Many neurological disorders have a number of common pathological features, suggesting that they may share a mechanistic origin. Abnormal dendritic morphology, which is thought to precede neuronal cell death, is one such feature (Kulkarni and Firestein, 2012; Kweon et al., 2017). Even amyotrophic lateral sclerosis (ALS), the pathological hallmark of which is motor neuron axonal degeneration (Fischer and Glass, 2007; Saberi et al., 2015; Taylor et al., 2016), has this characteristic. More than two decades ago, abnormal dendritic abnormality was first identified in ALS (Karpati et al., 1988; Nakano and Hirano, 1987; Takeda et al., 2014). Since then, some of the ALS-associated mutations in TARDBP and FUS have been linked to dendritic defects (Herzog et al., 2017; Tibshirani et al., 2017), but our understanding of the mechanism underlying these defects remains poor. Similarly, frontotemporal dementia (FTD), which is considered to be on the same disease spectrum as ALS because they have a shared set of genes and symptoms (Ling et al., 2013), also manifests dendritic abnormality. A study in 1999 showed that the complexity and number of dendritic branches were reduced in both pyramidal and non-pyramidal neurons from FTD postmortem brain samples (Ferrer, 1999). Surprisingly, very few studies since have examined the mechanism underlying dendritic pathology in FTD.

The most common genetic cause of both ALS and FTD is the intronic hexanucleotide (G4C2) repeat expansion mutation in C9orf72 (DeJesus-Hernandez et al., 2011; Renton et al., 2011). Repeat expansions greater than 30 are understood to be pathogenic, whereas fewer than 30 repeats are found in most healthy controls (Renton et al., 2011). These repeated sequences can be transcribed in both sense and antisense directions, and repeat RNA products of both often form nuclear foci where several RNA-binding proteins are sequestered (Lee et al., 2013; Xu et al., 2013). Some of these repeat RNAs exit the nucleus and undergo repeat-associated non-ATG (RAN) translation into five dipeptide repeat proteins (DPRs). Among the five DPRs, glycine-arginine (GR) and proline-arginine (PR) are most toxic; glycine-alanine (GA) is moderately toxic; and glycine-proline (GP) and proline-alanine (PA) are relatively benign (Freibaum and Taylor, 2017). Whether the arginine-rich DPRs can induce dendritic defects is not known.

Recently, Golgi outposts (GOPs) have been associated with dendritic defects in Drosophila models for Machado-Joseph Disease, spinocerebellar ataxia 1 and Huntington’s disease (Chung et al., 2017). In that study, nuclear polyQ proteins interfered with the CREB-binding protein (CBP)-CrebA pathway, thereby reducing the number of GOPs and decreasing plasma membrane (PM) supply in dendrites. Interestingly, a chimeric protein comprising green fluorescent protein (GFP) fused to a fragment of Golgi Complex Protein in COS-7 cells has been shown to form nuclear aggregates and recruit proteins such as CBP into the inclusion (Fu et al., 2005). This study suggests that when accumulated in the nucleus, aberrant non-polyQ proteins may also induce dysregulation of the CBP-CrebA pathway, leading to dendritic defects. Arginine-rich DPRs GR and PR are known to accumulate in the nuclei of many cell types (Kwon et al., 2014; Lee et al., 2016; Rudich et al., 2017; Wen et al., 2014), suggesting that they may induce dendritic defects via dysregulation of the CBP-CrebA pathway. Several previous studies have used Drosophila larval Class IV dendritic arborization (C4 da) neurons to study dendritic pathology (Chung et al., 2017; Kwon et al., 2018; Lee et al., 2011; Lin et al., 2015). The usefulness of the C4 da neuronal system stems from its sharing similar mechanisms for dendritic arborization with mammalian neurons and its amenability to in vivo single cell imaging (Jan and Jan, 2010; Kweon et al., 2017). In this study, we used C4 da neurons as a model in which to investigate abnormal dendritic morphology induced by DPR toxicity.

Drosophila stocks and gene switch-mediated expression

All flies were maintained at 25°C and 60% humidity. The following lines were obtained from Bloomington Drosophila Stock Center (USA): w¹¹¹⁸ (3605), UAS-PR36 (58694), UAS-GR36 (58692), UAS-GA36 (58693), UAS-PA36 (58695), UAS-RO288 (58681), UAS-CrebA (32572), UAS-GalT-RFP (65251), GMR-gal4 (1104), and elav-gal4 (8760). UAS-ManII-eGFP; UAS-CD4-tdTOM, UAS-CD4-tdGFP, ppk^1a-gal4, GS-ppk^1a-gal4, and ppk-UAS-CD4-tdTOM were provided by Yuh Nung Jan (University of California, San Francisco [UCSF], USA). To activate gene switch-mediated transcription of gal4, adult flies were fed food with RU486 (100 µM RU486 dissolved in 80% ethanol) for 15 days.

