|Home | About | Journals | Submit | Contact Us | Français|
Alzheimer's disease is the leading cause of cognitive loss and neurodegeneration in the developed world. Although its genetic and environmental causes are not generally known, familial forms of the disease (FAD) are due to mutations in a single copy of the Presenilin (PS) and Amyloid Precursor Protein (APP) genes. The dominant inheritance pattern of FAD indicates that it may be due to gain or change of function mutations. Studies of FAD-linked forms of presenilin in model organisms, however, indicate that they are loss of function, leading to the possibility that a reduction in PS activity might contribute to FAD and that proper psn levels are important for maintaining normal cognition throughout life. To explore this issue further, we have tested the effect of reducing psn activity during aging in Drosophila melanogaster males. We have found that flies in which the dosage of psn function is reduced by 50% display age-onset impairments in learning and memory. Treatment with metabotropic glutamate receptor (mGluR) antagonists or lithium during the aging process prevented the onset of these deficits, and treatment of aged flies reversed the age-dependent deficits. Genetic reduction of DmGluRA, the inositol trisphosphate receptor (InsP3R) or IPPase also prevented these age-onset cognitive deficits. These findings suggest that reduced psn activity may contribute to the age onset cognitive loss observed with FAD. They also indicate that enhanced mGluR signaling and calcium release regulated by InsP3R as underlying causes of the age-dependent cognitive phenotypes observed when psn activity is reduced.
Alzheimer's disease (AD) is a neurodegenerative disease characterized by progressive impairments in memory and cognitive abilities with a typical late age onset, although in cases of early onset familial Alzheimer's disease (FAD), the onset can be as early as the third decade of life. The histopathological hallmarks of AD are amyloid plaques and neurofibrillary tangles in the brains of afflicted patients (Alzheimer, 1907; Selkoe, 2002; Mattson, 2004; Hardy, 2006; Small and Gandy, 2006). The majority of FAD cases are linked to mutations in the Presenilin 1 and Presenilin 2 (PS1/PS2) genes, with additional cases linked to mutations in the Amyloid Precursor Protein (APP) gene. FAD cases exhibit dominant inheritance pattern where the disease is caused by a mutation in a single copy of one of these three genes (Mattson, 2004; Shen and Kelleher, 2007).
In humans as well as in animal and cell culture models of AD, FAD-linked Presenilin and APP mutations generally, although not in all cases, result in an enhanced ratio of Aβ 1–42 to Aβ 1–40 (Moehlmann et al., 2002; Schroeter et al., 2003; Qi-Takahara et al., 2005; Walker et al., 2005; Kumar-Singh et al., 2006; Sambamurti et al., 2006; DeStrooper, 2007; Isoo et al., 2007; Shen and Kelleher, 2007; Shioi et al., 2007). This finding and the dominant inheritance pattern of FAD has led to a model suggesting that FAD-linked mutations in PS1, PS2 and APP lead to gain of function or change of function forms of PS or APP protein (Hardy, 2006; Small and Gandy, 2006).
Studies in mice, Drosophila and C. elegans indicate that FAD-linked presenilin mutations have reduced function with respect to cleavage substrates (DeStrooper, 2007; Shen and Kelleher, 2007). Both loss-of-function mutations in psn and FAD-linked mutations lead to an increase in GSK-3β activity, which is normally negatively regulated by presenilin and may be a gamma secretase independent function (Baki et al., 2004; Serban et al., 2005; Shioi et al., 2007). Finally, the high number of different mutations in presenilin that give rise to FAD is more consistent with a loss of function mechanism as gain or change of function mutations for a protein are normally considered to be rare, particularly since the mutations in presenilin 1 are spread throughout different portions of the protein (Saura et al., 2004; DeStrooper, 2007; Shen and Kelleher, 2007). Taken together, these results suggest the possibility that a reduction in the functional activity of PS1 and PS2 contribute to the pathogenesis of FAD, including cognitive loss.
We investigate the effect of reducing psn function on cognition during aging as an approach to determine whether lowered psn function (but not completely absent function) may contribute to cognitive impairment. We find that reduction in psn function leads to age onset cognitive deficits. Through pharmacological and genetic studies we identify misregulation of mGluR signaling and InsP3R mediated calcium release, as causal for the observed age onset cognitive dysfunction. The implications of these studies toward understanding and possibly treating certain aspects of AD are discussed.
Drosophila strains used in this study are described in supplemental Table 1. The Drosophila strains were cultured as in McBride et al., 2005. Drugs were obtained from Tocris-Cookson (UK) and solubilized according to manufactures instructions. They were added to the fly food after cooling to the appropriate concentration. Concentrations used were based on dose response testing performed in a previous study on dfmr1 mutants (McBride et al., 2005). Vehicle for each drug was added to the appropriate control food for each experiment.
