Dendrite branching and self-avoidance are controlled by Turtle, a conserved IgSF protein in Drosophila

Developing neurons form dendritic trees with cell type-specific patterns of arborization, ranging from simple arbors with few branches to highly elaborate arbors that cover receptive territories with many branches. In neurons with even the most complex trees, dendrite branches growing from the same neuron(isoneuronal branches) avoid one another as they spread over a territory to receive sensory or synaptic inputs. Together, dendrite branching and self-avoidance are crucial for sculpting the particular architecture of a neuron's receptive field. Both processes are thought to be controlled by molecular recognition events that occur between isoneuronal branches, or between dendrites and the substrata along which they grow. However, few of the molecules participating in these recognition events have been described.

Cell surface recognition molecules that promote dendrite growth and/or branching include cadherins (Gao et al.,2000; Kimura et al.,2006; Shima et al.,2007; Sweeney et al.,2002), as well as those mediating responses to neurotrophins(Horch and Katz, 2002), B-type ephrins (Horch and Katz,2002), and cues that direct dendritic guidance, such as Semaphorins (Komiyama et al.,2007; Polleux et al.,2000), Slits (Dimitrova et al.,2008; Furrer et al.,2007; Godenschwege et al.,2002; Whitford et al.,2002) and Netrins (Furrer et al., 2003). In contrast to these examples, which promote or guide dendrite arborization, recognition molecules that prevent inappropriate or excessive dendrite branching in neurons with simple arbors remain unidentified.

Recognition mechanisms underlying dendrite self-avoidance have only recently emerged, with findings that the Dscam family of immunoglobulin superfamily (IgSF) proteins promote self-avoidance in Drosophila(Hughes et al., 2007; Matthews et al., 2007; Soba et al., 2007) and mice(Fuerst et al., 2008). It remains to be determined whether other families of cell surface proteins are also required to promote self-avoidance.

The identification of novel transmembrane proteins required for dendrite branching and self-avoidance is a key step in understanding molecular mechanisms that underlie dendrite patterning. The Drosophila protein Turtle (Tutl) and its mammalian orthologs, Dasm1 (Igsf9) in mice and IGSF9(KIAA1355) in humans (Doudney et al.,2002; Shi et al.,2004), are type 1 transmembrane proteins with an ectodomain comprising five immunoglobulin (Ig)-like domains and two fibronectin type III repeats (Fig. 1A). In Drosophila, mutations of the tutl gene impair responses to tactile stimuli and the execution of complex coordinated behaviors(Bodily et al., 2001), but the causes of these nervous system deficits are unknown. To date, no morphological defects have been reported for tutl mutants, despite the structural similarity of Tutl to the Neogenin, Deleted in Colorectal Carcinoma, Frazzled and Roundabout families of axon guidance receptors(Bodily et al., 2001).

In mice, the Tutl ortholog Dasm1 is selectively expressed in the developing hippocampus (Mishra et al.,2008; Shi et al.,2004). Dasm1 knockout mice have no observable defects in dendrite morphogenesis in the developing hippocampus, nor have defects of neuronal differentiation, synaptogenesis or behavior been seen in these mice(Mishra et al., 2008). Therefore, the role Dasm1 in the mammalian nervous system remains uncertain,and genetic approaches to study Dasm1 function in mice could be complicated by redundancy of Dasm1 with Igsf9b, a closely related protein(Mishra et al., 2008).

Here, we have used genetic approaches to study the effects of tutlmutations on dendritic arborization (da) neurons in the Drosophilaperipheral nervous system. We found that Tutl is expressed on dendrites of da neurons and, through loss-of-function and gain-of-function experiments in vivo, we demonstrate that Tutl cell-autonomously controls dendrite branching and self-avoidance. Tutl restricts branching in neurons with simple arbors and promotes self-avoidance in neurons with highly complex arbors. These results demonstrate that a member of the Tutl/Dasm1/IGSF9 family of proteins can influence dendrite morphogenesis in vivo, and that neurons of different classes employ Tutl as a common molecular component of mechanisms that sculpt dendrite arborization patterns of dramatically different complexity.

