Abnormal triaging of misfolded proteins by adult neuronal ceroid lipofuscinosis-associated DNAJC5/CSPα mutants causes lipofuscin accumulation

ABSTRACT Mutations in DNAJC5/CSPα are associated with adult neuronal ceroid lipofuscinosis (ANCL), a dominant-inherited neurodegenerative disease featuring lysosome-derived autofluorescent storage materials (AFSMs) termed lipofuscin. Functionally, DNAJC5 has been implicated in chaperoning synaptic proteins and in misfolding-associated protein secretion (MAPS), but how DNAJC5 dysfunction causes lipofuscinosis and neurodegeneration is unclear. Here we report two functionally distinct but coupled chaperoning activities of DNAJC5, which jointly regulate lysosomal homeostasis: While endolysosome-associated DNAJC5 promotes ESCRT-dependent microautophagy, a fraction of perinuclear and non-lysosomal DNAJC5 mediates MAPS. Functional proteomics identifies a previously unknown DNAJC5 interactor SLC3A2/CD98hc that is essential for the perinuclear DNAJC5 localization and MAPS but dispensable for microautophagy. Importantly, uncoupling these two processes, as seen in cells lacking SLC3A2 or expressing ANCL-associated DNAJC5 mutants, generates DNAJC5-containing AFSMs resembling NCL patient-derived lipofuscin and induces neurodegeneration in a Drosophila ANCL model. These findings suggest that MAPS safeguards microautophagy to avoid DNAJC5-associated lipofuscinosis and neurodegeneration. Abbreviations: 3-MA: 3-methyladenine; ACTB: actin beta; AFSM: autofluorescent storage materials; ANCL: adult neuronal ceroid lipofuscinosis; Baf. A1: bafilomycin A1; CLN: ceroid lipofuscinosis neuronal; CLU: clusterin; CS: cysteine string domain of DNAJC5/CSPα; CUPS: compartment for unconventional protein secretion; DN: dominant negative; DNAJC5/CSPα: DnaJ heat shock protein family (Hsp40) member C5; eMI: endosomal microautophagy; ESCRT: endosomal sorting complex required for transport; GFP: green fluorescent protein; HSPA8/HSC70: heat shock protein family A (Hsp70) member 8; INCL: infant neuronal ceroid lipofuscinosis; JNCL: juvenile neuronal ceroid lipofuscinosis; KO: knockout; LAMP1: lysosomal associated membrane protein 1; LAPTM4B: lysosomal protein transmembrane 4 beta; LN: linker domain of DNAJC5/CSPα; MAPS: misfolding-associated protein secretion; mCh/Ch: mCherry; mCi/Ci: mCitrine; MTOR: mechanistic target of rapamycin kinase; NCL: neuronal ceroid lipofuscinosis; PPT1: palmitoyl-protein thioesterase 1; PQC: protein quality control; SBP: streptavidin binding protein; SGT: small glutamine-rich tetratricopeptide repeat; shRNA: short hairpin RNA; SLC3A2/CD98hc: solute carrier family 3 member 2; SNCA/α-synuclein: synuclein alpha; TMED10: transmembrane p24 trafficking protein 10; UV: ultraviolet; VPS4: vacuolar protein sorting 4 homolog; WT: wild type.


Introduction
Neuronal ceroid lipofuscinosis (NCL) refers to a family of genetically inherited neurodegenerative lysosomal storage diseases that are associated with excessive accumulation of lipopigments (lipofuscin), which occurs in both neurons and non-neuronal tissues [1,2]. A key characteristic of lipofuscin is its autofluorescence, which allows detection by fluorescence microscopy [3]. These diseases can occur in either infants, juveniles, or adults. The infantile and juvenile variants of NCL (INCL and JNCL) are often more severe, associated with progressive vision loss, seizure, and brain death. The adult variant (ANCL also named CLN4) on the other hand has milder symptoms. Nevertheless, after diagnosis, ANCL patients usually die after 10 years [4].
Mutations in a collection of CLN genes have been linked to NCL diseases [4][5][6]. Many CLN genes encode proteins essential for lysosomal functions such as lysosomal proteases (e.g. CTSD) [4] or regulators governing the trafficking of lysosome resident proteins (e.g. CLN6 and CLN8) [7,8]. These observations suggested that lipofuscin accumulation and neurodegeneration in NCL might result from disrupted lysosome homeostasis.
Two ANCL-associated mutations have been reported [37][38][39], which are located next to each other in the CS domain, close to the interface of the upstream LN domain whose function is unknown. Recent studies suggest that these mutations reduce DNAJC5 palmitoylation while increasing its aggregation propensity [15,[40][41][42]. These mutations also cause mislocalization of the affected protein in cells [41]. Accordingly, the ANCL-causing DNAJC5 mutations are thought to reduce the DNAJC5 chaperoning function [42]. However, this notion does not explain why cells carrying ANCL-associated DNAJC5 mutations accumulate lipopigments in lysosomes.
In this study, we found that DNAJC5 used a J-domain independent activity to couple two protein quality control (PQC) processes: ESCRT-dependent endosomal microautophagy (eMI) and misfolding-associated protein secretion (MAPS). We identified SLC3A2/CD98hc as a DNAJC5 interactor that is critical for MAPS, but dispensable for microautophagy. Importantly, reducing SLC3A2-mediated protein secretion or expression of ACNL-associated DNAJC5 mutants uncouples these processes, resulting in lipofuscinosis and neurodegeneration.