Virgin male flies were collected under ether anesthesia within 4 hours of eclosion. Virgin XX, yf females were collected on the day of eclosion and kept in food vials in groups of 10–15. Flies were aged in a 12:12 LD before behavioral training and testing. All testing was performed during the relative light phase. Mated females were 5 days old and observed to mate with a male the night before training. The virgin females that were used as targets were 4 days old. All male subjects were transferred to fresh control food the day before testing. Male flies were assigned to random groups for behavior training and testing, which was performed blind (Siegel and Hall, 1979; Kane et al., 1997; McBride et al., 1999). The total amount of time a male was engaged in courtship activity while paired with an unanesthetized target female during a test period of ten minutes or until successful copulation occurred, was scored. A courtship index (CI) was calculated as the percentage of total observation time spent courting (Siegel and Hall, 1979). Binning of naïve courtship behavior was performed as in McBride et al., 2005, except that the percentage of flies advancing to a particular stage of courtship during the courtship interval was scored and compiled. Locomotor testing was done as in McBride et al., 2005.
Tunel assays were performed according to Ye and Fortini (Ye and Fortini, 1999) using a S7110 kit from Oncor. Brains were treated as following the protocol for imaginal discs. Acridine Orange stainings were performed according to Heriche et al. (Heriche et al., 2003). Analysis of stained brains was performed by 3D reconstruction of optic stacks taken at 0.5 um using a Leica scanning confocal microscope. The number of cell death foci and their relative position were tabulated for each genotype. Analysis of mushroom body morphology was performed as described in McBride et al., 2005.
Courtship indices (CIs) of tested males were subjected to arcsin square root transformations to approximate normal distributions (Joiner Ml and Griffith, 1997; McBride et al., 1999). ANOVAs were performed on pair-wise comparisons of arcsin transformed data to get critical p-values. All statistics were performed using Statview 3.0.
Drosophila has a single presenilin (psn) gene ortholog, the protein sequence of which is approximately 50% identical to human PS1 and PS2 (Boulianne et al., 1997; Hong and Koo, 1997). The encoded presenilin protein is expressed in a wide range of tissues including the brain (Ye and Fortini, 1998). Isolation and characterization of several loss-of-function mutants have revealed that psn is an essential gene, required for proper neurogenesis and Notch processing (Struhl and Greenwald, 1999; Ye et al., 1999). Biochemical studies of the Drosophila presenilin protein have shown that it is a component of the gamma secretase complex like its mammalian counterparts (DeStrooper, 2003; Hu and Fortini, 2003; Takasugi et al., 2003).
The underlying nature of the mutations in PS1 and PS2 that give rise to FAD are unclear (Saura et al., 2004; Qi-Takahara et al., 2005; Walker et al., 2005; Kumar-Singh et al., 2006; Sambamurti et al., 2006; DeStrooper, 2007; Hardy, 2007; Isoo et al., 2007; Shen and Kelleher, 2007; Wolfe, 2007). However studies in model organisms indicate that the FAD linked mutations may result in a reduction of PS1 and PS2 activity (DeStrooper, 2007; Shen and Kelleher, 2007). This possibility suggests that some phenotypes associated with Alzheimer’s disease, including age onset cognitive loss may be due to a reduction in overall Presenilin activity levels. To investigate this possibility further, we tested the effect of reducing psn activity using known psn loss of function alleles. We evaluated the learning and memory capabilities of young adult flies heterozygous for a given psn mutant allele (hereafter referred to as “psn-het flies”), with only ~50% of the psn found in wild type control flies (the wild type control strain is Oregon-R, (Ore R), the background strain from which the psn alleles we have used in this study were derived). We then re-evaluated their capabilities at an older age to determine if any age-related cognitive decline could be observed. We took advantage of the large number of psn alleles with diminished function that have been obtained from several genetic screens. In total, 10 different loss-of-function alleles of psn were examined, to ensure that any observed effects were not due to second site mutations or other background effects, and crossed each to our wild type (WT) Ore R line (see Experimental methods and Supplemental Table 1 for a description of the alleles used).
Before embarking on an examination of the effects of aging on psn-het flies, using the courtship based assays, we first needed to establish a timeline in which courtship based learning and memory were not affected in our control Ore R (WT) flies. The effect of aging on naïve courtship, courtship conditioned learning and memory has not been extensively characterized. The lifespan of Drosophila is roughly 60 days (post-eclosion, in adult form), therefore a 5 day-old adult fly is relatively “young”, a 30 day-old fly is “moderately aged” and a 45 day-old fly is relatively “old”.
Courting Drosophila males perform a characteristic sequence of behaviors when paired with a female: orienting toward and following the female, tapping her with his forelegs, vibrating one wing, licking her genitalia, and attempting to copulate (Hall, 1994). The percentage of time the male spends performing any of these behaviors toward a target during a defined period of time is referred to as the courtship index (CI) (Siegel and Hall, 1979). We tested progressively older naïve males and found that aged 45 day-old WT flies perform naïve courtship as robustly as young 5 day-old flies (Figure 1A). This indicates that there are no sensory impairments with age that diminish courtship. Furthermore, the quality of this courtship was not different between 5 and 45 day-old WT flies (Figure S1), nor was the level of locomotor activity (Figure S2).
