Multiple Lipid Scramblases Regulate the Quantity of Extracellular Vesicle Production From the Plasma Membrane

ABSTRACT

Extracellular vesicle (EV) biogenesis is linked to the disruption of lipid bilayer asymmetry during diverse processes from coagulation to programmed cell death. However, the specific contributions of lipid scramblase proteins to the regulation of EV production and the potential redundancy of diverse families of scramblases are not well understood. We developed a semi‐automated image analysis pipeline that accurately quantifies fluorescently labelled EVs on the cell surface of live Caenorhabditis elegans embryos. Using this method in a sensitized background with enhanced EV production, we performed an RNA interference (RNAi) screen targeting a subset of plasma membrane and intracellular scramblases. We found that knockdown of most scramblase genes we tested resulted in a minor but significant reduction in the number of EVs. We confirmed these significant decreases in EV abundance with anoh‐2 and atg‐9 scramblase mutants. However, loss‐of‐function egli‐1 or scrm‐3 scramblase mutants displayed no significant change in EV production, unlike egli‐1 or scrm‐3 knockdown. We also observed no effect on EV production in ced‐7 ABC transporter mutants. The discrepancy between the phenotypes of acute scramblase knockdown and chronic loss suggested the presence of compensatory mechanisms that maintain EV homeostasis in response to the permanent absence of a scramblase. RNA sequencing revealed changes in expression levels for conserved sets of genes in egli‐1 and scrm‐3 mutants. Together, our data suggest that multiple scramblases involved in diverse pathways regulate EV budding, potentially responding to distinct physiological or metabolic cues.

Cells are wrapped in a double‐sided plasma membrane made up of fatty lipids that can release lipid‐wrapped fragments called extracellular vesicles (EVs). Diverse scramblase proteins, which shuffle lipids across cellular membranes in both directions, can promote EV release. Scramblases act in opposition to lipid flippases, whose unidirectional lipid transport inhibits EV budding from the plasma membrane. This reveals that cells have many ways to regulate EV production by transporting lipids.

1. Introduction

Extracellular vesicles (EVs) are membrane‐bound organelles released by cells for many physiological and pathological functions, including signalling, metabolism, immune regulation, metastasis and cell death (Wang et al. 2024; Welsh et al. 2024). EVs can form from different cell membranes, including the direct budding of the plasma membrane called ectocytosis. However, the molecular mechanisms promoting EV biogenesis from the plasma membrane are poorly understood, making it challenging to determine how EV biogenesis is altered during homeostasis or disease.

Studies suggest that EV production is regulated by the disruption of plasma membrane lipid asymmetry. In resting cells, different phospholipids and sphingolipids are localized to specific leaflets of the plasma membrane, resulting in an asymmetry in lipid content in each layer (Lorent et al. 2020; Doktorova et al. 2023, 2025). This asymmetry is actively maintained by ATP‐dependent lipid transporters called flippases and floppases. Disrupting the flippase activity or localisation of the C. elegans P4‐ATPase TAT‐5 results in dramatic increases in EV production from the plasma membrane (Wehman et al. 2011; Beer et al. 2018). Increased EV release is also seen in cancer cell lines after knocking down a human homolog of TAT‐5, ATP9A (Naik et al. 2019). Disrupting the TAT‐5 flippase results in the surface exposure of specific lipids that are normally maintained in the cytosolic leaflet of the membrane bilayer (Wehman et al. 2011; Beer et al. 2018). Disrupting other proteins that result in similar exposure of cytofacial lipids also increases EV production (Beer et al. 2018), including the poorly characterized Dopey family protein PAD‐1 with conserved links to the TAT‐5 flippase (Molière et al. 2022). Interestingly, partially disrupting PAD‐1 through the insertion of a GFP tag in the pad‐1(babIs1) allele leads to an intermediate increase in EV production compared to pad‐1 knockdown by RNA interference (RNAi) (Fazeli et al. 2020). The pad‐1(babIs1) partial loss‐of‐function phenotype suggests that the graded activity of lipid transporters and their regulators modulates the quantity of EV budding from the plasma membrane. Furthermore, as strong loss of TAT‐5 or PAD‐1 function causes sterility and embryonic lethality in addition to high levels of EVs (Wehman et al. 2011; Beer et al. 2018), pad‐1(babIs1) is a viable background for enhancer‐suppressor screens for additional regulators of lipid asymmetry and EV budding.

Lipid asymmetry is intentionally disrupted during many physiological processes, such as cell fusion, cell division, and cell death (Doktorova et al. 2023; Jan and Jan 2025). Lipid channels called scramblases bidirectionally transport lipids between the inner and outer leaflets of a single bilayer, thereby destroying membrane asymmetry. Scramblase channels can be activated by ligands, ions, oligomerisation or mechanical forces to reversibly change conformation and disrupt lipid asymmetry in an ATP‐independent fashion (Rocha‐Roa and Vanni 2026). Scramblases can also be activated by reversible post‐translational modifications like phosphorylation in addition to irreversible modifications, such as the proteolytic processing of scramblase proteins in dying cells. Lipid scrambling can reduce lipid packing and increase the flexibility and ease of vesicle budding or blebbing (Wang et al. 2025). Indeed, patients with a bleeding disorder called Scott syndrome have mutations in the TMEM16F lipid scramblase and defects in EV budding from the plasma membrane of platelets (Sims et al. 1989; Fujii et al. 2015). The role of the TMEM16/Anoctamin family of calcium‐activated scramblases and ion channels is well conserved (Wang et al. 2013; Jan and Jan 2025), as targeting C. elegans anoh‐2 using RNAi decreased EV production from the plasma membrane of developing embryos (Wang et al. 2025). However, disrupting ANOH‐2 or TMEM16F does not completely prevent EV release, raising the possibility that additional scramblases contribute to EV regulation.

As opposed to lipid flippases that belong to a subfamily of ATPase proteins, many disparate families of proteins are known to scramble lipids at the plasma membrane. These include Anoctamins (TMEM16), Xk‐related (XKR) proteins, G‐protein coupled receptors (GPCR), serine incorporators (SERINC), osmosensitive calcium‐permeable channels (OSCA/TMEM63), pore forming channels (TMC) and the original phospholipid scramblase (PLSCR) family (Rocha‐Roa and Vanni 2026). Calcium‐activated PLSCR proteins promote phosphatidylserine (PS) exposure during apoptosis, including C. elegans SCRM‐1 and SCRM‐3 (Venegas and Zhou 2007; Wang et al. 2007). Apoptosis is associated with EV release (Mapes et al. 2012), but a role for SCRM proteins in EV biogenesis had not been investigated. In addition, lipids can be exposed by lipid translocases beyond ATP‐independent scramblases. For example, the ATP‐Binding Cassette (ABC) transporter Cell Death abnormality 7 (CED‐7) is required for the release of EVs that have lost lipid asymmetry during apoptosis (Mapes et al. 2012) but has not been examined for a role in EV release from living cells. Therefore, it is important to determine whether more transmembrane proteins at the plasma membrane promote lipid exposure for EV budding.

