Age- and stress-associated C. elegans granulins impair lysosomal function and induce a compensatory HLH-30/TFEB transcriptional response

The progressive failure of protein homeostasis is a hallmark of aging and a common feature in neurodegenerative disease. As the enzymes executing the final stages of autophagy, lysosomal proteases are key contributors to the maintenance of protein homeostasis with age. We previously reported that expression of granulin peptides, the cleavage products of the neurodegenerative disease protein progranulin, enhance the accumulation and toxicity of TAR DNA binding protein 43 (TDP-43) in Caenorhabditis elegans (C. elegans). In this study we show that C. elegans granulins are produced in an age- and stress-dependent manner. Granulins localize to the endolysosomal compartment where they impair lysosomal protease expression and activity. Consequently, protein homeostasis is disrupted, promoting the nuclear translocation of the lysosomal transcription factor HLH-30/TFEB, and prompting cells to activate a compensatory transcriptional program. The three C. elegans granulin peptides exhibited distinct but overlapping functional effects in our assays, which may be due to amino acid composition that results in distinct electrostatic and hydrophobicity profiles. Our results support a model in which granulin production modulates a critical transition between the normal, physiological regulation of protease activity and the impairment of lysosomal function that can occur with age and disease.


Introduction
Aging and stress are thought to enhance neurodegenerative disease risk through the accumulation of misfolded and aggregated proteins [1][2][3]. The lysosome is the key degradative organelle within the cell [4], and therefore plays a pivotal role in the maintenance of protein homeostasis. It contains specialized enzymes, called cathepsins, which work optimally at the acidic pH in this compartment and have a crucial role in processing and degrading proteins [5]. The transcription factor EB (TFEB) controls the expression of genes involved in lysosomal biogenesis and function [6,7]. TFEB dysregulation has been associated with neurodegenerative disease [8,9] and its overexpression may help to promote the clearance of protein aggregates [10,11]. Although, genetic and functional studies have implicated lysosomal dysfunction in the pathogenesis of multiple neurodegenerative diseases [12][13][14], understanding of the molecular basis of this phenomenon remains incomplete.
The progranulin (PGRN) protein can be proteolytically cleaved to liberate multiple cysteine-rich "granulin" peptides [26]. Granulins are highly conserved, disulfide-bonded miniproteins with unknown biological function [27][28][29][30][31]. Like progranulin, granulin peptides have been shown to localize to the endolysosomal compartment [32], and can be generated through the action of cysteine proteases on progranulin [32][33][34]. Owing to the twelve cysteines and six disulfide bonds found in each cleaved granulin, these peptides adopt a stacked β-sheet configuration that is compact, structurally stable and potentially protease resistant [35]. Several lines of evidence exist that cleaved granulin peptides oppose the function of the full-length protein. While progranulin has proliferative [35,36] and anti-inflammatory [37,38] properties, granulin peptides have been shown to inhibit cell growth [35] and stimulate inflammation [38]. In addition, we have previously demonstrated a role for C. elegans granulins in selectively promoting the accumulation of TDP-43, thereby exacerbating TDP-43 toxicity and potentially contributing to the pathogenesis of disease [39]. However, the mechanism by which granulins exert this specific regulation on TDP-43 metabolism remains unknown. C. elegans provides many advantages as a model system to study granulin function, including conservation of the progranulin gene, and the many available molecular and cell biology techniques.
In this study, we further investigate the molecular mechanisms of C. elegans granulins on lysosomal function and protein homeostasis. We show that C. elegans granulins localize to the endolysosomal fraction. Granulin production increases with age and stress, and granulin expression reduces animal fitness by impairing lysosomal protease expression and activity. This prompts cells to activate a compensatory transcriptional program involving HLH-30/ TFEB nuclear translocation and up-regulation of the transcription of HLH-30/TFEB-related genes. Overall, our findings highlight granulins as critical regulators of proteolytic lysosomal function and potential drivers of neurodegenerative disease pathogenesis.

