Glial expression of disease-associated poly-glutamine proteins impairs the blood–brain barrier in Drosophila

Abstract

Expansion of poly-glutamine (polyQ) stretches in several proteins has been linked to neurodegenerative diseases. The effects of polyQ-expanded proteins on neurons have been extensively studied, but their effects on glia remain unclear. We found that expression of distinct polyQ proteins exclusively in all glia or specifically in the blood–brain barrier (BBB) and blood–retina barrier (BRB) glia caused cell-autonomous impairment of BBB/BRB integrity, suggesting that BBB/BRB glia are most vulnerable to polyQ-expanded proteins. Furthermore, we also found that BBB/BRB leakage in Drosophila is reflected in reversed waveform polarity on the basis of electroretinography (ERG), making ERG a sensitive method to detect BBB/BRB leakage. The polyQ-expanded protein Atxn3-84Q forms aggregates, induces BBB/BRB leakage, restricts Drosophila lifespan and reduces the level of Repo (a pan-glial transcriptional factor required for glial differentiation). Expression of Repo in BBB/BRB glia can rescue BBB/BRB leakage, suggesting that the reduced expression of Repo is important for the effect of polyQ on BBB/BRB impairment. Coexpression of the chaperon HSP40 and HSP70 effectively rescues the effects of Atxn3-84Q, indicating that polyQ protein aggregation in glia is deleterious. Intriguingly, coexpression of wild-type Atxn3-27Q can also rescue BBB/BRB impairment, suggesting that normal polyQ protein may have a protective function.

Introduction

Aberrant poly-glutamine (polyQ) expansion has been associated with several human neurodegenerative diseases, such as spinocerebellar ataxia (SCA, types 1–3, 6, 7, 17), dentatorubro-pallidoluysian atrophy, spinobulbar muscular atrophy and Huntington’s disease (HD) (1). These polyQ diseases are all dominantly inherited and cause adult onset neuronal loss. A general feature of these polyQ disease genes is aberrant expansion of their CAG trinucleotide repeats, leading to polyQ expansion in the encoded proteins, which then form protein aggregates in various types of tissues (2). Whether these protein aggregates are disease-causing or have a protective role is still under debate (3,4). Although these polyQ-expanded proteins are widely expressed and form aggregates in neurons and glia in both human and mouse models (5–7), most studies have focused on neurons. PolyQ proteins affect different types of glia (8,9). Studies on animal models and induced pluripotent stem cells (iPSCs) of HD patients have shown that mutant Huntingtin (HTT), the causal agent for HD, can directly affect glia and the effects vary in different glial cell types (10–18). In Drosophila, targeted expression of polyQ disease-associated proteins in glia can cause degeneration of the nervous system and reduced lifespan (19–24). Here, we focus on the effect of the polyQ-expanded proteins, Huntingtin (Htt) and Ataxin 3 (Atxn3), in the glia that constitute the blood–brain barrier (BBB) and blood–retina barrier (BRB) of Drosophila. Spinocerebellar ataxia 3 (SCA3), also known as Machado–Joseph disease (MJD), is the most prevalent dominantly inherited form of ataxia and is caused by mutant Atxn3 (25,26). Htt with polyQ expansion is the causal agent for HD (27,28). HD and SCA3 represent two distinct types of neurodegenerative diseases. Neuronal expression of both Htt and Atxn3 with long polyQ stretches has been used to establish models of Drosophila neurodegeneration (29,30), and mutant Htt has been expressed in glia to study its effect on neurodegeneration (21–23). Our results suggest that these two proteins share a common pathogenic effect.

The BBB forms a diffusion barrier between the nervous system and the blood, such that large molecules in the blood cannot freely enter the nervous system. In mammals, the BBB is comprised of vascular endothelial cells surrounded by astrocytes (31). The Drosophila BBB is formed by an apical layer of extracellular matrix, followed by a neural lamella, the perineurial glia (PG) and the basal subperineurial glia (SPG) with septate junctions (SJs, the insect equivalent of vertebrate tight junctions) between SPGs (32,33). SJs provide the insulator function, whereas the PG and neural lamella act in barrier selectivity (32).

The eye is similarly shielded by the BRB. The mammalian eye has two types of BRB; an inner BRB composed of endothelium of the retinal blood vessels that is similar to the BBB, and an outer BRB composed of retinal pigment epithelium connected by tight junctions (34). The Drosophila eye is also surrounded by a BRB (35,36). The Drosophila and vertebrate BBBs are similar in many structural and functional aspects (37–40). Morphologically, the fly BBB and BRB are similar to the mammalian BRB in that they surround the organ being shielded.

BBB impairment has been found in many neurodegenerative diseases (reviewed in 41,42). Although mutant HTT has been shown to form aggregates in BBB components of HD patients (43), whether BBB impairment occurs in polyQ-induced neurodegenerative diseases remains controversial (43–46). Because mutant HTT is expressed in both neurons and glia in both R6/2 (47) and YAC128 (48) HD models, it has not been established whether BBB impairment is a direct effect of polyQ in BBB astrocytes. Because the Drosophila BBB and BRB are similar to the mammalian BBB in many respects, they represent appropriate experimental systems for understanding the link between the BBB and polyQ-induced neurodegeneration.

In this study, we expressed mutant human polyQ proteins in Drosophila glia and examined their effects on the nervous system. We used electroretinograms (ERGs) as an electrophysiological measurement of photoreceptor signal transmission in response to a flash of light (49), on the basis of extracellular recording of the graded potential from photoreceptor axons that are insulated from the signal by underlying lamina neurons (50,51). A reversed ERG phenotype was previously identified in the reverse polarity mutant, repo¹, which encodes a glia-specific homeodomain protein expressed in most glia and that is required for glial differentiation (52,53). We reveal that expression of polyQ-expanded proteins in glia can cause reversed ERG polarity, i.e. similar to the repo¹ phenotype, and show that this is due to a BBB/BRB glial cell-autonomous effect that impaired the BBB and BRB and non-autonomously caused neuronal degeneration.

Results

Glial expression of polyQ-expanded proteins induces reversed ERG polarity

To test the effect of polyQ-expanded proteins in glia, we expressed Atxn3 and Htt in Drosophila glia using the pan-glia repo-GAL4 driver (54). Wild-type Atxn3 carries 27 glutamines (Atxn3-27Q), whereas mutant Atxn3 contains 84 glutamines (Atxn3-84Q). Wild-type full-length Htt contains 16 glutamines (Htt^FL-16Q), whereas mutant Htt contains 128 glutamines (Htt^FL-128Q). Pan-glial expression of Atxn3-84Q-Myc (repo-GAL4 + UAS-Atxn3-84Q-Myc, abbreviated as repo > Atxn3-84Q-Myc) did not cause any noticeable morphological or behavioral abnormality in young flies, but they did exhibit reversed ERG polarity with complete penetrance (Fig. 1A and C). Similarly, repo > Htt^FL-128Q caused reversed ERG polarity (Fig. 1E). In contrast, glial expression of wild-type Atxn3-27Q-Myc or Htt^FL-16Q did not result in the same phenotype (Fig. 1B and D). Pan-neuronal expression of Atxn3-84Q-Myc or Htt^FL-128Q by the elav-GAL4 driver (elav > Atxn3-84Q-Myc and elav > Htt^FL-128Q, respectively) did not cause reversed ERG polarity (Fig. 1F and G). Expression of Htt^FL-128Q in both neurons and glia (repo + elav > Htt^FL-128Q) did not further exacerbate the reversed ERG polarity phenotype (Fig. 1H). Interestingly, this ERG polarity reversal is similar to the original phenotype identified for the repo¹ mutant (Fig. 1I) (52). Our results show that glia-specific expression of polyQ-expanded Atxn3 or Htt can cause a phenotype distinct from that via expression in neurons. In subsequent analyses, we used Atxn3-84Q-Myc as representing polyQ-expanded proteins.

