Protein-trap lines utilized in our screen were made and provided by the laboratories of Dr. Allan Spradling (Carnegie Institution for Science) (Buszczak et al., 2007) and Dr. Lynn Cooley (Yale University) (Quiñones-Coello et al., 2007). The Basigin null allele bsg^δ265 (Curtin et al., 2007) was provided by Dr. Kathryn Curtin (University of Arkansas). Transgenic lines for expression of full-length Basigin (UAS-bsg^FL, UAS-bsg^FL::GFP) and mutant variants (UAS-bsg^{KRR > NGG}::GFP, UAS-bsg^extra::GFP) (Besse et al., 2007) were provided by Dr. Anne Ephrussi (EMBL Heidelberg). The Basigin RNAi line was obtained from the Vienna Drosophila RNAi Center (Transformant ID: 105293) (Dietzl et al., 2007). IT(0871-Gal4), referred to as 871-Gal4 in the text, was provided by Dr. Thomas Clandinin (Stanford University). To generate the UAS-Bsg-G transgenic line, full-length cDNA of Bsg-G was first made by appending missing sequences to the partial Bsg-G cDNA obtained from the GH21853 cDNA clone (Drosophila Gold Collection, Berkeley Drosophila Genome Project). A c-terminal FLAG tag was added to the full-length Bsg-G cDNA and sub-cloned into a pUASTattB vector containing the mini-white gene. Plasmids were then injected into Drosophila embryos and transformants were selected based on eye color of adults. All other reagents were obtained from the Bloomington Drosophila Stock Center (Indiana University). Animals of either sex were used. Genotypes of animals analyzed for the experiments described herein are listed below. Animals were raised at 25°C, except those used for RNAi-based knockdown experiments, which were raised at 29°C.

Control: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A/FRT40A

bsg^δ265: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / bsg^δ265 FRT40A.

Control: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / FRT40A

bsg^δ265: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / bsg^δ265 FRT40A

Full-length Basigin: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / bsg^δ265 FRT40A; UAS-bsg^FL / +

Extracellular Basigin: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / bsg^δ265 FRT40A; UAS-bsg^extra::GFP / +

Basigin with KRR > NGG mutation: hsflp, GAL4^elav, UAS-mCD8GFP/+; TubP-Gal80 FRT40A / bsg^δ265 FRT40A; UAS-bsg^{KRR > NGG::GFP} / +

Dendritic localizations of full-length and mutated Basigin: ppk-Gal4/+; UAS-bsg^FL::GFP / +, ppk-Gal4/+; UAS-bsg^{KRR > NGG::GFP}/ +, ppk-Gal4 / bsg^δ265 FRT40A; UAS-bsg^{KRR > NGG::GFP}/ +.

Validation of IT (871)-Gal4 line: 871-Gal4/UAS-mCD8GFP

Control: 871-Gal4 / ppk-CD4tdGFP

Basigin-RNAi: UAS-bsg^RNAi / bsg^δ265; 871-Gal4 / ppk-CD4tdGFP

Basigin overexpression: 871-Gal4/UAS-Bsg-G.

The following antibodies were used: chicken anti-GFP (ab13970, Abcam, RRID:AB_300798, 1:500), goat anti-HRP (Sigma, 1:250), mouse anti-Coracle (c556.9 and c615.16, RRID:AB_1161644; developed by R. Fehon, 1:40), mouse anti-E-Cadherin (5D3, RRID:AB_528116; developed by B. Gumbiner, 1:100), rabbit anti-dsRed (632496, Clontech, RRID:AB_10013483:, 1:250), rat anti-Basigin (a kind gift from Anne Ephrussi (Besse et al., 2007), 1:100), rabbit anti-FLAG (Sigma, 1:100) and mouse anti-βPS integrin (CF.6G11, RRID:AB_528310; developed by D. Bower, 1:10). CF.6G11, 5D3, c556.9 and c615.16 were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology. Species-specific secondary antibodies (Jackson Immunoresearch) raised in donkey were used at 1:250. Permeabilization was done with 0.3% Triton-X100 except for rat anti-Basigin staining which required Tween-20.

Filleted 3^rd or 2^nd instar larvae were fixed in 4% paraformaldehyde for 18 min at room temperature on a tabletop shaker and stained using standard immunohistochemical techniques largely as described before (Grueber et al., 2002). Live-staining of Basigin was conducted by directly applying the primary antibody solution to filleted larvae and incubating for 15 min at room temperature. The animals were then rinsed thrice with 1x PBS every 3 min and fixed immediately using 4% paraformaldehyde. Subsequent staining procedures were identical to those used for fixed larval fillets.

