Arf6 and the 5'phosphatase of synaptojanin 1 regulate autophagy in cone photoreceptors

Abnormalities in the ability of cells to properly degrade proteins have been identified in many neurodegenerative diseases. Recent work has implicated synaptojanin 1 (SynJ1) in Alzheimer's disease and Parkinson's disease, although the role of this polyphosphoinositide phosphatase in protein degradation has not been thoroughly described. Here, we dissected in vivo the role of SynJ1 in endolysosomal trafficking in zebrafish cone photoreceptors using a SynJ1‐deficient zebrafish mutant, nrca14. We found that loss of SynJ1 leads to specific accumulation of late endosomes and autophagosomes early in photoreceptor development. An analysis of autophagic flux revealed that autophagosomes accumulate because of a defect in maturation. In addition we found an increase in vesicles that are highly enriched for PI(3)P, but negative for an early endosome marker in nrca14 cones. A mutational analysis of SynJ1 enzymatic domains found that activity of the 5'phosphatase, but not the Sac1 domain, is required to rescue both aberrant late endosomes and autophagosomes. Finally, modulating activity of the PI(4,5)P2 regulator, Arf6, rescued the disrupted trafficking pathways in nrca14 cones. Our study describes a specific role for SynJ1 in autophagosomal and endosomal trafficking and provides evidence that PI(4,5)P2 participates in autophagy in a neuronal cell type.


Introduction
Cells must balance synthesis of new proteins with the degradation of old and damaged proteins to maintain their cellular proteome. Neuronal cell types are particularly susceptible to breakdowns in protein homeostasis because they are postmitotic, long-lived cells. Many neurodegenerative diseases including Alzheimer's, Parkinson's and amyotrophic lateral sclerosis show cellular phenotypes that are consistent with abnormalities in protein turnover [1]. In the retina, degeneration of photoreceptors causes blindness [2]. Understanding the proteins and cellular processes that underlie protein turnover is vital to understanding the underlying pathology of these diseases as well as developing treatments.
Protein degradation is accomplished by two main pathways: the ubiquitin-proteasome system and the endolysosomal system. Membrane proteins are removed from the plasma membrane by endocytosis. The endocytosed proteins are delivered to early endosomes where they are sorted. Proteins destined for degradation will continue through the endolysosomal pathway, first through late endosomes and finally to the lysosome. Cytosolic proteins can be degraded either by the proteasome or by autophagy [3]. The term autophagy generally refers to macroautophagy, a process in which a double membrane structure forms in the cytoplasm, non-specifically engulfing cytoplasmic contents including proteins and organelles. In order to degrade their contents, autophagosomes must fuse with the lysosome. In yeast, autophagosomes fuse directly with the vacuole. In mammalian systems, autophagosomes can also fuse directly with the lysosome or first fuse with other endosomal compartments, forming an amphisome, before fusing with the lysosome [4].
Important regulators of the endolysosomal pathway are the phosphoinositide (PIP) lipids. The inositol head group of these phospholipids can be phosphorylated at three positions, giving rise to seven different PIP species. The differential subcellular distribution of PIPs in different membranes confers membrane identity and allows the spatial and temporal control of effector proteins to endosomal membranes [5]. The identities, levels and distributions of PIPs are tightly controlled through the actions of kinases and phosphatases.
Another PIP suggested to play a role in autophagy is phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ). PI(4,5)P 2 is found on the plasma membrane where it plays vital roles in cytoskeletal remodeling, signal transduction and endocytosis [10]. Recent studies have demonstrated that PI(4,5)P 2 is also involved in autophagy in mammalian cells, both in the initial formation of autophagosome precursors [11] as well as in autophagic lysosome reformation [12]. PI(4,5)P 2 is synthesized primarily by type 1 phosphadylinositol-4-phosphate 5-kinases (PIP5Ks). PIP5K has been demonstrated to be activated by the small GTPase Arf6 [13]. Arf6 is also able to stimulate PI(4,5)P 2 production indirectly by activating phospholipase D (PLD) that produces phosphatidic acid, a cofactor for PIP5K stimulation [14]. Arf6 has been found to play an important role in membrane trafficking between endosomes and the plasma membrane. Perturbing Arf6 activity through the expression of a constitutively active (CA) mutation results in the accumulation of PI(4,5)P 2 positive endosomes that sequester endocytic cargo [15]. Arf6 increases autophagy by regulating the delivery of membranes from the plasma membrane to autophagosome precursors linking the production of PI(4,5)P 2 at the plasma membrane to autophagic trafficking [11].
Further evidence for a role of PI(4,5)P 2 in autophagy and protein degradation comes from examining the phenotype associated with the loss of the polyphosphoinositide phosphatase synaptojanin 1 (SynJ1). SynJ1 has well characterized role in clathrin-mediated endocytosis where it dephosphorylates of PI(4,5)P 2 . Knocking out SynJ1 in neurons results in an inability to uncoat vesicles and a decrease in the pool of synaptic vesicles [16]. Loss of SynJ1 also leads to defects in protein degradative pathways in photoreceptor neurons. These defects include enlarged acidic vesicles, abnormal late endosomes, an increase in autophagosomes and an abnormal accumulation of synaptic proteins within photoreceptor cell bodies [17,18]. Alterations in SynJ1 function have been linked to Alzheimer's [19] and Parkinson's [20,21] diseases. Importantly, in an Alzheimer's mouse model, altering SynJ1 expression causes changes in delivery of amyloid beta to lysosomes [22]. Understanding the functional role of SynJ1 in protein turnover in the endolysosomal and autophagic pathways would help us understand the pathologies of these diseases.
The goal of this study was to define the specific trafficking steps altered in the SynJ1-deficient zebrafish mutant nrc a14 and to identify the functional role of SynJ1 in these pathways. To dissect the defect in the autophagic/endolysosomal pathway, we used multiple strategies that exploit the ability to image fluorescent molecules in live zebrafish. The zebrafish model is also a unique tool to characterize vesicle transport and associated diseases in the retina because of the rapid development of the eye. In zebrafish, photoreceptors develop and become functional long before SynJ1-deficient animals die [23][24][25].
We found that autophagic and late endosomal trafficking pathways are specifically altered in nrc a14 cones early in photoreceptor development, and misregulation of these pathways is not a general consequence of compromised photoreceptor function. The accumulation of autophagosomes is due to a defect in autophagosome maturation and an increase in the formation of autophagosome precursors. We also demonstrate that the 5′ phosphatase, but not Sac1, domain of SynJ1 is involved in regulating endolysosomal and autophagic trafficking in cones. Finally, we show that altering activity of the small GTPase Arf6a, which is involved in regulating endocytic membrane traffic through actions on PI(4,5)P 2 , can rescue the autophagy defects in nrc a14 cones, but that abnormal late endosomes in nrc a14 cones did not respond to modulating Arf6a activity in the same manner. Based on our data, we propose that SynJ1 negatively regulates the formation of autophagosome precursors through actions on membrane PI(4,5)P 2 .

