Hereditary spastic paraplegias (HSPs) are a group of neurodegenerative disorders characterized by degeneration of the longest motor neurons in the corticospinal tract, leading to muscle weakness and spasticity of the lower limbs. Pathogenic variants in genes encoding proteins that shape the endoplasmic-reticulum (ER) network are a leading cause of HSP, however, the mechanisms by which loss of ER-shaping proteins underpin degeneration of selective neurons in HSP remain poorly understood. To begin to address this, we have generated a novel in vivo model of HSP in Drosophila melanogaster by targeted knockdown of the ER-shaping protein Arl6IP1. Variants in the human homolog of this gene have recently been linked to HSP subtype SPG61. Arl6IP1 RNAi flies display progressive locomotor deficits without a marked reduction in lifespan, recapitulating key features of HSP in human patients. Loss of Arl6IP1 leads to fragmentation of the smooth ER and disrupted mitochondrial network organization within the distal ends of long motor neurons. Furthermore, genetically increasing mitochondrial fission, by overexpression of dynamin-related protein 1 (Drp1), restores mitochondrial network organization and rescues locomotor deficits in two independent Drosophila models of HSP. Taken together, these results propose a role for ER-shaping proteins in mitochondrial network organization in vivo and suggest that impaired mitochondrial organization may be a common mechanism underpinning some forms of HSP.
Hereditary spastic paraplegias (HSPs) are a group of neurodegenerative disorders characterized by degeneration of the long axons within the corticospinal tract leading to progressive lower limb spasticity and weakness. HSPs are classified as complicated or pure, depending on the presence or absence of additional, mainly neurological, symptoms which include distal peripheral neuropathy, cognitive defects or seizures (1). With a prevalence of 3–10 per 100 000 in most populations, HSPs are uncommon but not rare (2), yet there are currently no treatments to cure or even slow progression of axonopathy. HSPs are highly genetically heterogeneous and to date more than 55 spastic paraplegia genes (SPGs) have been identified (3). These genes encode proteins with roles in a range of different cellular processes including axonal transport, lipid metabolism and mitochondrial function. The largest class of SPGs, accounting for over half of all cases of autosomal dominant HSP, encode proteins that function in shaping the smooth endoplasmic reticulum (ER) network (2,4). This suggests that there are shared pathogenic mechanisms underpinning these axonopathies.
The smooth ER network is a highly dynamic membrane system which consists of tubules extending throughout all cells, from the perinuclear rough ER to the plasma membrane (5). This network is generated and maintained by integral membrane proteins that insert into the ER membrane via conserved hydrophobic hairpin-loop domains (6). Hairpin-loop domain proteins including: atlastin-1 (SPG3A), spastin (SPG4), protrudin (SPG33), reticulon (RTN2; SPG12) and receptor expression-enhancing proteins (REEP1 and 2; SPG31 and 72) shape the ER by regulating ER tubule curvature, fission and fusion (6–9). Pathogenic mutations in any one of these genes cause HSP, pointing to a crucial role for the smooth ER in axonal maintenance. Despite this, the mechanisms by which loss of these ER-shaping proteins give rise to neurodegeneration in HSP are not known.
Recently, another ER-shaping protein has been identified, ADP-ribosylation factor-like 6 interacting protein 1 (ARL6IP1). Although it shares no primary sequence homology with other ER-shaping proteins, ARL6IP1 contains hairpin-loop domains, characteristic of ER-shaping proteins, by which it localizes to smooth ER tubules (10). In vitro studies show that ARL6IP1, similar to reticulon family members, modulates ER morphology by increasing ER curvature to form highly curved smooth ER tubules (10,11). Whole-exome sequencing in a consanguineous family with HSP revealed homozygous loss of function variants in ARL6IP1 (SPG61; 12). This same group generated arl6ip1 knockdown zebrafish by morpholino injection which resulted in abnormal branching of spinal motor neuron axons and a curly tail phenotype, pointing to a role for ARL6IP1 in neuronal development.