Morphometric analysis was conducted by tracing dendritic arbors using Neurolucida (RRID:SCR_001775, MBF Bioscience, United States). Dendrite tracing was carried out in confocal stacks captured using identical z-sectioning parameters (class IV) or flattened projections (class I). For Basigin rescue and epidermal knock-down experiments, the dorsal posterior quadrant was selected and quantified as a representative of the entire class IV dendritic tree. Validity of this approach was ascertained by comparing branches between the dorsal posterior quadrant and full arbor for each neuron in the control (N = 6) and basigin mutant (N = 9) groups. As shown in Supplementary Figure 2, the number of nodes in the posterior dorsal quadrant of each neuron constituted an equivalent proportion of, and scaled down linearly relative to, those over the entire arbor in both control and mutant neurons (proportions: 27.67 ± 3.40% and 26.81 ± 4.90%, respectively; p = 0.696; Supplementary Figures 2C,D). Furthermore, the extent of decrease in branching in basigin mutant neurons relative to the control was almost identical when quantified over the full arbor (26.71%) or only in the dorsal posterior quadrant (28.17%) (Supplementary Figure 2E).

Dendrite field size of class IV neurons was measured as the area of the smallest convex polygon enclosing the full dendritic tree (Convex Hull method) using the Hull and Circle plug-in (Karperien, 2005) in ImageJ (RRID:SCR_003070). In order to determine differences in the extent to which each arbor filled its dendritic field, coverage density (CD) was calculated as follows: a flattened confocal or traced image of a class IV dendritic tree was overlaid with a grid of square boxes (Supplementary Figure 2B). The size of the square box depended on the overall dimension of the arbor such that the box area was equal to the square root of area of the rectangle that wholly contained the dendritic tree. CD was then calculated as the ratio of number of boxes containing dendrites to the total number of boxes, multiplied by 100. The Box Counting feature in FracLac plugin (Karperien, 2013) for ImageJ was used wherever possible (traced images) for automated detection of boxes containing dendrites (Supplementary Figure 2B); others (confocal images) were analyzed manually with the genotype of each neuron masked prior to analysis.

Statistical analysis was conducted using the R software package (R Project for Statistical Computing, RRID:SCR_001905). All data were checked for Gaussian distribution (Shapiro-Wilk test) and analyzed further by either Welch’s t-test or Wilcoxon Rank Sum Test, as appropriate, for two-sample comparisons. Multiple sample comparisons were done by performing Analysis of Variance (ANOVA) followed by a suitable post hoc test for pairwise comparisons as noted in figure legends. Statistical significance was inferred if p < 0.05. Data are presented as box plots in which the top and bottom box boundaries demarcate interquartile range (IQR) while whiskers represent 75^th percentile + 1.5^∗IQR and 25^th percentile – 1.5^∗IQR. Thick horizontal lines and black dots within boxes represent median and mean, respectively. Raw data are shown as points laid over box plots; placement of such points along the x axis was randomized within the constraints of group boundary to avoid visual occlusion when y values are similar.

To identify genes involved in dendrite morphogenesis, we examined the expression patterns of >250 genes in third instar Drosophila larvae using protein trap insertion lines generated previously (Morin et al., 2001; Buszczak et al., 2007; Quiñones-Coello et al., 2007). Each of the screened lines contained a GFP-coding exon inserted in a gene locus, which results in GFP-tagging of its protein products. Larval peripheral nervous system (PNS) expression data for all protein traps that we examined is provided in Figure 1. Approximately half of the lines showed expression in the larval PNS (Figures 1, 2). Of those with predominantly neuronal expression, some showed different class-specific levels of expression in da neurons. For instance, Jupiter::GFP showed strong expression in class I da neurons (Figure 2B). We observed punctate localization along dendrites in several lines (e.g., Tsp42Ee::GFP, ArgK::GFP, and Chrb: GFP) (Figures 2C–E). On the other hand, VAChT::GFP was expressed in all neurons, but at discrete levels anti-correlated with the branching complexity of da neurons—high in class I and low in class III and IV neurons (Figure 2F). A subset of lines showed ubiquitous nuclear expression (e.g., CB04957 for LamC, Figure 1) that may not reflect bona fide protein expression and localization due to enhancer trapping, as explained previously (Buszczak et al., 2007). However, some lines showed variable or non-nuclear GFP expression in addition to ubiquitous nuclear GFP (e.g., CB02121 for homer, Figure 1). A total of 78 lines showed expression in both neurons and epidermis. A subset of these lines featured dendritic GFP localization in neurons, and strong but intermittent epidermal enrichment of GFP adjacent to dendrites (e.g., Nrg::GFP in Figure 2G). We chose to focus further on the multifunctional immunoglobulin superfamily member Basigin (Bsg) based on its expression in both neurons and epidermal cells (Figure 3A).