C. elegans granulins impair organismal fitness and resistance to ER stress
We have previously shown that C. elegans progranulin (pgrn-1) null mutants exhibit enhanced resistance to endoplasmic reticulum (ER) unfolded protein stress [40]. As a genetic null, pgrn-1(-) animals produce neither full-length progranulin nor cleaved granulins; therefore, absence of either the holoprotein or the cleavage fragments could be responsible for the ER stress resistance. Based on our earlier finding that granulins could exacerbate TDP-43 toxicity [39], we hypothesized that the bioactive granulins were responsible for inhibiting ER stress resistance. Hence, to isolate granulin activity, we expressed individual C. elegans granulins 1, 2 and 3 at comparable levels in a pgrn-1 null background [39]. Granulin expression in a progranulin null background completely abolished the ER stress resistance phenotype (Fig 1A). In contrast, animals over-expressing C. elegans full-length progranulin in a progranulin null background remained ER stress resistant (Fig 1B). Over-expressed full-length progranulin was not cleaved under ER stress (S1A-S1D Fig), and could promote ER stress resistance in the presence of granulin (S1E Fig). Furthermore, transgenic expression of human tau protein and TDP-43 in a progranulin null background did not abrogate ER stress resistance (S1F Fig). Taken together, these data suggest that it is the granulins, and not full-length progranulin, that specifically inhibit ER stress resistance.
Given that granulins impair ER stress resistance, we wondered if they might more broadly impact protein homeostasis. Thus, we measured endogenous levels of heat shock protein HSP-4, the nematode homolog of human BiP/Grp78 [41]. HSP-4/BiP expression is upregulated during the unfolded protein response (UPR) [42]. We found that granulin-expressing animals displayed a trend for increased basal expression of HSP-4/BiP on day 1 of adulthood, reaching significance in animals expressing granulin 2 and 3 ( Fig 1C). Therefore, in the absence of the progranulin holoprotein, granulin expression upregulates HSP-4 and this is indicative of UPR induction and perturbed protein homeostasis.
While working with the granulin-expressing lines, we noted a decrease in overall animal fitness attributable to the granulins. Granulin production significantly reduced animal viability by lowering the number of eggs that hatched and slowing the development of animals to maturity ( Fig 1D). Granulin-expressing animals that did reach adulthood were smaller in size ( Fig  1E). Short-term associative learning can be assayed in C. elegans using a positive olfactory learning paradigm [43,44]. When granulin-expressing animals were tested in this assay they underperformed compared to controls (Fig 1F), suggesting that granulin expression may result in neuronal dysfunction. These data, coupled with previous work by others on the function of progranulin [35][36][37][38], suggest that granulins impair animal fitness, resistance to stress and neuronal function, while progranulin promotes these qualities.

Granulins localize to the endolysosomal compartment
To establish the trafficking and localization of granulin peptides within a whole organism, we utilized microscopy and biochemistry techniques. First, we determined the sub-cellular localization of full-length progranulin using a translational progranulin reporter, PGRN-1::RFP, and organelle-specific markers. As expected, in cells that secrete progranulin, such as the intestine, the reporter co-localized with both a Golgi marker, mannosidase II (Fig 2A), and a lysosomal marker, lysosomal-associated membrane protein 1 (LMP-1) (Fig 2B). However, in C. elegans granulins impair organismal fitness and resistance to ER stress. (A) Wild-type (N2) and pgrn-1(-) animals with and without granulin expression were subjected to ER stress with tunicamycin (5 μg / ml). The fraction developing to L4 stage was quantified (n = 50, 3 biological replicates). (B) Wild-type (N2) and pgrn-1(-) animals with and without C. elegans progranulin over-expression (OE) were subjected to ER stress with tunicamycin (5 μg / ml). The fraction developing to L4 stage was quantified (n = 50, 3 biological replicates). (C) Total worm lysates from synchronized day 1 adult granulin-expressing animals were immunoblotted with an anti-HSP-4/BiP antibody (3 biological replicates). Anti-actin was used as a loading control. (D) Wild-type and pgrn-1(-) animals with and without granulin expression were staged as embryos. Animals were scored for development to L4 stage (n = 50, 12 biological replicates). (E) Measurement of body length at day 1 adulthood (n = 12). (F) Measurement of short-term associative learning (three biological replicates). The glutamate receptor mutant nmr-1(ak4) was used as a positive control. Throughout, error bars show mean ± SEM, one or two-way ANOVA with post-hoc Tukey multiple comparisons test. Comparisons are to wild-type unless otherwise indicated ( � P<0.05, ��� P<0.001, ���� P<0.0001, ns = not significant, wt = wild-type).
Having established that progranulin can be trafficked from one tissue type to another, we next sought to better understand the subcellular localization of granulin peptides. To do so, we developed a protocol for subcellular fractionation of C. elegans. The purity of cytosolic, ER and endolysosomal fractions was confirmed with established markers (S2 Fig). Individual granulins that were transgenically expressed also demonstrated lysosomal localization (Fig 2F-2I). Therefore, C. elegans progranulin and granulins localize to the endolysosomal compartment.

