AD is the most common form of dementia with ∼50 million suffers worldwide which is set to double every 20 years with ageing populations. In the US alone the societal and economic cost of dementia is $818 billion (Alzheimer’s Association, 2016) and yet there remain no cure for AD, and only limited symptomatic treatments that includes anti-cholinesterases, a NMDA glutamate receptor antagonist and an Aβ clearing antibody, however these do not halt or reverse neurodegeneration. Loss of cholinergic neurons in the cortex and hippocampus is particularly pronounced leading to the hallmark behavioural changes such as personality changes and loss of sleep and memory which precede a shortening of life (Querfurth and LaFerl a, 2010; Selkoe and Hardy, 2016; Mukhopadhyay and Banerjee, 2021). Post-mortem examination of AD brains reveals neurodegeneration is highly correlated with accumulation of extracellular Amyloid-β (Aβ) plaques that are produced by the amyloidogenic cleavage of APP by secretases generating aggregating Aβ peptides such as the 42 amino acid (aa) peptide (Aβ42) (Selkoe and Hardy, 2016). The cognitive deficits of AD are most strongly correlated with the accumulation of intracellular tangles of hyperphosphorylated Tau. These are encoded by the microtubule associated protein Tau (MAPT) gene that can be alternatively spliced into six main isoforms of Tau which vary in numbers of both N-terminal domains (e.g., 0, 1, or 2N) and C-terminal aggregating tubulin-binding repeats (e.g., 3R or 4R). 4R isoforms are thought to display stronger tubulin binding and aggregation than 3R isoforms and are upregulated in the AD brain (Arendt et al., 2016). Tau aggregation is increased by hyperphosphorylation by several different kinases including glycogen synthase kinase-3β, cyclin-dependent kinase 5, JNK, microtubule-associated regulatory kinase, CaMKII and DYRK1A (Hanger et al., 1992; Ferrer et al., 2005; Plattner et al., 2006; Wang et al., 2007; Hooper et al., 2008; Dolan and Johnson, 2010; Ghosh and Giese, 2015; Toral-Rios et al., 2020).

Only 5% of the causes of AD are thought to be genetic, making it difficult to predict who will develop the disease or not, with diagnosis only usually possible after extensive and irreversible neurodegeneration has occurred frustrating attempts at early intervention and development of effective treatments (Querfurth and LaFerl a, 2010; Selkoe, 2012; Selkoe and Hardy, 2016; Congdon and Sigurdsson, 2018; Soeda and Takas hima, 2020). However, one large cohort of people almost certain to get AD are those with DS or trisomy 21 which is caused by three copies of HSA21 as opposed to the normal two, with AD causal gene, APP located on HSA21, hence three copies of APP leading to increased likelihood of AD pathology (Selkoe and Hardy, 2016; Herault et al., 2017). DS can be diagnosed in utero, is common (1:450-1:2,200 live births) and is characterised by a range of developmental abnormalities associated with loss of motor skills, learning ability and sleep (Lott and Dierssen, 2010; Malak et al., 2015). People with DS have an 90% chance of developing Alzheimer’s disease (AD)-like symptoms (AD-DS) that includes progressive dementia from about 40 years old preceded by Aβ plaques and neurofibrillary tangles (NFTs) of Tau accumulation resulting in neurodegeneration from about 10 to 20 years old (Zigman, 2013; Wiseman et al., 2015). This pathology leads to the AD symptoms of progressive motor, cognitive and health decline (Anderson-Mooney et al., 2016) and shortening of life by ∼30 years (O'Leary et al., 2018).

