Summary
Alzheimer’s disease (AD) is the most common form of dementia, impairing cognitive and motor functions. One of the pathological hallmarks of AD is neuronal loss, which is not reflected in mouse models of AD. Therefore, the role of neuronal death is still uncertain. Here, we used a Drosophila AD model expressing a secreted form of human amyloid-β42 peptide and show that it recapitulates key aspects of AD pathology, including neuronal death and impaired long-term memory. We found that neuronal apoptosis is mediated by cell fitness-driven neuronal culling, which selectively eliminates impaired neurons from brain circuits. We show that removal of less fit neurons delays amyloid-β42-induced brain damage and protects against cognitive and motor decline, suggesting that - contrary to common knowledge - neuronal death may have a beneficial effect in AD.
Introduction
Multicellular organisms have evolved mechanisms to maintain tissue homeostasis and integrity throughout development and ageing. Besides cell-intrinsic surveillance mechanisms, relative fitness levels within a cell population are constantly monitored, ensuring the removal of suboptimal but otherwise viable cells (Merino et al., 2016). The elimination of potentially dangerous or abnormal cells based on their fitness status is known as cell competition. Recent findings prove cell competition is a broad biological process proposed to constitute a quality control mechanism against developmental malformations (de la Cova et al., 2004; Gibson and Perrimon, 2005; Moreno et al., 2002), tumorigenesis (Alexander et al., 2004; Hogan et al., 2009; Kajita and Fujita, 2015; Martins et al., 2014; Menéndez et al., 2010) and ageing (Merino et al., 2015). On the other hand, cell competition machinery may be subverted by pre-cancerous cells to acquire a super-fit status, enabling them to expand, kill and invade surrounding wild-type tissue with a lower fitness status (Eichenlaub et al., 2016; Levayer et al., 2015; Moreno and Basler, 2004; Suijkerbuijk et al., 2016). However up to date, cell competition was not yet investigated during the course of ageing-associated disorders, particularly in neurodegenerative diseases.
In Drosophila the fitness status is translated at the cellular level by ‘fitness fingerprints’, which are encoded by distinct isoforms of the Flower protein located at the extracellular membrane (Petrova et al., 2012; Rhiner et al., 2010; Yao et al., 2009). Flower is a conserved protein with three isoforms in Drosophila that differ solely at the extracellular C-terminus: Flowerubi is expressed ubiquitously, while FlowerLoseB and FlowerLoseA are upregulated in suboptimal cells. The display of loser isoforms in a subset of cells is sufficient to target them for elimination by apoptosis, which is dependent on the transcription of the fitness checkpoint gene azot (Merino et al., 2015). Azot is an EF-hand calcium binding protein dedicated exclusively to cell competition-related apoptosis that integrates upstream relative fitness levels and targets suboptimal cells for death and subsequent engulfment by hemocytes (Portela et al., 2010, Casas-Tintó et al., 2015; Lolo et al., 2012). Mounting evidence demonstrates cell competition is a conserved process ranging from Drosophila to mammals that can also occur in post-mitotic cells and differentiated adult tissue such as epithelia or the neural system (Kolahgar et al., 2015; Tamori and Deng, 2013). The cell competition mediators flower and azot, for example, have been found to mediate elimination of injured or misconnected neurons (Merino et al., 2013; Moreno et al., 2015). The ‘flower code’ is cell type specific, since in the nervous system only FlowerLoseB, and not FlowerLoseA, is expressed in suboptimal neurons (Merino et al., 2013).
Neuronal loss is a key symptom of Alzheimer’s disease (AD), the most prevalent neurodegenerative disorder. AD is a slow progressive disease characterized by initial subtle memory problems that deteriorate to severe cognitive impairment, behavioural changes and difficulty to walk. The main pathological hallmarks of AD are brain deposition of extracellular amyloid-plaques and intracellular fibrils of hyperphosphorylated tau, exacerbated inflammation and finally neuronal damage and death (Braak and Braak, 1991). According to the amyloid cascade hypothesis, amyloid-β-related toxicity is considered the primary cause of the disease but the mechanisms mediating amyloid-induced neurodegeneration and cognitive decline are not fully elucidated (Ashe and Zahs, 2010; Huang and Mucke, 2012; Karran and De Strooper, 2016; Soldano and Hassan, 2014).
Post-mortem brain sections and structural MRI in AD patients show cerebral atrophy in regions involved in memory processing, such as the cortex and the hippocampus (Ossenkoppele et al., 2015; Seab et al., 1988). These findings suggest that the subpopulations of neurons primarily affected by AD, including the enthorhinal cortex and the hippocampal CA1 projection neurons, may be more vulnerable to cellular stress responses elicited by misfolded amyloid (Gómez-Isla et al., 1996; Saxena and Caroni, 2011; Wakabayashi et al., 1994). Although central to human pathology, mechanisms of neuronal loss have been understudied in vivo as AD mouse models do not recapitulate this aspect, showing little neuronal death (Ashe and Zahs, 2010; Karran et al., 2011).
