A central role for TOR signalling in a yeast model for juvenile CLN3 disease

Yeasts provide an excellent genetically tractable eukaryotic system for investigating the function of genes in their biological context, and are especially relevant for those conserved genes that cause disease. We study the role of btn1, the orthologue of a human gene that underlies an early onset neurodegenerative disease (juvenile CLN3 disease, neuronal ceroid lipofuscinosis (NCLs) or Batten disease) in the fission yeast Schizosaccharomyces pombe. A global screen for genetic interactions with btn1 highlighted a conserved key signalling hub in which multiple components functionally relate to this conserved disease gene. This signalling hub includes two major mitogen-activated protein kinase (MAPK) cascades, and centers on the Tor kinase complexes TORC1 and TORC2. We confirmed that yeast cells modelling CLN3 disease exhibit features consistent with dysfunction in the TORC pathways, and showed that modulating TORC function leads to a comprehensive rescue of defects in this yeast disease model. The same pathways may be novel targets in the development of therapies for the NCLs and related diseases.


INTRODUCTION
Yeasts have long been used as model systems to shed light on basic eukaryotic cell biology, and can provide a rapid and comprehensive route of investigation for genes of unknown function. This is particularly relevant for those genes that are conserved across diverse eukaryotic species and which can be presumed to play a fundamental biological role. A significant health challenge facing current research and drug development is the increasing incidence of age-related neurodegenerative disorders. Elucidating the mechanisms that underlie neurodegeneration is complex, particularly so in common age-related dementias that may have numerous contributing factors. There are, however, monogenic inherited neurodegenerative diseases that present a simpler alternative for investigation and real opportunities to determine these basic cellular changes, and some of the underlying genes are conserved even in yeasts. The genetic tractability of yeasts offers particular advantages to the challenges of understanding disease mechanisms in a relevant biological context and this knowledge can inform therapeutic development for such conserved genetic diseases.
The neuronal ceroid lipofuscinoses (NCLs) are such a group of monogenic neurodegenerative disorders that generally affect children [1]. The most common of these is juvenile CLN3 disease [2]. This disease is characterised by progressive neuronal atrophy that causes visual failure, seizures and a progressive decline in cognitive and motor function. This disease is accompanied by cellular features characteristic of many neurodegenerative conditions [3,4] that include the dysfunction of core cellular processes, such as reduced lysosomal and autophagic clearance [5,6] and mitochondrial abnormalities [7]. Moreover, CLN3 disease also leads to specific cellular pathologies characteristic of more common dementias. These include the accumulation of lipofuscin [1], as observed in aged neurons [4], aberrant amyloid-β processing [5], a feature of Alzheimer's disease [8] and α-synuclein accumulation [9], a feature of Parkinson's disease [10].
Juvenile CLN3 disease is caused by mutations in a single gene (CLN3) [2], whose function is unknown but which is highly conserved across eukaryotic species [11]. As a consequence, this disease is relatively straightforward to reproduce in experimental systems and, given its similarity to more common dementias, such experimental models are ideal paradigms to study the basic cellular changes that occur in neurodegenerative diseases. The fission yeast Schizosaccharomyces pombe contains a single orthologue of CLN3 (btn1) [11]. Work in this yeast has revealed a role for btn1 in many cellular processes. Like patient cells, yeast lacking btn1 (btn1Δ) have enlarged and less acidic vacuoles [12] and further work has supported roles for btn1 in vacuolar homeostasis [13]. The ability of btn1 modeling the most common CLN3 mutation (a 1 Kb deletion) to rescue these vacuolar defects found in the btn1Δ null strain has also revealed that this mutant allele retains some function, making juvenile CLN3 disease a mutation-specific disease rather than the consequence of a complete loss of function as originally assumed [14].
A host of further morphological defects exhibited in the btn1Δ strain support additional roles for btn1 in cytokinesis [12] and the organisation of cell polarity [15]. Furthermore, btn1Δ cells also display cell wall defects [16]. Importantly, work in this model has expanded our understanding of the disease by revealing that btn1 is involved in two independent pathways; one pH-dependent and one pH-independent, thus providing the first suggestion that Batten disease is more than a pH-related lysosome disorder [16]. Indeed, a comprehensive metabolomics approach has revealed that btn1 is required for the regulation of glycolysis and amino acid homeostasis [17]. The involvement of btn1 in numerous, apparently disparate, pathways may be a result of alterations at the Golgi apparatus, as the number, morphology, and location of thus organelle are affected by its deletion [13]. Lastly, this model has been used successfully to model disease mutations in Btn1p, an effort that has provided valuable insight into their consequences on protein trafficking and function. It was found that equivalent CLN3 disease mutations in btn1 affect the yeast phenotype in a way that can accurately predict the severity of disease, further establishing yeast as an accurate disease model despite its simplicity. Importantly, the observations reported in the fission yeast model have consistently been confirmed in mammalian systems [5,18,19]. Unfortunately however, despite these insights, the molecular processes that underlie cell death in this disease are poorly understood, the function of CLN3 is unknown, and there remains a significant need for protective therapeutic targets.
