RNA binding proteins hnRNP A2/B1 and CUGBP1 suppress Fragile X CGG premutation repeat-induced neurodegeneration in a Drosophila model of FXTAS

Abstract

Fragile X associated tremor ataxia syndrome (FXTAS) is a recently described neurodegenerative disorder of older adult carriers of premutation alleles (60-200 CGG repeats) in the fragile-X mental retardation gene (FMR1). It has been proposed that FXTAS is an RNA mediated neurodegenerative disease caused by the titration of RNA binding proteins by the CGG repeats. To test this hypothesis, we utilize a transgenic Drosophila model of FXTAS that expresses premutation length repeat (90 CGG repeats) from the 5’ UTR of the human FMR1 gene and displays neuronal degeneration. Here, we show that over-expression of RNA binding proteins, hnRNP A2/B1 and CUGBP1 suppress the phenotype of the CGG transgenic fly. Furthermore, we show that hnRNP A2/B1 directly interacts with riboCGG repeats and that the CUGBP1 protein interacts with the riboCGG repeats via hnRNP A2/B1.

Introduction

Fragile X Syndrome is the most common form of hereditary mental retardation, occurring in individuals lacking the fragile X mental retardation protein (FMRP). Large expansions of the CGG trinucleotide repeat (>200 repeats) in the 5’ untranslated region of the FMR1 gene leads to silencing of its transcript and the loss of FMR1 product, FMRP (Warren, 2001). Most individuals in the general population carry fewer than 60 CGG repeats while those individuals with CGG repeat expansions between 60 and 200 are referred to as premutation carriers. These alleles support transcription of FMR1 and hence premutation carriers are phenotypically normal with respect to the features of Fragile X syndrome. However, fragile X associated tremor/ataxia syndrome (FXTAS), a neurodegenerative disorder has recently been described in a subgroup of premutation adult carriers (Hagerman and Hagerman, 2004). FXTAS is characterized by tremor, gait problems, cerebellar dysfunction, cognitive decline and parkinsonism associated with generalized brain atrophy (Hagerman et al., 2001). Eosinophilic intranuclear inclusions in both neuronal and astroglial cells, Purkinje cell drop out and Purkinje axonal torpedos have been observed in postmortem examinations of the brains of premutation carrier males (Greco et al., 2002). The lack of association of features of Fragile X Syndrome with FXTAS (and vice versa) suggests an improbable relationship between the FMR1 protein and FXTAS. A distinguishing characteristic of premutation carriers is the production of FMR1 transcripts with extended CGG repeats, while patients with Fragile X Syndrome produce little or no FMR1 mRNA. Additionally, increased FMR1 mRNA levels have been observed in the premutation carriers and in a ‘knock-in’ CGG mouse model supporting the idea that FXTAS is an RNA-mediated neurodegenerative disorder (Tassone et al., 2000, Willemsen et al., 2003). A transgenic fly model expressing the 5’ UTR of the human FMR1 gene with 90 CGG repeats showed that the premutation length repeat out of the FMR1 mRNA context could cause neurodegeneration (Jin et al., 2003). Flies expressing these CGG repeats in the eye display disorganized ommatidia, de-pigmentation and progressive loss of photoreceptor neurons. As shown by temperature shift experiments these phenotypes are late onset and not a consequence of developmental abnormalities (Jin et al., 2003). The eye phenotype could be suppressed by over expressing hsp70, a chaperone involved in protein folding (Jin et al., 2003). These results suggest that transcription of the CGG90 repeats lead to an RNA-mediated neurodegenerative disease, possibly via influencing RNA binding proteins.

