Introduction

The epididymis is an essential component of the male reproductive system, as it provides a suitable luminal environment for sperm transport, concentration, maturation, storage and preservation1,2. The epididymal epithelium is pseudostratified and consists of various epithelial cell types that line the epididymal duct and create a unique luminal microenvironment via blood‒epididymis barriers for ion exchange, nutrients, solutes, and extracellular vesicle secretions3,4,5,6. In contrast, gap junctions play a critical role in cell‒cell communication7,8. The maintenance of homeostasis in the epididymal epithelium is critical for epithelial integrity and male fertility9.

Although researches have demonstrated the epithelial cell compositions and segmental-specific transcriptome profiles of the epididymal epithelium10,11,12,13, research on the ultrastructures of these cells has stagnated for decades. The morphology of the epididymis has long been deciphered by light microscopy (LM), confocal fluorescence microscopy and transmission electron microscopy (TEM). Roy et al. observed basal cell (BC) projections and axopodia extensions and retractions via intravital imaging14. To date, the ultrastructures and morphologies of epididymal epithelial cells, including principal cells (PCs), BCs, clear/narrow cells (NCs) and apical cells, have been imaged via TEM in rodents1. Notably, a vacuole-like ultrastructure has been observed in epididymal epithelium cells1,15. However, this structure remains unlabeled, the components are undefined, and the function has not been investigated to date owing to the lack of 3D hierarchy information.

Volume electron microscopy (vEM) is a novel technology that can reveal the intricate complexity of cells and tissues at the membrane and organelle scales16,17 and was recently highlighted as one of the “seven technologies to watch in 2023”18. In the present study, we revealed the ultrastructure and morphology of the epididymal epithelium at the mesoscopic level via vEM. We achieved large-scale 3D reconstruction of the initial segment (IS), caput, corpus and cauda epididymal epithelium of the mouse epididymis. On the basis of these large-scale datasets, for the first time, we identified the intercellular organelle reservoir (IOR) structures between epididymal epithelial cells. These vacuole-like string structures of the IOR were distributed along the lateral membrane between two epithelial cells and stabilized with desmosomes, facilitating the degradation of multiple intracellular organelles, such as protein aggregates, autophagosomes (APs), lysosome-related organelles (LROs) and mitochondrial residuals. Further investigation revealed fewer and smaller IORs in the epididymides of 21-month-old individuals than in those of 8-week-old individuals. Moreover, Rab27a mutant (Rab27aash/ash) mice presented a significant reduction of the IOR in the caput epididymis, suggesting its involvement in the formation of IORs. Interestingly, IORs were also found between intestinal epithelial cells, suggesting their broad distribution in the epithelium. Two amino acid transporters, SLC6A15 and SLC7A14, were distributed on the edge of the IOR, suggesting that the proteins within the IORs might be dynamic and regulated by the two transporters. These results suggest advanced physiological roles for homeostatic maintenance in epithelial cells.

Results

Segmentation of the IOR in the epididymis shown via focused ion beam scanning electron microscopy (FIB-SEM)

Although standard TEM has revealed the existence of a vacuole-like ultrastructure in the epididymal epithelium, reliance on two-dimensional ultrathin sections alone is inadequate for characterizing the occurrence and spatial distribution of these structures in cells. To overcome these limitations, we aimed to thoroughly investigate intact epithelial cells from different segments of the epididymis. Leveraging the advanced capabilities of vEM based on FIB-SEM, we achieved comprehensive three-dimensional (3D) imaging of entire cells or tissues at 10-nm resolution or higher.

To examine the presence of vacuole-like ultrastructures in specific cell types, we selected candidate blocks from the IS, caput, corpus, and cauda epididymides for FIB-SEM imaging. We observed these vacuole-like structures more frequently in the caput, corpus and cauda than in the IS, with surface areas smaller than 10 μm2 (Fig. 1a, Supplementary Fig. 1 and Supplementary Movies 1).

Fig. 1: FIB-SEM imaging and segmentation of epididymal epithelial cells.
figure 1

a Representative XY slice of the IS, caput, corpus and cauda epididymis. The delineated area in orange is the IOR. Lu luminal part. Scale bar: 10 μm. b Three-dimensional reconstruction of FIB-SEM images derived from the 56 d caput epididymis. The dimensions of the volume are indicated. Scale bar: 5 μm. c Reconstruction of adjacent PCs (gray and Tiffany blue) and IORs (orange) in the IS, caput, corpus and cauda epididymis. Scale bar: 3 μm. d FIB-SEM image of a mouse epididymis fixed with 2.5% GA with osmium impregnation (OTO) showing the presence of the IOR. Scale bar: 2 μm. e TEM image of the caput epididymis after high-pressure freezing (HPF) showing the presence of the IOR. Scale bar: 1 μm. f Representative image of the membrane thickness comparison for the IOR and the cell membrane. Orange arrowhead, IOR membrane (MIOR, orange circle); green arrowhead, single-cell membrane (sM, green circle); yellow arrowhead, double-cell membrane (dM, yellow circle); blue arrowhead, mitochondrial membrane (mM, blue circle). Scale bar: 1 μm. g Corresponding statistical analysis of membrane thickness (n = 5 images from 3 mice for dM and mM; n = 6 images from 3 mice for sM and MIOR). One-way ANOVA followed by Tukey’s multiple comparisons test. ****p < 0.0001. Data are presented as mean value ± SEM. h TEM images of PC junctions showing that desmosome structures were distributed at both ends of the IORs. Yellow arrowhead, desmosome; blue arrowhead, tight junction. Scale bars from left to right: 1 μm, 100 nm.

