Retromer subunit, VPS29, regulates synaptic transmission and is required for endolysosomal function in the aging brain.

Retromer, including Vps35, Vps26, and Vps29, is a protein complex responsible for recycling proteins within the endolysosomal pathway. Although implicated in both Parkinson's and Alzheimer's disease, our understanding of retromer function in the adult brain remains limited, in part because Vps35 and Vps26 are essential for development. In Drosophila, we find that Vps29 is dispensable for embryogenesis but required for retromer function in aging adults, including for synaptic transmission, survival, and locomotion. Unexpectedly, in Vps29 mutants, Vps35 and Vps26 proteins are normally expressed and associated, but retromer is mislocalized from neuropil to soma with the Rab7 GTPase. Further, Vps29 phenotypes are suppressed by reducing Rab7 or overexpressing the GTPase activating protein, TBC1D5. With aging, retromer insufficiency triggers progressive endolysosomal dysfunction, with ultrastructural evidence of impaired substrate clearance and lysosomal stress. Our results reveal the role of Vps29 in retromer localization and function, highlighting requirements for brain homeostasis in aging.


Introduction
The endolysosomal membrane system comprises a dynamic network of interconnected compartments that mediates sorting or degradation of endocytosed proteins (Klumperman and Raposo, 2014;Repnik et al., 2013). Many factors that regulate endolysosomal trafficking are implicated in human disease, including neurodegenerative disorders (Small and Petsko, 2015;Soukup et al., 2018). Among these, retromer is a complex that recycles selected protein cargoes from the endosome to the trans-Golgi network or the plasma membrane (Burd and Cullen, 2014;Lucas and Hierro, 2017). Mutations in VPS35-encoding a retromer core protein-are a rare cause of familial Parkinson's disease (Vilariño-Güell et al., 2011;Zimprich et al., 2011), and retromer has also been linked to endocytic trafficking and processing of the Amyloid Precursor Protein and Tau in Alzheimer's disease (Muhammad et al., 2008;Vagnozzi et al., 2019;Wen et al., 2011). Despite its emerging importance in neurodegenerative disease, the requirements of retromer in neurons and particularly within the aging nervous system remain incompletely understood.
Notably, among the retromer core proteins, the precise roles of each subunit remain incompletely defined, with especially scant data on VPS29. VPS29 binds the VPS35 C-terminus (Collins et al., 2005;Hierro et al., 2007;Kovtun et al., 2018). Deletion of Vps29 in yeast or Caenorhabditis elegans phenocopies Vps35 loss-of-function (Lorenowicz et al., 2014;Seaman et al., 1997). In mammalian epithelial cell culture, reducing VPS29 results in apparent destabilization and degradation of both VPS35 and VPS26 (Fuse et al., 2015;Jimenez-Orgaz et al., 2018). Reciprocally, pharmacological chaperones targeting the VPS35-VPS29 interface can stabilize the complex and enhance retromer function (Mecozzi et al., 2014;Young et al., 2018;Lin et al., 2018). Here, we have generated and characterized a Drosophila Vps29 null allele with a focus on in vivo requirements in the nervous system. We identify an unexpected requirement for Vps29 in the regulation of retromer localization, and further highlight a role in synaptic vesicle recycling and lysosomal function in the aging brain.