Figure 1.

Expressing polyQ-expanded proteins in glia elicits reversed ERG polarity. Flies were repeatedly stimulated by one second of light followed by 30 s dark to generate ERG responses. repo-GAL4 is a pan-glial driver; elav-GAL4 is a pan-neuronal driver. (A, B) Regular ERG polarity in flies driven by repo-GAL4 alone or expressing wild-type human Atxn3 (Atxn3-27Q-Myc). (C) Reversed ERG polarity (red arrowheads) under pan-glial expression of mutant human Atxn3 (Atxn3-84Q-Myc). Pan-glial expression of wild-type human Htt (Htt^FL-16Q) elicits regular ERG polarity (D), whereas pan-glial expression of mutant Htt^FL-128Q induces reversed ERG polarity (E, red arrowheads). In contrast, pan-neuronal expression of Atxn3-84Q-Myc (F) or Htt^FL-128Q (G) does not cause reversed ERG polarity. (H) Htt^FL-128Q expression in both neurons and glia induces reversed ERG polarity (red arrowheads), similar to that when expressed in glia alone (compare with C and E). (I) Expression of the hypomorphic allele repo¹ elicits reversed ERG polarity (red arrowheads). Scale bars for amplitude and time are indicated. ERGs were measured on approximately day 3 after fly eclosion. N = 10 for each group, penetrance = 100%.

We next tested whether other polyQ proteins can also induce this phenotype. Glial expression of Htt^EX1-97Q or SCA17-causing hTBP-109Q (55) consistently induced reversed ERG polarity, whereas Htt^EX1-20Q and wild-type hTBP-36Q did not (Supplementary Material, Fig. S1). Glial expression of pure long polyQ polypeptides (20, 41, 63 or 81 glutamines long) showed a polyQ length-dependent effect (Supplementary Material, Fig. S2); 20Q did not affect ERGs, 41Q reversed ERG polarity, and the longer polyQ polypeptides caused prepupal lethality. These results suggest that long polyQ sequences in proteins may cause ERG polarity reversal.

To address whether the reversed ERG polarity phenotype is polyQ-dependent, we first tested ERGs in degenerated retina caused by non-polyQ proteins. The Aβ42^E693G mutation is one of the causes of Alzheimer’s disease (AD) (56). We found that expression of Aβ42^E693G in the Drosophila developing retina (by GMR-GAL4) resulted in a rough eye phenotype with progressive loss of eye pigmentation (Supplementary Material, Fig. S3), indicating that Aβ42^E693G can cause retinal degeneration in the fly. However, glial expression of Aβ42^E693G in Drosophila did not cause reversed ERG polarity. We then used a triple transgenic Drosophila line that carries UAS-MAPT, UAS-APP, UAS-BACE1 (3XTG) to simultaneously express Tau, APP and BACE1 and mimic AD (57). Retinal expression of 3XTG for 10 days was sufficient to diminish the ERG response, demonstrating the neuronal toxicity of 3XTG (Supplementary Material, Fig. S3), but glial expression of 3XTG for 10 days did not cause reversed ERG polarity (Supplementary Material, Fig. S3). We also tested the human TAR DNA-binding protein 43 (hTDP-43), which is linked to amyotrophic lateral sclerosis and frontotemporal dementia (58). Neuronal expression of hTDP-43 in Drosophila causes neurodegeneration (59). Glial expression of hTDP-43 in Drosophila reduced ERG responses, but did not reverse ERG polarity (Supplementary Material, Fig. S3). Together, these results suggest that the reversed ERG polarity phenotype is specific to polyQ-expanded proteins.

Reversed ERG polarity is linked to BBB/BRB leakage

We wondered if specific types of glia contributed to the reversed ERG polarity. We tested a number of previously reported glial GAL4 drivers (60,61), among which only moody-GAL4- and 17A-GAL4-driven Atxn3-84Q-Myc induced complete penetrance of reversed ERG polarity (Supplementary Material, Fig. S4). We examined the expression patterns of all of these GAL4 drivers in adult Drosophila heads (Supplementary Material, Fig. S5) and in Drosophila at different developmental stages (Supplementary Material, Figs S6 and S7). The typical expression sites for moody-GAL4 and 17A-GAL4 in Drosophila are the SPG surrounding the brain (i.e. at the BBB) and the pseudocartridge glia located at the base and side of the retina (i.e. at the BRB). We found that expression of moody-GAL4 was restricted to the glia of the BBB and BRB, consistent with previous reports (60). These results suggest that the reversed ERG polarity phenotype might result from a defect in the BBB and/or BRB.

To examine BRB integrity, we injected high-molecular-weight fluorescence dextran into Drosophila abdomens and assessed dye penetration into the eye (62). Our positive control, the canonical mutant with reversed ERG polarity (repo¹), exhibited severe dye leakage into the eye (Fig. 2). All of the pan-glial expression genotypes that had exhibited reversed ERG polarity also showed complete penetrance of dye leakage into the eye (Fig. 2), confirming that reversed ERG polarity is tightly associated with BRB leakage.

Figure 2.

Flies exhibiting reversed ERG polarity also present dye leakage into the eye. High molecular weight fluorescence dye was injected into fly abdomens and fly eyes were photographed one hour later. (A) Fluorescent dextran dye signal is recorded in the body, antenna and proboscis, but not in the compound eye of control repo-GAL4 flies nor in flies with pan-glial expression of Atxn3-27Q-Myc, Htt^FL-16Q or hTBP-36Q. Fluorescent dextran signal can be detected in the eyes of flies with pan-glial expression of Atxn3-84Q-Myc, Htt^FL-128Q or hTBP-109Q, indicating dye leakage. Mutant repo¹ flies also exhibit severe dye leakage. The ERGs for each genotype are shown at right, with red arrowheads indicating instances of reversed ERG polarity. A color code for fluorescence intensity and scale bar are indicated. Multiple flies of each group are shown to represent the variability in dye leakage. (B) The relative fluorescent intensity of dextran in the compound eye (reflecting dye penetration) was normalized to the fluorescence intensity of antenna or notum of the same fly. The plot represents a quantification of dye leakage. All values represent mean ± SEM. Statistical analysis was performed by one-way ANOVA. ***P < 0.001. Each genotype is indicated. N = 12 for each group.

