An extracellular vesicle targeting ligand that binds to Arc proteins and facilitates Arc transport in vivo

  1. Peter H Lee  Is a corresponding author
  2. Michael Anaya
  3. Mark S Ladinsky
  4. Justin M Reitsma
  5. Kai Zinn  Is a corresponding author
  1. Division of Biology and Biological Engineering, California Institute of Technology, United States

Abstract

Communication between distant cells can be mediated by extracellular vesicles (EVs) that deliver proteins and RNAs to recipient cells. Little is known about how EVs are targeted to specific cell types. Here, we identify the Drosophila cell-surface protein Stranded at second (Sas) as a targeting ligand for EVs. Full-length Sas is present in EV preparations from transfected Drosophila Schneider 2 (S2) cells. Sas is a binding partner for the Ptp10D receptor tyrosine phosphatase, and Sas-bearing EVs preferentially target to cells expressing Ptp10D. We used co-immunoprecipitation and peptide binding to show that the cytoplasmic domain (ICD) of Sas binds to dArc1 and mammalian Arc. dArc1 and Arc are related to retrotransposon Gag proteins. They form virus-like capsids which encapsulate Arc and other mRNAs and are transported between cells via EVs. The Sas ICD contains a motif required for dArc1 binding that is shared by the mammalian and Drosophila amyloid precursor protein (APP) orthologs, and the APP ICD also binds to mammalian Arc. Sas facilitates delivery of dArc1 capsids bearing dArc1 mRNA into distant Ptp10D-expressing recipient cells in vivo.

Editor's evaluation

The manuscript addresses how extracellular vesicles (EV) are targeted to their recipient cells once they are produced and released. The study shows that a transmembrane protein Sas gets incorporated into EVs, and this protein binds to its receptor Ptp10D on target cells, thus targeting the EVs. The expression of dARC1 in the EV-producing cells leads to the increased expression of the protein dARC1 protein and mRNAs in the recipient cells.

https://doi.org/10.7554/eLife.82874.sa0

Introduction

Extracellular vesicles (EVs) are mediators of cell-cell communication that transport specific protein and RNA cargoes. They are a heterogeneous collection of vesicular structures that are exported from cells by a variety of mechanisms. Exosomes are 30–150 nm in diameter and are released into cell supernatants via fusion of multivesicular bodies (MVBs) with the plasma membrane. Exosomes and other EVs carry specific proteins and RNAs, and EVs derived from different cell types contain different cargoes. EV cargoes are biomarkers for specific diseases. Because EVs can encapsulate RNAs and protect them from degradation, and then deliver those RNAs to recipient cells, they represent a promising new type of therapeutic agent (O’Brien et al., 2020; Teng and Fussenegger, 2020).

Little is known about mechanisms involved in EV targeting to specific cell types. EVs are internalized into cells after receptor binding using a variety of endocytic mechanisms, resulting in the delivery of their cargoes into the recipient cells. They can also directly activate intracellular signaling without endocytosis by interacting with cell surface receptors. In this paper, we identify Stranded at second (Sas), a large Drosophila cell surface protein (CSP; Schonbaum et al., 1992), as an EV targeting ligand. Full-length Sas (SasFL) has an extracellular domain (ECD) containing a signal peptide, a unique N-terminal region, four von Willebrand factor C (VWFC) domains, and three Fibronectin Type III (FN-III) repeats (Figure 1a). A shorter isoform, Sasshort, which is translated from an alternatively spliced mRNA, lacks a 345 amino acid (aa) region of the ECD. The two Sas isoforms share a single transmembrane (TM) domain and a short cytoplasmic domain (ICD).

Figure 1 with 1 supplement see all
Localization of Sas isoforms.

(a) Schematic diagram of the SasFL protein. The Sasshort isoform lacks the EVT region. (b) SasFL moves away from expressing cells. V5-SasFL was expressed together with mCD8-GFP (transmembrane CSP) in Apterous neurons in late stage 16 embryos. (i,iv) double-labeling (V5, magenta; GFP, green); (ii, v) V5 channel; (iii, vi) GFP channel. (i-iii) Brain lobes. GFP labels cell bodies (arrowheads in iii) and axon tracts. V5 labels cell bodies only weakly (arrowheads in ii), strongly labels some axon tracts, and localizes to the periphery (sheath) of the brain lobes (arrows in ii). (iv-vi) Ventral nerve cord. GFP strongly labels Ap VNC cell bodies (arrowheads in vi) and weakly labels Ap axons (arrow in vi). V5 weakly labels cell bodies (arrowheads in v), and strongly labels segments of axons (arrows in v) and areas adjacent to the axons. These may be extracellular matrix and/or glial sheaths. Scale bar, 10 µm. (c) Western blot, showing that SasFL localizes to EVs. EVs were prepared from supernatants from S2 cells expressing the indicated protein, and equal amounts of cell lysate proteins and EV proteins were loaded on the gel. Top panel, anti-V5 blot. SasFL migrates slightly above the 250 kD marker, and Sasshort slightly below it. Middle panel, Wg (EV marker). Almost all of the Wg is in EVs. Bottom panel, anti-β-actin (cytoplasmic marker) blot. There is no signal in EVs. In these samples, 60% of SasFL and 10% of Sasshort is in EVs; the average over 4 experiments was 46% in EVs for SasFL and 10% for Sasshort. The absence of β-actin signal in the EVs shows that they are not heavily contaminated by cytosol. (d) Localization of endogenous Sas isoforms in the embryonic gut. Wild-type (wt) late stage 16 embryos were triple-stained for SasFL (using the [Schonbaum et al., 1992] antiserum, which primarily recognizes the EVT region; green), Sasshort (using our antipeptide antibody; red), and Crumbs (apical marker; blue). (i-iv) Foregut; (v-viii), hindgut. Note that in both gut regions SasFL colocalizes with Crumbs at the apical (luminal) cell surfaces (arrows), while anti-Sasshort labels the entire width of the gut wall (brackets). See Figure 1—figure supplement 1 for images of anti-Sasshort staining of wt and sas mutant embryos, demonstrating antibody specificity. Scale bar in d-i, 20 µm; in (d-v), 10 µm.

Figure 1—source data 1

Source data files for Figure 1 include raw and labelled images for the western blots shown in panel (c); source data files for Figure 1—figure supplement 1 include raw and labelled images for the western blots shown in panels (a) and (d).

https://cdn.elifesciences.org/articles/82874/elife-82874-fig1-data1-v2.zip

Sas is commonly used as a marker for the apical surfaces of epithelially derived cells, including tracheal cells in the respiratory system. sas mutant larvae die at or before second instar (hence the name stranded at second, which is derived from baseball terminology) and have tracheal phenotypes in larvae but no known phenotypes in embryos (Schonbaum et al., 1992). A tyrosine motif in the Sas ICD was found to bind to the PTB domain of Numb (Chien et al., 1998), an protein involved in endocytosis that is a negative regulator of Notch signaling. Sas has no obvious mammalian orthologs, but there are many mammalian CSPs that contain VWFC and FN-III domains.

In earlier work, we identified the receptor tyrosine phosphatase (RPTP) Ptp10D as a binding partner for Sas, and showed that Sas::Ptp10D interactions regulate embryonic axon guidance, as well as glial migration and proliferation (Lee et al., 2013). Ptp10D is one of the two Drosophila R3 subfamily RPTPs, which have ECDs composed of long chains of FN-III repeats. Sas::Ptp10D interactions also control the elimination of neoplastic epithelial clones by surrounding normal tissue. Sas is on normal epithelial cells, and it relocalizes to the parts of their cell surfaces that are adjacent to the neoplastic clone and binds to Ptp10D on the neoplastic cells. Ptp10D in turn relocalizes and dephosphorylates the EGF receptor tyrosine kinase, leading to death of the neoplastic cells (Yamamoto et al., 2017). The Sas ECD has other binding partners as well, because it interacts with cells that do not express Ptp10D in live embryo staining assays (Lee et al., 2013).

Mammalian Arc is a locally translated dendritic protein that regulates synaptic plasticity, in part by modulating endocytosis of AMPA receptors (Chowdhury et al., 2006; Shepherd et al., 2006). The Drosophila genome encodes two proteins distantly related to mammalian Arc, dArc1 and dArc2. The dArc2 gene, which encodes a truncated protein, was likely generated by a gene duplication, and the dArc1 and dArc2 genes are adjacent (Mattaliano et al., 2007). dArc1 functions in larval and adult brain neurons to regulate aspects of metabolism (Mattaliano et al., 2007; Mosher et al., 2015; Keith et al., 2021). Arc and dArc1 evolved independently from retrotransposon Gag proteins (Shepherd, 2018). Remarkably, they both form virus-like capsids that can encapsulate Arc mRNAs and are transported between cells via EVs (Ashley et al., 2018; Pastuzyn et al., 2018; Hantak et al., 2021). dArc1, but not dArc2, has a C-terminal Zn2+ finger that might be involved in nucleic acid binding (Pastuzyn et al., 2018; Erlendsson et al., 2020). Mammalian Arc lacks Zn2+ fingers, but RNA is required for normal capsid assembly (Pastuzyn et al., 2018). Drosophila dArc1 capsids bearing dArc1 mRNA move from neurons to muscles across larval neuromuscular junction (NMJ) synapses, and dArc1 transfer is required for activity-induced induction of morphological synaptic plasticity (Ashley et al., 2018).

Here we show that full-length Sas expressed in cultured Drosophila cells localizes to EVs that preferentially target to cells expressing Ptp10D. Sas binds to dArc1 in EVs via a tyrosine motif in its ICD that is conserved between Sas and the human and Drosophila APP orthologs. This finding led us to examine mammalian Arc as well, and we showed that it also binds to Sas and APP. The interaction between APP and Arc is of interest because several studies have implicated Arc in control of β-amyloid accumulation and Alzheimer’s disease (AD) (Wu et al., 2011; Landgren et al., 2012; Bi et al., 2018). APP also localizes to EVs (Laulagnier et al., 2018; Pérez-González et al., 2020).

Since Sas binds to dArc1, which can encapsulate its own mRNA (Ashley et al., 2018), we then investigated whether Sas EVs can target dArc1 mRNA to Ptp10D-expressing recipient cells in vivo. Here, we show that co-expression of dArc1 and full-length Sas in embryonic salivary glands (SGs) causes dArc1 mRNA to appear in distant tracheal cells that express Ptp10D.

Results and discussion

Sas is an EV targeting ligand

Sas has two isoforms, translated from alternatively spliced mRNAs. SasFL (PB/PD isoform) is a 1693 aa single-transmembrane cell-surface protein with a short (37 aa) ICD. It contains a 345 aa region (EVT) between the VWFC and FN-III domains that is lacking in the Sasshort (PA/PC) isoform (Figure 1a). We expressed SasFL tagged with an N-terminal V5 epitope tag (inserted immediately after the signal sequence) in embryonic late stage 16 Apterous (Ap) neurons, which consist of paired neurons (one per hemisegment) in the ventral nerve cord (VNC) and scattered neurons in the brain lobes. It was expressed together with mCD8-GFP, which is also a transmembrane CSP. The GFP signal was restricted to Ap neuron cell bodies, with faint staining on the axons. However, V5-SasFL was observed in sheaths around brain lobes and in areas adjacent to axons in the VNC, as well as in puncta throughout the VNC and brain (Figure 1b). The V5 signal in cell bodies was very weak, especially in the brain.

We also expressed V5-Sasshort in Ap neurons, and observed that V5 staining was restricted to cell bodies and axons in the VNC, and to cell bodies in the brain (Figure 1—figure supplement 1). Thus, although SasFL and Sasshort have the same TM and ICD, they differ in subcellular localization, with Sasshort being retained in expressing cells and SasFL moving away from these cells and into the extracellular matrix.