Here we sought to analyze a potential role of fitness-based neuronal elimination in the context of AD onset and progression in a Drosophila model where human amyloid-β is induced in the adult fly brain. We found a physiological mechanism that identifies and purges less fit neurons, delaying cognitive decline and motor disability.
Results
Expression of amyloid-β42 in the Drosophila nervous system affects neuronal fitness
First, we tested whether neurons transit through a stage of reduced fitness when overexpressing Aβ42 (Fig. 1A). We expressed a cassette containing two copies of the human amyloid-β42 (Aβ42) peptide fused to a signal peptide for secretion, under the control the GMR-Gal4 driver, known to produce a strong degenerative phenotype in the Drosophila eye (Fig. 1D) (Casas-Tinto et al., 2011), onwards abbreviated as GMR>Aβ42. To monitor cell fitness markers in the optic lobe, where GMR-Gal4 is expressed, we devised a sensitive reporter to detect FlowerLoseB, by knocking-in a flowerLoseB::mCherry tagged construct in the endogenous flower locus (Fig. 1B). FlowerLoseB::mCherry (indicator of low fitness) was strongly upregulated in the adult optic lobe of GMR>Aβ42 flies but not in the GMR>lacZ control (Fig. 1D,F).
To control if the secretion of a peptide per se is sufficient to downregulate fitness levels and induce flowerLoseB, we expressed the secreted form of a small peptide, EMAP (17kd), under the control of GMR-Gal4. Secreted EMAP is a chemotactic clue that attracts haemocytes to sites of cell competition (Casas-Tintó et al., 2015). We confirmed that secretion of EMAP alone does not upregulate FlowerLoseB::mCherry in the optic lobe, indicating that secretion of an innocuous peptide is not sufficient to decrease the fitness levels of neurons (Fig. 1D,F).
The FlowerLoseB isoform was particularly upregulated in neurons of the optic lobes as detected with the neuronal marker Elav (Fig. S1A). Accordingly, FlowerLoseB expression did not co-localize with cells expressing the glial marker Repo (Fig. S1B).
We then tested activation of another marker of low fitness azot, which is transcribed in cells destined to die based on previous fitness comparison (Merino et al., 2015). In order to visualize azot expression, we generated 1) azot::mCherry transgenic flies, which carry an extra copy of azot fused to mCherry, inserted in another chromosome (Fig. 1C) and 2) azot{KO;GFP} flies, wherein GFP was placed in the endogenous locus of the previously knocked-out (KO) azot gene (Fig. 1G). With both lines, we found that azot was not only upregulated in the optic lobes of GMR>Aβ42 adult flies but was already activated in neurons at previous developmental stages, including the eye discs of the larva and retinas of mid-pupa (Fig. 1E,H,I).
Expression of misfolding-prone toxic peptides linked to Huntington’s, but not to Parkinson, triggers neuronal competition
To further investigate neuronal fitness comparison in other types of neurodegenerative diseases, we turned to published human trangenes reported to induce degenerative phenotypes in the fly: HttQ128 and α-SynA30P. HttQ128 is a pathogenic form of the human huntingtin gene that encodes an expanded repeat of 128 poly-glutamines, causing reduction of viability, retinal death and abnormal motor behaviour in Drosophila (Lee et al., 2004). α-SynA30P is a mutant allele linked to familial Parkinson disease that originates premature loss of dopaminergic neurons, formation of brain inclusions similar to Lewis bodies and decrease of climbing ability when expressed in flies (Feany and Bender, 2000; Song et al., 2017). HttQ0 and α-SynWT, which carry a non-pathogenic form of huntingtin and the wild-type allele of α-synuclein, respectively, served as controls.
For this experiment we employed a previously published translational reporter, in which FlowerUBI, FlowerLoseA and FlowerLoseB are tagged with a specific fluorescent protein: YFP, GFP and RFP, respectively (Yao et al., 2009). We discovered that expression of HttQ128 from the GMR driver induces augmented levels of FlowerLoseB in the adult brain, contrary to the non-pathogenic form, HttQ0 (Fig. 2A,B). Surprisingly, levels of FlowerLoseB did not change with ectopic expression of the Parkinson-related peptides, α-SynA30P and α-SynWT (Fig. 2E,F). The same results were obtained using the FlowerLoseB::mCherry reporter to detect changes on cell fitness upon expression of these toxic peptides in the eye imaginal disc of the larva (Fig. S2A-C).
Although both HttQ128 and α-SynA30P induce neurodegeneration by accumulation of protein aggregates in Drosophila models, our results indicate that only HttQ128 triggers neuronal competition. This result may be explained by the difference in toxicity levels imposed to the tissue by each of these transgenes. We observed that HttQ128 expression leads to increased cell death in a larval epithelium (eye disc) in opposition to the α-SynA30P transgene, which did not lead to significantly increased apoptosis under the same conditions (Fig. 2C,D,G,H).