Lipid transporters on the membranes of intracellular organelles have also been shown to influence plasma membrane asymmetry (Takar et al. 2016; Wang and Kinoshita 2023), likely through the inter‐organelle flow of lipids by vesicular trafficking and lipid transfer proteins at membrane contact sites. The effect of intracellular transporters on the plasma membrane raises the possibility that intracellular scramblases could also play a role in EV budding from the plasma membrane. Intracellular scramblases have already been linked to viral infection, and viruses have been shown to hijack the biosynthetic pathways of EVs for viral particle production (Kumari and Banerjee 2026; Tang et al. 2022). Scramblases acting on the endoplasmic reticulum (ER) membrane include TMEM41 paralogs that connect to lipid transfer proteins and regulate the production of viral particles (Ghanbarpour et al. 2021; Huang et al. 2021; Li et al. 2021). The lipid scramblase ATG9 also cooperates with a lipid transfer protein for phagophore elongation during macroautophagy and is found on trans‐Golgi and intracellular vesicles (Matoba et al. 2020). Mammalian ATG9 homologs have also been shown to regulate viral particle production (Abernathy et al. 2019). Therefore, we sought to elucidate the role of a diverse subset of intracellular and plasma membrane scramblases in EV budding.

One challenge in studying EV regulation is the small size of EVs, rendering most EVs undetectable by light microscopy (Wang et al. 2024; Welsh et al. 2024). To overcome this limitation, we previously developed a degron‐based selective labelling approach to fluorescently label EVs and distinguish them from other membranes (Beer et al. 2019). The EV reporter is based on a fusion protein containing red fluorescent mCherry and the phosphoinositide lipid‐binding PH domain of phospholipase C that targets the reporter to the cytosolic face of the plasma membrane (Kim et al. 1996). The EV reporter also includes the C‐terminal phosphodegrons (CTPD) domain from OMA‐1, whose degrons cause protein degradation in the cytosol after the first embryonic cell division in C. elegans (Nishi and Lin 2005). The mCh::PH::CTPD reporter fluorescence is preserved in EVs that have separated from the plasma membrane prior to the onset of degradation (Beer et al. 2019). Quantification of fluorescent mCh::PH::CTPD puncta correlated well with EV abundance on electron micrographs (Beer et al. 2018, 2019; Fazeli et al. 2020), confirming the utility of this fluorescent EV reporter.

Equipped with the degron‐based EV reporter and sensitized pad‐1 mutant background with enhanced EV production, we set out to uncover the effects of disrupting different scramblases. We hypothesized that scramblase mutants would exhibit a significant decrease in EV production, as we previously observed in anoh‐2 knockdown embryos (Wang et al. 2025). Here, we describe the development of a semi‐automated MATLAB program to recognise and count EV puncta on the top surface of embryos with accuracy that recapitulates manual counts. We show that RNAi knockdown of several individual scramblases results in a significant decrease in EV numbers, but we did not find a complete suppression of EV production. Instead, using mutant alleles to constitutively remove scramblases, we find evidence for the role of different scramblases, as well as genetic compensation in scramblase‐mediated regulation of EV production. Together, these data suggest that many families of scramblases can regulate EV budding in response to different physiological cues.

2. Materials and Methods

2.1. Worm Culture

C. elegans hermaphrodite worms from the strains listed in Table S1 were fed OP50 bacteria on NGM agar at room temperature (22°C–23°C) according to standard methods (Brenner 1974). Before imaging, gravid adults and their progeny were incubated as indicated in the Supporting Information worksheet.

2.2. Genome Editing

The anoh‐2(wur37) deletion was created by CRISPR/Cas9‐mediated genome editing performed by InVivo Biosystems (Eugene, OR). Guide RNAs targeting the 4th exon of anoh‐2b (ACATTCATCGAAATTAGCCG) and the 12th exon of anoh‐2b (ACGAAGAAGTATCACACCAA) were injected into the COP1403 strain along with an oligonucleotide donor sequence with 35 bp homology arms for precise repair of anoh‐2 (TTACATTTACGGAGGGAACTGCAATTGTTCCACGGCAATGGTTCGGTTATCCAGTTTCACAAGAAAACCA). The 8724 bp deletion in anoh‐2 was verified by PCR and sequencing.

2.3. Genetic Crosses

Mutant strains were obtained from the Caenorhabditis Genetics Center, Knockout Consortium, and NBRP Japan (C. elegans Deletion Mutant Consortium 2012; Mitani 2017), crossed to the WEH642 strain and homozygosed. Insertions and deletions were detected by 3‐primer PCR using the listed oligonucleotides (Table S2). Regions of atg‐9 or ced‐7 containing point mutations were amplified using a 2‐primer PCR (Table S2) and detected with an MboI digest for atg‐9(bp564) or Tsp45I digest for ced‐7(n1996), designed using dCAPS Finder 2.0 (Neff et al. 2002).

2.4. Plasmid Cloning

Exons 4 and 5 of scrm‐3 were amplified from genomic DNA by PCR using the following primer sequences (cgatttccagcttttccactTCAACCA and gttcccgccgacctgGATGTCAA) and cloned into a t‐tailed pPD129.36 plasmid (Fraser et al. 2000). After verifying the sequence of LJ:Exon4‐5, the plasmid was transformed into HT115 bacteria for RNAi experiments.

2.5. RNA Interference (RNAi)

Gravid worms were bleached, and their progeny fed HT115 bacteria for 3 days at 23°C on NGM Lite plates supplemented with Ampicillin, Tetracycline and Lactose to induce dsRNA expression from plasmids. RNAi clones were identified on WormBase (Sternberg et al. 2024) isolated from genomic or cDNA libraries (Fraser et al. 2000; Rual et al. 2004), and verified by sequencing, specifically empty vector L4440, atg‐9: JA:sjj_T22H9.2, egli‐1: JA:sjj_D2013.10 or MV:mv_D2013.10, scrm‐3: JA:sjj_C04E12.7 and LJ:Exon4‐5. The scrm‐1 clone was a gift of Ding Xue (Wang et al. 2007). We did not perform ced‐1 RNAi because no phenotypes have been reported for ced‐1 knockdown in WormBase (Sternberg et al. 2024).

2.6. Imaging

Embryos were dissected from mutant or RNAi‐treated gravid adult hermaphrodites in Sulston's M9 buffer and mounted on 4% agarose pads for imaging. Z‐stacks of DIC and mCherry fluorescence were at the surface of live 3‐cell to 15‐cell embryos using a Zeiss Axio Observer 7 inverted microscope with a Plan‐APO 40 × 1.4 NA oil objective with Excelitas Technologies X‐Cite 120LED Boost illumination, and a Hamamatsu ORCA‐Fusion sCMOS camera controlled by 3i SlideBook6 software. Imaging settings such as exposure time and the light rheostat were held constant for all data sets to obtain consistent images; however, natural variation in fluorescence was observed and likely contributes to variation seen within data sets.

2.7. Manual Extracellular Vesicle Counts

The surface plane of an embryo's image stack was determined by DIC and by identifying the z‐plane that put EV puncta in focus while avoiding any planes including cell‐cell contacts in the embryo. Fluorescent puncta within the exported TIF were manually annotated using the Cell Counter function in FIJI (NIH) for quantitation.

2.8. Automated Extracellular Vesicle Counts

TIF images were analyzed using cmeAnalysis (Aguet et al. 2013) to detect fluorescent puncta within an editable region of interest (cell mask). Using custom MATLAB scripts available on GitHub, we analyzed the position and Signal‐to‐Noise Ratio (SNR) for each cmeAnalysis detection.

