The study subjects included individuals from 2 sources (1): chronically HIV-1 infected participants in Yun’cheng city, Shanxi province, China (2); newly infected MSM (men who have sex with men) in Beijing You’an hospital. Written informed consents were obtained from all participants. The overall study was reviewed and proved by the Ethics Committee of SHAPHC. HIV-infected Former blood donor cohort (FBD study): The 50 HIV-infected former blood donors donated plasma between 1995 and 1996 in three counties of Yuncheng city and were infected with HIV-1 from common-source of contaminated blood. The participants were screened and recruited at the beginning of 2009 (baseline of FBD study). The inclusion criteria were ART naïve subjects with no symptoms of AIDS or related opportunity illnesses. At the baseline, their ART-free statuses were confirmed by measuring the plasma concentration of antiretroviral drugs available in China. The loci of HLA-A, HLA-B, HLA-C and HLA-DR were also genotyped. For slow progressors, confirmative western blot analyses were conducted at two independent laboratories. After enrollment, participants attended quarterly visits unless subjected to antiretroviral medications. Absolute CD4⁺ T cell counts were measured at every visit and viral loads were assessed semi-annually. Immune activation was measured by the frequency of CD38 positive CD8⁺ T cells. The primary end point was defined as CD4⁺ T cell counts less than 350 cells/mL, initiation of long-term ART, AIDS or death (Supplementary Table 1 for detailed information). MSM study: The 54 newly HIV-1 infected MSM were selected from a previous established cohort at Beijing You’an Hospital, in which HIV-1-negative high-risk MSM were screened every 2 months for recent HIV-1 infection, and had been followed up for 2 years after the first positive point. Plasma samples spanning seroconversion (from pre-infection to early chronic infection) were collected. The timing of estimated infection was determined by the history of risk exposure and laboratory measurements. Most of the samples reached the end point within 12 months.

Cell lines. HEK293T (embryonic kidney cell line, ATCC #CRL-3216) was cultured in DMEM (Corning #10-013-CV) supplemented with 10% FBS (BI #04-001-1acs) and 1% (Corning #30-002-CI).

PAXgene Blood miRNA Kit (QIAGEN, Hilden, Germany) was used for extraction and purification of total RNA, including miRNA, from whole blood stabilized in PAXgene Blood RNA Tubes (PreAnalytix, BD, UK). The resultant total RNA containing miRNA was stored at -80°C until analysis. Total RNA samples were reverse transcribed to cDNA using Megaplex RT Primers and the TaqMan MicroRNA Reverse Transcription kit (Applied Biosystens, Foster City, CA), followed by 12-cycle preamplification reactions using the Megaplex PreAmp Primers and TaqMan PreAmp Master Mix (Applied Biosystens, Foster City, CA). miRNA expressions were then quantified by real-time PCR with the TaqMan Universal PCR Master Mix and TaqMan Array Human MicroRNA A+B Cards v3.0 (Applied Biosystems, ABI) on an 7900HT (Applied Biosystems). Expression levels of 754 miRNAs in each sample were quantified, with U6 used as the internal control. Relative miRNA expression levels were expressed as -ΔCt values, calculated by subtracting Ct values of U6 from those of each miRNA, and used for analyses. Statistical and hierarchical cluster analyses of expression data were performed using Biometric Research Branch (BRB)-Array Tools version 4.1.0 Beta 2. miRNAs showing minimal variation or the percentage of missing value more than 20% across the set of arrays were excluded from the analysis, and those whose expression differed by at least 1.5-fold from the median in either direction in at least 20% of the arrays were retained. Hierarchical clustering analysis was performed with median-centered correlation and complete linkage. Significance Analysis of Microarrays (SAM), which give 95% confidence that the false discovery rate was less than 5%, was applied to search for the differentially expressed miRNAs that are correlated with CD4⁺ T cell counts and HIV RNA levels.

The method for extracting total RNA from whole blood was described above. miRNA isolation from plasma was performed with miRNeasy RNA isolation kit (Qiagen). The RNA samples were reverse transcribed using stem-loop RT primer pools and TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems), followed by pre-amplification PCR according to the manufacturer’s instruction. The resultant cDNA products were diluted and subsequently subjected to real-time qPCR assay. All the assays were performed in duplicate and small nuclear RNA, RNU6B, was used as the normalization control.