Granulins are produced in an age-and stress-dependent manner and disrupt lysosomal morphology
In C. elegans and mammals, progranulin production increases with age [45, 46] and injury [47,48]. However, the degree to which granulin peptides are liberated has not been measured. We first asked if progranulin cleavage into granulins increases with age. Using our PGRN-1::RFP translational reporter, we found that granulin production does indeed increase in an agedependent fashion (Fig 3A and S3A and S3B Fig), suggesting that either an increase in expression and cleavage of progranulin, and/or an age-associated decline in granulin turnover, contributes to granulin accumulation. Granulin cleavage also increased in response to certain physiological stressors such as starvation (Fig 3B and S3C and S3D Fig). Thus, age and stressful stimuli, such as starvation, appear to promote the cleavage of full-length progranulin into granulins.
In order to determine the subcellular compartment in which cleaved granulin peptides are produced, we performed fractionation of fed or starved animals expressing the PGRN-1::RFP reporter. In fed animals, full-length progranulin was enriched in the endolysosomal fraction with very little lower molecular weight granulin observed in any fraction ( Fig 3C). Upon starvation, the cleaved granulins increased primarily in the endolysosomal fraction, confirming that the majority of the age and stress-induced granulins are, in fact, endolysosomal ( Fig 3C). Therefore, granulin peptides are produced in vivo in the endolysosomal compartment in a stress-responsive manner.
Given that granulins impair organismal fitness, localize to the endolysosomal fraction and impair stress resistance, we next investigated their impact on lysosomal morphology. In C. elegans, coelomocytes scavenge and detoxify the pseudocoelomic cavity and therefore have a well-developed endo-lysosomal system [49]. Although we could not image coelomocyte lysosomes in granulin 1-expressing animals due to the presence of a GFP co-expression marker, we found that both loss of progranulin and expression of granulins 2 and 3 grossly deformed these organelles (Fig 3D-3G). Lysosomes lost their spherical shape, more frequently exhibited membrane protrusions and tubular extensions (Fig 3D-3G), and became smaller in size, reaching significance for pgrn-1(-) animals and pgrn-1(-); granulin 3(+) animals ( Fig 3H). Together, these data suggest that granulin peptides accumulate in endolysosomes with age and starvation, where they, as well as loss of progranulin, may disrupt lysosomal morphology.

Granulins impair lysosomal protease activity
As we observed that expressed granulins disrupt lysosomal morphology, we next assessed their effect on lysosomal function by measuring the expression level and enzymatic activity of lysosomal proteases in lysates from granulin-expressing C. elegans. Granulin expression resulted in decreased protein levels of ASP-3, the nematode ortholog of mammalian cathepsin D (CTSD), reaching significance in granulin 2-expressing animals ( Fig 4A). Expression of all  (Fig 4B). This decrease in protease expression correlated with a decrease in protease activity (Fig 4C and 4D), reaching significance in granulin 2 and 3-expressing animals for ASP-3 activity and granulin 1 and 2-expressing animals for CPR/CPL-1 activity. Overall, our data suggest that granulin peptides disrupt C. elegans lysosomal protease activity in vivo.