2.9. Comparison of Manual Annotations and Automated Detections

Using custom MATLAB scripts available on GitHub, we calculated the straight‐line distance between the location of each cmeAnalysis detection and the location of each manually annotated puncta using the Euclidean distance formula. If a manually annotated puncta had a cmeAnalysis detection as the nearest neighbour (where each cmeAnalysis detection can only count as the nearest neighbour once), the computational detection was considered a ‘matched’ detection. Any cmeAnalysis detections that were more distant than other cmeAnalysis detections from any manual puncta were considered unmatched. A histogram of matches at different SNRs is in the Supporting Information, which was used to calculate the percent matches at or above different SNR cut‐offs.

2.10. RNA Sequencing and Analysis

Four biological replicates of gravid adult worms were flash‐frozen in liquid nitrogen and submitted to the TIGSS Molecular Genomics core for RNA isolation using a Qiagen RNeasy Plus kit. An mRNA library was generated using a Watchmaker Genomics prep kit and sequenced using an Illumina NextSeq 2000 P2 XLEAP‐SBS reagent kit. Raw FASTQ files were submitted to the TIGSS Bioinformatics core, assessed for sequence quality using FastQC and summarized with MultiQC. Adapter sequences and low‐quality bases were removed using Trim Galore. Transcript abundance was quantified using Salmon against the Caenorhabditis elegans Ensembl WBcel235 transcriptome. Transcript‐level abundance estimates were imported into R using tximport and summarized to gene‐level counts. Differential gene expression analysis was performed using DESeq2 and cut‐offs of |log2FoldChange|≥1 and p _adj < 0.05. Functional enrichment analysis was performed on all differentially expressed genes using clusterProfiler with a GeneRatio of 0.01 and an adjusted p value of 0.05 as cut‐offs.

2.11. Statistical Analysis

Data are reported as mean ± standard deviation. One‐tailed Student's t‐tests assuming unequal variance were corrected for multiple comparisons using a post‐hoc Bonferroni correction and adjusted p values are reported in the text. Gene expression data was compared using a Brown–Forsythe and Welch ANOVA. The number of independent experiments performed are indicated for each data set in the Supporting Information worksheet.

3. Results

To test the role of different lipid scramblases in EV release, we first needed to develop a robust method to quantify EV release from different C. elegans strains without human error or bias. We chose to develop a semi‐automated analysis pipeline using MATLAB and validate it by comparing its output to manual annotations in FIJI. To automatically detect EV puncta, we started with a clathrin‐mediated endocytosis detection algorithm, cmeAnalysis (Aguet et al. 2013), which was designed to detect small puncta in noisy fluorescence images. We used images of the cell surface of embryos expressing a previously characterized degron‐based EV reporter, mCherry::PH::CTPD (Figure 1A) (Beer et al. 2019). Using cmeAnalysis, we were able to detect EV puncta on the cell surface of C. elegans embryos (Figure 1B), but we also observed the detection of additional puncta in the fluorescence images. In images from a control strain expressing the EV reporter, cmeAnalysis detected 82 ± 59 puncta, a 30‐fold increase over the manual counts of 3 ± 2 EV puncta (n = 23) (Figure 1C). We trusted the manual counts of fluorescent EV puncta over the cmeAnalysis output, as EVs were rarely observed in electron micrographs of control embryos (Wehman et al. 2011; Beer et al. 2018; Fazeli et al. 2020). Indeed, the number of cmeAnalysis detections obscured the differences in EV numbers between embryos from wild type control hermaphrodites and embryos from pad‐1(babIs1) partial‐loss‐of function mutant hermaphrodites (Figure 1C). The manual counts showed a 15‐fold increase in EV release in pad‐1(babIs1) mutants (n = 40) compared to wild type control embryos, while cmeAnalysis did not detect a significant difference between the samples (p > 0.2). As the significant increase in EV release in pad‐1(babIs1) mutants had been previously documented in electron micrographs (Fazeli et al. 2020), these results suggested that cmeAnalysis was highly sensitive at detecting fluorescent puncta, but not selective enough to differentiate EV puncta in our fluorescence images.

FIGURE 1.

Increasing the signal‐to‐noise ratio (SNR) threshold improves automated extracellular vesicle (EV) detection by cmeAnalysis. (A) Representative image of mCh::PH::CTPD fluorescence labelling EVs on the surface of a 7‐cell C. elegans embryo bearing the pad‐1(babIs1) partial loss of function allele. (B) Same image as (A) annotated by cmeAnalysis with cell mask contour and circled SNR 2 (magenta), SNR 3.5 (cyan), or SNR 4+ (yellow) automated detections. (C) Comparison of manual, cmeAnalysis, SNR 3.5, and SNR 4 EV puncta counts for wild type control and pad‐1(babIs1) mutant embryos imaged in Denver, Colorado. (D) Percentage of cmeAnalysis detections matching manually annotated EV puncta at different signal‐to‐noise ratios (SNR). (E–H) Correlation graphs of SNR 3.5 (E, F) or SNR 4 (G, H) automated detections counted by the MATLAB scripts versus manually counted in FIJI from the different genotypes (coloured) listed in Table S1. The golden line represents an ideal match (x = y), and the dotted lines show a linear fit of the data: y = 0.79x + 8.2 with R ² = 0.90 in (E, F) and y = 0.64x + 5.8 with R ² = 0.89 in (G, H). (F) and (H) are insets of the lower left quadrant of (E) and (G).

3.1. Increasing the Signal‐to‐Noise Ratio Improves EV Detection

As cmeAnalysis appeared to be detecting weak fluctuations in fluorescence intensity in addition to brighter EV puncta (Figure 1B), we tried using intensity thresholding to only detect the brightest puncta. However, as the centre of the 25 µm thick embryos had higher overall fluorescence than the edges, this led to biases in EV detection in the centre versus the edges of the image. We next analyzed the Signal‐to‐Noise ratio (SNR) of the detected puncta compared to nearby background noise in the fluorescence images. Combining the analysis of 260 images of strains with a range of different EV counts, we found that the locations of automated detections below a Signal‐to‐Noise ratio of 3:1 (SNR 3) were rarely the nearest neighbour of a manually annotated EV puncta (22.1% at SNR 3, Figure 1D). When summing all matches below this threshold, only 10.5% of the 11,617 cmeAnalysis detections at or below SNR 3 were near the position of manual detections (Supporting Information), suggesting that counting detections above SNR 3 would improve the selectivity of cmeAnalysis detections for EV quantitation. Therefore, we decided to incorporate a Signal‐to‐Noise ratio cut‐off higher than 3:1 in the analysis pipeline.

At higher SNR, we observed higher selectivity for EV puncta (Figure 1D), but the cmeAnalysis detections never matched the manual annotations 100% at any SNR examined. This discrepancy may be due to human imprecision in our manual annotations or limited resolution in our images. To find an SNR cut‐off that would help us avoid counting noisy pixels (false positives) but minimise missed EV detections (false negatives), we first tested SNR 3.5. We chose this value because nearly half (44.6%) of SNR 3.5 detections were the nearest neighbour of a manual detection (Figure 1D). When summing all matches above this threshold, we observed 79% matching detections at or above SNR 3.5 in our initial comparison (Supporting Information). We plotted the manual and automated counts for 228 embryos from seven different genotypes with disparate levels of EVs (Figure 1E,F, Supporting Information). The SNR 3.5 data correlated with the manual data for embryos with moderate levels of EV production (Figure 1E, Table S3). For example, pad‐1(babIs1) partial loss‐of‐function embryos averaged 42 ± 33 EV puncta in manual counts and 47 ± 38 EV puncta in automated SNR 3.5 counts (Figure 1C), showing no significant difference between the two methods (p > 0.7). The SNR 3.5 cut‐off also detected the significant difference in EV counts between control and pad‐1(babIs1) mutants (p < 0.0001), demonstrating that the additional SNR cut‐off improved the accuracy of cmeAnalysis detections. However, the SNR 3.5 cut‐off tended to overcount wild type control embryos with few EVs (Figure 1F), with the automated counts averaging 8 ± 5 EV puncta, representing a three‐fold increase over manual detections (p < 0.001). We also noticed that the SNR 3.5 detections undercounted images with more than 150 EV puncta (Figure 1E). Therefore, although the SNR 3.5 cut‐off seemed accurate for some genotypes and was able to detect significant differences in EV production between some strains, it did not produce consistent EV counting results at either the low or high end of the spectrum.