To assess the spatial distribution of the IOR structures, we utilized FIB-SEM at 10 nm voxels to generate a collection of 1000–2000 electron microscopy (EM) datasets, encompassing a volume of 30.45–49.35 μm in width (x axis), 42.6–47.73 μm in length (y axis), and 10.89–13.08 μm in thickness (z axis) (Fig. 1b, Supplementary Fig. 2a and Supplementary Movies 2). We individually segmented the cell membrane and vacuole membrane via Amira software. After the datasets were reconstructed, we unexpectedly found that vacuole-like structures are present in the lateral intercellular space (LIS) of epithelial cells, including neighboring PCs, between PCs and BCs and between PCs and NCs, rather than inside the cells (Fig. 1c, Supplementary Fig. 2b and Supplementary Movies 35). Notably, protein aggregates and organelle membrane-like structures were enriched within these vacuoles, and we propose that this vacuole be referred to as an intercellular organelle reservoir (IOR). Furthermore, we quantified IORs at the 3D reconstruction level. FIB-SEM datasets revealed that the mean IOR number between paired PCs was a significant decrease in the IS compared to that of the caput, corpus and cauda, which is consistent with the datasets of 2D EM images (Supplementary Fig. 2c). Interestingly, quantification of the 3D datasets revealed that the volume of the IOR in the IS region was greater than that in the corpus regions (Supplementary Fig. 2di).

To exclude artefacts induced by fixation, we applied both conventional chemical and physical fixation. We observed IORs not only with glutaraldehyde fixation (Fig. 1d) but also with high-pressure cryofixation followed by freeze substitution (Fig. 1e), verifying that vacuole-like structures were present in the epididymis. Additionally, fluorescence light microscopy (fLM) of E-cadherin (Ecad) in the epididymis with prefusion fixation methods or different mouse lines (C57/B6J, C3H/HeJ and ICR/HaJ) revealed the presence of IOR structures among epididymal epithelial cells, although few IORs were observed with the nonperfusion (postfixation) procedures (Supplementary Fig. 3), suggesting that the IORs were not restricted to a specific mouse strain. Our fLM results consistently revealed that approximately 20% to 80% of the PCs displayed IORs across the four segments, with the highest abundance in the caput (Supplementary Fig. 4a, b). A comparison of the IOR number and area between two adjacent PCs in four segments revealed a dominance of PCs in the epithelium of the caput compared with the other segments (Supplementary Fig. 4ci). By combining the morphological observations of EM and fLM, we demonstrated that IORs were distributed among different cell types within the epididymis and were widely present, particularly in PCs in the caput and corpus of the epididymis.

Furthermore, to better understand the IOR structure, we investigated whether the IORs are membrane-bound or membrane-free structures. We assessed the thickness of its membrane and compared it to that of a single-cell membrane, two neighboring cell membranes, and the mitochondrial membrane. We found that the membrane of the IOR exhibited a thickness comparable to that of a single-cell membrane, albeit only half as thick as the mitochondrial membrane and two neighboring cell membranes (Fig. 1f, g). These observations suggest that the intercellular vacuoles are not individual vesicles but rather gap sites between epithelial cells. To further confirm the characterization of the IORs, we also used TEM to observe along the cell membrane and found that the IORs were stabilized by desmosomes (Fig. 1h).

Characterization of IORs and their composition

After characterizing the distribution of the IOR structure, we further explored their internal configurations. To elucidate the structural properties of IORs, we employed the reconstructed segmentation of 3D datasets to compare the structure, number, and size of IORs with those of intracellular vesicles (IVs). We found that the IORs presented a significantly greater contrast in the samples than the IVs did, indicating potential differences in content composition (Supplementary Figs. 2b and 5a). Moreover, we found that the volume of each IOR was substantially greater than that of the IVs inside various cell types, such as PCs/PCs, PCs/BCs, and PCs/NCs, at different epididymal segments (Supplementary Fig. 5b–d). In addition to protein aggregation, we observed many important organelle structures, including multivesicle bodies (MVBs), phagophores, autophagosome (APs), LROs and mitochondria (mito) (Fig. 2a, b), through segmentation and 3D reconstruction of membrane structures within the IORs (Fig. 2c–e and Supplementary Movies 6, 7). Importantly, 3D reconstruction revealed the occurrence of autophagy structure inside the IORs in the epididymis (Fig. 2d and Supplementary Movies 7). The quantification of EM images revealed the proportions of different contents in the caput, corpus, and cauda. We found that more than half of the IORs contained protein aggregates, MVBs, phagophore, and APs in three regions (Fig. 2f and Supplementary Fig. 6a, b).

Fig. 2: The characteristics and contents of IORs.
figure 2

a Representative XY orthoslice of major contents in IORs. Mito mitochondria, MVB multivesicular body, LRO lysosome-related organelles. Scale bar: 1 μm. b Representative TEM images of protein in the IOR after high-pressure freezing (HPF) showing the protein aggregates. Scale bars from left to right: 1 μm, 300 nm. c XY orthoslice of FIB-SEM images, 3D rendering of mitochondria (green and green arrowheads) in the IOR (orange). Scale bars from left to right: 500 nm, 1 μm. d XY orthoslice of FIB-SEM images, 3D rendering of side and top views of autophagic structures within IORs. Yellow, phagophore; red, AP; blue, autophagic cargo. Scale bars from left to right: 1 μm, 500 nm. e Three-dimensional reconstruction of contents in the IOR in the LIS of PCs. Orange, IOR; purple, protein; green, mitochondria; blue, MVB; black, LRO. Scale bar: 3 μm. f Analysis of the types of contents contained in the IOR (%) of the caput epididymis. g Immunofluorescence of LAMP1, LC3, p62 and TOMM20 (magenta) in IORs (labeled with Ecad, green). Lu luminal part. Scale bars from top to bottom: 5 μm, 3 μm. (5 independent experiments). h Representative IEM images following immunogold labeling of p62 in the IOR (n = 2 mice). Color-coded images, p62 (magenta). Scale bars from top to bottom: 500 nm, 100 nm.