To test this model and to define modifiers of the rCGG repeats-mediated eye phenotype, we carried out a genetic screen making use of a collection of candidate RNA binding proteins. CUGBP1, an RNA binding protein discovered for its ability to bind CUG repeats and implicated in myotonic dystrophy type 1 (DM1) (Timchenko et al., 1996, Timchenko et al., 2001, Timchenko et al., 2004) was identified in our screen as the sole modifier among the candidates. We show that over-expression of CUGBP1 is able to suppress the neurodegenerative eye phenotype of the CGG90 transgenic flies. Furthermore, we show that the CUGBP1 protein is able to interact with the CGG repeats via hnRNP A2/B1, a riboCGG binding protein. hnRNP A2/B1 is an RNA-binding protein that is present in intranuclear inclusions of FXTAS patients (Iwahashi et al., 2006). We demonstrate that hnRNP A2/B1 binds directly to the CGG repeats (accompanying manuscript, Jin et al. and in this paper), and that over-expression of hnRNP A2/B1 and its two Drosophila homologues suppress the rCGG repeat-mediated neurodegenerative eye phenotype.

Results

Over-expression of CUGBP1 suppresses the rCGG induced neurodegeneration

We carried out a genetic screen on the CGG90 neurodegenerative eye phenotype to identify potential RNA binding proteins that modify the rCGG-mediated toxicity. The screen involved directing the expression of premutation-length CGG repeats to the eye with the Gmr-GAL4 driver using the Drosophila GAL4/UAS system. This was followed by crossing Gmr-GAL4, UAS-(CGG)₉₀-EGFP transgenic flies with flies mutant in genes coding for approximately 60 different candidate RNA binding proteins which were either created by our group or were available from the Bloomington stock center (Table 1). The progeny were examined for potential suppression or enhancement of the disorganized eye phenotype by comparison to control rCGG flies (Table 1). Through this screen, we identified one modifier of rCGG-mediated neurodegeneration, CUGBP1. Figure 1 shows that over-expression of human CUGBP1 suppresses the neurodegenerative phenotype of the CGG90 transgenic fly. Given the role of CUGBP1 in myotonic dystrophy type 1 (DM1), we also examined the other major protein implicated in DM1, Muscle blind-like protein 1 (MBNL1) (Kanadia et al., 2003; Ranum and Day, 2004). We did not observe a genetic interaction between MBNL1 and rCGG-mediated neurodegeneration (Table 1 and data not shown). This is unlikely to be due to inadequate levels of expression of MBNL1 since the same lines have been shown to modify the eye toxicity and muscle degeneration phenotypes in a fly model of myotonic dystrophy (de Haro et al., 2006).

Candidate RNA binding proteins screened to identify modifiers of the rCGG-mediated neurodegenerative eye phenotype.

Putative loss of function and over-expression alleles of a collection of RNA binding proteins.

GENE	Loss-of-function allele	Overexpression allele
ALY	Aly02267
ARET	AretQBaretBG00614aretBG01566
CUGBP		uascugbpm2a*, uascugbpm2e*
ATX2	atx206490	atx2EP3145
BICC	bicC4
BICD	bicDr5
BL	blKG02524
BOL	bol1, bolEP584
CPO	cpo01432
DSX	dsx18
EIF4A	eif4ak01501	eif4aEP1011
EIF4E	eif4e07238	eif4eEP3371
FNE		P(uasfne.C)4-10B
FMR1	fmr1Delta113M	P(uasfmr1.Z)3
HEPH	heph03429
HOIP	hoipk07104
HOW	howEP858, howj5B5
HRB27C	hrb27c02647
HRB87F	Hrb87fKG02089
HRB98DE	Hrb98deEY01550,hrb98deZCL0558
IMP		impEP760, impEP1433
IRP-1B	irp-1bEP961
LARK	larkEP3259	P(uaslark-3HA)23A
L(3)02094	L(3)02094
MSI	msi1
MSL-3	msl-3EP3144, msl-3KG00939
MUB		mubEP3623
MUSCLEBLIND	mbnP15144	uasmbnl1(M10M)
MXC	mxcG48
NOS	nosL7, nosj3B6
PAPB	pabpk10109
PAPB2		pabp2EP2264
PIWI	piwi06843
PUM	pumbem	pumEP3461
PUF68		puf68EP3058
QKR58E-2	qkr58e-2KG07766
QKR58E-3		qkr58e-3EP2493, qkr58e-3EY02038
RBP97D	rb97d1
RBP9	rbp9BG02784
RIN		rinEP920
RM62		rm62EP 3607
RNPS1		rnpsEP1O82
SC35	sc35KG02986
SLBP	slbpEP3182
SMG	smg1
SnRNP70K	snRNP70k02107
SPEN	spenk06805
SQD	sqdj6E3	sqdEP746
STAU	Staury9, stau1
SXL	sxlf1
TBPH	tbphKG08578
TRA2	tra2KG08361
TSU		tsuEP567
U2A	u2a1
U2AF38	u2af3806751, u2af38k14504
VIG		vigEP812
YTR	ytrk13108, ytrEP2136
ZN72D	zn72DBG02677

