Defective Neuronal Development in the Mushroom Bodies of Drosophila Fragile X Mental Retardation 1 Mutants

dfmr1 mutants have a mushroom body β-lobe midline-crossing phenotype

The MBs are paired neuropils arranged symmetrically about the midline of the brain. In wild-type Drosophila melanogaster, axon-like fibers in the medially projecting lobes terminate near the midline but do not cross it (Strausfeld et al., 2003). We identified MB structural phenotypes in dfmr1 mutants by immunostaining for FasII of whole brains dissected from pharate adults. As previously described for sectioned material (Crittenden et al., 1998), confocal microscopy reveals strong immunolabeling of α/β neurons along the peduncles as well as distally as their axons bifurcate and project dorsally into the α lobes and medially into the β lobes. Weaker anti-FasII labeling is seen in γ neurons which are unbranched and project medially into the γ lobes (Fig. 1A, E). In contrast to wild-type MB morphology, brains of dfmr1 mutants show β-lobe fibers extending across the midline, sometimes sufficient to cause apparent fusion of the right and left β lobes (Fig. 1C). Using GAL4-201Y and a UAS-GFP reporter (see Materials and Methods) to visualize the γ lobes (Fig. 1B, D), it was evident that dfmr1 mutant γ-lobe fibers do not extend across the brain midline (Fig. 1D). In other words, dfmr1 mutations cause a selective defect in MB morphology involving neurons born during metamorphosis, whereas their γ-neuron neighbors, which are regenerating in close proximity at the same time (Lee et al., 1999), are unaffected.

Because the β-lobe midline-crossing phenotype was variable in degree, we developed a classification scheme based on the apparent fraction of β-lobe axons that cross the midline (see Materials and Methods) (Fig. 1F-H). Several dfmr1 mutant alleles were examined either as homozygotes or as transheterozygotes, including the P-element insertion allele, P[EP]3517, and four internal deletions, Δ83, Δ3, Δ113, and Δ50, derived by imprecise excision (Fig. 2F). With the exception of Δ50 homozygotes, many of which could not be scored (see below), the penetrance and expressivity of the β-lobe midline-crossing phenotype was high for all mutant genotypes tested, with >60% of brains showing severe midline crossing (Fig. 2A). Within a given genotype, there was no sexual dimorphism in penetrance or expressivity (data not shown). Although there was some variation in penetrance and expressivity among the alleles, there was no difference between the hypomorphic insertion allele, P[EP]3517, which has reduced protein expression (Morales et al., 2002) and Δ3, the largest deletion allele, which must be protein-null (Dockendorff et al., 2002). MBs of newly eclosed or 2-d-old adult dfmr1 homozygous mutants also showed β-lobe midline crossing in a majority of animals (see below).

In OreRC, a wild-type laboratory strain, we found a very low incidence and severity of β-lobe midline crossing: 7% mild and 7% moderate (Fig. 2B), which is similar to other analyses based on brain sections (Moreau-Fauvarque et al., 1998) and consistent with the occasional wild-type β-lobe axon seen crossing the midline in Golgi preparations (N. J. Strausfeld, personal communication). When Δ83 was placed over a wild-type third chromosome from an OreRC parent, the resulting heterozygotes showed no midline crossing (Fig. 2C). Similarly, Δ113 over a wild-type third chromosome from one of several sources showed normal β-lobe morphology (data not shown). Balancer siblings from the dfmr1 mutant stocks had either normal β lobes or a low incidence of midline crossing (Fig. 2C). Hence, the β-lobe midline-crossing phenotype is recessive.

The β-lobe midline-crossing phenotype maps to dfmr1

Approximately 70% of P[EP]3517 homozygous brains exhibited β-lobe midline crossing, with a great majority having a severe phenotype. If the phenotype results from P-element insertion into the dfmr1 gene, then the precise excision of this transposon should rescue β-lobe morphology. Excision lines were generated by standard procedures, and several independent precise excisions of the P element were identified by PCR amplification using primers flanking the insertion site (data not shown). Examination of the brains of homozygous precise-excision animals showed phenotypic reversion to wild-type or near-wild-type levels (Fig. 2B). For example, Ex(3517)-16 homozygotes had 14% β-lobe midline crossing but only of the mild or moderate class. This is comparable with the frequency and severity of the phenotype in OreRC (Fig. 2B).