Results
Loss of SynJ1 specifically disrupts endolysosomal trafficking early during cone photoreceptor development Cone photoreceptors from 5 days post-fertilization (dpf) nrc a14 zebrafish larvae, which lack SynJ1, have abnormal endolysosomal and autophagic trafficking [18]. At 5 dpf, cone photoreceptors are fully differentiated and functional [26]. To correlate the endolysosomal defects observed in nrc a14 cones [18] with initial stages of photoreceptor development, we examined late endosomes and autophagosomes in cone photoreceptors starting at 3 dpf. Retinal development is rapid in zebrafish; at 3 dpf, cone photoreceptors have begun to form outer segments (OSs) but have not formed fully functional synaptic connections. By 4 dpf, cone photoreceptors have formed OSs, established synaptic connections and can reliably respond to visual stimuli [26]. We analyzed fixed retinal sections of wild type (WT) and nrc a14 Tg(TαCP : GFP-map1lc3b) and Tg(TαCP : GFP-rab7) larvae ( Fig. 1A-C); these fish lines express the autophagosome marker green fluorescent protein (GFP)-LC3 or the late endosome marker GFP-Rab7, respectively, in cone photoreceptors [18].
In order to determine if an increase in autophagosomes is the primary phenotype of the nrc a14 mutation or a characteristic of dysfunctional photoreceptors, we examined the number of GFP-LC3 positive structures in 5 dpf pde6c w59 mutant larvae [29] (Fig. S1). The pde6c w59 mutation results in cone photoreceptor degeneration. In contrast to the dramatic accumulation of autophagosomes in nrc a14 cones, we observed no significant difference in the number of LC3 positive puncta between WT and pde6c w59 mutant cones (0.99 ± 0.08 vs. 0.77 ± 0.13 LC3 puncta/cell, Fig. S1B). This indicates that an accumulation of autophagosomes is not an indirect effect of the nrc a14 mutation on cone function.
We further investigated whether the defects in nrc a14 cones are specific to the late stages of endolysosomal trafficking, we examined the subcellular distribution of early endosomes in WT and nrc a14 cones. We generated the line Tg(TαCP : GFP-rab5a) that expresses the early endosome marker GFP-Rab5a in cones. We observed no apparent difference in the subcellular distribution of Rab5a positive early endosomes between WT and nrc a14 cones at 5 dpf (Fig. 1D). There was a slight, but not significant decrease in the number of Rab5a positive structures in nrc a14 cones (2.94 ± 0.24 vs. 2.16 ± 0.10 Rab5a puncta/cell, Fig. 1E). These data show that the trafficking defects in nrc a14 cones primarily affect the late endolysosomal pathway and are present at very early stages of cone development.

Maturation of autophagosomes is disrupted in cones lacking SynJ1
Autophagy is a dynamic process; the increased GFP-LC3 puncta in nrc a14 cones could result from increased formation of new autophagosomes, decreased autophagosome maturation and turnover, or a combination thereof. To examine autophagic flux, we generated the fish line Tg(TαCP : mCherry-GFP-map1lc3b) that expresses the tandem construct mCherry-GFP-LC3 in cones. LC3 protein accumulates in autophagosome membranes, resulting in the presence of punctate structures exhibiting both mCherry and GFP fluorescence. In order to degrade their contents, autophagosomes fuse with the acidic endosomes and finally the lysosome, becoming an autolysosome. The autolysosomes have primarily mCherry fluorescence because of quenching and loss of the pH sensitive GFP signal in the tandem construct [30]. In 5 dpf WT cones expressing mCherry-GFP-LC3, we observed GFP signal that was primarily cytosolic with a few punctate structures, similar to what we have previously observed in the Tg(TαCP : GFP-map1lc3b) fish line ( Fig. 2A & B, top panels, white arrowheads). The mCherry signal concentrated in punctate structures in the inner segments (ISs) of WT cones ( Fig. 2A & B, top panels, magenta arrowheads). As expected, nrc a14 cones expressing mCherry-GFP-LC3 displayed more GFP positive autophagosomes than WT cones (1.75 ± 0.32 vs. 1.19 ± 0.15 GFP+, mCherry+ puncta/20 μm 3 , Fig. 2C). Although we observed the presence of autolysosomes in nrc a14 cone ISs, the number of autolysosomes compared with WT was decreased (1.02 ± 0.17 vs. 1.50 ± 0.16 GFPÀ, mCherry+ puncta/20 μm 3 , Fig. 2C). We have previously reported that abnormal acidic compartments accumulate in nrc a14 cones, suggesting that the decrease in autolysosomes, we observe here is not due to defects in acidification of endolysosomal compartments in nrc a14 cones [18]. Together, these results indicate a defect, but not complete block, in the ability of autophagosomes to fuse with acidic compartments in nrc a14 cones. An accumulation of autophagosomes and defect in autophagosome-lysosome fusion could occur because of the inability of autophagosomes to move throughout the cell. To examine the dynamics of LC3 positive autophagosomes in WT and nrc a14 cones, we used live time-lapse confocal microscopy. We found that LC3-labelled autophagosomes were highly mobile in both WT and nrc a14 cones ( Fig. 3 and Supplementary Movies). The movements of autophagosomes in WT and nrc a14 cones were qualitatively similar. Vesicles traversed between the IS and synapse and also displayed short range movements within the IS or synapse. We did not detect autophagosome transport within the OS. We found that there was no difference in the average speed of LC3 positive vesicles (Fig. 3C) or the relative amount of time they spent moving (Fig. 3D). We did observe a change in the directionality of LC3 positive vesicles between WT and nrc a14 cones, whereas vesicles in WT cells often stopped or disappeared after moving towards the ISs   Bioessays 38: S119-S135, © 2016 The Authors. Bioessays published by WILEY Periodicals, Inc. S124 the block in autophagosome maturation observed in nrc a14 cones does not occur because of a loss of autophagosome mobility.

