A genetic screen in Drosophila uncovers a role for senseless-2 in surface glia in the peripheral nervous system to regulate CNS morphology

Abstract Despite increasing in mass approximately 100-fold during larval life, the Drosophila CNS maintains its characteristic form. Dynamic interactions between the overlying basement membrane and underlying surface glia are known to regulate CNS structure in Drosophila, but the genes and pathways that establish and maintain CNS morphology during development remain poorly characterized. To identify genes that regulate CNS shape in Drosophila, we conducted an EMS-based, forward genetic screen of the second chromosome, uncovered 50 mutations that disrupt CNS structure, and mapped these alleles to 17 genes. Analysis of whole genome sequencing data wedded to genetic studies uncovered the affected gene for all but 1 mutation. Identified genes include well-characterized regulators of tissue shape, like LanB1, viking, and Collagen type IV alpha1, and previously characterized genes, such as Toll-2 and Rme-8, with no known role in regulating CNS structure. We also uncovered that papilin and C1GalTA likely act in the same pathway to regulate CNS structure and found that the fly homolog of a glucuronosyltransferase, B4GAT1/LARGE1, that regulates Dystroglycan function in mammals is required to maintain CNS shape in Drosophila. Finally, we show that the senseless-2 transcription factor is expressed and functions specifically in surface glia found on peripheral nerves but not in the CNS to govern CNS structure, identifying a gene that functionally subdivides a glial subtype along the peripheral–central axis. Future work on these genes should clarify the genetic mechanisms that ensure the homeostasis of CNS form during development.


Introduction
Structure determines function is a unifying theme of biology.A thorough understanding of the genes that govern tissue or organ structure is, however, lacking.We use the Drosophila larval CNS as a model system to investigate the genetic factors that control tissue shape (Skeath et al. 2017).In response to the concerted action of hemocytes, neural activity, and glial cell function in a process called nerve cord condensation, the Drosophila larval CNS adopts its distinctive shape at the end of embryogenesis upon the shortening and thickening of the nerve cord (Olofsson and Page 2005;Karkali et al. 2022).Once its shape is established, the Drosophila CNS maintains its form throughout its massive growth during larval development.
The Drosophila larval CNS is consecutively and fully enwrapped by the CNS basement membrane and the perineurial and subperineurial surface glial cells (Stork et al. 2008;Meyer et al. 2014;Yildirim et al. 2019).The basement membrane is composed of independent Collagen IV and Laminin networks that are connected to each other by Perlecan; both networks establish connections to surface glial cells through interactions with the Integrin and Dystroglycan cell surface receptors (Yurchenco 2011).The tethering of the basement membrane to surface glial cells provides the CNS with structural and biomechanical strength essential to maintain its distinct shape.Interactions between the basement membrane and surface glia are, however, far from static.The tremendous growth of the CNS during larval life demands a dynamic basement membrane-surface glia interface that imparts structure to the CNS, while being continually remodeled to allow tissue growth and expansion.
In the larvae, the fat body secretes basement membrane proteins that are continuously deposited on the CNS basement membrane during development (Pastor-Pareja and Xu 2011), coordinating basement membrane deposition with tissue growth in the CNS.Remodeling of the basement membrane occurs at least in part through the action of the membrane-tethered and secreted extracellular proteases MMP1, MMP2, and AdamTS-A, which are expressed by surface glia (Page-McCaw et al. 2003;Meyer et al. 2014;Skeath et al. 2017).These proteases oppose the tissue stiffening action of Collagen IV by cleaving basement membrane proteins and in so doing help maintain a delicate balance between tissue stiffness and softness that maintains tissue structure while allowing tissue growth and expansion (Apte 2009;Miller et al. 2011;Rodriguez-Manzaneque et al. 2015;Skeath et al. 2017).In addition to the fat body, migratory hemocytes also express basement membrane proteins in larvae and may serve as https://doi.org/10.1093/g3journal/jkae152Advance Access Publication Date: 12 July 2024 Investigation a second source of basement membrane proteins and a basement membrane remodeling agent (Isabella and Horne-Badovinac 2015).
To gain more insight into the factors that regulate CNS shape, we undertook a large-scale forward genetic screen of the second chromosome to identify mutations that disrupt CNS structure.Our screen identified 17 genes, including characterized regulators of tissue structure, characterized genes with no known role in controlling tissue structure, such as Toll-2 (18-wheeler) and Rme-8, and previously uncharacterized genes, like a glucuronosyltransferase homologous to B4GAT1 and LARGE1/2, which regulate Dystroglycan in mammals, and the transcription factor Senseless-2.Our results highlight an unexpected role for perineurial glia in the peripheral nervous system (PNS) to help to establish and maintain CNS structure.

