Abstract
Nonspecific cognitive impairments are one of the many manifestations of neurofibromatosis type 1 (NF1). A learning phenotype is also present in Drosophila melanogaster that lack a functional neurofibromin gene (nf1). Multiple studies have indicated that Nf1-dependent learning in Drosophila involves the cAMP pathway, including the demonstration of a genetic interaction between Nf1 and the rutabaga-encoded adenylyl cyclase (Rut-AC). Olfactory classical conditioning experiments have previously demonstrated a requirement for Rut-AC activity and downstream cAMP pathway signaling in neurons of the mushroom bodies. However, Nf1 expression in adult mushroom body neurons has not been observed. Here, we address this discrepancy by demonstrating (1) that Rut-AC is required for the acquisition and stability of olfactory memories, whereas Nf1 is only required for acquisition, (2) that expression of nf1 RNA can be detected in the cell bodies of mushroom body neurons, and (3) that expression of an nf1 transgene only in the α/β subset of mushroom body neurons is sufficient to restore both protein synthesis-independent and protein synthesis-dependent memory. Our observations indicate that memory-related functions of Rut-AC are both Nf1-dependent and -independent, that Nf1 mediates the formation of two distinct memory components within a single neuron population, and that our understanding of Nf1 function in memory processes may be dissected from its role in other brain functions by specifically studying the α/β mushroom body neurons.
Introduction
Neurofibromatosis type 1 (NF1) is an autosomal, dominant genetic disorder that afflicts approximately 1 in every 3500 individuals. Like other clinical manifestations of NF1, expression and penetrance of cognitive phenotypes varies and may include deficiencies of visual–spatial processing, executive function, and attention (for review, see Ozonoff, 1999; North, 2000; Acosta et al., 2006). Homologs of human Nf1 in mouse and Drosophila melanogaster share significant identity at the protein level, and animal models in both species were developed shortly after the human Nf1 gene was cloned (Bernards et al., 1993; Jacks et al., 1994; The et al., 1997). Both models demonstrate cognitive phenotypes, and insights gained through animal studies have shed light on the genetic and biochemical basis of these defects.
Drosophila has been used extensively for expanding our basic understanding of memory, making it ideal for investigating NF1 cognitive deficits. After olfactory classical conditioning, Drosophila form protein synthesis-independent early memories (PSI-EM), comprising short-term memory (STM) tested at 3 min after training, middle-term memory often tested at 3 h after training, and protein synthesis-dependent long-term memory (PSD-LTM), tested at 24 h after conditioning. The nf1 mutant flies demonstrate deficiencies in PSI-EM and PSD-LTM (Guo et al., 2000; Ho et al., 2007). A current model postulates that Nf1 contributes to PSI-EM through stimulation of the rutabaga-encoded adenylyl cyclase (Rut-AC) (Guo et al., 2000). Stimulation of Gαs-dependent AC activity requires only the Nf1 C-terminal domain (Hannan et al., 2006). The PSI-EM phenotypes of nf1, rut-AC, and nf1/rut-AC mutants are similar (Guo et al., 2000), both genes are required at the time of learning (Guo et al., 2000; McGuire et al., 2003; Mao et al., 2004), and either ubiquitous expression of a constitutively active protein kinase A (hsPKA*) transgene (Guo et al., 2000) or neuronal expression of a Nf1 C-terminal domain transgene (Ho et al., 2007) rescues the nf1 phenotype. Furthermore, the current model also postulates that Nf1 contributes to PSD-LTM through regulation of Ras via its GAP-related domain (GRD) (Hannan et al., 2006; Ho et al., 2007). Stimulation of Ras-dependent AC activity is absent in nf1 mutants, but transgenic expression of the Nf1 GRD restores this activity (Hannan et al., 2006) and improves the PSD-LTM phenotype of nf1 mutants (Ho et al., 2007).
It is surprising that endogenous Nf1 expression has not been observed in adult mushroom body (MB) neurons (Walker et al., 2006). MB neurons are essential for olfactory memory formation (McGuire et al., 2003; Davis, 2005), and Rut-AC is preferentially expressed in these neurons (Han et al., 1992). Rescue experiments demonstrated that transgenic expression of rut-AC in α/β and γ MB neurons restores normal memory in homozygous mutants (Zars et al., 2000; McGuire et al., 2003; Mao et al., 2004; Akalal et al., 2006; Blum et al., 2009). If Nf1 indeed stimulates Rut-AC activity during learning, it is probably expressed, and required, in MB neurons.
Here, we explore whether Nf1 and Rut-AC are involved in the same operational phase of learning, whether they are expressed in the same neurons, whether both are required in the same neurons for rescue of PSI-EM, and whether the Ras-mediated function of Nf1 is required in overlapping neurons. We report a role for Rut-AC in memory stability that is Nf1 independent, observe nf1 expression in MB neurons, and demonstrate a requirement for nf1 expression in α/β MB neurons for both PSI-EM and PSD-LTM.
Materials and Methods
Fly culture and genetics.
