Using Drosophila amyloid toxicity models to study Alzheimer’s disease

Alzheimer’s disease (AD) is the most prevalent form of dementia and is characterised by a progressive loss of neurons, which manifests as gradual memory decline, followed by cognitive loss. Despite the significant progress in identifying novel biomarkers and understanding the prodromal pathology and symptomatology, AD remains a significant unmet clinical need. Lecanemab and aducanumab, the only Food and Drug Administration approved drugs to exhibit some disease-modifying clinical efficacy, target A β amyloid, underscoring the importance of this protein in disease aetiology. Nevertheless, in the absence of a definitive cure, the utilisation of preclinical models remains imperative for the identification of novel therapeutic targets and the evaluation of potential therapeutic agents. Drosophila melanogaster is a model system that can be used as a research tool to investigate neurodegeneration and therapeutic interventions. The short lifespan, low price and ease of husbandry/rearing make Drosophila an advantageous model organism from a practical perspective. However, it is the highly conserved genome and similarity of Drosophila and human neurobiology which make flies a powerful tool to investigate neurodegenerative mechanisms. In addition, the ease of transgenic modifications allows for early proof of principle studies for future therapeutic approaches in neurodegenerative research. This mini review will specifically focus on utilising Drosophila as an in vivo model of amyloid toxicity in AD.

. AD stands as the most prevalent cause of dementia and presents a significant and unmet medical challenge (Padmanabhan & Götz, 2023).Dementia encompasses a range of disorders characterised by cognitive function deficits.AD is primarily characterised by memory loss and progressive cognitive decline, although it can involve impairments in speech and visuospatial processing (Knopman et al., 2021).Disease progression is characterised by a general physical decline in which eventually the abilities to perform basic bodily functions, such as walking, swallowing and speaking, are lost.Patients live an average of 8 years following diagnosis but can live up to 20, requiring around-the-clock care in the end stages of disease (Alzheimer's Association, 2020).Globally, in 2019, the cost of dementia was estimated to be $2.8 trillion (Nandi et al., 2022).As AD is closely linked to the aging process, as global life expectancy continues to rise due to advancements in healthcare, the prevalence of AD is expected to increase accordingly, with cases almost tripling by 2050 (Nichols et al., 2019).In the United Kingdom, dementia and AD were responsible for the majority of deaths in 2022 (Death registration summary statistics, England & Wales, 2022).Specifically, they accounted for 65,967 deaths.Currently, there are no therapeutic strategies available that can halt the course of the disease or reverse its progression, although recent clinical trials have shown some initial promise.

THE GENETICS AND PATHOPHYSIOLOGY OF AD
AD is a proteinopathy disorder characterised by the accumulation of intracellular neurofibrillary tangles (NFTs) composed of hyper-phosphorylated Tau (p-Tau) and extracellular plaques of toxic amyloid-β (Aβ) peptides, the most common being Aβ1-42 (Padmanabhan & Götz, 2023).This leads to progressive neurodegeneration, with seminal studies indicating that neuronal cell death is initiated in the entorhinal cortex (Braak & Braak, 1992;Van Hoesen et al., 1991), propagating into the hippocampus and then spreading to the rest of the brain (Braak & Braak, 1991).However, recent studies have implicated pathological changes in the locus coeruleus as an even earlier occurrence in the disease course (Grudzien et al., 2007;Theofilas et al., 2017).

