To perform a screen for axonal growth and regeneration (Figures 1A,B), we selected genes which: (1) are associated with the Gene Ontology (GO) terms neural development and neurite morphogenesis, (2) had Gal4 inducible transgenes available at the time of the initiation of the study and (3) represent a diversity of molecular functions, including receptors, protein turnover, transcription factors, and chromatin modifiers. Three hundred and seven genes matching these criteria were first tested for their ability to induce developmental axonal over-growth in small Lateral Neurons ventral (sLNv), a small cluster of neurons with a highly stereotyped axonal morphology which can be readily quantified (Leyssen et al., 2005; Helfrich-Förster et al., 2007) and that has been previously used to investigate the molecular mechanisms underlying regeneration in the fly CNS (Ayaz et al., 2008; Figures 1A, 2A,B). The post-mitotic Pdf-Gal4 driver was used to express GFP together with each of the selected genes, and the length of axonal growth was quantified in comparison with controls. Expression of 13 genes (4.2%) promoted significantly increased axonal growth with no obvious adverse effects on neuronal survival or axonal trajectory (Figures 2A–E). In a second selection step (Figure 1B), these 13 genes were evaluated in an acute sLNv axonal injury model in Drosophila brains explanted and kept in culture (Ayaz et al., 2008; Koch, 2012). Given their superficial location, sLNv axons are easily accessible for injury, and were physically severed using an ultrasonic microchisel. Using the temperature dependency of the UAS/Gal4 system, high expression levels of candidate genes were induced in adult flies starting at 24 h before injury. Axonal regrowth was defined as the growth of novel sprouts from the site of injury within 4 days. We used three parameters to evaluate axonal regrowth following injury: capacity of regrowth (the percentage of brains that exhibited at least one axonal sprout grown de novo), total regrowth (defined as the sum of the lengths of novel sprouts), and the maximum projection distance (defined as the distance of the longest novel sprout from the site of injury to their terminus) (Figures 3A–D). Of the genes tested, seven (drl, Dscam1, faf, kay, pdm2, pum, and sens) showed enhanced regeneration in all three categories (Figures 3A–H). Kayak is the fly homolog of Fos, a key transcription factor downstream of JNK signaling, confirming that the screen can identify bona fide regeneration genes. Dichaete, D, is an example of a gene that promoted axonal outgrowth during development (Figures 2A,D), but failed to induce regeneration in most cases (Figures 3A,G) and often resulted in short sprouts with poor morphology, making it difficult to measure. Three genes (dimm, dac, and sqz) caused axonal phenotypes such as defasciculation, blebbing, or fragmentation, and were excluded from further analysis.

Of the identified seven genes, three in particular (Dscam1, faf and kay) appeared to consistently promote the most growth across all criteria. We therefore asked whether these genes might be acting together in a novel regeneration pathway linking the cell surface to the nucleus. We began by analyzing the de-ubiquitinating enzyme Faf since it promoted the highest levels of regeneration across all criteria (Figures 3A–C,H). First, we confirmed that faf is also able to induce axonal overgrowth in other CNS neuronal populations, such as the Dorsal Cluster Neurons (DCNs) (Figure 3—Supplementary Figure 1), confirming that Faf may be a general CNS axonal growth-promoting factor. Ubiquitin-dependent protein regulation is critical in regulating many neuronal events, including axonal growth (McCabe et al., 2004; Ambrozkiewicz and Kawabe, 2015). However, the signaling pathways operating downstream of these enzymes are still largely unknown. To test whether the axonal growth induced by Faf was dependent on its deubiquitinase activity, we mutated a critical cysteine 1,677 residue in the catalytic protease site to a serine (Chen and Fischer, 2000). In contrast to wild-type Faf, this mutated form of Faf was not able to significantly promote developmental axonal growth (Figures 4A,B,D). The mouse homolog of Faf, FAM/Usp9x, which can be active in Drosophila in other contexts (Wood et al., 1997; Chen et al., 2000), also induced robust sLNv axonal outgrowth (Figures 4C,D). Remarkably, even the yeast homolog of Faf, Ubp2, which only shares homology in the de-ubiquitination domain, induces sLNv axonal outgrowth very similar to Faf (Figure 4—Supplementary Figure 2). More importantly, both FAM and Faf, but not the enzymatic mutant Faf-Ser, induced significant axonal regeneration after injury (Figures 4E–H). These data suggest a conserved axonal growth and regeneration activity for Faf as a deubiquitinase enzyme.