Results
Vps29 is required for age-dependent retinal function Drosophila Vps29 is predicted to encode a 182 amino-acid protein that is 93% similar (83% identical) to human VPS29. Prior studies of Vps29 in flies have relied on RNA-interference knockdown approaches (Linhart et al., 2014). We instead generated a Vps29 null allele using a CRISPR-Cas9 strategy (Li-Kroeger et al., 2018). In the resulting mutant, Vps29 1 , the entire coding sequence was replaced by the visible marker gene, y wing2+ ( Figure 1A). Unexpectedly, Vps29 1 was homozygous viable, whereas both Vps35 and Vps26 mutants are lethal (Franch-Marro et al., 2008;Wang et al., 2014). Loss of the Vps29 genomic sequence in null animals was confirmed by PCR ( Figure 1B) and sequencing of the insertional breakpoints, and we were not able to detect any protein using an anti-Vps29 antibody on western blots from fly head homogenates ( Figure 1C). Although viable, Vps29 1 homozygotes are recovered at ratios below Mendelian expectation (Figure 1-figure supplement 1A). We also recovered viable Vps29 1 animals lacking both maternal and zygotic protein when crossing homozygous females to Figure 1. Vps29 is required for age-dependent retinal function. (A) The Vps29 genomic locus is shown, highlighting reagents used in this study. In the null allele, Vps29 1 , the gene coding sequence (blue, 2L: 1150852-1151420) is replaced by a ywing 2+ marker gene. Vps29 WT.GFP and Vps29 L152E.GFP are identical N-terminal tagged-Vps29 alleles, except for the L152E variant. A chromosomal deficiency Df(2L)Exel6004 is shown, with the deleted regions indicated by dashed lines. A bacterial artificial chromosome (BAC) (yellow) was used for transgenic genomic rescue (GR). (B) Genomic polymerase chain Figure 1 continued on next page heterozygous males. Notably, Vps29 1 mutant flies exhibit a modestly reduced survival (~50-60 days versus~75 days for controls), and this result was confirmed when Vps29 1 animals were crossed to the deficiency, Df(2L)Exel6004 ( Figure 1D). The reduced survival seen in Vps29 null animals was also rescued by a 23 kb P[acman] bacterial artificial chromosome (Venken et al., 2009) containing the Vps29 genomic locus, establishing specificity.
As introduced above, the retromer plays an essential role in the endocytic recycling of Rh1 in the Drosophila eye, thereby enabling proper phototransduction . Consequently, loss of either Vps35 or Vps26 leads to aberrant Rh1 trafficking, resulting in lysosomal stress and progressive retinal degeneration. In order to determine if Vps29 is similarly required, we generated somatic clones of either Vps29 1 or the Vps35 MH20 null allele in the fly eye using the eyeless (ey)-FLP/FRT system (Newsome et al., 2000), and monitored degeneration using electroretinograms (ERGs). Compared with control clones (FRT40A), removing either retromer component from photoreceptors disrupted light-induced depolarization in 10-day-old animals; however, the Vps29 phenotype was less severe than that due to loss of Vps35 ( Figure 1E,F). Consistent with other retromer components, the Vps29 1 retinal degeneration phenotype was suppressed when flies were raised in the dark, which eliminates phototransduction and the concomitant endocytosis of Rh1. Age-and lightdependent retinal degeneration in the absence of Vps29 was further confirmed by histology , and this phenotype was also reversed by introduction of the Vps29 genomic rescue construct. On further inspection of the ERG traces, we noticed that both Vps29 1 and Vps35 MH20 appear to disrupt synaptic transmission ( Figure 1E,G), based on progressive reduction in the on-and off-transient potentials in aging animals (Babcock et al., 2003). Interestingly, however, whereas raising animals in the dark rescued phototransduction, the ERG transient potential phenotypes persisted, suggesting a role for retromer at the synapse that is independent of Rh1 recycling.
In our ERG experiments, we also noticed that whereas the transient potentials are preserved in newly eclosed animals with Vps29 or Vps35 mutant eye clones, they were poorly sustained throughout the recordings. We therefore repeated our ERG experiments in retinae with either Vps29 1 or Vps35 MH20 mutant clones, but instead using a rapid light stimulation paradigm. Indeed, we documented a progressive decline in the on-and off-transient potentials following rapid stimulation, and consistent results were also obtained using the ey 3.5 -FLP driver (Mehta et al., 2005), which selectively targets presynaptic neurons ( Figure 1H and Figure 1-figure supplement 1C). This 'rundown' phenotype is characteristic of mutants that disrupt synaptic vesicle recycling (Harris and Littleton, 2015), suggesting that retromer is required for this process. We further demonstrated that the retinal synaptic transmission phenotype in Vps29 1 homozygotes was fully rescued using the pan- showing loss of Vps29 coding sequence in Vps29 1 homozygotes versus control (w) flies. P1, P2, and P3 denote expected PCR products from primer pairs targeting Vps29 genomic sequence. As an additional control, PCR was also performed for Vps35 genomic sequence. We also performed PCRs using primer pairs that span both sides of the breakpoint junctions abutting the inserted ywing 2+ marker gene cassette (not shown), and these products were Sanger sequenced to confirm the depicted molecular lesion. (C) Western blot from adult heads probed with anti-Vps29 antibody, confirming Vps29 1 is a protein null allele. (D) Vps29 1 homozygotes and Vps29 1 /Df transheterozygotes (Vps29 1 /Df(2L)Exel6004) show reduced survival that is rescued by the Vps29 genomic rescue strain (Vps29 1 /Df; GR/+). Quantitation based on n = 200-235 per group. See also Figure 1figure supplement 1A. (E) Representative ERG traces at 1-and 10 days after generation of ey-FLP clones from (i) control (FRT40A), (ii) Vps29 1 , or (iii) Vps35 MH20 . Flies were raised using an alternating 12 hr light/dark cycle (12h-LD) or in complete darkness. Loss of either Vps29 or Vps35 disrupted lightinduced depolarization (arrow) in 10-day-old flies. On-and off-transient ERG potentials (top and bottom arrowheads, respectively) were also lost. Raising flies in the dark restored ERG depolarization, but not the transients, indicating persistent defects in synaptic transmission. (F-G) Quantification (n = 6-8) of ERG depolarization and on-transient potentials. 'X' denotes undetectable on-transients in 20-day-old animals. See also Figure 1-figure supplement 1B. (H) Compared with controls (FRT40A clones), ERG transient potentials are extinguished by rapid stimulation in Vps29 1 clones. Consistent results were obtained using either ey-FLP or ey 3.5 -FLP, which targets presynaptic neurons only. Flies were raised in complete darkness and examined at 1 day. See also Figure 1-figure supplement 1C. (I) Rescue of Vps29 1 synaptic transmission defects by pan-neuronal expression (nSyb-GAL4 driver) of either Drosophila Vps29 or Vps35 (dVps29 and dVps35) or human Vps29 (hVps29-1,À2, or À3, representing three alternate isoforms). Quantification of ERG on-transients in 15-day-old flies (n = 9-12 per group) from the following genotypes: (1) Vps29 1 ;nSyb-Gal4/+; (2) Vps29 1 ;nSyb-Gal4/UAS-dVps29; (3-5) Vps29 1 ;nSyb-Gal4/UAS-hVps29; and (6) Vps29 1 ;nSyb-Gal4/UAS-dVps35. Statistical analysis was based on log-rank test with Bonferroni's correction (D) or one-way ANOVA (F, G, I). All error bars denote SEM. n.s., not significant; **, p<0.01; ***, p<0.001. The online version of this article includes the following figure supplement(s) for figure 1: Ye et al. eLife 2020;9:e51977. DOI: https://doi.org/10.7554/eLife.51977 neuronal expression driver (nSyb-GAL4) and a full-length Drosophila Vps29 cDNA (nSyb>Vps29) ( Figure 1I). Finally, similar rescue was observed when expressing three alternate human Vps29 isoforms, consistent with functional conservation. Surprisingly, we also found that pan-neuronal overexpression of Vps35 (nSyb>Vps35) partially restored synaptic transmission in Vps29 1 homozygous animals. In combination with the somewhat weaker phenotype, this result suggests that Vps29 may function primarily to potentiate the activity of other retromer subunits in the retina, such that Vps35 overexpression is compensatory. In sum, whereas Vps29 appears dispensable for retromer function during embryogenesis and development, these studies reveal a requirement in the aging nervous system for phototransduction and synaptic transmission.