Aβ toxicity
There are two forms of disease, a common late onset sporadic form (LOAD) and an early onset familial form (EOAD), with a mean age-of-onset below 60 (Kumar-Singh et al., 2006).Three genes carry autosomal dominant variants leading to EOAD: amyloid beta precursor protein (APP), presenilin 1 (PSEN1) and presenilin 2 (PSEN2) (Bekris et al., 2010).All the familial variants cause increased misprocessing of APP.This led to the formulation of the amyloid hypothesis, which states that AD is caused by the misprocessing of APP and the increased pro-duction of neurotoxic Aβ1-42 peptides (Haass & Willem, 2019) (Figure 1), (Rosenberg et al., 2016).APP is mostly cleaved sequentially by α-secretase and then by γ-secretase (a complex consisting of four individual proteins, including PSEN1 and PSEN2) leading to the formation of Aβ1-40 peptides.However, in AD, there is a significant upregulation of β-secretase (also known as β-site APP cleaving enzyme [BACE]) cleavage, leading to the formation of Aβ peptides that are 42 residues long (Aβ1-42).The inclusion of two additional hydrophobic residues in Aβ1-42 leads to the formation of β-sheet-rich structures, which further aggregate as neurotoxic plaques in the brain (Chen et al., 2017;Lemere et al., 1996;Quartey et al., 2021).
In addition to causative variants, over 600 genes have been identified as increasing susceptibility of developing AD (Knopman et al., 2021).A recent 2022 genome-wide association study (GWAS) identified 75 risk loci for AD with 42 being novel at the time of the study (Bellenguez et al., 2022).
Sporadic inheritance of the ε4 isoform of apolipoprotein E (APOE) is considered the most prominent genetic risk factor of LOAD (Bekris et al., 2010) and has been demonstrated to increase amyloid burden in the brains of AD patients (Castellano et al., 2011).Homozygote carriers of the APOE ε4 allele have an increased risk of developing AD 12-15 times more compared to APOE ε3 carriers (Van Der Lee et al., 2018).In addition, APOE ε4 carriers are associated with an earlier age of AD onset, possibly as a result of accelerated Aβ1-42 accumulation in the brain (Bales et al., 1999;Castellano et al., 2011).Although the exact neurobiological mechanism has not been elucidated, in vivo studies point to a potential impairment in Aβ1-42 F I G U R E 1 Amyloid toxicity and neurofibrillary tangle (NFT) pathology.Proteolytic processing of amyloid precursor protein (APP) can be amyloidogenic or non-amyloidogenic.Aβ1-42 peptides are generated from the cleavage of APP by secretases.α-Secretase drives the non-amyloidogenic pathway, whereas β-secretase drives the amyloidogenic pathway.α-Secretase cleaves APP into APPsα and αCTF (C-terminal fragment), and αCTF is further cleaved by γ-secretase to produce ACID and p3.In contrast, β-secretase cleaves APP into APPsβ and βCTF, and γ-secretase cleaves βCTF to generate Aβ1-42 and ACID.However, it is the imbalance of proteolytic processing favouring the β-secretase-mediated amyloidogenic pathway which is the predominant pathogenic factor.Although the cleavage product p3 from the non-amyloidogenic pathway appears to be nontoxic, the monomeric Aβ1-42 peptides generated from the amyloidogenic pathway have the propensity to aggregate and form neurotoxic oligomers.Tau is a microtubule stabilising protein, but certain toxic isoforms can undergo hyperphosphorylation. p-Tau exhibits reduced ability to bind to tubulin, resulting in reduced stability of microtubules.As a consequence, p-Tau monomers accumulate and oligomerise, eventually assembling into fibrous paired helical filaments, which give rise to the formation of neurotoxic neurofibrillary tangles (NFTs).The precise relationship between amyloid toxicity and the pathogenesis of NFTs is not yet fully understood.AICD, amyloid precursor protein intracellular domain; APP intracellular domain; APP, amyloid precursor protein; APPsα, secreted amyloid precursor protein-alpha; APPsβ, secreted amyloid precursor protein-beta; Aβ, amyloid beta; NFT, neurofibrillary tangle; p-Tau, hyperphosphorylated Tau αCTF, alpha C-terminal fragment; βCTF, beta C terminal fragment.Source: Created with BioRender.com.
clearance (Bales et al., 1999).Castellano et al. (2011) highlighted the most common genes, and their fly orthologues, in which variants are known to cause familial EOAD and the most common genes in which variants are risk factors for sporadic LOAD (Table 1).
Unsurprisingly, targeting the amyloidogenic APP processing has been a significant target for therapeutic interventions (Karran & De Strooper, 2022).The developments of monoclonal antibodies (mAb) against various forms Aβ, including monomers, protofibrils and plaques, have exhibited initial promise as the first disease-modifying treatments available for AD (Budd Haeberlein et al., 2022;Swanson et al., 2022;van Dyck et al., 2022).Aducanumab, a mAb targeting the aggregated Aβ in amyloid plaques, and lecanemab, a mAb targeting soluble Aβ protofibrils, have recently been approved by the Food and Drug Administration (FDA) for the treatment of AD (Cavazzoni, 2021)).In particular, lecanemab in a large phase 3 trial demonstrated a slower cognitive decline at 18 months, compared to the placebo group (van Dyck et al., 2022); however, it also led to severe adverse events in some participants (Budd Haeberlein et al., 2022;Swanson et al., 2022;van Dyck et al., 2022).Although both these mAb are not cures, for the first time, they demonstrate that interventions targeting Aβ can modulate disease progression, proving that indeed Aβ is a driver of disease, and validating the use of Aβ toxicity models for AD research.