However, another HSA21 gene, DYRK1A can also be considered a candidate causal gene for the pathological changes that occur in the DS and AD brain resulting in the associated cognitive and motor deficits (Ferrer et al., 2005; Wiseman et al., 2015; García-Cerro et al., 2017; Herault et al., 2017; Arbones et al., 2019). DYRK1A is highly expressed throughout the brain with increased levels in DS (Duchon and Herault, 2016; Kay et al., 2016). Therefore, the large alterations in DYRK1A gene-dosage are thought to have particularly negative consequences in DS, that may overshadow the more subtle effect it may have later in life that increase the risk of developing AD. DYRK1A has been genetically linked to late-onset AD (LOAD) within the Japanese population (Kimura et al., 2007). Furthermore people with DS having more than 100x risk of developing AD compared to rest of population, again largely thought to be due overexpression of APP and other chromosome 21 genes like DYRK1A throughout development and adulthood, leading to AD pathology including amyloid plaques and Tau tangles in young adults (Zigman, 2013; Wiseman et al., 2015; Selkoe and Hardy, 2016). However, in induced pluripotent stem cells from people with DS, the role of APP trisomy was demonstrated to be disconnected from tau pathology and neuronal cell death (Ovchinnikov et al., 2018). This suggests that DYRK1A and tau pathology is a larger contributor to AD-DS, than APP (Wiseman et al., 2018).

Regardless of the differing potential roles of APP, tau and DYRK1A in DS, AD-DS, and AD, there remains, no effective treatments for any of these diseases. This reflects the lack of knowledge of underlying mechanisms for these diseases and a deficit of new models, which is slowing progress and especially in development of new targets and treatments whose efficacy might translate to patients. Most of our understanding of the mechanistic changes that cause AD pathology comes from experiments performed in rodent familial AD models involving knock-ins or misexpression of different APP or Tau transgenes however these are not thought to recapitulate the sporadic disease which make up ∼95% of causes of AD. These factors have contributed to the large drug attrition of new drugs that although effective in these rodent models have not translated to any new AD treatments from clinical trials emphasising the need for new molecular models of sporadic AD (McGowan et al., 2006; Crews and Masliah, 2010; Van Dam and De Deyn, 2011; Gama Sosa et al., 2012; Guo et al., 2012; De Jager et al., 2014). Mouse models overexpressing Dyrk1a disrupt brain and eye development causing recapitulating cognitive and motor deficits seen in DS (Altafaj et al., 2001; Ahn et al., 2006; Guedj et al., 2012; Laguna et al., 2013; García-Cerro et al., 2014; Duchon and Herault, 2016). DYRK1A is also known to phosphorylate Tau at multiple sites including serine 262 that results in aggregation of pathological Tau neurofibrillary tangles associated with sporadic AD (Woods et al., 2001; Liu et al., 2008; Shi et al., 2008; Azorsa et al., 2010; Frost et al., 2011; Wegiel et al., 2011; Tenreiro et al., 2014). Increased DYRK1A also phosphorylates APP and promotes pathological processing of APP into neurotoxic Aβ peptides (Ferrer et al., 2005; García-Cerro et al., 2017; Arbones et al., 2019), overexpressing DYRK1A mice displaying AD pathology including neurodegeneration, disrupted synaptic plasticity and memory (Ferrer et al., 2005; Ahn et al., 2006; Liu et al., 2008; García-Cerro et al., 2017).

In Drosophila, neuronal overexpression of different human APP products [including human tandem oligomerising secreted Aβ42 (Chen et al., 2014a)] and mutants cause degeneration of the photoreceptor neurons of the fly eye, shortened lifespan, change in neuronal excitability as well as movement, circadian, sleep, and learning deficits (Iijima et al., 2004; Chiang et al., 2010; Spere tta et al., 2012; Chen et al., 2014a; Blake et al., 2015; Ping et al., 2015; Tabuchi et al., 2015; Higham et al., 2019a). Likewise, neuronal overexpression of AD-associated human Tau isoforms also result in degeneration of the photoreceptor neurons, central brain neurodegeneration, shortened lifespan, movement, changes in neuronal excitability, circadian rhythm, sleep, and learning defects (Wittmann et al., 2001; Folwell et al., 2010; Iijima-Ando and Iijima, 2010; Kosmidis et al., 2010; Beharry et al., 2013; Papanikolopoulou and Skoulakis, 2015; Sealey et al., 2017; Higham et al., 2019a; Higham et al., 2019b; Buhl et al., 2019; Lowe et al., 2019).

