Distinct roles of the RasGAP family proteins in C. elegans associative learning and memory

The Ras GTPase activating proteins (RasGAPs) are regulators of the conserved Ras/MAPK pathway. Various roles of some of the RasGAPs in learning and memory have been reported in different model systems, yet, there is no comprehensive study to characterize all gap genes in any organism. Here, using reverse genetics and neurobehavioural tests, we studied the role of all known genes of the rasgap family in C. elegans in associative learning and memory. We demonstrated that their proteins are implicated in different parts of the learning and memory processes. We show that gap-1 contribute redundantly with gap-3 to the chemosensation of volatile compounds, gap-1 plays a major role in associative learning, while gap-2 and gap-3 are predominantly required for short- and long-term associative memory. Our results also suggest that the C. elegans Ras orthologue let-60 is involved in multiple processes during learning and memory. Thus, we show that the different classes of RasGAP proteins are all involved in cognitive function and their complex interplay ensures the proper formation and storage of novel information in C. elegans.


Results
C. elegans RasGAPs are dispensable for chemotaxis to olfactory cues. Aversive olfactory associative learning and memory rely on chemotaxis because environmental cues are acquired by olfaction. To test the role of the gap genes in these processes, we first analysed the chemotaxis towards three chemoattractants (diacetyl, benzaldehyde, and isoamyl-alcohol) of single gap mutant worms or of double mutant animals carrying mutations in gap genes in different combination. Chemotaxis assays were conducted with chemicals at two different concentrations in order to be able to identify even mild sensory deficits of the mutant strains. The volatile compounds used in the assay represent bacterial metabolic molecules that can be nutriments, therefore, induce attraction 11 .
In summary, strains carrying rasgap mutations all respond to chemotactile stimuli. GAP-1 is involved in the chemosensation of isoamyl-alcohol and high concentration of diacetyl, while GAP-3 plays a role in the olfaction of high concentration of diacetyl only. For both single mutants we observed a statistically significant but only marginal decrease in the odor response. Futhermore, none of the mutations caused significant changes in chemosensation to low concentrations of diacetyl, and therefore these do not interfere with negative olfactory associative learning and memory tests. The only exception was the gap-1(ga133);gap-3(ga139) mutant, which had a chemosensory defect to both concentrations of all tested attractants.
We first tested the motility of well-fed worms on bacterial lawn to determine the baseline activity. Next we tested the food seeking behaviour by placing well-fed worms on empty plates. Finally, we investigated the starvation response of the different mutants by placing 1-hour-long starved worms on food source. No difference was observed in the mean body bends for fed ( Fig. 2A) and starved (Fig. 2C) animals having gap mutations compared to the N2 reference strain. Food seeking animals showed similar behaviour except the gap-1(ga133);gap-3(ga139) mutant, which showed decreased motility (p = 2.54 × 10 −12 , Fig. 2B). A possible explanation of this difference is the deteriorated food searching motivation rooting in the chemosensory defect described above, because no locomotor defect was found in the 'Fed' and 'Empty' conditions.

Complex regulation of learning and short-term associative memory by RasGAP proteins.
In the previous experiments we tested all gap single and double mutants for defects that would interfere with negative olfactory associative learning. Since all gap single mutants and all gap double mutant worms (except the gap-1;gap-3 mutants) showed no or minor defect in chemotaxis or motility, next the role of different RasGAPs during negative olfactory learning and short-term memory (STAM) was investigated 12 . Naïve worms are attracted to diacetyl, which can be turned into aversion during a 1-hour conditioning period that associates starvation with the previously attractive diacetyl. This association deteriorates over time, therefore the effectiveness of conditioning itself informs about learning, while the amount of deterioration characterizes the memory function.
First, we tested all three single mutants and compared chemotaxis to diacetyl in naïve or conditioned worms. The combination of a 1-h starvation period in the presence of diacetyl (conditioning stimulus) dramatically reduced the attraction towards the chemoattractant in wild-type animals and in gap-2 and gap-3 single mutants to similar extent, but we found acquisition defects in the gap-1(ga133) (p = 0.00139) worms ( Fig. 3 and Table 1). gap-1 mutant worms also showed a significant memory defect when compared to wild types (p = 9.7 × 10 −6 ), however, this may be due to the learning defect. To address that, we analyzed the rate of memory loss by comparing the difference between conditioned and 30 min delay attraction (also called recovery phase) to diacetyl in wild type and gap-1 mutant worms and we found a statistically not significant difference between the genotypes. Thus, the memory difference was likely a consequence of the acquisition defect and this suggests that gap-1 is playing a role in the learning process. In the case of the other two gap mutants, we observed a defect in both gap-2 and gap-3 only in memory ( Fig. 3) and this suggests that gap-2 and gap-3 are both regulating memory consolidation.
In order to exclude possible background mutations that could contribute to the observed phenotypes, we performed RNAi silencing of the different rasgaps in RNAi hypersensitive eri-1(mg366);lin-15B(n744) strain. As control, eri-1(mg366);lin-15B(n744) was fed with bacteria expressing GFP dsRNA (Fig. 4). Double stranded RNA targeting gap-2 and gap-3 mRNA caused a phenotype matching the gap-2 and gap-3 loss-of-function mutants, respectively (Fig. 4). In case of gap-1 RNAi we did not observe this effect, probably due to insufficient RNAi silencing. Thus, we performed a rescue experiment by reintroducing the gap-1 genomic fragment in gap-1(ga133) mutant worms and negative STAM was assessed in three parallel lines (see Materials and Methods for details). Genomic fragment encompassing the gap-1 locus restored learning and memory to levels comparable to wild type in all three lines (Fig. 5). Thus, the observed phenotypes are attributable to the loss-of-function of various rasgap genes. Interplay of RasGAP proteins during long-term associative memory (LTAM). Beside the role of RasGAPs in short-term memory, we also tested the impact of the different C. elegans rasgaps in long-term associative memory. LTAM was assessed by testing worms 16 and 24 hours after negative olfactory conditioning, as described previously 10 . In this experimental setup we used a reinforcement training consisting of three rounds of conditioning cycles. Briefly, starvation and diacetyl were associated for 1 hour, which turned the attraction towards diacetyl into aversion. Each conditioning step was followed by a feeding phase without diacetyl for 30 minutes to let the worms regenerate. Such rounds of conditioning and feeding were repeated 3 times. Using this training we could not detect a significant difference in acquisition in most gap mutant genotypes. A possible explanation of this phenomenon is that the naïve animals were conditioned 3 times in these assays compared to the single conditioning in assays testing STAM. Interestingly, the learning defect was still present in the gap-2(tm748);gap-3(ga139) double mutant strain (p = 1.24 × 10 −7 ).
prevented any assessment of learning and memory formation. Altogether, the rasgap genes are redundantly regulating long-term memory in nematodes.