PI(3)P positive vesicles lacking endosome markers accumulate in cones lacking SynJ1
To further dissect the vesicle trafficking defects in nrc a14 cones, we analyzed the subcellular distribution of PI(3)P and PI(3,5)P 2 . These two PIPs are found throughout the endolysosomal system and serve as markers of different endocytic compartments. We generated the transgenic fish lines Tg(TαCP : YFP-2XFYVE) and Tg(TαCP : mCherry-ML1NX2) that express a PI(3)P probe [7] and PI(3,5)P 2 probe [31] respectively under the cone specific TαCP promoter. In metazoan cells, PI(3)P is found primarily in the membranes of early endosomes and autophagosomes [7,32]. In WT cones, the PI(3)P probe localized to small punctate structures at the synapses and larger punctate structures in the IS ( Fig. 4A left panel). In nrc a14 cones, we observed a similar subcellular distribution of PI(3)P containing structures although the number of puncta, particularly at synaptic terminals, was increased ( Fig. 4A right panel).
We further characterized the subcellular distribution of PI(3)P in WT and nrc a14 cone photoreceptors by analyzing the colocalization of this PIP with our markers for autophagosomes, and early and late endosomes. We generated the fish line Tg(TαCP : tRFP-t-2XFYVE) that expresses a red-fluorescent protein-tagged version of the PI(3)P marker shown in Fig. 4A. We quantified colocalization as percent pixels of one probe that overlapped with the second probe after background intensity corrections (refer to Materials and Methods section for description of analysis) and found that the relative colocalization of the PI(3)P probe with LC3-positive autophagosomes, early endosomes and late endosomes in zebrafish photoreceptors agreed with the distribution of this probe in mammalian cells (Fig. 4F) [7]. We first looked at the colocalization between PI(3)P and our autophagosome marker, LC3 ( Fig. 4B top panels). The synthesis of PI(3)P occurs early in autophagosome biogenesis, whereas lipidation and concentration of LC3 in the membrane of autophagosomes occur at later stages [33]. The PI(3)P marker 2XFYVE has been found to colocalize with LC3 in Drosophila melanogaster [34] but to varying degrees in mammalian cells [35]. In WT zebrafish cones, we found that very few LC3 positive puncta were also positive for PI(3)P (Fig. 4B, green arrowheads). Interestingly, puncta that were positive for both PI(3)P and LC3 were often found at synaptic terminals (Fig. 4B, white arrowheads), whereas PI(3)P-negative, LC3-positive autophagosomes were more often found in the ISs where lysosomes reside (Fig. 4B, green arrowheads). These data suggest that the amount of PI(3)P signal on LC3-positive autophagosomes may reflect their maturation status. In nrc a14 cones, we observe a similar colocalization pattern of LC3 and PI(3)P (11.9 ± 1.3% colocalization of LC3 with 2XFYVE in WT vs. 15.3 ± 1.4% in nrc a14, Fig. 4E); most of the LC3-positive puncta were negative for PI(3)P (Fig. 4B, green arrowheads). Most of PI(3)P puncta in nrc a14 cones were LC3-negative similar to WT (4.2 ± 0.3% colocalization of 2XFYVE with LC3 in WT vs. 4.2 ± 0.2% in nrc a14 , Fig. 4B & F), indicating that these vesicles either represent structures in the endocytic system or immature autophagosomes that have not yet acquired LC3.
Next, we investigated colocalization between PI(3)P and our early endosome marker, Rab5a. As expected, PI(3)P is found on Rab5a early endosomes in both WT and nrc a14 cones (Fig. 4C, examples shown by white arrowheads). In nrc a14 cones, we observed that many of the PI(3)P positive puncta were negative for Rab5a compared with WT cones (Fig. 4C, examples shown by magenta arrowheads). We found a decrease in the percent colocalization of the 2XFYVE marker with Rab5a in nrc a14 cones compared with WT (9.9 ± 1.9% colocalization of 2XFYVE with Rab5a in WT vs. 6.9 ± 1.0% in nrc a14 , Fig. 4C & F) suggesting that the additional PI(3)P positive puncta are not early endosomes. In the endocytic pathway, PI(3)P is found primarily on early endosomes but is also found on the luminal vesicles of multivesicular bodies [7] and localizes to Rab7 positive membranes [36]. In both WT and nrc a14 cones, we observed that some Rab7 positive structures are positive for PI(3)P (Fig. 4D, white arrowheads). Overall, the same proportion of 2XFYVE structures was positive for Rab7 in WT and nrc a14 cells (18.6 ± 1.5% colocalization of 2XFYVE with Rab7 in WT vs. 22.0 ± 2.0% in nrc a14 , Fig. 4 D & F) consistent with an increase in the overall increase in Rab7 vesicles observed in nrc a14 cones. These results indicate that many of the additional PI(3)P puncta observed in nrc a14 cones are not early or late endosomes, suggesting that they likely represent early autophagosomal structures.
PI(3,5)P 2 is found on late endosomes and lysosomes and has been found to colocalize with markers for both immature and mature autophagosomes [6,8,37]. In WT photoreceptors, the PI(3,5)P 2 probe localized to punctate structures in the IS and showed diffuse cytoplasmic staining (Fig. 5A, left panel). In nrc a14 photoreceptors, we found a very similar subcellular distribution; the probe localized to punctate structures and also showed diffuse signal in the cytoplasm (Fig. 5A, right panel). We looked at the colocalization of PI(3,5)P 2 with autophagosomes and late endosomes. In WT cones, we rarely observed overlap of PI(3,5)P 2 and LC3 (Fig. 5B). In nrc a14 cones, LC3 and PI(3,5)P 2 colocalized more often (Fig. 5B, bottom panels, white arrowhead). These structures are most likely mature autophagosomes that have not fused with acidic compartments to quench the GFP-LC3 signal, consistent with the observed decrease in autolysosomes (Fig. 2C). We found that PI(3,5)P 2 positive structures colocalized with punctate Rab7 positive structures in both WT and nrc a14 cones (Fig. 5C, white arrowheads). We did not observe the accumulation of PI(3,5)P 2 on enlarged Rab7 structures in nrc a14 cones, indicating that an accumulation of PI(3,5)P 2 is not causing the formation of these abnormal structures.
