Mutations in Caenorhabditis elegans neuroligin-like glit-1, the apoptosis pathway and the calcium chaperone crt-1 increase dopaminergic neurodegeneration after 6-OHDA treatment

The loss of dopaminergic neurons is a hallmark of Parkinson’s disease, the aetiology of which is associated with increased levels of oxidative stress. We used C. elegans to screen for genes that protect dopaminergic neurons against oxidative stress and isolated glit-1 (gliotactin (Drosophila neuroligin-like) homologue). Loss of the C. elegans neuroligin-like glit-1 causes increased dopaminergic neurodegeneration after treatment with 6-hydroxydopamine (6-OHDA), an oxidative-stress inducing drug that is specifically taken up into dopaminergic neurons. Furthermore, glit-1 mutants exhibit increased sensitivity to oxidative stress induced by H2O2 and paraquat. We provide evidence that GLIT-1 acts in the same genetic pathway as the previously identified tetraspanin TSP-17. After exposure to 6-OHDA and paraquat, glit-1 and tsp-17 mutants show almost identical, non-additive hypersensitivity phenotypes and exhibit highly increased induction of oxidative stress reporters. TSP-17 and GLIT-1 are both expressed in dopaminergic neurons. In addition, the neuroligin-like GLIT-1 is expressed in pharynx, intestine and several unidentified cells in the head. GLIT-1 is homologous, but not orthologous to neuroligins, transmembrane proteins required for the function of synapses. The Drosophila GLIT-1 homologue Gliotactin in contrast is required for epithelial junction formation. We report that GLIT-1 likely acts in multiple tissues to protect against 6-OHDA, and that the epithelial barrier of C. elegans glit-1 mutants does not appear to be compromised. We further describe that hyperactivation of the SKN-1 oxidative stress response pathway alleviates 6-OHDA-induced neurodegeneration. In addition, we find that mutations in the canonical apoptosis pathway and the calcium chaperone crt-1 cause increased 6-OHDA-induced dopaminergic neuron loss. In summary, we report that the neuroligin-like GLIT-1, the canonical apoptosis pathway and the calreticulin CRT-1 are required to prevent 6-OHDA-induced dopaminergic neurodegeneration.


Introduction
Dopamine acts as a neurotransmitter to control attention, cognition, motivation and movement [1]. The progressive loss of dopaminergic neurons in the substantia nigra of the midbrain is a hallmark of Parkinson's disease, the second most common neurodegenerative disorder (for review [2,3]). The etiology of the disease is largely unknown and there is no causative treatment for Parkinson's disease to date (for review [4]). Less than 10% of Parkinson's patients report a positive family history while the majority of instances are sporadic [3]. Most Parkinson's disease cases appear to be evoked by the interplay of one or multiple genetic defects and environmental factors [4]. A major risk factor for both genetic and sporadic forms of Parkinson's disease is oxidative stress, the imbalance between the production of reactive oxygen species and the antioxidant response (for review [5]). Reactive oxygen species can damage cellular components and ultimately cause cell loss [5]. However, it is not clear which pathways mediate dopaminergic cell death after oxidative stress exposure (for review [6]).
The gradual loss of dopaminergic neurons can be modelled by exposure to the oxidative stress-inducing drug 6-hydroxydopamine (6-OHDA) (for review [7]). 6-OHDA is an hydroxylated dopamine analogue which is specifically taken up into dopaminergic neurons by the dopamine transporter and which inhibits complex I of the mitochondrial electron transport chain [7]. The resulting formation of reactive oxygen species is assumed to trigger 6-OHDAinduced neurodegeneration [7]. 6-OHDA was found in human brain samples [8] and in higher concentrations in the urine of Parkinson's patients [9].
6-OHDA exposure of C. elegans leads to the selective loss of dopaminergic neurons [10]. The components of dopaminergic synaptic transmission are highly conserved in C. elegans (for review [11]). The nematodes' dopaminergic neurons are not required for viability and overt movement but they mediate subtle motor behaviours [11]. C. elegans hermaphrodites possess eight dopaminergic neurons, all of which have mechanosensory roles [12]: four CEP (cephalic sensilla) and two ADE (anterior deirids) neurons in the head, and two PDE (posterior deirids) neurons in the midbody. These neurons can be visualised by expression of a fluorescent protein from the promoter of the dopamine transporter [11]. Upon 6-OHDA exposure, dopaminergic neurons display blebbed processes, dark and rounded cell bodies and chromatin condensation, and eventually degenerate [10]. The pathways mediating 6-OHDAinduced dopaminergic cell death are unknown. Interference with the core apoptosis machinery does not prevent 6-OHDA-induced dopaminergic neurodegeneration; however, affected cells do not show typical morphological features for necrotic cell death either [10].
We used the C. elegans 6-OHDA model of dopaminergic neurodegeneration to screen for genes that protect from oxidative stress. Previously, this approach led to the characterisation of the tetraspanin TSP-17 which acts in dopaminergic neurons and which alleviates 6-OHDAinduced neurodegeneration [13]. In this study, we report that the previously uncharacterised neuroligin-like gene glit-1 prevents 6-OHDA-induced dopaminergic cell death. glit-1 likely acts in the same genetic pathway as tsp-17: glit-1 and tsp-17 mutants are short-lived, exhibit similarly increased expression of oxidative stress reporters and show basically identical sensitivity upon exposure to oxidative stress-inducing compounds. We further found that the apoptosis pathway and the calcium chaperone CRT-1 protect dopaminergic neurons from 6-OHDA-induced cell death.

