Perlecan is an evolutionary conserved HSPG with a similar domain architecture in invertebrates, vertebrates, and the early multicellular eukaryote Trichoplax adhaerens (Warren et al., 2015). Drosophila Perlecan is encoded by the trol locus, which resides on the X chromosome and encodes 25 predicted Perlecan splice variants ranging in size from 2853 to 4489 amino acids (Figure 1A). To compare the relationship of Drosophila Perlecan to other secreted HSPGs, a phylogenetic tree was constructed using homologs of Perlecan, Agrin and Drosophila Carrier of Wingless (Cow). One of the longest isoforms of Drosophila Perlecan (Trol-RAT) was used for the analysis. Ciona intestinalis, Danio rerio, Mus musculus, Rattus norvegicus, Homo sapiens, Caenorhabditis elegans, and Trichoplax adhaerens homologs were identified with NCBI blast searches. FASTA sequences of the longest isoform from each species was used in a Clustal Omega multiple sequence alignment and visualized in Jalview as an average distance phylogenetic tree using the BLOSUM62 algorithm (Figure 1B). The Trichoplax Perlecan and Agrin homologs were the most distantly related, but still clustered within their specific subfamily. Drosophila Perlecan clustered in a leaf with other Perlecan homologs and distinct from the Agrin family. Although Agrin plays a key role in cholinergic NMJ development (Gautam et al., 1996; Nitkin et al., 1987; Sanes and Lichtman, 2001), Drosophila contains glutamatergic NMJs and lacks an Agrin homolog (Littleton and Ganetzky, 2000). Drosophila Cow was more closely aligned with Agrin homologs than the Perlecan family.

Given the absence of an Agrin homolog in Drosophila, Perlecan might play similar roles in organizing Drosophila synaptic proteins. Indeed, a previous study identified defects in GluRIIA receptor clustering and synaptic Wg diffusion in trol mutants (Kamimura et al., 2013). These data suggested Perlecan may regulate organization of Drosophila synapses, prompting us to further evaluate its function. To examine Perlecan localization within the larval PNS, an endogenous trol^GFP insertion allele from the FlyTrap protein-tagging library was characterized (Morin et al., 2001). Trol^GFP was enriched along nerve bundles (Figure 1C–F) and present at lower levels on the surface of body wall muscles (Figure 1C and G–I). The enrichment of Perlecan around nerve bundles is consistent with its localization within the neural lamella, a large ECM compartment that surrounds axons and glia of the CNS and PNS. Indeed, Perlecan has been previously observed within the neural lamella surrounding the VNC and peripheral nerves (Brink et al., 2012; Skeath et al., 2017). In addition, the localization of Viking (Vkg), the Drosophila secreted type IV Collagen homolog and a known component of the neural lamella (Yasothornsrikul et al., 1997), was disrupted in trol mutants (see below). Although immunogold electron microscopy (EM) identified Perlecan in the subsynaptic reticulum (SSR) surrounding NMJs (Kamimura et al., 2013), Trol^GFP did not display synaptic enrichment beyond the homogenous expression over the entire muscle surface (Figure 1G–I). To confirm Trol^GFP signal was specific to Perlecan, an RNAi construct targeting trol (UAS-trol-RNAi.1, Figure 1A) was recombined with Trol^GFP and driven with the ubiquitous tubulin-Gal4 driver. Trol^GFP was eliminated by co-expression of the RNAi, with no signal observed along nerve bundles or on the muscle surface (Figure 1D–I). The Trol^GFP line also displayed normal NMJ growth and maintenance (Figure 1—figure supplement 1), in contrast to trol mutants (see below), indicating endogenous Trol^GFP produces a functional Perlecan protein.

