A nanoscale atlas of extracellular vesicles and particles in Drosophila olfactory sensilla

  1. Pawel Vijayakumar
  2. Kalyani Cauwenberghs
  3. Muskaan Garg
  4. Jonathan Choy
  5. Shadi Charara
  6. Quintyn McKaughan
  7. Chih-Ying Su

  • Curated by eLife

    eLife logo

    eLife Assessment

    This important study presents an ultrastructural atlas of extracellular vesicles and non-vesicular particles in Drosophila olfactory sensilla, using cryofixation-based serial block-face electron microscopy to catalogue roughly 7,800 particles across four antennal regions. The evidence is compelling: the cryofixation minimizes fixation artifacts, and the rigorous, well-validated methodology and morphometric analyses provide strong support for the structural classifications and spatial distributions. This work provides a foundational resource for researchers studying extracellular-particle signaling in sensory organs.

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Extracellular particles, including non-vesicular extracellular particles (NVEPs) and extracellular vesicles (EVs), are emerging as key contributors to sensory signaling, yet their ultrastructural organization within native tissues remains underexplored. Native tissues preserve extracellular particle organization and heterogeneity that are often lost during dissociation. Using cryofixation-based serial block-face scanning electron microscopy, we generated a nanoscale atlas of NVEPs and EVs across 352 Drosophila antennal olfactory sensilla (∼70% coverage). Segmentation of more than 7,800 extracellular particles revealed distinct populations differing in morphology, size, electron density, and sensillum-class distribution. Analyses of EV biogenesis identified multivesicular bodies and putative membrane budding in auxiliary cells. Furthermore, sensilla housing degenerating neurons exhibited marked accumulation of EVs and NVEPs, accompanied by increased auxiliary-cell EV biogenesis. These findings provide a large-scale ultrastructural characterization of extracellular particles in native sensory tissues and establish a foundation for understanding how extracellular particles are distributed, generated, and function during sensory signaling and degeneration.

Article activity feed

  1. Version published to 10.7554/elife.112252.2 on eLife
  2. Version published to 10.7554/elife.112252 on eLife
  3. eLife

    eLife Assessment

    This important study presents an ultrastructural atlas of extracellular vesicles and non-vesicular particles in Drosophila olfactory sensilla, using cryofixation-based serial block-face electron microscopy to catalogue roughly 7,800 particles across four antennal regions. The evidence is compelling: the cryofixation minimizes fixation artifacts, and the rigorous, well-validated methodology and morphometric analyses provide strong support for the structural classifications and spatial distributions. This work provides a foundational resource for researchers studying extracellular-particle signaling in sensory organs.

  4. eLife

    Reviewer #1 (Public review):

    Summary:

    Using cryofixation and serial block-face electron microscopy (SBEM), P. Vijayakumar and K. Cauwenberghs characterize extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) within native Drosophila olfactory sensilla. The study provides a unique and valuable dataset comprising approximately 7,800 extracellular particles, systematically describing their morphology, size, density, and distribution. The ultrastructural analysis across different sensillum classes offers insights into the potential biogenesis and functions of these extracellular particles.

    Strengths:

    Cryofixation preserves EVs and NVEPs within native tissue conditions. The detailed quantification of a very large dataset provides a unique source of information on extracellular particle number, categories, distribution. The expertise of the group in the method and the tissue explored as well as their detailed quantification provide confidence in the dataset and observations.

    Weaknesses:

    All weaknesses previously observed have been addressed.

  5. eLife

    Reviewer #2 (Public review):

    Summary:

    This paper presents a large structural survey of extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) in the olfactory sensilla of Drosophila melanogaster. Using high-pressure freezing and serial block-face SEM, the authors avoid many of the artifacts associated with conventional fixation and analyze more than 7,800 particles across 352 sensilla. The manuscript maps the distribution of these particles, describes their morphological heterogeneity, and examines their likely origins across different sensillum classes in both normal and degenerating tissue.

    Strengths:

    The strongest aspect of the paper is the imaging. Preservation is painstakingly controlled. The cryofixation appears to preserve the sensillum lymph in a convincing native state than standard preparation methods, giving this work gravitas. Further the authors characterized thousands of particles, further making this data strong.

    The figures are strong. They are clear, easy to read, and generally well designed, I think people will use this paper as a model on how to present complex data in a concise and straightforward manner. The manuscript is careful in how it presents the dataset and does not overinterpret the descriptive observations. As an ultrastructural resource, this paper will be useful to the field. The identification of auxiliary support cells as major secretory sites, together with the striking accumulation of EVs in degenerating tissue, will provide a useful starting point for future work.

    Comments on revised version.

    The authors have completely addressed my concerns from the initial review. The revised text and Methods now clearly explain the vesicle categorization specifically how "dense," "cargo-filled," and "double EVs" are defined as strictly morphological groupings rather than verified biological classes. This is a rigorous and valuable ultrastructural resource.

  6. eLife

    Reviewer #3 (Public review):

    Using cryofixation-based serial block face electron microscopy of several subregions of the Drosophila antenna, the authors segment and assemble a high-resolution atlas of the anatomical structure, density, and spatial distribution of extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) in different Drosophila olfactory sensilla types. This systematic and thorough description is an important prerequisite to understanding the function of extracellular particles in intercellular signaling in the nervous system. Additionally, they describe examples of putative biogenesis events (budding/fusion), as well as neuronal and axillary cell degeneration events and measure the changes in particle accumulation in these altered microenvironments. Overall, this is a significant and comprehensive analysis and represents an invaluable resource to this burgeoning field. The authors assemble an important dataset and their claims match the level of evidence provided.

    Strengths:

    (1) The authors use segmentations from four different patches of the antenna to provide the systematic ultrastructural survey of extracellular particles in native insect sensilla. The dataset captures the diversity of sensilla types and reconstructs ~7800 particles.

    (2) We commend the authors from making the EM volumes and the segmented 3D models publicly available, through Dryad Digital Repository and the Cell Image Library, respectively. These datasets will be valuable resources to the community in future work.

    (3) The sample preparation technique appears to minimize typical artifacts associated with chemical fixation, as evidenced by the high reported sphericity of EVs.

    Comments on revised version.

    The authors have succinctly and effectively addressed all our major concerns, which mostly related to clarification of methods and analysis/categorization choices. The inclusion of continuous data visualizations to complement categorical descriptions (for particle position along the sensillum length and vesicle size) is very nice and gives the reader a fuller view of the dataset. The revised manuscript answers all our remaining questions and allows the reader to better understand the motivations for the specific categorical analyses applied.

  7. eLife

    Author response:

    The following is the authors’ response to the original reviews.

    Public Reviews:

    Reviewer #1 (Public review):

    Summary:

    Using cryofixation and serial block-face electron microscopy (SBEM), P. Vijayakumar and K. Cauwenberghs characterize extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) within native Drosophila olfactory sensilla. The study provides a unique and valuable dataset comprising approximately 7,800 extracellular particles, systematically describing their morphology, size, density, and distribution. The ultrastructural analysis across different sensillum classes offers insights into the potential biogenesis and functions of these extracellular particles.

