Development of olfactory projection neuron dendrites that contribute to wiring specificity of the Drosophila olfactory circuit

The antennal lobe (AL) of Drosophila is the first olfactory processing center in which olfactory input and output are spatially organized into distinct channels via glomeruli to form a discrete neural map. In each glomerulus, the axons of a single type of olfactory receptor neurons (ORNs) synapse with the dendrites of a single type of projection neurons (PNs). The AL is an ideal place to study how the wiring specificity between specific types of ORNs and PNs is established during development. During the past two decades, the involvement of diverse molecules in the specification and patterning of ORNs and PNs has been reported. Furthermore, local interneurons—another component of glomeruli—have been recently catalogued and their functions have been gradually dissected. Although there is accumulating knowledge about the involvement of these three cell types in the wiring specificity of the olfactory system, in this review, we focus especially on the development of PN dendrites.


INTRODUCTION
Drosophila is a versatile model organism that has been used in biological research for over a century to study a broad range of phenomena because of its short life span and cost-effectiveness.Moreover, sophisticated genetic tools and methods allow us to uncover molecular mechanisms of biological events, most of which can be applied to other organisms, including mammals, because of a high degree of homology.In this review, we focus on the recent findings in the development of olfactory projection neurons (PNs) that contribute to the wiring specificity of the Drosophila olfactory system; for more comprehensive reviews of the Drosophila olfactory system, see, for example, Jefferis and Hummel (Jefferis and Hummel, 2006), Brochtrup and Hummel (Brochtrup and Hummel, 2011) and for other recent topics in Drosophila, see other reviews (Kondo, 2014;Hakeda-Suzuki and Suzuki, 2014;Niwa and Niwa, 2014).
The olfactory system is indispensable for organisms to obtain information about foods, predators, and potential mates.From insects to mammals, olfactory input is spa-tially organized into distinct sensory channels that form a discrete map in the brain; the olfactory sensory neurons (OSN) that express the same odorant receptors converge to a common synaptic target, the glomerulus of the antennal lobe (AL; Fig. 1A, dotted circle) in insects (Vosshall et al., 2000;Couto et al., 2005;Fishilevich and Vosshall, 2005) or the olfactory bulb in mammals (Vassar et al., 1994;Mombaerts et al., 1996).Thus, the targets of OSN axons are organized as discrete units and form a spatial map (Luo and Flanagan, 2007).It is of great interest to uncover how this discrete map formation in the AL is established during development.In mammals, a single olfactory receptor (OR) gene that is selected to be expressed in each OSN has an instructive role in axon targeting (Mombaerts et al., 1996;Imai et al., 2006;Nakashima et al., 2013).OSN axons form glomerularlike structures first, and dendrites of secondary mitral and tufted cells participate in the formation of glomerular structure subsequently (Malun and Brunjes, 1996;Blanchart et al., 2006).In Drosophila, however, the dendrites of PNs (Fig. 1, A and B, green) prepattern the developing adult AL before the pioneering axons of olfactory receptor neurons (ORNs; Fig. 1, A and B, blue) reach the AL (Fig. 1B) (Jefferis et al., 2004).Traditionally, Drosophila olfactory sensory neurons (OSNs) are called Edited by Hiroshi Iwasaki * Corresponding author.E-mail: tchihara@mol.f.u-tokyo.ac.jp olfactory receptor neurons, ORNs.Furthermore, OR gene expression in ORNs begins after their axons have established type-specific connections with PNs and/or LNs.Thus the assembly sequence of OSNs/ORNs and secondary neurons is different between mammals and Drosophila.Here, we discuss recent findings on the genetic control of wiring specificity in the AL, especially focusing on the development of PNs that predominantly Fig. 1.Organization and specification of projection neurons.(A) Schematic view of the Drosophila olfactory system.Left: Odorants are perceived at the antennae and maxillary palps (light blue), which are innervated by olfactory receptor neuron (ORN, blue) dendrites.Olfactory information is transmitted from the ORNs to projection neurons (PNs, green) at the antennal lobe (AL, dotted gray circle).PNs relay olfactory information from the AL to higher olfactory neurons.Right: a magnified image of the AL.The AL (dotted circle) consists of about 50 glomeruli (gray circles) made up of synaptic contacts between 3 cell types: ORNs (blue), PNs (green), and local interneurons (LNs, red).Most of the dendrites of a single PN innervate into a specific glomerulus out of 50 glomeruli in the AL, while axons elongate and exhibit stereotypical branching at the mushroom body (MB) and the lateral horn (LH).