Faf has been shown to induce neuromuscular junction growth in Drosophila (DiAntonio et al., 2001) in a pathway that requires Wallenda (Wnd), a conserved MAPKK upstream of JNK signaling (Collins et al., 2006). Therefore, we tested whether Faf required Wnd to induce axonal growth. RNA interference knock-down (RNAi KD) of wnd inhibited Faf-mediated axonal outgrowth (Figure 4—Supplementary Figures 3A,B,I), whereas overexpression of wnd, but not a kinase-dead form of it, strongly promoted axonal outgrowth that essentially phenocopied faf overexpression (Figure 4—Supplementary Figures 3A,C,D). Moreover, overexpression of wnd also promoted axonal regrowth after injury (Figure 4—Supplementary Figures 3G,H,J). Therefore, Wnd likely acts downstream of Faf, to modulate axonal growth and regeneration in response to faf overexpression. Similarly, RNAi KD of the Drosophila homolog of JNK, basket (bsk), completely inhibited faf-mediated axonal outgrowth (Figures 4I,K). Conversely, co-expression of kay, the JNK pathway effector we identified as strong promoter of outgrowth in development (Figures 2A,C) and after injury (Figures 3A–C,F) enhanced Faf-mediated axonal outgrowth (Figures 4J,K). Together, these data suggest that Wnd and JNK act downstream of Faf to induce axonal outgrowth and regeneration.

How might faf activate JNK signaling to induce axonal regeneration? During fly eye development faf mediates the internalization of the Notch ligand Delta (Overstreet et al., 2004), and Notch signaling has been proposed to enhance regeneration of developing neurons (Kato et al., 2011), though it has also been shown to act as a repressor of axonal regeneration (El Bejjani and Hammarlund, 2012). To test if faf interacted with Delta in the context of sLNv axonal growth, we tested both a RNAi KD as well as a dominant negative (DN) transgene, and found that loss of Delta function in the sLNvs did not reduce the axonal outgrowth activity of faf (data not shown), suggesting an alternative mechanism in the context of axonal growth. Therefore, we reasoned that Faf might interact with different axon growth-promoting effectors.

Mammalian Dscam1 (Qu et al., 2013) has been shown to be a regulator of JNK signaling. Interestingly, our screen identified Dscam1 as one of the genes that most strongly and consistently promoted sLNv axonal outgrowth and regeneration (Figures 2A, 3A–D). This prompted us to investigate the molecular mechanisms underlying the growth and regeneration activity of Dscam1. The Dscam1 gene generates a large number of isoforms by alternative splicing of a plethora of extracellular domains and two transmembrane domains called TM1 and TM2 (Schmucker et al., 2000). We find that different isoforms containing either the TM1 (Figures 2A, 5A), or the TM2 domain (not shown), and different extracellular domains (UAS-Dscam1 1.30.30.1-GFP and UAS-Dscam1 1.34.31.1-HA, see Methods) can induce axonal outgrowth, suggesting that induction of axonal growth may be a general property of Dscam1-mediated signaling independent of its isoform specificity. It has previously been reported that isoforms containing TM1 are dendrite specific (Shi et al., 2007). However, we find that upon overexpression these isoforms localize to both cell bodies and axonal terminals (Figure 5—Supplementary Figure 4). Conversely, Dscam1 knock-down with two different RNAi lines (Watson et al., 2005) resulted in stunted sLNv axonal growth (Figures 5B,F). Finally, TM1-containing Dscam1 isoforms, UAS-Dscam1 1.30.30.1-GFP and UAS-Dscam1 1.34.31.1-HA induce robust axonal regeneration after injury, with the latter being the strongest line (Figure 5C).

Both mammalian and fly Dscam1 are known to interact with p21 activating kinase (Pak) (Schmucker et al., 2000) itself an upstream JNK Kinase. We find that inhibition of JNK activity, by using a dominant negative form of Bsk, completely abrogates Dscam1 mediated axonal growth (Figures 5D,F). Conversely, the expression of Kay reverses the loss of axon growth caused by Dscam1 RNAi knock-down (Figures 5E,F). These data suggest that Dscam1 acts upstream of JNK signaling to induce axonal growth.