C. elegans granulin domains display distinct electrostatic and hydrophobicity profiles that may contribute to functional specificity
As we observed differences between the three granulins in terms of the magnitude of their phenotypic effects within assays, we sought to determine whether these differences might be explained by variations in their amino acid sequence and physicochemical properties. C. elegans granulins 1, 2 and 3 share less than 50% sequence identity among themselves (Fig 5A), and less than 40% when excluding the highly conserved network of disulfide bonds. Electrostatic analysis ( Fig 5B) shows that granulin 3, located at the C-terminus of C. elegans PGRN-1, is positively charged at neutral pH, while granulin 1, the N-terminal granulin domain, remains negatively charged at all analyzed pH values (pH = 4 to 8). The central granulin 2 domain has little to no overall net peptide charge at neutral pH. A further comparison of granulin hydrophobicity ( Fig 5C) shows that the central region of granulin 2 (residues 202 to 221) and granulin 3 (residues 309 to 326) is predominantly hydrophobic, as measured by Kyte and Doolittle (K&D) hydrophobicity scores greater than zero. In contrast, the K&D score for the corresponding region of granulin 1 (res. 120 to 139) is slightly negative. While the functions of the individual C. elegans granulin domains remain to be further elucidated, these observed differences might suggest that each domain participates in unique protein-protein interactions (PPIs), and thus differing roles in the endolysosomal system.
We further compared the C. elegans granulin sequences with those of different species, including Homo sapiens (H. sapiens), Mus musculus (M. musculus) and Danio rerio (D. rerio) ( Fig 5A). We found that C. elegans granulins share higher identity scores to certain granulins from other species than among themselves. Similar to C. elegans granulins, differences in pH-dependent electrostatics ( Fig 5B) were noticeable for all species studied, with a recurring trend for the C-terminal granulin domains being the most positively charged. The low sequence identity and distinct physicochemical properties among the granulin domains were also observed for H. sapiens, M. musculus and D. rerio, contrasting with the highly conserved network of disulfide bonds. Taken together, these data highlight the importance of the amino acid residues situated outside of the well-conserved granulin sequence consensus for contributing to the charge and hydrophobicity profiles of each granulin domain. These may drive unique recognition patterns for PPIs that may ultimately be relevant in a disease context.

Granulins activate the lysosomal CLEAR response and induce nuclear localization of HLH-30/TFEB
To determine if granulin-induced disruption of lysosomal morphology and function promoted a transcriptional response, we performed RNA-seq profiling of wild-type, pgrn-1(-) and  We first compared pgrn-1(-) or pgrn-1(-); granulin animals to wild-type animals. Compared to wildtype, a total of 7084 differentially expressed genes (DEGs) were identified across all strains (Fig 6A and S4B Fig). The majority of DEGs identified for pgrn-1(-) animals were down-regulated compared to wild-type animals. These DEGs were enriched for GO terms associated with growth, development, cation and sugar binding (S4C and S4D Fig). In contrast, the majority of DEGs for granulinexpressing animals were up-regulated compared to both wild-type and pgrn-1(-) animals ( Fig  6A and S4B Fig). GO term analysis for DEGs in granulin-expressing animals showed a shared enrichment in genes associated with lysosomal function, including protein metabolic process and hydrolase activity acting on ester bonds (S4E-S4K and S5 Figs). Expression of granulin 2 resulted in the highest number of DEGs compared to both wild-type