A particular advantage of yeast model systems is the availability of genome-wide techniques. Synthetic genetic arrays (SGAs) have proved a particularly powerful means of exploring genetic interactions in yeast species [20]. This approach highlights genes involved in pathways parallel to, or converging with, the query gene. This provides information about functional relationships among genes, as well as processes that suppress the defects associated with a particular mutation. As they are hypothesis-free, SGAs are particularly valuable in the investigation of complex biological problems and those where gene function is unclear. These advantages are particularly relevant for neu-rodegeneration in general, due to the complexity of the problem, and for juvenile CLN3 disease in particular, due to the lack of a clear gene function. SGAs have previously been employed in budding yeast to investigate mutant huntingtin and α-synuclein toxicity [21]. Such an approach can place the gene under investigation within its biological context and thereby uncover much-needed protective pathways for neurodegenerative disease.
We applied SGA analysis to identify pathways that are altered as a consequence of loss of function of btn1 in an effort to better understand the molecular consequences of CLN3 disease, and to provide new candidate target pathways and processes for therapeutic development. A third of the genetic interactions that were identified centered on a set of conserved and connected signalling pathways. Manipulation of these pathways leads to a complete rescue of the pleiotropic array of btn1Δ phenotypes. This approach represents the most successful rescue of cellular dysfunction in any model for juvenile CLN3 disease to date.

Genome-wide analysis of genetic interactions with btn1 reveals a central role for TOR kinases
We applied an SGA approach as an unbiased, genome-wide strategy to probe the interactions of a conserved neurodegenerative disease gene (btn1). We identified 331 positive interactors and 131 negative interactors of btn1 (Table  S1A and B). A large number of negative interactions (n = 39; 29.7%) mapped to a set of highly interconnected signalling processes, which center on the Tor kinase complexes TORC1 and TORC2 (Fig. 1). Core components of both complexes display genetic interactions with btn1; tco89, encoding a component of TORC1, was found to interact positively with btn1, while tor1, encoding the Tor kinase of TORC2, was found to interact negatively with btn1. Such a pattern suggests that TORC2 signalling is beneficial to the fitness of btn1Δ cells, whereas TORC1 signalling is detrimental, an observation that is in keeping with the idea that the two complexes play opposing roles and undergo mutual repression [22][23][24]. btn1 was also found to negatively interact with genes encoding components of the connected cell wall integrity (CWI) and stress-associated protein kinase (SAPK) MAP kinase cascades, and a number of downstream stress regulated genes (Table S1A and B).

btn1Δ cells display features consistent with dysfunctional Tor signalling
The interaction of btn1 with core TORC components, and the link to surrounding signalling processes, provides compelling evidence for the importance of Tor signalling in cells lacking btn1. To validate this observation, we investigated btn1Δ cells for features consistent with Tor dysfunction. Repression of TORC1 activity is required to mount a correct response to nitrogen limitation [24]. As loss of TORC1 activity was beneficial in cells lacking btn1, we hypothesised that these cells may display features consistent with dysregulation of TORC1 repression, and respond poorly to nitrogen limitation. To test this idea, we grew wild-type and btn1Δ cells in minimal media, and minimal media lacking a nitrogen source, over 72 h. Viability was monitored at 24 h intervals using propidium iodide to label dead cells and calcofluor white to label the total cell population ( Fig.  2A). Cells lacking btn1 displayed a consistently lower viability in media lacking nitrogen, falling to 81.5 ± 2.2% after 24 h compared to 98.7 ± 0.27% in minimal media containing nitrogen (P < 0.0015, unpaired t test), and remaining lower throughout the time course. Wild-type cells displayed no change in viability when cultured in media lacking nitrogen.
TORC1 repression and TORC2 signalling is also required to mount a response to glucose limitation [25,26]. Cells that are defective in TORC2 function do not respond appropriately to glucose limitation by reducing their cell length [26]. This is a relatively uncommon phenotype, with only tor1 mutants and mutants in the Ca 2+ /calmodulindependent-like gene ssp1 known to display this phenotype [26,27]. To test the ability of btn1Δ cells to adapt to conditions of limited glucose, we assessed both cell growth and the morphological response of these cells when grown with an alternative carbon source (glycerol). Cells lacking btn1 displayed a clear growth defect under these conditions (Fig. 2B). Wild-type and btn1Δ cells grown for 6 h in media containing either glucose or glycerol as a carbon source were stained with calcofluor white to visualise the cell wall. Cell elongation was determined using a measure of cell circularity and given a score of between 0 and 1, where 0 represents a perfectly round cell ( Fig. 2C and D). Wild-type cells displayed a significant reduction in cell elongation upon growth in glycerol (0.25 ± 0.01 to 0.16 ± 0.01, P ≤ 0.01). Cells lacking btn1, however, displayed no significant change (0.32 ± 0.02 to 0.31 ± 0.02), consistent with a defect in their response to glucose limitation that is most likely linked to Tor1 function.