The fact that Faf and Dscam1 both promote axonal regeneration after injury and induce strikingly similar JNK-dependent axonal outgrowth phenotypes led us to hypothesize that Faf and Dscam1 interact in this context. Indeed, we find that Faf-induced axonal outgrowth required Dscam1, as Dscam1 knock-down almost completely abolished Faf-induced growth (Figures 6A,D). Consistent with this, co-overexpression of faf and Dscam1 in the sLNvs induces stronger axonal outgrowth than faf overexpression alone (Figures 6B,D). Importantly, Dscam1 knock-down also inhibits FAM/Usp9x mediated axonal outgrowth, indicating a conserved interaction (Figures 6C,D).

Faf antagonizes ubiquitination by cleaving the covalent bond between ubiquitin and a substrate protein (Huang et al., 1995), thereby leading to stabilization of proteins targeted for degradation. We asked if Faf might stabilize Dscam1 protein levels. Therefore, we expressed Dscam1 alone or together with faf, faf-Ser mutant or mouse FAM/Usp9x in Drosophila S2 cells. Both Faf and FAM/Usp9x, but not the Faf-Ser mutant lead to a ~30% increase in Dscam1 protein levels (Figures 6E,F), with no change in mRNA levels. However, we were unable to find evidence for Dscam1 ubiquitination in wild type or proteasome-inhibited S2 cells, nor a change in that status upon overexpression or knock-down of faf (data not shown). These data suggest that, at least in this context, Faf does not de-ubiquitinate Dscam1 directly.

Dscam1 has also been shown to be post-transcriptionally regulated via the Drosophila DLK1 homolog Wnd, in particular, through stabilization of its 3′-UTR. Specifically, using a Dscam1-3′-UTR>GFP reporter construct it has been shown that translation of the reporter protein is enhanced by Wnd (Kim et al., 2013). Since our data so far indicate that Faf acts upstream of both Dscam1 and Wnd (Figure 4—Supplementary Figure 2), we tested whether Dscam1 is also required for Wnd-induced axonal growth in sLNv neurons. We find that Dscam1 RNAi KD significantly decreased Wnd-induced growth (Figure 4—Supplementary Figures 3E,F). We then asked whether Faf enhances translation of Dscam1 through a mechanism that is mediated by the 3′UTR of Dscam1 in vivo. To this end, we expressed the Dscam1-3′-UTR>GFP reporter in the sLNv alone or together with faf. We found that GFP levels are significantly upregulated in sLNv upon Faf overexpression (Figures 6G–I).

We used the Gene Ontology tool (http://geneontology.org/) to select candidate genes annotated with the terms “Neurite Morphogenesis,” “Transcription Factors,” “Receptors,” “Chromatin Modifiers,” and “Ubiquitin Ligases.” We also included an additional set of genes previously implicated in axonal growth and/or involved in actin dynamics (indicated in Supplementary Table 1 by asterisks).

The final candidate gene sample only included genes for which appropriate gain- of-function fly lines were readily available at the stock centers at the time of the study (Supplementary Table 1).