In this study, we assess the function of HSP-associated ER-shaping proteins in motor neurons in vivo using the model organism Drosophila melanogaster. Drosophila is an ideal organism in which to model neurodegenerative disease because of the range of tools available to manipulate tissue-specific gene expression and the ability to recapitulate of molecular and behavioural disease phenotypes (13). Using two independent RNAi lines to knockdown the Drosophila homolog of ARL6IP1, Arl6IP1, we show that loss of this ER-shaping protein results in a progressive locomotor deficit and disruption of the smooth ER and mitochondria at the distal ends of motor neurons. Furthermore, genetically increasing mitochondrial fission, by overexpression (OE) of dynamin-related protein 1 (Drp1), rescues mitochondrial disruption and restores locomotor function in Arl6IP1 and Rtnl1 models of HSP. These findings provide evidence that loss of ER-shaping proteins contribute to HSP-associated neurodegeneration by disruption of the axonal mitochondrial network.
Loss of Arl6IP1 causes progressive locomotor defects in Drosophila
In order to model HSP-associated neurodegeneration, we have generated a novel model of HSP in Drosophila by knocking down the ER-shaping protein ARL6IP1. Drosophila have a single homolog to ARL6IP1, CG10326 (hereafter referred to as Arl6IP1) which shares 30% amino acid sequence identify with the human sequence, as revealed by a BLASTP search (Fig. 1A). To generate knock down flies, two independent Arl6IP1 RNAi lines were obtained for this study, one each from the GD and KK libraries maintained at the Vienna Drosophila RNAi centre (14; www.VDRC.at). Expression of either of these lines under the control of the ubiquitous driver da-GAL4 (15) results in robust knockdown of Arl6IP1 expression (Fig. 1B).
Ubiquitous loss of Arl6IP1 results in apparently healthy, viable flies (data not shown) however, to investigate the role of Arl6IP1 in neurons, we knocked down Arl6IP1 using the pan neuronal driver nSyb-GAL4 (16) and tested for evidence of degenerative phenotypes. Selective knockdown of Arl6IP1 in neurons significantly impairs larval locomotion (Fig. 1C), without greatly altering the life span of these flies (Fig. 1D; 17). These Drosophila therefore recapitulate key aspects of HSP and provide good models in which to investigate the molecular and cellular events underpinning this neurodegenerative disorder.
Loss of Arl6IP1 disrupts smooth ER staining at the distal ends of motor neurons
Given the observed neurodegenerative phenotypes caused by a loss of Arl6IP1, we investigated the effect of loss of Arl6IP1 on neuronal organisation. All neuromuscular junction (NMJ) analyses within this study were conducted by imaging muscles 6 and 7 of third-instar larvae (L3). Staining with the post-synaptic marker Discs-large (Dlg) revealed no change in the number or composition of synaptic boutons at the NMJ in Arl6IP1 RNAi larvae (10). To observe the effects of Arl6IP1 knockdown on smooth ER organisation in vivo we used a Rtnl1 exon trap insertion tagged with yellow fluorescent protein (YFP), which marks endogenous expression of the smooth ER protein Rtnl1. Rtnl1::YFP is highly expressed in neurons, particularly in axons and in neuron terminals where it displays a continuous staining pattern consistent with smooth tubular ER (17). In contrast, Rtnl1::YFP staining was strikingly disrupted within Arl6IP1 RNAi NMJs, displaying punctate staining within terminal synaptic boutons at the ends of both shorter motor neurons, synapsing at segment A3 (anterior), and longer motor neurons, synapsing at segment A7 (posterior) (Fig. 2). This suggests that Arl6IP1 is required for the maintenance of the axonal smooth ER network in vivo.
