Comparison of methods and characterization of small RNAs from plasma extracellular vesicles of HIV/HCV coinfected patients

Hepatitis C virus (HCV) and human immunodeficiency virus (HIV) hijack the host exosomal machinery as an additional mechanism of infection and evasion of the immune system, modifying the small RNA (smRNA) cargo during infection. We characterized the surface epitopes of extracellular vesicles (EVs) from plasma HIV/HCV-coinfected patients and their smRNA cargo profile, by comparing different isolation procedures. Six EVs isolation procedures were compared: ultracentrifugation, and five different polyethylene glycol-based methods (commercial, combined with a column purification step and two custom); and two RNA commercial kits (phenol and non-phenol based) were used. High-throughput sequencing of smRNAs was performed. Exosomal surface epitopes were analyzed by the MACSPlex Exosome Kit. Four miRNAs displayed differences among protocols (hsa-miR-205-5p and hsa-let-7a/b/f-5p). The selection of RNA isolation kit impacted on the detection of miRNAs and other smRNAs, where the phenol-based RNA isolation kit performed acceptably. EVs surface was enriched with HLA-DR/DP/DQ, CD81, and CD8. There were three liver-specific miRNAs overexpressed (let-7a-5p, miR-21-5p and hsa-miR-122-5p), thus, EVs cargo might reflect liver disease evolution. Other smRNAs such as piwi-interacting RNAs were also detected for the first time. Custom polyethylene glycol precipitation-based methods combined with an RNA phenol-based kit yielded the higher number of smRNAs for EVs isolated from plasma HIV/HCV patients.

. Statistical analysis showed slight differences between the six exosome isolation protocols, displaying four miRNAs significantly different between methods (Fold change (FC) ≥ 2; False discovery rate (FDR) ≤ 0.05): hsa-miR-205-5p, hsa-let-7a-5p, hsa-let-7b-5p and hsa-let-7f-5p. The protocols with the highest differences with respect to the others were NOR and TEI. We applied a relaxed filtering criterion (FC ≥ 1.5 and p-value ≤ 0.05) to explore which miRNAs displayed the bigger differences between protocols without an FDR correction, and these differences were represented for each protocol respect to the others in a Venn diagram ( Fig. 2A-F). Thus, we could identify that NOR protocol was the most different to the others, while the custom protocols, PEG6 vs. PEG10, did not display any dissimilarities.
When we analyzed differences between the two RNA isolation kits (miRNeasy of Qiagen vs. Norgen), only the hsa-let-7b-5p showed significant statistical differences (FC > 2 and FDR < 0.05) (Supplementary Table S3). The partial least squares discriminant analysis (PLS-DA) shows us that samples clustered according to the RNA isolation kit used (Fig. 3A), but the predictive ability of classification according to RNA isolation kit used was not significant (R2 > 0.99, Q2 = − 0.41). This analysis shows us that the miRNAs hsa-miR-16-5p, hsa-miR-382-3p, and hsa-miR-3613-5p displayed the highest variable importance in projection (VIP) values (Fig. 3B). The 30 miRNAs with the highest VIP score in the PLS-DA show us that there is a clear different pattern between kits irrespective of the exosome isolation protocol used (Fig. 3C). In general, the miRNAs that most differentiate between RNA isolation kits, showed higher counts with miRNeasy kit (Qiagen).
Other smRNAs. Regarding other smRNAs, 1,138 were identified after filtering and were selected for subsequent analysis (56.2% piRNAs, 12.5% rRNAs, 18.4% snRNAs, 12.9% snoRNAs). A total of 67 smRNAs displayed at least one count in all experimental procedures (39 piRNAs, 16 snRNAs and 12 snoRNAs). None of them showed significant differences between exosome isolation protocols with FC > 2 and FDR < 0.05. Next, we applied a relaxed filtering criterium (FC ≥ 1.5 and p-value ≤ 0.05) to explore small differences between protocols,  www.nature.com/scientificreports/ and these differences were represented for each protocol respect to the others in a Venn diagram ( Supplementary  Fig. S2).