We next determined an age range within which learning and memory would be intact in WT flies. Learning and memory can be examined in Drosophila by assaying conditioned courtship behavior. In conditioned courtship, a male fly learns to modify his courtship behavior after experience with an unreceptive female (Siegel and Hall, 1979; Hall, 1994). Virgin females generally respond to a courting male by mating. However, recently mated females are unreceptive and have an overlapping but altered pheromonal profile that naïve males find less provocative than that of virgin female targets (Ejima et al., 2007). A naïve male paired with a mated female will initially court her, but his courtship activity soon decreases. When this learning during training (LDT) is quantified, by comparing the CI during the first ten minutes with a mated female to the CI of the last ten-minute period of a one-hour pairing, wild type flies typically show a ≥40% decrease in courtship activity (Joiner Ml and Griffith, 1997; Kane et al., 1997). Hence LDT is a form of behavioral plasticity, but is distinct and separate from courtship suppression assayed post-training, which is a form of associative memory (Tompkins et al., 1983; Ackerman and Siegel, 1986). When a male is paired with a virgin female after 1 hour of experience with a mated female, his courtship remains depressed for 2–3 hours (Siegel and Hall, 1979). This effect is not a general suppression of all courtship activity, since trained males do not modify their courtship of other pheromonally distinct targets (Ejima et al., 2005; Siwicki et al., 2005). After training with a mated female, memory is measured as a decrease in CI towards virgin females in trained males relative to naïve controls.
In Drosophila, five phases of memory have been elucidated by a combination of genetic and pharmacological dissection. There is an immediate recall memory (immediate memory) at 0–2 minutes post training; short-term memory out to 1 hour; medium-term memory out to 6 hours; anesthesia-resistant memory out to two days; and long-term memory lasting up to 9 days post training that appears to be dependent on protein synthesis (Skoulakis and Grammenoudi, 2006). Intact short-term memory is dependent upon intact immediate recall. However, immediate recall and short-term memory are distinct from LDT. Therefore, intact memory can occur without LDT, and learning during training can occur without post-training memory (Joiner Ml and Griffith, 1997; Kane et al., 1997; McBride et al., 2005).
To assess LDT, a male fly was placed in a training chamber with a previously mated female for one hour, and the amount of time the male spent courting in the initial ten-minute interval was compared to the time spent engaged in courtship in the final ten-minute interval. Normal LDT in 5 day-old WT flies is illustrated by a significant ≥40% decrease in courtship during the training (Figure 1B); short-term memory is illustrated in 5 day-old WT flies by a significant decrease in courtship towards virgin females one hour after training (Figure 1D). We found that 45 day-old WT flies also demonstrated intact LDT and short-term memory (Figure 1C and 1E). These studies indicate that the courtship-based learning and memory paradigm can be used to study aging related issues of cognition and behavior in flies.
Courtship was robust in naïve males of all 10 psn-het genotypes examined as young adults at 5 days of age (Figure 1F). Naïve males of these psn-het genotypes also displayed the normal steps of courtship behavior and normal levels of locomotor activity (Figures S3 and S4). LDT at 5 days of age in all 10 psn-het genotypes was normal, as evidenced by decreases of greater than 40% from CI: initial to CI: final during training with mated females (Figure 1G–H). To examine the immediate recall memory at 5 days of age, we took males that had just completed a one-hour training session with a previously mated female, and immediately placed them in a new chamber with a virgin target female for a ten-minute courtship assay. This CI was then compared to the courtship level of naïve males that had been placed in a training chamber for one hour with no female, before being introduced to a virgin target female. Immediate recall was intact in all 10 psn-het genotypes examined (Figure 1I–J). To assay short-term memory the trained male was placed in a holding chamber for 60 minutes, then subsequently placed in a testing chamber with a virgin female target. Nine of the psn-het genotypes were tested and found to have normal short-term memory at 5 days of age (Figure 1K–1L).
We next examined naïve courtship and LDT in 30 day-old (late middle age) psn-het males. No significant effect on the level of naïve courtship (Figure 2A), its quality (Figure S5), locomotor activity (Figure S6), photo- or chemo-taxis (not shown) was observed. However males of all 10 different psn-het genotypes failed to demonstrate the typical decrease in courtship activity in LDT at 30 days of age (Figure 2B–C). Impaired LDT is a very rare phenotype in Drosophila, but has been seen before in young adult flies that have altered CaMKII or PKC expression or activity (Joiner Ml and Griffith, 1997; Kane et al., 1997). These results demonstrate an age-dependent impairment in LDT in psn-het male flies by 30 days of age.
We next examined immediate recall and short-term memory in the psn-het flies at 30 days of age. We found that males of all 10 different psn-het genotypes displayed intact immediate recall at 30 days of age (Figure 2D–E). Intact immediate recall of courtship memory without intact LDT has been observed previously in young adult Drosophila (Joiner Ml and Griffith, 1997; Kane et al., 1997). Although immediate recall was intact, examination of short-term memory in 9 of the heterozygous mutant presenilin lines, at 30 days of age, revealed that this form of memory was no longer intact (Figure 2F–2G). This is in contrast to 5 day-old psn-het and 45 day-old WT flies, which exhibited intact short-term memory. This demonstrates that there is an age-dependent impairment of short-term memory in heterozygous presenilin loss-of-function flies.