(B) Developmental events in the AL formation.PN dendrites start to innervate the developing adult AL at around 0 h after puparium formation (APF).By 18 h APF, PN dendrites occupy the coarse subregions of the AL according to their future glomerular position, while pioneering antennal ORNs have just reached the edge of the AL.At 32 h APF, pioneering axons from the maxillary palp ORNs reach the AL.Glomerular structures are visible from around 50 h APF.Each glomerulus is ensheathed by glia in the end.(C) PNs are mainly classified into 4 lineages according to the neuroblasts from which they are produced; anterodorsal (ad-, magenta), lateral (l-, blue), ventral (v-, green), and lateroventral (brown) lineages.Glomeruli in the AL can be uniquely identified by their position, size and shape.All identifiable glomeruli in the AL are shown in four anterior-to-posterior focal sections.The glomeruli targeted by uniglomerular ad-PNs, l-PNs, and v-PNs are labeled in magenta, blue, and green, respectively.The glomeruli with gray labels have not found their corresponding uniglomerular PNs.(D) Specification of ad-PNs produced from the ad-neuroblast.Acj6 (magenta) is expressed in all ad-PNs.From the ad-neuroblast, 40 types of ad-PNs are produced in an invariant sequence.The birth order strictly regulates the fate of each post-mitotic neuron, including programmed cell death (dotted circles).Krüppel (Kr) and Chinmo have been suggested to regulate the temporal fate change.Furthermore, the cell fate of ad-PNs is specified by Numb, which suppresses Notch signaling, and the other sibling undergoes programmed cell death induced by Notch (N) signaling (dotted circles).In Drosophila, about 50 glomeruli in the AL can be uniquely identified by their size, shape and relative position, and have been catalogued in atlases (Fig. 1C) (Laissue et al., 1999).Each glomerulus consists of the synaptic contacts between three cell types: afferent axons of ORNs, efferent dendrites of PNs, and the processes of LNs, which are ensheathed by glial processes (Jefferis and Hummel, 2006).Each side of the brain contains about 1,300 ORNs whose dendrites innervate the olfactory sensilla in the antennae or the maxillary palps (Fig. 1A, left).Drosophila antennae are covered with three main morphological types of sensilla-basiconic, trichoid, and coeloconicwhich house the ciliated dendrite endings of 1-4 ORNs (Couto et al., 2005;Fishilevich and Vosshall, 2005).Maxillary palp, basiconic, and trichoid ORNs express OR genes, while coeloconic ORNs express ionotropic glutamate receptor (IR) genes, another family of insect olfactory receptors (Benton et al., 2009).Axons of ORNs expressing the same type of OR or IR gene converge to a single glomerulus.To receive the direct input from one ORN type, most of the PNs send dendrites into a single glomerulus (uniglomerular PNs) and the other PNs with diverse morphologies send dendrites to multiple glomeruli (atypical multiglomerular PNs) (Lai et al., 2008;Yu et al., 2010;Lin et al., 2012).The cell bodies of 150-200 PNs are arranged in 4 distinct clusters around the AL (Ito et al., 2013;Yu et al., 2010;Lin et al., 2012).The PNs relay olfactory information from the AL to higher olfactory centers (the mushroom body and the lateral horn; Fig. 1A) where they connect with higher order neurons (Marin et al., 2002;Wong et al., 2002;Tanaka et al., 2004;Jefferis et al., 2007;Lin et al., 2007;Caron et al., 2013).LNs ramify their processes through large areas of both the ipsi-and contralateral AL and modify the transmission of the olfactory information at the AL.Although recent findings have systematically characterized and categorized LNs at the adult stage (Chou et al., 2010), little is known about the developmental window of patterning of LN processes compared to ORN axon and PN dendrite targeting.Taken together, the neural map formation in the AL of Drosophila is an ideal model to study the wiring specificity because of two aspects: it has about 50 identifiable specific targets, and the targeting patterns of PN dendrites and ORN axons are precisely and genetically controlled.In this review, we focus on the development of uniglomerular PNs, in which most of studies were performed.Technical terms used in this review are listed in Table 1.

