-HA transgene under a sev-Gal4 driver restricts expression to the Sev-expressing cells of the eye (R1/6, R3/4, R7, and cone cells). To enhance PntP2 activity, we performed the experiment in a sev^S11 background (supplementary material Fig. S4B). Subsequently, we performed ChIP-PCR experiments with those imaginal discs expressing PntP2-HA using an anti-HA antibody (Fig. 4F,G). To assess the antibody background, we also performed ChIP on wild-type discs lacking PntP2-HA (Fig. 4G). Our results showed an enrichment of the Appl ETS1 site bound to PntP2, whereas no enrichment was observed for the Appl ETS2 site (Fig. 4G). This result demonstrates that Appl is a direct target of Ras1/MAPK signaling through direct binding of PntP2 to Appl ETS1.

Our results raise the question of whether RTK signaling could also regulate expression of the APP family in vertebrates. Ligands of RTKs such as NGF, FGF and EGF increase APP mRNA levels in mammalian cell lines (Cosgaya et al., 1996; Ruiz-León and Pascual, 2001). Since most of these observations have been performed in cell culture, we used an alternative vertebrate developmental system to test APP transcriptional regulation by RTKs in vivo. We analyzed activation of appb, a fish ortholog of Appl, in a zebrafish transgenic line, which expresses a constitutively active form of the Xenopus Fibroblast Growth Factor Receptor 1 (fgfr1) in response to heat shock. Appb is expressed in the early developing central nervous system including the retina (Lee and Cole, 2007; Musa et al., 2001). Heat-shock treatment of ca-fgfr1 transgenic embryos resulted in a dramatic increase of appb compared to control embryos (supplementary material Fig. S5). This finding suggests an evolutionarily conserved mechanism of appb regulation by RTK. Further experiments will help to clarify whether deregulated RTK/Ras1 is implicated in the progression of neurodegeneration and aberrant accumulation of APP in pathological conditions.

In summary, two main conclusions can be drawn from this work. First, Drosophila Appl is involved in R7 axonal targeting. Moreover, our finding that the Appl loss-of-function defects are enhanced when combined with Nrt heterozygous mutant suggest that Appl acts at the membrane of R7, where it interacts with other proteins such as Nrt. Second, Appl activation downstream of the RTK/Ras1 is independent of neural specification, occurs in vivo, and is mediated by direct binding of PntP2 to ETS sequences in the Appl regulatory region.

Together, these findings may provide insights into the pathogenesis of neurological disorders such as Alzheimer's disease. The β-amyloid peptides, which accumulate in the amyloid plaques found in the brain of Alzheimer's disease patients, are produced after APP proteolysis. However, Alzheimer's disease has not only been associated to the production of the primary component Aβ by proteolysis of APP, but also by transcriptional regulation. Increased APP transcription underlies the phenotype in some cases of familial Alzheimer's disease (Querfurth et al., 1995). In addition, overexpression of APP appears to be responsible for the early onset of Alzheimer's disease in individuals with Down syndrome (Rumble et al., 1989). Thus, our results open the possibility to explore whether in some cases of Alzheimer's disease a burst of RTK/Ras1/MAPK occurs and whether this signaling activity ends with high APP accumulation.

Amyloid β peptides are known to be involved in vision dysfunction caused by age-related retinal degeneration in mouse models (Ding et al., 2008; Ning et al., 2008). Thus, our in vivo observations could be the basis for further research in mammalian models for neurodegenerative retinal disorders that share several pathological features with Alzheimer's disease.

Alleles used in this work: Sev^s11 and sev-Gal4sev^S11 (Basler et al., 1991); sev^d2 (Gerresheim, 1988); UAS-DER^DN (Freeman, 1996; Scholz et al., 1997); UAS-Ras^V12 (Scholz et al., 1997); Ras1^ΔC40b (Hou et al., 1995); Pnt^Δ88 (Scholz et al., 1993); fra³ (Kolodziej et al., 1996); fas3^E25 (Snow et al., 1989); CadN^m19 (Iwai et al., 1997); Nrt¹ (Hortsch et al., 1990); Df(1)w and Appl^d (Luo et al., 1992). To generate UAS-PntP2 transgenic flies, full-length Pnt-RB and three copies of the HA epitope were cloned between the XhoI and XbaI sites of the pUASTattB vector and inserted in the 68E site of integrase trangenesis system. ETS1·LacZ, ETS2·LacZ, ETS3·LacZ, and ETS4·LacZ transgenic flies were generated by cloning Appl genomic fragments that encompass the ETS predictions and surrounding nucleotides with high conservation scores across 12 Drosophila genomes, to avoid introduction of a potentially fragmenting enhancer. Fragments of ∼500 bp were cloned in the pLacZattB vector (http://flyc31.frontiers-in-genetics.org/sequences_and_vectors.php) upstream of the Hsp70 minimal promoter driving nuclear lacZ and inserted into a predefined genomic position 86Fb via ΦC31-mediated transgenesis.

Clones overexpressing UAS transgenes were induced by a 10-minute heat shock at 37°C to activate the hsFlp. Clones of homozygous mutant cells were induced by incubation at 37°C for 45 minutes. Clones were generated at 60 hours after egg laying. In all cases eye discs were analyzed at 120 hours after egg laying. Genotypes used for clonal analysis with FRTs in cis: w,hsFlp;Act FRT y+FRTGal4::UAS-GFP were used to drive the following transgenes: UAS-Ras^V12 or UAS-DER^DN. For clonal analysis of FRTs in trans: w, hsFlp; FRT82B UbiGFP and FRT82B Ras1^ΔC40b or FRT82B pnt^Δ88.

