Homeotic Gene teashirt (tsh) Has a Neuroprotective Function in Amyloid-Beta 42 Mediated Neurodegeneration

Background Alzheimer's disease (AD) is a debilitating age related progressive neurodegenerative disorder characterized by the loss of cognition, and eventual death of the affected individual. One of the major causes of AD is the accumulation of Amyloid-beta 42 (Aβ42) polypeptides formed by the improper cleavage of amyloid precursor protein (APP) in the brain. These plaques disrupt normal cellular processes through oxidative stress and aberrant signaling resulting in the loss of synaptic activity and death of the neurons. However, the detailed genetic mechanism(s) responsible for this neurodegeneration still remain elusive. Methodology/ Principle Findings We have generated a transgenic Drosophila eye model where high levels of human Aβ42 is misexpressed in the differentiating photoreceptor neurons of the developing eye, which phenocopy Alzheimer's like neuropathology in the neural retina. We have utilized this model for a gain of function screen using members of various signaling pathways involved in the development of the fly eye to identify downstream targets or modifiers of Aβ42 mediated neurodegeneration. We have identified the homeotic gene teashirt (tsh) as a suppressor of the Aβ42 mediated neurodegenerative phenotype. Targeted misexpression of tsh with Aβ42 in the differentiating retina can significantly rescue neurodegeneration by blocking cell death. We found that Tsh protein is absent/ downregulated in the neural retina at this stage. The structure function analysis revealed that the PLDLS domain of Tsh acts as an inhibitor of the neuroprotective function of tsh in the Drosophila eye model. Lastly, we found that the tsh paralog, tiptop (tio) can also rescue Aβ42 mediated neurodegeneration. Conclusions/Significance We have identified tsh and tio as new genetic modifiers of Aβ42 mediated neurodegeneration. Our studies demonstrate a novel neuroprotective function of tsh and its paralog tio in Aβ42 mediated neurodegeneration. The neuroprotective function of tsh is independent of its role in retinal determination.