Movement of V5-SasFL, and presumably of endogenous SasFL, away from its source could occur through cleavage of the Sas ECD from the cell surface or by release of intact Sas in EVs. To distinguish between these possibilities, we expressed V5-tagged SasFL and Sasshort in transiently transfected Drosophila Schneider 2 (S2) cells, which express endogenous Sas at almost undetectable levels. We prepared EVs from S2 cell supernatants using the Invitrogen Exosome Isolation Kit or by ultracentrifugation, and analyzed their contents by western blotting. EVs contained the Wg protein, which is a marker for EVs in S2 cells and is present at very low levels in lysates (Koles et al., 2012; Figure 1c, Figure 1—figure supplement 1). They lacked β-actin, showing that they are not heavily contaminated by cytosol. We found that about 50% of V5-SasFL localized to EVs, while ~90% of V5-Sasshort was retained in the cell lysate (Figure 1c). We did not observe any proteolytic cleavage products in EVs or unpurified supernatants.

EV preparations made using the Exosome Isolation kit have similar characteristics to those generated by ultracentrifugation (Skottvoll et al., 2019), and the kit requires much less material, allowing EVs to be purified from small-scale transfections. However, to confirm results obtained with the kit, we also purified EVs from S2 cells expressing V5-tagged SasFL using a standard ultracentrifugation protocol (Théry et al., 2006), and showed that SasFL and Wg are also present in these EVs (Figure 1—figure supplement 1).

The commonly used rabbit antiserum against Sas primarily recognizes the EVT region, so embryo staining reveals the localization of endogenous SasFL (Schonbaum et al., 1992). To visualize Sasshort, we made an anti-peptide antibody against a sequence spanning an exon junction in the PA/PC isoforms. This recognizes Sasshort, and staining with the antibody is eliminated in sas mutant embryos (Figure 1—figure supplement 1). Double-staining of the foregut and hindgut with the two Sas antibodies showed that SasFL localizes to apical cell surfaces, while Sasshort is distributed across the entire cell membrane (Figure 1d). These data imply that the EVT sequence lacking in Sasshort is required for both apical localization and targeting to EVs. Polarized cells can release EVs with different cargoes from their apical and basolateral surfaces (Matsui et al., 2021), so EV targeting could be downstream of apical localization in vivo. S2 cells are unpolarized, however, so this mechanism is unlikely to apply to trafficking of SasFL to EVs in cultured S2s.

Analysis of SasFL EVs by electron microscopy

To demonstrate that Sas is actually on EVs, we used immuno-EM and EM tomography to analyze purified EV preparations from V5-SasFL-expressing S2 cells. The tomographic images showed that the EVs span a range of sizes, from ~30 nm in diameter to >100 nm, and that they are a mixture of single and double-membrane vesicles (Figure 2c and d, Figure 2—figure supplement 1). For immuno-EM, we incubated EVs with anti-V5, followed by gold-labeled anti-mouse secondary antibody. Figure 2a shows a typical image, in which an EV is associated with multiple 10 nm gold particles. The distance between the EV membrane (yellow bracket: diameter of the vesicle) and a gold particle (white bracket: distance between membrane and a particle) can be more than 40 nm. This likely reflects the large size of the Sas ECD, in which the N-terminal V5 epitope is separated by 1590 aa from the TM domain. The distance is variable, however, because the Sas ECD, which is composed of a chain of domains separated by linkers, is likely to be flexible and able to adopt many different conformations. The region outside the membrane boundary is of higher density, probably because it represents the protein sheath around the EV membrane. To further characterize Sas localization, we performed an experiment in which EVs were incubated with both mouse anti-V5 and rabbit anti-Sas, which primarily recognizes the EVT region in the middle of the ECD, followed by 10 nm gold particle-labeled anti-mouse secondary antibody and 5 nm gold particle-labeled anti-rabbit secondary antibody. Figure 2b shows an EV that is associated with multiple 10 nm (arrow) and 5 nm (arrowhead) gold particles.

Figure 2 with 2 supplements see all
Analysis of EVs by electron microscopy and nanoparticle tracking analysis.

(a, b) Immuno-EM images of EVs from a purified EV prep from V5-SasFL-expressing S2 cells. EV outline (membrane) diameters are indicated by yellow brackets. White brackets, separation between EV outline and a gold particle. (a) Immuno-EM with 10 nm anti-V5 gold particles (arrows). (b) Immuno-EM with both 10 nm anti-V5 (large gold, arrow) and 5 nm anti-Sas (small gold, arrowhead). (c) EM tomogram of an empty double-membrane vesicle (arrows). Apparent EV sizes differ between immuno-EM and tomography, which use very different preparation methods. A low-mag view of a single slice from an EM tomogram of an EV preparation is shown in Figure 2—figure supplement 1. (d) EM tomogram of a double-membrane vesicle (arrows) with a capsid-sized denser object inside it (arrowhead). Figure 2—figure supplement 1 shows a 3D reconstruction of this EV. Empty and filled single-membrane EVs were also observed (Figure 2—figure supplement 1). Scale bars in (a–d), 50 nm. (e) Nanoparticle Tracking Analysis (conventional NTA with NanoSight) of purified EV preparations from untransfected (green curve) and SasFL-expressing (blue curve) S2 cells. The mean size distribution is indicated. Standard error indicated by red color around curves. Source data files include an Excel file of raw data for the NTA analysis, the conversion of the numbers from numbers of EVs per sample to numbers of EVs per cell, based on cell counts, and plots of the data.

To analyze the numbers and sizes of EVs from SasFL-expressing and control S2s, we examined purified EVs using Nanoparticle Tracking Analysis with a NanoSight instrument (NTA, System Biosciences, LLC). We observed that the distribution of EV diameters was shifted toward smaller values in cells expressing SasFL (mean diameter = 102 nm, vs. 129 nm for control cells) (Figure 2e). Expression of SasFL increased the number of EVs per cell in the exosome size range (30–150 nm in diameter) by 44%, and the number of EVs per cell of <100 nm in diameter by 72%, suggesting that the presence of high levels of SasFL increases the rate of EV production.

Many of the EVs from V5-SasFL-expressing S2 cells EVs had denser objects within their boundaries (Figure 2d, Figure 2—figure supplement 1). These were typically 30–40 nm in diameter. Such objects were also found within EVs from untransfected S2 cells (Ashley et al., 2018), so their presence does not require expression of SasFL.

SasFL EVs target to cells expressing Ptp10D

Having shown that SasFL moves away from expressing neurons in the embryo and is an EV component, we then asked whether it can be incorporated into distant cells in vivo, presumably through endocytosis of EVs. We expressed V5-SasFL in 3rd instar larval salivary glands (SGs) using an SG-specific GAL4 driver, Sage-GAL4. SG-specific expression of dsRed from this driver in whole larvae is shown in Figure 3—figure supplement 1. We then visualized V5 staining in other tissues. We found that V5-SasFL made in SGs is present in imaginal discs, which are separated from SGs by larval hemolymph (wing and haltere discs shown in Figure 3a–c). There was no V5 staining in imaginal discs from driver-alone larvae (compare Figure 3a and b). We also expressed V5-Sasshort in SGs, and this was not observed in imaginal discs (Figure 3c). This is consistent with the observation that V5-Sasshort does not move away from expressing Ap neurons and is restricted to cell lysates in S2 cells (Figure 1b and c). wt discs express the Sas receptor Ptp10D at low levels (see Figure 3—figure supplement 1), but we do not know whether Ptp10D is required for SasFL binding to discs. Overexpressed Ptp10D can stimulate SasFL accumulation in discs, however (see Figure 3—figure supplement 1).

Figure 3 with 1 supplement see all
Transfer of SasFL to recipient cells.

(a) A third instar wing disc, with a portion of the haltere disc (right side), from a Sage-GAL4/+ (SG-specific driver) larva, showing no V5 staining. (b) A wing disc, with a portion of the haltere disc, from a Sage >V5-SasFL larva, showing bright V5 staining. (c) A wing disc, with a portion of the haltere disc, from a Sage >V5-Sasshort larva, showing little or no V5 staining. Imaginal discs display no expression of GFP or mCherry reporters driven by Sage-GAL4. (d-f) Wing discs incubated with EVs from V5-SasFL-expressing S2 cells and stained with anti-V5. For anti-Ptp10D and anti-Numb staining, see Figure 3—figure supplement 1. (d) ap-GAL4/+; (e) ap >Ptp10 D; (f) ap >Ptp10 D+Numb. (d) Low levels of anti-V5 staining are observed. (e) Higher levels are observed in disc folds, which also express Ptp10D (Figure 3—figure supplement 1). (f) Bright anti-V5 staining is observed throughout the disc. This pattern matches anti-Numb staining (Figure 3—figure supplement 1). Scale bars in (a) and (d), 50 µm. (g) Transfer of SasFL from EVs into recipient S2 cells. Supernatants from S2 cells expressing V5-SasFL were incubated with cultures of untransfected S2 cells or cells expressing Ptp10D, Numb, or both, and cell lysates analyzed by western blotting with anti-V5. Note that V5-SasFL levels were elevated relative to control cells by expression of either Ptp10D or Numb, and that levels were further increased by coexpression of Ptp10D and Numb coexpression. (h) Quantitation of results from panels (d-f) and (g). Levels of transferred V5-SasFL were increased by ~fourfold relative to untransfected controls by Ptp10D+Numb coexpression in S2 cells (n=6), and by ~threefold relative to ap-GAL4/+control by Ptp10D+Numb coexpression in wing discs (n=5). Quantitation was done using Image J.

Figure 3—source data 1

Source data files include raw and labelled images for the western blots shown in panel (g), and an Excel file of the quantitation of the western blot and disc immunofluorescence signals used to generate panel (h).

https://cdn.elifesciences.org/articles/82874/elife-82874-fig3-data1-v2.zip

To examine the mechanisms involved in specific targeting of Sas EVs, we added supernatants (EV fraction) from V5-SasFL-expressing S2 cells to S2 cell cultures and analyzed recipient cell lysates by western blotting. We observed that expression of the Sas receptor Ptp10D in recipient cells increased V5-SasFL levels in these cells, as did expression of Numb, a regulator of endocytosis that binds to the Sas ICD (Chien et al., 1998). Expression of both Ptp10D and Numb produced a synergistic effect, increasing V5-SasFL by ~fourfold relative to untransfected recipient cells (Figure 3g–h). We speculate that binding of Numb to the Sas ICD increases Sas uptake and/or protects endocytosed Sas from degradation.

We then developed an assay to examine the effects of Ptp10D and Numb on Sas targeting in larval cells by incubating dissected 3rd instar wing imaginal discs with V5-SasFL supernatants. Wing discs from driver-alone (control) larvae displayed weak V5 staining after incubation with V5-SasFL EVs. Staining was increased in discs from larvae expressing Ptp10D in imaginal discs, and further elevated (~threefold increase relative to driver control) by expression of both Ptp10D and Numb (Figure 3d–f and h; Figure 3—figure supplement 1).

Sas binds to dArc1 and mammalian Arc via a conserved tyrosine motif

To examine whether Sas interacts with specific EV cargoes, we made EV preparations from S2 cells expressing V5-SasFL and from untransfected control cells, lysed them with nonionic detergent, incubated the lysate with anti-V5-coupled magnetic beads, and analyzed bead-bound proteins by mass spectrometry (Figure 4a, Supplementary file 1). We ranked the identified proteins by their degree of enrichment in the V5-SasFL sample relative to the control. Proteins that are present in the V5-SasFL sample should include EV cargoes that bind to SasFL and are therefore present in V5 IPs, plus proteins that nonspecifically bind to V5 beads. Proteins in the control sample would be only those that nonspecifically bind to V5 beads. We observed that the most highly enriched protein (after Sas itself) is dArc1 (22-fold) (Figure 4b). dArc2 is #7 on the list (6-fold). dArc1 mRNA, presumably encapsulated within dArc1 capsids, is known to be a prominent mRNA component of EVs from Drosophila cultured cells (Lefebvre et al., 2016; Ashley et al., 2018). Our data suggest that dArc1, and possibly dArc2, associate with SasFL, since they are enriched in V5 IPs from cells expressing V5-SasFL. We then went on to show that dArc1 binds directly to the Sas ICD (see below).

Interactions of Sas, Appl, and APP with Arcs.