In a parallel approach to mapping the β-lobe midline-crossing phenotype, we provided a single wild-type copy of dfmr1 (P[dfmr1⁺]) to mutant transheterozygotes. To be able to compare siblings, we crossed w; P[dfmr1⁺]; Δ3/TM6C, Tb Sb e to either w; +/+; Δ83/Δ83 or w; +/+; Δ113/Δ113. Each experiment yielded transheterozygous siblings (Δ3/Δ83 or Δ3/Δ113, respectively) that differed by the presence or absence of the wild-type dfmr1 transgene, which was detectable because of the effect of mini-w⁺ on eye color. For both Δ3/Δ83 and Δ3/Δ113, the presence of a single wild-type dfmr1 transgene greatly reduced the penetrance and expressivity of the β-lobe midline-crossing phenotype (Fig. 2D). Together, the phenotypic reversion after P-element excision and the partial rescue by a single copy of dfmr1⁺ demonstrate that β-lobe midline crossing maps to the dfmr1 gene. This MB phenotype is very similar to that of the mutant linotte (Moreau-Fauvarque et al., 1998; Simon et al., 1998), originally identified because of a memory phenotype (Dura et al., 1993).

Given the high penetrance and expressivity of the β-lobe midline-crossing phenotype of dfmr1 mutants, it is surprising that other investigators did not detect it using anti-FasII immunostaining (Dockendorff et al., 2002). Potential contributors to this discrepancy include both technical and biological issues. First, if lower-resolution microscopy methods were used, it would have been possible to confuse β-lobe midline crossing with the normal morphology of the adjacent ellipsoid body or simply to miss a mild phenotype. Second, because there is some late-pupal lethality of dfmr1 mutants (see Materials and Methods), the penetrance and severity of MB phenotypes in adults may be lower than in pharate adults. Our observation that only 42% of 2-d-old dfmr1 homozygous mutants have severe β-lobe midline crossing (compare Fig. 2A, E) is consistent with this idea. Finally, there could be spontaneous “correction” of β-lobe defects after eclosion, analogous to the report of transient abnormalities of cortical dendritic development in fmr1 mutant mice (Nimchinsky et al., 2001). However, we consider this very unlikely because the overall penetrance of β-lobe midline crossing in 2-d-old adult mutants is very high, 79% (n = 19).

Genetic background strongly affects the penetrance of β-lobe midline crossing

The normal β-lobe morphology of Δ83/+ brains demonstrates the recessive nature of the midline-crossing phenotype. Nonetheless, the modest penetrance of the phenotype in balancer siblings (≤30%) (Fig. 2C) was noteworthy. Moreover, when we replaced the third chromosome balancers to create dfmr1 stocks for different purposes, we noticed wide variation in the midline-crossing phenotype among balancer heterozygotes. These observations suggested the possibility of an interaction between dfmr1 mutations and other sites on the third chromosome. To test this directly, we mated Δ83/TM6B, Tb e to Δ83/TM3, Sb e and examined three classes of progeny siblings for the frequency and severity of β-lobe midline crossing (Fig. 2E). As previously, Δ83/ Δ83 brains showed 100% penetrance, the vast majority of which was severe, and Δ83/TM6B showed 30% penetrance, most of which was mild and moderate. However, Δ83/TM3 brains showed 90% penetrance, and 90% of that was severe. In other words, in the presence of the TM3 balancer, the dfmr1 β lobe midline-crossing phenotype becomes dominant. A similar enhancement was seen with the TM1, Me red e third chromosome balancer (data not shown). Thus, the β-lobe midline-crossing phenotype caused by dfmr1 mutations is highly susceptible to genetic background effects, as is the hippocampal phenotype of Fmr1 mutant mice (Ivanco and Greenough, 2002; Mineur et al., 2002).

Development of the β-lobe midline-crossing phenotype during metamorphosis

The α/β Kenyon cells are born during the first half of metamorphosis; as they differentiate, each neuron extends an axon-like process that branches into the ipsilateral α/β lobes (Lee et al., 1999). Given the neurite overgrowth phenotypes previously reported in dfmr1 mutants (Zhang et al., 2001; Dockendorff et al., 2002; Morales et al., 2002), the midline-crossing phenotype in dfmr1 mutant MBs at the end of metamorphosis might result from failure of β-lobe axons to terminate their growth as they approach the midline, thus extending into the contralateral β-lobe region. An alternative possibility is that wild-type β-lobe axons transiently cross the midline during their initial outgrowth and are subsequently pruned back, with the dfmr1 mutant phenotype reflecting failure of branch pruning. To resolve this question, we compared MB development in wild-type and mutant brains, using Δ83 homozygotes because they manifest complete penetrance of the midline-crossing phenotype, and several different control genotypes, including 201Y UAS-GFP/+, w(z) and Δ83/TM6B.