    Strengths:

    Cryofixation preserves EVs and NVEPs within native tissue conditions. The detailed quantification of a very large dataset provides a unique source of information on extracellular particle number, categories, and distribution. The expertise of the group in the method and the tissue explored, as well as their detailed quantification, provide confidence in the dataset and observations.

    We thank the reviewer for their positive assessment of our work. We are pleased that the value of cryofixation for preserving native ultrastructure, as well as the scale and rigor of our quantitative dataset, were recognized. We hope this resource will be useful to the broader community studying extracellular communication in sensory systems.

    Weaknesses:

    Major comments

    (1) As the authors state, the rare observation of EV budding or MVB release events suggests that these are transient processes, whereas EVs and NVEPs are retained for relatively long periods within the sensillum lumen. The current analyses may overinterpret steady-state vesicle abundance as differences in vesicle production.

    We agree with this important distinction. In our manuscript, particle number reflects steady-state prevalence within each sensillum class, not production rate. Biogenesis events were analyzed separately as transient indicators of production, with the caveat that EM captures static snapshots rather than dynamic rates. We have now clarified this explicitly in the first paragraph of the “EV and NVEP number” section in the Results. The revised text now reads:

    “To compare particle abundance across sensillum classes, we counted EVs and NVEPs within the sensillum lumen. These snapshot counts reflect the steady-state pool at fixation; they do not capture dynamic rates of production or turnover. Quantification of particle number per sensillum revealed class-dependent differences ...”

    (2) Given the above conclusion, differences between sensillum classes may be somewhat overstated.

    (a) The absolute number of EVs and NVEPs per sensillum is highly variable, even within the same sensillum class (Figure3D). For example, a substantial proportion of coeloconic sensilla have an empty lumen (Figure 3C). Consequently, expressing the data as ratios or proportions (Figures 3B, 4C, and 4E) may exaggerate differences between sensillum classes and should therefore be interpreted with caution.

    We thank the reviewer for the comment. To prevent ratio distortion, we excluded sensilla containing fewer than 5 particles from ratio calculations in Figure 3B, as noted in the figure legend. Ratios derived from low particle counts are inherently unstable and skew toward extreme values (0 or 1), producing unreliable estimates. This exclusion criterion reduces within-class variance without artificially inflating between-class differences. We have also revised the "Relative abundance" Results section to avoid overstating our findings. The revised text now reads:

    “The median EV ratios of large basiconic and trichoid sensilla did not differ significantly from 0.5, whereas intermediate, small basiconic, and coeloconic sensilla showed median ratios significantly below 0.5, indicating a population-level NVEP bias (Figure 3B). Nevertheless, EV ratios varied considerably among sensilla of the same class, and distributions overlapped broadly across sensillum classes. Consequently, no pairwise differences between classes survived FDR-corrected Dunn’s post hoc comparisons despite a trend across classes (Kruskal–Wallis test, p = 4.76 × 10-2). Thus, while NVEPs were generally more abundant than EVs at the population level in most sensillum classes, both particle types co-occurred in individual sensilla with highly variable relative proportions.”

    In Figures 4C and 4E, sensilla that lacked EVs in the lumen were excluded from the analysis so they would not skew the proportional data. We also updated the figure legend to state this explicitly.

    (b) The rate of EV/NVEP production is unknown. For a similar rate of production across sensillum classes, Figure 3E suggests that the differences in lumen morphology and size may largely explain variation in EV density and distribution.

    We appreciate the opportunity to clarify this point. As the reviewer pointed out, if particle production rates were similar across sensillum classes, differences in lumen size could in principle explain variation in total particle number. However, our data in Figure 3E argue against this possibility. If lumen volume were the primary determinant of particle abundance, volume-normalized densities should be roughly uniform across all sensillum types. Instead, we observed the opposite pattern: smaller sensilla (e.g., coeloconic) exhibited substantially higher particle densities than larger sensilla (e.g. trichoid or large basiconic).

    This inverse relationship indicates that particle abundance is not a simple function of available lumen space. Rather, it suggests that underlying factors—such as differential production, release, or clearance rates—are specific to each sensillum class. We have revised the "Volume-normalized EV and NVEP abundance" Results section to make this distinction more explicit. The revised text now reads:

    “Sensillum size varied markedly across classes (Figure 1B). If EV and NVEP production rates were comparable across sensillum classes, one would expect differences in lumen size alone to account for variation in total particle number. To test whether particle counts scale with sensillum size, we compared EV and NVEP numbers normalized to lumen volume….”

    “If lumen volume were the primary determinant of particle abundance, volume-normalized particle abundance should be roughly uniform across all sensillum types. However, this was not observed. Smaller sensilla (e.g., coeloconic) exhibited substantially higher particle concentrations than larger sensilla (e.g. trichoid or large basiconic) (Figure 3E). This inverse relationship indicates that particle abundance is not a simple function of available lumen space.

    Rather, it suggests that underlying factors—such as differential production, release, or clearance rates—are specific to each sensillum class.”

    That said, I agree that ab1 sensilla display a striking enrichment of large cargo-filled EVs compared with the other sensillum classes, while coeloconic sensilla display enrichment in small dense filled EVs (Figure 4C). Together, large and cargo-filled EV observation provides strong support for differences in EV biogenesis between ab1 sensilla and the other sensillum classes. In that context, I also think the EV size distribution shown in Figure 4 - Figure Supplement 1 should be moved into the main figure, as it demonstrates that the majority of EVs in the ab1 lumen are relatively large and are therefore likely to represent microvesicles.

    We thank the reviewer for the insightful comment. As suggested, we moved the EV size distribution from Figure 4—Figure Supplement 1 into the main figure (revised Figure 4D) to highlight that ab1 EVs are predominantly large and likely represent microvesicles.

    Could you clarify why ab1 sensilla are only included in Figure 4 and not analysed in Figure 3?

    In Figure 3, ab1 sensilla were included within the large basiconic class rather than shown separately. We grouped them this way because we did not observe marked differences in particle abundance or EV ratios between ab1 and the other large basiconic sensilla examined (ab2 and ab3). Figure 4, in contrast, focuses on subtype-specific differences in EV electron density/cargo, where ab1 is analyzed separately because it displayed a distinct profile.

    (3) Approximately 10% of ORNs appear to be degenerating in 6-8-day-old flies, which seems unexpectedly high. This contrasts with the relatively infrequent occurrence of auxiliary cell apoptosis or complete sensillum degeneration. In these "degenerating ORNs", the authors state that the hallmarks of ORN apoptosis are restricted to the dendrites. As hallmarks, they state dendrite truncation, fragmentation and blebbing. Rather than apoptosis, I wonder whether these observations might instead represent ciliary truncation and ectosome shedding, followed by degradation of the shed ciliary membrane into EVs. Ciliary truncation and ectosome shedding, followed by ciliary regrowth, are dynamic processes that have been described across multiple species. This interpretation could explain large EVs that remain in the lumen long after the cilium has regenerated. It would reconcile this article with the general agreement that cilia are a prime site for the budding of EVs across species. Additional evidence supporting apoptosis of the ORNs would help distinguish between these possibilities. Otherwise, I believe the author should reconsider their interpretation.

