Journal Pre-proof Fibrogenic signals persist in DAA-treated HCV patients after sustained virological response

Background & Aims: HCV SVR, achievable now by means of DAA therapy, identifies a new class of patients requiring medical surveillance to be designed in relation to the liver disease stage advancement. To this end, identification of both disease biomarkers and therapeutic targets appears necessary. Methods: Extracellular Vesicles (EVs) purified from plasma of 15 healthy donors (HD), and 16 HCV infected patients before (T0) and after (T6) DAA treatment have been utilised for functional and miRNA cargo analysis. EVs purified from plasma of 17 HD, 23 T0 and T6 patients have been employed for proteomic and western blot analysis. Functional analysis in LX2 cells measured fibrotic markers (mRNAs and proteins) in response to EVs. Structural analysis was performed by qPCR, label-free liquid chromatography-mass spectrometry (nLC-MS/MS) and Western blot. Results: On the basis of observations indicating functional differences (i.e. modulation of FN-1, ACTA2, Smad2/3 phosphorylation, collagen deposition) of plasma-derived EVs from HD, T0 and T6, we performed EVs structural analysis. We found consistent differences in terms of both miRNA and protein cargos: (i) antifibrogenic miR204-5p, miR181a-5p, miR143-3p, miR93-5p and miR122-5p were found statistically underrepresented in T0 EVs with respect to HD whereas miR204-5p and miR143-3p were found statistically underrepresented between HD and T6 (p-value<0.05) (ii) proteomic analysis highlighted, in both T0 and T6, the modulation of several proteins with respect to HD; among them, the fibrogenic DIAPH1 was confirmed upregulated by western blot (4.4 Log2 fold change). Conclusions: Taken together, these results highlight structural EVs modifications, conceivably causal for long-term liver disease progression in HCV patients that persists despite the DAA-mediated HCV SVR.


Introduction
New direct antiviral agents (DAAs)-based therapy, available with first generation compounds since 2011, efficiently reaches the Hepatitis C Virus (HCV) eradication, thus providing a therapeutic opportunity for 71 million people (WHO estimated) affected by chronic HCV infection who eventually develop cirrhosis or liver cancer [1]. HCV elimination does not always result into a healing of liver disease, particularly in patients with advanced fibrosis or cirrhosis; emerging clinical studies with DAAs in patients with liver cirrhosis stirred a heated debate about the risk of HCC occurrence and recurrence after viral cure [2][3][4]. In this scenario, early non-invasive prognostic tools useful to highlight the long-term DAA therapy clinical output appear necessary in addressing patients' management. Recently the scientific community has focused on circulating extracellular vesicles (EVs) for two reasons: i) EVs-delivered biological message may be causal to disease progression and ii) their analysis may provide prognostic and diagnostic evidence. Current technologies allow EV analysis for both informational content, (i.e. DNA, mRNAs, microRNAs, long non-coding RNAs and proteins) and functional analysis, assessed by means of target cell response. In fact, EV analysis has already provided an important contribution as in the case of EBV-associated tumours, where it was found that circulating EV cargo may include Epstein Barr Virus (EBV) miR-BART2-5p, a diagnostic and prognostic biomarker functionally able to protect latent cells from EBV reactivation [5].
Similarly, in antiretroviral therapy (ART)-treated HIV patients EVs were found to carry proteins related to immune activation and oxidative stress, functionally bringing immunomodulatory effects [6]. Specifically in the frame of DAA therapy, it has been previously described that the expression of specific miRNAs present in EVs is modulated by this therapy as these changes correlate with the EV-mediated NK cell degranulation capability [7].
In this study, we address the DAA therapy impact on liver fibrosis progression analysing the functional abilities and the structure of circulating EVs derived from HCV-infected and DAA-cleared patients.
J o u r n a l P r e -p r o o f 5 EVs functional analysis was performed on LX2 as recipient cells, an HSC/myofibroblast cell line partially mimicking the highly proliferative myofibroblast-like cells that derive from quiescent hepatic stellate cells (HSCs). Indeed, during chronic HCV infection, HSCs convert into highly proliferative myofibroblast-like cells expressing inflammatory and fibrogenic mediators responsible for ECM accumulation within the microenvironment, thus contributing to the fibrotic process leading to cirrhosis and liver failure in advanced stages [8,9]. EV structural analysis was performed for both miRNA and protein content. Data obtained show that EVs derived from HCV-infected patients, with respect to EVs derived from healthy donors (HD), increase the fibrogenic activity of LX2 cell line and that this functional data correlates with upregulation of HSC activators (e.g. of DIAPH1) and downregulation of some antifibrogenic miRNAs (e.g. miR204-5p, miR93-5p, miR143-3p, miR181a-5p and miR122-5p). Notably, longitudinal analysis highlights the persistent EV pro-fibrogenic activity in spite of the DAAmediated HCV eradication, this again in correlation with the EV informational content: upregulation of DIAPH1 and downregulation of miR204-5p and miR143-3p.