Concerning the RNA isolation kit analysis, 40 smRNA were differentially significant between Norgen and Qiagen (Supplementary Table S3). When we analyzed the total normalized counts and the number of identified smRNAs (Supplementary Table S2, Supplementary Fig. S1), the Norgen kit displayed a higher number of counts and miRNeasy the higher number of identified molecules. Figure 4A shows the PLS-DA of smRNA for RNA isolation kit, where samples clustered in two different groups (R2 = 0.99, Q2 = 0.19, three components). The RNA 5S ribosomal pseudogene 205 (RNA5sp205) showed the higher score for discriminating between RNA isolation kits (VIP = 3.4) (Fig. 4B). Finally, the top 30 smRNAs with the highest VIP in the PLS-DA were represented in a heatmap (Fig. 4C), which shows us two main clusters regarding the RNA isolation kit.
miRNA-related pathways. After normalization, a total of 45 miRNAs showed at least one count in all experimental procedures. We explored the putative regulated pathways by these miRNAs and identified that mainly fatty acid-related pathways ( Fig. 6) were represented. The key miRNAs related to this pathway were hsa-miR-103a-3p, hsa-miR-107, hsa-miR-15a-5p and hsa-miR-16-5p, which regulate fatty acid synthase (FASN), acyl-CoA synthetase long chain family member 3 (ACSL3) and 4 (ACSL4) and the 3-oxoacyl-ACP synthase (OXSM). The remaining pathways were mainly related to different signaling processes, where the extracellular matrix (ECM) receptor interaction pathway was well-represented by the hsa-let-7 family (a, b, c, f, g and i).

Discusion
This study aimed to characterize for the first time the surface markers and the smRNA cargo profile of plasma EVs of HIV/HCV coinfected patients. Twelve different procedures were compared to determine an optimal and transferable technique to clinical practice for biomarkers identification in these patients. As the different experimental procedures presented dissimilar performance, the method used for isolation of EVs should be carefully considered as it can largely influence the final result.
First of all, we have to account for the biophysical and molecular properties of EVs, which are sharing by HIV and HCV particles. This issue makes the precise separation between viruses and exosomes technically challenging. However, as our objective is the smRNA profiling, this is not a drawback. Regarding the smRNA cargo, miRNAs were highly abundant, followed by piRNAs, snoRNAs, snRNAs, and rRNAs. We observed that those based on custom (PEG6, PEG10) and commercial (TEI) polyethylene-glycol precipitation methods, combined with a phenol-based RNA isolation kit (miRNeasy, Qiagen), showed the highest number of different miRNAs and other smRNAs, while the non-phenol based kit of Norgen yielded the highest number of counts. Custom PEG protocols have been used for over fifty years to concentrate and purify viruses, which share some biophysical www.nature.com/scientificreports/ properties with EVs. Thus, similar procedures have been used to enrich and purify exosomes in an inexpensive and efficient alternative to commercial kits 13 . TEI contains a volume-excluding polymer, which additionally includes a proteinase K treatment to deal with plasma samples when the end goal is the analysis of exosomal RNA or protein cargo. Previous studies in serum samples have identified the best performance of PEG protocols for miRNAs identification, compared with other commercial kits. However, precipitated material may have protein aggregates, which suggests that precipitation methods such as PEG precipitate plenty of non-vesicular components that could interfere in EVs count. However, this is not a drawback if the goal is smRNAs profiling 14 .
In terms of counts, the commercial precipitation-based methods MIR and NOR yielded the highest expression for specific smRNAs, although they showed the lower diversity of smRNAs. Moreover, MIR includes a precipitation step with a thrombin treatment to enrich for EVs. Thrombin is a serine protease that converts fibrinogen to fibrin, theoretically making easier an efficient recovery of EVs. Thus, MIR has been recommended as an alternative to UCF for EVs analysis in serum samples 11 . However, others studies have pointed out that thrombin treatment seems to negatively affect the yield of EVs isolation procedures since thrombin-induced clotting seems to entrap EVs leading to a reduction in the number of isolated EVs 15 . This reduction in the yield is consistent with our results, as thrombin treatment could reduce the number of recovered EVs, and therefore the number of miRNAs detected. On the other hand, NOR is a precipitation-based method followed by a purification step in a column chromatography. Our results indicate that NOR Kit on its complete procedure (exosome isolation plus RNA extraction) is well optimized. However, if NOR exosome isolation is combined with the phenol-based kit miRNeasy, this combination gives us less yield.