J-domain independent translocation of DNAJC5 into the lumen of endolysosomes
If DNAJC5 chaperones misfolded proteins for secretion using endolysosomes as a secretory compartment, DNAJC5 should accompany cargos to the lumen of endolysosomes. To test this idea, we established a Keima-based fluorescence assay. Keima is a monomeric dual-excitable fluorescence protein (λ em max ~620 nm) with maximum excitation at 440 nm under neutral pH (6.0-8.0) or at 550 nm in an acidic environment (pH<6.0) ( Figure 1A) [43]. When fused to a cargo, Keima allows quantitative measurement of cargo trafficking to acidic endolysosomes by fluorescence microscopy or flow cytometry [44].
Confocal microscopy showed that HEK293T cells stably expressing Keima displayed only neutral cytoplasmic fluorescence. By contrast, in Keima-DNAJC5 expressing cells, we not only detected neutral Keima-DNAJC5 signals on the plasma membrane but also observed Keima-DNAJC5 on intracellular vesicles by excitation at either 440 nm or 550 nm ( Figure 1B). Thus, these vesicles are endolysosomes containing Keima-DNAJC5 both on the surface and in the lumen. Consistently, flow cytometry revealed an increased ratio of acidic to neutral Keima-DNAJC5 fluorescence when compared to Keima, and this phenotype was independent of the Kema-DNAJC5 expression level ( Figure 1C). As anticipated, treatment with bafilomycin A 1 (Baf. A1), a lysosome acidification blocker, dramatically reduced the acidic Keima-DNAJC5 signal while increasing the neutral Keima-DNAJC5 signal ( Figure 1D). Thus, DNAJC5 can enter the lumen of endolysosomes.
To test if endosomal translocation of DNAJC5 requires the J-domain, we measured the ratio of the acidic vs. neutral fluorescence in cells stably expressing either WT Keima-DNAJC5 or a Keima-DNAJC5 mutant lacking the J-domain (ΔJ). Intriguingly, deleting the J-domain further enhanced the endolysosomal translocation of DNAJC5 ( Figure 1E). This phenotype is likely achieved via a gain-of-function activity because the DNAJC5 ΔJ mutant was also more active in promoting MAPS (see below).
To confirm the endolysosomal translocation of DNAJC5 in neurons, we infected mouse primary neurons with lentiviruses expressing either Keima, Keima-DNAJC5, or Keima-DNAJC5 ΔJ under the SYN1 (synapsin1) promoter on DIV 3 (Days in vitro). The SYN1 promoter has been widely used to achieve neuronal specific gene expression at low levels. At DIV 10, we examined cells by confocal microscopy ( Figure 1F). Interestingly, some acidic Keima punctae were observed in Keima-expressing neurons, probably due to macroautophagy/ autophagy. Similar to non-neuronal cells, the acidic to neutral Keima ratio was increased in cells expressing Keima-DNAJC5 compared to Keima-expressing cells. Furthermore, neurons expressing Keima-DNAJC5 ΔJ had the highest acidic to neutral Keima ratio ( Figure 1F,G). As expected, Baf. A1 treatment diminished all acidic Keima signals ( Figure 1F,G). Collectively, these results demonstrate the translocation of DNAJC5 into the lumen of endolysosomes in neurons.

DNAJC5 promotes endolysosomal translocation of misfolded proteins via microautophagy
To test whether DNAJC5 could chaperone misfolded proteins into endolysosomes, we used a truncated GFP protein (GFP1-10) tagged with mCherry (mCh-GFP1-10) as a model substrate. We previously developed a photobleaching-based imaging assay, which allowed detection of a small fraction of mCh-GFP1-10 that are associated with endolysosomes in live cells because photobleaching reduced the cytosolic mCh-GFP1-10 background [35]. Expression of DNAJC5 increased the endolysosome association of mCh-GFP1-10, which was further enhanced in DNAJC5 ΔJ-expressing cells (Figure 2A, panels ii, iii vs. i; Figure 2B). As in WT DNAJC5-expressing cells, the mCh-GFP1-10-containing vesicles in DNAJC5 ΔJ-expressing cells also contain DNAJC5 ΔJ and the late endosomal protein RAB9 (Figure 2A, panels iv-ix). These results suggest that DNAJC5 recruits mCh-GFP1-10 to endolysosomes in a J-domain independent manner.
We next tested whether DNAJC5 could promote the translocation of misfolded proteins into the endolysosomal lumen using cells stably expressing Keima-SNCA, which is localized mostly in the cytoplasm with a few acidic Keima punctae, as demonstrated by confocal microscopy ( Figure 2D). When DNAJC5 WT or ΔJ was co-expressed, however, the number of acidic Keima-SNCA punctae was significantly increased ( Figure 2D,E).