Drosophila melanogaster stocks were kept on standard cornmeal media. For tissue-specific overexpression of the transgenes, we used the GAL4/UAS system (Brand and Perrimon, 1993). Lines with UAS insertion sites (i.e., UAS, EP, EPgy2, XP, and Mae-UAS) were received through the Bloomington or Szeged Stock Centres (or from specific laboratories when specified). Loss-of-function lines [Wnd RNAi GD8365, Dscam1 RNAi KK108835; Dscam1 RNAi (Watson et al., 2005); Bsk RNAi BL35594 and BL36643, and Bsk DN BL6409] were obtained from the Bloomington Stock Centre or from the Vienna Drosophila Research Centre (VDRC). UAS-Wnd kinase dead (KD) and UAS-Wnd E flies were a gift from C. Collins. The PDF-Gal4 line was obtained from P. Taghert. For Faf overexpression in flies, we used the EP3520 line (Szeged Stock Centrum), which was previously reported to induce Faf gain-of-function (DiAntonio et al., 2001). UAS-Faf and UAS-FAM lines were created in house by cloning Faf cDNA and FAM cDNA into a pUAST-attB vector, respectively (Bischof et al., 2007) and injected in an attP2 docking line (BL 8622). A UAS-Faf serine (Faf-Ser) mutant that harbors a cysteine to serine mutation at residue 1677 was also cloned using the same method. UAS-Dscam1 HA-FLAG and UAS-Dscam1 GFP flies, both containing the 1.30.30.1 isoform, were used for the injury experiments. The UAS-Dscam1 1.30.30.1 GFP flies have been described (Hughes et al., 2007) and the UAS-Dscam1 HA-FLAG flies were created by inserting a HA tag into the intracellular domain (after the 81st bp of exon 22) of isoform 1.34.31.1. The UAS-Dscam1 1.30.30.1 GFP flies have been used for the initial screen in development and after injury, and the UAS-Dscam1 1.30.30.1 HA has been used for additional confirmatory experiments in development and injury, as well as for epistasis experiments. Although both lines induced significant axonal growth in development and after injury, UAS-Dscam1 1.30.30.1 HA appeared to be the strongest of the two lines. For the genetic screen in development and after injury, pdf-Gal4, UAS-GFP; pdf-Gal4, UAS-2x eGFP/cyo flies were kept as a stock and used to drive expression of the various candidate genes, or crossed to wild-type Canton S (CS) flies. For the genetic epistasis experiments, pdf-Gal4, UAS-GFP; UAS-Dscam1 RNAi and pdf-Gal4, UAS-GFP; Faf (EP 3520) flies were maintained as a stock and crossed to overexpression lines to uncover genetic interactions.

To measure axonal outgrowth during development, flies were reared at 25°C and were dissected 2–10 days after eclosion. A minimum of five fly brain (10 sLNs projections) per genotype were stained with an anti-GFP antibody (to enhance the GFP signal), visualized under a fluorescent microscope equipped with a GFP filter and scored as “growth” (when sLNv axonal projections appeared considerably longer than in controls) or “no-growth” (when the length of sLNv projections was indistinguishable from controls or shorter). All genes that were scored growth promoting genes were confirmed as such in at least one independent experiment, and their growth inducing ability was analyzed by measuring the axonal sLNv dorsal axonal projections.

For the axonal regrowth analysis after injury, flies were reared at 18°C, in order to minimize overexpression effects during development, and shifted to 25°C the day before injury to allow optimal transgene expression.

Whole-brain explants on culture plate inserts were prepared and injured as described (Ayaz et al., 2008; Koch, 2012). In brief, Millicell low height culture plate inserts (Milipore) were coated with laminin and poly-lysine (BD Biosciences). Adult female flies were collected 2–10 days after eclosion and placed on ice. Fly brains were quickly and carefully dissected out in a sterile Petri dish containing ice cold Schneider's Drosophila Medium (GIBCO). Up to seven brains were placed on the membrane of one culture plate insert and culture medium (10,000 U/ml penicillin, 10 mg/ml streptomycin, 10% Foetal Bovine Serum, and 10 μg/ml insulin in Schneider's Drosophila Medium) was added. sLNv axonal injury was performed using an ultrasonic microchisel controlled by a powered device (Eppendorf). Culture dishes were kept in a plastic box in a humidified incubator at 25°C.

Freshly dissected brains of adult flies were fixed in 4% formaldehyde and processed for immunohistochemistry as described (Hassan et al., 2000). Cultured brains (4 days post-injury) were first fixed by replacing the culture medium in the Petri dish for 30 min. Then, 1 ml of fixative was carefully added on top of the filter for 1–2 h. Brains that detached from the membrane were excluded from further analysis. Immunostaining was performed as for freshly dissected samples. Primary antibodies were rabbit anti-GFP (A-6455, Molecular Probes), rat anti-HA (3F10, Roche); and anti-Pdh (gift from P. Taghert).

Image J software was used to measure the length of the dorsal axonal projections emanating from the sLNvs. The starting point was set as the point where axons turn medially and start to run parallel to the commissure. Axonal length was measured as a straight line (Computed Distance) from the starting point toward the midline (indicated by an asterisk) and as manual trace using Image J. The maximum computed distance was defined as the distance projected by the longest axonal sprout in a straight line and parallel to the commissure. The Average Length was the defined as the average length of the two longest axonal branches traced manually (freehand distance). Imaging was performed on an upright Zeiss Axioscope equipped with a CCD camera, or on a Zeiss 700 or Nikon AR1 confocal microscope. All measurements were performed using ImageJ.