Introduction
Alzheimer's disease (AD; OMIM: 104300), first described more than 100 years ago, is an age related, progressive neurodegenerative disorder characterized by the loss of neurons in the hippocampus and cortex. It results in the loss of cognition and memory and eventually leads to the death of the affected individual [1][2][3]. AD can be hereditary or acquired. Familial forms of AD have been associated with mutations in the amyloid precursor protein (APP). Among the multiple causes for AD that have been identified, the accumulation of amyloid plaques around neurons in the brain and the generation of neurofibrillary tangles (NFTs) due to hyperphosphorylation of microtubule binding protein Tau [4,5], are the two major causes for manifestation of AD. The amyloid plaques are formed by improper cleavage of the transmembrane protein APP. APP is proteolytically processed in the extracellular and intracellular domains by β and then γsecretase enzymes [2][3][4][5][6], which leads to the generation of a cytoplasmic fragment that has been implicated in intracellular and nuclear signaling. Normally, APP cleavage results in a forty amino acid long polypeptide (Aβ40), however, improper cleavage of APP results in a forty two amino acid long polypeptide, and hence called as amyloid-beta 42 (Aβ42). These extra two amino acids cause the molecule to become hydrophobic, resulting in the formation of amyloid plaques [2][3][4][5]. The "amyloid hypothesis" suggests that these Aβ42 plaques are toxic in nature and trigger aberrant signaling and disruption of normal cellular processes inside the neuronal cell which lead to loss of synaptic function and death of the neuron [2][3][4]6]. Thus, the accumulation of plaques is responsible for the gradual decline of mental cognition, awareness, and eventual death, of patients who suffer from AD. Therefore, it is important to understand how Aβ42 plaques trigger neurotoxicity and cell death in AD. The genetic mechanism behind the onset of this disease has not been fully understood. Several animal models, like the mouse, fly [7][8][9][10] etc., have been developed to understand the genetic underpinnings of this disease as the genetic machinery is conserved from insects to humans.
Drosophila melanogaster, fruit fly, because of its shorter life cycle, highly conserved signaling pathways, and less functional redundancy has proved to be an important complementary animal model for human diseases. The information generated in fly model system can be extrapolated and tested in mammalian model systems [7,11]. The Drosophila eye has been extensively used to model human neurodegenerative disorders [6,10,[12][13][14][15][16] as important signaling pathways required for the proper development and differentiation of the fly visual system are highly conserved with that of vertebrates [8]. Drosophila, a holometabolous insect, has a blue print for its adult organs housed inside the larvae referred to as the imaginal discs. The larval eye-antennal imaginal disc is a complex disc, which gives rise to the adult compound eye, antenna and head upon differentiation [17][18][19][20]. The retinal precursor cells in the eye imaginal disc undergo differentiation to form the photoreceptor neurons in the adult eye [21][22][23]. Retinal differentiation begins as a synchronous wave from the posterior margin of the eye imaginal disc and proceeds towards the anterior margin of the eye and is referred to as the Morphogenetic Furrow (MF) [21]. The MF leaves behind the differentiated photoreceptor neurons. Eight photoreceptor neurons and several support cells form a unit eye called as the ommatidium. The compound eye of the adult fly is comprised of about 800 ommatidia. In the pupal retina, the excessive cells other than the differentiated cells are eliminated by programmed cell death (PCD) [24]. There is no PCD during earlier stages of larval eye development. However, abnormal extracellular signaling due to inappropriate levels of morphogens may trigger cell death in the developing larval eye imaginal disc [25].
We have utilized the Drosophila eye to model AD [26]. Using a targeted misexpression approach [27], we misexpressed higher levels of human Aβ42 gene in the differentiating photoreceptors of the developing eye using a Glass Multiple Repeat (GMR) Gal4 driver [28]. Misexpression of human Aβ42 in the differentiating photoreceptor neurons of the fly retina exhibits the progressive neurodegenerative phenotypes which mimic the neuropathology of AD patients [26]. In order to discern the genetic basis of Aβ42 mediated neurodegeneration, we have employed our transgenic fly model to analyze the toxic effect of Aβ42 accumulation on signaling pathways through a forward genetic screen. These pathways include the core retinal determination (RD) pathway, which is made up of a cascade of genes viz., PAX-6 homolog eyeless (ey), eyes absent (eya), sine oculis (so) and dachshund (dac), which are responsible for the initiation and differentiation of eye development [18][19][20]29,30]. We also tested negative regulators of eye development like homothorax (hth) [31,32]. hth is known to suppress retinal differentiation in the eye and promote the head specific fate [33]. Another negative regulator of eye development is Wingless (Wg), a member of the highly conserved Wnt/ Wingless signaling pathway, which is responsible for regulating early growth, dorso-ventral (DV) lineage and, during the latter part of development, restricting the eye fate on outer dorso-ventral margins of the developing eye [34,35]. Wnt signaling is antagonized by another highly conserved TGF beta (TGFβ) signaling pathway, referred to as Decapentaplegic (Dpp) signaling in Drosophila [36][37][38]. Dpp signaling collaborates with Hedgehog (Hh) signaling to promote retinal differentiation in the developing eye as well as antagonize Wg signaling [39]. Another signaling pathway tested includes Notch (N) signaling pathway. N is a transmembrane receptor activated by "DSL" class ligands which is involved in: cell proliferation and differentiation during eye development, setting up the dorsal-ventral compartment boundary, planar polarity and spacing of the ommatidial clusters, and in cell fate specification [40]. In the developing eye imaginal disc, N activity is highest at the equator, the boundary between the dorsal and ventral halves of the eye. N pathway also employs the same secretases which are involved in the processing of APP for the cleavage of the protein [41]. We tested components of all these pathways in our forward genetic screen based on a premise similar to our earlier reported gain-of-function screen in the developing Drosophila eye [42].
We have identified teashirt (tsh) as a modifier of Aβ42 mediated neurodegeneration. A homeotic gene, tsh encodes a zinc finger transcription factor. The full length Tsh protein is 954 amino acids long and has three DNA binding zinc finger domains and a N-terminal PLDLS domain that is required for interaction with CtBP protein [31,[43][44][45][46][47]. In an enhancer trap screen for genes involved in embryonic development, tsh was first identified and assigned a role in specifying trunk identity during embryogenesis through its interactions with the Hox protein network [45,47]. However, tsh is also known to play additional roles in other tissues. The role of tsh in eye development was first reported in an enhancer trap screen for genes exhibiting domain specific expression of the mini-white reporter gene during eye development [31]. In the developing Drosophila eye imaginal disc, tsh exhibits domain specific function by promoting growth in the dorsal eye, whereas it suppresses eye fate in the ventral by ectopically activating Wg signaling [19,46,48]. tsh has also been shown to induce ectopic eyes [31], while its zinc fingers and PLDLS binding motif have been shown to be involved in its ectopic eye formation function [49]. The paralog of tsh, tiptop (tio) [50], encodes a 1025 amino acids long C 2 H 2 zinc finger protein of the Teashirt family [47]. It has been shown that tio and tsh have similar expression patterns in the eye [44,47,49]. It has also been suggested that tio exhibits functional redundancy with tsh through a mutual repression mechanism [43,44].
Here we report a novel neuroprotective function of tsh in Aβ42 mediated neurodegeneration. We demonstrate that upregulation of full length Tsh expression can significantly rescue the Aβ42 mediated neurodegenerative phenotype by blocking cell death of neurons. The structure function analysis determined that the PLDLS domain of Tsh acts as a suppressor of its neuroprotective function in Aβ42 mediated neurodegeneration. We have identified tio, a paralog of tsh [50], which can also rescue Aβ42 mediated neurodegeneration. Furthermore, the structure function analysis determined that the Zn4 and PLDLS domains of tio act as suppressors of the neuroprotective function of tio in Aβ42 induced neurodegeneration.
We have employed a Gal4/UAS system for targeted misexpression studies [27]. All Gal4/UAS crosses were maintained at 18°C, 25°C and 29°C, unless specified, to sample different induction levels. The adult flies were maintained at 25°C, while the cultures after egg laying (progeny) were transferred to 29°C for further growth. The misexpression of Aβ42 in the differentiating retina (GMRGal4>UAS-Aβ42) exhibits a stronger neurodegenerative phenotype at 29°C with no penetrance [26]. All the targeted misexpression experiments were conducted using the Glass Multiple Repeat driver line (GMR-Gal4) which directs expression of transgenes in the differentiating retinal precursor cells of the developing eye imaginal disc and pupal retina [28].