Endolysosomal translocation of misfolded proteins is dispensable for MAPS
We next asked whether DNAJC5-mediated microautophagy is required for MAPS. To this end, conditioned medium and cell lysate from HEK293T cells transfected with the MAPS substrate GFP1-10 [35] were analyzed by immunoblotting, which detected a fraction of GFP1-10 but not abundant cytosolic proteins such as HSPA8 and HSP90 in the medium ( Figure 2G and Figure S1B). Consistent with previous reports, co-expression of DNAJC5 enhanced GFP1-10 secretion [19,36]. Strikingly, the MAPS-stimulating activity of DNAJC5 ΔJ was almost 10-fold higher than that of WT DNAJC5 ( Figure 2G and Figure S1B). A similar observation was made for SNCA ( Figure S1C). Notably, a fraction of WT DNAJC5 was also secreted, and the level of secreted DNAJC5 ΔJ was much higher than that of WT DNAJC5 ( Figure 2G and Figure S1C). The correlation between endolysosomal translocation of DNAJC5 and its MAPS-stimulating activity appears to confirm endolysosomes as the secretory compartment for MAPS. However, when we examined the secretion of GFP1-10 in the presence of VPS4 DN, surprisingly, VPS4 DN dramatically increased GFP1-10 secretion under both basal and DNAJC5-overexpressing conditions ( Figure 2H,I). These results suggest that DNAJC5-mediated microautophagy and MAPS are two parallel but functionally coupled quality control processes.

ANCL mutations inhibit MAPS without affecting DNAJC5 endolysosomal translocation
To understand how the above-mentioned DNAJC5 functions were affected by ANCL-associated mutations, we first tested whether DNAJC5 L115R and L116Δ mutants still promote endolysosome association of misfolded SNCA using mCh-SNCAexpressing cells. Like WT DNAJC5, both mCitrine (mCi)tagged DNAJC5 L115R and DNAJC5 L116Δ stimulated endolysosome association of mCh-SNCA ( Figure 3A,B). Likewise, when the endolysosomal translocation of Keima-SNCA was analyzed, the disease-associated mutants were as active as WT DNAJC5 in promoting acidic Keima-SNCA positive puncta ( Figure 3C,D). Importantly, when cells were transfected with Keima-tagged WT DNAJC5 or the ANCL-associated mutants, both Keima-DNAJC5 L115R and L116Δ were translocated into endolysosomes more efficiently than WT DNAJC5 ( Figure 3E, F). These results suggest that the ANCL mutations do not inhibit the microautophagy activity of DNAJC5.
We next examined whether these ANCL mutations alter the DNAJC5 function in MAPS. When the secretion of GFP1-10 was examined, unlike WT DNAJC5, both DNAJC5 L115R and DNAJC5 L116Δ failed to promote GFP1-10 secretion ( Figure 3G,H). Likewise, these ANCL-associated DNAJC5 mutants also failed to induce the secretion of SNCA ( Figure  S1D). Additionally, compared to WT DNAJC5, the secretion of the DNAJC5 L115R and L116Δ mutants was also consistently reduced ( Figure 3 G and I). Because these ANCL mutations significantly reduce DNAJC5 palmitoylation [42], and because deleting the CS domain or treating cells with a palmitoyltransferase inhibitor both inhibit the function of DNAJC5 in MAPS [19,51], it appears that DNAJC5 palmitoylation is essential for MAPS but dispensable for DNAJC5mediated microautophagy. Importantly, our data exclude endolysosomes as a secretory intermediate compartment for MAPS, at least under conditions tested.

A linker domain targets DNAJC5 to LAMP1-negative perinuclear vesicles
Having excluded endolysosome as a MAPS compartment, we re-characterized the subcellular localization of DNAJC5 using HEK293T cells bearing a GFP at the carboxyl terminus of endogenous DNAJC5 by 3D confocal microscopy ( Figure  S2A), which revealed a fraction of DNAJC5 in perinuclear vesicles largely negative for LAMP1 or the lysosome-specific  dye LysoTracker ( Figure 4A and Figure S2B). The perinuclear localization of DNAJC5 became more prominent in U2OS cells over-expressing WT mCi-DNAJC5 ( Figure 4B, panels ivi). By contrast, the ANCL-associated mutants were largely absent from this compartment. As a result, both DNAJC5 L115R and L116Δ showed increased colocalization with LAMP1 ( Figure 4B panels vii-xii).
To determine the domain(s) responsible for targeting DNAJC5 to the LAMP1 negative compartment, we characterized the localization of a set of DNAJC5 deletion mutants by confocal microscopy ( Figure 4C,D). As previously reported [11], a DNAJC5 mutant lacking the CS domain (ΔCS) was entirely localized to the cytosol due to defects in both membrane binding and palmitoylation ( Figure S2C,D). By contrast, Ci-DNAJC5 ΔNT ( Figure S2C,D) and ΔJ ( Figure 4B, panels xiii-xv; Figure 4C) were localized to both LAMP1-positive and LAMP1-negative vesicles similarly as WT DNAJC5, suggesting that HSPA8-binding is dispensable for DNAJC5 membrane association. Intriguingly, a Ci-DNAJC5 mutant lacking the linker (ΔLN) remained associated with LAMP1-positive vesicles but was largely absent from perinuclear vesicles ( Figure 4B, panels xvi-xviii, Figure 4D). Keima-DNAJC5 ΔLN also showed increased endolysosomal translocation ( Figure 3F). The similar localization of ΔLN and ANCL mutants (table S1) prompted us to test the secretion of DNAJC5 ΔLN by immunoblotting, which showed markedly reduced level of secretion similarly as DNAJC5 ΔCS ( Figure 4E,F). These results suggest that the LN and CS domain act together to confer the perinuclear DNAJC5 localization, which is essential for DNAJC5-mediated MAPS.