Lastly, TORC2 and the connected CWI pathway are also required for resistance to high temperature [26]. This temperature sensitivity of TORC2 and CWI pathway components is osmoremedial, i.e. it can be rescued by hypertonic growth media [28]. We confirmed the temperature sensitivity of btn1Δ cells, and that it can be rescued with 2M Cell viability upon nitrogen limitation was determined over periods of up to 72 hours in wild-type (WT) and btn1Δ cells, using the cell impermeable nucleotide stain propidium iodide to stain dead cells and calcofluor white to stain the total cell population. Cells were cultured in either MM or MM lacking a nitrogen source (NH4Cl). 500 cells were scored for viability per data set, and data shown is a mean (± SEM) of 3 independent experiments. (B) WT cells and btn1Δ cells were serially diluted from a log-phase culture (1 x 10 6 cells/ml), and spotted onto plates containing either glucose or glycerol as a carbon source. Plates were then incubated at 30°C for 6 -7 days to determine growth on fermentative and non-fermentative carbon sources. Images are representative of three independent experiments. (C) The morphological response of WT and btn1Δ cells to growth on glycerol was analysed following 6 hours in culture using a measure of cell elongation, on a scale of 0 to 1, where 0 represents a perfectly round cell (1 − ( )). Data shown is a mean (± SEM) of 5 independent experiments. Statistical significance between each condition was determined using a one-way ANOVA with a Tukey's multiple comparison post-test (** = P < 0.01). (D) Representative images of experiments as performed in (C) are shown. Scale bar represents 10 μm. (E) WT and btn1Δ cells were also serially diluted from a log-phase culture (1 x 10 6 cells/ml) and spotted onto YES plates and YES plates containing 2M sorbitol. Plates were then incubated at 37°C for 3 -4 days to determine the growth at high temperature and the influence of osmotic stabilisation. Images are representative of three independent experiments. sorbitol ( Fig. 2E), consistent with previous observations [16]. These data together indicate that cells lacking btn1 display many features consistent with defects in Tor signalling and function. Further, ectopic expression of btn1 was able to rescue all these aspects of the mutant phenotype (Fig. S1A -D). Combined with the SGA results, these results suggest that Tor signalling is critical to the defects observed in FIGURE 3: Increasing TORC2-dependent signalling rescues aspects of btn1Δ phenotype. (A) Cells lacking btn1 were transformed with an expression vector containing gad8. Vacuole size was measured in these cells, in addition to WT and btn1Δ cells containing empty vector, following staining with the vital dye FM4-64. The diameter of 300 vacuoles was determined per data set, and data shown is a mean (± SEM) of 4 independent experiments. Statistical analysis was performed using a one-way ANOVA with a Tukey's multiple comparison post-test (** = P < 0.01, *** = P < 0.001). (B) Representative images of experiments as performed in (A) are shown. Scale bar represents 10 μm. (C) The morphological response of these populations to growth on glycerol was analysed following 6 hours in culture using a measure of cell elongation, on a scale of 0 to 1, where 0 represents a perfectly round cell (1 − ( )). Data shown is a mean (± SEM) of 5 independent experiments. Statistical significance between each condition was determined using a one-way ANOVA with a Tukey's multiple comparison post-test (*** = P < 0.001). (D) Representative images of experiments as performed in (C) are shown. Scale bar represents 10 μm. (E) These cells were serially diluted from a log-phase culture (1 x 10 6 cells/ml) and spotted onto YES plates. Plates were then incubated at 30°C or 37°C for 3 -4 days to determine growth at high temperature. They were also spotted onto plates containing either glucose or glycerol as a carbon source. Plates were then incubated at 30°C for 6 -7 days to determine growth under non-fermentative conditions. Images are representative of three independent experiments. (F) Viability upon nitrogen limitation was determined over periods of up to 72 hours in these cell populations, using propidium iodide to stain dead cells and calcofluor white to stain all cells. Cells were cultured in either MM or MM lacking a nitrogen source (NH4Cl). 500 cells were scored for viability per data set, and data shown is a mean (± SEM) of 3 independent experiments. btn1Δ cells.