As the goal of this study was to test whether scramblase disruption significantly decreases EV production, we wanted to avoid overcounting samples with few EV puncta. We tried a more stringent cut‐off, SNR 4, which had 58.3% of automated detections matching manual annotations (Figure 1D). Indeed, 86.5% of the 9105 cmeAnalysis detections at or above SNR 4 were close to the position of the manual annotations (Supporting Information). The SNR 4 cut‐off resulted in the undercounting of EV puncta for most genotypes (Figure 1G,H), but improved accuracy for embryos at the low end of the EV spectrum (Figure 1H). The wild type control strain averaged 4 ± 3 detections with the SNR 4 cut‐off, which was not significantly different from 3 ± 2 manual counts of EV puncta (p > 0.05) (Figure 1C). The SNR 4 counts from pad‐1(babIs1) partial loss‐of‐function embryos averaged 39 ± 33, which was 91% of manual counts and not significantly fewer (p > 0.9). As we were able to detect a significant difference in EV numbers between control and pad‐1(babIs1) mutant strains (p < 0.0001), we subsequently used SNR 4 as the threshold to run with cmeAnalysis to obtain robust results with our fluorescence images at the low and mid‐range of the EV spectrum.

3.2. Anoctamin ANOH‐2 Promotes EV Production

To confirm our previously published 47% decrease in EVs after RNAi‐mediated knockdown for the Anoctamin scramblase homolog ANOH‐2 (Wang et al. 2025), we asked whether mutating anoh‐2 would produce a similar decrease in EV production. We crossed an existing anoh‐2 mutant strain into the EV reporter strain in the sensitized pad‐1(babIs1) partial loss‐of‐function background. The 661 bp anoh‐2(tm4796) deletion removes part of an intron and an exon encoding one extracellular domain. The deletion is also reported to cause a frameshift that would remove transmembrane domains 6–8 (Wang et al. 2013). The scramblase domain of anoctamins has been shown to reside in transmembrane domains 4–7 (Yu et al. 2015). In embryos from the pad‐1(babIs1); anoh‐2(tm4796) mutant strain, we observed 43 ± 23 SNR4 puncta (Figure 2A). As our control reporter strain showed 55 ± 20 SNR4 puncta (n = 28), as previously published (Wang et al. 2025), this anoh‐2 deletion showed a non‐significant 21% decrease in the number of EVs released (n = 31, p > 0.1). A rule of thumb power test indicated that analysing over 23 embryos should have been sufficient to detect a significant 21% decrease, suggesting that our analysis was sufficiently powered and the small anoh‐2(tm4796) deletion does not significantly decrease EV production in pad‐1 mutants.

FIGURE 2.

RNAi knockdown of scramblases decreases EV number more than scramblase mutation. (A) SNR4 counts observed for loss‐of‐function mutant embryos for scramblases anoh‐2(tm4796), atg‐9(bp564), and egli‐1(gk278), as well as ABC transporter ced‐7(n1996) in the pad‐1(babIs1) background imaged in Denver, Colorado. One‐tailed t‐test with a Bonferroni correction for seven comparisons. (B) SNR4 counts from the surface of pad‐1(babIs1) embryos imaged in Colorado following RNAi knockdown of scramblases scrm‐1 (n = 27), scrm‐3 (JA n = 27, LJ n = 28), atg‐9 (n = 29), egli‐1 (JA n = 30, MV n = 27), or empty vector control (n = 44 embryos). One‐tailed t‐test with a Bonferroni correction for six comparisons. (C) SNR4 counts observed for deletion embryos for scramblases scrm‐1(tm805) scrm‐4(tm2239), scrm‐3(tm631), or anoh‐2(wur37) in the pad‐1(babIs1) background imaged in College Station, Texas. One‐tailed t‐test with a Bonferroni correction for seven comparisons. (D) SNR4 counts observed for pad‐1(babIs1); scrm‐3(tm631) deletion embryos treated with scrm‐3 LJ RNAi (n = 32) or empty vector control (n = 42 embryos) imaged in Texas. One‐tailed t‐test. (E) SNR4 counts observed for pad‐1(babIs1); egli‐1(gk278) deletion embryos treated with egli‐1 MV RNAi (n = 35) or empty vector control (n = 30 embryos) imaged in Texas. One‐tailed t‐test.

As the anoh‐2(tm4796) allele does not delete transmembrane domains 4 and 5, which have been shown in other Anoctamins to have a hydrophilic cleft that promotes lipid scrambling (Yu et al. 2015), the small anoh‐2 deletion may not disrupt lipid scrambling. Therefore, we used CRISPR/Cas9‐mediated gene editing to delete all transmembrane domains in anoh‐2. We crossed this 8.7 kb anoh‐2(wur37) deletion into the pad‐1(babIs1); EV reporter background and submitted the strain for whole genome RNA sequencing. We found a 6‐fold decrease in anoh‐2 transcript levels in pad‐1(babIs1); anoh‐2(wur37) mutants (Supporting Information). We also observed a frameshift after the deletion causing the last exon to encode a new sequence and terminate four amino acids after the original stop codon. We imaged embryos for surface EVs and observed 23 ± 10 SNR4 puncta (n = 31) in the large anoh‐2(wur37) deletion (Figure 2C), which represented a significant 34% decrease compared to the control pad‐1(babIs1) strain (p < 0.01). These findings suggest that the transmembrane domains of Anoctamin ANOH‐2 promote EV production, consistent with a role for scramblase activity. However, the large anoh‐2 deletion only partially suppressed the increased EVs in the pad‐1(babIs1) background from 15‐fold to 10‐fold, suggesting that additional factors beyond ANOH‐2 are disrupting lipid asymmetry to promote EV budding from the plasma membrane of early C. elegans embryos.

3.3. Environmental Regulation of EV Quantity

After a lab move, we noticed that our pad‐1(babIs1) control data collected in College Station, Texas (n = 41) had significantly fewer SNR4 puncta (35 ± 18) than our pad‐1(babIs1) control data (55 ± 20 SNR4 puncta) collected on the same strain on the same microscope in Denver, Colorado (p < 0.0001). This observation led us to plot the Colorado data (Figure 2A,B) separately from the Texas data (Figure 2C–E). This finding implies that there are environmental factors impacting the quantity of EV production from the plasma membrane, consistent with our previous findings on the impacts of temperature (Wang et al. 2025).

3.4. PLSCR Scramblase Knockdown but Not Knockout Reduces EV Production

To evaluate the effects of other lipid scramblases on EV release, we knocked down a subset of scramblases prioritized based on published links to EV production or EV‐associated processes like cell death or viral particle production. We used three‐day RNAi treatment on pad‐1(babIs1) partial loss‐of‐function mutant worms expressing the mCherry::PH::CTPD reporter starting from the first larval stage and examined the embryos from treated adults. As the labelled EVs are released from the plasma membrane before mCherry::PH::CTPD is degraded at the 2‐cell stage of embryogenesis, we focused on scramblase genes with known mutant phenotypes and showing maternal expression in early embryos.