Retromer regulates synaptic vesicle endocytosis and recycling
To further explore the requirement of retromer at synapses, we next turned to the larval neuromuscular junction (NMJ). Based on previous work, Vps35 is present at both the NMJ pre-and post-synapse, and loss of Vps35 causes synaptic terminal overgrowth and altered neurophysiology suggestive of synaptic vesicle recycling defects (Inoshita et al., 2017;Korolchuk et al., 2007;Malik et al., 2015;Walsh et al., 2019). We therefore examined whether Vps29 mutants show similar phenotypes. Indeed, NMJ preparations from Vps29 1 homozygous larvae revealed a modest but significant increase in synaptic bouton numbers, consistent with synaptic overgrowth (Figure 2A). However, both amplitude and frequency of NMJ miniature excitatory junction potentials (mEJPs) were unaffected following loss of Vps29 ( Figure 2B). Similarly, excitatory junction potentials (EJPs) were also preserved ( Figure 2C). These results suggest that Vps29 is neither required for spontaneous nor evoked synaptic vesicle release. However, high-frequency stimulation (10 Hz for 10 min) provoked a marked synaptic depression at the Vps29 1 NMJ ( Figure 2D). These data provide further support for the role of Vps29 in synaptic vesicle recycling.
The NMJ synaptic run-down phenotype is characteristic of mutants causing defective endocytosis (Bellen et al., 2010;Verstreken et al., 2002;Winther et al., 2013). To further investigate if synaptic vesicle endocytosis requires the retromer, we performed dye uptake studies in animals lacking Vps29 or Vps35 ( Figure 2E and Figure 2-figure supplement 1). Larval preparations were stimulated with potassium chloride in the presence of extracellular buffer containing the fluorescent dye FM 1-43, permitting quantification of endocytic flux (Verstreken et al., 2008). In control animals, FM 1-43 dye was rapidly internalized at the presynaptic membrane. However, NMJs from either Vps29 1 or Vps35 MH20 homozygous animals significantly reduced FM dye uptake. These results are consistent with models whereby retromer either directly or indirectly supports vesicle recycling at the presynapse.
Vps29 regulates retromer localization and is required in the aging nervous system The requirement of retromer during Drosophila embryogenesis has hindered systematic characterization of its function in the adult nervous system, with the exception of studies in the retina relying on clonal analysis of Vps35 and Vps26 . Although Vps29 is dispensable for embryonic development, our results (above) establish a largely conserved Vps29 requirement for retromer function in the retina and at the NMJ synapse. We therefore studied Vps29 null adult animals to further explore retromer requirements in the adult brain. Except for the aforementioned retinal requirement, we did not observe obvious morphological defects or apparent evidence of neurodegeneration in the brains of 45-day-old Vps29 1 /Df adults based on hematoxylin and eosin staining of paraffin sections (Figure 3-figure supplement 1A). In order to examine nervous system function, we next tested the startle-induced negative geotactic response (climbing) (Davis et al., 2016;Rousseaux et al., 2018). Although locomotor behavior was normal in newly eclosed Vps29 null adults, climbing ability significantly declined with aging ( Figure 3A). Our results are consistent with published work using RNA-interference to target Vps29 (Linhart et al., 2014). Vps29 1 homozygous flies manifested a stronger locomotor dysfunction phenotype than Vps29 1 /Df, suggesting the presence of a potential modifier; however, both genotypes were fully rescued by the genomic BAC transgene. Similar results were obtained for the Vps29 retinal degeneration phenotype (Figure 1figure supplement 1B). Locomotor defects in Vps29 1 homozygous animals were also partially rescued by pan-neuronal Vps29 expression (nsyb>Vps29), and human VPS29 showed comparable  Figure 3B). Overall, these results suggest Vps29 is required for the maintenance of nervous system function in aging animals.
In prior studies of HeLa cell cultures, VPS29 knockdown caused reduced levels of both VPS35 and VPS26, likely due to destabilization of the retromer trimer complex and subunit turnover (Fuse et al., 2015;Jimenez-Orgaz et al., 2018). Reciprocally, enhancing interactions between VPS29 and VPS35 promotes retromer stability (Mecozzi et al., 2014;Young et al., 2018). However, Vps35 and Vps26 protein levels were unaffected in Vps29 null animals (either Vps29 1 homozygotes or Vps29 1 /Df), based on western blots from either whole larvae or adult Drosophila heads ( Figure 3C and Figure 3-figure supplement 1B,C), including from 30-day-old animals. Furthermore, we confirmed that Vps35 and Vps26 remain tightly associated in the absence of Vps29, based on coimmunoprecipitation assays ( Figure 3D). Thus, Vps35-Vps26 complex assembly and stability in the nervous system are preserved in Drosophila lacking Vps29.
Besides expression and stability, retromer function is tightly regulated by its subcellular localization (Follett et al., 2016;Lucas et al., 2016). We examined retromer expression and localization in the adult fly brain using endogenously tagged fluorescent protein alleles, encoding either Vps35 RFP or Vps35 GFP (Koles et al., 2016), as well as Vps29 GFP ( Figure 1A) fusion proteins. As expected, Vps29 and Vps35 appeared to colocalize, and were broadly expressed in the Drosophila brain, including the antennal lobes and the mushroom body (a-/b-lobes and peduncles) ( Figure 4A,B). We confirmed that most of the observed staining in the adult brain, including within neuropil regions, derived from neuronal Vps35, as this signal was lost following neuronal-specific knockdown . However, in Vps29 1 homozygous animals, Vps35 showed a striking redistribution, shifting from neuropil to enrichment in soma, and forming large perinuclear puncta ( Figure 4C). Moreover, Vps35 was mislocalized in brains from 1-day-old adults and this result did not significantly change with aging. Consistent results were obtained in the lamina, where retinal photoreceptors synapse on second-order neurons in the visual pathway ( Figure 4-figure supplement 1B). Specifically, in Vps29 1 homozygotes, both Vps35 and Vps26 accumulated in the lamina cortex and staining was attenuated in the lamina neuropil. Lastly, we also found that Vps35 was depleted from the Vps29 1 homozygous larval brain neuropil and at the NMJ synaptic bouton; however, we did not detect any apparent enrichment in motor neuron cell bodies ( Figure 4-figure supplement 2). Overall, these data suggest that Vps29 is required for normal retromer localization within neurons.
