A network of interacting Drosophila cell surface proteins has been defined in which a 21-member IgSF subfamily, the Dprs, binds to a nine-member subfamily, the DIPs. The structural basis of the Dpr-DIP interaction code appears to be dictated by shape complementarity within the Dpr-DIP binding interface. Each of the six dpr and DIP genes examined here is expressed by a unique subset of larval and pupal neurons. In the neuromuscular system, interactions between Dpr11 and DIP-γ affect presynaptic terminal development, trophic factor responses, and neurotransmission. In the visual system, dpr11 is selectively expressed by R7 photoreceptors that use Rh4 opsin (yR7s). Their primary synaptic targets, Dm8 amacrine neurons, express DIP-γ. In dpr11 or DIP-γ mutants, yR7 terminals extend beyond their normal termination zones in layer M6 of the medulla. DIP-γ is also required for Dm8 survival or differentiation. These findings suggest that Dpr-DIP interactions are important determinants of synaptic connectivity (Carrillo, 2015).

This study has defined a network of interacting Drosophila IgSF CSPs in which 21 Dpr proteins bind to 9 DIPs. The structure of the Dpr-DIP complex resembles that of neural and immune cell adhesion complexes. Each of the six dpr and DIP genes examined in this study is expressed by a different subset of neurons in the larval VNC and pupal OL. In the larval neuromuscular system, Dpr11 and its binding partner DIP-γ regulate presynaptic terminal development and neurotransmission. In the pupal OL, they are required for normal formation of synapses between a Dpr11-expressing sensory neuron and a DIP-γ expressing interneuron (Carrillo, 2015).

The crystal structure shows that Dprs and DIPs belong to a group of IgSF CSPs that interact via their N-terminal Ig domains. These include immune cell receptors such as CD2, CD58, JAML, CAR, B7-1, and CTLA-4, and Nectin/Nectin-like (Necl) proteins. The nine Nectin/Necls interact with each other, forming a small network. Although DIPs resemble Nectins/Necls, their closest vertebrate counterpart is the five-member IgLON subfamily, which is also expressed in neurons. Dprs have no clear mammalian orthologs. DIPs and Dprs are distinguished from IgLONs and Nectins in that their interactions are across subfamilies, not within a subfamily. The closest structural homolog of the Dpr-DIP complex is the SYG-1-SYG-2 complex, known to be involved in synapse specification (Carrillo, 2015).

The Dpr-DIP complex has an interface involving no charge pairs, suggesting that binding specificity is encoded through shape complementarity. The Dpr-DIP interaction code may be created by substitution of larger or smaller residues within the binding interface in order to create more or less complementary surfaces between individual interacting Dpr-DIP pairs. This differs substantially from the electrostatic complementarity model, in which receptor-ligand specificity is created primarily through hydrogen bonding interactions and salt bridges. Interestingly, for Dscam homophilic interactions, where each of the many thousands of possible variants binds primarily to itself, both electrostatic and shape complementarity play crucial roles. Each Dscam variant has to find a single binding solution, which is a task that can be solved in many ways. By contrast, the complex cross-reactivity observed for Dpr-DIP interactions may impose restrictions on encoding of specificity that mandate the selection of shape complementarity as the primary mechanism (Carrillo, 2015).

The larval neuromuscular system is a genetic model system for glutamatergic synapses in mammals. In mutants lacking either Dpr11 or DIP-γ, NMJs contain many small clustered boutons called satellites. The satellite bouton phenotypes are rescued by either pre- or postsynaptic expression of the proteins. mEPSP amplitude and frequency are increased to similar extents in dpr11 and DIP-γ mutants. These data, together with the fact that the two loci genetically interact, indicate that the two proteins have linked functions, and suggest that the phenotypes are due to loss of Dpr11-DIP-γ adhesion complexes (Carrillo, 2015).

BMPs are trophic factors for mammalian neurons, and retrograde BMP signaling controls NMJ arbor growth in Drosophila. Satellites are observed in mutants in which BMP signaling is upregulated. Consistent with this, presynaptic pMad staining, which reports on the magnitude of the BMP signal, is increased in dpr11 mutants, and dpr11 and DIP-γ interact with genes encoding BMP signaling components (Carrillo, 2015).

Each dpr and DIP examined is expressed in a unique subset of neurons that project to specific layers in the OL neuropils. Identifying these neurons can define relationships between dpr/DIP expression and synaptic connectivity, because detailed synaptic maps for units of the first two areas of the OL, the La and Me, have been created using electron microscopic reconstruction (Carrillo, 2015).

Axons of UV-sensitive R7 photoreceptors synapse in layer M6 of the Me onto Dm8, Tm5a, Tm5b, and other targets. dpr11 is selectively expressed by yR7s, which express Rh4 opsin and are in ~70% of ommatidia. Dpr11 is the first cell surface protein to be associated with a subclass of R7s. DIP-γ is expressed by Dm8s, which arborize in M6 and receive more R7 synapses than any other neurons (Carrillo, 2015).

To examine whether formation of synapses between yR7s and Dm8s involves interactions between Dpr11 and DIP-γ, a marker for existing active zones, Brp-shortmCherry, was expressed in yR7s. In control animals, yR7 terminals are bulb-shaped and regularly arranged in M6. In dpr11 and DIP-γ mutants, the main bodies of yR7 terminals have altered shapes, and active zone and membrane markers are found in extensions projecting into deeper Me layers. These data suggest that synapses between yR7 and its M6 targets do not form normally in the absence of Dpr11 or DIP-γ. Because most M6-projecting DIP-γ-positive cells seen in the FLP-out analysis are Dm8s, and because Dpr11's other partner, DIP-β, does not label M6, it is infered that the loss of Dpr11 or DIP-γ is likely to primarily affect yR7-Dm8 synapses in M6 (Carrillo, 2015).

In DIP-γ mutants, there are large gaps in M6 labeling by DIP-γ or Dm8 reporters. The number of OrtC2b+, DIP-γ+ cells is reduced by >3-fold, suggesting that most DIP-γ-expressing Dm8s die. Alternatively, they might turn off expression of the OrtC2b-GAL4 driver, although this seems less likely. This effect on cell fate suggests that DIP-γ is required for reception of a neurotrophic signal. Since dpr11 mutants have no DIP-γMiMIC M6 gaps, implying that they have normal numbers of OrtC2b+, DIP-γ+ cells, this signal might be communicated through Dprs 15, 16, and/or 17, the other Dprs that bind to DIP-γ. Other OL neurons are also dependent on trophic factors for survival. R cell growth cones secrete the Jelly Belly (Jeb) ligand, which binds to its receptor Alk on L3 neurons, and L3s die in the absence of Jeb or Alk. The functions of DIP-γ in mediating normal development of yR7-Dm8 connectivity, as assayed by displacement of the active zone marker in yR7s, may be distinct from its roles in Dm8 survival, because about half of the overshoots in DIP-γ mutants appear to grow through a Dm8 arbor labeled by the DIP-γ reporter (Carrillo, 2015).

dpr11 is expressed by subsets of direction-selective T4 and T5 neurons that arborize in the Lop layers activated by front-to-back and back-to-front motion, and DIP-γ is expressed by three LPTCs, which receive synaptic input from T4s and T5s. These data suggest that Dpr11 and DIP-γ expression patterns might have evolved to facilitate assembly of synaptic circuits for specific sensory responses: near-UV vision for yR7-Dm8 connections and movement along the anterior-posterior axis for T4/T5 subset-LPTC connections. In a conceptually similar manner, a specific type of vertebrate amacrine neuron, VG3-AC, forms synapses on W3B retinal ganglion cells, which are specialized for detecting object motion. Both VG3-ACs and W3B-RGCs selectively express the IgSF protein Sidekick2 (Sdk2), and Sdk2-mediated homophilic adhesion is required for their connectivity (Carrillo, 2015).