First, we used the 201Y driver to show that the medial lobes of late larval γ neurons in Δ83 or Δ113 homozygous mutant MBs (n = 20) do not cross the midline (data not shown). Moreover, retraction of vertical and medial branches of γ neurons occurs on schedule and to its full extent in young mutant pupae (n = 6) (data not shown). Given the normal midline phenotype of mutant larval and young pupal γ neurons, the late-stage β-lobe midline-crossing phenotype cannot be blamed on faulty γ-neuron morphology earlier in development. Second, we used FasII immunostaining to track the development of α/β neurons. In wild-type pupae at HE + 16-17 hr, the α and β lobes appear as thin bundles projecting dorsally and medially (Fig. 3A). These bundles lengthen and thicken over the next several hours, and by HE + 22 hr, the tips of the β lobes reach their maximum medial extension, approaching but not crossing the brain midline (Fig. 3C). The appearance is similar at HE + 24 hr, except for a slight enlargement of lobe diameter (Fig. 3E). Over the next 24 hr, the two lobes continue to increase in diameter as fibers of newly born α/β neurons are added (Fig. 3G). Between 48 and 72 hr after HE, α/β lobes resemble their appearance in the pharate adult (data not shown). We saw no evidence in the wild type of transient midline crossing of β-lobe fibers.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

The β-lobe midline-crossing phenotype in dfmr1 mutants develops during metamorphosis. Projection of sequential optical sections through the MB lobe regions at successive stages during metamorphosis is shown. Times indicated are hours after HE, the beginning of the pupal period. The α and β lobes were visualized by FasII immunostaining, which also labels many other pathways, especially during early pupal development. Frontal views; dorsal is at the top. The distal end of the peduncle (p) is visible in some images. The white arrowheads indicate the position of the brain midline. Pairs of images compare wild-type in the left column (A, C, E, G) with Δ83/Δ83 mutants in the right column (B, D, F, H) at the same developmental stage. A, B, HE + 16-17 hr. In both wild-type (201Y UAS-GFP/+) and mutant brains, the medial tips of the β lobes (arrows) can be seen approaching the midline. The lobes are thin relative to their mature morphology. The apparent length difference between the wild-type (A) and mutant (B) α lobes is attributable to differences in the orientation of the samples during mounting. C, D, HE + 22 hr. Wild-type [w(z)] and mutant β-lobe termini are adjacent to the midline (arrows). In C, the image of the β-lobe termini is projected onto part of the ellipsoid body (eb). The wild-type α/β lobes have thickened compared with earlier stages. D, In the mutant, α/β-lobe growth is somewhat delayed, and the left α lobe (asterisk) is misdirected out of the plane of the image. (For additional examples of α-lobe misdirection, see Fig. 5.) The apparent reduced length of the right α lobe is attributable to variation in sample mounting. E, F, HE + 24 hr. The γ lobes have become FasII-positive and can be seen adjacent to the midline in both wild-type and mutant brains. E, The wild-type (201Y UAS-GFP/+; Δ83/TM6B) α/β lobes look similar to those at the previous stage. F, In the mutant (201Y UAS-GFP/+; Δ83/Δ83), β-lobe fibers have crossed the midline (arrow), with the fused region almost as thick as the adjacent β lobes. G, H, HE + 48 hr. G, Wild-type (201Y UAS-GFP/+; Δ83/TM6B) α/β lobes are nearing maturation, although they are still thinner than in the adult. H, The mutant brain (201Y UAS-GFP/+; Δ83/Δ83) shows severe β-lobe midline crossing (arrow), with full fusion of right and left sides.

In Δ83/Δ83 mutant brains, the α and β lobes also appear as thin FasII-immunoreactive bundles at HE + 16-17 hr (Fig. 3B), with progressive thickening over the next 30 hr. The medial tips of the β lobes approach the midline at the correct time (HE + 22 hr) (Fig. 3D), but immediately thereafter, fibers can be seen to extend from the β-lobe termini across the midline (Fig. 3F). By 48 hr, the right and left β lobes appear fused and very similar to the pharate-adult mutant phenotype (Fig. 3H). Thus, the midline-crossing phenotype is attributable to β-lobe axons failing to stop at the normal lobe terminus and, instead, growing across the brain midline into the contralateral β lobe. Consistent with what we saw in the pharate adults, the phenotype usually develops symmetrically.