Regarding the RNA isolation kit, we observed that miRNeasy kit showed higer number of miRNAs. Similar results were found by Prendergast et al. 16 in serum samples. Phenol-based methods such as miRNeasy kit usually performed well but seemed to produce some bias 17 , as structured smRNAs with low GC content are recovered inefficiently when using low RNA quantities.
Reports on smRNAs in EVs have been usually limited to miRNAs, but there is a variety of smRNAs, such as piRNAs, snoRNAs, snRNAs, and rRNAs, among others. PiRNAs were the most abundant after miRNAs. These molecules are a bit longer than miRNAs (24-30 nucleotides), and they are mainly expressed in germline cells but also in endothelial cells where their function is still unknown. PiRNAs are involved in DNA methylation and www.nature.com/scientificreports/ also mediate epigenetic changes; however, their primary function is suppressing transposable and other repetitive elements by binding to the PIWI subfamily of Argonaute proteins, keeping the genome integrity 18 . These smRNAs have been very little studied, but a putative implication in antimicrobial immunity has been explored in different organisms 19,20 . In this line, piwi proteins have recently been detected as inhibitors of HIV replication and other mobile genetic elements in activated T cells 21 , but additional studies need to be carried out to fully understand the piRNAs role in HIV and/or HCV infection.
Regarding the characterization of plasma EVs from HIV/HCV patients, we also analyzed the surface epitope markers of the plasma EVs, and the pathways regulated by their miRNA cargo. HIV-infected individuals have previously shown abundant plasma exosomes, enriched with CD9, CD63, and HSP70 22 , but there is no information about coinfection with HCV. On the one hand, we identified higher expression of exosome surface markers such as CD81, and molecules related to HIV and HCV infection, such as HLA-DR/DP/DQ and CD8. This is consistent with the higher immune activation of HIV/HCV coinfected patients. The MHC class II human leukocyte antigen isotypes (HLA-DR/DP/DQ) present antigens to immune cells, and they are widely present in exosomes, being also closely related to HIV and HCV infection. HLA-DP variants are related to the presentation of viral peptides, such as HIV envelope proteins 23 . HIV-induced immune activation leads to the expansion of CD8 T cells expressing HLA-DR, and the HLA-DR is incorporated into the envelope of HIV-1. The HLA-DR/ DP/DQ molecules are also crucial determinants in the immune response to HCV infection, through the effective presentation of viral antigens to T cells. Thus, several studies have identified genomic variants at HLA region significantly associated with HCV persistence 24 . The CD81 is a tetraspanin involved in cell proliferation, especially enriched in the membrane of exosomes. This molecule is highly present in hepatocytes, being critical for HCV entry. CD81 is associated with the HCV envelope glycoprotein E2, which promotes the HCV transmission by exosomes 25 within permissive cells, such as hepatocytes and peripheral blood mononuclear cells. CD81 is also involved in HIV assembly, co-localizing with the HIV-1 Gag protein in infected cells, exosomes, and virions 26 . Therefore, the surface markers identified in plasma EVs of HIV/HCV coinfected patients are associated with infectivity aspects of both viral infections.