Bsg-GFP (Line ID: CA06978) contains a GFP exon cassette inserted in an intron in the basigin locus. Staining of filleted Bsg-GFP third instar larvae with anti-GFP revealed signal in diverse tissues. Strong expression was observed in the larval photoreceptors and NMJ (Figure 3B), both of which are known to require Basigin for proper development (Besse et al., 2007; Curtin et al., 2007). In the larval PNS, we observed Bsg::GFP in da neuron cell bodies, axons, and dendrites (Figure 3A). Additionally, we observed expression in epidermal cells that lie in close proximity to da neuron dendrites. The sub-cellular localization of GFP in the epidermal cells overlapped with that of Coracle (Figure 3A), a septate junction resident protein that marks regions where da neuron dendrites become enclosed within epidermal cell invaginations (Kim et al., 2012). We examined GFP expression in two other independently generated protein-trap lines (Bsg-GFP-2 and Bsg-GFP-3) that harbor GFP-coding exon insertions at different sites within the basigin locus. Both lines showed similar expression in neurons and epidermis as described above (Supplementary Figures 1A,B). We additionally labeled wild-type larvae with Basigin antibody. Labeling was observed in da neurons, including along dendrites and also at the NMJ (Figure 3C). We did not observe specific epidermal localization in fixed tissue, however, labeling was apparent when the primary antibody was applied to unfixed preparations (Figure 3D). The reason for this discrepancy is not known, but it is conceivable that epidermal expression is masked by fixation, or that the expression pattern of the three Bsg-GFP traps is somehow aberrant. To examine epidermal localization independent of GFP tagging we generated a transgenic FLAG-tagged Basigin (UAS-Bsg-G) line and drove expression in epidermal cells using the InSITE line {IT.GAL4}871 (Gohl et al., 2011). Specificity of the driver was verified using a fluorescent reporter line. Strong GFP signal was observed across the epidermis of 871-Gal4/+;UAS-mCD8::GFP/+ larvae (Supplementary Figures 1C,D) but none was detected in da neurons (Supplementary Figure 1D’), indicating the epidermal specificity of 871-Gal4 in larvae. Strong FLAG signal was observed along epidermal cell boundaries and along epidermis adjacent to dendrite branches (Figure 3E), similar to the Bsg::GFP localization patterns observed in the protein trap lines. Taken together, these data support expression of Basigin in da neurons and also in epidermal cells in the larval body wall, and localization of these sources of Basigin in adjacent intercellular regions.

Based on Basigin expression in the PNS, we next investigated whether Basigin is involved in regulating neuronal morphogenesis in sensory neurons. We performed loss of function analysis using the Mosaic analysis with a repressible cell marker (MARCM) approach (Lee and Luo, 1999). We examined the null allele bsg^δ265 in which a part of the Basigin locus including the start codon is deleted (Curtin et al., 2005). Loss of Basigin in class IV neurons (ddaC) resulted in aberrant dendritic arbors in third instar stages. In contrast to the complex space-filling dendrites of control class IV neurons (Figure 4A), bsg^–/– class IV neurons (Figure 4B) had significantly fewer branches (Control: 903 ± 82, N = 6; bsg^δ265: 662 ± 136, N = 9; p = 0.0009) and reduced total dendrite length (Control: 18846 ± 751 mm, N = 6; bsg^δ265: 14297 ± 2484 mm, N = 9; p = 0.0004) (Figures 4C,D). By contrast, there was no significant difference in dendritic field area between control and mutant neurons (Supplementary Figure 2A). These results suggest that mutant neurons are able to scale their dendritic territory to an appropriate size for the third instar stage and that the dendrite phenotype reflected a defect in space filling. Indeed, we found significantly lower dendrite coverage density (a dimensionless metric; see “Materials and Methods” section) in mutant neurons compared to control neurons (Control: 41.7 ± 2.9, N = 6; bsg^δ265: 35.81 ± 2.21, N = 9; p = 0.002) (Figure 4E).

