SUMOylation in astrocytes induces changes in the proteome of the derived small extracellular vesicles which change protein synthesis and dendrite morphology in target neurons

Emerging evidence highlights the relevance of the protein post-translational modi�cation by SUMO (Small Ubiquitin-like Modi�er) in the central nervous system for modulating cognition and plasticity in health and disease. In these processes, astrocyte-to-neuron crosstalk mediated by extracellular vesicles (EVs) plays a yet poorly understood role. Small EVs (sEVs), including microvesicles and exosomes, contain a molecular cargo of lipids, proteins, and nucleic acids that de�ne their biological effect on target cells. Here, we investigated whether SUMOylation globally impacts the sEV protein cargo. For this, sEVs were isolated from primary cultures of astrocytes by ultracentrifugation or by the use of a commercial sEV isolation kit. SUMO levels were regulated: 1) via plasmids that over-express SUMO, or 2) via experimental conditions that increase SUMOylation, i.e., by using the stress hormone corticosterone, or 3) via the SUMOylation inhibitor 2-D08 (2′,3′,4′-trihydroxy-avone, 2-(2,3,4-Trihydroxyphenyl)-4H-1-Benzopyran-4-one). Corticosterone and 2-D08 had opposing effects on the number of sEVs and on their protein cargo. Proteomic analysis showed that increased SUMOylation in corticosterone-treated or plasmid-transfected astrocytes increased the presence of proteins related to cell division, transcription, and protein translation in the derived sEVs. When sEVs derived from corticosterone-treated astrocytes were transferred to neurons to assess their impact on protein synthesis using the �uorescence non-canonical amino acid tagging assay (FUNCAT), we detected an increase in protein synthesis, while sEVs from 2-D08-treated astrocytes had no effect. Our results show that SUMO conjugation plays an important role in the modulation of the proteome of astrocyte-derived sEVs with a potential functional impact on neurons.


INTRODUCTION
SUMOylation is a post-translational modi cation where a SUMO family protein is covalently conjugated to a lysine residue of a target protein [1,2].It regulates neuronal function with possible consequences on cognitive functioning and neural plasticity [3,4], while its dysregulation is proposed to be associated with disease states [5].Moreover, astrocytes are central players in maintaining neuronal homeostasis, which is in part mediated by paracrine signaling, including their derived extracellular vesicles (EVs) [6][7][8].Experimental small extracellular vesicles (sEV) preparations, which are typically 30 to 200 nm in diameter, contain both exosomes and microvesicles: exosomes are released after fusion of multivesicular bodies with the plasma membrane, while microvesicles are generated directly from the plasma membrane [9,10].The levels of common sEV protein markers, such as CD63, CD9, Flotillin or TSG101, vary according to the vesicle type and their biogenesis pathway [11,12].In addition to the biogenesis pathway, the molecular content (i.e., lipids, proteins and nucleic acids) of sEVs depends on the physiological or pathophysiological state of the donor cell [7].When sEVs are transferred to target cells, they regulate cell function in a cargo-dependent manner, e.g., by in uencing physiological or pathophysiological processes [13,14].
There are three extensively studied SUMO genes with a molecular weight of ~ 10 kDa termed SUMO-1, SUMO-2, SUMO-3 in mammals [15].SUMO-2 and SUMO-3 share a high degree of sequence similarity and are therefore referred to as SUMO2/3.In contrast to SUMO-2, SUMO-1 lacks a SUMO consensus motif around lysine 11 and cannot form conjugate SUMO chains.SUMO conjugation impacts protein function and the formation of multi-protein complexes via non-covalent interactions with the SUMO interactive motif, also known as SIM [16].In that way, SUMOylation regulates the activity, stability, and subcellular localization of proteins and affects physiological processes such as translation, transcription, replication, chromosome segregation, DNA repair, differentiation, apoptosis, senescence, cell cycle, nuclear transport, and signal transduction, among others [15,17].SUMOylation regulates the ability of sEVs to export speci c proteins, such as α-synuclein and the ribonucleoprotein hnRNPA2B1 with consequent loading, in the last case, of miRNAs [18,19].Additionally, the glycolytic enzyme Fructose-bisphosphate Aldolase C is found exclusively in its SUMOylated form in sEVs derived from brain astrocytes [20], while the export of its SUMOylated form to sEVs is regulated by exposure of animals to stress.
We hypothesized that SUMOylation dynamically and globally affects the sorting of proteins into sEVs in astrocytes.We thus identi ed sEV proteins by mass spectrometry and found that stimulating SUMOylation in sEV donor astrocytes (by SUMO overexpression or corticosterone treatment), proteins related to cell division, transcription, and translation are enhanced in sEVs.At the functional level, sEVs derived from corticosterone-treated donor astrocytes increase de novo protein synthesis in target neurons.