Flies were obtained from stock centers at Bloomington[tutl⁰¹⁰⁸⁵, tutl^Df (Df(2L)ed-dp)] and Harvard(tutl^f03096 and tutl^f02770), and from published sources (GAL4^109(2)80, GAL4²²¹,ppk1.9-GAL4, UAS-mCD8::GFP). We generated tutl²³ by FLP/FRT-mediated recombination (Parks et al., 2004) to delete the intervening DNA between tutl^f03096 and tutl^f02770, leaving a reconstituted WH PBac element at the deletion site(Fig. 1B), as confirmed by PCR and DNA sequencing. We generated UAS-tutl by cloning into the pUAST vector a 5.2-kb EcoRI/BamHI fragment from the EST RE40452,which encodes full-length tutl corresponding to tutl-RD in FlyBase(Fig. 1B). For UAS-tutlΔcyto, we used PCR (oligonucleotide primers:5′-ACG ACT CAC TAT AGG GCG-3′ and 5′-CGC CTCTAGACT ATA CGG CAC AAA C-3′) to amplify a fragment from EST RE40452. The second primer introduced a stop codon and an XbaI site (underlined). The fragment was cut with EcoRI and XbaI and introduced into EcoRI/XbaI-restricted pUAST. UAS-tutlΔcyto encodes a truncated Tutl protein comprising Tutl amino acids 1-879 predicted by Tutl-RD.

For tutl MARCM, virgin females of the stock elav^C155-GAL4,UAS-mCD8::GFP,hs-FLP; FRT40A,tub-GAL80 were crossed to males that were either elav^C155-GAL4,UAS-mCD8::GFP,hs-FLP;FRT40A or elav^C155-GAL4,UAS-mCD8::GFP,hs-FLP;FRT40A,tutl²³. For cut MARCM, flies of the stock FRT19A,tub-GAL80, hs-FLP; GAL4^109(2)80, UAS-mCD8::GFP were crossed to flies carrying FRT19A, cut^c145/FM7c. Embryos were collected for 2 hours, incubated at 25°C for 2-3 hours, then heat shocked at 38°C for 1 hour and incubated at 25°C until they were analyzed just prior to pupation. Larvae mutant for tutl were cultivated on agar plates. Mutant animals were selected with the aid of balancer chromosomes CyO,twi-Gal4,UAS-GFP (for tutl, ab and kn) or TM3,twi-Gal4,UAS-GFP (for ss).

Larvae were dissected in PBS and GFP-positive da neurons were imaged with confocal microscopy using a Yokogawa spinning disk system (Perkin-Elmer) on an Eclipse TE2000-U microscope (Nikon). z-series images were collected using Metamorph software (Molecular Devices) and prepared for publication in Photoshop by converting images to grayscale and adjusting brightness and contrast. Reconstruct software (Fiala,2005) was used to quantify the numbers of branch termini, branch points and crossing points, as well as dendritic length (classes I-III) and dendritic field area [polygon method(Grueber et al., 2002)]. The dendritic arbors of class IV neurons were traced and measured with Imaris software (Bitplane). The data were tested for normal distribution using the Shapiro-Wilk test, and statistical analysis was performed using Analyse-It software for Microsoft Excel.

Anti-Tutl polyclonal antiserum was raised in rabbits to a GST-Tutl (amino acids 1-421) fusion protein corresponding to immunoglobulin (Ig) domains 1-3 of the Tutl ectodomain, then affinity-purified and pre-absorbed using standard methods. For anti-Tutl immunofluorescence, embryos or third instar larvae were dissected in PBS and fixed in 4% paraformaldehyde(Ou et al., 2008), then anti-Tutl antibody (1:25, 4°C) was detected with Rhodamine Red-X-conjugated secondary antibody (1:300). In double labeling of embryos for Tutl and HRP, we also added Cy2-conjugated anti-HRP (1:500, 4°C). Prior to mounting samples from third instar larvae, muscles overlying the dorsal cluster da neurons were removed by dissection for better visualization.