To further verify the distribution of these organelles in IORs, we also used double immunostaining for Ecad and lysosome/autophagic markers, including LAMP1, p62 and LC3, as well as TOMM20, a mitochondrial marker, with fLM (Fig. 2g, Supplementary Fig. 6c and Supplementary Movies 8), and pre-embedding immunoelectron microscopy (IEM). Quantification of fLM images revealed that the distributions of these organelles differed across the caput, corpus, and cauda. The LAMP1 levels in the IOR were similar in the three segments, however, LC3 was expressed at higher levels in the corpus, whereas p62 was abundantly expressed in the caput (Supplementary Fig. 6df). IEM further confirmed the locations of these organelle-related processes within the IORs (Fig. 2h and Supplementary Fig. 6gi).

Manifestation of the IOR during epididymal development and ageing

Next, we examined whether IORs and their autophagy-related contents changed at different developmental stages and within the epididymal caput of 21-month-old (21 m) mice via immunofluorescence labeling and FIB-SEM. Through confocal immunofluorescence analysis of Ecad, we found that IORs were relatively rare at postnatal day 32 (P32), but their numbers gradually increased in fully differentiated epididymal epithelial cells at P38 and P56. The increase in the number of IORs in the caput of the epididymis was consistent with the timing at which the sperm entered the epididymis for maturation. Notably, a substantial decrease in IORs was observed in 21 m mice (Fig. 3a–e and Supplementary Fig. 7). We then assessed differences in the expression of the autophagy-related protein p62 in the epididymal caput of P38, P56 and 21 m mice. Our results revealed that the aged mice presented a notable decrease of p62 expression between cells, while the intracellular expression remained relatively stable (Fig. 3a, f, g). Consistent with the confocal imaging observations, the FIB-SEM datasets revealed a significant decrease in IOR numbers among caput epididymal epithelial cells at 21 m compared with those in P56 mice, with occasional membranous structures in a stacked pattern noted in the IORs (Fig. 4a–g and Supplementary Movies 9). Overall, we postulate that the IOR undergoes dynamic changes that initially increase and subsequently decrease as the epididymides mature and ageing.

Fig. 3: Dynamic presence of IORs during epididymal development and ageing.
figure 3

a Confocal image showing the dynamics of IORs from 38 days to 21 months in the caput epididymis indicating Ecad (green) and p62 (magenta) expression. White arrowheads: IORs. Lu: luminal side of the epididymal tubule. Scale bars from top to bottom: 5 μm, 3 μm. b Corresponding analysis of the percentage of IORs in a single tubule (n = 7 for 32 days, n = 3 for 38 days, n = 4 for 56 days, n = 5 for 21 months). c IOR numbers between neighboring PCs from 32 days to 21 months (n = 25 cell-pair for 32 days, n = 13 cell-pair for 38 days, n = 19 cell-pair for 56 days, n = 14 cell-pair for 21 months). Comparison of the IOR area (d) and frequency distribution of the IOR area (e) across different ages of the epididymis (n = 47 IORs from 7 fields for 32 days, n = 25 IORs from 4 fields for 38 days, n = 56 IORs from 5 fields for 56 days, n = 21 IORs from 6 fields for 21 m). Quantification of p62 in IORs (f, n = 10 for 38 days, n = 12 for 56 days, and n = 9 for 21 months) and intracellular signals (g, n = 10 for 38 days, n = 12 for 56 days, and n = 9 for 21 months). Data are presented as mean value ± SEM. Statistical significance between groups was determined via one-way ANOVA followed by Tukey’s test or two-tailed unpaired t-test. For all figures: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, exact p values are provided in Source Data. The source data is provided as a source data file.

Fig. 4: FIB-SEM imaging and segmentation of the IOR alteration during ageing.
figure 4

a XY orthoslice of the caput epididymis of FIB-SEM showing a decrease in IORs at 21 m compared with the 56 days caput epididymis. Scale bar: 10 μm. b Three-dimensional reconstruction of the FIB-SEM image of the 21-month caput epididymis. Scale bar: 5 μm. Corresponding analysis of IOR numbers (c) (n = 16 cell-pair for 56 days, n = 24 cell-pair for 21 months, ****p < 0.0001), IOR area (d) (****p < 0.0001), frequency distribution of the IOR area (e) (n = 96 for 56 days, n = 64 for 21 months), frequency distribution of the IOR area at 56 days (f) and 21 months (g). Data are presented as mean value ± SEM. Statistical significance between groups was determined via two-tailed unpaired t-test. Exact p values are provided in Source Data. The source data is provided as a source data file.