Results
gt1981 mutants exhibit highly increased dopaminergic neurodegeneration after 6-OHDA exposure Animals carrying mutations in neuroprotective genes are expected to exhibit increased cell death when exposed to a low dose of a neurodegenerative drug. To find genes that prevent dopaminergic neuron death, a C. elegans strain with GFP-labelled dopaminergic neurons ( Fig  1A) was mutagenised with ethyl methanesulfonate. The F2 progeny of these mutagenised animals was exposed to the neurodegenerative drug 6-hydroxydopamine (6-OHDA) at a dose that does not induce cell death in wild-type animals. We treated C. elegans synchronised at the first larval stage (L1), the first developmental phase after egg hatching at which the ADE and CEP dopaminergic head neurons are already fully differentiated. 72 hours after 6-OHDA treatment, the now adult population was screened for mutant animals exhibiting excessive loss of dopaminergic head neurons. We isolated the gt1981 mutant line, which exhibited highly increased 6-OHDA-induced dopaminergic neurodegeneration when L1, L2 or L3 stage larvae were exposed to 6-OHDA ( Fig 1B). Later stages were not severely affected by gt1981 mutation ( Fig  1B). We further found that the general healthiness of gt1981 mutants is not compromised following 6-OHDA exposure as determined by measuring animal development: gt1981 mutants exhibited slightly slower growth than wild-type animals, but this delay was not exacerbated following 6-OHDA treatment (S1A Fig). Thus, the gt1981 mutation affects dopaminergic neuroprotection, but not the overall development of C. elegans larvae after 6-OHDA exposure.
Inhibition of neuronal 6-OHDA uptake by mutation of the C. elegans dopamine transporter DAT-1 prevents 6-OHDA-induced dopaminergic neurodegeneration in wild-type animals [10]. We thus tested if DAT-1 is also required for 6-OHDA hypersensitivity conferred by glit-1 mutation and found that this is the case as glit-1;dat-1 double mutants were resistant to 6-OHDA (S1D Fig). glit-1 is expressed in pharynx, intestine and several cells in the head, including dopaminergic neurons To determine where GLIT-1 is expressed, we generated GFP reporter constructs. A Pglit-1::gfp transcriptional construct revealed that GLIT-1 is expressed in the pharynx, the intestine, and in several cells in the head including dopaminergic neurons (S6F- S6H Fig). N-terminal GFPtagging of GLIT-1 protein confirmed this expression pattern (Fig 2A and 2B and S6A-S6E Fig). GFP::GLIT-1 expression could be clearly detected in dopaminergic neurons in L4 stage larvae and adult animals, but such expression could not be ascertained at earlier developmental stages, possibly due to low levels of GLIT-1 expression (Fig 2C-2F, S6C-S6E Fig, S1 Movie). We note that N-terminal GFP-tagging appears to render GLIT-1 non-functional, likely by compromising the N-terminal membrane targeting sequence: GFP::GLIT-1 localised to the cytoplasm and not the plasma membrane. Accordingly, the construct did not rescue the 6-OHDA hypersensitivity of glit-1(gt1981) mutants (S7A Fig). Furthermore, a targeted chromosomal GFP insertion at the GLIT-1 C-terminus did not reveal any expression. In summary, transcriptional and translational reporter constructs show that GLIT-1 is expressed in the pharynx, the intestine and in several cells in the head, including dopaminergic neurons.
CEP but intact ADE neurons as 'only ADE' (dark grey bar) and those with complete loss of dopaminergic head neurons as 'no ADE + CEP' (black bar). Error bars = SEM of 2 biological replicates, each with 25-40 animals per stage and strain. Total number of animals per condition n = 50-80 (****p<0.0001, n.s. p>0.05; G-Test comparing BY200 wild-type and mutant data of the same lifecycle stages). (C) Dopaminergic head neurons 24, 48 and 72 hours after treatment with 10 mM 6-OHDA and 72 hours after treatment with ascorbic acid only ('72h Ctr') for BY200 wild-type or glit-1 mutant animals. Error bars = S.E.M. of 2 biological replicates for glit-1(ok237) and 3 biological replicates for all the other strains, each with 60-115 animals per strain and concentration. Total number of animals for the '72h Ctr' experiment n = 30-100 and for all other conditions n = 130-340 (****p<0.0001; G-Test comparing BY200 wild-type and mutant data of the same time point). (D) GLIT-1 protein structure prediction based on homology modelling with acetylcholinesterase (PDB ID: 2W6C). The gt1981 point mutation leads to a proline to glycine conversion (P113G) and is indicated in red. The amino acids replacing the acetylcholinesterase catalytic triad are indicated in blue. (E) glit-1 gene structure with positions of the glit-1(gt1981) point mutation, the glit-1(gk384527) splice site mutation and the glit-1(ok237) deletion. For the point mutations the nucleotide and amino acid changes are indicated in brackets. The ok237 deletion spans 5'UTR and first exon of glit-1 and 5'UTR and first exons of dnj-14 (DNaJ domain (prokaryotic heat shock protein)) and is indicated with a red bar. Also, glit-1 is located in an operon (grey bar) with the ribosomal protein rpl- 25 Oxidative stress protection by the C. elegans neuroligin-like glit-1 We next aimed to determine in which tissues GLIT-1 acts to protect against toxicity conferred by 6-OHDA. An integrated, single-copy construct driving glit-1 expression under its own promoter almost completely prevented dopaminergic neuron loss ( Fig 3A). In contrast, single-copy expression of glit-1 in dopaminergic neurons, the pharynx and the intestine, respectively, did not alleviate 6-OHDA-induced neurodegeneration of glit-1 mutants (Fig 3B  and 3C). We conclude that GLIT-1 likely acts in several tissues to protect against dopaminergic neurodegeneration after 6-OHDA exposure.