Four da neuron classes (I-IV) of increasing dendritic complexity and size can be readily observed in the larval body wall of Drosophila,sandwiched in two dimensions between muscles and epidermis(Grueber et al., 2002). We focused on the dorsal-most cluster of peripheral sensory neurons, which contains at least one representative from each class. We examined Tutl expression with immunofluorescence and found that Tutl was expressed in the cell bodies and along the dendrites of class I neurons ddaD and ddaE in third instar larvae (Fig. 1C′,D′). Class I neurons have the least complex arbors of all da neurons. Tutl was also readily observed in the cell body of the class IV neuron ddaC (Fig. 1C′,D′), the arbors of which are the most complex. Tutl was also expressed in the class II neuron ddaB and the class III neurons ddaA and ddaF (Fig. 1C′),in addition to in other sensory neurons in the dorsal cluster (see Fig. S4 in the supplementary material). The specificity of the antibody for Tutl was confirmed by the absence of expression in tutl²³ mutants(Fig. 1E′), which carry a novel tutl allele (Fig. 1B).

To examine the phenotypical consequences of tutl mutations in da neurons, we studied tutl²³ and other available P-element insertion (tutl⁰¹⁰⁸⁵) and deficiency(tutl^Df) alleles. We began by examining dendrite morphology of the class I da neuron ddaE, using the class I driver GAL4²²¹ and UAS-mCD8::GFP as a reporter to reveal the dendritic tree (Grueber et al.,2003). In control larvae at third instar, ddaE neurons have a simple, comb-like appearance (Fig. 2A) and 24.1±0.8 (mean±s.e.m.) branch termini per cell (Fig. 2C). One of the two or three primary dendrites projects dorsally and gives rise to several lengthy interstitial secondary branches that grow in a posterior direction toward the segment boundary (Fig. 2A). By contrast, the dendritic trees of ddaE neurons in homozygous or hetero-allelic tutl mutants had a number of defects, including severely shortened interstitial branches and irregular patterns of curled or tortuous growth,which often lacked directed orientation(Fig. 2B; see Fig. S1A-C in the supplementary material). We scored these defects while blind to genotype, and found them in 20/20 ddaE neurons from tutl⁰¹⁰⁸⁵/tutl²³ mutants, but in only 2/20 wild-type controls, which indicated a high penetrance of the tutlmutant phenotype. In addition, compared with wild-type ddaE neurons, we found significantly more branch termini (Fig. 2B; Fig. S1A-C in the supplementary material), increasing to 36.7±0.9 in tutl²³ homozygotes(Fig. 2C). This was as severe as in tutl²³/tutl^Df hemizygotes(33.5±1.1; Fig. 2C; Fig. S1B in the supplementary material), supporting our molecular and immunochemical data that tutl²³ is a null allele of tutl. Heterozygotes (tutl²³/+) had a degree of branching that was intermediate between controls and homozygotes(Fig. 2C), suggesting that ddaE branching is sensitive to the levels of Tutl. Overall, the mutant genotypes had increases in ddaE branches that ranged from 126-152% of wild-type controls(Fig. 2C). Similar observations were made for ddaD (not shown), another class I da neuron.

Although Tutl is expressed in dendrites of class I da neurons, the phenotypes we observed could result from either cell autonomous or non-cell autonomous Tutl activity. To investigate whether tutl is required cell autonomously in class I da neurons, we generated single mutant neuron clones using the MARCM system (Lee and Luo, 1999). Control ddaE MARCM clones exhibited normal morphology and branching (24.0±1.2; Fig. 2C,E), but the dendrites of tutl²³ ddaE MARCM clones showed increased branch termini to the same level as that found in tutl mutant animals (34.7±2.6; Fig. 2C,F). Other features of the tutl homozygous mutant phenotype (i.e. shortened interstitial branches, irregular patterns and directions of growth) were not readily observed in tutl²³ ddaE MARCM clones, and thus we are unable to ascribe them to cell autonomous loss-of-Tutl function. Therefore, we focused on the role of Tutl in the control of dendrite branching, where MARCM analysis pointed to a specific and cell autonomous role for Tutl in preventing excessive dendrite branching.