Animals
Pregnant Sprague Dawley rats were acquired at day 10 after crossings in the Animal Care Facility of Ponti cia Universidad Católica de Chile.The progeny was used at postnatal day one.Rats were housed in individual cages under a 12:12 light:dark cycle under ad libitum access to food and water.
The care of the animals used in this study to perform cell cultures was in accordance with the recommendations of the National Institute of Health Guide for the Care and Use of Laboratory Animals and the ARRIVE Guidelines.The protocol was approved by the Universidad de los Andes Bioethical Committee in the frame of the Fondecyt Project 1140108 and (# CEC202039).In total, the progeny of n = 15 female rats were used.This study was not pre-registered.
No blinding and sample size calculations were performed.

Antibodies
The antibodies used in this study are listed in Table 1.Cell culture Astrocytes were obtained from the telencephalon of Sprague Dawley rats at postnatal day one as previously described [21].For this, P1 rats were anesthetized under CO 2 for 3 minutes between 10:00 to 11:00 AM and decapitated.
Astrocytes were maintained in DMEM (#12100046, ThermoFisher, Massachusetts, USA) containing 10% Fetal bovine serum (#26140079, ThermoFisher, Massachusetts, USA), with 100 units/ml of penicillin and 100 µg/ml of streptomycin (#15140122, ThermoFisher, Massachusetts, USA) and incubated at 37°C, with 5% CO2 and 95% humidity.On days 4 and 8 in vitro (DIV), the total media volume was changed.After 15 DIV, the astrocytes were replated to decrease microglia and seeded at a con uence of 70-80%.For pharmacological treatments, the culture medium was completely changed to sEV free culture medium (i.e., FBS was depleted from EVs by ultracentrifugation before addition).The following compounds were added 1 hour later:

Transfections
Astrocytes were transfected by using Lipofectamine 2000 in accordance with the manufacture's instruction.Brie y, Transfections were performed after reaching 60% con uency using a 3:1 ratio of DNA: Lipofectamine (#11668030, ThermoFisher, Massachusetts, USA), in Opti-MEM (#31985062, ThermoFisher, Massachusetts, USA).pCDNA3.1 HIS-SUMO-1 and HIS-SUMO-2 plasmids were generously donated by Dr. Ronald T. Hay, University of Dundee, UK [22].We used GFP plasmid as a control.After 6 h incubation, the medium was exchanged with sEV free culture medium (DMEM supplemented with 100 units/ml of penicillin, 100 µg/ml of streptomycin and 10% sEV depleted FBS).After 24h post transfection the e ciency of transfection was determined by GFP expression by uorescent microscopy.

Astrocyte immunostaining and image acquisition
Astrocytes were xed on coverslips with 4% paraformaldehyde (#158127, Sigma, Missouri, USA) in PBS for 10 minutes, washed with PBS and permeabilized with 0.2% Triton X-100.Then, astrocytes were incubated with blocking solution (10% BSA in PBS) for one hour followed by overnight incubation with anti-GFAP from ThermoFisher, Massachusetts, USA or anti-SUMO-1 from Cell Signaling, Massachusetts, USA, in blocking solution.Finally, astrocytes were incubated with the corresponding secondary antibodies in blocking solution, washed with PBS and mounted with a mounting medium with DAPI (#104139, Abcam, Cambridge, UK).The images were acquired with a Leica TCS SP8 laser scanning confocal microscope using LAS X 3.5.2.18963 software at 20x magni cation, Zoom: 1, laser lines: 405, 488 and 552nM.