and pgrn-1(-) animals, followed by granulin 3 and then granulin 1 (Fig 6A and S4B Fig). The observed overlap in enriched GO terms on granulin 2 and 3 expression further suggests similarities between these two granulins compared to granulin 1, and also reflects the phenotype severity observed in development and behavioral assays. Interestingly, the upregulated DEGs identified in pgrn-1 (-); granulin 3(+) animals were significantly enriched for genes whose promoters contained the putative TFEB binding site E-box sequence 5'-CACGTG-3' (P = 0.011). This trend was also observed in the upregulated DEGs for pgrn-1(-); granulin 1(+) (P = 0.149) and pgrn-1(-); granulin 2(+) (P = 0.097) but did not reach statistical significance. TFEB is the master lysosomal transcription factor that regulates lysosomal biogenesis and autophagy [6,7], and the C. elegans TFEB is HLH- 30 [50].
In response to starvation, stressful stimuli and aging, HLH-30/TFEB translocates from the cytosol to the nucleus to activate its transcriptional targets [6,7,50,51]. This program, known as the Coordinated Lysosomal Expression and Regulation (CLEAR) response induces expression of genes involved in lysosomal function and autophagy, including progranulin. We assessed HLH-30/TFEB cytoplasmic versus nuclear localization in control, pgrn-1(-) and granulin expressing animals. Granulin expression promoted nuclear localization of HLH-30/TFEB, reaching significance in granulin 3-expressing animals (Fig 6B and 6C). This effect was not seen in pgrn-1(-) animals where a much lower number of DEGs were identified, and was also not observed in pgrn-1(-) animals expressing human tau or TDP-43 protein (S6A Fig). These results suggest that the disruption of lysosomal morphology and protein homeostasis seen in granulin-expressing animals leads to a specific compensatory translocation of HLH-30/TFEB from the cytosol to the nucleus.
When granulin-expressing animals were crossed into a wildtype background, the presence of wildtype progranulin partially mitigated the negative effects of granulin-expression on development ( We speculate that ER stress may promote the cleavage of endogenous PGRN, resulting in even higher levels of cleaved granulins (endogenous and transgenic granulins) and enhanced ER propKa 3.1. A 3-color scale is used for the granulin domains, colored from red (negative) to blue (positive) using a percentile scale. (C) Kyte and Doolittle (K&D) hydrophobicity scores for granulin domains of C. elegans. The central regions of granulins 2 and 3 are highlighted with circles since most of these residues have positive K&D scores, which is suggestive of a potential hydrophobic patch.
https://doi.org/10.1371/journal.pgen.1008295.g005 stress sensitivity. These data further suggest a reciprocal relationship between full-length progranulin and cleaved granulins, and highlights that their relative levels may be important for normal animal development and fitness.
To determine if the upregulation of TFEB target genes was a compensatory transcriptional response in granulin-expressing animals, we crossed these animals into an hlh-30(-) null background. When lacking hlh-30, granulin-expressing animals had further impairments in overall fitness, with fewer growing to adulthood ( Fig 6D) and more arresting at early larval stages ( Fig  6E). Together, these data demonstrate that granulin expression, even in the absence of stress or starvation, is sufficient to activate a compensatory CLEAR response and induce expression of genes containing TFEB binding sites. Overall, the ability of granulins to 1) impair a proteotoxic stress response, 2) disrupt lysosomal morphology, 3) direct TFEB to the nucleus and 4) induce a CLEAR response indicates that granulin-dependent impairment of lysosomal function negatively impacts cellular protein homeostasis (Fig 6F).