Rab27a is involved in the formation of IOR structures

Microtubules (MTs) and actin filaments (AFs) play crucial roles in facilitating the transport of IVs and organelles19,20,21. Interestingly, Rab27a was reported to regulate MVB size, exosome secretion22, transport of LROs23 and coordination of the activity of AF motors24. To elucidate the potential mechanism of IOR formation, we used the caput epididymides of Rab27aash/ash mice which the expression of Rab27a was absent25 to observe IOR dynamics. Compared with those of control littermates, significantly fewer Ecad-labeled IORs were observed in Rab27aash/ash mice, as shown by fLM images (Fig. 5a–g). Furthermore, we found that more intracellular LROs labeled by LAMP1 in the epithelial cells but fewer in the IORs of Rab27aash/ash mice (Fig. 5a, h, i). Consistent with the fLM results, both XY orthoslice and 3D reconstruction of FIB-SEM datasets revealed that the IOR number and area between PCs were significantly decreased in Rab27aash/ash mice (Fig. 6a–c, Supplementary Fig. 8a, b and Supplementary Movie 10). Of note, the sperm motility was decreased in the Rab27aash/ash mice compared to the control and further affected the number of the offspring pups (Supplementary Fig. 8c, d). Additionally, the quantification of the proportions of different structure types derived from FIB-SEM images revealed an obvious decrease LROs and MVBs in the IORs in Rab27aash/ash mice compared with control mice (Fig. 6d–h and Supplementary Fig. 8e). Moreover, we found a significant increase in the perimeter of intercellular membrane structures in Rab27aash/ash mice compared with control mice (Fig. 6i–l). These results indicate that Rab27a is involved in the formation of the IOR structure.

Fig. 5: IORs were decreased in Rab27aash/ash mice.
figure 5

a Representative images of the localization of LAMP1 (magenta) in IORs labeled with Ecad (green) in the control and Rab27aash/ash caput epididymides. White arrowheads: IORs. Lu luminal part. Scale bar: 5 μm. b Percentage of IORs in a single tubule from confocal images in (a) (n = 8 for control, n = 7 for Rab27aash/ash), ****p < 0.0001. c IOR numbers between neighboring PCs (n = 16 cell-pair for control, n = 11 cell-pair for Rab27aash/ash). ****p < 0.0001. Comparison of the IOR area (d) and frequency distribution of the IOR area (eg) (n = 74 IORs for control, n = 35 IORs for Rab27aash/ash; 6 fields in each group). ****p < 0.0001. h Quantification of IORs with positive LAMP1 signaling in control and Rab27aash/ash caput epididymides (n = 6 fields in each group). **p < 0.01. i Quantification of intracellular LAMP1 in control and Rab27aash/ash caput PCs (n = 19 cells from 6 fields for each group). ***p < 0.001. Data are presented as mean value ± SEM. Statistical significance between two groups was determined via two-tailed unpaired t-test. Exact p values are provided in Source Data. The source data is provided as a source data file.

Fig. 6: FIB-SEM imaging and segmentation of IOR reduction in Rab27aash/ash mice.
figure 6

a XY orthoslice of the caput epididymis from FIB-SEM revealed fewer IORs in Rab27aash/ash mice than in control mice. Scale bar: 10 μm. b Corresponding analysis of the number of IORs between neighboring cells (n = 41 cell-pair for the control dataset, n = 47 cell-pair for the Rab27aash/ash dataset). ****p < 0.0001. c IOR area comparison between control and Rab27aash/ash epididymides (n = 57 IORs for control, n = 51 IORs for Rab27aash/ash). ***p < 0.001. d Representative FIB-SEM orthoslice of contents in the IOR of the control and Rab27aash/ash epididymides. Scale bar: 3 μm. e Comparison of lysosomes in IORs derived from FIB-SEM images (n = 4 fields for control and Rab27aash/ash), **p < 0.01. fl Corresponding analysis of vesicle structures in IORs. Frequency distribution of the vesicle area (f), frequency distribution of the vesicle area of the control (g), and Rab27aash/ash (h), comparison of the vesicle perimeter in IORs (i) (****p < 0.0001), frequency distribution of (j) (n = 73 for control, n = 76 for Rab27aash/ash), and frequency distribution of the vesicle perimeter of control (k) and Rab27aash/ash groups (l). Data are presented as mean value ± SEM. Statistical significance between groups was determined via two-tailed unpaired t-tests. Exact p values are provided in Source Data. The source data is provided as a source data file.

The presence of IORs in small intestinal epithelial cells

To elucidate the complex cytological process of IORs, we performed single-cell RNA sequencing (scRNA-seq) analysis derived from our previous published results10 to evaluate the molecular expression linked to protein degradation and organelle autophagy pathways in various epididymal epithelial cell types. Unexpectedly, both the macroautophagic and protein degradation pathways presented low levels within PCs in the epididymis (Fig. 7a), which was corroborated by immunostaining of the late autophagic molecule LAMP1, as well as EM images of the lysosomal structure (Supplementary Fig. 9). Furthermore, the analysis of scRNA-seq data from 17 mouse organs26 (including our scRNA-seq data from the epididymis), comprising 248,029 cells, revealed notably low expression of autophagic genes in the epididymis, as well as the intestine (Fig. 7b), suggesting the potential existence of IORs in the intestine.

Fig. 7: The presence of IORs in the small intestine.
figure 7

The lower expression of autophagy-related genes in PCs (a) and the epididymis and intestine (b) shown by scRNA-seq analysis. c Three-dimensional reconstruction of FIB-SEM images derived from the 56 days crypt. The dimensions of the volume are indicated. Scale bar: 5 μm. d XY orthoslice of FIB-SEM images and 3D rendering of IOR in 56 days crypt. White dash line is cell membrane. The delineated area in orange is the IOR. Scale bar: 2 μm for slice, 500 nm for Zoom. e Confocal imaging showing the presence of IORs in the crypt of the small intestine, indicating the presence of Ecad (green) and p62 (magenta) in the IORs. White dash line is the IOR. White arrowheads: p62 signal. Scale bars from left to right: 20 μm, 2.5 μm, and 1 μm.