The Drosophila ortholog of DYRK1A is mnb, and produces five mnb isoforms, E-I, which all contain a highly conserved kinase domain (Hong et al., 2012; Gramates et al., 2017) and have conserved function in neuronal morphology, growth, brain development and cognition (Tejedor et al., 1995; Guimerá et al., 1996; Hämmerle et al., 2003; Degoutin et al., 2013; Ori-McKenney et al., 2016; Shaikh et al., 2016). Mnb is presynaptically localised at developing synapses and reducing its expression changed presynaptic structure and impaired recycling of transmitter vesicles with mnb-F overexpression ameliorating the effects of reduced mnb expression (Chen et al., 2014b). Gal4-mediated neuronal overexpression of mnb-H resulted in a significant increase in mnb-H expression, which is the isoform with the longest coding region (Hong et al., 2012; Gramates et al., 2017; Zerbino et al., 2018). Gal4-mediated neuronal overexpression of mnb-H was shown to cause motor impairment during development and ageing, shortened lifespan and resulted in age-related neurodegeneration with synaptic analysis showing increased number of glutamatergic boutons, enhanced spontaneous vesicular transmitter release, and slowed recovery from short-term depression (Lowe et al., 2019).

DYRK1A kinase antagonists have shown therapeutic potential for treatment for DS and AD in animal models (Smith et al., 2012; Duchon and Herault, 2016; Arbones et al., 2019). SM07883 DYRK1A inhibitor reduced pathological phosphorylation of Tau (including S212) in mice overexpressing MAPT P301L (associated with Frontal Temporal Lobe Dementia) reducing aggregation of Tau, neurodegeneration and improving behavioural deficits (Melchior et al., 2019). Leucettine DYRK1A inhibitor treatment of DYRK1A overexpressing mice rescued their cognitive deficits via correcting brain connectivity and expression of synaptic proteins (Nguyen et al., 2018). In 3xTg-AD mice that overexpress mutant Swedish APP (AD causal mutation), MAPT P301L and Presenilin M146V (AD causal mutation), a DYRK1A benzimidazole-like inhibitor reversed cognitive deficits via decreasing Aβ42 aggregation and decreasing phosphorylation of insoluble Tau (Branca et al., 2017).

Due to human DYRK1A and fly mnb sharing 82% amino acid identity which is even greater the kinase domains, many of the phenotypes and pharmacological sensitivity is conserved across species. For instance, conserved loss of function mutations reduced brain size in humans and flies, hence the name minibrain (Hämmerle et al., 2003; Lee et al., 2020a). With the DYRK1A-E396term being demonstrated to be loss of function mutation as it decreases the canonical phosphorylation of human Tau by DYRK1A. When the conserved mutation (mnb-D401term) was expressed in flies it removed the additive neurotoxic effect of co-expression of mnb and human Tau on eye degeneration (Lee et al., 2020a). Furthermore, a DYRK1A ATP-binding site competitive inhibitor, called CX-4945 reversed DYRK1A overexpression mediated increases in phosphorylation of Tau (S212), APP and PS1 in vitro and decreased degeneration of fly eyes overexpressing human Tau and decreased lethality caused by pan-neuronal developmental overexpression of mnb-H. The drug also suppressed pathological increases in Tau phosphorylation of mice overexpressing DYRK1A (Kim et al., 2016). This supports the potential of DYRK1A inhibition to treat DS and AD pathology. A recent highly potent and specific ATP-binding site small molecule DYRK1A inhibitor has been reported that crosses the blood brain barrier with 1 and 100 μM PST-001 significantly reducing DYRK1A activity reversing cognitive deficits of mouse DS models (Stensen et al., 2021a). Here we compare the effects of neuronal overexpression of fly mnb-H, human Tau (0N4R) and human tandem oligomerizing secreted Aβ42 on degeneration of adult photoreceptor neurons, longevity, motor performance, sleep, and memory and test the ability of a PST-001 DYRK1A inhibitor to suppress these phenotypes.