TORC2 and CWI pathway: modulating distinct signalling nodes elicit different levels of correction in btn1Δ cells
Given the negative genetic interaction between btn1 and tor1, increasing TORC2 activity could alleviate aspects of the btn1Δ phenotype. We initially explored this approach through overexpression of the downstream kinase Gad8, which is a multicopy suppressor of aspects of the loss of TORC2 [29]. A key phenotype in cells lacking btn1, and one that is strongly linked to the phenotype of CLN3 disease, is a change in vacuole homeostasis [12]. One of the clearest manifestations of this feature in btn1Δ cells is an increase in vacuole size [12]. We investigated vacuole size in wildtype and btn1Δ cells containing empty vector alone (as a control), as well as btn1Δ cells expressing gad8 from the nmt1 promoter of the pREP41 plasmid ( Fig. 3A and B). Consistent with previous reports, btn1Δ cells exhibited significantly larger vacuoles than wild-type cells (0.65 ± 0.03 μm compared to 0.39 ± 0.01 μm, P ≤ 0.001). The overexpression of gad8 in btn1Δ cells significantly rescued vacuole size (0.40 ± 0.03 μm) (P ≤ 0.01), restoring them to near that of wild-type. These data provide the first indication that increasing TORC2 function significantly rescues a disease-relevant phenotype of btn1Δ cells.
As discussed above, the ability of cells to respond to glucose-limitation with a change in cell size requires TORC2 function [26]. To investigate whether the inability of btn1Δ cells to undergo this morphological response can be rescued by increasing TORC2 pathway activity, we assessed the elongation of btn1Δ cells overexpressing gad8 following growth for 6 hours in media containing either glucose or glycerol as a carbon source ( Fig. 3C and D). Wild-type and btn1Δ cells containing empty vector alone displayed comparable responses to cells without vector (WT glucose -0.28 ± 0.01, WT glycerol -0.19 ± 0.01, P ≤ 0.001; btn1Δ glucose -0.32 ± 0.02, btn1Δ glycerol -0.28 ± 0.01, ns). The overexpression of gad8 was able to significantly rescue this morphological response, despite these cells being elongated under growth in glucose (glucose -0.36 ± 0.01, glycerol -0.19 ± 0.01, P ≤ 0.001).
Despite this rescue, expression of gad8 was unable to significantly rescue the growth of btn1Δ cells on glycerol or at high temperature, suggesting that the ability of these cells to adapt to heat and nutrient stress is still impaired (Fig. 3E). Similarly, expression of gad8 was unable to rescue the viability of btn1Δ cells grown in the absence of a nitrogen source (Fig. 3F).
Given that the phenotypes not rescued by increasing TORC2 activity were linked to stress and nutrient adaptation, we reasoned that modulating processes more closely linked to the CWI pathway might be more effective in correcting these phenotypes. These processes seem particularly relevant in this instance, as this pathway is known to respond to both heat and glucose limitation. In particular, mutants of the pmk1 MAP kinase, like btn1Δ mutants, display a partial growth defect on glycerol. Further, this pathway is required for proper activation of Sty1 (the MAP kinase of the SAPK pathway) under glucose-limitation, inte-grating these processes with further identified btn1 interactors [30].
To increase CWI activity, we chose to increase Rho GTPase levels in these cells, which represent a key hub in CWI pathway regulation, by overexpressing the btn1 interactor rho5 and its essential paralogue rho1. Overexpression of rho1 significantly decreased vacuole size in btn1Δ cells (P ≤ 0.001) to near that of wild-type cells containing empty vector alone (WT -0.30 ± 0.02 μm, btn1Δ -0.47 ± 0.04 μm, btn1Δ with p41rho1 -0.32 ± 0.01 μm) ( Fig. 4A and B). The overexpression of rho5 led to an over-correction of this phenotype (0.19 ± 0.01 μm), producing vacuoles significantly smaller than those of btn1Δ cells (P ≤ 0.001) and wild-type cells containing vector alone (P ≤ 0.05). Consistent with these observations, over-expression of pmp1, which inhibits the activity of pmk1 [31], caused an increase in vacuole size in btn1Δ cells (btn1Δ with p41pmp1 -0.80 ± 0.08 μm vs btn1Δ -0.52 ± 0.04 μm, P ≤ 0.01) but not in wild-type cells (WT with p41pmp1 -0.24 ± 0.01 μm vs WT -0.23 ± 0.02, ns) (Fig. 4C). Such data indicate that increasing CWI pathway activity is able to correct the vacuole defect of btn1Δ cells, and directly inhibiting this pathway exacerbates the vacuole defect. The over-correction by rho5 could indicate a strong dose-dependence in this particular rescue. Indeed, previous work has demonstrated such dose dependence in the relationship between btn1 and vacuole size [12].
We next investigated whether the Rho GTPases were able to rescue the morphological response of these cells to glycerol, as seen with gad8. We looked just at rho1 in this case, as rho5 overexpression has a pronounced effect on cell shape and septation [32], making morphological comparisons difficult (Fig. S2). Overexpression of rho1 elicited a rescue of this response, with cells displaying a significant reduction in cell elongation when grown in glycerol containing media (0.28 ± 0.02 to 0.16 ± 0.01, P ≤ 0.01) (Fig. 4D and E).