We then used moody-GAL4 to drive expression of polyQ-expanded proteins solely in the BBB/BRB glia. BBB/BRB-specific expression of Atxn3-84Q-Myc, Htt^FL-128Q or Htt^EX1-97Q caused reversed ERG polarity and dye leakage into the eye (Fig. 3). Therefore, BBB/BRB-specific expression of polyQ-expanded proteins is sufficient to generate the reversed ERG polarity and dye leakage phenotypes. We next investigated whether BBB/BRB-specific expression is necessary for the polyQ-expanded protein-induced phenotypes. To do this, we generated a moody-GAL80 driver that can specifically and efficiently block GAL4 activity in the moody-expressing carpet glia of the Drosophila eye disc (Supplementary Material, Fig. S8). Accordingly, Atxn3-84Q-Myc, Htt^FL-128Q or hTBP-109Q could be expressed in all glia by repo-GAL4, but were specifically excluded from the BBB/BRB glia by moody-GAL80 and, in all cases, the reversed ERG polarity phenotype was not observed (Fig. 4). These results demonstrate that expression of polyQ-expanded proteins specifically in BBB/BRB glia is both sufficient and necessary to cause dye leakage and reversed ERG polarity phenotypes, suggesting a BBB/BRB glial cell-autonomous effect.

Figure 3.

PolyQ-expanded protein expression in BBB/BRB glia is sufficient to elicit reversed ERG polarity. Expression of Atxn3-84Q-Myc, Htt^FL-128Q, or Htt^Ex1-97Q exclusively in BBB/BRB glia, driven by moody-GAL4, is sufficient to elicit the reversed ERG polarity and dye leakage phenotypes. Scale bar, genotypes, frequency of reversed ERG polarity and sample sizes are indicated. The color spectrum in photos at right indicates the fluorescence intensity of dextran dye leaked into compound eyes during dye injection experiments. Only 40% of moody>Htt^FL-128Q flies exhibited dye penetration into the eye, whereas 95% of moody>Htt^EX1-97Q flies presented dye penetration into the eye. Genotypes, scale bar, penetrance of reversed ERG polarity, and the color-coded spectrum of fluorescence intensity are indicated.

Figure 4.

PolyQ-expanded protein expression in BBB/BRB glia is necessary to elicit reversed ERG polarity. Fly lines expressing mutant polyQ proteins in all glia apart from BBB/BRB glia (by using moody-GAL80) do not exhibit the reversed ERG polarity and dye leakage phenotypes. Scale bar, genotypes, frequency of reversed ERG polarity, and sample sizes are indicated. The color spectrum in photos at right indicates the fluorescence intensity of dextran dye leaked into compound eyes during dye injection experiments.

To assess BBB integrity, we further examined fixable dextran-biotin dye leakage in adult Drosophila head sections (see Materials and Methods). In moody > Atxn3-84Q-Myc adults, the dye penetrated into the retina 2 h after injection, but only penetrated into the brain much later (Fig. 5). These findings show that both BRB and BBB integrity are impaired by expression of polyQ-expanded proteins and that they have a more pronounced effect on the BRB than the BBB.

Figure 5.

moody-GAL4 flies expressing mutant Atxn3 show dye leakage into the CNS. (A) Heat maps showing the fluorescence intensity of biotin-labeled 10 kDa dextran in adult head sections. There is very little dye leakage into the retina (R) or brain (B) of control moody-GAL4 adult flies. In contrast, there is significant dye leakage in retinas two hours after dye injection in moody>Atxn3-84Q-Myc adult flies, but little dye leakage in the brain area. The fat body surrounding the brain area, which has a high level of dye penetration, serves as a control to normalize levels of dextran inside the CNS. Quantification and statistical analysis is shown in (A′). (B) Dye leakage was detected in both the retina and brain of adult moody>Atxn3-84Q-Myc flies one day after dye injection. Quantification and statistical analysis is shown in (B′). Color-coded spectrums of fluorescence intensity are indicated. All scale bars are 50 μm. Sample sizes for each group are indicated in the plots. All values in the plots are attributed units (AU) and represent mean ± SEM. Statistical analysis was performed by unpaired Student’s t-test. ***P < 0.001. n.s. represents not significant, P > 0.05. Scale bar, 50 μm. Genotypes, color-coded spectrums of fluorescence intensity and penetrance are indicated.

To directly examine the structural integrity of the BRB, we expressed mCD8-GFP to outline the shape of the BRB. Expression of Atxn3-84Q-Myc, but not Atxn3-27Q-Myc, in BRB glia impaired the integrity of the BRB, as visualized by the discontinuity in mCD8-GFP signal (Fig. 6). Neuroglian (Nrg) and Gliotactin (Gli) are markers for SJs (63,64). Expression of Atxn3-84Q-Myc caused a dramatic reduction in both Nrg and Gli signal in BRB (Fig. 6). Our results demonstrate that BRB glia-specific expression of polyQ-expanded proteins disrupts BRB structural integrity.

Figure 6.

PolyQ-expanded protein expression in BRB glia impairs BRB integrity. (A, B) moody-GAL4-driven Atxn3 and mCD8-GFP expression. The fly head is sectioned to show the retina and optic lobe. The GFP signal (green) indicates the BRB glial membrane. Nrg: orange (A); Gli: orange (B); DAPI, blue; phalloidin, red. Expression of Atxn3-27Q-Myc does not affect BRB integrity. Expression of Atxn3-84Q-Myc caused gaps (white arrows) in the BRB glial layer (white arrowheads) and decreased Nrg expression (white arrows). (B) The SJ marker, Gli, is specifically abolished in the BRB layer (white arrow), but not in the non-BRB layer (white arrowhead). These data indicate that expression of a polyQ-expanded protein in BRB glia impairs BRB integrity. Scale bars are 50 μm. Genotypes and the staining reagent are indicated.

Reversed ERG polarity is caused by BBB/BRB impairment

The earlier results reveal a strong correlation between reversed ERG polarity and BBB/BRB impairment, but their causal relationship is not clear. The reversed ERG polarity phenotype is thought to be due to disruption of the electrical insulation between the retina and brain (53). We tested this hypothesis by directly impairing the BBB/BRB. Specific knockdown in BBB/BRB glia of Moody, a G-protein-coupled receptor critical for BBB formation (62,65,66), led to reversed ERG polarity and dye leakage (Fig. 7B). Knockdown of the SJ component Neurexin IV (NrxIV) (67) in all glia, rather than in BBB/BRB glia alone, elicited reversed ERG polarity and dye leakage (Fig. 7C and D), suggesting that SJs might need to form with other non-BBB/BRB glial cells to establish the barrier. Similarly, BBB/BRB-specific knockdown of Megatrachea, an essential protein for SJ formation (68), also led to reversed ERG polarity and dye leakage (Fig. 7E). The gap junction component Innexin 1, encoded by inx1 (also known as optic ganglion reduced, ogre), is known to play a role in nutrient responses in BBB glia (69). The inx1 mutant (ogre^MI14846) did not exhibit reversed ERG polarity (Fig. 7F), suggesting that gap junctions, in contrast to SJs, are not involved in the barrier function of the BBB. Transcriptomic analyses show that several genes involved in formation of SJs as well as other junctional components are expressed in adult BBB glia (39) and in larval carpet glia (Supplementary Material, Table S1). These data indicate that polyQ-expanded protein expression in BBB/BRB glia impairs BBB/BRB structural integrity, which reverses ERG polarity, thereby establishing the causal relationship between BBB/BRB leakage and reversed ERG polarity.

