PICALM rescues glutamatergic neurotransmission, behavioural function and survival in a Drosophila model of Aβ42 toxicity

Abstract Alzheimer’s disease (AD) is the most common form of dementia and the most prevalent neurodegenerative disease. Genome-wide association studies have linked PICALM to AD risk. PICALM has been implicated in Aβ42 production and turnover, but whether it plays a direct role in modulating Aβ42 toxicity remains unclear. We found that increased expression of the Drosophila PICALM orthologue lap could rescue Aβ42 toxicity in an adult-onset model of AD, without affecting Aβ42 level. Imbalances in the glutamatergic system, leading to excessive, toxic stimulation, have been associated with AD. We found that Aβ42 caused the accumulation of presynaptic vesicular glutamate transporter (VGlut) and increased spontaneous glutamate release. Increased lap expression reversed these phenotypes back to control levels, suggesting that lap may modulate glutamatergic transmission. We also found that lap modulated the localization of amphiphysin (Amph), the homologue of another AD risk factor BIN1, and that Amph itself modulated postsynaptic glutamate receptor (GluRII) localization. We propose a model where PICALM modulates glutamatergic transmission, together with BIN1, to ameliorate synaptic dysfunction and disease progression.

Here we demonstrate that over-expression of the Drosophila PICALM orthologue, lap (like-AP180), ameliorates Aβ42-induced shortened lifespan and locomotor defects in a fly AD model, importantly without affecting Aβ42 levels. Because these findings implicated a gene involved in endocytosis/exocytosis in Aβ42 toxicity, we next performed a small scale, targeted, genetic screen of endocytic/exocytic genes (Table 1) and identified Rab5 and EndoA as partial suppressors of Aβ42-induced toxicity. Given that Rab5 and EndoA are involved in cargo translocation from the plasma membrane to the early endosome, our findings suggested that early steps in endocytosis, including those mediated by lap, are crucial to AD progression. Aβ expression also led to the accumulation of vesicular glutamate transporters (VGlut) at the pre-synaptic region, and increased lap expression restored their wild type localisation. Concordantly, lap reduced the increased glutamate release during spontaneous activity associated with Aβ expression. lap also restored the localisation of the BIN1 orthologue, Amph, which was disrupted upon Aβ expression. Moreover, Amph modulated the localisation of glutamate receptors (GluRII), which was also disrupted by Aβ expression. We therefore propose a novel model where PICALM and BIN1 can co-operate to restore the distribution of pre-and post-synaptic proteins involved in glutamatergic neurotransmission and thus ameliorate aberrant glutamatergic transmission and neurotoxicity in the presence of Aβ42.