sEV isolation
Cell cultures were grown for 72 hours in an sEV free culture medium (DMEM supplemented with 100 units/ml of penicillin, 100 µg/ml of streptomycin and 10% sEV depleted FBS via depletion of vesicles after ultracentrifugation for 2 hours at 100,000 g).The conditioned media was harvested to isolate sEVs by serial centrifugations as previously described [21,23].After 30 minutes at 2,000 g, the supernatant was recovered and centrifuged for 45 minutes at 10,000 g.Then, the supernatant was centrifuged for 2 hours at 100,000 g, and the new supernatant was eliminated.The sEV enriched pellet was washed once in PBS (#14190, ThermoFisher, Massachusetts, USA), centrifuged for 2 hours at 100,000 g and re-suspended in PBS to store at -80°C until use.Alternatively, and as indicated in each case, the Total Exosome Isolation kit (#4478359, ThermoFisher, Massachusetts, USA), was used following the provider's instructions.The sEV pellet was resuspended in PBS and stored at -80°C until use.For characterization of astrocyte sEVs, iodixanol density gradients were performed.For this, astrocyte sEVs (1mg) were resuspended in 1 mM EDTA and 36% iodixanol/PBS pH 7.4.The solution was placed at the bottom of the tube and covered by solutions of descending concentrations of iodixanol in PBS (30%, 24%, 18% and 12%) to centrifuge at 163,000 g for 24 hours in a TH-641 Swinging Bucket rotor.Fractions (1ml) were collected to measure their density and assess their protein pro le.

Nanoparticle tracking analysis
The nanoparticle tracking analysis (Nanosight NS300, Malvern Instruments, Malvern, UK) was used to determine particle concentration and size distribution.Samples were diluted 5 or 10 times in PBS to obtain ≥ 80 particles per eld for analysis, and three videos were recorded for 30 seconds each.

Western blots
Cells were lysed in RIPA buffer (150 mM NaCl, 25 mM Tris-HCL pH 7.4, 0.5% NP-40, 0.5% sodium deoxycholate, 20 mM NEM (N-Ethylmaleimide (NEM, #34115, Sigma, Missouri, USA).Protein from the homogenates and sEVs (sEV resuspended in PBS containing 0.1% SDS) was quanti ed using the bicinchoninic acid method [24].The samples were boiled with loading buffer and 20 mM NEM for 5 minutes at 100°C to prevent de-SUMOylation.As depicted, each lane was loaded with either the same amount of total protein or the same number of vesicles.Proteins were separated by polyacrylamide gel electrophoresis under denaturing conditions (SDS-PAGE).Gels were stained with Coomassie dye or transferred to nitrocellulose membranes to proceed with standard Western blot procedures.Brie y, membranes were incubated with blocking solution (10% non-fat milk in PBS) for one hour followed by overnight incubation with anti-ALDOA, anti-ALIX, anti-EF-2, anti-SUMO-1 and anti-CD63 from Santa Cruz, Texas, USA.Anti-FLOTILLIN, and anti-GM130 from BD Transduction Labs, New Jersey, USA).Anti-SUMO-2/3 (was generously donated by Dr. Ronald T. Hay, University of Dundee, UK [22]).Then, membranes were washed in PBS-T (PBS containing 0.05% Tween20) and incubated 1 hour with the corresponding secondary antibodies.Finally, membranes were washed in PBS-T and reveal with chemiluminescence kit (#32106, ThermoFisher, Massachusetts, USA).

Proteomics
sEV proteins were separated using SDS-PAGE.Each lane was divided into 8 sections to perform in-gel digestion.
Liquid chromatography followed by tandem-mass spectrometry (MS/MS) of the sample fractions was performed on a hybrid dual-pressure linear ion trap/orbitrap mass spectrometer (LTQ Orbitrap Velos Pro, Thermo Scienti c) equipped with an EASY-nLC Ultra HPLC (Thermo Scienti c).Peptide samples were dissolved in 10 µL of 2% acetonitrile/0.1% tri uoric acid and fractionated on a 75-µm i.d., 25-cm PepMap C18-column, packed with 2 µm of resin (Dionex, Germany).Separation was achieved by applying a gradient of 2-35% acetonitrile in 0.1% formic acid over 150 minutes at a ow rate of 300 nL/min.The LTQ Orbitrap Velos Pro MS was exclusively used for CID fragmentation when acquiring MS/MS spectra, which consisted of an orbitrap full mass spectrometry (MS) scan followed by up to 15 LTQ MS/MS experiments (TOP15) on the most abundant ions detected in the full MS scan.The essential MS settings were as follows: full MS (resolution, 60,000; mass to charge ratio range, 400-2000); MS/MS (Linear Trap; minimum signal threshold, 500; isolation width, 2 Da; dynamic exclusion time setting, 30 seconds; and singly charged ions were excluded from the selection).Normalized collision energy was set to 35%, and activation time was set to 10 milliseconds.Raw data processing and protein identi cation were performed by Peaks Studio 8.0 (Bioinformatics solutions Inc.).The false discovery rate was set to < 1%.The bio-informatic analysis was done by DAVID Bioinformatics Resources 6.8.pValue was represented as -Log 10 .
Data availability: The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identi er PXD029940.For reviewers, the account details are the following: Username:reviewer_pxd029940@ebi.ac.uk and Password:MhY1L1qp Incubation of neurons with sEVs A total of 1000 vesicles were added per neuron, and this was repeated 24 hours later.After 72 hours, the cells were used for uorescence non-canonical amino acid tagging assay (FUNCAT).

