Identification of novel modifiers of Aβ toxicity by transcriptomic analysis in the fruitfly

The strongest risk factor for developing Alzheimer's Disease (AD) is age. Here, we study the relationship between ageing and AD using a systems biology approach that employs a Drosophila (fruitfly) model of AD in which the flies overexpress the human Aβ42 peptide. We identified 712 genes that are differentially expressed between control and Aβ-expressing flies. We further divided these genes according to how they change over the animal's lifetime and discovered that the AD-related gene expression signature is age-independent. We have identified a number of differentially expressed pathways that are likely to play an important role in the disease, including oxidative stress and innate immunity. In particular, we uncovered two new modifiers of the Aβ phenotype, namely Sod3 and PGRP-SC1b.

study, we expressed the wild type Ab 42 coding sequence whereas, for gene-specific RNAi knock-down or over-expression experiments, we extended our study by validating the original observations in flies expressing the familial AD-linked Arctic (E22G) variant of Ab 42 . We investigated the changes in transcriptome profiles over time for both control flies and those expressing Ab 42 . The use of such an early onset model allowed us to distinguish between changes in gene expression due to AD and those due to ageing.

Results
Transcriptome analysis of AD and control flies over time. To investigate the differences between the processes of ageing and AD we used microarrays to measure changes in gene expression over time in control and Ab 42 -expressing flies (hereafter referred to as Ab flies). The Ab flies used in these experiments carried 2 copies of a transgene expressing human Ab 42 (elavGAL4 . UAS-Ab 42 ), the 2 3 Ab 42 model. Ab flies have a much shorter lifespan with a median survival (50% flies still alive) of 23 vs. 63 days in control flies (see Supplementary Fig. S1, for the climbing, survival, and molecular phenotypes of Ab flies used in our experiments). Therefore, in order to clearly distinguish Ab and age-related changes in gene expression, we compared Ab and control flies using a two-pronged strategy. In the first experiment, we age-matched flies according to their chronological age and extracted RNA samples from fly heads at days 3, 10 and 20 for both Ab flies and the control cohort ( Fig. 1). At these time points the survival of the flies is approximately 100% and so we can match samples from flies according to their chronological age.
Beyond day 20, mortality in Ab flies begins to increase and it is conceivable that increasing mortality itself could be associated with changes in gene expression. To compare gene expression changes associated with the increase in mortality in Ab flies with those associated with normal ageing, we continued to extract RNA samples from Ab and control flies but at different times such that the % survival of each group was the same (80% and 20% survival, this corresponded to days 21, and 25 in Ab flies and days 56, and 68 for control flies). We used this data for a separate analysis of the gene expression changes over time according to survival. All RNA samples were extracted from one cohort of flies (see Methods) to reduce the effect of biological variability, but the age-matched and survival-matched samples were analysed separately and we refer to them as separate experiments. The day 3 (100% survival) sample is common to both analyses.
Analysis of the gene expression profiles from the two experiments (see Methods for details) identified 233 (day-matched) and 636 (survival-matched) differentially expressed genes, with a total of 712 genes combined (and an overlap of 157 genes, see Supplementary  Table S1 online). We clustered average expression levels of the significantly differentially expressed genes using fuzzy c-means clustering (R mfuzz package) 12,13 for each of the two experiments ( Fig. 2: ''a-e'' day-matched and ''f-j'' survival-matched).
In each experiment, we identify two categories of clusters. The first category of clusters (for both experiments), represents genes that are differentially up-or down-regulated in Ab flies compared to controls, but do not change with time (Clusters ''b'', ''c'', ''f'' and ''i''; 69, 47, 95 and 85 genes respectively -note that as fuzzy clustering is used, the number of genes assigned to each cluster does not total the number of significant genes). These genes presumably represent a direct response of the flies to the Ab aggregation insult. By contrast the second category of expression profile clusters, for both experiments, represents genes with expression profiles that change over time (clusters ''a'', ''d'', ''e'', ''g'', ''h'' and ''j' ' -46, 35, 39, 121, 169, and 171 genes respectively). Based on these clusters, it appears that very few genes change expression over time in Ab flies compared to controls, and that changes over time are more pronounced in control flies.
