The endo-lysosomal system drives lumen formation in a human epiblast model

Abstract

Epiblast lumen formation is a critical step that occurs during human embryo implantation. It begins with the apicosome, an apical compartment partly derived from endocytosed material. However, how the endo-lysosomal system contributes to apicosome formation and later lumen development remains unclear. Using a human pluripotent stem cell-derived model of epiblast formation, we show that apicosome formation is accompanied by transient expansion of early endosomes, late endosomes, and lysosomes, as well as the formation of hybrid compartments with features of both early and late endo-lysosomal stages. These changes depend on the RAB GTPases RAB35 and RAB7, which are required for proper apicosome formation and lumen morphology of the human epiblast model. Our findings identify RAB35- and RAB7-dependent endo-lysosomal remodeling as a key mechanism driving apical membrane morphogenesis during human epiblast development.

Rengarajan et al. identify RAB35- and RAB7-regulated endo-lysosomal dynamics as important drivers of the formation of the apicosome and lumen formation in a model of human epiblast formation. This study presents new insights into peri-implantation human embryo morphogenesis.

Introduction

A key step in early human embryonic development is the formation of the epiblast, a radially organized cyst with an apical central lumen and basal exterior^1–5. This polarized epiblast cyst is essential for subsequent critical developmental events such as amniogenesis and gastrulation^6–8. Previous studies using three-dimensional (3D) human pluripotent stem cell (hPSC)-derived model of human epiblast formation^9–11 have shown that the first sign of lumen formation in this hPSC-epiblast model is the formation of an intracellular structure called an apicosome (^1,12, Fig. 1A). In an aggregate of cells, apicosomes from multiple cells fuse to initiate lumen formation in the center of the aggregate. Each individual apicosome has characteristics of the apical lumen; its membrane contains microvilli and a primary cilium, and its lumen has a high concentration of calcium¹². Apicosomes are seen in several epiblast cells in implanting mouse blastocysts¹², suggesting that apicosomes may contribute to epiblast formation in vivo.

Fig. 1. Early endosome and LE/lysosome compartments transiently enlarge during apicosome formation.

A A schematic overview of apicosome formation in hPSC and classification of stages. Pre-apicosome stage: multiple PODXL vesicles (green) accumulate peri-nuclearly. Proto-apicosome: EZRIN and aPKCζ (red circle) accumulate on a single large PODXL vesicle. Mature apicosome: a single large apicosome with apical characteristics (e.g., microvilli (green projections), primary cilium (yellow projection), high Ca²⁺ concentration (not shown)). B, C Enlarged early endosomes accumulate in the pre-apicosome stage. Representative confocal images of H9 hESC harvested at 3- (B) and 6- (C) hr timepoint, which were stained with the indicated markers. Arrows indicate small peripheral PODXL-labeled EE vesicles. D Quantification for EEA1 vesicle volume (left, 318 (3-h) and 182 (6-h) vesicles out of 15 cells each), EEA1/PODXL co-localization coefficient per cell (middle, 20 cells) and number of EEA1 vesicles per cell (right, 15 cells) at pre- (3-h) and proto- (6-h) apicosome stages. E, F Enlarged lysosomes accumulate during the pre-apicosome stage. Representative confocal images of H9 hESC harvested at 3- (E) and 6- (F) hr timepoints, which were stained with the indicated markers. In E, arrowheads indicate PODXL structures that contain RAB7; asterisks indicate highly enlarged LE/lysosome compartments labeled by RAB7, LAMP1 and/or lysotracker, which lack PODXL. G Quantification of LAMP1 vesicle volume (left, 155 (3-h) and 34 (6-h) vesicles out of 15 cells each) and number of LAMP1 vesicles per cell at pre- (3-h) and proto- (6-h) apicosome stages (right, 15 quantified cells at each timepoint). Brackets indicate highly enlarged LE/lysosomes (>4 μm³). At least three independent experiments were performed. In D and G, asterisk indicates a statistically significant difference based on two tailed Student’s t-test (D: P = 0.0009, P < 0.0001 and P = 0.0408 for vesicle volume (left), co-localization coefficient (center) and vesicle number (right), respectively; G: P = 0.0017 and P < 0.0001 for vesicle size and number, respectively), the mean and the ± standard error of the mean (SEM) are shown. For B, C, E and F, two additional representative cells are shown in Supplementary Fig. 2A–D (EE/RE) and Supplementary Fig. 3A–D (LE/lysosome). Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

The process of apicosome formation takes place in three distinct stages (Fig. 1A). First, vesicles containing apical proteins such as podocalyxin (PODXL), a member of the CD34 sialoprotein family that is a critical driver of lumenogenesis in several systems^13–18, accumulate peri-nuclearly (pre-apicosome stage, first 3 h after initiation). Then, actin cytoskeleton and additional apical proteins (e.g., EZRIN, aPKCζ) accumulate to form nascent apicosomes (proto-apicosome stage, between 6 and 12 h after initiation). Finally, the proto-apicosome matures through an increase in volume and the formation of microvilli and a primary cilium (mature apicosome stage, between 12 and 20 h after initiation^1,12, Fig. 1A).

We previously reported that the apicosome forms near markers of the early, recycling, late endosomes/lysosome (LE/lysosome)^1,12,19. Moreover, the forming apicosome accumulates endocytosed material¹², suggesting that the endo-lysosomal system may contribute to apicosome formation. However, the actual contribution of that endo-lysosomal pathway to the genesis of the apicosome structure remained unclear. Here, we explored endo-lysosomal dynamics during apicosome formation using immunofluorescence (IF) and spatial proteomics using the engineered ascorbate peroxidase APEX2^20,21. We uncovered a series of key trafficking events, driven by a RAB5/RAB35/RAB7 hybrid endosome, that occur at the pre-apicosome stage to ensure that only one apicosome is formed per cell. Moreover, our work reveals RAB35 as a resident component of the apicosome. Overall, these results present a model of apical morphogenesis, which depends on the endo-lysosomal pathway that is used during early human development.

Results

The endo-lysosomal system undergoes dynamic remodeling during apicosome formation

To gain insight into the roles of the endo-lysosomal system during apicosome formation, we first used IF to monitor the localization of endo-lysosomal organelles at different apicosome formation stages (pre-apicosome, 3-h (hr); proto-apicosome, 6-h; mature apicosome, 20-h, see Supplementary Fig. 1A, B for IF images and quantitation for different apicosome stages at several timepoints) in control cells (H9 human embryonic stem cells). Importantly, apicosome formation is asynchronous (Supplementary Fig. 1B) because the cells are still undergoing mitosis. Therefore, we identified apicosome stages based on the size, morphology and number of PODXL vesicles and/or the presence or absence of pERM (phosphorylated EZRIN/RADIXIN/MOESIN) as described in Methods. Wheat germ agglutinin (WGA), which binds to glycosylated proteins like PODXL^22–24, was also used to label some vesicles in the endo-lysosomal pathway.

We first monitored the early endosome (EE) using antibodies against EEA1 (early endosome antigen 1) and RAB5 (Fig. 1B, C). In the pre-apicosome stage, cells contained two to three enlarged (≥1 μm³ in volume) EEs as well as several small peripheral EEs, which were almost always positive for PODXL (Fig. 1B, D (EEA1 vesicle volume and PODXL/EEA1 co-localization analysis), also see Supplementary Fig. 2A, B for additional sets of images), suggesting that apical proteins transit through or accumulate in the EE at this stage.

At the proto-apicosome stage, although both enlarged and small EEs were still seen (Fig. 1C, D), these EEs were significantly reduced in number compared to the pre-apicosome stage (Fig. 1C). Moreover, compared to the pre-apicosome stages, significantly fewer EEs contained PODXL, regardless of whether the EEs were adjacent to the proto-apicosome or more peripheral (Fig. 1C, D) (see Supplementary Fig. 2C, D for additional sets of images). Finally, at the apicosome stage, enlarged EEs were absent (Supplementary Fig. 2E, F). Small EEs were found both clustered near the apicosome and more peripherally (Supplementary Fig. 2E, F). Similar to the proto-apicosome stage, EE at the apicosome stage did not contain PODXL (Supplementary Fig. 2E, F). These data suggest that the EEs transiently enlarge during apicosome formation, and that PODXL transits through the EEs at the earliest stages of apicosome formation. These data are consistent with prior research showing that a substantial portion of the PODXL in the apicosome derives from PODXL originally localized on the plasma membrane at the onset of apicosome formation¹².

We next examined the association of the recycling endosome (RE) with the forming apicosome, using an antibody against RAB11. At both the pre- and proto-apicosome stages, the REs were proximal to the enlarged PODXL-positive EEs but did not accumulate high levels of PODXL (Fig. 1B, C, also see Supplementary Fig. 2A–D). At the apicosome stage, the REs were closely associated with one side of the apicosome, likely near the base of the primary cilium (Supplementary Fig. 2E, F). These findings suggest that the REs do not dramatically enlarge or relocalize during apicosome formation, nor does PODXL accumulate in the RE.