Loss of Arl6IP1 disrupts mitochondrial network organization at the distal ends of long motor neurons
The ER has extensive interactions with mitochondria which are essential for several functions shared by these two organelles including mitochondrial morphology, calcium homeostasis and lipid synthesis (18). We therefore looked at mitochondrial network organisation within motor neurons in response to loss of Arl6IP1. GFP-tagged mitochondria were expressed under the control of the motor neuron driver OK6-GAL4 (19) and mitochondria within control and Arl6IP1 RNAi motor neurons were imaged. To investigate whether loss of Arl6IP1 has a length-dependent effect on motor neurons, we analysed mitochondrial organisation in both shorter (anterior) motor neuron axons and longer (posterior) motor neuron axons (Fig. 2A). There was no evidence of gross mitochondrial transport defects (i.e. clumped or accumulated mitochondria) and within anterior portions of long motor neurons axonal mitochondria were normal (Fig. 3A). In contrast, mitochondrial morphology was notably altered within posterior portions of Arl6IP1 RNAi long motor neurons. Quantification of mitochondrial circularity revealed that loss of Arl6IP1 causes elongation of axonal mitochondria compared with controls (Fig. 3B). Furthermore, loss of Arl6IP1 leads to a significant reduction in mitochondrial staining within terminal synaptic boutons of these long motor neurons whereas mitochondria within terminal synaptic boutons of shorter NMJs are unchanged from controls (Fig. 3C and D). Similarly, loss of the ER-shaping protein Rtnl1 causes increased elongation in axonal mitochondrial and decreased mitochondrial staining in the distal ends of posterior, but not anterior, motor neurons (
Increased mitochondrial fission by OE of Drp1 increases mitochondrial localisation in distal NMJ boutons of HSP models
The ER regulates mitochondrial morphology by constricting the mitochondrial membrane and marking sites for mitochondrial fission (20). Given our findings that disruption of the tubular ER network gives rise to mitochondrial defects, we hypothesized that this could be due in part to alteration of mitochondrial fission dynamics at the distal ends of long motor neurons. To test this hypothesis, we investigated the effect of overexpressing the mitochondrial fission protein Drp1 in our loss of ER-shaping protein Drosophila (
Drp1 OE partially rescues mitochondrial phenotypes caused by a loss of Arl6IP1. Drp1 OE in Arl6IP1 RNAi larvae restored mitochondrial morphology such that mitochondria were less elongated (Fig. 4A). Furthermore, Drp1 OE in Arl6IP1 RNAi larvae restored mitochondrial localisation within terminal synaptic boutons of posterior NMJs (Fig. 4B). Similarly, Drp1 OE in Rtnl1 RNAi motor neurons rescued mitochondrial network organisation within posterior motor neurons (
Increased mitochondrial fission by overexpressing Drp1 disrupts smooth ER staining within Arl6IP1 RNAi motor neurons
While the smooth ER is known to regulate mitochondrial organisation, to our knowledge, no studies have investigated whether altered mitochondrial dynamics affect the axonal smooth ER network. Unexpectedly, we found that Rtnl1::YFP staining in Drp1 OE larvae was disrupted, displaying fragmented staining within synaptic boutons at the ends of both shorter and longer motor neurons (Fig. 5). Similarly, Drp1 OE in Arl6IP1 RNAi larvae results in punctate or absent smooth axonal ER staining in NMJ terminal boutons (Fig. 5). This suggests that the observed repair of the mitochondrial network by Drp1 OE in loss of Arl6IP1 larvae is not due to restoration of the smooth ER network.
Drp1 OE partially rescues neurodegenerative phenotypes in Drosophila models of HSP
Finally, we investigated whether disruption of the mitochondrial network caused by loss of ER-shaping proteins contributes to the neurodegenerative phenotype in our models of HSP. Neuronal Drp1 OE using nSyb-GAL4 does not significantly alter Drosophila climbing behaviour compared with controls (Fig. 6). Nonetheless, Drp1 OE in Arl6IP1 RNAi or Rtnl1 RNAi Drosophila results in a striking rescue of the progressive climbing deficit caused by loss of these ER-shaping proteins, in some cases back to control levels (Fig. 6A and B;Fig. 6C and D;
Mitochondrial morphology is controlled by the opposing processes of fission and fusion. Given the rescue observed by increasing mitochondrial fission, we sought to determine whether decreasing mitochondrial fusion would similarly effect the mitochondrial network disruption and behavioural phenotypes in Arl6IP1 RNAi Drosophila. Mitochondrial fusion was decreased by RNAi-mediated knockdown of Marf, the Drosophila orthologue of mammalian Mitofusins 1 and 2 (21), which encodes a protein that mediates fusion of the mitochondrial outer membrane. Marf RNAi causes increased axonal mitochondrial circularity and staining in distal NMJ boutons (
Loss of function mutations in ER-shaping proteins are the most common cause of the neurodegenerative disorder HSP. However, the mechanisms by which loss of ER-shaping proteins give rise to axonopathy in HSP are not known. Here, we report on a novel in vivo model of HSP by targeted knockdown of the ER-shaping protein Arl6IP1, the Drosophila homolog of SPG61. Arl6IP1 knockdown flies display progressive locomotor deficits and disrupted smooth ER organisation at the distal ends of motor neurons. We suggest that these flies offer a good model in which to study the role of smooth ER in motor neuron function and, by comparison with other models of HSP, to identify common cellular mechanisms underpinning HSP-associated axonopathy.