If Rab7 activation mediates the disruption in retromer localization and function in Vps29 mutants, our model predicts that reduction in Rab7 may rescue the neuronal requirements for Vps29. Indeed, although complete loss-of-function for Rab7 is lethal, removing one copy (Rab7 Gal4-KO /+) (Cherry et al., 2013) partially rescued the age-dependent locomotor impairment and synaptic transmission defects manifest in Vps29 1 /Df animals ( Figure 5A,B). Reduction of Rab7 also restored synaptic vesicle endocytosis at the larval NMJ in Vps29 1 homozygotes, based on the FM 1-43 dye uptake assay ( Figure 5C). In mammalian cell culture studies, VPS29 regulates Rab7 activity via recruitment of the GTPase-activating protein (GAP), TBC1D5 (Jia et al., 2016;Seaman et al., 2009). Drosophila has a single ortholog of TBC1D5 that is well-conserved (28% identity/47% similarity) (Gramates et al., 2017). Consistent with our model, pan-neuronal overexpression of dTBC1D5 (nsyb>dTBC1D5) normalized Rab7 protein levels ( Figure 5D) and partially rescued the synaptic transmission defect in Vps29 mutants ( Figure 5E). Next, we introduced a point mutation at the Vps29 genomic locus causing the single amino-acid substitution, Vps29 L152E ( Figure 1A); mutation of this conserved residue has been shown to disrupt the interaction with TBC1D5 in mammalian cells (Jia et al., 2016;Jimenez-Orgaz et al., 2018). Vps29 L152E failed to complement Vps29 1 ( Figure 5F-H and Figure 5-figure supplement 1), causing age-dependent photoreceptor synaptic transmission defects and locomotor impairment. The residual, albeit weak on-transient ERG potential in Vps29 1 /Vps29 L152E adults suggest that Vps29 L152E is a hypomorphic allele. Moreover, Vps29 L152E was expressed at higher levels and Rab7 was also increased, as in Vps29 1 . Overall, these results indicate that Vps29 regulates retromer localization coordinately with Rab7 and TBC1D5. 1A. (C) In the absence of Vps29, Vps35 GFP (green) protein is redistributed in the adult brain, shifting from neuropil to soma and forming large perinuclear puncta. Neuronal nuclei are labeled with anti-Elav (red). Within the boxed region of interest, including antennal lobe and surrounding nuclei, the GFP intensity ratio (perinuclear to neuropil) was quantified (n = 3) in Vps29 1 , Vps35 GFP homozygotes or Vps35 GFP controls. Scale bars, 100 mm. See also     Loss of Vps29 causes lysosomal dysfunction during aging As introduced above, retromer mediates the recycling of protein cargoes from the endolysosomal system to the trans-Golgi network or plasma membrane. Moreover, retromer loss may impair lysosomal biogenesis and function, for example by disrupting delivery of key hydrolases (Maruzs et al., 2015). In the Drosophila retina, besides aberrant trafficking of Rh1 , loss of Vps35 causes accumulation of ceramide lipid species (Lin et al., 2018), which contributes to lysosomal expansion, cellular stress, and ultimately photoreceptor loss. We confirmed that glucosylceramide similarly accumulates in Vps29 1 homozygous photoreceptors ( Figure 6A), consistent with the retinal degeneration phenotype. In order to better understand the proximal consequences of retromer mislocalization that lead to neuronal dysfunction, we performed a series of experiments to assay lysosomal function in the adult brains of Vps29 mutants. We first examined the lysosomal proteases, cathepsin D (CTSD) and cathepsin L (CTSL). These enzymes are each synthesized as propeptides in the endoplasmic reticulum (ER) and become active following cleavage in the acidic, lysosomal milieu. Lysosomal dysfunction therefore commonly leads to reduced levels of mature cathepsins and increased levels of the immature proforms (Hasanagic et al., 2015;Li et al., 2017). We detected normal levels of mature CTSL and increased mature CTSD in adult head homogenates from 1-dayold Vps29 1 /Df animals ( Figure 6B and Figure 6-figure supplement 1A). However, in 30-day-old animals, we found that the uncleaved proforms of both cathepsins were sharply increased, consistent with progressive lysosomal dysfunction with aging.
We next examined well-established markers of lysosomal autophagy, which plays a critical role in nervous system homeostasis and neurodegeneration (Martini-Stoica et al., 2016;Menzies et al., 2015). Substrate degradation is triggered when Atg8-decorated autophagosomes fuse with lysosomes, leading to turnover of Ref (2)P/p62 and concomitant clearance of polyubiquitinated proteins. We did not observe any accumulation of these markers in heads from either 1-or 30-day-old Vps29 1 /Df flies (Figure 6-figure supplement 1B,C), consistent with preserved autophagic flux. Interestingly, however, at more extreme ages (45 days), Vps29 mutants showed evidence of accelerated autophagic failure, with more significant elevation of both p62 and Atg8, and accumulation of polyubiquitinated proteins ( Figure 6C). Notably, reduction of Rab7 (Rab7 Gal4-KO /+) improved autophagic clearance in Vps29 mutant brains.
(C) Autophagic flux is reduced in aged animals lacking Vps29, and this phenotype is suppressed by reduction of Rab7. Western blots of adult head homogenates were probed for autophagic markers, including p62, Atg8, polyubiquitin (FK1), or Tubulin (loading control), and quantified based on Figure 6 continued on next page other apparent evidence of synaptic degeneration. Lastly, we examined TEM of adult brain sections in 30-day-old animals, focusing on cortical regions with densely packed neuronal cell bodies in the dorsal-posterior brain. Consistent with findings in the retina, we discovered significantly increased numbers of aberrant lysosomal structures (multilamellar bodies) in Vps29 mutants. Collectively, our data indicate that retromer dysfunction following loss of Vps29 is associated with progressive lysosomal structural and functional degeneration in the aging nervous system.