DNAJC5 interacts with SLC3A2 via the linker domain
To identify factor(s) that regulate the association of DNAJC5 with the LAMP1-negative compartment, we performed tandem affinity purification using cells stably expressing DNAJC5 ΔJ bearing FLAG-and SBP (Streptavidin binding protein) tags ( Figure S3A,B). We used DNAJC5 ΔJ because of its strong MAPS-stimulating activity and also to avoid HSPA8 and its associated proteins. Mass spectrometry analyses identified SLC3A2/CD98hc, a cofactor of heterodimeric amino acid transporters [52], as a potential DNAJC5 binding partner ( Figure 5A and Figure S3C,D). Additional co-immunoprecipitation showed that endogenous SLC3A2 interacts with not only DNAJC5 ΔJ but also WT DNAJC5, although WT DNAJC5 had a lower affinity to SLC3A2 than DNAJC5 ΔJ ( Figure 5B). Reciprocal pulldown using a cell line bearing a GFP tag at the endogenous SLC3A2 locus further confirmed the interaction ( Figure 5C and Figure S3E). Importantly, coimmunoprecipitation detected an interaction between endogenous DNAJC5 and SLC3A2 in mouse primary neurons ( Figure S3F). Furthermore, when we transfected DNAJC5 variants into SLC3A2::GFP cells and performed GFP pulldown, DNAJC5 ΔLN and DNAJC5 ΔCS showed significantly reduced binding to SLC3A2 compared to other variants ( Figure 5C).
To further confirm the interaction between SLC3A2 and DNAJC5, we examined the localization of SLC3A2::GFP by confocal microscopy. In addition to the plasma membrane, SLC3A2-GFP was detected on intracellular vesicles clustered around the nucleus ( Figure S4A), a pattern similar to that of endogenous DNAJC5. Like DNAJC5, SLC3A2-GFP was localized to LAMP1-and LysoTracker-positive vesicles, but many SLC3A2-positive signals around the nucleus were free of LAMP1 ( Figure S4B to B"). As anticipated, cells expressing GFP and mCherry on endogenous DNAJC5 and SLC3A2, respectively, showed extensive colocalization of these proteins on vesicles ( Figure 5D,E; movie S1). Altogether, these results demonstrate a specific interaction between SLC3A2 and DNAJC5, which requires the DNAJC5 linker (LN) and CS domains.

SLC3A2 is required for the perinuclear association of DNAJC5 and MAPS
The observation that DNAJC5 ΔLN, defective in SLC3A2 interaction, is also absent from the LAMP1-negative perinuclear compartment ( Figure 4B) suggests SLC3A2 as a potential regulator of DNAJC5 localization. To test this idea, we immunostained endogenous DNAJC5 and LAMP1 in SLC3A2depleted or control knockout cells. In control cells, the presence of DNAJC5 in the LAMP1 negative compartment was readily visible. By contrast, in SLC3A2 KO cells, DNAJC5 was almost completely absent from this compartment (Figure 5F-H; Figure S5A,B), while its colocalization with LAMP1 became more prominent. By contrast, the plasma membraneassociated DNAJC5 was not affected in SLC3A2-depleted cells ( Figure S5A,B). Flow cytometry showed that SLC3A2 deficiency or overexpression did not significantly affect the endolysosomal translocation of Keima-DNAJC5 ( Figure S5C,D). Thus, SLC3A2 is specifically required for the perinuclear association of DNAJC5 but dispensable for its endolysosomal translocation.
We next tested whether SLC3A2 is required for DNAJC5mediated MAPS using SNCA and GFP1-10 as model substrates. Indeed, the secretion of these proteins under basal conditions or in DNAJC5-overexpressing cells was significantly reduced when SLC3A2 was knocked down by 60-   Figure S6D). These results suggest a role for SLC3A2 in DNAJC5-mediated MAPS.
The link of the perinuclear DNAJC5 to unconventional protein secretion prompted us to investigate whether DNAJC5 is colocalized with TMED10, a recently identified unconventional protein secretion regulator that is localized to ERGIC vesicles [53]. We endogenously tagged DNAJC5 with GFP and TMED10 with mCherry. Dual-color fluorescence microscopy showed that the perinuclear DNAJC5 is not colocalized with TMED10 in U2OS cells ( Figure S5J).