    We thank the reviewer for suggesting the possibility that ciliary truncation and ectosome shedding could account for the affected dendritic membrane morphology. However, after careful reconsideration, we believe our original interpretation is better supported by the ultrastructural data. Below, we address the reviewer's specific concerns.

    (1) Number of degenerating ORNs: The reviewer's estimate of "~10% of ORNs" likely conflates sensilla with individual neurons. Among 352 sensilla analyzed, we identified 32 sensilla containing at least one degenerating ORN. However, because each sensillum typically houses 2–4 ORNs and not all ORNs within these sensilla were affected, the actual number of degenerating ORNs was 47 out of 802 surveyed ORNs—less than 6%, not ~10%. We have now clarified this calculation in the revised text to prevent confusion.

    (2) Morphological evidence: While ciliary ectosome shedding and truncation are well-documented across multiple species, the structural hallmarks we observe differ in two key respects.

    (1) Extent of membrane loss: Ciliary ectosomes are small vesicles (~100 nm) shed locally from the ciliary tip or base. By contrast, the degenerating ORNs in our data show severe fragmentation or truncation affecting >1/3 of the total dendritic length (Figure 7)—a scale of disintegration inconsistent with localized membrane turnover.

    (2) Absence of regrowth: Ciliary truncation is a transient event typically followed by regrowth. If the dendritic truncation we observed reflected this process, we would expect to find ORNs containing both truncated cilia and nascent regrowth. However, we found no evidence of concurrent regeneration in our datasets.

    (3) Consistent with prior reports of ORN apoptosis: Basal ORN apoptosis has been documented in 1-week-old Drosophila (Fernández-Hernández et al., 2020) and can be induced by innate immune signaling (Takeuchi et al., 2022). The ~6% frequency we observe is therefore biologically plausible. We cannot exclude that our genetic background (carrying the UAS-APEX2 transgene) may modestly elevate this baseline rate: APEX2 is a peroxidase, and even low-level leakage expression could increase reactive oxygen species and sensitize cells to stress-induced apoptosis. As stated originally in the text, this survey was not intended as a quantification of ORN degeneration. In the revision, we now clarified that the absolute numbers should be interpreted with caution given the transgenic background. The revised text now reads:

    “In addition to sensilla with intact ORNs, we identified 32 sensilla containing ORNs with morphological features of apoptosis (Figure 7). Consistent with previous observations of ORN apoptosis (Chihara et al., 2014; Fernández-Hernández et al., 2020; Takeuchi et al., 2022), affected dendrites exhibited hallmark features of apoptotic degeneration, including truncation, fragmentation, and membrane blebbing (Nagata, 2018; Xu et al., 2019). These abnormalities were observed across sensillum classes, including 6 coeloconic, 10 small basiconic, 7 large basiconic, 1 intermediate, and 8 trichoid sensilla. In some sensilla, both ORNs appeared affected, whereas in others only a single ORN exhibited signs of degeneration. Together, these sensilla accounted for 47 affected ORNs out of 802 surveyed (~6%). The observation of degenerating ORNs in these SBEM volumes provided a unique opportunity to examine the ultrastructure of apoptotic bodies and degeneration-associated extracellular particles in situ within native olfactory sensilla.”

    “We note that this survey was not intended to provide an exhaustive quantification of ORN degeneration or to estimate the frequency of degenerating ORNs. Our flies carry the UAS-APEX2 transgene, and even low-level leakage expression of this peroxidase might elevate reactive oxygen species and thereby increase susceptibility to apoptosis; therefore, these frequencies should be interpreted with caution. Rather, these observations demonstrate that ORN degeneration can occur across multiple sensillum classes and provided a unique opportunity to compare the morphological properties of EVs and NVEPs in sensilla with either degenerating or intact ORNs….”

    We appreciate the reviewer's thoughtful suggestion of an alternative model, but the severity of dendritic disintegration and the absence of regrowth signatures together argue against ciliary truncation and ectosome shedding. We have added a brief discussion acknowledging ciliary EV shedding as a physiological process in other systems, which reads:

    “Ciliary extracellular vesicle shedding is well documented across evolutionarily distant systems, from the flagella of Chlamydomonas to the sensory cilia of C. elegans and mammalian primary cilia (Wang & Barr, 2016, 2018). It is therefore notable that we did not capture active EV budding from ORN dendritic cilia in our adult Drosophila preparations. This absence likely reflects both biological and technical constraints: ciliary EV production may be developmentally regulated and reduced in 1-week-old adults relative to earlier stages, and static volume EM may miss rare, transient release events….”

    Reviewer #2 (Public review):

    Summary:

    This paper presents a large structural survey of extracellular vesicles (EVs) and non-vesicular extracellular particles(NVEPs) in the olfactory sensilla of Drosophila melanogaster. Using high-pressure freezing and serial block-face SEM, the authors avoid many of the artifacts associated with conventional fixation and analyze more than 7,800 particles across352 sensilla. The manuscript maps the distribution of these particles, describes their morphological heterogeneity, and examines their likely origins across different sensillum classes in both normal and degenerating tissue.

    Strengths:

    The strongest aspect of the paper is the imaging. Preservation is painstakingly controlled. The cryofixation appears to preserve the sensillum lymph in a more convincing native state than standard preparation methods, giving this work gravitas. Further, the authors characterized thousands of particles, further making this data strong. The figures are strong. They are clear, easy to read, and generally well designed; I think people will use this paper as a model for how to present complex data in a concise and straightforward manner. The manuscript is careful in how it presents the dataset and does not overinterpret the descriptive observations. As an ultrastructural resource, this paper will be useful to the field. The identification of auxiliary support cells as major secretory sites, together with the striking accumulation of EVs in degenerating tissue, will provide a useful starting point for future work.

    We thank the reviewer for this thorough and encouraging assessment. We hope this ultrastructural resource will prove useful to the field.

    Weaknesses:

    The main point that could use more clarification is the vesicle categorization. In particular, the distinction between "dense," "cargo-filled," and "double EVs" is not always easy to follow from a biological perspective. Some additional discussion of how the authors think these categories relate to one another, and whether they are intended as purely morphological groupings or as distinct biological classes, would strengthen the manuscript.