Materials and Methods
Genetic screen: To identify mutations that disrupt larval CNS morphology, we performed a standard, autosomal recessive forward genetic screen of the second chromosome using a chromosome isogenic for the M[3xP3-RFP.attP]ϕC31 "landing pad" transgene inserted in cytological location 51D.The 3xP3-RFP construct in this transgene expresses RFP in most glia and allows rapid visualization of CNS morphology in live larvae (Zhu et al. 2024).As outlined in Fig. 1, we mutagenized isogenic M[3xP3-RFP.attP] males with 25 mM EMS for 8 h, allowed them to recover from EMS treatment overnight, and then mated them en masse to CyO P[Tb 1 -RFP]/wg Sp1 P [Hs-hid] virgin females, discarding mutagenized males on day 5.The CyO P[Tb 1 -RFP] chromosome carries the larval marker Tubby (Tb), facilitating the identification of larvae that carry or do not carry this balancer based on larval body shape.The wg Sp1 P[Hs-hid] chromosome carries the hid transgene under the control of the hsp70 promoter, which induces activation of the proapoptotic hid gene upon heatshock, killing all flies that carry this transgene.17,790 F1 males of the M[3xP3-RFP.attP]***/CyOP[Tb 1 -RFP] genotype, where "***" denotes the mutagenized chromosome, were individually crossed to 3-5 CyO P[Tb 1 -RFP]/wg Sp1 P [Hs-hid] virgin females and allowed to mate for 7 days before adults were discarded.On days 8 and 9, each vial was heatshocked for 30-35 min at 37°C to ensure that only F2 progeny of the following genotype survived: M[3xP3-RFP.attP]***/CyOP[Tb 1 -RFP]; 12,211 of the 17,790 single male crosses were successful; for each of these crosses, F2 progeny were sib-mated to expand the stock and resulting F3 progeny were screened for larval-, pupal-, or semilethal phenotypes; 3,180 lines displayed such a phenotype.For each of these lines, F4 larvae homozygous for the mutagenized chromosome, easily identified by their lack of the Tb marker, were visually screened for disrupted CNS morphology, e.g. an elongated CNS, under a fluorescent dissecting microscope.Fifty lines were identified that harbored second chromosomal mutations that when homozygous disrupted the morphology of the larval CNS.
Complementation analysis: The 50 mutations were grouped into 3 phenotypic classes: those with an elongated CNS (n = 34), a herniated or misshapen CNS (n = 15), or a wider CNS (n = 1).Exhaustive complementation crosses among lines within each group identified 9 complementation groups and 9 "singleton" mutations.Crosses of each of these lines to second chromosomal mutations known to disrupt CNS morphology, such as Laminin B1 (LanB1) and the tightly linked viking and Collagen Type IV alpha1 (Col4a1) genes, which are uncovered by the deficiency, Df(2L)BSC172, identified 9 alleles of LanB1 and one each of viking and Col4a1, leaving 8 complementation groups and 6 singletons.
Whole genome sequencing: At least 3 alleles per complementation group, all singleton mutations, and the isogenic target chromosome were subjected to whole genome sequencing.For each allele, 10-15 late-third instar larvae homozygous for the relevant mutagenized M[3xP3-RFP.attP]*** chromosomes were collected, washed, and frozen.Genomic DNA extractions were performed using the Qiagen DNeasy Blood & Tissue Kit and and Illumina Nextera Flex DNA Library Prep Kit was used to prepare indexed next-generation sequencing libraries following the manufacturer's protocol.Genomic DNA was then provided to GTAC (Washington University) for whole genome sequencing at ≥30× coverage.Sequence reads were mapped to Drosophila reference BDGP6 using NovoAlign (Novocraft Technologies).Sentieon software was used to remove duplicate reads, realign reads around indels, recalibrate quality scores, and call variants.The Sentieon DNAscope tool was used to identify variants.Joint variant calls were made using the mutant lines and a reference line.Variants were annotated using SnpEff and the variants were filtered using SnpSift to select unique coding variants within previously determined genomic intervals.Ultimately, this process identified nucleotide differences or variants in coding regions and splice sites and small deletions that differed between the mutagenized chromosome and the isogenic target chromosome; these variants are likely EMS-induced mutations.On average, our sequencing and bioinformatics approach identified 19.4 likely EMS-induced mutations in coding regions or splice sites per mutagenized second chromosome (range 7-32), using a minimum read depth of 20 as a cutoff and requiring homozygosity or near homozygosity for the variant.For singletons, this approach identified a list of putative EMS-induced mutations, one of which is likely the causative lesion.For complementation groups, this approach identified the likely causative lesion for all complementation groups, as only one gene in each complementation group harbored independent EMS-induced mutations in its coding region in all sequenced alleles.When possible, we confirmed correspondence between gene and mutant phenotype via complementation crosses with appropriate deficiencies and genespecific alleles and phenotypic analysis.All putative causative lesions were confirmed via PCR-based methods and Sanger sequencing (GeneWiz Inc.).
Genbank accession information for whole genome sequence information: BioProject ID PRJNA1128589.
Genetic mapping and complementation crosses: To identify the causative lesion in singleton alleles, we took 2 approaches.For 3 lines, we used complementation crosses against deficiencies or gene-specific alleles that uncovered mutations likely to cause the associated CNS phenotype to confirm the correspondence between gene and mutant phenotype.This approach identified mutations in diaphanous, pvf3, and lethal giant larvae as the causative to the CNS phenotype.For the remaining 3 singletons, we used the approach of Zhai et al. (2003) to map the causative lesion in each line on the genetic map relative to 4 P[w+] element inserts located at genetic map positions 13, 52, 62, and 86 in the second chromosome.Using this approach, we mapped 1 singleton to genetic map positions 18-23 and 53-57 in the second chromosome; the results for the third singleton were inconclusive likely due to the presence of 2 or more lethal mutations in the chromosome.Complementation crosses against appropriate deficiencies and gene-specific alleles followed by phenotypic analysis identified Tango1 as the singleton on 2L and worniu as the singleton near genetic map positions 53-57.The 2 putative causative lesions were confirmed by PCR-based methods and Sanger sequencing.
Transgene construction: The UAS-senseless-2 transgene was generated by amplifying nucleotides 276-2,529 of the sens-2 cDNA, FI 20031 (accession number BT33368).The resulting PCR product was cloned into NotI and KpnI sites of pUAS-T.DNA was coinjected with helper plasmid into embryos of the genotype w1118 by Rainbow Transgenic Flies, Inc.
Due to the large size of ppn, the UAS-Ppn[RG] transgene was constructed piecemeal through PCR-based approaches amplified defined regions of partial cDNA clones and genomic DNA; these individual regions were then stitched together via Gibson cloning.First, we used the GH25513 cDNA clone, which contained a portion of the 5′ coding region of Ppn (accession number BT100225) to amplify nucleotides 1-2,471 of the coding region of the Ppn[RG] transcript starting with the ATG start codon in exon 1 and continuing into exon 6.Second, we used genomic DNA from a strain of flies isogenic for a wild-type third chromosome to amplify nucleotides 2,452-5,205 within exon 6 nucleotide 5,162-5,384 (exons 7) of the Ppn[RG] transcript.Third, we used the partial 3′ GH05059 cDNA clone (accession number AY060635) to amplify in separate reactions nucleotides 5,344-6,602 (exons 8-9) and nucleotides 6,556-8,526 (exons 10-17, including the ATT stop-codon) of the Ppn[RG] transcript.The PCR products were assembled via Gibson cloning and inserted into the KpnI and XbaI sites of the phiC-based construct pJFRC28, which was then injected into the P[CaryP] attp2 fly line (BDSC: 8622) with phiC integrase by Rainbow Transgenic Flies, Inc.