Regarding the pathway enrichment analysis, fatty acid-related pathways were the main pathways targeted by the miRNA contained in EVs of HIV/HCV coinfected patients. Some of the miRNAs involved were hsa-miR-107, www.nature.com/scientificreports/ hsa-miR-103a-3p, hsa-miR-16-5p and hsa-miR-15a-5p, which targeted key fatty-acid genes such as FASN, ACSL3, OXSM and ACSL4. How these exosomal miRNAs are involved in modifying gene expression on target cells is still unknown, but they could be predictive biomarkers of liver disorders in HCV-related liver diseases. The interaction among HIV, HCV, and the antiretroviral treatment (ART) is associated with an increased risk of dyslipidemia, atherosclerosis, and cardiovascular diseases, among other metabolic abnormalities 27 . This deregulation may directly affect the innate antiviral response, liver disease progression, and the response to antiviral therapies. Additionally, HCV can hijacks and manipulates fatty acid flux to create specific lipid-enrichment microenvironments to promote its life cycle 28 , which will be reflected in the EVs content produced by the infected cells. We also identified the well-known miR-122 in plasma exosomes of HIV/HCV patients, which is a highly liver-specific miRNA implicated in fatty acids and cholesterol biosynthesis, among others. Additional significant pathways implicated in the fibrogenesis process were identified, such as are the glycosaminoglycan biosynthesis and the ECM receptors interactions pathways. The correct degradation of ECM is essential for the maintenance of tissue homeostasis, where the imbalance between the tissue inhibitors of metalloproteinases and matrix metalloproteinases causes liver damage and the progression or regression of fibrosis 29 . These molecules will, therefore, activate hepatic stellate cells (HSCs), which are the central mediator of fibrotic processes that will produce and accumulate extracellular matrix. This could explain, in part, the faster progression to fibrosis of HIV/HCV co-infected patients, since HIV creates a favorable milieu for the profibrogenic activation of HSCs 29 . This is the first report exploring the exosome surface markers and the smRNA profile of HIV/HCV coinfected plasma-derived EVs. Currently, research on both viral infections with exosome interaction is preliminarily, and their mechanism of interaction needs further investigation. www.nature.com/scientificreports/ conclusions PEG precipitation-based protocols combined with a phenol-based RNA isolation kit is a suitable method for extracellular vesicles smRNA profiling of clinical samples. The RNA isolation kit also significantly impacts on the detection of miRNAs and other smRNAs, where the miRNEasy kit was appropriated in terms of the number of identified molecules. The plasma EVs of HIV/HCV patients are enriched with HLA-DR/DP/DQ, CD81 and CD8 surface molecules, and their miRNA cargo mainly modulates fatty-acid metabolism genes, among others.

Material and methods
Twelve different experimental procedures were carried out for the purification of exosomes and other extracellular vesicles, and the isolation of the RNA for transcriptome analysis of smRNAs. Extended information on experimental design is included in Supplementary Material and Methods.
patients characteristics. Samples were recruited under the COVIHEP group as previously described 30 .
Samples included in the present study were obtained from Hospital Universitario La Paz (Madrid) and processed at National Center for Microbiology, Institute of Health Carlos III, Madrid (Spain). The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Institute of Health Carlos III review committee (CEI PI 67_2015-V4), and written informed consent was obtained from all patients involved. Four HIV/HCV coinfected patients of Caucasian origin were recruited. All of them were receiving suppressive ART and showing CD4 + T-cells counts ≥ 500 cel/mm 3 for at least one year before sample collection. Besides, patients had an active HCV-chronic infection, and they were naïve to any HCV treatment (positive PCR and positive HCV antibodies).

Biological material.
Whole peripheral blood was isolated and processed within the first 4 h after extraction. Density gradient centrifugation with Ficoll plaque was performed in order to separate peripheral blood mononuclear cells. Plasma fraction was pre-clarified at 15.000 rpm 15 min at 4 °C, and pooled and stored at -80 °C until use. Median fluorescence signal intensity (MFI) for all 39 capture bead subsets were background corrected by subtracting respective MFI values from the negative control. The normalization factor was performed by calculating the MFI of CD9, CD63, and CD81 markers for each sample. exosome isolation protocols. In total, six different protocols were analyzed for its performance in miRNA and others smRNA recovery, from 1 ml of a pooled pre-clarified plasma previously frozen at -80 °C.

Bead
Ultracentrifugation (UCF). Plasma was centrifuged at higher speed (12,000×g), in order to remove macrovesicles and apoptotic bodies. Afterward, the plasma was ultracentifuged twice at 43,000 rpm using a Beckman Optima L-90 K Ultracentrifuge (SW55Ti rotor). The pellet of purified exosomes was resuspended in 100 μl of cold phosphate-buffered saline (PBS). This protocol was modified based on the contributions of Rekker et al. 31 and Rani et al. 32 .
Total exosome isolation (TEI). The total exosome isolation kit (from plasma) (ThermoFisher Scientific) was used according to the manufacturer´s instructions, with modifications recommended by Lane et al. 33 and Van Deun et al. 34 . Proteinase K treatment was performed to maximize the purity of the exosomes to eliminate plasma proteins and dissociating the extracellular ribonucleoprotein complexes.