Discussion
We have previously shown in C. elegans that expression of granulin peptides enhances TDP-43 toxicity and prevents its degradation [39]. In this study, we sought to understand the mechanism by which granulins exert their effects and determine if they more broadly impacted protein homeostasis. We found that granulins are produced in an age and stress-dependent manner, and consequently impair lysosomal protease expression and activity. Their expression negatively impacts cellular protein homeostasis and drives a compensatory lysosomal stress response in an attempt to up-regulate HLH-30/TFEB-regulated genes. These effects manifest as an overall decrease in animal fitness.
This study contributes a new dimension to our understanding of the regulation of lysosomal proteostasis via the identification of C. elegans granulins as age and stress-produced peptides that impair overall animal fitness by reducing lysosomal function. C. elegans granulins, similar to the human peptides, localize to the endolysosomal compartment [32]. Granulins are composed of evolutionarily conserved stacked beta hairpins stabilized by disulfide bonds, which are often found in natural protease inhibitors [52]. This highly compact and stable structure is thought to confer resistance to denaturation and protection against proteolytic cleavage in the lysosomal environment [53]. Indeed, a role for granulins in regulating protease maturation has previously been demonstrated in plant cysteine proteases that incorporate a granulin domain C-terminal to the catalytic domain, such as RD21 in A. thaliana [54]. In further support of granulins as regulators of protease activity, homozygous progranulin mutation carriers develop a progressive myoclonic epilepsy syndrome that phenocopies loss of function mutations in another lysosomal protease inhibitor, cystatin B [18,55]. Recent studies have shown that human full-length progranulin and individual granulin domains may physically interact with CTSD and stimulate the enzymatic activity of the protease [25,[56][57][58]. However, in the absence of full-length protein, C. elegans granulins promote a distinct phenotype of animals expressing HLH-30::GFP (scale bar = 200 μm). (C) Percentage of animals with nuclear localized HLH-30::GFP (n = 120 animals from 3 biological replicates). Wild-type, pgrn-1(-) and granulin-expressing animals with and without hlh-30 expression were staged as embryos, and animals were scored for (D) development to gravid adult (n = 50, 3 biological replicates), and (E) the number of larvae arresting at L1 and L2 stage (n = 50, 3 biological replicates). Throughout, values shown are mean ± SEM, one-way ANOVA and Tukey multiple comparisons test. Comparisons are to wild-type unless otherwise indicated ( � P<0.05, �� P<0.01, ��� P<0.001, ���� P<0.0001, ns = not significant). The~20% larval arrest observed in the pgrn-1(-) strains with wild-type hlh-30 reach significance when compared pairwise with wildtype using a Student's t-test ( � P<0.05). (F) Genetic model for progranulin and granulin function in lysosomal function, protein homeostasis and stress resistance.
Granulins likely play a normal physiological role in regulating protease expression and activity. Given their ability to promote the CLEAR program, granulins may serve as a signal for stress or impaired health that requires regulated checks on protease activity, perhaps to limit inflammation. This would be consistent with the role of progranulin in complement-mediated synaptic pruning by microglia [59]. We speculate that under conditions of progranulin haploinsufficiency, the normal balance between progranulin and granulins becomes skewed towards excessive granulins. In excess, the inhibitory effect of granulins upon protease activity impairs the function of lysosomes; with age, the natural compensatory mechanisms such as the CLEAR program become overwhelmed, resulting in cellular dysfunction. When this occurs in neurons and/or support cells such as microglia, the end result may be neurodegeneration. Because granulins increase with age, it remains possible that accumulation of granulins directly contribute to the proteostatic pressures associated with increasing age. Comprehensive measures of progranulin-to-granulin ratios with age and in progranulin mutation carriers are needed.
The lentiviral delivery of progranulin to degenerating brain regions protects against neurotoxicity and cognitive defects in mouse models of Parkinson's disease [60] and Alzheimer's disease [61]. As such, efforts to increase progranulin production in patients are underway [62][63][64][65]. However, a more recent study has suggested that progranulin delivery to brain promotes in T-cell infiltration and neuronal and glial degeneration [66]. Progranulin cleavage and granulin levels were not measured in these studies, and may account for differences in the observed results.
Progranulin is a highly conserved protein [27,29,30]. The number of granulin domains has increased through phylogeny from one in Dictyostelium discoideum and plants, three in nematodes to seven-and-a-half in humans [29,54]. It is intriguing to speculate that this expansion in cleavage fragments could lead to regulation of additional proteases. In support of this, we find that the amino acid residues situated outside of the well-conserved granulin sequence consensus contribute to distinct charge and hydrophobic profiles for each granulin domain. These unique characteristics may be important for driving specific protein-protein interactions and thus different roles in the cellular environment. Indeed, the distinct effects of granulin 2 and 3 on protein homeostasis, lysosomal function and TDP-43 toxicity [39], as compared to granulin 1, may suggest functional differences between granulins.
Our results establish age-regulated granulins as modulators of lysosomal function, and suggest that a toxic gain of granulin function, rather than or in addition to simply loss of full-length progranulin, may contribute to FTLD disease pathogenesis. This could explain why progranulin loss-of-function mutations are transmitted in an autosomal dominant fashion. The presence of granulins only in the haploinsufficiency state could explain why TDP-43 pathology is not seen in the null state [18]. Several lysosomal proteases that cleave progranulin have recently been identified [32][33][34], although how those proteases decide when and where to cleave progranulin remains unknown. This study prompts several important follow up questions regarding the rate and order in which granulins are liberated from progranulin, how pH changes impact the predicted association of granulins with lysosomal proteases and whether increased granulin impact other neurodegenerative disorders such as Alzheimer's disease. The current study also has implications for therapeutic progranulin repletion efforts, as care should be taken to determine whether replacement progranulin is processed into granulins. Finally, our findings suggest that in addition to progranulin repletion, prevention of progranulin cleavage into granulins could represent a rational therapeutic target in neurodegeneration.