The distributions of PI(4,5)P 2 and PI(4)P are not altered in the absence of SynJ1 We analyzed the subcellular distribution of the canonical vertebrate SynJ1 substrates, PI(4,5)P 2 and PI(4)P, to determine if a change in distribution of these two PIPs could account for the trafficking defects in nrc a14 photoreceptors. We transiently expressed the genetically encoded probes PLCδ-PH [38] and FAPP1-PH [39] to detect PI(4,5)P 2 and PI(4)P, respectively. In WT photoreceptors, PI(4,5)P 2 was localized throughout the plasma membrane and concentrated at synaptic terminals ( Fig. 6A & B). Interestingly, although we observed the PLCδ- PH signal extending up until the ends of the calycal processes at the apical end of the IS, we did not detect PI(4,5)P 2 in photoreceptor OSs (n > 100 cells). Between 3 and 6 dpf, the subcellular distribution of PI(4,5)P 2 in WT and nrc a14 photoreceptors appeared the same (Fig. 6B). We did not observe an accumulation of PI(4,5)P 2 on intracellular structures that could be attributed to abnormal late endosomes or autophagosomes in nrc a14 photoreceptors. PI(4)P is found predominantly on the Golgi apparatus [40]. We transiently expressed the PI(4)P probe, RFP-FAPP1, in cones of WT and nrc a14 larvae expressing a Golgi marker. This line, Tg(crx : man2a-GFP), expresses GFP fused to the medial Golgi targeting sequence from mannosidase2a (Man2a) in photoreceptors and retinal bipolar cells [18]. Although we observed variability in the distribution of the RFP-FAPP1 PI(4)P probe in transient expression experiments, we did not observe consistent differences in PI(4)P subcellular distribution between WT and nrc a14 photoreceptors (Fig. 6C). The probe localized to the Golgi and partially overlapped with the medial Golgi marker in both WT and nrc a14 photoreceptor cells (Fig. 6C). Thus, misaccumulation of PI(4)P or PI(4,5)P 2 on endosomal structures within cones does not explain SynJ1 trafficking defects.
The activity of the 5′phosphatase domain, but not Sac1 domain, of SynJ1 is required for proper autophagic and endolysosomal trafficking.
Synaptojanin 1 contains two enzymatic PIP phosphatase domains. To define the enzymatic activity of SynJ1 involved in regulating the endolysosomal and autophagic systems in photoreceptors, we performed rescue experiments in nrc a14 cones with WT SynJ1 or SynJ1-containing mutations in the two conserved phosphatase domains. The 5′phosphatase domain has specific activity towards 5′phosphates, and the less-specific Sac1 domain has both 3′-phosphatase and 4′-phosphatase activities [41]. We transiently expressed FLAGtagged zebrafish SynJ1 constructs (SJCs) in cones of nrc a14 Tg(TαCP : GFP-map1lc3b) and Tg(TαCP : GFP-rab7) larvae. Injection of DNA constructs into zebrafish larvae results in mosaic expression [42], allowing us to examine both SJC positive and negative cones within the same retina. Rescue of the autophagosome and late endosome phenotypes by the SJCs in nrc a14 cones was assessed qualitatively by examining the morphology of Rab7 vesicles and quantitatively by counting the total number of LC3 puncta or synaptic Rab7 puncta per cell. The number of puncta in the SJC-expressing cells was normalized to the number of puncta in the non-SJC expressing cells in the same larvae.
We observed that WT and mutant SynJ1 localized to the plasma membrane and were concentrated at synaptic terminals ( Fig. 7 and Fig. S2). We did not observe that any of the SJCs localized to intracellular structures. In addition, expression of WT and mutant SynJ1 did not affect the appearance or number of autophagosomes or late endosomes in WT cones (Fig. S2).
Expression of WT SynJ1 in nrc a14 cones significantly rescued the number of LC3 puncta (64 ± 4% reduction in LC3 puncta, Figure 6. PI(4,5)P 2 and PI(4)P distributions are not altered in nrc a14 cones. A and B: The PI(4,5)P 2 probe GFP-PLCδ-PH was expressed in photoreceptors using the crx promoter. A: In wild type (WT) cells at 5 days post-fertilization (dpf), this probe localized to the plasma membrane and concentrated at synapses, but did not extend above the mitochondria (magenta) into the outer segment (OS). C: The PLCδ-PH probe showed the same distribution in nrc a14 photoreceptors from 3 to 6 dpf. PI(4)P was visualized with the probe RFP-FAPP1-PH. In both WT and nrc a14 photoreceptors at 5 dpf, the RFP-FAPP1-PH probe (magenta) overlapped with the medial Golgi marker Man2a-GFP (green). There was no difference in the subcellular distribution of PI(4)P between WT and nrc a14 photoreceptors. Scale bar = 2 μm in all images. IS, inner segment.  , synaptic Rab7 puncta (57 ± 6% reduction in synaptic Rab7 puncta, Fig. 7B & D) and the abnormally enlarged Rab7 structures in nrc a14 ISs. Late endosomes appeared punctate in nrc a14 cones expressing the WT SJC, similar to the Rab7 structures observed in WT cones (Fig. 7B). Further, we observed significant rescue of both aberrant vesicles in ISs and synaptic morphology by electron microscopy (EM) analysis of nrc a14 mutant cones stably expressing a GFP-SynJ1 transgene (Fig. S3). Next, we determined whether SynJ1 with a point mutation in the Sac1 catalytic domain could rescue the endolysosomal defects in nrc a14 cones. The Sac1 domain of zebrafish SynJ1 contains a conserved CX 5 R(T/S) catalytic motif [43]. Mutating the conserved cysteine to serine results in a catalytically dead Sac1 domain in yeast [44], mouse [45] and human [20] synaptojanin homologues. We generated the homologous C385S mutation in the Sac1 domain of zebrafish SynJ1. We found that the Sac1 mutant SJC was able to significantly rescue the total number of LC3 puncta (42 ± 5% reduction in LC3 puncta, Fig. 7C). The degree of rescue compared with WT SynJ1 was slightly reduced; however, this difference was not statistically significant. The Sac1 mutant also rescued the number of synaptic Rab7 puncta (56 ± 7% reduction in synaptic Rab7 puncta, Fig. 7D) in nrc a14 cones, as well as the appearance of abnormal Rab7 structures in cone ISs (Fig. 7A & B (SynJ1)constructs (SJCs) in cones. Wild type SynJ1 was able to significantly rescue both the late endosome and autophagosome abnormalities in nrc a14 cones. SynJ1 lacking SacI catalytic activity (C385S) also significantly rescued the late endosome and autophagosome defects. In contrast, the D732A construct was not able to rescue either the late endosome or autophagosome phenotypes. Arrows point to SJCexpressing cells (magenta in merge). Scale bar = 2 μm in all images. Results were quantified by counting the number of total C: LC3 puncta or D: Rab7 synaptic puncta in the SJC-expressing cells and normalized to the number of puncta in the non-SJC expressing cells in the same nrc a14 larvae. Graph in C shows average normalized LC3 puncta per cell, and graph in D shows average normalized synaptic Rab7 puncta per cell. Error bars are SEM and n ≥ 50 cells for all conditions (exact n shown on graph). The rescue of both the total LC3 puncta and Rab7 synaptic puncta was significant for the wild type and C385S SJCs, but not for the D732A SJC ( *** p-value < 0.001 by one-way analysis of variance followed by Dunnett's multiple comparison correction).
A. A. George et al.

