We previously reported that recombinant murine Arc elutes from gel filtration columns in several distinct but overlapping peaks, consistent with the presence of at least three low-order oligomeric species and one high-order (greater than 20 mer) species (Byers et al., 2015). A portion of the high-order species are likely to represent the relatively homogeneous population of ∼32 nm virus-like particles that were reported to encapsulate mRNA in a relatively non-selective manner (Ashley et al., 2018; Pastuzyn et al., 2018). RNA binding is important for high-order Arc oligomerization, as stripping of RNA from bacterially expressed human Arc reduces the A_260/280 ratio from ∼1.0 to ∼0.68 and inhibits the formation of virus-like structures (Pastuzyn et al., 2018). Likewise, it was recently reported that Arc oligomerization increases upon addition of Arc mRNA (Eriksen et al., 2020).

In the present study, we sought to examine the membrane remodeling properties of nucleic acid-free Arc. Fortuitously, we found that the A_260/280 of Arc decreased from ∼1.0 to ∼0.57 (as expected for a nucleotide-free protein) if bacterial lysates were precipitated with 35% ammonium sulfate (AS) prior to Arc purification (Figure 1A). The supernatant obtained after AS precipitation had an A_260/280 of ∼2.0 (Figure 1A), a value indicative of pure RNA. AS treatment had no discernible effect on the electrophoretic patterns of our Arc preparations (Figure 1B). Importantly, Arc obtained from AS-precipitated bacterial lysates failed to display the high-order oligomeric peak that eluted near V_o but did not alter the elution positions of the internal peaks (Figure 1C). Thus, the Arc preparations that were analyzed in all GUV experiments presented below are nucleotide-free and do not form high-order oligomers (i.e., virus-like structures) in solution.

Arc contains α-helical N- and C-terminal domains connected by an intrinsically disordered linker region (Eriksen et al., 2020) (Supplementary Figure S1A). The C-terminal domain folds into a structure resembling the HIV Gag capsid domain (Zhang et al., 2015). The N-terminal domain contains determinants of self-assembly and lipid binding (Eriksen et al., 2020). We previously reported that Arc binds to small unilamellar vesicles formed from a brain lipid mixture approximately 5-fold more tightly than to vesicles formed from 100% phosphatidylcholine (PC). Because HIV-1 preferentially buds from membranes containing phosphatidylinositol 4,5-bisphosphate (PIP₂), we tested whether phosphoinositides were responsible for the higher affinity of Arc to the more complex lipid mixtures. Co-sedimentation assays showed that Arc bound with similar affinities to pure PC liposomes and to PC liposomes containing 3 mol% phosphatidylinositol 4-phosphate (PI4P), phosphatidylinositol 3-phosphate (PI3P) or PIP₂ (Supplementary Figure S1B). Thus, in contrast to HIV-Gag, the interaction of Arc with membranes is not appreciably enhanced by the presence of phosphoinositides. Hence, our initial studies of Arc-mediated effects on membrane morphology were conducted using GUVs generated from dioleoyl PC (DOPC) and labeled with LAURDAN as previously described (see, for example (Bagatolli et al., 2003; Pott et al., 2008; Malacrida and Gratton, 2018; Castro-Castillo et al., 2019)).