^*CUGBP1 over-expression was confirmed for this study by western blot analysis (data not shown).

Figure 1. CUGBP1 modulates rCGG-mediated neurodegenerative eye phenotype.

Over-expression of CUGBP1 suppresses rCGG-mediated eye toxicity. From left to right; fly expressing (CGG)₉₀-EGFP (GMR-GAL4 UAS: CGG90/UAS: EGFP) has disorganized fused ommatidia; middle, fly expressing (CGG)₉₀-EGFP and CUGBP1 (GMR-GAL4 UAS: CGG90/+; UAS: CUGBP/+) has restored organized ommatidia, fly expressing CUGBP1 (GMR-GAL4/+; UAS: CUGBP/+) has organized ommatidia. The crosses were performed at 22°C and Gmr-GAL4 was used.

CUGBP1 interacts with rCGG repeats via hnRNP A2/B1

Similar to the DM1 model, it has been proposed that the lengthy rCGG repeats sequesters RNA binding proteins from their normal function(s) which leads to neuronal degeneration (Hagerman and Hagerman, 2002; Jin et al., 2003). Therefore to understand how CUGBP1 modulates rCGG-mediated toxicity, we tested whether CUGBP1 could bind rCGG repeats. We did not detect a direct interaction between CUGBP1 and rCGG repeats (data not shown), which is consistent with a previous report (Timchenko et al., 1996). Next, we reasoned that CUGBP1 might interact indirectly with rCGG repeats via rCGGBPs, thus we turned to interactions between CUGBP1 and rCGGBPs. hnRNPA2/B1 and Pur α are the two RNA binding proteins identified to bind rCGG repeats in a parallel biochemical screen (accompanying paper, Jin et al.). Therefore, we tested the interactions between Pur α or hnRNP A2/B1 with CUGBP1. We found that endogenous Pur α did not co-immunoprecipitate with CUGBP1 (data not shown). However, endogenous hnRNP A2/B1 co-immunoprecipitated with endogenous CUGBP1 from mouse brain lysates and this interaction is RNA independent (Figures 2A, B and C), which suggests a distinct complex is formed between CUGBP1 and hnRNP A2/B1 independent of Pur α. Also, we showed that CUGBP1 does not interact with other RNA binding proteins, hnRNP A3 and FMRP, which suggests a specific interaction of CUGBP1 with hnRNP A2/B1 (Figure 2A). To confirm that the interaction between CUGBP1 and hnRNP A2/B1 is relevant to FXTAS pathogenesis, we investigated whether a complex is formed between all three components, CUGBP1, hnRNP A2/B1 and the rCGG repeats. Using in vitro translated protein, we found that CUGBP1 bound to rCGG repeats in the presence of hnRNP A2/B1 (Figure 2D). These observations suggest that the modulation of rCGG-mediated eye phenotype by CUGBP1 is through hnRNP A2/B1. To further investigate whether this physical interaction is relevant in vivo, we tested for possible genetic interactions between CUGBP1 and hnRNP A2/B1. We found that co-expression of Drosophila homologues of hnRNP A2/B1 proteins suppress the rough-eye phenotype caused by a high expression of CUGBP1 (Figure 3). This suggests that these proteins might antagonize each other's biological function/s although this relationship may or may not be relevant to FXTAS pathogenesis.