We therefore analysed the expression profiles of all genes present on the array for Ab and control flies separately (see Methods) to identify all genes that change expression significantly over time in each group. For Ab flies we identified 144 genes, whereas for control flies we identified 612 (with an overlap of 90 genes, Supplementary Table 1 online). For both Ab and control flies, the genes changing over time were involved in similar pathways based on enriched Gene Ontology (GO) terms, in particular immune response and metabolic processes (data not shown). Therefore we conclude that the dysregulation is gene specific rather than pathway specific. For each of these genes, we tested the correlation between gene expression level and percent survival. Of the 144 genes whose expression level changes significantly over time in Ab flies, there were only four (IM23, CG14933, CG7830, CG8036) whose expression level correlated with survival (Pearson's Product Moment Correlation Coefficient (r), 0.8 # r # 20.8). There is little information available on the function of these four genes, so we are unable to explain why the expression of these genes in particular correlates with the decrease in survival. Based on the fact that only 144 genes change expression over time in Ab flies, we conclude that the transcriptional response to Ab expression in our fly model is mostly not age dependent.
By contrast, 195 out of 612 genes changing expression correlated with the decreased survival in control flies using the same threshold. Of the four genes whose expression correlated with survival in Ab flies, two were also correlated in controls (CG14933 and IM23) and the expression of both genes increased over time for both Ab and control cohorts. We therefore conclude that there is no common mortality signature (i.e. genes that change in expression over time in both AD and control cohorts and whose expression levels correlated with survival in both), and that the gene expression changes occurring with normal ageing are distinct from those associated with Ab expression in Drosophila. At the transcriptional level, the Ab expression-associated signature is a constant change in the relative level of transcription of a certain set of genes over all time points measured, rather than the signature of ageing that manifests as a change in the level of transcription over time.
The 612 genes whose expression changes significantly over time in our control flies constitute a transcriptional signature of normal ageing in control flies. We compared this list of genes to a previous ageing study in Drosophila 14 . 44% (282/612) of genes changing  expression with age in our control flies were also identified as ageingrelated genes by Landis et al 14 , significantly more than expected by chance (hypergeometric test for enrichment, p , 1 3 10e-19). Furthermore, 63% of our 195 ageing signature genes for which expression level correlated with decreased survival with age were identified as ageing genes in the same study 14 (hypergeometric test for enrichment, p , 1 3 10e- 22). An analysis of the correlated genes for the control cohort using the Flymine database 15 revealed that the  pathways in which these genes are involved are significantly enriched in a number of pathways and GO annotations related to xenobiotic metabolism, glutathione metabolism and immune response pathways 16,17 . This is consistent with current theories of ageing [18][19][20] . The transcriptional changes associated with ageing in our control flies are therefore consistent with expected changes with age in Drosophila, both at the single gene and pathway level. The fact that our analysis produces expected results for control flies lends weight to the conclusions drawn from our analysis of the Ab flies.
We performed Principal Components Analysis (PCA) of the expression data ( Supplementary Fig. S2) and found that, while across the first three sample points Ab and control groups cluster together, in the final two (where samples did not have the same age in days) they do not, confirming that the downstream effects of Ab expression in this model are not similar to normal ageing, at least at the level of transcription. Supplementary Table S1 online lists all the differentially expressed genes, what method determined their significance, the cluster to which they belong for both experiments, their Pearson correlation with survival and their normalised expression level (see Methods) for each data point and for each of the replicates.