Morphological analysis
Sholl analysis was used to analyze neurons incubated with astrocyte-derived sEVs obtained in the different experimental via plugins from ImageJ software (National Institute of Health, USA).The dendritic tree was examined in 3 µm increments.The following parameters were obtained: longest dendrite length (i.e., the largest radius at which there is an intersection with a neuronal process) and the total number of intersections (i.e., the sum of all intersections with each different radius) [21].Images were processed using ImageJ software (National Institutes of Health) and and Adobe Photoshop 2020, and assembled using Adobe Illustrator 2020.

FUNCAT assay
To visualize newly synthesized proteins in cells, the FUNCAT method was used according to Daniela Dieterich [25].Cultures were incubated for 3 hours with methionine-free HibernateA supplemented with non-canonical amino-acid AHA (L-azidohomoalanine), which was incorporated into newly synthesized proteins instead of methionine.As a negative control, methionine was used (4 mM).After metabolic labeling neurons were washed with cold PBS-MC (

Characterization of astrocyte sEVs and SUMO over-expression
We initially aimed to characterize the vesicles obtained by two methods of sEV puri cation: the ultracentrifugation method (UC) and a commercial vesicle isolation kit (IK) (Supplementary Fig. 1).This is because sEVs obtained by IK provide a rapid method for precipitating EVs with a high yield.This is despite EVs being contaminated with proteins from the cell culture medium, since in our case this would not in uence the results when transferred to neuronal cultures and when using the proper controls (e.g., EVs obtained from stimulated and not stimulated cultured astrocytes).In contrast, EVs for mass spectrometry were obtained by UC.Our Western blots revealed that the sEV marker proteins Alix and CD63 were enriched in the vesicles obtained by both methods compared to the donor astrocytes.We detected similar Flotillin 1 (FLOT1) levels in UC vesicles and donor astrocytes, while it was present at lower levels in the IK sEVs.Independently of the preparation method, we did not detect any GM130 in sEVs (a protein that should not be present in sEVs) (Supplementary Fig. 1a).We found that both methods produced particles of similar size (152 ± 7.8 nm and 156 ± 10.7 nm from UC and IK, respectively (mean ± SEM, n = 6)) using nanoparticle tracking analysis (NTA) (Supplementary Figs.1b and c).Additionally, sEVs obtained by UC oated at the expected density for small extracellular vesicles including exosomes, i.e., 1.119 and 1.135 g/ml for fractions 6 and 7, respectively (Supplementary Fig. 1d), and these same fractions also contained Aldolase A (ALDOA) and Eukaryotic elongation factor 2 (EF-2, Uniprot ID # P05197) proteins (Supplementary Fig. 1d).
We transfected astrocytes with HIS-SUMO-1 (S1) or HIS-SUMO-2 (S2) to determine whether SUMOylation affects the protein content of astrocyte-derived sEVs (Fig. 1).SUMO conjugated proteins increased after SUMO 1 or SUMO 2 overexpression (Fig. 1a).The mean sEV size or number of harvested EVs after SUMO transfection in UC-derived sEVs did not differ among conditions (Figs. 1b and 1c), albeit a tendency to fewer EVs was observed after SUMO overexpression (n = 8; S1 p = 0.068 and S2 p = 0.133).However, the average protein content of each vesicle (in fg per vesicle) increased in S2-derived sEVs (n = 8; p = 0.046), and a similar tendency was observed in S1-derived sEVs (n = 8, p = 0.221) (Fig. 1d).We identi ed these proteins by mass spectrometry (n = 2 samples per condition, Supplementary Table 1) and then analyzed our results via Venn and Gene Ontology analysis (GO) using the DAVID database.We found 512 proteins signi cantly enriched in classical exosome components and common to all three experimental conditions (Fig. 1e and GO data not shown), con rming that the analyzed fractions were enriched with sEVs.We then analyzed the proteins exclusive to S1 sEVS (Fig. 1f) and S2 sEVs (Fig. 1g) and found they were enriched with proteins related to translation and protein transport, respectively.They were also both enriched with proteins involved in the glycolytic process.To con rm this nding, we looked for a more physiological/pathophysiological condition that could increase protein SUMOylation in cells and, eventually, change the protein content in their exported sEVs.Given our previous experience in which SUMOylated Aldolase C increased in rat serum sEVs after stress by movement restriction, we tested whether corticosterone (CORT) could regulate SUMOylation in donor cells [20].

