Expression profiling of human milk derived exosomal microRNAs and their targets in HIV-1 infected mothers

Despite the use of antiretroviral therapy (ART) in HIV-1 infected mothers approximately 5% of new HIV-1 infections still occur in breastfed infants annually, which warrants for the development of novel strategies to prevent new HIV-1 infections in infants. Human milk (HM) exosomes are highly enriched in microRNAs (miRNAs), which play an important role in neonatal immunity. Furthermore, HM exosomes from healthy donors are known to inhibit HIV-1 infection and transmission; however, the effect of HIV-1 on HM exosomal miRNA signatures remains unknown. In this study, we used nCounter NanoString technology and investigated miRNAs expression profiles in first week postpartum HM exosomes from HIV-1 infected and uninfected control mothers (n = 36). Our results indicated that HIV-1 perturbed the differential expression patterns of 19 miRNAs (13 upregulated and 6 downregulated) in HIV-1 infected women compared to healthy controls. DIANA-miR functional pathway analyses revealed that multiple biological pathways are involved including cell cycle, pathways in cancer, TGF-β signaling, FoxO signaling, fatty acid biosynthesis, p53 signaling and apoptosis. Moreover, the receiver operating characteristics (ROC) curve analyses of miR-630 and miR-378g yielded areas under the ROC curves of 0.82 (95% CI 0.67 to 0.82) and 0.83 (95% CI 0.67 to 0.83), respectively highlighting their potential to serve as biomarkers to identify HIV-1 infection in women. These data may contribute to the development of new therapeutic strategies in prevention of mother-to-child transmission (MTCT) of HIV-1.

Sample acquisition and preparation. HM samples were self-collected into sterile tubes within the first week and at one, three, and six months post-partum, and immediately shipped on ice for processing in our laboratory. The samples were separated into lipid, skim milk supernatant, and cellular fractions and stored at − 80 °C and liquid nitrogen, respectively as previously described 26,27 . Exosome isolation from human milk. Exosomes were isolated from the skim milk supernatants using the Total Exosomes Isolation reagent (from other body fluids) as per manufacturer's recommendations (Thermo Fisher, Canada). Briefly, 500 µl volume of each HM sample was centrifuged at 2000×g for 10 min (1st spin). Without disrupting the pellet, supernatant was transferred to a new tube and centrifuged again at 10,000×g for 30 min (2nd spin). The supernatant was transferred to a new tube and centrifuged at 10,000×g for 10 min (3rd spin). To the clear supernatant, 500 µl of 1 × PBS and 500 µl of exosome isolation reagent was added, vortex-mixed and incubated for 30 min at room temperature. After, incubation, the samples were centrifuged at 10,000×g for 10 min and the supernatant was removed carefully and discarded. The exosomes in the pellets were dissolved in 50 µl of exosome resuspension buffer (Thermo Fisher, Canada), vortex-mixed and again centrifuged at 10,000×g for 5 min at room temperature. Without disturbing the non-organic particulate matter in the pellet, the supernatants containing the purified HM exosomes were transferred to a new tube and stored at -20 °C until further use.
Transmission electron microscopy (TEM). HM derived exosomes morphology was evaluated by TEM through negative staining as described 28 . Briefly, HM exosomes were placed onto formvar grids, fixed with 2.5% glutaraldehyde, and contrasted with 1% uranyl acetate and finally visualized with a JOEL-1200EX transmission electron microscope located at McMaster Electron Microscopy facility. The images with × 40,000-× 300,000 magnifications were taken using AMTV600 computer program.
Western blotting. Exosomes were isolated from the HM samples as described above. Protein fraction was isolated, quantified using DC™ protein assay kit (Bio-Rad) and run on SDS-PAGE gel. Western blot analysis was performed with the primary antibody against CD81 (sc-166029; Santa Cruz) and HRP-labeled goat anti-mouse IgG 1706516 (Bio-Rad) as secondary antibody as described 26,27 . Exosome RNA isolation. Total RNA was extracted from the HM exosomes using Total Exosome RNA and Protein Isolation Kit as per manufacturer's instructions (Invitrogen, Carlsbad, CA). Briefly, the isolated exosomes were dissolved in pre-warm 2 × denaturing solution followed by acid-phenol: chloroform extraction. The upper aqueous phase was precipitated with ethanol and total RNA was eluted with preheated (95 °C) elution buffer. The concentration of RNA was determined using the Nanodrop spectrophotometer (Nanodrop Technologies, Inc, Wilmington, Germany) as described 29 and were stored at − 80 °C until further use. www.nature.com/scientificreports/ expression profiling was performed using the nCounter Human ver 3.0 miRNA Panel on nCounter Analysis System (NanoString Technologies) as described 30 . A total of three cartridge chips were run at the same time each consisting of 12 samples (9 HIV-1 positive and 3 negative control per chip). For data analysis, HIV-1 positive and control samples were separately pooled. Raw NanoString counts were pre-processed and differential counts derived using the R package 'edgeR' (PMID: 19910308) as described 29,31 . Briefly, counts were normalized using trimmed mean of M-values (TMM) method and miRNA that were less than the geometric mean of negative control probes for more than half of the samples were removed; the final miRNA count for differential expression analysis was 267. Differential expression between groups was calculated using the function exactTest, which is analogous to Fisher's exact test, but adapted for overdispersed data (PMID: 19910308). Adjusted p-values were derived using Benjamini-Hochberg's procedure for controlling false discovery rate. In order to predict the role of these miRNAs as a biomarker, Receiver Operating Characteristic (ROC) curves were generated for the top five miRNAs and their areas under the curve (AUC) were calculated using the R package "plotROC" (P < 0.001).
In-silico bioinformatic analysis. For functional classification of miRNAs, DIANA-mirPath v3.0 was used for Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) annotation analyses as described 32 . DIANA-Tarbase and "Pathways Union" options were selected to perform a KEGG pathway analysis using this database of experimentally validated targets. For GO analysis, miRNAs belonging to specific GO categories based on the experimental findings with "Categories Union" was conducted. The significance of each functional annotation term was generated using a modified Fisher's exact test with P-value threshold of < 0.05  www.nature.com/scientificreports/ for KEGG pathways and < 1e−20 for GO analysis, respectively as described 32 . The candidate target genes of the identified miRNAs were predicted using TarBase v8.0 33 . The untranslated region (UTR) location was predicted by TargetScan v7.2 34 . In order to identify the potential interactions of the differentially expressed miRNAs in HM from HIV-1 infected women, miRNA-mRNA Network analysis was performed using Network Analyst software 29 where miRNA to gene interaction data were collected from well-annotated databases such as miR-TarBase v7.0, TarBase v7.0 and miRecords as described 35 . For degree, betweenness and shortest Path, "all but miRNA nodes" filter option was selected for the analysis.
Ethics approval and consent to participate. Written as well as informed consent for the collection of demographics, behavioral data, and biological samples were obtained from all study participants. The study was approved by the McMaster Research Ethics Board (REB Approval #08-176), CCI of Children's Hospital, Los Angeles, the institutional review boards of the University of Manitoba Hospital ethical review committee, University of Maryland Baltimore and Plateau State Specialist Hospital Nigeria Institutional Review Boards as described previously 25,27 . All clinical investigations were conducted according to the principles of the Helsinki Declaration.