To analyze the role of the candidate genes in axonal regrowth after injury we imaged cultured brains at two different time points after injury: approximately 5 h and 4 days. Comparison between these two timepoints allowed us to define the location at which the injury took place, in order to define de novo growth. Morphometric analysis of axonal regrowth was always performed 4 days after injury, following fixation and GFP staining of the brains in culture. Capacity of regrowth was defined as the ability of the injured sLNv projection to regrow at least one new axonal sprout. Without the support of the head cuticle, brains will flatten and therefore undergo slight morphological changes during the culture process. To be conservative and account for potential inaccuracies in defining the injury point, only regrown axons with a minimum length of 12 μm were defined as de novo growth and taken into account for analysis. To quantify axonal regrowth, newly grown axons were measured in a straight line and manually traced using ImageJ. In this case, the maximum computed distance was defined as the average of the distance of the two longest axonal sprouts in a straight line in any direction. Maximum growth was defined as the sum of the freehand lengths of all de novo grown axons. Images 5 h after injury were acquired on an upright Nikon microscope equipped with a Hamamatsu CCD camera ORCA-R2. Imaging 4 days after injury was performed on a Zeiss 700 or a Nikon A1R confocal microscope after GFP immunostaining. See figure legends for details of individual experiments, including statistical tests used and see Supplementary Table 2 for the number of samples tested.

Drosophila Schneider's (S2) cells were maintained in Sf-900 II SFM medium (Gibco). To achieve transgene overexpression in Schneider's (S2) cells we electroporated a UAS construct in combination with PMT-Gal4, according to previously developed methods (Klueg et al., 2002). For Faf and FAM overexpression, we created a UAS-Faf and a UAS-FAM construct as described in “Fly stocks and Genetics.” For Dscam1 overexpression, the UAS-Dscam1 1.30.30.1 GFP construct was used (Hughes et al., 2007).

Cells were electroporated using an Amaxa Nucleofector KitV (Lonza), according to the manufacturer's instructions. Cells were harvested 72–96 h after copper induction, briefly washed with PBS and pellets frozen until cells were lysed in a 1% NP40 buffer in Tris-HCL. Protein concentration was determined by a modified Lowry assay (Peterson, 1977). Western blotting was performed with a SDS-PAGE Electrophoresis System (Biorad). Briefly, protein samples were diluted in SDS containing sample buffer and 15 μg per sample was loaded onto a 3–8% Tris-Acetate mini gel (Novex, Life Technologies). Samples were blotted using tank transfer to a nitrocellulose membrane (GE Healthcare), blocked with milk and probed with primary antibodies against Dscam1 (1:1,000) (Watson et al., 2005) or against actin (1:5,000, ab3280, Abcam), which was used as a protein loading control. Anti-rabbit or anti-mouse horseradish peroxidase conjugated secondary antibodies (Amersham) were then added, and proteins were detected using enhanced chemiluminescence (ECL Plus, GE Healthcare) on a FUJI LAS imager system (Fuji). Values for Dscam1 were normalized to the values of the loading control (actin) and quantified using the blot analysis function for IMAGE J. Kruskall Wallis test was used to compare the different conditions. Data is shown as mean ± SEM and significance was set at p ≤ 0.05.

RNA was extracted with Trizol. 1 μg of total RNA was reverse transcribed using the Quantitect RT kit (Qiagen). qPCR was performed using the Taqman Real Time protocol (Applied Biosystems) and probes. Data is shown as mean ± SEM.

Adult brains were dissected and immediately prepared for imaging. Confocal stacks of all sLNv and lLNv cell bodies in each side of the brain were performed. The optimal confocal settings were first adjusted for wild type brains and kept unchanged to allow comparison between genotypes. A maximum projection was created for each brain side and each image was quantified for GFP intensity using the “Image Analysis” module of Zeiss Zen 2.0 software. All quantifications were done by an investigator blind to experimental conditions. Student t-test was used to compare both genotypes. Data is shown as mean ± SEM and significance was set at p ≤ 0.5.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.