Our results indicate that ER-shaping proteins play an important role in mitochondrial network organisation within long motor neurons. The mitochondrial network is highly dynamic, constantly changing shape, size, number and location throughout neurons and it is the opposing processes of mitochondrial fission and fusion that control the changes in mitochondrial morphology. The ER has a highly conserved role in regulating mitochondrial fission; ER contacts with mitochondria mark the sites for mitochondrial constriction and division (20). Fission factors, including Drp1, localize to these ER-mitochondrial contact sites where they oligomerize and complete fission of the outer and inner mitochondrial membranes by GTP hydrolysis (22,23). In this study we report a striking disruption of the smooth ER network at the distal ends of motor neurons and a similar phenotype has recently been reported in loss of Atlastin models of HSP (24). We hypothesize that disruption of the smooth ER network by loss of ER-shaping proteins could impair mitochondrial fission by reducing either the frequency of ER-mitochondrial contacts or the capacity for the ER to constrict the mitochondrial membrane (20). This fission is regulated by short ER tubules which bud out from the rough ER. Such short-range tubules are unlikely to function effectively in long motor neurons in which axons project great distances from the cell body and comprise up to 99% of the total cell volume (2). In our models, mitochondrial defects are specifically observed at the distal ends of axons, pointing to a length-dependent sensitivity of motor neurons to loss of ER-shaping proteins.
Results from this study suggest that mitochondrial disruption, occurring as a result of loss of ER-shaping proteins, underpins much of the neurodegenerative phenotype in our models of HSP given that the locomotor deficits in these models are rescued by Drp1 OE, without restoration of the smooth ER network (25) and supresses the degenerative phenotypes caused by expression of the Parkinson’s disease genes PTEN-induced kinase 1 (PINK1) and parkin (21,26–28). Several lines of evidence support an important role for mitochondria in HSP-associated axonopathy; mutations in genes encoding paraplegin, chaperonin 60, DDHD domain containing 1 (DDHD1) and C12orf65, with direct links to mitochondrial function or localisation, cause HSP (SPG7, SPG13, SPG28 and SPG55, respectively) (29–32). More recently, the ER-shaping protein REEP1 (SPG31) has been shown to localize to, and modulate, ER-mitochondrial contacts (33). Mitochondrial structural and functional defects are a common feature in many in vitro and in vivo models of HSP in which the genetic defect is not directly associated with mitochondria (17,34–36). Furthermore, mitochondrial defects have been detected in fibroblasts from patients with mutations in REEP1, cytochrome P450 family 2 subfamily U member 1 (CYP2U1) and DDHD1 (SPG31, SPG28 and SPG56, respectively) (37,38). Together, this raises the possibility that strategies which restore mitochondrial load and/or function within affected motor neurons could offer novel therapeutic approaches to ameliorate some of the symptoms of HSP. These findings do not preclude the possibility that other functions associated with the ER, such as ER stress, may also be disrupted in HSP. In fact, drugs that modulate ER stress have recently been shown to partially rescue degenerative phenotypes in spastin models of HSP (39).