Figure 2. CUGBP1 interacts with rCGG repeats through hnRNP A2/B1.

Immunoprecipitation (IP) was performed on mouse brain lysates using anti-CUGBP1 (A) or anti-hnRNP A2/B1 (B) antibody with IgG as a negative control. The immunoprecipitates were analyzed by immunoblotting with antibodies specific to hnRNP A2/B1, hnRNP A3, FMRP or CUGBP1. FMRP and hnRNP A3 westerns were performed using different gels. 10% of the input used for IP was run in parallel. C. Immunoprecipitation (IP) was performed on mouse brain lysates with RNase using anti-CUGBP1 and analyzed by immunoblotting to hnRNP A2/B1 D. In vitro translated protein were incubated with biotinylated rCGG repeats, western blot data shows that CUGBP1 interacts with rCGG repeats in the presence of hnRNP A2/B1 (lanes 3 and 4)

Figure 3. Over-expression of Drosophila hnRNP A2/B1 homologues suppress the CUGBP1 rough eye phenotype.

From left to right: fly expressing CUGBP1 (GMR-GAL4/+; UAS: CUGBP/+) has disorganized ommatidia, fly expressing CUGBP1 and Hrb87F (GMR-GAL4/+; UAS: CUGBP/Hrb87F) or fly expressing CUGBP1 and Hrb98DE (GMR-GAL4/+; UAS: CUGBP/Hrb98DE) have more well organized ommatidia. The crosses were performed at 25°C.

hnRNP A2/B1 interacts with the rCGG repeats

We confirmed the interaction between rCGG repeats and hnRNP A2/B1 by performing gel-shift assays using the antibody against hnRNP A2/B1 in mouse cerebellar lysates, the brain region most affected in FXTAS (Greco et al., 2006). The addition of antibody against hnRNP A2/B1 leads to a supershift, indicating that hnRNP A2/B1 is present in the rCGG-protein complex (Figure 4A). We also carried out binding reactions of both cytoplasmic and nuclear mouse cerebellar lysates, with the CGG₁₀₅ as well as with a normal length CGG₃₀ biotinylated RNA. The captured proteins were analyzed using antibody against hnRNP A2/B1. We found that both normal (30 repeats) and premutation (105 repeats) length rCGG repeats could bind to hnRNP A2/B1. However, this interaction is most evident in the cytoplasmic mouse cerebellar lysates, despite the finding that hnRNP A2/B1 is present in both the cytoplasm and nuclei (Figure 4B). Nuclear hnRNPA2/B1 shows little or no interaction with rCGG repeats suggesting that some modification(s) of the protein in either the nucleus or cytoplasm alters its ability to bind the rCGG repeats.

Figure 4. Interaction between rCGG repeats and hnRNP A2/B1 modulates rCGG-mediated neurodegeneration.