We next explored the later compartments of the endo-lysosomal system (Fig. 1E, F). To do this, we used antibodies against RAB7, a small GTPase that localizes to the LE/lysosome, LAMP1, a transmembrane protein of the late endosome and lysosome, and lysotracker, which stains the acidic LE and lysosome^25–27. At the pre-apicosome stage, we observed several enlarged LE/lysosomes with an average volume between 1 to 2 μm³ (Fig. 1E (arrowheads), see quantitation in Fig. 1G, and Supplementary Fig. 3A, B for additional images). Some smaller peripheral LE/lysosomes (less than 1 μm³ in volume) were also observed at this stage (Fig. 1E, G). Notably, most cells contained one or more highly enlarged structures (>4 μm³) that were labeled with RAB7, LAMP1 and lysotracker, but were devoid of PODXL (Fig. 1E (asterisks), see bracket in Fig. 1G). These data suggest that the lysosome enlarges and PODXL transits through or accumulates in the LE/lysosome during apicosome formation.

In the proto-apicosome stage, the number of LE/lysosomes was lower than in the pre-apicosome stage (Fig. 1F, G). Enlarged LE/lysosomes that contained PODXL were still apparent; however, their prevalence was notably reduced (Fig. 1F, G, see Supplementary Fig. 3C, D for additional images). Moreover, the highly enlarged LE/lysosomes were mostly absent (Fig. 1F, G). Small LE/lysosomes, which are often found adjacent to the proto-apicosome, were also observed at this stage, although they were fewer in number (Fig. 1F, G). Finally, at the apicosome stage, enlarged LE/lysosomes were absent, and smaller LE/lysosomes were often observed adjacent to the apicosome (Supplementary Fig. 3E, F). These results indicate that, like the EEs, the LE/lysosomes undergo a transient enlargement and that PODXL can be transiently observed in a subset of enlarged LE/lysosomes.

Finally, given the similar morphologies of the enlarged EEs and LE/lysosomes in the pre- and proto-apicosome stages, we examined the relationship between markers of the EE and LE/lysosomes. At the pre-apicosome stage, there was extensive co-localization of EE and LE/lysosome markers on PODXL vesicles (Fig. 2A, arrowhead, see Supplementary Fig. 3A, B for additional images), suggesting that, at this stage, there is a hybrid organelle containing markers of both the EE and LE/lysosome. The highly enlarged LE/lysosomes, which lack PODXL, lacked markers of the EE (Fig. 2A, asterisks). These observations indicate that EEs and hybrid organelles contain PODXL at this time point, but many cells contain one or two enlarged LE/lysosomes that lack both markers of the EEs and PODXL. At the proto-apicosome stage, the number of hybrid vesicles is reduced (Fig. 2B).

Fig. 2. Hybrid organelles fuse with the LE/lysosome compartment during apicosome formation.

A, B Hybrid organelles accumulate in pre- and proto-apicosome stages. Representative confocal images of H9 hESC harvested at pre- (3-h, A) and proto- (6-h, B) apicosome stages, which were stained with the indicated markers. In A, arrowhead indicates a hybrid vesicle that contains PODXL, RAB5 and RAB7; asterisks indicate highly enlarged LE/lysosome compartments labeled by RAB7 and lysotracker, which lack PODXL; arrows indicate structures that contain PODXL, RAB7 and lysotracker. For A and B, two additional representative cells are shown in Supplementary Fig. 3A–D. Scale = 20 µm. Blue pseudocolor indicates DNA in all images. C Hybrid organelles fuse with a highly enlarged LE/lysosome. Representative high-resolution time-lapse confocal images of an H9 hESC expressing PODXL-mCherry (shown with red or gray coloring to aid in signal visualization), which were stained using lysotracker dye (green). Asterisks indicate a lysotracker-labeled highly enlarged LE/lysosome compartment that lacks PODXL-mCherry; arrows indicate a hybrid organelle (labeled with PODXL and mCherry) that fuses into the LE/lysosome structure. An increase in mCherry fluorescent signal is seen at the contact site (brackets) within 5 min. Three optical sections with a 0.5 µm interval along the Z-axis are shown at each timepoint. Merged panel (merge + DIC) includes green (lysotracker), red (PODXL-mCherry) and DIC images. Scale = 5 µm. D Quantification of the averaged PODXL-mCherry signals in the whole area (X-Y plane) of highly enlarged lysotracker-labeled LE/lysosome compartments after hybrid organelle fusion (0-min). Ten total highly enlarged LE/lysosomes from 4 distinct cells were measured. A statistical significance was examined using a linear mixed-effects model, which does not require adjustments for multiple comparisons (two-sided analysis, P = 0.000432). E A schematic summarizing molecular and trafficking characteristics of the endo-lysosomal system during the earliest stages of apicosome formation.

To understand the fate of PODXL in the hybrid organelles, we monitored PODXL-mCherry and lysotracker in live cells at the pre-apicosome stage. At this point, all PODXL structures that contain Lysotracker are hybrid organelles. Consistent with the observation from fixed cells, we observed many enlarged lysotracker-positive structures that also contained PODXL-mCherry (Fig. 2C, arrows), as well as highly enlarged lysotracker-positive structures that lacked PODXL-mCherry (Fig. 2C, asterisks). We observed many instances (more than 30 cells in 4 different runs) where at least two to three PODXL-mCherry and lysotracker-positive structures (Fig. 2C, indicated by arrows at 0-min) appear to fuse with an existing highly enlarged lysotracker-positive structure during 2-h long imaging (Fig. 2C, asterisks in merge). Next, we measured mCherry intensity in the highly enlarged lysotracker vesicles with clearly attached (or fusing) mCherry-positive hybrid vesicles over time (5-min interval) by using the time of attachment at 0-min (Fig. 2D, 10 total vesicles across 4 cells). This quantitation was analyzed using a linear mixed-effects model²⁸ to test whether changes across the three timepoints in all ten vesicles are statistically significant (Fig. 2D). Indeed, the mCherry signal significantly increased in the highly enlarged lysotracker vesicle at an average of 0.55 a.u. per minute over the next 10 min (Fig. 2D, P < 0.001). This suggests that PODXL in the hybrid organelles is often delivered to the highly enlarged LE/lysosome, where it is degraded.

Together, these data suggest that the endo-lysosomal system undergoes dynamic remodeling during apicosome formation. The EE and LE/lysosome, but not the RE, enlarge. PODXL transits through the EE, where a fraction of it is delivered to the apicosome, whereas the remainder is targeted to the LE/lysosome from the hybrid organelles (Fig. 2E).

The PODXL-proximity proteome reveals machinery important for endo-lysosomal dynamics

To gain additional insight into machinery controlling endo-lysosomal dynamics during apicosome formation, we examined the proteome of the PODXL⁺ endosomes and apicosomes at 30-h, a timepoint when all three stages of apicosome maturation can be observed (Supplementary Fig. 1B). To do this, we used a previously developed doxycycline (DOX)-inducible APEX2-based proximity labeling system (Fig. 3A, B^19,29).

Fig. 3. APEX2-based spatial biotinylation of the PODXL proximal territory during apicosome formation.

A Design of APEX2 fusion construct based on a DOX-inducible piggyBac transposon system. GoI, gene of interest; TR, terminal repeats; TRE, tetracycline-responsive element; pA, polyadenylation sequence; rEf1α, rat elongation factor 1 α promoter; rtTA, reverse tetracycline transactivator; IRES, internal ribosome entry site; Puro, puromycin. Resulting proteins are shown on the right¹ : 1) PODXL-FLAG-APEX2; 2) FLAG-APEX2-NES. B Flow diagram of APEX2-based proximity biotinylation of PODXL and cytosolic territory proteins. H9 hESC expressing APEX2 fused to PODXL and NES were incubated with BP (biotin-phenol) for 1 h, followed by brief H₂O₂ treatment to trigger biotinylation. The purple-shaded area indicates biotinylated territory. C APEX2 fusion constructs localize correctly. Representative optical section of 20-h cells expressing PODXL- or NES-APEX2 constructs with or without DOX, stained with indicated markers. With DOX treatment, the FLAG signal is seen in the pERM⁺ apicosome in the PODXL-APEX2 cells, while abundant cytosolic FLAG staining is seen in the NES-APEX2 cells. D APEX2 fusion biotinylate proteins proximal to the fusion protein. Representative confocal images of biotin-labeled PODXL- or NES-APEX2 cells from pre-apicosome, proto-apicosome and mature apicosome stages stained with the indicated markers. Arrowheads indicate streptavidin-labeled PODXL vesicles. In the fluorescent images, insets ((i)-(iv) in (C); (i)-(vi) in (D) indicate magnified regions in the merge. Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

We established and validated stable cell lines with APEX2 constructs: PODXL-APEX2, to label the developing apicosome, and APEX2-NES for background subtraction during ratiometric analysis. To validate that the APEX2 constructs localize correctly, we performed IF on cells expressing the transgene using the FLAG epitope, which is fused to each APEX construct. PODXL-APEX2 was enriched at the apicosome territory, whereas APEX2-NES was observed throughout the cytoplasm based on FLAG staining (Fig. 3C). We next validated that the APEX2 constructs efficiently biotinylated proteins in the correct region of the cells. APEX2 labeling was performed by incubating cells with biotin-phenol (BP) for 1 h, followed by 1.5 min H₂O₂ labeling and harvesting (Fig. 3B). In control samples (with no BP or H₂O₂), weak diffuse staining was observed (Supplementary Fig. 4). In contrast, in cells expressing PODXL-APEX2, strong biotin signal was detected in the peri-nuclear vesicles (pre-apicosome, Fig. 3D, arrowheads) and at the proto- and mature apicosomes (Fig. 3D), whereas, in cells expressing APEX2-NES, strong biotin signal was detected in the entire cytoplasmic domain (Fig. 3D). For PODXL-APEX2, although labeling was concentrated at the apicosome, additional diffuse cytosolic localization of biotin signal was observed, likely due to diffusion of the peroxidase activated label, or diffusion of labeled cytosolic proteins (Fig. 3D). These analyses demonstrate that this selective APEX2-based biotinylation can label and enrich PODXL-proximity proteins in apicosome-forming cells.