Total RNA was extracted from cells using a RNA isolation kit from Qiagen, and 500ng of purified total RNA was reverse transcribed using oligo (dT)-15mer primer and AMV Reverse Transcriptase (Promega) according to manufacturer’s protocol. The synthesized cDNA was subsequently analyzed by SYBR Green-based quantitative real-time PCR assay using a Promega kit and the following gene-specific primers: ACTB, forward primer, 5’- GCATGGGTCAGAAGGATTCCT-3’, revers primer, 5’- TCGTCCCAGTTGGTGACGAT-3’; T-bet, forward primer, 5’-CGGCTGCATATCGTTGAGGT-3’, revers primer, 5’-TAGGCAGTCACGGCAATGAA-3’; IFNG, forward primer, 5’-TGGCTTTTCAGCTCTGCATC-3’, revers primer, 5’- CCGCTACATCTGAATGACCTG-3’; STAT1, forward primer, 5’-CAGCTTGACTCAAAATTCCTGGA-3’, revers primer, 5’- TGAAGATTACGCTTGCTTTTCCT-3’;

Primary peripheral blood mononuclear cells (PBMCs) were isolated from whole blood by Ficoll separation. The contained total CD4⁺ T cells or naïve CD4⁺ T cells were then isolated using human total CD4⁺ T cell isolation kit and human naïve CD4⁺ T cell isolation Kit (Stemcell), respectively. Cell purity was >95% as determined by flow cytometry with surface marker CD3+ CD4+ CD8- and CD3+ CD4+ CD8- CD45RA+ CCR7+ for total and naïve CD4+ T cells, respectively.

Primary T cells were resuspended at 5× 10⁶ cells/mL in serum-free culture media X-VIVO15 (Lonza), 200µL of the resulting cell suspension was transferred to each well of a 96-well round-bottom microtiter plate. Transfection mixture was prepared by mixing 120 pmol of miR-31 antagomir (antagomiR-31) or control antagomir (antagoNC) and 0.6µL of TurboFect transfection reagent (Invitrogen) into X-VIVO15 to a final volume of 20µL, followed by incubation at room temperature for 20 min before adding to each well. Cells were then cultured in a humidified 37°C, 5% CO2 incubator for 48 hours before being collected for analysis. For the assessment of the effect of miR-31 on 3’ UTR of STAT1, 3’ UTR of STAT1 or its miR-31 site mutated version was subcloned into pMIR-REPORT firefly plasmid (Ambion) and co-transfected with a Renilla luciferase plasmid, and miR-31 mimics (Mimic-miR-31) or negative control microRNA (mimic-control) into HEK293T cells using TurboFect. The cells were collected at 24 hours post transfection for measurement of luciferase activities using the dual luciferase reporter assay system (Promega).

Naïve CD4+ T cells isolated from PBMCs of healthy donors were transfected with antagomiR-31 or antagoNC as described above. Cells were collected 48 hours later in RNAzol (Molecular Research Center) and stored at -80°C until processed to RNA isolation using the Direct-zolTM RNA MiniPrep Kit (ZYMO). The RNA-seq library preparation, sequencing and analysis were performed by BioWavelet Ltd. (Chongqing, China). In brief, the library was constructed using EASY-RNAseq method (22) and then sequenced on Illumina HiSeq platform using the 150-bp pair-end configuration. Raw reads were parsed with Trimmomatic for adaptor and quality trimming before alignment to the human genome reference GRCh38.p7. Only reads assigned to a unique genomic locus were counted using FeatureCounts, and the resulting read counts were converted to FPKM values to measure gene expression. Gene set enrichment analysis of the expression data was performed using GSEA software (Broad Institute) as described (23); hierarchical cluster analysis was conducted based on average linkage and Euclidean distance metric using Multiexperiment Viewer (version 4.9, http://www.tm4.org) (24).