Further to a rescue of the morphological response to glycerol, both rho1 and rho5 were able to correct the growth defect of btn1Δ cells on glycerol as a carbon source. In addition, rho1 was able to rescue the temperature sensitivity phenotype of these cells (Fig. 4F). The overexpression of rho5 did not correct this phenotype, although, given the link between heat-sensitivity and septation defects in btn1Δ cells [15] and the enhanced septation defect of cells overexpressing rho5 (Fig. S2), this is not wholly unexpected.
In addition to these phenotypes, overexpression of rho1 and rho5 rescued the viability of btn1Δ cells when grown in the absence of a nitrogen source, with viability remaining higher than 95% throughout the course of the experiment (Fig. 4G). These data provide evidence for a profound positive interaction between CWI pathway signalling and btn1, indicating that such processes are intimately linked with the defects that occur in btn1Δ cells.
Reducing TORC1 activity corrects defects in btn1Δ cells TORC1 activity is a key signal of a favourable environment, promoting proliferation and suppressing stress-responsive cells overexpressing rho1 and rho5 were serially diluted from a log-phase culture (1 x 10 6 cells/ml) and spotted onto YES plates. Plates were then incubated at 30°C or 37°C for 3-4 days to determine growth at high temperature. They were also spotted onto plates containing either glucose or glycerol as a carbon source. Plates were then incubated at 30°C for 6-7 days to determine growth under non-fermentative conditions. Images are representative of three independent experiments. (G) Viability upon nitrogen limitation was determined over periods of up to 72 hours in these cell populations, using propidium iodide to stain dead cells and calcofluor white to stain all cells. Cells were cultured in either MM or MM lacking a nitrogen source (NH4Cl). 500 cells were scored for viability per data set, and data shown is a mean (±SEM) of 3 independent experiments. processes [24]. As a consequence, it is antagonistic to many of the processes explored in this study that improved the btn1Δ phenotype. TORC1 represses TORC2 function [23], and negatively regulates Sty1 [33], which displays extensive cross-talk with the CWI pathway [30]. This relationship is highlighted by the fact that the core TORC1 component gene tco89 is a positive interactor of btn1.
In order to test whether repression of TORC1 activity could rescue the phenotypes of btn1Δ cells, we first overexpressed a dominant-negative form of the upstream activator of TORC1, Rhb1 (rhb1 D60K ) [34]. Upon overexpression of dominant-negative rhb1 (dnrhb1) in btn1Δ cells, we observed a significant reduction (P ≤ 0.001) in vacuole size compared to btn1Δ cells containing vector alone (btn1Δ with p41dnrhb1 -0.40 ± 0.04 μm, btn1Δ with empty vector -0.65 ± 0.03 μm) ( Fig. 5A and B). This result was comparable to the vacuole size of wild-type cells (0.39 ± 0.01 μm), indicating a rescue of vacuole morphology upon repression of TORC1 activity.
We next investigated the ability of TORC1 repression to rescue the nutrient-sensing defects of cells lacking btn1. The overexpression of dnrhb1 led to a significant rescue of the morphological response of these cells to glucoselimitation, following 6 hours growth in glycerol as a carbon source (Fig. 5C and D). Cell elongation fell from 0.29 ± 0.02 to 0.18 ± 0.01 when grown in glycerol as opposed to glucose, a response comparable to that observed in wild-type cells (glucose -0.28 ± 0.01, glycerol -0.19 ± 0.01). This also corresponded to a rescue of the growth defect of these cells on glycerol (Fig. 5E). Further to the glycerol growth Data shown is a mean (± SEM) of 5 independent experiments. Statistical significance between each condition was determined using a one-way ANOVA with a Tukey's multiple comparison post-test (*** = P < 0.001). (D) Representative images of experiments as performed in (C) shown. Scale bar represents 10 μm. (E) These cells were serially diluted from a log-phase culture (1 x 10 6 cells/ml) and spotted onto YES plates. Plates were then incubated at 30°C or 37°C for 3-4 days to determine growth at high temperature. They were also spotted onto plates containing either glucose or glycerol as a carbon source. Plates were then incubated at 30°C for 6-7 days to determine growth under non-fermentative conditions. Images are representative of three independent experiments. (F) Viability upon nitrogen limitation was determined over periods of up to 72 hours in these cell populations, using propidium iodide to stain dead cells and calcofluor white to stain all cells. Cells were cultured in either MM or MM lacking a nitrogen source (NH4Cl). 500 cells were scored for viability per data set, and data shown is a mean (± SEM) of 3 independent experiments. Scale bar represents 10 μm. defect, dnrhb1 overexpression also rescued the heat sensitivity of btn1Δ cells (Fig. 5E). Finally, this construct was also able to elicit a complete rescue of viability under nitrogen limiting conditions, with viability remaining above 95% throughout the course of the experiment (Fig. 5F).