To examine a role for Perlecan in synaptic development and function, a previously generated null mutant that deletes the trol locus (Figure 1A) was characterized (Voigt et al., 2002). Male trol^null larvae are smaller than their heterozygous female trol^null/+counterparts and display disrupted locomotion and lethality during the 3rd instar stage (Datta and Kankel, 1992; Voigt et al., 2002). To examine synaptic morphology in trol^null and heterozygous control 3rd instars, immunostaining was performed at muscle 4 NMJs for presynaptic Complexin (Cpx) and postsynaptic Discs-large (Dlg). In trol^null/+controls, Cpx and Dlg colocalized at NMJ boutons (Figure 2A). Although some trol^null/y larvae had intact NMJs, many trol NMJs displayed Dlg +boutons that lacked presynaptic Cpx (Figure 2A). The presence of postsynaptic ‘footprints’ lacking presynaptic material is a defining feature of mutants that undergo synaptic retraction (Eaton et al., 2002; Graf et al., 2011; Koch et al., 2008; Pielage et al., 2008; Pielage et al., 2005). Trol^null mutants also displayed fewer synaptic boutons (Figure 2B), in addition to increased synaptic footprints (Figure 2C), compared to control NMJs in abdominal segments A3-A5 (bouton number p values: 0.0379 for A3, 0.0012 for A4,<0.0001 for A5). Consistent with other retraction mutants (Graf et al., 2011), the trol^null phenotype was more severe in posterior abdominal segments (Figure 2B and C). NMJs showing severe synaptic retraction, defined by retraction footprints and decreased number of synaptic boutons two standard deviations compared to controls, were only observed in trol^null larvae, with increasing severity in posterior abdominal segments (Figure 2D). Together, these data suggest NMJs are lost over development in larvae lacking Perlecan.

To determine if synaptic loss is specific to the absence of Perlecan and not the trol^null genetic background, NMJs were examined in trol^null larvae in trans to a deficiency (Df(1)ED411) that removes the trol locus. Df(1)ED411/+ heterozygous NMJs appeared normal and lacked postsynaptic footprints. In contrast, Df(1)ED411/trol^null NMJs had significantly reduced bouton number (p=0.0048), with 50% of NMJs showing postsynaptic footprints (Figure 2—figure supplement 1A–C). In addition, expression of UAS-trol-RNAi.2 with the ubiquitous tubulin-Gal4 driver resulted in reduced bouton number, with >65% of NMJs showing postsynaptic footprints compared to RNAi or Gal4 only controls (Figure 2—figure supplement 1D–F).

The experiments described above examined synaptic retraction of type Ib glutamatergic MNs. However, larval muscles are also innervated by type Is glutamatergic and type II and III neuromodulatory MNs. The two glutamatergic MNs have distinct morphology and physiology, with tonic-like (Ib) or phasic-like (Is) properties (Aponte-Santiago and Littleton, 2020). To assay if Perlecan is required for stability of other MN subtypes, synaptic retraction was quantified at Is NMJs in trol^null larvae. Similar to Ib, trol^null Is NMJs displayed a significant reduction in bouton number (p=0.0011 A3, p=0.0238 A4, p=0.0068 A5), with >30% showing severe retraction phenotypes compared to heterozygous controls (trol^null/+, Figure 2—figure supplement 2A–C). In addition to glutamatergic MNs, type II and III neuromodulatory MNs also displayed missing NMJs (data not shown). Together, these data indicate synaptic retraction occurs across all MN subtypes in 3rd instar larvae lacking Perlecan.

A previous study identified changes in pre- and postsynaptic Wg levels at trol mutant NMJs, suggesting Perlecan restricts Wg diffusion within the synaptic cleft. The absence of Perlecan resulted in enhanced presynaptic and reduced postsynaptic Wg signaling (Kamimura et al., 2013). Although synaptic retraction has not been previously associated with the Wg pathway at Drosophila NMJs (Franco et al., 2004; Mathew et al., 2005; Mosca and Schwarz, 2010; Packard et al., 2002; Restrepo et al., 2022), genetic interaction studies were conducted to assay whether increased presynaptic Wg output might contribute to synaptic retraction phenotypes observed in trol^null mutants. A UAS construct expressing a constitutively active (CA) form of the Drosophila GSK3 serine/threonine kinase Shaggy (Sgg, UAS-sgg^S9A), which dominantly blocks Wg signaling (Cook et al., 1996; Siegfried et al., 1992), was expressed in glutamatergic MNs of trol^null larvae using vGlut-Gal4. NMJ bouton number and synaptic retraction were quantified at muscles 6 and 7 in UAS-sgg^S9A controls, in trol^null, vGlut-Gal4, and in trol^null mutants expressing CA-Sgg (trol^null; vGlut-Gal4 >sggS^S9A). Like muscle 4, muscle 6/7 NMJs showed reduced bouton number (P=0.0174) and postsynaptic footprints in >90% of trol^null larvae, indicating synaptic retraction is not restricted to MNs innervating muscle 4 (Figure 3A–C). Blocking presynaptic Wg signaling with CA-Sgg did not suppress the loss of boutons or prevent synaptic retraction in trol^null mutants (Figure 3A–C, p=0.8814 between trol and trol, sgg^CA bouton numbers). These data indicate Perlecan’s function in controlling synaptic stability is independent of its effects on Wg signaling.