Flies were reared on standard cornmeal medium, at 25°C, 60% relative humidity, and a 12 h light/dark cycle. For Gene-Switch experiments, an appropriate volume of RU486 (11β-[p-(dimethylamino)phenyl]-17β-hydroxy-17-(1-propynyl)estra-4,9-dien- 3-one) stock solution was mixed into molten standard medium at 65°C, along with food coloring, to a final concentration of 200 μm. Flies were reared and maintained on this altered food source throughout development and adulthood or were transferred from standard food to food containing RU486 at eclosion and were trained at 5 d after eclosion. A. Bernards (Massachusetts General Hospital, Boston, MA), A. Sehgal (University of Pennsylvania, Philadelphia, PA), and M. Stern (Rice University, Houston, TX) provided fly stocks. All stocks used in this study were outcrossed to our w(CS10) stock (w1118 flies outcrossed to Canton-S for 10 generations) for six generations, except for the rut2080 stock, which contains a P-element insertion bearing the rosy gene. As a result, rut2080 is maintained in a ry background. K33 flies contain a P-element in the E(spl) complex, which was mobilized to produce the nf1P2 allele. Subsequently, an imprecise excision of the nf1P2 P-element produced the nf1P1 allele (The et al., 1997). K33 flies do not show a behavioral phenotype relative to w(CS10) flies, so both of these lines were used interchangeably as controls (labeled as “control” in figures). PBac{PB}Nf1c00617 (nf1c00617) flies (Thibault et al., 2004) were obtained from the Bloomington Drosophila Stock Center (Bloomington, IN). We confirmed the insertion site of nf1c00617 via inverse PCR using primer sequences generated and made publicly available by Exelixis through the Bloomington Drosophila Stock Center. As reported on the FlyBase website, the stock nf1c00617 contains an insertion of the PBac{PB} transposon in the seventh intron of the nf1 genomic sequence, at position 21811820 of the D. melanogaster genome (R5.5).
Behavioral assays.
Olfactory learning and memory experiments were conducted using an olfactory classical conditioning paradigm (Beck et al., 2000). Standard training procedures were used for all 3 min and 3 h memory experiments. Two- to 7-d-old flies were exposed to a single training trial, in which they were sequentially presented with methylcyclohexanol (MCH) and benzaldehyde (BA) odors for 1 min each. During presentation of the first odor [conditioning stimulus (CS+)], flies were simultaneously exposed to 12–1.25 s pulses of 90 V electric shock. After a 30 s delay, the second odor (CS−) was presented in the absence of electric shock. Flies were then transferred to a T-maze and allowed to choose between the two trained odors, each contained within one arm of the maze. Avoidance of the CS+ during testing was calculated as a performance index (P.I.), defined as the fraction of flies preferring the CS− minus the fraction of flies preferring the CS+. A P.I. of 1.0 indicates that all flies avoided the CS+, and a P.I. of 0.0 indicates a 50:50 distribution between the T-maze arms and therefore no learning. To control for possible naive odor bias, each trial comprised two groups in which the first group was trained with MCH as the CS+, and the second group was trained with BA as the CS+. To control for visual distraction, all experiments were performed in a darkroom illuminated with dark red light. Environment within the darkroom was also controlled at 23–25°C and 65–75% humidity. For 24 h memory assay, flies were 1-to 4-d-old, and either a 5×-massed or 5×-spaced protocol was followed. During massed training, flies were presented with five training cycles, as described above, with a 30 s intertrial interval. During spaced training, flies were presented with five training cycles with a 15 min intertrial interval, to elicit PSD-LTM (Yu et al., 2006). For memory acquisition and stability assays, a modified training paradigm known as the short program was used (Beck et al., 2000). In this schedule, odor exposures were reduced to 10 s with a single 1.25 s electric shock presented at the eighth second and a 30 s delay between odor presentations. When multiple training trials were presented, there was a 30 s intertrial interval.
RNA in situ hybridization.
Probe template containing an 875 bp region, including exons 6 and 7, was amplified from NF1mini plasmid (A. Bernards) by PCR and cloned into pCRII–TOPO plasmid (Invitrogen). Digoxigenin (DIG)-labeled RNA probes were transcribed from linearized plasmid in the antisense orientation, using the DIG RNA Labeling kit (SP6/T7) (Roche). Both w(CS10) and nf1P1 fly heads were cryosectioned (15 μm) and fixed in 4% paraformaldehyde. Sections were denatured with 0.2N HCl, treated with Proteinase K, postfixed in 4% paraformaldehyde, and acetylated. After a 1 h prehybridization at 50°C, denatured DIG-labeled probes were hybridized to sections at 50°C for 16–24 h. Hybridization buffer contained 50% formamide, 5× SSC, 5× Denhardt's solution, 250 μg/ml yeast tRNA, 500 μg/ml salmon sperm DNA, 50 μg/ml heparin, 2.5 mm EDTA, 0.1% Tween 20, and 0.25% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate. After a number of washes to decrease salt concentration, slides were incubated with α-DIG–alkaline phosphate (Roche) and visualized using nitroblue-tetrazolium-chloride/5-bromo-4-chlor-indolyl-phosphate (Roche) solution. When staining was complete, slides were washed in PBST (PBS and Triton X-100) and mounted in Glycergel (Dako).
