PDZD8 binds Protrudin and is recruited to late endosomes
The ER protein, PDZD8, was recently identified as an ER- mitochondria membrane contact site protein, however, the molecular nature of this MCS is not known27. To address this and further investigate the mechanism of action of PDZD8, we identified interacting proteins by mass spectrometry based analysis of endogenous PDZD8, isolated from crosslinked human HCT116 cell extracts using validated anti-PDZD8 antibodies28. Mass spectrometry analysis of crosslinked endogenous PDZD8 immunopurifications revealed that the predominant interactor was Protrudin, an integral ER protein previously shown to reside at ER-late endosome MCSs22 (Fig. 1a, left panel; list provided in Supplementary Dataset 1). Consistently, mass spectrometry analysis of reciprocal immunopurifications of endogenous Protrudin from crosslinked cell extracts using a validated anti-Protrudin antibody revealed that its predominant interactor was PDZD8 (Fig. 1a, right panel; list provided in Supplementary Dataset 2). Immunoprecipitation and western analysis of extracts from HEK293 cells co-overexpressing PDZD8-GFP and Protrudin-mCherry also indicated that these components interact, which further validated the PDZD8-Protrudin interaction detected using antibodies directed against the endogenous tag-less versions of these proteins (Fig. 1b). Thus, our data indicate that PDZD8 and Protrudin interact in cells.
Published work indicates that Protrudin localizes to ER- late endosome MCSs primarily via its PIP lipid-binding FYVE domain, where it functions to facilitate the regulation of endosomal motility22,29. Given the stable interaction we observed between Protrudin and PDZD8, we considered whether PDZD8 was also recruited to endosomal MCSs. To address this question, we transiently transfected GFP-tagged PDZD8 and performed live-imaging of GFP-tagged PDZD8 (PDZD8-GFP) in human U2OS cells. For all live-cell imaging of transiently expressed tagged constructs, we transfected cells with the minimal amount of plasmid sufficient for microscopy detection to avoid protein overexpression artifacts. Under these conditions, we estimate that PDZD8-GFP was ~10-fold overexpressed compared to endogenous protein (Supplementary Fig. 1). PDZD8-GFP primarily localized diffusely in the ER (Fig. 1c; whole-cell projection), however, at a lower frequency, we also observed PDZD8-GFP localized at a significantly higher intensity to spherical structures, suggesting an enrichment of PDZD8 at specific ER subdomains (Fig. 1c; single plane image). We co-overexpressed PDZD8-GFP with mCherry tagged markers of early endosomes (Rab5), late endosomes (Rab7), and lysosomes (LAMP1) to test whether these PDZD8-enriched subdomains were associated with endosomes. PDZD8-GFP did not co-localize with either Rab5 or LAMP1-labeled endosomal structures (Fig. 2a). In contrast, in cells co-expressing PDZD8 and Rab7, we observed that PDZD8-GFP spheres co-localized with mCherry-Rab7-labeled endosomes and that there was a significant increase in the number of PDZD8-GFP labeled spheres per cell, as compared to the diffusely ER-localized and relatively rare PDZD8-GFP spheres observed in cells overexpressing PDZD8 alone (compare Fig. 1c with Fig. 2a, Supplementary Fig. 2, 2.5 ± 1.5 per cell compared to 65.7 ± 27.4 per cell, n = 20 cells). To test whether Rab7-colocalized PDZD8 spheres reflected the recruitment of PDZD8-GFP enriched ER co-localized with late endosomes, we imaged cells co-expressing PDZD8-GFP and mCherry-Rab7 and the general ER marker, BFP-Sec61β. Consistently, at regions of co-localization of PDZD8-GFP with mCherry-Rab7-labeled endosomes, we also observed BFP-Sec61β-labeled ER spheres tightly associated with late endosomes (Fig. 2b). Although the increased number of PDZD8-enriched ER subdomains associated with late endosomes was dependent on overexpressing both PDZD8 and Rab7 (Figs. 1c and 2a, and Supplementary Fig. 2), endogenous PDZD8 in U2OS cells, as detected by indirect immunofluorescence of PDZD8, using a validated anti-PDZD8 antibody28, was also observed in spherical structures, and in cells overexpressing Rab7-mCherry PDZD8- structures colocalized with Rab7-labeled endosomes (Fig. 2c and Supplementary Fig. 3). In addition, we observed that Rab7 peptides were significantly enriched in immunopurifications from crosslinked cellular fractions of endogenous PDZD8 and Protrudin, (Fig. 2d). These data suggest that ER-associated PDZD8 is recruited to late endosomes in a Rab7-dependent manner.