Depletion of SLC3A2 causes lipofuscin-like structures in mammalian cells
Given the link between DNAJC5 and lipofuscinosis, we determined whether SLC3A2 deficiency caused accumulation of lipofuscin-like autofluorescent storage materials (AFSMs).
Indeed, confocal microscopy analyses of more than 200 cells showed that ~19% of SLC3A2 KO HEK293T cells contained one or two large spherical puncta detectable by excitation with a UV light ( Figure 6F and Figure S6A). By contrast, AFSMs were only detected in less than 0.5% of WT cells. Importantly, AFSMs were barely detected when SLC3A2 was re-expressed in the KO cells, suggesting that the AFSM accumulation was caused by SLC3A2 depletion ( Figure 6F). Knockdown of Slc3a2 in mouse primary neurons also caused AFSM accumulation ( Figure S6B-D). 3D confocal analysis combined with dual-color fluorescence microscopy showed that AFSMs in SLC3A2 knockout cells were surrounded by membranes enriched for DNAJC5 and RAB9, consistent with an origin from endolysosomes ( Figure 6G and Figure S6E). Additional immunostaining detected Saposin A1, a known marker of lipofuscin [54] in all AFSMs observed in SLC3A2 KO cells ( Figure 6H), confirming their identity as lipofuscin. Additionally, cells derived from late INCL patients bearing CLN2 mutations also contained large lipofuscin-   like AFSMs that were wrapped by DNAJC5-positive membranes ( Figure 6I) and morphologically indistinguishable from those in SLC3A2 KO cells [55]. Interestingly, cells overexpressing DNAJC5 ΔLN, L115R or L116Δ also contained an increased number of Saposin A1-positive AFSMs ( Figure 6J), although they were smaller in size compared to those in SLC3A2 KO cells, possibly because of the short expression time. Why knockdown of SLC3A2 only generates AMSFs in a fraction of the cells remains to be elucidated. However, because CLN2 encodes a lysosomal peptidase, our findings suggest that lipofuscin biogenesis is associated with either defective lysosomal degradation or excessive flow of proteins and membranes into endolysosomes (see discussion).

AFSM accumulation and SLC3A2 deficiency contribute to neurodegeneration in a fly ANCL model
We used a recently established fly model to further evaluate the role of DNAJC5 and SLC3A2 in neurodegeneration. We expressed WT human DNAJC5 (HsDNAJC5) or the ANCLassociated DNAJC5 L116Δ mutant in photoreceptor cells of Drosophila larval eyes using the GMR-Gal4 driver. As reported previously [41], DNAJC5 L116Δ overexpression caused massive neuronal cell death, resulting in a severe rough eye phenotype in adult flies raised at 25°C ( Figure 7A). By contrast, WT DNAJC5 did not change the eye morphology ( Figure 7A). Interestingly, confocal microscopy analyses of larval eye discs showed that DNAJC5 L116Δ -expressing tissues contained many autofluorescent punctae in apical cytoplasm within photoreceptor cells ( Figure 7B-D; Figure S7A).
Our data suggests that SLC3A2 regulates the DNAJC5 localization and MAPS. If disruption of MAPS while maintaining DNAJC5-mediated microautophagy contributes to AFSM production and neurodegeneration, knockdown of SLC3A2 should enhance DNAJC5-induced neurodegeneration. We, therefore, crossed GMR>HsDNAJC5 WT and GMR>HsDNAJC5 L116Δ flies to a strain bearing a SLC3A2targeting shRNA downstream of the UAS regulatory element, or as controls to strains bearing either a luciferase shRNA or mCherry shRNA. Expression of SLC3A2 shRNA resulted in ~60% depletion of endogenous SLC3A2 mRNA ( Figure S7B). When raised at 28°C, flies expressing WT HsDNAJC5 together with a control shRNA suffered a modest rough eye phenotype, but no significant pigment loss ( Figure 7E' and E"; Figure  S7C,D). By contrast, although flies expressing SLC3A2 shRNA alone had normal eyes ( Figure 7E and Figure S7E), flies expressing WT HsDNAJC5 together with SLC3A2 shRNA had more severe rough eyes ( Figure 7E"') with a significant loss of pigments ( Figure S7C,D). Likewise, knockdown of SLC3A2 also enhanced the rough eye phenotype in hDNAJC5 L116Δ -expressing flies raised at 25°C (Figure S7E,F). Thus, depletion of SLC3A2 enhances neuronal cell death induced by DNAJC5 overexpression, probably because it disrupts the balance between DNAJC5-mediated MAPS and microautophagy, causing a disproportionally increase in the flow of misfolded proteins and membranes to endolysosomes.