Discussion
Although strongly implicated in pathogenesis of human neurodegenerative disease, the requirements for retromer in the aging brain remain poorly defined. Our discovery that Vps29 is dispensable for Drosophila embryogenesis provides an unexpected opportunity to examine the neuronal consequences of retromer dysfunction in vivo. In the absence of Vps29, we document age-dependent impairment in lysosomal proteolysis and autophagy, leading to apparent accumulation of undigested substrates and expansion of the endolysosomal compartment. These cellular changes are accompanied by progressive nervous system dysfunction, with synaptic transmission especially vulnerable. Unexpectedly, Vps29 loss did not affect the expression or stability of other retromer components but influenced their localization via a Rab7-and TBC1D5-dependent regulatory pathway.

VPS29 is required for retromer recruitment via Rab7 and TBC1D5
The retromer core initiates cargo recognition and mediates the engagement of other factors that enable proper endocytic trafficking (Cullen and Steinberg, 2018). However, the differential requirements of each protein within the heterotrimer, including VPS35, VPS26, and VPS29, remain incompletely defined. Structural studies suggest the retromer core has an elongated, flexible structure, with VPS35 serving as a central scaffold, binding VPS26 at its N-terminus, and VPS29 at its C-terminus (Hierro et al., 2007;Kovtun et al., 2018;Lucas et al., 2016). Consistent with this, loss of VPS35 destabilizes the retromer complex, resulting in reduced VPS26 and VPS29 protein, and this result has been confirmed both in vitro and in vivo (Cui et al., 2019;Liu et al., 2014;Wang et al., 2014). Similarly, knockdown or knockout of VPS29 in mammalian epithelial cell culture compromises retromer stability, causing loss of the other components (Fuse et al., 2015;Jimenez-Orgaz et al., 2018). Surprisingly, in the absence of Vps29 in Drosophila, we found that Vps35 and Vps26 are present at normal levels and remain tightly associated. Instead, Vps35 appears to be mislocalized, shifting from neuropil to soma in the adult brain, and likely disrupting its normal function. Loss of Vps29 also causes apparent hyperactivation (and mislocalization) of Rab7, and a mutation previously shown to disrupt the VPS29-TBC1D5 interaction shows similar phenotypes. Importantly, reduction of Rab7 or overexpression of TBC1D5 potently suppresses Vps29 loss-of-function. In the adult brain, perinuclear accumulation of Vps35/Rab7 strongly suggests aberrant localization to the late endosome and lysosome, based on the overlap with Arl8 and Spinster. Our in vivo findings recapitulate a two-step molecular mechanism for retromer recruitment and release in neurons (Figure 8). First, the retromer core is recruited to the endosomal membrane by Rab7-GTP, allowing it to participate in cargo recycling. Subsequently, Vps29 engages TBC1D5, which activates GTP hydrolysis of Rab7, thereby releasing retromer from the endosome. We propose that in the absence of Vps29, retromer is   The reciprocal interplay and interdependence of retromer and Rab7 that we document in the Drosophila nervous system is consistent with prior studies in cell culture (Jimenez-Orgaz et al., 2018;Liu et al., 2012;Seaman et al., 2009). By contrast, the viability of Vps29 mutants as well as our finding of a more selective, regulatory role for Vps29 in complex localization is unexpected. Unlike the other retromer core subunits, Vps29 is dispensable for embryogenesis and loss-of-function causes milder retinal phenotypes than Vps35 or Vps26. Moreover, Vps35 overexpression partially rescued selected Vps29 phenotypes, consistent with a context-dependent, regulatory role. Since Vps35 can weakly bind TBC1D5 (Jia et al., 2016), it may be capable of inactivating Rab7, compensating in part for the absence of Vps29. Our finding that retromer  . Model for Vps29-dependent retromer recruitment and release. In neurons, the retromer core Vp26-Vps35-Vps29 is recruited by Rab7 to the endosomal membrane. Vps29 engages TBC1D5, which promotes inactivation of Rab7 and release of Vps35. In the absence of Vps29, the residual retromer complex is trapped at the endosomal membrane. retains some residual activity in flies lacking Vps29 is reminiscent of results from C. elegans (Lorenowicz et al., 2014). Moreover, in S. cerevisiase, interaction of Vps26 and Vps35 are similarly preserved following loss of Vps29 (Reddy and Seaman, 2001). Importantly, these data suggest that the specific requirement(s) for Vps29-and by extension, retromer activity-may be context dependent, varying with cell-type (e.g. neuron) and/or specific cargo (e.g. Rh1, ceramides, or others). While Vps29 mutants altered Vps35 subcellular localization in both the adult and larval brains causing relative loss from neuropil regions, potential differences were also noted: the apparent redistribution to soma was only seen in adults. It is also possible that tissue-or species-specific differences might relate to Vps29 participation in other endosomal sorting machinery besides retromer, such as the recently described 'retriever' (McNally et al., 2017); however, this complex has not yet been studied in Drosophila. Although we have found that human VPS29 is capable of functional substitution in flies, it will be important to determine whether our findings also apply in the mammalian nervous system context.