To test if GLIT-1 overexpression protects wild-type animals against dopaminergic neuron loss after 6-OHDA exposure, we analysed two independent transgenic lines carrying Pglit-1:: glit-1 extrachromosomal arrays. The extent of neurodegeneration in these transgenic animals was comparable to that of wild-type animals, suggesting that overexpression of glit-1 does not confer protection against 6-OHDA (S7B Fig). GLIT-1 acts in the same genetic pathway as TSP-17 Previously, we have shown that mutation of the C. elegans tetraspanin tsp-17 renders dopaminergic neurons hypersensitive to 6-OHDA [13]. To determine if GLIT-1 and TSP-17 act in the same genetic pathway, we generated the glit-1(gt1981);tsp-17(tm4995)double mutant and found  that it exhibits a similar extent of 6-OHDA-induced dopaminergic neurodegeneration as the glit-1 and tsp-17 single mutants ( Fig 3D). Hence, GLIT-1 and TSP-17 are likely to function in the same genetic pathway to protect from dopaminergic neuron loss after exposure to 6-OHDA.
We used qRT-PCR to determine if expression of tsp-17 and glit-1 is induced after 6-OHDA treatment. However, we did not find altered tsp-17 and glit-1 transcript levels in wild-type and tsp-17 and glit-1 mutant larvae (S7C- S7F Fig), suggesting that tsp-17 and glit-1 provide constitutive protection against oxidative stress. This hypothesis is consistent with the fact that we did not determine a protective effect of glit-1 overexpression (S7B Fig). GLIT-1 does not appear to affect intestinal permeability and acts independently of NLG-1 and NRX-1 The glit-1 Drosophila homologue Gliotactin is localised at tricellular junctions and necessary for the development of septate junctions, an invertebrate analogue of tight junctions in epithelial cells [17,18]. To address if glit-1 mutation might compromise the epithelial barrier function of the C. elegans intestine, we tested if a blue dye ('Smurf' assay) or fluorescein leak into the body cavity after ingestion [19,20]. We did not find an increased intestinal permeability in glit-1 mutants as compared to wild-type animals (S8 Fig and S9 Fig). We note that the molecular weight of fluorescein (Mw = 376.2 g/mol), the smaller one of the two dyes, is still slightly higher than the molecular weight of 6-OHDA (Mw = 205.64 g/mol). Thus, we cannot fully exclude the possibility of increased organismal 6-OHDA uptake in glit-1 mutants, albeit this appears unlikely.

glit-1 mutation does not appear to affect dopamine signalling
As glit-1 and tsp-17 are both expressed in dopaminergic neurons, we aimed to determine if the hypersensitivity of glit-1 mutants to 6-OHDA is altered by mutations affecting dopamine metabolism (simplified cartoon in S11F Fig, for review [23]). However, we did not find that Ctr'). Error bars = SEM of 2-3 biological replicates, each with 60-115 animals per strain. Total number of animals per condition n = 180-350 (*p<0.05, n.s. p>0.05; G-Test). The '72 h Ctr' was conducted twice with 30 animals per strain, resulting in a total n = 60. https://doi.org/10.1371/journal.pgen.1007106.g003 Oxidative stress protection by the C. elegans neuroligin-like glit-1 dopamine receptor mutations specifically alter 6-OHDA-induced dopaminergic neurodegeneration in the glit-1 mutant background (S11A, S11D and S11E Fig, S1A Text). We further found that some mutants defective in dopamine synthesis can lead to increased 6-OHDAinduced neurodegeneration specifically in glit-1 mutants (S11B-S11E Fig). At present, we cannot explain the partly contradictory effects we found (S1B Text).
To assess in more detail if the signalling of dopaminergic neurons is altered by glit-1 and tsp-17 mutation, we tested several behaviours that are mediated by dopamine. We firstly found that dopamine signalling appears to be functional in the mutants [13] (S12A Fig) by analysing the basal slowing response, the dopamine-mediated slowing of C. elegans when encountering a lawn with bacterial food [24]. Secondly, we found that glit-1 mutants (S12B Fig), unlike tsp-17 mutants [13], do not exhibit premature swimming-induced paralysis, a behaviour indicative of increased extracellular dopamine levels [25]. Thirdly, we tested dopamine-induced paralysis: Overstimulation of C. elegans with exogenous dopamine eventually leads to animal paralysis [26]. We found that glit-1 and tsp-17 mutants stopped moving at much lower dopamine concentrations than wild-type animals ( Fig 4A). This increased sensitivity to exogenous dopamine depends on dopamine signalling as it was rescued by mutation of the dopamine receptors dop-2 and dop-3 ( Fig 4B). We conclude that glit-1 mutants, in contrast to tsp-17 mutants, do not exhibit signs of altered intrinsic dopamine signalling. However, both tsp-17 and glit-1 mutants exhibit increased dopamine-induced paralysis.