To identify the PODXL-proximity proteome, we isolated biotinylated proteins from PODXL-APEX2 and NES-APEX2 cells in duplicate, a non-biotinylated APEX2 negative control (cells expressing PODXL-APEX2 without H₂O₂ labeling), and a no APEX2 negative control (unmodified H9 hESC with H₂O₂ labeling, Fig. 4A, B). Isolated proteins were subjected to on-bead digestion followed by tandem mass tag (TMT)-based tagging. The TMT-tagging allows a ratiometric analysis of samples to improve protein identification from complex samples^30,31.

Fig. 4. Ratiometric profiling of PODXL-proximity proteome during apicosome formation.

A Workflow of PODXL territory proteome analysis using duplicates of PODXL and NES samples and negative controls (PODXL-APEX2 without H₂O₂, unmodified H9 treated with BP and H₂O₂). B Experimental design of TMT 6-plex-based profiling. Row 1, specific TMT tags; row 2, APEX2 lines; rows 3 and 4, proximity labeling conditions. POI, protein of interest. C Diagram of filters used in ratiometric analyses to obtain proteins specific to PODXL territory. Brackets: Number of proteins that passed each filter. D Representative histograms and receiver operating characteristic (ROC) curves of PODXL replicate 1 and 2, illustrating how filters 1 and 2 were applied. Sensitivity is derived from the true positive rate, while specificity is the true negative rate, which is calculated by subtracting the false positive rate from 1. See Methods for additional cutoff analysis and specificity/sensitivity details. E Functional classification of the apicosomal PODXL territory proteins. F–H Gene Ontology analysis of the PODXL territory proteins: Biological Process (F, top 10 shown), Molecular Function (G, top 10) and Cellular Component (H, top 25).

All samples were then combined, fractionated, and analyzed as a pooled mixture by LC-MS/MS (Fig. 4A). We identified 4,915 unique proteins (Fig. 4C, see Supplementary Data 1A for the raw dataset). We used a previously described pipeline to normalize samples using known mitochondrial matrix soluble proteins (Supplementary Data 1B^20,21,29,), remove proteins biotinylated in negative control samples (Supplementary Data 1C–E), and remove proteins that are highly abundant in the cytoplasm, as determined from the NES-APEX2 sample (Supplementary Data 1F, G). We determined cut-off values for true and false positives by monitoring the enrichment of proteins with cell membrane annotation as previously described¹⁹. This analysis identified 322 proteins that were enriched in both replicate samples (Supplementary Data 2A). The enrichment values were highly correlated between the two datasets (Supplementary Fig. 4E, F); moreover, it accurately identified 14 out of 51 known PODXL binding partners that are expressed in hPSC (Supplementary Data 2B³²), suggesting that the identified proteins represent a bona fide PODXL-proximal proteome.

The PODXL proximity proteome is enriched for proteins associated with various trafficking and organelle activities (Fig. 4E, also see Supplementary Data 2C). In Gene Ontology (GO) enrichment analysis (Fig. 4F–H and Supplementary Data 2D–F), we found terms associated with the endolysosomal system, including the Cellular Component terms early endosome (FDR = 2.80e-35), late endosome (FDR = 3.34e-18) and lysosome (FDR = 1.53e-20), consistent with our imaging data (Figs. 1 and 2). GTPase activity was the most highly enriched Molecular Function term (FDR = 1.61e-24), reflecting many Rab- and Rho-type GTPases in the PODXL-proximal proteome (Fig. 4G and Supplementary Data 2E). Notably, the Cellular Component term recycling endosome was also highly enriched (FDR = 1.89e-16, Supplementary Data 2F). Proteins with this annotation included RAB11A and RAB11B, which were both more than five-fold enriched in the PODXL-proximal proteome compared to NES-control (Supplementary Data 2A). This suggests that PODXL may transit through the RE during apicosome formation, even if it does not accumulate there.

We next sought to understand whether the association of PODXL with the endo-lysosomal system detected in our proteomics reflected transient remodeling of its traffic during apicosome formation or its behavior at steady state. To do this, we compared the 30-h PODXL-proximity proteome with our prior PODXL-proximal proteome from hPSC-epiblast cells, in which the mature apicosome has further matured into an apical lumen. We identified 186 proteins unique to the 30-h timepoint, 257 unique to the hPSC-epiblast and 136 shared between the two datasets (Supplementary Data 2A–I, respectively). In GO enrichment analysis, we found that the 30-h PODXL-proximity proteome was enriched in proteins annotated with terms associated with the endo-lysosomal system, including the Cellular Component terms early endosome, late endosome and lysosome (FDR = 2.79e-20, 2.06e-16 and 9.48e-17, respectively, Supplementary Data 2J–L). In contrast, the hPSC-epiblast PODXL-proximity proteome was less enriched for these terms (early endosome, FDR = 0.00034; late endosome, FDR = 0.001; lysosome, FDR = 4.29e-06), and was, instead, dominated by terms associated with the actin cytoskeleton (Supplementary Data 2M–O). Together, these results indicate that the traffic of PODXL through the endo-lysosomal system is actively remodeled during apicosome maturation.

RAB35 regulates apicosome formation

Our IF and proteomics data suggest that dynamic changes in endo-lysosome traffic occur during apicosome formation. To probe the functional significance of these changes, we sought to impair traffic at different points in the endo-lysosomal system. To probe dynamics in the early endo-lysosomal system, we targeted RAB35, a known regulator of traffic in the early endo-lysosomal system that is known to control PODXL trafficking in MDCK cells^17,33–35.

We first examined the localization of RAB35 during apicosome formation using a DOX-inducible GFP-RAB35 cell line. In cells at the pre-apicosome stage, RAB35 localized to enlarged endosomes that contained EEA1 and PODXL (Fig. 5A), suggesting that RAB35 localizes to the enlarged EE. RAB35 also co-localized with RAB7 in endosomes that contained PODXL (Fig. 5A, see arrowheads for RAB7⁺ PODXL vesicles), but not the highly enlarged endosomes that lacked PODXL (Fig. 5A, asterisks), suggesting that it is found in the hybrid organelle but not the highly enlarged LE/lysosome. In cells with a proto-apicosome or mature apicosome, RAB35 was localized to the apicosome membrane, as defined by pERM staining (Fig. 5B, C). The localization of RAB35 to the pERM-positive proto-apicosome structure is distinct from other tested RABs (Fig. 1). These results identify RAB35 as a component of the hybrid endosome and apicosome membrane.

Fig. 5. RAB35 is required for single apicosome formation.

A–C RAB35 localizes to small and large early endosomes during pre- and proto-apicosome stages, and localizes to the apicosome. H9 hESC carrying DOX-inducible GFP-RAB35-WT transgenic construct were treated with DOX starting at 0-h, harvested at indicated timepoints (3-h, (A); 6-h, (B), 20-h, (C)), and were stained with indicated markers. Representative confocal images of these samples are shown. Arrowheads indicate RAB7/RAB35-labeled PODXL vesicles. Asterisks in A indicate highly enlarged LE/lysosome compartments that lack PODXL. D RAB35 is needed for apicosome formation. Representative confocal images of control, RAB35-KO #1 and RAB35-KO #2 cells, as well as RAB35-KO cells expressing a DOX-inducible GFP-RAB35-WT construct (RAB35-KO + GFP-RAB35) at 20-h, stained with the indicated markers. Formation of several apicosomes (labeled by both PODXL and pERM) was seen in the absence of RAB35, which was rescued when RAB35 was reintroduced. E Quantification of the formation of the multi-apicosome phenotype in control (unmodified H9 hESC, black), RAB35-KO (red), RAB35-KO + GFP-RAB35-WT without (green) or with (purple) DOX treatment. RAB35 reintroduction in the RAB35-KO background leads to a significant reduction in multiple apicosome formation (n = 100 cells in each of the three independent experiments in each background, statistical significance based on one-way ANOVA (P < 0.0001) as well as by Fisher’s LSD test (asterisks indicate statistical significance of selected pairwise comparisons, P < 0.0001, two-tailed), the mean and the ±SEM are shown). F Loss of RAB35 impairs cilia formation. Representative confocal images of control, RAB35-KO #1 and RAB35-KO #2, as well as RAB35-KO cells expressing a DOX-inducible GFP-RAB35-WT construct (RAB35-KO + GFP-RAB35) at 20-h, stained with the indicated markers. Insets indicate magnified regions in the merged images. Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