A. Gel super shift assay shows the binding of hnRNP A2/B1 to the rCGG repeats. Lane 1, 100-bp RNA ladder; Lane 2, rCGG probe only; Lanes 3, rCGG probe with mouse cerebellar lysates; Lane 4-6, rCGG probe and mouse cerebellar lysates with increasing amounts of anti-hnRNP A2/B1 antibody. B. Mammalian hnRNP A2/B1 binds to rCGG repeats. Western using an antibody against hnRNP A2/B1 is shown. The biotinylated rCGG repeats were incubated with mouse cerebellar lysates and captured by Dynal beads. FT represents flow through while E indicates the eluted proteins that bind to rCGG repeats. Both nuclear and cytoplasmic lysates were used and indicated. 30, 105 rCGG repeats or beads alone were used in the binding reactions. The last two lanes are nuclear protein lysates and cytoplasmic lysates as inputs. C. hnRNP A2/B1 directly interacts with rCGG repeats and shows increased binding to 105 rCGG repeats. Equal amount of in vitro translated hnRNP A2/B1 was incubated with identical molar quantities of either biotinylated 30 rCGG repeats or 105 rCGG repeats; beads alone as control. Input and bound fractions are shown. D. Over-expression of hnRNP A2/B1 and its Drosophila orthologs suppress the rough-eye phenotype caused by the expression of CGG repeats, and this suppression is maintained when co-expressed with CUGBP1 (supplemental figure 2). From left to right, GMR-GAL4 UAS: CGG90/UAS: EGFP, GMR-GAL4 UAS: CGG90/+; UAS: Hrb98DE/+, GMRGAL4 UAS: CGG90/+; UAS: Hrb87F/+, GMR-GAL4 UAS: CGG90/UAS: EGFP, GMR-GAL4 UAS: CGG90/UAS: hnRNP A2/B1. Flies shown in the first three panels are grown at 25°C and the last two panels at 22°C.

To test that hnRNP A2/B1 directly interacts with rCGG repeats, we isolated the full-length cDNA encoding human hnRNP A2/B1 (A2 form) for in vitro translation. This protein was incubated with identical molar quantities of biotinylated normal (30 repeats) or premutation (105 repeats) length rCGG repeats. We demonstrated that hnRNP A2/B1 directly interacts with the rCGG repeats, and that increased binding of the protein to the premutation length rCGG repeats is observed. (Figure 4C). Also, we note that transgenic flies expressing normal length rCGG repeats do not have a rough eye phenotype (supplemental figure 1).

Over-expression of human and two Drosophila homologues of hnRNP A2/B1 suppress the rCGG-mediated phenotype

To further explore the functional significance of the association between hnRNP A2/B1 and rCGG repeats, we generated transgenic lines expressing members of the hnRNP A2/B1 family of proteins. Drosophila genes Hrb87Fand Hrb98DE encode the fly proteins HRB87F and HRB98DE (also known as hrp36 and hrp38, respectively) (Zu et al., 1996), which are the closest Drosophila homologues for mammalian hnRNP A2/B1 protein (55% identity). We generated UAS lines that over-express human hnRNP A2/B1, Hrb87F or Hrb98DE in the presence of a GAL4 driver, and crossed those lines with transgenic flies expressing 90 CGG repeat. We found that over-expression of either human hnRNP A2/B1, Hrb87For Hrb98DE suppresses the rCGG-mediated eye toxicity (Figure 4D). This result suggests that hnRNP A2/B1 is also involved in rCGG-mediated neurodegeneration. Furthermore, it supports the model that titration of these proteins by the rCGG repeats prevents them from carrying out their functions since over-expression of these proteins suppress the neurodegenerative phenotype of flies expressing rCGG repeats. Additionally, flies that are homozygous null for Hrb87F present with a mild rough eye phenotype and loss of photoreceptor neurons (Sengupta and Lakhotia, 2006) which is in agreement with the titration model.

Discussion

Fragile X-Associated Tremor/Ataxia Syndrome (FXTAS) is a neurodegenerative disorder recognized in some fragile X syndrome premutation carriers with FMR1 alleles containing 55–200 CGG repeats (Hagerman et al., 2004). Previous studies have demonstrated that FXTAS is RNA-mediated, caused by expanded fragile X premutation rCGG repeats (Arocena et al., 2005; Jin et al., 2003; Willemsen et al., 2003). Our studies have tested the hypothesis that RNA binding proteins are involved in FXTAS pathogenesis, and we report that CUGBP1 and hnRNPA2/B1 are two such RNA binding proteins.