Rab proteins serve as molecular switches that are active while bound to GTP and are inactivated upon GTP hydrolysis to GDP. Activated GTP-bound Rab7 is recruited to the late endosome membrane and recruits additional proteins to the membrane that act as effectors and control endosomal maturation and motility9. We further tested the role of Rab7 in recruiting PDZD8-enriched ER domains to endosomes by assessing the guanine nucleotide specificity of Rab7 in PDZD8 endosome recruitment. ER-localized PDZD8 was recruited to endosomes in cells expressing the constitutively GTP-bound Rab7Q67L, but not the GDP-bound Rab7T22N (Fig. 2e). We also assessed the Rab7 guanine nucleotide specificity by examining the interaction between PDZD8 and Rab7 by immunoprecipitation in cells overexpressing PDZD8-GFP and different versions of Rab7: only Rab7 and the GTP-bound Rab7Q67L, and not the GDP-bound Rab7T22N, were detected by western analysis of anti-GFP immunoprecipitates from cell extracts, consistent with our cytological data (Fig. 2f). Together these data indicate that ER-localized PDZD8 is recruited to late endosomes in a manner dependent on GTP-activated Rab7. These observations are consistent with a recent study published while this work was in review reporting that PDZD8 and Rab7-GTP are enriched at ER-late endosome contact sites30.
Distinct PDZD8 domains bind Protrudin and late endosomes
We exploited the increased number of PDZD8-enriched ER subdomains associated with late endosomes under conditions of co-overexpression of PDZD8 and Rab7 to robustly determine the PDZD8 domains important for ER recruitment to endosomes. Given that PDZD8 is a multidomain containing protein, we generated and analyzed truncated versions of PDZD8-GFP with different domain compositions (Fig. 3a, schematic). As expected, expression of truncated constructs containing the single N-terminal transmembrane (TM) domain were localized to the ER in cells whereas constructs lacking the TM were localized to the cytosol (Fig. 3b, ΔCC:1-1000, ΔC1ΔCC:1-470 and TM + SMP:1-300 versus ΔTM:27-1154, ΔTMΔSMP:300-1154, C1CC:800-1154, SMP + PDZ:27-470, and SMP:90-300, respectively). However, in the absence of the TM, constructs containing the C1 and coiled-coil domains were also localized to spherical structures in cells, similar to those observed at low frequency in cells overexpressing full length PDZD8-GFP (Figs. 1c and 3b, ΔTM:27-1154, ΔTMΔSMP:300-1154, C1CC:800-1154). In contrast, constructs lacking the TM and the C1 and coiled-coil domains, were strictly localized in a diffuse manner in the cytosol (Fig. 3b, SMP + PDZ:27-470, SMP:90-300). Expression of either the C1:800-1000 or CC:1000-1154 regions was not detected in cells. Our observations suggest that the C1CC:800-1154 portion of PDZD8 is sufficient for PDZD8 recruitment to late endosomes. Consistent with this, we observed that PDZD8 C1CC:800-1154 co-localized with mCherry-Rab7-labeled endosomes in cells (Fig. 3b). In addition, the CC domain of PDZD8 was also necessary for the recruitment of PDZD8 to Rab7-labeled late endosomes, as a construct lacking the CC domain localized diffusely to the ER membranes in cells under conditions of Rab7 overexpression (Fig. 3b, ΔCC:1-1000). Consistent with this, we observed a decrease in the efficiency of co-immunoprecipitation of PDZD8ΔCC with Rab7Q67L by western blot analysis (Fig. 2f). Thus, our data suggest that recruitment of an ER PDZD8 subdomain to late endosomes is mediated via the PDZD8 C terminal C1/CC regions in a GTP-Rab7-dependent manner.