To investigate the role of RAB35 in apicosome formation and endosome dynamics, we used two independent hPSC lines lacking RAB35 (one line previously established in¹⁹, and a new line established for this study, Supplementary Fig. 5). In both RAB35-KO lines, apicosome formation was severely disrupted. At 20-h, RAB35-KO cells failed to form a single apicosome structure (Fig. 5D). Instead, several apicosome-like structures that were positive for PODXL and pERM were formed (Fig. 5D, quantitation shown in Fig. 5E and Supplementary Fig. 6A). In addition, in the RAB35-KO background, primary cilia formation was impaired. In controls, a well-formed primary cilia (visualized with ARL13B) was present in nearly every mature apicosome (Fig. 5F). However, in the RAB35-KO cells, although ARL13B staining was usually observed associated with one of the pERM⁺ structures, the staining was often diffuse or the cilia were notably shorter compared to controls (Fig. 5F). Both the multiple apicosome and primary cilia phenotypes were rescued when a wild-type form of RAB35 fused to GFP at its N-terminus (GFP-RAB35) was expressed in the RAB35-KO background (Fig. 5D–F). These data indicate that RAB35 is critical for apicosome formation and maturation.

Loss of RAB35 impairs late endo-lysosomal dynamics at the pre-apicosome stage

We next explored endosome dynamics in RAB35-KO cells using IF. At the pre-apicosome stage, components of the EE were associated with enlarged PODXL vesicles in RAB35-KO cells, similar to control cells (Fig. 6A, 3-h cells are shown). Similarly, LE/lysosome components were associated with PODXL vesicles in the RAB35-KO cells (Fig. 6B). However, in the RAB35-KO cells, the highly enlarged structures that were labeled with LE/lysosome markers but lacked PODXL were absent (Fig. 6B), although a much smaller structure that was labeled with LE/lysosome markers but lacked PODXL was occasionally seen (see Supplementary Fig. 6B for LAMP1⁺ vesicle diameter quantitation). These results suggest that traffic to the LE/lysosome may be impaired in the RAB35-KO cells. Consistent with this hypothesis, levels of PODXL were elevated in the RAB35-KO cells, whereas levels of RAB7 were unaffected (Fig. 6C–E). Moreover, the number of PODXL⁺ vesicles per cell was increased in the RAB35-KO cells at 3-h (Fig. 6F). Together, these results reveal that the multiple apicosome formation in the RAB35-KO cells is associated with defective formation of the highly enlarged LE/lysosome at the pre-apicosome stage and increased PODXL levels. Thus, RAB35 may function to directly or indirectly drive PODXL traffic to the lysosome.

Fig. 6. Loss of RAB35 impairs LE/lysosome dynamics during apicosome formation.

A, B RAB35 disrupts the formation of highly enlarged LE/lysosomes in the pre-apicosome stage. Representative confocal images of 3-h control and RAB35-KO cells stained with indicated markers (EE and RE compartments (A); LE/lysosome compartments (B)). Asterisks in B indicate highly enlarged LE/lysosomes that are labeled by RAB7, LAMP1 and/or lysotracker. C Loss of RAB35 elevates total PODXL levels at 3-h. Western blot analysis of 3-h control, RAB35-KO #1 and RAB35-KO #2 cells, blotted for PODXL, RAB7, EZRIN and β-actin. Three independent replicates are shown for each sample group. D, E Quantification of levels of PODXL (D) and RAB7 (E), normalized to total protein, using samples in (C, three biological replicates per condition, with the mean and the ±SEM). One-way ANOVA analysis showed a statistical significance in D (PODXL, P = 0.0016), but not in E (RAB7, P = 0.4696); asterisks indicate a statistical significance based on Fisher’s LSD test (D, P = 0.0199 (control and KO#1) and P = 0.0005 (control and KO#2)). F Quantification of the number of PODXL vesicles (widest diameter larger than 2 µm) in 3-h control and RAB35-KO cells (n = 18 cells per background, asterisk indicates a statistically significant difference based on Student’s t-test (P = 0.004), two-tailed; mean and ±SEM are shown). Insets indicate magnified regions shown as merged images. Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

We next explored the effects of the loss of RAB35 on the endosomes and lysosomes at 20-h (Supplementary Fig. 6C, D). At this timepoint, components of the EE, RE, LE and lysosomes were adjacent to the multiple enlarged PODXL and pERM positive apicosome-like structures in cells lacking RAB35 (Supplementary Fig. 6C, D). Therefore, although the RAB35-KO cells are unable to form a single mature apicosome structure, the pERM-positive apicosome-like structures do not display endosomal characteristics.

To test whether RAB35 GTPase activity contributes to these functions, we examined apicosome formation in RAB35-KO cells that inducibly express wild-type (WT), constitutively active (CA – Q67L), or dominant negative (DN – S22N) forms of GFP-RAB35 fusion protein. We found that expression of either RAB35-WT or -CA constructs restored single apicosome formation, while expression of RAB35-DN did not restore single apicosome formation (Fig. 7A, quantitation shown in Fig. 7B and Supplementary Fig. 7A). Similarly, RAB35-WT and -CA restored the formation of the highly enlarged LE/lysosomes (Fig. 7C–E, see asterisks in Fig. 7D, E), whereas RAB35-DN expression did not (Fig. 7F, G). Together, these results show that the active GTP-bound form of RAB35 is required for the formation of a single apicosome and for the formation of the highly enlarged LE/lysosome compartment at the pre-apicosome stage.

Fig. 7. RAB35 GTP binding is required for the formation of single apicosomes and highly enlarged lysosomal compartments.

A Representative confocal image of RAB35-KO cells carrying DOX-inducible GFP fused RAB35-WT, -CA (constitutively active, Q67L) and -DN (dominant negative, S22N) transgenic constructs treated with DOX for 20 h, stained with pERM and PODXL. In the “no DOX” condition (top), DOX-untreated 20-h RAB35-KO + RAB35-WT cells were harvested at the 20-h timepoint. Note that the expression of the RAB35-DN construct does not lead to single apicosome formation. B Quantification of 20-h cells with more than one apicosome for the indicated background (n = 100 cells in each of the three independent experiments in each background. A statistical significance of the sample group was seen based on one-way ANOVA (P < 0.0001) as well as by Fisher’s LSD test (asterisks indicate statistical significance of selected pairwise comparisons, P < 0.0001; ns, P = 0.1039)). C–F RAB35 GTP binding is needed for the formation of highly enlarged LE/lysosomes. Representative optical section of RAB35-KO cells carrying inducible GFP-RAB35-WT (DOX-untreated in (C); DOX-treated in (D)), -CA (E) and -DN (F) constructs, harvested at the 3-h timepoint (with 3-h DOX treatment) and stained with indicated markers. Note that, while seen with the expression of RAB35-WT or -CA, the expression of the RAB35-DN construct does not lead to the formation of highly enlarged RAB7/LAMP1 double-positive LE/lysosomes. Asterisks indicate the highly enlarged LE/lysosome compartments (D, E). G Quantification of the number of PODXL vesicles (widest diameter larger than 2 µm) in 3-h cells from the indicated sample group (n = 18 cells per background; numbers indicate total counted vesicles). The number of PODXL vesicles is reduced to levels similar to controls when the RAB35-WT or -CA construct is expressed in the RAB35-KO background, which is not seen with the expression of the RAB35-DN. A statistical significance of the sample group was seen based on one-way ANOVA (P = 0.0004) as well as by Fisher’s LSD test; selected pairwise comparisons are shown (asterisks, P = 0.0049 (control/RAB35-KO, control/RAB35-DN, RAB35-KO/RAB35-WT with DOX), 0.0259 (control/RAB35-WT without DOX), 0.001 (RAB35-KO/RAB35-CA)). The mean and the ±SEM are shown. Insets indicate magnified regions in the merged images. Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

RAB7 drives formation of the LE/lysosome compartments and the apicosome

We next explored the contribution of LE/lysosome traffic to the formation of the apicosome. To do this, we generated two hPSC lines that each carry a unique frameshift mutation in one of the RAB7A loci, which display reduced RAB7 (Supplementary Fig. 7B, hereafter referred to as RAB7^+/−). Strikingly, similar to RAB35-KO lines, the RAB7^+/− cells failed to form normal apicosomes at 20-h, and instead showed multiple enlarged PODXL structures that also contained pERM (Fig. 8A, B). Moreover, at 3-h, lysotracker staining showed that the formation of the large LE/lysosome compartment was reduced in the RAB7^+/− cells compared to the controls (Fig. 8C, see Fig. 8D for quantitation). These results show that both RAB35 and RAB7 are critical for apicosome formation.

Fig. 8. RAB7 is required for single apicosome formation and lysosome dynamics.