Immunohistochemistry
Eye-antennal imaginal discs were dissected from wandering third-instar larvae in 1X PBS and stained following the standard protocol [46]. Eye-imaginal discs were fixed in 4% paraformaldehyde and stained with a combination of antibodies using the standard protocol. Primary antibodies used were

Detection of Cell Death
Cell death was detected using TUNEL assays [26,[66][67][68]. TUNEL assays were used to identify the cells undergoing cell death where the cleavage of double and single stranded DNA is labeled by a Fluorescein. The fluorescently labeled nucleotides are added to 3' OH ends in a template-independent manner by Terminal Deoxynucleotidyl Transferase (TdT). The fluorescent label tagged fragmented DNA within a dying cell can be detected by fluorescence or confocal microscopy. Eyeantennal discs after secondary antibody staining [69] were blocked in 10% normal donkey serum in phosphate buffered saline with 0.2% Triton X-100 (PBT) and labeled for TUNEL assays using a cell death detection kit from Roche Diagnostics.
The TUNEL positive cells were counted from five sets of imaginal discs and were used for statistical analysis using Microsoft Excel 2010. The P-values were calculated using onetailed t-test and the error bars represent Standard Deviation from Mean [26].

Scanning Electron Microscopy
All the flies were prepared through a series of increasing acetone concentration treatments following the standard protocol [70]. Each sample was then treated in 1:1 acteone/ HMDS (Hexa Methyl Di Silazane, Electron Microscopy Sciences) solution for 24 hours. This was followed by treatment with 100% HMDS and allowed to air dry in the hood. Dehydrated flies were mounted on Electron microscopy stubs and coated with gold using a Denton vacuum sputter coater. Images were captured using a Hitachi S-4800 High Resolution Scanning Electron Microscope (HRSEM). The final images and figures were prepared using Adobe Photoshop CS4 software.

Adult Eye Imaging
Adult flies were prepared for imaging by freezing at -20ºC for approximately 2 hours followed by mounting the fly on a dissection needle. The needle with fly was suspended horizontally over a glass slide using molding putty. Images were captured on a MrC5 color camera mounted on an Axioimager.Z1 Zeiss Apotome using Z-sectioning approach. Final images were generated by compiling the individual stacks from the Z-sectioning approach using the extended depth of focus function of Axiovision software version 4.6.3.