In contrast to class IV neurons, bsg^–/– class I dendritic arbors were comparable to those of control class I neurons (Figures 4F,G) with statistically identical branch number (Control: 29 ± 6, N = 3; bsg^δ265: 30 ± 1, N = 3; p = 0.875) and total dendrite length (Control: 1418 ± 219 mm, N = 3; bsg^δ265: 1425 ± 165 mm, N = 3; p = 0.965) (Figures 4H,I). Thus, our data indicate that Basigin is cell-autonomously required for morphogenesis of complex space-filling dendritic arbors in da neurons.

Drosophila larvae show a drastic increase in body size from the first to third instar stages. Although class IV neurons establish their complete tiling pattern by the end of the first instar stage, they continue to elaborate branches to maintain full coverage of their territories as the animal grows (Parrish et al., 2009). Our observation of reduced dendrite branching and coverage density in bsg^–/– neurons at late third instar in the absence of change in dendritic field size suggested a possible defect in addition of new branches during development. To determine if bsg^–/– mutant class IV neurons have aberrant branch elaboration programs, we examined their dendrite morphology at 72 h after egg-laying (AEL), a stage that marks the end of second instar and is characterized by active dendrite elaboration. Consistent with previous reports (Parrish et al., 2009), we found that control class IV neurons have highly branched dendritic arbors that completely innervate their dendritic fields at this stage (Figures 5A,A’). In contrast, bsg^–/– class IV neurons at the same stage had simpler dendrite arborization with ∼34% fewer branches (Figures 5B,B’) compared to control neurons (No. of nodes, Control: 704 ± 70, N = 3; bsg^δ265: 465 ± 62, N = 4; p = 0.039). Dendrite arborization nevertheless continued to increase in bsg^–/– neurons, with significant change from late second to late third instar (No. of nodes, 465 ± 62, N = 4 at 2^nd instar vs. 662 ± 136, N = 9 at 3^rd instar; p = 0.029), as in control neurons (No. of nodes, 704 ± 70, N = 3 at 2^nd instar vs. 903 ± 82, N = 6 at 3^rd instar; p = 0.071). The average number of branches added between late 2^nd and late 3^rd instars was nearly equivalent between bsg^–/– and control neurons (197 and 199 branches, respectively), indicating that mutant neurons show no defects in branch addition at late larval stages (Figure 5C). Although these experiments do not rule out possible contribution of late-stage branch maintenance defects, our data are consistent with a primary defect in bsg^–/– neurons throughout larval development, which, in wild-type animals, features prolific dendrite elaboration to keep up with the expanding body wall (Parrish et al., 2009).

Basigin is a single-pass transmembrane protein with two predicted Ig domains in its N-terminal extracellular region and a short intracellular C-terminal ending. Its transmembrane region is highly conserved and the juxtamembrane KRR motif, a putative binding site for cytoskeletal organizers (Yonemura et al., 1998), is important for regulation of NMJ morphology in flies (Besse et al., 2007). The N-terminal end of Basigin contains a signal sequence, suggesting that it may be released extracellularly in some form, likely upon cleavage of the full-length protein. Indeed, evidence from vertebrate studies indicates that Basigin is secreted in microvesicles by human uterine epithelial cells, and the secreted peptides induce MMP expression in human uterine fibroblast cells (Braundmeier et al., 2012). To determine the domain requirements for Drosophila Basigin during dendrite morphogenesis, we examined the ability of mutated variants of Basigin (Figure 6A) to rescue the dendrite morphogenesis defects described above. Full-length Basigin protein and variants with mutations or truncations in putative functional domains were expressed exclusively in bsg^–/– cells by adding the respective transgenes into the crossing scheme used to generate bsg^–/– MARCM clones. For quantification of dendritic branching we focused analysis on single quadrants of class IV arbors. We validated this method by examining the loss-of-function clones described above (Figure 4), and found that the number of nodes in the posterodorsal quadrant constituted equivalent proportions of nodes over the entire arbor in control and mutant animals (Supplementary Figure 2C). Furthermore, no significant difference was observed when the extent of reduction in branching in bsg^–/– class IV neurons was quantified over the whole arbor or the posterodorsal quadrant (Supplementary Figure 2E). Together, these results indicated that the posterodorsal quadrant could be used as a proxy for branching across the arbor for these genotypes.