Flow cytometry was performed on a LSRFortessa Flow Cytometer (BD Biosciences) and the data were analyzed using FlowJo software. The following antibodies and regents were used: CD8 PB (clone RPA-T8) was from BD Biosciences; CD45RA PE-Cy7 (clone H2100), CCR7 APC (clone 3D12), CD38 PE-Cy7 (clone HIT2) were from eBioscience; CD3 PE/Dazzle594 (clone UCHT1), CD25 PE (clone BC96), HLA-DR ECD (clone L243), T-bet PE (clone 4B10) were from BioLegend; p24 FITC (clone KC57) was from Beckman; Live/Dead™ blue dye (Invitrogen) was used for distinguishing live and dead cells. Measurements of cytokines in cell culture medium were performed with BD™ Cytometric Bead Array Th1/2/17 kit according to the manufacturer’s protocol.

Isolated CD4+ T cells were transfected with antagomiR-31 or antagomir-NC for 48 hours, followed by culturing in R10 medium (RPMI-1640 medium supplemented with 10% FBS) under stimulation with anti-CD3 and anti-CD28 for 3 days. The cells were subsequently washed and spin-infected (1200xg for 2 hrs at 25°C) with HIV-1 IIIB at 0.01 multiplicities of infection (MOI). After washing, cells were cultured in R10 supplemented with IL-2 (20 IU/ml) for 11 days, cells and supernatants were then collected for infectivity determination. Cells were stained with live/dead, CD3, and CD4 antibodies and after fixation and permeabilization, samples were stained with anti-p24 antibodies before submission to flow cytometry analysis. The concentrations of p24 in the supernatant were determined by ELISA (bioMerieux Industry) according to the manufacturer’s protocol.

Western blotting analyses were performed with whole cell lysates of antagomiR-31 or antgoNC treated naïve CD4+ T cells according to previously published procedures (25). Following antibodies were used: mouse anti-actin antibody (1:3,000; Cell Signaling Technology, #3700S), rabbit anti-STAT1 Ab (1:1,000; Cell Signaling Technology, #9172S), and rabbit anti-phospho-STAT1 Ab (Cell Signaling Technology, #9167S). The immunoblots was imaged and analyzed using Image Studio (LI-COR Biosciences). The anti-human IFN-γ antibody (clone B27) used for neutralization of IFN-γ in the cell culture medium and the IgG1κ isotype control antibody (Clone MG1-45, Cat. No. 401409) were purchased from BioLegend.

The service of ATAC-seq was provided by Active Motif Inc., using procedures as described previously (26). In brief, 200,000 cells in duplicates were used for transposition reaction. After purification with MinElute PCR purification kit (Qiagen), transposed DNA was amplified by PCR with a total of twenty-five cycles, which was determined as optimal by qPCR. Library was then size selected using AMPure XP beads to remove fragments greater than 800 bp or smaller than 100 bp. The eluted library was validated for size distribution by DNA ScreenTape Assay (Agilent, USA) and quantified with Qubit dsDNA High-sensitivity Assay Kit (Thermo Fisher Scientific) on a Qubit Fluorometer. Finally, all the samples were pooled and sequenced on a NextSeq 500 in a pair-end modality rendering 2 x 150bp sequences. After removing low-quality reads and duplicate reads, the ATAC-seq data were aligned to the human reference genome hg38 and the alignments were visualized on the UCSC genome browser.

All statistical analyses were performed using GraphPad Prism 7.0 (GraphPad Software). Survival curve differences were assessed using the log-rank-test. The Mann-Whitney U test was used to compare the difference between two groups, data from the same individuals were compared by the Wilcoxon matched-pairs t test. The degree of association between two variables was measured by Spearman correlation. In all cases, P values below 0.05 were considered statistically significant; those below 0.1 were considered to approach significance. All P values reported were 2-sided.

The study was reviewed and approved by the Ethics Committee of Shanghai Public Health Clinical Center (SPHCC). Written informed consents were provided by all participants.