Accumulation of amyloid peptides in the brain is thought to be the first step in Alzheimer’s pathogenesis. While Aβ42 is the main component of amyloid plaques found in human patients, Aβ40 is a shorter isoform of the human amyloid peptide, which is less amyloidogenic. Aβ40 does not deposit as soluble oligomeric aggregates in Drosophila and does not cause neurodegeneration in vivo (Iijima et al., 2004; Speretta et al., 2012). When overexpressed in the Drosophila brain using GMR-Gal4, Aβ40 did not alter the levels of the FlowerLoserB reporter (Fig. 2I-L). We decided to focus on Aβ42-associated toxicity from now on because the levels of azot and flower were most affected by this peptide.
Aβ42-producing clones are eliminated over time from a neuronal epithelium
To determine if Aβ42 induces cell elimination when expressed in clones, we induced its expression by heat-shock in clones marked by GFP in the neuro-epithelium of the eye disc. We registered that Aβ42-producing clones are progressively excluded from the tissue and detected a higher proportion of dying cells marked by DCP1 inside these clones (Fig. 3A,D,E,F). FlowerLoseB::mCherry and Azot::mCherry were majorly detected inside Aβ42-producing clones, but some signal was also present outside of clones borders (Fig. 3B,C,E,F). We found that Aβ42 diffused largely out of clone borders and accumulated at the basal side of the eye discs, explaining non-autonomous induction of flower and azot (Fig.S2D). FlowerLoseB::mCherry and Azot::mCherry were not detected in control clones expressing an innocuous transgene (Fig.3C). As expected the cleaved form of Drosophila Caspase Protein 1 (DCP1) co-localized with FlowerLoseB::mCherry, showing that unfit cells affected by Aβ42 were undergoing apoptosis (Fig. 3E, Fig.S2E).
flower and azot are necessary for amyloid-β42 induced neuronal death
Next, we analysed Aβ42-associated toxicity and neuronal loss in the adult brain. GMR-driven Aβ42 induces cell death in the optic lobe over time, eliciting a 2.8-fold increase in the number of positive cells for activated DCP1, which co-localized with the neuronal marker ELAV, compared to control flies at two weeks after eclosion (Fig. 4B,C).
The presence of FlowerLoserB isoforms at the cell membrane of a particular neuron does not imply that the cell will die (Merino et al., 2013; Moreno et al., 2015; Rhiner et al., 2010) (Fig. 4A). Only if relative fitness differences with neighbouring neurons persist, cell death is initiated (Levayer et al., 2015; Merino et al., 2013; Rhiner et al., 2010), which requires downstream transcriptional activation of azot (Merino et al., 2015) (Fig. 4A).
To check if neuronal fitness comparisons mediate Aβ42-induced death, we modulated flower and azot genetic dosages. We found that suppressing relative differences of FlowerLoseB levels among cells by silencing LoseA/B isoforms via a long hairpin (Merino et al., 2013), is sufficient to induce a strong decrease in total apoptosis detected upon Aβ42-expression in the adult brain, bringing it back to almost wild-type levels (Fig. 4B,C). Silencing azot with RNAi also reduced the cell death observed in the presence of Aβ42 alone, bringing it back to almost wild-type levels (Fig. 4B,C).
As a positive control for inhibition of apoptosis we used UAS-dIAP1 (>dIAP1), an antagonist of the apoptotic pathway in Drosophila (Hay et al., 1995). Total cell death in GMR>Aβ42 adult optic lobes was suppressed by over-expression of dIAP1 (Drosophila inhibitor of apoptosis1) (Fig. 4B,C). We confirmed that part of apoptotic cells marked by DCP1 co-localize with Azot::mCherry in a GMR>Aβ42 background (Fig. S1C). These results were further supported by a similar experiment conducted in the eye imaginal disc of the larva (Fig. S3A,B), showing that Aβ42-associated cell death is mediated by flower and azot. Moreover we found that the cell fitness-based neuronal elimination induced by neurodegeneration is not specific to Aβ42, and also occurs in the case of HttQ128-associated degeneration (Fig. S3C,D).
In order to study how neuronal fitness is affected over time we monitored cumulative azot expression during ageing of GMR>Aβ42 brains with the reporter line azot{KO;GFP}. This reporter allows visualization of impaired cells (GFP+) that activate the azot promoter (Merino et al., 2015). Using this tool, we observed that GFP signal is only detected in FlowerLoseB positive cells (Fig. S1D). Optic lobes overexpressing Aβ42 accumulate GFP-positive cells at an increased rate compared to control brains of identical ages lacking Aβ42 (53%, 70% and 96% increase over the wild-type of the same age at 5, 15 and 30 days, respectively) (Fig. 4D,E). Altogether, this shows that Aβ42 expression leads to a progressive generation of neurons which will be targeted to death via azot (Fig. 4A).