2.2
The link between Aβ and Tau p-Tau aggregates in the brain and can lead to the formation of neurotoxic NFTs (Kondo et al., 1988) (Figure 1) which promote neurodegeneration through the disruption TA B L E 1 A summary of the genes, and their fly orthologues, in which pathogenic variants are known to cause familial EOAD and the most common genes in which pathogenic variants are risk factors for sporadic LOAD.Neural progenitor cell proliferation regulator (Caillé et al., 2004).Learning and memory (Bourdet et al., 2015) APP-L 24%

PSEN2
Variants cause autosomal dominant forms of EOAD (familial).i.e.N141I Proteolytic processing of APP and Notch signaling pathways components (De Strooper et al., 1998;Steiner et al., 1999). of multiple processes.The association between Aβ1-42 and Tau pathology in relation to AD pathogenesis remains unclear but according to the amyloid hypothesis, p-Tau NFTs are downstream of the amyloid cascade.The first in vivo evidence emerged in 2001 when Science published two sequential articles that demonstrated that the injection of Aβ1-42 (Mikol et al., 2001) and a pathogenic variant of APP (Lewis et al., 2001) induced NFT pathology in mice (Bloom, 2014).Further in vivo studies have suggested that Tau pathology is a requisite of Aβ1-42 toxicity since eliminating Tau in murine models carrying Pathogenic variants of APP protected against amyloid neuronal toxicity (Leroy et al., 2012;Roberson et al., 2007).Additionally, a study with human AD patients reported that the presence of cortical Tau NFTs in positron emission tomography scans was consistently observed only in the presence of Aβ (Pontecorvo et al., 2017).However, a defined interaction between Aβ1-42 and Tau and their relative contribution to AD remains a controversial topic of research.For example, a recent in vivo study proposed that Tau is not required for Aβ1-42-induced memory deficits (Puzzo et al., 2020).

Psn
Untangling the neurobiological and clinical relationship between Aβ1-42 and Tau remains a necessary research objective.
In summary, the aetiology of AD is remarkably complex.The pathology encompasses a broad range of factors, including, but not limited to, genetic risk factors and molecular and cellular biochemical changes (Knopman et al., 2021).