Binding between DIP and Dpr neuronal recognition proteins has been proposed to regulate synaptic connections between lamina and medulla neurons in the Drosophila visual system. Each lamina neuron was previously shown to express many Dprs. This study demonstrates, by contrast, that their synaptic partners typically express one or two DIPs, with binding specificities matched to the lamina neuron-expressed Dprs. A deeper understanding of the molecular logic of DIP/Dpr interaction requires quantitative studies on the properties of these proteins. A quantitative affinity-based DIP/Dpr interactome was generated for all DIP/Dpr protein family members. This revealed a broad range of affinities and identified homophilic binding for some DIPs and some Dprs. These data, along with full-length ectodomain DIP/Dpr and DIP/DIP crystal structures, led to the identification of molecular determinants of DIP/Dpr specificity. This structural knowledge, along with a comprehensive set of quantitative binding affinities, provides new tools for functional studies in vivo (Cosmanescu, 2018).

Brains from flies to humans comprise vast numbers of different types of neurons interconnected by networks of precisely patterned synaptic connections. Currently, the molecular mechanisms underlying the specification of neural circuit assembly are poorly understood. The predominant model, based on Roger Sperry's 'chemoaffinity hypothesis,' postulates that neurons make specific connections with their targets based on interactions between specific cell surface molecules. Cell-cell recognition proteins are often members of families diversified in evolution by gene duplication to yield numerous members, each bearing a canonical binding interface characteristic of the family. For such protein families, binding between members is often promiscuous, and it is the distinctive strength of binding, or binding affinity, that underlies the differential biological functions of each protein. Understanding the logic underlying the patterning of neural networks will require determination of the binding affinities of cell recognition proteins, their expression patterns, their signaling properties, and gain- and loss-of-function genetic analyses (Cosmanescu, 2018).

In Drosophila, two families of cell-recognition proteins of the immunoglobulin superfamily (IgSF), the 21-member Dpr (defective proboscis extension response) and the 11-member DIP (Dpr interacting proteins) families, have many of the properties expected of proteins controlling synaptic specificity. Members of each family are expressed in subsets of neurons throughout the developing nervous system. Within the Drosophila visual system, the five lamina monopolar neurons, L1-L5, as well as the R7 and R8 photoreceptor cells each express unique combinations of Dpr proteins. Cognate DIPs were found to be expressed in some of their synaptic partners in the medulla, suggesting a potential role in synaptic targeting. It remains unclear from these earlier studies, however, whether each medulla neuron type expresses many DIPs as observed for Dpr expression in lamina neurons or only a more limited repertoire of them (Cosmanescu, 2018).

In the accompanying paper, single Dm12 neurons with DIP-α null mutations exhibit robust defects in target-layer specificity in a wild-type background, and misexpression of cognate Dpr ligands dramatically re-specifies these connections (Xu, 2018). In addition to targeting phenotypes, DIP/Dpr interactions also play a role in cell survival. Loss of DIP-γ as well as its binding partner Dpr11 in R7 neurons (Xu, 2018) leads to a reduction in the number of Dm8 neurons. In both cases, cell loss results from apoptosis during development (Xu, 2018), consistent with the idea that DIP/Dpr interactions may influence the regulation of apoptosis. The number of DIP and Dpr paralogs, their patterns of expression within the brain, and the complexity of the DIP/Dpr interactome allude to a widespread and complex role in patterning neural circuitry (Cosmanescu, 2018).

High-throughput in vitro binding experiments using an ELISA-based assay revealed a heterophilic interaction network between members of the two families, where all but two members of the DIP family were found to interact with individual or subsets of Dprs. Although such assays are effective at identifying heterophilic binding, technical constraints of the method often select against the detection of homophilic interactions. Furthermore, these assays utilized multimerized chimeras to increase binding affinities so as to enable robust detection; as a consequence, however, this method inherently obscures the native molecular binding affinities, yielding binary results that provide a yes or no answer as to whether an interaction takes place (Cosmanescu, 2018).

Do binding affinities of adhesion proteins significantly impact interactions between cells? Differential affinities can have clear effects on signaling between adherent cells: for example, T cells bearing receptors with different affinities for peptide-major histocompatibility complex (MHC) complexes on antigen-presenting cells adopt different developmental fates. With respect to selectivity of cellular interactions, type I classical cadherin family proteins provide a typical example of the role of affinity: each type I cadherin family member binds to all other type I family members, yet the differences in affinity of each pairwise interaction dictate their distinct adhesive and cell-patterning functions. Thus, for protein families with promiscuous binding, in which selectivity is dictated mainly through the differential pairwise binding affinities of different family members, quantitative measures are required to understand their function (Cosmanescu, 2018).

In the nervous system, binding affinities of cell-cell recognition proteins have been shown to control the targeting of neurites to their appropriate partners. For example, members of the two-protein family of Ig-like sidekick (Sdk) proteins are expressed in specific layers within the inner plexiform region of the mouse retina during synapse formation. In vitro, Sdk1 and Sdk2 bind heterophilically through a canonical interface, but their homophilic affinities are stronger. Despite their heterophilic binding, the higher affinity of the respective homophilic interactions appears to determine their synaptic targeting activities. By contrast, within this same region of the retina, the type II cadherin family members cadherin-8 and cadherin-9, which show distinctive heterophilic and homophilic affinities to other type II cadherin family members, appear to rely on heterophilic rather than homophilic binding for proper layer-specific targeting. Thus, differential molecular binding affinities of both cadherins and Ig superfamily proteins contribute to synaptic patterning (Cosmanescu, 2018).

DIP-Dpr binding specificity is controlled by interactions between their immunoglobulin-like extracellular domains. The extracellular regions of Dpr family members consist of two tandem Ig-like domains, and the extracellular region of DIP family members consists of three tandem Ig-like domains. The crystal structure of a two-domain fragment of DIP-α in complex with the membrane-distal Ig1 domain of Dpr6 revealed the Ig1-Ig1 interaction to be characterized by a buried core of hydrophobic residues and an extensive network of hydrogen bonds. The interaction topology of this complex shares a strong resemblance to other complexes of Ig-like cell adhesion molecules, including those of vertebrate nectins and C. elegans SYGs, both of which have roles in nervous system development. Interestingly, members of the nectin and SYG-related protein families exhibit both homophilic and heterophilic binding (Cosmanescu, 2018).

As a step toward understanding how DIP and Dpr protein families contribute to neural circuit assembly, this study sought to extend understanding of both the binding affinities of DIPs/Dprs and the neuron-specific localization of DIPs in the Drosophila visual system. The multi-color flip out (MCFO) technique was used to provide a more extensive map of DIP expression in the medulla. To assess the biophysical properties of interactions between protein family members, surface plasmon resonance (SPR) was used to determine binding affinities for all DIP-Dpr interactions, identified DIPs and Dprs that form homodimers, and identified specificity-determining residues in DIP-Dpr interfaces that had not previously been noticed. This new knowledge was used to design site-directed mutants with defined intermolecular binding affinities for in vivo functional experiments reported in the accompanying paper (Xu, 2018). Biophysical studies raise the intriguing possibility that DIP/Dpr interactions function over a wide range of affinities to regulate neural circuit assembly throughout the Drosophila nervous system (Cosmanescu, 2018).

Developing axons and dendrites encounter the processes of perhaps hundreds of different neuronal cell types and must select appropriate synaptic partners from a myriad of neuronal processes. RNA sequencing technologies have revealed that developing neurons express hundreds of cell surface proteins, many of which bind in vitro to proteins known to be expressed on neighboring cells. Identifying which interactions are important and understanding how their expression patterns and binding interactions contribute to the specificity, complexity, and function of neural circuits remains a central challenge in developmental neuroscience (Cosmanescu, 2018).