We first knocked down Phospholipid scramblase 1 (SCRM‐1), which showed the highest expression in early embryos of the eight PLSCR family scramblases (Hashimshony et al. 2015). We used an RNAi clone previously shown to reduce PS exposure on apoptotic cells and produce a phenotype similar to scrm‐1 mutants (Wang et al. 2007). Treatment of pad‐1(babIs1) mutants with scrm‐1 RNAi resulted in 53 ± 21 SNR4 puncta (Figure 2B), showing no significant effect on EV production (n = 27, p > 1). As RNAi knockdown typically leads to a partial loss‐of‐function, we also examined a double mutant for deletions predicted to disrupt scrm‐1 and a PLSCR homolog within 10 kb that showed weak early embryonic expression, scrm‐4 (Hashimshony et al. 2015). The scrm‐1(tm805) deletion removes two exons encoding the last three beta‐strands of the 12‐stranded transmembrane beta‐barrel, while the scrm‐4(tm2239) deletion removes two exons encoding the first seven beta‐strands of the beta‐barrel and is predicted to cause a frameshift with a premature stop codon (Sternberg et al. 2024). Deleting scrm‐1 and scrm‐4 did not significantly decrease EV production in the pad‐1(babIs1) background (n = 30, p > 0.9) (Figure 2C). This finding suggests that not every scramblase expressed in early embryos is used to regulate EV release from the plasma membrane in living cells, consistent with the known differences in activation mechanisms and physiological function for different scramblases (Rocha‐Roa and Vanni 2026).

We next tested phospholipid scramblase 3 (SCRM‐3), a PLSCR family scramblase shown to regulate PS exposure on dying cells (Venegas and Zhou 2007), which is expressed in early embryos (Hashimshony et al. 2015). Using a genomic clone targeting exons 1–4 of scrm‐3 (JA), we observed 51 ± 27 SNR4 puncta (Figure 2B), showing no significant effect on EV abundance in pad‐1(babIs1) mutants (n = 27, p > 1). However, the first three exons of scrm‐3 contain sequence similarity with scrm‐1 and scrm‐2, raising the possibility that the scrm‐3 JA clone could have off‐target effects and may only weakly knock down scrm‐3. Therefore, we cloned a new RNAi construct targeting the fourth and fifth exons of scrm‐3 (LJ), which did not share significant sequence similarity with scrm‐1. After treatment with scrm‐3 LJ RNAi, we observed 27 ± 18 SNR4 puncta in pad‐1(babIs1) mutants, showing a significant 51% decrease in the amount of EVs (n = 28, p < 0.001). These data provide evidence that SCRM‐3 can promote EV budding from the plasma membrane of living cells, similar to ANOH‐2.

To confirm the role of SCRM‐3, we crossed an existing deletion for scrm‐3 into the pad‐1(babIs1) background and examined its effects on plasma membrane EVs. The scrm‐3(tm631) allele deletes parts of two exons encoding the first four beta‐strands of the beta‐barrel (Sternberg et al. 2024). Whole genome RNA sequencing revealed a 4‐fold decrease in scrm‐3 mRNA levels in the scrm‐3(tm631) mutant and a frameshift with a premature stop codon that would truncate the protein at Proline 25 of 251 amino acids. Deleting scrm‐3 in pad‐1(babIs1) mutants resulted in 37 ± 20 SNR4 puncta (Figure 2C), similar in number to controls (p > 0.9). This shows that deleting scrm‐3 did not decrease EV production, making the role of SCRM‐3 in EV regulation less clear.

As the scrm‐3(tm631) deletion did not confirm the scrm‐3 LJ RNAi results, we wondered whether scrm‐3 LJ RNAi was targeting scrm‐3 or other genes. To test whether the effect of scrm‐3 LJ RNAi depended on the scrm‐3 gene, we performed RNAi on the pad‐1(babIs1); scrm‐3(tm631) mutant strain. We observed 44±18 SNR4 puncta after scrm‐3 LJ RNAi treatment, which was not significantly different from 44 ± 16 SNR4 puncta after empty vector control RNAi (p > 0.4) (Figure 2D). Given that the RNAi and deletion target non‐overlapping parts of the scrm‐3 sequence, these results demonstrate that knocking down scrm‐3 was only able to decrease EV production if SCRM‐3 was functioning, providing evidence against off‐target effects. Instead, the results suggest that there are genetic differences in the scrm‐3(tm631) mutant background that prevent the suppression of pad‐1‐induced EV production.

As RNAi is a shorter‐term perturbation than mutation, we considered that mutant strains could compensate for deletions by upregulating related genes, including other scramblases. We performed RNA sequencing on gravid adults from a series of strains expressing the EV reporter: a wild‐type control strain, the pad‐1(babIs1) EV‐overproducing strain and the pad‐1(babIs1); scrm‐3(tm631) deletion strain. Of the significantly differentially expressed genes (Table S7), we observed downregulation of three scramblase genes in the pad‐1(babIs1); scrm‐3(tm631) double mutant strain: scrm‐3 itself, an Anoctamin anoh‐1, and a TMEM41A homolog bus‐19 (Table S11). However, we found that both the pad‐1 and the pad‐1; scrm‐3 deletion strains showed increased expression of scrm‐4 compared to the wild type control strain (Figure 3A), raising the possibility that SCRM‐4 overexpression could compensate for the absence of SCRM‐3 (or partially suppress EVs in pad‐1 mutants).

FIGURE 3.

Known scramblases do not compensate for EV release in scramblase mutants. (A) scrm‐4 mRNA expression levels from normalized DESeq2 counts in Supporting Information. Brown‐Forsythe and Welch ANOVA: ^* p < 0.05, ^** p < 0.01. (B) SNR4 counts observed for double or triple scramblase mutant embryos in the pad‐1(babIs1) background imaged in Texas. One‐tailed t‐test: n.s. p > 0.05.

To disrupt potential genetic compensation by SCRM‐4, we crossed the scrm‐4(tm2239) scrm‐1(tm805) double deletion strain to the scrm‐3(tm631) deletion strain to generate a triple deletion strain in the pad‐1(babIs1) background. We observed 31 ± 14 SNR4 puncta in the pad‐1; scrm triple deletion strain (Figure 3B), which is similar in number to the pad‐1(babIs1) control (Figure 2C). Therefore, disrupting scrm‐4 did not significantly decrease EVs (p > 0.1). We also did not observe an increase in EVs in the pad‐1(babIs1); scrm‐4(tm2239) scrm‐1(tm805) double deletion strain compared to the pad‐1(babIs1) controls (Figure 2C), suggesting that increased SCRM‐4 expression in pad‐1(babIs1) mutants does not partially suppress EV overproduction. These results suggest that although scrm‐4 expression showed signs of dynamic regulation, SCRM‐4 was not significantly compensating for related scramblases SCRM‐3 or SCRM‐1 to regulate EV production from the plasma membrane.