Figure 7.

Impairment of BBB/BRB and SJ components can reverse ERG polarity. (A) moody-GAL4 flies exhibit normal ERG polarity and no dye penetration into the eye in dye injection experiments. (B) Moody knockdown by RNAi in BBB/BRB glia causes reversed ERG polarity and dye leakage, but with incomplete penetrance. (C) Knockdown of the SJ component NrxIV in BBB/BRB glia does not induce reversed ERG polarity or dye leakage phenotypes. (D) Pan-glial knockdown of NrxIV induces the reversed ERG polarity and dye leakage phenotypes, but with incomplete penetrance. (E) BBB/BRB-specific knockdown of Mega results in reversed ERG polarity, but with incomplete penetrance. (F) The inx1 mutant exhibits normal ERG polarity and no dye leakage into the eye. (G) Specific knockdown of Repo in BBB/BRB glia is sufficient to induce the reversed ERG polarity and dye leakage phenotypes. (H) The repo hypomorph mutant, repo¹, exhibits reversed ERG polarity and dye leakage phenotypes. The red arrowheads indicate reversed ERG polarity. The color spectrum in photos at right indicates the fluorescence intensity of dextran dye leaked into compound eyes during dye injection experiments. Genotypes, scale bars, penetrance and color-coded spectrums of fluorescence intensity are indicated.

Reduced Repo expression is important in causing BRB/BBB impairment

Interestingly, specific knockdown of repo in BBB/BRB glia caused reversed ERG polarity and dye leakage (Fig. 7G), as found for the repo¹ mutant (Fig. 7H), revealing that the canonical reversed ERG phenotype found in the repo¹ mutant is due to a BBB/BRB deficit. These results suggest that BBB/BRB impairment can lead to reversed ERG polarity and that reduced Repo expression plays an important role in this phenotype. Moreover, we found three Repo-binding motifs in the moody enhancer (70) (Supplementary Material, Fig. S9), indicating that Repo may regulate transcription of moody, which is required for BBB formation (62,65,66).

PolyQ-expanded proteins affect BBB/BRB formation at the early pupal stage

To determine the critical time-window during which polyQ-expanded proteins cause reversed ERG polarity, we combined tub-GAL80^ts (71) with repo-GAL4 to temporally control Atxn3-84Q-Myc expression (abbreviated as repo^ts>Atxn3-84Q-Myc). Atxn3-84Q-Myc expression can be turned on at 29°C and turned off at 18°C. Expression of Atxn3-84Q-Myc from the embryonic through to larval stages did not cause reversed ERG polarity (Fig. 8A). Adult-specific expression of Atxn3-84Q-Myc also did not cause reversed ERG polarity or dye leakage, but the flies exhibited motor defects (data not shown), suggesting that glial expression of polyQ-expanded proteins in adults has some deleterious effects unrelated to the BBB/BRB defect. When we expressed Atxn3-84Q-Myc exclusively during the pupal stage, we observed high penetrance of reversed ERG polarity and dye leakage (Fig. 8A). We then combined BBB/BRB-specific moody-GAL4 and the tub-GAL80^ts drivers to temporally control expression of mutant Atxn3 (abbreviated as moody^ts>Atxn3-84Q-Myc) in distinct pupal stages. Expression of Atxn3-84Q-Myc specifically in BBB/BRB glia at the early, but not mid or late, pupal stage was sufficient to cause reversed ERG polarity with high penetrance (Fig. 8B). Together, these findings suggest that polyQ-expanded proteins impede BBB/BRB development during the early pupal stage, with only a marginal effect when the BBB/BRB has already been established.

Figure 8.

The early pupal stage is the critical period for Atxn3–84Q-induced reversal of ERG polarity. (A) repo-GAL4-driven Atxn3-84Q-Myc was combined with tub-GAL80^ts (abbreviated as repo^ts>Atxn3-84Q-Myc), and ERG polarity was normal when the respective flies were raised at 18°C but reversed when they were raised at 29°C. Temperature shift between 18°C (black line) and 29°C (red line) at different developmental stages (indicated at left) allowed temporal control of GAL4 activity. Glial expression of Atxn3-84Q-Myc resulted in reversed ERG polarity and dye leakage, but only when expressed at the pupal stage and not at embryonic, larval or adult stages. (B) moody-GAL4-driven Atxn3-84Q-Myc was combined with tub-GAL80^ts (abbreviated as moody^ts>Atxn3-84Q-Myc) and temperature shift between 18°C and 29°C allowed temporal control of GAL4 activity. Expression of moody^ts>Atxn3-84Q-Myc in early pupae induced ERG polarity reversal with almost complete penetrance. Expression of moody^ts>Atxn3-84Q-Myc in the mid-pupal stage induced ERG polarity reversal with 33% penetrance, whereas expression in the late pupal stage did not affect ERG polarity. The color spectrum in photos at right indicates the fluorescence intensity of dextran dye leaked into compound eyes during dye injection experiments. Genotypes, time windows for temperature manipulation, scale bars, color-code spectrums of fluorescence intensity and penetrance are indicated.

Atxn3-84Q-Myc and Htt^EX1-97Q forms aggregates in glia

Protein aggregation has long been thought a key hallmark for polyQ-related diseases (72). Under pan-glial expression, Atxn3-84Q-Myc formed large aggregates that colocalized with Repo⁺ nuclei, whereas Atxn3-27Q-Myc could be detected in both Repo⁺ nuclei and cellular processes, suggesting that it is distributed in both nuclei and the cytosol (Fig. 9). Western blotting analysis of fly head extracts further showed that Atxn3-84Q-Myc, but not Atxn3-27Q-Myc, formed high-molecular-weight species that were stuck in the stacking gel (Fig. 10A). Semi-denaturing detergent agarose gel electrophoresis (SDD-AGE) (73) and filter trap analysis (74) also demonstrated significant aggregation of Atxn3-84Q-Myc, but not Atxn3-27Q-Myc (Fig. 10B and C, respectively). Similarly, polyQ-expanded forms of Htt^EX1-97Q and hTBP-109Q also formed protein aggregates in glia (Supplementary Material, Figs S10 and S11). Taken together, these results indicate that proteins containing longer stretches of polyQ can indeed form aggregates in Drosophila glia.

Figure 9.

Atxn3–84Q-Myc forms aggregates in glia. Immunohistological staining of paraffin sections of fly heads with glial expression of Atxn3. Atxn3-27Q-Myc (anti-Myc, green) is evenly distributed in glia, whereas Atxn3-84Q-Myc forms aggregates in glial nuclei. Repo staining (orange) represents glial nuclei. Phalloidin, red; DAPI, blue. Inserts are enhanced views of the respective panel. Scale bar, 50 μm. Each genotype and staining reagent is indicated.

Figure 10.