To validate the potential presence of IOR in the intestine, we further perform both vEM and immunostaining experiments. IORs were presented between epithelial cells at the crypt (Fig. 7c, d and Supplementary Movie 11) and villus (Supplementary Fig. 10a–c and Supplementary Movie 11) of the intestine. Immunostaining also revealed p62 expression within the Ecad-labeled IORs (Fig. 7e and Supplementary Movie 11) in the crypt. Using mito-Dendra2DDX4/+ transgenic P100 mice and confocal imaging, we detected mitochondria within IORs between crypt epithelial cells (Supplementary Fig. 10d and Supplementary Movie 12).

Localization of amino acid transporters on the IOR

The minimal level of autophagy in epididymal epithelial cells, as indicated above, cannot be dismissed as an evolutionary tactic for the breakdown and reutilization of exceptionally effective proteins and organelles to maintain cellular stability. Given the extensive presence of protein markers and ultrastructures linked to organelle autophagy and protein degradation within IORs, we questioned whether particular amino acid transporters were present. To explore this, we selected SLC7A14 and SLC6A15, two recently discovered transporters that facilitate the conveyance of amino acids with varying polarities. Double immunolabelling with Ecad revealed the clustering of SLC7A14 and SLC6A15 within the membrane encircling the IOR (Fig. 8a). As expected, both SLC6A15 and SLC7A14 also clustered at the IOR margin among intestinal epithelial cells (Fig. 8b). Together, these results suggest that the IOR structure is involved in the organelle turnover and/or protein recycling.

Fig. 8: Amino acid transport on the edge of the IOR.
figure 8

Confocal imaging showing that SLC7A14 and SLC6A15 were localized at the membrane segments of the IOR labeled with the indicated antibodies in epididymal epithelial cells (a) and intestinal epithelial cells (b). The left two images show the whole tubule distribution of SLC7A14 and SLC6A15 (magenta). The epithelial cells were labeled with Ecad (green). The nuclear was labeled with DAPI (blue). White arrowheads are SLC7A14 or SLC6A15 signal. Scale bars (a) from left to right: 20 μm, 5 μm, and 3 μm. Scale bars (b) from left to right: 20 μm, 5 μm, and 1 μm.

Discussion

With vEM, we achieved large-scale 3D reconstruction (49.35 × 42.6 × 13.08 μm for the IS, 35.96 × 44.19 × 11.8 μm for the caput, 30.45 × 45.68 × 10.9 μm for the corpus and 41.2 × 47.73 × 10.89 μm for the cauda) of the epididymal epithelium with ~10-nm isotropic voxels to reveal major cell types of the epididymal epithelial cells, including PCs, BCs, and NCs, providing the framework of the epididymal epithelial cells of the different epididymal segments. Importantly, this high-resolution 3D ultrastructure reconstruction, for the first time, revealed a IOR between epididymal epithelial cells, which was a unlabeled vacuole-like structure under TEM in textbooks1,15. These IORs exhibited a pearl-string-like distribution and contained the autophagic, LRO and mitochondrial residual structures identified via FIB-SEM, as indicated by the presence of P62, LC3, LAMP1 and TOMM20, indicating the occurrence of extracellular autophagy and lysosome-related degradation in the intercellular space of epithelial cells. Additionally, we examined IOR structural alterations during epididymal development and ageing as a model. Specifically, IORs are rare in the undifferentiated epididymal epithelium and occur between differentiated epithelial cells. These IORs were reduced in the epididymides of aged mice (21 months old), suggesting the physiological importance of homeostasis in the epididymal epithelium.

Epithelial cells are long-lived, highly differentiated cells that maintain cellular homeostasis by ensuring internal stability. One of the important ways of maintaining cellular homeostasis is through autophagy, a process that removes misfolded or damaged biomolecules and dysfunctional organelles27. Autophagy is an intracellular process in which proteins and organelles are delivered to lysosomes for degradation28; subtypes, such as mitophagy and ER-phagy, clear excess or impaired organelles29. These processes safeguard cellular survival under stress by recycling cellular constituents30,31. In the male reproductive tract, autophagy is critical during spermatogenesis32,33,34,35, but its role in the epididymis is less understood. Ultrastructural studies using EM36 and cryoelectron tomography (cryo-ET)37,38 have revealed the progression of AP biogenesis and phagophore formation. Consistent with prior observations, 3D reconstructions of IOR components in our study not only revealed autophagy-like structures but also captured the dynamic communication of autophagy-like structures in the IOR (Fig. 2). We therefore detected key autophagy molecules (P62, LC3, and LAMP1) in the IOR, indicating potential intercellular degradation activities within the IOR. Furthermore, our scRNA-seq results revealed lower expression of autophagy-related genes in PCs than in other cell types, suggesting a limited supply of organelle clearance machinery. IORs in the LISs of these cells may facilitate efficient and cooperative elimination of damaged or surplus cytoplasmic components. Additionally, we found low levels of the autophagy machinery in the intestine, indicating potential IORs as well. Corroborating this finding, similar IORs have been observed in the small intestine (Fig. 7).