Results
nf1c00617 mutants exhibit size and cognitive phenotypes
We first explored whether the putative allele nf1c00617 conferred the small size and learning phenotypes typical of nf1P1 and nf1P2 mutants (The et al., 1997; Guo et al., 2000). To measure body size, female and male adult flies were measured from the anterior tip of the antennae to the posterior tip of the abdomen. As expected, both nf1P1 and nf1P2 males and females were significantly smaller than control males and females, respectively (The et al., 1997). In contrast, only nf1c00617 males were smaller than control males, but the nf1c00617 males were still significantly larger than nf1P1 and nf1P2 males. There was no significant size difference between nf1c00617 and control females (Fig. 1A).
nf1 mutants exhibit size and memory phenotypes. A, Female and male adults were measured from the anterior tip of the antennae to the posterior tip of the abdomen. There was no significant difference between nf1c00617 and control females, but nf1c00617 males were significantly smaller than control males (*p < 0.005). Both genders of nf1P2 and nf1P1 were significantly smaller than both controls and nf1c00617, and nf1P2 flies were also significantly smaller than nf1P1. For control, w(CS10) was used, n = 100, and for all nf1 mutants, n = 50. Means ± SEM are shown. *p < 0.005, **p < 0.0001. B, Learning and memory phenotype of nf1c00617 homozygous mutants compared with other nf1 alleles. Performance was assayed 3 min, 3 h, and 24 h after training. All mutant groups showed impaired performance relative to controls. nf1c00617 mutants performed significantly better than nf1P1, nf1P2, and rut2080 mutants when tested at 3 min after training but not when tested 3 or 24 h after training. A 5×-spaced protocol was used to elicit 24 h memory. For 3 min and 3 h memory, w(CS10) was used as a control. K33 was used as a control for 24 h memory. C, Expression of nf1 in all neurons rescues the 3 h memory phenotype of nf1c00617 mutants. Both control and elav/uas–dnf1;nf1c00617 flies performed significantly better than both elav/+;nf1c00617 and nf1c00617 flies, demonstrating a complete rescue of the nf1c00617 memory phenotype. There was no significant difference between the performance of control and elav/uas–dnf1;nf1c00617 flies or between elav/+;nf1c00617 and nf1c00617 flies. There was also no significant difference between the performance of nf1c00617 and uas–dnf1/+;nf1c00617 flies, as shown in Figures 5⇓–7. Control flies were w(CS10). For 24 h memory, n = 14. For all other data, n = 6. Means ± SEM are shown. *p < 0.05, **p < 0.001 relative to controls.
In addition to a mild size phenotype, nf1c00617 homozygous mutant flies also exhibited 3 min, 3 h, and 24 h memory phenotypes (Fig. 1B). nf1c00617 performance at all time points was significantly poorer than the control, but, when tested 3 min after training, these flies performed significantly better than nf1P1, nf1P2, and rut2080 homozygous mutants. We also sought to confirm that the cognitive phenotype of nf1c00617 mutants is attributable to a disruption of Nf1 function. We expressed a transgene containing full-length Drosophila nf1 (uas–dnf1) in the nf1c00617 homozygous mutant background using the elav–gal4 driver, which promotes gene expression in all neurons. Restoration of nf1 expression in all neurons fully rescued the 3 h memory phenotype of nf1c00617 homozygous mutants, confirming that the nf1c00617 insertion does indeed disrupt Nf1-dependent memory (Fig. 1C).
nf1 mutants are defective in memory acquisition but not memory stability
Although a requirement for neurofibromin during olfactory conditioning was established previously (Guo et al., 2000), former experiments did not address which operational phase of learning is impaired. Any deficit in olfactory memory may represent a failure to associate the odor and shock stimuli, an increased rate of memory decay, or failure of memory retrieval (Cheng et al., 2001). Memory acquisition and stability were assayed for nf1c00617, nf1P2, and rut2080 homozygous null mutants. Flies were presented 1–15 training trials with a 30 s intertrial interval, and performance was assayed immediately after the last training trial (Fig. 2A). Control performance improved with increasing number of training trials and reached a plateau at a ceiling level with as few as five training trials. After a single training trial, all mutants performed poorly relative to the control, but nf1c00617 flies did not continue to display a learning deficit when presented with additional training trials. In contrast, nf1P2 and rut2080 flies performed poorly until after 15 training trials, at which point their performance was not significantly different from the control. Our results suggest that both Nf1 and Rut-AC are required for the acquisition of olfactory memory.