In an effort to identify novel pathways affected due to the reduction in psn levels, we looked for genetic interactions between psn and other genes known to affect cognition and behavior which might play a role in presenilin signaling based on the literature. Interestingly, we identified a strong genetic interaction between psn and the Drosophila fragile X mental retardation gene, dfmr1. At 5 days of age, psn and dfmr1 heterozygous single mutant flies display normal naïve courtship levels relative to their given genetic background (Figure S7). However, 5 day old flies that are transheterozygous mutant for both psn and dfmr1 display a severe reduction in naïve courtship levels, which is rescued by the introduction of one wild type copy of a dfmr1 genomic rescue construct. Such heterozygous synergistic genetic interactions are rare and suggest that both genes act in one or more common pathways important for normal naïve courtship activity.
To explore the possibility of pharmacologically rescuing some of the age dependent psn-het phenotypes, and in light of the genetic interaction between psn and dfmr1 described above, we focused on treatments that rescue naïve courtship and cognitive deficits in dfmr1 mutants. One pathway that is known to be affected in both fly and mouse models of fragile X is metabotropic glutamate receptor (mGluR) signaling (Huber et al., 2002; McBride et al., 2005; Yan et al., 2005; Dolen et al., 2007; Bolduc et al., 2008). In previous studies we found that several of the phenotypes displayed by the dfmr1 mutants, including naïve courtship and memory, could be rescued by treatment with antagonists of DmGluRA and LiCl (McBride et al., 2005). More recent studies have also observed this and have also found that similar rescue is obtained by genetic reduction of DmGluRA activity, the sole Drosophila mGluR (Bolduc et al., 2008; Pan et al., 2008).
Our working model is that treatment of flies with both lithium and antagonists of DmGluRA will increase cAMP signaling and may also decrease InsP3R mediated calcium release (Figure 3A). Lithium inhibits GSK-3β activity by competing with the magnesium-binding site and increasing phosphorylation at serine 9 (Gould and Manji, 2005; Huang and Klein, 2006; Jope and Roh, 2006). There is an antagonistic relationship between PKA and GSK-3β, therefore this acts to upregulate the downstream effect of cAMP signaling because GSK-3β can inhibit CREB mediated gene transcription (Bullock and Habener, 1998; Grimes and Jope, 2001; Mai et al., 2002; Tanji et al., 2002; Hansen et al., 2004; Gould and Manji, 2005). Lithium also inhibits IPPase (Acharya et al., 1998) and IMPase (Berridge, 1993) thereby reducing the pool of InsP3 available to stimulate InsP3R mediated calcium signaling (Berridge, 1993; Takei et al., 1998; Williams et al., 2002). Previous heterologous cell-based studies have shown that DmGluRA can couple to heterotrimeric Gi-protein leading to decreases in cAMP levels (Parmentier et al., 1996). More recent in vivo studies have shown that DmGluRA also has properties typically found associated with Gq-coupled Group I mGluRs such as regulating synaptic morphology and AMPA receptor presentation (Bogdanik et al., 2004; Pan and Broadie, 2007; Pan et al., 2008). Thus antagonizing DmGluRA activity will increase cAMP levels (by relieving inhibition of cAMP after synaptic stimulation) and will possibly decrease InsP3R mediated calcium signaling (Figure 3A; McBride et al., 2005). We therefore explored the possibility that the age dependent cognitive deficits could be rescued by treatment with DmGluRA antagonists as well as lithium.
For this experiment, we used the competitive Group II antagonist (LY341495) and non-competitive mGluR5 antagonist (MPEP), as well as lithium at concentrations previously found to be optimal for treating dfmr1 mutants (McBride et al., 2005). LY341495 is a very specific Group II antagonist that has been shown to antagonize DmGluRA in the nanomolar range (Bogdanik et al., 2004). Although MPEP was developed as a non-competitive antagonist of the vertebrate Group I mGluR5, we have included it in our studies as the MPEP binding pocket of mGluR5 is more highly conserved with DmGluRA then any other vertebrate mGluRs and in previous experiments has similar activity to the known DmGluRA antagonists LY341495, MPPG, MTPG and shows similar effects to that obtained by genetically reducing DmGluRA activity (Parmentier et al., 1996; Pagano et al., 2000; Malherbe et al., 2003; Bogdanik et al., 2004; McBride et al., 2005; Bolduc et al., 2008; Pan et al., 2008). However as it has not been formally shown to be an antagonist of DmGluRA it is therefore a putative DmGluRA antagonist.