Corticosterone treatment and SUMOylation in astrocyte sEVs
For this, we incubated astrocytes with CORT while, contrariwise, we inhibited SUMOylation with the E2 SUMO conjugation enzyme inhibitor 2-D08 [26] (Fig. 2).Equal quantity of proteins were loaded per lane (controlled by protein determination and Coomassie stainings, not shown).In astrocytes incubated with CORT, SUMOylation did not change compared to control in the case of SUMO-1 (1.1 ± 0.11 times over control, n = 5, p = 0.9728).In the case of SUMO-2, it increased 1.4 ± 0.13 times over control (n = 5, p = 0.04).In turn, 2-D08 decreased SUMO conjugation over control to 0.5 ± 0.05 (n = 6, p = 0.016) and to 0.5 ± 0.04 (n = 5, p = 0.011) in the cases of SUMO-1 and SUMO-2, respectively (Fig. 2a, 2b and 2c).We conducted immuno uorescence assays using the astrocyte marker glial brillary acid protein (GFAP) to check whether our experimental treatments affected astrocyte morphology and reactivity (Fig. 2d).2-D08 plus corticosterone severely affected astrocyte morphology and decreased cell density from 30.4 ± 2.9 to 15.8 ± 1 per 0.068 mm 2 (n = 3, p = 0.0003, not shown).We hypothesized that this condition was toxic to astrocytes.The sEVs we derived from these cells would be contaminated by apoptotic or other vesicle types; therefore, we did not use this experimental condition for transfer experiments to neurons or for proteomic analysis.2-D08 also decreased astrocyte size (Fig. 2e, n = 4, p = 0.031) and GFAP immunoreactivity (Fig. 2f, n = 3, p = 0.0093), but not survival (not shown).In addition, we observed a decrease in the relative SUMO-1 intensity in the cytosol (cytosol/nuclei) of astrocytes treated with 2-D08 and double treated with 2-D08 and corticosterone (Fig. 2g and h, p = 0.0012 and p = 0.0006, respectively).In contrast, the size of their nuclei did not change (Fig. 2i).Therefore, in the presence of the SUMOylation inhibitor 2-D08, SUMO or SUMO conjugated proteins concentrated in the nucleus.
To test whether the differential SUMO-dependent protein loading was independent of the cell type, we performed pharmacological treatments both in astrocytes and in HeLa cells (Supplementary Fig. 2).We loaded 10 6 sEVs from each experimental condition of astrocytes and HeLa cells on an SDS-PAGE gel to observe the general protein pattern (Supplementary Fig. 2a and 2b).We did not observe any differences in the average size of the sEVs derived from each experimental condition via NTA (Fig. 3a and Supplementary Fig. 2c).We then quanti ed the number of sEVs released per million cells when we collected them (Fig. 3b and Supplementary Fig. 2d) and found the cells treated with 2-D08 to release substantially more sEVs (n = 9, p < 0.0001 for astrocytes; n = 9, p = 0.0001 for HeLa cells).However, in the case of astrocytes, each vesicle contained less protein (n = 9, p = 0.0002) while a similar tendency was observed for HeLa cells (n = 9, p = 0.103).At the same time, we also found that astrocytes treated with CORT released sEVs containing more protein (Fig. 3c, n = 9, p = 0.0008), with a similar tendency in HeLa cells (Supplementary Fig. 2e, n = 9, p = 0.198).