let-60 is involved in the gap-related learning and memory phenotypes. RasGAPs are known
to increase the intrinsic GTPase activity of LET-60 and by that they regulate different biological processes in C. elegans 13 . Based on these earlier observations we hypothesized that gap mutations combined with a reduction of function let-60 mutation would result in a learning and memory phenotype similar to the wild type. The let-60 mutation itself causes a strong chemosensory defect 14 , which was in part restored in all gap;let-60 double mutant. Furthermore, in agreement with our assumptions, we found that the let-60(n2021) hypomorph mutation restored the learning and memory defects observed in the rasgap mutants (Fig. 7). This suggests that all three rasgaps act at least in part through let-60 to regulate learning and memory in C. elegans.

Discussion
Here, we investigated the role of each member of the RasGAP family in the learning and memory process in C. elegans. We found that all RasGAP forms are regulating cognitive functions, and strikingly, they exhibit specific roles during learning and memory. While the exact molecular machinery connecting RasGAPs to learning and memory formation in C. elegans is unclear yet, our results show that RasGAPs are likely involved in multiple distinct processes during acquisition and storage of new informations. In C. elegans, synaptic cytoskeletal reorganization is widely accepted as a form of synaptic plasticity, and was suggested recently as a molecular process of forgetting 15 . Fast activation and inactivation of Ras is important for the olfactory behaviour 16 , although the negative feedback loop for inactivation has not been elucidated yet.
In humans, analysis of molecular signaling networks revealed potential cross-talks between the Ras/ MAPK pathway and the possible cytoskeletal rearrangments 17 , which is further supported by the earlier finding that RasGAPs control Rho-mediated cytoskeletal reorganization 18 . In rat hippocampal neurons, synaptic Ras GTPase activating protein (SynGAP), an orthologue of C. elegans GAP-2, can be phosphorylated by the calcium/calmodulin-dependent protein kinase II (CaMKII) upon long-term potentiation induction. This leads to the dispersion of SynGAP from the dendritic spines 19 . Ras family proteins, the major effectors downstream of GAPs, have also been associated with Ca 2 + -dependent synaptic crosstalk after NMDA receptor activation in rat hippocampal slices 20 . Neurofibromin 1 (NF1), a GAP orthologue not found in C. elegans yet, can also inactivate Ras in rat hippocampal dendritic spines 21 . In mouse model systems, homozygous knock-out of syngap leads to postnatal lethality, while heterozygous mice exhibit specific defects in hippocampal long-term potentiation and glutamate receptor trafficking, although the molecular relation between these two processes is difficult to assess at phenotypic level 22 . In a mouse schizophrenia model, reduced expression of SynGAP leads to non-habituating mice showing persistent hyperactivity, lack of social memory, impaired fear conditioning and working memory, probably due to defected interaction between SynGAP and NMDA receptor 23 . Neurofibromin 1 also regulates GABA release, long-term potentiation and learning in mice 24 . Furthermore, in zebrafish, loss of Neurofibromin 1 results in learning and memory defects 25 . Altogether, these findings suggest a) a conserved role for different RasGAPs in learning and memory, b) a significant role for SynGAP in signal transmission during learning and memory formation. In addition to the NF1 and SynGAPs we show here that RASAL and p120 RasGAP subfamily members are also regulating learning and memory and that different RasGAP subclasses are redundantly modulating different parts of the acquisition and storage processes.
Furthermore, our results are in good agreement with clinical findings. Rasopathies are human pathological conditions associated with germline mutations of the Ras/MAPK pathway 26,27 . The overlapping phenotypic features are developmental and cutaneous abnormalities, predisposition to malignancies and varying degree of neurocognitive impairment including learning disability. Rasopathies involving RasGAPs are neurofibromatosis type 1 (NF1, also known as von Recklinghausen disease) 28,29 and capillary malformation -arteriovenous malformation syndrome (CM-AVM) 30 , caused by germline mutations in the nf1 and rasa1 genes, respectively. The exact molecular mechanisms underneath the symptoms are poorly understood yet.