Given the detection of autophagic and lysosomal signals within these IORs, dysfunctional or surplus organelles and proteins may be processed through secretory autophagy39,40,41,42,43 or lysosomal exocytosis, which release their contents into the extracellular space. However, the fate of these secretory vesicles upon extracellular release remains largely unexplored. In contrast to previous reports, our results revealed a unique intercellular “organelle” IOR in which the autophagic and lysosomal processes exist between epithelial cells. Rab27, a small GTPase, is known to be involved in regulating secretory processes within the cell, particularly in the context of exosome release, lysosomal exocytosis and secretory granule exocytosis in various cell types44,45,46. Intriguingly, Rab27a has been reported to regulate the docking of multivesicular endosomes (MVEs) at the plasma membrane22, as well as the secretion of extracellular vesicle47 and exosome48,49. This suggests that the release of autophagic-structure and MVBs into IORs may be mediated via a Rab27a-dependent pathway. In the present study, we observed that the presence of IORs was significantly reduced in the Rab27a-deficient ashen epididymis compared to the control, as well as the decrease of LRO and MVBs within IORs. These findings suggest a potential mechanism by which Rab27a regulates IOR formation. While intracellular transport of vesicles and organelles has traditionally been viewed as microtubule-dependent for long distances and actin-dependent for short distances50, recent studies have challenged this paradigm. Emerging evidence demonstrated that actin/myosin networks facilitated long-distance organelle transport51,52. Furthermore, Rab27a has been shown to play a pivotal role in actin-dependent long-distance transport coordination by regulating organelle-bound motors and tracking assembly proteins24. Considering the capacity of actin to facilitate long-range organelle transport and the established role of Rab27a in this process, our results suggest a model in which Rab27a might directly regulate IOR formation. Further studies are warranted to elucidate the detailed mechanisms underlying this process. Additionally, we observed the budding process of mitochondrial residuals in our FIB-SEM results (Fig. 2c). This process may be dependent on the tyrosine kinase SPE-8-AF-mediated mitopherogenesis, which allows the export of healthy mitochondria from cells53. Given these findings, the transportation or secretion of mitochondrial residuals may be facilitated by AFs.

Epithelial tissue is traditionally understood to have minimal to no intercellular space containing extracellular materials between cells54,55,56. Early studies reported the presence of the LIS for ion transport in the skin epithelium, with such LISs considered to connect with the basal space and facilitate ion exchange57,58,59. A recent study by Dumortier et al. revealed that the accumulation of pressurized fluid causes the formation of intercellular microlumens in the blastocoel lumen60. Similarly, our study revealed that the IOR forms numerous closed vacuole-like spaces within the LISs of epithelial cells, which are organized in a string-of-pearls-like structure throughout the caput, corpus, and cauda segments of the mouse epididymis. Interestingly, both microlumina and IORs were identified with Ecad (Cdh1), suggesting a similarity in their formation and characterization, such as hydraulic fracturing of cell‒cell contacts and/or junction formation. In epididymal epithelial cells, cellular junctions play critical roles in maintaining epithelial integrity and functions, including adherens junctions61,62, tight junctions6, gap junctions63 and desmosomes8. Notably, Cyr et al. used TEM to show that the IS of the rat epididymis had the most extensive tight junctions, with fewer tight junctions but more desmosomes observed in other regions64. Our findings clearly revealed that IORs were stabilized by desmosomes and that few existed in the IS, providing a explanation for why more desmosomes appeared along the cells.

Another observation is that the IOR contains multiple degraded or degrading organelles. This finding raises the question of the fate of the degraded molecules. The most plausible explanation is that these molecules can be recycled into epithelial cells. In support of this hypothesis, amino acid transporters or protein transporters may act as potential candidates for the recycling of reutilizable molecules. Amino acid transporters facilitate the transfer of amino acids into and out of cells or organelles, providing raw materials for cellular processes. We selected two amino acid transporters, SLC7A14 and SLC6A15. SLC7A14, a sodium-independent transporter of cationic amino acids (CAAs), has been implicated in retinitis pigmentosa65 and hepatic insulin sensitivity regulation through lysosomal GABA imports66. A recent study also linked SLC7A14 mutation to auditory neuropathy via lysosomal dysfunction67. SLC16A15, a sodium-dependent symporter for neutral amino acids, regulates the intracellular concentrations of neutral amino acids and contributes to brain function68. Our results revealed that SLC7A14 and SLC6A15 were located at the edge of the IOR. These findings suggest a potential role for these transporters in facilitating the uptake of amino acids degraded within the IORs, thereby contributing to epithelial homeostasis. Our findings suggest that the IOR between lateral epithelial cells provides a cooperative extracellular mechanism for eliminating and recycling cellular material, complementing intracellular organelle degradation processes.

Although we have discovered IOR structures using FIB-SEM technology, there are still some limitations. Firstly, the dual-beam FIB-SEM we used requires chemical fixation, which may introduce certain artifacts. To validate our findings, we observed the IOR structures using various chemical fixation methods and high-pressure freezing substitution, indicating that the formation of IOR is not due to the fixation methods. Nonetheless, we will attempt to combine high-pressure freezing substitution with continuous GridTape TEM to observe the characteristics of IOR structures under non-fixed conditions in the following studies for three-dimension reconstruction. Secondly, the volume size obtained in this study is still insufficient to simultaneously achieve the goal of capturing the entire lumen and maintaining high resolution. Besides FIB-SEM, current vEM technologies include automated serial TEM (TEMCA and GridTape TEM), fast ion beam technologies (plasma and gas cluster ion beam systems), and sequential imaging of serial sections on a substrate (array tomography)69,70. By using these technologies in the future, we might observe the distribution of IORs in the entire lumen or millimeter-scale and examine their potential impact on epididymal epithelial cells. Thirdly, current vEM techniques offer limited functional insights as they capture fixed moments. We aim to use transgenic mouse models combined with correlative light and volume electron microscopy (vCLEM) to further explore IOR formation, functions, and mechanisms in future studies.

Taken together, our findings elucidated epithelial cell homeostasis, where organelles or proteins undergo degradation and recycling via a extracellular “organelle” termed the IOR. Shared IORs play critical roles in maintaining epithelial cell fitness and facilitating cell‒cell communication between the lateral intercellular spaces of epithelial cells, providing a foundation for a better understanding of epithelial cell homeostasis within epithelial tissues.