Proteomics of astrocyte sEVs
We next identi ed the sEV protein content by mass spectrometry derived from control, corticosterone-or 2-D08-treated astrocytes (Supplementary Table 2).In total, we identi ed 149 proteins exclusive to sEVs from corticosterone-treated astrocytes (Fig. 3d).We compared protein synthesis-related hits (biological processes) between conditions and found rRNA transcription and transcription by RNA polymerase I to be over-represented in sEVs from corticosterone-treated astrocytes (Fig. 3e).In Fig. 3F, we list the 19 proteins directly related to protein synthesis (biological process GO term), exclusively present in CORT-treated astrocytes.We found predicted SUMOylation sites in 15 out of 19 proteins using the GPS-SUMO database [27] (Fig. 3f).In summary, increased SUMOylation raised the number of protein synthesisrelated proteins in sEVs, thus suggesting that these sEVs may affect target cell function.

Effect of astrocyte-derived sEVs on protein synthesis in neurons
sEVs can be taken up by neurons [21]; therefore, we investigated whether astrocyte-derived sEVs could modulate protein synthesis in target cells, i.e., neurons.We incubated mature cortical neurons (21DIV) with sEVs derived from astrocytes treated with 2-D08, CORT or DMSO (control).Newly synthesized (AHA-containing proteins) in neurons were observed by TAMRA staining (Fig. 4a).Protein synthesis was increased in the soma of neurons incubated with sEVs derived from CORT-treated astrocytes when compared with control (Fig. 4b, n = 3, p = 0.0224).We did not observe any changes in the FUNCAT signal from neurons incubated with control astrocyte sEVs or 2-D08 astrocyte treated sEVs compared to untreated neurons, thus con rming that sEVs derived from corticosterone-treated astrocytes contain more proteins involved in stimulating new protein synthesis.We used immuno uorescence for MAP2 to visualize neurites in culture and Sholl analysis to quantify dendritic arborization (Fig. 4c).In accordance with previous results, we found that neurons incubated with the sEVs derived from control astrocytes had a signi cantly lower number of dendritic branch intersections than untreated neurons, and this was more prominent between 40 and 80 µm from the cell soma (Fig. 4d and 4e, n = 7, p = 0.036 and 0.0098, respectively) [21].This effect was abolished after addition of CORT sEVs.We did not observe any signi cant differences in total dendrite length between conditions (not shown).
We detected two proteins present in astrocyte sEVs known to affect protein synthesis and metabolism directly (EF-2 and ALDOA) (Fig. 4f).We found that CORT increased loading of the low and high EF-2 molecular weight forms when compared with 2-D08 (Fig. 4g, n = 5, p = 0.0261 and p = 0.0454, respectively).In turn, as already observed, only the 55 kDa form of ALDOA could be detected in sEVs and its presence decreased with 2-D08 treatment when compared with CORT (Fig. 4h, n = 5, p = 0.030).Note that the molecular weight of non-modi ed ALDO isoforms is 36 kDa.The molecular weights of the proteins detected in sEVs are at least compatible with mono-or poly-SUMOylation-like posttranslational modi cations.This had been shown by us for de ALDOC isoform of aldolase [20].The conclusions of our study are presented as a graphical abstract in Fig. 5.