Our results indicate that endogenous PDZD8 and Protrudin interact (Fig. 1a). Protrudin has been previously shown to reside at ER-late endosome MCSs22. To test whether ER-localized PDZD8 and Protrudin are recruited together to late endosomes in Rab7-dependent manner, we co-expressed BFP-Rab7, PDZD8-GFP, and Protrudin-mCherry in cells. Consistently, both PDZD8-GFP and Protrudin-mCherry were co-localized with Rab7-labeled late endosomes (Fig. 4a). Thus, our results suggest that PDZD8, Protrudin, and Rab7 interact together at ER subdomains associated with late endosomes in cells. We tested which domains of PDZD8 are required for its interaction with Protrudin using anti-GFP immunoprecipitations from HEK293T cells co-expressing PDZD8-GFP constructs (Fig. 3a, schematic) and Protrudin-mCherry (Fig. 4b). Consistent with the observed reciprocal interactions of endogenous PDZD8 and Protrudin in immunoprecipitates (Fig. 1), full-length overexpressed PDZD8-GFP efficiently co-immunoprecipitated with Protrudin-Cherry (Fig. 4b, compare no PDZD8-GFP, lane 1 with (full length)-GFP, lane 2) from cell extracts. In contrast, PDZD8 constructs lacking the TM domain were not efficiently co-immunoprecipitated with Protrudin from cell extracts (Fig. 4b, compare (full length)-GFP, lane 2 with (27-1154)-GFP and (27-470-GFP), lanes 3 and 7, respectively). The PDZD8 construct containing only the N-terminal TM and SMP domains was also efficiently co-immunoprecipitated with Protrudin (Fig. 4b compare (full length)-GFP, lane 2, with (1-300)-GFP, lane 4). Given that this construct lacks the C1 and CC regions that were sufficient for the recruitment of PDZD8 to endosomes (Fig. 3b), this observation suggests that PDZD8 utilizes distinct domains for interacting with Protrudin and GTP-Rab7-labeled late endosomes. These data also suggest that the PDZD8-Rab7-endosome interaction occurs in a Protrudin-independent manner: the PDZD8 construct lacking the TM domain failed to co-immunoprecipitate with Protrudin, but retained its ability to be recruited in a Rab7-dependent manner to the late endosomal membrane (Figs. 3b and 4b, compare (full length)-GFP, lane 2, with (27-1154)-GFP, lane 3) and the PDZD8 1-1000 lacking the C-terminal CC domain efficiently co-immunoprecipitated with Protrudin, but failed to be recruited to endosomes in a Rab7-dependent manner (Figs. 3b and 4b, compare (full length)-GFP, lane 2, with (1-1000)-GFP, lane 6).
To more directly test whether PDZD8 and Protrudin are independently recruited to late endosomes, we examined the localization of Protrudin-GFP in a previously characterized HeLa PDZD8 knock-out (KO) cell line generated using CRISPR-CAS928. Consistent with independent targeting pathways, Protrudin-GFP was efficiently recruited to Rab7-labeled endosomes in PDZD8 KO cells (Fig. 4c, left top two panels). Consistent with published data22, in the absence of PDZD8, the recruitment of Protrudin to Rab7-late endosomes was primarily dependent on its PIP-binding FYVE domain, as Protrudin-FYVE4A-GFP, in which alanine replaced each of the four basic residues within the FYVE domain, localized diffusely to the ER in PDZD8 KO cells (Fig. 4c, right top two panels). This observation indicates that in PDZD8 KO cells, overexpression of Rab7 alone is not sufficient to recruit Protrudin-FYVE4A-GFP to late endosomes. However, in PDZD8 KO cells expressing BFP-Rab7 and PDZD8-mCherry, Protrudin-FYVE4A-GFP co-localized to Rab7-endosomes (Fig. 4c, compare right top two panels with right (+PDZD8 + Rab7) panel). This PDZD8-dependent mode of Protrudin-FYVE4A-GFP recruitment to BFP-Rab7-labeled endosomes required the PDZD8 CC region, consistent with our data indicating a role for CC domain in the recruitment of PDZD8 to late endosomes (Figs. 3b and 4c, compare right top two panels with right (+PDZD8ΔCC + Rab7) panel). We also observed that Protrudin-GFP was not as efficiently recruited to BFP-Rab7-labeled endosomes in PDZD8 KO cells expressing PDZD8 lacking the CC region, as Protrudin-GFP was localized diffusely to the ER (Fig. 4c, compare left top three panels with left (+PDZD8ΔCC) and (+PDZD8ΔCC + Rab7) panels). This observation is consistent with our immunoprecipitation data indicating that PDZD8 lacking the CC region retained its ability to interact with Protrudin at the ER membrane (Fig. 4b, lane 6). Collectively, these results support a model where, at the ER membrane, PDZD8 and Protrudin interact and influence each others recruitment to late endosomes, but each can also be independently targeted to ER-late endosomes subdomains via Rab7-GTP and PIP lipids, respectively.