    We thank the reviewer for this important point. Our vesicle classifications are strictly morphological groupings based on electron density and membrane morphology in EM images. We do not have molecular markers that would allow us to assign these morphological classes to distinct biological functional categories. As suggested, we have revised the first paragraph of the "Variation in EV properties across sensillum classes" section to clarify that these are morphological categories and to describe how their interior electron densities relate to potential macromolecular content. The revised text now reads:

    “Among the 1,397 EVs described in the previous section, ultrastructural analysis revealed three categories based on intraluminal electron density: dense (n = 730), lucent (n = 421), and cargo-filled (n = 246) (Figure 4A). These are morphological categories strictly based on ultrastructure, not verified biological classes. Dense and lucent EVs, both displaying uniform intraluminal density, were defined relative to the surrounding sensillum lymph: dense EV interiors were similar to or more electron-dense than the lymph, whereas lucent EV interiors were less electron-dense than the surrounding lymph. In contrast, cargo-filled EVs showed heterogeneous intraluminal density: a light interior with discrete, electron-dense inclusions. These categories were readily distinguishable in SBEM images and observed across sensillum classes. They likely reflect differences in luminal content: dense and cargo-filled EVs appeared more macromolecule-rich than lucent EVs.”

    We hope this morphological classification will encourage future studies to correlate these ultrastructural categories with specific molecular markers and biological functions.

    Reviewer #3 (Public review):

    Summary:

    Using cryofixation-based serial block face electron microscopy of several subregions of the Drosophila antenna, the authors segment and assemble a high-resolution atlas of the anatomical structure, density, and spatial distribution of extracellular vesicles (EVs) and nonvesicular extracellular particles (NVEPs) in different Drosophila olfactory sensilla types. This systematic and thorough description is an important prerequisite to understanding the function of extracellular particles in intercellular signaling in the nervous system. Additionally, they describe examples of putative biogenesis events (budding/fusion), as well as neuronal and axonal cell degeneration events and measure the changes in particle accumulation in these altered microenvironments. Overall, this is a significant and comprehensive analysis and represents an invaluable resource to this burgeoning field. The authors assemble an important dataset and their claims match the level of evidence provided.

    Strengths:

    (1) The authors use segmentations from four different patches of the antenna to provide a systematic ultrastructural survey of extracellular particles in native insect sensilla. The dataset captures the diversity of sensilla types and reconstructs ~7800 particles.

    (2) We commend the authors for making the EM volumes available in the public Cell Image Library with accession numbers. It would be helpful to the community to also make the segmentations for this great resource easily accessible.

    (3) The sample preparation technique appears to minimize typical artifacts associated with chemical fixation, as evidenced by the high reported sphericity of EVs.

    We thank the reviewer for this positive and thorough assessment. As suggested, we also made the segmentations publicly available alongside the EM volumes. The information is available in the Materials and Methods, which reads:

    “The segmented 3D models (including EVs, NVEPs, ORNs, and sensilla) are available separately through the Dryad Digital Repository (DOI: 10.5061/dryad.3j9kd521p). Segmented structures were grouped according to their SBEM volume of origin (volumes 1–4) and organized into structural groups using the built-in object list dialog in IMOD (accessed via "Edit" and then "Object list" in the Model View window). This grouping allows users to extract morphometric information for selected structural groups. Detailed information about the object list function is provided in the IMOD user guide (https://bio3d.colorado.edu/imod/doc/3dmodHelp/objectList.html).

    Specific points:

    (1) In Figure 3C, the authors should include a continuous measure of particle distribution in the sensilla. Currently, the authors define three categories of particle localization. In the five examples shown in Figure 3A, the spatial distribution of these particles appears quite distinct across classes. For example, EVs/NVEPs in large and small basiconic sensilla are largely restricted to the area proximal to the base, with a limited number located more distally. In contrast, intermediate sensilla show a marked concentration of particles more distally.

    We thank the reviewer for this suggestion. Continuous measurements of extracellular particle position along the sensillum length are provided in Figure 3—figure supplement 1, where each particle is plotted as a fraction of its distance from the sensillum base.

    We agree that the five examples in Figure 3A reveal distinct spatial patterns across classes. However, these reconstructions were selected to illustrate morphological diversity rather than to represent the average distribution for each sensillum type. We observe considerable within-class variability, which is why we used categorical bins in Figure 3C to highlight the most consistent finding: that coeloconic sensilla uniquely concentrate most extracellular particles below the cuticle base. We have revised the text accordingly, which now reads:

    “The position of extracellular particles relative to the cuticle base is likely to inform their cellular targets and sites of action, particularly in the lumen above the cuticle base, where ORN outer dendrites reside. We therefore examined particle position distribution by plotting each particle’s relative distance from the sensillum base (Figure 3—figure supplement 1). This continuous distribution analysis revealed distinct, though variable, class-specific spatial patterns.”

    “To summarize these distributions, we additionally classified sensilla into three categories: no particles, particles present only below the cuticle base, or particles present above the base (including those also with particles below, as sensilla with particles exclusively above the base were rare) (Figure 3C). Most sensilla contained EVs and/or NVEPs, while few were entirely particle-free. Of note, coeloconic sensilla—the smallest sensillum type (Figure 1B)—showed the highest proportion entirely lacking particles, and when present, EVs and NVEPs were largely restricted below the cuticle base (Figure 3C; Figure 3—figure supplement 1) Thus, while most sensilla harbored particles above the base, coeloconic sensilla showed either no particles or particles restricted below the base, suggesting class-specific differences in sites of action.”

    (2) The conclusion of different EV ratios across sensillum classes stems from a Kruskal-Wallis of p = 0.0476, with none surviving pairwise comparisons. This is not a strongly supported conclusion and is probably better characterized as a trend.

    We thank the reviewer for the thoughtful comment. As suggested, we revised the "Relative abundance" Results section to avoid overstating our findings. The revised text now reads:

    “The median EV ratios of large basiconic and trichoid sensilla did not differ significantly from 0.5, whereas intermediate, small basiconic, and coeloconic sensilla showed median ratios significantly below 0.5, indicating a population-level NVEP bias (Figure 3B). Nevertheless, EV ratios varied considerably among sensilla of the same class, and distributions overlapped broadly across sensillum classes. Consequently, no pairwise differences between classes survived FDR-corrected Dunn’s post hoc comparisons despite a trend across classes (Kruskal– Wallis test, p = 4.76 × 10−2). Thus, while NVEPs were generally more abundant than EVs at the population level in most sensillum classes, both particle types co-occurred in individual sensilla with highly variable relative proportions.”

    (3) The statement "selective enrichment of large, cargo-filled vesicles within the ab1 lumen suggests specialized EV-mediated communication adapted to the coordination demands of this neuronal population" seems speculative for a Results section without supporting functional evidence. It would seem better suited for the Discussion.

    As suggested, we moved the statement from the Results section to Discussion.

    (4) Figure 4: Criteria for defining the classes of EVs.

    (a) The authors should explain the rationale for classifying EVs using relative density rather than absolute density? We would expect EVs with similar contents to have similar electron density (similar darkness in the images). Would classifying them relative to the background, which itself might vary across sensilla or regions, create a possible confound, especially when comparing across sensillum classes?