DISCUSSION
We have shown that SUMOylation facilitates the presence of transcription and translation-related proteins into sEVs, which is consistent with the predominant role of SUMO in the regulation of nuclear processes [28].Further studies are needed to determine whether the recruited proteins are directly selected by SUMOylation, by non-covalent interaction with SIM domains or by other indirect mechanisms.Notably, the high molecular weight forms of ALDOA and EF-2 are preferentially loaded into sEVs when compared to unmodi ed proteins.Moreover, classical mass spectrometry detects SUMOylated proteins to a low degree because of technical constraints [29].However, increased SUMOylation in donor cells changes the proteome of the derived EVs which may contribute to modify target cell function (e.g., protein synthesis).In turn, other molecular components involved in protein synthesis, such as miRNAs, could be associated to the increased SUMOylation in donor cells [18].
The effect of SUMOylation on the sEV proteome SUMOylation of proteins is a reversible modi cation due to the action of SUMO-speci c isopeptidases and, thus, it may regulate cellular processes in a highly dynamic way [30].Consistent with such a regulatory role, we did observe a signi cant effect of increased SUMOylation in donor cells on the sEV proteome.We determined that this mechanism affects the protein content of sEVs derived from HeLa cells and astrocytes.It is unknown whether this change in protein composition is related to increased protein stability of SUMO-conjugated proteins, of a speci c effect on loading (i.e., selection of proteins into sEVs during biogenesis), and whether in this case, this depends on the SUMOylation of the incorporated substrate or on indirect incorporation mediated by the formation of multiprotein complexes that depend on the interaction of SUMOylated proteins with SIM domains [31,32].SUMOylation poses a particular challenge for mass spectrometry as the absence of a basic residue near to the SUMO C-terminus results in a signi cant 27 or 32-amino-acid sequence branch conjugated to the substrate peptide.MS/MS analyses of these branch peptides generally reveal abundant fragment ions resulting from cleavage of the SUMO tail, but which obscure those needed for characterizing the target peptide sequence.In spite of this limitation leading to an under-estimation of SUMOylated proteins, we observed increased levels of proteins related to gene expression in these sEVs.This is consistent with the established relationship between SUMO and chromosome stabilization, DNA replication, mRNA splicing, and transcription and translation in cells [33][34][35].Interestingly, our work shows that a SUMO-dependent regulatory action can be transported via sEVs between cells, because protein synthesis is stimulated in target neurons.
The idea that SUMOylation of speci c proteins participates in their loading on sEVs is supported by previous work showing that hnRNPA2B1 [18] and α-synuclein [19] sorting depend on SUMOylation.In different cell types, the lysine residues of proteins modi ed by SUMO have been identi ed [36,37].In them, ALDOA (similar to other ALDO isoforms) reveals to contain 4 SUMOylation sites, suggesting that these are involved in loading into sEVs.However, the exact residue(s) that determine the loading of the 55 kDa form need(s) to be resolved.ALDOA plays a key role in glycolysis and gluconeogenesis, but it also has non-glycolytic, termed moon-like, roles, e.g., in regulating the actin cytoskeleton [38].We detected this protein with a molecular weight of approximately 55 kDa in sEVs, consistent with one SUMO conjugation and increased content in sEVs under stress conditions [20,39,40].We used CORT to emulate stress in vitro and 2-D08 to inhibit SUMOylation [41], and these treatments modulated ALDOA levels in sEVs, con rming a role of this modi cation as a mechanism that promotes the incorporation of selective proteins into sEVs.
In turn, EF-2 (which was detected by MS/MS), catalyzes the translocation of peptidyl-tRNA during the elongation phase.Several SUMO modi ed lysine residues of EF-2 have been reported [39,42]; and EF-2 was pulled down by noncovalent interaction assays with the SIM domain [43].The loading of the unmodi ed EF-2 form (100 kDa, unlikely to be SUMOylated) is regulated by CORT (which stimulates mainly SUMO-2 incorporation into proteins), suggesting that SUMO-2 SUMOylation affects indirectly the loading into sEVs.EF-2 is SUMOylated during anti-apoptotic and cytoprotective responses [44].Additionally, similar to other RNA-binding proteins, SUMOylation might favor the interaction of EF-2 with RNA and facilitate translation [45].The transfer of EF-2 from astrocytes to other CNS cells, including neurons, could contribute to protein synthesis-dependent neuronal plasticity [46,47].
In general, only few studies focus on the total protein content in sEVs and the number of sEVs released under different experimental conditions.Using a different post-translational modi cation, the relationship of protein modi cation and total sEV protein content was shown [48].However, these authors did not normalize for the number of secreted sEVs to parent cell numbers, as we did.We found that SUMO-1 and SUMO-2 globally increase protein content in the derived sEVs.We also found that sEVs derived from CORT-treated astrocytes (which increases SUMOylation), had a larger average protein content.In contrast, inhibition of SUMOylation using 2-D08 in both astrocytes and HeLa cells signi cantly increase the number sEVs that they produced; however, these vesicles contain less proteins.This is a novel nding as until now, the theoretical content of protein per vesicle, in the range of fg, had not been shown to be regulated [49].It should be determined whether these sEVs are similarly depleted of other molecular cargoes (such as RNA species).The release of other vesicle types, such as synaptic vesicles or insulin-containing vesicles in pancreatic ß cells, is also regulated by SUMOylation [50,51].Additional work is required to understand how SUMOylation in uences exocytosis of MVBs, and whether it shifts the endocytosis-exocytosis balance of small vesicles.
Functional effect of astrocyte-derived sEVs on neurons Our proteomic analysis of sEVs after corticosterone treatment revealed increased levels of proteins related to gene expression, especially transcription, compared to controls and 2-D08 treatment.When neurons were incubated with sEVs derived from control astrocytes, we observed a decrease in dendritic complexity.This could be due to decreased growth or increased pruning and therefore enhanced network maturation [21].In contrast, sEVs derived from CORTtreated astrocytes did not affect dendritic length or complexity compared to non-treated neurons but stimulated protein synthesis.We did not identify these newly synthesized proteins, but it may affect the stabilization of many essential neuronal functions such as metabolism, synaptic stabilization or growth, and transport mechanisms.Interestingly, the protein synthesis-promoting action opposes the inhibitory effect of control sEVs on the complexity of the dendritic tree, and this may promote neuronal plasticity [52].
2-D08 abolished this effect on protein synthesis, suggesting SUMOylation is necessary for loading proteins related to protein synthesis into sEVs.The relationship between SUMOylation and protein translation has previously been described [33][34][35], and it is also involved in chromosome stabilization, DNA replication, mRNA splicing, and transcription [53,54].The interesting new nding is that this capacity can be transferred to target cells via sEVmediated signaling.We did not use sEVs from astrocytes treated with CORT and 2-D08 simultaneously because of the abnormal morphology of these cells, while the loss of nuclear SUMO localization is consistent with cell toxicity.
Astrocytes treated with CORT alone did not display these abnormalities.However, we could not distinguish whether the effect of CORT on the sEV proteome was the result of activation of speci c sorting mechanisms, of changes in parent cell gene expression, or of increased protein stability [55,56].
SUMOylation of sEV proteins as disease biomarkers.
Circulating sEVs, which can be obtained from a blood sample, are powerful biomarkers for CNS disorders because they can cross the blood-brain barrier [20,57].Further studies are needed to identify speci c SUMOylated sEV proteins that indicate disease, or, on the contrary, resilience to disease [58].The search of SUMOylated proteins could increase the sensitivity and speci city of biomarkers.On the other hand, this nding could have other uses such as speci cally loading some therapeutically relevant proteins through the addition of a SUMO group to therapeutically relevant sEV proteins.Figures    Figure 5