Primary Research ■
Bioessays 38: S119-S135, © 2016 The Authors. Bioessays published by WILEY Periodicals, Inc. S128 middle panels). This observation indicates that the Sac1 phosphatase activity of SynJ1 is not required for proper endolysosomal and autophagic trafficking in cones. This rules out a direct role of SynJ1 in regulating PI(4)P or PI(3)P in endolysosomal and autophagic trafficking.
Finally, we generated a catalytically dead D732A mutation in the 5′phosphatase domain of zebrafish SynJ1. The conserved motif XWXGDXN(F/Y)R is found in many 5′phosphatases [46] and mutating the aspartate to alanine results in the loss of 5′ phosphatase activity of yeast [47], worm [48] and mouse [45] synaptojanin. In contrast to the Sac1 C385S point mutation, the D732A mutation was not able to rescue either the LC3 (Fig. 7C) or Rab7 (Fig. 7D) phenotype in nrc a14 cones. The nrc a14 cones expressing the D732A SJC still exhibited an accumulation of GFP-LC3 puncta as well as abnormal Rab7 structures (Fig. 7A &  B bottom panels). This result indicates that SynJ1 is required to regulate a PIP species with a 5′phosphate, and that defects in this process are responsible to the observed late endosome and autophagosome abnormalities observed in nrc a14 cones.
The PI(4,5)P 2 regulator, Arf6, affects autophagy in cone photoreceptors and rescues trafficking defects caused by loss of SynJ1 Our mutational analysis of SynJ1 has indicated that SynJ1 regulates a PIP with a 5′phosphate. Although our earlier experiments did not detect an obvious change in the subcellular distribution of PI(4,5)P 2 , this PIP is the canonical substrate of SynJ1 in vertebrate cells and is a likely candidate to explain the abnormalities we observe in nrc a14 cones. In order to determine if alterations in the level of PI(4,5)P 2 are involved in the trafficking defects observed in nrc a14 cones, we altered the activity of the small GTPase Arf6. Arf6 is involved in controlling PI(4,5)P 2 levels by activating PIP5K at the plasma membrane [13,15]. Additionally, this protein has been implicated in the formation of pre-autophagosomal membranes through its effects on PI(4,5)P 2 [11]. This small GTPase also has characterized mutations that result in CA or dominant negative (DN) activity.
Similar to our SynJ1 rescue experiments, we cloned the zebrafish Arf6 homologue, Arf6a, from zebrafish cDNA. We then generated CA and DN constructs, transiently expressed these constructs in cones of nrc a14 Tg(TαCP : GFP-map1lc3b) and Tg(TαCP : GFP-rab7) larvae and assessed the number of LC3 or Rab7 positive structures. When expressed in cones, the localization of the Arf6a constructs agreed with published data [14,15,49,50].
Previous studies have shown that the small GTPase Arf6 is involved in the formation of autophagosome precursors through both PIP5K and PLD [11]. We found that expressing WT Arf6a had no significant effect on the number of LC3positive autophagosomes in either WT or nrc a14 cones (Fig. 8). Expression of a GTP-hydrolysis defective, CA mutant Arf6a-Q67L, has previously been reported to decrease the number of pre-autophagosomal structures and autophagosomes in cell culture, presumably by sequestering PI(4,5)P 2 in endosomal structures [11]. When expressed in WT cones, the Arf6a Q67L mutant had no effect on the number of LC3 positive autophagosomes (Fig. 8A & C). In nrc a14 cones, expression of Arf6a Q67L resulted in a significant decrease in the number of autophagosomes (38 ± 7% reduction in LC3 puncta, Fig. 8B & C). We found that expression of a GDP-bound DN Arf6 construct, Arf6 T44N, was able to significantly reduce the number of autophagosomes in both WT (33 ± 9% reduction) and nrc a14 (29 ± 6% reduction) cones (Fig. 8C). Finally, we tested whether an Arf6a mutant with defective PLD binding, N48R, would affect the autophagy phenotype in nrc a14 mutants. The Arf6 N48R mutation has been shown to decrease the formation of LC3-II in Bafilomycin-treated cells [11]. We found that Arf6a N48R had no effect on autophagy in WT cones, but significantly rescued the number of LC3 puncta in nrc a14 cones (36 ± 6% reduction). These results indicate that Arf6a and SynJ1 are acting in the same pathway to modulate autophagy in cone photoreceptors. As Arf6 has been shown to positively regulate levels of the canonical SynJ1 substrate PI(4,5)P 2 through actions on PIP5K and PLD, it is likely that both Arf6 and SynJ1 are acting on autophagy through this lipid.
We also examined the effects of Arf6 on the late endosome phenotype in nrc a14 mutants (Fig. 9). Arf6 has been shown to play a role in sorting and maturation steps of late endosomes [51]. We found that expression of WT Arf6 had no effect on the number of synaptic Rab7 vesicles in either WT or nrc a14 cones. Expression of the Arf6 Q67L mutant resulted in a significant increase in the number of Rab7 positive vesicles at the synapses of WT cones (65 ± 22% increase in synaptic Rab7 puncta, Fig. 9C). Expression of Arf6 Q67L had the opposite effect in nrc a14 cones; it resulted in a significant decrease in the number of synaptic Rab7 positive vesicles (35 ± 7% reduction, Fig. 9C). To our knowledge, the effect of the Arf6 Q67L of Rab7 positive vesicles has not been studied. It has been reported that expression of Arf6 Q67L results in the entrapment of cargo in Arf6-Q67L positive endosomal structures [15,52]. It is possible that the presence of this trapped cargo results in an up-regulation of the degradative pathway, resulting in the presence of an increased number of Rab7 positive compartments in WT cones and possibly ameliorating the protein degradation block observed in nrc a14 cones. Further work will need to be carried out to test this idea and understand the effects of this Arf6 construct on late endosomes. Despite significantly rescuing the autophagy phenotype, expression of the dominant negative mutants Arf6 T44N or Arf6 N48R had no significant effects on the number of synaptic Rab7 vesicles in either WT or nrc a14 cones (Fig. 9).
These Arf6 experiments revealed that Arf6 and SynJ1 act in the same pathway to modulate autophagy, implicating PI(4,5)P 2 in this process. Additionally, we observed that the dominant negative T44N and N48R mutations rescue the autophagy phenotype, but not the Rab7 phenotype suggesting that the two phenotypes are distinct and not caused by one another. However, caution needs to be taken in interpreting these results, as expression levels of DN constructs as well as the sensitivity of the LC3 and Rab7 assays may differ.
Together, our data show that SynJ1 is acting on a PIP species with a 5′phosphate to cause changes in both the formation of PI(3)P positive structures, which may be pre-autophagosomes, as well as the maturation of autophagosomes. Because altering the activity Arf6a is able to rescue the nrc a14 autophagy phenotype, we propose that the most likely identity of the D-5 SynJ1 substrate is PI(4,5)P 2 . We have established a specific role for SynJ1 in endolysosomal trafficking, defined the functional activity required to regulate this process and implicated