A RAB7 is needed for apicosome formation. Representative optical sections of 20-h control, RAB35-KO, as well as RAB7^+/− clone #1 and #2 cells, stained with the indicated markers. Similar to RAB35-KO, formation of several apicosomes (labeled by both PODXL and pERM) was seen in the RAB7 heterozygous background. B Quantification of the formation of the multi-apicosome phenotype in control (unmodified H9 hESC, black), RAB35-KO (red), RAB7^+/− (dark purple) and RAB7^+/−(purple, n = 100 cells in each of the three independent experiments in each background). C RAB7 is needed for the formation of highly enlarged acidified LE/lysosomes. Representative optical sections of control, RAB7^+/− clone #1 and RAB7^+/− clone #2 cells at the pre-apicosome stage, stained with the indicated markers. Highly enlarged LE/lysosomes labeled with lysotracker and RAB7 (indicated by asterisks in controls) are not seen in RAB7^+/− cells. D Quantification of the diameter at the widest point of PODXL negative LAMP1⁺ LE/lysosome vesicles in control, RAB7^+/− clone #1 and RAB7^+/− clone #2 cells at 3-h. Twenty pre-apicosome cells were counted for each background. Only six and four PODXL-negative LE/lysosome compartments were found in the RAB7^+/− #1 and #2 cells, respectively (among 20 total cells); twenty-eight LE/lysosome compartments were found among 20 cells in the control background. In B and D, statistical significance is based on one-way ANOVA (P < 0.0001) as well as by Fisher’s LSD test (asterisks indicate statistical significance of selected pairwise comparisons – B, P < 0.0001 and ns, P = 0.3762; D, P = 0.0003 (control/RAB7^+/−#1), P = 0.0014 (control/RAB7^+/−#2)). The mean and the ±SEM are shown. Insets indicate magnified regions in the merged images. Scale = 20 µm. Blue pseudocolor indicates DNA in all images.

RAB35 acts upstream of RAB7 to promote single apicosome formation

RAB35 and RAB7 are generally thought to function in distinct locations in the endo-lysosomal system. Depending on the cell type, RAB35 has been linked to recycling to the plasma membrane, functioning either in the early endosomal system (EE or RE) or the plasma membrane, whereas RAB7 is largely thought to function in fusion in the late endo-lysosomal system^36–45. Nonetheless, Rab35 is known to function upstream of Rab7 in a phagosomal pathway in C. elegans^46,47. Given the observations that the loss of RAB35 or reduced RAB7 level impairs the formation of the highly enlarged LE/lysosomes, and that multiple apicosome formation is seen in both genetic backgrounds, we explored whether RAB35 acts upstream of RAB7 during apicosome formation. To test this hypothesis, we inducibly expressed WT, constitutively active (CA – Q67L), or dominant negative (DN – T22N) forms of GFP-RAB7 fusion protein in the RAB35-KO background. Consistent with the idea that RAB7 functions downstream of RAB35, expression of GFP-RAB7-WT or -CA (but not DN) in RAB35-KO restored the formation of single apicosomes (Fig. 9A, quantitation in Fig. 9B and Supplementary Fig. 7C), highly enlarged LE/lysosomes (Fig. 9C–F, asterisks), as well as PODXL vesicles (Fig. 9G). This result suggests that RAB35 and RAB7 likely control sequential steps in a degradative route important for apicosome formation (Fig. 9H).

Fig. 9. RAB7 acts downstream of RAB35 in apicosome formation and lysosome expansion.

A Representative optical section of RAB35-KO cells inducibly expressing GFP-RAB7-WT, -CA (Q67L) or -DN (T22N) construct at 20-h (with 20-h DOX treatment), stained with the indicated markers. In the “no DOX” condition (top), DOX-untreated 20-h RAB35-KO + RAB7-WT cells were harvested at the 20-h timepoint. B Quantification of the percentage of cells with more than one apicosome in the indicated genetic background (n = 100 cells in each of the three independent experiments in each background). A statistical significance of the sample group was based on one-way ANOVA (P < 0.0001) and Fisher’s LSD test (P < 0.0001 for all comparisons except for RAB7-WT without DOX and RAB7-DN (P = 0.0249)). C–F Representative confocal images of RAB35-KO cells inducibly expressing GFP-RAB7-WT (minus DOX in (C), plus DOX in (D)), -CA (Q67L, (E)) or -DN (T22N, (F)) construct at 3-h (with or without 3-h DOX treatment), stained with indicated markers. G Quantification of the number of PODXL vesicles (widest diameter larger than 2 µm) in 3-h cells from the indicated sample group (n = 18 cells per background; numbers indicate total counted vesicles). Statistical significance: one-way ANOVA (P < 0.0001) and by Fisher’s LSD test (P = 0.0062 (control/RAB35-KO), P = 0.0012 (control/RAB7-WT without DOX), P = 0.0046 (RAB7-WT without DOX/RAB7-WT with DOX), P = 0.0084 (RAB7-WT without DOX/RAB7-CA), P = 0.0886 for ns)). H Proposed model of apicosome formation. At the earliest stages of apicosome formation, enlarged EEs containing RAB5, RAB35 and PODXL form via macropinocytosis and/or fusion of small EE vesicles. A subset of enlarged EEs recruits RAB7 and LAMP1 (top), becoming hybrid organelles. Subsequently, these hybrid vesicles shed RAB35 and RAB5, and either become or fuse with an acidic LE/lysosomal compartment, where PODXL is degraded. It remains unclear whether proteins can recycle from the hybrid endosome to the apicosome. As apicosome maturation proceeds, a single enlarged EE becomes the apicosome (bottom). RAB35 may contribute to single apicosome formation by promoting fusion of enlarged EEs, and/or by facilitating maturation of the hybrid endosome. Asterisks indicate statistical significance. The mean and the ±SEM are shown. Scale = 20 µm.

Loss of RAB35 or RAB7 impairs hPSC-epiblast morphogenesis

The apicosome is the first stage in epiblast formation, a critical structure in early development^1,12. To determine whether the defects in apicosome formation seen in RAB35-KO or RAB7^+/− impair later stages in epiblast formation, we investigated the roles of RAB35 and RAB7 in a multi-cellular hPSC-epiblast cyst model. To do this, we cultured RAB35-KO and RAB7^+/− for 2 to 4 days, in a culture condition that induces the formation of cysts with a central lumen in controls^12,32. At d2, controls formed round cysts that are composed of radially polarized cells as well as a small central lumen labeled with pERM and PODXL (Fig. 10A). This lumen further expands over the next 24 (d3, Fig. 10B) and 48 h (d4, Fig. 10C), which results in cysts with a large central lumen.

Fig. 10. RAB35 and RAB7 drive epithelial morphogenesis in a hPSC-epiblast model.

A–C Representative confocal images of developing control, RAB35-KO and RAB7 heterozygous hPSC-epiblast cysts, harvested at d2 (A), d3 (B) and d4 (C), which were then stained with indicated apical (pERM, PODXL) and membrane markers (E-CADHERIN). Scale = 20 µm. Blue pseudocolor indicates DNA in all images. D–G Quantification of apical lumen morphogenesis in the indicated genetic backgrounds at d4. D formation of one central circular lumen (left), one central but disorganized lumen (middle, e.g., non-circular, jagged) and one central lumen with small satellite lumens/apical vesicles (right). E 3D sphericity (20 cysts each). F circularity (membrane roughness). G roundness (aspect ratio measurement). In F and G, 404, 296, 382, 341 and 353  confocal sections with lumens along the z-axis (1μm interval) were counted for control, RAB35-KO #1 and #2 and RAB7^+/− #1 and #2, respectively (out of 20 total cysts for each background). The mean and the ±SEM are shown. Statistical significance is based on one-way ANOVA (P < 0.0001 in D, F, G and P = 0.0014 in (E)) as well as by Fisher’s LSD test (asterisks indicate statistical significance of selected pairwise comparisons, P < 0.0001 for all comparisons in (D), for all comparisons in F and G except for the comparison between control and RAB7-KO#2 (P = 0.0001, P = 0.0002, respectively); E, P = 0.001 (control/RAB35-KO#1), P = 0.008 (control/RAB35-KO#2), P < 0.0001 (control/RAB7-KO#1) and P = 0.023 (control/RAB7-KO#2)).

In the absence of RAB35, cyst organization is severely impaired at d2; RAB35-KO hPSC-epiblast cysts showed several structures labeled by pERM and PODXL (Fig. 10A), which likely reflect multiple apical domains or unfused apicosomes. When these cells were cultured for three or four days, central lumens formed; however, cysts often contained ectopic or satellite lumens, and the main lumen was often deformed (Fig. 10B–D, see Fig. 10E–G for lumen shape quantitation). Compared to controls, d4 RAB35-KO cysts showed more than four-fold increase in the number of cysts with ectopic lumens, and a significant decrease in the 3D lumenal sphericity, as well as in the lumenal membrane morphology and roundness (based on circularity and aspect ratio measurements, respectively) (Fig. 10C–G), suggesting that RAB35 contributes to lumen formation.