glit-1 and tsp-17 mutants are sensitive to oxidative stress at an organismal level As oxidative stress elicited by 6-OHDA led to increased dopaminergic neurodegeneration in glit-1 and tsp-17 mutant animals, we aimed to determine if these mutants are also hypersensitive to oxidative stress at an organismal level. To this end we exposed animals for 1 hour to paraquat, an oxidative stress-inducing herbicide linked to the development of Parkinson's disease [3]. Paraquat, unlike 6-OHDA, is not known to be specifically taken up into dopaminergic neurons. We found that after incubation with different concentrations of paraquat, glit-1 and tsp-17 mutants exhibit a very similar developmental delay compared to wildtype animals ( Fig 5A). To follow up on this result, we also analysed the second glit-1 allele glit-1(gk384527), as well as the tsp-17(tm4995);glit-1(gt1981) double mutant after exposure to very low concentrations of paraquat. We found that tsp-17 and glit-1 single and double mutants exhibit an almost identical developmental delay phenotype ( Fig 5B). Furthermore, glit-1 and tsp-17 mutants also developed much slower than wild-type animals after incubation with hydrogen peroxide (H 2 O 2 ), another oxidative stress-inducing compound ( Fig 5C). In summary, glit-1 and tsp-17 mutants are hypersensitive to oxidative stress at the organismal level and likely act in the same pathway.
Increased oxidative stress has been implicated with premature aging (for review [27]), thus we asked if the increased oxidative stress sensitivity exhibited by glit-1 and tsp-17 mutants is linked to a shorter lifespan. We found that glit-1 and tsp-17 mutants die earlier than wild-type animals ( Fig 5D, S13 Fig), indicating that the overall fitness of the mutants is decreased.

Oxidative stress reporters are hyper-induced in glit-1 and tsp-17 mutants
We aimed to determine if glit-1 and tsp-17 mutants exhibit an altered oxidative stress response. Using qRT-PCR, we measured transcript levels of the oxidative stress reporter gst-4 (glutathione S-transferase), gcs-1 (gamma glutamylcysteine synthetase) and gst-1 (glutathione S-transferase) in wild-type and mutant L1 stage larvae after 1 hour of incubation with 10 mM 6-OHDA in comparison to mock treatment. GST-1 was previously shown to mildly inhibit the druginduced degeneration of C. elegans dopaminergic neurons [28]. We found that gst-4 transcripts were approximately 1.5-fold upregulated in wild-type animals after 6-OHDA treatment in comparison to approximately 5-fold in tsp-17 and glit-1 mutants (Fig 6A and S15A Fig). Furthermore, gcs-1 and gst-1 transcripts were not induced by 6-OHDA in wild-type larvae, but an approximately 2-3-fold induction occurred in tsp-17 and glit-1 mutants (Fig 6B and 6C and S15B and S15C Fig). Comparing the basal level of gcs-1 and gst-1 and gst-4 expression in wildtype and tsp-17 and glit-1 mutants, we found that there is no difference in gcs-1 and gst-1 expression, while gst-4 appeared upregulated in tsp-17 and glit-1 mutants. (S14 and S15D-S15F Fig). In addition, we measured reporter expression after exposure to paraquat and found a modest level of hyperinduction of the reporters in tsp-17 and glit-1 mutants (S16 Fig). In summary, we found increased levels of oxidative stress signalling in glit-1 and tsp-17 mutants, indicating that the two genes protect from various oxidative-stress inducing compounds.
We tested additional C. elegans stress response pathways for their involvement in 6-OHDA protection. However, we found that mutation of the DAF-2/DAF-16 insulin/insulin-like growth factor signalling pathway (for review [34]), the JNK-1 and KGB-1 JNK (Jun N-terminal kinase) pathways and the PMK-1 and PMK-3 p38 MAPK (mitogen-activated protein kinase) pathways (for review [35]) did not overtly alter 6-OHDA-induced dopaminergic neuron loss (S17 Fig, S18 Fig). In addition, we tested if 6-OHDA induces the expression of reporters for the endoplasmic reticulum unfolded protein response, hsp-4, and the mitochondrial unfolded protein response, hsp-6. While hsp-4 transcript levels were slightly elevated after 6-OHDA, we did not detect a change in hsp-6 transcript levels (S19A-S19D Fig). Furthermore, we did not find a difference in hsp-4 and hsp-6 transcript levels between wild-type, and tsp-17 and glit-1 mutant larvae (S19A-S19D Fig). Thus, of the stress pathways we tested, only hyperactivation the skn-1 pathway leads to detectable protection from neurotoxicity conferred by 6-OHDA.
Mutations affecting the apoptosis pathway sensitise dopaminergic neurons to 6-OHDA It is not known which cell death pathway mediates 6-OHDA-induced neurodegeneration, so we revisited the role of the canonical apoptosis pathway. The conserved core apoptosis pathway is required for the vast majority of C. elegans apoptotic cell deaths and comprised of proapoptotic EGL-1 BH3-only domain protein, the Apaf1-like protein CED-4 and the CED-3 caspase. The apoptosis pathway also includes the Bcl2-like protein CED-9, which protects from apoptosis. CED-13 in contrast solely contributes to apoptosis induced by DNA damaging agents such as ionising irradiation [36] (simplified overview in Fig 7G, for review [37]). In addition to mediating apoptosis, CED-3, CED-4 and CED-13 were also reported to be part of a hormetic response mediating lifespan extension upon limited mitochondrial oxidative stress [38]. We found that compared to wild-type animals, egl-1, ced-13, ced-4 and ced-3 loss-of-function (lof) mutants and the ced-9 gain-of-function (gof) mutant exhibit increased 6-OHDAinduced dopaminergic neurodegeneration (Fig 7A-7F). Thus, the apoptosis pathway appears to protect dopaminergic neurons after 6-OHDA exposure. In contrast, we found that increased neurodegeneration in glit-1 mutants is not further enhanced by ced-3, ced-4 and ced-13 mutations (Fig 7B-7D, S20 Fig). It is conceivable that the increased neurodegeneration conferred by apoptosis gene mutations becomes undetectable in the glit-1 mutant background. Alternatively, it is possible that those apoptosis genes and glit-1 act in the same genetic pathway to protect dopaminergic neurons. In summary, our results show that the apoptosis pathway helps to prevent 6-OHDA-induced dopaminergic neurodegeneration.