(a) Protocol for mass spectrometry analysis. Purified EVs from control S2 cells or S2 cells expressing V5-SasFL were lysed and IP’d with anti-V5, followed by protease digestion and mass spectrometry analysis. (b) Mass spectrometry results. The 7 proteins present at the highest levels in IPs from V5-SasFL EVs relative to IPs from control EVs (>6 fold ratio) are listed. Sas itself was the most highly enriched protein, as expected. dArc1 and dArc2 were enriched by 22-fold and 6-fold, respectively. (c) Co-IP/western blot analysis of association between Sas and Arc fusion proteins in transfected S2 cells. S2 cells were transfected with the V5-mCD8ECD-SasTM-ICD fusion protein construct, or with equivalent constructs in which the Sas ICD was replaced by the Appl or APP ICD, with or without Myc-tagged dArc1 or mammalian (rat) Arc (rArcFL) constructs. Lysates (Input 1%) were blotted with anti-Myc and anti-V5 (left), and IP’d with anti-Myc nanobody and blotted with anti-V5 and anti-Myc (right). Anti-V5 bands of the correct size were observed in anti-Myc IPs when dArc1-Myc was expressed with Sas or Appl ICD constructs (red symbols and boxes), and when Myc-rArcFL was expressed together with Sas or APP ICD constructs (blue symbols and boxes)(n=6). (d) Direct binding of purified GST-dArc1 and GST-rArcFL fusion proteins to Sas, APP, and Appl ICD peptides, as well as to scrambled and YDNPSY deletion mutant Sas peptides. Biotinylated peptides were bound to streptavidin magnetic beads, which were incubated with GST-Arc proteins, followed by western blotting of bead-bound proteins with anti-GST. GST-dArc1 bound to the wild-type, but not to scrambled or YDNPSY deletion mutant Sas ICD peptides, while GST-rArcFL bound to wild-type Sas and APP ICD peptides. (e) Sequences of the complete Sas, APP, and Appl ICDs, corresponding to biotinylated peptide sequences. The conserved tyrosine motif is boxed, with tyrosines in red. *, stop codons.

Figure 4—source data 1

Source data files include raw and labelled images for the western blots shown in panels (c) and (d) and an Excel file of the table in panel (b).

https://cdn.elifesciences.org/articles/82874/elife-82874-fig4-data1-v2.zip
Figure 4—source data 2

Co-IP analyses raw data.

https://cdn.elifesciences.org/articles/82874/elife-82874-fig4-data2-v2.zip
Figure 4—source data 3

Peptide binding assay raw data.

https://cdn.elifesciences.org/articles/82874/elife-82874-fig4-data3-v2.zip
Figure 4—source data 4

MS analysis result table.

https://cdn.elifesciences.org/articles/82874/elife-82874-fig4-data4-v2.xlsx

The presence of Arc proteins in EVs is consistent with published data from both mammalian and Drosophila cell cultures. EVs from media of short-term cultures of mouse cortical neurons were shown to contain denser objects whose size (~30 nm in diameter) was consistent with mammalian Arc capsids, and which were associated with anti-Arc gold particles (Pastuzyn et al., 2018). For dArc1, capsid-like structures that bound to anti-dArc1 gold particles were detected in lysed preparations of EVs from S2 cells (Ashley et al., 2018). The dArc1 capsid is 37 nm in diameter (Erlendsson et al., 2020; Hallin et al., 2021). As described above, EM tomography of EVs from SasFL-expressing S2 cells showed denser objects of 30–40 nm in diameter within many of them (Figure 2d, Figure 2—figure supplement 1), consistent with the idea that these are dArc1 capsids. A video of a 3D reconstruction of the EM tomogram of the EV in Figure 2d is included, and Figure 2—figure supplement 1 shows a still image from this video.

Other proteins observed in the mass spectrometric analysis that were present at higher levels in the IP from V5-SasFL-expressing EVs included small ribonucleoproteins (SmE and SmF), a ribosomal protein (NHP2), and a collagen (Vkg; Figure 4b). Proteins in these categories were found to be major EV components in a proteomic analysis of S2 and Kc167 cell EVs (Koppen et al., 2011). We think it likely that some or all these proteins are abundant contaminants that do not actually interact with Sas but happened to be present at higher levels in the IP from Sas-expressing cells vs. the IP from control cells. We did not further examine any of these proteins.

Since dArc1 was enriched in SasFL preparations purified from EV lysates with anti-V5, we then investigated whether it binds to the Sas ICD (which would be in the EV interior) by co-IP in S2 cells. We coexpressed Myc epitope-tagged dArc1 with a fusion protein in which the V5-tagged ECD of mouse CD8 was attached to the TM domain and the 37 aa ICD of Sas. We then IP’d cell lysates with anti-Myc, and detected V5-mCD8ECD-SasTM-ICD and dArc1-Myc by western blotting. We observed that dArc1 co-IP’d with the Sas ICD fusion protein (Figure 4c).

The Sas ICD sequence contains the sequence motif YDNPSY, which is a PTB-binding motif (NPXY) that overlaps by two amino acids with an SH2-binding motif (YXXP) that is also a potential Abl tyrosine kinase substrate sequence (Colicelli, 2010; Figure 4e). The NPXY motif is the target for binding of the Numb PTB (Li et al., 1998). This suggests that an SH2 protein and a PTB protein might compete for binding to this sequence, if the first tyrosine was phosphorylated to create an SH2 docking site. The PTB domain of Numb does not require tyrosine phosphorylation to bind to its NPXY target. Interestingly, in an earlier mass spectrometric analysis, we found that the Shc protein, which contains a phosphotyrosine-binding SH2 domain, was associated with Sas purified from S2 cells treated with pervanadate to induce high-level tyrosine phosphorylation.

We searched for other Drosophila CSPs containing a sequence with similar properties in their ICDs, and found only one, Appl, which has the sequence YENPTY but is otherwise unrelated to the Sas ICD. Human APP, the mammalian ortholog of Appl, contains the same sequence in its short ICD (Figure 4e), as do the two APP paralogs, APLP1 and APLP2. We then replaced the Sas ICD in the V5-mCD8ECD-SasTM-ICD construct with the Appl and APP ICDs, and found that the Appl ICD protein co-IP’d with dArc1-Myc (Figure 4c), implicating the Y(D/E)NP(S/T)Y sequence in binding to dArc1. This sequence contains the consensus motif for binding of mammalian Arc to TARPγ2, CaMKII, and NMDA receptor peptides, which is X-P-X-(Y/F/H)(Zhang et al., 2015; Nielsen et al., 2019). Arc binds to the NMDA receptor as a monomer (Nielsen et al., 2019). The TARPγ2 Arc-binding peptide is RIPSYR, which is similar to the sequences in Sas (PSYK) and APP (PTYK). Accordingly, we expressed Myc-tagged mammalian Arc (rArcFL) in S2 cells and examined whether it could co-IP with the V5-mCD8-ICD fusion proteins. We observed that Myc-rArcFL was able to co-IP with the Sas and APP ICDs (Figure 4c).

The co-IP data indicate that the Sas, APP, and Appl ICDs associate with dArc1 and Arc, but does not show that the proteins bind directly to each other. To evaluate this, we made the complete Sas, APP, and Appl ICDs (Figure 4e), as well as a scrambled version of the Sas ICD and a deletion mutant of the Sas ICD that lacks the YDNPSY sequence, as biotinylated peptides, and bound these to streptavidin-coupled magnetic beads. To make purified Arc proteins for binding, we expressed dArc1 and mammalian Arc as GST fusion proteins in E. coli. We mixed the beads with purified GST-dArc1 and GST-rArcFL proteins and examined whether we could observe specific binding. As a positive control, we made purified Numb PTB domain, and showed that it bound as expected to the Sas, APP, and Appl peptides, which all contain the NPXY PTB-binding motif, but not to the scrambled Sas peptide or the YDNPSY deletion mutant. In the peptide binding assay, we observed that dArc1 directly bound to the Sas ICD peptide, but not to the other peptides. Mammalian Arc also bound to the Sas ICD peptide, as well as to the APP ICD peptide (Figure 4d).

Mammalian and Drosophila Arc are not orthologs, and are apparently derived from independent Ty3/gypsy retrotransposon lineages (Ashley et al., 2018; Pastuzyn et al., 2018; Hantak et al., 2021). The fact that both proteins mediate intercellular communication suggests that they may be products of convergent evolution. Our results implicate the Y(D/E)NP(S/T)Y sequence as a determinant of binding to Arcs (Figure 4e), and show that fly and mammalian Arcs can bind to the same peptide sequences.

Our data also suggest that APP might be a CSP that has a relationship to Arc which is similar to that of Sas to dArc1. APP also localizes to EVs (Laulagnier et al., 2018; Pérez-González et al., 2020). The connection between APP and Arc will be of interest to explore in future studies, since Arc has been implicated in AD pathogenesis (Wu et al., 2011; Landgren et al., 2012; Bi et al., 2018). The first Y in the YENPTY motif in APP has been reported to be a substrate for the Abl tyrosine kinase (Zambrano et al., 2001). If YENP was phosphorylated, it would become a docking site for a class of SH2 domain proteins, and binding of this protein(s) could occlude Arc binding to the adjacent PTYK sequence. The Abl inhibitor imatinib (Gleevec), which would be expected to block phosphorylation of this site, inhibits formation of β-amyloid peptide (Aβ)(Netzer et al., 2003), and binding of Arc to APP could be relevant to this effect.

Excel file of mass spectrometry data. Contains two tabs: (1) Raw MaxQuant File; (2) Peptide and LFQ values.

Sas can facilitate intercellular transfer of dArc1 and its mRNA in vivo

Sas is not required for loading of dArc1 capsids into EVs, since dArc1 mRNA is a normal component of EVs from cell lines that do not express Sas. If it behaves like mammalian Arc in its interactions with peptides (Nielsen et al., 2019), dArc1 might bind to Sas as a monomer. Perhaps Sas recruits dArc1 monomers (possibly bound to mRNA via their Zn2+ fingers) to nascent EVs during their biogenesis, and they then assemble into capsids. Binding of Sas to dArc1 may help to increase the probability that Sas-bearing EVs contain dArc1 capsids bearing dArc1 mRNA. One function of Sas might then be to deliver the EVs and their dArc1 capsid cargo, including encapsulated RNA, to specific recipient cells that express the Sas binding partner Ptp10D.

Having shown that SasFL can move within larvae and that it binds to dArc1, which is a known component of EVs that mediates intercellular communication, we then examined whether it can cause dArc1 to move from source cells into recipient cells in vivo. To establish an assay system for dArc1 capsid movement in embryos, we first expressed V5-SasFL in late stage 16 embryonic SGs together with RFP, and observed that V5 signal moved to the gut and tracheae, while RFP was retained in the SGs as expected. When V5-Sasshort was expressed in SGs, however, V5 staining was only observed in the SGs (Figure 5—figure supplement 1). This is consistent with the failure of V5-Sasshort to move from SGs to imaginal discs in larvae (Figure 3c).

To examine dArc1 protein transport, we needed to express untagged dArc1 and visualize it with antibody against dArc1 (Ashley et al., 2018), because we were unsuccessful in detecting movement of tagged versions of dArc1. dArc1 is made at very low levels in embryos. In late stage 16 control embryos (Sage-GAL4/+ or Sage-GAL4 >SasFL), we observed faint staining throughout the embryo, with higher levels in the gut (Figure 5—figure supplement 2). We then expressed dArc1 from a UAS construct that contained only the dArc1 open reading frame (ORF), flanked by heterologous 5’ and 3’ UTR sequences. The 3’ UTR was derived from SV40. When we expressed dArc1 alone in SGs, we observed bright anti-dArc1 staining in the SGs and increased staining relative to controls in the gut and in dots in the body wall. When SasFL and dArc1 were expressed together, dArc1 staining in the gut was further increased (Figure 5—figure supplement 2).

To localize dArc1 staining in the body wall and compare it to Ptp10D staining, we examined dissected ‘fillets’ at high magnification. For reference, Figure 5—figure supplement 1 shows the evolution of Ptp10D expression in fillets from stage 14 to late stage 16. VNC expression continuously increases during this time period, while tracheal expression begins in stage 14, decreases in stage 15, and re-emerges at stage 16, at which time Ptp10D is expressed in the main tracheal trunk and major tracheal branches. Figure 5d’ shows that, in late stage 16 embryos expressing both SasFL and dArc1 in SGs, there were many bright puncta stained with anti-dArc1 in the dorsal tracheal trunk, which expresses Ptp10D. These puncta appeared similar to those previously observed at larval NMJs (Ashley et al., 2018). They were not detectable in control embryos (Sage-GAL4/+; Figure 5a’).