Generation of Drosophila transgenic construct

To control for gal4 dilution, transgenic flies carrying empty pACU2 vector (provided by Yuh Nung Jan; UCSF) were generated by BestGene (USA). The empty vector is inserted in the second chromosome at VK02 site. This transgene (UAS-empty) was used as a control in Drosophila retinal imaging study.

Drosophila retinal imaging

Leica SP5 was used to take Drosophila retinal images. Upon dissection, the retinas (left only) were imaged at 160× magnification.

Immunohistochemistry

Third instar larvae or adult flies were dissected and fixed with 3.7% formaldehyde for 20 min at room temperature. After washing with phosphate-buffered saline (PBS), the larval fillets were incubated in a blocking buffer for 45 min. Next, the fillets were incubated in the primary antibody overnight at 4°C. The following primary antibodies were used: α-polyPR (23979-1-AP, 1:20; Proteintech, USA) and α-RFP (#ab62341, 1:400; Abcam, UK). Fillets were then washed with washing buffer three times for 5 min each. Then, they were further incubated in the secondary antibody for 3 h at room temperature. The following secondary antibodies were used: goat α-mouse Alexa 555 (A28180, 1:400; Thermo Fisher Scientific, USA), goat α-rabbit Alexa 647 (A21244, 1:200; Thermo Fisher Scientific) and goat α-Horseradish Peroxidase (123-545-021, 1:200; Jackson ImmunoResearch, USA). Fillets were washed five times with washing buffer prior to mounting on a slide glass for imaging.

Confocal microscopy

Images of C4 da neurons in larvae or adult flies after immunohistochemistry and in larvae in vivo were obtained using Zeiss LSM700 or LSM780 confocal microscopy (Zeiss, Germany). All images of the C4 da neurons were acquired from abdominal A2-A6 segments.

Reverse transcriptase polymerase chain reaction (RT-PCR)

Total RNA was extracted from the heads of Drosophila with the following genotypes: w¹¹¹⁸/elav-gal4, UAS-PR36/elav-gal4, and UAS-GR36/elav-gal4. Concentration of total RNA of each sample was measured by NanoDrop spectrophotometer (Thermo Fisher Scientific). 18S rRNA was used as a reference for normalizing CrebA mRNA level.

Primer sequences for CrebA:

5’- CAACTACCTCAGCACCTATACGAC

3’- GTTACCTTCGGAATCATCGCTGG

Dendrite analysis

All images of dendrites were first converted into skeletal images using ImageJ. Then, those images were used to measure dendritic length and branch points.

Quantification of CD4-tdGFP intensity

Images obtained from confocal microscopy were processed using Zen black program, Adobe Photoshop CC and ImageJ. ImageJ→Plugin→NucMed→Lookup table→Blue Green Red was used for GFP intensity quantification. Then, the whole dendritic area of C4 da neurons was selected and its CD4-tdGFP intensity was analyzed using histogram plug-in in the ImageJ software. All images presented in the same panel were processed in the same manner.

Statistical analysis

Statistical analysis was done using GraphPad Prism 7 (GraphPad Software, USA). Depending on the data, we applied Student’s t-test followed by Tukey’s post-hoc analysis. In all figures, not significant (N.S.), *, **, ***, and **** represent P > 0.05, P < 0.05, P < 0.01, P <.001, and P

Arginine-rich DPRs induce loss of dendritic branch points in Drosophila C4 da neurons