Methods

Animals

C57BL/6J, C3H/HeJ and ICR/HaJ mice were purchased from GemPharmatech Co., Ltd. (Nanjing, PR. China), and maintained at the Laboratory Animal Center of Nantong University. All procedures followed the guidelines for animal experimentation with approval from the Animal Ethics Committee of Nantong University (Reference number: 20181229-001). PhAM-floxed male mice (Jackson Laboratory, Strain #: 018385) were bred with DDx4-Cre female mice (Shanghai Model Organisms Center, Shanghai, PR. China) to generate chimeric transgenic mito-Dendra2DDX4/+ mice with fluorescence in the mitochondria. Rab27a mutant mice (C3H/HeSn-Rab27aash/J, ashen mice, Jackson Laboratories) were gifted by Professor Shumin Duan46 at Zhejiang University. C3H/HeSnJ mice were used as controls.

Tissue sample isolation and fixation

The mice were anaesthetized via an intraperitoneal injection of 2.5% avertin (346 mg/Kg) and transcardially perfused with 0.9% NaCl followed by 4% paraformaldehyde (PFA, P804536-500g, Macklin). The epididymides were then collected for further fixation at 4 °C overnight.

For various fixation methods, the mice were anaesthetized via an intraperitoneal (i.p.) injection of 2.5% avertin (346 mg/Kg), and perfusion or nonperfusion protocols were performed. The groups were as follows: 4% PFA perfusion; nonperfusion and postfixation with 4% PFA; nonperfusion and postfixation with 4% PFA + 0.1% glutaraldehyde (GA, 18426, Ted Pella, Inc.); nonperfusion and postfixation with 4% PFA + 0.5% GA; and nonperfusion and postfixation with 4% PFA + 2.5% GA at 4 °C overnight.

Sample preparation for FIB-SEM

Epididymis or small intestine were trimmed into different segments (2 mm × 2 mm × 2 mm) and fixed for 24 h in 2.5% GA (18426, Ted Pella, Inc.) with 0.1 M PBS (pH 7.4) as previously described71. The sample blocks were postfixed with an osmic acid solution (2% osmic acid, 18459; Ted Pella: 3% potassium ferricyanide, 455989; Sigma = 1:1) at 4 °C for 1 h. The samples were incubated in 1% thiocarbohydrazide (223220, Sigma) for 20 min at room temperature (RT) and fixed with 2% aqueous OsO4 for 30 min at RT. After incubation with 1% aqueous uranyl acetate overnight at 4 °C, the samples were incubated in 0.66% lead nitrate (203580, Sigma) diluted in 0.03 M/L aspartic acid (11189, Sigma) (pH 5.5) for 30 min at 60 °C. The samples were then dehydrated in alcohol gradients (50%, 70%, 90%, 100%, 10 min each), penetrated with resin (EPON812, 14900, EMS, USA) and flat‐embedded for 48 h at 60 °C.

FIB-SEM image collection

Resin blocks of epididymal segments or small intestine were carefully trimmed via a trimmer (EM TXP, Leica). The resin blocks were then coated with platinum using an ion sputtering apparatus (EM ACE200, Leica). A scanning electron microscope (Teneo vs., Thermo Fisher) with an ultramicrotome in its specimen chamber was employed to obtain the target area. After the target area was pinpointed, the resin blocks were transferred to a 45° prefilled SEM stub and coated with platinum via an ion sputtering apparatus (EM ACE200, Leica) for 300 s. Imaging was performed via dual-beam SEM (FIB Helios G3 UC, Thermo Fisher). Data were collected in serial surface view mode with a slice thickness of 5 nm at 30 kV and 0.79 nA. Each serial surface was then imaged with a 2 kV acceleration voltage and 0.2 nA current in backscatter electron mode with an in-column backscatter electron detector (ICD). The tilt angle of the sample stage was 7° during milling. The image frame size was set to 4096 × 3536 pixels with a dwell time of 2 μs per pixel.

Three-dimensional reconstruction and analysis of cells and vesicles

All 3D segmentations were generated via Amira software (Version 6.5.0, Thermo Fisher, USA). Raw Tiff images (1300 ~ 1500/sample) were imported into Amira for alignment. The resolution and image quality were improved by the command “filter sandbox” with the property parameters “3*3*3”. Direct 3D reconstruction, volume rendering and visualization of the nucleus, cell membranes, and IOR were conducted manually upon careful visual inspection with Amira software (Version 6.5.0, Thermo Fisher, USA).

Segmentation of autophagy-like structures in the IOR

IORs containing autophagy-like structures were segmented with subvolume extraction created in Amira. A Gaussian filter was applied (standard deviation: x:1, y:1, z:1; kernel size factor: 1) prior to segmentation. The “magic wand” tool was used with masking thresholds for automated selection (phagophore: 90-140; AP: 60-170). Any incorrect selections were manually corrected by adding a “draw limit line” and using the “brush” function. The 3D rendering produced by “volume rendering” was further smoothed with the cubic mode.

Immunofluorescence

All samples were postfixed and subsequently dehydrated with 10%, 20% and 30% sucrose solutions (Sangon, Shanghai, China) overnight at 4 °C. The samples were embedded in optimum cutting temperature compound (O.C.T., Tissue-Tek® OCT compound, Sakura, USA) and frozen in liquid nitrogen. Sections were microtomed on a Leica CM1950 clinical cryostat (Leica Biosystems) at a thickness of 10 μm and collected onto adhesion microscope slides (Citotest, Nantong, China). The tissue sections were heat-fixed at 37 °C for at least 1 h, and the slides were stored at −80 °C.