Suppression of azot-dependent removal of Aβ42-damaged neurons aggravates accumulation of degenerative vacuoles and decreases lifespan
Next, we asked what are the consequences of blocking fitness-based elimination of Aβ42-damaged cells for ageing, locomotion and cognition.
First we established a model where expression of Aβ42 is restricted to adult neurons: To this end, we generated flies containing the Aβ42-cassette under the control of the inducible promoter elav-GeneSwitch (elavGS) (Fig. 5A) (Poirier et al., 2008; Roman et al., 2001). We detected Aβ42 accumulation in the optic lobes and mushroom body calyx of elavGS>Aβ42 adult flies fed for 5 days on the GeneSwitch-activator RU486 but not in the brain of uninduced flies (Fig. 5B, Fig. S4A). Aβ42 aggregates stained positive for aggresome markers (Fig. S4G), confirming the amyloidogenic nature of human Aβ42 when secreted by Drosophila neurons.
Apoptosis was increased in the optic lobes of 10 days-old adults raised on RU486 (Fig. S4B) compared to uninduced flies. TUNEL-positive cells co-localized with ELAV, indicating that Aβ42 caused neuronal death (Fig. S4B). RU486 did not cause apoptosis on its own (Fig. S4C,D). Increased cell death was not accompanied by elevated proliferation, consistent with the fact that little regeneration occurs in uninjured adult brains (Fernández-Hernández et al., 2013) (Fig. S4E,F).
Brains of induced elavGS>Aβ42 flies showed hallmarks of neurodegeneration, such as increased number of degenerative vacuoles (Fig. 5C,D). In induced elavGS>Aβ42 flies, the total number of vacuoles was the double of elavGS>lacZ control flies of the same age (Fig. 5C,D). Interestingly, azot knock-down in induced elavGS>Aβ42 further aggravated brain degeneration and caused a 57% increased in the total number of neurodegenerative vacuoles (Fig. 5C,D) Conversely, when induced elavGS>Aβ42 flies were provided with a third functional copy of azot, which is known to accelerate the elimination of unfit cells (Merino et al., 2015), brain architecture was restored and the number of vacuoles dropped 30% (Fig. 5C,D). Finally, we suppressed apoptosis by overexpressing dIAP1 together with Aβ42 in adult neurons and observed that brains deteriorated faster than in induced elavGS>Aβ42 flies alone (Fig. 5C,D).
To rescue brain morphology, we made use of the azot{KO;hid} transgenic line, which contains the coding sequence of the pro-apoptotic gene hid inserted in the azot KO locus, leading to hid transcription under the control of azot endogenous enhancer sequences (Merino et al., 2015). The total size of vacuoles in the brain of induced elavGS>Aβ42/ azot{KO;hid} flies, which lack Azot protein but still eliminate unfit cells via transcription of hid, was significantly decreased at two weeks time, proving azot has a role mainly dedicated to apoptosis regulation in the course of neurodegeneration (Fig. S5A,B).
The observation that suppression of apoptosis led to accelerated vacuole formation in the brain, made us speculate that cells may be undergoing alternative forms of cell death, such as necrosis. To test this hypothesis, we followed a protocol using propidium iodide (PI) that can penetrate compromised membranes of necrotic cells (Liu et al., 2014; Yang et al., 2013). We detected an increased number of cells permeable to PI in the brains of elavGS>Aβ42 flies two weeks after induction, comparing to non-induced flies, indicating that Aβ42 can trigger necrosis in the brain (Fig. S5C). However, blocking apoptosis either by overexpression of dIAP1 or knock-out of azot did not lead to a further increase in the levels of necrosis, making necrosis an unlikely cause of accelerated vacuole formation in these genotypes (Fig. S5C).
When analysing life expectancy of elavGS>Aβ42 flies, we found that secreted Aβ42 is detrimental for longevity (Fig. 5E,F): induced flies lived on average 40 days versus 52 days for uninduced flies. Life expectancy of induced elavGS>Aβ42 flies dropped to 20days when azot expression was silenced by RNAi (representing a 51% decrease in mean survival comparing with elavGS>Aβ42 flies carrying a wild-type dose of azot) (Fig. 5E,F). It was not possible to determine a clear effect of diap1-overexpression on longevity per se or in combination with Aβ42 (data not shown). Apoptosis is involved in many biological processes with potentially opposing consequences for lifespan.
An extra copy of azot is sufficient to restore motor coordination and improve long-term memory formation
Furthermore, we studied the consequences of fitness-based neuronal-culling on walking behaviour. Using tracking software (Colomb et al., 2012) we extracted several behavioural parameters from 5 min walking sessions of individual flies (Fig. 6A). elavGS>Aβ42 flies induced for two weeks on RU486 showed decreased activity time, shorter walks and ataxia (Fig 6B), compared to uninduced elavGS>Aβ42 flies (Supplementary Videos 1,2). azot RNAi significantly exacerbated behavioural and locomotor dysfunctions caused by Aβ42 alone (Fig. 6A,B). On the contrary, an extra copy of azot was sufficient to completely restore the behavioural defects observed in elavGS>Aβ42 flies, including lengths of walks, activity time and ataxia (Fig. 6A,B, Supplementary Video 3). Finally, blocking apoptosis with UAS-diap1 in elavGS>Aβ42 individuals, further compromised walking performance (Fig. 6A,B), whereas dIAP1-overexpression alone did not result in impaired locomotion (Fig. S6).