SPECIFICATION OF PNs
Visualization of neurites at a single neuron level is required to observe the precise targeting patterns and connections of over 3,000 neurons (~2,600 ORNs from both sides, 150-200 PNs, ~200 LNs) in each AL.A clonal analysis, called mosaic analysis with a repressible cell marker (MARCM), allows one to label and manipulate the genotype of a single PN (single-cell clone) or a group of PNs that share the same lineage (neuroblast clone) in vivo (Lee and Luo, 1999).Moreover, the improved version of MARCM, twin-spot MARCM, enables the labeling of sister clones derived from a common progenitor in different colors (Yu et al., 2009).The use of these genetic clonal methods has revealed that PNs were prespecified by lineage and birth order.PNs are classified into 4 lineages (anterodorsal, lateral, ventral, and lateroventral lineages) depending on which neuroblasts they are generated from (Ito et al., 2013).Anterodorsal (ad-) PNs are generated from the ad-neuroblast, lateral (l-) PNs from l-neuroblast, ventral (v-) PNs from v-neuroblast, and lateroventral (lv-) PNs from lv-neuroblast.Land v-neuroblasts generate other types of cells such as LNs in addition to PNs, while ad-neuroblast generates only ad-PNs.The lv-neuroblast was recently identified, and its lineage analysis has not yet been performed.Although the details of how different lineage identities are acquired are still poorly understood, transcription factors such as Acj6 and Drifter seem to confer lineage iden- tity for the targeting specificity of PN dendrites (Komiyama et al., 2003).Acj6 is expressed in all ad-PNs and is required for proper dendrite targeting of ad-PNs.PNs homozygous for acj6 mutation fail to innervate glomeruli innervated by wild-type ad-PNs.Moreover, Acj6 misexpression in l-PNs causes dendrite innervation into ectopic glomeruli that are targets of ad-PNs.As for l-PNs, Drifter is expressed from their precursors and required for proper dendrite targeting.Thus Acj6 and Drifter are required for the specification of dendrite targeting of ad-and l-PNs, respectively.However, recent findings revealed that Acj6 is also expressed in some atypical PNs in l-lineage (Lai et al., 2008;Lin et al., 2012), suggesting that the targeting specification by Acj6 is limited inside uniglomerular PNs.Inside lineages, the birth order strictly dictates the fate of each postmitotic neuron.Generally, in holometabolous insects, neurons are produced in two periods; embryonically born neurons are required for the wiring of larval circuitry, which may be remodeled during metamorphosis to contribute to the adult circuitry, and larval born neurons are adult-specific (Ito and Hotta, 1992).PNs that innervate the adult AL are produced both at embryonic and larval stages (Marin et al., 2005;Jefferis et al., 2001).From a complete lineage analysis among neurons derived from the ad-neuroblast, it was revealed that the ad-neuroblast serially generates 40 types of PNs (Fig. 1D) (Yu et al., 2010).Eighteen types of PNs are born during the embryonic stage, and 22 additional types of PNs are generated in the early larval stage.Embryonic born PNs that participate in the larval olfactory system undergo stereotyped pruning of their dendrites and axon terminal branches locally during early metamorphosis by receiving the nuclear hormone ecdysone, and persist in the adult olfactory circuit (Marin et al., 2005).Although many types of PNs are born from a common progenitor serially, the birth order can specify the fate of each post-mitotic neuron, including the fate of programmed cell death.Furthermore, every neuron type has a unique and invariant cell number.Thus the fate of individual neurons in the ad-PN lineage is tightly regulated on the basis of the time window of their birth.Another study showed that the birth order is also important in l-and v-lineages that produce not only PNs but also other types of neurons, such as LNs (Lai et al., 2008;Lin et al., 2012).To ensure the temporal specification of the types of neurons, various transcription factors can also be sequentially and transiently expressed in neuroblasts that produce PNs, since the transcription factor cascade (early to late: Hunchback → Krüppel → Pdm → Castor) is known to specify the first several neurons across neuroblast lineages in the embryonic ventral nerve cord (Isshiki et al., 2001).Indeed, Krüppel governs the temporal fate of VA7l ad-PNs that target VA7l glomerulus; loss of Krüppel causes the loss of VA7l PNs and VA7l precursors yields the next-born VA2 PNs instead (Fig. 1D) (Kao et al., 2012).Since the loss of Hunchback, Pdm, or Castor did not cause any loss of PNs, it seems Krüppel might repress the next temporal identity factor in the transcriptional cascade that consists of different factors from the embryonic ventral nerve cord.Chinmo, another transcriptional factor, affects the temporal identity of PNs in two time windows.Loss of Chinmo causes fate change of DM4, DL5, and VM3(a) PNs to DM3 PNs and VM3(b), DL4, DL1, DA3, and DC2 PNs to D PNs (Fig. 1D) (Kao et al., 2012).These results suggest that Chinmo operates in newborn neurons within two time windows to suppress the following temporal fate.