Neurofibromin is required for memory acquisition, not memory stability. A, nf1 mutants are defective in the acquisition of memories. All groups were trained with 1–15 trials (T), and performance was assayed immediately after conditioning. With one training trial exposure, all mutants performed significantly poorer than controls. When exposed to 15 training trials, all mutants performed at the same level as controls. For all other experiments, both nf1P2 and rut2080 mutants performed significantly poorer than controls, whereas nf1c00617 mutants did not. Across experiments, the performance of nf1P2 and rut2080 mutants paralleled that of controls but was delayed. Control flies were w(CS10) for all experiments. For the five trial experiment, n = 9 for each group. For all other experiments, n = 6 for each group. Means ± SEM are shown. *p < 0.05. B, Memory stability is normal in nf1 mutants but defective in rut2080 mutants. Mutant performance at 3 min after training was normalized to control performance by varying the number of training trials, as indicated. Performance of nf1 homozygous mutants was not significantly different from control performance at the time points tested. Whereas rut2080 and control performances were indistinguishable 3 min after training, they were significantly different at both 30 and 60 min after training. Control flies were w(CS10) for all experiments. For 30 min decay of nf1P2 performance, n = 11. For all other experiments, n = 6 for each group. Means ± SEM are shown. *p < 0.005.
Immediate performance of each mutant was then normalized relative to the control so that memory decay could be compared (Fig. 2B). When nf1c00617 mutants and controls were both trained with three trials, immediate memory was identical. Likewise, memory tested at two subsequent time points was indistinguishable. Normalization of nf1P2 performance required training with seven trials when the control was trained with three trials. Memory of nf1P2 mutants, when tested at two subsequent time points, was also indistinguishable from performance of controls, suggesting that Nf1 is not required to maintain olfactory memories over time. Finally, the immediate performance of rut2080 and control flies was normalized by 10 and 3 training trials, respectively. As early as 30 min after training, the memory of rut2080 flies was abolished, suggesting that Rut-AC plays an important role in maintaining memories over time. Furthermore, our results suggest that the role of Rut-AC in memory stability is Nf1 independent.
nf1 is expressed broadly in the brain and in mushroom body neurons
Walker et al. (2006) reported widespread expression of Nf1 in the CNS of Drosophila larvae but a notable absence of expression in third-instar mushroom body neurons. Additionally, the authors also indicated an absence of Nf1 expression in the adult mushroom body neurons but expression elsewhere in the adult CNS. However, the putative association of Nf1 with Rut-AC in learning (Guo et al., 2000), the preferential expression of Rut-AC in mushroom body neurons (Han et al., 1992), and the clear evidence indicating that Rut-AC is required in adult mushroom body neurons for olfactory learning (Zars et al., 2000; McGuire et al., 2003; Mao et al., 2004; Akalal et al., 2006; Blum et al., 2009) necessitates Nf1 expression in adult mushroom body neurons.
We therefore probed nf1 gene expression by RNA in situ hybridization on w(CS10) brain sections. Figure 3 illustrates that nf1 is expressed in many, or all, cell body regions of the central brain. Expression of nf1 was apparent in cell body regions surrounding the antennal lobes, protocerebrum, lateral horn, and mushroom body calyces. Additional examination revealed nf1 RNA expression in many mushroom body cell bodies (Fig. 3E). Hybridization of the same probe with nf1P1 homozygous mutant brains did not result in staining, confirming the specificity of the antisense probe. Our results indicate that the nf1 gene is broadly expressed in the adult CNS, including the mushroom body neurons.
nf1 is endogenously expressed in mushroom body neurons. A–D, Series of frontal sections of w(CS10) adult fly heads arranged from anterior to posterior. A, Staining of an anterior section, at the level of the superior medial protocerebrum and antennal lobes. Intense staining was present in cell body regions, with no staining of the antennal nerve or neuropil. Scale bar, 100 μm. B, A more posterior section showing continued staining of cell body regions of the adult brain, with no staining of neuropil areas. C, Section is at the level of the central complex, showing staining of cell body regions, which surround the internal neuropil. D, Posterior section at the level of the calyx shows staining in cell body regions, including the cell body region of mushroom body neurons. E, Higher magnification of the region within the white box of D. Staining was visible in cells dorsal and lateral to the calyx neuropil, in the region known to contain cell bodies of the mushroom body neurons. Scale bar, 10 μm. F, In situ hybridization to nf1P1 homozygous mutant shows no staining in cell body regions. Representative image was taken at the same magnification and at approximately the same level as B.
Considering the robust expression of nf1 RNA in mushroom body cell bodies, it is surprising that attempts to visualize the Nf1 protein in adult mushroom body neurons, with the same antibodies used for larval immunohistochemistry, have not been successful (Walker et al., 2006) (our unpublished observations). However, three different Nf1 isoforms exist, which encode peptides of 2746 (Nf1-RC), 2764 (Nf1-RD), and 2802 (Nf1-RB) amino acids. It is not known which of these is expressed in the adult CNS. We have developed RNA probes against a sequence region that is common among all transcripts. Walker et al. (2006) used monoclonal antibodies (DNF1-21) generated against the C-terminal 450 amino acids of Nf1-RB (The et al., 1997). Only 195 of these residues are present in Nf1-RC, and these are followed by 199 resides that are not in common with Nf1-RB. Furthermore, only 411 of the 450 amino acids in Nf1-RB are present in Nf1-RD, and this is followed by one residue that is not in common with Nf1-RB. It seems probable that the DNF1–21 does not react with the isoform that is expressed in mushroom body neurons.