Discussion
Our study fails to establish endolysosome as an unconventional secretion compartment for MAPS. Instead, we find that DNAJC5-mediated MAPS and eMI are two tightly coupled PQC mechanisms essential for endolysosomal homeostasis. ANCL-associated DNAJC5 mutations abolish DNAJC5 function in MAPS but maintain its microautophagy-stimulating activity ( Figure 7F), causing lipofuscinosis and neurodegeneration. Specifically, we show that a fraction of DNAJC5 can be efficiently translocated into endolysosomes together with its clients. It was reported previously that selective eMI is mediated by HSPA8 and ESCRT proteins [47,48,50]. Although DNAJC5-mediated microautophagy is not dependent on its HSPA8-interacting J-domain, whether HSPA8 or other DNAJC5-associated chaperones such as SGT [14] function in substrate recruitment in this process remains to be tested. Intriguingly, recent studies showed that ANCLassociated mutations reduce DNAJC5 palmitoylation [42]. These mutants are also more prone to aggregation [40,41,56]. Nevertheless, these mutants are more efficiently translocated into endolysosomes compared to WT DNAJC5, suggesting that DNAJC5 palmitoylation is not essential for microautophagy. Indeed, it is known that DNAJC5 palmitoylation defective mutants can still bind to membranes in cells [11].
A fraction of DNAJC5 is also localized to a perinuclear membrane compartment, which is largely free of LAMP1 and not stained well by a LysoTracker dye. However, this compartment can be weakly labeled by LysoTracker after prolonged staining (movie S2), suggesting that it might be a prelysosomal compartment. The localization of DNAJC5 to this compartment does not require its J-domain but depends on palmitoylation of DNAJC5 since both the ANCL-associated disease mutants (this study) and a CS-domain deleted mutant fail to associate with this compartment [19]. Additionally, we identified a DNAJC5 binding partner named SLC3A2, which regulates the peri-nuclear localization of DNAJC5. SLC3A2 is a type II membrane glycoprotein capable of interacting with six light chain molecules, forming a set of amino acid transporters on the plasma membrane [52]. Moreover, additional adaptors such as LAPTM4B can retain a SLC3A2-containing transporter in endolysosomes, regulating the amino acid balance between endolysosome and cytoplasm and thus the MTOR signaling [57]. Whether SLC3A2-dependent regulation of the DNAJC5 localization requires additional components awaits further characterization.
Although the DNAJC5-associated pre-lysosomal compartment does not overlap with TMED10, a cis-Golgi protein recently implicated in unconventional protein secretion, several lines of evidence suggest that this compartment functions in MAPS. First, DNAJC5 mutants defective in binding to this compartment all fail to promote MAPS. Additionally, knockout of SLC3A2 impairs the association of DNAJC5 to this compartment, which also reduces the secretion of misfolded proteins. In S. Cerevisiae, unconventional protein secretion under stress conditions is mediated by a Golgi-derived compartment termed CUPS [58,59]. The pre-lysosomal DNAJC5-positive compartment may be functionally analogous to CUPS. In yeast, protein translocation into CUPS is thought to be mediated by the ESCRT machinery and some autophagy regulators [59]. By contrast, unconventional secretion of misfolded proteins in mammalian cells is not dependent on ESCRT and cannot be blocked by the autophagy inhibitor 3-MA [35]. Noticeably, previous studies have suggested a role for HSPA8 in USP19-stimulated MAPS [19,36]. The fact that neither the peri-nuclear localization of DNAJC5 nor its MAPS-stimulation activity depends on the J-domain raises the possibility that HSPA8 may collaborate with USP19 directly during substrate recruitment considering the reported interaction of USP19 with HSPA8 [60]. How misfolded proteins enter the DNAJC5-associated perinuclear compartment and the role of HSPA8 in this process remainto be elucidated.
DNAJC5-mediated microautophagy and MAPS appear to operate in parallel to maintain protein homeostasis. Conceivably, a potential benefit from these coupled PQC processes is to prevent overflow of toxic materials to endolysosomes. These processes, when properly tuned, should reduce misfolded proteins and improve cell homeostasis. By contrast, deregulation of these processes may lead to the accumulation of misfolded proteins in either endolysosomes or the cell exterior. The fact that DNAJC5 lacking the J-domain has significantly increased activities in both endolysosomal translocation and protein secretion suggests an autoinhibitory mechanism that tightly controls these processes. Consistently, structural studies have illustrated phosphorylation-dependent conformational changes, which disrupt a J-domain-mediated intermolecular interaction [61].
Our findings suggest that ANCL-associated DNAJC5 mutations cannot be classified as simple loss-or gain-offunction alleles. Instead, while these mutations abolish the pre-lysosomal localization of DNAJC5 and the corresponding MAPS function, they increase the translocation of DNAJC5 into lysosomes. One presumed consequence is the abnormal flow of misfolded proteins and membranes into endolysosomes, damaging this organelle over time and resulting in undigested remnants in the form of lipofuscin. This model is consistent with the finding that mutations in PPT1 (palmitoyl protein thioesterase 1) also result in similar disease phenotypes [62]. Presumably, endolysosomeassociated DNAJC5 is processed by PPT1, which may regulate its function in proteostasis regulation. Additionally, recent studies showed that ANCL-associated DNAJC5 mutants, albeit lacking MAPS activity, can still promote exosome biogenesis [51,63], further consolidating the proposed link between gain of DNAJC5 function in eMI and ANCL.
Endogenous tagging was performed as described previously [66] (http://www.pcr-tagging.com). Briefly, the PCR cassettes were amplified from pMaCTag-05 plasmid (a gift from Michael Knop; Addgene, 119,984) by the AccuPrime™ Pfx DNA polymerase (Invitrogen, 12,344) using the primers The PCR products were gel purified using a QIAGEN Gel Extraction Kit. HEK293T cells were transiently transfected with 1 µg of a PCR cassette and 1 µg of pCAG-enAsCas12a (E174R/S542R/K548R)-NLS(nuc)-3xHA (a gift from Keith Joung & Benjamin Kleinstiver; Addgene, 107,941) using TransIT293 (Mirus) according to the manufacturer<apos;>s protocol and GFP-or mCherry-positive cells were sorted two weeks later by FACS. For making double-tagging cells (DNAJC5::GFP SLC3A2::mCherry), The PCR cassette for SLC3A2 described above was transfected to DNAJC5::GFP cell and cells showing double fluorescences were sorted by FACS. Endogenous tagged cell lines used in this study were all validated by immunoblottings ( Figure S2A and Figure S3E). All expression constructs, si-RNAs, chemicals, and antibodies are listed in Table 1.