ADVANTAGES OF THE POWERFUL DROSOPHILA MELANOGASTER MODELLING SYSTEM: PRACTICALITY AND GENETICS
Genetic and epidemiological studies have provided a number of candidate pathways involved in disease (Karch & Goate, 2015); however, proving causality and identifying good targets for therapeutic intervention is crucial.This is where model organisms, such as Drosophila, are of paramount importance.Precise genetic engineering in Drosophila allows the targeting and isolation of specific genes.This enables an observation of causality between a genetic variant and the resulting phenotype.
Pathological processes in neurodegenerative disease often initiate decades before symptomatic onset which severely limits the use of human post-mortem tissue in investigating the full disease course (Lu & Vogel, 2009).Brain imaging studies on individuals carrying the autosomal dominant variants that result in early-onset familial AD have reported that Aβ accumulation occurs 15 years before the estimated age of symptomatic onset (Benzinger et al., 2013;Sperling et al., 2014).In contrast, disease progression in Drosophila models can be studied from initiation to the terminal stages because of their relatively short lifespan of 70-100 days (Piper & Partridge, 2018), and this is especially useful in late onset disease, such as AD.A primary advantage of Drosophila research is the ability to generate with relative ease large populations for widescale experiments, enabling statistical analysis.Females lay roughly 50 eggs per day at peak fecundity (Novoseltsev et al., 2003), and the generation time from a fertilised egg to an enclosed adult is approximately 10 days when reared at 25 • C (Fernández-Moreno et al., 2007).This is another advantage over rodent models in which small sample sizes can provide statistical challenges.This provides a model that is highly sensitive in detecting modifiable changes in lifespan.Drosophila can also be fed a well-defined diet of sugar, yeast and agar (Piper & Partridge, 2018) in which pharmacological agents can be mixed into and thus facilitate easy administration of drugs to the flies.
However, arguably, the ease of genetic engineering in Drosophila is its major draw.The entire genome of the Drosophila melanogaster was sequenced in 2000 (Bergkvist et al., 2020) and is a powerful genetic tool kit to study human diseases and pathogenic mechanisms.Drosophila is very genetically tractable, and the evolution of sophisticated genome engineering techniques has facilitated intricate and precise genetic modification to be achievable.Thousands of Drosophila stocks, carrying variants, overexpression or RNAi constructs, can be easily acquired for a low cost from stock centres and as invertebrates, the use of Drosophila in medical research does not require regulatory licences or oversight (Piper & Partridge, 2018).Drosophila exhibits low genetic redundancy, which allows to uncover gene function more easily, and carry orthologues of around 75% of disease-causing genes in humans (Verheyen, 2022).Expression of human disease-associated genes in fly neurons usually results in deleterious phenotypes, analogous to those in humans.This makes flies an ideal model to uncover the function of disease-associated genes.
Fly disease models take advantage of the GAL4-UAS system of expression (Figure 2a), or the modified GS system (Figure 2b) which enables spatial and temporal control of transgene expression in Drosophila (Osterwalder et al., 1993).Recently, the developments of novel expression systems, such as the QF-QUAS and LexA-opA systems (Figure 2c,d), which can be combined with the GAL4-UAS system, allow the manipulation of genes in two tissues at the same time (Del Valle Rodríguez et al., 2012).This, for example, allows us to look at the interaction between Aβ in neurons and other genetic manipulations in glia, which will be invaluable as the field is increasingly appreciating The UAS-GAL4 system.A GAL4 driver is cloned downstream of a tissue-specific promoter, leading to restricted production of the driver to the tissue of interest.The gene of interest is cloned downstream of an upstream activating sequence (UAS) promoter.When a fly carries both the GAL4 and UAS constructs, GAL4 is produced, it binds to its target UAS in the F1 offspring which drives transgene expression.At 18 • C, GAL80 forms a complex with GAL4 (Continues) to repress transgene transcription.GAL80-mediated transcription repression can be removed by exposing flies to 29 • C. (b) The GS UAS-GAL4 system.GS GAL4 is a chimeric protein consisting of a GAL4 fused to a tissue specific promoter and the transcriptional activation domain of a progesterone receptor.Equivalent to the GAL4/UAS system, mating between a fly carrying GS GAL4 and a fly carrying a UAS-linked transgene of interest enables GAL4 to bind to its target UAS in the F1 offspring.Dissimilarly, transgene activation is only induced when the drug RU486 (mifepristone) binds to the progesterone receptor component, thus allowing temporal control via desired timing of RU486 application which can readily be applied to fly food.(c) The LexA-opA system.The LexA-opA system consists of the DNA binding domain of the transcription factor LexA, derived from Escherichia coli, and its target DNA sequences known as LexA operator sites (LexAop).LexA driver lines consist of the LexA transcription factor bound to a tissue-specific promoter to enable spatial control of the transgene.LexA driver line is mated with a line carrying a LexAop sequence fused upstream of the transgene of interest enabling LexA-LexAop binding which drives transgene expression tissue specifically.(d) The Q system.Similarly, mating between a fly carrying the transcriptional activator QF and a fly carrying transgene linked to a QUAS promoter enables QF to bind to its target QUAS in the F1 offspring which drives transgene expression.The repressor QS can bind to QF and inhibit activation of transgene expression.In the presence of the drug QA, QS-mediated repression of QF is inhibited.DNA, deoxyribonucleic acid; Elav, embryonic lethal abnormal vision; F1, filial 1; GS, gene-switch; QA, quinic acid; UAS, upstream activating sequence.
the importance of interactions among different cell types to disease development (Jonson et al., 2018).