    We chose relative electron density to normalize for variation in staining and contrast across SBEM volumes. Absolute pixel values would be confounded by image-level differences in acquisition and processing. Using the local lumen as an internal reference provides a consistent, scalable metric for classifying thousands of particles as electron-lucent or electron-dense.

    We acknowledge that background lumen density can vary subtly between sensillum types. For this reason, we treat relative density as a qualitative binary classification (lucent vs. dense) rather than a continuous numeric scale, and we have clarified this rationale in the Methods, which reads:

    “Classification of EV luminal density and morphology

    The inner luminal electron density of individual extracellular vesicles (EVs) was visually classified into three categories: dense, lucent, and cargo-filled. To account for variations in staining intensity and contrast across EM images, EV luminal density was evaluated relative to the surrounding lumen background of the same image. Dense EVs exhibited luminal density similar to or greater than that of the surrounding sensillum lumen, whereas lucent EVs displayed lower luminal density. Both dense and lucent EVs were defined by the uniformity of that density throughout the EV interior. On the other hand, cargo-filled EVs contained heterogeneous, electron-dense material within the lumen….”

    (b) The two example images (in Figure 4A) of the cargo-filled EVs appear to have different densities themselves. Do the cargo-filled ones also display systematic differences in density and, if so, why is this another class instead of being a subcategory within the dense and lucent classes (i.e. dense with/without cargo, lucent with/without cargo)? The dense and lucent classes are defined by their density, whereas this is a more structural property.

    We classified EVs by the uniformity of their internal density rather than by overall density alone. Lucent and dense EVs exhibit homogeneous interiors—uniformly light or uniformly dark, respectively. Cargo-filled EVs are defined by heterogeneity: a lighter background containing discrete, highly electron-dense inclusions. Because this structural distinction reflects fundamentally different cargo packaging states, we treated cargo-filled EVs as a separate class rather than as a subcategory of lucent or dense EVs We have clarified in the Methods (see above) and Results (see below) that uniformity versus heterogeneity is the operative criterion.

    “Among the 1,397 EVs described in the previous section, ultrastructural analysis revealed three categories based on intraluminal electron density: dense (n = 730), lucent (n = 421), and cargo-filled (n = 246) (Figure 4A). These are strictly morphological categories based on ultrastructure, not verified biological classes. Dense and lucent EVs, both displaying uniform intraluminal density, were defined relative to the surrounding sensillum lymph: dense EV interiors were similar to or more electron-dense than the lymph, whereas lucent EV interiors were less electron-dense than the surrounding lymph. In contrast, cargo-filled EVs showed heterogeneous intraluminal density: a light interior with discrete, electron-dense inclusions. These categories were readily distinguishable in SBEM images and observed across sensillum classes. They likely reflect differences in luminal content: dense and cargo-filled EVs appeared more macromolecule-rich than lucent EVs.”

    (c) Regarding "Double" and "Ball-and-Socket" EVs, does the density vary between the two particles involved (e.g., does the inner structure consistently differ in density from the outer)?

    We thank the reviewer for this question. We found no consistent density differences between the inner structure and the EV lumen for either “Double” or “Ball-and-Socket” EVs.

    Double EVs are classified by the presence of an internal, membrane-bound vesicle within an outer vesicle. The luminal densities of the inner and outer compartments are typically similar, though occasional variations occurred.

    Ball-and-socket EVs are classified by their structural coupling—one vesicle fitting into a concave invagination of another. The “ball” and “socket” typically share similar densities (e.g., both lucent or both dense). While density variations can occur, they are not a consistent or defining feature of this morphological class.

    We revised the text accordingly, which reads:

    “Beyond density differences, we identified structural variants of EVs, including double EVs and ball-and-socket EVs. Double EVs, defined as vesicles that enclose one or more smaller vesicles…. The luminal densities of the inner and outer compartments were typically similar, though occasional variations occurred.”

    “In contrast, ball-and-socket EVs consist of two vesicles physically coupled in a joint-like configuration, with one vesicle partially invaginated into the other…. The “ball” and “socket” typically share similar densities (e.g., both lucent or both dense), although density variation is not a defining feature of this class….”

    (5) What was the rationale for the 200μm and 1000μm size cutoffs? A continuous distribution of maximum particle sizes would provide a clearer understanding of the data.

    We thank the reviewer for this question. The 200 nm and 1000 nm cutoffs were chosen to represent commonly used EV categories in the literature: <200 nm as the conventional upper limit for small EVs/exosome-like particles (endosomal/MVB origin), 200–1000 nm for medium EVs/microvesicle-like particles (plasma-membrane budding), and >1000 nm for large EVs or apoptotic bodies. These cutoffs facilitate comparison with prior studies and interpretation in the context of known biogenesis pathways.

    We agree that continuous size distributions are informative and have therefore shown them in Figure 8B. The continuous data reveal a modest but significant overall shift toward “smaller” NVEPs and EVs in degenerating sensilla at the population level.

    However, categorical analysis reveals an additional, biologically relevant pattern that continuous distribution summaries can obscure: because the mean is disproportionately weighted by the most abundant particle type, the predominance of small EVs can mask the appearance of a less numerous but distinct larger population. Intact sensilla frequently contain mainly small EVs (<200 nm) (27%), a category entirely absent in degenerating sensilla (Figure 8C), whereas very large EVs (>1000 nm) are rare in intact sensilla (5%) but common in degenerating sensilla (~38%). Thus, although medium-to-very-large particles are outnumbered by small EVs and therefore contribute little to the overall mean, they represent a clear emergent population in degenerating sensilla. In other words, degeneration produces more particles across sizes and, importantly, adds a population of medium-to-very-large particles.

    We have revised the Results to (1) state the literature-based rationale for the 200 nm and 1000 nm cutoffs, (2) point readers to the continuous distribution in Figure 8B, and (3) explicitly explain how the continuous and categorical analyses provide complementary insights into degeneration-associated changes in particle size and abundance. The revision now reads:

    “However, a striking pattern emerged when sensilla were categorized by the size of their largest EVs. Approximately 27% of intact sensilla only contained small EVs (<200 nm, the likely size range of exosomes), whereas this category was absent in degenerating sensilla (Figure 8C). Conversely, only ~5% of intact sensilla contained very large EVs (>1000 nm, large EVs and apoptotic bodies), compared to ~38% of degenerating sensilla.”

    “The categorical analysis reveals an additional, biologically relevant pattern that the continuous distribution alone can obscure: degeneration-associated extracellular particle accumulation is primarily driven by increased particle abundance across all size ranges. Notably, all degenerating sensilla contained EVs >200 nm, and a substantially larger fraction contained very large EVs (>1000 nm). Although apoptotic bodies are highly heterogeneous in size and can range broadly from ~50 nm to several microns (Elsner et al., 2023), these large EVs fall within the upper size range commonly associated with apoptotic bodies (Battistelli & Falcieri, 2021; Shahi et al., 2024).”