Mutation of the calcium-storing chaperone crt-1 exacerbates 6-OHDAinduced dopaminergic neurodegeneration Since dopaminergic neurons do not undergo apoptosis after 6-OHDA exposure, we aimed to determine if they die via a necrosis-like cell death. Mutation of the C. elegans calreticulin CRT-1, a calcium-binding and calcium-storing chaperone in the endoplasmic reticulum, prevents several necrosis-type cell deaths [39,40]. In contrast, we found that crt-1 mutation led to a highly increased loss of dopaminergic neurons after exposure to 6-OHDA in an otherwise wild-type background (Fig 8A-8C). We further found that the crt-1;glit-1 and the crt-1;tsp-17 double mutants are synthetic lethal. We used a fluorescently labelled balancer chromosome to keep crt-1 in a heterozygous state in a glit-1 homozygous background. In the next generation, genetic segregation of the crt-1 and the fluorescent balancer chromosome is expected. However, we did not find any non-fluorescent animals, confirming that crt-1 homozygosity is lethal in the glit-1 mutant background. We furthermore found that synthetic lethality occurs when tsp-17(tm4995) and crt-1(ok948) are combined. Although it is possible that this synthetic lethality is related to the function of CRT-1 in calcium-induced necrosis, synthetic lethality might also be caused by developmental roles of crt-1: crt-1 mutation was shown to be synthetic lethal with vab-10 and other genes required for mechanical resilience of the epidermis [41]. Using qRT-PCR, we also tested if crt-1 expression is altered by 6-OHDA but we found crt-1 transcript levels remain unchanged after 6-OHDA treatment in wild-type, and in tsp-17 and glit-1 mutant animals (Fig 8D, S21B Fig). crt-1 expression also remains constant after paraquat treatment (S16 Fig). Furthermore, basal expression of crt-1 is not altered in tsp-17 or glit-1 mutants as compared to wild-type (Fig 8E, S21C Fig). To follow up on a more general role of calcium homeostasis in 6-OHDA-induced dopaminergic neurodegeneration, we also analysed itr-1 (inositol triphosphoate receptor) and cnx-1 (calnexin) mutants. ITR-1 and CNX-1 were proposed to play similar roles as CRT-1 in the calcium-induced necrotic cell death of C. elegans PLM (posterior lateral microtubule cells) neurons [39]. However, we found that unlike crt-1 mutation, itr-1 and cnx-1 mutations do not alter the extent of dopaminergic neuron loss after 6-OHDA exposure (S21A Fig). In summary, we report that the calreticulin orthologue CRT-1 is required for protecting against 6-OHDA-mediated dopaminergic neurodegeneration.

Discussion
We show that GLIT-1 protects C. elegans dopaminergic neurons from 6-OHDA-induced degeneration, likely in a similar way as the previously identified tetraspanin TSP-17 [13]. We found that glit-1 and tsp-17 mutants are equally hypersensitive to 6-OHDA and paraquat, and that double mutants do not exhibit an enhanced phenotype. We previously showed that tsp-17 mutants display phenotypes consistent with excessive dopaminergic signalling and we hypothesised that TSP-17 protects dopaminergic neurons by inhibiting the uptake of 6-OHDA into dopaminergic neurons [13]. In this study, we uncovered that both TSP-17 and GLIT-1 protect against oxidative stress at an organismal level. Consistent with increased oxidative stress levels in glit-1 and tsp-17 mutants, we find increased transcription of oxidative stress reporters and hypersensitivity after exposure to oxidative stress-inducing compounds. In addition, using a reverse genetics approach, we found that the calcium chaperone CRT-1, as well as canonical apoptosis pathway and CED-13, a protein previously implicated in DNA damage-induced apoptosis, are also required to protect dopaminergic neurons from 6-OHDA-mediated degeneration.
We are considering several possibilities as to how the GLIT-1 and TSP-17 transmembrane proteins might confer the various phenotypes we observe. glit-1 and tsp-17 are both expressed in dopaminergic neurons. However, in contrast to TSP-17, which also appears to partially function in dopaminergic neurons [13], GLIT-1 expression is likely required in multiple tissues to confer protection against 6-OHDA. Furthermore, tsp-17 mutants, but not glit-1 mutants display subtle defects in a behaviour mediated by dopamine, the swimming-induced paralysis (this study and [13]). Both the neuroligin-like glit-1 mutant and the tetraspanin tsp-17 mutant exhibit increased dopamine-induced paralysis. This dopamine sensitivity phenotype could point towards an increased level of intrinsic dopamine signalling, possibly accounting for an increased level of 6-OHDA uptake via the DAT-1 transporter. It is also conceivable that the increased sensitivity to dopamine-induced paralysis and oxidative stress is due to a higher organismal drug uptake, which might be caused by decreased endothelial barrier function. The GLIT-1 homologue in Drosophila, Gliotactin, localises to tricellular junctions and is necessary for the development of septate junctions, an invertebrate analogue of epithelial tight junctions [17,18]. However, using dye leakage assays, we did not find indications for increased intestinal permeability in glit-1 mutant animals.