Accessory mushroom body phenotypes of dfmr1 mutants

In addition to the high-penetrance, recessive β-lobe midline-crossing phenotype of dfmr1 mutants, other MB abnormalities were observed at low frequency with some alleles (Figs. 2, 4). These accessory phenotypes most often involve the α lobe, but β-lobe involvement was occasionally seen as well. There were two general varieties of accessory phenotypes: lobe(s) missing or lobe(s) misdirected. For ease of graphical display (Fig. 2), we grouped “missing” and “misdirected” accessory phenotypes but separated them according to which lobes were involved. Note that the combination of highly penetrant β-lobe midline crossing and low-penetrance α-lobe-missing has been previously reported in the memory mutant linotte (Simon et al., 1998). Because the accessory phenotypes could be found in the absence of β-lobe midline crossing, we consider them independent phenotypes. Of course, in the extreme case of both β lobes missing or misdirected (as in some Δ50 homozygotes), the midline phenotype cannot be scored.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Accessory phenotypes of the MB α and β lobes associated with some dfmr1 alleles. Projections of sequential optical sections through the MB lobe regions of FasII-immunostained pharate adult brains (except C, which is from a 2-d-old adult) are shown. Frontal views; dorsal is at the top. A, α lobe misdirected without β-lobe midline crossing (Δ83/TM6B). The right α lobe projects parallel to the ipsilateral β lobe (arrow). B, α lobe misdirected with β-lobe midline crossing (Δ3/Δ83). The right α lobe projects parallel to the ipsilateral β lobe (arrow). The β lobes are virtually fused across the midline (arrowhead). C, β lobe misdirected (Δ83/Δ83). The left β lobe projects parallel to the ipsilateral α lobe (arrow). The contralateral α/β lobes appear morphologically normal. D, α lobes missing with β-lobe midline crossing (Δ50/Δ50). Neither α lobe is present (stars), and the β lobes are fused across the midline (arrowhead). E, β lobe missing (Δ50/Δ50). The left β lobe is not present in its normal location (star), and the ipsilateral α lobe is thicker than usual (arrow), suggesting β-lobe misdirection and α/β-lobe fusion. F, β lobe reduced, with probable partial misdirection (Δ83/TM3). The right β lobe is greatly reduced in thickness; its image is projected onto the ellipsoid body (eb). The ipsilateral α lobe is enlarged (arrow) relative to the contralateral one, suggesting that part of the right β lobe is misdirected vertically and fused with the α lobe. G, α and β lobes missing unilaterally, with contralateral β-lobe midline crossing (Δ50/Δ50). FasII-immunoreactive mushroom body fibers have formed a large clump (arrow) ipsilateral to the missing lobes (stars). The weakly labeled fibers represent the peduncle (p) and the γ lobe. On the contralateral side, β-lobe fibers cross over the midline a short distance (arrowhead). H, α and β lobes missing bilaterally (Δ50/Δ50). No recognizable α or β lobes are visible (stars), but the peduncles and γ-lobe fibers are present. FasII-immunoreactive mushroom body fibers form large clumps bilaterally (arrows).

The accessory phenotypes are dramatic and very diverse. In some brains, the aberrant α or β lobe appeared to project parallel to its ipsilateral β or α lobe, respectively (Fig. 4A-C). In other words, the α/β axons had branched, but the two bundles projected side by side either vertically or medially. In other preparations with an apparently missing lobe, the remaining ipsilateral lobe was much thicker than normal, suggesting that α/β-lobe fusion had occurred (Fig. 4E, F). In cases of a missing lobe without evidence of misdirection (Fig. 4D), it is possible that α/β axons failed to branch at the bottom of the peduncle and then selected one of two possible projection paths to generate a single α or β lobe. It is also possible that all missing lobes are really misdirected but that we were not always able to discern the path or location of the misdirected lobe (Fig. 4G, H). It is worth considering the detection sensitivity for accessory phenotypes, especially compared with the β-lobe midline-crossing phenotype. If a small subset of α/β neurons had misdirected projections, e.g., both branches projecting medially within the β lobe, this might be very difficult to detect by our methods. Nor would single-neuron analyses such as Golgi impregnations (Strausfeld et al., 2003) or MARCM (mosaic analysis with a repressible cell marker) clones (Lee and Luo, 1999) reliably find such defects if they affected a small percentage of α/β neurons. Hence, our analysis may underestimate the true penetrance of accessory phenotypes. In all cases in which accessory phenotypes were observed in pharate adults, both γ lobes were present, albeit somewhat reduced in size in the most severely affected MBs, and projected in the normal medial direction.