Oxidative stress-related changes in gene expression with AD. The transcript most highly up-regulated (,8-fold; Fig. 3a) in Ab flies in our experiments was Sod3 (CG9027) which encodes an extracellular Cu/Zn superoxide dismutase 21 . In Drosophila there are 4 Sod3 transcripts (and 3 protein products), and two of these, Sod3-RD and Sod3-RE are specifically up-regulated in the Ab flies (see Fig. 3a for Sod3-RD/RE and Fig. 3b for Sod3-RA/RB profiles). Superoxide dismutase enzymes have recently been linked to inflammation, with Sod3 proposed to contribute to this process both by scavenging free radicals and also, more directly, by affecting immune responses and signal initiation 22 . Interestingly, in 2009, before Sod3 had been discovered in flies 21 , Rival et al 10 showed that overexpression of Sod1 was associated with a decrease in lifespan in Ab flies.
Cyp6a20, encoding a cytochrome P450 enzyme, was another gene with significantly altered expression in Ab flies compared to control flies. We found that Cyp6a20 expression was significantly reduced in Ab flies in an age-independent manner (see Supplementary Table S1 for expression data). Cyp6a20 was also identified in a genetic screen as a modifier of the survival phenotype in Ab flies 10 and cytochrome P450s have previously been identified as being a group of enzymes that are up-regulated with age 23,24 . On the other hand, genes such as Prx2540-2 ( Fig. 3c), encoding peroxidases involved in the clearance of hydrogen peroxide 25 , appear to change with time in control flies. This change was not observed in Ab flies. Increased expression with age in control, but not Ab, flies was also observed for genes involved in glutathione metabolism, for example several of the glutathione-Stransferases increased expression over time only in control flies (Supplementary Table S1). Interestingly, one member of this family, GstE9, did increase expression over time in Ab flies. We again conclude that the same processes are important in Ab toxicity and ageing, but with different specific genes affected by each process.
Immune response-related changes in gene expression with AD. Two important processes that showed significantly altered gene expression profiles in our AD model are the innate immunity and defence response pathways. Many genes in these pathways normally become up-regulated during ageing 23,26 . In control flies, increased expression with age was observed for genes involved in the antibacterial humoral response 14 , such as Dpt, CecA1, CecA2 and Drs.  Similar changes were not observed in Ab flies. For example, immune response genes, such as that encoding the proteoglycan recognition protein PGRP-SC1b were significantly altered in our experiment (Fig. 3e), with increased transcript levels in Ab flies vs. controls at all time points. PGRP-SC1b is a catalytic PGRP that is likely to be involved in down-regulation of the Imd innate immunity pathway in response to injury 27 .
Cellular transport and chaperone related changes in gene expression with AD. Intracellular transport, specifically endocytic processing, has been implicated in AD using GWA studies for sporadic AD 28 . In our study, sec31, a gene coding for an essential component of the COPII coat for ER to Golgi transport 29 , was increased in expression in Ab flies compared to controls (Fig. 3f). CG14715, another gene implicated in intracellular transport and protein folding, encodes the Drosophila ortholog of FKBP2/FKBP13, a prolyl-isomerase thought to function as an ER chaperone 30,31 . The expression of CG14715 was down-regulated in an age-independent manner in the Ab flies (Fig. 3d).
Identifying modifiers of the Ab phenotype using gene-specific RNAi. Our microarray analysis identified significant expression changes (between control and Ab flies) in 712 genes. These genes are potential modifiers of the Ab phenotype. In order to determine whether the observed expression changes are relevant to the onset or development of the disease or whether they are simply correlative, we manipulated the levels of four of these genes in the Drosophila AD model (Data summarised in Table 1). We chose two genes that are highly up-regulated in an age-independent manner in Ab flies: Sod3, which was identified as the most differentially expressed gene between the Ab and control flies (,8 fold increased levels in Ab flies, Fig. 3a) and PGRP-SC1b (,2 fold increased levels in Ab flies, Fig. 3e). These genes are involved in oxidative stress and innate immunity, respectively. We also investigated sec31, encoding a protein involved in intracellular transport (,1.5 fold increased levels in Ab flies, Fig. 3f) and CG14715, coding for a putative ER chaperone (,1.7 fold decreased levels in Ab flies, Fig. 3d).