Identification of genetic modifiers of Aβ42 mediated neurodegeneration
We have generated a transgenic fly model where GMR enhancer drives the expression of the human Aβ42 gene (GMR>Aβ42) in differentiating photoreceptor neurons of the developing retina [26,28]. Accumulation of Aβ42 triggers aberrant signaling mechanism(s) and impairs the basic cellular processes leading to the death of neurons in the developing neural retina of pupa and the adult eye ( Figure 1, Table 1) [26]. However, the genetic basis of Aβ42 mediated neurodegeneration has not been fully understood. We performed a forward gain-of-function genetic modifier screen using a candidate gene approach [42] to identify the downstream targets or genetic modifiers of Aβ42 mediated neuropathy in the developing Drosophila eye. We looked for modifiers of the GMR>Aβ42 phenotype when we individually misexpress the member genes of various highly conserved signaling pathways ( Figure 1, Table 1). The premise of the screen was based on the observation that the neurodegenerative phenotypes of the GMR>Aβ42 exhibits no penetrance. Therefore, any deviation in this phenotype can be attributed to misexpression of the gene(s) of interest.
We tested several candidates which included candidates of retinal determination (RD), negative regulators of eye development homothorax (hth), and members of highly conserved signaling pathways such as Wg, Dpp, Hh and N signaling pathways (Table 1). In comparison to the wild type compound eye ( Figure 1A), the adult eye of GMR>Aβ42 is highly reduced in size with a slit like appearance and a glazed surface with black necrotic spots where extensive cell death has occurred ( Figure 1B). Targeted misexpression of tsh with Aβ42 (GMR>Aβ42+tsh) resulted in significant rescue of this neurodegenerative phenotype ( Figure 1C, Table 1). The frequency of the rescue phenotypes due to misexpression of the tsh was significantly higher. Nearly 75% of the GMR>Aβ42+tsh flies showed the strong rescue in the adult eye. The remaining 25% flies showed a weaker rescue phenotype. Members of the core retinal determination pathways, such as ey (GMR>Aβ42+ey; Figure 1D, Table 1), eya (GMR>Aβ42+eya; Figure  1E, Table  1), so (GMR>Aβ42+so; Figure  1F, Table  1), and dac (GMR>Aβ42+dac; Figure 1H, Table 1) did not show any significant rescue of the neurodegenerative phenotype of GMR>Aβ42. Furthermore, misexpression of so alone (GMR>so) results in reduced eye phenotype due to roughening of the anterior half of the eye and elimination of retinal tissue in the posterior half [71]. It has been shown that GMR>eya had rough eyes [72]. Misexpression of dac in particular resulted in the most severe enhancement ( Figure 1H) as compared to the other RD genes. The resulting GMR>Aβ42+dac phenotype was so strong that most of the progeny died as pharate and failed to hatch out of their pupal cases. Although GMR>Aβ42+ey and GMR>Aβ42+eya showed some overall increase in the size of the eye field, these eyes had a glazed surface morphology and lacked any ommatidia and pigment cells which allowed us to conclude that the neurodegenerative phenotype of GMR>Aβ42 was not rescued. It is known that Eya physically associates with So to perform a retinal differentiation function in the developing eye [51]. Targeted misexpression of both eya and so together (GMR>Aβ42+eya+so) did not result in rescue of the neurodegenerative phenotype ( Figure 1G, Table 1). We also tested negative regulators of the eye such as homothorax (hth) (GMR>Aβ42+hth; Figure 1I, Table 1), which resulted in a subtle increase in the eye field but lacked any ommatidia. Misexpression of a dominant negative allele of hth, hth ENR , (GMR>Aβ42+ hth ENR ), exhibited a significant increase in the eye field but these eyes had a neurogenic phenotype with increased numbers of bristles and lacked any pigment cells ( Figure 1J, Table 1). Thus, there was only an increase in the number of cells with no restoration of the eye phenotype. Therefore, the members of the core retinal determination pathway and negative regulators of the eye did not have any major role in Aβ42 mediated neurodegeneration. We also tested the controls to verify our results (data not shown).
Furthermore, misexpression of wg (GMR>Aβ42+wg, Figure  1K, Table 1) resulted in a strong enhancement of the Aβ42 mediated neurodegenerative phenotype. All adult eyes were highly reduced in size with a glazed surface. The role of Wg signaling in Aβ42 mediated neurodegeneration was further validated using shaggy (sgg), an antagonist of Wg signaling [54]. Misexpression of sgg (GMR> Aβ42+sgg) resulted in a significant rescue, as evident from the increased size of the adult eye along with near wild-type looking ommatidia and bristles ( Figure 1L; Table 1). Thus, modulating Wg signaling can modify Aβ42 mediated neurodegeneration. Interestingly, Dpp signaling is known to antagonize Wg signaling in the eye as well [37][38][39]. Misexpression of dpp (GMR>Aβ42+ dpp) resulted in a rescue ( Figure 1M, Table 1), which is similar to what was observed upon blocking Wg signaling by misexpression of sgg ( Figure 1L). The frequency of the rescue phenotypes in the adult eyes were siginificantly higher. Other candidates tested were optomotor blind (omb) (GMR> Aβ42+omb, Figure 1O; Table 1) and N (GMR> Aβ42+N, Figure  1N; Table 1), and they did not affect the neurodegenerative phenotype. Blocking Hh signaling pathway by using misexpression of patched (ptc) (GMR> Aβ42+ptc, Figure 1P, Table 1) resulted in a subtle increase in eye field with no rescue of neurodegeneration. Our screen resulted in identification of a homeotic gene, tsh, and members of Wg and Dpp signaling pathways as modifiers of Aβ42 mediated neurodegeneration. Surprisingly, tsh was able to provide one of the strongest rescues ( Figure 1C; Table 1), and we therefore pursued to verify the neuroprotective role of tsh in Aβ42 mediated neurodegeneration. We also tested the controls to verify our results (data not shown).