Families of cell surface proteins with related ectodomains and differences in binding specificity provide one way of generating diverse patterns of connectivity. As opposed to Dscams and protocadherins, which are expressed stochastically to provide neurons with single cell identities that form the basis of self-avoidance, this study envisions that selective recognition between synaptic partners relies on deterministic mechanisms of gene regulation to ensure the appropriate cell-type-specific pairing of ligands and receptors. Indeed, it is the deterministic expression of matching DIP/Dpr pairs in some synaptic partners in the visual system that led to the idea that DIP/Dpr interactions might influence synaptic specification. A previous study demonstrated through mRNA sequencing and genetic tagging methods that Dprs were expressed in a dynamic and complex way in developing lamina neurons (Tan, 2015). Each lamina neuron expresses a discrete combination of numerous Dprs. It was also shown that some synaptic partners of lamina neurons, specific medulla neurons, express cognate-interacting DIP proteins. This study extended these observations through a systematic analysis of eight of the 11 DIPs using the MCFO technique. Of the 60 neuronal cell types analyzed, 26 expressed a single DIP, 12 expressed two DIPs, and one expressed four DIPs (i.e., 39/60 or 65% of the neurons express at least one of the eight DIPs). Assuming the remaining three DIPs, for which gene-trap GAL4s are not yet available (i.e., DIP-ι, -κ, and -λ), are expressed in a similar fashion, it is estimated some ~90% of the 60 different medulla neuron types considered in this study express one or, less frequently, two DIPs (Cosmanescu, 2018).

By comparing the synaptic connectivity maps between lamina and medulla neurons, the expression patterns of DIPs and Dprs, and the DIP/Dpr interactome, many DIP/Dpr pairs expressed in synaptic partners were identified. Lamina neurons form synapses on many different medulla neuron types; for instance, lamina L3 neurons express many Dprs and form synapses with over 10 different medulla neurons, many of which express DIPs that bind to Dprs expressed in L3. It appears then that lamina neuron outputs diverge to synapse with multiple partners. By contrast, medulla neurons express a more limited set of DIPs. For instance, Dm4 neurons only express DIP-α and form synapses with on the order of 20 L3 neurons, which express, among other Dprs, Dpr6 and Dpr10, high-affinity ligands for DIP-α. L3 is by far the predominant input to Dm4. The inputs into Dm4, therefore, are convergent. Indeed, information from multiple lamina neurons of the same type frequently converge onto a single DIP-expressing Dm neuron (Cosmanescu, 2018).

A clear pattern emerges whereby multiple Dprs on lamina neurons may promote connections to multiple targets, whereas a single DIP expressed on Dm neurons, for instance, accommodates convergence of many different neurons of the same type onto a single partner. Overall, this arrangement mirrors the interactome, where a single type of DIP tends to interact with high affinity to multiple Dprs, but in general a single Dpr exhibits high-affinity binding to one type of DIP. A similar trend is seen with both Tm and TmY neurons; they typically form connections with more different types of neurons than Dm, but fewer than lamina neurons. Interestingly, about half of the Tm and TmY neurons analyzed (10/21) express more than one DIP, whereas only one of 18 Dm neurons expressed more than one DIP (Cosmanescu, 2018).

Quantitative biophysical and structural studies enabled the identification of residues in DIPs and Dprs that control their binding specificity. Grouping DIPs and Dprs according to their cross-family binding interactions facilitated the identification of resides at positions in the sequence that were correlated with the binding preferences of different specificity groups. Most of these specificity residues are charged or polar in contrast to the conclusion of Carrillo (2015) that shape complementarity was the dominant determinant of inter-subgroup specificity. As discussed above, part of the discrepancy is due to their focus on the hydrophobic core of the interface although most distinct specificity determinants are located in the periphery (Cosmanescu, 2018).

The specificities of DIP-Dpr interactions are partially overlapping and grouped by phylogeny, with interaction affinities spanning approximately two orders of magnitude. Three main DIP affinity groups and DIP-δ, which forms a one-member group, emerge with cognate Dpr interactions mainly falling within a single DIP group, with sparser and weaker interactions between groups. These groupings became clear only when binding affinities were incorporated and false positive and negative interactions removed (e.g., removal of Dpr16/Dpr17 with DIP-ε and addition of DIP-κ and DIP-λ interactions). Quantitative binding affinities were also crucial for assigning the 'primary' DIP-binding specificities (the DIP[s] with highest interaction affinity) for groups of Dprs, which was used in the identification of specificity determinants. Dprs with similar binding preferences are closely related with a few exceptions, and DIPs within each of the three main groups are also close in phylogeny, with the exception of DIP-γ and DIP-κ. Indeed, single mutants in dpr6 and dpr10, which are phylogenetic nearest neighbors with similar DIP-binding profiles, show weaker phenotypes than null mutations, inactivating their common binding partner, DIP-α (Cosmanescu, 2018).

Like other families of cell surface proteins with related ectodomains, DIPs and Dprs bind through canonical interfaces common to all family members. Because interactions between members of such diversified protein families rely on a common binding mode, many family members might be expected to bind one another, albeit with different affinities. Thus, DIP and Dpr proteins engage in promiscuous interactions, as has also been observed for other protein families implicated in targeting, e.g., type I and type II cadherins, sidekicks, nectins, synCAMs, and Drosophila irre cell recognition module (IRM) proteins. The binding properties of these protein families differ significantly from the strict homophilic recognition observed for stochastically expressed multi-domain repulsion proteins (i.e., Dscam and clustered protocadherins). These achieve recognition only when all interacting domains are matched with their cognate partners, leading to an all-or-none binding specificity. Multi-domain interfaces may be required to achieve precise fine-tuning to avoid the promiscuity that is characteristic of two-domain interfaces. In contrast, wide-ranging affinities in protein families, such as Dprs and DIPs, may be exploited by developing neurons to sculpt neural circuitry in different ways (Cosmanescu, 2018).

The demonstration that some DIPs and Dprs form homodimers adds another layer to the potential regulatory complexity of interactions between these proteins. DIP homodimerization affinities are in the range of 22-35 μM, with Dpr homodimerization affinities ranging from 39 to 71 μM. The homodimerization affinity of a DIP can be significantly weaker than with its heterophilic binding to Dpr partners (DIP-α), equivalent to the strongest heterophilic interactions of its group (DIP-ζ), or stronger than its heterophilic interactions (DIP-η). For Dprs, in each case the homodimer affinities that were determined were substantially weaker than their heterophilic DIP interactions. Crystal structure and mutational analyses reveal that DIP/DIP and DIP/Dpr interfaces are largely overlapping. The Dpr/Dpr dimer structure has not yet been determined. Although AUC was used to identify homodimers, in principle, heterophilic DIP-DIP and Dpr-Dpr interactions could also form, though this study did not seek to identify such potential interactions. Indeed, Özkan, 2013, detected Dpr3-Dpr7 and Dpr5-Dpr6 heterophilic interactions in their high-throughput interaction study (Cosmanescu, 2018).