DROSOPHILA MODELS AS A TOOL FOR SCREENING
The facile genetics and the ease of carrying out experiments at scale, mean large-scale forward genetic screens can be used to rapidly identify disease modifiers in Drosophila models (Ugur et al., 2016).Typically, screens involve high throughput analysis of a plethora of variant Drosophila stocks, which are investigated for alterations in a pre-defined phenotype (Lenz et al., 2013); in the case of neurodegeneration, there are a number of established methodologies for assessing phenotypes (Figure 3).

Eye morphology
The most commonly screened phenotype is the morphology of the fly eye.The Drosophila compound eye is a remarkable biological structure comprising around 750  ommatidia, meticulously arranged with remarkable precision, requiring 2/3 of essential genes for their correct development (Thaker & Kankel, 1992).They are neuronal in origin and provide a visible and easily scored metric for rapid identification of cell abnormalities or death, making them particularly suitable for neurodegeneration research (Iyer et al., 2016), even though eye morphology is often scored according to a qualitative score, leading to variable interpretations and weak modifiers may induce subtle variations that are undetectable by the naked eye (Iyer et al., 2016).However, recent technological advancements have led to the development of machine learning tools that can not only quantitatively assess the degeneration of the Drosophila eye but also categorise images into morphological classes using a pre-trained algorithm (Diez-Hermano et al., 2020), allowing for an unbiased quantitative assessment of eye morphology.Eye screens are a powerful screening tool; however, the morphology is mostly defined during development.Most neurodegenerative diseases are late onset, and therefore, phenotypes associated with ageing are more pertinent.

Lifespan assays
Lifespan assays, where the survival of 100-150 flies is scored, allow to monitor genetic modifiers of longevity.
Lifespan assays are frequently employed to demonstrate genetic/pharmacological modifier mediated rescues of the shortened lifespans exhibited in Drosophila amyloid toxicity models (Catterson et al., 2021;Finelli et al., 2004;Niccoli et al., 2016).As previously mentioned, Drosophila generally exhibits a lifespan of 70-100 days, which is much shorter than mammalian model organisms that typically live up to 3-5 years (Gorbunova et al., 2008).This characteristic makes Drosophila an ideal model organism for conducting large-scale longevity studies within a relatively short timeframe.

