Short- and long-term memory are modulated by multiple isoforms of the fragile X mental retardation protein

Abstract

The diversity of protein isoforms arising from alternative splicing is thought to modulate fine-tuning of synaptic plasticity. Fragile X mental retardation protein (FMRP), a neuronal RNA binding protein, exists in isoforms as a result of alternative splicing, but the contribution of these isoforms to neural plasticity are not well understood. We show that two isoforms of Drosophila melanogaster FMRP (dFMR1) have differential roles in mediating neural development and behavior functions conferred by the dfmr1 gene. These isoforms differ in the presence of a protein interaction module that is related to prion domains and is functionally conserved between FMRPs. Expression of both isoforms is necessary for optimal performance in tests of short- and long-term memory of courtship training. The presence or absence of the protein interaction domain may govern the types of ribonucleoprotein (RNP) complexes dFMR1 assembles into, with different RNPs regulating gene expression in a manner necessary for establishing distinct phases of memory formation.

Figure 1.

Alternative splicing produces FMRP isoforms in mammals and Drosophila. A, Schematic of alternative splicing patterns in the human and D. melanogaster FMR1 genes that lead to diversity in C-terminal peptide domains. In hFMRP, the presence of a nuclear export sequence is dependent on inclusion of exon 14 in the mature transcript, and choice of the 3′ acceptor site in exon 15 determines the presence of a serine residue that is a phosphorylation substrate. An alternative 3′ acceptor site is present in exon 17. Skipping of exon 14 creates a +1 frame shift (Ashley et al., 1993), resulting in FMRP isoforms with novel C-terminal peptides. A bar over exons 13–17 of hFMRP denotes a region that interacts with kinesin light chain (Dictenberg et al., 2008). B, In D. melanogaster, a 541 nt intron that separates exons 9 and 10 of dfmr1 is alternatively spliced to produce the dFMR1 Q/N+ and Q/N− isoforms. C, Sequence of the 541 nt intron between exons 9 and10 of dfmr1. The retained 3′ segment of this intron contains several consensus hexanucleotide sequences (AAUAAA) that recruit the cleavage and polyadenylation machinery. The Q/N− isoform (designated PB in FlyBase; http://flybase.bio.indiana.edu/) lacks a glutamine/asparagine-rich peptide encoded by exon 11. D, The 529 aa Q/N− isoform of dFMR1 is detected via Western blot of protein extract from heads and constitutes a modest fraction of total dFMR1 protein. E, C-terminal peptides of human FMR/FXR and dFMR1 are related. A ClustalW alignment reveals that an 81 aa peptide, encoded by exons 16 and 17 of human FMR1, shares 38% identity/similarity with the terminal 81 residues of dFMR1. Amino acids of hFMRP depicted in gray denote residues that arise from use of an alternate splice acceptor site in exon 17. ClustalW alignment of the terminal 81 residues of dFMR1 with human FXR1 and FXR2 shows 33 and 43% identity/similarity between the peptides. A Blosum scoring matrix, opening and end gap penalties of 10, and extending and separation gap penalties of 0.05, which are the default values for the ClustalW program, were used for the analyses.

Figure 2.

The C-terminal Q/N domain of dFMR1 has protein interaction properties. A, Fusion of codons specifying the dFMR1 Q/N domain to GFP results in aggregation of GFP signal, although none is observed with native GFP. Scale bars, 10 μm. B, The dFMR1 Q/N domain does not confer a metastable state of activity to a reporter protein. Fusion of the dFMR1 Q/N domain to a constitutively active glucocorticoid receptor (GR⁵²⁶) transcription factor results in a high level of fluctuation in LacZ expression, resulting in colonies that vary in the degree of X-gal hydrolysis. Cultures derived either blue or white colonies give a mix of blue and white when plated to X-gal media. Arrowheads point to white colonies derived from blue colony parental material. GPD, Glyceraldehyde-3-phosphate dehydrogenase.

Figure 3.

Generation of dfmr1ΔQ/N alleles and fly stocks. A, Domain map of dFMR1, depicting the extent of the Q/N domain deletion. B, Protein sequence coded by the terminal exon of dfmr1. Q/N residues are denoted in bold, and the extent of the deletion is highlighted in black. C, The Q/N domain deletion proteins are stable in a steady state. Two transgenic lines are analyzed, one with a dFMR1 expression level equal to that of wild type (ΔQ/N 8), and the other as an ∼1.8-fold overexpression (ΔQ/N 1B). WT, res, Wild-type rescue. D, E, Spatial expression of wild type (WT) and the ΔQ/N protein in the CNS are quite similar. Enhanced dFMR1 expression is noted in the antennal lobes (al), ventrolateral protocerebrum (vl pr), and lateral horn (l ho). Scale bars, 100 μm.

Figure 4.