Expression of full-length wild-type Basigin (Bsg^FL) rescued dendrite defects in bsg^–/– class IV neurons (No. of nodes in posterior dorsal quadrant, Control: 249.67 ± 36.57, N = 6; bsg^δ265: 179.33 ± 56.54, N = 9; Bsg^FL: 224.25 ± 56.51, N = 4). By contrast, Basigin lacking its transmembrane and cytoplasmic regions (Bsg^extra) failed to rescue dendritic branching defects in bsg^–/– class IV neurons (Bsg^extra: 171 ± 18.18, N = 5, Figures 6A,B). Likewise, full-length Basigin with point mutations that substituted the juxtamembrane KRR basic residues in the cytoplasmic tail to NGG (Bsg^{KRR > NGG}) failed to rescue the branching defects of bsg^–/– neurons (Bsg^{KRR > NGG}: 162 ± 12.29, N = 3, Figures 6A,B). Since motifs in the cytoplasmic tail of cell surface proteins may be essential for proper sub-cellular localization, we examined if the lack of rescue by the mutant Basigin variant could be explained by a defect in proper subcellular localization. GFP-tagged full-length Basigin expressed under the class IV neuron-specific ppk-Gal4 driver showed robust localization to dendrites (Figure 6C), consistent with data from the Bsg-GFP trap line as well as anti-Basigin staining of wild-type animals (Figure 3). Likewise, GFP-tagged Bsg^{KRR > NGG} showed stable expression in class IV neurons with strong localization to dendrites including fine terminal branches (Figure 6C). Therefore, gross mislocalization of Bsg^{KRR > NGG} proteins within neurons is unlikely to account for their inability to rescue dendrite elaboration defects. Our data do not eliminate the possibility that lack of rescue by Bsg^extra – despite its ability to partially rescue some NMJ phenotypes (Besse et al., 2007)– may be attributable to aberrant trafficking of the truncated protein. Taken together, our results indicate that the function of Basigin in regulating dendrite elaboration of class IV neurons requires membrane-tethering and an intact intracellular KRR motif.

Given our evidence for Basigin expression in epidermal cells and their close association with da neuron dendrites, we next examined possible cell non-autonomous roles for Basigin in regulating dendrite morphogenesis. For this experiment, we used the 871-Gal4 line (Supplementary Figures 1C,D) to drive UAS-bsgRNAi in the epidermis. Our results showed that epidermal knock down of Basigin had no effect on epithelial cell shape or average cell size at the third instar stage (Figures 7A,B). Moreover, expression and localization patterns of epidermal markers such as coracle, βPS integrin and dE-cadherin appeared unaffected by knock down of Basigin (Figures 7A,C). Class IV dendrites were visualized by tdGFP expressed under the control of the ppk promoter (ppk-CD4tdGFP) (Han et al., 2011) in 871-Gal4; bsg^+/–; UAS-bsgRNAi background. Epidermis-specific knock down of Basigin in this manner resulted in aberrant morphogenesis of neuronal dendrites. Compared to class IV neurons in control larvae, those in larvae expressing RNAi transgenes against Basigin had significantly lower dendrite coverage at 3^rd instar (Coverage density, Control: 61.80 ± 3.06, N = 16; Bsg-RNAi: 55.75 ± 2.64, N = 15; p = 0.017; Figures 7D,E). However, we did not detect a significant change in dendrite branching at the same stage (No. of nodes per quadrant: Control: 244.06 ± 58.97, N = 17 and Bsg-RNAi: 224.77 ± 62.20, N = 13, p = 0.398.) Epidermal knock down of Basigin had no effect on class I ddaE dendrite length, similar to the lack of effect of bsg mutation in class I neurons (Control: 1337.78 ± 181.82 mm, N = 8; Bsg-RNAi: 1225.58 ± 176.84 mm, N = 8, p = 0.231; Figures 7F,G). Likewise, no change was observed in branching of class I neurons (Control: 16.5 ± 2.78, N = 8; Bsg-RNAi: 18.75 ± 2.19, N = 8, p = 0.095; Figures 7F,H). Thus, in addition to cell autonomous roles in class IV neurons, our results suggest cell non-autonomous roles for epidermal-derived Basigin in dendrite morphogenesis.