Many other genes, required for learning and memory, are preferentially expressed in mushroom body neurons (Nighorn et al., 1991; Han et al., 1992; Skoulakis et al., 1993; Cheng et al., 2001; Davis, 2005), so expression of Nf1 in the adult mushroom bodies makes sense. Additionally, the broad nature of nf1 expression in the central brain is not surprising because of its role in other behavioral outputs, i.e., circadian rhythm (Williams et al., 2001) and escape latency (The et al., 1997), both of which may require Nf1 expression in distinct neurons of the adult brain.
nf1 expression in mushroom body neurons, only during adulthood, rescues 3 h memory
Previous work using a ubiquitously expressed, heat-shock-inducible, wild-type transgene (hsnf1) demonstrated that nf1 is required in adulthood for normal olfactory associative learning (Guo et al., 2000). Although these experiments identified a physiological role for Nf1 in memory formation, they did not identify the neurons that require Nf1 for normal olfactory learning. After this work was completed, the Gene-Switch system was developed in our laboratory for simultaneous temporal and spatial control of gene expression (Mao et al., 2004). Gene-Switch is a Gal4-based, RU486-inducible regulator of uas transgene expression. With this system, we have been able to simultaneously identify time and space requirements of nf1 expression for normal olfactory learning.
Expression of uas–dnf1 was first induced by Gene-Switch line 12-1 during both development and adulthood, to confirm that this system could induce sufficient nf1 expression to rescue the nf1c00617 3 h memory phenotype. Gene-Switch line 12-1 uses a mushroom body enhancer that induces expression in both α/β and γ mushroom body neurons (Mao et al., 2004), and performance of this line does not differ from w(CS10) controls (data not shown). Animals were reared on food laced with RU486 from embryogenesis, through training and testing as adults. Expression of uas–dnf1 in mushroom body neurons during development and adulthood resulted in complete rescue of the 3-h memory phenotype (Fig. 4A). Next, flies were reared on normal food and transferred to food containing RU486 at eclosion. Our results indicate that expression of nf1, only in adult mushroom body neurons, is sufficient for complete rescue of the nf1 3 h memory phenotype (Fig. 4B).
nf1 expression in mushroom body neurons, only during adulthood, rescues nf1c00617 memory. nf1 expression was restored in mushroom body neurons of homozygous nf1c00617 mutants by expressing uas–dnf1 under the control of the Gene-Switch system. A, All groups were fed RU486 during development and adulthood, except as indicated, and performance was assayed 3 h after training. Expression of nf1 during development and adulthood resulted in complete rescue of the 3 h memory phenotype. For all groups, n = 6. Means ± SEM are shown. *p < 0.005. B, All groups were fed standard food during development and were transferred to food containing RU486 for 5 d after eclosion, except as indicated. Performance was assayed 3 h after training. Expression of nf1, only during adulthood, completely rescued the 3 h memory phenotype. For all groups, n = 6. Means ± SEM are shown. *p < 0.05.
nf1 expression in α′/β′ or γ neurons does not restore 3 h memory
Recent efforts, in our laboratory and others, have begun to distinguish the specific roles of individual mushroom body neuron subtypes in learning and memory (Yu et al., 2005, 2006; Akalal et al., 2006; Krashes et al., 2007). Some evidence suggests that synaptic output from α′/β′ neurons is required during acquisition and consolidation for the formation of stable memories (Krashes et al., 2007; Wang et al., 2008). Because our data implies a role for neurofibromin in memory acquisition, we decided to see whether nf1 expression in α′/β′ neurons would be sufficient to rescue the 3 h memory phenotype.
nf1 was expressed in α′/β′ neurons of nf1c00617 flies with the c305a–gal4 driver (Krashes et al., 2007; Aso et al., 2009). It was not possible to test for rescue of the nf1P1 phenotype, because w;c305a;nf1P1 animals have severely reduced fecundity. However, expression of nf1 in this subset of mushroom body neurons was not sufficient to rescue the 3 h memory phenotype of nf1c00617 homozygous mutants (Fig. 5A). This result suggests that Nf1-dependent signaling pathways are not required in α′/β′ neurons for acquisition and early memory formation.
nf1 expression in α′/β′ or γ neurons does not restore 3 h memory. A, Expression of nf1 was restored in α′/β′ neurons of homozygous nf1c00617 mutants by expressing a uas–dnf1 transgene under control of the c305a–gal4 driver. Performance was assayed 3 h after training. K33 was used as the control. All genotypes showed significant memory impairment relative to control flies. For all groups, n = 6. Means ± SEM are shown. *p < 0.0001. B, Expression of nf1 was restored in γ neurons of homozygous nf1c00617 or nf1P1 mutants by expressing uas–dnf1 under the control of the NP1131–gal4 driver. Performance was assayed 3 h after training. Control flies were w(CS10). All genotypes show significant memory impairment relative to control flies. For all groups, n = 6. Means ± SEM are shown. *p < 0.05.