Atxn3–84Q-Myc expressed in glia forms protein aggregates. (A) Western blotting analysis of Drosophila adult head extracts stained with anti-Myc showing that Atxn3-84Q-Myc, but not Atxn3-27Q-Myc, forms high-molecular-weight species (arrowhead) in the stacking gel. (B) SDD-AGE stained with anti-Myc shows that Atxn3-84Q-Myc, but not Atxn3-27Q-Myc, forms high-molecular-weight species (arrowhead). (C) Filter trap analysis demonstrating that Atxn3-84Q-Myc, but not Atxn3-27Q-Myc, formed high-molecular-weight (HMW) species that were trapped on the filter membrane. Quantitations by densitometry of the HMW protein species are shown below each set of results (normalized to α-tubulin). Statistical analysis was conducted by one-way ANOVA. N = 3; ***P < 0.0001. Atxn3-84Q-Myc or Atxn3-27Q-Myc expression in each sample was driven by repo-GAL4. Applied antibodies are indicated.

Rescue of polyQ-expanded protein-induced phenotypes

We found that coexpression of the heat-shock proteins human HSP40 (hHSP40 or hDNAJ), fly HSP40 (dHSP40 or dDNAJ), or human HSP70 (hHSP70 or HSPA1L) could completely rescue the reversed ERG polarity and BBB/BRB leakage phenotypes induced by Atxn3-84Q-Myc, whereas fly HSP23, HSP27, HSC70cb, HSC70-4, HSP83 and yeast HSP104 (75) all failed to rescue these phenotypes (Fig. 11 and Supplementary Material, Fig. S12). These results demonstrate specificity of effect, even within the HSP70 family. Coexpression of GFP did not rescue the phenotypes, indicating that the rescue by hHSP40, dHSP40 and hHSP70 is not due to a titration effect by the extra UAS transgene. We further examined the rescue effect in the nervous system. Expression of Atxn3-84Q-Myc in retina, driven by GMR-GAL4, did not affect ERG polarity in young flies (1 day after eclosion), but did cause retinal degeneration that resulted in no ERG response in old flies (10 days after eclosion) (Supplementary Material, Fig. S13). This age-dependent elimination of the ERG response could be rescued by coexpression of hHSP40 (Supplementary Material, Fig. S13), suggesting that the rescue mechanism is not glia-specific, consistent with previous findings in mammalian and fly neurons (74,76).

Figure 11.

Rescue of Atxn3–84Q-induced reversed ERG polarity. The reversed ERG polarity and dye leakage induced by moody>Atxn3-84Q-Myc expression can be rescued by coexpression of dHSP40, hHSP40, hHSP70 or Atxn3-27Q-Myc, but not by GFP. Coexpressing Repo also rescued the reversed ERG polarity phenotype (red arrowheads) with complete penetrance, but ERG amplitude is reduced. Sample sizes and penetrance are shown. Images of dye injection to represent BBB dye leakage are shown. Fluorescence intensities in the eye are color-coded. Genotype, scale bars, and penetrance are indicated.

Because hHSP40 can function as a protein chaperon, we tested whether it rescues the ERG and BBB/BRB phenotypes by inhibiting the nuclear protein aggregates formed by Atxn3-84Q-Myc in glia, which are associated with a pronounced reduction of Repo levels in the same glial cells (Fig. 12A, upper panels). When hHSP40 was coexpressed with Atxn3-84Q-Myc, Repo levels were restored (Fig. 12A, lower panels), even when Atxn3-84Q-Myc aggregates were still detectable in BRB cell nuclei. Western blot analysis revealed that repo-GAL4-driven coexpression of hHSP40 markedly increased the soluble form of Atxn3-84Q-Myc (Fig. 12B and C). Filter trap assay also demonstrated that hHSP40 dramatically reduced Atxn3-84Q-Myc protein aggregation (Fig. 12B and C). Similarly, aggregates of Htt^Ex1-97Q or hTBP-109Q were also reduced by coexpression of hHSP40 (Supplementary Material, Figs S10 and S11). Moreover, hHSP40 partially rescued the shortened lifespan caused by pan-glial expression of Atxn3-84Q-Myc or Htt^FL-128Q (Fig. 13B and C). These results also imply that aggregates of mutant polyQ-expanded proteins in nuclei may impede the function of transcription factors like Repo. As chaperons such as hHSP40 can inhibit formation of these protein aggregates, Repo levels can be restored, thereby rescuing the reversed ERG polarity and BBB/BRB leakage phenotypes.

Figure 12.

Rescue of Atxn3–84Q-induced effects by hHSP40. (A) In repo>Atxn3-84Q-Myc adult fly head sections, large Myc⁺ (green) aggregates (white arrows) can be detected in the BRB layer. Repo (orange) levels are strongly reduced (white arrows) in cells of the BRB layer. Upon hHSP40 coexpression, the large Myc⁺ aggregates become smaller (white arrowheads) and Repo levels are restored (white arrowheads). Scale bar, 20 μm. (B) In our immunoblotting analysis (upper) and filter trap assay (lower) of fly head extracts, coexpression of hHSP40 reduced the high-molecular-weight species of Atxn3-84Q-Myc (white arrowhead) and increased the level of soluble Atxn3-84Q-Myc (black arrowhead). (C) Quantitation by densitometry analysis of (B), normalized to α-tubulin. The white bars show the amount of soluble Atxn3-84Q-Myc in the Western blot, whereas the black bars show the amount of insoluble Atxn3-84Q-Myc in the filter trap assay. N = 3. Statistical analysis was conducted by one-way ANOVA. ***P < 0.0001. Each genotype is indicated.

Figure 13.

Rescue by hHSP40 of the shortened lifespan induced by Atxn3–84Q expression in glia. (A) Glial expression of Htt^FL-128Q or Atxn3-84Q-Myc results in shortened lifespan. Conversely, glial expression of wild-type Htt^FL-16Q or Atxn3-27Q-Myc extends lifespan. (B, C) Coexpression of hHSP40 partially rescues the shortened lifespan caused by glial expression of Atxn3-84Q-Myc or Htt^FL-128Q. Statistical analysis was conducted by Log-rank test. ***P < 0.001. Each genotype is indicated.

We then tested whether the moody > Atxn3-84Q-Myc phenotypes may be caused by apoptosis in the BBB/BRB glia. Coexpression of anti-apoptotic P35 (77) did not rescue the reversed ERG polarity and dye leakage phenotypes (Supplementary Material, Fig. S14), suggesting that they are not due to apoptosis in BBB/BRB glia. Intriguingly, we found that BBB/BRB-glia-specific expression of Atxn3-84Q induced vacuolization in BRB glia and retinas of aged flies (Supplementary Material, Fig. S14), suggesting a non-autonomous toxicity effect on neurons induced by BRB impairment. Coexpression of P35 in BBB/BRB glia did not rescue BRB degeneration (Supplementary Material, Fig. S14), indicating that BRB degeneration is not due to apoptosis. Intriguingly, P35 expression in BRB rescued retinal degeneration.