The accessory phenotypes are also striking because of their genotype distribution and mode of inheritance (Fig. 2). The P-element insertion allele P[EP]3517 had a low incidence (∼10%) of accessory phenotypes, and these were eliminated when the P element was excised precisely (Fig. 2A, B). The penetrance and expressivity of accessory phenotypes of Δ50 homozygotes were exceptionally high. Accessory phenotypes were seen with all Δ83 or Δ50 genotypes, including Δ83 and Δ50 homozygotes and, with lower penetrance, in any transheterozygous mutants with one of those alleles (i.e., Δ50/Δ113, Δ83/Δ3, and Δ83/Δ113) (Fig. 2A, E) and Δ83/Df(3R)by62 (see below). Moreover, they were seen in Δ83 and Δ50 heterozygotes in which the other third chromosome was wild type for dfmr1 (Δ83/+, where + came from OreRC, Δ83/TM6B, Δ83/TM3, and Δ50/TM6B), even in the absence of the β-lobe midline-crossing phenotype (Fig. 2C, E). In the genomic rescue experiment with Δ83/Δ3 (Fig. 2D), the dfmr1⁺ transgene reduced the frequency of β-lobe midline crossing, but the accessory phenotypes were still present at an unimproved frequency. In contrast, Δ3 and Δ113 homozygotes, and Δ3/Δ113 transheterozygotes never displayed accessory phenotypes. In other words, the accessory phenotypes are restricted to a subset of dfmr1 alleles and have a semidominant pattern of inheritance. Thus, although all five alleles tested have loss-of-function behavior with respect to β-lobe morphology at the midline, three of them (Δ83, Δ50, and, to a lesser extent, P[EP]3517) have gain-of-function properties with respect to α- and β-lobe formation and projection.

A deficiency chromosome reveals regulators of MB maturation during metamorphosis

As part of our efforts to map the β-lobe midline-crossing phenotype, we placed Δ83 in combination with Df(3R)by62, the only available cytologically visible deletion in the dfmr1 region. One-third of Δ83/Df(3R)by62 MBs showed accessory phenotypes, in keeping with the semidominant, gain-of-function effect of Δ83 described above. However, only 8% exhibited the β-lobe midline-crossing phenotype, and it was only mild or moderate in severity (n = 39). On closer examination of the FasII-labeled MBs in Δ83/Df(3R)by62 pharate adults, we recognized that the α and β lobes appeared very immature, resembling those seen in wild-type brains at earlier stages of metamorphosis (Fig. 5A-C). In many cases, the lobes appeared younger than those of pupae whose β-lobe fibers were just approaching the midline. Similar results were seen in Δ113/Df(3R)by62 and P[EP]3517/Df(3R)by62 pharate adults (data not shown). Hence, this new phenotype precludes the use of Df(3R)by62 for mapping MB phenotypes. Nonetheless, as explained below, it has revealed novel information about the genetic basis of MB maturation during metamorphosis.

The MB α/β immaturity phenotype is characterized by abnormally thin α and β lobes that are very irregular along their lengths (Fig. 5A-C), whereas the γ lobes appear normal. Because lobe thickness and contour vary somewhat in mature wild-type brains, and because the phenotypes of Δ83/Df(3R)by62 brains were variable, we coded a large number of MB images from different genotypes and, thus blinded, developed a categorical maturity scoring scheme (see Materials and Methods): WNL, moderate, and severe (Fig. 5A-C). Subsequent scoring was done blind to genotype. Genetic features of the α/β immaturity phenotype were revealed by comparison of Df/Δ83 and Df/Balancer siblings generated in two independent experiments by somewhat different mating schemes (Fig. 6; for details, see legend). Figure 6, A and B, depicts the data graphically and demonstrates a dominant effect of Df(3R)by62. In each experiment, 80-90% of both Δ83/Df(3R)by62 and Df(3R)by62/Balancer brains showed α/β-lobe immaturity. In contrast, all MBs of Δ83/Balancer siblings were WNL (Fig. 6B). In a third experiment, we examined the Df(3R)by62/+ progeny of Df(3R)by62/TM6C and OreRC parents (Fig. 6C). Again, >90% of the MBs of deficiency heterozygotes were immature, confirming the dominant effect of the deficiency.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