Scientific RepoRts | 5:15084 | DOi: 10.1038/srep15084 are in accordance with the clinical features of Rasopathies, further supporting that the molecular mechanisms of learning, memory and the role of the Ras/MAPK pathway in these are well conserved. This work also opens perspectives for neurocognitive and neurobehavioural studies of RasGAPs in C. elegans.
For the gap-1 rescue, the fosmid clone WRM0629aG09 was digested with AvrII/SbfI restriction endonucleases and the 9.5 kb fragment encompassing the gap-1 gene was microinjected at a concentration of 100 ng/μ l into both arms of the syncytial gonads of gap-1(ga133) worms as described earlier 32 . Sur-5::dsRed at 10 ng/μ l concentration was coinjected as transformation marker.
Motility assay. Motility was characterized by the number of body bends per minute as described earlier 36,37 . Briefly, well fed single young adult worms were transferred onto seeded NGM plates (baseline activity), onto empty plates (food searching activity) or animals were starved for 1 hour and transferred to seeded NGM plates (feeding activity). Body bends were counted for 1 minute after a 2-minute resting phase. At least 20 animals were recorded per condition and per genotype.
C. elegans behavior assays. Chemotaxis to different compounds was assessed as described previously 38 . Briefly, worms were washed three times in CTX buffer (5 mM KH 2 PO 4 /K 2 HPO 4 pH 6.0, 1 mM CaCl 2 , 1 mM MgSO 4 ) and were allowed to settle down by gravity. 50 to 200 worms were placed to the middle of 10 cm CTX test plates (5 mM KH 2 PO 4 /K 2 HPO 4 pH 6.0, 1 mM CaCl 2 , 1 mM MgSO 4 , 2% agar). Worms were given a choice between a spot of attractant (diacetyl, benzaldehyde and isoamylalcohol) in ethanol at the indicated dilutions with 20 mM sodium-azide and a counter spot with ethanol and sodium-azide. The distribution of the worms over the plate was determined after 1 hour and the chemotaxis index was calculated as described earlier 38 . Figure 7. let-60 is required for the gap(lf) learning and memory phenotypes . Naïve gap(lf);let-60(n2021hf) double mutants (A) were conditioned to assess learning (B) and short term associative memory (C). Naïve mutants are characterized by lowered chemotaxis index, e.g. chemosensory defect due to the let-60(n2021hf) mutation. Conditioned worms have no significant learning defect (B) and recovery phase has not revealed significant memory defect either (C). Both conditioned and recovery phases were assessed by calculating learning indices (LI = [CI conditioned -CI naïve ]/CI naïve ) to ensure comparability with the N2 wild type.
Scientific RepoRts | 5:15084 | DOi: 10.1038/srep15084 Negative olfactory associative conditioning was performed with modifications of the original protocol 12 . For the conditioning, 1 hour long starvation was coupled with 2 μ l of undiluted diacetyl dropped on the lid of 10 cm CTX plates. A subpopulation of worms was washed for half an hour in CTX buffer after conditioning to allow recovery. Chemotaxis of naïve, conditioned and recovery worms were tested with diluted diacetyl (1:1000) as described above.
Long-term associative memory assays were performed as described earlier 10 . Briefly, worms were conditioned three times by repeating the cycles of conditioning described above and by allowing the worms to regenerate for 30 mins in presence of food after each conditioning. Memory function was assessed after 16 and 24 hours as described for the negative olfactory associative conditioning.
Learning index was calculated as the difference of chemotaxis indices of conditioned and naïve worms normalized by the chemotaxis index of naïve animals: Welch's test 39 , as implemented in the statistical module of SciPy 0.13.3, was used to calculate statistical significances. P-values always refer to results of two-tailed tests, in multiple comparisons p-values are always Bonferroni corrected; *p ≤ 0.05, **p < 0.01, ***p < 0.001. Two-way ANOVA with Bonferroni corrected post-hoc t-test was used for interaction analysis of learning and memory assays, these tests were performed with the R project (http://www.r-project.org/). Error bars represent standard deviation. All computational tools are open source and were designed and implemented by experts to follow the best practices and to ensure scientific reproducibility.