An important question which remains to be addressed is, how does loss of mitochondria lead to selective axonopathy of the longest neurons? Mitochondria play crucial roles in presynaptic function including: (i) provision of ATP, required to maintain synaptic membrane potential and reload synaptic vesicles, and (ii) regulation of calcium homeostasis by direct calcium uptake, preventing excitotoxicity. Maintenance of a healthy mitochondrial load is therefore vital for synaptic function and neuronal maintenance. One mechanism by which mitochondrial density may be reduced in the distal ends of our HSP models is by impaired mitochondrial quality control by mitophagy. Mitophagy maintains a healthy pool of mitochondria in neurons by targeted degradation of depolarized mitochondria. Inhibition of mitochondrial fission, by knockdown of Drp1, attenuates mitophagy in vitro (40). To our knowledge, mitophagy has not been investigated in models of HSP; however, PINK1 and parkin have been shown to maintain mitochondrial health in distal neuronal axons through promoting mitophagy (41). Alternatively, elongated and enlarged mitochondria, observed in our models of HSP and patient-derived fibroblasts, could impair mitochondrial trafficking (37,38). Although the relationship between mitochondrial morphology and transport is not yet clear shorter mitochondria are more frequently in motion than longer ones (42). Furthermore, loss of Drp1 or the HSP-causing protein Kif5A have been shown to disrupt mitochondrial trafficking within neurons (43,44). Post-mitotic neurons, and particularly neurons with very long distances between cell body and synapse, may be more susceptible to defects in these processes. Notably, in some patients with pathogenic variants in the ER-shaping proteins ARL6IP1 and Atlastin 1 (SPG61 and SPG3A respectively), both the long sensory afferent neurons and the long upper motor neurons degenerate (1).These sensory neurons share similar extended nerve fibres, suggesting that these structures are particularly susceptible to damage caused by disruption of the smooth ER network.
The study of pathogenic mechanisms underpinning neurodegeneration in particular neurons in HSP is extremely challenging to investigate in human patients or to model effectively in vitro. By studying in vivo models which recapitulate characteristics of disease, we have identified an important role for the smooth ER in regulating mitochondrial network organisation in long motor neurons. Moreover, our findings from two independent models of HSP, generated by loss of the ER-shaping proteins Arl6IP1 or Rtnl1, suggest that disruption of the mitochondrial network by loss of ER-shaping proteins may be a common mechanism contributing to HSP-associated neurodegeneration.
Materials and Methods
Flies were raised on standard yeast, dextrose, cornmeal and agar food at 25 °C with a 12:12 light–dark cycle and transferred to fresh vials every 2–3 days. For knockdown and OE experiments, the fly lines used were: UAS-Al6IP1-RNAi GD and KK (line numbers 5894 and 10790, respectively); UAS-Rtnl1-RNAi (II) and (III) (line numbers 33919 and 7866, respectively); UAS-Marf-RNAi (line number 40478) and the w1118 control stock 60 000 (all obtained for the Vienna Drosophila RNAi Centre, www.vdrc.at) (14) and UAS-Drp1 (25). UAS lines were crossed to either da-GAL4 (15), nSyb-GAL4 (16) or OK6-GAL4 (19) as indicated in the text. Other fly stocks used were Rtnl1::YFP (45) and mito::GFP (46).
Assessment of motor neuron function in adult flies was determined using the negative geotaxis assay. Male flies were grouped into vials of 10 individuals per genotype, kept at 25 °C and tested under the same conditions every 6 days. A minimum of three vials per experiment and three independent experiments were used throughout. During testing, the mean number of flies climbing to the top of a vertical glass vial (10 cm length, 2.5 cm diameter) within 15 s in three trials was quantified. For survival assays, male flies from climbing assays were transferred into fresh media every 2–3 days and mortality was scored daily. Deficits in larval locomotor behaviour were assessed by a larval locomotor assay. Third-instar larvae were transferred to a 15 cm petri dish filled with 2% agarose and allowed to habituate for 30 s. The number of 1 cm2 grid lines crossed in a 60-s period was counted. A minimum of 10 larvae per genotype were assayed per experiment.
Total RNA was purified from ten third instar stage larvae via treatment with TRIzol reagent according to the manufactures protocol (Invitrogen, Paisley, UK). cDNA was produced using the SuperScript III First Strand Kit (Invitrogen, Paisley, UK) and 2 µg of DNase treated RNA. Amplification of Rp49 mRNA was used to control for cDNA concentration in the PCR reaction. Primers used were: Rp49-L: CCGACCACGTTACAAGAACTCTC; Rp49-R: CGCTTCAAGGGACAGTATCTGA; Arl6IP1-L: GGTGCTGTGGTACCTGGACT; Arl6IP1-R: CCATAGCCAAAAGACCCAAA; Rtnl1-L: GCTGTGGCACATATCAATGG; Rtnl1-R: TCGATTGCTTGTTGTTCTCG. PCR conditions were 94 °C for 45 s, 60 °C for 45 s and 72 °C for 1 min, repeated for 23 cycles for Rp49, 26 cycles for Arl6IP1 and 25 cycles for Rtnl1. Following PCR, products were visualized via a gel containing ethidium bromide.