Discussion
We have presented evidence that SynJ1 plays a novel role in autophagic and endolysosomal pathways. Zebrafish cones lacking SynJ1 display dramatic accumulations of autophagosomes and abnormal late endosomes. These abnormalities appear early in photoreceptor development, prior to the onset of synaptic activity. In addition, we have shown that these defects are specific to protein degradation pathways and not a general consequence of altered photoreceptor function. We found evidence for both a defect in maturation as well as an increase in formation of autophagosomes. The proper regulation of both the autophagic and late endosomal phenotypes requires activity of the 5′phosphatase domain of SynJ1. Finally, we performed rescue experiments in nrc a14 cones with the PI(4,5)P 2 regulator Arf6 and found that modulation of Arf6 activity could rescue the accumulation of autophagosomes by inhibiting their formation. These findings support the idea that SynJ1 negatively regulates the formation of autophagosome precursors through actions on membrane PI(4,5)P 2 .
Synaptojanin 1 is well known for its key role in neurons in clathrin-mediated endocytosis both pre-synaptically and postsynaptically [16,53]. However, recent studies have also uncovered roles for this PIP phosphatase in cellular functions other than clathrin-mediated endocytosis. In an Alzheimer mouse model, SynJ1 knockdown resulted in accelerated delivery of amyloid beta to lysosomes [22]. Further, our recent work shows that in photoreceptors, SynJ1 plays a role in vesicle trafficking distinct from synaptic vesicle recycling and is specifically involved in the autophagic and endolysosomal pathways [18]. In addition, the involvement of SynJ1 in protein degradation pathways could help explain the pathology observed in a recently identified human synj1 mutation in a young patient with a taupathy [54] as well as the involvement of SynJ1 in Alzheimer's and Parkinson's disease [19][20][21]55]. The study we present here represents a detailed mechanistic analysis of SynJ1 function in protein degradation trafficking pathways.
In order to link the enzymatic activity of SynJ1 and autophagic and late endosomal trafficking pathways, we conducted a mutational analysis of the two SynJ1 phosphatase domains. The 5′phosphatase specifically removes phosphates from the 5′ position of the inositol ring. The Sac1 domain of SynJ1 is less specific and acts on phosphates at both the 3′ and 4′ positions on the inositol ring [41,45]. Previous studies examining the roles of the different SynJ1 domains have focused on the contribution of these domains in synaptic vesicle recycling. While both enzymatic domains play roles in this process, loss of 5′ phosphatase activity had more severe consequences than loss of Sac1-like activity and caused impaired endocytosis defects over a broader range of stimuli [45,48]. Further, recent studies examining patients with early onset Parkinson's disease have discovered linked mutations in the Sac1 domain of SynJ1, and thus, prolonged loss of this activity appears to also have dire effects [20,21]. Our work indicates that loss of the 5′phosphatase activity of SynJ1 is more detrimental than the loss of Sac1 activity in endolysosomal trafficking in cone photoreceptors. We found that constructs lacking 5′phosphatase activity could not rescue the accumulation of late endosomes and autophagosomes in nrc a14 cones; however, constructs lacking activity of the Sac1 domain could significantly rescue these phenotypes. Our data revealed a slight (but not significant) impairment in the ability of SynJ1-lacking Sac1 activity to rescue the accumulation of autophagosomes relative to WT SynJ1. Overall, our results indicate that a PIP with a 5′phosphate is likely involved in the pathology characterized in this study.
The plasma membrane has been found to be a source of membranes for the formation autophagosome precursors that are positive for ATG16L, a protein involved in the initial steps of autophagosome biogenesis, but negative for LC3 [56]. Increasing PI(4,5)P 2 on the plasma membrane leads to an increase in both LC3 negative autophagosome precursors as well as LC3 positive autophagosomes [11]. As PI(4,5)P 2 is a known substrate of SynJ1, it is possible that an increase in PI(4,5)P 2 in nrc a14 cones photoreceptors could be the cause of the defects in autophagy. We did not detect any changes in PI(4,5)P 2 using the PLC-δ-PH probe; however, changes in levels of PI(4,5)P 2 are difficult to quantify with this probe. As an alternate strategy to determine if PI(4,5)P 2 is involved in the disrupted trafficking pathways in nrc a14 cones, we altered the activity of the small GTPase Arf6. This small GTPase has been shown to play numerous roles in endocytic trafficking [15,49,57] and has recently been linked to the formation of autophagosomes through its actions on PI(4,5)P 2 levels [11].
We found that altering Arf6 activity significantly rescued the autophagy defects observed in nrc a14 cones. Overexpression of the constitutively active Arf6 mutant Q67L results in the formation of endosomal structures that accumulate PI(4,5)P 2 and sequester endocytic cargo [15]. The Arf6 Q67L mutant also decreases the number of autophagosomes in both basal and starvation conditions on cell culture [11]. We observed a similar decrease in the number of autophagosomes in nrc a14 cones expressing Arf6a Q67L. Knocking down Arf6 protein or expression of a PLD-binding deficient mutant has also been shown to decrease autophagy [11]. We overexpressed a GDP-locked Arf6a mutant, T44N, as well as a PLD-binding deficient mutant Arf6a, N48R, and observed rescue of the nrc a14 autophagy phenotype.
These results indicate that the increased accumulation of autophagosomes in nrc a14 cones results from a pathway that can be altered by both Arf6 and SynJ1, implicating PI(4,5)P 2 . Further support of a model in which SynJ1 negatively regulates the delivery of endocytic membranes to forming autophagosomes comes from a recent study of the role of endophilin in clathrin-independent endocytosis [58]. They observed that overexpression of synaptojanin inhibited recruitment of endophilin to the plasma membrane, whereas knockdown of synaptojanin had the opposite effect. They concluded that synaptojanin negatively regulates this endocytic pathway by dephosphorylating plasma membrane PI(4,5)P 2 .
We also investigated the role of Arf6a activity in the nrc a14 late endosome phenotype. The role of Arf6 in late endosome trafficking is less well characterized; however, recent studies have linked the actions of Arf6 to sorting [51] and acidification [59] at late endosomes. We found that only the Q67L mutation had any effect on late endosomes. Expression of Arf6a Q67L in nrc a14 cones rescued the accumulation of Rab7 positive late A. A. George et al.
Primary Research ■ endosomes at synaptic terminals. Interestingly, this mutation had the opposite effect in WT cells and resulted in a significant increase in the number of Rab7 positive structures at WT cone synapses. The observation the DN Arf6 constructs rescued the autophagosome but not the late endosome; phenotype in nrc a14 cones supports the idea that the accumulation of autophagosomes is not directly caused by defects in late endosome function. This observation along with the increased number of PI(3)P structures in nrc a14 cones strongly suggest that there is an increase in the formation of autophagosomes in the absence of SynJ1. Definitive proof of this awaits the development of reagents that recognizes pre-autophagosomal structures in zebrafish [60].
The role of SynJ1 in the proper regulation of late endosome dynamics is less clear. Our SynJ1 rescue experiments indicate that the 5′phosphatase activity of SynJ1, and therefore a PIP with a D-5 phosphate, is involved this process. However, we found that modulating activity of Arf6a did not have the same effects on the late endosome phenotype as the autophagy phenotype suggesting that plasma membrane PI(4,5)P 2 levels are not regulating these two processes in the same manner. It is also possible that the late endosome phenotype is less sensitive to alterations in Arf6a activity, and significant changes could not be detected using our methods. In either case, the origin of the abnormal Rab7 structures is unknown and could arise from dysfunction in either the proper formation, fusion, maturation or turnover of late endosomes. PI(4,5)P 2 has been demonstrated to positively regulate homotypic vacuole fusion in yeast [61], suggesting that an increase in PI(5,4)P 2 caused by SynJ1 deficiency could cause the enlarged structures we observe in nrc a14 cones. Further work will be required to dissect the role of SynJ1 in late endosome dynamics.
Together, we present a model in which SynJ1 negatively regulates the formation of autophagosomes by decreasing plasma membrane PI(4,5)P 2 . Our study has defined a novel role for SynJ1 in a degradative pathway in cone photoreceptors and provides a framework for understanding the contribution of SynJ1 to neurodegenerative disease pathology.

