Proteomic analysis of plasma extracellular vesicles reveals mitochondrial stress upon HTLV-1 infection

Extracellular vesicles (EVs) can participate in intercellular communication and pathogenesis. EVs contain many cargos, including proteins, and the composition of EVs differs between cell-types and activation levels. Thus, plasma EVs can be used as a biomarker of systemic response to infection and/or disease progression. In this study, we aimed at describing alterations in the protein content of plasma EVs upon infection with the human T-lymphotropic retrovirus type 1 (HTLV-1). HTLV-1 is the etiological agent of a lymphoproliferative disease (ATL) and a series of inflammatory diseases, including a neurodegenerative inflammatory disease (HAM/TSP). We found that plasma EVs are more abundant and smaller in HTLV-1 asymptomatic carriers or HAM/TSP patients when compared to uninfected healthy donors. Moreover, EVs from HTLV-1 infected donors contain markers of metabolic and mitochondrial stress.

We collected by size exclusion chromatography the plasma EVs with a size comprised between 50 and 150 nm in diameter (Fig. 1A). On average, the size of the EVs isolated from the plasma of HTLV-1 infected individuals was significantly smaller than those of NI donors (Fig. 1A,B).
We also found that plasma from HTLV-1 infected individuals contained significantly more EVs than those from NI donors (Fig. 1C). The plasma EV concentration was the highest for HACs.

Proteome distinguished between infected and non-infected donors. EV protein content was ana-
lyzed by mass spectrometry. For each sample, a large number of peptides (more than 20,000) was detected, which correspond to more than 2,000 identified proteins per sample. A large overlap in exosome protein composition was observed among samples from individuals with similar clinical status (Supplementary Figure 1), demonstrating that inter-individual variations are minor.
Almost a third of all the identified proteins (32,5%) was annotated as belonging to extracellular exosomes (Gene ontology GO:0070062, false discovery rate (FDR) = 10 −88 ). Similarly, we detected 92 of the top 100 most frequently identified exosomal proteins, as defined by the Exocarta database. Viral proteins were not detected in any sample.
Next, the LFQ intensities of the EV protein content from the different samples were examined by pairwise correlation (Fig. 2A). Samples from NI individuals displayed high correlation values among them, and presented lower correlation coefficients with samples from HTLV-1 infected individuals. Intriguingly, high correlation values were obtained for the samples derived from HTLV-1 infected individuals, irrespective to the clinical status. Similarly, a hierarchical cluster analysis using Ward's method and a correlation-based distance was not able to segregate samples from HTLV-1 infected individuals according to the clinical status (HAC vs HAM/TSP) (Fig. 2B). Together, our data analysis of the LFQ intensities suggests that there is a significant protein signature of HTLV-1 infection in the collected EVs. The differences between EV protein composition in HACs and HAM/ TSP patients are subtler. Indeed, the strong correlation among samples from HTLV-1 infected individuals indicates that quantified proteins that move specifically in HAC or HAM/TSP represent a small proportion of all the quantified proteins. Proteins involved in blood coagulation (25 proteins present in GO:0007596, FDR = 1.5 10 −10 ) and complement activation (10 proteins present in GO:0006956, FDR = 3.6 10 −9 ) were more abundant in samples from NI (Fig. 3C). These proteins are also components of platelet granules (GO:0031091), suggesting that the decrease we observe was due to a lower proportion of platelet granules among plasma EVs. This hypothesis was supported by the fact that 97 out of 221 proteins that are more abundant in NI had been previously described in platelet-derived vesicles (as presented by Exocarta database). Interestingly, the deficit in platelet-derived proteins is most prevalent in HACs than in HAM/TSP, which mirrors the variations in exosome count (Supplementary Figure 2A Figure 2B). Moreover, these proteins are often more abundant in samples from HACs than of HAM/TSP patients.
As this trend follows the overall number of EVs, our data suggest that lysosomal and mitochondrial proteins are present together in the small EVs observed upon HTLV-1 infection.