Strains
C. elegans strains were cultured at 20˚C according to standard procedures [67]. Some strains were provided by the Mitani Laboratory (National Bioresource Project, Japan) at the Tokyo Women's Medical University and the Caenorhabditis Genetics Center (CGC) at the University of Minnesota. Strain descriptions are at www.wormbase.org. The N2E control strain was used as the wild-type strain. The pgrn-1(tm985) strain has a 347 bp deletion in the pgrn-1 gene resulting in a null allele [45]. The following C. elegans strains were used in this study: CF3050

Generation of transgenic C. elegans
To generate strains expressing individual granulins, each granulin was amplified separately from wild-type C. elegans progranulin cDNA as previously described [39].

ER stress assays
ER stress assays were performed as previously described [40].

Animal viability
L4 stage animals were allowed to lay eggs overnight. Fifty synchronized eggs were transferred to seeded plates. After three days, the fraction of animals that developed to the L4 stage was quantified.

Body size
L4 animals were staged, grown at 20˚C overnight and imaged the following day as day 1 adults. Animals were mounted on a 2% agarose pad with 25 mM sodium azide (Spectrum Chemical, #SO110) and imaged using a Zeiss AxioImager microscope at 10 x. Body size was measured in ImageJ software using the skeletonize function.

Short-term associative learning
Short-term associative learning assays were performed as previously described [43,44].

Immunoblotting
Sixty L4 stage animals were allowed to lay eggs overnight (~sixteen hours). Adult worms and hatched larvae were washed off the plates with M9 buffer. Eggs were collected with a cell scraper and transferred to a newly seeded plate by chunking. These eggs were allowed to develop to early L4 stage and 200 μl of 20 mM FUDR (Fisher Scientific, #AC227601000) was added to prevent development of progeny and overgrowth of plates. At each time point, animals were collected from plates with ice cold M9 and washed once to remove food.

Confocal microscopy
Animals were mounted on microscope slides with 2% agarose pads containing 30 mM levamisole hydrochloride (Fisher Scientific, #AC187870100) and imaged using a Zeiss LSM 700 laser-scanning confocal microscope using 488 nm and 561 nm lasers and 63x and 100x objectives. L1 animals were imaged 1-2 h after hatching. Z-stacks were taken every 0.7 μm. Image processing was carried out using ImageJ software. A maximum intensity projection of the z-stack for each animal was created. Images at 488 nm and 561 nm were overlaid and analyzed for co-localization.

Subcellular fractionation
Thirty L4 stage animals were picked to 60 x 10 cm plates per strain. Plates were confluent with mixed stage animals after four days growth at 20˚C. Progranulin cleavage was observed after starving animals for an additional seventy-two hours at 20˚C. A lysosomal fraction was isolated from a light mitochondrial-lysosomal fraction as previously described [68] with the following modifications. Animals were collected in 0.25 M sucrose (pH 7.2) and washed twice with 0.25 M sucrose. Lysosomes and mitochondria were separated using a discontinuous Nycodenz (Progen Biotechnik, Germany, #1002424) density gradient. Lysosomes were collected from the 19.8% / sucrose interface and the 26.3 / 19.8% interface and pooled. Lysosomes were diluted five times with 0.25 M sucrose, and pelleted at 37,000 × g for 15 minutes. Cytosolic, ER and lysosomal fractions were confirmed by immunoblotting for specific subcellular fraction markers (LAMP-1, HSC-70, calnexin).

Protease activity measurements from total worm lysates
Protease activity was measured using commercially available kits (BioVision Cathepsin D Activity Fluorometric Assay Kit, #K143-100 and BioVision Cathepsin L Activity Fluorometric Assay Kit, #K142-100). Animals were staged as for immunoblotting, but without the addition of 20 mM FUDR. At day 1 of adulthood, worms were collected from plates with ice cold M9 and washed twice to remove food. Worm pellets were resuspended in 1% NP-40 buffer (Fisher Scientific) without protease inhibitors and frozen at -80˚C overnight. Pellets were thawed and sonicated for 4 cycles of 1 min on and 2 min off (BioRuptor, Diagenode). Lysates were centrifuged for 5 minutes at 13,000 rpm at 4˚C and supernatant was transferred to a fresh tube. 0.25 μg total protein per sample was used per assay and samples from one strain were run in triplicate. Fluorescence measurements were taken every minute at 25˚C (Infinite M200, Tecan). As controls, 250 nM Pepstatin A (for pan-aspartyl protease inhibition in CTSD assay, BioVision) or 10 μM CA-074 (for Cathepsin B inhibition, EMD Millipore, #205530) and 10 μM CTSLiII (for Cathepsin L inhibition, EMD Millipore, #219426) were added to the lysate and pre-incubated for 10 minutes on the bench at room temperature. Linear regression was performed on at least 30 minutes of data to calculate the rate of enzyme activity.