In principle, some DIPs and Dprs could function in cell-cell recognition driven by homophilic rather than heterophilic interactions. In support of this possibility, genetic rescue studies indicate that, in some contexts, homophilic interactions can substitute for heterophilic binding. For example, DIP-α overexpression in DIP-α-interacting neurons reduces Dm4 cell loss by apoptosis in Dpr6/10 null mutants (Xu, 2018). In some contexts, competition between homophilic and heterophilic binding partners could play a regulatory role in controlling interactions between neurons, as has been suggested for Sdks and nectins. Interestingly, germline knockin mutants of a homophilic binding-deficient form of DIP-α designed in this study led to a 50% increase in synapse number for Dm4 neurons. These findings are consistent with the notion that complex regulatory roles may modulate DIP/Dpr interactions during circuit assembly, and these, in turn, may regulate cell number and neuronal morphogenesis, as well as the distribution, number, and specificity of synaptic connections (Cosmanescu, 2018).

Drosophila Dpr (21 paralogs) and DIP proteins (11 paralogs) are cell recognition molecules of the immunoglobulin superfamily (IgSF) that form a complex protein interaction network. DIP and Dpr proteins are expressed in a synaptic layer-specific fashion in the visual system. How interactions between these proteins regulate layer-specific synaptic circuitry is not known. This study establishes that DIP-alpha and its interacting partners Dpr6 and Dpr10 regulate multiple processes, including arborization within layers, synapse number, layer specificity, and cell survival. This study demonstrated that heterophilic binding between Dpr6/10 and DIP-alpha and homophilic binding between DIP-alpha proteins promote interactions between processes in vivo. Knockin mutants disrupting the DIP/Dpr binding interface reveal a role for these proteins during normal development, while ectopic expression studies support an instructive role for interactions between DIPs and Dprs in circuit development. These studies support an important role for the DIP/Dpr protein interaction network in regulating cell-type-specific connectivity patterns (Xu, 2018).

The DIP/Dpr protein families exhibit complex biochemical interactions. Some DIP and Dpr proteins bind homophilically and all paralogs bind heterophilically, albeit with different affinities and degrees of specificities. Furthermore, these proteins are expressed in cell-type-specific patterns and high-affinity interactors are frequently expressed on synaptic partners. These findings, the cellular complexity of the visual system, and the specificity of synaptic connectivity led to the proposal that DIP/Dpr proteins contribute to the establishment of layer-specific neural circuitry. As a step toward critically addressing this possibility, this study reports that DIP-α and its high-affinity binding partners Dpr6 and Dpr10 regulate interactions between processes within the M3 layer (Xu, 2018).

As the phenotypes in DIP-α homozygous mutants and DIP-α homozygous mutant neurons in a wild-type background are different from each other, for clarity they are summarized here before discussing the results in more detail. In homozygous animals lacking either DIP-α or both Dpr6 and Dpr10, there is a reduction in the number of Dm4 and Dm12 neurons. Layer-specific targeting of these cell types is unaffected. There is no obvious change in the morphology Dm12 neurons in DIP-α or dpr6/10 homozygous animals. There is, however, an increase in the number of columns covered by each Dm4 neuron within the M3 layer. As both wild-type and mutant Dm4 neurons tile, the increase in the number of columns covered may reflect the decrease in cell number and argues that column coverage is governed by homotypic interactions (i.e., Dm4/Dm4 interactions independent of either DIP-α or Dpr6/10) (Xu, 2018).

This study explored the role of DIP-α and Dpr6,10 in controlling cell number in depth in the context of Dm4. DIP-α and Dpr6,10 heterophilic interactions promote cell survival by antagonizing a hid-activated cell death pathway. Developmental studies, antibody staining, and knockin mutant and chimeric rescue experiments support the notion that the interactions between this DIP/Dpr pair occur between axonal processes as they first encounter one another within the incipient M3 layer. The simplest interpretation of these data is that Dm4 neurons are generated in excess during normal development and interactions between them and L3 afferents (and perhaps other Dpr6- and Dpr10-expressing processes in M3) act as a source of limited trophic support, thereby determining the number of Dm neurons surviving into the adult. As Dm4 neurons tile, the number of Dm4 neurons indirectly sets the number of columns covered. This is consistent with the decrease in the extent of Dm4 arborization in animals with more Dm4 neurons as a consequence of Diap1 expression (Xu, 2018).

That interactions between DIP-α and Dpr6/10 regulate other aspects of Dm4 and Dm12 development was seen in genetically mosaic animals, in which DIP-α was selectively removed from single Dm4 or Dm12 neurons in an otherwise wild-type background. Different phenotypes in DIP-α mutant Dm4 and Dm12 neurons were observed: (1) There was a 30% decrease in the number of columns covered by mutant Dm4 neurons in mosaic animals. This is different from the number of columns covered by mutant neurons in a mutant background; presumably extension of processes within the layer is promoted by these DIP/Dpr interactions, such that mutant neurons compete less effectively for territory within the layer with their wild-type counterparts. There was no defect in layer-specific targeting. (2) By contrast to Dm4, in mosaic animals single DIP-α null mutant Dm12 neurons exhibited a robust mistargeting to another layer. Although all mutant Dm12 neurons targeted to M3, 60% of these sent additional processes to M8, where they arborized within this layer. A modest (~10%) reduction in column coverage in M3 was observed. (3) Removal of DIP-α from Dm12 led to a 30% reduction in the density of presynaptic sites. By contrast, the removal of DIP-α did not lead to a change in the density of presynaptic sites in Dm4. (4) DIP-αhet-homo mutant neurons, in which DIP-α heterophilic and homophilic interactions were disrupted, led to phenotypes in Dm4 and Dm12 indistinguishable from those seen in DIP-α null mutant neurons. (5) DIP-αhomo mutant neurons in a wild-type background led to an increase in the number of presynaptic sites in Dm4, but not in Dm12. Together, these data support a role for interactions between DIP-α on the surface of Dm4 and Dm12 neurons on mediating interactions with the processes of other neurons, notably L3, and perhaps other neurons within the developing M3 layer that are important for establishing neural circuitry (Xu, 2018).

Gain-of-function studies provide additional strong support for this conclusion. Misexpression of Dpr10 (or Dpr6) in a different layer from the expression in wild-type animals led to a nearly complete re-specification of targeting to this layer of both Dm4 and Dm12 axonal processes. This finding and the dependence of mistargeting upon DIP-α provide compelling evidence that binding observed in vitro occurs in vivo and contributes to the establishment of layer-specific circuitry. These observations are also consistent with the overlapping expression patterns of Dpr6/10 and DIP-α proteins in the developing neuropil. Thus, gain- and loss-of-function mutations leading to defects in arborization, layer-specific targeting, synapse number, and cell survival provide compelling evidence that interactions between DIP-α and Dpr6/10 on the surface of neurons in the developing M3 layer are necessary for normal circuit development. The mechanisms by which these interactions regulate these specific developmental processes, however, remain poorly understood (Xu, 2018).

It remains plausible that these different phenotypes result from a range of effects on cell viability, from death to compromised cell function. But as P35 expression in single mutant Dm12 neurons does not rescue the targeting defects, the view is favored that these wiring phenotypes are independent of cell survival. Other ligand/receptor pairs regulating both wiring and cell survival have been described, including the classical neurotrophins, and clustered protocadherins. Whether the interactions between DIP/Dpr proteins directly control survival, targeting, or synapse number, or whether they facilitate interactions between other cell surface proteins that, in turn, directly regulate specific effector functions, is not known (Xu, 2018).

Striking differences between the Dm12 mutant phenotypes in different genetic contexts were observed. Mistargeting of Dm12 was seen in sparsely distributed DIP-α mutant neurons in a wild-type background, whereas the targeting of mutant Dm12 neurons in a whole-animal DIP-αnull mutant was unaffected. Whether this difference reflects the activation of compensatory mechanisms in homozygous animals or whether the juxtaposition of single null neurons with wild-type neighbors artificially uncovers redundancy by creating neighboring neurons with different 'competitive' fitness is not known. In addition to the aforementioned targeting differences, mutant Dm12 neurons nestled within an otherwise wild-type background showed a reduced density of presynaptic sites compared to wild-type or mutant neurons in an all mutant background. Indeed, there was no difference in the density of synapses seen in wild-type Dm12 neurons compared to mutant Dm12 neurons in an all mutant background. As each L3 neuron receives input from three different Dm12 neurons, these data are consistent with DIP-αmutant Dm12 neurons in mosaics being at a competitive disadvantage relative to wild-type Dm12 neurons synapsing on the same L3. That is, a compensatory increase in the number of synapses in the two remaining wild-type Dm12 partners would be anticipated (Xu, 2018).