Discussion
We found that, in HTLV-1 carriers, small EVs are accumulated in the blood stream. These vesicles are enriched in mitochondrial and lysosomal proteins.
It is very likely that these mitochondria-and lysosome-rich EVs are not released from the infected cells, but rather from uninfected cells (or hematopoietic origin or not) in response to viral detection or inflammation. First, HTLV-1 infected cells represent a small portion of the circulating cells (ranging from <0,05 to 10%), and would not be responsible for such an alteration in vesicles count. Moreover, there is no correlation between the amount of altered EVs and the HTLV-1 proviral load (the amount of vesicles in lower in HAM/TSP patients when compared to HTLV-1 carriers). Finally, the composition of these vesicles does not correspond to the composition of EVs secreted by infected cells, as previously determined 7 .
Viral proteins were not detected. The absence of detection of viral proteins was anticipated: HTLV-1 viral expression in vivo is low, and, as most of the isolated EVs may have other origins (e.g. platelets, neutrophils, uninfected lymphocytes). However, a meeting abstract suggested that both Tax and HBZ could be detected in exosomes from HAM/TSP patients (but not HACs) 8 . It is unclear whether Tax was not detected in exosomes in our study because of lower proviral loads, differences in disease progression, or technique sensitivity limits.
High levels of EVs are often described upon viral infection. For example, an increase in EV concentration has been reported in HIV viremic individuals 9 . We have also found that EVs present in the plasma of HTLV-1 infected individual are smaller than in NI donors. Changes in the size of an EV population are not uncommon. Smaller EVs have been previously observed in the context of inflammation, cellular stress and tissue damage 10,11 . These smaller vesicles could be either secreted by a particular cell type (that remains to be identified) or correspond to a change in secretion pattern by cells that otherwise secrete larger vesicles. Indeed, a given cell type can release subpopulations of exosomes with different morphology, and properties 12,13 ; the relative proportion of these EVs subpopulations may be modified in response to alterations in the microenvironment.
The proportion of platelet-derived vesicles was found diminished on HTLV-1 infection. The decrease in platelet granules is astonishing as HTLV-1 infection has been associated with increased platelet count 14 . Thus, we believe that the apparent diminution in platelet-derived proteins is not due to a deficit in the secretion by platelets, and is rather due to a dilution of these vesicles among virus-induced smaller EVs. This is consistent with the fact that the apparent depletion of platelet-associated proteins is more important for plasma from HAC than of HAM/ TSP. Of note, the apparent overexpression of complement and coagulation in HAM/TSP compared to HAC had been previously suggested 8 .
Lipoproteins have been found enriched, when compared to control donors. This is reminiscent of a previous study that found elevated levels of triglyceride and very low-density lipoproteins (VLDL) in HTLV-1 infected women compared with uninfected individuals 15 . Our cohort is mostly composed of female donors. It is unclear whether the fluctuations in lipoproteins occur within exosomes: VLDL overlaps with the size range of EVs or may be adsorbed at the surface of EVs in vitro 16 .
The most important protein signature in HTLV-1 infected individuals corresponded to mitochondrial proteins. Such enrichment has not been previously reported, neither on samples from HIV-infected patients 17 nor in the context of inflammatory diseases (e.g. inflammatory bowel disease, multiple sclerosis) 11,18 . Of note, Dutta et al. have not reported such alterations in the context of HTLV infection, likely because they compared the composition of exosomes isolated from the serum of HACs, HAM/TSP and ATL patients, but did not consider the composition of exosomes from NI healthy donors 8 .
Functional mitochondria have been shown to be transferable from cell to cell through vesicles both in vitro and in vivo. The transfer can occur in stress conditions, and can rescue mitochondrial functions 19 . However, this transfer is usually mediated through large vesicles 19 . In our case, the EVs are much smaller and they also contain lysosomal proteins. Thus, we speculate that these vesicles correspond to extruded mitochondrial in response to inflammation, or cellular stress. In vitro it has been reported that, under TNFα-stimulation, Jurkat cells release mitochondria like vesicles, which can be later recognize by innate immune cells and act as a danger signal 20 . This may occur in HTLV-1 infected individuals as they display high plasma levels of inflammatory cytokines 21 . Additionally, oxidative stress may participate in such mitochondrial extrusion: on the one hand, Hela cells  22 ; on the other hand, serum antioxidant capacity is reduced in HTLV-1 infected patients 23 .
However, these mechanisms do not fully explain mitochondrial extrusion in our case. Indeed, while HAM/TSP patients display higher levels of inflammatory cytokines, the concentration in mitochondria-and lysosome-rich EVs is lower in HAM/TSP patients. This suggests that, during HAM/TSP pathogenesis, mitochondrial extrusion is inhibited, or some sort of exhaustion occurs, which eventually counteract inflammatory-induced mitochondria extrusion.
In conclusion, we have demonstrated that EV protein cargo is altered in vivo upon HTLV-1 infection, and that they may mirror cellular stress. Such alteration may participate in pathogenesis.