tsh Is a Genetic Modifier of Aβ42 Mediated Neurodegeneration
Targeted misexpression of full length tsh in the GMR>Aβ42 background (GMR> Aβ42+tsh; Figure 2C) significantly rescued the Aβ42 neurodegenerative phenotype of a highly reduced eye field with glazed appearance ( Figure 2B). Even though a strong rescue was observed in the adult eyes of GMR>Aβ42+tsh flies ( Figure 2C), the Aβ42 mediated neurodegenerative phenotype was not completely restored to the wild type adult eye phenotype (Figure 2A). Furthermore, we found that the rescue of neurodegeneration by tsh is spatial in nature as the tsh mediated rescue was restricted to the anterior half of the adult eye ( Figure 2C; marked by yellow dotted line) whereas neurodegeneration still persists in the posterior half of the adult eye. In terms of the chronology of differentiation, the  Figure S1a), however, the adult eye was slightly reduced in size and lacking neurodegeneration on the posterior margin ( Figure S1B).
We investigated the status of tsh transcription using a lacZ reporter [63] and Tsh protein levels in the GMR>Aβ42 background. During early eye development, both tsh lacZ and Tsh protein are expressed in the entire early eye primordium [46,73]. During retinal differentiation in the third instar eye imaginal disc, tsh lacZ ( Figure 2D, D') and Tsh protein ( Figure  2D, D") expression retracts anterior to the MF and Tsh is not localized in the differentiating photoreceptor neurons ( Figure  2D-D"). However, the tsh lacZ domain exhibited less retraction as compared to the Tsh protein, probably due to perdurance of the lacZ protein, which serve as a reporter of tsh transcription. We found that tsh lacZ ( Figure 2E, E') and Tsh protein ( Figure  2E, E") levels were not affected in the GMR>Aβ42 background ( Figure 2E, E"). Therefore, Tsh is absent in the developing neural retina when GMR>Aβ42 mediated neurodegeneration occurs. However, targeted misexpression of tsh with GMR>Aβ42 (GMR>Aβ42+ tsh) resulted in the rescue of Aβ42 mediated neurodegeneration and showed strong accumulation of Tsh protein in the differentiating photoreceptor neurons ( Figure 2F, F"). Furthermore, we found that tsh lacZ expression was not induced in the differentiating neurons suggesting that Tsh protein, when misexpressed in a developing eye field, can provide neuroprotection. It also indicates that Tsh protein does not regulate its own transcription in the developing eye.
Since the neurodegeneration phenotype of GMR>Aβ42 is progressive over the course of development, we analyzed Tsh levels in the pupal retina ( Figure 2G-I). In the wild-type pupal retina, the nuclei of the differentiated neurons, marked by pan neural marker Elav ( Figure 2G, G'), exhibited weak expression of Tsh ( Figure 2G, G"). However, in the GMR>Aβ42 background, Tsh expression appears diffused probably due to the fact that photoreceptor nuclei are being disintegrated ( Figure 2H, H' H"). Furthermore, there is clumping of the ommatidial nuclei in the pupal retina (Fig, 2H, H", marked by white arrow head), which results in holes in the pupal retina ( Figure 2H', marked by yellow dotted line). However, in GMR>Aβ42+tsh, where robust expression of Tsh protein is observed, a significant rescue of the neurodegenerative phenotype is seen ( Figure 2I, I', I"). The neuroprotective function of Tsh is evident from the regularly placed photoreceptor nuclei in the ommatidia and lack of holes in the retina of GMR>Aβ42+tsh ( Figure 2I, I', I").

Tsh can block induction of cell death
To further validate our hypothesis of a neuroprotective role of tsh in Aβ42 mediated neurodegeneration, we investigated the rate of cell death using a TUNEL staining approach. TUNEL staining marks the fragmented ends of DNA of the dying cells nuclei [67,68]. In comparison to the wild-type eye imaginal disc, which exhibits a few cells undergoing cell death based on few TUNEL positive cells ( Figure 3A, A'), the GMR>Aβ42 eye imaginal disc exhibits a significantly higher number of TUNEL positive cells ( Figure 3B, B', D). A higher number of TUNEL positive cells can explain the highly reduced eye size of the GMR>Aβ42 adult eye [26]. We counted the number of dying cells in the wild type as well as GMR>Aβ42 eye disc and found that the number of dying cells increases by more than three fold in the GMR>Aβ42 disc as compared to the wild-type eye ( Figure 3D). Furthermore, targeted misexpression of tsh with GMR>Aβ42 (GMR>Aβ42+tsh) results in a significant rescue of the neurodegenerative phenotype where the number of TUNEL positive dying cells was reduced to half the number of cells undergoing cell death in the GMR>Aβ42 eye disc ( Figure 3C, C', D). These results further validate our hypothesis that tsh can provide neuroprotection against GMR>Aβ42 mediated neurodegeneration.