Results lap Reduces Aβ42 Pathology
To explore the role of PICALM in AD aetiology in vivo, we examined the role of the Drosophila homologue lap (40) in Aβ pathology. We generated an adult-onset model of Aβ toxicity that expressed Elav-GS is induced by the drug RU486 (42), which was added to the fly food only after eclosion, thus restricting expression of Aβ to adult neurons. The Aβ 42 -expressing flies had a shortened lifespan and displayed locomotor deficits (Fig 1A -1B), suggesting that wild type Aβ 42 is toxic to adult neurons, as previously reported (43). In humans, two SNPs near the gene PICALM, rs3851179 and rs541458, which are associated with decreased levels of AD occurrence in a number of patient cohorts (44,45) are associated with higher levels of expression of PICALM (using the LIBD eQTL browser (46)), potentially suggesting that PICALM has a protective role in AD. To test this, we generated flies over-expressing lap under the control of the UAS promoter and confirmed the over-expression under induction of ElavGS by qPCR ( Fig 1D). Co-over-expression of lap in Aβ 42 -expressing flies attenuated Aβ 42 toxicity, ameliorating both the reduction in lifespan and the impaired locomotion as assessed by negative geotaxis (climbing) assays ( Figure 1A-1B). Importantly, increased lap expression did not alter Aβ 42 protein levels ( Figure 1C), suggesting that lap acts downstream of Aβ 42 generation or degradation.
Conversely, inhibition of lap by RNA interference enhanced Aβ 42 toxicity, leading to further shortening of lifespan ( Figure 1F).
The protective allele for SNP rs3851179 is also enriched in Italian centenarians, suggesting a role for PICALM/lap in healthy ageing as well as AD (47). We therefore examined the effect of lap on healthy ageing, by both over-expressing and down-regulating its expression in neurons in the absence of Aβ 42 . In contrast to our findings with Aβ42 expression, lap over-expression on its own shortened lifespan, while RNAi against lap extended lifespan ( Figures 1E and 1G), indicating that the neuroprotective effect of lap over-expression was specific to Aβ toxicity and not due to a broader effect on ageing.
AD is associated with glutamate excitotoxicity (23,48) and Aβ oligomers lead to increased glutamate release (25). To check if this was also the case in our fly model, we used the fluorescent extracellular glutamate reporter iGluSnFR to detect extracellular glutamate levels (49,50). We expressed UAS-iGluSnFR in larval motor neurons, and observed transient, local bursts of fluorescence, presumably associated with spontaneous local release of glutamate (Figure 2A and E). We occasionally also observed waves of fluorescence running along the anterior-posterior axis, similar to those previously reported (50) (Movie 1). Expression of Aβ led to a dramatic increase in the intensity of local bursts, (Movie 2, Figure 2B and F), suggesting that Aβ 42 increased release of glutamate, consistent with a previous study observing increased glutamate neurotransmission in APP/PS1 transgenic mice (51).
However, we did not observe any change in the number of glutamate release events per minute with Aβ expression ( Figure 2J). Strikingly, lap co-expression reduced the intensity of local fluorescence bursts back to control levels ( Figure 2D, H and I). Taken together, these findings suggest that Aβ compromises components of glutamatergic signalling, leading to increased glutamate release, and that lap over-expression acts to re-instate healthy levels of glutamatergic signalling.
Next we investigated the molecular mechanisms by which lap reduced Aβ 42 toxicity. PICALM plays a major role in endocytosis (52,53), which is important for accurate neurotransmitter signalling. To determine whether endocytosis could play a wider role in the rescue of Aβ toxicity, we performed a targeted genetic screen of well characterized components regulating endocytosis (Table 1). Of the genes tested, only over-expression of Rab5 and EndoA ameliorated the shortened lifespan ( Figure 3A, B). However Rab5 did not ameliorate climbing ( Figure 3C) while EndoA worsened the overall climbing ability of Aß expressing flies, but slowed down the rate of decline, suggesting it was slowing the development of Aß toxicity but possibly had some direct detrimental effect on climbing . In contrast, over-expression of Rab4, Rab7, Rab8, Rab10 and Rab11 exacerbated Aβ42 toxicity (Table 1). These findings suggest that up-regulation of early steps of clathrin mediated endocytosis up to the early endosome could play some part in amelioration of Aβ toxicity, consistent with a study in yeast (54).
However, given the contrasting effects we observed for different Rab genes in our small screen, we hypothesised that lap's rescue could be mediated by additional key factors.

lap Mediates VGlut Localisation
Defects in endocytosis caused by Aβ have been reported to disrupt the trafficking of transmembrane proteins to their proper destination (54), and alterations in endocytic trafficking can disrupt the delivery or recycling of synaptic proteins (55). lap collaborates with clathrin to recycle synaptic vesicles, regulating the efficiency of synaptic vesicle endocytosis and vesicle size (56). lap is also required for recruitment of synaptic vesicle protein (40, 57) and is known to bind to vesicular glutamate transporters (VGlut) (57). We confirmed this by co-immunoprecipitating VGlut with lap in heads of wild type adult flies ( Figure 4D). VGlut is involved in loading glutamate into synaptic vesicles and regulates glutamate release events. We therefore determined whether the protective role of lap was mediated by VGlut.
VGlut expression was not altered by Aβ 42 or lap expression ( Figure 4C). Next we assessed VGlut distribution. Adult Drosophila fly brains have a high density of neurons, making it difficult to monitor individual synapses. We therefore turned to the larval neuromuscular junctions (NMJ), which provide an excellent model system for monitoring individual synapses, and are extensively used to analyse cellular and molecular mechanisms of synaptic development and neurotransmission (58). We observed 10 that VGlut was abnormally accumulated at the pre-synaptic terminal of the NMJ upon Aβ 42 expression ( Figure 4A, B), potentially impairing glutamatergic synaptic transmission. This accumulation was reduced by lap co-over-expression ( Figure 4A, B), suggesting that lap over-expression acts to re-instate wild type glutamatergic signalling by directly binding ( Figure 4D) and regulating the localisation of vesicular transporters ( Figure 4) and therefore affecting the release of glutamate as previously observed ( Figure 2).