Even more interestingly, though pan-glial expression of Htt^FL-128Q or Atxn3-84Q-Myc led to shortened lifespans, expression of wild-type Htt^FL-16Q or Atxn3-27Q-Myc alone extended lifespans (Fig. 13A). Coexpression of Atxn3-27Q-Myc completely rescued the reversed ERG polarity and BBB/BRB leakage phenotypes caused by BBB/BRB glia-specific expression of Atxn3-84Q-Myc. This rescue effect was specific to Atxn3, because neither Htt^FL-16Q nor 20Q could rescue the Atxn3-84Q-Myc phenotype (Supplementary Material, Fig. S12). A previous study also showed that wild-type Atxn3 could protect neurons (78). Our study suggests that wild-type Atxn3 also exerts a protective effect on glia.

Discussion

Glia subtype-specific effects of polyQ-expanded proteins

Specific and distinct neuronal subtypes are impacted in different neurodegenerative diseases, rather than all neurons being uniformly degenerated. Analyses of responses in distinct neuronal and glial cell types are important for a full understanding of such diseases. Our results show that the SJs of the BBB/BRB are most sensitive to expression of polyQ-expanded proteins. This is a clear in vivo demonstration of a glial defect that is cell-autonomously induced by mutant polyQ proteins. Pan-glial expression of Atxn3-84Q and Htt^FL-128Q also caused other neurological and behavioral phenotypes that are not induced by BBB/BRB-specific glial expression, suggesting that non-BBB/BRB glial subtypes are involved in other polyQ-induced phenotypes.

Our results show that polyQ-expanded protein expression in the BRB glia caused its degeneration. In addition, it non-autonomously caused retinal degeneration, suggesting that the damaged BRB glia may send a degeneration signal to the retina. BRB degeneration cannot be rescued by P35, so it is not due to P35-dependent apoptosis. However, P35 expression in BRB glia can non-autonomously rescue retinal degeneration, suggesting that the degeneration signal from BRB glia can be cell-autonomously blocked by P35. If P35 acts by its canonical mechanism to block the activity of its effector caspases DrICE and Dcp-1 (79), then generation of the degeneration signal from BRB glia may be dependent on these effector caspases.

Phenotypic rescue by HSP40 and HSP70 chaperons

We found that the chaperons dHSP40, hHSP40 and hHSP70 can ameliorate the effects caused by human Htt and Atxn3, whereas the fly chaperons HSP23, HSP27, HSC70cb, HSC70-4, HSP83 and yeast HSP104 failed to do so. Because either fly dHSP40 or human hHSP40 can rescue the phenotype in Drosophila, there is no apparent human/fly specificity. A loss-of-function mutation of fly HSP70 enhanced the neurodegenerative phenotype caused by human MJDtr-Q61 (80). Genetic and pharmacological activation of fly HSP70 activity can reduce protein aggregation and neurodegeneration caused by polyQ-expanded human androgen receptor (AR) (81). Moreover, retinal expression of fly HSP40 and HSP70 can reduce the phenotypes caused by Atxn3 (82). The rescue effect of wild-type Atxn3 on Atxn3-84Q-induced phenotype is dependent on dHSP40 (83). Retinal expression of fly HSP40 and human HSP40 can rescue the degeneration caused by Htt-120Q and 127Q (84). Together with our results, these findings suggest that HSP40 and HSP70 play a protective role in glia, preventing protein aggregation and the neurodegeneration caused by polyQ-expanded proteins in a mechanism(s) conserved from fly to human.

Reduced repo expression causes BRB impairment and leads to reversed ERG polarity

The reversed ERG polarity phenotype was first identified in the repo¹ mutant. The repo¹ hypomorphic mutation is a P element insertion in the intron of the repo gene and reduces repo RNA levels (53). Intriguingly, our results show that BBB/BRB-specific knockdown of Repo (a glia-specific homeodomain protein) (53), is sufficient to robustly elicit both the reversed ERG polarity and dye leakage phenotypes. Our results confirm that the original repo¹ reversed ERG polarity phenotype is specifically due to a BBB/BRB defect and proves the hypothesis that the phenotype is due to disruption of the electrical insulation between the retina and brain (53). Because Repo is a glia-specific transcription factor, it may transcriptionally regulate target genes required for BBB/BRB integrity. We found that the moody enhancer has three Repo-binding sites, suggesting that moody may be transcriptionally regulated by Repo. Thus, the reduced Repo expression in the repo¹ mutant may inhibit Moody expression, thereby causing the BBB/BRB defect.

We also found that Repo levels in BBB/BRB glia are reduced due to mutant polyQ-expanded protein expression and that this can be remedied by coexpression of hHSP40. It is possible that the polyQ-expanded proteins in BBB/BRB glia caused a reduction in Repo levels, which then reduced Moody expression and indirectly affected the expression of SJ components, or that reduced Repo levels may have a general effect on BBB/BRB glia. This latter scenario is supported by our finding that Repo can rescue the reversed ERG polarity phenotype. The rescue effected by hHSP40 and hHSP70 may act by inhibiting formation of polyQ protein aggregates and restoring Repo levels.

The weak repo¹ mutant is viable and shows age-dependent ERG defects and lamina degeneration, with reversed ERG polarity being the earliest phenotypic manifestation of the mutation (53). Therefore, BBB/BRB glia are the most susceptible glial subtypes to reduced Repo levels, so the reversed ERG polarity phenotype is a sensitive assay for BBB/BRB defects.

Potential BBB/BRB impairment in mutant polyQ patients and its implications

Our results show that BBB/BRB glia may be the most sensitive glial subtype to polyQ-expanded proteins. In mammals, BBB impairment is apparent in two mouse HD models and, importantly, in HD patients (11,43). It is difficult to define the origin of pathogenesis in these mouse HD models and human HD patients because the mutant HTT protein is broadly expressed in many cell types. Our results in Drosophila provide clear in vivo evidence that polyQ-expanded proteins can cell-autonomously cause BBB/BRB impairment, thereby defining the cell type potentially responsible for pathogenesis. The recent finding that iPSC-derived brain microvascular endothelial cells from HD patients have defective barrier functions supports this scenario, suggesting a BBB defect in HD patients (85).

The R6/2 HD mouse model was found to exhibit BBB leakage in two studies (46), but not in another study (45). In a YAC28 HD mouse model, BBB leakage was only observed after chronic immune stimulation (46). These studies suggest that BBB integrity in the HD mouse is likely to be vulnerable and sensitive to environmental or physiological perturbations so, upon stress conditions such as immune challenge, BBB integrity might fail. Our results suggest that the BBB defect is not specific to HTT, but also occurs due to other polyQ-expanded proteins. We hypothesize that the integrity of the BBB of patients hosting mutant polyQ proteins may also only be weakly affected, only becoming apparent under certain stress conditions and suggesting that detection of BBB impairment in such patients would require sensitive methods.

Our study also demonstrates that the BBB defect caused by polyQ-expanded protein expression in glia occurred during early development, probably affecting the formation of SJs. If the same scenario also applies in human, then polyQ pathogenesis may begin very early, because BBB formation in mouse is complete by embryonic stage E15.5 (86)—roughly correlated to late second to early third trimester in the human fetus—although the actual time of BBB formation in human has not been determined. HD patients are suggested to exhibit early developmental defects (87). More recently, it has been revealed that BBB breakdown occurred early in a HD mouse model expressing several tight junction components (88). Our results suggest that BBB/BRB impairment starts even earlier and is a general feature of polyQ-expanded protein-related diseases. Although diagnoses of these diseases can be on the basis of genomic analysis of CAG repeat length in patients, the degree of BBB/BRB impairment may be used for early diagnosis of disease progression, perhaps even before clinical symptoms appear. Although assessment of BBB impairment may be difficult, establishing BRB impairment by non-invasive methods, such as optical coherence tomography (OCT), is feasible (34) and may offer easy and early detection of polyQ-expanded protein-induced neurodegeneration.