Tsh can rescue axonal targeting defects
During development of the Drosophila visual system connections are generated from the neural retina to the brain by means of axonal targeting [74,75]. The differentiating neurons send out axons that lead by their growth cones and precisely trace their appropriate synaptic targets. In the developing eye imaginal disc, each differentiated photoreceptor neuron send axons [74]. The axons from the photoreceptors of ommatidia fasciculate together to form an ommatidial bundle. The ommatidial bundle pierces through the basement membrane of the eye disc and then extends to the posterior edge of the eye disc, and through the optic stalk innervates the  Note that the number of dying cells increased more than three folds in the GMR>Aβ42 background as compared to the wildtype eye imaginal disc. The number of dying cells in GMR>Aβ42+tsh background is reduced to half as compared to GMR>Aβ42. Although the number of dying cells in GMR>Aβ42+tsh background it is still more than the wild-type eye disc. (E-G) Photoreceptor cells projections in third instar larva visualized using Chaoptin (MAb24B10, green channel) [65] staining. MAb24B10 marks the retinal axons from the neural retina to the brain, and proneural marker Elav (red channel). (E') In the wild type larva, retinal axons projection pattern from the photoreceptors in the retina to the optic lobes in the brain. Note that ommatidial axonal bundle from eye field contacts the brain at two locations in a highly organized fashion in the wild type. (E', F') Misexpression of (F') Aβ42 (GMR>Aβ42) results in aberrant retinal axon targeting from the neural retina to the brain (E") as compared to the wild type. (G') Targeted misexpression of tsh in the Aβ42 background (GMR>Aβ42+ tsh) results in a strong restoration of retinal axonal targeting. The magnification of confocal images of (A-C) the Eye-antennal imaginal disc is 20X, and (E-G) the retinal axon is 60X. doi: 10.1371/journal.pone.0080829.g003 different layers of the brain [76,77]. The axons from photoreceptors (PRs) 1-6 terminate in the lamina whereas PR7-PR8 terminates in separate layer of medulla after passing through lamina. The retinal axons can be marked by Chaoptin (MAb24B10) [65]. In the wild-type eye imaginal disc MAb24B10 marks the axons which innervate the different layers of the optic lobe of the brain ( Figure 3E, E'). However, in the GMR>Aβ42 eye imaginal disc, the retinal axon targeting becomes impaired as we can no longer observe axons innervate properly in the optic lobes of the brain ( Figure 3F, F'). Misexpression of tsh in the GMR>Aβ42 background (GMR>Aβ42+tsh) can not only restore the size of the eye field but can also significantly restore the retinal axon targeting phenotype (Figure 3 G, G').