At room temperature, GUVs from DOPC lipids remain in the fluid phase. In the absence of Arc, GUVs from DOPC lipids typically did not show any internal structures. Exemplary LAURDAN fluorescence images are shown in Figures 2A–C. In contrast, dramatic alteration of the GUV structure was observed within 2–60 min after addition of Arc (2–9 µM) (Figures 2D–F). Many GUVs exhibited internal structures, either in the form of smaller vesicles or as irregularly shaped lipidic structures. To visualize Arc, we labeled the protein with Alexa 594 and repeated the experiment. Spectrally resolved images are shown in Figures 2G–I. As with unlabeled Arc, the formation of small internal vesicles was observed. Alexa 594 fluorescence remained strong in the solution outside the GUVs, indicating that labeled Arc had not accumulated significantly within the vesicles in less than 1 h. These observations suggest that Arc binds to the GUV membrane, which causes parts of the outer membrane to pinch off and to be released into the interior of the GUV as small vesicles. Binding of Arc to the GUV membrane is shown in Figures 2J,K, with panel (J) showing only the blue-green LAURDAN fluorescence and panel (K) showing only the red fluorescence from Alexa 594-labeled Arc. The difference in fluorescence intensity due to membrane-bound Arc and Arc in solution was quantified in Figure 2L and found to be significant, demonstrating that Arc accumulates on the GUV membrane. Arc binding to the GUV surface is more evident in Supplementary Figure S2A, in which labeled Arc has been added to GUVs while they are still attached to the platinum wire upon which they were electroformed.

To quantify the effect of Arc on the GUV structure, we repeated several experiments with 15–91 GUVs per group (Supplementary Table S1) and in each determined the percentages of GUVs with internal structures (Figure 3). The presence of internal lipids was observed in only 16.5% ± 2.4% (mean ± SE) of GUVs formed in the absence of Arc compared to 63.6% ± 10.5% of vesicles treated with unlabeled Arc. A similarly high fraction of GUVs with internal structures was observed for Arc labeled with Alexa 594. To ensure that time alone did not lead to the formation of internal vesicles, we re-imaged the control group after 60 + min and observed no significant changes. We also added the buffer solution without protein to exclude changes in osmolarity as a potential cause of lipid internalization. Finally, as a negative control and example of an inert protein, EGFP (5 µM) was shown not to induce internal vesiculation, and neither did Glutathione S-transferases (GST) protein (Supplementary Figure S2B).

In a prior study, we showed that approximately 50% of Arc in mouse brain synaptosomes is resistant to extraction by cold Triton X-100 and is recovered in low buoyant density fractions following density gradient centrifugation (Barylko et al., 2018). These properties are consistent with the distribution of Arc to so-called “lipid rafts,” which are operationally defined as membrane microdomains enriched in cholesterol, sphingolipids, and saturated phospholipids (Pike, 2006). Lipid rafts have also been classified as liquid-ordered (L_o) membrane domains, in contrast to the more fluid liquid-disordered (L_d) domains which are enriched in unsaturated lipids (Bieberich, 2018). To test whether Arc-dependent vesicle formation occurs preferentially from L_d or L_o domains, LAURDAN-labeled GUVs were prepared from a ternary lipid mixture of DOPC, DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine) and cholesterol at a 1:1:1 ratio, which allows the separation of liquid phases (Ma et al., 2016). LAURDAN is one of the family of 2,6-naphthalene modified probes designed by Weber (Weber and Farris, 1979) to be sensitive to the polarity of its environment. LAURDAN’s emission is significantly red-shifted in disordered lipid environments (Parasassi et al., 1986). As shown in Figure 4A, the addition of unlabeled Arc to these GUVs induced the formation of internal vesicles, similar to those formed from DOPC vesicles. LAURDAN emission associated with these internal vesicles were generally greener in appearance (i.e., red-shifted), which is indicative of a more disordered lipid environment, relative to the bluer LAURDAN emission from the outer GUV lipid bilayer. After LAURDAN staining of the lipids, we added Alexa 647-labeled Arc to the GUVs. Examples of LAURDAN and Alexa 647 fluorescence images are shown in Figures 4A–C and Figures 4E,F, respectively. We note that for the GUV shown in Figure 4B, the Alexa 647 image overlay (Figure 4G) shows an accumulation of Arc inside the vesicle, confirmed by comparing the Alexa 647 fluorescence intensity inside and outside the GUV (Figure 4H). In rare cases, we also observed examples of positive membrane curvature (Figure 4G), arrows. The ability of Arc to protrude as well as to invaginate the GUV membrane may reflect the flexibility of the hinge region connecting its N- and C-terminal domains (Hallin et al., 2018) and may be relevant for its dual activities in extracellular vesicle budding and endocytosis.