Large-scale GFP trap collections have been instrumental in identifying proteins that are at the right place to be involved in many different cellular processes, and can complement insights gained from forward genetic screens (Morin et al., 2001; Buszczak et al., 2007; Quiñones-Coello et al., 2007). By screening pre-existing GFP trap lines from three collections we identified diverse candidate regulators of dendrite morphogenesis that can be followed up systematically using mutant analysis, including transcription factors, cell adhesion molecules, cytoskeletal regulators, and signaling proteins. One advantage of a protein expression pattern-based approach is that follow up can be hypothesis-driven, since candidates can be picked based on biological processes or putative molecular function. Indeed, we previously showed that one candidate from this screen, the Scalloped transcription factor, acts at the top of a repressive transcriptional cascade to diversify sensory neuron morphology (Corty et al., 2016). Here we followed up on Basigin-GFP, which showed expression in dendrites and substrate epidermal cells, to study factors that could be involved in dendrite-substrate interactions. Interactions between dendrites and their extracellular environment are increasingly recognized as a central driving force in dendrite morphogenesis and it will be important to further examine how Basigin fits into the multitude of cues that have been identified so far (Parrish et al., 2009; Han et al., 2012; Kim et al., 2012; Dong et al., 2013; Salzberg et al., 2013; Jiang et al., 2014; Cherra and Jin, 2016; Díaz-Balzac et al., 2016; Meltzer et al., 2016; Tenenbaum et al., 2017; Poe et al., 2017; Hoyer et al., 2018).

Our MARCM analysis of Basigin revealed a decrease in dendrite coverage, dendrite branching, and total dendrite length of Basigin-deficient class IV sensory neurons. Basigin has been implicated in diverse biological processes in vertebrates and invertebrates from embryonic membrane apposition (Reed et al., 2004), embryo implantation (Igakura et al., 1998), tumor invasion (Muramatsu and Miyauchi, 2003), synapse formation (Besse et al., 2007), to cell surface localization of lactate transporters (Kirk et al., 2000). A common theme spanning known functions of Basigin is mediation of cell-cell or cell-substrate interaction, which are critical processes for tissue development and integrity. Such a role is also well-suited to mediate neuronal growth over both small and large spatial domains to achieve precise innervations. Our Basigin loss of function analysis fits with this perspective.

Our experiments suggest that Basigin function in dendrites involves engagement of its extracellular Ig domains by extrinsic effectors, which may be molecules residing on, or released from, epithelial cell surfaces. Non-neuronal Basigin may be one such extracellular effector, since knockdown of Basigin in epidermis also led to defects in class IV dendrite morphogenesis (Alizzi et al., 2020; this study). Similarly, at the fly NMJ, Basigin is required both in the postsynaptic muscles and presynaptic motor neurons for synapse development and function (Besse et al., 2007). Electron micrographs of the larval body wall show that da neuron dendrites are in apposition to epithelial cell surfaces (Han et al., 2012; Kim et al., 2012). Interestingly, Drosophila Basigin has been reported to cause cell aggregation when expressed in S2 cells suggesting a homophilic binding capacity (Besse et al., 2007). Isoform-specific homophilic binding has also been reported for vertebrate Basigin (Hanna et al., 2003). Likewise, we propose that Basigin mediates dendrite-substrate interactions to promote complex dendrite morphogenesis. Understanding the factors that regulate the expression and localization of Basigin will be important future directions. The bsg^–/– phenotypes that we observed were consistent with a report that identified Basigin as a target of the RNA binding protein Found in neurons (Fne) in the control of space-filling dendrite growth of class IV neurons, which provides insight into the regulation of Basigin expression (Alizzi et al., 2020). Our data also indicate that the localization of Basigin aligns well with the septate junction protein Coracle and so it could conceivably be a component of junctional complexes.

How might Basigin function to promote dendrite patterning? In one scenario, interactions with the epidermis may coordinate addition of new branches throughout the dendritic arbor, thereby maintaining the arbor’s space-filling property. Signals derived from the epidermis regulate scaling growth of dendritic territories as the body wall expands during larval growth (Parrish et al., 2009). The phenotypes we observed in bsg^–/– class IV neurons are unlikely to reflect defects in scaling of overall dendritic territory, as the dendritic field continues to expand in bsg^–/– neurons with no net difference in dendritic area compared to control neurons. Instead, our results may point to an “intra-arbor scaling” process involving Basigin that coordinates dendrite coverage density with the growth of the substrate. In this model, lack of Basigin impairs space-filling growth within the dendritic tree, which normally serves to maintain coverage over the expanding body wall. Comparison between 2^nd and 3^rd instar revealed that the failure results from inadequacy, rather than inability, of branch addition through these stages. It remains unclear, however, whether primary and higher order branches are differentially affected. Future studies employing live imaging approaches would be well positioned to offer a nuanced understanding of space filling defects with branch order resolution in Bsg mutants.