PLDLS domain is a negative regulator of the neuroprotective function of Tsh
tsh encodes a C 2 H 2 zinc finger transcription factor protein which consists of a PLDLS domain and three zinc finger domains [19,[44][45][46][47][48][49]73,78]. We performed a structure function analysis to determine the role of various domains of the Tsh protein in its neuroprotective function by utilizing various deletion constructs of tsh. These constructs were generated by individually removing each domain from the full length Tsh protein [49]. We used these constructs to misexpress truncated Tsh protein with Aβ42 and then screen for the domain required for its neuroprotective function. We found that misexpression of truncated Tsh lacking the Zn1 domain (GMR>Aβ42+tsh∆Zn1) ( Figure 4A, B, C), Zn2 domain (GMR>Aβ42+tsh∆Zn2) ( Figure  4A, D, E) or Zn3 domains (GMR>Aβ42+tsh∆Zn3) (Figure 4A, F, G) was able to rescue the GMR>Aβ42 phenotype as seen in the eye antennal imaginal disc (Figure 4C, E, G) as well as the adult eye (Figure 4 B, D, F), respectively. Interestingly, the rescue by truncated tsh lacking these three zinc finger domains were comparable to the full length tsh as seen in (Figure 1C, 2C, F). Therefore, removing zinc finger domains does not affect the neuroprotective function of tsh. Misexpression of truncated Tsh lacking the PLDLS domain (GMR>Aβ42+tsh∆PLDLS) showed a stronger rescue of Aβ42 mediated neurodegeneration in the eye imaginal disc and the adult eye ( Figure 4A, H, I). Interestingly, the rescue by truncated Tsh lacking the PLDLS domain was much stronger than the rescue by the full length Tsh, suggesting that the PLDLS domain acts as a suppressor of the neuroprotective function of tsh. Our results strongly suggested that DNA binding domains of tsh such as Zn1, Zn2 and Zn3 are either functionally redundant or play no significant role in the neuroprotective function of tsh. tio can rescue Aβ42 mediated neurodegeneration tsh shares a regulatory relationship with its paralog -tio [44,47,49], which is expressed in a similar pattern and encodes a protein with four zinc finger domains and a N-terminal CtBP domain ( Figure 5A). Targeted misexpression of full length tio in the GMR>Aβ42 background (GMR>Aβ42+tio) showed a significant rescue of the Aβ42 mediated neurodegeneration phenotype as seen in the ( Figure 5B) adult and ( Figure 5C) the imaginal disc. We observed a significantly stronger rescue in terms of increased eye size with normal looking, regularly arranged, ommatidia and little or no patches of necrosis in the anterior region ( Figure 5B). Furthermore, the rescue of Aβ42 mediated neurodegeneration by tio is similar to tsh ( Figure 2C).
To determine the role each domain of tio plays in its neuroprotective function, we carried out a structure function analysis using deletion constructs of tio generated by individually removing each domain from the full length protein [49]. We used these constructs to misexpress truncated Tio protein with Aβ42 and then screened for the domain required for its neuroprotective function. We found that misexpression of truncated Tio lacking the Zn1 domain (GMR>Aβ42+tio∆Zn1) ( Figure 5A, D, E), Zn2 domain (GMR>Aβ42+tio∆Zn2) ( Figure  5A, F, G) or Zn3 domain (GMR>Aβ42+tio∆Zn3) ( Figure 5A, H, I) was able to rescue the GMR>Aβ42 phenotype as seen in the eye antennal imaginal disc ( Figure 5 E, G, I) as well as the adult eye ( Figure 5 D, F, H), respectively. Interestingly, the rescues by truncated tio lacking Zn1 or Zn2 or Zn3 domains were comparable to the full length tio as seen in (Figure 5A,B, C), suggesting that these domains are not required for the neuroprotective function of tio. We found that misexpression of truncated Tio lacking the Zn4 domain (GMR>Aβ42+tsh∆Zn4; Figure 5A, J, K) or PLDLS domain (GMR>Aβ42+tsh∆PLDLS; Figure 5 A, L, M) domain showed a significant rescue of Aβ42 mediated neurodegeneration as seen in the adult eye ( Figure  5A, J, L) and the eye imaginal disc ( Figure 5A, K, M), respectively. Therefore, removal of the Zn4 ( Figure 5A, J, K) or the PLDLS ( Figure 5A, L, M) domain from the full length Tio increased the intensity of rescue as compared to the full length tio ( Figure 5A, B, C), suggesting that the Zn4 and the PLDLS domain act as suppressor of the neuroprotective function of tio.

Discussion
Several signaling pathways may play a role in GMR>Aβ42 mediated neurodegeneration. Generally, accumulation of GMR>Aβ42 plaques triggers some aberrant signaling response which finally triggers abnormal signaling leading to generation of stress in the neurons and finally culminating in the death of the neurons [2][3][4][5][6][7]26,79]. One of the most important facets of this process is to understand the downstream targets of amyloid beta mediated neurodegenerative response. A genome wide forward genetic screen can be labor intensive and therefore, we employed a candidate gene approach where we picked up the candidates of the various signaling pathways and tested them individually. Our candidate genes approach for the forward genetic screen resulted in identification of homeotic gene tsh as a neuroprotective agent. Interestingly, tsh has been shown to be involved in patterning, growth and retinal development [31,45,46,73,78]. However, its role as a neuroprotective agent has not been fully understood.

Tsh has a neuroprotective role in Aβ42 mediated neurodegeneration
Our data suggests that misexpression of Tsh in the differentiating photoreceptor neurons of the fly retina can rescue the neurodegenerative phenotype of GMR>Aβ42 mediated neurodegeneration. We found that the neuroprotective function of tsh is mediated through prevention of induction of cell death ( Figure 3). Earlier, we showed that GMR>Aβ42 mediated cell death is both caspase dependent and caspase independent [26]. Since tsh can significantly restore the neurodegenerative phenotype ( Figure 2C), it is expected that tsh mediated rescue might affect either one or both of them significantly. It will be interesting to look for the mechanism by which tsh can prevent induction of neurodegeneration in the GMR>Aβ42 background. Tsh is a homeotic gene and is involved in several signaling pathways to regulate patterning and growth [19,31,45,46,73,78]. Tsh is known to be involved in regulating retinal development in Drosophila and has the capability to induce ectopic eyes [31]. Since our disease model is restricted to the retina of the fly, we wanted to test if the neuroprotective function of Tsh is mediated through its role in the retinal differentiation pathway [19,31,46,48,78].