Two-color microarray design and analyses were carried out as previously reported (Beltran et al., 2007; Blanco et al., 2010). Total RNA was extracted from eye-antenna imaginal discs from wandering blue-gut staged early third instar larvae using the RNeasy Protect Mini Kit (Qiagen Inc., Valencia, CA, USA). At least two independent total RNA extractions were carried out from sev^S11 and sev^d2 strains. Total RNA from w¹¹¹⁸ was used as a common reference. We obtained lists of genes that were differentially expressed (had an FDR-corrected P<0.05 and at least two spots from the four replicate arrays that passed the quality filters) over 1.5- or 2.0-fold in the mutants compared to the reference strain. Raw and normalized data are deposited in the Gene Expression Omnibus (GEO) database [http://www.ncbi.nlm.nih.gov/projects/geo/] with the accession number GSE37793.

We extracted 1000 nucleotides upstream of the transcription start site and the intron sequences of our set of 233 target genes, according to RefSeq annotations in the University of California, Santa Cruz (UCSC) genome browser. We used the MatScan program (Blanco et al., 2006) to analyze these sequences with Jaspar and Transfac predictive models for ETS. We converted these predictions into the UCSC custom track format to map them along the D. melanogaster genome. Using the Conservation track (multiple alignment of Drosophila species), we filtered out the predictions that were not conserved in at least five species (including D. pseudoobscura or more distant species).

Adult brains (1 to 3 days old) and eye imaginal discs where processed for antibody staining using conventional techniques. Primary antibodies: anti-GFP rabbit serum (1∶1000, Santa Cruz Biotechnology, Inc., Santa Cruz, CA), rabbit anti-β-galactosidase (1∶1000, Cappel, ICN/Cappel, Aurora, OH), mouse anti-Prospero, mouse anti-Elav and mouse anti-Chaoptin (1∶20, 1∶100 and 1∶10, respectively) from Developmental Studies Hybridoma Bank (Iowa City, IA), guinea pig anti-senseless (1∶800; H. Bellen, Baylor College of Medicine, Houston, USA), Rabbit anti-Appl (1∶1600, L. Torroja, Universidad Autónoma de Madrid, Spain), Rhodamine phalloidin (1∶40, Molecular Probes, Invitrogen). Images were obtained with a Leica SPE confocal microscope and processed with ImageJ and Adobe Photoshop 7.0 software.

In situ hybridization was carried out on eye imaginal discs fixed with 4% formaldehyde using digoxigenin-labeled antisense RNA probes according to standard protocols. Riboprobes for Appl were synthesized using a cDNA clone from DGC (GH04413), sequenced using primers from the SP6 and T7 promoters and linearized with EcoRI for antisense probe. Discs were then analyzed with a Leica DMLB fluorescence microscope. In experiments using Flp/Gal4 clones, anti-digoxigenin and anti-GFP incubations were done simultaneously. Secondary antibody incubation was performed after color development with NBT/BCIP.

Whole-mount in situ hybridization using digoxigenin-labeled antisense RNA probes was carried out using standard protocols. Riboprobes for Appb were synthesized from cDNA of embryos at 6 days post-fertilization. The Tg(hsp70:ca-fgfr1) transgenic line (Marques et al., 2008) was used to overexpress a constitutively active form of the Xenopus fgfr1 by applying a heat shock at 39°C for 1 hour.

The UV/green spectral preference was carried out using a T-maze. Two arms of the T-maze were illuminated with either UV or green LEDs (370 nm from Optosource and 525 nm from Ledman Optoelectronic). Light intensity was calibrated such that control flies preferred the UV side. Flies were kept in darkness for 24 hours and then pre-adapted to white light for 1 hour before testing. They were then locked into a small compartment in the T-maze. Subsequently, the flies were allowed to enter the vials illuminated by either test light for 30 seconds. After each test, the flies in each vial were counted and the Choice Index (CI) calculated from the numbers choosing UV (N_UV) or green (N_G) light by the following formula: CI = (N_UV−N_G)/(N_UV+N_G). The few flies remaining in the center compartment were discarded. For each genotype, three trials (∼30 3- to 5-day-old flies per trial) were carried out. All genotypes for the heterozygous combinations were kept in a Df(1)w background. Flies Df(1)w are viable in homozygosis, carry the w⁻ and y⁻ markers and were used as controls for behavioral studies (Luo et al., 1992).

Detailed procedure has been described before (Perez-Lluch et al., 2011). Third-instar larva eye imaginal discs isolated from w; sev^S11sev-GAL4/UAS-PntP2-HA were immunoprecipitated with anti-HA; three independent replicates were performed. As controls we used two independent replicates for discs lacking PntP2-HA. For PCRs, 2 µl of a 50 µl DNA extraction was amplified with specific primers. PCR bands were quantified using Fujifilm MultiGauge software, normalized against the negative control and depicted as fold enrichment. Specific primers: ETS1 FW 5′-TTCTTCTGACCCACTGCTC-3′, ETS1 RV 5′-GATGAGGGTACGCTGGTAG-3′, ETS2 FW 5′-CCGAGTGTGTGAGCGTGAG-3′, ETS2 RV 5′-AAGCTCTGGACTACGAATGG-3′. The histograms in Fig. 4G represent the mean of the replicates. Standard error of the mean was applied for each experiment.

We thank G. Jimenez, L. Torroja, S. Araujo, S. Pérez, M. Morey and H. Bellen for reagents and discussions, and A. Mateo for technical support. We thank S. Beltran and A. Sànchez for their help with microarray analysis. We thank B. Hassan for insightful suggestions. We also thank the Drosophila Transformation and Bioinformatics (E. Blanco) Platforms of The Consolider Project.