Among the five types of DPRs produced by RAN translation from G4C2 repeat RNA, we tested four of them (PA, GA, PR, and GR) for their effect on dendritic morphology in Drosophila C4 da neurons. GP was not tested due to lack of an available fly line and because it is known to be benign in various model systems (Freibaum and Taylor, 2017). Additionally, we also tested G4C2 RNA only (RO288) transgene to examine the effect of RNA toxicity on dendritic morphology. Expression of arginine-rich DPRs PR36 and GR36 in C4 da neurons via a ppk^1a-gal4 driver led to a noticeable decrease in the number of dendritic branch points compared to the controls, whereas expression of PA36, GA36, and RO288 did not (Fig. 1A). However, when we quantified the number of dendritic branch points in DPR-expressing neurons, we found that even expression of PA36, GA36, and RO288 led to a significant decrease compared to wildtype controls (Ctrl) (Fig. 1B). Notably, C4 da neurons expressing PR36 and GR36 showed considerably fewer dendritic branch points compared to the Ctrl or those expressing PA36, GA36, and RO288 (Fig. 1B). Likewise, expression of PA36, GA36, and RO288 significantly decreased the total dendritic length compared to Ctrl, whereas PR36 and GR36 expression led to an even shorter total dendritic length (Fig. 1C). However, expression of DPRs induced no morphological abnormality in the pattern of axon scaffold in the larval ventral nerve cord (Supplementary Fig. S1).

Because ALS/FTD is an adult onset neurodegenerative disease, we decided to test whether arginine-rich DPR expression can induce dendritic defects in fully developed adult flies. To this end, we conditionally expressed PR36 in adult C4 da neurons using GS-ppk^1a-gal4—a RU486-inducible gene switch gal4. At day 1, we observed no dendritic changes in flies with the PR36 transgene compared to the controls (Fig. 1D) because the transcriptional activity of gal4 was not yet induced. After 15 days of induction, the adult C4 da neurons expressing PR36 showed a significant reduction in the number of dendritic branch points (Fig. 1E). These results suggest that PR36 expression can induce non-developmental dendritic phenotypes in the adult Drosophila C4 da neurons.

Arginine-rich DPRs disrupt PM supply and induce a loss of GOPs in Drosophila C4 da neurons

We also noticed a significant decrease in the fluorescence intensity of CD4-tdGFP in C4 da neurons expressing arginine-rich DPRs compared to the controls (Figs. 2A and 2B, Supplementary Fig. S1). Expression of PA36 and GA36 did not alter the fluorescence intensity of CD4-tdGFP (Supplementary Fig. S2). Interestingly, decrease in fluorescence intensity of transgenic PM marker proteins has been closely associated with reduced PM supply (Chung et al., 2017; Murthy et al., 2005), suggesting that arginine-rich DPRs may reduce PM supply.

The supply of proteins and lipids to the PM in dendrites may in part be mediated by GOPs (Horton et al., 2005; Ye et al., 2007). GOPs are also essential for proper dendritic morphology (Horton et al., 2005; Ori-McKenney et al., 2012; Ye et al., 2007), and alteration in their number or location has been associated with dendritic defects in animal models for Parkinson’s and several polyglutamine (polyQ) diseases (Chung et al., 2017; Lin et al., 2015). Based on these facts, we hypothesized that GOPs might be involved in dendritic defects caused by arginine-rich DPR toxicity. To test this, we expressed PR36 in C4 da neurons along with the medial-Golgi marker, ManII-eGFP, and compared the number of GOPs to controls (Fig. 2C). We found that the number of GOPs marked by ManII-eGFP decreased significantly in PR36-expressing neurons compared to the controls (Fig. 2D). Similar results were obtained when we expressed the trans-Golgi marker GalT-RFP instead of ManII-eGFP (Supplementary Fig. S3). GR36 expression also led to a significant reduction in the number of GOPs (Supplementary Figs. S4A and S4B), similar to the level caused by PR36 expression. These data suggest that arginine-rich DPR toxicity leads to a reduction in the number of GOPs.

Nuclear PR toxicity induces down-regulation of CrebA mRNA expression

Previous studies have shown that GOPs can be regulated by secretory pathway-related genes (Chung et al., 2017; Iyer et al., 2013; Ye et al., 2007). CrebA is known to be a master regulator of the secretory pathway (Fox et al., 2010), and its expression is down-regulated by polyQ toxicity (Chung et al., 2017). Interestingly, a few studies have suggested that expression of aberrant proteins in the nucleus can lead to generalized toxicity, one sign of which is the sequestration of CBP (Fu et al., 2005; McCampbell et al., 2000; Nucifora et al., 2001; Steffan et al., 2000); CrebA is a downstream molecule of CBP (Chung et al., 2017). Because PR36 localizes predominantly in the nucleus of C4 da neurons (Figs. 3A and 3B), we hypothesized that nuclear toxicity induced by PR36 may alter the level of CrebA mRNA. To test this, we expressed PR36 in Drosophila brains via the elav-gal4 driver and measured the amount of CrebA mRNA in fly heads with respect to controls. RT-PCR showed that PR36 expression led to a significant decrease in the level of CrebA mRNA compared to controls (Figs. 3C and 3D). However, unlike PR36, GR36 expression did not induce changes in CrebA mRNA level (Supplementary Figs. S4C and S4D).