Figure 5 with 2 supplements see all
Sas facilitates transfer of dArc1 capsids bearing dArc1 mRNA into distant cells in vivo.

(a-d) Localization of dArc1 protein and Ptp10D in high-magnification views of body walls from fillets of late stage 16 embryos (anterior to the left, dorsal up). (a) Control (Sage-GAL4/+); (b) Sage >dArc1; (c), Sage >SasFL; (d), Sage >SasFL + dArc1. (a–d) show double-staining with anti-dArc1 (red) and anti-Ptp10D (blue). (a’-d’) show the dArc1 channel alone. (a’’-d’’) show the Ptp10D channel alone. Arrows, dorsal tracheal trunk. There are numerous bright dArc1 puncta in the tracheal trunk when SasFL and dArc1 are expressed together. Fewer and weaker puncta are observed when SasFL or dArc1 are expressed alone, and no puncta are seen in Sage-GAL4/+ controls. (e–i) dArc1 mRNA from the endogenous gene (+strand), detected by FISH with a 3’ UTR minus-strand (antisense) probe. (e) Control (Sage-GAL4/+); (f) Sage >dArc1; (g) Sage >SasFL; (h) Sage >SasFL + dArc1. There is weak expression of dArc1 mRNA in the SGs in controls. When dArc1 (from an ORF construct) is expressed alone, bright SG staining is observed, indicating that exogenous dArc1 increases expression of endogenous dArc1 mRNA. There are also scattered dArc1 mRNA puncta elsewhere in the embryo. When SasFL and dArc1 are expressed together, bright dArc1 mRNA FISH staining of the entire tracheal system is observed (arrows indicate dorsal tracheal trunks), as well as the foregut (arrowhead) and esophagus (n>100 embryos examined for each genotype; representative results are shown). i, i’, high-magnification views of dArc1 mRNA in the tracheae in an obliquely mounted (anterior to the left, dorsal up) embryo expressing SasFL and dArc1 in SGs. i’ is a higher-magnification inset (yellow dotted outline) from (i). Arrow in (i), SG; arrowhead, foregut loop. Arrow in i’, dorsal tracheal trunk; arrowhead, transverse connective. Scale bar in (e) (applies to e–h), 50 µm; scale bar in (a) (applies to a–d), 10 µm; scale bar in (i), 50 µm; scale bar in (i’), 50 µm.

There were lower numbers of fainter dArc1 puncta in tracheal trunks of the two other genotypes (Sage >dArc1 and Sage >SasFL)(Figure 5b’ and c’). Endogenous Sas is expressed at low levels in SGs, and endogenous dArc1 mRNA is also present in SGs (Figure 5g), although dArc1 protein is not detectable. Endogenous SasFL may be able to transport some of the overexpressed dArc1, and overexpressed SasFL might transport some endogenous dArc1, giving rise to the observed puncta. It is also interesting that dArc1 (and dArc1 mRNA; see below) is observed in tracheal cells, but not in VNC neurons, which also express Ptp10D at high levels. There is a glial sheath around the VNC at late stage 16, and this might block access of EVs to Ptp10D-expressing neurons. Alternatively, perhaps there are cofactors required for EV binding and/or internalization that are not expressed in neurons.

Dramatic effects of SasFL on dArc1 expression and localization were observed when endogenous dArc1 mRNA was examined by fluorescence in situ hybridization (FISH) in embryos expressing the UAS-dArc1 ORF construct in SGs. To detect mRNA, we used the 700 nt antisense 3’ UTR probe employed in the (Ashley et al., 2018) paper to visualize dArc1 mRNA puncta at the NMJ. Note that this probe does not recognize overexpressed dArc1 mRNA made from the UAS construct, because that contains only the dArc1 ORF and no dArc1 3’ UTR sequences. In late stage 16 control embryos (Sage-GAL4/+), we observed faint FISH signals in the SGs and a few puncta elsewhere in the embryo (Figure 5e). A similar pattern was seen in Sage >SasFL embryos (Figure 5g). When dArc1 was expressed from the UAS-dArc1 ORF construct, we observed bright FISH signals in SGs with the endogenous dArc1 3’ UTR probe (Figure 5f). There were also scattered puncta in other parts of the embryos. This shows that exogenous dArc1 induces expression of endogenous dArc1 mRNA (or stabilizes the mRNA). No signal was observed when a sense dArc1 3’ UTR probe was used for FISH (Figure 5—figure supplement 2). Finally, when SasFL and dArc1 were expressed together, we observed a completely different pattern, in which the entire tracheal system is lit up by the FISH signal for the endogenous dArc1 3’ UTR (Figure 5h). The foregut and esophagus also stain brightly. By contrast, in embryos expressing Sasshort and dArc1 in SGs, the dArc1 FISH signal is observed in the SGs, with only a few puncta elsewhere in the embryo (Figure 5—figure supplement 2), consistent with the fact that Sasshort cannot move within the embryo (Figure 5—figure supplement 1).

When an antisense probe for the SV40 3’ UTR sequence in the UAS-dArc1 construct was used for FISH, we observed signal only within the SGs, even when SasFL was coexpressed (Figure 5—figure supplement 2). This suggests that dArc1 mRNA lacking the endogenous dArc1 3’ UTR sequences is not efficiently loaded into dArc1 capsids that can move elsewhere in the embryo. This is consistent with the findings of Ashley et al., 2018, who showed that the dArc1 3’ UTR is required for dArc1 mRNA transfer at the NMJ. No signal was observed when a sense SV40 3’ UTR probe was used for FISH (Figure 5—figure supplement 2).

Figure 5i and i’ show the tracheae and SGs at higher magnification in side views of an embryo expressing both SasFL and dArc1 in SGs. The dorsal tracheal trunk (arrow) and the transverse connective (arrowhead) both display bright dArc1 FISH signals. Note that, because this is a confocal image (optical section), the cells at the edges of the tracheal trunk are bright, while the hollow lumen is dark. The brightness of the tracheal FISH signal suggests that it represents not only dArc1 mRNA transferred from capsids, but dArc1 mRNA synthesized in these cells in response to dArc1 protein made from the dArc1 mRNA transported in the capsid. If this is correct, it would represent an amplification mechanism in which translated dArc1 mRNA from EVs can induce expression of much more dArc1 mRNA in the recipient cells. Finally, we examined whether the Sas ICD is required for dArc1 mRNA transport by expressing dArc1 together with a protein (SasECD-TM-GFP) in which the Sas ICD was replaced by GFP. This protein is present in EVs when expressed in S2 cells, but it does not produce any dArc1 FISH signal outside of the SGs (Figure 5—figure supplement 2), indicating that it cannot facilitate transport of dArc1 capsids to tracheal cells.

Conclusions

Our results on movement of Sas EVs containing dArc1 capsids are summarized in the diagram of Figure 6. These findings contribute to the understanding of intercellular communication mechanisms by showing that SasFL is an EV targeting ligand that can direct internalization of EVs into cells expressing the Sas receptor Ptp10D. Endogenous SasFL is likely to move between cells in vivo, because tagged SasFL moves away from both neuronal and non-neuronal cells when ectopically expressed.

Schematic diagram of the processes involved in movement of EVs bearing SasFL and dArc1 capsids from salivary glands to tracheal cells.

Steps 1 and 2, the presence of the dArc1 protein made from a UAS-dArc1 ORF construct increases expression of endogenous dArc1 mRNA in embryonic SGs. EVs with SasFL on their surfaces bearing dArc1 capsids containing endogenous dArc1 mRNA diffuse or are transported through the hemolymph (Step 3) and bind to Ptp10D-expressing tracheal cells (Step 4). The EVs internalize into the tracheal cells and release dArc1 mRNA (Step 5), and dArc1 protein translated from that mRNA induces expression of more endogenous dArc1 mRNA (Steps 5 and 6). Nuclei, orange circles in 1–2 and blue circles in 4–6.

dArc1 is related to retrotransposon Gag proteins, and it forms a capsid that contains dArc1 mRNA and is loaded into EVs (Ashley et al., 2018). Overexpressed SasFL can facilitate transfer of dArc1 capsids into distant Ptp10D-expressing recipient cells in vivo. Whether endogenous SasFL also transfers dArc1 between cells cannot be determined from our data, and would be difficult to evaluate, since both the sas and dArc1 genes are broadly expressed. The Sas ICD binds directly to dArc1.

Mammalian Arc also forms capsids that are transported via EVs (Pastuzyn et al., 2018), and it binds to the Sas and APP ICDs, which share a tyrosine motif. Full-length APP and some of its proteolytic products are localized to EVs, and EVs from N2a cells bearing tagged APP are internalized into cultured neurons, but not into glia (Laulagnier et al., 2018). It will be interesting to determine if APP EVs contain Arc capsids, and if the presence of APP on Arc-containing EVs causes Arc to be preferentially delivered to a specific population of neurons. This could have implications for Alzheimer’s disease research, because APP is the source of β-amyloid peptide and Arc has been linked to β-amyloid accumulation and AD pathogenesis (Wu et al., 2011; Landgren et al., 2012; Bi et al., 2018).

Methods

Fly stocks and genetics

The following stocks were used: yw for wild-type control, ap-GAL4 (Bloomington 50156), UAS-mCD8::GFP (Bloomington 5130), UAS-myr::mRFP (Bloomington 7118), UAS-mCherry.NLS (Bloomington 38424), sas15 (null mutant)(Bloomington 2098), Sage-GAL4 (a gift from Deborah J. Andrew), Ptp10DEP1172 (Bloomington 11332), UAS-dArc1 (Bloomington 37532), UAS-Numb (a gift from Yuh Nung Jan), UAS-SasFL and UAS-V5-SasFL (Lee et al., 2013), Arc1esm18 (Bloomington 37530). Crosses and embryo collections were performed at room temperature. For overexpression experiments, embryos were shifted to 29 °C for at least 120 min prior to fixation and staining and 3rd instar larvae were shifted to 29 °C for overnight for further analysis. For the EV targeting experiments, imaginal discs from 3rd instar larvae were harvested at room temperature and incubated in 200 µl of S2 supernatant overnight at 29 °C before fixation and staining. There are 10,000–50,000 cells in a 3rd instar imaginal disc. Given the results from the NTA analysis, we can conclude that ~140,000 EVs are present in 200 µl of supernatant from V5-SasFL-expressing S2 cells. We used 5 wing discs per incubation, so the ratio of EVs to cells is ~0.5 to~2. The relative V5 signal intensities on the imaginal discs were measured by densitometry analysis using ImageJ software.

Immunohistochemistry

Embryos and larval tissues were stained with standard immunohistochemical procedures. The following antibodies were used: rabbit anti-V5 (1:1000, Invitrogen); mouse anti-GFP (1:1000, Invitrogen); rabbit-anti-SasFL (1:2000, gift of D. Cavener); rat-anti-Sasshort (1:50, GenScript USA Inc); mAb Cq4 against crumbs (1:100, DSHB); guinea pig-anti-Numb (1:1000, gift from J. Skeath); rabbit-anti-dArc1 (1:100, gift from T. Thomson); mAb 8B2 against Ptp10D (1:5, DSHB); mAb MR1A against Prospero (1:40, DSHB); rat-anti-Repo (1/2000, gift from S. Banerjee); rabbit anti-Evi (Wntless, 1:5000, gift from K. Basler); FITC-conjugated phalloidin (1:1000, Thermo Fisher Scientific); AlexaFluor 488 anti-mouse, AlexaFluor 488 anti-rat, AlexaFluor 568 anti-rabbit, AlexaFluor 568 anti-rat and AlexaFluor 647 anti-mouse (1:1000, Invitrogen). Rat anti-Sasshort antibody was generated against a synthetic peptide, HSSIPANGANNLQP, flanking the EVT region (intron is between the N and G residues) and the KLH-conjugated antibody was purified by protein G column (GenScript USA Inc). Samples were mounted in VECTASHIELD (Vector Laboratories) and analyzed on a Zeiss LSM 880.