In these experiments, we used a modified version of the fly AD model (elavGAL4 . UAS-Ab 42 arc) in which flies expressed a single copy of the Ab peptide containing the familial E22G (arctic) mutation that increases the aggregation propensity of the Ab 42 peptide 32 . It has been shown previously 11,32,33 that these models are equivalent and that their effect on the flies' lifespan is proportional to the aggregation propensity of the Ab variant. Luheshi et al. (2010) investigated the effect of mutations (including the arctic mutation) in the sequence of Ab 42 and showed that the aggregation propensity and in vivo toxicity as quantified by locomotor and survival assays were correlated. Finally, as shown below, the two models show similar changes in expression of specific genes. Although, we cannot exclude the possibility that the brain pathology due to expression of the alternative Ab transgenes is different, this evidence suggests that they have similar effects.
We confirmed the up-regulation of Sod3 using qRT-PCR. Sod3 has 4 alternative transcripts. Increases in Sod3 mRNA levels in the microarray experiment (neuronal expression of two copies of Ab 42 ) were specific to two transcripts, Sod3-RD and Sod3-RE. qRT-PCR quantification of total Sod3 levels suggested a 50% increase in Sod3 transcription in the heads of Ab flies at day 20 (Fig. 4a). However, transcript-specific primers showed a much larger increase in Sod3-RD (,100-fold) and Sod3-RE levels (,20-fold) confirming the array data (Fig. 4b, 4c). In Abarc flies (expressing neuronal Ab 42 arc peptide), total Sod3 mRNA levels were actually decreased (Fig. 4d). However as in the 2 3 Ab 42 flies, both Sod3-RD and Sod3-RE mRNA levels were significantly increased in heads from Abarc flies (Fig. 4e, 4f). Thus, both Ab and Abarc flies show a specific increase in the levels of Sod3-RD and Sod3-RE transcripts.
As Sod3 transcript levels are also increased in Abarc flies, we tested whether reducing Sod3 levels using RNAi could ameliorate the Ab phenotype. Ubiquitous RNAi against Sod3 resulted in effective knock-down of total Sod3 expression levels in head mRNA (down to 10% of control) and substantial knock-down of Sod3-RD and Sod3-RE expression levels (down to 30-40% of control: Supplementary Fig. S3), confirming that indeed the Sod3-RNAi line is substantially knocking down Sod3 transcript levels.
Targeting Sod3 RNAi specifically to the nervous system resulted in a consistent (though not statistically significant) decrease in total Sod3 RNA levels in head RNA from both Sod3 RNAi control and Abarc flies (vs. non-RNAi controls, Fig. 4d-f). Since Sod3 levels were measured from whole head RNA extracts, which also include other cell types apart from neurons, our measurement of total head Sod3 mRNA is likely to be an underestimate of the actual knock-down in the nervous system.
Sod3 RNAi improved climbing ( Fig. 4g and Supplementary Fig.  S3) and survival in AD flies; median survival for Abarc Sod3-RNAi flies was 35 days vs. 31 days for control Abarc flies (Log-rank test, P , 0.0001: Fig. 4h and Supplementary Fig. S3), and this appeared to correlate with significantly decreased Sod3-RE mRNA levels (Fig. 4f). Ubiquitous knock-down of Sod3 has previously been reported to be detrimental for Drosophila lifespan 21 . However, we did not observe any significant effect of nervous system-specific Sod3-RNAi in control (i.e. no Abarc) flies ( Fig. 4g and 4i and Supplementary Fig. S3). Thus, ablating Sod3 up-regulation in Abarc flies ameliorates the Ab phenotype.