Mircury (MIR). The Mircury exosome isolation kit-Serum and Plasma, (Exiqon) was used according to the manufacturer´s instructions. This is a precipitation-based method which includes pretreatment of the plasma with thrombin to remove possible fibrin residues.
Norgen (NOR). We used the commercial kit Norgen´s Plasma/Exosome Serum and Free-Circulating RNA Isolation Mini Kit (Norgen, Biotek corp.), to isolate exosomes by a precipitation-based method followed by a purification step in column chromatography. The kit was used according to the manufacturer's instructions.
Two custom polyethylene glycol (PEG) based protocols. Two custom protocols were used with different molecular weights of PEG: 6KDa (PEG6) and 10KDa (PEG10). PEG6 was performed as described by Andreu et al. 14  RnA extraction. Total RNA including smRNAs were extracted from previously isolated EVs using two different kits: 1. miRNeasy Mini kit (Qiagen), which is a phenol-based procedure, following manufacturer´s instructions. DNAse treatment was performed. 2. Plasma/Serum Exosome and Free-Circulating RNA Isolation Mini Kit (Norgen), which works for isolating any size of RNA from extracellular vesicles such as miRNAs and free-circulating RNAs, were used following the manufacturer´s instructions.
The RNA concentration was measured by Nanodrop, and RNA size distribution was evaluated by the Bioanalyzer 2100 with Agilent RNA 6000 Pico kit (Agilent, catalog no. 5067-1513).
Sequencing. SmRNA library synthesis and sequencing were performed at the Centre for Genomic Regulation (CRG) at Barcelona (Spain). SmRNA library was performed with a high sensitivity kit for miRNA analysis in exosomes, the TailorMix miRNA Sample Preparation Kit (SeqMatic, ref. TM302), according to the manufacturer's protocol. Sequencing was performed in an Illumina HiSeq2500, single read, 50nts.
Bioinformatic analysis. The miRNA data analysis was performed as previously described 28 . SmRNA data (snRNA, snoRNA, rRNA, and piRNAs) was extracted with Oasis 2 software 36 from trimmed sequences.

MiRNA quantification by qRT-PCR. MiRNAs were quantified in exosomes samples extracted by both the
Qiagen and Norgen kits of exosomes isolated by all the different protocols as previously described 28 . Sequences of the miRNA nucleotides were extracted from the miRBase Release 21 (www.mirba se.org) 37 (Supplementary  Table S4).
Statistical analysis. The statistical analyses were carried out with the R statistical package version v3.4.1 (R Foundation for Statistical Computing, Vienna, Austria).
The R-package "edgeR" (v 3.18.1) 38 was used for the smRNA differential expression analysis. Trimmed mean of M-values normalization method was used, and a negative binomial generalized log-linear model to read counts for each gene. We used a GLM approach for multiple groups for determining differential expression among exosome protocols (two replicates per protocol), and among RNA isolation kits (six replicates per kit). It was performed with glmQLFit() and glmQLFTest() functions. Significantly differentially expressed (SDE) miRNAs were identified by a statistically significant p value < 0.05 adjusted by FDR using Benjamin-Hochberg correction, irrespective of the FC. We filtered out those miRNAs and smRNAs with less than the threshold of 12 counts among all samples.
We also performed a supervised multivariate analysis (multiple dependent variables) with a PLS-DA to create a regression model that classifies the variables according to their ability to classify each sample in the correct group. The optimal number of components for the model was determined with the leave-one-out crossvalidation method using R2 and Q2 values as performance measures. All variables were previously normalized, log-transformed (generalized logarithm transformation), and auto-scaled (mean-centered and divided by the standard deviation of each variable). The PLS-DA also provides the VIP, which estimates the importance of each variable in the projection used in a PLS model. The VIP is used for ranking variables, where a VIP score greater than 1 is considered to enable discrimination between 2 groups.
A Venn diagram was performed to determine the overlapping SDE miRNAs between comparisons with InteractiVenn 39 . miRNA-based target prediction and pathway enrichment analysis of the target genes. The web-based computational tool DIANA-miRPath v3.0 40 was used for the in silico target identification of the SDE miRNAs. This tool also performs a pathway union analysis of miRNAs targets, which is performed with the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways 41 .