We compared the effect of neuronal overexpression of the fly ortholog of DYRK1A called mnb, human Tau or human Aβ42 and found they caused a range of AD and AD-DS relevant phenotypes in Drosophila including degeneration of the photoreceptor neurons, shortened lifespan, motor impairment, sleep, and memory loss. The human AD causal genes were more phenotypically extreme than fly mnb overexpression. We found treatment with the DYRK1A inhibitor PST-001 (Stensen et al., 2021a), caused an in vivo reduction of phosphorylation of panneuronally expressed human Tau at S262, a site previously identified as being phosphorylated by DYRK1A leading to AD pathology (Azorsa et al., 2010; Frost et al., 2011; Tenreiro et al., 2014). Therefore PST-001 is an inhibitor of the kinase activity of DYRK1A and its fly ortholog, mnb. PST-001 was effective at suppressing the disease relevant phenotypes across models without any detectable adverse effect on control flies. Consistent with PST-001 leaving enough endogenous mnb activity for wildtype functions in the assays studied but still being effective at inhibiting the high levels of kinase activity in the mnb overexpressing flies thereby rescuing its mutant phenotypes. Likewise, PST-001 was effective at ameliorating the disease-relevant phenotypes of human Tau overexpressing flies due to the reduction in pathological phosphorylation of S262 by mnb. Likewise, the abililty of PST-001 to rescue the same mutant phenotypes caused by human Aβ42 overexpression is likely due to kinase inhibition of mnb rescuing the pathological phosphorylation, processing and neurotoxic effects of the Aβ42 peptide, as occurs in other systems (Ferrer et al., 2005; García-Cerro et al., 2017; Arbones et al., 2019). The Western and phenotypic suppression data is consistent with 334 μM PST-001 not being able to completely inhibit mnb hence leaving enough endogenous mnb activity for wildtype functions in the assays studied hence the lack of deficits in the treated controls. Previous work suggests that PST-001 shows inhibition across the DYRK-family (human DYRK1A, DYRK2 and DYRK3) as well as being effective associative memory deficits in Ts65 Dn DS model mice (Stensen et al., 2021b), this and our data therefore is consistent PST-001 being highly effective against Drosophila DYRK1A (the product of the mnb gene).

This study provides important information on not only the function of mnb, but also the relative pathological consequences of overexpression of AD-associated human Aβ42 and Tau and the potential of DYRK1A inhibitor to treat these deficits. We found that overexpression of AD-associated human Tau 0N4R caused the greatest degeneration of the photoreceptors causing a reduction of ∼60% in eye size compared to control, with Aβ42 also being neurotoxic decreasing the eye by ∼50%, while overexpression of fly mnb caused more modest cell loss reducing the eye size by ∼25%. This confirms previous studies suggesting a link between misexpression of DYRK1A with eye size and neurodegeneration (Wegiel et al., 2008; Wegiel et al., 2011; Laguna et al., 2013; Duchon and Herault, 2016; García-Cerro et al., 2017; Watson-Scales et al., 2018), which then drives the accelerated decline in motor and cognitive function in DS and AD-DS and shortened lifespan all strongly correlated with such histopathological changes in the brain (Buchman and Bennett, 2011; Choong et al., 2015; Wiseman et al., 2015; Anderson-Mooney et al., 2016; Arbones et al., 2019). We found that treating flies overexpressing Tau, Aβ42 or mnb in the eye with a DYRK1A inhibitor suppressed degeneration such that their eyes were ∼20%–30% bigger than the equivalent untreated genotype. Therefore, not only did the DYRK1A inhibitor suppress neurodegeneration caused by the overexpression of mnb, but it was also able to suppress amyloid and Tau pathology via decreasing endogenous mnb phosphorylation, reducing the neurotoxicity of these genes. This suggests that the PST-001 DYRK1A has therapeutic potential for AD-DS.

We found DYRK1A inhibition was also able to suppress the shortened life associated with pan-neuronal overexpression of AD-associated Tau or Aβ42, again this was only a partial rescue with the treated flies living longer than their equivalent untreated genotype control but not as long as completely normal flies. Intriguingly this lifespan extension was not present in flies overexpressing mnb treated with the DYRK1A inhibitor PST-001, suggesting that the DYRK1A inhibitor may be able to directly suppress human Tau and Aβ42 pathological effects on senescence, further work will be required to elucidate these mechanisms.