The discrepancy in phenotypes between mutant neurons in an all mutant background and mutant neurons with wild-type neighbors is similar to recent observations on the effects of neurexin knockouts in climbing fiber synapses on Purkinje neuron dendrites. Here, severe synaptic phenotypes were observed in sparsely labeled triple mutant neurons in a largely wild-type background, compared to only very weak phenotypes observed in sparsely labeled triple mutant neurons with many triple mutant neighbors. Similar observations were made on the dendritic targeting behavior of Dscam4 mutant lamina L4 neurons. Phenotypes were seen in homozygous Dscam4 mutant neurons with wild-type neighbors, but not in homozygous neurons in a homozygous mutant background. These studies suggest that genetic mosaic analyses may establish artificial competitive interactions between neurons, which, in turn, uncovers gene function (Xu, 2018).

Correlating the expression patterns and binding specificities of different DIPs and Dprs revealed that many cognate DIP/Dpr pairs are expressed on synaptic partners throughout the visual system. That matched expression patterns reflect function is supported by the finding that two DIPs (DIP-α and DIP-γ) and their high-affinity ligands (Dpr6/10 and Dpr11, respectively) regulate layer-specific circuit assembly. The role of the DIP and Dpr families in regulating specificity more broadly is likely to be complex as binding affinities between different DIPs and Dprs vary over two orders of magnitude and some Dprs and DIPs also exhibit homophilic binding (Xu, 2018).

The increase in synapses in Dm4 neurons seen in DIP-αhomo mutants raises the interesting possibility that homophilic interactions may inhibit and thereby modify heterophilic interactions (Xu, 2018).

In principle, homophilic interactions regulating synapse number could occur between DIP-α proteins acting in cis (i.e., in Dm4), or alternatively in trans with other cells with arbors within the layer (e.g., in Dm12) to antagonize a synaptogenic signal generated by heterophilic interactions between synaptic partners. In other contexts, however, such as cell survival, homophilic interactions in trans may also promote similar responses to heterophilic binding, as was observed targeted expression of DIP-α substituting for a cell survival phenotype seen in dpr6/dpr10 mutants. The inconsistencies of these results, the anti-synaptogenic signal and pro-cell survival signals, could also reflect differences in expression levels. In the former case, the homophilic binding-deficient form was expressed from the endogenous locus, acted cell autonomously at normal levels, and with the same spatiotemporal pattern as wild-type. By contrast, overexpression of DIP-α under the control of the GAL4/UAS system was sufficient to promote survival in a cell-non-autonomous fashion. Together these data raise the exciting possibility that circuit organization, in part, reflects different types of interactions between various DIPs and Dprs on neuronal processes leading to a variety of functional outputs (Xu, 2018).

Although there are many Dprs and DIPs, they represent only a small number of the vast array of cell surface proteins expressed in neurons in the developing visual system. There are over 100 neuronal cell types contributing processes, axons, and dendrites to the medulla neuropil, and each neuron type makes a characteristic pattern of connections. Different types of neurons express many cell surface proteins in common (e.g., hundreds) and they also express others in a cell-type-enriched fashion. Many of these proteins exhibit homophilic or heterophilic binding or both, and thus may interact with proteins expressed on the surface of other neurons in the developing neuropil. It is proposed that DIPs and Dprs act with other specificity molecules in a combinatorial and partially redundant fashion to allow axons and dendrites to discriminate between the diverse neuronal cell surfaces they encounter during visual circuit assembly. As DIPs and Dprs are expressed in a cell-type-specific fashion throughout the developing CNS, it seems likely that these proteins will act in different combinations to contribute to wiring specificity beyond the developing visual system (Xu, 2018).

In stereotyped neuronal networks, synaptic connectivity is dictated by cell surface proteins, which assign unique identities to neurons, and physically mediate axon guidance and synapse targeting. Two groups of immunoglobulin superfamily proteins in Drosophila, Dprs and DIPs, have been identified as strong candidates for synapse targeting functions. This study uncovers the molecular basis of specificity in Dpr-DIP mediated cellular adhesions and neuronal connectivity. First, five crystal structures of Dpr-DIP and DIP-DIP complexes are presented, highlighting the evolutionary and structural origins of diversification in Dpr and DIP proteins and their interactions. It was further shown that structures can be used to rationally engineer receptors with novel specificities or modified affinities, which can be used to study specific circuits that require Dpr-DIP interactions to help establish connectivity. This study investigated one pair, engineered Dpr10 and DIP-alpha, for function in the neuromuscular circuit in flies, and reveal roles for homophilic and heterophilic binding in wiring (Cheng, 2019)

Recent advances in connectomics and transcriptomics have the potential to enhance mechanistic understanding of neuronal wiring, especially if such datasets can be matched by accurate neuronal protein interaction datasets, and a structural and evolutionary understanding of how common molecular tools across animal taxa have been repeatedly used, and regularly expanded to create more complex neuronal networks. Previous evidence shows that Dprs and DIPs may be representative of neuronal surface proteins that have expanded in the arthropod line to help wire complex but stereotyped brains (Cheng, 2019)

The interaction network created by the Dprs and DIPs demonstrates how gene duplication events have led to diversity in molecular recognition and function in neuronal surface molecules. While the distant gene duplication events have given rise to the five Dpr and five DIP subclasses and have resulted in specialization of interactions, the more recent duplication events have only created mostly redundant molecular interactions. A comprehensive analysis of other arthropod Dprs and DIPs may reveal evolutionary forces that have resulted in repeated gene duplications in these families, and it is intriguing to speculate that the complexity of neural networks and the numbers of Dprs, DIPs and other neuronal surface receptors may correlate in arthropod species (Cheng, 2019)

The Dpr and DIP complex structures show a two-fold pseudo-symmetric architecture. This study also shows the presence of DIP-η (DIP-eta) and DIP-α homodimers in solution and present a symmetrical DIP-η homodimer structure that closely mimics heterodimeric Dpr-DIP complexes. This raises the question of whether the homophilic or the heterophilic interaction evolved first. Since Dpr and DIP IG1 sequences can be aligned with identities well above any random IG domain sequences, and Dpr and DIP IG1 domains are nearly identical in structure (RMSD values ≤ 1 Å), it is believed that Dpr and DIP IG1 domains may be the result of an ancient duplication event of a homodimeric IG domain. Following this logic, the DIP-η and DIP-α complexes may represent homodimers that were retained through multiple gene duplications. As heterophilic binding allows for higher diversity in neuronal recognition than homophilic would (i.e. 21 x 9 possible heterodimers > 30 possible homodimers), heterophilic binding must have been favored for specifying neuronal connections in complex structures such as the fly optic lobe. This is corroborated by observations that heterodimers have higher affinities than the homodimers (Cheng, 2019).