Methods
Samples and ethic statement. Plasmas were obtained in the context of a Biomedical Research Program approved by the Committee for the Protection of Persons, Ile-de-France II, Paris (2012-10-04 SC). All individuals gave informed consent. Experiments were performed in accordance with guidelines and regulations.
Vesicle isolation and characterization. Extracellular vesicles (EVs) were isolated from 1 mL of human plasma by size exclusion chromatography (SEC), using qEV columns (Izon, UK) 24 . Fractions 7 to 9 were pooled, as recommended by the manufacturer. The retrieved vesicles were characterized for concentration and size by Tunable Resistive Pulse Sensing technology, using the qNANO device (IZON) 25 .
Peptide preparation and Mass Spectrometry. Protein samples were solubilized in urea 8 M, Tris 100 mM pH7.5, then disulfide bonds were reduced with 5 mM dithiothreitol (DTT) for 30 min at 23 °C and alkylated with 20 mM iodoacetamide for 30 min at room temperature in the dark. Subsequently, LysC (Promega) was added for the first digestion step (500 ng) for 3 h at 30 °C. Then the sample was diluted to 1 M urea with 100 mM Tris pH 7.5, and trypsin (Promega) was added to the sample (1 µg for 16 h at 37 °C, then 500 ng for 3 h). Proteolysis was stopped by adding 2% formic acid. Resulting peptides were desalted using Sep-Pak SPE cartridge (Waters) according to manufactures instructions.
LC-MS/MS analysis of digested peptides was performed on an Orbitrap Q Exactive Plus mass spectrometer (Thermo Fisher Scientific, Bremen) coupled to an EASY-nLC 1000 (Thermo Fisher Scientific). Peptides were loaded and separated at 250 nl.min −1 on a home-made C18 50 cm capillary column picotip silica emitter tip (75 μm diameter filled with 1.9 μm Reprosil-Pur Basic C18-HD resin, (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany)) equilibrated in solvent A (0.1% FA). Peptides were eluted using a gradient of solvent B (ACN,0.1% FA) from 2% to 5% in 5 min, 5% to 22% in 150 min, 22% to 45% in 60 min (total duration of the chromatographic run was 250 min including high ACN level steps and column regeneration). Mass spectra were acquired in data-dependent acquisition mode with the XCalibur 2.2 software (Thermo Fisher Scientific, Bremen) with automatic switching between MS and MS/MS scans using a top-10 method. MS spectra were acquired at a resolution of 70,000 (at m/z 400) with a target value of 3 × 10 6 ions. The scan range was limited from 300 to 1,700 m/z. Peptide fragmentation was performed using higher-energy collision dissociation (HCD) with the energy set at 28 NCE. Intensity threshold for ions selection was set at 1 × 10 6 ions with charge exclusion of z = 1 and z > 7. The MS/MS spectra were acquired at a resolution of 17,500 (at m/z 400). Isolation window was set at 1. The following search parameters were applied: carbamidomethylation of cysteines was set as a fixed modification, oxidation of methionine and protein N-terminal acetylation were set as variable modifications. The mass tolerances in MS and MS/MS were set to 5 ppm and 20 ppm respectively. Maximum peptide charge was set to 7 and 7 amino acids were required as minimum peptide length. A false discovery rate of 1% was set up for both protein and peptide levels. A match between run features was selected between biological replicates with a matched time window set at 1 min and alignment time window set at 20 min. Proteins identified in the reverse and contaminant databases were discarded from the list of identified proteins.
Quantification of the identified proteins was performed using the label free quantification (LFQ) algorithm in MaxQuant 26 .
Statistical Analysis. The pairwise correlation matrix represents the Pearson correlation coefficients between each pair of samples computed using all complete pairs of LFQ intensity values measured in these samples. The hierarchical cluster analysis has been conducted via multiscale boostrap resampling (1000 bootstrap replications) with the Ward's method and a correlation-based distance measure thanks to the pvclust function of the R package pvclust 27 , after log2 transformation of the LFQ intensities, imputation of the missing values with the imp.norm function of the R package norm and a standardization of the values.
For the statistical analysis of one condition versus another, proteins exhibiting fewer than 3 LFQ values in both conditions were discarded from the list to avoid misidentified proteins. Thus a protein is later referred as quantified in a condition when LFQs have been measured in at least 3 samples of the given condition. After log2 transformation of the leftover proteins, LFQ values were normalized by median centering within conditions (normalizeD function of the R package DAPAR) 28 . Remaining proteins without any LFQ value in one condition have been considered as proteins quantitatively present in a condition and absent in another. They have therefore been set aside and considered as differentially abundant proteins. Next, missing values were imputed using the ScIEntIfIc REPoRts | (2018) 8:5170 | DOI:10.1038/s41598-018-23505-0 imp.norm function of the R package norm. Proteins with a log2 (fold-change) inferior to 1 have been considered as proteins with no significant difference in abundance. Statistical testing of the remaining proteins (having a log2 (fold-change) superior to 1) was conducted using a limma t-test 29 , thanks to the R package limma 30 . An adaptive Benjamini-Hochberg procedure was applied on the resulting p-values thanks to the function adjust.p of R package cp4p 31 using the robust method of Pounds and Cheng 32 to estimate the proportion of true null hypotheses among the set of statistical tests. The proteins associated to an adjusted p-value inferior to a FDR level of 1% have been considered as significantly differentially abundant proteins. Finally, the proteins of interest are therefore the proteins that emerge from this statistical analysis supplemented by those being quantitatively absent from one condition and present in another.
Funtional Enrichement analysis. Enrichment in molecular functions or cellular localization of cargos was analyzed using Funrich 33 and STRING 34 databases. The lists of proteins composing exosomes and platelet granules were from Exocarta 35 .