These neurotoxic degenerative effects caused by overexpression of Tau, Aβ42 or mnb also had behavioural consequences when expressed in neurons. Firstly, the ability for flies to perform a co-ordinated negative geotaxis climbing response was greatly reduced by the disease associated genes, with DYRK1A inhibition being able to improve the performance of all genotypes. The drug was shown to completely behaviourally rescued the Tau and mnb overexpressing flies, such that their performance was indistinguishable from wildtype flies. Again, we do not know why DYRK1A inhibition showed differential benefits on the different models, but it is possible that this may be due to the transgenes being expressed in different types of neurons e.g., eyes (GMR-GAL4) compared to all neurons (elav-GAL4), suggesting flies may display selective vulnerability to the neurotoxic effects of the different transgenes. It should also be noted that pan-neuronal Aβ42 overexpression caused the most extreme (∼60%–70%) reduction in motor performance, and DYRK1A inhibition caused the greatest benefit (∼40% improvement), therefore there may be a potential floor effect, whereby the magnitude of the deficit caused by Aβ42 was too great for the drug treatment to completely return the genotype to wildtype.

We also found a similar differential sensitivity to DYRK1A treatment between the genotypes for treatment of the sleep loss caused by clock wide overexpression of the disease-associated genes. Clock expression of Tau, Aβ42 or mnb all caused sleep loss both in the day and night with PST-001 treatment fully rescuing sleep loss of mnb or Tau overexpression. Interestingly when the total sleep was split into day and night sleep, PST-001 was found to suppress day and night sleep loss of mnb flies, as opposed to treating just the nocturnal sleep loss of tauopathic flies. Strikingly people with AD also have similar difficulty sleeping at night (Aldrich et al., 1989; Vitiello and Borson, 2001) making them susceptible to nocturnal wandering (Logsdon et al., 1998). Our work also complements and extends previous work on the effect of clock expression of Tau and Aβ42 on circadian rhythms and sleep disruption in flies (Chen et al., 2014a; Blake et al., 2015; Cassar and Kretzschmar, 2016; Gerstner et al., 2016; Kim et al., 2018; Arnes et al., 2019; Buhl et al., 2019; Cassar et al., 2020). For instance, flies pan-neuronally expressing the Arctic mutant of Aβ42 sleep less during the day and night (Tabuchi et al., 2015), while expression of tandem Aβ42 caused behavioural arrhythmia (Chen et al., 2014a).

Our work also demonstrates the potential of DYRK1A inhibitors as sleep correctors for DS and AD-DS which may slow pathology and boost cognition, the high throughput capability of flies will allow further characterisation and screening of DS and AD circadian and sleep drugs (Dissel et al., 2017; Wang et al., 2020). People with DS also display disrupted sleep including sleep apnoea which can negatively impact their cognition and motor control (Chawla et al., 2020). Likewise, many people with AD in addition to the disrupted sleep at night mentioned previously also display sun-downing, where upon the patient experiences more anxiety and confusion in the evening (Vitiello et al., 1992; McCurry et al., 1999; Volicer et al., 2001). Insomnia, nocturnal wandering and not remembering where one is, are the primary reasons for eventual institutionalisation of people with AD, resulting in loss of independence, support networks and increased healthcare costs (Pollak and Perlick, 1991).

Furthermore, a robust circadian clock and sleep schedule improves memory function and is required for consolidation of long term memory (Gerstner and Yin, 2010) with post-mortem AD brain slices revealing neurodegeneration of the suprachiasmatic nucleus (SCN) of the hypothalamus which is the location of the mammalian circadian clock. Transgenic mice models of AD also show SCN degeneration (Sterniczuk et al., 2010; Stevanovic et al., 2017). Therefore, flies and mice expressing human 0N4R Tau exhibit behavioural dysfunction and neurophysiological changes, including elevated neuronal activity, which precedes neurodegeneration (Wittmann et al., 2001; Sterniczuk et al., 2010). Like flies (Buhl et al., 2019; Cassar et al., 2020), tauopathic mice show disrupted circadian rhythms and sleep showing neuronal hyperexcitability and inability to sleep at night (Stevanovic et al., 2017). The sleep-wake cycle is known to regulate brain interstitial fluid Aβ42 and tau in mice and cerebrospinal fluid Aβ42 and Tau in humans. Levels increase during the day and are removed at night with sleep deprivation further increasing pathological Tau seeding and spreading (Musiek and Holtzman, 2016; Holth et al., 2019). This reiterates the importance of healthy circadian rhythms and sleep for cognition and improving DS and AD symptoms as well slowing or preventing neurodegeneration in AD and AD-DS.