Locomotor assays
Neuronal dysfunction and degeneration significantly affect locomotor ability, and thus, locomotor behaviour can be used as a more specific indicator of neurodegeneration in Drosophila.The most common assay is the climbing assay, and this is based on the negative geotaxis response, where flies will rapidly climb up the vial when tapped to the bottom of a vial (Chakraborty et al., 2011).
Traditionally, climbing performance is measured as height achieved after a set time (Gargano et al., 2005), scored manually, recently the development of FreeClimber (Spierer et al., 2021), an open-source Python-based system allows the simultaneous monitoring and scoring of climbing speed of a population of flies, vastly speeding up the assay.Drosophila activity monitor is another automated system for locomotor activity that produces continuous data collection throughout a set period of time  (Pfeiffenberger et al., 2010).An 'activity monitor' holds individual flies in sealed tubes in which an infrared beam detects every cross of the tube midpoint.The number of crossings per fly can be summed up over a continuous period of time and reveal changes in normal locomotor activity (Pfeiffenberger et al., 2010).These tools are efficient computational methodologies that facilitate standardised and accurate analysis in addition to alleviating inherent systematic biases associated with conventional manual measurements.The significance of such tools is highly valuable in the context of modifier screens, as they enable rapid and reliable analysis of substantial volumes of data.

Histological assays
It is widely accepted that the presence of vacuoles in the Drosophila brain reflects neuronal loss (Cook et al., 2012;Da Cruz et al., 2008;Loewen & Ganetzky, 2018).Quantification of the number and size of vacuoles in the Drosophila brain is a conventional method to assess the degree of neurodegeneration (Behnke et al., 2021;Herman et al., 1971;Ulian-Benitez et al., 2022) and is commonly employed in studies utilising Drosophila models of amyloid toxicity (Coelho et al., 2018;Iijima et al., 2008;Ray et al., 2017).
As well as models of Aβ1-42 toxicity, flies have been used to prove toxicity of novel Aβ species identified in human brains.For example, Aβ1-43 was shown for the first time in flies to trigger aggregation of Aβ1-40, which is usually not toxic (Burnouf et al., 2015), and pyroglutamate-modified Aβ (Aβ pE3-42 ), an N-terminal modified Aβ species, was shown to be more toxic than Aβ1-42 itself (Sofola-Adesakin et al., 2016).
In summary, a number of fly models of AD have been developed to explore amyloid toxicity in Drosophila, some of the more common ones are displayed in Table 2. To note, this is not an exhaustive list of the models for amyloid toxicity that have been created.It can be considered a representation of the variety of models available.