Ethics statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol, 3113-01, was approved by IACUC of the University of Washington.

Cloning and plasmids
The tandem mCherry-GFP-LC3 clone was created by inserting mCherry with a C-terminal linker and EcoRI cut site into a pDONR221 vector. GFP-tagged zebrafish LC3 [62] was then amplified with an N-terminal EcoRI site and inserted into the pME mCherry vector. The PLCδ pleckstrin homology domain was purchased from Addgene (Cambridge, MA, USA) (plasmid #21262). The PLCδ PH and FAPP1 PH (Tim Levine, [39]) were PCR amplified and inserted into Gateway pDONRP2R-P3 vectors using standard Gateway cloning protocols (Invitrogen, Carlsbad, CA, USA). The PI(3)P probe 2XFYVE has previously been characterized [7]. The YFP-2XFYVE (Harold Stenmark) clone was created by inserting YFP with a C-terminal linker and EcoRI cut site into a pDONR221 vector. 2XFYVE was then amplified with an N-terminal EcoRI site and inserted into the pME YFP vector. The mCherry-ML1NX2 (Haoxing Xu) clone has been previously characterized [31] and was inserted into a pDONR221 vector. The full-length zebrafish synJ1 gene cDNA has previously been cloned by our lab [17]. An N-terminal FLAG tag was added using overhang polymerase chain reaction (PCR) methods and the fulllength clone, FLAG-tagged clone, was inserted into a pCR8/GW Gateway vector (Invitrogen). The C385S and D732A mutations were created using standard site-directed mutagenesis (primer sequences in Table S1). Full-length zebrafish arf6a was cloned from larval zebrafish cDNA. Constitutively, active and dominant negative mutations were created using standard site-directed mutagenesis (primer sequences in Table S1). Expression constructs and other fluorescent protein fusions were generated using the MultiSite Gateway system (Invitrogen) and the Tol2 kit. Expression was driven by the cone transducin alpha promoter (TαCP) [63] or the cone-rod homeobox promoter (crx) [64].