To assess long-term memory (LTM) formation we used courtship suppression assays (Keleman et al., 2007; Siegel and Hall, 1979). Courtship conditioning is a form of associative learning by which male flies have to recall that they were previously rejected by a mated/unreceptive female and reduce courtship activity when re-exposed (Fig. 7A). Because prolonged Aβ42 expression resulted in locomotion defects, we reduced Aβ42 induction by one week to ensure that all naïve control males reached a courtship index of 0.6-0.8 (courting 60%-80% of the observation period) (Fig. 7B), normally seen in wild-type sham controls (Nichols et al., 2012). Presence of RU486 did not affect LTM formation of elavGS flies without the Aβ42 transgene (data not shown). We measured a significant difference in courtship index between sham and trained males for all genotypes except for elavGS>Aβ42, azotKO-/- flies (Fig 7B). One week-induced elavGS>Aβ42 flies showed impaired LTM formation compared to uninduced flies (Supplementary Video 4,5), which was strongly aggravated in the absence of azot (Fig. 7B,C) (Supplementary Video 6). Additional expression of >diap to block cell death had a detrimental effect on LTM (but not statistical significant) (Fig 7C and Supplementary Video 7). Conversely, introduction of an extra copy of azot, which increases the efficiency of cell culling (Merino et al., 2015), was sufficient to restore robust LTM formation in Aβ42-expressing flies (Fig 7C,D and Supplementary Video 8), resulting in a significant improvement of memory compared to elavGS>Aβ42 / >lacZ flies. These results underline that azot-mediated clearance of neurons is beneficial for motor and cognitive functions affected by adult onset Aβ42-expression. Moreover, introduction of a single extra copy of azot was sufficient to completely prevent Aβ42-induced motor and cognitive decline.
In flies, the mushroom body (MB) is important for learning and memory (Aso et al., 2014). To investigate if the above observed memory defects were caused by altered MB architecture, we revealed MB structure by anti-Fasciclin II (FasII) immunohistochemistry, which strongly labels the α and β lobes (Crittenden et al., 1998; Fushima and Tsujimura, 2007). We analyzed all genotypes after one week of Aβ42 induction when memory phenotypes were evident, but did not detect marked differences in MB structure. In particular, elavGS>Aβ42, azotKO-/- flies did not exhibit severe morphological defects despite the strong memory impairment (Fig S6). We observed a modest variability between individuals of the same genotype and depicted mild alterations in lobes of the MB, which were comparable among genotypes (Fig S6). This result suggests that memory differences between genotypes are not a result of MB malformation but rather a consequence of a genetic interaction between Aβ42 and azot.
Discussion
Here we found that expression of misfolding-prone toxic peptides linked to Alzheimer’s and Huntington’s diseases in Drosophila neural tissues affects neuronal fitness and triggers cell competition, leading to increased activation of the FlowerLoseB isoform and Azot. Our results demonstrate that fitness fingerprints are important physiological mediators of cell death occurring during the course of neurodegenerative diseases. However this mechanism is specific to certain neurodegeneration-causing peptides or particular stages of the disease, since it is not elicited by expression of Parkinson-related α-Synuclein, for instance. Interestingly, our results suggest that that the toxic effects of a given peptide correlate directly with the level of neuronal competition and death it induces.
Surprisingly we found that neuronal death had a beneficial effect against β-amyloid-dependent cognitive and motor decline. This finding challenges the commonly accepted idea that neuronal death should be detrimental at all stages of the disease progression. We found that most amyloid-induced neuronal apoptosis is beneficial and likely acts to remove damaged and/or dysfunctional neurons in an attempt to protect neural circuits form aberrant neuronal activation and impaired synaptic transmission.
One curious observation in our study is the fact that Aβ42 induces cell death both autonomously and non-autonomously in clones of the eye disc. Dying cells co-localize with FlowerLoseB reporter both inside and outside of GFP-marked clones. We observed that Aβ42 peptide is secreted to regions outside of clones and accumulates at the basal side of the eye disc. The neurons of the eye disc, which project their axons into the optic stalk through the basal side of the disc, are likely affected by the accumulation of this toxic peptide, explaining the induction of cell death outside of clones.
We detected that blocking apoptosis in amyloid-β42 expressing flies either by azot silencing or overexpression of dIAP1 increases the number of vacuoles in the brains of these flies. This seems to be a counterintuitive observation as one would expect that a reduction in apoptosis would result in less cells being lost and a reduction of neurodegenerative vacuoles. However this observation can be conciliated with our model: we suspect that less fit neurons have impaired dendritic growth and inhibit the expansion of neighbouring neurons. This inhibition would disappear once the unfit neuron is culled, allowing compensatory dendritic growth and neuropil extension.
