The carboxypeptidase D homolog silver regulates memory formation via insulin pathway in Drosophila

What is the molecular basis of memory formation? Many genes have been implicated in this process, including those involved in neural cell adhesion, mRNA transport, translation control, and cAMP-PKA signaling. Drosophila, with easy accessibility to genetic, molecular, and behavioral analyses, was also employed in olfactory learning studies and many key genes underlying memory formation were identified in these studies (Bellen et al., 2010). The majority of identified genes have been shown to be intensively expressed in the mushroom body (MB) of the Drosophila brain (Davis, 2005), including rut (adenylyl cyclase), DCO (the catalytic subunit of PKA), and AKAP Yu (Heisenberg, 2003; Davis, 2005; Lu et al., 2007). However, these genes are far from enough to understand memory formation comprehensively. In order to identify more genes involved in memory formation, we previously generated 2,667 enhancer trap mutants, each of which contains a p-element (P{GawB}) insertion (Liu et al., 2008). The insertion includes a Gal4 sequence that can be used to label the expression pattern of the disrupted gene and 368mutants were selected because their perturbed-gene expression enriched in the mushroom body. These selected mutants were then screened for 3 h memory performance in a well-defined olfactory conditioning paradigm, and one strain (No. 1021) was found defective. Plasmid rescue showed that the p-element of this mutant located in the first intron of the gene silver (svr) (Fig. 1A). Western blot analysis revealed that protein expression level of Svr was significantly decreased in this mutant (svr, Fig. 1B). Svr has been reported to involve in viability and behaviors such as cold and ethanol sensitivity, as well as long-term memory in courtship behavior (Sidyelyeva et al., 2010). To explore the function of Svr in olfactory memory formation, we tested the performance index of svr in different time points after training. Memory of the svr mutant (svr) exhibited a significant impairment at both 3 min and 3 h after one-cycle training (Fig. 1C). And 24 h memory of the mutant was also disrupted in svrmutant after spaced training (Fig. 1C). The memory defect was unlikely to be caused by deficiency in sensorimotor system since no abnormality in shock or odor avoidance was observed (Table S1). Finally, in order to confirm that the 3 min memory defect is not a result of random background mutation, we conducted a genetic complementation experiment, using svr and another independent P-element insertion allele, svr. Indeed, although both heterozygotemutants (svr/+ and svr/+) had normal 3 min memory, such memory was impaired in doubleheterozygote mutant (svr/svr) (Fig. 1D). One previous study showed that svr is involved in Drosophila development (Sidyelyeva et al., 2010). To figure out whether the disrupted memory formation of silver mutant resulted from the abnormal development or the interference of physiological process of neural system, we acutely manipulated the expression of svr transgene in adult flies with the TARGET system. In this system, the Gal4-induced expression is suppressed by a ubiquitously expressed Gal80 protein at the permissive temperature (18°C), but not at the restrictive temperature (30°C). The gene svr contains three carboxypeptidase domains, and has five endogenous transcriptional forms as a result of alternative splicing (Sidyelyeva et al., 2010). Previous findings suggested that the functions of CPD domain 1 and 2 are largely redundant, and the inactive CPD domain 3 is required for fully rescuing the mutant’s phenotype (Sidyelyeva et al., 2010). As a result, we used a longer form containing all three CPD domains (UAS::svr1B-2-3-t1 construct) to restore memory (Sidyelyeva et al., 2010). Acute expression of svr transgene in svr-labeled-neurons of svr mutant flies (svr/Y; UAS-svr/+; Gal80/+) effectively rescued the perturbed memory in svr mutant to a comparable level in the control group (UAS-svr/ +; Gal80/+) (left panel, Fig. 1E). On the other hand, no significant difference between svr acute expression group (svr/Y; UAS-svr/+; Gal80/+) and svr mutant group (svr/Y) was detected in un-induced conditions (right panel, Fig. 1E). These results suggest that Svr interferes the physiological process of memory formation. Next, inorder tofindout the functioningneural circuit ofSvr in fly brain, we visualized the Gal4 expression pattern of svr using GFP labeling. Confocal imaging of the GFP signal in svr/+; UAS-mCD8::GFP/+ revealed a preferential expression in two major compartments: the MB and a small group of


Isolation of memory-defective mutants
Transposon mutagenesis was performed as described previously (Dubnau et al., 2003;Liu et al., 2008). 2,667 homozygous strains were subjected to expression pattern analysis. 368 lines specifically expressed in the MB were considered as candidates for screening 3hr memory defect after one-cycle training. Based on molecular, genetic and behavioral characteristics, we identified a svr mutant line，svr 1021 ，which had memory defect.

Fly stocks
All flies were reared at 25°C and 60% humidity on standard cornmeal medium with 12/12 hr light/dark cycle. The svr KG02090 and UAS-InR del (expressed a constitutively active InR with most of its α subunit deleted) lines were obtained from Bloomington Stock Center (Bloomington, IN). UAS::svr1B-2-3-t1 (termed as UAS-svr) transgenic fly was kindly provided by Dr. Lloyd D. Fricker. dilp2-GAL4 was a gift from Dr. Ping Shen. The OK107, elav-Gal4 and tubulin-Gal80 ts (Gal80 ts ) lines were stocks in our lab. To eliminate genetic background differences, all strains used in behavioral experiments were outcrossed with w 1118 (isoCJ1) wild-type flies for at least five generations.