Several other transcription factors are also involved in the control of precise dendrite targeting of PNs; Islet, Lim1, Cut, Squeeze, Lola, Empty spiracles, Jing, homothorax, and extradenticle are expressed in all or specific types of PNs and support the proper targeting of PN dendrites (Komiyama and Luo, 2007;Spletter et al., 2007;Das et al., 2008;Lichtneckert et al., 2008;Nair et al., 2013;Ando et al., 2011).Thus, distinct PN types are at least partially defined by the combinatorial expression of transcription factors that regulate their targeting specificity.
Lastly, Notch/Numb signaling also contributes to the fate specification of PNs (Lin et al., 2010(Lin et al., , 2012)).In adlineage, only one in each pair of postmitotic neurons survives into the adult stage.Strikingly, Notch signaling promotes programmed cell death in the missing sibling in the ad-lineage (Fig. 1D, dotted circles).Moreover, Notch/ Numb is also involved in the specification of PNs in l-and v-lineages; however, their specification mode is different.In l-lineage, where l-neuroblast produces LNs and other types of neurons in addition to l-PNs, inactivation of Notch signal by Numb is required for the specification of l-PNs.In contrast, in v-lineage, Notch signal is required for the specification of v-PNs.

COARSE TARGETING OF PNs
After cell fate is determined during the embryonic and larval stages, PN dendrites initially undergo targeting to the coarse subregions of the developing adult AL at the early pupal stage.In this step, PN dendrites seem to utilize positional information along the AL axes, and molecules governing PN dendrite targeting along the dorsolateral (DL)-ventromedial (VM) and mediolateral (ML) axes have already been reported (Komiyama et al., 2007;Sweeney et al., 2011;Sekine et al., 2013).
Graded expression of the Semaphorin family molecules, Sema-1a/2a/2b contributes to the targeting of PN dendrite along DL-VM axis (Komiyama et al., 2007;Sweeney et al., 2011).Sema-2a/2b are secreted from degenerating larval ORNs and form a continuous high-VM-low-DL gradient in the developing adult AL by 0 h after puparium formation (APF) (Blue in Fig. 2A), while transmembrane Drosophila olfactory projection neurons Sema-1a starts to be expressed in PN dendrites at around 6 h APF and forms a high-DL-low-VM gradient at the developing adult AL (Green in Fig. 2A).Interestingly, Plexin A or B, well-known receptors for Semaphorin, are not involved in the coarse targeting of PN dendrites, and instead, Sema-1a acts as a receptor for Sema2a/2b.Sema-2a/2b binds to Sema-1a and directs the dendrite targeting of adult PNs; the dendrites of PNs expressing high levels of Sema-1a are repelled by larval ORN-derived Sema-2a/2b, and thus target the DL region of the AL.Consistent with this, sema-2a -/-sema-2b -/-double mutants have a phenotype similar to the loss of sema-1a in PN; PN dendrites that originally target DL region of the AL mistarget to VM region, while the dendrites that originally target VM region do not show significant change (Fig. 2B).In addition, Sema-2a overexpression in larval ORNs is sufficient to rescue the VM mistargeting phenotype of sema-2a -/-sema-2b -/-double mutants.Thus, secreted Sema-2a and Sema-2b from degenerating larval ORNs provide instructive spatial cues for the Sema-1a-dependent PN dendrites targeting along the DL-VM axis; those expressing a high level of Sema-1a target the DL region of the AL, whereas those with a low level of Sema-1a target the VM region.