Given the substantial rescue of btn1Δ cells by dnrhb1mediated activation of TORC1, we wanted to confirm the interaction between btn1 and TORC1 by another means. We chose to use two pharmacological inhibitors of TORC1, rapamycin (allosteric) and caffeine (competitive inhibitor of ATP binding) [35]. TORC1 antagonism leads to cell rounding, a response that mimics the morphological change ordinarily exhibited by wild-type cells in response to nitrogen and amino acid starvation. Therefore, we examined the effect of TORC1 antagonism on cell rounding [36] and on vacuole size, heat sensitivity, and growth in glycerol. As expected, wild-type cells displayed a significant reduction in cell elongation upon rapamycin treatment for 6 hours (0.26 ± 0.02 to 0.15 ± 0.01, P ≤ 0.01) ( Fig. 6A and B). This reduction in cell elongation did not change further upon exposure to both rapamycin and caffeine [35]. Cells lacking btn1, however, did not respond to rapamycin, displaying only a slight morphological change (0.30 ± 0.01 to 0.25 ± 0.02) that was slightly enhanced by the addition of caffeine and rapamycin, but again not significantly (0.22 ± 0.03) (Fig. 6A and B). The addition of caffeine and rapamycin to wild-type cells did not significantly affect vacuole size (vehicle -0.37 ± 0.05 μm, rapamycin -0.31 ± 0.03, caffeine and rapamycin 0.24 ± 0.01) (Fig. 6C and D). Rapamycin alone had no effect on the vacuole size of cells lacking btn1, but, in contrast, a combination of caffeine and rapamycin significantly (P ≤ 0.001) reduced vacuole size (vehicle -0.58 ± 0.03 μm, rapamycin -0.55 ± 0.04, caffeine and rapamycin 0.30 ± 0.02) (Fig. 6C and D). Further, treatment with rapamycin alone elicited a rescue of both the heat sensitivity defect and glycerol growth defect of cells lacking btn1 (Fig. 6E). The ability of these pharmacological treatments to rescue the defects in vacuole size, heat sensitivity and growth in glycerol of btn1Δ cells appears to be closely linked to their ability to target TORC1 signalling itself, as opposed to changes (E) WT and btn1Δ cells were serially diluted from a log-phase culture (1 x 10 6 cells/ml) and spotted onto YES plates either containing or lacking rapamycin (100 μg/ml). Plates were then incubated at 37°C for 3-4 days to determine growth at high temperature. They were also spotted onto plates containing glycerol as a carbon source either containing or lacking rapamycin (100 μg/ml). Plates were then incubated at 30°C for 6-7 days to determine growth under non-fermentative conditions. Images are representative of three independent experiments. in the activity of the connected SAPK pathway, as expression of a constitutively active wis1 mutant (wis1DD) [37] did not rescue these phenotypes (Fig S3A -C). These data confirm that activation of the TORC1 pathway pharmacologically in btn1Δ cells rescues most of the phenotypes arising from loss of btn1 function. However, the pharmacological inhibition of the TORC1 pathway is not completely equivalent to the inhibition of the TORC1 pathway using dnrhb1 in btn1Δ cells.

DISCUSSION
In this study, SGA analysis was used as an unbiased approach to identify the genetic interactors of btn1, and place btn1 into its biological context within the whole cell. This approach highlighted a key role for the Tor kinase complexes, TORC1 and TORC2, in the dysfunctional changes that occur in fission yeast lacking btn1, as one third of the interactions connected directly with pathways in which these complexes are active. Further, we have shown that btn1Δ cells display defects in their stress response to nitrogen and glucose limitation, in addition to osmoremedial heat sensitivity, consistent with TORC1 and TORC2 dysfunction in these cells (Fig. 7).
Specifically, the SGA identified negative interactions in tor1, the core kinase component of TORC2, multiple components of the stress-activated protein kinase (SAPK) pathway, and a component of the linked CWI pathway. In addition, tco89, encoding a core component of TORC1, which antagonises both the SAPK cascade and TORC2 signalling, was identified as a positive genetic interactor of btn1. btn1 was also found to interact negatively with a number of genes that display stress-responsive expression.
TORC1 negatively regulates a number of processes associated with cellular catabolism and adaptation to stress [24]. One of these roles is in the repression of autophagy. Autophagy similarly appears important to the fitness of btn1Δ cells, as atg14 and atg12 were identified as negative interactors of btn1, and both are required for autophagosome formation [38]. In support of this work, and the use of yeast as a model, defective autophagy has previously been linked to CLN3 disease [39]. In repressing stressresponsive processes, TORC1 is known to negatively regulate sty1, the mitogen-activated protein (MAP) kinase and component of the stress-activated protein kinase (SAPK) cascade [33]. Both wis4 and wis1, encoding the MAP kinase kinase kinase and MAP kinase kinase of the SAPK pathway, The left panel shows the response of healthy cells to stresses applied in this study, such as low nitrogen or glucose or raised temperature. These cells are able to respond via interconnected signaling pathways such as CWI, TORC and SAPK. Activation of these pathways allows the cell to respond to the effects of stress and restore cell homeostasis and integrity. In contrast, cells lacking btn1 are unable to respond to the applied stresses (represented by '?' and 'X') and undergo a variety of previously reported responses that include an increase in vacuole size, swollen and disorganized Golgi and mistrafficking of vacuole enzymes such as carboxypeptidase Y, increased septation index and longer cell cycle and cell length, monopolar growth, defective cell wall, absence of cell rounding (in response to low glucose). Genetic or pharmacological activation of the three signalling pathway provides partial or full rescue of these pleiotropic phenotypes of btn1Δ cells. G, Golgi; Sec, secretory vesicles; V, vacuole; SI, division septum; M, monopolar growth. respectively, interact negatively with btn1, as does the transcription factor gene pap1, which is a downstream target of sty1 [40]. Furthermore, btn1 interacts negatively with a number of genes that display stress-responsive expression.