Gene rescue and in vivo RNAi phenocopy assays
To restrict UAS-linked transgene expression specifically to glia, we used the repo-GAL4 driver line.To restrict UAS-linked transgene expression specifically to neurons, we paired the elav-GAL4 driver line, which activates transgene expression strongly in all neurons and moderately in glia, with repo-GAL80, which blocks GAL4-dependent activation in glia (Awasaki et al. 2008).Gene rescue experiments were performed in the sens-2-T2A-GAL4/sens-2 js4 background.
Allele nomenclature: Alleles were given with their final designation (e.g.js1, js2) after they were placed in complementation groups.Within each complementation group alleles were named based on the order in which they were identified in the screen.The 1 singleton mutation was designated jsz4.
Gene expression analysis: Gene expression analysis was performed essentially as described in Patel (1994).Briefly, the larval CNS was dissected in PBS, fixed in 3.7% formaldehyde for 35 min, and washed in PTx (1×PBS, 0.1% TritonX-100) 5 times over 1 h prior to primary antibody incubation.Fixed tissues were incubated in primary antibody with gentle rocking overnight at room temperature, then washed 5 times in PTx over the course of an hour, incubated in the appropriate secondary overnight for at least 4 h to overnight at room temperature, and then washed again in PTx at least 5 times over 1 h.The CNS was then dissected and mounted in PTx or in DPX mountant after dehydration via an ethanol series and clearing in xylenes (Truman et al. 1994).All imaging was performed on a Zeiss LSM-700 Confocal Microscope, using Zen software.The following antibodies were used in the study: Deadpan anti-Sens-2 (this paper; 1:500); rabbit anti-Ppn (this paper; 1:250); goat Anti-GFP Dylight TM 488 Conjugated,preadsorbed (Rockland;.Secondary antibodies were obtained from the following sources.The Alexa Fluor Plus 488 secondary antibodies of appropriate species specificity were obtained from ThermoFisher and used at a dilution of 1:400: e.g.Donkey anti-Rat IgG Alexa Fluor Plus 488; catalog number Cat #A48269.The Cy5 secondary antibodies of appropriate species specificity were obtained from Jackson Immunoresearch and used at a dilution of 1:400: e.g.Cy5 Donkey Anti-Rat IgG, #712-175-153.Please see the Supplementary, Reagents Table for a complete description of secondary antibodies used in this study.

Antibody generation
To generate the Senseless-2 antibody, DNA encoding amino acids 1-199 of Senseless-2 was inserted into pET-29a (Novagen) for protein expression and purification.Protein-specific antibody responses were mounted in rabbits (Pocono Rabbit Farm and Laboratory, PA, USA) and the resulting sera were used at a 1:500-1:1,000 dilution.As all senseless-2 mutations arise C-terminal to the epitopes against which the antibody was generated, we confirmed the specificity of this antibody to senseless-2 by driving UAS-senseless-2 expression in subperineural glia within the ventral nerve cord, which normally do not express senseless-2, and detecting high levels of Senseless-2 protein in these cells (Supplementary Fig. 1).
Papilin antibody generation: YenZym (CA, USA) was used as a commercial source to generate affinity-purified antibodies against 2 distinct synthetic peptides that corresponded to amino acids 27-47 (RFPGLRQKRQYGANMYLPEC) and 2,670-2,691 (TRPVTQRPSYPYRPTRPAYVPE) of Ppn-PG.Briefly, each peptide was conjugated to KLH, used as an immunogen in rabbits to generate a peptide-specific antibody response, and antibodies specific to the peptide were affinity purified.The 2 affinity-purified antibodies were combined and used at a dilution of 1:250-1:500 for immunofluorescence analysis.

Generation of senseless-2 CRIMIC T2A-GAL4 line
To generate sens-2-GAL4-DBD, we used a modified version of the strategy developed by Kanca et al. (2019).We used the Genewiz company to synthesize a DNA fragment into the EcoRV site of the pUC-GW-Kan vector.This fragment is made of the left and right homology arms (HA) which are immediately adjacent to the cut site and restriction enzyme sites (SacI-KpnI) between these arms.We then directionally cloned the attp-FRT-splitGAL4-FRT-attp fragment (see below) into the middle of left and right HAs using the SacI and KpnI sites.We used the genome sequence of the injected flies (vasa-cas9, BDSC-51324) for design.
HA flanking sequences: Left HA: 5′cgtgtgtgagagaga 3′aacggtcttttccct; Right HA: 5′gggtggggcagcgcc 3′tatctatcatagata SacI-attp-FRT-splitGAL4-FRT-attp-Kpn1 fragment: We used the Gibson cloning method to clone 3 fragments into pBS-KS digested with SacI and KpnI.Primer pairs and templates are shown below: Note we also included ECoRI and EcoRV sites for replacing the trojan exon between attp-FRT sites as a back-up strategy when needed.Guide RNAs were identified with CRISPR target Finder (Gratz et al. 2014) with maximum stringency and minimal off-target effects.We used pCFD4 vector (Port et al. 2014)  All constructs were injected into vasa-cas9 (BDSC #51324), which were then crossed to Tubulin-GAL4-AD, UAS-TdTomato/ CyO, and scored for TdTomato expression to identify positive lines.Verification of targeting was confirmed via PCR-based sequencing methods.After we generated the sens-2-GAL4-DBD line, we used the strategy described in Diao et al. (2015) to replace the split-GAL4 insert with a T2A-GAL4 and then conducted all experiments with the sens-2-T2A-GAL4 line, termed sens-2-GAL4.
The papilin-T2A-GAL4 line was generated from the ppn MI03189 insert following the genetic cross scheme outlined in Diao et al. (2015).
Fly strains: Please see Supplementary Table 1 in Supplementary File 1 and Reagents Table (Supplementary File 2) for a complete list of fly stocks used in this paper.