To gain insight into the roles of miRNAs during HIV-1 infection, we compared the miRNA profiles of whole blood samples from a cohort of HIV-infected former blood donors (FBDs), who donated plasma during 1995-1996 in Yuncheng city, China and contracted HIV-1 infection from common-source of contaminated blood (27, 28). The blood samples were collected during one follow-up visit in November 2010 (14-15 years post HIV infection), when the entire cohort was ART naïve; the cohort contained 4 EC (with undetectable plasma viral load), 5 VC (with plasma viral load of 50–2000 copies/mL) and 14 viremic patients (with plasma viral load >2000 copies/mL) (Figure 1A and Table 1). The miRNA profiles were clearly clustered into two major groups and all the samples from EC and VC, with only one exception, were assigned to the left group, suggesting a potential connection between miRNA expression with viral control. When patients were classified based on an immunologic definition, 6 belonged to LTNP (maintaining CD4+ T cell count >500 cells/μL over 10 years) and the rest were progressors (experiencing sustained decline in CD4+ T cell counts). The clustering of all six LTNPs in the left group (data not shown) indicated a correlation between miRNAs and disease progression. In search of miRNAs involved in both viral control and disease progression, we identified the differentially expressed miRNAs between low viral load group (<2000 copies/mL) and high viral load group (>10000 copies/mL), and those between high CD4+ T cell count group (>450 cells/μL) and low CD4+ T cell count group (<250 cells/μL). High CD4+ T cell count traditionally defines as count of 500 cells/μL or higher; we instead used 450 cells/μL as the cutoff point as 3 patients with CD4 + T cell count in the range of 477-494 cells/μL were assigned to the high CD4+ T cell count group after taking into consideration that their CD4 + T cell counts were greater than 500 cells/μL either before or after the visit. As shown in Figure 1B, 15 miRNAs emerged as potential miRNA candidates correlating with HIV disease from both immunologic and virological perspectives. Among the 15 miRNA candidates, 10 miRNAs, as marked in Figure 1A, showed expression patterns significantly correlated with CD4+ T cell counts and viral loads, according to microarray data. For validation, these miRNAs were quantified in 50 baseline samples from the FBD study (containing 5 EC, 6 VC and 39 viremic patients) by stem-loop quantitative PCR (qPCR) assays and their correlation with disease progression were subsequently assessed using the Kaplan-Meier survival analyses. To this end, we defined an absolute CD4+ T cell count of < 350 cells/μL, initiation of long-term antiretroviral therapy, or progression to AIDS and death as the endpoints. During the 5-year follow-up period, the patients in the FBD study showed divergent clinical outcomes with 33 of 50 (66.0%) reaching endpoint (Table 1). Among the 10 miRNAs analyzed, miR-31 has the most significant correlation with CD4+ T cell count (Figure 1C), and at the same time showed a correlation with disease progression as low levels of miR-31 were significantly associated with higher risk of disease progression with a Hazard Ratio (HR) of 23.57 (P = 0.0001) (Figure 1D).

To further explore the role of miR-31 in HIV-1 infection, we investigated an additional HIV patient cohort, an acute-phase prospective MSM cohort comprising 54 homosexual men newly infected with HIV-1 genotypes AE, B’ or BC, which circulate prevalently in China. The first HIV-positive samples were collected individually at 12~97 days post infection (dpi); 31 of the 54 (57.4%) participants experienced a precipitous drop in the CD4+ T cell count during the 2-year observation period after HIV-1 infection, and 16 consented to start ART (Table 1). The subgroup with higher levels of miR-31 showed substantially higher percentage of nonprogressors than that showing low expression of miR-31. This trend was discernible when using miR-31 levels assessed before HIV infection, but the p value was insignificant (Figure 1E, HR = 2.99, P = 0.09). The miR-31 high expression group and low expression group showed significant difference in disease progression when using the measurement of miR-31 during acute infection phase (Figure 1F, HR = 2.43, P = 0.019), or the early phase of chronic infection (Figure 1G, HR = 2.56, P = 0.013). The relatively small size of samples collected before HIV infection is likely to account for the failure to reach significance despite a high HR value (Figure 1E). These data substantiated the negative correlation between miR-31 level and HIV disease progression, suggesting a potential protective role of miR-31 in HIV disease.

To explore the mechanism by which miR-31 slows HIV-1 disease progression, we examined the expression of miR-31 in immune cell types in peripheral blood by analyzing a miRNA RTqPCR data from Rossi et al’s work (29). The analysis indicated that miR-31 was expressed at a much higher level in naïve T cell subsets than in non-naïve T cell subsets (Figure 2A). Characterization of blood samples from FDB cohort evidenced the link between miR-31 and naïve state of CD4+ T cells. When split into two halves according to the expression level of miR-31, the abundance of naïve CD4+ T cells was 2.8-fold higher in the group with high miR-31 expression, denoted as miR-31high, relative to the group showing relatively lower miR-31 level, namely miR-31low group (Figure 2B; P < 0.0001). We also observed a negative correlation between miR-31 expression with levels of two T cell activation markers, CD38 and HLA-DR (Figures 2C, D). Following this, we interrogated the microarray data sampling whole-blood RNA of 8 individuals from each of the miR-31high and miR-31low groups. Among the many differentially expressed genes between miR-31high and miR-31low groups, gene ontology analyses showed enrichment in the categories of cell death, cell cycle regulation, and T Cell activation (Supplementary Figure 1), indicating that miR-31 might regulate T cell activity through multiple mechanisms.