The Df(3R)by62-associated α/β immaturity phenotype is dominant and enhanced by dfmr1 mutations. A-C, Three independent experiments in which FasII-labeled pharate adult brains were scored for α/β-lobe maturity as in Figure 5. All samples were scored blind with respect to genotype. A, Progeny siblings from the mating of Δ83/TM6B to Df(3R)by62/TM1. In Df(3R)by62/Balancer [Df(3R)by62/TM6B] brains, most of the α/β lobes were outside the normal maturity range. In Δ83/Df(3R)by62 siblings, the immaturity phenotypes were shifted toward greater severity. B, Genotypes associated with the mating of 201Y UAS-GFP; Δ83/TM6B to Df(3R)by62/TM6C. Δ83/Df(3R)by62, Progeny siblings. Df(3R)by62/Balancer includes Df(3R)by62/TM6B progeny siblings as well as the parental genotype Df(3R)by62/TM6C. Once again, the presence of Δ83 enhances the phenotypic severity of α/β-lobe immaturity associated with Df(3R)by62. In contrast, the α/β lobes of Δ83/Balancer (Δ83/TM6C) siblings are 100% within the normal range. C, Df(3R)by62/+ progeny from the cross of Df(3R)by62/TM6C to the wild-type OreRC strain. Less than 10% of brains showed α/β maturity in the normal range, confirming the dominant effect of Df(3R)by62.

A comparison of Df(3R)by62 over a dfmr1⁺ chromosome (balancer or OreRC) with Df(3R)by62 over the dfmr1 allele Δ83 reveals a clear difference in the severity profiles of MB immaturity. For Df/+, in three independent experiments scored blind, the majority of samples showed moderate immaturity phenotypes. In contrast, when the deficiency is paired with a dfmr1 mutation, 40-50% of the brains show severe MB immaturity. Thus, although the effect of Df(3R)by62 is dominant, the severity of the immaturity phenotype is enhanced by Δ83, demonstrating genetic interaction between dfmr1 and the deficiency chromosome. This synergistic interaction is likely to involve one or more loci in the deficiency interval flanking dfmr1, but other sites on the third chromosome may participate as well.

The MB immaturity phenotype could result from either arrest or delay of α- and β-lobe development. To distinguish between these two mechanistic possibilities, we compared MB development in Δ83/Df(3R)by62 and wild-type brains at successive times during the first half of metamorphosis (Fig. 5D-I). MB lobe development of Δ83/Df(3R)by62 brains appears immature beginning at HE + 18 hr (Fig. 5E), and the disparity with the wild type gets progressively worse over time thereafter (Fig. 5F-I). However, there is a progressive, albeit very slow, increase in lobe maturation in the Δ83/Df(3R)by62 brains, indicating that the developmental defect is one of delay and not arrest. Taken together, the data suggest a role for dfmr1 in the overall rate of α/β-lobe maturation.

dFMRP limits neurite extension in a learning and memory brain center

We have demonstrated that dfmr1 mutations cause a strongly penetrant neuronal overextension phenotype, in a highly plastic brain region well known for its role in multimodal sensory integration, olfactory associative learning, and visual context generalization (Zars, 2000; Roman and Davis, 2001). The β-lobe midline-crossing phenotype has recessive, loss-of-function characteristics and arises during the de novo differentiation of α/β neurons during metamorphic development of the MBs. Neuronal excess is a recurring theme in the fmr1 mutant phenotypes of both flies (Zhang et al., 2001; Dockendorff et al., 2002; Morales et al., 2002; Lee et al., 2003; this study) and mammals (Comery et al., 1997; Irwin et al., 2001), although some neuron types show reduced axon extension (Morales et al., 2002). From an evolutionary perspective, midline fusion of MB lobes represents an atavistic change: a reversion to a more primitive form (Strausfeld et al., 1998).