The retromer and endolysosomal trafficking in neurons
Our characterization of flies lacking Vps29 points to important roles for retromer in both lysosomal and synaptic homeostasis in neurons, interdependent processes that similarly require proper endocytic trafficking. Following Vps29 loss-of-function, we detected evidence of progressive deterioration in lysosomal function in the adult brain, based on markers of cathepsin maturation and autophagic flux. In both cases, lysosomal function was preserved in newly eclosed animals, but deficits emerged gradually with aging. This result suggests that neurons can initially compensate for retromer insufficiency, whereas adaptive mechanisms may breakdown during aging. Vps35 was strongly mislocalized in 1-day-old animals, suggesting that the manifestation of age-dependent vulnerability likely involves more downstream consequences of retromer dysfunction, such as the accumulation of undigested substrates and resulting lysosomal expansion. Indeed, we found ultrastructural evidence of lysosomal stress in the aged retina and brain, including significantly enlarged lysosomes that were filled with electron-dense and/or multilammellar material. One caveat for interpretation is that TEM data was not included for newly eclosed animals. Nevertheless, our results are consistent with evidence from other experimental systems, that protein and lipid cargoes accumulate gradually in the late endosome and lysosomes in the absence of retromer activity (Lin et al., 2018;Tang et al., 2015a;Wang et al., 2014;Wen et al., 2011). Neuronal compensation may entail increasing capacity for substrate turnover, since we found that mature forms of CTSD and markers of autophagic flux were increased in young animals ( Figure 6-figure supplement 1). It is also possible that the production of critical membrane proteins dependent on retromer for recycling can be boosted, at least temporarily. The initial resilience of the brain to Vps29 deficiency that we uncover in Drosophila is consistent with relative preservation of autophagy following VPS29 knockout in HeLa cells (Jimenez-Orgaz et al., 2018). However, there is ample evidence from animal model studies that aging is accompanied by reduced degradative capacity in neurons, which may ultimately contribute to decompensation and increased proteotoxic stress (Mattson and Magnus, 2006).
We additionally found that Vps29 is required for synaptic transmission in both the Drosophila retina and larval NMJ, producing a phenotype characteristic of other genes that regulate endocytosis and synaptic vesicle recycling (Bellen et al., 2010;Harris and Littleton, 2015). In Drosophila S2 cells, Vps35 has been shown to participate in endocytosis (Korolchuk et al., 2007), and our findings confirm and extend prior work (Inoshita et al., 2017) implicating Vps35 in synaptic structure and function at the NMJ. Whereas both genes appear similarly required for synaptic vesicle endocytosis, loss of Vps35 but not Vps29, also affected spontaneous NMJ activity, causing increased miniature excitatory junction potentials (Inoshita et al., 2017). This is consistent with our other data suggesting Vps29 causes overall milder defects. Vps29 synaptic phenotypes were suppressed by either reduction of Rab7 or overexpression of TBC1D5, similar to our studies of lysosomal phenotypes. Interestingly, ERGs also revealed age-dependent, progressive decline in synaptic transmission. Despite evidence of significant degradation of synaptic function with aging, TEM studies of the Drosophila lamina did not reveal overt synaptic structural changes, suggesting a more subtle, predominantly functional defect. Since retromer is present throughout embryonic and larval development, we cannot exclude the possibility of developmental defects due to Vps29 loss-of-function. Indeed, Vps35 was mislocalized in the Vps29 mutant larval brain and we documented increased numbers of NMJ synaptic boutons. Nevertheless, the adult brain appeared grossly normal, and electroretinograms and climbing were unaffected in newly eclosed flies.
In sum, we consider two non-exclusive models for potential retromer requirement(s) at the synapse. First, retromer may regulate the trafficking of a key factor required for synaptic membrane endocytosis and/or replenishing of the vesicle pool from synaptic endosomes. Second, Vps29related synaptic dysfunction may instead be related, more broadly, to progressive lysosomal dysfunction. In fact, recent studies have linked synaptic endosomes, lysosomes, and autophagosomes to presynaptic proteostasis, including during vesicle recycling and release (Jin et al., 2018;Wang et al., 2017). For example, another Rab GTPase activating protein, TBC1D24-Skywalker in Drosophila-has been shown to regulate synaptic vesicle recycling and quality control via a sorting endosome at NMJ terminals (Fernandes et al., 2014;Uytterhoeven et al., 2011). Interestingly, in the Drosophila visual system, we found that raising Vps29 or Vps35 mutants in the dark rescued retinal degeneration but not synaptic transmission defects, consistent with an uncoupling of retromer requirements in these two spatially distinct processes. Thus, it is possible-if not likely-that synaptic endolysosomal trafficking and function can be regulated independently from that of the neuronal soma.
Additional studies will be required to further dissect and refine our understanding of retromer-dependent synaptic mechanisms. For example, it may be informative to perform neurophysiology studies in mutants for other genes encoding retromer partners (e.g. TBC1D5) or regulators of late endosome-lysosome fusion, such as components of the HOPS complex (Fernandes et al., 2014;Lund et al., 2018). It will also be important to examine synaptic physiology following retromer loss in the mammalian brain. Retromer components are present at synaptic terminals in mouse primary neuronal culture (Munsie et al., 2015;Vazquez-Sanchez et al., 2018) and in cortical synaptosome preparations (Tsika et al., 2014). However, functional studies to date have largely focused on the post-synaptic membrane, where evidence supports retromer requirements for trafficking of glutamatergic and adrenergic neurotransmitter receptors (Choy et al., 2014;Tian et al., 2015).

Retromer and neurodegenerative disease
Retromer dysfunction has been linked to pathogenesis of Parkinson's and Alzheimer's disease (Small and Petsko, 2015). In both neurodegenerative disorders, synaptic dysfunction is an early pathologic feature, and lysosomal autophagy is also strongly implicated (Ihara et al., 2012;Wang et al., 2017). Our findings of age-dependent, progressive synaptic and lysosomal dysfunction in the adult brain of Drosophila lacking Vps29 may therefore be relevant for our understanding of retromer in human diseases. In fact, variants at numerous other genetic loci related to endocytic trafficking and synaptic maintenance have been discovered as risk factors for Alzheimer's disease (BIN1, PICALM, CD2AP, SORL1) (Karch et al., 2014) and Parkinson's disease (SNCA, LRRK2, SH3GL2/ EndoA, RAB7L1) (Soukup et al., 2018). Genetic links have also recently emerged between Parkinson's disease and lysosomal storage disorders, comprising a heterogeneous group of recessive disorders arising from defects in lysosomal biogenesis and/or function (Ysselstein et al., 2019). These disorders, which can include neurodegenerative features, are characterized by the accumulation of undigested substrates and lysosomal expansion (e.g. glucosylceramide in Gaucher's disease). Importantly, retromer participates in the trafficking of lysosomal hydrolases (Cui et al., 2019;Seaman et al., 1997), and in the absence of Vps29, we documented accumulation of glucosylceramides along with ultrastructural evidence of brain lysosomal pathology similar to lysosomal storage disorders. Based on mouse models, synaptic dysfunction may be an early manifestation in lysosomal storage disorders similar to Vps29 mutant flies (Mitter et al., 2003;Ohara et al., 2004;Sambri et al., 2017). Aberrant lysosomal morphology has also been described in both Alzheimer's and Parkinson's disease models (Giasson et al., 2002;Stokin et al., 2005) and in human postmortem brain (García-Sanz et al., 2018;Piras et al., 2016). Targeting the retromer for therapeutic manipulation in neurodegenerative disease is one promising approach, given the availability of small molecule chaperones that bind Vps29 (Mecozzi et al., 2014;Vagnozzi et al., 2019;Young et al., 2018). However, since the endolysosomal system has many functions, including within neurons and other cell types, it will be essential to pinpoint the specific retromer mechanism(s) responsible for age-related brain diseases.