Developmental and circadian locomotion phenotype of ΔQ/N flies. A, Hatch rates of eggs deposited by females expressing dFMR1 ΔQ/N protein are greater than those of dfmr1 null females and do not differ significantly from those of wild-type (WT) females. n > 290 for all genotypes tested; and p ≤ 0.0001, χ² test for homogeneity. B, The ΔQ/N flies do not exhibit a frequency of midline crossing of mushroom body β-lobe axons that differs from wild type. n > 30 for all genotypes examined; p ≤ 0.001, χ² test for homogeneity. C, An excess of neuromuscular junction boutons are present in larvae expressing the ΔQ/N allele as the sole source of dFMR1 protein compared with those expressing a wild-type allele of dfmr1. **p ≤ 0.01; ***p ≤ 0.001, ANOVA. D, E, The ΔQ/N flies have a partial loss-of-function phenotype in rhythmic circadian locomotion activity. D, The percentage of ΔQ/N flies judged to be rhythmic [based on a power(1)/significance(1) difference >10] is lower than wild type and higher than that of flies with a deletion null allele of dfmr1 (*p ≤ 0.05; ***p ≤ 0.001, Fisher's exact test). E, The strength of circadian locomotion rhythms in flies harboring the ΔQ/N allele is intermediate to those with wild-type or null alleles of dfmr1 (*p ≤ 0.05, ANOVA). power(1) and significance(1) values were derived from Clocklab software (Actimetrics).

Figure 5.

Social behavior and short-term memory phenotypes of dfmr1 ΔQ/N mutants. A, The ΔQ/N mutants have a significant decrease in naive courtship activity compared with flies with a wild-type (WT) dfmr1 allele. B, C, dfmr1ΔQ/N mutants exhibit immediate recall (0–2 min after training) but not 1 h short-term memory of conditioned courtship training. Flies with wild-type, null, or ΔQ/N alleles of dfmr1 were paired with a nonreceptive female for 1 h (training) and subsequently paired with a receptive virgin female at the above times for testing memory of training. The significant reduction in courtship index seen at 0–2 min after training compared with the naive controls with wild-type and ΔQ/N flies demonstrates memory of the training. In contrast, the 1 h posttraining courtship indices of ΔQ/N mutants rise to a level similar to the naive control, whereas the wild-type flies retain suppressed courtship activity. ***p ≤ 0.001, ANOVA.

Figure 6.

Flies unable to produce the dFMR1 Q/N- isoform have 1 h STM of courtship training but are deficient in 4 d long-term memory of training. A, Schematic of the dfmr1^L allele. A cDNA clone was used to replace dfmr1 genomic DNA in which the alternative splicing that produces the Q/N− isoform occurs, thus eliminating Q/N− synthesis. B, Expression analysis of stocks harboring dfmr1^L transgenes as the sole source of dFMR1 protein. The two transgenic stocks examined (C51L × L1, C51L × L32) express the Q/N+ isoform at levels ranging from 90 to 100% of wild type (WT). The measured expression levels of dFMR1 were normalized to that of β-tubulin. C, Flies expressing the dfmr1^L allele have naive courtship levels that equal or exceed those of wild-type flies and have functional 1 h STM of courtship training, as judged by the repressed courtship index after training. Three independent insertion lines were tested (C51L, L32, and L1). ***p ≤ 0.001, ANOVA. D, Flies expressing either the dfmr1^L or ΔQ/N allele are defective in 4 d long-term memory of courtship training. For each genotype tested, males were either paired with a nonreceptive female for 7 h or sham trained (also referred to as naive trained) in the absence of a female for the same period. The males from both training classes were then kept in isolation for 4 d, then paired with a virgin female, and monitored for courtship activity. The relative courtship index is the ratio of the courtship indices from trained males to sham-trained males. A value of 100 is demonstrative of no memory of training. Two dfmr1^L lines were tested (C51L and L32). Because the single transgenes in these lines provide approximately a haploid dose of dFMR1 protein, dfmr1^L stocks were crossed to each other to increase the protein dose to that of a diploid (C51L × L1, C51L × L32). ***p ≤ 0.001, ANOVA. FS, Frame-shift null allele.

Figure 7.

A model for functions of dFMR1 Q/N+ and Q/N− isoforms in neurons based on results from this study and other published observations on FMRP function. The deficit in 1 h STM in ΔQ/N flies parallels the requirement for protein synthesis in mGluR-induced hippocampal LTD (Huber et al., 2000, 2002). The C-terminal peptides of human FMRP facilitate an interaction with kinesin (Dictenberg et al., 2008) and are related to the C-terminal peptide of dFMR1 (Fig. 1E). The dFMR1 Q/N domain may then allow for interactions with RNP complexes that promote dendritic RNA trafficking and/or regulation of local translation of mRNA species that are critical for early stages of memory formation. Production of the dFMR1 Q/N− isoform is critical for 4 d LTM of courtship training. The Q/N− isoform could contribute to regulated expression of genes that further consolidate memory of training. This might occur through translation regulation mechanisms within dendrites or spines. Alternatively, de novo transcription is required for certain forms of LTM. FMRP is detected in the nucleus, and roles for FMRP in chromatin regulation, RNA processing, and nucleocytoplasmic transport of RNA have been reported or proposed (Deshpande et al., 2006; Lai et al., 2006; Didiot et al., 2008). Disruption of isoform-specific roles for any of these steps in nuclear gene expression may contribute to the observed deficit in LTM.