Normal polyQ proteins protect glia as well as neurons

Wild-type Atxn3 has been shown to suppress the neuronal toxicity elicited by neuronal expression of Atxn3-84Q (78). Here, we found that wild-type Atxn3 expression in glia can significantly rescue the reversed ERG polarity and BBB/BRB leakage phenotypes caused by glial expression of Atxn3-84Q-Myc. Interestingly, our results show that Htt^FL-16Q failed to rescue these Atxn3-84Q-induced phenotypes, suggesting specificity in the rescue effect. Our data show that normal polyQ proteins play a protective role in glia, as well as in neurons. Because polyQ-linked diseases are all dominantly inherited, we hypothesize that polyQ-expanded proteins may compete with wild-type polyQ protein for binding to some essential cellular factor, perhaps by sequestering it into non-functional protein aggregates. Increasing the dosage of wild-type protein, as done in our rescue experiments, rescued the reversed ERG polarity and barrier leakage phenotypes. Thus, increasing the expression level of endogenous polyQ proteins or identification of the cellular factor that is competitively inhibited by polyQ-expanded proteins represent potential therapeutic strategies for associated neurodegenerative diseases.

Materials and Methods

Fly stocks

UAS-Atxn3-27Q-Myc (BDSC#33609), UAS-Atxn3-87Q-Myc (BDSC#33610), UAS-Htt^FL-128Q (BDSC#33808), UAS-Htt^FL-16Q (BDSC#33810), Mi[ET1]thus^MB02475 (BDSC#23811), R10D10-GAL4 (BDSC#48261), R29A12-GAL4 (BDSC#49478), 43H01-GAL4 (BDSC#47931), 32H04-GAL4 (BDSC#49734), 17A-GAL4 (BDSC #8474), NP222 (DGRC#112830), dEaat1-GAL4 (BDSC#8849), UAS-3XTG (UAS-MAPT, UAS-APP, UAS-BACE1) (BDSC#33799), UAS-moesin-GFP (BDSC#109119), UAS-Moody-RNAi (NIG#4322R-2), UAS-HDAC6-RNAi (BDSC#31053), UAS-NrxIV-RNAi (BDSC#38192), UAS-Repo-RNAi (BDSC#50735), UAS-Megatrachea-RNAi (VDRC#50306), UAS-P35 (BDSC#5073), UAS-dHSP40 (BDSC#30553), UAS-hHSP40 (BDSC#66267), UAS-HSP23 (BDSC#30541), UAS-HSP27 (F002489), UAS-HSP83 (F003554), UAS-HSC-70cb (BDSC#53728), UAS-hHSP70C-4 (BDSC#5846), UAS-hHSP70 (BDSC#7454), UAS-Aβ2^E693G (BDSC#33773), UAS-20Q-HA (BDSC#30549), UAS-41Q-HA (BDSC#30540), UAS-63Q-HA (BDSC#30544), UAS-CAT (BDSC#24621), UAS-hSOD1 (BDSC#33606), UAS-REPO (F001843), UAS-NLS-Red (BDSC#6282) and ogre^MI14846 (BDSC#60968) were obtained from the Bloomington Drosophila Stock Center (BDSC), the Kyoto Drosophila Stock Center (DGRC), the Vienna Drosophila Resource Center (VDRC), the Zurich ORFeome Project (FlyORF) and the Japan National Institute of Genetics (NIG) Stock Center. The following stocks were kindly provided by the indicated person: moody-GAL4 (62,89), C135-GAL4, Mz97-GAL4, alrm-Gal4 (61), repo-GAL4, UAS-mir-34 (from Dr. Yu-Chen Tsai); UAS-HSP104 (75), UAS-RPI-RNAi, UAS-108Q (from Dr. Horng-Dar Wang); tub-GAL80^ts (from Dr. Cheng-Ting Chien); UAS-Htt^Ex1-20Q (21), UAS-Htt^Ex1-97Q (90), UAS-hTBP-36Q, UAS-hTBP-109Q and UAS-hTDP-43 (91) (from Dr. McCabe, B.D.).

Generation of transgenic flies

The moody-GAL80 construct was generated by polymerase chain reaction (PCR) of the 5 kb enhancer of moody (65) from the w¹¹¹⁸ genome with the primer set: 5′-aagtacccgccagccaagcgtacaaaatat-3′ and 5′-aagcgctttagaataaaaaagaaa-3′. The PCR product was subcloned into the pCaSpeR-GAL80 vector (92). The DNA construct was confirmed by DNA sequencing and used to generate transgenic flies according to standard procedures.

Electroretinography

ERGs were generated according to a modification of a published methodology (93), using glass electrodes filled with buffer (NaCl 130 mM, KCl 2 mM, CaCl₂ 2 mM, MgCl₂ 5 mM, HEPES 10 mM). The recording probe was placed on the surface of the Drosophila compound eye and the reference probe was inserted into the fly head. The light source was a projector (model 765; Newport Electronics, Santa Ana, CA) with a 100 W quartz tungsten-halogen lamp. ERG signals were amplified using a Digidata 1440A Data Acquisition System and Axoclam 900A Microelectrode Amplifier (Molecular Device), and recorded using pCLAMP 10 Electrophysiology Data Acquisition and Analysis Software (Molecular Device). ERGs were recorded in flies about 3 days after eclosion because the compound eyes of newly eclosed flies are too soft. Flies were kept in an incubator with a 12 h dark and 12 h light cycle. Just prior to ERG measurement, flies were exposed to room lighting.

Western blot analysis

Western blot analysis of fly head extracts was conducted according to a modification of a previously described protocol (94). Briefly, 20 fly heads were collected and lysed in 200 μl 8M urea buffer (62.5 mm Tris pH 6.8, 1 mm ethylenediaminetetraacetic acid, 2 mm ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid, 2% sodium dodecyl sulphate (SDS), 50 mm dithiothreitol, 0.5% β-mercaptopropionic acid, 8 M urea, 10% glycerol), with a cocktail of phosphatase and protease inhibitors (Roche). Each lane was loaded with one fly head extract (10 μl). PVDF membranes were then probed with antibodies—rabbit anti-MYC (Santa Cruz, 1:1000), mouse anti-hTBP (Abcam 1TBP18 #ab818, 1:1000), mouse-anti-aggregated Htt (Minipore #MAB5374, 1:1000), anti-tubulin α (Sigma, 1:10 000)—before being visualized using Immobilon Western Luminol (Millipore). Images were captured with a Fujifilm LAS 4000 system (GE). Data were quantified with ImageQuantTL (GE) and statistics were performed with one-way ANOVA or Student’s t-test, and graphically represented using GraphPad Prism (GraphPad Software).