3.5. ABC Transporter CED‐7 Does Not Promote EV Production From Embryos

Given the variability in PLSCR scramblase results, we asked whether EV production was altered by an ATP‐dependent regulator of lipid exposure. The ABC transporter CED‐7 was an attractive candidate because CED‐7 is required for the release of EVs that have lost lipid asymmetry from dying cells (Mapes et al. 2012). We disrupted CED‐7 using an early premature stop codon allele (R5*), ced‐7(n1996) (Wu and Horvitz 1998), crossed into the pad‐1(babIs1) background. We detected 45 ± 23 SNR4 puncta in embryos from pad‐1; ced‐7 mutants (Figure 2A), showing a non‐significant 19% reduction in EVs produced (n = 33, p > 0.2). These data suggest that CED‐7 is not a major regulator of EV budding from the plasma membrane in the context of living embryonic cells.

3.6. ER Scramblase TMEM41 Homolog Knockdown but Not Knockout Reduces EV Production

Given that transporters on organelle membranes can also alter plasma membrane lipid asymmetry (Takar et al. 2016; Wang and Kinoshita 2023), we tested whether scramblases located on intracellular organelles could regulate EV release from the plasma membrane. As ER scramblases from the TMEM41 family are linked to viral particle production (Tang et al. 2022), we examined the TMEM41B homolog EGLI‐1, which is expressed in early embryos (Hashimshony et al. 2015). Knocking down egli‐1 in pad‐1(babIs1) mutants resulted in a 32%–56% decrease in EVs using two different RNAi clones (Figure 2B). Treating pad‐1 worms with the genomic clone (JA) resulted in 24 ± 19 SNR4 puncta (n = 30), while the cDNA clone (MV) resulted in 37 ± 21 SNR4 puncta (n = 87), each resulting in a significant decrease in EVs (p < 0.001). These knockdown results suggest that the intracellular scramblase EGLI‐1 can promote EV production from the plasma membrane.

As both egli‐1 RNAi clones had shown consistent suppression of EVs after egli‐1 knockdown, we disrupted EGLI‐1 using a deletion, egli‐1(gk278), which removes 470 bp of the promoter and the first half of the gene encoding the first four of six transmembrane domains (C. elegans Deletion Mutant Consortium 2012). We crossed egli‐1(gk278) into the pad‐1(babIs1) background and performed whole genome RNA sequencing. We found a 38‐fold decrease in egli‐1 transcripts in egli‐1(gk278) mutants (Supporting Information). The rare remaining transcripts started hundreds of bases before the usual egli‐1 transcript, which contained an open‐reading frame starting from Methionine 158 that would encode only the last two of six transmembrane domains. Deleting egli‐1 resulted in 45 ± 22 SNR4 puncta in pad‐1(babIs1) mutants (Figure 2A), which represented a non‐significant 19% reduction in EVs (n = 33, p > 0.1). These data show that the egli‐1 deletion had no significant effect on EV quantity in early embryos, again demonstrating that scramblase knockdown and deletion can produce quantitatively distinct results on EV production, similar to what we observed with scrm‐3.

Given the disparity between the egli‐1 RNAi and deletion results, we considered whether egli‐1 RNAi only targeted egli‐1 or additional genes. To test whether egli‐1 knockdown was able to decrease EV numbers in an egli‐1 mutant background, we performed RNAi on the pad‐1(babIs1); egli‐1(gk278) deletion strain. We observed 38 ± 18 SNR4 puncta after egli‐1 MV RNAi treatment and 33 ± 13 SNR4 puncta after empty vector control RNAi treatment (Figure 2E). We did not observe a significant change in EV numbers after egli‐1 knockdown (p > 0.1) and any trend was toward a non‐significant increase instead of the expected decrease in EVs. These results suggest that egli‐1 RNAi is only effective in the presence of the first half of the egli‐1 gene, despite the RNAi clone targeting the full gene. Furthermore, these findings reveal that the egli‐1 RNAi clone is unlikely to target other genes that regulate EV production, raising the possibility that egli‐1 mutants have genetic differences that compensate for the absence of EGLI‐1.

As we found similar patterns of inconsistent RNAi and deletion phenotypes with egli‐1 and scrm‐3, we examined their relative expression levels in either mutant. We found that scrm‐3 expression is increased in pad‐1; egli‐1 deletions compared to pad‐1 mutants, although this 2.5‐fold increase was not considered significant in the context of the RNA‐seq dataset (Supporting Information). To test for compensation or redundancy, we crossed the egli‐1 and scrm‐3 deletion strains together to create a double mutant strain in the pad‐1(babIs1) background. We observed 36 ± 23 SNR4 puncta in pad‐1; egli‐1; scrm‐3 double deletions (Figure 3B), which is similar in number to pad‐1(babIs1) control or pad‐1; scrm‐3 single deletion strains (p > 0.6). These data show that the non‐significant increase in scrm‐3 expression could not compensate for egli‐1 deletion. Furthermore, EGLI‐1 did not show redundancy with one of the PLSCR scramblases in the embryonic context.

3.7. Autophagy‐Associated ATG‐9 Promotes EV Production

To continue our examination of intracellular scramblases, we next investigated the trans‐Golgi/vesicular scramblase autophagy‐related protein 9 (ATG‐9), which is expressed in early embryos (Hashimshony et al. 2015) and whose homologs have been linked to viral particle production (Abernathy et al. 2019). After atg‐9 knockdown in pad‐1(babIs1) mutants, we observed 39 ± 26 SNR4 puncta (Figure 2B), showing a significant 31% decrease in the quantity of EVs (n = 29, p < 0.03). We disrupted ATG‐9 using a premature stop codon mutation, atg‐9(bp564) (Lin et al. 2013), which prevents autophagosome elongation in early embryos (Kolli et al. 2024). Embryos from the pad‐1; atg‐9 mutants averaged 35 ± 19 SNR4 puncta (Figure 2A), a significant 36% reduction in EVs (n = 27, p < 0.002). Therefore, we were able to recapitulate the effects of atg‐9 knockdown with a mutation in atg‐9, confirming that ATG‐9 has a role in fine‐tuning the quantity of EV production.

3.8. Potential Pathways of Genetic Compensation for Scramblase Deletion

In total, the significant suppression of pad‐1‐induced EVs from four scramblase RNAi knockdowns could only be recapitulated for half of the scramblase mutant strains we examined. We therefore compared our RNA sequencing results, starting with the mutant strain that showed significant suppression in EV numbers, the pad‐1(babIs1); anoh‐2(wur37) deletion. We found that the pad‐1; anoh‐2 deletion strain showed more similar gene expression to the wild type control strain and had more genes showing a significant change in expression by two‐fold or more when compared to pad‐1 mutants than the wild type control (Table 1). Indeed, the EV overproducing pad‐1 strain without any scramblase mutations significantly changed the expression of more genes compared to the wild type control strain than when anoh‐2 was deleted (Table 1). This suggests that there is little genetic compensation occurring in the large anoh‐2 deletion.

TABLE 1.

Differentially expressed genes in lipid transporter mutants.

Compared strain	Baseline strain	Differentially expressed genes	Upregulated genes	Downregulated genes
pad‐1(babIs1)	WT control	246	202	44
pad‐1(babIs1); anoh‐2(wur37)	WT control	74	40	34
pad‐1(babIs1); egli‐1(gk278)	WT control	2289	168	2121
pad‐1(babIs1); scrm‐3(tm631)	WT control	2108	204	1901
pad‐1(babIs1); anoh‐2(wur37)	pad‐1(babIs1)	199	18	181
pad‐1(babIs1); egli‐1(gk278)	pad‐1(babIs1)	1964	99	1865
pad‐1(babIs1); scrm‐3(tm631)	pad‐1(babIs1)	1754	59	1695

Note: Number of differentially expressed genes based on DESeq2 |log2FoldChange|≥1 and p _adj < 0.05. Gene lists are available in Tables S4–S10.