Despite the simultaneous process outgrowth of newly born α/β neurons and regenerating γ neurons, the dfmr1 midline-crossing phenotype is restricted to the β lobes, with regrowing γ-lobe axons respecting the brain midline. This differential sensitivity of γ and α/β neurons to dfmr1 mutations parallels their distinct cognitive functions. γ lobes are required for short-term olfactory memory formation (Zars et al., 2000), whereas short-term memory is not affected by lack of α and β lobes (Pascual and Préat, 2001). Rather, α/β lobes mediate olfactory memory retrieval (McGuire et al., 2001) and long-term memory formation (Pascual and Préat, 2001). On the basis of these studies and others (Heisenberg, 2003) linking MB structure with cognitive functions, we predict that dfmr1 mutants will show memory phenotypes that help clarify the specific function of α/β neurons.

The midline-crossing defect of dfmr1 mutant MBs implies either the lack of a midline repulsive signal or, alternatively, a failure by β-lobe axons to sense, recognize, or respond to that “stop” signal. Other mutant conditions that cause MB axon overextension across the midline provide clues to the underlying mechanism(s). The memory mutant linotte (lio) (Dura et al., 1993) also has a recessive, highly penetrant β-lobe midline-crossing phenotype that is sensitive to genetic background and sometimes accompanied by missing α lobes (Moreau-Fauvarque et al., 1998; Simon et al., 1998). lio is expressed in the developing MBs, especially the α and β lobes (Moreau-Fauvarque et al., 1998), as well as in a transient midline glial “ring” whose structure it regulates (Simon et al., 1998). lio is probably identical to the atypical receptor tyrosine kinase-encoding derailed gene (Moreau-Fauvarque et al., 2002) (but see Bolwig et al., 1995), which mediates repulsive signaling via the ligand Wnt5 in the embryonic nerve cord, preventing midline crossing by developing axons (Bonkowsky et al., 1999). Mutations of castor, which encodes a transcription factor expressed in a subset of CNS progenitor cells, also cause β-lobe midline crossing and interact with linotte mutations (Hitier et al., 2001). Overexpression of the actin-binding protein Ciboulot induces β-lobe midline crossing (Boquet et al., 2000). The involvement of regulators of transcription (Castor), translation (dFMRP), cell surface signaling (Linotte/Derailed), and the actin cytoskeleton (Ciboulot) highlights the complex pathway controlling MB lobe morphogenesis. Translational repression of ciboulot by dFMRP would explain the observed phenotypes. Finally, mutations in the dFMRP interactor CYFIP cause midline-crossing defects in the embryonic CNS (Schenck et al., 2003). By analogy with mechanisms controlling axon behavior there, dFMRP may regulate Slit-Roundabout signaling (Guthrie, 2001).

Genetic complexities of the dfmr1 locus: parallels to mammalian Fmr1

Three types of genetic complexity associated with dfmr1 provide challenges to understanding its biology, as well as opportunities to discover loci involved in its function. First, the recessive dfmr1 overextension phenotype is very sensitive to genetic background, such that related third-chromosome balancers enhance the phenotype to varying degrees. The very features that make balancers such powerful tools, breakpoints that prevent recombination and markers that can be followed through successive generations (Greenspan, 1997), may act in trans as modifiers of mutations under study. While the source of genetic background variation is different in murine inbred strains, the phenomenon of background-dependent phenotype modification is common and, as in the case of Fmr1, may be dramatic. The size of particular terminal fields within the hippocampus is significantly reduced, and associated with poor spatial learning, in mice whose Fmr1 mutation was on a C57Bl/6 background (Mineur et al., 2002), whereas that structure was significantly enlarged when the same mutation was crossed onto an FVB background (Ivanco and Greenough, 2002). Similarly, MB phenotypes due to mutations of mushroom bodies miniature vary greatly depending on genetic background (deBelle and Heisenberg, 1996).