Mitochondria are recruited to PDZD8-Rab7 contact sites
Membrane contact sites are unique cellular structures mediated by proteins in adjacent organelle membranes that tether organelles in close proximity (estimated 10–30 nm apart) to enable non-vesicular transfer of lipids and solutes. To test whether the PDZD8-GFP ER subdomains colocalized with Rab7-labeled endosomes are bona-fide membrane contact sites, we analyzed regions of co-localization at high resolution using correlated light and electron microscopy (CLEM). In areas where PDZD8-GFP and mCherry-Rab7 were co-localized, as inferred from the fluorescence signals, we observed regions of ER in contact with late endosomes, as evidenced by the presence of multivesicular bodies (Fig. 5a inset 1). In addition, the distance between the ER and late endosome membrane measured ~15 nm, characteristic of a membrane contact site (Fig. 5a, inset 1, arrow heads). Thus, our data indicate that the PDZD8-GFP ER subdomains associated with late Rab7-labeled endosomes represent ER-endosome membrane contact sites.
In many regions of PDZD8-GFP and mCherry-Rab7 co-localization, where we observed an ER-late endosome contact site, we also observed a mitochondrion (Fig.5a, inset 2), whose shape was co-altered with associated ER adjacent to an endosome (Fig. 5a, inset 3). This observation suggests that PDZD8-late endosomal MCSs may recruit mitochondria in cells. To further test this possibility, we assessed the proximity of mitochondria to Rab7-labeled endosomes in U2OS cells overexpressing PDZD8-GFP or PDZD8ΔCC-GFP, which either contain or lack PDZD8 ER subdomains localized to Rab7-labeled endosomes, respectively (Fig. 3b). Mitochondrial proximity was quantified by the number of overlaps observed between thresholded images of mCherry-Rab7 labeled late endosomes and mitochondria, labeled using Mitotracker Deep Red FM (Thermo Fisher Scientific; Supplementary Fig. 4, left and right panels, respectively, representative thresholded images, n = 8 cells for each condition). Based on this analysis, cells expressing PDZD8-GFP, had a higher fraction of the Rab7-labeled endosomes in proximity to mitochondria as compared to cells expressing PDZD8ΔCC-GFP (Supplementary Fig. 4, bottom panel).
We also performed Grazing Incidence Structured Illumination Microscopy (GI-SIM) super resolution live-cell imaging and examined the relative movement of PDZD8-labeled ER and Rab7-labeled endosomes. Analysis of movement of Rab7-labeled endosomes and the ER using either the general ER marker mEmerald-Sec61β, PDZD8-GFP or PDZD8ΔCC-GFP revealed that PDZD8-GFP, but not PDZD8ΔCC-GFP overexpression, significantly decreased endosome motility relative to endosomes in cells expressing mEmerald-Sec61β, consistent with our observation that co-overexpression of PDZD8 and Rab7 results in the formation of multiple highly associated ER-endosomal regions in cells (Supplementary Fig. 5, temporally color-coded map over 50 s, upper panel: mEmerald-Sec61β; middle panel: PDZD8-GFP; lower panel: PDZD8ΔCC-GFP, and corresponding Supplementary Movies M1–M3). In addition, GI-SIM imaging also indicated that PDZD8-labeled ER subdomains associated with Rab7-labeled endosome were stably localized adjacent to mitochondria (Fig. 5b, arrows, and Supplementary Movie M2), consistent with CLEM analysis (Fig. 5a) and proximity analysis (Supplementary Fig. 4). In total, our data suggest that overexpression of PDZD8 and Rab7 facilitates the formation of contacts between both ER-endosomes and ER-mitochondria in the context of a three way MCS, consistent with previous work indicating that PDZD8 is a component of an ER-mitochondria tether27.