Cell culture and preparation of EVs and cell lysates

EVs and cell lysates were prepared from S2 cells (ATCC, CRL-1963) that were cultured for four days at 22℃ in Schneider’s medium (Gibco) supplemented with 10% exosome-free FBS (#EXO-FBSHI-50A-1, SBI) to avoid contamination from Bovine serum exosomes. S2 cells were authenticated and confirmed to be free of mycoplasma by ATCC. DNA constructs were transiently transfected into S2 cells using Effectene (Qiagen). EVs for western blot analysis and electron microscopy were collected using Total Exosome Isolation reagent (#4478359, Invitrogen) from the supernatants of S2 cultures. This kit has been found to produce exosomes of equivalent quality from mammalian cells (with respect to the presence of exosome markers and the depletion of non-exosome proteins) to those generated using ultracentrifugation (Skottvoll et al., 2019). One part of the reagent and two parts of supernatant were mixed and incubated at 4℃ overnight. Pellets of EVs were collected after centrifugation at 10,000 x g for 60 min at 4℃. The EV pellets were resuspended in PBS for western blot analysis.

To prepare EVs using the ultracentrifugation protocol of Théry et al., 2006, S2 cells were cultured in a medium with 10% exosome-free FBS (SBI) and the cultures were grown in 100 mm Petri dishes at 22℃. Cells were collected at 1–1.5×106 cells/ml. The supernatant from a 50 ml culture was filtered using a 0.22 µm filter to eliminate dead cells and large debris prior to further purification by ultracentrifugation (Optima MAX-XP, Beckman Coulter). The filtered supernatants were spun at 100,000 x g for 70 min and the collected EV pellets were resuspended in PBS. Then, the samples were spun again at 100,000 x g for 70 min. The final EV pellet was resuspended in 60 µL of PBS and processed for further analyses or stored at –80℃.

For the EV targeting experiments between S2 cells, supernatants from transiently transfected donor cells were collected and filtered using 0.22 µm PVDF membranes before resuspension and incubation with the recipient cells. Two days before the supernatant swap between EV donor and recipient cell cultures, the recipient cells were transiently transfected with DNA constructs. The recipient cells were incubated in the supernatants with EVs from donor cells for 2 hr at 22℃. For western blot analyses, cell lysates were prepared using RIPA cell lysis buffer. To measure the size and number of EV particles from S2 cell culture, collected EV pellets were subjected to NTA by System Biosciences, LLC (Palo Alto, CA, USA) using a NanoSight instrument. The NTA measurements rely on light scattering to extract particle size and the number of particles in a sample and the NTA software (Version 2.3) collects data on multiple particles to calculate the hydrodynamic diameter of each particle using the Stokes-Einstein equation (System Biosciences, LLC).

Mass spectrometry analysis

Samples were lyophilized and proteins were trypsin-digested as previously described (Pierce et al., 2013). A total of 200 ng of digested peptides were analyzed as previously described (Sung et al., 2016). Briefly, peptides were loaded onto a 26 cm analytical HPLC column (75 µm inner diameter) packed with ReproSil-Pur C18AQ 1.9 µm resin (120 Å pore size; Dr. Maisch, Ammerbuch, Germany). Peptides were separated with a 120 min gradient at a flow rate of 350 nl/min at 50 °C (column heater) using the following gradient: 2–6% solvent B (7.5 min), 6–25% B (82.5 min), 25–40% B (30 min), 40–100% B (1 min), and 100% B (9 min), where solvent A was 97.8% H2O, 2% ACN, and 0.2% formic acid, and solvent B was 19.8% H2O, 80% ACN, and 0.2% formic acid. Samples were analyzed using an EASY-nLC 1000 coupled to an Orbitrap Fusion operated in data-dependent acquisition mode to automatically switch between a full scan (m/z=350–1500) in the Orbitrap at 120,000 resolving power and an MS/MS scan of higher-energy collisional dissociation fragmentation detected in the ion trap (using TopSpeed). The automatic gain control (AGC) targets of the Orbitrap and ion trap were 400,000 and 10,000.

Mass spectrometry data

Raw data were searched using MaxQuant (version 1.5.3.30)(Cox and Mann, 2008; Wagner et al., 2011) against the Uniprot D melanogaster database. Fragment ion tolerance was 0.5 Da. Precursor mass tolerance was 4.5 ppm after automatic recalibration. Searches were permitted up to two missed tryptic peptide cleavages. Cysteine carbamidomethylation was designated as a fixed modification while Methionine oxidation and N-terminal acetylation were designated as variable modifications. False discovery rates were estimated to be <1% using a target-decoy approach. Complete data are in Supplementary file 1.

Protein expression and purification

To express and purify Arc proteins in the E. coli system, the cDNAs of dArc1, dArc2 and rArc were subcloned into the pGEX-4T-1 vectors together with GST-6xHis-tags and TEV protease cleavage site. Arc proteins were expressed in E. coli strain BL21 (DE3) grown in LB broth by induction of log-phase cultures with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and incubated overnight at 23 °C. Cells were pelleted and resuspended in B-PER lysis buffer (#78243, Thermo Fisher Scientific) before centrifugation to collect cell lysates.

Tagged Arc proteins were pulled down using Ni-NTA resin column and the eluates with GST-6xHis-dArc1 and GST-6xHis–rArc proteins were used for peptide binding assays.

Western blotting

Proteins were separated by SDS–PAGE, transferred at 200 mA for 60 min to nitrocellulose membranes using a Bio-Rad Wet Tank Blotting System in Tris-Glycine Transfer Buffer with 10% methanol. Blocked membranes were incubated with primary antibodies in 0.5% milk PBS-0.1% Tween for overnight. HRP-conjugated antibodies (anti-V5-HRP (#RV5-45P-Z, ICL), anti-mouse IgG HRP (#sc-516102, Santa Cruz Biotechnology), anti-beta-actin-HRP (#HRP-60008, Proteintech), anti-rabbit IgG HRP (#65–6120, Invitrogen), anti-rat IgG HRP (#35470, Invitrogen), anti-alpha-tubulin-HRP (#HRP-66031, Proteintech), anti-cMyc-HRP (#RMYC-45P-Z, ICL), and anti-GST-HRP (#MA4-004-HRP, Invitrogen)) were used at 1:10,000 for 60 min. Blots were developed using ECL Western Blotting Substrate (#32109, Pierce), and imaged on a MINI-MED 90 X-Ray Film Processor (AFP Manufacturing Co.).

Electron tomography and immuno-EM

For imaging of EVs by electron tomography (ET), EVs were prepared using the Exosome Isolation Kit as described above. Supernatant was removed and replaced with ~10 ml 10% Ficoll, 5% sucrose in 0.1 M sodium cacodylate trihydrate with minimal disturbance of the pellet. Pellets were transferred to brass planchettes (type A/B; Ted Pella, Inc) and ultra-rapidly frozen with a HPM-010 high-pressure freezing machine (Bal-Tec/ABRA). Vitrified samples were transferred under liquid nitrogen to cryo-tubes (nunc) containing a frozen solution of 2.5% osmium tetroxide, 0.05% uranyl acetate in acetone and placed in an AFS-2 Freeze-Substitution Machine (Leica Microsystems, Vienna). Samples were freeze-substituted at –90 °C for 72 hr, warmed to –20 °C over 12 hr, held at –20° for 12 hr, then warmed to room temperature. Samples were rinsed 3 x with acetone and infiltrated into Epon-Araldite resin (Electron Microscopy Sciences). Resin was polymerized at 60 °C for 24 hr.

Serial semi-thin (170 nm) sections were cut with a UC6 ultramicrotome (Leica Microsystems) using a diamond knife (Diatome Ltd., Switzerland). Sections were collected onto Formar-coated copper/rhodium slot grids (Electron Microscopy Sciences) and stained with 3% uranyl acetate and lead citrate. Colloidal gold particles (10 nm) were placed on both surfaces of the grid to serve as fiducial markers for subsequent image alignment. Grids were placed in a dual-axis tomography holder (Model 2040; Fischione Instruments, Inc) and imaged with a Tecnai T12 transmission electron microscope (Thermo-Fisher Scientific) at 120 k eV. For dual-axis tomography, grids were tilted +/-62° and images acquired at 1° intervals. The grid was rotated 90° and a similar tilt-series was recorded about the orthogonal axis. Tilt-series data was acquired automatically using the SerialEM software package. Tomographic data was calculated, analyzed and modeled on iMac Pro and M1 computers (Apple, Inc) using the IMOD software package.

For immuno-EM, EV pellets were prepared as per above. Supernatant was removed and pellets fixed with 4% paraformaldehyde in PBS for 1 hr. Pellets were then infiltrated with 2.1 M sucrose in PBS over 24 hr, with >3 changes of the infiltration solution during that time. Pellets were placed onto aluminum sectioning stubs, drained of excess liquid and frozen in liquid nitrogen. Cryosections (100 nm) were cut at –140 °C with a UC6/FC6 cryoultramicrotome (Leica Microsystems) using cryo-diamond knives (Diatome Ltd). Cryosections were collected with a wire loop containing 2.3 M sucrose in PBS and transferred to Formvar-coated, carbon-coated, glow-discharged 100-mesh copper/rhodium grids (Electron Microscopy Sciences) at room temperature. Nonspecific antibody binding sites were blocked by incubating the grids with 10% calf serum in PBS for 30’. Sections were then labeled with 1° antibodies (diluted in 5% calf serum/PBS) for 2 hr, rinsed 4 x with PBS, then labeled with 10 nm and/or 15 nm gold-conjugated 2° antibodies (diluted in 5% calf serum/PBS) for 2 hr. Grids were rinsed 4 x with PBS, 3 x with dH2O then simultaneously negatively stained and stabilized with 1% uranyl acetate, 1% methylcellulose in dH2O. Immuno-EM samples were imaged as per the tomography samples, above.

Immunoprecipitation

For the Myc-co-IP assay, transiently transfected S2 cells using Effectene (Qiagen) were cultured in Schneider’s medium at 22  °C for 4 days. Tagged expression constructs (V5-mCD8ECD-sasTM-ICD, V5-mCD8ECD-APPICD, V5-mCD8ECD-ApplICD, dArc1-Myc and Myc-rArc) were cloned in pAc5.1B vector according to standard cloning procedure. For IP analysis, cell lysates were prepared using IP Lysis buffer (#87787, Pierce) and the lysates were incubated in Myc-Trap agarose (#yta-20, Chromotek) following the manufacturer’s protocol and the eluates were analyzed by standard western blot analysis.

Peptide binding assay

For the peptide binding assay, biotinylated peptides (wt SasICD, SasICD variations (scrambled and ΔYDNPSY), APPICD and ApplICD) made by RS Synthesis, Inc, were incubated with Streptavidin magnetic beads (#88817, Pierce) for 45 min at 4 °C and the beads were extensively washed with TBST. Purified GST-6xHis-dArc1 and rArc proteins were added to the beads with bound biotinylated peptides and incubated at 4 °C overnight. Similar experiments were performed with Numb PTB domain protein purified from E. coli. The beads were carefully washed with TBST and eluates prepared for western blot analysis following the standard protocol described above.

Fluorescent in situ hybridization (FISH)

The FISH protocol was a modification of protocols from Kosman et al., 2004. Fixed L16 whole embryos were prepared using standard protocols and rinsed with ethanol quickly four times. Then the embryos were permeabilized twice with a mixture of xylenes and ethanol (1:2, v/v) and washed three times with ethanol for 5 min each. To rehydrate the embryos, the embryos were washed with 100%, 50% and 0% methanol in PBT sequentially for 30 min each step. The rehydrated embryos were permeabilized again using proteinase K (20 µg/mL in PBT) for exactly 7 min and washed three times for 5 min each in PBT followed by a second fixation (5% paraformaldyhyde and 1% DMSO in PBT) for 25 min and washed three times in PBT for 5 min each. Then the embryos were prepared for pre-hybridization by incubation in 50% hybridization buffer (50% formamide, 5 x SSC, 100 μg/ml fragmented salmon testes DNA, 50 μg/ml heparin, 0.1% Tween-20) in PBT for 5 min. For pre-hybridization, embryos were incubated in hybridization buffer for more than 90 min at 55℃ while changing the buffer every 30 min. The pre-hybridized embryos were incubated in DIG-tagged dArc1 mRNA probe for 18 hr at 55℃ for annealing. The embryos were washed with hybridization buffer three times for 30 min each at 55℃, after which the buffer was replaced with replaced the buffer with PBT containing rhodamine-conjugated sheep anti-DIG antibody (#11207750910, SigmaAldrich) overnight at 4℃. Then the embryos were washed and mounted for confocal microscopy.