The observations reported in this study, including the lack of intracellular regions and the flexible nature of the ectodomain, have led the authors to believe that Dprs and DIPs may not be signaling receptors, and would require binding to co-receptors or secreted ligands for relaying signal to the cytoplasm upon formation of homo or heterodimers. It is also unclear if cis dimers can form, and signal. As cis dimers would inhibit productive trans cell-adhesive structures, their presence has significant functional relevance. It is believed that interdomain flexibility and long low-complexity 'stalk' regions linking the IG domains to the membrane would enable cis dimerization for homodimeric DIPs, such as DIP-α and DIP-η. In fact, surface plasmon resonance (SPR) experiments where DIP-η is captured on solid support at high densities reports much higher apparent KD values for the Dpr1-DIP-η interaction (23 μM vs 4.0 μM measured when non-dimerizing Dpr1 is captured on SPR chip), as the cis DIP-η homodimer formation on the chip likely competes with Dpr1 binding. The cis homodimerization may actually be the result of a strategy to inhibit cellular adhesions resulting from relatively weak trans interactions, which would not be able to overcome the cis homodimers. This would lead to more stringent selectivity for intercellular interactions, and would prevent non-specific synapses. These interactions were examined using engineered mutations in the NMJ, and evidence was found for functional relevance for both cis homodimeric and trans heterodimeric interactions, supporting this view (Cheng, 2019)

The requirement of the homomeric DIP-α-DIP-α interaction for proper synaptic targeting presents a layer of complexity to what at first appearance was a straightforward binary model. It is now known that DIP-α is required for proper synapse wiring, as a wild type UAS-DIP-α transgene in the mutant background can restore connectivity. However, when a UAS-DIP-α transgene with a mutation that breaks the DIP-α-DIP-α interaction was introduced in the same mutant background, the mutant form is unable to rescue the loss of connectivity. This does not appear to be a trafficking defect, as DIP-αI83A appears at similar wild-type levels in 1 s terminals as it does on other muscles. DIP-αI83A binds Dpr10, promiscuous binding of DIP-αI83A to Dpr10 on other muscles cannot be ruled; however, overexpression of UAS-DIP-αI83A with either DIP-α-GAL4 or Eve-GAL4, which also drives in MNISN-1s, does not reveal a GOF phenotype. Instead, the data support a model in which weak trans interactions with other molecules are resisted by homodimeric DIP-α complexes. This mode of targeting would allow for motor neuron growth cones to bypass non-specific or very weak interactions on non-target muscles and only synapse on bona-fide muscle targets. Interestingly, the concurrent study demonstrates that Dpr10 is expressed in specific muscles during embryonic development synchronous with growth cone exploration of those muscles, and thus overcome DIP-α homodimerization in favor of the stronger Dpr10-DIP-α heterodimer (Cheng, 2019)

During the late revision stages of this manuscript, two articles from the Shapiro, Honig and Zipursky groups were published (Cosmanescu, 2018; Xu, 2018). The results in this manuscript and the accompanying manuscript (Ashley, 2019) are in general agreement. The structures presented in this study and in Cosmanescu (2018) show a conserved mode of binding, now observed crystallographically across three DIP homodimers and five Dpr-DIP heterodimers. The conservation of the hydrophobic core and the variable polar periphery is another shared observation. The amino acids chosen to disrupt DIP-α and Dpr10 complexes, DIP-α I83 and Dpr10 Y103, were common to both studies. Finally, both sets of studies demonstrate phenotypes when DIP-α homodimers or Dpr10-DIP-α heterodimers are affected via mutagenesis (Cheng, 2019)

The Drosophila larval neuromuscular system provides an ideal context in which to study synaptic partner choice, because it contains a small number of pre- and postsynaptic cells connected in an invariant pattern. The discovery of interactions between two subfamilies of IgSF cell surface proteins, the Dprs and the DIPs, provided new candidates for cellular labels controlling synaptic specificity. This study shows that DIP-alpha is expressed by two identified motor neurons, while its binding partner Dpr10 is expressed by postsynaptic muscle targets. Removal of either DIP-alpha or Dpr10 results in loss of specific axonal branches and NMJs formed by one motor neuron, MNISN-1s, while other branches of the MNISN-1s axon develop normally. The temporal and spatial expression pattern of dpr10 correlates with muscle innervation by MNISN-1s during embryonic development. A model is presented whereby DIP-alpha and Dpr10 on opposing synaptic partners interact with each other to generate proper motor neuron connectivity (Ashley, 2019).

The proper wiring of neural circuits is essential for animal behavior, and alterations in connectivity are linked to neurological disease phenotypes in humans. Thus, identifying cell-surface molecules involved in neural wiring is critical for understanding biological mechanisms in normal development and in diseased states. Using genetics to uncover these mechanisms has been difficult, partially due to the fact that achieving the necessary precision appears to require partially redundant biochemical interactions (Ashley, 2019).

One of the simplest and most accessible systems in which to study the genetic determination of synaptic connectivity patterns is the Drosophila larval neuromuscular system. In each larval abdominal hemisegment, 35 identified motor neurons innervate a set of 30 muscle fibers. Each motor neuron chooses one or more specific muscle fibers as synaptic targets, and the map of connections is almost invariant. Drosophila neuromuscular junction (NMJ) synapses are glutamatergic and use orthologs of mammalian AMPA receptors for synaptic transmission. Many scaffolding and regulatory proteins that modulate these receptors are conserved between insects and vertebrates. The sizes and strengths of Drosophila NMJs are regulated by retrograde signaling from their postsynaptic muscle targets. In addition to this developmental plasticity, NMJ synapses also exhibit short-term and homeostatic plasticity. These features make the Drosophila NMJ a useful genetic model system for excitatory glutamatergic synapses in the mammalian brain (Ashley, 2019).

Although many molecules involved in axon guidance, NMJ morphology, and synaptic activity have been identified through genetic and reverse genetic experiments, understanding of the mechanisms by which individual larval muscle fibers are recognized as synaptic targets by Drosophila motor axons is still lacking. Gain-of-function (GOF) experiments suggest that individual muscles are labeled by cell-surface proteins (CSPs) that can define them as targets for motor axons. 30 CSPs have been identified that cause motor axons to mistarget when they are ubiquitously expressed in muscles. These proteins contain a variety of extracellular domain (XCD) types, including immunoglobulin superfamily (IgSF) domains and leucine-rich repeat (LRR) sequences. Some of these proteins are normally expressed on subsets of muscles in embryos, suggesting that they could act as molecular signatures during motor axon targeting (Ashley, 2019).

However, none of the CSPs identified thus far are required for innervation of the muscles that express them, suggesting that they have partially redundant functions. In loss-of-function (LOF) mutants lacking CSPs expressed on muscle fibers or the receptors or these proteins on motor axons, innervation occurs normally in most cases. There are no published LOF mutations in CSP genes that cause high-penetrance failures of innervation of specific muscle fibers (Ashley, 2019).

A network of new candidates for synaptic targeting molecules was recently identified through a global in vitro 'interactome' screen. In this network, the 'Dpr-ome', a set of 21 proteins with two IgSF domains, the Dprs, interact in a complex pattern with a set of 11 proteins with three IgSF domains, called DIPs. The expression patterns of many Dprs and DIPs have been studied, and each is expressed in a small and unique subset of neurons in the larval ventral nerve cord and pupal brain (Ashley, 2019).

The functions of one Dpr-DIP binding pair, Dpr11-DIP-γ, have been studied in both the larval neuromuscular system and the pupal optic lobe. Loss of either dpr11 or DIP-γ produced phenotypes affecting NMJ morphology and retrograde bone morphogenetic protein (BMP) signaling, but did not alter NMJ connectivity patterns. DIP-γ is expressed in most motor neurons, so it is unlikely to be involved in recognition of targets by specific motor neurons. In the optic lobe, however, DIP-γ is selectively expressed in amacrine neurons that are postsynaptic to photoreceptor neurons that express Dpr11, suggesting that Dpr11-DIP-γ interactions may be important in determining synaptic connectivity patterns. For several other Dpr-DIP in vitro binding pairs, optic lobe neurons expressing a Dpr are also synaptically connected to neurons expressing the cognate DIP . In the antennal lobe, Dprs and DIPs regulate adhesion and sorting of axons of olfactory receptor neurons (Ashley, 2019).