The Rab7 effector, WDR91, marks PDZD8-associated endosomes
The early to late endosome maturation process consists of an intertwined complex series of steps, which include endosomal acidification, a Rab5 to Rab7 switch, an interdependent conversion of membrane PIP from PI(3)P to PI(3,5)P2 and an alteration of the lipid composition as late endosomes become depleted in phosphatidylserine (PS) and rich in the unique phospholipid lysobisphosphatidic acid (LBPA)31,32,33. To gain insight into what endosomal maturation stage the PDZD8-mediated ER-late endosome MCS forms, we co-expressed in U2OS cells PDZD8-GFP and BFP-Rab7 with either mCherry-Rab5 or LAMP1-mCherry. LAMP1 is a membrane protein that is targeted from the Golgi to lysosomes through a lysosomal targeting motif34. Rab7 and LAMP1 have been shown to co-localize on late endosomes and lysosomes9. In cells overexpressing both PDZD8 and Rab7, however, Rab7-labeled endosomes that possessed regions of associated PDZD8-labeled ER were not labeled by LAMP1 (Fig. 6a). When PDZD8 and Rab7 were co-expressed with the early endosomal marker Rab5, the endosomal population that was labeled by PDZD8 and Rab7 also showed no overlap with Rab5 labeled early endosomes (Fig. 6b). We next co-expressed PDZD8-GFP and BFP-Rab7 with mCherry-WDR91, a Rab7 effector that controls the phosphoinositide conversion step in endosomal maturation and serves as a marker for the PIP conversion step35. When co-expressed in cells, WDR91 and PDZD8 were localized together on late endosomes uniformly labeled with Rab7 in a mutually exclusive pattern (Fig. 6c), which has not been previously observed for WDR9135. The localization of WDR91 and PDZD8 to the same population of late endosomes suggests that the PDZD8 ER-endosome MCS may form at an endosomal maturation stage during the PIP conversion step in which the Rab5 to Rab7 switch is complete.
Although we do not know the mechanism underlying the mutually exclusive localization pattern of PDZD8-GFP and mCherry-WDR91 associated with Rab7-lableled endosomes, we exploited it to resolve the localization of mitochondria relative to the PDZD8 ER subdomains associated with Rab7-labeled endosomes. In cells co-expressing BFP-Rab7, mCherry-WDR91, and PDZD8-GFP, mitochondria were observed more frequently in close proximity to the PDZD8-associated portion of Rab7-endosomes (Fig. 6c, bottom panel). We analyzed the fraction of total BFP-Rab7 and Mitotracker pixels that overlapped, and then calculated how many of these pixels also overlapped with either PDZD8-GFP or mCherry-WDR91 pixels (Fig. S6, magenta colored pixels in left and right panels, respectively). This analysis revealed that a majority of the contact area between Rab7-labeled-endosomes and mitochondria occurs adjacent to PDZD8-GFP co-localized with Rab7-labeled endosomes (Supplementary Fig. 6, bar graph), consistent with our previous observations (Fig. 5). GI-SIM live imaging of cells expressing mCherry-WDR91, BFP-Rab7, and PDZD8-GFP supported this conclusion as mitochondria are persistently localized over time to portions of Rab7-endosomes that are associated with PDZD8-ER subdomains (Supplementary Movie M4). Thus, these data support the conclusion that mitochondria localize with PDZD8 ER subdomains associated with Rab7-labeled endosomes in cells.