The fact that introduction of a single extra copy of azot was sufficient to completely prevent Aβ42-induced motor and cognitive decline may suggest new venues for AD treatment that aim to support elimination of dysfuncional neurons at early stages of AD pathology. For example, in patients at early symptomatic stages when cognitive impairment is first detected, enhancing physiological apoptotic pathways using Bcl-2 or Bcl-xL inhibitors, or promoting the cell competition pathway described here, may have strikingly beneficial effects.
Contributions
D.S.C., C.R. and E.M. designed the experiments. D.S.C. performed and analysed the experiments. C.R. obtained and analysed data shown in Fig.6. S.S. obtained and analysed the data shown in Fig.7. M.M.M obtained preliminary data. B.H., B.T. and C.T. generated the molecular biology reagents. D.S.C., C.R. and E.M. wrote the manuscript.
Competing interests
The authors declare no competing financial interests.
STAR methods
Contact for Reagent and Resource Sharing
Further information and requests for resources and reagents should be directed to the Lead Contact, Eduardo Moreno: eduardo.moreno{at}research.fchampalimaud.org
Key Resources Table
Experimental Model and Subject Details
Fly husbandry and stocks
Flies were maintained on standard cornmeal-molasses-agar media and reared at 25°C under 12h alternating light-dark cycles. Stocks used in this study are listed in the Key Resources Table.
Method Details
Generation of flowerLoseB::mCherry and azot::mCherry reporter
The flowerLoseB::mCherry knock-in was made by genomic engineering (Baena-Lopez et al., 2013; Huang et al., 2009). The genomic engineering by Huang is a 2-step process consisting of ends-out gene targeting followed by phage integrase phi31-mediated DNA integration. A founder knock-out line was established with a genomic deletion of the flower locus at position 3L: 1’5816’737-15810028. A knock-in construct containing the deleted flower locus fused with mCherry after exon 6 (specific for flowerLoseB isoform) was integrated in the KO line. The knock- in construct was done by site directed mutagenesis to remove the stop codon of exon 6 and add a restriction site to clone mCherry. The knock-out of flower and the knock-in flowerLoseB::mCherry were proven by PCR. Vectors used for generating the flowerLoseB::mCherry were as following: pGX-attP: (Knock-out vector), pGEM-T (used for the site directed mutagenesis and insertion of mCherry) and pGEattBGMR (Knock- in vector).
Primers used were the following: AAGCGGCCGCAGCAGCAACAACAGCAGCAACG and AAGCGGCCGCACCGTTCAATATGCAGGCGGC (5’ arm flower amplification), GGAGATCTGGATGATTCCTGAGCTGCGGTAT and AACTGCAGATGGGGACACCTAAAGAGGCACC (3’ arm flower amplification),
AACTATATTGGGCCGGCCAAGCTAACCGAATGCAAGAGGAACCGGAACCTA and GCATTCGGTTAGCTTGGCCGGCCCAATATAGTTTCTCACTAAAAATATATGCTTGC (mutagenesis primers), TAGGGCCGGCCATGGTGTCCAAGGGCGAAG and ATGGCCGGCCCTTATTTATACAGCTCGTCCATGC (mCherry amplification), ACATAGATCTATAAAAGCTTTCAATGTACACAAATTTG and AGCTGGCGCGCCAAAAAGCATGCCCCACAATAGTTAC (for flower KI).
The azot::mCherry reporter was generated by fusion PCR to combine mCherry coding sequence to azot genomic region, including 2430bps upstream of the start codon, the full azot exon and 175 bp at the 3’ UTR. azot genomic region was amplified from the Bac clone CH321-21G13 (http://pacmanfly.org/) using the following primers: TTGCTTAGACTGTGGCCAGAG and CTCTTCGCCCTTGGACACCATTCGCATTGTCATCATGTTGACGA for 5’ region and azot exon; GACATCTTCTCGCCCAGGTTG and ATGGACGAGCTGTATAAATAACCTCCATGTGAGTACTCGTA for 3’ UTR; GAGATCTCGACGTTCATACGGACGGACAGGCAGACGGAAGGAC and ACTGCATATAACATGCGCGAGA for the promoter region of azot. mCherry was amplified from c5_stable2_neo vector with primers TCAACATGATGACAATGCGAATGGTGTCCAAGGGCGAAGAG and ACGAGTACTCACATGGAGGTTATTTATACAGCTCGTCCATG. The final construct was obtained by two rounds of fusion PCRs (first with primers TAGGCGCGCCCCGCTCATTGTTTCCAAAGTGATTTTC and GCCGCTAGCGTATGAACGTCGAGATCTCGG; second with primers ACTGCATATAACATGCGCGAGA and TAGGCGCGCCCCGCTCATTGTTTCCAAAGTGATTTTC) and was cloned in pGEattBGMR with the restriction sites NheI and AscI.