Post-synaptic Loss of Amph is Rescued by lap
BIN1, another AD modifier identified by GWAS in humans (59), also plays a role in endocytosis (60) and has recently been shown to regulate neurotransmitter release in mouse glutamatergic neurons (22).
Over-expression of Amph, the fly homologue of BIN1, led to a slight rescue of in lifespan ( Figure 5A) and climbing ( Figure 5B), without affecting Aß levels ( Figure 5C).
In flies, lap is expressed pre-synaptically (40, 56), whereas Amph is expressed post-synaptically (61, 62), implying no direct interaction between lap and Amph. To investigate the interplay between them, we assessed Amph localisation at the larval NMJ. We expressed Aβ either pre-synaptically, with D42-Gal4, or post-synaptically, with Mef2-Gal4, and saw that, in both cases, Amph abundance was significantly decreased upon Aβ 42 expression ( Figure 5D-E). Over-expression of Aß at the NMJ does not to affect bouton size or dendritic branching (63). We therefore we hypothesised that the changes in signalling were responsible for changed in Amph localisation via a yet unknow mechanism.
Over-expression of lap pre-synaptically rescued Amph localisation at the NMJ ( Figure 5D), suggesting that pre-synaptic lap can affect the post-synaptic localisation of Amph, potentially through the effects on VGlut and glutamate release observed above. GluRIIC localisation remained unchanged ( Figure 6E). These findings suggest that Aβ expression can alter the composition of GluRII receptors post-synaptically.
We next assessed whether increased Amph expression could rescue these deficits in the composition of GluRII receptors. However, we found that over-expression of Amph led to a dramatic decrease in the localisation of all the GluRII subunits at the NMJ ( Figure 6), indicating that Amph likely modulates glutamate receptor localisation rather than composition.
Lap and Amph therefore both regulate the localisation of key components of glutamatergic signalling, which could contribute to their rescue of Aβ toxicity.
Given Amph expressionion post-synaptically affects the localisation of GluRIIA, while Aβ 42 and lap expression pre-synaptically affects Amph localisation post-synaptically, we checked whether expression of Aß and lap pre-synaptically affected GluRIIA. Indeed expression of Aβ 42 pre-synaptically increased GluRIIA levels post-synaptically, and co-overepxression of lap reduced GluRIIA levels ( Figure 7), suggesting that lap, via Amph, can affect GluRIIA localisation. How this is mediated will 12 require further investigation.