To corroborate these observations in vitro, we sorted naïve CD4+ T cells from healthy donors and subsequently treated them with either antagomiR-31 or antagomir mismatch control (antagoNC), followed by mRNA isolation 48 hrs later for RNA-seq analysis. The knockdown efficiency was >90% as determined by quantitative RT-PCR (Supplementary Figure 2). As compared to antagoNC-treated cells, the antagomiR-31-treated cells showed a transcriptional profile tilted more toward activation-like status, exemplified by consistent upregulation of activation markers, including IL2RA (CD25), CD38, CD69, IL2RB, and CD74, and downregulation of CCR7 and IL7R (CD127), which characterize no or low effector functions (Figures 2E, F). For validation, we assessed the membrane expression of CD25, CD38, CD69, CCR7 and CD127 by flow cytometry. Consistent with the RNA-seq data, the CD25+ subset was significantly expanded upon antagomiR-31 treatment relative to antagoNC treatment (Figure 2G), while a weak but discernible elevation was observed in CD38 and CD69 expression (Supplementary Figures 3A, B). Contrastingly, the expression of two stemness markers, CCR7 and CD127, were downregulated upon miR-31 inhibition (Supplementary Figures 3C, D). Thus, miR-31 inhibition promoted the acquisition of an activation-like status in naïve CD4+ T cells, substantiating a connection between miR-31 level with T cell homeostasis.

Activated, proliferating CD4+ T cells are highly susceptible to infection and support efficient HIV-1 replication (30). If our notion that the level of miR-31 gates the activation status of CD4+ T cells, we would predict that impairing miR-31 expression would render enhanced susceptibility of CD4+ T cells to HIV-1 replication, recapitulating the relationship we observed with HIV patients (Supplementary Figure 4). To this end, we isolated CD4+ T cells from healthy donors and then transfected them with antagomiR-31 or antagoNC, followed by 72-hr stimulation with anti-CD3/CD28 antibodies before infection with HIV-1 at a multiplicity of infection (MOI) of 0.01 (Figure 2H). As detected by flow cytometry, the percentage of cells expressing HIV P24 antigen increased over time for both antagomiR-31- and antagoNC-treated cells, however the former showed an approximately 3-fold elevation relative to the latter at 11 days post infection (Figure 2I). The difference in infection rate was corroborated by significantly increased level of p24 protein in the supernatants of antagomiR-31-treated cells (Figure 2J). This in vitro result strengthened our postulation that miR-31 is a major determinant of susceptibility of CD4+ T cells to HIV infection via governing their activation status.

Strong TCR signaling favors the development of Th1 cells (31), and a driving force of this process is provided by the action of transcription factor T-bet, which undergoes transient upregulation in naïve T cells after TCR stimulation (32, 33). Given the above demonstration of connection of miR-31 to T cell activation status, we sought to characterize whether inhibition of miR-31 leads to the induction of a Th1-biased response. Thus, we sorted naïve CD4+ T cells from healthy donors and treated these cells with antagomiR-31- and antagoNC same as above. As analyzed by flow cytometry, the expression of T-bet was significantly higher in antagomiR-31-treated cells than antagoNC-treated cells, with 5.4-fold (P = 0.0078) and 1.4-fold (P = 0.0078) increase in percentage of T-bet-positive cells and mean fluorescence intensity (MFI) respectively (Figure 3A). The production of IFN-γ, both in protein level in the supernatant assessed by Cytometric Bead Array (CBA) and in cellular mRNA level determined by quantitative RT-PCR, was concomitantly increased by antagomiR-31 treatment (Figures 3B, C). Two other Th1 cytokines, IL-2 and TNF, were also notably upregulated (Figures 3D, E). The upregulation of IFN-γ signaling was further revealed by gene set enrichment analyses (GSEA), within which antagomiR-31-treated naïve CD4+ cells, in comparison to those receiving antagoNC, showed significant enrichment of IFN-γ pathway (Figure 3F), exemplified by a few well characterized IFN stimulated genes (ISGs) including IRF1, IRF8, ICAM1, STAT1, IFI35, TAP1, SOCS1, and SOCS3 (Figure 3G). Taken together, these data suggested that a major impact of downregulation of miR-31 on naïve CD4+ T cells is enhancement of T-bet signaling and consequently induction of a Th1-biased response, which might account for the enhanced susceptibility to HIV.