RNAi against CG14715 (which encodes a prolyl-isomerase thought to function as an ER chaperone 30,31 ) in the fly nervous system of Abarc flies resulted in a significant increase in survival (median survival for two Abarc CG14715 RNAi lines was 34 days vs. 31 days for Abarc, Log-rank test, P , 0.0001; Supplementary Fig. S5) but no improvement in climbing ability (compared to non-RNAi controls; Supplementary Fig. S6). By contrast, overexpression of CG14715, in Abarc flies, resulted in a significant increase in climbing ability in early adulthood but had no effect on survival (Supplementary Fig.  S5). In control (non-Abarc) flies, RNAi knock-down or overexpression of CG14715 resulted in either no effect or a slight deficit in climbing ability; neither treatment affected survival (Supplementary Fig. S5). Thus manipulating CG14715 expression can modify the Ab phenotype, but the effect is complex.
Overexpression or RNAi knock-down of sec31 in the fly nervous system resulted in decreased climbing ability and longevity in control (non-Abarc) and either no, or a negative, effect in Abarc flies. (Supplementary Fig. S6). The sec31 protein is an essential component of the COPII tracking complex 29 so it likely that sec31 levels are critical for normal cellular function. Thus it is unclear whether sec31 has any specific modifying effect on the Ab phenotype.
These results suggest that modifying expression of Sod3, PGRP-SC1b and CG14715 in Ab flies can suppress the locomotor and survival defects associated with toxic Ab 42 expression.

Discussion
In this study, we have used time-course transcriptomic analysis to identify 712 D. melanogaster genes that are differentially expressed between Ab-expressing and control flies. Our results suggest that Ab-expressing flies are more similar to young than to old control flies. Since Ab and control flies remained transcriptionally distinct as their mortality increased, we therefore conclude that the expression of Ab in this model does not equate to an increased rate of ageing.
Cluster analysis revealed that differentially expressed genes can be separated into those that change expression over time and those whose expression is constant. These results are consistent with the very aggressive nature of the model in which very high levels of Ab are expressed at all times. We suggest that any change at the molecular level that correlates with the phenotype in the Ab flies lies downstream of the transcriptome and that, by day 3, the Ab flies are already in an essentially ''terminal'' transcriptional state. In this state,  certain pathways such as the oxidative stress response pathway appear to be dysregulated and drive degeneration, which is expressed as both decreased locomotor activity and increased mortality.
We have further investigated some of the genes that were consistently over expressed in Ab flies and we identified two modifiers of Ab toxicity: Sod3 and PGRP-SC1b. In particular we found that increased levels of two Sod3 transcripts in Ab flies were not accompanied by a compensatory increase in expression of either catalase or peroxidases. An imbalance in the relative levels of these three types of enzymes may result in an increased level of toxic H 2 O 2 in Ab flies, contributing to disease pathology. The expression of an RNAi for Sod3 resulted in a reduction in mRNA levels for at least one Sod3 transcript, and was accompanied by improved locomotor ability and survival in Ab, but not control, flies. This suggests that decreasing Sod3 enzyme levels in our Ab flies alleviated a toxic H 2 O 2 overload. Rival et al 10 , found similar results regarding the toxicity of H 2 O 2 . In particular, they observed increased survival of Ab flies when a dominant negative mutant Sod1 was expressed and reduced survival when wild-type Sod1 was expressed. By contrast, they found that the median survival was increased when Cat (encoding catalase) was overexpressed, suggesting that overproduction of H 2 O 2 by the Sod1 enzyme can overwhelm catalase resulting in toxicity and a decrease in lifespan of the Ab flies. Moreover, a previous study 34 in C. elegans showed that loss of the sod-4 gene, (the C. elegans ortholog of Sod3) had no effect on lifespan in wild-type worms, but increased the survival of daf-2 (insulin receptor) mutants. Doonan and colleagues 34 suggested that the SOD-4 enzyme may be generating H 2 O 2 , which acts as a signalling molecule and activates IIS (insulin/IGF-like signalling) by inactivating redox-sensitive phosphatases 35 ; consequently, its loss in IIS mutants would enhance their long-lived phenotype. Increased H 2 O 2 in Ab flies could also act as a signal. Thus, by reducing Sod3 enzyme levels, we would decrease both the toxic overload and affect the signalling role of H 2 O 2 to enhance lifespan. There have also been a number of reports on the activation of autophagy by H 2 O 2 through the PI3K/Beclin1 and the PI3K/Akt/ mTOR pathways 36 . In the case of PGRP-SC1b, it is not clear why RNAi results in an increase in lifespan and improved locomotor abilities in AD flies. We speculate that it could be due to a dysregulation of the pathway, similar to what we observed for Sod3.