CrebA overexpression restores PM supply and the number of GOPs in C4 da neurons showing PR toxicity

If down-regulation of CrebA is involved in PR36-induced dendritic defects, its overexpression might restore some of those defects. We first confirmed that CrebA overexpression alone increased the number of GOPs (Figs. 4A and 4B) in C4 da neuronal dendrites, consistent with the results from a previous study (Chung et al., 2017). When CrebA was overexpressed in PR36-expressing C4 da neurons, the number of GOPs was significantly increased compared to when PR36 was expressed alone (Figs. 4A and 4B). The quantitative analysis suggests that the effect of CrebA and PR36 on GOPs seems to be non-additive. Restoration of the GOP number suggests that the PM supply may also have been restored. To test this hypothesis, we examined the fluorescence intensity of CD4-tdGFP as a proxy for PM supply. CrebA overexpression alone increased the fluorescence intensity of CD4-tdGFP in C4 da neuronal dendrites (Figs. 4C and 4D), consistent with the previous study (Chung et al., 2017). When CrebA was overexpressed in PR36-expressing C4 da neurons, the fluorescence intensity of CD4-tdGFP was recovered to a normal level (Figs. 4C and 4D), implying that PM supply had been restored. However, the quantitative analysis indicates that the effect of CrebA and PR36 on the PM supply seems to be additive, calling into question the relevance of CrebA to the PR36-induced PM supply defects. CrebA overexpression also did not change the number of dendritic branch points and the length of total dendrites in C4 da neurons expressing PR36 (Supplementary Figs. S5). In addition, CrebA overexpression in adult fly retina could not mitigate PR36-induced degeneration (Supplementary Figs. S6). Notably, a similar pattern of restoration was observed in a polyQ disease fly model, where overexpression of CrebA restored GOP number and the fluorescence intensity of the PM marker, but not dendritic morphology in C4 da neurons (Chung et al., 2017). Taken together, these data suggest that multiple mechanisms underlie dendritic defects induced by PR toxicity and that CrebA overexpression alone cannot resolve all dendritic problems as was previously shown for polyQ toxicity (Chung et al., 2017).

In this study, we show that C9orf72-associated arginine-rich DPRs reduce the number of GOPs and dendritic branch points as well as PM supply in C4 da neurons. Mechanistically, PR36 expression decreases CrebA mRNA level, the overexpression of which restores PM supply and the number of GOPs. To our knowledge, this is the first time that dendritic defects and loss of GOPs have been linked to arginine-rich DPR toxicity.

Golgi pathology has been observed in a number of neurodegenerative diseases and is considered to be one of the early features of neurodegeneration (Caracci et al., 2019; Gonatas et al., 2006). However, whether GOPs, which are unique to neuronal dendrites, are associated with neurodegenerative diseases has been understudied. Two recent studies showed GOP abnormalities in Drosophila models for Parkinson’s and polyglutamine diseases (Chung et al., 2017; Lin et al., 2015), but whether they trigger neurodegeneration remains uncertain. In this study, we showed that although CrebA overexpression restored the number of GOPs and PM supply, it could not rescue PR-induced neurodegeneration of Drosophila retina (Supplementary Figs. S6). These data suggest that neurodegeneration may be caused by something other than defects in GOPs and PM supply. Previously, a link between PM supply (a function of Golgi) and neurodegenerative diseases such as Huntington’s and Charcot–Marie–Tooth was suggested (Pfenninger, 2009). However, to our knowledge, no direct evidence indicating that defects in PM supply directly cause neurodegeneration has been reported. Therefore, further investigation of the role of GOPs and PM supply in neurodegeneration and dendritic pathology of other neurodegenerative diseases is warranted.

We show that PR36 expression decreases the level of CrebA mRNA, but the mechanism by which this happens remains unknown. Previous studies showed that nucleus-accumulated polyQ (McCampbell et al., 2000; Nucifora et al., 2001; Steffan et al., 2000) and non-polyQ proteins (Fu et al., 2005) sequester CBP, which may reduce the transcription of its downstream factors such as CrebA (Chung et al., 2017). PR also localizes mostly in the nucleus (Figs. 3A and 3B), suggesting that PR may interfere with the function of CBP and subsequently reduce CrebA mRNA transcription. Another molecule through which CrebA might be down-regulated by PR toxicity is Ataxin-2 (Atx2). A knockdown of Atx2 was recently shown to reduce the level of CrebA mRNA in Drosophila lateral neurons at dawn (Xu et al., 2019). Interestingly, Atx2 physically interacts with PR (Lee et al., 2016; Yin et al., 2017), suggesting that PR may interfere with Atx2 function through direct interaction and lead to a subsequent reduction in CrebA mRNA level.