A close association between nf1 and rut-AC during learning has been well established (Guo et al., 2000; Ho et al., 2007), and expression of rut-AC in γ neurons of the mushroom bodies partially rescues the learning phenotype of rut mutants (Zars et al., 2000; McGuire et al., 2003; Mao et al., 2004; Akalal et al., 2006; Blum et al., 2009). Therefore, we examined whether uas–dnf1 expression in γ neurons might rescue the phenotype of nf1 mutants. Driver NP1131–gal4 was used to express uas–dnf1 in γ neurons (Akalal et al., 2006; Aso et al., 2009). However, expression of nf1 in this subset of mushroom body neurons was not sufficient to rescue the 3 h memory phenotype of either mutant allele (Fig. 5B). This result suggests that Nf1-dependent signaling pathways are not required in γ neurons during memory acquisition and early consolidation. Furthermore, any role of Rut-AC in γ neurons, during PSI-EM processing, is Nf1 independent.
nf1 expression in α/β neurons restores protein synthesis-independent memory
Because expression of rut-AC in α/β neurons partially rescues the learning phenotype of rut mutants (Zars et al., 2000; McGuire et al., 2003; Mao et al., 2004; Akalal et al., 2006; Blum et al., 2009), we next expressed uas–dnf1 in both nf1P1 and nf1c00617 homozygous mutant backgrounds using the Mz1081–gal4 driver. With this driver, transgene expression is promoted throughout the central brain, including antennal lobe and α/β neurons. Expression of uas–dnf1 in these neurons was sufficient to rescue the 3 h memory phenotype of both nf1 homozygous mutant alleles (Fig. 6A). Next, we used the NP9–gal4 driver, which shows a more limited expression pattern relative to Mz1081–gal4. NP9 is also broadly expressed in the central brain, including antennal lobe and α/β neurons. Expression of uas–dnf1 with this driver was also able to fully rescue the 3 h memory phenotype of both nf1P1 and nf1c00617 homozygous mutant alleles (Fig. 6B). We next expressed uas–dnf1 using the c739–gal4 driver, which promotes robust transgene expression in the α/β mushroom body neurons and weak transgene expression in other neurons of the central brain (Aso et al., 2009). Once again, we saw complete rescue of the 3 h memory phenotype, providing more substantial evidence that the expression of nf1 only in α/β neurons is sufficient for restoring normal memory in nf1 mutants (Fig. 6C). Finally, we expressed uas–dnf1 using 17d–gal4, which promotes transgene expression in the central core of α/β mushroom body neurons (Aso et al., 2009). Expression with this driver did not restore normal 3 h memory of either mutant allele (Fig. 6D), suggesting that either the expression level induced by this driver is not sufficient or that it does not promote expression in a sufficient number of α/β neurons to support learning.
nf1 expression in α/β neurons rescues 3 h memory. Expression of nf1 was restored in α/β neurons of homozygous nf1c00617 and nf1P1 mutants by expressing a uas–dnf1 transgene under the control of either Mz1081–gal4 (A), NP9–gal4 (B), c739–gal4 (C), or 17d–gal4 (D). Performance was assayed 3 h after training. A–C, Expression of nf1 in α/β neurons resulted in complete 3 h memory rescue of both alleles. K33 was used as the control. All negative control groups showed significant memory impairment relative to control flies. D, Expression of nf1 under control of 17d–gal4 failed to rescue 3 h memory. Control flies were w(CS10). All genotypes showed significant memory impairment relative to control flies. For all experiments, n = 5–9. Means ± SEM are shown. *p < 0.05.
We also tested whether uas–dnf1 expression in α/β neurons would rescue 3 min memory, the earliest testable form of PSI-EM. We used the c739–gal4 driver to promote expression of uas–dnf1 in α/β neurons of nf1P1 homozygous mutant flies. We observed full rescue of the nf1 phenotype at 3 min after training (Fig. 7A). Additionally, we tested the effect of overexpressing nf1 in normally performing flies using c739–gal4. Consistent with a previous report (Guo et al., 2000), we observed no enhancement of normal performance when nf1 was overexpressed in α/β neurons (Fig. 7B).
nf1 expression in α/β neurons also rescues 3 min and 24 h memory but does not enhance control performance. A, Three minute memory was rescued by nf1 expression in α/β neurons. Expression of nf1 was restored in α/β neurons of homozygous nf1P1 mutants by expressing a uas–dnf1 transgene under control of the c739–gal4 driver, and performance was assayed 3 min after training. Complete rescue of the phenotype was observed, consistent with experiments showing rescue of 3 h memory. Control flies were w(CS10). For all groups, n = 6. Means ± SEM are shown. *p < 0.005. B, nf1 expression in α/β neurons rescued 24 h memory. Expression of nf1 was restored in α/β neurons of homozygous nf1P1 mutants by expressing a uas–dnf1 transgene under control of the c739–gal4 driver. Performance was assayed 24 h after 5×-spaced training, which produced long-lasting memory that was significantly different from that elicited by 5×-massed training. K33 was used as the control. For spaced versus massed, n = 12. For rescue, n = 15. Means ± SEM are shown. *p < 0.05. C, Control performance is not affected by neurofibromin overexpression in α/β neurons. Overexpression of uas–dnf1 in α/β neurons of heterozygous nf1P1 flies was driven by c739–gal4, and performance was assayed 3 h after training. For all groups, n = 5. Means ± SEM are shown. *p < 0.005.