Although postsynaptic footprints are a hallmark of synaptic retraction, it is possible that Perlecan loss disrupts early MN or synaptic development such that synapses classified as retracted never contained presynaptic material. To dynamically visualize NMJ development, serial intravital imaging of larval muscle 26 NMJs was performed over 4 days in trol^null and heterozygous controls following brief anesthesia as previously described (Akbergenova et al., 2018). Larvae containing endogenously tagged nSynaptobrevin (nSyb^GFP) (Guan et al., 2020) and a construct expressing GluRIIA^RFP (Schmid et al., 2008) were used to visualize presynaptic vesicles and postsynaptic receptors, respectively (Figure 4A). Although control and trol^null larvae had similar NMJ area at the 2nd instar stage when imaging began (Figure 4B), only control NMJs continued to grow on subsequent days of imaging (Figure 4C). In contrast, both pre- and postsynaptic area declined in trol^null larvae during the 3rd instar imaging window. Control NMJs had greater pre- than postsynaptic area, while trol^null larva had smaller nSyb^GFP than GluRIIA^RFP area by the final day of imaging, consistent with loss of presynaptic material and lingering postsynaptic footprints (Figure 4C). Several patterns of presynaptic loss were observed in trol^null NMJs during serial imaging (Figure 4A). In some cases, an entire branch of an axonal arbor was lost between imaging days. At other NMJs, presynaptic material was absent from internal boutons in an axon branch, with proximal and distal boutons from the same axon containing nSyb^GFP. These findings confirm that trol^null MNs form NMJs that are subsequently retracted during development.

In vivo imaging indicated trol^null NMJs are morphologically intact prior to retraction. However, loss of Perlecan could cause functional disruption of synaptic output earlier in development. To assay synaptic function at intact versus retracting NMJs in trol^null 3^rd instars, two-electrode voltage-clamp (TEVC) electrophysiology was performed at muscle 6. Following physiological recordings, dissected trol^null larvae were bathed with fluorescent anti-Hrp to visualize NMJs and determine if they were intact or retracted at the recording site (Figure 4D). At intact trol^null NMJs, nerve stimulation resulted in evoked release amplitude similar to controls, indicating normal presynaptic output prior to retraction (Figure 4E and F, p=0.3189). In contrast, fully retracted trol^null NMJs completely lacked evoked release (Figure 4E and F, p=0.0001). Quantal imaging with postsynaptic membrane-tethered GCaMP7s (Akbergenova et al., 2018; Melom et al., 2013) revealed a similar loss of spontaneous release in retracted NMJs (data not shown). When evoked responses from intact and retracted trol^null NMJs were combined and averaged, a significant reduction in evoked excitatory junctional current (eEJC) was observed compared to controls (Figure 4F, p=0.027). The gradual loss of NMJs is consistent with other cell types where Perlecan is dispensable for the initial formation of BMs and only required for their maintenance (Costell et al., 1999; Matsubayashi et al., 2017).

The largest source of Perlecan in Drosophila comes from the larval fat body, where it is secreted into the hemolymph and incorporated into the ECM surrounding most cell types (Pastor-Pareja and Xu, 2011). Although abundant in the fat body, trol mRNA is also present at lower levels in larval MNs and muscles (Jetti et al., 2023). To determine the cell type(s) responsible for Perlecan secretion that controls synaptic stability, UAS-trol RNAi.2 was driven with a panel of cell-type-specific Gal4 drivers. As indicated above, ubiquitous knockdown of trol mRNA with tubulin-Gal4 abolishes Perlecan expression (Figure 1D–G) and causes synaptic retraction (Figure 2—figure supplement 1). In contrast, knockdown of Perlecan with UAS-trol-RNAi.2 (Figure 1A) driven by multiple Gal4 lines expressed in specific cell populations, including pan-neuronal (elav^C155), neuronal and muscle (elav^C155 and mef2-Gal4), glial (repo-Gal4), fat body (ppl-Gal4, Lsp2-Gal4), hemocytes (Hml-Gal4), and fat body and hemocytes (c564-Gal4), failed to trigger synaptic retraction or reduce synaptic bouton number (Figure 5A and B). In addition, knockdown of Perlecan in the three cell types that form or surround NMJs (MNs, muscles and glia) failed to reduce Trol^GFP signal around larval nerves or on the muscle surface (Figure 5C–G). Although it is possible that these individual drivers are weaker than tubulin-Gal4, no phenotypes were observed when trol-RNAi was driven by elav^C155, elav^C155 and mef2, and repo at 29 °C to increase Gal4 activity (data not shown). Together, these data suggest Perlecan secretion from multiple cell types is required to stabilize NMJs.