Immunohistological staining

Eye discs were dissected out in phosphate-buffered saline (PBS) and then fixed in 4% paraformaldehyde for 10 min. After permeabilization with PBS/0.1% Triton X-100 for one hour, samples were blocked in buffer (5% bovine serum albumin, 5% goat serum/PBS/0.1% Tween 20) for one hour, and then incubated overnight with primary antibody diluted in blocking buffer at 4°C. After washing with PBS three times, the samples were incubated for one hour with the appropriate secondary antibody at room temperature, and counterstained with 4′,6-diamidino-2-phenylindole (DAPI). For paraffin sections, fly heads were fixed overnight in 4% paraformaldehyde containing 0.1% Triton X-100 at 4°C. Tissues were dehydrated with a gradient of ethanol and xylene, and then embedded in paraffin. Samples were cut to a thickness of 5 μm, and sections were then hydrated and stained as described previously. Images were obtained using an LSM 780 confocal microscope (Zeiss). The primary antibodies were mouse anti-Repo (8D12, DHSB), mouse anti-Nrg (DHSB), rabbit anti-GFP (Invitrogen) and rabbit anti-MYC (A14, Santa Cruz). The secondary antibodies were anti-mouse-647, anti-rabbit-647 and anti-rabbit-488. DAPI and phalloidin were employed as counterstains.

Filter trap assay

Our filter trap assay was modified from a previous report (74). We used 2% SDS/TBS buffer to wash and rinse the cellulose acetate membrane (0.2 μm; Advantec). Fly head extracts in 8 M urea were diluted by 2% SDS/TBS buffer to 200 μl and loaded into the well of slot blot apparatus (PR648, Hofer Scientific). Extracts were then vacuum-sucked through the membrane. After drying, the membrane was stained by standard Western blotting procedures.

Semi-denaturing detergent agarose-gel electrophoresis

SDD-AGE was modified from a previous study (73). Fly heads were lysed in 8 M urea buffer. One to three fly head extracts were loaded in a 1.5% agarose gel and electrophoresed in TAE buffer containing 0.1% SDS. The electrophoresed proteins were then transferred from the agarose gel to a nitrocellulose membrane (PerkinElmer, 0.45 μm) by wet transfer in TBS buffer and subjected to 200 mA overnight at 4°C. The membrane was then blocked in 5% milk/TBS and antibody-stained by standard western blotting procedures.

BBB and BRB permeability assay

The BBB permeability assay was modified from a previous study (62). FlyNap (triethylamine)-anesthetized adult flies were injected with thin borosilicate needles containing 50 mg/ml tetramethylrhodamine dextran (MW 10 000, Molecular Probes, #D1816) or biotin-labeled dextran (MW 10 000, Invitrogen, #D1956) under a dissecting microscope. Approximately 20 nl of dye was injected using a micromanipulator into third instar larvae or 40 nl of dye into the soft tissue between the exoskeleton of two abdominal segments of adults. After 1-h recovery, larvae were dissected to collect their brains for fixation, whereas eyes of live adult flies were examined and photographed by fluorescence microscopy or confocal microscopy. For histological analysis, adult heads were collected from flies injected with biotin-labeled dextran and fixed overnight in 4% paraformaldehyde at 4°C, before being embedded in paraffin and stained with rabbit anti-biotin antibody (Abcam, #ab53494). Quantification of dye leakage into the eye was by measured by the average fluorescence intensity over the whole eye and normalized against the fluorescence intensity of the antenna or notum using Image J. Dye leakage in the brain was quantified by measuring the fluorescence intensity in the brain normalized against the fluorescence intensity of nearby fat tissues.

Longevity assay

Around 20 newly eclosed flies were collected and kept in a vial supplied with standard fly food at 25°C. Four repeat vials, comprising 80 flies in total, were established for each genotype. Dead flies were scored and live flies were transferred to a new vial every two to three days. Statistical analysis was performed by Log-rank (Mantel-Cox) test and plotted by GraphPad Prism (GraphPad Software).

Scanning electron microscopy

Adult fly heads were collected in PBS, fixed in Bouin’s solution overnight, and then dehydrated with an ethanol gradient before transferred to 100% acetone at 4°C overnight. Critical point drying with liquid CO₂ was carried out, and followed by sputter-coating with gold. The images were observed and captured with the JEOL JSM-5600 electron microscope at the Electron Microscope Core, Institute of Molecular Biology, Academia Sinica, Taiwan.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We are grateful to Dr Bon-chu Chung (Institute of Molecular Biology, Academia Sinica, Taiwan) for allowing use of a microinjection manipulator and fluorescence microscope, Drs Chun-Fang Wu, Benny Shilo, Cheng-ting Chien, Chi-Kuang Yao, Hui-Yu Ku, Yijuang Chern and Ya-Hui Chou for very helpful suggestions, the Pathology Core of the Institute of Biomedical Science (Academia Sinica, Taiwan) for paraffin embedding, and the Taiwan Fly Core for ordering fly stocks. We also thank those people who shared their fly stocks: Dr Tara Edwards (Department of Biology, Dalhousie University, Canada) for R29A12-GAL4; Dr Talila Volk (Weizmann Institute of Science, Israel) and Dr Ulrike Gaul (Department of Anatomy, University of California) for moody-GAL4; Dr Marc Freeman (Department of Neurobiology, University of Massachusetts Medical School) and Dr Chun-Hong Chen (The Institute of Molecular and Genomic Medicine, NHRI, Taiwan) for alrm-GAL4; Dr Nancy Bonini (Department of Biology, University of Pennsylvania) for UAS-HSP104; Dr Horng-Dar Wang (College of Life Science, National Tsing Hua University, Taiwan) for UAS-RPI-RNAi, UAS-108Q and UAS-hHSP40; Dr. Cheng-Ting Chien (Institute of Molecular Biology, Academia Sinica, Taiwan) for tub-GAL80^ts; Dr Hitoshi Okazawa (Department of Neuropathology, Tokyo Medical and Dental University, Japan) for UAS-Htt^EX1-20Q; Dr Yu-Chen Tsai (Department of Life Science, Tunghai University, Taiwan) for UAS-mir-34; Dr Ming-T Su (Department of Life Science, National Taiwan Normal University) for UAS-hTBP-109Q and hTBP-36Q; Dr McCabe, B.D. (Department of Physiology and Cellular Biophysics, Columbia University) for UAS-hTDP-43; Dr Ya-Hui Chou (Institute of Cellular and Organismic Biology, Academia Sinica) for the Pcasper-GAL80 DNA vector and Dr Yijuang Chern (Institute of Biomedical Science, Academia Sinica, Taiwan) for sharing the filter trap equipment

Conflict of Interest statement. None declared.

Funding

Supported by grants (NSC 99-2321-B-001-016, NSC 100-2321-B-001-012, NSC 101-2321-B-001-004, NSC 100-2321-B-001-012, NSC 102–2321-B-001-002, MOST 103-2311-B-001-035-MY3) to Y.H.S., and by postdoctoral fellowship to P.A.Y. (NSC 101-2811-B-001-012, NSC 101-2811-B-001-099, NSC 101-2811-B-001-097, MOST 103-2811-B-001-059) from the National Science Council and Ministry of Science and Technology, Taiwan, Republic of China.