The comparative analysis of gene expression between AD and ageing revealed changes at the single gene level, rather than the pathway level. Nevertheless, if the individual genes that are regulated very differently in Ab flies vs. wild-type controls are considered, a number of important inferences may be made concerning AD. One particular example is the oxidative stress pathway, which is up-regulated with age in control flies (in our experiment and Landis et al) 14 .
In the wild-type fly, manipulating cellular antioxidant defences (using transgenes) is not necessarily beneficial or detrimental to the health of the organism 18 . In other words, physiological levels of ROS can be dealt with by the insect's powerful enzymatic and nonenzymatic detoxification routes. However, mitochondrial dysfunction is observed in AD and this may be exacerbated with age 37 ; therefore it is possible that ROS generated as a consequence of this mitochondrial dysfunction overwhelms cellular detoxification pathways.
It is becoming clear that the cascade of events that originates from the aggregation of Ab and tau involves major stress response pathways, and all these stress pathways appear to be inter-related. However, their co-regulation and inter-relationships have been poorly characterised to date. In this study, we observed the dysregulation of a number of genes that belong to pathways that appear to be related, even though we were not able to assess their precise regulatory relationship. In all, this study has allowed us to investigate the processes that change in Ab flies and dissect these from the processes that change with normal ageing. We observed a large number of dysregulated processes in AD flies. In particular, we highlight a number of genes involved in redox stress, innate immune response and pathogen defense response and intracellular transport. We have shown that either knocking-down or over-expressing some of these genes increased lifespan and improved locomotor (climbing) ability in Ab flies compared to control flies. This suggests that the processes of oxidative stress and the immune response are likely to play an important role in the disease. The insights into time/age-dependent gene expression levels in AD that have been gained using an insect model may prove valuable in the design of strategies to combat this economically and socially important disease.

Methods
Fly stocks and maintenance. The UAS-Ab 42 -51D and UAS-Ab 42 arc-51D flies were generated using the PhiC31 method as previously described 38 . ElavGAL4 C155 and tubGAL4 were used for neuronal-specific expression and ubiquitous expression of transgenes respectively. For microarray experiments, w 1118 , elavGAL4 . UAS-Ab 42 -51D (32 copies) flies and controls w 1118 , elavGAL4/1; 51D (32 copies, empty insertion site) were used. For RNAi experiments, we replaced UAS-Ab 42 -51D (32 copies) flies with flies carrying a single UAS-Ab 42 arc-51D transgene. The presence of one, as opposed to two, transgenes in this line facilitated the use of the other transgenes required in these experiments. Ab 42 carrying the E22G (arctic) mutation was used, since a single copy of the wildtype Ab 42 transgene was found not to result in a significant locomotor or survival defects in our experiments. The phenotypes caused by expression of the wild-type Ab transgene and the arctic Ab transgene are quantitatively and qualitatively similar, as can be seen from the relative effects of each transgene on locomotor performance and survival ( Supplementary Fig. S1). Median survival for the arctic model is 31 days in the Ab flies compared to 24 days in the 2 3  Over-expressed in Ab flies. Fig. 3a Improved climbing and survival in Abarc flies. No effect on survival of control flies, slight increase in climbing performance. Figure 4

PGRP-SC1b