The majority of proteins that control synaptic stability at Drosophila NMJs function cell autonomously within the neuron (Eaton et al., 2002; Graf et al., 2011; Koch et al., 2008; Pielage et al., 2008; Pielage et al., 2005). To determine whether MN secretion of Perlecan is sufficient to stabilize synapses, UAS-trol constructs encoding two different Perlecan isoforms (Figure 1A) were overexpressed with vGlut-Gal4 in the trol^null background. Overexpression of Perlecan specifically in MNs did not rescue the reduction in bouton number or synaptic retraction phenotypes in trol^null larvae (Figure 5—figure supplement 1A–C, p=0.54 for bouton number with trol^RG rescue), suggesting neuronally secreted Perlecan is insufficient for maintaining synaptic stability. Together, these data indicate Perlecan acts non-cell autonomously from multiple cell types to stabilize larval NMJs.

Studies of Perlecan’s role within the ECM of other Drosophila cell types and in several mammalian tissues indicate the protein helps withstand mechanical stress during tissue development (Arikawa-Hirasawa et al., 2002; Costell et al., 1999; Pastor-Pareja and Xu, 2011; Skeath et al., 2017; Töpfer et al., 2022). Although Perlecan is not enriched at synaptic boutons, trol^null NMJs could retract due to a failure to withstand mechanical stress from repeated contractions during larval crawling that would normally be buffered by the small amount of Perlecan normally on muscles. To test this model, a mutation in Myosin heavy chain (Mhc^S1) that causes a dominant hypercontractive muscle phenotype (Montana and Littleton, 2006; Montana and Littleton, 2004) was brought into the trol^null background to assay if synaptic retraction phenotypes were enhanced. Despite increased muscle contraction in trol^null; Mhc^S1/+ larvae, no enhancement of synaptic retraction or decreases in bouton number were observed (Figure 6A and B, p=0.5975). Given hypercontraction in Mhc^S1 mutants requires synaptic transmission (Montana and Littleton, 2004) and retracted trol^null NMJs lack evoked release (Figure 4C–E), we cannot exclude the possibility that Mhc^S1 mutants only enhance early stages of synaptic retraction prior to loss of presynaptic output. To examine if muscle hypercontraction increases Perlecan NMJ abundance as a protective mechanism to withstand elevated muscle contraction force, endogenous Trol^GFP was brought into the Mhc^S1 mutant background. No enhancement of Perlecan staining was observed around axons, at NMJs, or on muscles in Trol^GFP; Mhc^S1/+ larvae compared to Trol^GFP alone (Figure 6C–E). Together with the lack of Perlecan enrichment around boutons, the failure of muscle hypercontraction to increase instability of trol^null NMJs suggest the protein is unlikely to play a mechanical role within the extracellular space around boutons to directly stabilize NMJs.

Drosophila MN cell bodies reside within the VNC and their axons exit in segmental nerve bundles that also contain incoming sensory neuron axons. Nerve bundles are wrapped by several layers of glial cells and surrounded by the neural lamella (Edwards et al., 1993; Stork et al., 2008), a specialized ECM structure containing Perlecan. Although trol^null mutants display synaptic retraction phenotypes at larval NMJs, Perlecan is primarily expressed within the neural lamella surrounding larval nerves and not at NMJs (Figure 1D–I). As such, where Perlecan acts to regulate synaptic stability is unclear. If loss of Perlecan compromises the function of the neural lamella as a physical and protective barrier for encapsulated axons over time, NMJ retraction might occur for all axons within a nerve bundle in a temporally coordinated manner across larval hemisegments. To examine if synapses within each segmental nerve bundle showed evidence of coordinated loss, abdominal bodywall hemisegments of trol^null larvae expressing vGlut-Gal4; UAS-10xGFP were examined. NMJ area on muscle 6/7, 4, and 1 was quantified for each hemisegment and compared to controls. NMJs were often completely absent in one hemisegment, while fully intact in others. Indeed, Trol^null NMJs within each individual hemisegment displayed similar decreases in synaptic area (Figure 7A–C), indicating entire hemisegments undergo coordinated synaptic loss while others remain intact.