Probe preparation

Probes for detection of endogenous dArc1 mRNA were designed against a 760 nt region of the dArc1 mRNA 3’ UTR sequence, which was used for FISH in a previous study (Ashley et al., 2018). Probes for detetion of dArc1 mRNA from the UAS-dArc1 ORF construct were designed against a 616 nt region of SV40 3’ UTR sequence. To generate antisense and sense probes for dArc1 mRNA, cDNA sequences from dArc1 were PCR amplified and purified to use as positive and negative probe templates. The same procedure was used for antisense and sense probes for the SV40 3’ UTR. The DNA templates were heated to 55℃ for 2 min and then put back on ice. Transcription reactions were set up to label probes with digoxigenin (DIG, # 11277073910, Roche) and incubated at 37℃ for 2 hr. Probes were precipitated and resuspended in hybridization buffer and stored at –20℃.

The following primers were used to generate endogenous dArc1 mRNA probes:

  • dArc1 probe forward primer: GATTTTTCGTCTGATCCTGGTC

  • dArc1 probe reverse primer: CCGTTTCTGAGTTTAATGGTTG

These primers were used to generate SV40 3’ UTR mRNA probes:

  • SV40 3’ UTR probe forward primer: TCTAGAGGATCTTTGTGA

  • SV40 3’ UTR probe reverse primer: TGCTATTGCTTTATTTGT

Data availability

Key data generated or analysed during this study are included in the manuscript and supporting files.

References

Decision letter

  1. K VijayRaghavan
    Senior and Reviewing Editor; National Centre for Biological Sciences, Tata Institute of Fundamental Research, India
  2. Jason D Shepherd
    Reviewer; University of Utah, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "An extracellular vesicle targeting ligand that binds to Arc proteins and facilitates Arc transport in vivo" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and K VijayRaghavan as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Jason D Shepherd (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Extracellular vesicles (EVs) are emerging as important mediators of cell-to-cell signaling. In this paper the authors aim to demonstrate that Stranded at second (Sas), a Drosophila cell surface protein, binds to dArc1 and Ptp10D to mediate intercellular transport of dArc1 via EVs. dArc1 protein has been shown to form virus-like capsids that carry dArc1 mRNA from neurons to muscle, but little is known about this new intercellular communication pathway. Similarly, not much is known generally about how EVs are targeted to specific cell types, or how specific EV cargo can be delivered. Thus, this work is of interest to cell biologists and neuroscientists. However, the jumbled description of the results and lack of rigor in some experiments diminish the impact and interpretability of the conclusions. Moreover, almost all experiments rely on gain-of-function and over-expression of Sas, thus the relevance to normal physiological signaling is unclear. There are other concerns detailed below that need to be addressed before acceptance. The authors are urged to address concerns raised that can be dealt with using Drosophila. The manuscript should also be toned to make it more 'Drosophila-centric' as generalizations will require additional experiments ( such as with mammalian cell culture). While re-writing, a substantial reduction in the introduction is also suggested. In particular, there is no need to summarise the results in this section in such detail.

We detail below a composite summary of the major points raised by the reviewers for your attention,

The authors' EV isolation protocol is generally not a preferred procedure because it precipitates proteinaceous non-vesicular material, resulting in a mixed preparation of vesicles and protein aggregates. While the authors point to Skotvoll et al., 2019, this paper compares only one approach (ultracentrifugation) to the EV isolation kit. While there is an overlap in the protein components of these approaches in their report, there are considerable differences in the identified proteins. The consensus in the field is that these kits fall under what MISEV2018 Guidelines describe as "high recovery, low specificity" methods, and very likely, these samples have a greater incidence of non-vesicular materials. In this case, the authors' samples are more "secretome-like" and this is observed in their EM of S2 EVs (Figure S2 a,b,d). Using a more robust and specific EV isolation approach, such as a combination of filtration or ultracentrifugation, followed by density gradient fractionation would absolutely improve the quality of their manuscript by providing more convincing evidence that the proposed cargoes are indeed trafficked within EVs. The authors should provide further experimental validation of Sas's EV enrichment, isolating EVs from cultured cells using a more accurate approach.

The description of the results omits some data in the figures and is not in a logical order. This makes the manuscript hard to read and follow. There is also a lack of rigor and quantification in some experiments. Specifically:

1. Figure 1

Lack of quantification: 1 image is just not enough.

Calculate the ratio lysate/EV, also with EVi as a marker for EVs. Only the lysate is shown. EVs should also!

The authors claim (line 153) that most of FL Sas are in EVs? ` It looks like 60%, but it depends on the loading. Please quantify.

Further:

In Figure 1b, the authors suggest that full-length Sas is a predominant EV component, but not Sas-short. The figure would be strengthened by including data describing the in vivo distribution of the V5-tagged short Sas isoform in conjunction with mCD8-GFP. One would expect the Sas-short isoform to be less broadly distributed compared to Sas-FL.

In Supplemental Figure 1, the full blots should be shown for panel A. Why are the control β-Tub bands shown side-by-side, but not the bands for Sas-FL and Sas-short? Also, the authors could include protein samples from wild-type and Sas-15 embryos to enhance the validation of their antibody reagents.

2. Figure 2: Nice IEM and tomography, but many details are needed. Could more IEM profiles (more examples) be provided to be sure of `BG and specificity (in a gallery). Are the capsid reconstitution with purified dARC1 and 2 performed in the presence of darc1 rRNA? Any RNA (figure 2)? Description of dArc1 putative capsids is absent from the Results section (2f,g) until describing Figure 4 data (line 362). Given that there is no immuno-EM labeling of dArc1 protein, it is not clear if Sas and dArc1 are localized to the same EVs. Nor is it clear if the double membrane EVs are actually EVs that contain capsids. Overall, the EM data lacks quantification. How many EVs on average show Sas labeling? How many EVs have double membranes? The immuno-EM data would be strengthened by adding quantifications in the form of distance measurements of gold particles of Anti-V5 and Anti-Sas in relation to EV membranes. The dense protein staining surrounding EVs seems unusual, is this due to an artifact of the purification? EV kits are generally non-specific and isolate non-EV membranes, Corroboration using ultracentrifugation or size exclusion chromatography methods would be beneficial. SAS-FL overexpression results in more EVs, which confounds subsequent experiments suggesting that Sas targets EVs to specific cell types/regions.

3. Figure 3: There are no data showing the expression of Sas in SG cells using the GAL4 lines. Is this expression restricted to just SG cells? (The results jump from a-b to f-g. c-e are out of order. It will be easier if this is ordered). The quantification in g should be broken into two and paired with the actual data (c-e, and f). It is not clear how the quantification in g was performed. How many western blots (WBs) were analyzed? There seems to be a bubble in the first lane of f, which would preclude quantification. Why is d not quantified, and there seems to be an overall increase in background staining in e that is not specific to discs? The source data files are not labeled and these data should be incorporated into annotated supplemental figures. Is transfer in a-b due to Ptp10D? How many WBs were quantified in g? What is the specificity for full-length Sas? The expression of short Sas should not lead to its incorporation in EVs and their overnight addition should not lead to the same effect (Figure 3). This should be better investigated as short Sas is a good control for full-length Sas.

In the salivary gland (SG) expression data presented in Figure 3a-b, the authors should have included V5-Sas-short as a negative control for cross-tissue transport. Similarly, in Figure 3c-e, EVs isolated from Sas-Short transfected cells should have been included as a control in the overnight EV-imaginal disc incubation assays. Does over-expression of Numb on its own lead to increased uptake of Sas-FL EVs? Also, does the need for Ptp10D and Numb over-expression for efficient capture of Sas-FL EVs in Figure 3c-e raise doubts about the data presented in Figure 3a-b, in which the transfer from SGs seems to be efficient without the need for Ptp10D or Numb over-expression?

Also

a: Are wing discs the only recipient tissues upon S2 cells EVs exposure? The tissues in Figure 5 are the trachea and guts. Why switch?

b: What about Ptp10D mutant disc (as in section 2)?

Better quantification? Number of discs etc.

4. Figure 4: C and d, IP data has no inputs for IPs, no sizing markers, and no IgG controls for antibody specificity. These data would also be more convincing if done with FL Sas and included co-Ips from cell lysates.

The IPs presented in Figure 4C with myc-tagged constructs are not controlled properly. Western blots of protein extracts prior to IPs should be shown and a control antibody should be used to show equal loading in the WB for either dArc1-Myc or Myc-rArc-FL expressing cells. Showing the IPs one on top of each other is very confusing and WBs are critical to showing that the expected proteins are expressed in the transfected cells, even if their IP signature is negative.

5. In general, the WBs in the figures show very white backgrounds with high contrast. Please check if any contrast enhancement (or other such changes) were made and if so, they need to be documented and justified. Else, please revert to raw images of blots. Total protein controls are also missing.

6. Figure 5: Ashley et al. (Cell 2018) showed that dArc1 mRNA transfer required the 3'UTR so it is puzzling that the authors used heterologous UTRs. The results using FISH on endogenous dArc1 mRNA are dramatic. The authors should show definitively that their probe does not pick up over-expressed dArc1. In general, a better quantitative analysis should be provided. For instance, there is no quantitative data for Figure 5, but this point should be examined in other figures as well. The dAC1 increased expression in the target cells upon dARC1 increased production in SG(Figure 5) becomes an important part of the paper (and the model) but is not investigated! How does it work? Does the delivery of darc1 mRNAs packaged in capsids simply lead to more dARC1 translation? Is it proportional?

Or, is there also stimulation of darc1 transcription? Is there also an increase in the mRNA level (we cannot see the SG control of 5o (sage>+) supporting the authors' claim on line 562!).

Also:

a: What is the dARC1 LOF mutant phenotype in SG and trachea?

b: Do EV produced upon FL Sas v5 expression in an SG not produce dARC1 in the recipient trachea?

c: Does this also depend on the same synergetic presence of Ptp10D and Numb in the trachea? Ptp10D LOF?

The authors' data suggest that dArc1 protein and mRNA are transferred by Sas-FL expressing EVs to Ptp10D enriched tissues in fly embryos. Could the authors conduct immuno-EM or nano-flow cytometry studies to assess the general proportion of Sas and dArc1 co-expression in purified EVs? If the general model is accurate, one would expect a higher degree of co-occurrence compared to control EV cargos. Additionally, evaluating the dArc1 protein and mRNA transfer properties in different genetic backgrounds (i.e. Sas-FL and Sas-Short in Ptp10D over-expressed or in Ptp10D null embryos) would significantly enhance the paper.

Also in Figure 5, the results suggesting that transgenic over-expression of dArc1 leads to a strongly increased expression of endogenous dArc1 should be validated using a complementary strategy (e.g. RT-qPCR on dissected tissue specimens, RNA-FISH with probes specific to the endogenous mRNA).

- For all the tissue imaging studies, including those in Figures 5-6,, the authors should provide an indication of how many specimens were analyzed in total, presuming the samples shown in the figures were representative of observations made across multiple tissue specimens co-labeled in the same experiment.

7. Figure 6: what is the red circle (and blue)?

What is the black circle, an SG cell?

8. Most (all) experiments are performed with over-expression of FL Sas or ICD. Does endogenous Sas bind endogenous Ptp10D and dARC1? ICDs? Also full-length APP? Many of the conclusions would be strengthened by the loss of function experiments, especially showing a requirement for Sas in dArc1 transfer.

9. Use of more extensive fly genetics using specific Ptp10D LOF in wing discs and trachea (to show the converse of the GOF). Does Ptp10D acts as the MAIN receptor to FL Sas? Numb LOF, a combination of LOF and GOF? Does Ptp10D GOF compensate for Numb and vice versa?