Plasmid rescue of P-element insertion
Plasmid rescue was carried out by digesting total genomic DNA from svr 1021 mutant with either EcoRI or XbaI, ligating the resulting restriction fragments and then transforming isolated ampicillin-resistance plasmids to E.coli. PCR. Sequencing of genomic DNA flanking and genomic blast confirmed the insertion site of P{GawB} transposon.

Pavlovian olfactory aversive conditioning
2 Training and test were performed at 25°C and 70% relative humidity as described previously (Tully et al., 1994). During one-cycle training, around 100 flies were sequentially exposed to two aversive odors, 3-octanol (OCT) (Fluka, 1.5×10 -3 diluted in heavy mineral oil) and 4-methylcycloohexanol (MCH) (Fluka, 1×10 -3 diluted in heavy mineral oil) for 60s with 45s flush of fresh air after each odor. Flies received electric foot shock (twelve 1.5s pulse of 60V) during the presence of the first odor (CS+) but not the second (CS-). In spaced training, flies were exposed to ten training sessions with 15 min rest intervals.
During test, trained flies were allowed to choose between CS+ and CS-in a T-maze for 120s. A performance index (PI) was calculated from the distribution of flies in the two T-maze arms. It was the result of the difference between fly numbers in the two T-maze arms divided by the total fly number of the two arms. To eliminate odor bias, each experiment (n = 1) consisted of two reciprocal groups, with one trained to associate OCT with shock and the other to associate MCH with shock. The final PI was the average of PIs from the two groups.

Sensorimotor responses
To test the olfactory acuity, odor-avoidance responses were quantified by exposing groups of about 100 untrained flies to the test odor (either OCT or MCH, 1.5×10-3 and 1×10-3 dilution, respectively) versus fresh air in the T-maze. To test the shock reactivity, groups of 100 untrained flies were exposed to two T-maze arms with 60V electric foot shock delivered to one of the arms but not the other. In both tests, flies were allowed to make a choice between the two arms for 120s and PI was calculated as described in the memory test.

Heat shock regimen
Flies carrying the tubulin-Gal80 ts were raised and maintained in 18°C to minimize any potential "leaky" expression during development. 2-5 days old progenies were collected and divided into two groups. The induced group was transferred to a 30°C incubator for 3 days, whereas the uninduced control group was kept at 18°C. Both 3 groups were allowed to recover at 25°C for at least 1hr before behavioral experiments.

Generation of antibody and western blot
The anti-Svr antibody was generated by injecting a purified polypeptide fragment (DRLEQSHVHQLR) into rabbits. Subsequently the antiserum was affinity-purified by Protein A spin chromatography kit (Pierce, Rockford, IL). For Western blot, approximately five adult brains of each genotype were collected in a microcentrifuge tube. Each brain sample was homogenized with 25 µl working loading buffer in an ice bath using a pellet pestle. Immediately after heating the homogenate for 10 min at 70°C, the debris was spun down and 20 µl supernatant was loaded into 3-8% TA gel (Invitrogen) and blotted to a nitrocellulose membrane (Millipore). The blot was probed with primary antibodies against Svr (1:200) and Actin (1:4,000) (Zhongshan Gold Bridge Biotech) overnight at 4 °C and then with HRP-conjugated goat anti-rabbit IgG (Zhongshan Gold Bridge Biotech) for 2 h at room temperature.
Finally, the exposed films were scanned and analyzed by Image J software (NIH).

Confocal imaging
5 days old adult brains were dissected in PBS, transferred to fixative buffer (4% paraformaldehyde in PBS), and vacuumed for 30 min at room temperature. For GFP expression detection, the fixed and vacuumed brain was mounted in FocusClear (Pacgen Biopharmaceuticals). The brains were imaged on a Zeiss 710 laser scanning confocal microscope. Image stacks was obtained with Imaris 7.0 software (Bitplane, Zurich, Switzerland).

Statistics
All data were presented as means ± SEM and analyzed by Student's t tests or ANOVA followed by Sidak's post hoc test in Graphpad Prism (Graphpad Software, Inc.).
Asterisks indicate statistically critical value (*p < 0.05; **p < 0.01; ***p < 0.001); 'n.s.' denotes no significant difference (p > 0.05). a Flies were challenged with 60V shock versus no shock, or an odor versus fresh air and were required to make a binary choice. b No significant difference was detected between svr mutants and control group (p>0.526, n ≥ 6, ANOVA).

Figure S1. The memory formation was impaired in the svr-GAL4 mutant.
Memory retention at 3 minutes after one-cycle training were significantly decreased in svr-GAL4 mutant svr NP3600 (t-test, p=0.0008 compared to w 1118 ). n=10-12.