We consider it likely that both glit-1 and tsp-17 protect constitutively against oxidative stress at an organismal level, as several oxidative stress reporters are hyper-induced in glit-1 and tsp-17 mutants. The specific sensitivity of dopaminergic neurons to 6-OHDA is likely due to the specific uptake of 6-OHDA via the dopamine transporter. In line with this, we never observed dopaminergic neurodegeneration in wild-type or mutant animals after treatment with paraquat and H 2 O 2 , even at doses at which animal development is severely delayed. Both TSP-17 and GLIT-1 are homologous to vertebrate tetraspanins and neuroligin-like molecules, but do not have clear mammalian orthologues. As the genetic screen for factors that protect dopaminergic neurons from oxidative damage is far from saturation, we consider it likely that further studies like ours will help to uncover additional pathways that protect neurons from oxidative damage.
Our data clearly indicate that the apoptosis pathway protects dopaminergic neurons from 6-OHDA-induced degeneration. The increased neuronal loss in egl-1 loss-of-function (lof), ced-9 gain-of-function, ced-3 (lof) and ced-4 (lof) mutants could be ascribed to a defect in the core apoptosis pathway defined by these genes. However, the ced-13 (lof) mutant also displays higher dopaminergic neurodegeneration. CED-13 has previously been found to mildly affect radiationinduced germ cell apoptosis, but not to alter developmental apoptosis as mediated by the core apoptosis pathway [36]. In addition, the core apoptosis pathway and CED-13 were reported to form a protective, lifespan-prolonging pathway that senses mild oxidative stress (conferred by the partial inhibition of mitochondrial respiration) [38]. Remarkably, glit-1 expression was increased under these oxidative stress conditions (supplementary information in [38]). It is still not known how apoptosis proteins sense oxidative stress [38], and it certainly remains to be established how they confer protection of dopaminergic neurons after exposure to 6-OHDA (our observations). There is mounting evidence that 'apoptosis proteins' adopt functions that are not related to apoptosis. For instance, it was demonstrated that ced-4 is required for DNA damage-induced cell cycle arrest [42] and we previously showed that CED-4 and CED-9 are expressed in both dying and surviving cells, consistent with roles beyond apoptosis [43].
We further found that the calreticulin CRT-1 protects against 6-OHDA-induced dopaminergic neurodegeneration, in contrast to the calnexin CNX-1 and the putative inositol phosphate receptor ITR-1. CRT-1 is a calcium-binding and calcium-storing chaperone of the endoplasmic reticulum (ER) which is predominantly expressed in the C. elegans intestine and induced by organismal stress [44]. CRT-1 is required for neuronal cell death induced by hyperactivated MEC-4(d) (mechanosensory abnormality) and DEG-1(d) (degeneration of certain neurons) ion channels, as well as by a constitutively active Gαs subunit [39]. Furthermore, crt-1 mutation alleviates dopaminergic cell death caused by gain-of-function mutations of the TRP-4 (transient receptor potential) channel [40]. ITR-1 and CNX-1 are less well characterised than CRT-1, but mutation of itr-1 also reduces neuron loss as induced by hyperactivity of MEC-4(d) and TRP-4, albeit to a lesser extent than mutation of crt-1 [39,40]. cnx-1 mutation only suppressed MEC-4(d), but not TRP-4-induce cell death [39,40]. CRT-1, ITR-1 and CNX-1 are thought to be required for calcium release from the ER that is needed for necrotic cell death [39]. When necrosis is linked to increased calcium influx across the plasma membrane, as conferred by the hyperactivated DEG-3(d) channel, necrosis is not blocked by crt-1 mutation [39]. Finally, CRT-1 also appears to influence the regeneration of axons after laser cut for ALM neurons [45], but not PLM neurons [46]. In the ALM axon regeneration model, the CED-3 and CED-4 apoptosis proteins also promote neuronal repair [45]. Is it possible to reconcile the differential influence of crt-1 mutation on different types of necrosis and axonal repair with the increased sensitivity to 6-OHDA-induced dopaminergic neurodegeneration? It has been speculated that depending on the levels and duration of the calcium signal, distinct pathways driving neuronal regeneration or cell death can be triggered [47]: Low calcium transients would promote neuronal repair, whereas high levels of increased calcium over a longer period of time would trigger necrosis. 6-OHDA might thus induce a different injury as compared to locally applied laser cuts or the aforementioned necrosis models.
The apoptosis pathway and calreticulin are conserved between C. elegans and humans. It is thus conceivable that both apoptosis and necrosis play a role in neurodegeneration in Parkinson's disease. To the best of our knowledge, apoptosis and necrosis genes have not been defined as Mendelian Parkinson's loci, and have not been associated with an increased risk of Parkinson's disease based on genome-wide association studies. However, it is noteworthy that the inheritance of Parkinson's disease is considered to be comparably low, and that the vast majority of disease cases appear to be defined by environmental factors. Thus, apoptosis or oxidative stress signalling linked to apoptosis proteins, as well as necrosis, could well have an important role in protecting dopaminergic neurons in idiopathic Parkinson's disease.

Methods
We note that experiments shown in this paper were done in parallel to experiments shown in the accompanying paper '6-OHDA-induced dopaminergic neurodegeneration in Caenorhabditis elegans is promoted by the engulfment pathway and inhibited by the transthyretin-related protein TTR-33' and thus share the wild-type control.