Discussion
Several publications have described a role of PICALM in Aβ 42 production and clearance. However, its role in Aβ 42 toxicity remains less explored. Our findings provide a novel link between the Drosophila homologues of PICALM and BIN1, lap and Amph respectively, and glutamatergic transmission in an Alzheimer's disease model.
We showed that Aβ expression leads to an increase in spontaneous local burst of glutamate, possibly because of an increase in presynaptic VGlut, which could lead to excessive glutamate release (69). Based on the known effects of VGlut over-expression in increasing vesicle size rather than altering glutamate concentration within the vesicle (Daniels et al 2006), and the ability of lap to reduce vesicle size (Zhang et al 1998), it is likely that increased presynaptic vesicle size is one mechanism by which A increases glutamate release. In addition, we find that Aβ expression increases post-synaptic GluRIIA, which can lead to an alteration in post-synaptic sensitivity of the GluRII receptor (67). Both of these changes could lead to aberrant glutamatergic signalling and neurotoxicity. Defects in glutamatergic signalling are a key feature of Alzheimer's disease (23), with expression of glutamate receptors and transporters altered in sporadic AD patients (28) and mouse AD models (70).
Furthermore, treatment of mouse hippocampal slices and cultured neurons with Aß oligomers leads to excess extracellular glutamate (71,72). We have shown that lap and Amph interact from both sides of the synapse to restore wild type levels and localization of effectors of glutamatergic synaptic transmission. lap decreases VGlut levels pre-synaptically, whereas Amph lowers GluRIIA levels post-synaptically. Moreover, lap can restore post-synaptic Amph localization, which is disrupted by Aβ 13 expression, suggesting that lap, via Amph, could possibly modulate GluRIIA localization too.
lap has been shown to bind VGlut and mediate its endocytosis from the plasma membrane (57).
We also showed that Rab5 and EndoA, which play a role in mediating the formation of clathrin coated pits at the plasma membrane (73), as well as in fusion of endocytic vesicles with early endosomes (74,75), could also rescue Aβ 42 shortened lifespan. This finding suggests that lap's endocytic function may contribute to its ability to rescue Aβ 42 toxicity, by removing excess VGlut from the synaptic terminal.
In contrast, PICALM endocytic function has not been directly linked to synaptic vesicle proteins or to glutamatergic signalling, and it will be interesting to determine whether PICALM plays a similar role in mammalian AD models and can directly modulate Aβ 42 toxicity. It is interesting to note the opposing effect of lap in normal ageing as opposed to a disease context. It is possible that, since glutamate receptors have been shown to decrease during ageing (76,77), a decrease in lap and possibly also endocytosis, helps maintain glutamate signalling in the context of ageing, but in a pathological context characterized by excess glutamate signalling, an increase in lap is beneficial. One caveat in our experiments is we over-expressed a single lap isoform, whereas in flies there are 9 isoforms; it will be interesting to determine the effect of over-expressing the other isoforms.
Amph also plays an important role in endocytosis, but its role in Aβ toxicity remains unexplored.
Our study highlights a role for Amph in regulating the localisation of glutamate receptor GluRII at the synapse. GRIK4, the human homologue of GluRIIA, is increased in AD patients (78), and it would therefore be interesting to verify whether BIN1 can modulate its localization.
In summary, we identified a novel role of two prominent AD-associated GWAS hits, PICALM and BIN1, as modulators of glutamatergic signalling, which could contribute to their role in AD aetiology. It would be interesting to investigate whether this role is conserved in mammalian models of AD, thus potentially opening the possibility of targeting PICALM and BIN1 as modulators of Aβ toxicity in sporadic AD.

Use of mifepristone (RU486) to induce transgene expression by elav-GS
For all experiments involving RU486 addition to fly food, the compound was dissolved in a stock solution of ethanol and added to the fly food while it was still liquid but had cooled to 50ºC.

Negative geotaxis assay
The climbing assay was performed as previously described (81) Statistical analysis was performed in R using ordinal logistics package.

RT-qPCR
Total RNA was extracted from 20-25 fly heads per sample using TRIzol® (GIBCO) according to the manufacturer's instructions. The concentration of total RNA purified for each sample was measured using an Eppendorf biophotometer. One microgram of total RNA was then subjected to DNA digestion using DNase I (Ambion), immediately followed by reverse transcription using the SuperScript® II system (Invitrogen) with oligo(dT) primers. Quantitative PCR was performed using the PRISM 7000

Image Acquisition and Analysis
All images were acquired as stacks on a Zeiss LSM700 inverted confocal microscope using a 63x objective and are shown as maximum intensity projections of the complete Z-stack. The 10 µm stack was taken from muscles 7 and 6 of segments A2-A4. All images for one experiment were taken at the same microscope settings to reduce variability, and all larvae dissected were imaged. All images from one experiment were processed in the same way, setting the threshold so that all the back-ground intensities would be the same across all samples. Mean fluorescence intensity of each slice of the NMJ was measured with ImageJ. Values shown are the averages for 5-10 NMJ ± SEM. Samples were compared by one-way ANOVA followed by Tukey's post hoc test.

Live imaging of glutamate release
Heterozygous D42-Gal4>UAS-iGluSnFR L2 larvae were prepared for live imaging as described previously (50) and imaged on a Zeiss LSM880 airyscan confocal microscope. We immobilized larvae by gently squeezing them under a cover glass in halocarbon oil. Images were collected at a rate of 0.92 frame/s. A single plane was taken from the ventral nerve cord (VNC). All images for one experiment were taken at the same microscopy settings and motion-corrected using the Fiji plug-in, MoCo (82). Samples were compared by one-way ANOVA followed by Tukey's post hoc test.

Statistical Analysis
Statistical Analysis is described in each section above. Interaction between genotype and RU were analysed by Cox-Proportional Hazards (for lifespans) and ordinal logistics analysis (for climbings) in R.
One way ANOVAs were carried using GraphPad Prism v8.0 software. Parameters are reported in the figure legends.

Conflict of Interest Statement,
The authors declare no conflict of interest