Two different psn-het genotypes (psn B3 and psn I2) were treated with different drugs or non-drug containing control vehicle solutions from days 6 to 29 of adulthood (post-eclosion). Flies were then transferred to control food containing no drug for one day before being tested. Flies were treated with 5 mM NaCl (ionic control for LiCl), 0.04% DMSO (the vehicle for LY341495), 5 mM LiCl (lithium), 8.6 µM MPEP or 400 nM LY341495. Remarkably, treatment with LiCl, LY341495 or MPEP improved LDT performance in psn-het flies (Figure 3H–J). In contrast, no improvement was observed in flies treated with NaCl or DMSO (Figure 3K–L). These experiments demonstrate that the impairment in LDT exhibited by the untreated psn-het flies can be prevented by treatment with lithium, LY341495 and MPEP.
This result with LDT raised the possibility that treatment with lithium, LY341495 and MPEP might also prevent the age dependent impairment in short-term memory that is exhibited by the psn-het males. Using the same set of pharmacologic treatments, we found that psn-het flies treated from 6–29 days post-eclosion with either 5 mM lithium, 8.6 µM MPEP or 400 nM LY341495 demonstrated intact short-term memory (Figure 3M–O), whereas those treated with 5 mM NaCl or 0.04% DMSO did not (Figure 3P–Q). This result demonstrates that the impairment in short-term memory exhibited by 30 day-old psn-het flies can be prevented by treatments aimed at reducing DmGluRA activity. In contrast, prolonged treatment of OreR (WT) flies with lithium, LY341495 and MPEP impaired LDT and STM, whereas vehicle treatment did not (Figure 3B–G).
The rescue that we observe when psn-hets are treated with LY341495, MPEP or lithium is analogous to the results we obtained when we treated dfmr1 mutants with these same drugs (McBride et al., 2005). The similarity of these results suggests that both psn-hets and dfmr1 mutants display cognitive deficits that are due to enhanced DmGluRA signaling. However, another possibility is that these drug treatments represent some type of panacea that can rescue any cognitive deficit observed in flies. With this second possibility in mind, we tested the effect of treating another disease model that displays age onset loss of short-term memory with LY341495, MPEP and lithium.
We have found that expression of human wild type alpha synuclein in the mushroom bodies of the fly leads to age dependent loss of short-term memory. The mushroom bodies are required for short-term memory formation in the conditioned courtship paradigm (Joiner and Griffith, 1999; McBride et al., 1999). In controls for these flies, short-term memory is intact at 30 days of age (Figure S8A). In flies expressing human wild type alpha synuclein in the mushroom bodies, short-term memory is intact at 5 days of age but is impaired at 30 days of age (Figure S8B). Treatment of these flies with lithium, LY341495 or MPEP failed to rescue this deficit (Figure S8C–S8D), indicating that rescue of cognitive deficits in the psn-hets by the drug treatments is disease specific and may not necessarily be extrapolated to other disease models displaying age-dependent cognitive impairments.
The age-dependent impairments in LDT and short-term memory observed in the psn-het flies could possibly be due to the loss of critical neurons due to cell death. There is a body of literature describing loss of synapses and neurons in the brains of Alzheimer's patients and in related mouse models (Selkoe, 2002; Mattson, 2004; Walsh and Selkoe, 2004; Hardy, 2006; Small and Gandy, 2006). To determine whether neuronal loss could account for the cognitive loss, we performed TUNEL staining on aged psn-het and control flies. Additionally, we examined the morphology of the mushroom bodies (MBs) in aged psn-het brains, which are required for short-term memory in the conditioned courtship paradigm (Joiner and Griffith, 1999; McBride et al., 1999). In the brains of 50 day-old psn-het and OreR flies, we failed to detect any consistent neuronal loss by TUNEL staining, or any gross defects in MB morphology, indicating that the observed cognitive loss is most likely not attributable to an age-dependent loss of neurons, or breakdown in MB integrity (Figure 4A–F). Consistent cell death was observed in 5 day old (positive control) Adar mutant brains, which have previously been shown to undergo neuronal cell loss (Figure 4C–D; (Palladino et al., 2000)).
Since the brain tissue of the aged psn-het flies appears to be morphologically normal, we considered the possibility that the age-related behavioral phenotypes could be reversed after they have already become established. Since the age-dependent impairments in LDT and short-term memory could be prevented by treatment with LY341495, MPEP or lithium, we next examined if these phenotypes could be reversed by these treatments. Reversal of the deficits would further indicate that the age-dependent cognitive impairments were likely the result of physiological neuronal defects (disrupted intracellular signaling possibly followed by synapse loss) and not cell death. Utilizing the same two genotypes (psn-hets with decreased presenilin activity) that were subjected to treatments as young adults (Figure 3), we now began treatments at day 30 and treated until day 44, testing at 45 days of age. As above, flies were treated with 5 mM NaCl, 0.04% DMSO, 5 mM lithium, 8.6 µM MPEP or 400 nM LY341495. Treatment with LY341495 and the putative DmGluRA antagonist MPEP, as well as lithium reversed the LDT (Figure 4G–I) and short-term memory (Figure 4L–N) phenotypes in both lines tested. In contrast, treatment with 5 mM NaCl and the DMSO vehicle from days 30–44 failed to restore normal LDT (Figure 4J–K), or short-term memory (Figure 4O–P) in either of the two lines. This demonstrates that the impairments in LDT and short-term memory exhibited by untreated psn-het flies can be reversed by treatments aimed at reducing DmGluRA activity. This finding is significant considering that our results may be relevant to the pathogenesis of Alzheimer's disease in humans which is only recognized after the onset of cognitive impairments, emphasizing the need to develop therapeutic strategies that can be implemented after the appearance of overt disease symptoms.