Neuroprotective function of Tsh is independent of RD gene function
It has been shown that tsh can induce eya and so to form an ectopic eye in the antenna [31]. Based on these results, tsh was assigned to the category of genes which are involved in eye development even though they do not belong to the core retinal determination pathway [19]. We investigated the role of RD genes in the neuroprotective function of tsh by testing levels of the RD genes eya and dac in the third instar eye imaginal disc. Eya, a tyrosine phosphatase, is involved in the retinal differentiation process and is expressed both in the differentiating photoreceptor neurons of the neural retina as well as the retinal precursor cells anterior to the MF ( Figure 6A, A') [17,18,30]. In the GMR>Aβ42 background third instar eye imaginal disc we found that Eya expression is not affected ( Figure 6B) and in the GMR>Aβ42+tsh background, we did not see any ectopic induction or increased levels of Eya in the differentiating neurons ( Figure 6C). We also investigated levels of dac, another RD gene [52], which is expressed in two different domains one anterior to the MF and another posterior to the MF ( Figure 6D). In the GMR>Aβ42 ( Figure 6E) as well as the GMR>Aβ42+tsh background ( Figure 6F), we did not find any change in the levels of Dac expression. Our data strongly suggests that the neuroprotective role of Tsh is independent of its role in RD gene function. This data is consistent with the result from the genetic screen where we found that increasing levels of the RD gene members did not affect the GMR>Aβ42 mediated neurodegenerative phenotype ( Figure 1, Table 1). Therefore, the neuroprotective role of tsh is a novel function of this genetic locus.

Tsh neuroprotective role is downstream of FE65 mediated regulation of APP cleavage
Aβ42, the major component of amyloid plaques, is a byproduct of the improper cleavage of Amyloid Precursor Protein (APP) protein [80]. Tsh has been shown to be involved in processing of Amyloid Precursor Protein (APP) cleavage. It has been shown that APP can bind to FE65 protein and this complex can regulate gene expression. FE65 is an adaptor protein with two phosphotyrosine binding (PTB) domains and a single WW domain [81,82]. In a yeast two hybrid screen using the PTB domain of FE65 protein as a bait, Tsh was identified as an FE65 interacting protein in the neurons [82]. The coimmunprecipitation studies showed a direct interaction of FE65 and Tsh 3, one of the vertebrate homologs of Tsh, with the promoter region of Caspase 4 [82]. It was demonstrated that Tsh can work in a protein complex to trigger cell death in neuritic plaques of AD. Interestingly, in Drosophila there is a no homolog of FE65 present. Therefore, Tsh cannot physically associate with FE65. Therefore, in our studies the neuroprotective function of tsh is due to prevention of neurodegeneration caused by accumulation of human Aβ42 in the system. The mechanism by which Tsh can prevent neurodegeneration is yet to be fully understood.  [49]. The full length Tio protein consists of a PLDLS domain and four DNA binding Zinc finger domains [49]. A membrane specific marker Disc large (Dlg, white channel) was used to mark the outline of cells and pan neural marker Elav (red channel) marks nuclei of the photoreceptor neurons. Targeted

Whether Tsh blocks or delays the onset of neurodegeneration
Our studies also raised an interesting question of whether Tsh mediated neuroprotective phenotype which has spatial component as the rescues were restricted to the anterior half of the adult eye ( Figure 2C), has some temporal component. There were no examples of rescues in the posterior part of the adult eye. In the Drosophila eye, differentiation initiates from the posterior margin of the developing eye imaginal disc and moves in a synchronous fashion towards the anterior margin of the eye field. Therefore, in terms of chronology, the ommatidia on the posterior margin are older in comparison to the ones towards the anterior margin of the eye imaginal disc [21,22]. Interestingly, targeted coexpression of tsh with Aβ42 (GMR> Aβ42+tsh) results in significant rescue of the eye but the rescue is restricted to the anterior half of the adult eye ( Figure  2C). It raises an interesting possibility that either it takes time for the tsh levels to build up and as a result it cannot rescue neurodegeneration in the older retinal neurons of the posterior half of the adult eye, or tsh alone can only delay the onset of neurodegeneration in the neural retina and as a result the adult eye exhibits rescue in the anterior half. We therefore analyzed the adult eye phenotype of GMR>Aβ42+tsh adults of different ages. We found that the size of the eye field reduces with age. In comparison to the freshly eclosed one day old fly, the thirty day old fly eye is reduced along with increased loss of pigment cells and more necrotic spots ( Figure S2).
We also investigated whether a paralog of tsh, tio, can rescue the neurodegenerative phenotype and we found functional redundancy between tsh and tio. The different functional domains of Tsh and Tio tested for the neuroprotective function showed that PLDLS domain act as suppressor in both Tio and Tsh (Figure 4, 5). These results further validate that the functional redundancy observed between Tsh and Tio during retinal development [44,49] may also hold true in terms of their neuroprotective functions. Our