pGST.parallel expression vectors were gifts of Dr. Hong Zhang (UTSW). His-MBP-TEV protease was a gift of Dr. Elizabeth Goldsmith (UTSW). Reagents for electrophoresis were from Bio-Rad (Hercules, CA, United States). Glutathione agarose was from Pierce (Dallas, TX, United States). Lipids were from Avanti Polar Lipids (Alabaster, AL, United States). Other reagents, buffers, and protease inhibitors and LAURDAN were from Sigma-Aldrich (St. Louis, MO, United States).

The GST-Arc construct was expressed in E. coli Rosetta 2 cells. Cells were harvested after growing for 20 h at 16°C. GST-Arc was extracted from the bacterial pellet with solution A (20 mM HEPES, pH 8.0, 100 mM NaCl, 5 mM DTT, protease inhibitor cocktail consisting of 10 µg/ml each of N-p-tosyl-L-lysine chloromethyl ester, N-p-tosyl-L-arginine methyl ester, N-p-tosyl-L-lysine chloromethyl ketone, leupeptin, pepstatin A, and 0.2 mM phenylmethylsulfonyl fluoride (PMSF) and 0.05 mg/ml lysozyme). The extract was centrifuged at 100,000 x g for 1 h, Arc in the supernatant was precipitated with ammonium sulfate at 35% saturation, and the precipitates were resuspended and dialyzed against solution A to remove ammonium sulfate. To purify GST-Arc, solutions containing GST-Arc were incubated with glutathione resin for at least 5 h and the resin was first washed with solution A, then with solution A containing 0.2% Triton X-100, and finally with solution A containing 2 M NaCl (without detergent). Washed resin was incubated with His₆-tagged TEV protease to release Arc (60:1 M ratio for 5 h). TEV was removed by incubation of Arc samples with His-tag purification resin (Roche). Purified Arc was dialyzed against solution B (20 mM HEPES, pH 7.5, 100 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine (TCEP), and PMSF), aliquoted, frozen in liquid N₂ and stored at –70°C.

Gel filtration chromatography of Arc was carried out by FPLC on a HiLoad 16/600 Superdex 200 column (GE Healthcare). Elution patterns were monitored by absorbance at 280 nm (A₂₈₀). A₂₆₀ was also measured and fractions were analyzed by SDS–PAGE and Coomassie blue staining. Calibration standards are listed in the legend to Figure 1.

Liposomes consisting of 100% brominated PC or 97% brominated PC plus 3% PI4P, PI3P, or PIP₂ were prepared as described by Lee and Lemmon 2001 (Lee and Lemmon, 2001). Lipids were dissolved in chloroform and dried under a stream of nitrogen followed by overnight drying under vacuum. Dried lipids were resuspended in 20 mM HEPES (pH 7.4) and 100 mM NaCl, followed by 10 freeze/thaw cycles in liquid nitrogen and sonication in a bath sonicator (Laboratory Supplies Company, Model G112SPIT). Liposomes were then extruded 10 times through 0.1 μm filters using a Mini–Extruder (Avanti Polar Lipids). To remove aggregates, Arc solutions were centrifuged for 1 h at 314,000 x g at 25°C immediately before incubation with liposomes. Binding assays were carried out by incubating 3.5 μM Arc with liposomes for 15 min at 25°C in 20 mM HEPES (pH 7.4) and 100 mM NaCl. Samples were then centrifuged at 300,000 x g for 1 h at 25°C in a TL-100 tabletop ultracentrifuge. Supernatants were removed and pellets were resuspended in initial sample volumes. Equal volumes of supernatants and pellets were electrophoresed on SDS-polyacrylamide gels and proteins were visualized by Coomassie blue staining.