When normalized for dendritic length, ddaE neurons in tutl⁰¹⁰⁸⁵/tutl²³ mutants and tutl²³ MARCM clones retained increased numbers of branch termini relative to controls (Fig. 2D; see also Fig. S2A in the supplementary material), which suggests that tutl mutations increase dendrite branching complexity independently of ddaE dendrite growth. We analyzed the branching defect of tutl mutants in more detail by counting (1) branch points on primary dendrites that project directly from the cell body, and (2) second or third order branch points situated more distally on the arbor(Fig. 2G). Although the number of branch points on primary dendrites was unchanged in tutl mutants and tutl²³ MARCM clones, there was a clear increase in the number of second and third order branch points.

To determine whether Tutl cell autonomously inhibits branching in other classes of da neurons, we used MARCM to examine class II (ddaB), class III(ddaA, ddaF) and class IV (ddaC) neurons(Grueber et al., 2002). In tutl²³ MARCM clones for neurons of class II(Fig. 3A-D) and class III(Fig. 3E-H), we found no significant changes of branch number nor did we observe any effects on the pattern, growth or targeting of their dendritic trees. In the class IV neuron ddaC, we observed defects in dendrite self-avoidance (see below), but branch number was unaffected (see Fig. S2B in the supplementary material, MARCM data). Together, these results indicate that tutl has class-specific effects on dendrite branching in vivo.

Dendrite self-avoidance involves repulsive interactions between isoneuronal branches following transient contact(Hughes et al., 2007; Matthews et al., 2007; Soba et al., 2007). To investigate the role of tutl in dendrite self-avoidance, we examined the complex arbor of the class IV da neuron ddaC in tutl mutants and tutl²³ MARCM clones. In control third instar larvae, the high-order branches of ddaC seldom overlap with one another(Fig. 4A′,C′). However, in tutl mutants (Fig. 4B′) and tutl²³ MARCM clones(Fig. 4D′), there were increased numbers of crossing points between isoneuronal dendrites. This was not due to any increases in the total length of dendrites(Fig. 4E), nor was it due to any increases in number of branch termini or changes in the area of the dendritic field (see Fig. S2B,C in the supplementary material). In fact, when the number of crossing points was normalized to total dendritic length(Fig. 4E), tutlmutants and tutl²³ MARCM clones showed significant increases in isoneuronal self-crossing to 150% and 127%, respectively, of controls (Fig. 4F). The milder effect using the MARCM technique may be due to the persistence in ddaC clones of residual Tutl protein from precursor cells. The phenotypes induced in tutl mutants and tutl MARCM clones indicate that Tutl functions within class IV da neurons to prevent overlap between isoneuronal dendrites.

Class IV neurons also exhibit dendritic `tiling', which is the complete and non-redundant coverage of receptive fields by neurons of a similar functional type (Grueber et al., 2002; Parrish et al., 2007). Class IV neurons show tiling with other class IV neurons, even though they overlap extensively with dendrites of class I-III neurons. Like self-avoidance, tiling is thought to be caused by mutual repulsion between dendrites(Parrish et al., 2007), and the processes are related through a common requirement for the nuclear Dbf2-related (NDR) protein kinase Tricornered (Trc) and its putative adaptor protein Furry (Emoto et al.,2004). To determine whether tutl is required for dendritic tiling between class IV neurons, we examined the borders between ddaC and another neighboring class IV neuron (v'ada) for dendritic overlap(Fig. 5A,A′). We found no evidence that tutl is required for tiling, as the branches between these different class IV neurons approached one another but did not overlap in tutl mutants (Fig. 5B,B′).