RAB7^+/− hPSC-epiblast cysts also showed disorganized morphogenesis. At d2, although RAB7^+/− cysts contained a defined central lumen labeled by pERM and PODXL, ectopic PODXL foci were seen throughout the developing cysts (Fig. 10A). Satellite lumens were also seen in d4 RAB7^+/− cysts (Fig. 10D). Moreover, similar to the lumens seen in RAB35-KO cysts, lumens of the RAB7^+/− hPSC-epiblast cysts were less spherical and were more jagged compared to controls (Fig. 10E–G). Together, these results reveal that endo-lysosomal traffic contributes both to the formation of the apicosome and the subsequent formation of a normal lumen.

Discussion

In this study, we examined the dynamics and role of the endo-lysosomal system during apicosome formation in a human epiblast model. We identified a transient enlargement of the early and LE/lysosomes accompanied by the formation of a hybrid compartment containing RAB5, RAB35 and RAB7. We also demonstrated that RAB35 is a marker of the apicosome and that RAB35 is critical for driving single apicosome formation as well as the formation of the highly enlarged LE/lysosomes. These findings provide key insights into how LE and lysosomal compartments contribute to human epiblast lumen formation and expand the list of machinery required for apicosome formation (Figs. 9H^1,19,48,49,). Moreover, because previous spatial proteomic studies probed apical polarization and vesicle dynamics only in established epithelia^19,50–52, which may reflect its behavior at steady state, this study presents a valuable proteomic resource that captures a transient remodeling state of an apical vesicle population during an active phase of epithelial morphogenesis.

Our results are consistent with a model whereby the apicosome derives from one or more of the enlarged RAB5/RAB35 endosomes observed at early timepoints. At the earliest stages of apicosome formation, PODXL is associated with RAB5 and RAB35. This finding is consistent with our prior work, which suggests that the apicosome contains endocytosed materials¹². Since the RAB5 endosome often serves as the first way-station for endocytosed cargo²⁷, it is likely that the co-localization with RAB5 reflects the transit of PODXL through the early endosome. The enlarged nature of a subset of the RAB5 endosomes remains notable; however, whether this reflects their formation via a process like micropinocytosis, fusion of multiple smaller endosomes, or enlargement of individual endosomes via other processes remains unknown.

The hybrid enlarged RAB5-RAB35-RAB7 organelle is a striking aspect of apicosome formation. At early stages of apicosome formation, many of the enlarged vesicles that contain PODXL and RAB5 also contain RAB35 and RAB7. RAB7 generally defines the LE and is often recruited by machinery that is recruited by RAB5⁵³. Therefore, we postulate that a subset of the enlarged RAB5 endosomes recruits RAB35 and RAB7, and that these are destined for destruction, either by conversion of a fully active lysosome or fusion with the highly enlarged LE/lysosome (Fig. 9H). The remaining RAB5 endosomes then fuse in a process that likely depends on RAB35 to generate the apicosome.

It remains unclear how the RAB5-RAB7 organelle forms and how it resolves. In many systems, RAB5 endosomes convert to a RAB7 endosome^54,55. During this conversion, there is a transient period of overlap⁵⁶. However, in this hPSC-based system, IF analysis revealed the presence of abundant RAB5-RAB7 organelles, and our time-lapse imaging showed that the hybrid organelles appear notably long-lived. Therefore, it remains unclear whether the conversion from RAB5 to RAB7 is mechanistically different in hPSC.

We envision two models for RAB35. RAB35 may induce traffic from the hybrid organelle to the EE or apicosome. This removal of apical components progressively converts the hybrid organelle to the LE/lysosome, which then rapidly fuses with the highly enlarged LE/lysosome. Alternatively, RAB35 may play a role more directly in endosome maturation. One possible mechanism involves the ESCRT proteins. ESCRT proteins have been recently linked to the RAB5-RAB7 conversion^57,58. The ESCRT protein HGS (also known as HRS) binds RAB35 in neuronal cells, where it promotes degradation of synaptic vesicle proteins⁵⁹. Therefore, RAB35 may promote conversion through an ESCRT-mediated mechanism. Consistent with this hypothesis, the ESCRT-0 proteins STAM1 and STAM2 are also enriched in the PODXL-proximity proteome. Future work will be needed to resolve these possible mechanisms.

Apicosome formation is only the first step of lumen formation in the hPSC-epiblast model. After apicosomes form in single cells within a cell aggregate, the apicosomes relocate to the center of the aggregate to initiate the formation of a central lumen in a cyst of cells^12,60. Our data suggest that RAB35 is particularly important at this stage. Cell aggregates lacking RAB35 accumulate multiple apicosome-like structures at timepoints when wild-type cell aggregates have formed an ordered lumen. This defect is consistent with the known role of RAB35 in promoting fusion of PODXL-containing structures with one another or with the cell surface^34,35,61.

Our results indicate that RAB7 is not required for lumen formation per se, but that it contributes to the speed of lumen formation, and to final lumenal morphology. We hypothesize that its role here is to remove excess apical material by promoting fusion at the LE/lysosome (Fig. 9H), and, in the absence of this activity, several RAB5/RAB35 positive organelles persist, leading to the multiple apicosome phenotype. This role is consistent with the ability of increased levels of active RAB7 to restore apicosomes in the RAB35-KO background. In this model, elevated RAB7 removes excess hybrid organelles by promoting fusion at the LE/lysosome. Such a function is consistent with the localization of RAB7 to the hybrid organelle and LE/lysosome, and its known function in other systems.

An alternative possibility is that RAB7 substitutes for RAB35 fusion reactions at the apicosomes, consistent with its mislocalization to the apicosome membrane when over-expressed (Fig. 9A). Notably, in other lumen formation systems such as the fly trachea⁶² and zebrafish vasculature⁶³, RAB7 has a proposed role in forming secretory lysosomes that are needed for cell-cell separation. Therefore, RAB7 could mediate the fusion of some hybrid organelles with the apicosome membrane. Future work is needed to fully elucidate the role of RAB7 in this system.

Taken together, in addition to epiblast lumenogenesis, this study opens an avenue to further explore the role of LE and lysosome compartments during apical membrane morphogenesis in additional developmental processes (e.g., neural tube^64,65, gut tube^66,67) and in disease (e.g., microvillus inclusion disease^68,69, polycystic kidney disease^70,71).

Methods

Cell lines used in this study

H9 hESC line was used in this study (WA09, female (XX), P30, WiCell; National Institute of Health (NIH) registration number: 0062). All protocols for the use of the hESC line were approved by the Human Stem Cell Research Oversight Committee at the Medical College of Wisconsin. H9 hESC were maintained in a feeder-free system for at least 20 passages and characterized as karyotypically normal at several passage numbers, including P30, P46 and P54. Karyotype analysis was performed at Cell Line Genetics. All hPSC lines tested negative for mycoplasma contamination (LookOut Mycoplasma PCR Detection Kit, Sigma-Aldrich). All transgenic and KO hPSC lines in this study used H9 hESC as the parental line. hESC were maintained in a feeder-free culture system with 50%/50% mix of mTeSR1 and mTeSR plus (STEMCELL Technologies). H9 cells were cultured on 1% (v/v) Geltrex (Thermo Fisher Scientific) or with Cultrex SCQ (Bio-Techne) coated six-well plates (Nunc). Cells were passaged as small clumps every 4 to 5 days with Dispase (Gibco). All cells were cultured at 37 °C with 5%CO₂. The media was changed every day. hESCs were visually checked every day to ensure the absence of spontaneously differentiated mesenchymal-like cells in culture. Minor regions of differentiated cells were scratched off the plate under a dissecting scope when identified. The quality of all hESC lines was periodically examined by immunostaining for pluripotency markers and successful differentiation to three germ layer cells.

Apicosome and hPSC-epiblast model formation assays

Methods for these assays are previously described^12,32,72. Briefly, singly dissociated cells were prepared using Accutase (Sigma-Aldrich) and were plated on coverslips coated with 1% Geltrex at 1 × 10⁴ cells/cm². Cells were plated in the 50/50 mTeSR1/mTeSR plus media containing Y-27632 (STEMCELL Technologies) and 2% Geltrex. Apicosome formation initiates spontaneously after plating. To generate hPSC-epiblasts, singly dissociated cells were plated at 2 × 10⁵ cells/cm² in the same media without Geltrex for 24 h to form cell aggregates, after which cells were cultured in the absence of Y-27632 but in the presence of 2% Geltrex overlay to allow aggregates to initiate radial organization and central lumen formation for the next 24 h. Beyond day 2, cells were cultured in the media with 1% Geltrex (without Y-27632).

Apicosome staging

Pre-, proto- and mature apicosome stages were identified based on the number, size, morphology and molecular characteristics (e.g., pERM labeling) of PODXL/WGA vesicles. In the pre-apicosome cells, three to four PODXL vesicles that are between 2–2.5 µm in diameter at the widest point are seen. These PODXL vesicles lack pERM. In the proto-apicosome cells, in addition to one to two PODXL vesicles that are 2–2.5 µm in diameter, one larger PODXL vesicle (3–3.4 µm in diameter) that is labeled with pERM is present, which is the proto-apicosome. Mature apicosomes (positive for both PODXL and pERM) are larger (ranging between 6.7–8 µm in diameter) and show a central open domain.