    Recommendations for the authors:

    Reviewer #1 (Recommendations for the authors):

    (1) Figure 1: For readers outside the field, a schematic illustrating the organization of a sensillum would be extremely helpful. Specifically, showing the positions of auxiliary cells relative to the ORN soma, inner dendrite, outer dendrite, and sensillum lumen would facilitate interpretation of the possible origins and fate of EVs and NVEPs.

    We thank the reviewer for this helpful suggestion. We have added Figure 1—figure supplement 1 to the revised manuscript as suggested.

    (2) Figure 4 (and possibly Figure 2C): The size distributions of dense, lucent, and cargo-filled EVs appear to be multimodal, suggesting at least two distinct populations. This could indicate heterogeneity in EV biogenesis. Could the authors analyse these distributions further?

    We thank the reviewer for the careful observation. As suggested, we analyzed the distributions further with Hartigan's dip test for multimodality.

    Hartigan's dip test measures the maximum vertical distance (the "dip") between the sample distribution and the closest unimodal distribution function. This non-parametric approach directly tests whether a dataset contains a single peak (unimodal) versus multiple peaks (multimodal) without assuming a standard Gaussian curve or requiring specific parametric distribution shapes. High p-values (p > 0.05) indicate that the data is consistent with a single, unimodal population.

    For Figure 2C, Hartigan's dip test confirmed unimodality for overall EV (p = 0.990) and NVEP (p = 0.981) diameter distributions. For Figure 4B, EV density categories similarly showed unimodality for dense (p = 0.993), lucent (p = 0.993) and cargo-filled EVs (p = 0.951).

    We have updated the figure legends accordingly to reflect these changes in the revision.

    (3) Figure 4C and Figure 4-figure supplement 1: Most EVs in ab1 sensilla are large cargo-filled vesicles, whereas coeloconic sensilla contain the highest proportion of small dense EVs and are associated with more MVBs. These observations are, in my opinion, among the strongest pieces of evidence supporting differences in EV biogenesis between sensillum classes, yet they receive relatively little discussion.

    We thank the reviewer for raising this point. We have revised the Discussion accordingly, which now reads:

    “EVs are thought to arise through endosomal release via MVBs or direct plasma membrane budding (Doyle & Wang, 2019; Malkin & Bratman, 2020; Neyroud et al., 2022; Remans et al., 2025; Singh et al., 2024; Welsh et al., 2024). Our data suggest both pathways are active but differentially favored across sensillum classes. Of note, coeloconic sensilla contain abundant small, dense EVs alongside frequent intracellular MVBs, consistent with endosomal release. Conversely, ab1 sensilla are enriched in large, cargo-filled vesicles with fewer MVBs, suggesting a preference for direct plasma membrane budding. The scarcity of both MVBs and budding figures in intact sensilla likely reflects the transient nature of EV biogenesis.”

    (4) Figure 4D-F: The ball-and-socket EVs are particularly interesting. I assume these structures are observed only among the larger EVs (>200 nm), but this information is not provided. Is there any evidence that they represent intermediates between complex EV budding events and ultimately double EVs?

    As the reviewer pointed out, ball-and-socket EVs are predominantly >200 nm in diameter. However, we found no evidence that they represent budding intermediates. These structures are typically located distally in the sensillum lumen toward the sensillum tip, distant from the plasma membranes of ORNs or auxiliary cells where active budding or exocytosis would occur. They are also morphologically distinct from double EVs: in ball-and-socket configurations, the two participating vesicles are of comparable size, whereas double EVs consist of a small intraluminal vesicle fully enclosed within a substantially larger outer vesicle.

    While static EM cannot capture membrane dynamics, the distance from cell membranes and the size symmetry between the coupled vesicles suggest that ball-and-socket EVs represent physical interactions between mature, extracellular vesicles rather than biogenetic intermediates. We have clarified these structural distinctions in the revised Results, which reads:

    “In contrast, ball-and-socket EVs consist of two vesicles physically coupled in a joint-like configuration, with one vesicle partially invaginated into the other (Figure 4D). These structures are predominantly >200 nm and typically located distally in the sensillum lumen, distant from the plasma membranes of ORNs or auxiliary cells where active budding would occur.... Unlike double EVs, which comprise a small intraluminal vesicle enclosed within a substantially larger outer vesicle, ball-and-socket configurations involve two similarly sized vesicles. Their size symmetry and distance from cell membranes suggest they represent physical interactions between mature extracellular vesicles rather than budding intermediates....”

    (5) Figures 4 and 5: The origin of NVEPs is not addressed. One possibility is that EVs and NVEPs share a common origin, which would be consistent with their frequent co-occurrence (approximately a 0.5 ratio). Could the authors investigate whether NVEPs originate from MVBs or cargo-filled EVs by comparing the size of the dense particles they contain? Figure 5A suggests that MVBs contain densely packed material that may be informative.

    We thank the reviewer for this suggestion. NVEPs lack a bounding membrane, whereas MVBs contain membrane-bound intraluminal vesicles, making a direct MVB origin structurally inconsistent. We also saw no evidence—such as ruptured membranes or degradation intermediates—that cargo-filled EVs release their dense contents as free particles. We have therefore avoided speculative biogenetic models for NVEPs. The ~0.5 co-occurrence ratio more likely reflects shared extracellular accumulation than a common cellular origin.

    (6) In C. elegans, ectosomes are observed that are directly captured by auxiliary glia surrounding the cilia base; did the authors check for membrane budding at the level of the cellcell contact between dendrite and auxiliary cells?

    We thank the reviewer for this comparison with C. elegans. We examined ORN–auxiliary cell contact sites for membrane budding and for evidence that auxiliary cells capture EVs, but found neither at these interfaces.

    (7) This dataset is exceptional for follow-up and comparative studies. I noticed that the raw dataset is available through the Cell Image Library. Are these SEM volumes segmented or annotated for EVs and NVEPs?

    We thank the reviewer for this question. The Cell Image Library hosts the raw, unsegmented SBEM volumes. As suggested, we have made the segmented 3D models—including EVs, NVEPs, ORNs, and sensilla—available separately through the Dryad Digital Repository.

    We have updated the Materials and Methods to direct readers to both repositories.

    “The SBEM image volumes are available in the Cell Image Library (http://www.cellimagelibrary.org/home) under the following accession numbers: CIL:57519 (Volume 1), CIL:54614 (Volume 2), CIL:54609 (Volume 3), CIL:57579 (Volume 4). The segmented 3D models (including EVs, NVEPs, ORNs, and sensilla) are available separately through the Dryad Digital Repository (DOI: 10.5061/dryad.3j9kd521p). Segmented structures were grouped according to their SBEM volume of origin (volumes 1–4) and organized into structural groups using the built-in object list dialog in IMOD (accessed via "Edit" and then "Object list" in the Model View window). This grouping allows users to extract morphometric information for selected structural groups. Detailed information about the object list function is provided in the IMOD user guide (https://bio3d.colorado.edu/imod/doc/3dmodHelp/objectList.html).”