To determine if the neural lamella surrounding peripheral nerves was disrupted in the absence of Perlecan, the expression and localization of the neural lamella-localized Collagen IV homolog Vkg was assayed. In control segmental nerves, Vkg showed a similar localization to Perlecan and surrounded axon bundles exiting the VNC and at nerve branch points to muscles (Figure 8A and E). In contrast, Vkg expression in trol^null nerve bundles was dimmer with gaps in staining around axonal segments, along with abnormal aggregation at specific sites along the nerve (Figure 8A, B and E). To quantify neural lamella disruption, fluorescence intensity of Vkg and anti-Hrp (to label axons) was determined for segmental nerve bundles. Unlike controls, Vkg staining in axonal cross sections from trol^null larvae showed a thinner, or in some cases absent, neural lamella surrounding Hrp +axon bundles (Figure 8B). Anti-Hrp staining was also brighter in trol^null axon bundles, suggesting there may be greater antibody penetration in the absence of a functional neural lamella (Figure 8A and B). The mean fluorescence of Vkg and Hrp signal was calculated in axon bundles traveling over muscle 4. Vkg signal around axons was significantly reduced in trol^null larvae, with the Vkg/Hrp ratio significantly lower in trol^null axons than controls (Figure 8C and D, p=0.0014 and 0.0004). This phenotype was independent of whether NMJs had undergone retraction, as nerves for both intact and retracted NMJs displayed a reduced neural lamella. Compared to controls, Vkg staining was also non-evenly distributed along nerves, with multiple sites showing extracellular accumulation of Vkg beyond the traditional boundaries of the neural lamella (Figure 8E). In some protrusions, Hrp +axonal material protruded from its normal boundary to co-localize with Vkg. As such, Perlecan may play a role in capturing or retaining Vkg within the neural lamella surrounding larval nerves.

Prior studies indicated disruptions to the microtubule cytoskeleton often proceed NMJ loss in other synaptic retraction mutants (Eaton et al., 2002; Pielage et al., 2005). Given loss of Perlecan compromises the neural lamella, we examined axonal morphology in trol^null mutants by visualizing the axonal and synaptic microtubule network with immunostaining for Futsch, the Drosophila homolog of microtubule associated protein 1B (MAP1B) (Hummel et al., 2000). Microtubule bundles in MN axons innervating muscle 4 were examined in larvae expressing vGlut-Gal4; UAS-10xGFP or UAS-myrRFP. Microtubules in heterozygous trol^null/+ control larvae formed filamentous tracks within axons that extended into synaptic boutons (Figure 9A). In contrast, trol^null axons contained fragmented and non-continuous microtubule tracks or lacked Futsch staining altogether at branch points where the axon exited towards the muscle (Figure 9A). Quantification of Futsch staining intensity in trol^null axons revealed a significant reduction (Figure 9B, p<0.0001). Similar defects were observed within synaptic terminals, where NMJs undergoing retraction lacked Futsch staining or displayed fragmented microtubules (Figure 9A). These data indicate disruptions to the microtubule cytoskeleton within trol^null axons and NMJs accompany synaptic retraction.

To determine if axons showed more severe defects in morphology, axon bundles and microtubules were imaged at different developmental timepoints. In mature 3^rd instar controls, axon bundles have defined boundaries with smooth tracks of Futsch + microtubules (Figure 9C). In contrast, trol^null axon bundles displayed progressive defects throughout larval development. 2^nd instar trol^null larvae had some axonal swellings, small RFP + protrusions beyond the normal nerve boundary, and slightly twisted and disorganized Futsch + microtubule tracts (Figure 9C). By early 3^rd instar, nerve bundles were disorganized, with numerous protrusions and tangled microtubules (Figure 9C). At the mature 3^rd instar stage, some axon bundles were entirely severed, with large tangled nets of disorganized Futsch at the ends of severed nerves (Figure 9C–D), similar to previously described retraction balls that form after axonal injury in mammals (Cajal, 1928). When nerve bundles in a hemisegment were severed, NMJs in that hemisegment displayed severe retraction. Single motoneuron labeling with a MN1-Ib Gal4 driver (Aponte-Santiago et al., 2020) expressing UAS-CD4-TdT to label axons in trol^null larvae revealed discontinuous axon membrane labeling (Figure 9E). The time course of these deficits suggests that axonal damage and breakage occurs upstream of synapse loss, given that 2^nd instar trol^null larvae have normal synaptic area but display altered axonal structure (Figure 9C–D, Figure 3B).