Immunohistochemistry and image acquisition
Wandering third instar larvae were collected and eye imaginal discs dissected. For clone induction, larvae were given a heat shock at 37ºC 48h or 72h before dissection. For pupal dissections, white prepupae (0hr) were collected and maintained at 25°C for 40h. Dissections were performed in chilled PBS, samples were fixed for 30min in formaldehyde (4% v/v in PBS) and permeabilized with PBT 0,4%Triton.
The primary antibodies and fluorescent reagents used in this study are listed in the key resource table. TUNEL staining (Roche) was performed according to the supplier’s protocol and modified as previously (Lolo et al., 2012). For detection of protein inclusions, brains were fixed and permeabilized as described above and incubated for 1h30min with the proteostat aggresome dye (Enzo Life Sciences) before mounting. For the necrosis assay, brains were dissected in PBS 1X and incubated for 30min at room temperature with 10 μg/ml propidium iodide (PI) (Sigma-Aldrich) in Schneider medium, following by washing and standard fixation ((Liu et al., 2014; Yang et al.,2013)). Samples were mounted in Vectashield (Vectorlab) and imaged on a Leica confocal SP5 or a Zeiss LSM 880 using a 20X dry objective or a 40X oil objective.
Longevity Assays, brain morphology and neuronal manipulation
To minimize disturbing neural development and reduce differences on the genetic background, the RU486-inducible GeneSwitch system was employed (Osterwalder et al., 2001; Roman et al., 2001). The stock solution of RU486 (Mifepristone, Sigma, prepared in 80% ethanol) was diluted in MiliQ water to a final concentration of 100μM and 300μl of the diluted solution was added to the surface of the fly food and allowed to dry at room temperature for 48-36h (Poirier et al., 2008). For the mock solution, 80% ethanol was diluted 10X in water. For survivorship analysis, newly eclosed flies were transferred to bottles and allowed to mate for 2 days at 25ºC (He and Jasper, 2014; Linford et al., 2013). Females were then sorted into groups of 15-20 (more than 100 flies in total were used per genotype) and placed at 29ºC into vials containing standard food supplemented either with RU468 or mock solution. Flies were transferred to new food every 2-3 days and dead/censored animals were counted. For brain morphology analysis, males were subjected to the same protocol, aged at 29ºC until the required stage and dissected. To quantify number of azot-expressing cells with azot{KO;GFP}, newly eclosed males were collected and kept at 25ºC to age for 5, 15 or 30 days.
Behavioral assays
Detailed protocol and further description of the Buridan’s arena can be found elsewhere (Colomb and Brembs, 2014; Colomb et al., 2012). Shortly, 2 weeks-old females, kept in a 12/12 hours light/dark regime, were raised on standard or RU486-containing food. The day before measurements, flies were CO2 -anaesthetized (max 5 min) and their wings were cut with surgical scissors at two thirds of their length. For recordings, flies were placed in the center of the Buridan’s platform in a dark room. The walking activity of each individual fly was recorded for 5 minutes with the Buritrack software (http://buridan.sourceforge.net). Individual tests were re-initalized when flies jumped from the platform or exhibited grooming behaviour. Walking behaviour was analyzed with the CeTrAn software V4 (https://github.com/jcolomb/CeTrAn/releases/tag/v.4).
Long-term memory (Courtship suppression assay): The repeat training courtship assay was used to assess 24 hour long-term memory formation as published previously (Fitzsimons et al., 2013). Briefly, a training session was conducted by coupling individual males with a freshly mated female for a period of seven hours, while sham males were housed alone and served as controls to verify that courtship activity of a specific genotype was intact. Males were induced on RU486 food for one week, which was previously shown to be sufficient to induce the elavGS driver and elicit Aβ42 expression in fly heads (Rogers et al., 2012). After 24 hours, all males, trained and sham, were coupled with new mated females and courtship activity was measured over a period of ten minutes as the percentage of time spent courting (courtship index, CI)(Reza et al., 2013). A memory index (MI) was then calculated as the ratio between the CI of every trained male and the mean value of the CI of the sham males of the same genotype. A range of scores between zero and one was obtained, with zero indicating good memory and one indicating memory similar to a sham control(Ejima and Griffith, 2007), e.g. no memory. Normal memory is generally characterized by a MI of 0.5-0.7 (Fitzsimons et al., 2013). Collected flies were flipped onto fresh food every two days and kept at 25°C in a 12 hr light/dark cycle. elavGS>Aβ flies on standard food served as uninduced control. In all experiments, the experimenter was blind to the genotype of the flies. Experiments were performed under ambient light at 25°C with 65-70% relative humidity and recorded for 10min using a camcorder (Sony Handycam HDR PJ410).