    (8) The discussion would benefit from some additions. Potential differences in EV biogenesis between ab1 and coeloconic sensilla. Comparison with other species: the absence of observable EV budding from ORN cilia is surprising given reports in Chlamydomonas, C. elegans, and cultured mammalian cells, yet this discrepancy is not discussed. The introduction and discussion suggest the existence of EV-mediated communication between ORNs and auxiliary cells, but only one example is provided. Is there additional evidence supporting such communication in Drosophila? If not, it may be helpful to discuss evidence from other model organisms where EV-mediated communication between sensory neurons and supporting cells has been reported (e.g. C. elegans).

    We thank the reviewer for these suggestions. We have expanded the Discussion to address these points. The additional Discussion sections are shown below.

    “EVs are thought to arise through endosomal release via MVBs or direct plasma membrane budding (Doyle & Wang, 2019; Malkin & Bratman, 2020; Neyroud et al., 2022; Remans et al., 2025; Singh et al., 2024; Welsh et al., 2024). Our data suggest both pathways are active but differentially favored across sensillum classes. Of note, coeloconic sensilla contain abundant small, dense EVs alongside frequent intracellular MVBs (Figures 4C and 5F), consistent with endosomal release. Conversely, ab1 sensilla are enriched in large, cargo-filled vesicles with fewer MVBs, suggesting a preference for direct plasma membrane budding. The scarcity of both MVBs and budding figures in intact sensilla likely reflects the transient nature of EV biogenesis.”

    “Ciliary extracellular vesicle shedding is well documented across evolutionarily distant systems, from the flagella of Chlamydomonas to the sensory cilia of C. elegans and mammalian primary cilia (Wang & Barr, 2016, 2018). It is therefore notable that we did not capture active EV budding from ORN dendritic cilia in our adult Drosophila preparations. This absence likely reflects both biological and technical constraints: ciliary EV production may be developmentally regulated and reduced in 1-week-old adults relative to earlier stages, and static volume EM may miss rare, transient release events. Instead, our data suggest that auxiliary cells are active contributors to the luminal EV pool in adults, a view independently supported by the recent identification of support-cell-specific GAL4 drivers that label EV-like structures originating from tormogen and trichogen membranes (Ballou et al., 2026). EV-mediated exchange between sensory neurons and supporting cells appears to be evolutionarily conserved. In Drosophila olfactory sensilla, where ORNs are intimately ensheathed by auxiliary cells, EV-mediated crosstalk may similarly coordinate sensillar physiology, though future functional studies with cellspecific genetic manipulation will be needed to test this directly.”

    Reviewer #2 (Recommendations for the authors):

    There are a few places where the presentation could be improved:

    (1) Figure 3B: The numbers showing analyzed sensilla at the bottom of each bar are difficult to read because the white text overlaps with white data points. This should be an easy fix.

    As suggested, we have revised Figure 3B to improve readability.

    (2) Figure 3C: The category "particles above cuticle base" appears to combine cases where particles are found only above the cuticle base with cases where they are present both above and below. These should be separated for clarity.

    We thank the reviewer for this suggestion. In our dataset, sensilla with particles exclusively above the cuticle base were rare; “above” particles were almost always accompanied by “below” particles. A separate "exclusively above" category would yield near-zero frequencies, adding visual clutter to Figure 3C without additional biological insight. We have retained the current grouping and clarified in the revised Results section, which reads:

    “To summarize these distributions, we additionally classified sensilla into three categories: no particles, particles present only below the cuticle base, or particles present above the base (including those also with particles below, as sensilla with particles exclusively above the base were rare) (Figure 3C).”

    (3) Figure 4C: Since the text and plot separate large basiconic sensilla into ab1 and ab2/3, it would help to state in the main text that ab1 sensilla contain four neurons whereas ab2/3 contain two.

    We have revised the Results section accordingly, which now reads:

    “Among large basiconic sensilla—ab1 (four ORNs), ab2 (two ORNs), and ab3 (two ORNs)—ab1 showed the greatest enrichment of cargo-filled EVs.”

    (4) Vesicle categorization: A representative EM image showing examples of the overlaps summarized in the Figure 4B Venn diagram would be helpful.

    We thank the reviewer for this suggestion. We have revised the Venn diagram in Figure 4 to include representative EM images showing examples of these overlapping morphologies.

  8. Version published to 10.64898/2026.05.30.728996v4 on bioRxiv
  9. Version published to 10.64898/2026.05.30.728996v3 on bioRxiv
  10. Version published to 10.7554/elife.112252.1 on eLife
  11. Version published to 10.64898/2026.05.30.728996v2 on bioRxiv
  12. eLife

    eLife Assessment

    This important study presents an ultrastructural atlas of extracellular vesicles and non-vesicular particles in Drosophila olfactory sensilla, using cryofixation-based serial block-face electron microscopy to catalogue roughly 7,800 particles across four antennal regions. The evidence is compelling: the cryofixation minimizes fixation artifacts, and the rigorous, well-validated methodology and morphometric analyses provide strong support for the structural classifications and spatial distributions, though some conclusions require moderation or additional evidence. The work provides a potentially foundational resource for researchers studying extracellular-particle signaling in sensory organs.

  13. eLife

    Reviewer #1 (Public review):

    Summary:

    Using cryofixation and serial block-face electron microscopy (SBEM), P. Vijayakumar and K. Cauwenberghs characterize extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) within native Drosophila olfactory sensilla. The study provides a unique and valuable dataset comprising approximately 7,800 extracellular particles, systematically describing their morphology, size, density, and distribution. The ultrastructural analysis across different sensillum classes offers insights into the potential biogenesis and functions of these extracellular particles.

    Strengths:

    Cryofixation preserves EVs and NVEPs within native tissue conditions. The detailed quantification of a very large dataset provides a unique source of information on extracellular particle number, categories, and distribution. The expertise of the group in the method and the tissue explored, as well as their detailed quantification, provide confidence in the dataset and observations.

    Weaknesses:

    Major comments

    (1) As the authors state, the rare observation of EV budding or MVB release events suggests that these are transient processes, whereas EVs and NVEPs are retained for relatively long periods within the sensillum lumen. The current analyses may overinterpret steady-state vesicle abundance as differences in vesicle production.

    (2) Given the above conclusion, differences between sensillum classes may be somewhat overstated.
    a) The absolute number of EVs and NVEPs per sensillum is highly variable, even within the same sensillum class (Figure 3D). For example, a substantial proportion of coeloconic sensilla have an empty lumen (Figure 3C). Consequently, expressing the data as ratios or proportions (Figures 3B, 4C, and 4E) may exaggerate differences between sensillum classes and should therefore be interpreted with caution.
    b) The rate of EV/NVEP production is unknown. For a similar rate of production across sensillum classes, Figure 3E suggests that the differences in lumen morphology and size may largely explain variation in EV density and distribution.