The mechanism by which Basigin promotes dendrite coverage likely involves the positively charged KRR motif in the intracellular tail given the necessity of this region for rescue of the Basigin mutant phenotype. Although the molecules that bind to this motif in Basigin are unknown, evidence from studies on other transmembrane proteins identifies the KRR motif as a binding site for cytoskeletal regulators, specifically those of the Ezin/Radixin/Moesin (ERM) family (Yonemura et al., 1998). Therefore, Basigin may impact the neuronal cytoskeleton via regulators that bind to the KRR motif. This assertion is consistent with prior results showing that Bsg is important for stabilization of the neuronal cytoskeleton (Alizzi et al., 2020). Although the effects exerted by such a mechanism are likely to be local, our findings do not rule out other pathways that result in more global effects over the entire dendritic tree. For example, vertebrate Basigin is known to induce global cellular changes such as activation of signaling pathways (e.g., ERK1/2 signaling) and gene expression (Belton et al., 2008).

Proteins of the Ig superfamily are implicated in nearly all aspects of neural circuit development, including axon growth (Usardi et al., 2016), dendrite targeting (Yamagata and Sanes, 2012) and synapse specificity (Carrillo et al., 2015). Ig superfamily proteins have diverse structures with varying number of Ig domains with or without other identifiable functional motifs. In mice, there are two Basigin isoforms with two and three Ig domains each (Muramatsu, 2016), while humans have two additional isoforms each with a lone extracellular Ig domain (Liao et al., 2011). In Drosophila, Basigin is a 2-Ig protein (Besse et al., 2007), but a third Ig domain is also predicted in the longest isoform (B. Shrestha, personal observation). Thus, Basigin is structurally similar to the small IgSF protein family, of which the 2-Ig domain family is best known and implicated in various aspects of neural development. The latter includes the Beat proteins, involved in axon guidance, in Drosophila (Pipes et al., 2001), and the ZIG proteins in C. elegans, some of which are important for axon and synapse maintenance (Howell and Hobert, 2016). Notably, and potentially analogous to the role of Basigin in dendrite development, ZIG-10 in C. elegans is expressed in the epidermis and the motor neurons that physically contact them (Cherra and Jin, 2016) and loss of ZIG-10 in either cell decreases synapse number. Additionally, this function of ZIG-10 requires membrane tethering. Thus, our findings linking Drosophila Basigin to dendrite development contribute to a growing body of literature implicating small 2- or 3-Ig containing IgSF members in neuronal morphogenesis.

It will be important to further dissect roles for other small Ig proteins in dendrite development and to extend these analyses to vertebrate systems. Vertebrate Basigin is most closely related to Embigin and Neuroplastin, and the family comprising these three proteins collectively mediate processes ranging from tumor metastasis to embryo implantation and synapse formation (Muramatsu, 2016). At the molecular level, these proteins have diverse functions: as chaperones for monocarboxylate transporters, aiding their cell-surface localization (Kirk et al., 2000), as inducers of MMP expression (Toole, 2003), organizers of cellular cytoskeleton (Curtin et al., 2005; Besse et al., 2007), and as auxiliary subunits of plasma membrane Ca²⁺-ATPases (Schmidt et al., 2017). Embigin induces motor nerve terminal sprouting at vertebrate NMJs (Lain et al., 2009), while Neuroplastin is important for synapse development and function (Sarto-Jackson et al., 2012; Herrera-Molin et al., 2014; Carrott et al., 2016). Basigin itself exhibits broad CNS expression in both mice and humans [Allen Mouse Brain Atlas, Allen Human Brain Atlas (Lein et al., 2007; Hawrylycz et al., 2012)]. Basigin has been shown to regulate the expression and surface localization of monocarboxylate transporters in retinal pigment epithelia in mice, and mice lacking Basigin exhibit degeneration of photoreceptors and are blind (Philp et al., 2003). Our findings in Drosophila that Basigin functions in dendrite morphogenesis raise the possibility that it plays similar roles in development of the vertebrate nervous system.