Loss of distal axons following damage is a well-known trigger for neurodegeneration in both invertebrates and vertebrates (Gerdts et al., 2016; Perlson et al., 2010). Neurite loss, synaptic retraction and eventual neuronal death following axonal damage often proceeds through a defined molecular cascade known as Wallerian degeneration (Coleman and Höke, 2020; Conforti et al., 2014; Llobet Rosell and Neukomm, 2019; Luo and O’Leary, 2005; Sambashivan and Freeman, 2021; Wang et al., 2012). Inhibiting this protein cascade promotes distal axon survival following injury in multiple systems (DiAntonio, 2019; Fang et al., 2012; Figley et al., 2021; Gerdts et al., 2015; Gilley et al., 2017; Gilley et al., 2015; Llobet Rosell et al., 2022; Neukomm et al., 2017; Sasaki et al., 2016). To determine whether NMJ loss in trol^null mutants utilizes the Wallerian degeneration signaling cascade, an established RNAi inhibitor of an upstream component of the pathway, dSarm (Gerdts et al., 2015; Gerdts et al., 2013; Osterloh et al., 2012), was expressed in trol^null MNs. Inhibition of this pathway did not rescue distal axon maintenance, as trol^null larvae expressing dSarm RNAi still had reduced synaptic bouton number (P=0.5387 between trol and Wallerian degeneration inhibition conditions), with >80% of NMJs displaying postsynaptic footprints associated with synaptic retraction (Figure 9F–H). These data indicate Drosophila MNs undergo axonal degeneration and synaptic retraction in trol^null mutants independent of the Wallerian degeneration pathway. Together, we conclude that loss of Perlecan disrupts the neural lamella, leading to axonal damage that causes cytoskeletal disruption and synaptic retraction in a temporally coordinated manner across individual hemisegment nerve bundles during larval development (Figure 10).

Flies were maintained at 18–25°C and cultured on standard medium. 3rd instar larvae were used for all experiments unless otherwise noted. Larvae lacking Perlecan were collected at 1st or 2nd instar stage and separated from wild type counterparts and placed on petri dishes containing standard medium to facilitate survival to 3rd instar stage. Male and female larvae were used depending upon genetic background; see figure legends for genotypes. Canton-S (CS) was used as the wild type background in heterozygous controls except in the case of Df(1)ED411/+, where the wild type background was w¹¹¹⁸ (BDSC# 3605). trol strains used include trol^null (Voigt et al., 2002; provided by Brian Stramer and Yutaka Matsubayashi), GFP-tagged trol (referred to in text as Trol^GFP, KDSC #110–836), trol deficiency (Df(1)ED411, BDSC #8031, 36516), trol overexpression constructs (UAS-trol.RD BDSC #65274, UAS-trol.RG BDSC #65273), and trol RNAis (UAS-trol-RNAi.1 VDRC #22642, UAS-trol-RNAi.2 VDRC #24549). Gal4 drivers used were: tubulin-Gal4 (BDSC #5138), vGlut-Gal4 (provided by Aaron DiAntonio), elav^C155-Gal4 (BDSC #8765), mef2-Gal4 (BDSC #27390), repo-Gal4 (Lee and Jones, 2005), ppl-Gal4 (BDSC #58768), Lsp-Gal4 (BDSC #6357), Hml-Gal4 (BDSC #6395), c564-Gal4 (BDSC #6982), and MN1-Ib-Gal4 (BDSC #40701). Overexpression of constitutively active Shaggy was performed using UAS-sgg^S9A (BDSC #5255). Presynaptic labeling for intravital imaging was performed using CRISPR-generated nSyb^GFP (Guan et al., 2020). PSD labeling for intravital imaging was performed using GluRIIA-RFP inserted onto chromosome III under the control of its endogenous promoter (provided by Stephan Sigrist). The hypercontractive Mhc^S1 mutation (Montana and Littleton, 2004) was used to assess mechanical stress. UAS-dSarm-RNAi (VDRC #105369, provided by Aaron DiAntonio) was used to inhibit Wallerian degeneration in the trol^null background. UAS-10xGFP (Poukkula et al., 2011), UAS-CD4-TdTomato (BDSC #77139), and UAS-myrRFP (BDSC #7118) were used to visualize individual motoneuron axon anatomy. Viking^GFP (DGRC #110692, G00454 original FlyTrap identifier, provided by David Bilder) was used to visualize the neural lamella surrounding peripheral nerves.