To test whether genetic compensation could be occurring in the mutant strains that failed to recapitulate the knockdown phenotypes, we compared the RNA sequencing results from the pad‐1(babIs1); egli‐1(gk278) deletion and the pad‐1(babIs1); scrm‐3(tm631) deletion. The egli‐1 and scrm‐3 deletion strains showed thousands of genes significantly changing their expression levels by two‐fold or more (Table 1), mostly genes down‐regulated in the deletion strains compared to wild‐type controls or pad‐1 mutants. Comparing the identity of the differentially expressed genes, most expression changes were shared between the pad‐1; egli‐1 deletion and the pad‐1; scrm‐3 deletion strains (Figure 4A,B, Supporting Information), suggesting that these deletions could fail to suppress EV release due to similar compensatory pathways. The order of magnitude difference in gene expression changes between the EV‐suppressing anoh‐2 deletion and the two non‐EV‐suppressing scramblase deletions suggests that there could be complex genetic pathways compensating for some scramblase deletions.

FIGURE 4.

Genetic compensation occurs in scramblase mutants. (A, B) Venn diagrams of differentially expressed genes compared to (A) control worms or (B) pad‐1(babIs1) mutants. Individual example genes are indicated in (A), and gene lists are available in Supporting Information. (C–E) GO term analysis of differentially regulated genes between the pad‐1(babIs1) EV‐overproducing strain and (C) pad‐1(babIs1); anoh‐2(wur37), (D) pad‐1(babIs1); egli‐1(gk278) or (E) pad‐1(babIs1); scrm‐3(tm631) mutant strains.

To identify potential compensating pathways involved in EV regulation in scramblase mutants, we performed Gene Ontology analysis on all differentially expressed genes. Consistent with the low number of differentially expressed genes, only a couple of small pathways showed significant differences between pad‐1 and the pad‐1; anoh‐2 deletion strain (Figure 4C). In contrast, a dozen pathways showed significant differences for the pad‐1; egli‐1 and pad‐1; scrm‐3 strains (Figure 4D,E), with the largest group of genes belonging to the GO term transmembrane transport. This GO term included the ABC transporter ced‐7, as well as a potential scramblase, vdac‐1. However, neither gene showed a significant change in expression in our analyses (Table S4–S11). Ion transport and transmembrane proteins involved in cell adhesion also showed significant enrichment in the egli‐1 and scrm‐3 deletion strains (Figure 4D,E), but these GO terms did not include any currently predicted scramblases. However, several scramblase families were originally identified as ion transporters (Rocha‐Roa and Vanni 2026), so we looked for ion transporters whose expression increases consistent with compensation for loss of a scramblase. We found that copper transporter genes homologous to SLC31A are significantly upregulated in the pad‐1; egli‐1 and pad‐1; scrm‐3 strains (Supporting Information). As a common feature of scramblase proteins is transmembrane domains, the differentially expressed genes from multiple GO terms could implicate transmembrane proteins characterized for other functions as potential scramblases in addition to other regulators of EV production.

The most significantly altered pathways in the pad‐1; egli‐1 and pad‐1; scrm‐3 strains included proteins that are unlikely to be lipid scramblases, such as uncharacterized apple domain containing proteins (GO term anatomical structure morphogenesis) or intraflagellar transport (IFT) genes from the cilium assembly GO term (Figure 4D,E, Supporting Information). In addition, although fatty acid metabolism was only a significant GO term for scrm‐3 mutants (Figure 4E), some of the most significantly altered genes were associated with lipid remodelling, such as the fatty acid desaturase fat‐7 in both egli‐1 and scrm‐3 mutants (Figure 4A, Supporting Information). We also observed gene regulation in pad‐1 mutants that could be related to intracellular phenotypes like enlarged multivesicular endosomes (Beer et al. 2018), such as the lysosomal lipase lipl‐1 downregulated in all four strains compared to the control (Figure 4A, Supporting Information). This RNA sequencing data set is likely to provide foundational insights into future studies on scramblases and lipid regulation to control EV release and other dynamic membrane processes.

4. Discussion

Our MATLAB‐based approach allowed us to reduce bias in quantitating levels of EVs using cmeAnalysis with the addition of a more selective SNR threshold. We developed a pipeline for automated detection in mid‐range EV‐producing strains, such as a partial loss of PAD‐1 function (Figure 1), facilitating the use of pad‐1(babIs1) strains for enhancer‐suppressor screens of EV production. Setting the position of the SNR threshold required balancing sensitivity with selectivity. Images with single EVs on the surface of the embryo may have less reporter fluorescence, making the EV fluorescence harder to distinguish from other biological sources of fluorescence. A more selective cut‐off (SNR 4 in this study) recapitulated manual detections in low EV producing strains (Figure 1H), which is important to screen for factors promoting EV release from the plasma membrane. However, the correlation coefficient for low EV strains indicates a need for further improvement (Table S3). Furthermore, if a perturbation decreased loading of the mCh::PH::CTPD reporter into EVs, this could also decrease the EV signal and lead to fewer SNR‐based detections. The validation of AI‐based puncta counters will likely be a fruitful strategy to detect mutations that decrease EV abundance or cargo loading.

In high EV producing strains, we also observed decreases in cmeAnalysis detections at less selective SNRs (Figure 1E). This undercounting could be due to local SNR decreases when two bright fluorescent puncta are close to each other. Indeed, mutants that overproduce EVs show clusters of EVs in electron micrographs (Wehman et al. 2011; Beer et al. 2018). Given the limited lateral resolution of widefield imaging, clusters of EVs will appear merged together into larger, brighter fluorescent puncta, which make it unlikely that each fluorescent puncta in EV overproducers represents a single EV, further exacerbating the undercounting of EVs from fluorescence images. Therefore, improvements to imaging resolution, sampling, and/or analysis pipelines would enable the robust use of automated analysis for quantitation of severe EV‐overproducing strains.

Together, our data suggest that EV production is promoted by a variety of lipid scramblases. Using a mid‐range EV‐overproducing strain, we observed a small but significant role of multiple scramblases in vesicle budding based on knockdown RNAi experiments (Figure 2B), consistent with previous studies suggesting that disruption of lipid asymmetry alters membrane packing to influence membrane blebbing (Wang et al. 2025). However, it is worth noting that the largest effects of scramblase knockdown were only a two‐fold decrease in EVs, which is in contrast to the order‐of‐magnitude increases observed when disrupting lipid flippases or their regulators like PAD‐1 (Fazeli et al. 2020; Pitts et al. 2022, 2023). Most flippases belong to a single family of ATPase proteins, while scramblase activity exists across diverse families of proteins (Sakuragi and Nagata 2023). The sheer number of scramblases could explain the relatively minor impact shared among diverse scramblases. Our study was designed to analyse changes in EV number by over 30%, meaning that we were unlikely to detect transporters contributing to less than 20% of EV production in the pad‐1(babIs1) background. The natural variation in EV overproduction in pad‐1(babIs1) mutants (±50%) also makes the detection of minor contributors challenging. Despite these caveats, our candidate screen with a half dozen scramblases suggests that there are many scramblases that act in opposition to flippases to disrupt lipid asymmetry and promote EV production. This likely includes additional scramblases that were not examined in this study (Table S11). There are also likely to be proteins whose scramblase activity remains to be identified. A common feature of scramblase proteins is a hydrophilic groove in their transmembrane domains to escort the hydrophilic headgroups of lipids across the hydrophobic barrier of the bilayer membrane, leading to the hypothesis that many transmembrane proteins can change conformations to reveal a hydrophilic groove and scramble lipids. Indeed, half a dozen new scramblases have been identified since we carried out the initial RNAi screen, and more scramblases are likely to be revealed by future studies (Rocha‐Roa and Vanni 2026).