Fly stocks and husbandry
A complete list of fly strains used in this study is included in the Key Resources Table. Detailed information on experimental genotypes can also be found in figures and figure legends. The Vps29 1 null allele was generated using CRISPR/Cas9-mediated gene replacement, as previously described (Li-Kroeger et al., 2018). Briefly, a dominant marker ywing 2+ flanked by homology arms on each side, along with sgRNA expression plasmids, were injected into embryos expressing Cas9 (y 1 ,w*, . G0 animals were next crossed to y w, and progeny were screened for the presence of the yellow + wing marker to identify Vps29 1 (Vps29 ywing2+ ). Studies of Vps29 mutant flies used either Vps29 1 homozygotes or the transheterozygous genotype, Vps29 1 /Df(2L)Exel6004 (referred to as Vps29 1 /Df). To generate the Vps29 genomic rescue strain, a 23 kb P[acman] genomic fragment (CH322-128G03) including the Vps29 locus was injected into y,w,FC31; VK33 attP embryos, followed by selection of F1 transformants. All studies using the transgenic Vps29 BAC examined the genomic rescue construct (GR) in heterozygosity (either Vps29 1 ;GR/+ or Vps29 1 /Df;GR/+).
In order to generate the Vps29 WT.GFP and Vps29 L152E.GFP alleles, we again used the CRISPR/Cas9 gene-replacement system, starting with the Vps29 1 strain. The GFP coding sequence was cloned inframe and proximal to the Vps29 cDNA using NEBuilder HiFi DNA Assembly Cloning Kit (NEB). Using Q5 Site-Directed Mutagenesis Kit (NEB), the L152E mutation was introduced to the GFP:: Vps29 plasmid. As above, the GFP::Vps29 or GFP::Vps29 L152E constructs were each injected into y 1 M{nos-Cas9.P}ZH-2A w*; Vps29 1 /CyO flies. G0 flies were crossed to y w, and F1 progeny were screened for loss of the yellow + marker to establish the Vps29 WT.GFP and Vps29 L152E.GFP strains. The sequence of all CRISPR/Cas9-generated strains were confirmed by genomic PCR followed by Sanger sequencing. Genomic DNA from a single fly was prepared using 'squish buffer' (10 mM Tris-Cl pH 8.0, 1 mM EDTA, 25 mM NaCl, and 0.2 mg/ml Proteinase K). Total genomic DNA was extracted using the PureLink Genomic DNA Kit (Thermo Fisher Scientific). PCR primers used for additional confirmation of the Vps29 1 allele ( Figure 1B) are listed in the Key Resources Table. As part of this study, we also generated transgenic strains carrying UAS-dVps29-myc, UAS-dTBC1D5-myc, and UAS-hVps29-FLAG. The full-length Drosophila Vps29 cDNA (Drosophila Genomics Resource Center (DGRC) plasmid: GH25884) was cloned proximal to c-myc coding sequencing, using primers dVps29-myc-F and dVps29-myc-R (Key Resources Table). The full-length cDNA for TBC1D5 (DGRC plasmid: BS16827) was similarly cloned proximal to the c-myc coding sequencing using primers dTBC1D5-myc-F and dTBC1D5-myc-R. The Vps29-myc and TBC1D5-myc PCR products were each digested with BglII and XbaI restriction enzymes (NEB) and ligated to the pUAST-attB vector using T4 ligase (NEB). Plasmids containing FLAG-tagged Human VPS29 cDNAs (OHu00442D, OHu02289D, and OHu05688D, respectively) were obtained from GenScript. For PCR cloning, we used hVps29-R along with each of the following: hVps29-1-F for hVps29 transcript variant 1 (NM_016226.4), hVps29-2-F for hVps29 transcript variant 2 (NM_057180.2), and hVps29-3-F for hVps29 transcript variant 3 (NM_001282150.1). PCR products were digested by XhoI and XbaI restriction enzymes (NEB), and then ligated to the pUAST-attB vector. The resulting constructs were purified, verified by Sanger DNA sequencing, and injected into y,w,FC31; VK33 embryos to generate transgenic flies. The generation of w;Vps35 TagRFP-T (Vps35 RFP ) with a Rippase-switchable TagRFP-T to GFP tag was described in Koles et al. (2016). w;Vps35 GFP was generated by crossing Vps35 RFP to bam-Gal4;;UAS-Rippase::PEST @attP2 flies to isolate germline ripouts of the TagRFP-T cassette, leaving a C-terminal EGFP knockin.
All Drosophila crosses were raised on molasses-based food at 25˚C unless otherwise noted. To induce neuronal specific knockdown of Vps35 GFP , we used the deGradFP system (Caussinus et al., 2012;Nagarkar-Jaiswal et al., 2015). Crosses were established and maintained at 18˚C, and following eclosure, adults were shifted to 29˚C for 7 days before brain dissection.