INSIGHTS FROM DROSOPHILA MODELS OF AMYLOID TOXICITY
Drosophila models offer valuable insights into establishing the causal effects of variants identified through human GWAS studies in influencing disease outcomes or uncovering potential downstream effectors of these variants.For instance, PICALM, a well-characterised risk factor for AD, has been demonstrated to affect synaptic glutamatergic signalling in an Aβ fly model (Yu et al., 2020).Similarly, TREM2, a potent disease modifier in humans, has been found to influence Tau toxicity but not amyloid toxicity in fly models.This observation suggests a potential specific role for TREM2 variants in the development of AD (Sekiya et al., 2018).Furthermore, Drosophila models of amyloid toxicity enable researchers to delve into the fundamental neurodegenerative processes at the molecular and genetic levels.They help identify novel connections or expand upon findings from human studies, ultimately enhancing our understanding of disease mechanisms.For example, Drosophila models have demonstrated that the upregulation of glial engulfment receptors can rescue Aβ accumulation and toxicity, emphasising the significant role of glial cells in Aβ clearance (Ray et al., 2017).Additionally, the modulation of mitochondrial dynamics, achieved by overexpressing the regulator Miro, has been shown to mitigate neurodegeneration in AD fly models, suggesting a direct involvement of mitochondria in disease progression (Panchal & Tiwari, 2020).
Fly models have also helped to shed light on other crucial aspects of AD pathogenesis, stemming from observations of patients' symptoms.For instance, fly models have been used to show that the upregulation of glucose import in neurons can ameliorate amyloid toxicity, indicating that the drop in glucose metabolism observed in patient brains is a direct driver of disease (Niccoli et al., 2016).Similarly, considering the impact of gut microbiota dysbiosis on brain inflammation in humans, Drosophila Aβ models have revealed that enterobacteria infection exacerbates the progression of AD (Wu et al., 2017).Drosophila has also been employed to explore emerging research areas related to amyloid toxicity, offering insights that can potentially lead to the identification of novel therapeutic targets.For example, a recent paper demonstrated that the retention of proteins in the endoplasmic reticulum rescued amyloid toxicity in a Drosophila model but interestingly in the absence of diminishing Aβ expression levels (Catterson et al., 2021).
Drosophila models of amyloid toxicity have also been used to identify and screen potential therapeutic targets.Drosophila has been employed to investigate the impact of approved acetylcholine esterase inhibitors (aChEIs), such as rivastigmine, on amyloid toxicity and recently synthesised aChEIs (Siddique & Naz, 2022;Uras et al., 2021).
Arguably, the major advantage of Drosophila disease models lies within their ability to easily undergo largescale screens for genetic modifiers, and this has been successfully applied to amyloid toxicity.The most commonly screened phenotype is the eye morphology, as this provides a quick, easily scorable and reliable method for measuring neurodegeneration.Cao et al. (2008) screened 1963 strains for modifiers of the Aβ1-42 model eye phenotype and found 23 modifiers, identifying proteostasis pathways and chromatin transcription regulation as important regulators of Aβ toxicity.Tan et al. (2008) using a similar screen, detected that loss of function of Toll, the Drosophila homolog of the mammalian Interleukin-1 (IL-1) receptor, strongly suppresses hAβ1-42-induced neuropathology.This finding implicates Aβ pathology in neuroinflammation, and following this study, inhibition of IL-1β, a pro-inflammatory cytokine, was demonstrated to prevent memory impairments in rodents (Williamson et al., 2011).Furthermore, treatment with fenamate non-steroidal antiinflammatory drugs prevented memory deficits in an Aβ1-42 rodent model via inhibition of IL-1β secretion (Daniels et al., 2016).Screening via climbing and lifespan assays have also effectively identified candidate modifiers of Aβ1-42 neuropathology in Drosophila (Belfiori-Carrasco et al., 2017;Sanokawa-Akakura et al., 2010;Xu et al., 2023).Automated computational analysis of phenotypes is now becoming increasingly integrated into genetic screens, increasing the power and speed (Belfiori-Carrasco et al., 2017;Liu et al., 2015;Rival et al., 2009).Yang et al. ( 2023) used an automated image analysis tool to screen for naturally occurring variants as genetic modifiers of Aβ1-42 and Tau-induced eye phenotypes, identifying 14 unique variants, implicating neuronal and developmental processes in disease pathogenesis.

LIMITATIONS OF DROSOPHILA MODELS OF AMYLOID TOXICITY
It is important to acknowledge that as an invertebrate species, a major drawback of Drosophila model systems are the substantial physiological and anatomical differences that limit translatability to humans.The Drosophila brain is considerably smaller in size with substantially fewer neurons in comparison to the human brain, which limits the capacity to accurately translate to the study of defined human brain regions, such as the hippocampus, and neural connectivity.Moreover, Drosophila lack an adaptive immune system (Padmanabhan & Götz, 2023), which limits their ability to accurately model neuroinflammation, a factor that is increasingly recognised as a significant contributor to AD pathogenesis (Uddin et al., 2021).Additionally, some common solvents for drug compounds are highly toxic to Drosophila, in particular dimethyl sulfoxide (Cvetković et al., 2015), which can interfere with pharmacological testing in Drosophila model systems.In contrast, they tolerate ethanol well up to 4%, as they naturally feed on rotting fruit with high ethanol concentrations; however, high concentrations of ethanol are also toxic (Bayliak et al., 2016).Therefore, it is good practice to incorporate vehicle-alone controls when performing drug-experiments.
Intriguingly, the direct expression of hAβ1-42 in fly neurons results in pronounced neurodegenerative phenotypes, despite the absence of plaques containing amyloid fibril structures (Iijima et al., 2004(Iijima et al., , 2008)).Staining of the fly brain has revealed that hAβ1-42 overexpression presents as diffuse deposits (Iijima et al., 2004(Iijima et al., , 2008) ) which is in contrast to the hAPP/hBACE model where the formation of amyloid plaques with fibrillar structures is observable (Chakraborty et al., 2011;Greeve et al., 2004).Furthermore, it has been observed via staining that fly brains pan-neuronally expressing hAβ1-42 exhibit mainly intracellular localisation of amyloid deposits.There is a growing body of evidence in rodent models of amyloid toxicity that extracellular plaque deposition is preceded by intraneuronal Aβ1-42 accumulation (Billings et al., 2005;Wirths et al., 2001).Although this finding may diverge from the classical pathological hallmark of extracellular plaques composed of amyloid fibrils, it should not be perceived as a fundamental limitation.Instead, it indicates that the Drosophila model directly driving hAβ1-42 in neurons potentially reflects an earlier stage of amyloid toxicity.
It is crucial to acknowledge that Drosophila will never attain flawless recapitulation of an entire complex mammalian disease like AD.The principal advantage of Drosophila resides in their genetic manipulability, enabling the modelling of specific aspects of the disease, such as amyloid toxicity.