Because Mz1081–gal4, NP9–gal4, and c739–gal4 drivers all promote some expression outside of the α/β neurons, i.e., in antennal lobe neurons, we wanted to further demonstrate that expression of nf1 in α/β neurons, rather than these other sites of expression, rescued the memory phenotype. Our laboratory previously introduced an MB{Gal80} transgene into the chromosome carrying c772–gal4, by recombination (Liu et al., 2007). The MB{Gal80} transgene contains the gal80 repressor gene for Gal4 (Lee and Luo, 1999) downstream of a mushroom body enhancer, which drives expression predominantly in the mushroom body neurons (Zars et al., 2000; Mao et al., 2004). Combining the MB{Gal80} transgene with c772–gal4 specifically suppresses uas transgene expression in the mushroom body neurons, whereas expression in antennal lobe neurons is unaffected (Liu et al., 2007). Expression promoted by c772–gal4 by itself is strongest in α/β and γ mushroom body neurons, moderate in antennal lobe neurons, and weak in α′/β′ mushroom body neurons (Liu et al., 2007; Aso et al., 2009).
As expected, the nf1c00617 3 h memory phenotype was fully rescued when c772–gal4 was used to express uas–dnf1 (Fig. 8A). However, flies carrying the combined c772–gal4; MB{Gal80} driver along with uas–dnf1 exhibited a performance score that was indistinguishable from nf1c00617 homozygous mutants (Fig. 8B). Therefore, rescue of the 3 h memory phenotype of nf1 mutants requires expression in α/β mushroom body neurons.
Inhibiting nf1 expression in α/β neurons prevents memory rescue. A, Expression of nf1 was restored in α/β neurons of homozygous nf1c00617 mutants by expressing a uas–dnf1 transgene under control of the c772–gal4 driver. No significant difference was observed between controls and c772/uas–dnf1;nf1c00617 flies, but all other genotypes showed a significant memory impairment relative to controls. Control flies were w(CS10). Performance was assayed 3 h after training. For all groups, n = 15. Means ± SEM are shown. *p < 0.005. B, When MB{Gal80} was introduced to inhibit Gal4 activity in mushroom body neurons, expression of nf1 by c772–gal4 did not rescue the 3 h memory phenotype of nf1c00617. Similar expression of uas–dnf1 in a heterozygous nf1c00617 genetic background did not alter performance relative to control, but all other genotypes showed significant memory impairment by comparison. Performance was assayed 3 h after training. Control flies were w(CS10). For all groups, n = 8. Means ± SEM are shown. *p < 0.05.
nf1 expression in α/β neurons restores protein synthesis-dependent long-term memory
Recent work in our laboratory suggests that a memory trace can be visualized in the α/β neurons of mushroom bodies 24 h after adult flies are subjected to a 5×-spaced training procedure (see Materials and Methods) that produces PSD-LTM (Yu et al., 2006). In contrast, a 5×-massed training procedure produces neither PSD-LTM nor a memory trace in α/β neurons. A deficit in LTM has been reported for nf1 mutants trained with a spaced protocol (Ho et al., 2007). It is possible that this phenotype is attributable to the requirement for nf1 in α/β neurons. We therefore used c739–gal4 to promote expression of uas–dnf1 in α/β neurons of nf1P1 homozygous mutant flies and tested for rescue 24 h after 5×-spaced training. We observed full rescue of the nf1 phenotype at 24 h (Fig. 7C). Because expression of nf1 in α/β neurons can restore both PSI-EM and PSD-LTM, our results suggest that Nf1 mediates multiple types of memory processing, through at least two different biochemical pathways, within the same population of neurons.
Discussion
Regardless of species being studied, neurofibromin is involved in many different brain activities, including, but not limited to, cognitive processes, circadian rhythms, cortical development, and glial development. Even within the cognitive realm, Nf1 function depends on the context of specific training conditions. Protein synthesis-independent short- and middle-term memories appear to require an activation of Rut-AC by Nf1, whereas protein synthesis-dependent long-term memory requires an additional modulation of Ras activity (Hannan et al., 2006; Ho et al., 2007). By rescuing the performance of homozygous mutants, we have demonstrated that expression of Nf1 in adult α/β mushroom body neurons is sufficient to support all forms of Nf1-dependent memory. We have also revealed a requirement for Nf1 during acquisition. Together, our observations expand our current understanding of Nf1 and Rut-AC functions and challenge current models of mushroom body neuron activity in olfactory memory formation.