Second, the dfmr1 accessory phenotypes causing misdirection or absence of α and β lobes, presumably because of pathfinding errors, are semidominant, incompletely penetrant, and associated with a subset of dfmr1 mutations. The phenotypic constellation of β-lobe midline crossing plus α/β misdirection has been reported for mutants of two other genes, linotte, which is quite similar to dfmr1 in its MB phenotype (Moreau-Fauvarque et al., 1998), and alpha-lobes-absent, which shows highly penetrant α/β-lobe misdirection and infrequent β-lobe midline crossing (Pascual and Préat, 2001). Thus, the accessory phenotypes of dfmr1 cannot be easily dismissed. Nonetheless, their allele specificity and semidominant nature are perplexing and suggest two alternative explanations. All of the imprecise excision alleles fail to produce a protein product detectable by a monoclonal antibody (Wan et al., 2000; Zhang et al., 2001; Dockendorff et al., 2002; Morales et al., 2002) that recognizes an epitope in the N-terminal 40% of the protein (T. Jongens, personal communication). However, the alleles that produce accessory phenotypes, the P[EP]3517 insertion allele, and the two imprecise excision alleles with the least loss of dfmr1 coding sequence (Δ83 and Δ50) might be capable of directing expression of an altered dfmr1 transcript or protein, with gain-of-function properties and undetectable by the monoclonal antibody. The rescue of accessory phenotypes in the precise-excision line is consistent with this model. Moreover, there is a precedent for gain-of-function Fmr1 phenotypes in mammals. Elevated levels of Fmr1 RNA associated with fragile-X premutation alleles cause premature ovarian failure in females and a neurodegenerative disorder in older males (Oostra and Willemsen, 2003). Alternatively, if all the excision alleles are truly nulls, then Δ83, Δ50, and P[EP]3517 chromosomes may share a linked allelic variant that acts alone, or in concert with dfmr1 mutations, to induce α/β pathfinding defects.

Third, a large chromosomal deficiency that flanks dfmr1, Df(3R)by62, has a dominant effect on the MB α/β lobes, causing developmental delay that is enhanced by dfmr1 mutations. The genetic interaction between dfmr1 and the deficiency chromosome indicates that dfmr1 participates in a second aspect of α/β development, boosting the rate of lobe maturation during metamorphosis. It also makes loci in the vicinity of dfmr1 (85D10-11 through 85F8-11) candidate target genes of dFMRP. Several mechanisms could explain the apparent inhibition of β-lobe midline crossing in most dfmr1/Df(3R)by62 animals: (1) lobe development is so compromised that fibers do not progress far enough to cross; (2) a critical period “permissive” for midline crossing is missed because of developmental delay; or (3) phenotypic suppression is caused by an interacting dosage-sensitive locus.

D. melanogaster as a model for fragile X syndrome

The Drosophila genetic model system is valuable for studying molecular pathophysiology of human disease (Bonini, 2000; Lasko, 2002; Driscoll and Gerstbrein, 2003). Perhaps most exciting is its use to identify novel therapeutic strategies for neurological disorders with few treatment options (Warrick et al., 1999; Kazantsev et al., 2002). The circadian rhythm and neuronal morphology phenotypes of dfmr1 mutants (Zhang et al., 2001; Dockendorff et al., 2002; Inoue et al., 2002; Morales et al., 2002; Lee et al., 2003; this study) are reminiscent of human fragile X sleep disorders (Gould et al., 2000) and cortical neuropathology (Irwin et al., 2001), respectively. Genetic and biochemical approaches in Drosophila have identified two dFMRP targets, a MAP1B homolog (Zhang et al., 2001) and Rac1 (Lee et al., 2003; Schenck et al., 2003). In addition, experiments in Drosophila have been central to demonstrating the involvement of small interfering RNA and micro RNA in the translational control mechanism of FMRP (Caudy et al., 2002; Ishizuka et al., 2002; Schenck et al., 2003; Jin et al., 2004). The sensitivity of dfmr1-regulated MB structure to genetic interactions will allow screening for dFMRP targets and interacting proteins relevant to cognition and other behaviors.

The dfmr1 model of fragile X syndrome should also be viewed in the wider context of phylogenetic conservation of genes controlling cognition. Fundamental molecular mechanisms of learning and memory are shared between vertebrates and invertebrates (Mayford and Kandel, 1999). Of the ∼300 molecularly identified human mental retardation genes, 87% have a homolog, and 76% have a single-candidate functional ortholog in Drosophila (Inlow and Restifo, 2004). Some of these Drosophila genes have already revealed neurological phenotypes, including learning and memory defects (Inlow and Restifo, 2004, and references therein).

Drosophila MBs manifest not only developmental reorganization (Technau and Heisenberg, 1982; Lee et al., 1999), but also experience-dependent plasticity (Technau, 1984; Heisenberg et al., 1995; Barth and Heisenberg, 1997). Hence, MB phenotypes in Drosophila mental retardation gene mutants are of particular interest. Drosophila Lissencephaly-1 was the first of these to be shown to regulate MB neuronal structure (Liu et al., 2000). Our demonstration of dfmr1 MB defects suggests that others are likely to follow as additional mental retardation genes are studied in Drosophila.