In contrast to Sema-1a/2a/2b, which regulates PN dendrite targeting along the DL-VM axis, Meigo is required for proper dendrite targeting along the ML axis (Sekine et al., 2013).Loss of meigo causes PN dendrite spillover Fig. 2. The coarse targeting of PN dendrites.(A) The graded expression of Sema-1a (green) and Sema-2a/2b (blue) in the developing adult AL (dotted ovals).At 0 h after puparium formation (APF), the larval ORN axons (dark blue) occupy the larval AL (blue).As pupal development proceeds from 0 to 16 h APF, larval ORN and larval AL degenerate and secrete Sema-2a/2b, which is distributed in a gradient along the dorsolateral (DL)-ventromedial (VM) axis in the developing adult AL.Around the same period, PN dendrites expressing transmembrane Sema-1a, which acts as a receptor for Sema2a/b, innervate the developing adult AL and are repelled from the Sema-2a/2b gradient and form the counter-gradient of Sema-1a in the developing adult AL; PNs with the high Sema-1a expression target the DL region of the AL, whereas PNs expressing low Sema-1a target the VM region.(B) Loss of Sema-1a in PN or sema-2a -/- sema-2b -/-double mutation causes the same targeting defects; the PN dendrite that originally targets the DL glomerulus (dotted green) of the AL is shifted to the VM side (green).(C) Loss of Meigo causes mistargeting of the PN dendrite that originally targets the lateral side of the AL (dotted green) to the medial side and also causes spillover from a glomerulus (green).from the target glomeruli and shifts the PN dendrites that originally target the lateral side of the AL to the medial side (Fig. 2C), suggesting the positional information along the ML axis exists in the developing adult AL.Meigo is a unique nucleotide sugar transporter (NST) localized to the endoplasmic reticulum (ER) and transports nucleotide sugars from the cytoplasm to the lumen of the ER, which is an essential step for proper protein glycosylation.Interestingly, overexpression of ephrin, a well-known axon guidance molecule, drastically suppresses the dendrite spillover phenotype of meigo -/- PNs.Indeed, Meigo positively regulates the Nglycosylation of Ephrin that is essential for its proper function-glomerular refinement of PN dendrites.Considering the molecular nature of Meigo, an ER-localized NST, more membrane-bound or secreted molecules, in addition to Ephrin, are likely to be regulated by Meigo.Further investigation would identify molecules that specify PN dendrite targeting along the ML axis, which clarifies the molecular mechanism for the coarse targeting of PN dendrites.

GLOMERULAR TARGETING OF PNs
After the coarse targeting of PN dendrites along the AL axes, PN dendrites must select individual specific glomerular targets from about 50 glomeruli.Leucine-rich repeat transmembrane proteins, Capricious and Tartan are involved in this step (Hong et al., 2009).Capricious is differentially expressed and localized at the dendrites of specific subsets of PNs, representing a salt-and-pepper like pattern in the developing adult AL, and cell-autonomously instructs glomerular-specific targeting of PN dendrites (Fig. 3).Loss of Capricious causes mistargeting of Capricious-positive PN dendrites to the glomeruli originally targeted by Capricious-negative PNs.Furthermore, misexpression of Capricious in Capricious-negative PNs instructs dendrites to target glomeruli originally targeted by Capricious-positive PNs.Tartan, which shares 67% sequence identity in its extracellular domain with Capricious, has partial redundant function with Capricious.Thus, Capricious and Tartan instruct the targeting of PN dendrites to discrete glomeruli.
Forward genetic screens and candidate approaches have uncovered many other molecules that are cellautonomously required for the targeting of PN dendrites to specific glomeruli-Rpd3 (major histone deacetylase), Bap55 (chromatin remodeling factor), Pasha and Dicer-1 (members of the microRNA processing pathway), Verloren (small ubiquitin-like modifier protease), cohesin SA and SMC1 (multisubunit complex required for sister-chromatid cohesion during mitosis and meiosis), and Unc-51 (regulator of vesicle trafficking along microtubule)-are all reported to regulate proper dendrite targeting of PNs (Tea et al., 2010;Tea and Luo, 2011;Berdnik et al., 2008Berdnik et al., , 2012;;Schuldiner et al., 2008;Mochizuki et al., 2011).In addition to these intrinsic machineries, interactions between PNs and other cell types also contribute to proper PN dendrite targeting and the synaptic matching between dendrites of specific type of PNs and axons of corresponding ORNs.Next, we would like to discuss the molecules involved in PN-ORN matching and LN-PN interaction.