Patients' selection
This study received the approval of the I.R.C.C.S. National Institute for Infectious Diseases L.

HCV RNA detection
Hepatitis C virus (HCV) RNA plasma quantification was measured using the ABBOTT Real Time HCV Assay (ABBOTT Molecular Inc., Des Plaines, IL, USA) with a reported LLOQ = 12 IU/mL.

Blood sampling
Peripheral blood in K2-Ethylenediaminetetraacetic acid (EDTA) BD Vacutainer® blood collection tubes (BD Biosciences, Franklin Lakes, NJ, USA) was centrifuged at 3500 g for 15' to obtain the plasma, then aliquoted and stored at INMI Biobank at −80°C until EV purification.

Extracellular vesicle (EV) purification
The isolation of plasma/cell-derived EVs was performed by positive selection using Microbeads recognizing the tetraspanin proteins CD9, CD63 and CD81 following manufacturer's instructions (Miltenyi Biotec, BG, Germany). Briefly, 1 mL of plasma sample/cell conditioned medium was centrifuged at 2000 g for 30' and at 10000 g for 45' to remove cell debris and larger vesicles. After EV labelling with microbeads, EVs contained in the sample were firstly magnetically separated and used in subsequent investigations.

Statistical analysis
J o u r n a l P r e -p r o o f 9 All statistical analyses were performed using GraphPad Prism 8 software. Data on LX2 cells were analyzed using Student's t-tests. Mann-Whitney test and Wilcoxon's non-parametric test were applied for EV miRNA and protein content analysis. Perseus software (version 1.6.7.0) after log2 transformation of the intensity data was applied to proteomic study. Statistical analysis was carried out on proteins identified in 100% of the samples. Results were considered statistically significant at p≤0.05. To improve visualization, a z-score plot and a cluster heat map were generated. Gene Ontology enrichment analysis of biological processes, molecular functions and cellular components were performed by PANTHER software using Fisher's exact test and applying the False Discovery Rate calculation as a correction for multiple testing.

Functional analysis of circulating EVs
Previous studies characterized LX2 cells as similar to "activated HSCs" [13] that retain a transcriptional reprogramming plasticity in response to co-culture with hepatocytes [14]. Here, the properties of the EVs derived from non-tumorigenic hepatocyte cell lines have been first tested on recipient LX2 cells and shown to induce a down-regulation of fibrogenic markers (i.e. FN-1, ACTA2, COL1α1 and TGFβ1), thus providing proof that LX2 cells represent a suitable read out for EV functional studies. These data compared with those obtained with EVs derived from HepG2 tumoral cell line are shown in Supplementary Figure S1. Therefore, this procedure has been carried out with EVs purified from plasma of HCV-infected patients before (T0) and after DAA treatment (T6), in comparison with those derived from HD. EVs purification protocol and their protein content are described in materials and methods, and shown in Supplementary Figure S2 and in Table 2.
Incubation of LX2 with EVs derived from HD plasma results in a significant reduction of FN-1, ACTA2, COL1α1, PAI-1 and CTGF mRNA expression levels with respect to LX2 untreated cells ( Fig. 1A). FN-1 and ACTA2 downregulation has been confirmed also at protein level (Fig. 1B).
J o u r n a l P r e -p r o o f 10 Moreover, extracellular matrix (ECM) deposition, investigated by confocal analysis and shown in Figure 1C, was found reduced in LX2 cells treated with EVs derived from HD in comparison with non-treated cells.
Notably, HCV-derived EVs do not share this ability. As shown in figure 1D, fibrogenic markers FN-1, ACTA2, COL1α1 and TGFβ1 mRNAs are significantly higher in LX2 treated with EVs of both T0 and T6 with respect to HD EVs. This observation has been confirmed at protein level for FN-1, ACTA2 and for the p-SMAD2/3 (markers of TGFβ pathway activation) (Fig 1 E-F). Finally, ECM deposition has been investigated by confocal analysis; as shown in Figure 1G, a significant collagen deposition has been observed in LX2 cells treated with T0 and T6 EVs, in striking contrast with what was observed with HD EVs (Fig. 1C).
Overall, these data provide evidence for an EV-mediated fibrotic stimulus in HCV patients and, notably, as highlighted by the longitudinal study, for its persistence in spite of the DAA-mediated viral clearance (Supplementary Table S1).