C. elegans strains and maintenance
Strains were grown at 20˚C. wdr-23(tm1817) and tsp-17(tm4995) mutants were generated and kindly provided by Shohei Mitani of the National Bioresource Project (NBRP) (http://shigen. nig.ac.jp/c.elegans/). Additional information about alleles in this study is provided by Worm-Base (www.wormbase.org). Strains from the Caenorhabditis Genetics Center (CGC) or the NBRP were outcrossed to the BY200 strain to eliminate unlinked mutations and to introduce a green fluorescent protein (GFP) marker for dopaminergic neurons. The isolated gt1981 mutant was outcrossed six times to the BY200 wild-type strain.

Mutagenesis and mapping
L4 and young adult staged BY200 animals were mutagenised with 25 mM ethyl methanesulfonate (EMS) in M9 buffer for 4 hours at 20˚C, washed and grown at 15˚C overnight on NGM plates seeded with OP50 E. coli. The animals were bleached once they started to lay eggs and the F1 progeny let hatch overnight in M9. This synchronised F1 population was left to develop on seeded plates until reaching the early adult stage. The F1 adults were then washed off in M9 again to lay a synchronised population of F2 L1 larvae which was treated with 10 mM 6-OHDA and screened for dopaminergic neuron loss after 72 h. Candidates were transferred to separate plates, backcrossed at least three times and re-tested for their hypersensitivity to 6-OHDA. The genomic DNA was sent for whole-genome sequencing and SNP mapping was performed as previously described [14,15].

Generation of transgenic animals and imaging
For generation of transgenic constructs, genomic DNA was amplified using primers containing additional 8 base pair restriction sites for insertion into the pCFJ151 vector [48]. AscI_Pdat-1 -ATTAggcgcgccAATGTTTCTAGTCGTTTTTGTATTTTAAAG; Pdat-1_FseI-AGAAAAggc cgg cca CATGGCTAAAAATTGTTGAGATTCG. The glit-1 coding sequence was amplified from cDNA (the AT-rich glit-1 introns rendered cloning unfeasible). The vector was adapted to include the following restriction site architecture: AscI_SgfI_NotI_FseI_PacI. The strains were generated by microinjections of unc-119(ed3) mutants. For microscopy, animals were transferred into a drop of 25 mM levamisole solution on top of a 2.5% (w/v) agar pad on a microscope slide. Images were acquired using a DeltaVision microscope (Applied Precision) and analysed with the softWoRx Suite (Applied Precision) and ImageJ [49].

C. elegans assays
All C. elegans assays were performed in a blinded manner. To obtain synchronized L1 larvae for oxidative stress assays, 1-10 adult animals were incubated to lay eggs in 70 μl M9 without food on a benchtop shaker at 20˚C, 500 rpm for 24-30 h. Due to the swimming-defect of tsp-17(tm4995), all L1 larvae for experiments that included this strain were obtained by filtering a mixed population of animals.
For 6-OHDA assays, 10 μL 200 mM ascorbic acid and 10 μL of the respective 6-OHDA 5 x stock concentration were added to 30 μL of L1-stage larvae in M9. 6-OHDA concentration was chosen such that differential sensitives of various strains could be efficiently analysed. After 1 hour incubation at 20˚C and shaking at 500 rpm, 150 μL M9 buffer was added to oxidise and inactivate the 6-OHDA. The animals were pipetted to one half of an NGM plate containing a stripe of OP50 bacteria on the opposite half of the plate and adult animals and eggs were removed to prevent growth of animals that were not treated at L1 stage. Plates were incubated at 20˚C before blind scoring of the 6-OHDA-induced degeneration using a Leica fluorescent dissecting microscope. Unless indicated otherwise, animals were scored 72h after treatment in technical duplicates. For treatment of different developmental stages, the respective stages were selected from a mixed population of animals after 6-OHDA treatment. Replicates for strains shown in the same graph were performed in parallel.
For paraquat and H 2 O 2 assays, 10 μL of the respective 5x stock concentration was added to 40 μL of L1-stage larvae in M9. After 1 hour incubation in the shaker, 150 μL M9 buffer was added and the L1 larvae were counted after pipetting them to a seeded NGM plate. Surviving animals were scored again after 24 h.
Dye leakage assays were performed as described previously [19,20]. A mixed, well-fed population of animals was washed off the plate and incubated for 3 hours in 2.5% (wt/vol) Brilliant Blue FCF (Sigma) and 2.5% (wt/vol) fluorescein sodium salt (Sigma), respectively. The dyes were solubilised in a saturated solution of OP50 E. coli in LB (Lysogeny Broth) medium. After incubation, the animals were pipetted onto NGM plate seeded with OP50 E. coli. For microscopy, single animals were transferred immediately into a drop of 25 mM levamisole solution on top of a 2.5% (w/v) agar pad on a glass slide. For the blue dye ('Smurf') assay, images were taken using a camera mounted to a Leica stereo microscope. To determine fluorescein uptake, a bright-field image and a fluorescence image using the GFP channel were taken with a Zeiss Axio Scope. Bright-field and fluorescence image channels were merged using ImageJ [49].
For lifespan assays, 50 L4 stage animals were selected per strain and transferred to new plates using a platinum wire every day during the first 6 days and then only to avoid mixing with the progeny. Animals that did not move after touch with a platinum pick were scored as dead. Dead animals were counted and removed every day. Bag of worms, dry animals and burst animals were censored. The data was analysed using OASIS (Online application for Survival Analysis, http://sbi.postech.ac.kr/oasis/) [50].
Dopamine paralysis assays were carried out as previously described [51]. Staged adult animals were incubated for 20 minutes on a plate containing dopamine and assessed for their ability to complete body movement through minimum and maximum amplitude during a 5 second observation period. Two plates containing 25 animals each were assessed for each condition.