Innate immune response 27 Over-expressed in Ab flies. Fig. 3e Improved climbing and survival in Abarc flies. No effect in control flies. Figure 5  sec31 ER to Golgi transport 29 Under-expressed in Ab flies. Fig. 3f Increased survival of Abarc flies, no effect on climbing ability. No effect on survival or climbing performance in control flies. Supplementary Figure S6  CG14715 Intracellular transport and protein folding 30,31 Over-expressed in Ab flies, Fig. 3d Negative effect on climbing and survival in both control and Abarc flies. Supplementary Figure S5 www. Analysis of microarray data. The raw data were filtered to remove any probes that were rejected in over 50% of samples and were quantile-normalised across all arrays using Limma 39 . One array (hybridised to a time point 3 sample) was removed from further analysis at this stage. Any missing values were imputed using the impute package (R package version 1.32.0. http://CRAN.R-project.org/package5impute). Differentially expressed genes were identified using two methods. Firstly, Limma was used to fit a linear model to the entire time course and genes identified as significantly differentially expressed were those with an F statistic p-value , 0.05 following FDR correction. Secondly, the maSigPro 40 package was used to identify genes with significantly (p , 0.05 after FDR correction) different changes in expression over time. For both methods, the samples were matched by % survival. The results from maSigPro and Limma analysis were combined and the expression of each significant gene averaged over all replicates and standardised (to have a mean of 0 and standard deviation of 1). These data, matched by % survival, were then clustered using the R package Mfuzz 13 , which implements fuzzy c-means clustering 12 . Two clustering parameters are required; the fuzzifier m and the number of clusters c. The appropriate value of m was determined using the Mfuzz function ''m.estimate'', c was determined by examining the effect of c on the minimum centroid distance 13,41 . maSigPro 40 was used to identify genes changing expression over time in Ab and control flies separately (p , 0.05 after FDR correction). Each of these genes was tested for correlation with % survival using a Pearson Product Moment Correlation Coefficient in R.
Lifespan. Flies were reared at standard density, allowed to mate for 24 h, sorted by sex, and then transferred to experimental vials at a density of ten female flies per vial. Flies were transferred to fresh vials three times a week, and deaths were scored three to five times a week. Lifespan data were subjected to survival analysis (Log-rank tests) using GraphPad Prism 5 Software (GraphPad Software, Inc).
Locomotor/climbing assays. The locomotor ability of the flies was assessed in a 1 min negative geotaxis assay as previously described 9 . Ten flies were placed in a plastic 25-ml pipette and knocked to the bottom of the pipette. The number reaching the 10 ml line of the pipette (n top ) and the number remaining at the bottom (below the 2-ml line) (n bottom ), after 1 min, were measured. The performance (mobility) index was then calculated as (n top 2 n bottom 1 n total )/2n total . Three to four replicates were used per genotype. Climbing in each pipette was assessed three times and the average performance index for each pipette calculated. Assays were carried out in a well-lit room at a temperature of 23-24uC. The mean of the independent biological replicates for each genotype was plotted with the s.e.m.. Two-tailed Student's t-tests (f-test for equal variance) were used to identify significant differences at specific time points.
qRT-PCR. Total RNA was extracted from 20 adult heads per genotype using standard Trizol (Invitrogen, Paisley, UK) protocols. RNA was DNase-treated (Fermentas, Thermo Scientific, UK) and cDNA was prepared using oligo-d(T) primers and a Promega Reverse transcription kit (#A3500) according to the manufacturer's protocol (Promega, Southampton, UK). qRT-PCR was performed using a Biorad iQ machine and KAPA (KK4608) SYBR green PCR master mix (Biorad, Hemel Hempstead, UK). Relative quantities of transcripts were determined using the relative DDCt method and normalised to Act5C. Two to five independent RNA extractions were used for each genotype. Primer sequences are available upon request.