Structural analysis: circulating EVs miRNA content
In order to correlate the functional properties above described to structural evidence conceivably providing a mechanistic insight, EVs isolated from the plasma of the same HD and the same naïve, viraemic HCV patients have been firstly analysed and compared for their miRNA informational cargo. Indeed, different levels of fibrogenic markers may reflect the presence/absence of specific microRNAs (miRNAs) with known antifibrogenic properties (listed in Table 1).
A crucial aspect for a qPCR analysis of these miRNAs has been the identification of miRNAs to be used as normalizing reference factors; to this end, taking advantage of previous findings, geometric mean of miR26a, miR22-5p and miR191 has been here used [11,12].
Next, miRNA analysis was extended to EVs isolated from the plasma of the same patients after 6 months of DAA therapy (T6). Consistently with functional analysis on LX2 cell line, DAA treatment correlates with a statistical difference between HD and DAA-treated patients for miR204-5p and miR143-3p, thus indicating that EV-mediated signals still lack antifibrogenic informational content in spite of viral eradication (Fig 2). This result appears conceivably causal for the functional observation previously described, since both bioinformatic predictions and previous reports pinpoint these miRNAs as regulators of fibrogenic markers. In order to formally acquire mechanistic insights on each miRNA-specific role, a miRNA mimic-based analysis has been performed. miRNA mimic pool (miR204-5p, miR181a-5p, miR143-3p and miR93-5p) and each single miRNA mimic have been transfected into LX2 cells, and potential direct/indirect target mRNAs (FN-1, ACTA2, COL1α1, TGF-β1, TGF-βR1, CTGF) have been quantified. As shown in Supplementary Figure S4, this analysis provides functional evidence recapitulating the EV-based observation and allows to identify miRNA specific targets (Table1).
Taken together, these results indicate that circulating EVs derived from HD, HCV-infected and DAAtreated patients have a different miRNA content with a conceivable causal role on fibrosis progression also in spite of viral clearance.

Structural analysis: circulating EV proteomic analysis
Next, we aimed at EV protein content characterization; to this end, EVs have been purified as above from 3 HD and 3 HCV patients (at T0 and at T6). Following proteolytic digestion, proteins were extracted and separated in 8 fractions based on their hydrophobicity before being analysed by labelfree liquid chromatography-mass spectrometry (nLC-MS/MS). Several proteins have been found 12 coherently expressed at the same level in all the 9 samples (see Table 2). Among them, we identified two categories of EV markers in accordance with the "minimal experimental requirements for definition of EVs and their functions" [34]. The first category includes transmembrane proteins localized on plasma membrane and/or endosomes, proof of the lipid-bilayer structure specific of EVs Next, in order to identify differentially expressed proteins among the three groups, computational analysis has been carried out; firstly, principal component analysis (PCA) of 393 detected proteins indicated that HCV patients (T0-T6) are distributed in a distinct group from HD (Fig 3A). In the differential expression analysis (DEA) comparing HD to T0, 74/393 EV-associated proteins were found differentially expressed with FDR <0.05 (19 expressed higher in control and 55 expressed higher in T0, see Table 3A and B). Conversely, non-statistically significant differences in EV proteomic profile were observed between HCV patients before (T0) and after DAA treatment (T6) ( fig 3B). Indeed, as for PCA analysis, also for differentially expressed proteins, hierarchical clustering classified the subjects into only two groups, one comprising both T0 and T6 datasets, and the other containing HD. Heat map representation of these results is shown in Figure 3C. Furthermore, gene ontology (GO) enrichment analysis of the 74 differentially expressed proteins in T0 with respect to HD is reported ( fig 3D, Table 4). Among the proteins found upregulated in T0, new potential biomarkers and therapeutic targets causal to liver disease progression may be found.
So far, we focused on the protein diaphanous homolog 1 (DIAPH 1) belonging to the family of formin proteins and recently described to promote myofibroblastic activation of HSC [36][37][38]. Indeed, this protein was found as the most upregulated in T0, and worthy of note, also in T6 samples in spite of viral clearance. With the intent to validate this evidence by means of a different approach and in order J o u r n a l P r e -p r o o f 13 to extend it to a broader patients' number, western blot analysis was performed on 14 HD and 20 longitudinally enrolled HCV patients' (T0 and T6) EV samples. As shown in figure 4 and coherently with proteomic data, DIAPH 1 EV level was found significantly higher in T0 as well as in T6 with respect to HD. Taken together, these results provide evidence indicating that circulating EVs derived from HCV-infected and DAA-treated patients differ in protein composition with respect to HD.
Moreover, specific cargo content conceivably indicates an EV causal role on liver disease progression.
Overall, our data highlight a correlation among EV function and miRNA and protein EV content, proposing EVs as relevant players in HCV-related pathogenesis. Evidence shows a number of persistent modifications, a warning that a long-term fibrosis is progressing in spite of DAA-mediated viral clearance.