PN-ORN MATCHING
While studying the function of Down syndrome cell adhesion molecule (Dscam) in PNs, the existence of a synaptic-partner-matching system was implied (Zhu et al., 2006).Overexpression of Dscam in a subset of PNs caused shifts of target glomeruli.Interestingly, the partner ORN axon also followed this shift and the original ORN-PN connection was maintained.Recently, Teneurinm (Ten-m) and Teneurin-a (Ten-a) were found to contribute to this synaptic-partner-matching system (Hong et al., 2012).Ten-m and Ten-a are expressed in a salt-and-pepper like pattern in the developing adult AL.Ten-m and Tena are differentially expressed in both matching PN and ORN types (Fig. 4A-1).For example, the axons of Or67d ORNs make connection with the dendrites of DA1 PNs.Both Or67d ORNs and DA1 PNs have a high expression of Ten-a, while Ten-m is highly expressed with other ORN-PN pairs.Loss of Ten-a in DA1 PNs causes mismatching of DA1 PNs with Ten-m-expressing ORNs (Fig. 4A-2).Furthermore, overexpression of Ten-m in Ten-aexpressing PNs causes mismatching of these PNs with Ten-m-expressing ORNs (Fig. 4A-3), while Ten-m overexpression in Ten-m-expressing PNs does not cause any defects in PN-ORN matching.Similar results are observed when Ten-a is overexpressed in Ten-m-or Tena-expressing PNs.Interestingly, when Ten-m is expressed in both small subsets of PNs and ORNs that do not make connections in the wild type, ectopic connections were observed in vivo, indicating that Ten-m promotes PN-ORN homophilic attractions.Thus, Ten-m and Ten-a mediate mutual selection and direct matching between individual pre-and postsynaptic partners.Since PN dendrites innervate the developing adult AL prior to the arrival of ORN axons, ORN axons that express Ten-m/a may actively search for the proper partner through contacting multiple PN dendrites.

LN INFLUENCES ON PN DENDRITE TARGETING
Although research to reveal the development and function of LNs has just begun, the contribution of LNs to PN dendrites targeting has been recently suggested (Melnattur et al., 2013).In the Dichaete mutant brain, PN dendrites target different or ectopic glomeruli.Interestingly, Dichaete expression is not detected in any mature or developing PNs, and PNs homozygous for the Dichaete mutation do not show any defects.Furthermore, no Dichaete expression is detected in developing antennal ORNs.Indeed, Dichaete expression is detected in several types of LNs located lateral and adjacent to the  ,VA1d,VA1lm,DC3,and DA3).At 48 h after puparium formation (APF), when individual glomeruli just become identifiable, elevated Ten-a and Ten-m expression is evident in selected glomeruli.Ten-a is highly expressed at DA1 and DA3 glomeruli, while Ten-m is highly expressed in VA1d, VA1lm, and DA3 glomeruli.Matching PNs and ORNs express the same level of Ten-a and Ten-m.For example, DA1 PNs and Or67d ORNs express a high level of Ten-a but a low level of Ten-m.(2) When ten-a is knocked down in DA1 PNs that originally expresses a high level of Ten-a, they mismatch with Or47b ORNs that do not express a high level of Ten-a.AL.From these observations, it is suggested that LNs affect PN dendrite targeting non-cell-autonomously.Since LN morphology does not seem to be defective in Dichaete mutants; this indicates that LNs participate in a signaling pathway that influences PN dendrite targeting (Fig. 4B).The actual signaling molecule responsible for intercellular effects from LNs to PNs remains to be characterized.

FIELD FORMATION
As stated above, PN dendrite targeting seems to occur in a stepwise fashion.Initially, PN dendrites undergo coarse targeting along several axes in the developing adult AL.Then various molecules and strategies sustain PN type-specific dendrite sorting to specific glomeruli.In the end, PN dendrites ramify to form the terminal arborization in the glomerulus, and the boundaries between glomeruli are sharpened through dendritedendrite interaction.Glycyl-tRNA synthetase (GARS) and three cell-surface molecules (Dscam, DN-cadherin and Ephrin) were shown to regulate the final field formation of PN dendrites (Chihara et al., 2007;Zhu and Luo, 2004;Zhu et al., 2006;Sekine et al., 2013).