HnRNP A2/B1 and CUGBP1 in FXTAS

HnRNP A2/B1 was identified to bind rCGG repeats in a parallel biochemical screen (accompanying paper, Jin et al.). We confirmed this interaction between rCGG repeats and hnRNP A2/B1 in our studies. We found that both normal (30 repeats) and premutation (105 repeats) length rCGG repeats could bind to hnRNP A2/B1. However, nuclear hnRNPA2/B1 shows little or no interaction with rCGG repeats suggesting that some modification(s) of the protein in either the nucleus or cytoplasm alters its ability to bind the rCGG repeats. Indeed, the protein kinase, Casein kinase 2 is able to phosphorylate hnRNP A2/B1 in vivo (Pancetti et al., 1999). The differential binding abilities of hnRNPA2/B1 to rCGG repeats depending upon the protein's cellular location suggests that cytoplasmic hnRNP A2/B1 has an independent or distinct role from its function in the nucleus. An alternative possibility to explain the differential binding of hnRNP A2/B1 is that other proteins compete with it for rCGG binding in the nuclear fraction. Iwahashi et al., 2006 identified more than twenty of such proteins in the nuclear inclusions of FXTAS patients.

Also, we identified CUGBP1 has a modifier of the rCGG mediated eye phenotype through our candidate based genetic screen. We show that CUGBP1 is able to interact with the rCGG repeats in the presence of hnRNP A2/B1. Furthermore, we show that over-expression of either CUGBP1 or hnRNPA2/B1 suppresses the eye phenotype caused by the expression of fragile X premutation rCGG repeats. These data support the model that titration of these proteins by the rCGG repeats prevents them from carrying out their functions since over-expression of these proteins suppress the neurodegenerative phenotype of flies expressing rCGG repeats.

The roles of hnRNP A2/B1 and CUGBP1 in the pathogenesis of FXTAS are intriguing. In DM1, total CUGBP1 steady-state levels are increased, with increased nuclear CUGBP1 and decreased CUGBP1 in the cytoplasm (Ranum and Cooper, 2006). In our study, we found that increased levels of CUGBP1 suppress the neurodegenerative phenotype caused by the rCGG repeats. This is an interesting and surprising observation as increased levels of CUGBP1 are implicated in DM1. However, the findings that cytoplasmic hnRNP A2/B1 interacts strongly with rCGG, and that CUGBP1 is able to interact with rCGG repeats via hnRNP A2/B1 suggest that principally cytoplasmic functions are compromised. This is consistent with the reduction of cytoplasmic CUGBP1 found in DM1.

MBNL1, the other major protein implicated in myotonic dystrophy, was found in the nuclear inclusions of FXTAS patients (Iwahashi et al., 2006). However we did not detect a modification on the rCGG repeat eye phenotype by modulating MBNL1 levels. This result raises the issue of the functional significance of a protein that is found in inclusions. The observation of widespread nuclear ubiquitin and hsp70 positive inclusions in affected tissues of FXTAS (Greco et al., 2002) suggested that these play a direct role in pathology, possibly by sequestering factors in a manner to that proposed in DM1. The presence of FMR1 mRNA in inclusions (Tassone et al., 2004) argues that rCGG repeats may nucleate inclusions that contain rCGGBPs. However, it is unlikely that the majority of FMR1 mRNA is confined to inclusions, and the soluble RNA may also interact aberrantly with RNA binding proteins. Thus, presence of a protein in inclusions may not be an indicator of a direct role in pathogenesis.

In summary, through both biochemical and genetic assays, we implicate the RNA binding proteins CUGBP1, hnRNP A2/B1 and Pur α (Pur α results are described in the accompanying manuscript by Jin et al.) in the pathogenesis of FXTAS. Our data support the model of FXTAS pathogenesis whereby these RNA binding proteins are titrated from their biological functions. These results could also explain the incomplete penetrance of FXTAS as polymorphisms in expression levels of these proteins could offer a protective effect to a fraction of premutation carriers. More importantly, the existing data suggest two distinct ways by which the CGG repeat exerts its toxicity. In the nuclear inclusions rCGG repeats sequester RNA binding proteins such as hnRNP A2/B1 (observed in FXTAS patients). However, our observations show that the interaction between hnRNP A2/B1 and the rCGG repeats is most evident in the cytoplasm. The data suggest another mechanism such that FMR1 mRNA not confined to nucleus may also interact aberrantly with RNA binding proteins.