    That said, I agree that ab1 sensilla display a striking enrichment of large cargo-filled EVs compared with the other sensillum classes, while coeloconic sensilla display enrichment in small dense filled EVs (Figure 4C). Together, large and cargo-filled EV observation provides strong support for differences in EV biogenesis between ab1 sensilla and the other sensillum classes. In that context, I also think the EV size distribution shown in Figure 4 - Figure Supplement 1 should be moved into the main figure, as it demonstrates that the majority of EVs in the ab1 lumen are relatively large and are therefore likely to represent microvesicles. Could you clarify why ab1 sensilla are only included in Figure 4 and not analysed in Figure 3?

    (3) Approximately 10% of ORNs appear to be degenerating in 6-8-day-old flies, which seems unexpectedly high. This contrasts with the relatively infrequent occurrence of auxiliary cell apoptosis or complete sensillum degeneration. In these "degenerating ORNs", the authors state that the hallmarks of ORN apoptosis are restricted to the dendrites. As hallmarks, they state dendrite truncation, fragmentation and blebbing. Rather than apoptosis, I wonder whether these observations might instead represent ciliary truncation and ectosome shedding, followed by degradation of the shed ciliary membrane into EVs. Ciliary truncation and ectosome shedding, followed by ciliary regrowth, are dynamic processes that have been described across multiple species. This interpretation could explain large EVs that remain in the lumen long after the cilium has regenerated. It would reconcile this article with the general agreement that cilia are a prime site for the budding of EVs across species. Additional evidence supporting apoptosis of the ORNs would help distinguish between these possibilities. Otherwise, I believe the author should reconsider their interpretation.

  14. eLife

    Reviewer #2 (Public review):

    Summary:

    This paper presents a large structural survey of extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) in the olfactory sensilla of Drosophila melanogaster. Using high-pressure freezing and serial block-face SEM, the authors avoid many of the artifacts associated with conventional fixation and analyze more than 7,800 particles across 352 sensilla. The manuscript maps the distribution of these particles, describes their morphological heterogeneity, and examines their likely origins across different sensillum classes in both normal and degenerating tissue.

    Strengths:

    The strongest aspect of the paper is the imaging. Preservation is painstakingly controlled. The cryofixation appears to preserve the sensillum lymph in a more convincing native state than standard preparation methods, giving this work gravitas. Further, the authors characterized thousands of particles, further making this data strong.

    The figures are strong. They are clear, easy to read, and generally well designed; I think people will use this paper as a model for how to present complex data in a concise and straightforward manner. The manuscript is careful in how it presents the dataset and does not overinterpret the descriptive observations. As an ultrastructural resource, this paper will be useful to the field. The identification of auxiliary support cells as major secretory sites, together with the striking accumulation of EVs in degenerating tissue, will provide a useful starting point for future work.

    Weaknesses:

    The main point that could use more clarification is the vesicle categorization. In particular, the distinction between "dense," "cargo-filled," and "double EVs" is not always easy to follow from a biological perspective. Some additional discussion of how the authors think these categories relate to one another, and whether they are intended as purely morphological groupings or as distinct biological classes, would strengthen the manuscript.

  15. eLife

    Reviewer #3 (Public review):

    Using cryofixation-based serial block face electron microscopy of several subregions of the Drosophila antenna, the authors segment and assemble a high-resolution atlas of the anatomical structure, density, and spatial distribution of extracellular vesicles (EVs) and non-vesicular extracellular particles (NVEPs) in different Drosophila olfactory sensilla types. This systematic and thorough description is an important prerequisite to understanding the function of extracellular particles in intercellular signaling in the nervous system. Additionally, they describe examples of putative biogenesis events (budding/fusion), as well as neuronal and axonal cell degeneration events and measure the changes in particle accumulation in these altered microenvironments.

    Overall, this is a significant and comprehensive analysis and represents an invaluable resource to this burgeoning field. The authors assemble an important dataset and their claims match the level of evidence provided.

    Strengths:

    (1) The authors use segmentations from four different patches of the antenna to provide a systematic ultrastructural survey of extracellular particles in native insect sensilla. The dataset captures the diversity of sensilla types and reconstructs ~7800 particles.

    (2) We commend the authors for making the EM volumes available in the public Cell Image Library with accession numbers. It would be helpful to the community to also make the segmentations for this great resource easily accessible.

    (3) The sample preparation technique appears to minimize typical artifacts associated with chemical fixation, as evidenced by the high reported sphericity of EVs.

    Specific points:

    (1) In Figure 3C, the authors should include a continuous measure of particle distribution in the sensilla. Currently, the authors define three categories of particle localization. In the five examples shown in Figure 3A, the spatial distribution of these particles appears quite distinct across classes. For example, EVs/NVEPs in large and small basoconic sensilla are largely restricted to the area proximal to the base, with a limited number located more distally. In contrast, intermediate sensilla show a marked concentration of particles more distally.

    (2) The conclusion of different EV ratios across sensillum classes stems from a Kruskal-Wallis of p = 0.0476, with none surviving pairwise comparisons. This is not a strongly supported conclusion and is probably better characterized as a trend.

    (3) The statement "selective enrichment of large, cargo-filled vesicles within the ab1 lumen suggests specialized EV-mediated communication adapted to the coordination demands of this neuronal population" seems speculative for a Results section without supporting functional evidence. It would seem better suited for the Discussion.

    (4) Figure 4: Criteria for defining the classes of EVs.
    a) The authors should explain the rationale for classifying EVs using relative density rather than absolute density? We would expect EVs with similar contents to have similar electron density (similar darkness in the images). Would classifying them relative to the background, which itself might vary across sensilla or regions, create a possible confound, especially when comparing across sensillum classes?
    b) The two example images (in Figure 4A) of the cargo-filled EVs appear to have different densities themselves. Do the cargo-filled ones also display systematic differences in density and, if so, why is this another class instead of being a subcategory within the dense and lucent classes (i.e. dense with/without cargo, lucent with/without cargo)? The dense and lucent classes are defined by their density, whereas this is a more structural property.
    c) Regarding "Double" and "Ball-and-Socket" EVs, does the density vary between the two particles involved (e.g., does the inner structure consistently differ in density from the outer)?

    (5) What was the rationale for the 200μm and 1000μm size cutoffs? A continuous distribution of maximum particle sizes would provide a clearer understanding of the data.

  16. Mentioned by @biorxivpreprint.bsky.social on Bluesky

    A nanoscale atlas of extracellular vesicles and particles in Drosophila olfactory sensilla https://www.biorxiv.org/content/10.64898/2026.05.30.728996v1

  17. Mentioned by @biorxiv-neursci.bsky.social on Bluesky

    A nanoscale atlas of extracellular vesicles and particles in Drosophila olfactory sensilla https://www.biorxiv.org/content/10.64898/2026.05.30.728996v1

  18. Version published to 10.64898/2026.05.30.728996v1 on bioRxiv