Results and discussion
The cellular and molecular processes that control CNS shape are poorly understood.To gain insight into these processes, we performed a large-scale EMS-based, forward genetic screen of the second chromosome to identify mutations that disrupt the morphology of the larval CNS (Fig. 1).Briefly, we used a target chromosome that contains a 3xP3-RFP transgene insert in the second chromosome that labels most glia (Zhu et al. 2024), enabling rapid visualization of the CNS in live larvae.We set up 17,790 single male F1 crosses, screened 12,211 F3 lines for second chromosomal mutations that when homozygous were larval-, pupal-, or semilethal, identified 3,180 such lines, and visually screened each for defects in overall CNS morphology.We identified 50 mutations that disrupt CNS shape, including 11 mutations in LanB1, viking, or Col4a1, genes which have well-characterized roles in regulating tissue and CNS structure (Pastor-Pareja and Xu 2011; Kim et al. 2014;Skeath et al. 2017;Dai et al. 2018).The 39 remaining mutations were grouped into 3 phenotypic classes: those with a wider CNS (n = 1), a herniated or misshapen CNS (n = 15), and an elongated CNS (n = 23).Complementation crosses identified 8 complementation groups and 6 "singleton" mutations among these 39 lines, and analysis of whole genome sequencing data of most lines at ≥30× coverage identified the causative lesion(s) in and affected gene for 13 of the 14 genes.When possible, phenotypic analysis with appropriate deficiencies and genespecific alleles was used to confirm correspondence between gene and mutant phenotype.The CNS phenotype, molecular nature, and known function of these genes are summarized in Table 1 and Fig. 2.
Below, we outline the genes identified in the screen and their associated mutations and CNS phenotypes, starting with the gene/mutation that yielded a wider CNS and ending with those that yielded an elongated CNS.We describe in greater detail the function of the senseless-2 gene, which encodes a Zinc finger transcription factor that may act as a genetic switch to distinguish perineurial glial found in the PNS from those in the CNS.

Gene that yields a widened CNS phenotype when mutated diaphanous:
We identified 1 mutant allele that when homozygous yielded a wider CNS.This allele identified a frameshift mutation found immediately after the codon for amino acid 51 in the diaphanous coding region.Larvae homozygous for this allele or transheterozygous for it and dia 5 or Df(2L)ED1315, a deficiency of the region, resulted in an expansion of the thoracic region of the CNS that became more pronounced as larvae were about to pupariate (Table 1; Figs. 2 and 3b).diaphanous encodes the sole Dia-class formin protein in Drosophila; it is required for cytokinesis and induces polyploidy in follicle cells and mitotic neuroblasts (Castrillon and Wasserman 1994).Using antibodies specific to Deadpan and ELAV to label neuroblasts and neurons, respectively, we observed a massive increase in neuroblast size likely caused by defects in cytokinesis and increased ploidy of these cells (Fig. 3c-f).Our results indicate that the expanded nature of the thoracic ventral nerve cord in diaphanous mutants likely results from the expanded size of neuroblasts and potentially other cell types in the CNS.

Genes that yield a misshapen CNS phenotype when mutated
We identified mutations in 3 genes-Dark, Toll-2, and 1 singleton mutation-that gave rise to a misshapen CNS (Table 1).We identified 11 alleles of Dark.Homozygosity for each allele or transheterozygosity for most combinations of these alleles yielded a mild, partially penetrant CNS phenotype in which small hernias arise in the CNS (Fig. 2).This phenotype was also recapitulated when each allele was placed over a deficiency of the region, Df(3R)BSC331.We sequenced the 11 alleles, each of which contained an apparently identical 370-bp deletion within the coding region (Table 1).The identity of the molecular lesion in the alleles suggests the mutations arose in mitotic spermatagonia or spermatocyte cells and are not independent of each other.Dark encodes Zinc finger C2H2 transcription factor a homolog of the Apaf-1/Ced-4 proapoptotic genes and has been shown to promote hyperplasia in the CNS (Rodriguez et al. 1999).
Toll-2: Three allelic mutations identified the Toll-2 (18-wheeler) gene with 2 of the mutations being premature stop codons and the third being a missense mutation in the leucine-rich repeat domain (Table 1; Fig. 4a).Homozygosity for each allele or transheterozygosity for all combinations of these alleles yielded a phenotype in which the CNS was reduced in size and the ventral nerved cord was misshapen and marked by irregular protrusions, hernias, and indentations (Figs. 2 and 4b-i).These mutant phenotypes were recapitulated when the Toll-2 alleles were placed over Df(2R)BSC594, a deficiency of the region.The CNS phenotype of the Toll-2 mutants grossly resembles that of perlecan, which encodes one of the key components of the CNS basement membrane (Yurchenco 2011), hinting that Toll-2 receptors may function in surface glia to regulate CNS morphology by interfacing with components of the basement membrane.Cell death may also contribute to the observed defects in CNS morphology, as loss of Toll-2 function has been shown to drive cell death in neurons and neuroblasts (Li et al. 2020;Hermanstyne et al. 2022).
The jsZ4 allele represents the sole mutation that we were unable to map to a defined gene (Table 1).Its partially penetrant phenotype is highlighted by an elongated nerve cord and the presence of hernias and protrusions in it (Fig. 2).As we sequenced the genome of this allele at ≥30× coverage, our failure to map its mutant phenotype to a specific gene could arise due to the phenotype being caused by 2 or more mutations, a noncoding mutation, or other reasons.