Based on these findings, a survey of DIP expression in the larval neuromuscular system was carreid out in order to identify DIPs whose expression is restricted to subsets of motor neurons. Remarkably, DIP-α is expressed by only two motor neurons in each hemisegment. There are two types of glutamatergic motor neurons in the larval neuromuscular system: 1b (big boutons) and 1s (small boutons). Larval muscle fibers are divided into four fields: the ventral, ventrolateral, lateral, and dorsal fields. Each 1b motor neuron innervates one or two muscle fibers. The three 1s motor neurons have multiple branches, and each 1s neuron forms branches on most or all of the fibers within a specific muscle field. DIP-α is expressed in MNISN-1s, which synapses on dorsal muscles, and in MNSNb/d-1s, which synapses on ventral and ventrolateral muscles. Fate determination and axon guidance of MNISN-1s have been extensively studied in embryos, where it is known as RP2 (Ashley, 2019).

A subset of muscles innervated by MNISN-1s axon branches are muscles 4, 3, and 2, which are arranged in a ventral->dorsal sequence. In DIP-α mutant larvae, the interstitial axon branch onto muscle 4 (m4) is always missing, and the branch onto m3 is usually absent. The branch onto m2, however, is always present. MNISN-1s filopodia are observed in the m4 target area in both wild-type and DIP-α mutant embryos, but 1s boutons never form on m4 in mutants. This suggests that nascent axonal projections onto m4 fail to stabilize and convert into NMJs in the absence of DIP-α (Ashley, 2019).

The 'Dpr-ome' revealed that DIP-α binds to Dpr6 and Dpr10. This study examined phenotypes in the larval neuromuscular system caused by loss of these Dprs, and found that in dpr10 null mutant larvae the MNISN-1s axon branch onto m4 is missing, mimicking the DIP-α mutant phenotype. In 3rd instar larvae, dpr10 is expressed in almost all muscle fibers. However, during motor axon outgrowth in embryos, dpr10 expression initiates in two muscle fibers 140 in the immediate vicinity of m4, and then comes on in m4 itself around the time at which axon branches appear on this muscle. These results suggest that Dpr10 is a muscle recognition cue whose binding to DIP-α on the motor axon triggers recognition and stabilization of the MNISN-1s filopodia on specific muscles (Ashley, 2019).

The accompanying paper (Venkatasubramanian, 2019) shows that DIP-α and Dpr10 have expression patterns in adult leg motor neurons and muscles that are qualitatively similar to those seen in the larval neuromuscular system, and that loss of DIP-α or Dpr10 causes failure of DIP-α-expressing leg motor neurons to innervate a subset of their normal muscle targets. Thus, in both of these neuromuscular systems, interactions 150 between DIP-α and Dpr10 control formation of synapses on specific muscle targets (Ashley, 2019).

This paper shows that interactions between DIP-α and its in vitro binding partner, Dpr10, are essential for innervation of a specific subset of larval muscle fibers by branches of the MNISN-1s motor axon. DIP-α is expressed by only two motor neurons, and the protein localizes to the NMJs of those neurons. MNISN-1s innervates most of the muscles in the dorsal muscle field, but only the proximal (most ventral) branches of its axon are affected in DIP-α mutants. The branch innervating m4, m4-1s, is absent in 100% of hemisegments in mutants. DIP-α is required in the MNISN-1s neuron to direct innervation of m4. Examination of the MNISN-1s axon during embryonic development shows that its filopodia explore the surface of m4 and surrounding muscles in a normal manner in DIP-α mutants, but a stable m4 NMJ never forms (Ashley, 2019).

Innervation of muscles near m4 is also reduced in DIP-α mutants, while innervation of more dorsal muscles is increased. One of DIP-α's two binding partners, Dpr10, is expressed at high levels in muscles and can localize to the postsynaptic side of NMJs, and the m4-1s branch is also absent in dpr10 mutants. RNAi knockdown experiments showed that Dpr10 is required in muscles. By examination of the] temporal and spatial expression patterns of dpr10 in embryos, it was found that its earliest expression is on muscles flanking m4, some of which also lack 1s NMJs in DIP-α mutants. This suggests that recognition of Dpr10 on these muscles by DIP-α on the 465 MNISN-1s growth cone is a cue for branch formation or stabilization (Ashley, 2019).

A number of mutant screens for alterations in the morphologies and patterning of NMJs in the larval neuromuscular system have been performed. LOF mutations in a few genes, including those encoding the cell- 471 surface IgSF domain protein Sidestep and its binding partner, Beaten Path, cause motor 472 axons to fail to arborize normally onto any muscle fibers, resulting in large-scale alterations 473 in innervation patterns. However, there are no prior reports of LOF mutations in single 475 genes that cause high-penetrance changes in targeting of single larval motor axons to 476 individual, or groups of, muscle fibers (Ashley, 2019).

The failure to find genes required for innervation of specific muscles in LOF screens has suggested that individual muscles may be labeled by multiple targeting cues, and that neurons express receptors for more than one of these cues. Loss of any one neuronal receptor or muscle targeting cue does not cause strong phenotypes because they have partially redundant functions. The remaining receptors and cues may substitute for the missing proteins in mutants and allow normal muscle targeting to occur (Ashley, 2019).

It has been difficult to identify neuronal receptors whose expression is specific to particular subsets of motor axons. Neuronal CSPs that have been previously studied in the context of motor axon guidance and arborization onto muscles (e.g., Receptor Tyrosine Phosphatases (RPTPs), Beaten Paths, Fasciclin II, Netrin receptors, Semaphorin receptors) are usually expressed by most or all motor neurons. Not surprisingly, then, mutations in genes encoding these proteins usually produce guidance or arborization phenotypes that affect many motor axons and muscles. By contrast, DIP-α is expressed in only two of the 35 motor neurons that innervate muscles in each larval abdominal hemisegment. These are the 1s motor neurons MNISN-1s (RP2) and MNSNb/d-1s. This finding suggested that any phenotypes caused by loss of DIP-α might be specific to the axons of these two motor neurons (Ashley, 2019).

Like other motor axons, the two DIP-α-expressing axons probably express binding partners for many muscle cell-surface proteins. Neuronal and muscle binding partners could act as signaling receptors, ligands for receptors, or adhesion molecules. The 1s motor axons that express DIP-α have multiple branches, and each axon innervates most of the muscles within a muscle field. MNISN-1s innervates muscles in the dorsal field. One might expect that targeting phenotypes would be observed in DIP-α LOF mutants only if binding of DIP-α to one of its Dpr binding partners was essential for recognition of specific muscle fibers by individual branches of a 1s motor axon. In fact, the loss of DIP-α was observed to cause a high-penetrance loss of branches onto a particular group of dorsal muscle fibers innervated by MNISN-1s. These are internal muscles 4 and 3, and external muscles 19 and 20, which are underneath m4 and m3 in embryonic and larval 'fillet' preparations. However, the branches of the MNISN-s axon onto internal dorsal muscle 2 and external muscles 9 and 10, which lie underneath 510 m2, are still present in DIP-α mutants (Ashley, 2019).