Table 1
List of primary and secondary antibodies used for Western blot and immuno uorescence.
USA), an inhibitor of the SUMO-conjugating enzyme UBC9, or DMSO as a control, (#85190, ThermoFisher, Massachusetts, USA).The addition of these compounds was repeated 24 and 48 hours later, and the conditioned medium was collected 72 hours after the rst addition.The medium was collected to isolate sEVs, and the cells were trypsinized (Trypsin, #15090046, ThermoFisher Massachusetts, USA), resuspended in phosphatebuffered saline (PBS #14190, ThermoFisher, Massachusetts, USA).An aliquot was counted using a Neubauer chamber, and the remaining cells were homogenized and stored at -80°C until further use.
Carl Zeiss group) was used together with its user interface ZEN 2012 SP5 (Carl Zeiss group).All the images from the FUNCAT procedure were taken in Z-stacks.The FUNCAT signal was visualized with ImageJ as a re lookup table, to visually favor the expression range of the newly synthesized proteins through a range of colors from blue to white, where blue represents the absence of new protein synthesis.Average values are expressed as mean ± standard error mean (SEM).Statistical signi cance of results was assessed using one sample Student's t-test (when assessing whether fold change over control values in Western blots were different from 1); two-tailed Student's t-test or one-way ANOVA followed by a posthoc Dunnetts test using GraphPad Prism version 5 for Windows, GraphPad Software, La Jolla, California USA, as indicated in each case.Differences were considered statistically signi cant if p < 0.05.Throughout the manuscript, independent biological replicates are de ned as independently performed experiments on material derived from different animals.
and incubated with anti-MAP2 from Millipore Massachusetts, USA, in B-block+ 0.2% TritonX-100 for 2 hours at room temperature, washed with PBS pH 7.4 and incubated with secondary antibody in B-block solution for 1 hour at room temperature.Finally, the cells were incubated for 10 min with 300 mM DAPI 4′,6-diamidino-2-phenylindole (D9542, Sigma Missouri, USA) in PBS pH 7.4, washed with PBS pH 7.4, and mounted using MOWIOL mounting media MOWIOL (#81381, ThermoFisher, Massachusetts, USA).Immunohistochemistry staining in combination with FUNCAT were analyzed via confocal laser scanning microscopy (CLSM).For this, an Axio Z1 microscope provided with a LSM 710 confocal system (