Genes that yield an elongated CNS phenotype when mutated
Our largest phenotypic class-an elongated CNS-was identified by 23 mutations and 9 genes.Two of the 9 genes are known tumor suppressors-lethal giant larvae (lgl) and brain tumor (brat).We identified a single mutation in lgl, a frameshift mutation which when homozygous yields an elongated ventral nerve cord and greatly enlarged brain lobes (Table 1; Fig. 2).Complementation crosses against the lgl 4 allele and a deficiency that uncovers lgl, Df(2L)ED50001, confirmed correspondence between the gene and mutant phenotype.The elongated CNS phenotype likely arises from elevated cell proliferation since lgl plays a well-characterized role in restricting cell proliferation (Gateff and Schneiderman 1974).Like lgl, loss of brat function is known to increase neuroblast and cell proliferation and result in tumorous growths in the brain (Gateff 1994).We identified 3 mutations in brat that yielded a moderate CNS elongation phenotype and enlarged brain when homozygous or trans-heterozygous to each other or Df(2L)Exel8040, a deficiency of the region (Table 1; Fig. 2).
worniu: We identified a nonsense mutation, L117*, in worniu that yielded a moderately elongated nerve cord (Table 1; Fig. 2).We confirmed correspondence between worniu and the mutant phenotype based on the worniu js1 allele failing to complement the worniu 1 and worniu 2 alleles as well as Df(2L)Exel8034, which uncovers worniu, for the elongated nerve cord phenotype.worniu encodes a C2H2-type Zinc finger transcription factor that is expressed in most neuroblasts and has been shown to regulate CNS structure (Ashraf et al. 2004).
Tango1: We identified a missense mutation, T60I, in the Tango1 coding region that when homozygous resulted in a moderately elongated CNS (Table 1; Fig. 2).We confirmed correspondence between this mutation and the observed CNS elongation phenotype by complementation against 2 overlapping deficiencies that each uncover Tango1, Df(2L)BSC6, and Df(2L)BSC188.Tango1 functions at ER exit sites to promote protein secretion and is required to promote the specialized secretion of the large Collagen and Laminin proteins (Saito et al. 2009  Rme-8: Three mutations in the Rme-8 gene also gave rise to a moderately elongated CNS phenotype when homozygous or trans-heterozygous to each other (Table 1; Figs. 2 and 5).This phenotype was recapitulated when each mutation was placed over Df(2R)BSC279, a deficiency of the region.Rme-8 encodes a DnaJ domain-containing protein of the Hsp40 chaperone family that also contains 4 IWN repeat (Zhang et al. 2001;Norris et al. 2022); Rme-8 promotes endocytic recycling of transmembrane proteins, such as Notch (Gomez-Lamarca et al. 2015).Using an Rme-8-T2A-GAL4 CRIMIC insert in the fourth intron of Rme-8 to drive a UAS-linked nuclear-RFP transgene (Lee et al. 2018;He et al. 2019), we found that Rme-8 is broadly expressed in the CNS in neurons and glia, with high-level expression in surface glia (Supplementary Fig. 2).
Pvf3: We identified a single mutation in Pvf3-a splice site donor mutation just prior to the region that encodes its PDGF domain (Table 1; Fig. 6a).Larvae homozygous for this mutation yield a highly elongated CNS, which is recapitulated when this allele is placed in trans to Df(2L)ΔPvf2-3, Df(2L) BSC291, or 2 P element inserts in Pvf3 (Fig. 6c-e), confirming correspondence between Pvf3 and the elongated CNS phenotype.Pvf3 acts together with Pvf2 to promote hemocyte survival and migration through activation of the Pvr receptor (Parsons and Foley 2013).Loss of Pvr function or blockade of neural activity has been shown to inhibit nerve cord condensation and promote nerve cord elongation (Olofsson and Page 2005).Our observation that loss of Pvf3 function alone yields a highly elongated CNS phenotype suggests that Pvf3 plays a nonredundant role relative to Pvf2 to promote hemocyte survival, migration, and/or function.Four mutant alleles identified CG9171, which encodes a predicted glucuronosyltransferase that promotes O-linked protein mannosylation.All 4 mutations reside in its extracellular glycosyltransferase domain, likely disrupting enzymatic function (Fig. 7a).Homozygosity or trans-heterozygosity for all possible combinations of the 4 mutations yielded a highly elongated CNS phenotype, but neuroblast formation and neuronal patterning appeared grossly normal (Fig. 7b-g).This phenotype was recapitulated when each allele was placed in trans to Df(2L)ED334, a deficiency of the region.The most closely related mammalian homologs of CG9171-B4GAT1 and LARGE1/2-act in tandem to drive the O-mannosylation of Dystroglycan (Praissman et al. 2014).B4GAT1 adds a single glucuronic acid residue onto an Xylose acceptor on Dystroglycan.The