Western blot analysis and immunoprecipitation
For western blots, adult fly heads or whole larvae were homogenized in 2X Laemmli Sample Buffer (Bio-Rad) with 5% b-mercaptoethanol (Calbiochem) using a pestle (Argos Technologies). The lysates were heated at 95˚C for 5 min, followed by centrifugation at 21,130 Â g at 4˚C for 15 min before SDS-PAGE analysis. Samples were loaded into 12% Bis-Tris gels (Invitrogen), separated by SDS-PAGE. For immunoprecipitation of Vps35 GFP , 300 fly heads from 1-to 2 day-old animals were homogenized on ice in 700 mL of lysis buffer containing 10 mM Tris pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40, and 1X complete protease inhibitor (Roche). Homogenized samples were centrifugated at 21,130 Â g for 30 min at 4˚C. 3% (by volume) of the supernatant were reserved for total protein input. The remaining supernatant was incubated for 1 hr at 4˚C with 50 mL of protein A/G agarose slurry (Thermo Fisher) to reduce non-specific binding. Following pre-clearing, the lysate was incubated for 3 hr at 4˚C with 35 mL of GFP-Trap agarose beads (Allele Biotechnology) with mild agitation. The beads were washed three times with lysis buffer, boiled in 70 mL of 2X Laemmli sample buffer (Biorad), and subjected to SDS-PAGE (BOLT12% Bis-Tris Gel, Invitrogen). Gels were transferred to PVDF membrane (Millipore), and blocked in 5% bovine serum albumin (Sigma) in 1X TBST (Tris-buffered saline + 0.1% Tween-20). We used the following primary antibodies and dilutions: HRPconjugated secondary antibodies were used at 1:5000. Antibodies against Drosophila Vps35 were generated by immunization of guinea pigs with a bacterially expressed peptide comprising the C-terminal 338 amino acids of Vps35. Bacterial expression as a GST fusion protein was facilitated by the pGEX 6 P-2 vector. The GST tag was used for protein purification and was cleaved off using PreScission Protease (GE Healthcare Life Sciences) prior to immunization.

Survival and climbing assays
For survival analyses, around 200 flies per genotype (approximately equal numbers of females and males) were aged in groups of no more than 30 flies per vial, transferring to fresh vials and food every 2-3 days. For the startle-induced negative geotaxis assay (climbing), four to five groups consisting of approximately 15 flies each were placed in a plastic cylinder. Flies were gently tapped to the bottom of the cylinder, and locomotor activity was videotaped and quantified as the percentage of animals climbing past the 9 cm line within a 15 s interval.

Electroretinogram (ERG)
All crosses were initially maintained in the dark to prevent flies from being exposed to light before eclosion. Newly-eclosed adults were either shifted to a 12 hr light/dark cycle (3,000 lux for light exposure) or maintained in light-sealed boxes for constant darkness. During aging, the positions of light-exposed vials were randomly shuffled within racks. ERG recordings were performed as previously described (Chouhan et al., 2016). In brief, adult flies were anesthetized and glued to a glass slide, with electrodes placed on the corneal surface and the thorax. Flies were maintained in the dark for at least 1 min prior to stimulation using a train of alternating light/dark pulses. Retinal responses were recorded and analyzed using LabChart software (ADInstruments). At least eight flies were examined for each genotype and timepoint.

NMJ electrophysiology and FM 1-43 dye studies
Animals were dissected at the wandering third instar stage in HL3 buffer (above), without addition of calcium. Electrophysiologic recordings were performed with 0.5 mM extracellular Ca 2+ buffer concentrations, as described in Ojelade et al. (2019). Larval motor axons were severed and miniature excitatory junction potentials (mEJPs) were recorded from muscle 6 of abdominal segments A2 and A3 for 2 min. EJPs were evoked at 0.2 Hz. EJPs and mEJPs were analyzed using pClamp6 (Molecular Devices) and Mini Analysis Program (Synaptosoft) software, respectively. High-frequency stimulation was recorded at 10 Hz for 10 min. The recorded EJP data were binned at 0.5 min intervals and normalized to the average EJP. EJP amplitudes were corrected for nonlinear summation as previously described (Martin, 1955). The FM 1-43 dye uptake assay was performed as described in Verstreken et al. (2008). Larval NMJ preparations were stimulated for 2 min in dark conditions using HL3 buffer containing 90 mM K + , 1.5 mM Ca 2+ , and 4 mM FM 1-43 dye. Larval preparations were subsequently washed five times with HL3 solution (without 90 mM K + and calcium), and FM dye uptake was imaged using a Leica Sp8 confocal system. Signal intensity of FM dye per bouton was normalized to each bouton area. Six boutons were sampled per animal. For each genotype, 7-11 animals were assayed.

Transmission Electron Microscopy (TEM)
For analysis of adult eye ultrastructure, fly heads were dissected by removing proboscis and air sacs, then severed from the thorax. For analysis of central brain ultrastructure, fly heads were dissected by removing cuticles and eyes to ensure complete penetration of fixative, leaving each head affixed to the thorax. All samples were processed for TEM as previously described (Chouhan et al., 2016) using a Ted Pella Bio Wave processing microwave with vacuum attachment. Briefly, dissected samples were fixed (4% paraformaldehyde, 2% glutaraldehyde, and 0.1 M sodium cacodylate, pH 7.2) at 4˚C for 48 hr, and then submerged in 1% osmium tetroxide for 45 min. The fixed samples were dehydrated using an ethanol series followed by propylene oxide, and then embedded using Embed-812 resin (EMS). For brain samples, each head was removed carefully from the thorax prior to embedding. Transverse sections of brains (50 nm) and tangential sections of eyes (50 nm) were prepared with a Leica UC7 microtome, and post-stained with 1% uranyl acetate and 2.5% lead citrate. All TEM images were acquired using a JEOL 1010 Transmission Electron Microscope. TEM images of photoreceptor sections were prepared from three different animals per genotype. TEM images of brain sections were prepared from four animals per genotype. Images were acquired from the dorsal-posterior cortical brain region (Figure 7-figure supplement 1D), including at least 50 cells per brain.

Quantification and statistics
Confocal images were processed and analyzed using ImageJ software (NIH). Sample size for all comparisons is included in each figure legend (also noted above). For survival curves, we performed logrank test with Bonferroni's correction. For other statistical analysis, we performed two-tailed, unpaired t-tests or Analysis of Variance (ANOVA) followed by Tukey's post-hoc test for multiple comparisons, as specified in all figure legends. The significance threshold for all analyses was set to p<0.05. Otherwise, results are noted as 'not significant ' (n.s.). Error bars in all analyses represent the standard error of the mean (SEM).