Rut-AC is required in both α/β and γ mushroom body neurons for complete rescue of rut STM deficits (Akalal et al., 2006), yet we have shown that Nf1 is required only in the α/β mushroom body neurons. It is unclear why Rut-AC activation would only require Nf1 in one subset of neurons. One possibility is that the Nf1 stimulation of Rut-AC in α/β neurons during acquisition may indirectly facilitate, through unknown signals, Rut-AC activity in the γ neurons. Previous results have been interpreted to suggest that there are communication loops that exist between certain types of mushroom body neurons and with extrinsic mushroom body neurons for normal learning and consolidation (Yu et al., 2006; Krashes et al., 2007). A similar process could allow Rut-AC activation in γ neurons to be indirectly dependent on Nf1 in α/β mushroom body neurons. Alternatively, it could be that the Rut-AC is dependent on Nf1 in the α/β mushroom body neurons for its role in learning but Nf1 independent in γ mushroom body neurons.
For rescue of PSD-LTM, both Rut-AC (Blum et al., 2009) and Nf1 expression are required only in α/β neurons, suggesting that their interaction is necessary to support this form of memory as well. A recent study concluded that the Nf1-GRD, which has been shown to mediate an adenylyl cyclase activity (Hannan et al., 2006), is necessary and sufficient for Nf1-dependent LTM (Ho et al., 2007). In contrast to Rut-AC, this adenylyl cyclase activity is stimulated by Ras and is Gαs independent. It is important to note, however, that the Nf1-GRD domain only partially rescued the LTM phenotype of nf1 mutants. Full rescue of LTM required a full-length nf1 transgene. Together, these data and ours suggest that Nf1 simultaneously mediates the activation of both AC signaling pathways in α/β neurons to facilitate new protein synthesis and the formation of long-lasting memory.
Early work on the role of Rut-AC in olfactory associative memory suggested that this adenylyl cyclase plays a role in behavioral acquisition (Tully and Quinn 1985; Dudai et al., 1988). Using an olfactory avoidance assay, it was suggested that rutabaga mutants could obtain normal performance with more intense training (Dudai et al., 1984). We have also observed a delay in the acquisition of olfactory memory in rutabaga mutants, which require three times the amount of training as controls to overcome. A similar delay in acquisition was discovered for nf1 mutants, consistent with the hypothesis that Nf1 is required for G-protein activation of Rut-AC during learning (Guo et al., 2000; Ho et al., 2007). Our results also demonstrate that Rut-AC is essential for the stability of olfactory memory. However, this function is independent of an interaction with Nf1. We believe that the association of Nf1 and Rut-AC may be transient, only required for the initial activation of Rut-AC in its role as a molecular coincidence detector in α/β neurons (Tomchik and Davis, 2009). If this model were true, memory stability would therefore require continued stimulation of Rut-AC molecules via an independent and perhaps spatially distinct mechanism that does not require Nf1.
Recent efforts, in our laboratory and others, have attempted to assign temporal and operational phases of olfactory memory processing to distinct regions within the adult olfactory system. During pairing of odor and electric shock, new projection neuron synapses are recruited to the odor representation (Yu et al., 2004). Pairing dopamine application with neuronal depolarization in adult brain preparations results in a Rut-AC-dependent synergistic increase of cAMP in both α and α′ lobes (Tomchik and Davis, 2009), and memory acquisition requires synaptic transmission from α′/β′ neurons (Krashes et al., 2007). Although we have demonstrated that both Nf1 and Rut-AC are required for memory acquisition, neither of these need be expressed in α′/β′ neurons (Zars et al., 2000; McGuire et al., 2003; Mao et al., 2004; Akalal et al., 2006; Blum et al., 2009). We therefore propose that memory acquisition cannot be thought of as a specific event involving a distinct neuronal subset. Rather, we envision a model in which the pairing of odor and electric shock induces a change on the neuronal systems level. Each individual neuron subset may register this change in a different way, but every change is in some way necessary for memory acquisition as a whole. Additional work will be required to determine whether memory consolidation, retrieval, or processing of longer-term memories also require plasticity throughout the entire olfactory system.
It is clear from the data herein that Nf1 function is required in the adult brain, in α/β neurons defined by the c739–gal4 driver, for PSI-EM formation and for PSD-LTM formation. By identifying a minimal region in which Nf1 expression is required, we are now able to isolate its role in memory formation from others that may occur in the brain. This mapping promises a more accurate analysis of Nf1-dependent memory and insights into both memory processing as a whole and into the cognitive deficits associated with neurofibromatosis type 1.
Footnotes
These studies were supported by National Institutes of Health Grant NS19904 (R.L.D.). M.E.B. was supported by a Children's Tumor Foundation Young Investigator Award and National Institute of General Medical Sciences Grant T32GM008307. We thank Curtis Wilson for technical assistance.
- Correspondence should be addressed to Ronald L. Davis at his present address: Department of Neuroscience, The Scripps Research Institute, Jupiter, FL 33458. rdavis{at}scripps.edu