Previous studies have shown that PR and GR disrupt a number of the same cellular processes including membraneless organelle dynamics (Lee et al., 2016), nucleocytoplasmic transport (Hayes et al., 2020), and RNA decay (Sun et al., 2020). This suggests that PR36 and GR36 might share a similar cellular mechanism that underlies dendritic defects in C4 da neurons. Indeed, we found that both PR36 and GR36 expression led to a significant reduction in the number of GOPs (Figs. 2A and 2B, Supplementary Figs. S4A and S4B). However, only PR36 induced changes in CrebA mRNA level (Figs. 3C and 3D). These results suggest that although the cellular processes disrupted by PR and GR repeats may converge, the molecular mechanism underlying the disruptions likely diverge.

Defective vesicular trafficking in ALS/FTD has been observed in a number of disease models including those with mutations in VAPB, CHMP2B, TMEM106B, and C9orf72 (Burk and Pasterkamp, 2019). For instance, a mutation of Vap33 (a Drosophila ortholog of VAPB) in Drosophila has been shown to induce accumulation of phosphatidylinositol-4-phosphate in Golgi and lead to increased production of endosomes and impaired lysosomal degradation (Mao et al., 2019). In mouse cortical neurons, a mutation of CHMP2B has been shown to disrupt neuronal endolysosomal trafficking in such a manner that can be rescued by knockdown of TMEM106B (Clayton et al., 2018). Moreover, C9orf72 has been shown to interact with key components of the COPII pathway, such as Rab1 and Sec24, and may modulate the pathway (Burk and Pasterkamp, 2019; Farg et al., 2014; Goodier et al., 2020). Interestingly, synergistic toxicity between C9orf72 haploinsufficiency and DPR toxicity has recently been suggested (Boivin et al., 2020; Shi et al., 2018; Zhu et al., 2020). Thus, we propose that haploinsufficiency and arginine-rich DPRs may synergistically target GOPs and lead to defective vesicular trafficking in dendrites.

Spinocerebellar ataxia 36 (SCA36) is caused by a hexanucleotide expansion repeat in the intron of NOP56 gene (Kobayashi et al., 2011), the nucleotide sequence of which is TG3C2—remarkably similar to G4C2. A recent study showed that the expression of TG3C2 repeats in mice induced a loss of dendritic arbor in Purkinje cells (Todd et al., 2020), and the repeats were found to undergo RAN translation to produce PR and GP repeats in SCA36 patients (McEachin et al., 2020; Todd et al., 2020). Among the non-coding repeat expansion disorders, only SCA36 and C9-ALS/FTD manifest motor neuron degeneration (Ikeda et al., 2012; Kobayashi et al., 2011; Swinnen et al., 2020), suggesting that they may share a pathogenic mechanism. Although we have not tested GP repeats, GP expression is known to be mostly non-toxic in a number of model systems (Freibaum and Taylor, 2017). Thus, we propose that PR repeats may be the shared pathogenic driver underlying dendritic defects in these diseases. We envisage that further studies comparing the mechanistic similarity between SCA36 and C9-ALS/FTD using PR repeats will aid in developing new therapeutic targets for these diseases.

This work was funded by Basic Science Research Program through the National Research Foundation of Korea, funded by the Ministry of Science and ICT (2018R1A2B6001607 and 2019R1A4A1024278) (to S.B.L.); the Development of Platform Technology for Innovative Medical Measurements Program from the Korea Research Institute of Standards and Science Grant (KRISS-2019-GP2019-0018) (to S.B.L.); and KBRI basic research program through Korea Brain Research Institute funded by Ministry of Science and ICT (20-BR-04-02) (to S.B.L.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

J.H.P., C.G.C., and J.H.K. wrote the manuscript. J.H.P., C.G.C., and J.H.K. performed experiments. J.H.P., C.G.C., and J.H.K. analyzed the data. J.S., B.H.L., Y.S.L., and S.B.L. provided expertise and feedback. S.B.L. supervised the research.

The authors have no potential conflicts of interest to disclose.