Image analysis and Statistics
Image quantification was done with Fiji. The number of positive cells in the adult brain for DCP1, TUNEL, FlowerLoseB::mCherry, FlowerLoseB::RFP or Azot::mCherry was counted on 40-µm-wide maximum projections including the anterior part of the optic lobe. Noise signal was removed using a Gaussian blur filter (sigma =1) and/or applying a background subtraction (rolling=20). GFP expressing cells in azot{KO;GFP} flies were assumed to be GFP-positive particles wider than 9pixels on a 25μm-thick projection (showing a 141μm2 field) of the optic lobe. Measure of death induction in eye imaginal discs was done by counting the number of TUNEL positive particles in 10μm-thick maximum projections. Spaces between phalloidin staining with an area >25μm2 were assumed to be neurodegenerative vacuoles. Presence of vacuoles was quantified two weeks after eclosion at a 10μm deep ventral plane located in the central brain (next to the mushroom body).
Survival curves: For statistical analysis, a log-rank test (Mantel Cox) was applied to determine significant differences between survival curves.
Walking behaviour data was analyzed by an ANOVA model, which was validated posthoc with Tukey-Anscombe plot and QQ plot of the residuals. p values were calculated comparing all experimental genotypes with each other and corrected for multiple testing using Holm’s method (Holm, S, 1979). The variables distance, activity time, and turning angle were chosen for analysis based on previous test experiments.
Courtship suppression assay, raw data was subjected to arcsine transformation in order to obtain a normal distribution and the memory indexes of each genotype were subjected to a one-way ANOVA followed by Bonferroni and Holm’s correction by comparing genotypes to elavGS>Aβ induced controls. When comparing only two genotypes, the Student’s T-test (two-tailed, unpaired) was used. Significance was set at P < 0.05.
The distribution of the number of positive cells (for DCP1, FlowerLoseB::mCherry, FlowerLoseB::RFP or Azot::mCherry) in the optic lobes of adult flies was analyzed for statistical significant differences between groups with a Kruskal-Wallis test and a Dunn’s test was applied for multiple comparisons between genotypes. The number of brain vacuoles per hemisphere was analyzed for homogeneity between genotypes with a Levene test and p-values were calculated with one way ANOVA and a Dunnett’s posthoc test. To determine statistical differences between genotypes for the number of TUNEL positive cells in eye discs, a one-way ANOVA test, followed by a Dunnett’s posthoc were applied. When only two groups were compared and data did not follow a normal distribution assessed by d’Agostino-Pearson omnibus test, statistical significance was accessed with a Mann-Whitney U non-parametric test (for example in the quantification of TUNEL positive cells in the adult brain, GFP positive cells in the optic lobe and clone area). The number of PH3 positive cells was analyzed with an unpaired t-test with Welch’s correction. All graphs are displayed as mean ± standard error.
Supplemental Videos legends
Buridan arena
Video 1 Locomotion of an uninduced elavGS>Aβ42 / >lacZ individual.
Walking trajectories of heterozygous elavGS>Aβ42/+ females after two weeks on normal food.
Video 2 Locomotion of an induced elavGS>Aβ42 / >lacZ individual.
Walking behavior of heterozygous elavGS>Aβ42/+ females after two weeks on RU486.
Video 3 Locomotion of elavGS>Aβ42, azot+/+;+ individual.
Walking trajectories of heterozygous elavGS>Aβ42, azot+/+;+ females after two weeks on RU486.
Courtship suppression assay
Representative 3 min extracts of 10 min long videos recorded to evaluate long-term memory 24 hours after exposure to unreceptive female (trained) or no female in the same test chamber (sham controls).
Videos 4 Courtship activity of trained elavGS>Aβ42 / >lacZ flies, induced for one week on RU486.
Videos 5
Courtship activity of trained elavGS>Aβ42 / >lacZ flies raised for 1 week on normal food (no induction).
Videos 6
Courtship activity of trained elavGS>Aβ42, azotKO-/- flies, induced for one week on RU486.
Video 7
Courtship activity of trained elavGS>Aβ42 / >diap fly induced for one week on RU486.
Video 8
Courtship activity of trained elavGS>Aβ42, azot+/+;+ fly induced for one week on RU486.
Acknowledgements
We thank to Troy Littleton, Sergio Casas-Tintó, Richard Baines, the Vienna Drosophila Resource Center, the Bloomington Stock Center and the Developmental Studies Hybridoma Bank for sending stocks and reagents. We also acknowledge Julien Colomb for providing equipment and software (CeTrAn V4) used in the walking behaviour assay. We are grateful to the fly community at Champalimaud Research for criticial feedeback and for sharing antibodies and fly stocks. We thank technicians of the Champalimaud Fly Platform for support with the fly work and stock maintenance and Gil Costa for helping with scientific drawings. D.S.C. was supported by an EMBO long-term fellowship (ALTF 979-2014). Work in our laboratory is funded by the European Research Council, the Swiss National Science Foundation, the Portuguese Science Foundation, the Josef Steiner Cancer Research Foundation and the Swiss Cancer League.