To test whether tutl is sufficient to inhibit dendrite branching,we overexpressed full-length Tutl in da neurons using a UAS-tutltransgene. In class I neurons (ddaE), branching was unaffected (not shown),suggesting that neurons with small simple arbors and substantial levels of endogenous Tutl along their dendrites are unaffected by adding more Tutl. We then tested the effects of UAS-tutl on the large and highly complex dendritic arbors of the class IV da neuron ddaC(Fig. 6A). In contrast to class I ddaE neurons, overexpression of Tutl in ddaC neurons inhibited dendrite branching (Fig. 6B,C).

One idea consistent with the branching and self-avoidance defects caused by tutl mutations is that Tutl could promote repulsion between isoneuronal dendrite branches. Repulsion could conceivably induce branch collapse in class I da neurons, or could steer dendrites away from one another in class IV da neurons. To explore whether Tutl is sufficient to repel dendrites away from one another, we exploited the fact that the dendrites of neurons of different classes overlap with one another in the body wall. We used the driver C161-GAL4 to overexpress UAS-tutl in class I, class II and class III neurons simultaneously, but we found no evidence for Tutl-induced repulsion among dendrites of different classes (see Fig. S3 in the supplementary material).

Mutations in trc cause excessive branching of class I ddaE neuron dendrites (Soba et al., 2007)and defects of dendrite self-avoidance, particularly at distal branches of the complex arbors of class IV neurons (Emoto et al., 2004). Because mutations of trc and tutldisplay these similar features, it is possible that they could act in a common molecular pathway to govern the processes of branch inhibition (class I) and self-avoidance (class IV). We looked for genetic interactions in animals doubly heterozygous for the null mutations trc¹ and tutl²³, but there were no enhanced branching (class I) or self-avoidance (class IV) defects compared with controls heterozygous for tutl²³ alone (see Table S1 in the supplementary material).

In tutl mutants, many isoneuronal dendritic branches still avoid one another appropriately (Fig. 4B,D), suggesting that Tutl is part of a multi-component system that ensures the proper distribution of dendrites over receptive territories. One component of this self-avoidance system involves the diverse family of receptors encoded by the Dscam gene. Dscam mutants show defects of self-avoidance in all four da neuron classes(Hughes et al., 2007; Matthews et al., 2007; Soba et al., 2007). We examined self-avoidance in class IV neurons, where both genes are required,but found no evidence for genetic interactions between tutl²³ and Dscam³³, a strong mutant allele of Dscam (Hummel et al.,2003) (data not shown).

Class-specific patterns of da neuron dendrite morphogenesis are regulated by key transcription factors. For example, mutations of the genes encoding the transcription factors Abrupt (Ab) or Spineless (Ss) cause ectopic branching in class I neurons (Kim et al.,2006; Li et al.,2004; Sugimura et al.,2004), resembling the effects we have found for tutlmutants. To test whether Tutl expression in class I neurons is regulated by Ab or Ss, we examined Tutl expression in ab^k02807 or ss^D115.7 mutants with immunochemistry. Both of these mutant alleles are known to cause defects in dendrite morphogenesis, yet we found no obvious changes in Tutl immunoreactivity (see Fig. S4A-B″ in the supplementary material). Overexpression of Tutl in class IV neurons inhibits dendrite branching, as do mutations in genes encoding the transcription factors Knot (also known as Collier)(Crozatier and Vincent, 2008; Hattori et al., 2007; Jinushi-Nakao et al., 2007)and Cut (Grueber et al.,2003). To explore the possibility that Tutl expression is normally suppressed to endogenous levels in class IV da neurons by Knot or Cut, we examined Tutl protein expression in ddaC neurons in kn^KN2mutants and cut^c145 MARCM clones. Changes of Tutl immunoreactivity were not detected in da neurons in either case (see Fig. S4C-D″ in the supplementary material).