Mutation in gars gene that is essential for protein translation exhibits strikingly specific phenotypes: individual PNs appear to have normal dendrite and axon growth and guidance, but they fail to elaborate and maintain their terminal arborizations at their targets (Chihara et al., 2007).This implies that dendrite and axon terminal arborization are much more sensitive to perturbation of protein translation than the growth and guidance of their stalks.Interestingly, the Drosophila gars gene (as its yeast and human homologs) encodes both cytoplasmic and mitochondrial GARS proteins.Further molecular genetic analyses of gars provide a comprehensive view of the function of protein translation in different neuronal compartments.For instance, the terminal arborization defects seen during development are mainly a result of cytoplasmic protein translation.On the other hand, mitochondrial protein translation is essential for the maintenance of arborization in aging animals, with a highly preferential requirement in dendrite but not axon compartments.Importantly, the gars gene is the Drosophila homolog of human causal gene for Charcot-Marie-Tooth disease type 2D (CMT2D).Using Drosophila PNs as a model, it has been shown that wild-type human GARS gene can functionally rescue the Drosophila gars mutant phenotypes, and that human GARS with CMT2D disease-causing point mutations significantly reduce this rescuing capability, suggesting that the human disease mutation is likely a loss-of-function mutation.These provide novel insight into the pathogenesis of CMT2D, and other neurological diseases related to mitochondrial function and protein translation.At the same time, this study also presents the possibility of PNs as a model system to understand the mechanisms underlying neurological diseases.
Furthermore, Dscam controls the elaboration of the dendrite field, since loss of Dscam in PNs or LNs causes a marked reduction in their target field size (Zhu et al., 2006).DN-cadherin is ubiquitously expressed in the developing adult AL throughout its development, and mediates dendrite-dendrite interactions between PNs to restrict dendrites to single glomeruli (Zhu and Luo, 2004).Loss of DN-cadherin causes a spill over phenotype; the dendrites target the correct glomerulus, but the targeting is diffuse and partially innervates neighboring glomeruli.Moreover, wild-type dendrites also spill over beyond their glomerular target when surrounded by PN dendrites that are homozygous for DN-cadherin mutation.Thus, DN-cadherin mediates dendrite-dendrite interactions between PNs, which is required for the refinement of PN dendrites to the final field.As described in the section of PN coarse targeting, Meigo-mediated Ephrin function in PNs is also required to restrict the dendrite field size (Sekine et al., 2013).It is crucial to uncover how Ephrin in PNs recognizes and distinguishes PN dendrites in the neighboring glomeruli and the requirement of Eph, a potential binding partner of Ephrin in the dendrite field formation.

CONCLUSION
As we summarized in this review, the development of PNs has been studied intensively; PN specification by lineage and birth order, coarse and glomerular targeting of PN dendrites, synaptic partner matching between PN and ORN, and the dendrite field formation.Although technical improvements and tremendous efforts have been made by many researchers, it would require a great investment of time to uncover how about 150-200 PNs specifically target 50 glomeruli.For example, we still do not know whether there are molecular links among PN specification, coarse and glomerular targeting of PN dendrites, and synaptic partner matching.Inside lineages, PNs seem to be specified by birth order according to the sequential expression of transcription factors in the neuroblast.It might be interesting to identify what kinds of transcriptional factors are expressed other than Krüppel in the neuroblasts and whether Capricious, Teneurin, etc. could be targets of those transcription factors to elucidate the molecular link between PN specification and PN dendrite targeting.In addition to identifying molecules that work in PNs, it would be intriguing to characterize the development and function of LNs in the formation of olfactory circuits.It has been suggested that LNs are diverse in their neurotransmitter profiles, connectivity, and physiological properties (Chou et al., 2010), which might produce an acceptable variation of the olfactory circuit (neural circuit individuality).This individuality in the neural circuit might be beneficial (or could be harmful) for an animal to adapt to environmental changes.Recent advances in genetic methods and resources such as the generation of specific Gal4 drivers, Q-system and CRISPR methods could be helpful to identify identify key molecules that regulate PN dendrite targeting (Potter et al., 2010;Gratz et al., 2013;Kondo, 2014).Identification of additional molecular players, the detailed characterization of LNs, and the investigation of interaction among PNs, ORNs, and LNs would provide a better understanding of the molecular logic in the formation of the Drosophila olfactory circuit.
map in the developing adult AL.