– The introduction section would benefit from some sentence restructuring, synthesizing major ideas into a more coherent train of thought. For instance, very short, matter-of-fact statements such as "EV cargoes are biomarkers for specific diseases" are not helpful, especially since no specific examples of diseases are provided. Also, more primary sources of literature should be cited, as many statements are lacking supporting evidence, particularly those describing EV biology and characteristics.

– While the authors use "comparatively" to describe the current knowledge of EV-biogenesis pathways as "well understood," this is not an appropriate description and can be misleading. EV-biogenesis pathways are only beginning to be unraveled.

– When describing their major finding (Lee et al., 2013) that Ptp10D is a binding partner for Sas, The authors should stress that this was done in a previous study. Simply adding a world like "previously" will make this clear.

– The paragraph from lines 85-90 stating that 'Sas localizes to EVs, as demonstrated…' appears to be out of context and is more appropriate for the 'Results' section.

– The authors should mention what device they used for NTA, not just provide the name of the outsourcing company. From their description, it appears they used NanoSight, but don't provide acquisition parameters.

– In Figure 2 f,g the authors provide EM images of capsid-like structures for dArc1 and dArc2, however, they do not address these until much later in the manuscript, when they describe how these capsids were generated. The authors should consider reorganizing their figures to improve the flow of their narration and not group them into categories, as this makes the manuscript more difficult to follow.

– There may be referencing errors in the manuscript when mentioning figures, for instance when they describe the detection of V5-mCD8ECD-SasTM-ICD by western blotting, they reference Figure 4d, but it appears they actually mean Figure 4c. Another such error appears in the text describing Figure 5n, which depicts Drosophila embryos co-expressing SasFL and dArc1, and which authors describe as Figure 5m. I recommend that they carefully double-check the manuscript for any such errors, which are confusing for the reader.

– When appropriate the authors should add statistical analysis to their quantitative data, for example, Figure 3g.

https://doi.org/10.7554/eLife.82874.sa1

Author response

Essential revisions:

The authors' EV isolation protocol is generally not a preferred procedure because it precipitates proteinaceous non-vesicular material, resulting in a mixed preparation of vesicles and protein aggregates. While the authors point to Skotvoll et al., 2019, this paper compares only one approach (ultracentrifugation) to the EV isolation kit. While there is an overlap in the protein components of these approaches in their report, there are considerable differences in the identified proteins. The consensus in the field is that these kits fall under what MISEV2018 Guidelines describe as "high recovery, low specificity" methods, and very likely, these samples have a greater incidence of non-vesicular materials. In this case, the authors' samples are more "secretome-like" and this is observed in their EM of S2 EVs (Figure S2 a,b,d). Using a more robust and specific EV isolation approach, such as a combination of filtration or ultracentrifugation, followed by density gradient fractionation would absolutely improve the quality of their manuscript by providing more convincing evidence that the proposed cargoes are indeed trafficked within EVs. The authors should provide further experimental validation of Sas's EV enrichment, isolating EVs from cultured cells using a more accurate approach.

We have done this. We isolated EVs from S2 cells expressing V5-tagged SasFL using a standard ultracentrifugation protocol (Thery et al., 2006). We observed that SasFL was also present in EVs isolated in this manner (new Supp. Figure 1). The presence of dArc1 and dArc1 mRNA in EVs from S2 cells is not in question, since this has been published by several other groups.

Figure 1

Lack of quantification: 1 image is just not enough

Calculate the ratio lysate/EV, also with EVi as a marker for EVs. Only the lysate is shown. EVs should also!

The authors claim (line 153) that most of FL Sas are in EVs? ` It looks like 60%, but it depends on the loading. Please quantify.

We quantitated the relative amounts of SasFL and Sasshort in EVs and lysates, and these data are now reported in the paper. On average, 46% of SasFL and 10% of Sasshort are in EVs. We also now use Wg as an EV marker instead of Evi, because Wg is almost absent from lysates. Wg was originally shown to be a marker for EVs from S2 cells by Koles et al., 2012. We observed that Wg was present in EVs isolated with the kit and by ultracentrifugation.

In Figure 1b, the authors suggest that full-length Sas is a predominant EV component, but not Sas-short. The figure would be strengthened by including data describing the in vivo distribution of the V5-tagged short Sas isoform in conjunction with mCD8-GFP. One would expect the Sas-short isoform to be less broadly distributed compared to Sas-FL.

We have done this, and the data are in the new Figure 1—figure supplement 1. The data show that Sasshort fails to move into the extracellular matrix when it is expressed in Ap neurons.

The authors claim (line 153) that most of FL Sas are in EVs? ` It looks like 60%, but it depends on the loading. Please quantify. In Supplemental Figure 1, the full blots should be shown for panel A. Why are the control β-Tub bands shown side-by-side, but not the bands for Sas-FL and Sas-short? Also, the authors could include protein samples from wild-type and Sas-15 embryos to enhance the validation of their antibody reagents.

We have done the quantification. On average, 46% of SasFL is in EVs, and 10% of Sasshort. In the blot shown in Figure 1, 60% of SasFL is in EVs, and 10% of Sasshort. We replaced the blot of Figure 1—figure supplement 1 to show the two isoforms side-by-side. It is not possible to do western blots on sas mutant embryos because it is a lethal mutant, so there is no way to generate a pure population of embryos.

In Supplemental Figure 1, the full blots should be shown for panel A. Why are the control β-Tub bands shown side-by-side, but not the bands for Sas-FL and Sas-short?

We have replaced this blot. We now show 3 sections of adjacent lanes of the same gel.

Figure 3: There are no data showing the expression of Sas in SG cells using the GAL4 lines. Is this expression restricted to just SG cells?

We now include data (Figure 3—figure supplement 1) showing that Sage-GAL4 expresses only in SGs.

Figure 3: It is not clear how the quantification in g was performed. How many western blots (WBs) were analyzed? There seems to be a bubble in the first lane of f, which would preclude quantification. Why is d not quantified, and there seems to be an overall increase in background staining in e that is not specific to discs? The source data files are not labeled and these data should be incorporated into annotated supplemental figures. Is transfer in a-b due to Ptp10D? How many WBs were quantified in g? What is the specificity for full-length Sas? The expression of short Sas should not lead to its incorporation in EVs and their overnight addition should not lead to the same effect (Figure 3). This should be better investigated as short Sas is a good control for full-length Sas.

Six WBs were used for quantification, and five discs for each genotype. Transfer in Figure 3b may involve Ptp10D, since Ptp10D is expressed at low levels in wt discs (Figure 3—figure supplement 1). Transfer does not require Ptp10D, however, since there is transfer to untransfected S2 cells, which lack Ptp10D. Our earlier data (Lee et al., 2013) shows that Sas also has additional, unidentified binding partners. There would be no point in adding supernatant from S2 cells transfected with Sasshort, since there is almost no Sasshort in such supernatants.

Figure 3, contd. In the salivary gland (SG) expression data presented in Figure 3a-b, the authors should have included V5-Sas-short as a negative control for cross-tissue transport. Similarly, in Figure 3c-e, EVs isolated from Sas-Short transfected cells should have been included as a control in the overnight EV-imaginal disc incubation assays. Does over-expression of Numb on its own lead to increased uptake of Sas-FL EVs? Also, does the need for Ptp10D and Numb over-expression for efficient capture of Sas-FL EVs in Figure 3c-e raise doubts about the data presented in Figure 3a-b, in which the transfer from SGs seems to be efficient without the need for Ptp10D or Numb over-expression?

Also

a: Are wing discs the only recipient tissues upon S2 cells EVs exposure? The tissues in Figure 5 are the trachea and guts. Why switch?

b: What about Ptp10D mutant disc (as in section 2)?

Better quantification? Number of discs etc

We have added an experiment showing that when Sasshort is expressed in SGs, it does not move to discs (new Figure 3c). The experiment of adding supernatant from Sasshort-expressing cells is not worth doing, for the reasons explained above. We did some experiments with Numb alone, and it had no effect, so we did not include this. The transfer experiments don’t raise questions about Figure 3b for the reasons explained above. Ptp10D stimulates transfer but is not required. There is low-level Ptp10D expression in discs, and Sas has other binding partners. Transfer is much more efficient when SasFL is expressed in SGs in vivo than when supernatant is added to discs. We used discs because they can be easily dissected away from larvae. Tracheae are embedded in the body wall, and the larval gut is huge and hard to visualize. In embryos, tracheae and gut can be visualized in whole-mounts, but that is not possible in 3rd instar larvae. It is not worth doing Ptp10D LOF experiments. These would be time-consuming, requiring months to combine the LOF allele with drivers and reporters, and would likely not produce clear results, since Sas has other binding partners. We now report the number of discs measured (5 per condition).

Figure 4: C and d, IP data has no inputs for IPs, no sizing markers, and no IgG controls for antibody specificity. These data would also be more convincing if done with FL Sas and included co-Ips from cell lysates.

The IPs presented in Figure 4C with myc-tagged constructs are not controlled properly. Western blots of protein extracts prior to IPs should be shown and a control antibody should be used to show equal loading in the WB for either dArc1-Myc or Myc-rArc-FL expressing cells. Showing the IPs one on top of each other is very confusing and WBs are critical to showing that the expected proteins are expressed in the transfected cells, even if their IP signature is negative.

5. In general, the WBs in the figures show very white backgrounds with high contrast. Please check if any contrast enhancement (or other such changes) were made and if so, they need to be documented and justified. Else, please revert to raw images of blots. Total protein controls are also missing.

We have replaced the WBs with new blots showing lysate blotted with each antibody. The CD8 chimera is better than SasFL because it isolates the binding region to the ICD and also produces cleaner blots than SasFL, being a much shorter protein.

Figure 5: Ashley et al. (Cell 2018) showed that dArc1 mRNA transfer required the 3'UTR so it is puzzling that the authors used heterologous UTRs. The results using FISH on endogenous dArc1 mRNA are dramatic. The authors should show definitively that their probe does not pick up over-expressed dArc1. In general, a better quantitative analysis should be provided. For instance, there is no quantitative data for Figure 5, but this point should be examined in other figures as well. The dAC1 increased expression in the target cells upon dARC1 increased production in SG(Figure 5) becomes an important part of the paper (and the model) but is not investigated! How does it work? Does the delivery of darc1 mRNAs packaged in capsids simply lead to more dARC1 translation? Is it proportional?

Or, is there also stimulation of darc1 transcription? Is there also an increase in the mRNA level (we cannot see the SG control of 5o (sage>+) supporting the authors' claim on line 562!).

Also:

a: What is the dARC1 LOF mutant phenotype in SG and trachea?

b: Do EV produced upon FL Sas v5 expression in an SG not produce dARC1 in the recipient trachea?

c: Does this also depend on the same synergetic presence of Ptp10D and Numb in the trachea? Ptp10D LOF?

– For all the tissue imaging studies, including those in Figures 5-6,, the authors should provide an indication of how many specimens were analyzed in total, presuming the samples shown in the figures were representative of observations made across multiple tissue specimens co-labeled in the same experiment.

We used the dArc1 ORF construct so that we could distinguish the mRNA from endogenous dArc1 mRNA, and because it was reported to express well. The probe we used cannot detect the transgene RNA because that does not contain any of the dArc1 3’ UTR sequences in the probe. We now include new FISH experiments with an SV40 3’ UTR probe, which picks up only the mRNA from the transgene, and this confirms the results of Ashley et al., showing that the transgene mRNA cannot move away from the SGs because it is not loaded into capsids. Regarding the other criticisms of Figure 5: (1) it is beyond the scope of this paper to define the mechanism by which dArc1 protein increases production of the endogenous dArc1 mRNA. It could be due to de novo transcription or to mRNA stabilization. (2) there are no embryonic dArc1 LOF phenotypes that have been detected. (3) It would be difficult and time-consuming to incorporate Ptp10D LOF mutations into a genetic background that already contains several transgenes and a driver. In any case, it is likely that Sas has other binding partners, as demonstrated by our original paper (Lee et al., 2013), so we would not expect that transfer would be eliminated. (3) The FISH experiments are representative of the results from >100 embryos. Such experiments in Drosophila are normally not quantitated. (4) We show that translocation of dArc1 mRNA requires expression of both SasFL and dArc1. dArc1 capsids are a normal EV component, but our data show that without SasFL they cannot move to tracheae. It is difficult to determine what percentage of EVs have both SasFL and dArc1, because in immuno-EM experiments there are always many antigens that do not have gold particles, and in any case we were not able to do immuno-EM with anti-dArc1. (5) We now show data with both dArc1 3’ UTR and SV40 3’ UTR probes, showing that the transgenic construct expresses in SGs but its mRNA is not observed outside of the expressing cells.