CONCLUDING REMARKS
In AD research, Drosophila models offer quick, consistent and well-characterised phenotypes, allowing for highthroughput screening of gene and drug modifiers.While lacking mammalian features in the CNS, they remain invaluable for narrowing possibilities and pointing to new avenues that can then be further explored in mammalian and human systems.

A U T H O R C O N T R I B U T I O N S
Elli Tsintzas wrote the first draft, Teresa Niccoli reviewed and revised the article.All the authors read, reviewed, and approved the final version of the article.

F
Established methodologies for assessing neurodegenerative phenotypes in Drosophila.(a) Assessment of eye morphology.A decrease in eye size and a rough eye appearance are morphological phenotypes that are widely accepted as a qualitative assessment of retinal neurotoxicity and neurodegeneration.GMR driver lines can be used to drive transgenic expression, for example, hAβ1-42 peptides, in the differentiating retinal photoreceptor neurons of the developing eye.(b) The climbing assay.Neuronal dysfunction and degeneration result in reduced locomotor behaviour in Drosophila, which can be observed as a reduction in climbing in the climbing assay.Flies are placed in a vial, mechanical stimulation via tapping the flies to the bottom of vial induces negative geotaxis, leading to a 'startle-response' of rapid climbing up the vial that is severely impaired in neurodegenerative phenotypes.(c) The lifespan assay.Modifiers of longevity can be easily assessed by scoring lifespan and statistically analysing the survivorship curve.(d) The Drosophila activity monitor (DAM) assay.The DAM assay is an automated tool that can assess locomotor behaviour over a continuous period of time in the absence of locomotion induction via the negative geotaxis response.Individual flies in sealed tubes are held in an 'activity monitor' in which an infrared beam detects every cross of the tube midpoint.Changes in normal locomotor behaviour can be detected by summing up the number of crossings per fly over a continuous period of time.GMR, Glass Multimer Repeat; hAβ1-42, human amyloid-beta 1-42 peptide.
We thank our funders: ARUK Senior Fellowship ARUK-SRF2018A-003 (TN), an MRC grant MR/V003585/1 (TN), and a Dunhill Studentship (ET).We thank the associate editor and the reviewers for their useful feedback that improved this paper.

Gene Description Implicated core biological processes Fly orthologue with % similarity to the human gene
14691809, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ahg.12554by University College London UCL Library Services, Wiley Online Library on [05/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14691809, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ahg.12554by University College London UCL Library Services, Wiley Online Library on [05/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14691809, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ahg.12554by University College London UCL Library Services, Wiley Online Library on [05/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License TA B L E 2 The most common Drosophila models of amyloid toxicity.