Abbreviations AL Antennal lobe PN Projection neuron OSN Olfactory sensory neuron ORN Olfactory receptor neuron LN Local interneuron OR Olfactory receptor IR Ionotropic glutamate receptor MARCM Mosaic analysis with a repressible cell marker ad-Anterodorsal l-Lateral v-Ventral lv-Lateroventral DL Dorsolateral VM Ventromedial ML Mediolateral APF After puparium formation Sema Semaphorin NST Nucleotide sugar transporter ER Endoplasmic reticulum Dscam Down syndrome cell adhesion molecule Ten Teneurin GARS Glycyl-tRNA synthetase CMT2D Charcot-Marie-Tooth disease type 2D

Fig. 3 .
Fig. 3. Capricious is differentially expressed in different types of PNs.Schematic representation of glomeruli expressing Capricious.Gray glomeruli are innervated by Capricious-positive PNs, while white glomeruli are innervated by Capricious-negative PNs.The glomeruli with broken lines are innervated by PNs whose Capricious expression has not yet been uncovered.

Fig. 4 .
Fig. 4. Interaction among PNs, ORNs, and LNs during development.(A-1) Schematic representation of the expression pattern of Teneurin-a (Ten-a, pink) and Teneurin-m (Ten-m, light blue) in five glomeruli (DA1, VA1d, VA1lm, DC3, and DA3).At 48 h after puparium formation (APF), when individual glomeruli just become identifiable, elevated Ten-a and Ten-m expression is evident in selected glomeruli.Ten-a is highly expressed at DA1 and DA3 glomeruli, while Ten-m is highly expressed in VA1d, VA1lm, and DA3 glomeruli.Matching PNs and ORNs express the same level of Ten-a and Ten-m.For example, DA1 PNs and Or67d ORNs express a high level of Ten-a but a low level of Ten-m.(2) When ten-a is knocked down in DA1 PNs that originally expresses a high level of Ten-a, they mismatch with Or47b ORNs that do not express a high level of Ten-a.(3) Ten-m misexpression in DA1 PNs causes mismatching of DA1 PNs with Or88a and 47b ORNs that express a high level of Ten-m.(B) Model of Dichaete function in LNs for PN dendrite targeting.Dendrite targeting of PNs is defective in Dichaete mutants (right), although Dichaete expression is not observed in PNs.Dichaete is expressed in LNs and seems to be essential for proper PN dendrite targeting.Since LN development does not seem to be drastically altered in Dichaete mutants, LNs could provide a developmental signal to PNs to perform proper targeting (left).
Fig. 4. Interaction among PNs, ORNs, and LNs during development.(A-1) Schematic representation of the expression pattern of Teneurin-a (Ten-a, pink) and Teneurin-m (Ten-m, light blue) in five glomeruli (DA1, VA1d, VA1lm, DC3, and DA3).At 48 h after puparium formation (APF), when individual glomeruli just become identifiable, elevated Ten-a and Ten-m expression is evident in selected glomeruli.Ten-a is highly expressed at DA1 and DA3 glomeruli, while Ten-m is highly expressed in VA1d, VA1lm, and DA3 glomeruli.Matching PNs and ORNs express the same level of Ten-a and Ten-m.For example, DA1 PNs and Or67d ORNs express a high level of Ten-a but a low level of Ten-m.(2) When ten-a is knocked down in DA1 PNs that originally expresses a high level of Ten-a, they mismatch with Or47b ORNs that do not express a high level of Ten-a.(3) Ten-m misexpression in DA1 PNs causes mismatching of DA1 PNs with Or88a and 47b ORNs that express a high level of Ten-m.(B) Model of Dichaete function in LNs for PN dendrite targeting.Dendrite targeting of PNs is defective in Dichaete mutants (right), although Dichaete expression is not observed in PNs.Dichaete is expressed in LNs and seems to be essential for proper PN dendrite targeting.Since LN development does not seem to be drastically altered in Dichaete mutants, LNs could provide a developmental signal to PNs to perform proper targeting (left).