Responses to the remaining criticisms in the composite review

2. Figure 2: Nice IEM and tomography, but many details are needed. Could more IEM profiles (more examples) be provided to be sure of `BG and specificity (in a gallery). Are the capsid reconstitution with purified dARC1 and 2 performed in the presence of darc1 rRNA? Any RNA (figure 2)? Description of dArc1 putative capsids is absent from the Results section (2f,g) until describing Figure 4 data (line 362). Given that there is no immuno-EM labeling of dArc1 protein, it is not clear if Sas and dArc1 are localized to the same EVs. Nor is it clear if the double membrane EVs are actually EVs that contain capsids. Overall, the EM data lacks quantification. How many EVs on average show Sas labeling? How many EVs have double membranes? The immuno-EM data would be strengthened by adding quantifications in the form of distance measurements of gold particles of Anti-V5 and Anti-Sas in relation to EV membranes. The dense protein staining surrounding EVs seems unusual, is this due to an artifact of the purification? EV kits are generally non-specific and isolate non-EV membranes, Corroboration using ultracentrifugation or size exclusion chromatography methods would be beneficial. SAS-FL overexpression results in more EVs, which confounds subsequent experiments suggesting that Sas targets EVs to specific cell types/regions.

1) In immuno-electron microscopy of embedded material, regardless of the methodology used, antibodies do not penetrate beyond the surfaces of the physical section. Because of this, only epitopes that are exposed at the section surfaces are available to and labeled by the antibodies. Thus, if a portion of an EV is at the section surface it can be labeled. However, if the entirety of the EV is contained within the volume of the section and no part of it is exposed at the surface, it is not accessible to the antibodies and will not be labeled (despite the fact that densities corresponding to the structure can still be seen). Because of this, overview images and/or galleries may show both labeled and unlabeled structures, but those that are unlabeled cannot be interpreted or quantified as being negative for the specific probe. (2) The capsids made from dArc1 and dArc2 made in E. coli do not have added RNA but probably incorporate E. coli RNA. In any case, we have removed these data, as they repeat published data on the capsids and don’t add value to the paper. (3) We have not been able to obtain immuno-EM labeling for dArc1. The antibody is of low quality and we had only a very small quantity of it. However, it is already known from Ashley et al., 2018 and others that EVs from untransfected S2 cells, which express SasFL only at very low levels, often contain dArc1 capsids and dArc1 mRNA. (4) There is no point in doing distance measurements since the Sas ECD is a flexible structure that would not have a defined length. In fact, the observed distance is highly variable, probably because the Sas ECD, which is composed of a chain of domains separated by linkers, is likely to be flexible and able to adopt many different conformations. (5) It is true that SasFL expression modestly increases EV production, but the effects observed in Figure 5 are all-or-nothing, so this cannot account for translocation of dArc1 mRNA to other cells. 6) We have purified EVs using ultracentrifugation, and these data are reported in Figure 1. We did not think it worthwhile to do a new series of EM experiments just to visualize S2 EVs purified in this manner, since this has already been published.

The authors' data suggest that dArc1 protein and mRNA are transferred by Sas-FL expressing EVs to Ptp10D enriched tissues in fly embryos. Could the authors conduct immuno-EM or nano-flow cytometry studies to assess the general proportion of Sas and dArc1 co-expression in purified EVs? If the general model is accurate, one would expect a higher degree of co-occurrence compared to control EV cargos. Additionally, evaluating the dArc1 protein and mRNA transfer properties in different genetic backgrounds (i.e. Sas-FL and Sas-Short in Ptp10D over-expressed or in Ptp10D null embryos) would significantly enhance the paper.

Also in Figure 5, the results suggesting that transgenic over-expression of dArc1 leads to a strongly increased expression of endogenous dArc1 should be validated using a complementary strategy (e.g. RT-qPCR on dissected tissue specimens, RNA-FISH with probes specific to the endogenous mRNA).

See above. We were not able to do immuno-EM with antibody against dArc1 and in any case this technique is not quantitative. One could not draw any conclusions from counting EVs that appeared to be labeled for both antigens vs. only one (see discussion above). We did do FISH with probes specific for the endogenous mRNA, as discussed above.

Most (all) experiments are performed with over-expression of FL Sas or ICD. Does endogenous Sas bind endogenous Ptp10D and dARC1? ICDs? Also full-length APP? Many of the conclusions would be strengthened by the loss of function experiments, especially showing a requirement for Sas in dArc1 transfer.

8. Use of more extensive fly genetics using specific Ptp10D LOF in wing discs and trachea (to show the converse of the GOF). Does Ptp10D acts as the MAIN receptor to FL Sas? Numb LOF, a combination of LOF and GOF? Does Ptp10D GOF compensate for Numb and vice versa?

1) Ectopic expression of Sas and dArc1 is the only way to evaluate transfer from one cell type to another, since both genes are broadly expressed. (2) There is no way to determine if endogenous Sas binds to dArc1. Endogenous Sas, however, is found in cells adjacent to those expressing Ptp10D (Lee et al., 2014; Yamamoto et al. 2017), suggesting that the two proteins might form a complex. (3) There would be no reason to express full-length APP in S2s, since we have already demonstrated that dArc1 and Arc binding can be localized to the APP ICD using the CD8 chimera. In any case, the rest of APP is on the outside of the cell/EV and therefore not accessible to Arcs. (4) One cannot show a requirement for Sas in dArc1 transfer. This would be possible to test if it was observed that dArc1 mRNA moved to other cells in the absence of SasFL coexpression in SGs (since one could then remove Sas and ask if transfer still occurred), but since dArc1 mRNA remains in the SGs unless SasFL is coexpressed this experiment cannot be done. (5) It would be very time-consuming to introduce Ptp10D LOF mutations into genetic backgrounds containing 3 or more transgenes (as in Figure 5). (6) Numb is an early embryonic lethal, so numb mutant embryos cannot be examined.

https://doi.org/10.7554/eLife.82874.sa2

Article and author information

Author details

  1. Peter H Lee

    Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    hlee@caltech.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2411-6094
  2. Michael Anaya

    Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  3. Mark S Ladinsky

    Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States
    Contribution
    Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1036-3513
  4. Justin M Reitsma

    Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States
    Present address
    AbbVie, Illinois, United States
    Contribution
    Resources, Software, Investigation, Methodology
    Competing interests
    is affiliated with AbbVie. The author has no other competing interests to declare
  5. Kai Zinn

    Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    zinnk@caltech.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6706-5605

Funding

National Institute of Neurological Disorders and Stroke (NS28182)

  • Kai Zinn

National Institute of Neurological Disorders and Stroke (NS096509)

  • Kai Zinn

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Mass spectrometry work was performed at the Caltech Proteome Exploration Laboratory. Imaging was done at the Caltech Biological Imaging facility. EM work was done at the Caltech Cryo-EM facility. We thank Violana Nesterova for figure preparation. We thank the following colleagues for reagents and Drosophila lines: Jason Shepherd (University of Utah) for pGEX-dArc and rArc constructs; Travis Thomson and Vivian Budnik (University of Massachusetts) for rabbit anti-Arc1; Douglas Cavener (Penn State) for rabbit anti-SasFL; Deborah Andrew (Johns Hopkins) for Sage-GAL4; James Skeath (Washington University) for guinea pig anti-Numb; Swati Banerjee (UTHSC, San Antonio) for rat anti-Repo, and Yuh-Nung Jan (UCSF) for UAS-Numb. We thank Simon Erlendsson, Fernando Bazan, Paul Worley, and Tino Pleiner for discussions about Arc purification and Arc and Sas structures. This work was supported by NIH RO1 grants NS28182 and NS096509 to KZ, and by Howard Hughes Medical Institute support to R Deshaies, who was JMR’s faculty supervisor when he was a postdoctoral fellow at Caltech.

Senior and Reviewing Editor

  1. K VijayRaghavan, National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

Reviewer

  1. Jason D Shepherd, University of Utah, United States

Publication history

  1. Received: August 21, 2022
  2. Preprint posted: September 8, 2022 (view preprint)
  3. Accepted: June 15, 2023
  4. Accepted Manuscript published: June 16, 2023 (version 1)
  5. Version of Record published: June 23, 2023 (version 2)

Copyright

© 2023, Lee et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 316
    Page views
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Peter H Lee
  2. Michael Anaya
  3. Mark S Ladinsky
  4. Justin M Reitsma
  5. Kai Zinn
(2023)
An extracellular vesicle targeting ligand that binds to Arc proteins and facilitates Arc transport in vivo
eLife 12:e82874.
https://doi.org/10.7554/eLife.82874

Further reading

    1. Cell Biology
    2. Computational and Systems Biology
    Matthew V Cowley, Johanna Pruller ... Christopher RS Banerji
    Research Article Updated

    Facioscapulohumeral muscular dystrophy (FSHD) is an incurable myopathy linked to the over-expression of the myotoxic transcription factor DUX4. Targeting DUX4 is the leading therapeutic approach, however, it is only detectable in 0.1–3.8% of FSHD myonuclei. How rare DUX4 drives FSHD and the optimal anti-DUX4 strategy are unclear. We combine stochastic gene expression with compartment models of cell states, building a simulation of DUX4 expression and consequences in FSHD muscle fibers. Investigating iDUX4 myoblasts, scRNAseq, and snRNAseq of FSHD muscle we estimate parameters including DUX4 mRNA degradation, transcription and translation rates, and DUX4 target gene activation rates. Our model accurately recreates the distribution of DUX4 and targets gene-positive cells seen in scRNAseq of FSHD myocytes. Importantly, we show DUX4 drives significant cell death despite expression in only 0.8% of live cells. Comparing scRNAseq of unfused FSHD myocytes to snRNAseq of fused FSHD myonuclei, we find evidence of DUX4 protein syncytial diffusion and estimate its rate via genetic algorithms. We package our model into freely available tools, to rapidly investigate the consequences of anti-DUX4 therapy.

    1. Cell Biology
    2. Developmental Biology
    Mengjing Bao, Ruth E Dörig ... Beat Suter
    Research Article

    Microtubules (MTs) are built from α-/β-tubulin dimers and used as tracks by kinesin and dynein motors to transport a variety of cargos, such as mRNAs, proteins, and organelles, within the cell. Tubulins are subjected to several post-translational modifications (PTMs). Glutamylation is one of them, and it is responsible for adding one or more glutamic acid residues as branched peptide chains to the C-terminal tails of both α- and β-tubulin. However, very little is known about the specific modifications found on the different tubulin isotypes in vivo and the role of these PTMs in MT transport and other cellular processes in vivo. In this study, we found that in Drosophila ovaries, glutamylation of α-tubulin isotypes occurred clearly on the C-terminal ends of αTub84B and αTub84D (αTub84B/D). In contrast, the ovarian α-tubulin, αTub67C, is not glutamylated. The C-terminal ends of αTub84B/D are glutamylated at several glutamyl sidechains in various combinations. Drosophila TTLL5 is required for the mono- and poly-glutamylation of ovarian αTub84B/D and with this for the proper localization of glutamylated microtubules. Similarly, the normal distribution of Kinesin-1 in the germline relies on TTLL5. Next, two Kinesin-1 dependent processes, the precise localization of Staufen and the fast, bidirectional ooplasmic streaming, depend on TTLL5, too, suggesting a causative pathway. In the nervous system, a mutation of TTLL5 that inactivates its enzymatic activity decreases the pausing of anterograde axonal transport of mitochondria. Our results demonstrate in vivo roles of TTLL5 in differential glutamylation of α-tubulins and point to the in vivo importance of α-tubulin glutamylation for cellular functions involving microtubule transport.