To validate the pharmacological studies described above, we tested the effect of genetically reducing the level of DmGluRA, the only metabotropic glutamate receptor in the Drosophila genome (Bogdanik et al., 2004), in the psn-het background. Given the reported binding specificities of MPEP, the ability of MPEP treatment to phenocopy genetic loss of function in DmGluRA mutants, the reported actions of lithium, and the fact that LY341495 has been shown to be a very efficient antagonist of DmGluRA, our drug treatments should reduce DmGluRA signaling activity (Bogdanik et al., 2004; McBride et al., 2005). If our observed rescue is occurring through reduction of DmGluRA signaling activity, our expectation is that genetic reduction of DmGluRA should similarly provide some level of rescue of the age onset cognitive deficits. To carry out this test we used the previously described null allele of DmGluRA (DmGluRA112) and precise excision wild type allele (DmGluRA2b) (Bogdanik et al., 2004).
In testing DmGluRA heterozygous flies alone, we found that at 30 days of age they display a defect in LDT as well as in short-term memory (Figure 5A and D), which is similar to what we observed when OreR control flies were treated with LY341495, MPEP and LiCl (Figure 3B–G). In contrast, the wild type controls for these lines displayed intact LDT and short-term memory at 30 days of age (Figure 5A and D). Consistent with our hypothesis, we find that reducing the gene dosage of the DmGluRA receptor by 50% in the psn-het background prevents the age onset loss of LDT and short-term memory previously observed at 30 days of age, whereas introduction of the precise excision allele has no effect (Figure 5B–C and 5E–F). These results validate the pharmacological studies described above and again indicate that enhanced DmGluRA signaling contributes to the age onset cognitive deficits. The results obtained from the DmGluRA heterozygous flies and OreR WT flies treated with MPEP, LY341495 and lithium, as described above, all indicate that LDT and short-term memory are sensitive to a reduction in the activity of this receptor as well.
The results of our pharmacological and genetic analysis indicate that reducing DmGluRA signaling rescues age onset cognitive deficits in psn-hets. Activation of DmGluRA through the Gi coupled pathway lowers cAMP signaling and through potential Gq coupling should increase InsP3R mediated calcium signaling (Figure 3A; McBride et al., 2005). Lithium as well as DmGluRA antagonists should act to increase cAMP mediated signaling by increasing PKA activity and may also act to lower InsP3R mediated calcium signaling (Figure 3A).
Previous studies indicate that presenilin proteins bind to the inositol triphosphate receptor and expression of mutant forms of presenilin protein lead to elevated Ca2+ signaling (Stutzmann et al., 2004; Cheung et al., 2008). The mechanism linking mGluR activity to increased InsP3R activity is shown in Figure 3A. Decreased presenilin function by FAD mutations has been shown to increase the sensitivity of the InsP3R to InsP3. This means that in response to glutamate release after synaptic stimulation the mGluR will be activated, thereby activating Gq and Gi. The Gq activation will in turn generate InsP3. Since there is less wild type presenilin present to basally inhibit the sensitivity of the InsP3R to InsP3, there will be enhanced InsP3R mediated calcium release in response to physiologic synaptic stimulation. To test if the rescue we observe with genetic and pharmacological manipulation aimed at reducing DmGluRA activity is possibly working by lowering enhanced InsP3R activity, we have undertaken manipulations to directly lower InsP3R activity. Toward this goal, we tested the effect of genetically reducing the levels of the inositol polyphosphate 1-phosphatase (IPPase) locus. IPPase as well as IMPase are directly inhibited by lithium and are required for InsP3 recycling and synthesis (Berridge, 1993). Genetic reduction of IPPase has been shown to reduce InsP3R signaling in flies (Acharya et al., 1998). We tested two loss of function alleles (IPPase1 and IPPase3) (Acharya et al., 1998) as heterozygotes and found that at 30 days of age, heterozygous males for one allele (IPPase1) lacked significant LDT, but heterozygote males for both alleles displayed normal short-term memory (Figure 6A and D). When crossed in the psn-het backgrounds the IPPase3 allele prevented loss of LDT in both psn alleles and the IPPase1 allele prevented this loss of LDT in the psnB3-het background (Figure 6B–C). Both alleles of IPPase prevented loss of short-term memory in both psn-hets lines (Figure 6E–F). These results indicate that reduction of IPPase activity can prevent the age onset cognitive deficits observed with the psn-het flies. Next we directly tested whether genetic reduction of the InsP3R gene itself could modulate the age dependent psn-het phenotypes. To test the effect of genetically reducing InsP3R activity, we used three different strength alleles; a hypomorph (wc361), a molecular null (90B0) and an antimorphic allele (wc703) (Venkatesh and Hasan, 1997; Deshpande et al., 2000). At 30 days of age heterozygous males for these alleles display weak (wc703) or no detectable LDT (wc361 and 90B0) and all display normal short-term memory (Figure 7A and E). When introduced into the psn-het background, two of these alleles prevented loss of LDT in the B3 allele of psn (Figure 7B–D) and all three alleles prevent the loss of short-term memory in both psn alleles at 30 days of age (Figure 7F–H). The 90B0 allele continues to prevent this loss of short-term memory at 40 days of age (Figure 7I). These results suggest that a reduction in InsP3R mediated Ca2+ release is also another potential route to ameliorate the age onset cognitive deficits associated with reduced psn levels.