related glycosyltransferase LARGE recognizes this epitope and extends it by adding many copies of a repeating disaccharide to form matriglycan.Matriglycan links Dystroglycan to the basement membrane by binding to Laminin and other proteins and plays a key role in disease, as  Prior work on the glucuronyltransferase, dGlcAT-P, the fly ortholog of B3GAT1, revealed that it acts in hemocytes to regulate CNS structure (Pandey et al. 2011).Loss of dGlcAT-P function results in a highly elongated CNS phenotype similar to that observed for mutations in CG9171, with additional studies indicating the dGlcAT-P phenotype arises due to mechanical stretching of the CNS (Pandey et al. 2011).Forced expression of dGlcAT-P in hemocytes, but not in glia or neurons, fully rescued its CNS mutant phenotype, indicating that dGlcAT-P acts in hemocytes to govern CNS structure.These results support the idea that CG9171 acts in hemocytes to control CNS structure; expression studies support this idea, as they show strong CG9171 expression in hemocytes (Ozturk-Colak et al. 2024).
C1GalTA: Like CG9171, the 2 alleles in Core 1 Galactosyltransferase A (C1GalTA; CG9520) yielded a highly elongated CNS phenotype (Table 1; Figs. 2 and 8), which was recapitulated when each allele was placed over a deficiency of the region, Df(2L)BSC204.Unlike the CG9171 phenotype in which the mutant ventral nerve cord is supple and flexible, C1GalTA mutant larvae exhibited a highly elongated, very brittle nerve cord and flat and misshapen brain lobes (Fig. 8b-e).These phenotypes are most obvious in larvae homozygous for the phenotypically stronger C1GalTA js1 allele.C1GalTA promotes protein glycosylation by adding galactose in a β1,3 linkage to N-acetylgalactosamine (GalNAc) on proteins.In  a prior study showed that dC1GalTA is required for glycosylation of Laminin and the presence of T antigen (Gal β1,3 GalNAc) on hemocytes, and that loss of C1GalTA function drives CNS elongation (Lin et al. 2008).
The C1GalTA mutant phenotype resembles that of papilin (ppn), a third chromosomal gene that we were working on in parallel to the genetic screen due to its sequence similarity to AdamTS genes (Campbell et al. 1987;Kramerova et al. 2000;Keeley et al. 2020).Ppn is a key component of the basement membrane, where it has been shown to facilitate Collagen IV removal to promote basement membrane remodeling (Keeley et al. 2020).Like the C1GalTA mutant phenotype, the reduction of ppn function yields a highly elongated and brittle nerve cord and flat, misshapen brain lobes (Fig. 9a,b).Ppn-specific antibodies showed that ppn is expressed in hemocytes, the cells that assemble and disassemble basement membranes, but not in most other tissues (Fig. 9c,d), including the fat body-the major producer of basement membrane proteins in larvae (Pastor-Pareja and Xu 2011).We hypothesize that hemocytes deposit Ppn on most tissues during development.
Most basement membrane components, like Col4a1, Vkg, Nidogen, and Perlecan, are produced by fat body cells (Pastor-Pareja and Xu 2011), but Ppn is not expressed in the fat body, suggesting there may be a physiological reason for the exclusion of Ppn expression in the fat body.To test this model, we misexpressed ppn in the fat body and asked if it altered basement membrane protein localization by visualizing the localization of a GFP protein trap in Viking (Morin et al. 2001;Buszczak et al. 2007).Upon forced ppn expression in the fat body, we observed a massive, cell-autonomous retention of Viking-GFP in fat body cells (Fig. 9e-g), demonstrating the incompatibility of fat body ppn expression with appropriate Viking secretion and suggesting that Ppn physically associates with Viking in vivo to help mediate the removal of the basement membrane components.The similarity of the ppn and C1GalTA mutant phenotypes suggests that C1GalTA acts with Ppn to promote Viking/Collagen IV removal from the basement membrane and that failure to do so results in a brittle, elongated CNS.Future work that determines how these genes functionally interface with each other should help clarify the molecular basis of basement membrane remodeling and the control of tissue structure.
infertile crescent: Three alleles identified infertile crescent (ifc; Table 1), which encodes the sole Drosophila dihydroceramide desaturase that converts dihydroceramide to ceramide in the last step of the de novo ceramide biosynthesis pathway (Jung et al. 2017;Hannun and Obeid 2018).Loss of ifc results in an elongated CNS marked by bulges in peripheral nerves (Fig. 2); we characterized the function of ifc in detail in a separate study and found it functions primarily in glia to promote glial homeostasis and to guard against neuronal cell death (Zhu et al. 2024).