3rd instar larvae were filleted in Ca²⁺-free HL3.1 solution (in mM: 70 NaCl, 5 KCl, 4 MgCl₂, 10 NaHCO₃, 5 trehalose, 115 sucrose, 5 HEPES, pH 7.18) and fixed in 4% paraformaldehyde for 15 min, washed in Ca²⁺ -free HL3.1 twice and 0.1-PBT (1 x PBS with 0.1% Triton X-100) once, then blocked in 5% normal goat serum (NGS) in 0.5-PBT (1 x PBS with 0.5% Triton X-100) for 30 min at room temperature or overnight at 4 °C. Samples were incubated overnight at 4 °C in blocking solution containing primary antibodies, and then washed three times with 0.1-PBT. Samples were incubated for 2 hr at room temperature in blocking solution containing fluorophore-conjugated secondary antibodies. Primary antibodies used in this study were mouse anti-Brp at 1:500 (Nc82 DSHB, Iowa City, IA), rabbit anti-GluRIIC at 1:2000 (Jorquera et al., 2012), rabbit anti-Cpx at 1:5000 (Huntwork and Littleton, 2007), mouse anti-Dlg at 1:500 (4F3 DSHB, Iowa City, IA), mouse anti-Futsch (22C10 DSHB, Iowa City, IA) and mouse anti-GFP at 1:1000 (#A-11120, Thermo Fisher Scientific, Waltham, MA). Secondary antibodies used in this study were goat anti-mouse Alexa Fluor 488-conjugated IgG at 1:500 (#A-32723, Thermo Fisher Scientific, Waltham, MA), goat anti-rabbit Alexa Fluor 568-conjugated IgG at 1:500 (#A-11011, Thermo Fisher Scientific, Waltham, MA), goat anti-mouse 555-conjugated IgG at 1:500 (#A-32727, Thermo Fisher Scientific, Waltham, MA) and goat anti-mouse 647-conjugated IgG at 1:500 (#A-32728, Thermo Fisher Scientific, Waltham, MA). For Hrp staining, samples were incubated in DyLight 649 or 488 conjugated Hrp at 1:500 (#123-605-021, #123-485-021; Jackson ImmunoResearch Laboratories, West Grove, PA, USA). For Phalloidin staining, samples were incubated in Texas Red-X Phalloidin at 1:500 (#T7471, Thermo Fisher Scientific, Waltham, MA). Samples were mounted in Vectashield Vibrance hard setting antifade mounting medium (#H-1700, Vector Laboratories, Burlingame, CA).