Discussion
The major result of our work is that it provides evidence on circulating EVs derived from plasma of both HCV-infected (T0) and DAA-treated patients (T6) impact on HSCs transdifferentiation/activation. The comparison between HD-derived EVs and both T0 and T6derived EVs highlighted functional differences correlating with structural EV properties.
Functionally, EVs derived from HD plasma display an antifibrogenic ability (i.e. decrease of LX2 fibrogenic markers) that is lost by HCV T0/T6 EVs; coherently, HCV EVs stimulates SMAD2/3 phosphorylation and collagen deposition. These functional differences correlate to specific cargo signatures in terms of miRNAs and proteins.
Our data indicating HCV-derived EV ability to induce fibrogenic markers expression are in line with previous reports highlighting activation of LX2 upon HCV EV uptake. In particular, EVs were found to activate TGFβ pathway (through miR19a and miR192) [39,40].
The antifibrogenic miRNAs miR204-5p, miR181a-5p, miR143-3p, miR93-5p and miR122-5p were found decreased in viraemic HCV patients and miR204-5p and miR143-3p were found decreased in J o u r n a l P r e -p r o o f 14 HCV T6 with respect to HD. Previous studies focused on intracellular miRNA profile in relation to HCV infection [41]. Specifically, among the miRNAs associated with inflammation, fibrosis and cancer progression, miR143-3p appears upregulated in HCV-infected cells. This should not be considered in contrast with our results since it is well established that EV miRNA content does not reflect the cellular level of ncRNA repertoire and that specific RNA-binding proteins (RBPs) participate in specific miRNA EV-loading [42,43]. This combined evidence suggests the need to investigate HCV influence on RBP loading activity.
With respect to EV protein cargo, a proteomic-based approach pinpointed the liver pathogenetic protein DIAPH1, increased in both T0 and T6 EV datasets (log (2x) 4.4-fold change) with respect to HD EVs. This latter result has been validated in a significant number of samples. DIAPH1 has a key role in myofibroblast differentiation and is involved in actin nucleation/elongation and ACTA2 induction [37,38]. Notably, it has been recently shown that DIAPH1 stimulates myofibroblastic activation of HSCs promoting the endocytosis of TGFβRII [36]; it may also contribute to HSC activation by favouring the formation of ACTA2-positive stress fibres. Moreover, DIAPH1 inactivation suppresses the HSC tumour-promoting effects in a tumour implantation mouse model [36].
With respect to other HCV-induced EV hit proteins here suggested as potential biomarkers, it is worth of note that ANXA3 and ACTR2 have been previously described as poor prognosis signatures in J o u r n a l P r e -p r o o f 15 HCV cirrhotic patients [44]. This evidence strongly supports the notion of fibrogenic/pathogenic EV informational content persistence in spite of viral clearance.
The risk of developing liver-related complications in DAA-treated patients reaching SVR appears as yet debated [45]. Interestingly, evidence has revealed that the liver fibrosis process is not unidirectional and permanent, but instead plastic and potentially reversible [46][47][48]. At the moment there are no "golden standard" therapies for liver fibrosis, and HSCs are now considered a main drug target as highlighted by selective Rilpivirine (RPV)-mediated HSC ablation able to ameliorate liver fibrosis [49].
In this context, circulating plasma and serum EV-content analysis could allow the monitoring of biomarkers (e.g. DIAPH1) associated with liver fibrosis. This in the light of supporting the long-term monitoring and clinical management of patients [50] also in the case of clinical picture amelioration that may be observed after SVR.
In a recent work [51] a quantitative proteomic analysis has been performed on microparticles (PMP)        Table 1 Anti-fibrogenic miRNAs analyzed in this study.
List of miRNAs analysed with their direct/indirect target genes and the relative references; direct/indirect target genes here validated by mimics are listed in the last column.   Table 3 Upregulated proteins identified in HCV-T0 and HD.