Tutl and its mammalian orthologs have a conserved ectodomain comprising five Ig-like and two FnIII domains. They also have lengthy but divergent cytoplasmic tails. To begin to investigate the molecular basis for Tutl function in vivo, we wondered whether the cytoplasmic tail of Tutl was dispensable for its function in dendrite morphogenesis. Testing this in class IV da neuron self-avoidance is complicated by the gain-of-function effect of Tutl in ddaC (i.e. branch inhibition). Instead, we examined class I da neurons where Tutl has no gain-of-function effect. We assessed the requirement for the cytoplasmic tail in a rescue assay for branch inhibition in the class I neuron ddaE. To do this, we used GAL4²²¹ to specifically express either full-length Tutl or a truncated form lacking the cytoplasmic tail(TutlΔcyto) in class I neurons of tutl mutants(Fig. 7; see also Fig. S5 in the supplementary material). The full-length form of Tutl fully restored the number of branch termini per cell (ddaE) to wild-type levels(Fig. 7B,D), providing further confirmation that the phenotype observed in tutl mutants was specifically due to the loss of Tutl. Importantly, we found that TutlΔcyto was equally capable of rescuing tutl mutants(Fig. 7C,D), which indicates that the cytoplasmic tail of Tutl is indeed dispensable for dendrite branch inhibition in class I da neurons.

Hypomorphic mutants of tutl have behavioral defects(Bodily et al., 2001), but strong alleles are lethal. Because Tutl is expressed broadly in the nervous system (Bodily et al., 2001),we used the neural-specific driver elav^C155-Gal4 to test whether full-length Tutl could rescue viability in tutl²³/tutl¹⁰⁸⁵ mutants, and whether the cytoplasmic tail was also dispensable for viability. Rescue with full-length Tutl was nearly complete, as 53 of an expected 57 (93%) mutants were rendered viable. The rescued larvae showed normal behavioral responses to tactile stimuli, and could right themselves when overturned (data not shown). By contrast, the rescue with TutlΔcyto was only partial, with 30 viable adults of an expected 152 (20%). Similar results for both full-length Tutl and TutlΔcyto were observed for other tutl mutant genotypes,including tutl²³ homozygotes (data not shown). Together,the data indicate that although the cytoplasmic tail is dispensable for the role of Tutl in limiting dendrite branching in da neurons, it is required for Tutl function in other cell types within the nervous system.

Dendrite branching and self-avoidance are two important cellular mechanisms that shape the receptive fields of neurons during development. Here, we have investigated the role of Tutl in these processes using the da sensory neurons of Drosophila, an excellent system in which to study dendrite arborization at a single cell level in vivo. Tutl is a member of the Tutl/Dasm1/IGSF9 family of evolutionarily conserved transmembrane proteins. We have found that Tutl inhibits excessive branch formation in neurons with simple dendrites (class I), and contributes to the processes that prevent crossing of isoneuronal dendrite branches in neurons with complex arbors(class IV), which demonstrates that Tutl influences the architecture of dendrites and their coverage of receptive territories. In contrast to our results for class I and class IV neurons, our MARCM studies found no evidence of a cell-autonomous role for Tutl in class II or class III neurons, despite detectable Tutl expression in their cell bodies. The reasons for a lack of apparent effects on class II or class III da neuron dendrites in tutlMARCM clones remain unclear. Sufficient Tutl protein, inherited from precursors, could have remained in MARCM clones to promote normal outgrowth. Alternatively, there might be no role for Tutl in these cells. Nevertheless,it is clear from our results for class I and class IV da neurons that Tutl is required for the arborization of dendritic trees with dramatically different complexity.

Tutl cell-autonomously inhibits dendrite branching in vivo, providing a means by which da neurons with the simplest architecture suppress the formation or stabilization of supernumerary dendrite branches during development. We observed a clear increase in the number of second and third order branch points on tutl mutant ddaE neurons. This finding suggests that tutl regulates branching only at certain locations along the growing arbor, perhaps by inhibiting branch additions or promoting branch retractions.