TORC2 is a positive regulator of stress-responsive processes [29], and is also involved in cytoskeletal organization, vacuolar morphology and glucose sensing [22]. In addition to a negative interaction with the gene encoding the Tor kinase component of TORC2 (tor1), btn1 also negatively interacts with genes encoding components of the connected CWI MAP kinase cascade. The relationship between TORC2 and the CWI pathway is complex. TORC2 is required for the response to cell wall stressors [41], and is a direct regulator of the CWI pathway through the activation of the Rho guanine-nucleotide exchange factor (GEF) ROM2 in budding yeast [42]. In fission yeast, it has recently been shown that Tor1, acting as part of TORC2, enhances Pck2 synthesis to activate the CWI under both cell wall stress and glucose limitation conditions [43].
Despite this positive relationship, recent work in fission yeast has indicated a negative regulation of TORC2 by pmk1, the MAP kinase component of the fission yeast CWI pathway, at least in certain conditions [25]. In fission yeast both Rho1 and Rho2 are well established as regulators of the CWI pathway [44,45]. Rho5 acts as a functional homologue of rho1, sharing the role of regulating cell wall homeostasis [46]. More recently, rho5 has also been shown to be an upstream regulator of the CWI [47]. The regulation of this pathway is complex as Rho1 both positively [45] and negatively [48] regulates the downstream MAP kinase cascade. This potentially intricate relationship is reflected in the interactors of btn1 that fall within this pathway, that exhibit both positive and negative effects upon signalling. btn1 exhibits a negative interaction with the rho1 homologue gene rho5, and a number of genes involved in downstream cytoskeletal and polarity components. Despite this, the Rho GEF gene rgf1, which is a positive regulator of the CWI pathway [49], is a positive interactor of btn1 whereas rdi1 (encoding a GDP dissociation factor) is a negative interactor. This multifaceted pattern of interactions is consistent with current understanding of this pathway, given the complex interactions between these signalling nodes.
Although a process that maintains CWI might seem fungal-specific, and not relevant to disease in a higher eukaryote, previous work in budding yeast revealed similar dysfunctional processes in cells overexpressing α-synuclein as a model for Parkinson's disease, suggesting that there is a mammalian counterpart to the CWI pathway in yeast. Such observations are of specific relevance to CLN3 disease, as α-synuclein is also upregulated in this condition [9]. RHO1 signalling and the CWI pathway are disrupted by overexpression of α-synuclein [50]. Further, it was demonstrated that similar changes in downstream kinase cascades could be observed in mammalian cells overexpressing α-synuclein. Affected kinases included c-Jun, which has previously been linked to CLN3 disease, and is part of a pathway closely related to the SAPK pathway in yeast [51]. This is not the only instance in which Rho signalling has been linked to neurodegeneration. Leucine-rich repeat kinase 2 (LRRK2), mutations in which represent the most common cause of familial late-onset Parkinson's disease, has been shown to bind Rac1, and to a lesser extent RhoA and Cdc42 [52]. All of these binding partners are members of the Rho GTPase family. The binding of LRRK2 to Rac1 led to its activation and relocalisation, and was lost when disease-causing mutations were introduced into LRRK2. Further, the overexpression of Rac1 was shown to be protective in these cells. Other studies have also demonstrated a protective effect of Rho signalling, focusing both on Rho GEFs [53] and the Rho GTPases themselves [54]. Such studies highlight how processes as seemingly specialized as CWI in yeast can be highly informative of processes relevant to disease in higher order eukaryotes. Further, they demonstrate that Rho-dependent signalling could be a vital neuroprotective process in more common neurodegenerative disease, as well as being a key pathway in the neurodegenerative disease model presented in this study.
Importantly, the validation of these interactions during this work revealed interventions that produced a complete rescue of all phenotypes investigated in the btn1Δ strain, including phenotypes linked to TORC1 function, TORC2 function, CWI and vacuole homeostasis. Among the interactions investigated, increasing Rho1 levels, and thereby upregulating the main signalling node in the CWI pathway, or repressing TORC1 function (pharmacologically or genetically) led to complete rescues.