Immunostaining

Samples were fixed using 4% paraformaldehyde at indicated timepoints for 40 to 60 min, then were rinsed with PBS three times, and permeabilized with 0.1% SDS (Sigma-Aldrich) solution for 40 min. The samples were blocked in 4% heat-inactivated goat or donkey serum (Gibco) in PBS for 1 h to overnight at 4 °C. The samples were incubated with primary antibody solution prepared in blocking solution at 4 °C overnight, washed three times with PBS (10 min each), and incubated in blocking solution with goat or donkey raised Alexa Fluor–conjugated secondary antibodies (Thermo) at room temperature for 2 h. Counterstaining was performed using Hoechst 33342 (nucleus, Thermo) or Alexa Fluor–conjugated WGA (membrane, Thermo). For LysoTracker (LysoTracker Red DND-99, 500 nM in DMSO) staining, cells were treated with LysoTracker for 30 min before harvesting, and were processed for immunostaining. All samples were mounted on slides using Fluoromount-G (Thermo). Antibodies for IF staining are found in Supplementary Dataset 3.

Confocal microscopy

Confocal images were acquired using a Zeiss LSM980 laser scanning confocal microscope. Zen (Zeiss) as well as Photoshop (Adobe) were used to process images.

Time-lapse imaging

Singly dissociated PODXL-mCherry hESC were plated on a 35 mm glass-bottom culture dish (MatTek) to initiate apicosome formation for one hour. Lysotracker Green DND-26 (500 nM, Thermo) was added for 15 min before cells were imaged in an environmentally controlled chamber (37 °C with 5% CO₂) for two to three hours using a Zeiss LSM980 confocal microscope.

Constructs and cell lines

Generation of the PODXL-APEX2, APEX2-NES and PODXL-mCherry constructs has been previously described¹⁹. piggyBac-based DOX inducible RAB35 (Addgene#47424, 47425, 47426 (gift of Peter McPherson⁷³)) and RAB7 (Addgene#12605 (gift of Richard Pagano⁷⁴), 28048, 28049 (gift of Qing Zhong⁷⁵)) constructs were generated by PCR amplification (forward primer: Clo-dTOPO-EGFP-fw; reverse primers: Clo_dTOPO_hRab35_rv (RAB35), Clo_dTOPO_EGFP-RAB7A-rv (RAB7)) and by subcloning the PCR products into the pENTR-dTOPO (Thermo), followed by Gateway cloning (Thermo) into the PB-TA-ERN (Addgene#80474, gift of Knut Woltjen⁷⁶).

piggyBac-based transgenic and genome-edited hESC lines

The pBACON-puro-hRAB35 construct used to generate an additional RAB35-KO has been established previously¹⁹. pBACON-puro-hRAB7A construct was established in this study to generate RAB7^+/− cells (primers: CRISPR_hRAB7A_s and CRISPR_hRAB7A_as). To generate transgenic or genome-edited hESC lines, piggyBac constructs (4 µg) and pCAG-ePBase (1 µg; gift from Ali Brivanlou) were cotransfected into H9 hESC (60,000 cells/cm²) using GeneJammer transfection reagent (Agilent Technologies). To enrich for cells expressing the construct, drug selection (puromycin, 2 µg/ml; neomycin, 250 µg/ml) was performed 48 to 72 h after transfection. Selected pools were plated sparsely, and small colonies were picked for further expansion to establish clonal lines. hESC stably expressing each construct maintained the expression of pluripotency markers.

During pBACON-based genome editing, puro-selected cells were cultured at low density (300 cells/cm²) for clonal selection. Established colonies were manually picked and expanded for screening indel mutations using PCR amplification of a region spanning the targeted gRNA region for hRAB35 (primer pair: Seq_hRAB35_fw and Seq_hRAB35_rv) and hRAB7A (primer pair: Seq_hRAB7A-Fw 1 and Seq_hRAB7A-Rv 1), which were subcloned into pPBCAG-GFP (gift of Joseph Loturco⁷⁷) at Eco RI and Not I sites, and sequenced (Seq-3′TR-pPB-Fw). Genomic DNA was isolated from individual clones using DirectPCR Lysis Reagent (Tail) (VIAGEN). At least 12 to 15 bacterial colonies were sequenced to confirm genotypic clonality. Control cells are H9 hESC in all loss-of-function experiments.

APEX2 labeling and sample preparation

To examine APEX2 labeling, singly dissociated PODXL- or NES-APEX2 cells were plated on glass coverslips at 1 × 10⁴ cells/cm² in mTeSR1/mTeSR plus media containing DOX (2 µg/ml) and Geltrex (2%) for 29 h. In the last 1 h, cells were incubated in mTeSR/mTeSR plus media containing 2% Geltrex and biotin-phenol (BP, biotinyl tyramide, 500 µM, AdipoGen) without DOX. Hydrogen peroxide (H₂O₂) was then added directly into the medium to a final concentration of 1 mM for 90 s at room temperature to initiate biotinylation, and cells were immediately fixed using 4% paraformaldehyde for fluorophore-conjugated streptavidin staining as well as immunostaining for microscopic analysis.

To prepare proteomics samples, the 6 samples were prepared individually (shown in Fig. 4A) in 10 tissue culture-treated 100 mm dishes (thermo, precoated with 1% Geltrex). For each plate 1.0 × 10⁶ hPSC (DOX-inducible stable lines) were plated to obtain approximately 1.0 to 1.2 × 10⁷ cells per sample at 30-h, sufficient for 3.0 mg of total protein. At 30-h, APEX2 labeling was performed as described above. In addition, followed by the H₂O₂ treatment step, cells were washed 3x using quencher solution (10 mM sodium ascorbate (Spectrum Chemical), 10 mM sodium azide (Sigma-Aldrich) and 5 mM Trolox (Sigma-Aldrich) in Dulbecco’s PBS (Gibco)). After quenching, APEX2-labeled samples from 10 dishes were then resuspended as a pool in quencher buffer and centrifuged at 500 g for 5 min to collect the cell pellet for lysis in radioimmunoprecipitation assay (RIPA) lysis buffer (Pierce) containing 1x Halt protease inhibitor cocktail (Thermo), 1 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich), 10 mM sodium azide, 10 mM sodium ascorbate and 5 mM Trolox. Cell lysates were centrifuged at 15,000 g for 15 min at 4 °C, and the supernatant was collected for enriching biotinylated proteins using streptavidin beads.

The Pierce 660-nm assay (Pierce) was used to quantify protein concentrations in sample supernatants. To isolate biotinylated proteins, streptavidin-coated magnetic beads (Pierce) were first washed twice with RIPA lysis buffer. For each sample, 3.0 mg of total protein was incubated with 500 µl of streptavidin beads overnight at 4 °C with gentle rotation. Beads were subsequently washed on a MagnaRack (Thermo Fisher Scientific). Beads were washed twice with RIPA lysis buffer, once with 1 M KCl, once with 0.1 M Na₂CO₃, once with 2 M urea in 10 mM tris-HCl (pH 8.0), then twice with RIPA lysis buffer, and three times with PBS. Last, PBS was removed as much as possible, and beads were frozen in −80 °C before performing on-bead digestion, TMT labeling, peptide pooling, fractionation, and LC-MS/MS. All these steps were performed at 4 °C unless otherwise noted.

Western blot

SDS-PAGE gels (10% or gradient gel, 4 to 20%, Bio-Rad) and polyvinylidene difluoride membranes were used. Membranes were blocked using Intercept (TBS) Blocking Buffer (LI COR), total protein quantification was performed by using Revert 700 Total Protein Stain (LI COR), and primary antibody overnight incubation was performed at 4 °C, followed by 1-h IRDye (LI-COR) secondary antibody incubation. Biotinylated proteins were detected and quantified by streptavidin-IRDye conjugate (LI-COR). Blots were imaged using the LI-COR Odyssey Infrared Imaging system.

Quantitative MS

Proteins bound to streptavidin beads were digested by trypsin following the standard on-bead trypsin digestion workflow⁷⁸. Samples were proteolyzed and labeled with TMTsixplex by following the manufacturer’s protocol (Thermo Fisher Scientific) with minor modifications. Briefly, upon reduction [10 mM DTT in 0.1 M Triethylammonium bicarbonate (TEAB); 45 °C, 30 min] and alkylation (55 mM 2-chloroacetamide in 0.1 M TEAB; room temperature, 30 min in dark) of cysteines, the proteins were digested overnight with trypsin (1:25; enzyme:protein) at 37 °C, with constant mixing using a thermomixer. Proteolysis was stopped by adding 0.2% trifluoroacetic acid, and peptides were desalted using a SepPak C18 cartridge (Waters Corp). The desalted peptides were dried in Vacufuge (Eppendorf) and reconstituted in 100 µl of 0.1 M TEAB. The TMTsixplex reagents were dissolved in 41 µl of anhydrous acetonitrile, and labeling was performed by transferring the entire digest to the TMT reagent vial and incubating at room temperature for 1 h. The reaction was quenched by adding 8 µl of 5% hydroxyl amine and further 15-min incubation. Labeled samples were mixed together and dried using a vacufuge. An offline fractionation of the combined sample into six fractions was performed using a high-pH reversed-phase peptide fractionation kit according to the manufacturer’s protocol (Pierce; catalog no. 8488). Fractions were dried and reconstituted in 12 µl of 0.1% formic acid/2% acetonitrile in preparation for LC-MS/MS analysis.