Senseless-2 acts in perineurial glia to regulate CNS structure
Four mutant alleles identified the senseless-2 (sens-2) gene, which encodes a member of the Zinc finger C2H2 superfamily of proteins and possesses 6 C2H2 Zinc finger domains (Figs. 2 and 10a).Three of the alleles-sens-2 js2 , sens-2 js3 , and sens-2 js4 -introduce, respectively, an early splice site mutation, a missense mutation in a conserved cysteine in the Zinc finger domain, and an early frameshift mutation (Table 1).These alleles when homozygous or transheterozygous to each other or a deficiency of the region, Df(2L) ED6569, yielded an essentially identical highly elongated CNS phenotype (Fig. 10b,c) and are likely null or strong hypomorphic alleles.The fourth allele, sens-2 js1 , yields a modest CNS elongation phenotype; larvae trans-heterozygous for sens-2 js1 and sens-2 js4 exhibit an intermediate CNS elongation phenotype between the 2 alleles, identifying sens-2 js1 , which contains a missense mutation in the zinc finger domain (Table 1; Fig. 10a), as a weak hypomorphic allele of sens-2.The sens-2 CNS elongation phenotype manifests first instar larvae and is maintained throughout larval life and into the pupal stage at which point sens-2 homozygous flies die.In live third instar larvae, the CNS appears to be under tension from both laterally and posteriorly projecting peripheral nerves, with the posterior nerves being shorter in length in sens-2 mutant larvae relative to wild-type (Figs. 2 and 10b).The overall shape of the CNS is dramatically altered, but the overall pattern and number of neuroblasts and neurons appeared grossly normal (Fig. 10c).
To clarify the cellular basis for the sens-2 CNS elongation phenotype, we tracked sens-2 expression in larvae by generating a Sens-2 specific antibody and a sens-2-T2A-GAL4 CRIMIC line (sens-2-GAL4) and leveraging a sens-2-GFP transgene surrounded by its endogenous genomic locus (Kudron et al. 2018; see Materials and Methods).These tools revealed that sens-2 is expressed in subsets of glia and neurons in the larval nervous system (Fig. 10d-f; Supplementary Figs. 3 and 4).We focused on sens-2 expression in the nervous system, as its function in glia, neurons, or both likely contributes to the observed CNS phenotype.Double-labeling experiments with REPO to mark all glia and ELAV to label all neurons revealed that sens-2 is expressed in a small number of CNS neurons and most glia in peripheral nerves, but that sens-2 is not detected in any glia within the CNS itself (Fig. 10d,e; Supplementary Figs. 3 and 4).Recent scRNA analysis of the larval CNS supports the restriction of sens-2 expression to surface glia and subsets of neurons (Nguyen et al. 2024).More detailed analysis of sens-2 expression within peripheral nerves of late-third instar larvae suggested that sens-2 is expressed in all perineurial glia and weakly in some subperineural glia, but that its expression appears to be excluded from wrapping glia and some subperineurial glia (Fig. 10f; Supplementary Figs. 3 and 4).We conclude that sens-2 expression demarcates PNS perineurial glia (sens-2-positive) from CNS perineurial glia (sens-2-negative), identifying sens-2 as one of few markers to distinguish a glial subtype along the CNS-PNS axis.
To determine if sens-2 function is required in glia, neurons, or both to regulate CNS structure, we used the GAL4/UAS system together with either a UAS-sens-2-RNAi transgene or a UAS-sens-2 transgene (Fig. 10g).Depletion of sens-2 function specifically in neurons via RNAi using elav-GAL4 together with repo-GAL80 had effect on CNS structure or morphology (n = 11/11; Fig. 10g), but depleting sens-2 function specifically in glia using repo-GAL4 recapitulated the elongated nerve cord phenotype observed in sens-2 mutant larvae (n = 11/11; Fig. 10g).We note that using elav-GAL4 in the absence of repo-GAL80 resulted in elongated nerve cords (n = 11/12) due to the ability of elav-GAL4 to drive significant gene expression in glia (Berger et al. 2007).We observed similar results with the sens-2-GAL4 line: RNAi-mediated depletion of sens-2 in all sens-2 expressing cells elicited an elongated nerve cord (n = 12/12), but when repo-GAL80 was used to block sens-2 RNAi in glia, the phenotype disappeared, and nerve cords appeared wild-type (n = 11/11).Gene rescue experiments corroborated the above results (Fig. 10g): Expression of a wild-type sens-2 transgene under control of the sens-2-GAL4 line in otherwise sens-2 mutant larvae fully rescued the sens-2 CNS phenotype (n = 10/10), but when repo-GAL80 was used to block transgene expression specifically in glia, the sens-2 elongated CNS phenotype reappeared (n = 10/ 10).We conclude that sens-2 function is required solely in PNS perineurial and perhaps subperineural glia to control CNS structure, and that loss of sens-2 function in glial cells alone is sufficient to yield the observed elongated nerve cord phenotype.
To assess the function of sens-2 in different glial subtypes, we leveraged the GAL4/UAS system and GAL4 lines specific for each glial subtype to drive sens-2 expression and that of Myr-GFP, which outlines cell morphology, in each glial We observed no gross change in cell morphology upon sens-2 misexpression in astrocyte-like, cortex, and ensheathing glia.Forced expression of sens-2 by 2 different perineurial-specific GAL4 lines, which also drive strong gene expression in the gut, resulted in early larval lethality inhibiting our ability to assess the impact of sens-2 misexpression in this glial subtype.Misexpression of sens-2 in subperineural glia, however, drove a clear change in cell morphology (Fig. 10h).Subperineural glia normally form a thin, flat cell layer that fully encircles the circumference of the CNS and peripheral nerves and resides immediately interior to the perineurial glial cell layer (Yildirim et al. 2019).Upon sens-2 misexpression in subperineural glia, these glial cells still fully encircle the CNS, but they now also extend cell membranes into the interior of the CNS to fully or partially enwrap individual neurons (arrows, Fig. 10h).sens-2 misexpression in subperineural glia then modifies the behavior of this glial cell type, bestowing on it the ability to enwrap neuronal cell bodies in addition to the entire CNS, suggesting that sens-2 alters the membrane properties of surface glia to facilitate their ability to wrap peripheral nerves.
Our work on sens-2 suggests that it acts as a genetic switch to distinguish the functional properties of surface glia found In the PNS from those found in the CNS.A key difference between these 2 tissues is their diameter-peripheral nerves are tiny in diameter relative to the much larger CNS.In this context, the ability of forced sens-2 expression to alter the behavior of subperineural glia so that they inappropriately enwrap adjacent neuronal cell bodies highlights a profound impact of sens-2 expression on key functional properties of surface glia-their ability to enwrap (or not enwrap) small diameter structures, such as cells or nerves.Our work did not clarify the molecular basis through which sens-2 dictates such functional properties.In the future, it will be important to identify the transcriptional targets of sens-2 to clarify the exact mechanism through which it governs the development and differentiation of surface glia in the PNS.
Summary: Prior work from many laboratories has highlighted the importance of interactions between the basement membrane and glial cells in dictating CNS structure (Kim et al. 2014;Meyer et al. 2014;Skeath et al. 2017).Initial results from our screen reinforce these findings.Tango1, C1GalTA, ppn, and pvf3 all appear to act on basement membrane proteins, form components of the basement membrane, or regulate the survival or migration of hemocytes, the bricklayers, and tuckpointers of the basement membrane.Continued work on these genes, especially C1GalTA and ppn, which likely act in the same pathway, holds the promise of clarifying our understanding of basement membrane function and remodeling on tissue structure.Conversely, sens-2 and ifc act in glial cells to regulate CNS structure (this paper; Zhu et al. 2024).We expect Rme-8 and Toll-2 also act in glia to regulate CNS shape, with future work needed to confirm this prediction and to clarify if and how such factors interface with the basement membrane to govern CNS morphology.