Primary mouse neuron cultures
Primary cortical or hippocampal neuron cultures were prepared from P0-1 murine pups following an animal study protocol approved by NIDDK. Hippocampi were dissected in Hanks' balanced salt solution (HBSS) and washed with MEM (Thermo, 11,095,080) twice. The hippocampi were incubated with 0.05% trypsin-EDTA (Thermo, 25,300,062) containing 100 µg/mL DNase-I (Sigma, 10,104,159,001) at 37°C for 10 min. Trypsin was then inactivated by addition of MEM containing 10% FBS. After washing with MEM three times, the tissues were dissociated by pipetting several times in MEM containing 100 µg/mL DNase-I. Cells were then centrifuged at 300 x g for 5 min and resuspended in the plating medium (MEM containing 10% FBS, 1 mM sodium pyruvate (Sigma, S8636), 2 mM L-glutamine (Sigma, 59202C), 100 µg/mL primocin (Invitrogen, ant-pm-05) and 0.6% glucose). The cell solution was passed through a 70-µm strainer (VWR, 76,327-100) once to filter out any cell clumps. Cells were seeded into plates or µ-slide 4 well chamber with glass bottom (ibidi, 80,427) pre-coated with 50 µg/ml poly-D-lysine (Sigma, P7280) and 2 µg/mL laminin (Sigma, L2020). Generally, we seed hippocampal cells obtained from a pup to cover 4 cm 2 or 8 cm 2 of surface area for Immunoblotting and imaging purpose, respectively. We seed cortical cells obtained from a pup to cover 15 cm 2 surface area for immunoblotting. After incubation at 37 °C with 5% CO 2 for overnight, the culture medium was changed to the Neurobasal media (Thermo, 21,103,049) containing 2% B27 (Thermo, 17,504,044), 0.5 mM L-glutamine and 100 µg/mL primocin to support the growth of hippocampal neurons. The culture media was replaced with fresh media once a week or when needed.

Protein samples preparation and Immunoblotting
Conditioned media and cell lysates were prepared as described previously [19,35]. Briefly, cells (2.5 × 10 5 ) seeded in a poly-D-lysine-coated 12-well plate were grown for 24 h, and then transfected with 250 ng plasmids expressing the indicated MAPS substrates together with 250 ng plasmids expressing a MAPS regulator (e.g., DNAJC5). We replaced the medium with 1.5 mL fresh DMEM medium 24 h posttransfection. Cells were grown for another 16 h before conditioned media were collected. The media were subjected to sequential centrifugation, first at 1000 × g for 5 min to remove contaminated cells, and then at 10,000 × g for 30   figurelegends, protein secretion is normalized by the expressed protein in cell lysates. Immunoblottings were performed using the standard protocols. All primary antibodies were diluted in 5% BSA (Sigma, A9418) in phosphatebuffered saline (PBS; Corning, 21-040) as described in Table 1. To quantify secreted protein in conditioned media, HRP-conjμated secondary antibodies were used. Immunoblotting signal was detected by the enhanced chemiluminescence method (ECL) and recorded by a Fuji LAS-4000 imager or Bio-Rad Chemidoc. The intensity of the detected protein bands was quantified by ImageGauge v3.0, Bio-Rad ImageLab, or ImageJ. Protein secretion levels were determined by normalizing the level of the secreted proteins by the amount of the same protein in cell lysates. For other immunoblotting quantifications, fluorescently labeled secondary antibodies were used. Immunoblots were scanned by a LI-COR Odyssey scanner or Bio-Rad Chemidoc.

Immunoprecipitation
For co-immunoprecipitation assays, pre-equilibrated FLAG M2 agarose beads (Sigma, A2220) or GFP-Trap beads (Chromotek, gta-20) were incubated with lysates containing tagged proteins for 1 h at 4°C. The beads were washed with the lysis buffer, and the bound proteins were eluted in 1x Laemmli buffer at 95°C and resolved by SDS-PAGE for immunoblotting analysis. For co-IP of endogenous DNAJC5 with SLC3A2 from neurons, lysates of cortical neurons from 3 wells of a 6 well plate were combined. Cell lysates were prepared in 1.5 mL volume using either CHAPS lysis buffer (1% CHAPS [Sigma, 10,810,118,001], 50 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM DTT, protease inhibitors) or NP40 lysis buffer. The lysates were pre-cleared with 75 ul of protein A agarose (Sigma, IP02) equilibrated with 3% BSA, for 4 h at 4°C. After centrifugation at 2,500 x g for 2 min, 45 ul of lysates was saved as input and the remaining samples were used for immunoprecipitation with 2.5 µg of rabbit IgG or SLC3A2 antibody (Thermo, PA596401) overnight at 4°C. The immunocomplexes were affinity isolated using 20 μL of preequilibrated protein A beads for 2 h at 4°C, followed by washing in the same lysis buffer. The immunoprecipitates were eluted by 1 x Laemmli sample buffer, heat-denatured at 95°C for 5 min before SDS-PAGE electrophoresis and immunoblotting. To avoid IgG bands on the IP blot, protein G-HRP (abcam) was used as secondary antibody.