To date, all cases of early onset familial forms of Alzheimer's disease are due to mutations in a single copy of PS1, PS2 or APP. The large number of different mutations in the PS1 and PS2 genes that cause AD as well as analysis of the protease function of FAD mutants is consistent with a loss of function of PS1 and PS2 as a contributing cause of FAD (DeStrooper, 2007; Hardy, 2007; Shen and Kelleher, 2007). This interpretation of the data suggests that the levels or activity of PS1, PS2 and APP are crucial for maintaining normal cognition throughout life. Based on these studies we have explored whether this dosage sensitivity of presenilin also exists in Drosophila, by examining the cognitive capabilities of psn-het flies in young and progressively aged psn-het flies. Using a conditioned courtship paradigm, we have found that deficits in LDT and short-term memory develop in the psn-het flies with age. These results further strengthen the hypothesis that some aspects of FAD and possibly AD may be caused by reduced levels of presenilin activity.
Previous Drosophila AD models have revealed phenotypes that have parallels with those observed in mouse models and human patients. These fly models have been based on ectopic expression of either human tau or human Aβ 40 or Aβ 42 peptides, and have been shown to exhibit either hyperphosphorylated tau or aggregations of Aβ that result in neurodegeneration (Wittmann et al., 2001; Iijima et al., 2004; Mershin et al., 2004; Crowther et al., 2005). Additionally, it has been demonstrated that expression of either human tau, Aβ 40 or Aβ 42 leads to memory deficits (Iijima et al., 2004; Mershin et al., 2004). In these models, memory deficits are already present in young adult flies. In contrast, we have uncovered cognitive impairments that are age dependent in the psn-het flies and occur prior to any detectable loss of neurons. In studying the psn-het flies, we detect a clear deficit in LDT and short-term memory that appeared by 30 days of age. Our examination for cell death was done on psn-hets that were at least 50 days old, e.g. 20 days older than when the defects in LDT and short-term memory are detected. Consistent with this finding we were able to reverse the cognitive deficits by initiating treatment at day 30, which would likely not occur if neuronal loss were the cause of the cognitive decline.
Although neuronal death is typical of AD, cognitive impairment precedes neuronal death in humans and animal models. Indeed, recent studies have indicated that synapse loss is better correlated with memory impairment than the histopathological hallmarks of plaques and tangles in AD patients (DeKosky and Scheff, 1990; Terry et al., 1991; Coleman and Yao, 2003). In fact, memory impairment in Alzheimer's disease can occur in the absence of plaques and tangles in animal models (Oddo et al., 2003; Gong et al., 2004; Iijima et al., 2004; Mershin et al., 2004; Billings et al., 2005) or in FAD patients (Raux et al., 2000; Amtul et al., 2002; Dermaut et al., 2004; Halliday et al., 2005). Additionally, within 2–4 years of AD onset 25–35 % loss of synapses is observed in biopsies of frontal cortex in afflicted patients (Davies et al., 1987). Furthermore, synaptic function may be impaired even before synaptic loss occurs in animal models of AD as well as in patients (Oddo et al., 2003; Palop et al., 2003; Westphalen et al., 2003; Yao et al., 2003). Thus, our results add to a growing body of literature that indicates that loss of synaptic plasticity can occur prior to or without plaque or tangle formation or neuronal loss. Our results also demonstrate that this loss of synaptic plasticity can occur solely due to a reduction in psn activity, in the absence of Aβ accumulation, which does not occur in Drosophila.
In a genetic search for pathways affected by a reduction in psn levels, we identified a strong genetic interaction between psn and dfmr1. Since this genetic interaction revealed a phenotype in courtship that was previously rescued in dfmr1 mutant flies by treatment with DmGluRA antagonists and lithium, we explored the possibility that these drugs might have efficacy in treating the age onset phenotypes of the psn-hets. Consistent with our findings, recent studies of mouse Alzheimer's disease models and of afflicted humans have suggested that calcium signaling is enhanced and cAMP signaling may be decreased (Selkoe, 2002; Gong et al., 2004; Mattson, 2004; Walsh and Selkoe, 2004). We propose that these two signaling pathways are downstream of DmGluRA signaling (Figure 3A).
In our studies, we demonstrated that treatment with lithium, the DmGluRA antagonist LY341495 and the putative DmGluRA antagonist MPEP prevented age-dependent impairments in psn-het flies at 30 days of age and can reverse LDT and short-term memory impairments if treatment is begun at 30 days of age, a time-point after the deficits appear. These treatments should function to counteract decreased cAMP signaling and may counteract increased InsP3R-mediated calcium release (Figure 3A).