The sections were permeabilized with 100% methanol for 12 min and blocked with 10% bovine serum albumin (4240GR100, BioFroxx) for 1 h at RT. The samples were then incubated with diluted primary antibodies (Table S1) overnight at 4 °C. The next day, the samples were washed 6 times with PBST, and the secondary antibody (Table S1) was added at RT for 1 h. The nuclei were stained with DAPI solution (C1006, Beyotime) for 30 min. Slides were mounted in Abberior Mount Solid Antifade (MM-2013-2X15ML, Abberior) and imaged on a Leica Thunder wide-field microscope (Thunder 3D Assay, Leica). Confocal images were acquired with a TCS SP8 microscope (Leica, Germany) and deconvolved with Huygens software (Scientific Volume Imaging). Z-stack images were acquired at 350 nm intervals over a thickness of 1–2.5 μm. Imaging and image processing were performed with ImageJ (2.0.14, NIH). All the images are displayed as raw data unless otherwise noted. ImageJ (2.0.14, NIH) was used to outline the cell and IOR, after which the positive signal fluorescence intensity was calculated.

Pre-embedding immunolabelling for TEM

The epididymis were perfused and postfixed with 4% PFA + 0.5% glutaraldehyde in 0.1 M Sorensen’s phosphate buffer (pH = 7.2). Subsequently, 100 μm sections were treated with 50 mM glycine (G8200, Solarbio) and washed in PBS. The sections were then permeabilized with 0.1% Triton X-100 for 30 min and blocked in 0.1% BSA in PBS, followed by an overnight incubation at 4 °C with the following primary antibodies: rabbit anti-LAMP1 (1:50, Cell Signaling Techno, #99437), rabbit anti-LC3 (1:250, Cell Signaling Technology, #12471), rabbit anti-p62 (1:100, ProteinTech, #18420-1-AP), and rabbit anti-TOMM20 (1:250, Servicebio, #GB111481) in a humidifying chamber. After being washed with PBS, the sections were incubated with Nanogold-Fab’ Goat Anti-Rabbit IgG conjugated with 1.4-nm gold particles (Nanoprobes, #2003) for 1 h at RT, followed by an overnight incubation at 4 °C. After another PBS wash, the sections were fixed in 2.5% glutaraldehyde for 2 h. Silver enhancement was performed via an HQSilver kit (Nanoprobes, #2012) according to the manufacturer’s instructions. The sections were subsequently fixed in 1% osmium tetroxide and contrast-stained with 0.5% uranyl acetate. The samples were dehydrated in a gradient series of ethanol, infiltrated and embedded in a low viscosity embedding kit (Electron Microscopy Sciences, #14300). After carbon evaporation, the grids were imaged via an FEI Tecnai Spirit T 120 kV transmission electron microscope (FEI, Thermo Fisher, USA).

High-pressure freezing (HPF) and substitution

The corpus epididymides were dipped into an external cryoprotectant (1-hexadecene) and carefully loaded into HPF sample carriers (0.2 mm depth, 16770141 Type A, Leica). The lid (16770142 Type B, Leica) was also dipped in 1-hexadecene and placed on top of the sample carrier. The samples were frozen using a Leica EM ICE HPF device.

For freeze substitution, the corpus epididymides were cryosubstituted in EM AFS2 (Leica). The sample chamber was filled with ethyl alcohol to keep the temperature stable. The sample carriers were transferred into caps containing 1% osmium tetroxide (OsO4) in acetone. The freeze substitution program was as follows: Step 1: −108 °C for 1 h; Step 2: −108 °C to −90 °C for 2 h; Step 3: −90 °C for 60 h; Step 4: −90 °C to −60 °C for 15 h; Step 5: −60 °C for 24 h; Step 6: −60 °C to −30 °C for 15 h; Step 7: −30 °C for 15 h; Step 8: −30 °C to 0 °C for 3 h.

For Epon embedding, the samples were washed with acetone three times for 30 min each at 4 °C, followed by two washes with acetone for 30 min each at RT. Acetone was then substituted with a 7:3 mixture of acetone and Epon resin, and the samples were incubated for 8 h. This solution (7:3 acetone/Epon) was replaced with a 3:7 acetone/Epon mixture for 12 h, followed by the addition of pure Epon for another 12 h. The Epon resin was replaced with fresh pure Epon and incubated for another 12 h. The samples were polymerized in an oven at 60 °C for 2 days. Thin 70-nm sections were obtained via an ultramicrotome (EM UC7, Leica) with an ultradiamond knife (DiAtome, USA). After carbon evaporation, the grids were imaged with an FEI Tecnai Spirit T 120 kV transmission electron microscope (FEI, Thermo Fisher, USA).

scRNA-seq data analysis

To investigate the expression of autophagy-related genes across various mouse tissues, we utilized single-cell profiling data from the GSE132042 dataset, which encompasses a broad range of mouse tissues26. Additionally, we incorporated our previous single-cell data specific to mouse epididymal cells from the GSE159713 dataset10. The analysis was conducted via the standard scRNA-seq analysis pipeline of the SynEcoSys database (accessible at https://synecosys.cn/). For data integration and visualization, Seurat software (version 4.1.1)72 facilitated the merging of datasets, enabling comprehensive analysis across tissue types. The results were visually represented through a dot plot.

Statistics

GraphPad Prism 9 software was used to analyze the experimental data. Box-and-whisker plots represent the maximum (upper whisker) and minimum (lower whisker) values, 75th (top of the box) and 25th (bottom of the box), the average (magenta line) and median (blue line in the box). The sample size was not predetermined. If not otherwise stated, each experiment was repeated independently three times. Unpaired t-tests were used to analyze two-group data. Ordinary one-way ANOVA was used to analyze significance in more than two groups. Error bars of all the data are presented as the means ± SEM. Values of *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001 were considered significant.

Reporting summary

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