Fig. 1 .
Fig. 1.Diagram of the genetic cross scheme used to execute the forward genetic screen.

Fig. 2 .
Fig. 2. Representative examples of CNS mutant phenotypes were identified in the genetic screen.Ventral views of wild-type and mutant late-third instar larvae of the indicated genotype showing 3xP3 RFP staining in grayscale to highlight CNS morphology.Anterior is up; scale bar is 500 μm.

Fig. 3 .
Fig. 3. Mutations in diaphanous increase the width of the CNS.a) Schematic of Diaphanous protein with location and nature of the dia js1 allele.b) Ventral views of wild-type or dia js1 mutant late-third instar larvae with 3xP3 RFP expression (grayscale) highlighting the CNS.Anterior is up; scale bar is 250 μm.c-f) Ventral views of dissected CNS (c, d) or high magnification view of the brain (e, f) from wild-type and dia js1 mutant late-third instar larvae labeled for neuroblasts with Deadpan (Dpn is green in panels c and d and gray in panels e and f) and for neurons with ELAV (magenta in c, d).Arrows highlight enlarged neuroblasts; dotted lines demarcate the presumed boundary of 1 enlarged neuroblast.Anterior is up; scale bar is 100 μm in panels c and d and 50 μm in panels e and f.

Fig. 4 .
Fig. 4. Loss of Toll-2 function results in a misshapen CNS.a) Schematic of Toll-2 protein with location and nature of identified Toll-2 alleles.b-e) Ventral views of wild-type or Toll-2 mutant late-third instar larvae in which 3xP3 RFP expression highlights the CNS.Arrows highlight hernias, bulges, or dysmorphology in the CNS.Anterior is up; scale bar is 500μm.f-i) Ventral views of dissected CNS of wild-type (f) and Toll-2 mutant (g-i) late-third instar larvae labeled for neuroblasts with Deadpan (Dpn; green) and neurons with ELAV (magenta).Arrows point to areas of the CNS that display dysmorphology.Anterior is up; scale bar is 200 μm.

Fig. 5 .
Fig. 5. Loss of Rme-8 function promotes CNS elongation.a) Schematic of Rme-8 protein showing the location of the lipid binding, IWN repeat, and DnaJ domains as well as the location and nature of the 3 identified Rme-8 alleles.b-e) Ventral views of wild-type or Rme-8 mutant late-third instar larvae in which 3xP3 RFP expression highlights the CNS.Anterior is up; scale bar is 500 μm.f-g) Ventral views of dissected CNS of wild-type (f) and Rme-8 mutant (g) late-third instar larvae labeled for neuroblasts with Dpn (green) and neurons with ELAV (magenta).Anterior is up; scale bar is 200 μm.

Fig. 6 .
Fig. 6.Loss of Pvf-3 function greatly increases the length of the ventral nerve cord.a) Schematic of Pvf-3 protein with location and nature of the Pvf-3 js1 allele.b,c) Ventral views of wild-type or Pvf-3 js1 mutant late-third instar larvae in which 3xP3 RFP expression highlights the CNS.Anterior is up; scale bar is 500μm.d,e) Ventral views of dissected CNS of wild-type (d) and Pvf-3 js1 /Df(2L)Pvf2-3mutant (e) late-third instar larvae labeled for neuroblasts with Deadpan (Dpn; green) and for neurons with ELAV (magenta).Anterior is up; scale bar is 200 μm.

Fig. 7 .
Fig. 7. Loss of CG9171 drives CNS elongation.a) Schematic of CG9171 protein with location and nature of the 4 identified CG9171 alleles.b-e) Ventral views of wild-type or CG9171 mutant late-third instar larvae in which 3xP3 RFP expression highlights the CNS.Anterior is up; scale bar is 500 μm.f-g) Ventral views of dissected CNS of wild-type (f) and CG9171 mutant (g) late-third instar larvae labeled for neuroblasts with Dpn (green) and neurons with ELAV (magenta).Anterior is up; scale bar is 200 μm.

Fig. 8 .
Fig. 8. Loss of dC1GalTA function promotes CNS elongation.a) Schematic of C1GalTA protein with the location of its Fringe-like glycosyltransferase domain and the nature of the 2 identified alleles.b,c) Ventral views of wild-type or C1GalTA js1/js2 mutant late-third instar larvae in which 3xP3 RFP expression highlights the CNS.Anterior is up; scale bar is 500 μm.d,e) Ventral views of dissected CNS of wild-type (d) and C1GalTA js1/js2 mutant (e) late-third instar larvae labeled for neuroblasts with Deadpan (Dpn; green) and for neurons with ELAV (magenta).Anterior is up; scale bar is 100 μm.

Fig. 9 .
Fig. 9. Papilin regulates CNS structure and is sufficient to drive the accumulation of Vkg-GFP in the fat body.a,b) Ventral view of CNS of wild-type (a) and ppn MI03189 mutant late-third instar larvae (b) labeled for ELAV.Anterior is up; scale bar is 100 μm.c,d) Stage 15-16 embryos labeled for Ppn (gray, c; magenta, d-d′) and Collagen 4a1 (green, d).Anterior is to the left; scale bars are 50 μm.Arrowheads in d-d′ point to Ppn + hemocytes immediately adjacent to the fat body.e-g) Vkg-GFP localization (gray in e, f; green in g) in wild-type fat body (e), fat body in which Ppn[RG] is expressed in all fat body cells (f), and fat body in which small cell clones, marked in magenta, ectopically express Papilin protein (g).Scale bar is 50 μm for panels e and f and 25 for panel g.

Fig. 10 .
Fig. 10.sens-2 acts in peripheral glia to regulate CNS structure.a) Schematic of Sens-2 protein with location and nature of the 4 identified sens-2 mutations.b) Ventral views of late-third instar larvae of indicated genotype showing 3xP3 RFP expression in the CNS and nerves.Anterior is up; scale bar is 500 μm.c) Ventral views of photomontages of late-third instar larvae of indicated genotypes labeled for neuroblasts with Deadpan (Dpn; green) and for neurons with ELAV (magenta).Anterior is up; scale bar is 200 μm.d-f) Ventral views of low (d) and high magnification (e, f) images of the CNS(d, e) and peripheral nerves (f) of late-third instar larvae labeled for sens-2-GFP (green and grayscale) to mark sens-2 expressing cells and REPO (magenta) to mark glia.Arrows in d-e identify peripheral glia that express sens-2-GFP; arrowheads identify central glia that lack GFP expression.Arrowhead in f marks sens-2-GFP-negative wrapping glia in peripheral nerve.g) Ventral views of the dissected CNS from late-third instar larvae of indicated genotype labeled for ELAV to mark neurons and highlight the structure of the CNS.Anterior is up; scale bar is 200 μm.h) Ventral view and Y-Z and X-Z projection views of late-third instar larvae of indicated genotype labeled for Myr-GFP (green/grayscale) and ELAV to label neurons (magenta).Arrows point to Myr-GFP + membrane invaginations that enwrap neurons.Anterior is up; scale bar is 20 μm.

Table 1 .
Molecular nature, phenotypic class, and causative lesions of identified genes.