The successful treatments of the psn-hets are consistent with results from two studies in mouse models of AD, both of which demonstrated improved cognition through pharmacological treatment without affects on Aβ plaque burden. First, Gong et al. (Gong et al., 2004), demonstrated that increasing cAMP levels can rescue behavioral plasticity, and second, Malm et al. (Malm et al., 2007), showed that indirectly inhibiting GSK-3β activity rescues behavioral plasticity. The results from our studies demonstrate that treatment with either lithium, LY341495 or MPEP can rescue cognitive loss due to reduced psn activity, a condition we propose exists in FAD and possibly sporadic AD. We therefore suggest that lithium and mGluR antagonists should be considered for further study as potential agents for therapeutic treatment of cognitive deficits associated with Alzheimer's disease.
We wanted to further elucidate the pathways downstream of DmGluRA that might be responsible for the rescue observed when the DmGluRA activity is reduced, genetically or pharmacologically. Lithium treatment and decreases in mGluR activity both lower InsP3R mediated calcium signaling in vertebrates. In our genetic tests, we found that genetic reduction of InsP3R mediated calcium signaling provides similar rescue to that observed when DmGluRA activity is reduced genetically. By testing multiple alleles of IPPase as well as InsP3R we found that, in general, loss of function mutants of both genes provided rescue of age onset deficits of LDT and short-term memory in the psn-hets.
The InsP3R is an endoplasmic reticulum (ER) localized Ca2+ channel that releases Ca2+ from the ER into the cytoplasm in response to InsP3 binding. It was recently demonstrated that FAD mutations of presenilin expressed in cells resulted in increased sensitivity to InsP3 mediated Ca2+ release, leading to enhanced cytoplasmic concentrations of Ca2+ (Cheung et al., 2008). These results fit with earlier studies where both FAD mutations and loss of functional activity of wild-type presenilin have been linked to increases in InsP3R mediated calcium signaling in cell culture and transgenic mouse models of AD (Peterson et al., 1988; Huang et al., 1991; Ito et al., 1994; Tatebayashi et al., 1995; Hirashima et al., 1996; Etcheberrigaray et al., 1998; Stutzmann et al., 2004). Our results demonstrating that pharmacological treatment with drugs that should reduce DmGluRA activity, as well as genetic reduction of DmGluRA, IPPase and the InsP3R are consistent with the hypothesis that reduction in this Ca2+ release mechanism might ameliorate the age onset cognitive deficits associated with AD. Moreover, these results provide the first test of the pathogenic nature of enhanced InsP3R mediated calcium signaling in AD. Previous studies have speculated on the pathogenic nature of this calcium signaling, but could not rule out the possibility that these changes were compensatory in AD. Herein we were able to demonstrate a rescue of cognitive deficits by decreasing this signaling, indicating that upregulation of this pathway leads to cognitive loss over time.
In conclusion, our study of Drosophila psn haplo-insufficiency contributes to the understanding of age onset cognitive loss and provides a new model to study aspects of Alzheimer’s disease. We examined several phases of learning and memory in young and older Drosophila adults and identified age dependent cognitive impairments in learning during training and short-term memory. We have demonstrated that lithium treatment can rescue cognition in an animal model of AD. These results are in contrast to a previous attempt with lithium treatment in a mouse model of AD that failed to rescue working memory impairments (Caccamo et al., 2007). We have additionally identified and demonstrated the efficacy of mGluR antagonists as a novel therapeutic target for the treatment of the cognitive deficits associated with AD. Our results also indicate that lowering IPPase activity or lowering InsP3R mediated calcium signaling can rescue cognitive loss. Finally, this study demonstrates that reversal of memory impairment after onset can be obtained by treatment with either mGluR antagonists or lithium. These results suggest novel therapeutic targets that may have relevance for treatment of AD to be explored in other animal models.
We would like to thank Evan Braunstein, Joseph Hinchey, Sean Campbell, Suzanne Zukin, Peter Davies, Myles Akabas, Nancy Carasco, Lloyd Fricker, Kami Kim, Matt Tremblay, Mike Gertner, Kevin Foskett and Goran Periz for critical comments during this project. We thank Oliver Schipper for help with fly husbandry. We would like to acknowledge Mel Feany, Gaiti Hasan and Charles Zuker for reagents. We would like to acknowledge funding from the FRAXA Research Foundation to S.M.J.M., C.H.C. and T.A.J., NIH funds from NS046573 and an ADC pilot award from AG-10124, as well as an ADR pilot award, a program of the American Health Assistance Foundation, for support of this research to T.A.J. We would like to acknowledge funding from the Autism Speaks to T.V.M and S.M.J.M., as well as funding to E.K. from The National Fragile X Foundation. Additionally, we would like to thank the Albert Einstein College of Medicine MSTP grant for funding S.M.J.M. and NIH grants and GM087650 and AG14583 (M.E.F).