More notably, a neuroprotective role of TORC1 repression is already well established in a number of different systems [55]. A protective effect of the TORC1 antagonist rapamycin has been demonstrated in models for Parkinson's disease [57], Huntington's disease [56], spinoceribellar ataxia type 3 [58] and retinal degeneration mediated by mitochondrial dysfunction [59]. In addition, Tor signalling is also elevated in Alzheimer's disease models [60].
It is important to note, however, that targeting TORC1 in disease presents some technical challenges. Although in some studies an increase in lysosome function has been observed upon rapamycin treatment [61], other work indicates that rapamycin is an incomplete antagonist of TORC1 and a poor inducer of lysosomal functions [62], in concert with our observations. Another complication of targeting TORC1 in disease is the antagonistic role of TORC2. TORC2 signalling and the connected CWI pathway were both protective in our model for CLN3 disease. Further to the protective role of Rho signalling in a Parkinson's disease model [50], TORC2/AKT signalling has also been shown to be neuroprotective in a separate study of Parkinson's disease [63]. As a result, general Tor antagonists, such as Torin1, which target both TORC1 and TORC2, do not produce the same positive outcomes that TORC1 specific antagonists are able to elicit [57]. In addition, chronic rapamycin exposure also leads to a reduction in TORC2 activity. This fact could explain why rapamycin has been shown to be detrimental in certain studies, for example in one study using a fruit fly model for Alzheimer's disease [64].

Conclusion
The data presented here highlight a set of interconnected pathways that, when modulated correctly, produce a profound rescue of the dysfunctional changes observed in a yeast model for neurodegeneration caused by mutations in a single gene (Fig. 7). Links between CLN3 and the response to stress have previously been reported in other model organisms [51,65], supporting the relevance of this yeast work to human CLN3 disease. Importantly however, a link between CLN3 and Tor signalling specifically, as reported here, is a novel mechanistic and therapeutic insight into CLN3 disease, the most common paediatric neurodegenerative disease. Furthermore, the pathways identified display strong parallels with known protective pathways in more common dementias, providing not only strong biological plausibility for the importance of these interactions, but also an indication that common processes may be exploited in therapeutic development for seemingly disparate neurodegenerative diseases.

Synthetic genetic arrays (SGAs)
SGA analysis was performed as described in [20], using a Singer RoToR robot (S. pombe settings) and the Bioneer haploid library (v2.0) and analysed as described previously [72]. A control SGA using ade6::NatMX4 as a query was performed simultaneously. The ade6Δ mutant does not alter the fitness of the deletion collection, and arising double mutants are used to determine Bioneer strain fitness and for colony size normalisation. After germination, haploid double mutants were pinned to YES + G418 + Nat in a 1536 format. After three days growth at 30°C, agar plates were imaged using a MultiDoc-It Imaging System (UVP) and processed for analysis using the Gitter image analysis R package [73]. Using Excel, the colony sizes produced by Gitter were normalized to the median colony size of each plate to account for within-plate positional growth effects and differential plate growth. These normalized values were used to establish the ratio of query btn1Δ mutant size to control (ade6:NatMX4) colony size. Ratios from the four btn1Δ isolates used for the screen were used to calculate a mean ratio. Using this mean value, colony-size ratio scores of < 0.8 were defined as negative interactions, and ratios of > 1.2 as positive interactions. Genetic interactions were scored from two independent biological experiments. Only interactions common to both replicates were considered robust enough for follow-up. Negative and positive interaction sets were tested using PANTHER [74].

Microscopy
All images were obtained using a Leica SPE2 true scanning confocal microscope with 63 x 1.4 NA oil immersion objective. Images were recorded using Leica AF software. Calcofluorwhite (Sigma-Aldrich) staining was performed as described previously [68]. Calcofluor-white was used to allow cell segmentation for assays of cell morphology. For assays of cell viability, propidium iodide (Sigma-Aldrich) was added to the cells (15 μg/ml final concentration) at the same point as the calcofluor-white. Staining was performed in a volume of 1 ml, allowed to proceed at room temperature for 10 min, and the cells were washed twice in MM before visualisation. No cell fixation was applied. The vital dye FM4-64 (Molecular Probes) was used to visualise vacuolar morphology. The staining procedure has been described previously [12].

Image analysis
All microscopy image analysis was performed using ImageJ software. For analysis of cell morphology, cells were segmented using the BOA plugin of the quantitative imaging of membrane proteins (Quimp) package [75]. Cell elongation was then determined using the equation: 1 − ( ). Thirty cells were counted per dataset.

Statistical analysis
All statistical analysis was performed using Prism software, version 6.0C (Graphpad). In instances where only two samples were compared, significance was determined using an unpaired t test. In instances where multiple columns were compared, an ordinary one-way ANOVA was used with a Tukey's multiple-comparison post-test. All error bars represent mean ± standard error of the mean (SEM). In all cases, the statistical test used, p-values and the number of samples analysed are highlighted in the figure legend for clarity.