Histology and immunomicroscopy
For larval neuromuscular dissections, third-instar larvae were dissected in chilled Ca2+-free HL3 solution and fixed in 4% formaldehyde solution for 30 min (47). Larvae were arranged dorsal side up and minutien pins were placed between the mouth hooks and in the tail between the posterior spiracles; during this step larvae were stretched out lengthwise to aid cutting. Using spring scissors a horizontal incision was made anterior to the posterior pin on the tail, thus allowing a vertical cut to be made along the dorsal midline up to the mouthpiece. Last, small transverse cuts were made near the mouthpiece allowing the larvae to be stretched out and kept in place via minutien pins. Tissues and organs were then removed to reveal the central nervous system and ventral nerve cord, axon bundles and muscles. After fixation, larval preps were permeabilized in 0.1% Triton X-100 in PBS (PBT), blocked in PBT serum (PBT + 4% fetal calf serum) and incubated with the antibody against Dlg (4F3; Developmental Studies Hybridoma Bank in Iowa, USA) overnight at 4 °C. After four washes in PBS, larval preps were incubated with anti-mouse 594 nm for 2 h at room temperature. Fixed and stained preparations were mounted in Vectashield (Vector Laboratories, Peterborough, UK) and viewed using an Olympus FluoView FV100 confocal microscope. Images were acquired using a 60x/1.5 NA objective and using the FV10-ASX version.04.01 software. Motor axons were imaged passing through segment A2 (‘anterior’) and segment A7 (‘posterior’) (Fig. 2A). NMJs were imaged at muscle 6/7 of segment A3 (‘anterior’) and segment A7 (‘posterior’) (Fig. 2A).
To quantify smooth ER organization within terminal synaptic boutons at NMJs, Rtnl1::YFP staining was classified as continuous, i.e. having a linear staining pattern, or disrupted, i.e. staining was fragmented or absent. Staining was categorized on images in a blinded fashion, without knowing which image corresponded to which sample, and then subsequently ‘un-blinded’.
Mitochondrial staining intensity per bouton area was quantified in ImageJ. Mean grey intensity values of mito::GFP within terminal synaptic boutons were normalized to bouton area.
For mitochondrial circularity, images were first thresholded and converted to binary images in ImageJ. The watershed function was used on binarized images to separate adjoined mitochondria. Mitochondrial circularity was quantified using the Shape Descriptors option in the ImageJ/Analyze menu which uses the equation 4 π × [Area]/[Perimeter]2 to generate a value between 0 and 1, where 1 is a perfect circle and as the value approaches 0 it becomes increasingly elongated.
Number of synaptic boutons per NMJ were counted manually, with type 1b and 1s boutons differentiated by intensity of Dlg staining.
All data were exported to Prism 5 (GraphPad Software, Inc.) for statistical analysis. For ER and mitochondrial analysis, statistical significance was determined using one-way ANOVA and Dunnett’s post-tests. Statistical significance for adult climbing assays were determined using two-way ANOVA and Bonferroni post-tests and life span assays were analysed using the log-rank Mantel-Cox test. Unless otherwise specified, significance from control is indicated on all graphs as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
We thank M. Feany, J. Roote, the Developmental Studies Hybridoma Bank and the Bloomington and Vienna Drosophila Stock Centres for antibodies and stocks. We thank J. Simpson for assistance with confocal microscopic imaging and D. Costello for comments on the article.
Conflict of Interest statement. None declared.
This work was supported by a University College Dublin Career Development Award (SF992). P.C.F. is supported by a University College Dublin School of Biomolecular and Biomedical Science PhD Scholarship. N.C.O S. is an Assistant Professor at University College Dublin.