To improve quantitation accuracy, we used multinotch-MS3⁷⁸, which minimizes the reporter ion ratio distortion resulting from fragmentation of co-isolated peptides during MS analysis. Orbitrap Fusion (Thermo Fisher Scientific) and RSLC Ultimate 3000 nano-UPLC (Dionex) were used to acquire the data. The sample (2 µl) was resolved on a PepMap RSLC C18 column (75 µm inside diameter × 50 cm; Thermo Fisher Scientific) at a flow rate of 300 nl/min using 0.1% formic acid/acetonitrile gradient system (2 to 22% acetonitrile in 110 min; 22 to 40% acetonitrile in 25 min; 6-min wash at 90% followed by 25-min re-equilibration) and directly sprayed onto the mass spectrometer using EasySpray source (Thermo Fisher Scientific). The mass spectrometer was set to collect one MS1 scan (Orbitrap; 120,000 resolution; AGC target, 2 × 10⁵; max IT, 50 ms) followed by data-dependent, “Top Speed” (3 s) MS2 scans (collision-induced dissociation; ion trap; NCD 35; AGC, 5 × 10³; max IT, 100 ms). For multinotch-MS3, top 10 precursors from each MS2 were fragmented by higher-energy-collisional-dissociation (HCD) followed by Orbitrap analysis [NCE 55; 60,000 resolution; AGC, 5 × 10⁴; max IT, 120 ms;100 to 500 m/z (mass/charge ratio) scan range].

Ratiometric analysis of proteomic data

Proteome Discoverer (v2.1; Thermo Fisher Scientific) was used for initial data analyses. MS2 spectra were searched against the SwissProt human protein database (downloaded on 4 December 2018; 20331 reviewed entries) using the following search parameters: MS1 and MS2 tolerance was set to 10 parts per million and 0.6 Da, respectively; carbamidomethylation of cysteines (57.02146 Da) and TMT labeling of lysine and N termini of peptides (229.16293 Da) were considered static modifications; oxidation of methionine (15.9949 Da) and deamidation of asparagine and glutamine (0.98401 Da) were considered variable. Identified proteins and peptides were filtered to retain only those that passed a ≤1% FDR threshold and ≥2 unique peptides. Quantitation was performed using high-quality MS3 spectra using the Reporter Ion Quantifier Node of Proteome Discoverer (average signal-to-noise ratio of 10 and <30% isolation interference). Specific TMT ratios were normalized using known 495 mitochondria matrix soluble proteins²¹: 265 were found in the PODXL territory dataset for ratiometric analyses in Filter 1 [PODXL-APEX2 #1/negative (126/130), PODXL-APEX2 #2/negative (127/130) and in Filter 2 [PODXL-APEX2 #1/APEX2-NES #1 (126/128), PODXL-APEX2 #2/APEX2-NES #2 (127/129). In each TMT ratio, the median ratio of 265 mitochondrial matrix soluble proteins was calculated (PODXL-APEX2 #1/negative, 3.09; PODXL-APEX2 #2/negative, 2.69; PODXL-APEX2 #1/APEX2-NES #1, 1.06; PODXL-APEX2 #2/ APEX2-NES #2, 0.93; all proteins in each TMT ratio were divided using these values to generate a normalized PODXL territory dataset (ratios in log2 scale). In Filter 1 (F1) and Filter 2 (F2), true positive (TP-F1 and TP-F2, proteins with UniProt “cell membrane” and “plasma membrane” annotations) and false positive [FP, 265 mitochondrial matrix soluble proteins (Filter 1) and all proteins in the list lacking UniProt “cell membrane” and “plasma membrane” annotations (Filter 2)] were defined. TP-F1 and TP-F2 are proteins that are known to localize in the membrane territory; FP-F1 are proteins that are predicted to be nonbiotinylated by APEX2 fusion constructs in this study; FP-F2 consists of proteins that can be biotinylated by our APEX2 constructs but are not predicted to be proximal to the PODXL territory. True-positive rate (TPR) and false-positive rate (FPR) were calculated for each ratio [TPR = TP/(TP + FN); FPR = FP/(FP + TN); FN, false negative; TN, true negative]; these values were used to generate ROC curves to test the suitability of TP and FP for each ratiometric analysis based on the area under the curve (AUC; a commonly used statistic that calculates the area under the ROC curve and quantifies the probability in which a randomly chosen positive case outranks a randomly chosen negative case), as well as to determine cutoffs at which the largest difference between TPR and FPR was observed (PODXL-APEX2 #1/negative [log₂(126/130) – AUC = 0.97, cutoff = 0.551]; PODXL-APEX2 #2/negative [log₂(127/130) – AUC = 0.97, cutoff = 0.702]; PODXL-APEX2 #1/APEX2-NES #1[log2(126/128) – AUC = 0.7, cutoff = 0.19]; PODXL-APEX2 #2/APEX2-NES #2 [log2(127/129) – AUC = 0.74, cutoff = 0.062]) (Fig. 4D). TPR is sensitivity, and true-negative rate (TNR) is specificity (TNR = 1-FPR). TPR-FPR is equivalent to the Youden index⁷⁹, which is a statistic commonly used to represent the performance of a dichotomous test. Larger values of the index mean better performance. For example, a value of 1 would mean the performance is perfect, as there are no false positives or false negatives. During the analysis of filter 2, proteins that passed filter 1 were used. Given a threshold parameter T, we have AUC and cutoff (CO) formulas as

$$TPR\left(T\right):T\to y\left(x\right)$$ 1

$$FPR\left(T\right):T\to x$$ 2

$$AUC={\int }_{x=0}^{1}TPR\left({FPR}^{-1}\left(x\right)\right)dx$$ 3

$$CO=arg{\max }_t\left(TPR\left(t\right)-FPR\left(t\right)\right)$$ 4

GO enrichment analysis

The final apical (322) proteomes were uploaded to the STRING database [string-db.org⁸⁰, analysis performed on December 2024, v12.0]: The top GO terms (ranked by FDR) on Cellular Component, Biological Process, and Molecular Function were plotted (Fig. 4F–H).

Quantitative analyses

Graphs were generated using Prism 6 (GraphPad Software), and Student’s test (Figs. 1D, 1G, 6F, Supplementary Fig. 6B) or one-way ANOVA (Figs. 5E, 6D, E, 7B, G, 8B, D, 9B, G, 10D-G, Supplementary Fig. 6A and Supplementary Fig. 7, followed by Fisher’s LSD post-hoc test where appropriate) was performed to examine statistical significance. P > 0.05 was considered nonsignificant; P ≤ 0.05 was considered significant and marked with an asterisk. For Supplementary Fig. 1B, 100 cells were counted from three independent samples (a total of 300 cells) per timepoint. For Figs. 5E, 7B, 8B, 9B, Supplementary Fig. 6A and Supplementary Fig. 7, 100 cells were counted from three independent samples (total 300 cells) per experimental condition. In Figs. 6F, 7G and 9G, 18 cells were counted per experimental background. For Fig. 8D and Supplementary Fig. 6B, PODXL-negative LAMP1 vesicle diameter was measured from 20 cells per genetic background using Imaris (Oxford Instruments). In our quantification analysis of the time-lapse imaging data (Fig. 2D), we used a linear mixed-effects model to test whether changes across the three timepoints in all ten vesicles from four distinct cells were statistically significant by setting time as a continuous predictor and vesicle as a random effect. At least three independent experiments were performed in all cases. Means and SEM are shown in all graphs, where appropriate. Surface rendering feature in Imaris (Oxford Instruments) was used to generate 3D surfaces (1.0 μm surface level) and to measure lumenal sphericity in Fig. 10E. Lumenal circularity and roundness in Fig. 10F, G were measured from all optical sections with an open central lumen labeled by PODXL using Image Analysis module in ZEN (Zeiss).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

A.R., S.W., M.C.D., and K.T. designed experiments; A.R., S.W., C.-W.L., L.E.E., J.C.W., A.E.C., L.E.T., N.S., B.E.S., and K.T. performed experiments; A.R., M.C.D., and K.T. analyzed data and wrote the manuscript; M.C.D. and K.T. supervised the project; all authors contributed to the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by NIH grants R01-HD098231, R01-HD102496, and R01-GM129255; MCW CBNA Start-up funds; and MCW Cancer Center Graduate Fellowship. We thank the University of Michigan Proteomics Resource Facility. Biorender was used to generate the images of microcentrifuge tubes in Fig. 4A.

Data availability

Raw proteomics dataset has been deposited in jPOST (⁸¹, accession number JPST004663, [https://repository.jpostdb.org/entry/JPST004663.0]). Other Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77140-9.