For the swimming-induced paralysis (SWIP) assay 5-15 L4 stage larvae were transferred into a glass well in 40 μL water. The number of paralysed animals scored using a Leica dissecting microscope every minute for 30 minutes [25].
Basal slowing assays were performed as previously described [52]. NGM plates were prepared and half of them seeded with HB101 bacteria before incubating them at 37˚C overnight. Also, mid-L4 larvae were transferred on separate plates. The next day, these staged young adult animals were put in 40 μL M9 for 2 minutes to clean them from bacteria and then transferred to the centre of a plate. 6 animals each were put on a plate with and without bacteria, respectively. After a 2 minutes adaptation time, the locomotion rates of each animal was quantified by counting the number of body bends completed in five consecutive 20 s intervals. The experiment was performed twice on different days.

qRT-PCR
L1 stage animals were filtered from two 9 cm plates containing mixed stage animals. Animals were washed twice and then split in half for control treatment (addition of ascorbic acid and water only) and 6-OHDA treatment. After treatment, RNA extraction was performed based on He, F. (2011). Total RNA Extraction from C. elegans. Bio-protocol Bio101: e47. DOI: 10.21769/BioProtoc.47. The following changes were introduced: The chloroform extraction was repeated once more and 0.1 volumes of 3 M NaAc were added to the propanol extraction to help RNA precipitation. The RNA pellet was resuspended in 15 μL H2O. The cDNA was digested using the Turbo DNA-free kit (Ambion Life Technologies) in 18 μL total volume. 1 μg of RNA was reverse transcribed to cDNA with 15 bp oligo dT primers using the M-MLV Reverse Transcriptase (Promega) according to manufacturers' instructions. The cDNA was analysed by qRT-PCR using the StepOnePlus Real-Time PCR System (Applied Biosystems) applying melting temperature of 62˚C. The following primers were designed using Quant-Prime [53] . All reactions were run in triplicates. Relative abundance was calculated using the ΔΔCt method [54]. To control for template levels, values were normalised against the control genes Y45F10D.4 and pmp-3 [55,56].
For phylogenetic analysis of GLIT-1, NCBI Protein BLAST (Basic Local Alignment Search Tool) was used to search the 'reference proteins' database of the indicated organism against the GLIT-1 protein sequence. A multiple alignment was performed using COBALT with the 250 best-matching sequences from all organisms.
Protein alignments to visualise conservation of GLIT-1 domains and residues alignment of protein sequences was performed in CLC workbench Version 7.6.4 using the default parameters: Gap open cost 10.0, Gap extension cost 1.0, End gap cost as any other, Alignment very accurate (slow) and create a the tree (Tree construction method: Neighbour Joining, Protein distance measure: Jukes-Cantor, Bootstrapping: 100 Replicates).

Statistical analysis
The two-tailed t-test and were performed with the Analysis ToolPak Add-In in Microsoft Excel 2010. To choose a t-test assuming equal or unequal variances, respectively, the same Add-in was used to conduct an F-test.
To compare 6-OHDA treatment data, a G-Test was performed with R Studio 1.0.44 Desc-Tools descriptive statistics tools package. p-values were calculated for each biological replicate and for the pooled biological replicate data, and the least significant of these p-values is indicated in the graphs. Unless otherwise indicated, data is only marked as significant if all replicates and the pooled replicate data were found to be significant. p>0.05; G-Test comparing BY200 wild-type to mutant animal data). A significant p-value indicated if all or all but one replicate were found to be significant. (F) Cartoon illustrating dopamine signalling cartoon (adapted from [23]). Dopamine synthesis in dopaminergic (DAergic) neurons starts with tyrosine (Tyr), which is converted to L-DOPA (L-3,4-dihydroxyphenylalanine) by the tyrosine hydroxylase CAT-2. L-DOPA is then converted to dopamine (DA) by the aromatic amino acid decarboxylase BAS-1 (biotenic amine synthesis related). Dopamine is packed into vesicles by the vesicular monoamine transporter CAT-1. Postsynaptic dopamine signalling is stimulated by the D1-like dopamine receptor DOP-1 and inhibited by the D2-like dopamine receptor DOP-3. The D2-like receptor DOP-2 can be postsynaptic, or presynaptic and act as an autoreceptor. The D1-like (stimulatory) and the D2-like (inhibitory) dopamine receptors are indicated in blue and red, respectively. Dopamine is transported back into the dopaminergic neurons via the dopamine transporter DAT-1. Cholinergic neurons signal via the neurotransmitter ACh (acetylcholine), which in turn stimulates muscle action. (TIF) S12 Fig. Dopamine-mediated behaviours in glit-1 and tsp-17 mutants. (A) Animal speed (measured in body bends/20s) on plates without and with food ('+Food'). Indicated is the average speed per young adult animal (grey symbols) and the average speed per strain (black bar). cat-2 (abnormal catecholamine distribution) is a tyrosine hydroxylase mutant and defective in dopamine synthesis. Error bars = SEM of 2 biological replicates, each with 6 animals per strain and state. Total number of animals per condition n = 12 ( ÃÃÃÃ p<0.0001, ÃÃ p<0.0001; twotailed t-test). (B) Percentage of swimming L4 stage animals. Error bars = SEM of 3-4 biological replicates with 12-18 animals per strain. Total number of animals per strain n = 50-60 ( Ã p<0.05, n.s. p>0.05; two-tailed t-test comparing wild-type and mutant data at 30 minute time point). (TIF) S13 Fig. glit-1 and tsp-17