Fish husbandry and generation of transgenic zebrafish
Zebrafish were reared and maintained in the University of Washington fish facility as previously described [65]. Embryos were maintained in EM at 28°C on a 14/10 hour light/dark cycle prior to experimentation or rearing in the fish facility. Homozygous nrc a14 mutants were identified by the optokinetic response assay (OKR) or by genotyping as previously described [18,66]. Because WT and nrc a14 heterozygotes appear indistinguishable in every phenotypic assay we have performed, we refer to all OKR-positive larvae as WT. The Tg(crx : Man2a-GFP), Tg(TαCP : GFP-rab7) and Tg(TαCP : GFP-map1lc3b) fish lines have been previously described [18]. We created other transgenic fish lines as previously described [18,67]. Newly identified transgenic fish was assayed by OKR at 5 dpf to ensure that transgene expression did not affect visual responses. List of all transgenic lines used in this study is in Table S2.

Immunohistochemistry
All zebrafish larvae were fixed at the indicated developmental stage between 12 and 1 PM to minimize variation caused by circadian effects. Retinal slices were prepared as previously described [68]. To detect FLAG-tagged SJCs, slices were incubated with 1:1,000 anti-FLAG M2 (Sigma, St. Louis, MS, USA) primary antibody at 4°C overnight followed by secondary anti-mouse Alexa 568 (Invitrogen, A11002) at 1:200 for 1 hour at room temperature. Nuclei were counter stained with 5 μM Hoechst (Invitrogen). Slides were mounted with a coverslip and fluoromount-G (Southern Biotech, Birmingham, AL, USA). Imaging of retinal sections was performed on an Olympus FV1000 confocal microscope (Tokyo, Japan) with a 60× 1.35NA oil immersion objective. Fluoview software (Olympus version 2.0c) was used to acquire images.

Live imaging
Larvae were treated with 0.003% 1-phenyl-2-thiourea (PTU) in EM at~24 hours post-fertilization to prevent melanization. Larvae were anaesthetized in Tricaine (Sigma) and mounted in warm 0.5-1% low mount agarose. Embedded larvae were covered in EM containing PTU and tricaine and imaged. Imaging was performed on an Olympus FV1000 at room temperature using a 40× 0.8NA water immersion objective. Fluoview software (Olympus version 2.0c) was used to acquire images.

Transmission electron microscopy
Transmission electron microscopy was performed at the UW Vision Core as previously described [26]. For quantification, the entire retina was imaged.

Image processing and data analysis
Images were processed using NIH ImageJ (v1.49). Representative images in Figures are 2 μm MAX projections of confocal stacks (Figs. 1-3, 4A, 5A, 7-9), average projections of a single cell (Fig. 6) or single optical slices (Figs. 4B and 5B). Contrast was adjusted, and a 1 pixel median filter was applied to representative images in ImageJ; any adjustments were applied to in the exact same manner to WT and nrc a14 images. All puncta scoring was performed using blind manual counting on individual slices of unprocessed z-stacks. For qualitative comparisons between WT and nrc a14 larvae, at least six larvae of each genotype were analyzed. For quantitative data, the number of larvae analyzed is included in the text and

Analysis of LC3 vesicle dynamics
Images were acquired in 5 second intervals for 10 minutes. Timelapses were analyzed using the SpotTracker2D plugin in ImageJ [69]. Images were processed using the StackReg [70] and SpotEnhancing Filter2D plugins. Images were assigned random numbers prior to tracking. Particles were chosen randomly in the first frame and tracked until the particle disappeared, or the time-lapse ended. Particles that did not persist for at least 12 frames (1 minute) were discarded. A pause was defined as a movement of less than 0.475 μm between frames (0.095 μm s À1 ).

Colocalization analysis
Background correction was performed using a modified flatfield correction method [71]. After correction, images from each channel were thresholded and converted to a binary image using an automated method to avoid user bias. Using the Calculator Plus plugin, we created binary images that represent overlap from the two channels and quantified the number of voxels in each image. We express colocalization as percent colocalization. Figure 4E shows the percent colocalization of the indicated marker with 2XFYVE defined as (umber of voxels positive for both the marker and 2XFYVE/the number of voxels positive for the marker) × 100%. Figure 4F shows the percent colocalization of the indicated marker with 2XFYVE defined as (number of voxels positive for both the marker and 2XFYVE/the number of voxels positive for 2XFYVE) × 100%.