To explore the mechanism by which miR-31 regulate T-bet signaling, we performed ATAC-Seq to compare the chromatin accessibility landscape of antagomiR-31-treated naïve CD4+ T cells to that of antagoNC-treated control cells. Among differentially accessible chromatin regions (peaks) displayed by antagomiR-31-treated cells, an informative one was mapped to ~11.7 kb upstream of TBX21 gene transcription start site (Figure 4A), corresponding to a previously identified distant enhancer element featuring STAT1 binding motifs (34). This data, which can be explained by recruitment of activated STAT1 to the mapped distant enhancer element, was in line with that STAT1 was among the most significant differentially expressed genes associated with antagomiR-31-treatment (Figure 3G). Linking these observations to the fact that STAT1 is required for the upregulation of T-bet during naïve CD4+ T cell activation (35), we thus hypothesized that miR-31 might control the activation status of CD4+ T cell via the STAT1-T-bet-IFN-γ circuit.

Following this, we first analyzed the relationship between STAT1 and miR-31 levels in whole blood samples of HIV-1-infected individuals. A significant negative correlation was detected, providing the first evidence supporting our hypothesis (Figure 4B). We also assessed the correlation between miR-31 and specific phenotypic markers, and found that miR-31 was negatively correlated with two stemness-related markers, CCR7 and IL7R (Supplementary Figure 5). These correlation assessments, combined with the positive correlation between miR-31 and activation marker CD38 and HLA-DR as shown in Figures 2C , D, substantiated the notion that the naïve state of CD4+ T cells is preserved at least in part by miR-31. The regulation of STAT1 by miR-31 was further validated in antagomiR-31-treated naïve CD4+ T cells, which exhibited increased STAT1 mRNA and protein abundance as well as upregulated phospho-STAT1 level when compared to antagoNC-treated control cells (Figure 4C). To investigate whether miR-31 directly regulate STAT1, we used bioinformatics to predict the potential miR-31 targeting of STAT1. One algorithm provided by miRWalk was able to identify a putative miR-31 target site in the 3’ untranslated region (UTR) of STAT1 mRNA (Figure 4D). Subsequently, we made two luciferase reporter constructs containing either wild-type STAT1 3’ UTR or its mutant with the predicted miR-31 target site being eliminated for evaluation of the effect of miR-31 co-expression (Figure 4E). A significant miR-31-mediated repression of luciferase reporter gene was observed with the construct containing wild type 3’ STAT1 UTR but was rescued by mutation of the predicted miR-31 target site (Figure 4F). Together, these experiments proved that STAT1 is a cognate target of miR-31.

Next, we evaluated the contribution of STAT1 downregulation to the activity of miR-31 in modulating CD4+ naïve T cells. To this end, we employed 5’-cholesterol modified siRNA to inhibit STAT1. Treatment with STAT1-specific siRNA (si-STAT1) successfully reduced, although not abolished, the antagomiR-31-induced upregulation of STAT1 and phospho-STAT1 (Figure 4G). As assessed by flow cytometry, the antagomiR-31-induced T-bet upregulation was substantially prevented by si-STAT1 treatment (Figure 4H), so was the elevated IFN-γ protein level in the supernatant as determined by CBA (Figure 4I). Finally, we determined the requirement of IFN-γ signaling for the antagomiR-31-mediated activation of STAT1. In the presence of an IFN-γ neutralizing antibody, the increment of total STAT1 and phospho-STAT1 induced by antagomiR-31 were both largely blocked (Figure 4J). We thus concluded that the regulatory role of miR-31 in the homeostasis of naïve CD4+ T cells is at least in part attributable to its gating action on the STAT1/T-bet/IFN-γ positive feedback loop via directly targeting STAT1 (Figure 5).