Experimental Procedures

RNA-binding assays and Protein Identification

To prepare brain lysates, mouse brains were washed twice with phosphate-buffered saline (PBS) and homogenized in lysis buffer (10 mM Tris-HCl, pH 7.6, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT). The samples were then centrifuged for 5 min at 10,000 rpm to pellet nuclei and the supernatant (cytoplasmic fraction) was collected and used for RNA-binding assays. Radioactive-, biotin-, or fluorescent-labeled CGG repeat RNAs were synthesized using the RNAMaxx High Yield Transcription Kit (Stratagene). The RNA probe (100ng) was incubated with 20 μg brain lysates or with 50 ng in vitro-translated protein(s) at room temperature for 30 minutes in 1X binding buffer (20 mM Tris-HCl, pH 7.6, 100 mM KCl, 5 mM MgCl₂, 5 mM DTT and 10% glycerol). The binding reaction was loaded and separated on native polyacrylamide gel, which was analyzed via Storm 840 phosphorimager (Amersham Biosciences Corp). For the binding reaction with biotinylated RNAs, DynaBeads M-280 Streptavidin (Dynal, Invitrogen) were used to capture the rCGG-protein complex. The beads were washed before use and resuspended in binding buffer. The captured proteins were subjected to Western blots with anti-hnRNP A2/B1 (Santa Cruz Biotechnology) and anti-CUGBP1 (Upstate) antibodies.

Immunoprecipitation

Mouse brains were collected and homogenized in 1 ml ice-cold lysis buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 30 mM EDTA, 0.5% Triton X-100) with 2X complete protease inhibitors. All further manipulations of the brain lysates were performed at 4°C or on ice. Nuclei and debris were pelleted at 10,000 X g for 10 minutes; the supernatant was collected, and pre-cleared for 1 hr with 100ul recombinant protein G agarose (Invitrogen). Anti-CUGBP1 antibody (Upstate) or Anti-hnRNP A2/B1 antibody (Santa Cruz Biotechnology) was incubated with recombinant protein G agarose at 4°C for 2hr and washed three times with lysis buffer. The precleared lysates were immunoprecipitated with antibody coated recombinant protein G agarose at 4°C overnight. The precipitated complexes were used for western blot. Anti-hnRNP A2/B1 antibody (Santa Cruz Biotechnology), Anti-hnRNP A3 (Santa Cruz Biotechnology), Anti-CUGBP1 (Upstate) antibody or Anti-FMRP (kind gift from J.L Mandel) was used for western blot. For RNase treatment, the pre-cleared lysate was incubated with RNase A (1 mg/ml) at 37 degree for 15 min. Each sample was subjected to additional centrifugation at 10,000 X g before immunoprecipitation.

Drosophila Genetics

The pUAST constructs were generated by cloning full-length cDNAs, including 2 lines each of Drosophila Hrb87F, Hrb98DE and human hnRNP A2/B1 into the pUAST transformation vectors. The constructs were confirmed by DNA sequencing and then injected in a w¹¹¹⁸ strain by standard methods. All the other UAS lines, insertions and GAL4 lines used in this study were obtained from the Bloomington Drosophila stock center or generated in the lab. Fly lines were grown on standard medium with yeast paste added. Genetic screen was carried out at 22°C, all other crosses were performed at 25°C unless indicated in text.

Microscopy

For scanning electron microscopy (SEM) images, whole flies were dehydrated in ethanol, dried with hexamethyldisilazane (Sigma), and analysed with an ISI DS-130 LaB6 SEM/STEM microscope.