Computational analysis of granulin domains
Sequences for C. elegans (Q9U362), Homo sapiens (P28799) and Mus musculus (P28798) PGRN were extracted from Uniprot (The Uniprot Consortium, 2019), while Danio rerio PGRNb (AAH96854.1) sequence was obtained from National Center for Biotechnology Information (NCBI) Protein database (https://www.ncbi.nlm.nih.gov). Amino acid multiple sequence alignment was performed using the MAFFT online service (version 7, https://mafft. cbrc.jp/alignment/server/) [69]. The EMBOSS Needle server was used for pairwise sequence alignment between C. elegans granulin 1, granulin 2 and granulin 3 and individual granulin domains from H. sapiens, M. musculus and D. rerio (https://www.ebi.ac.uk/Tools/psa/emboss_ needle/) [70]. Identification of granulin domains from the full-length sequences was based on sequence similarity to H. sapiens granulin A using the Basic Local Alignment Search Tool protein (BLASTp) server (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Granulin A (PDB ID: 2JEY.A) was used as a reference for homology modeling of all granulin domains [71] using the Prime software. Electrostatic analysis ranging from pH 4 to 8 was performed on the in silico models with the software propKa 3.1 [72]. Kyte & Doolitle (K&D) hydrophobicity scales were obtained from the ExPASy Bioinformatics Resource Portal (https://web.expasy.org) for PGRN sequence of all species here studied. For the K&D per-residue score, a window size of 5 was used, i.e. the final score for a given residue i is the sum of the scale values for i and i-2, i-1, i+1 and i+2.

RNA-sequencing analysis
Total RNA was isolated from wild-type (N2E), pgrn-1(-), pgrn-1(-); granulin 1(+), pgrn-1(-); granulin 2(+) and pgrn-1(-); granulin 3(+) expressing animals synchronized at day 1 of adulthood. Animals were collected from plates with ice cold M9 and washed three times to remove OP50 food. After harvesting, the animals were resuspended in QIAzol (Qiagen #79306) and flash frozen in liquid nitrogen. RNA was extracted and purified using a Qiagen miRNeasy kit (Qiagen #217004). Samples were extracted in quadruplicate (four biological replicates for each strain), for a total of 20 samples. Total RNA was quantified using the RiboGreen assay (Ther-moFisher, #R11490) and RNA quality was checked using an Agilent TapeStation 4200 (Agilent). RNA Integrity Numbers (eRINs) were >8 in all the samples. Libraries for RNA-seq were prepared using the Illumina TruSeq library preparation protocol (Illumina Inc), multiplexed into a single pool and sequenced using an Illumina HiSeq 4000 sequencer across 4 PE 2 x 75 lanes on a single flowcell. After demultiplexing, we obtained between 13 and 32 million reads per sample, each one 75 paired end bases long. Quality control was performed on base qualities and nucleotide composition of sequences. Alignment to the C. elegans genome (ce11) was performed using the STAR spliced read aligner [73] with default parameters. Additional QC was performed after the alignment to examine the following: level of mismatch rate, mapping rate to the whole genome, repeats, chromosomes, and key transcriptomic regions (exons, introns, UTRs, genes). Between 92 and 93% of the reads mapped uniquely to the worm genome. Total counts of read fragments aligned to candidate gene regions within the C. elegans reference gene annotation were derived using HTS-seq program and used as a basis for the quantification of gene expression. Only uniquely mapped reads were used for subsequent analyses. Following alignment and read quantification, we performed quality control using a variety of indices, including sample clustering, consistency of replicates, and average gene coverage. One sample for pgrn-1(-); granulin 1(+) was excluded from analysis as a quality control outlier. Differential expression analysis was performed using two parallel approaches, the EdgeR Bioconductor package [74], and voom [75]. Differentially expressed genes (DEGs) were selected based on False Discovery Rate (FDR, Benjamini-Hochberg adjusted p-values) estimated at � 5%. There was a large overlap between DEGs identified by edgeR and voom (edgeR: 89.0% common DEGs with voom (6307/7084), voom: 93.9% common DEGs with edgeR (6307/6714)). Clustering and overlap analyses were performed using the Bioconductor packages within the statistical environment R (www.rproject.org/). Gene Ontology annotation was performed using DAVID (david.abcc.ncifcrf.gov/) and GOrilla [76,77].

TFEB binding site analysis
The promoter regions of all differentially regulated transcripts were analyzed for the presence of the C. elegans TFEB/HLH-30 binding site E-box sequence 5'-CACGTG-3'. Enrichment of TFEB binding sites was tested by comparison to the expected distribution based on 10,000 random permutations. A permutation test was used to calculate p-values.

HLH-30/TFEB imaging
Forty L4 animals were picked, grown at 20˚C overnight and imaged the following day as day 1 adults. The nuclear localization of HLH-30::GFP was imaged using a Zeiss AxioImager microscope at 10x. Animals were imaged within 5 minutes of mounting on a 2% agarose pad with 25mM sodium azide (Spectrum Chemical, #SO110). Data from three independent experiments were pooled.