The tutl phenotype is distinct from that of mutants of Neuroglian (Nrg), which also encodes a cell surface IgSF protein that affects dendrite branching. Loss of Nrg reduces the number of branches on the dendritic arbors of class I da neurons, and increases branching along their axons, suggesting a role for Nrg in correctly distributing neurites but not as a branching inhibitor(Yamamoto et al., 2006). The tutl mutant phenotype is also distinct from the dendrite overgrowth phenotype observed in mutants of the IgSF receptor Robo(Dimitrova et al., 2008), or of the cadherin Flamingo (also known as starry night)(Gao et al., 2000; Kimura et al., 2006; Sweeney et al., 2002). In vertebrate systems, no recognition molecules have yet been shown to inhibit dendrite branching in vivo. However, it is noteworthy that inhibition of axon branching has been demonstrated in the chick visual system, where inappropriate arborization of retinal ganglion cell (RGC) axon terminals is thought to be inhibited by EphA (Yates et al., 2001) and Ryk (Schmitt et al., 2006) receptors. In zebrafish, RGC axons are inhibited from branching by Robo2 (Campbell et al.,2007), an IgSF protein with which Tutl shares homology.

After Dscam, Tutl is the only cell surface protein that has been shown to be required for dendrite self-avoidance in either invertebrates or vertebrates. As in Dscam mutants, the dendrites of tutlmutant neurons cross one another with increased frequency, leading to uneven coverage of the receptive field. Unlike Dscam, which promotes self-avoidance in all four da neuron classes (Hughes et al., 2007; Matthews et al.,2007; Soba et al.,2007), Tutl does so only in the highly complex arbors of class IV neurons. We observed no genetic interactions between Dscam and tutl mutations, and have yet to find any evidence that Dscam and Tutl could act in a common molecular pathway to control dendrite self-avoidance. Future studies could reveal whether and how these seemingly distinct pathways converge but, based on our findings, we speculate that the molecular mechanisms ensuring dendrite self-avoidance will prove to be more complex than is appreciated currently.

Neither Tutl nor Dscam affect dendritic tiling among neurons of a similar functional type, illustrating that self-avoidance and tiling are likely to be mediated by distinct recognition molecules on the surfaces of dendrites.

The full-length form of Tutl is a transmembrane protein with a five Ig/two FnIII ectodomain and a cytoplasmic tail, which suggests that it could act as a signaling receptor. Alternative splicing also gives rise to a membrane-tethered form that lacks the cytoplasmic tail(Bodily et al., 2001). This suggests that Tutl could also function as a membrane-bound ligand for an unknown receptor or, alternatively, as a co-receptor in a multiprotein receptor complex. These possibilities are not mutually exclusive, because Tutl could conceivably act as a ligand or a co-receptor in one cellular context,and as a signaling receptor in another. We found that the cytoplasmic tail was completely dispensable for the inhibition of dendrite branching in class I da neurons. This is consistent with a model in which Tutl acts as a ligand or a co-receptor in dendrites. By contrast, we found that the cytoplasmic tail was required to fully rescue viability in tutl mutants, suggesting that Tutl acts as a signaling receptor in this context.

It is currently unclear how Tutl controls dendrite branching and self-avoidance because our studies have not revealed a connection between Tutl and known regulators of dendrite morphogenesis such as Trc. We sought evidence for genetic interactions between trc and tutl and found none. These results alone cannot exclude the possibility that Trc and Tutl act in a common pathway to govern dendrite branching or self-avoidance, but it is noteworthy that the phenotypes of trc and tutl mutants also show some differences that could suggest they work through independent molecular pathways. Unlike tutl, trc is required for dendritic tiling among different class IV neurons, and tutl mutants do not display the excessive terminal branching in class IV neurons that is characteristic of trc mutants (Emoto et al.,2004).

The transcription factors Abrupt, Spineless, Knot and Cut each regulate patterns of dendrite branching in keeping with tutl mutations or Tutl overexpression (Crozatier and Vincent,2008; Grueber et al.,2003; Hattori et al.,2007; Jinushi-Nakao et al.,2007; Kim et al.,2006; Li et al.,2004; Sugimura et al.,2004). However, in immunohistochemical studies of loss-of-function mutants for these transcription factors, we found it to be likely that Tutl expression is influenced by a regulatory program that is distinct from those involving Abrupt, Spineless, Knot or Cut.