Tandem affinity purification and mass spectrometry of FLAG-SBP-DNAJC5 precipitates
A schematic flow chart of the purification procedure is presented in Figure S3A. Once membrane permeabilization is confirmed by trypan blue staining, the lysates were spun at 16,000 x g for 5 min. The resulting membrane pellets were washed with 1x PB buffer followed by centrifugation. The washed pellets were resuspended in 0.33% formaldehyde in PB buffer and incubated at 37°C for 25 min with rocking for crosslinking. After top-speed centrifugation for 5 min, the pellets were further lysed in 8 mL of RIPA lysis buffer (50 mM Tris pH7.5, 1% Nonidet P-40, 0.1% SDS, 2 mM EDTA pH8.0, 0.5% sodium deoxycholate [Sigma, D6750], 150 mM NaCl) with 1 mM DTT and a protease inhibitor cocktail for 1 h at 4°C. After centrifugation at 16,000 x g for 5 min, the supernatants were collected for the following tandem affinity purification steps.
To purify DNAJC5 and its interacting proteins, the RIPAsoluble fractions were incubated with 200 μl of pre-washed FLAG M2 agarose bead slurry with rocking for 1 h at 4°C. The beads were washed with 1 mL of RIPA buffer three times and 1 mL of streptavidin binding buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 M NaCl) once. After spinning down at 500 x g for 5 min, the bound proteins were eluted by incubation with 250 μl of 3 x FLAG peptide solution (200 μg/mL; Sigma, F4799) for 15-20 min twice. After . Eluted peptides were subjected to electrospray ionization and then injected into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo, FSN05-10,001). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein databases with the acquired fragmentation pattern by the software program, Sequest (Thermo). All databases include a reversed version of all the sequences and the data was filtered to between a one and two percent peptide false discovery rate (FDR).

Imaging analysis
For immunostaining, cells cultured on #1.5 thickness coverslips precoated with poly-D-lysine ( Photobleaching experiments were performed as described previously [35]. Cell imaging experiments were conducted on either a LSM780 laser scanning confocal microscopy (Zeiss) or a SoRa spinning disk super-resolution microscopy (Nikon) equipped with heating and a CO 2 incubation system. Images were further processed for dot number counting or fluorescence ratio determination using ImageJ. For colocalization assay, images were denoised using Nikon NIS-element plugin Denoise.ai and the colocalization tools in the NISelements software (Nikon), colocalization tools in the Zeiss Zen (black) software, or JACoP plugin in ImageJ were used.

Fly experiments
Fly strains bearing shRNA-expressing cassettes downstream of UAS sequences are 31,603, 35,785, 57,746 from the Bloomington Drosophila Stock Center. The flies expressing either WT human DNAJC5 or DNAJC5 L116Δ were described previously [41]. Unless otherwise specified, cultures were maintained in 25°C incubators on BDSC cornmeal food (Homemade based on the recipe from Genesee Scientific). For imaging fly eyes, 5-10 adult flies were fixed in PBS containing 4% formaldehyde for 1 h, rinsed with PBS. The flies were then dehydrated by soaking sequentially in 30%, 50%, 70%, 90%, and 100% ethanol. Dried flies were mounted in an Ibidi imaging chamber and scanned by a Nikon C1 spinning disk confocal microscope using Ex 488 /Em 520 nm. Shown is the maximum projection view of the scanned Z-section images. Two methods were used to score the eye phenotypes, which generate similar results. One analysis done in the Zinsmaier lab was as described previously [41]. Briefly, eye phenotypes of 1 -to 4-day-old adults were digitally imaged using a Nikon stereomicroscope. A semi-quantitative assessment of the eye phenotypes was achieved by serially coding the obtained images and blind scoring using naïve researchers. Eye phenotypes were given a relative score in comparison to known scoring classes: (0) normal WT-like eye. (1) Mild: slightly 'rough' eye surface and/or slight dis-colorization; (3) Moderate: significant dis-colorization and rough eye surface, slightly disorganized ommatidia and/or reduced size; (5) Severe: Discolored and deformed eye due to loss of or malformed ommatidia, significant reduction in eye size. Scores were collected for at least 5 individual flies per genotype derived from at least 4 individual crosses. Alternatively, eye images were processed by ImageJ to measure the size of the area that have lost the eye pigments.
To detect AFSM in photoreceptor cells, imaginal eye discs were dissected from third instar larvae, fixed in 4% formaldehyde in PBS for 20 min at room temperature. Eye discs were washed three times with PBS, then permeabilized in a PBSbased staining solution containing 0.2% saponin and 10% FBS for 10 min. The discs were then stained by Alexa Fluor 594labeled phalloidin (Thermo, A12381) in the staining solution for 30 min at room temperature. The discs were washed twice by PBS and then imaged by a Zeiss LSM780 laser scanning confocal microscope.

Statistical analysis
All experiments were repeated two or more times. For statistical analyses, at least three independent experiments were carried out. For quantification of microscope images, at least 30 randomly selected cells or at least 10 randomly selected fields were analyzed. Statistical analysis for experiments with two treatment groups used one or two-tailed t-test. For more than two groups, we used one-way ANOVA followed by multiple comparison analyses of variance by the Dunnett<apos;>s test where all groups were compared back to a single control group. Tukey<apos;>s multiple tests were used where all groups were compared. Differences were considered significant at the 95% level of confidence. Details of statistical tests used, number of biological replicates (n), and P values for each experiment are included in figure legends. All statistical analyses and graphing were performed using GraphPad Prism 9.