GUVs were formed by electroformation (Angelova and Dimitrov, 1986) as previously reported (Bagatolli et al., 2003; Pott et al., 2008; Hedde et al., 2017; Castro-Castillo et al., 2019). Briefly, a set of Pt wires and a Teflon growth chamber were cleaned by sonication in ethanol and in chloroform for 60 min each. During sonication, 200 mM solutions of sucrose and glucose in demineralized water were carefully prepared to achieve equiosmolality between the two solutions. Lipid mixtures of either DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine) alone or DOPC, DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine) and Cholesterol (Avanti Polar Lipids, Alabaster, AL, United States) at a 1:1:1 volumetric ratio were prepared in chloroform at final concentrations of 0.3–0.4 mM of lipids. For fluorescence labeling, LAURDAN was added at a 1/200–1/500 dye-to-lipid ratio. All lipid samples were prepared in glass vials using Hamilton glass syringes cleaned with chloroform. After insertion of the Pt wires into the growth chamber, 3–4 µL of lipid solution (0.3 mM total lipids) was added to each wire, and the solvent was evaporated in a vacuum chamber for 30 min. After solvent evaporation, the camber was heated above 50°C to ensure that the lipids were above the melting temperature. To each of the three wells of the growth chamber, 350 µL of sucrose solution preheated to the same temperature was added. GUVs were formed by applying a 10 Hz sinusoidal electrical signal to the Pt wires with an amplitude of 2 V for 60–90 min. GUVs were released from the Pt wires by reduction of the frequency to 1 Hz for 10 min and transferred to 1.5 ml plastic tubes (Eppendorf, Hamburg, Germany). Glass bottom imaging chambers (Nunc Lab-Tek, ThermoFisher, Waltham, MA, United States) were prepared by coating with 1 mg/ml of BSA followed by the addition of 300 µL glucose solution to each well. GUVs were introduced to each chamber by adding 70 µL of solution. The higher density of the sucrose caused the GUVs to settle at the bottom of the imaging well.

GUV samples were imaged on a laser scanning microscope (LSM710, Zeiss, Oberkochen, Germany). Fluorescence was excited via a two photon process using a pulsed femtosecond Ti:Sapphire laser system (Mai Tai, Spectra Physics, Santa Clara, CA, United States) tuned to 780 nm for both LAURDAN and the Alexa dyes. The excitation beam was reflected off a 690-nm short pass dichroic mirror and fluorescence was detected with the LSM710 spectral detector in 32 channels in a wavelength range of 415–726 nm. Images of 256 × 256 pixels or 512 × 512 pixels were raster scanned with a pixel dwell time of 6.3 µs in regions of 35–354 µm. Each line was scanned four times and the signal averaged to improve the signal-to-noise. For consistency, images were generally taken at the GUV equatorial plane.

Arc was fluorescently labeled by reacting the protein with Alexa Fluor 594/647 NHS Ester (A20004, ThermoFisher, Waltham, MA, United States) overnight at 4°C. Labeled protein (∼1:1 labeling ratio) was separated from free dye using a Sephadex G-25 column (GE Healthcare, Marlborough, MA, United States). Arc-Alexa 594/647 was added to the GUV-containing imaging wells to a final protein concentration of 2–9 µM. EGFP was added to the GUVs at a final concentration of 5 µM. Fluorescence images were acquired within 90 min after protein addition.

Images of fluorescently labeled GUVs were quantified by visually determining the presence of internal lipid structures. If multiple substructures were present, only the largest was counted as one vesicle. Vesicles smaller than 5 µm were not considered in the analysis. Images were visualized with a linear rainbow color code from purple to deep red to visualize the 415–726 nm detection range of the 32 channel LSM710 spectral detector (Figures 2, 3) or by creating red-green-blue (RGB) image overlays (Figure 4) of the blue (449/10 nm), green (527/10 nm) and red channels (683/10 nm).