The same MNISN-1s branches are lost with high penetrance in dpr10 null mutants, indicating that Dpr10 is the DIP-α binding partner relevant to innervation of these muscles. In larvae, dpr10 is expressed at high levels in most muscle fibers. Knockdown of dpr10 by 515 RNAi in all muscles affects formation of the same MNISN-1s branches that are eliminated in 516 dpr10 mutants. Therefore, binding of neuronal DIP-α to muscle Dpr10 is likely to underlie 517 recognition of specific muscles as targets for proximal MNISN-1s axon branches. In 518 embryos, dpr10 expression is initiated in m20 and m5, which flank m4. NMJs on m20 are 519 also absent in DIP-α mutants; m5 does not receive 1s innervation. dpr10 expression begins 520 in m4 around the time of exploration of this muscle by filopodia 521 emerging from the MNISN-1s axon (Ashley, 2019).

The gene encoding DIP-α's other binding partner, Dpr6, is expressed by most motor neurons, but is not detectably expressed by muscle fibers. Although dpr6 is expressed by MNISN-1s in embryos, dpr6 mutants do not have m4-1s phenotypes. This suggests tha Dpr6 does not play a direct role in the targeting of MNISN-1s to m4. DIP-α can also bind 527 homophilically, but with reduced affinity relative to its heterophilic binding affinities for Dpr10 and Dpr6. However, during normal 5development DIP-α would not have the opportunity to mediate homophilic interactions between motor axons, since it is expressed only on the MNISN-1s axon and not on the other motor axons with which it fasciculates during outgrowth (Ashley, 2019).

Correct innervation of m4 and the other muscles in its immediate vicinity may require a balance between the expression levels of DIP-α's binding partner Dpr10 on muscles vs.axons. As described above, knocking down Dpr10 in muscles eliminates innervation of m4, suggesting that transsynaptic interactions between neuronal DIP-α and muscle Dpr10 are essential for recognition of this muscle by an interstitial MNISN-1s growth cone. There is also a 50% reduction in m4 innervation when Dpr10 is knocked down in all neurons, while knockdown in MNISN-1s produces no innervation defects. This suggests 540 that interactions between DIP-α on MNISN-1s axons and Dpr10 on other axons with which 541 it fasciculates also contribute to correct m4 innervation (Ashley, 2019).

Driving high-level expression of DIP-α or Dpr10 in all neurons abolishes m4 innervation by MNISN-1s. Normally DIP-α is not expressed in other axons in the ISN fascicle, so upon DIP- 545 α expression ectopic axon-axon interactions mediated by homo- and heterophilic binding may alter MNISN-1s connectivity. Interestingly, these GOF phenotypes are also seen when Dpr10, but not DIP-α, is increased in MNISN-1s only. High-level expression of DIP-α or Dpr10 on muscles also eliminates (DIP-α) or reduces (Dpr10) innervation of m4. Some of these phenotypes may be due to cis Dpr10-DIP-α interactions on the same membrane, which could reduce the amount of DIP-α or Dpr10 that is available to interact with its partner in trans. Excessive adhesion between the MNISN-1s axon and the other axons in its 552 bundle (in the case of Dpr10 overexpression in all neurons), or between the MNISN-1s ]]axon and the muscles it traverses during its outgrowth (in the case of Dpr10 overexpression in muscles) may also affect the ability of a branch to separate from the axon and form an NMJ. Overexpression of Dpr10 in muscles may similarly cause excessive MNISN-1s adhesion to distal muscles, and this model is supported by the exuberant 557 number of branches on m2 (Ashley, 2019).

Knockdown or overexpression of DIP-α or Dpr10 in neurons or in muscles does not reduce 560 the frequency of innervation of the most dorsal muscles by MNISN-1s, indicating that these 561 muscles are recognized as targets via other cues. Interestingly, however, m1, which is adjacent to m2 and rarely innervated by MNISN-1s, gains innervation in DIP-α mutants, and the 1s NMJ on m2 becomes larger. These results suggest that MNISN-1s is normally specified to make a certain number of synaptic boutons, and that loss of boutons on proximal muscles m4, m3, m19, and m20 results in an increased number of boutons on distal muscles (Ashley, 2019).

Using these results, a model was constructed that can explain how interactions between DIP-α and Dpr10 specify targeting of MNISN-1s axon branches to m4 and the other muscles in its vicinity. DIP-α begins to be expressed in MNISN-1s (RP2) in st14 mbryos, during the period of motor axon guidance. The MNISN-1s axon reaches its terminus in the vicinity of m1/m2 at stage 16, before it forms interstitial branches onto m20, where Dpr10 is already expressed. After the m20-1s branch forms, Dpr10 appears on m4, and binding of DIP-α on MNISN-1s to Dpr10 on m4 and surrounding muscles results in the formation of stable branches that differentiate into NMJs. During this process, DIP-α on MNISN-1s might switch from interacting with Dpr10 on fasciculated axons within the ISN bundle to binding to Dpr10 on muscles (Ashley, 2019).

The Dpr-ome binding network was defined by a global in vitro 'interactome' screen for binding interactions among all Drosophila cell-surface and secreted proteins containing three common extracellular domain types: IgSF, Fibronectin Type III, and LRR. There are 21 Dpr proteins, each containing two IgSF domains, 11 DIP proteins, each containing three 584 IgSF domains, and an LRR protein called cDIP that binds to many Dprs and DIPs (Ashley, 2019).

Analysis of expression of individual dpr and DIP genes revealed remarkable and unprecedented patterns in the larval ventral nerve cord and pupal brain. Each dpr and DIP is expressed by a small and unique subset of interneurons. In the pupal optic lobe, neurons expressing a particular Dpr are often presynaptic to neurons expressing a DIP to which that Dpr binds in vitro. These findings suggested that Dpr-DIP interactions might be important for formation of synaptic circuits during brain and ventral nerve cord development (Ashley, 2019).

In earlier work, the expression and function of DIP-γ and its binding partner Dpr11 were examined. Dpr11 is selectively expressed in 'yellow' R7 photoreceptors, which make the Rh4 rhodopsin, and DIP-γ is expressed in a subset of Dm8 amacrine neurons in the optic lobe medulla. Dm8s receive more synapses from R7 than any other neuron. DIP-γ is required for survival of the Dm8 neurons that express it. The fact that oss of DIP-γ causes loss of brain neurons that express these proteins suggests that DIP-Dpr interactions can transmit trophic signals. This does not appear to be the case for either DIP in the larval or adult neuromuscular system, however, since there are no missing motor neurons in DIP-γ or DIP-α mutants (Ashley, 2019).

The expression patterns of DIP-γ and DIP-α suggest that they may be involved in similar processes during optic lobe development. In addition to yellow R7s, Dpr11 is expressed in a subset of motion-sensitive T4 and T5 neurons, which synapse onto large cells called Lobula Plate Tangential Cells (LPTCs). Dpr11-expressing T4 and T5 cells project to the layers 1 and 2 of the lobula plate, and DIP-γ is expressed in a small number of LPTCs that arborize in those layers. In the optic lobe lamina, L3 and L5 neurons express Dprs 6 and 10, while L2 expresses only Dpr6. These L cells are synaptically connected to Dm4, Dm12, and Dm1 cells in the medulla, which express DIP-α. Loss of DIP-α or of both Dprs 6 and 10 causes death of some Dm4 neurons and affects synaptic targeting of Dm12 neurons (Ashley, 2019).

In the larval neuromuscular system, however, the functions of DIP-γ appear to be very different from those of DIP-α. DIP-γ and Dpr11 are both expressed by most or all motor neurons. In DIP-γ and dpr11 LOF mutants, there are no alterations in muscle targeting, but NMJs have phenotypes characterized by the presence of small clustered boutons called satellites. Retrograde BMP signaling is upregulated in these mutants. By contrast, DIP-α is expressed by only two motor neurons, and its interactions with Dpr10 expressed on muscles control formation and/or targeting of a specific set of interstitial axon branches (Ashley, 2019).