Tutl remains somewhat enigmatic because we have yet to find evidence for a genetic or regulatory connection between tutl and genes with similar mutant phenotypes. Nevertheless, our discovery that Tutl regulates dendrite morphogenesis and the coverage of receptive territories underscores the fact that the molecular mechanisms that underlie dendrite morphogenesis remain incompletely understood. We can only speculate as to why tutl mutants have class-specific effects on dendrite morphogenesis, despite Tutl expression in all da neuron classes. Perhaps an unidentified Tutl-interacting protein,such as a receptor required for Tutl function, might be differentially expressed among da neurons and could thus account for the specificity of the phenotype. Other explanations may also exist. For example, it is possible that our MARCM experiments failed to show cell-autonomous defects in certain da neuron classes (classes II and III) because the requirement for Tutl in these cells was met by perdurance of sufficient Tutl protein inherited from the precursor cells of MARCM clones. Alternatively, Tutl in class II and class III da neurons might function non-cell autonomously to influence neighboring cell types.

It is intriguing that the two processes of branching and self-avoidance are related by a common requirement for tutl. Both phenotypes are consistent with the idea that Tutl promotes repulsion, perhaps between isoneuronal dendrite branches, or between dendrites and the substrata along which they grow. However, there is no direct evidence at this time for a repulsive role for Tutl. Simultaneous overexpression of Tutl in different da neuron classes was insufficient to induce branch repulsion among their dendrites. Together with our rescue experiments showing the dispensability of the cytoplasmic tail for dendrite branching, these data suggest that Tutl could function as a ligand or a co-receptor in complexes with one or more unidentified proteins at the cell surface. Such proteins might not be expressed in all da neurons, which could explain why tutl mutations do not affect da neuron classes II and III, and why Tutl cannot induce repulsion when overexpressed in overlapping neurons of Classes I-III. The tutl mutant phenotypes remain the strongest evidence of a repulsive role for Tutl, and it is likely that direct evidence for repulsion must await the identification of the relevant Tutl-interacting proteins.

If it is true that Tutl mediates repulsion, we speculate that the nature or degree of that repulsion could be influenced by the size of the dendritic arbor, leading to class-specific effects. Class I dendrites remain relatively small with Tutl protein distributed along the entire arbor, where Tutl-mediated repulsion could promote the collapse of transient interstitial branches that are known to extend during development(Gao et al., 1999) (branch inhibition). In large class IV arbors where Tutl is distributed more sparingly, Tutl-mediated repulsion could be one part of a multi-component system to redirect isoneuronal branches away from one another (self-avoidance)and thereby ensure proper distribution of dendrites over receptive territories(Hughes et al., 2007; Matthews et al., 2007; Soba et al., 2007). In this way, neurons of different classes could employ a common repulsive mechanism involving Tutl to sculpt arborization patterns of dramatically different complexity.

Our findings that Tutl inhibits dendrite branching in Drosophilacontrast with initial observations in cultured rodent neurons, in which RNAi-knockdown experiments suggested that the Tutl ortholog Dasm1 was required to promote dendritic outgrowth (Shi et al., 2004). However, it was recently argued that these RNAi findings were due to off-target effects(Mishra et al., 2008). The role of Dasm1 in mammalian dendrite morphogenesis is currently unclear, as Dasm1 knockout mice have no observable dendritic defects(Mishra et al., 2008). However, the possibility has been raised that Dasm1 function in dendrites is redundant with the function of Igsf9b, a closely related protein that is coexpressed in the developing hippocampus, the expression of which is unaltered in the brains of Dasm1 knockout mice(Mishra et al., 2008). Loss-of function studies for both Dasm1 and Igsf9b should reveal whether Tutl-like proteins in mammals share with Tutl an evolutionarily conserved role in dendrite morphogenesis.