We also saw cognitive deficits, with MB memory neuron overexpression of Tau, Aβ42 or mnb reducing associative memory performance with PST-001 treatment increasing the memory of Tau overexpressing flies and improving the performance of the other mutant genotypes. This validates the use of PST-001 as a cognitive enhancer in preclinical animal models of DS, AD-DS, and AD. This suggests that DYKR1A inhibitors may be beneficial to people with DS prior to displaying memory impairment associated with AD-DS, allowing targeted treatment prior to the onset of irreversible neurodegeneration, thereby correcting the causes as opposed to just the symptoms of AD.

The results of our study are also consistent with finding that mice overexpressing human DYRK1A or mouse Dyrk1a, have similar motor and cognitive deficits reinforcing triplication of DYRK1A is likely to contribute to these behavioural deficits in DS and AD-DS (Altafaj et al., 2001; Martínez de Lagrán et al., 2004; Ortiz-Abalia et al., 2008; Arque et al., 2013; Souchet et al., 2014; García-Cerro et al., 2017; Watson-Scales et al., 2018). Reiterating that mnb is both molecularly and functionally conserved with DYRK1A. Likewise, this work also reinforces that neuronal overexpression of AD-associated Tau (Wittmann et al., 2001; Folwell et al., 2010; Iijima-Ando and Iijima, 2010; Kosmidis et al., 2010; Beharry et al., 2013; Papanikolopoulou and Skoulakis, 2015; Sealey et al., 2017; Higham et al., 2019a; Higham et al., 2019b; Buhl et al., 2019; Lowe et al., 2019) and Aβ42 (Iijima et al., 2004; Chiang et al., 2010; Spere tta et al., 2012; Chen et al., 2014a; Blake et al., 2015; Ping et al., 2015; Tabuchi et al., 2015; Higham et al., 2019a) result in neurodegeneration leading to a range of disease relevant phenotypes in flies.

It would be interesting to understand further how Tau, Aβ42 and mnb interact to influence neuronal function causing the behavioural changes seen. In addition to neurodegeneration, overexpression of these disease associated genes disrupts intrinsic and synaptic plasticity (Higham et al., 2019a; Higham et al., 2019b; Buhl et al., 2019; Lowe et al., 2019), contributing the loss of memory and sleep reported. Together this work also shows that the fly models replicate many of the key features of rodent transgenic Tau, Aβ42 and DYRK1A models including changes in excitability, increased Ca²⁺ signaling, neurodegeneration, and impaired synaptic plasticity, memory, and sleep (Selkoe, 2012; Spillantini and Goedert, 2013; Spires-Jones and Hyman, 2014; Wang and Mattson, 2014; Booth et al., 2016a; Arendt et al., 2016; Booth et al., 2016b; Kay et al., 2016; Selkoe and Hardy, 2016; Biundo et al., 2018; Arbones et al., 2019). This body of work also highlights how well these behaviours and mechanisms are conserved between flies and mammals including humans, allowing the well-established assays, genetics, and rapid ageing of Drosophila to study these processes.

We showed that Drosophila which overexpress mnb, human Tau or Aβ42 in different neuronal populations cause a range of AD and AD-DS relevant phenotypes and pathology including degeneration of the photoreceptor neurons, shortened lifespan, motor impairment, sleep, and memory loss. We demonstrate that the PST-001 DYRK1A inhibitor decreases phosphorylation of human Tau at S262 and is able to suppress disease relevant phenotypes caused by overexpression of mnb, human Tau or Aβ42. This allows researchers to exploit the genetic tractability and short generation time of this non-protected species to study the role of these genes in DS and AD processes and then rapidly characterise them as potential targets for new drugs to reverse these disease-relevant deficits.