Images of fixed NMJs were acquired on a Zeiss Pascal confocal microscope (Carl Zeiss Microscopy, Zena, Germany) using a 63 X Zeiss pan-APOCHROMAT oil-immersion objective with a 1.3 NA. 3D image stacks were merged into a maximum intensity projection using Zen (Zeiss) software. Abdominal segments and muscle numbers imaged are listed in figure legends or in results text. Boutons were counted manually using the parameters specified in each experiment (e.g. Dlg +or Hrp/GluRIIC+). Severe retraction was calculated by taking the mean bouton number of control genotype(s) in an individual experiment and calculating the standard deviation. A bouton number of <mean – 2*SD and presence of postsynaptic footprints indicated severe retraction. Line profiles of fluorescence intensity across axons or boutons were generated in Volocity 3D Image Analysis software (PerkinElmer) using the ‘measure line profile’ algorithm in Volocity 3.2 or 5 software. Lines were drawn between 4.17–4.22 µM for axon line profiles and 2.4–2.6 µM boutons were chosen and lines extended 2 µM in either direction for bouton line profiles. Mean Vkg and Hrp intensity were calculated from maximum intensity projections using Volocity 3D Image Analysis software with the “Find Objects” algorithm and a Vkg threshold that identified neural lamella area in both genotypes. Images of live NMJs and live and fixed axons were acquired on a Zeiss Axio Imager 2 with a spinning-disk confocal head (CSU-X1; Yokagawa) and ImagEM X2 EM-CCD camera (Hamamatsu) using an Olympus LUMFL N 60 X water-immersion objective with a 1.10 NA. Low magnification images of fixed larvae were acquired on a Zeiss LSM 800 confocal microscope using a 10 x objective. A 3D image stack was acquired for each axon, NMJ, or larvae imaged. Futsch nets were counted manually. Mean fluorescence intensity of Futsch signal was calculated from the maximum intensity projection using Volocity 3D Image Analysis software. A 5 μm² ROI at the point of individual axon exit from the nerve bundle (muscle branch point) was generated and Volocity calculated the mean intensity within this ROI.

Larvae were anesthetized with SUPRANE (desflurane, USP) from Amerinet Choice. Larvae were covered with a thin layer of halocarbon oil and incubated with a small paper towel containing Suprane for 1–2 min in a fume hood. Anesthetized larvae were arranged ventral side up on a glass slide between spacers made by tape. Larvae were covered with a fresh thin film of halocarbon oil and then with a cover glass. After each imaging session, larvae were placed in numbered chambers with food at room temperature. The same data acquisition settings were used to visualize NMJs at each session. Larvae were imaged every 24 hr. Area of pre- and postsynaptic material was calculated from maximum intensity projections using Volocity 3D Image Analysis software with the ‘Find Objects’ algorithm and a pre- or postsynaptic threshold that identified presynaptic boutons or postsynaptic receptor fields in both genotypes. The same threshold was used to analyze images from each day of analysis.

Postsynaptic currents were recorded with a –80 mV holding potential. Experiments were performed in room temperature HL3.1 saline solution as previously described with final [Ca²⁺] adjusted to 0.3 mM (Jorquera et al., 2012). Recordings were performed at muscle 6 of segments A3 and A4 in 3^rd instar larvae. Motor axon bundles were cut and individual bundles were suctioned into a glass electrode for stimulation. Action potentials were stimulated at 0.33 Hz using a programmable stimulator (Master8, AMPI; Jerusalem, Israel). Data acquisition and analysis was performed using Axoscope 9.0 and Clampfit 9.0 software (Molecular Devices, Sunnyvale, CA, USA). Inward currents are labeled on a reverse axis. For post-hoc Hrp staining, samples were incubated for 5 min in DyLight 488 conjugated Hrp (2 µL of stock solution applied directly to filleted larvae and washed twice with HL3.1 before imaging) (#123-485-021; Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Live imaging was performed as described above.

NCBI BLAST and NCBI Gene search were used to identify Perlecan homologs and to select the longest isoforms of Perlecan, Agrin, and Carrier of Wingless (Cow) in the genomes of D. melanogaster, C. elegans, C. intestinalis, D. rerio, M. musculus, H. sapiens, and T. adhaerens. Clustal Omega multiple sequence alignment (with default parameters) was used to align all of the sourced sequences (Madeira et al., 2022). Jalview was used to generate a phylogenetic tree using a BLOSUM62 matrix and average distance clustering. Sequences used for alignment and phylogenetic tree included:

Graphing and statistical analysis were performed using GraphPad Prism (San Diego, CA, USA).

For comparisons between two groups, statistical significance was determined using a Student’s t-test. For comparisons between three of more groups, statistical significance was determined using a one-way ANOVA followed by multiple comparisons with p-values corrected for multiple hypothesis testing using either Tukey, Šidák, or Dunnett’s multiple comparisons tests (individual test chosen based on Prism recommendation). Figures depict the mean of each distribution and individual data points. N indicates the number of individual NMJs analyzed unless otherwise noted. Number of larvae per group, mean ± SEM, n, and the p values are indicated in figure legends. Asterisks in the figures denote p-values of: *, p≤0.05; **, p≤0.01; ***, p≤0.001; and ****, p≤0.0001. The Source Data and Statistical Analysis file contains spreadsheets for each figure and includes all primary source data and statistical comparisons.