The physiological context of EV biogenesis is also likely to influence the relative roles of different lipid transporters in EV production. Although the ABC transporter CED‐7 promotes EV release from the plasma membrane of apoptotic cells (Mapes et al. 2012), we found no role for CED‐7 in EV release in the early embryo (Figure 2B). However, our findings do not preclude a role for CED‐7 or cell death‐associated scramblases in promoting EV release from living cells in other tissues or in response to other stimuli beyond apoptosis. Indeed, the Xkr8 scramblase is activated by caspase‐mediated cleavage of its autoinhibitory C‐terminus during apoptosis, but Xkr8 autoinhibition can also be relieved by phosphorylation of its C‐terminus in living cells (Sakuragi et al. 2019). In embryos, we found consistent roles promoting EV production for a calcium‐activated Anoctamin ANOH‐2. Early embryos experience a wave of calcium after fertilisation (Toperzer et al. 2023), which could activate scramblases like Anoctamin channels after fertilisation. We also found an embryonic role for macroautophagy‐associated ATG‐9, whose scramblase activity is thought to be induced in response to an autophagy‐associated lipid transfer protein delivering lipids to a single leaflet of the bilayer (Ghanbarpour et al. 2021). As early embryos form many ATG‐9‐dependent autophagosomes for the clearance of sperm organelles (Kolli et al. 2024), this could similarly explain the importance of ATG‐9 lipid scrambling in EV release from early embryos. As scramblases have disparate activation mechanisms and physiological roles, their timing of activation and relative importance for lipid exposure and EV regulation are likely to differ in tissues experiencing distinct stimuli.

Although Anoctamins like TMEM16F and ANOH‐2 were already known to impact EV production (Sims et al. 1989; Fujii et al. 2015; Wang et al. 2025), it was unexpected that the only other scramblase to show consistently significant results with both RNAi knockdown and mutant alleles would be macroautophagy‐associated ATG‐9. Given that wild‐type embryos average only a couple of EVs on their surface, the biological significance of ATG‐9 protein acting to increase EV numbers by one‐third seems unclear. However, reducing neuronal EV production by one‐half was sufficient to rescue cell morphology when EVs were overproduced (Park et al. 2024), demonstrating that minor alterations in EV number can be biologically significant. It remains to be determined, however, whether the scramblase activity of ATG‐9 disrupts plasma membrane asymmetry to regulate EV production in a direct fashion. It is also possible that the effect of atg‐9 disruption on EV abundance is secondary to its canonical role regulating autophagosome formation and autolysosomal degradation or its non‐canonical roles in protein secretion (Peng et al. 2025; Zhang et al. 2025). ATG‐9 can be trafficked from trans‐Golgi network vesicles to the plasma membrane in C. elegans neurons (Yang et al. 2022), but the localisation and dynamics of ANOH‐2 or ATG‐9 have not been investigated in early embryos, leaving it unclear where ANOH‐2 and ATG‐9 are activated for EV regulation.

While we observed a significant reduction in EV production after three‐day RNAi knockdown experiments of four different scramblases (Figure 2B, Wang et al. 2025), we observed no significant change in EV counts in half of the corresponding scramblase mutant strains (Figure 2A,C). It is unlikely that this is due to off‐target effects of the RNAi clones, as treating the deletion strains with the corresponding RNAi did not decrease EVs (Figure 2D,E). Furthermore, the only RNAi clone with predicted off‐targets, scrm‐3 JA, did not alter EV production, while the more specific clone, scrm‐3 LJ, had significant effects (Figure 2B). The absence of a significant EV phenotype in half of the mutant strains could be explained by compensation from other genes in response to constitutive loss of one scramblase. Indeed, our RNA‐seq data revealed that the expression of thousands of genes was similarly altered in scramblase knockouts that did not recapitulate the EV phenotypes of scramblase knockdowns (Figure 4A,B). As increased expression of the related scramblase scrm‐4 was not necessary to compensate for the loss of scrm‐3 (Figure 3A,B), the mechanisms of scramblase compensation remain unclear. Most gene expression changes in the scrm‐3 and egli‐1 strains involved downregulation (Table 1), not overexpression as hypothesized for scramblases or their activators. Regardless, these data sets should provide the community with candidate regulators for future studies of EV release from the plasma membrane.

In addition to genetic regulators, our work also reveals evidence for environmental regulation of EV release from the plasma membrane. Imaging performed before and after a lab move using the same strain and microscope revealed significantly different baseline levels of EVs. Membrane fluidity and lipid composition are impacted by differences in air pressure and oxygen availability (Brooks 2014; Nakazawa et al. 2016), differences to be expected between the high plains of Denver, CO and the near sea level savannah of College Station, TX. A difference in room temperature could also be a contributing factor, as we previously discovered that each degree Celsius can contribute eight additional EVs in the pad‐1(babIs1) strain within a few hours of incubation (Wang et al. 2025). These findings reveal the importance of controlled conditions as well as highlight potential explanations for lab‐to‐lab variation in EV counts. Furthermore, inconsistent results between RNAi and mutant conditions could reflect experimental differences, such as metabolite differences between the HT115 and OP50 bacterial strains fed to the worms (Neve et al. 2020). Indeed, metabolic effects on EV production can impact their usefulness as biomarkers (Welsh et al. 2024). Therefore, it will be important to investigate environmental regulators of EV production in addition to the proteins that regulate EV abundance. In summary, many physiological stimuli are likely to use lipid regulation to fine‐tune EV release from the plasma membrane.

Author Contributions

Bianca L. Guerra: investigation, data curation, formal analysis, validation, visualization, writing – review and editing, methodology, conceptualization. Tanner Chase White: conceptualization, methodology, investigation, validation, visualization, writing – original draft, formal analysis, data curation. Alissa Rumin: data curation, formal analysis, methodology, validation, visualization, investigation, conceptualization, writing – original draft. Lauren Pitts: methodology, investigation, data curation. Alexander Nguyen: software, investigation, validation, data curation, visualization. Vanshika Chauhan: software, investigation. Dinah Loerke: methodology, software, data curation, investigation, validation, supervision, formal analysis, visualization, writing – review and editing. Ann M. Wehman: writing – original draft, writing – review and editing, conceptualization, methodology, software, data curation, supervision, resources, project administration, formal analysis, validation, visualization, funding acquisition, investigation.

Conflicts of Interest

 The authors declare no competing interests.

Funding

This research was supported by the National Institutes of Health (NIH) grant R35GM152234 and an Allen Distinguished Investigator Award, a Paul G. Allen Frontiers Group‐advised grant of Allen Family Philanthropies.

Supporting information

Supporting Information: jex270197‐sup‐0001‐SuppMat.xlsx

Data Availability Statement

The EV count data and differential gene expression data are available as a Supplemental Data spreadsheet. Raw sequences and analysis are available in the GEO repository GSE346039. The MATLAB scripts we developed to work with cmeAnalysis from the DanuserLab are available on GitHub under the username WehmanLab. Raw fluorescence image data will be provided upon request.