Results
Clinical characteristics of women participants. HIV-1 infected and uninfected women were recruited from the Plateau State, Nigeria as described previously 27,36 A total of 36 HM samples from the first week postpartum (27 HIV-1 positive and 9 HIV-1 negative as controls) were processed for HM exosome miRNA profiling, as shown in the study layout in Fig. 1. The characteristics of the study population included in the current analyses  Table 1. Of the 27 HIV-1 positive mothers, 22 were infected with HIV-1 for 4-15 years whereas, 5 women were infected with HIV-1 for only 3 years. All HIV-1 positive women were receiving ART according to the regimen set by Nigerian Government and the WHO 37 and had CD4+ count ≤ 300/mm 3 with undetectable viral load. It was impossible to obtain samples from ART-naïve HIV-1 positive women. Number of years on ART were counted, the day a woman was diagnosed positive for HIV-1 and placed on ART. As shown in Table 1, 13 out of 22 HIV-1 positive women showed high infant mortality compared to uninfected. Additionally, infants born to HIV-1 positive women were found to be stunted and underweight similar to what has been described previously 25 .
Milk exosome characterization and RNA quality check. Exosomes were isolated from individual HM samples and confirmed by TEM. HM derived exosomes were 30-100 nm in size and largely spherical in shape ( Fig. 2A). HM derived exosomes either freshly isolated or kept at room temperature for two days were confirmed using protein marker CD81 in a western blot analysis (Fig. 2B, Suppl Fig. S1). Next, exosomal RNA was isolated from HIV-1 positive and negative HM samples with an average yield of 40 ng/µl. The isolated RNA showed distinctive spikes of noncoding RNA bands at > 25 nucleotides (Fig. 2C,D).
GO and KEGG pathways. KEGG pathway analysis on the predicted targets led to the identification of 31 significant pathways in which the predicted miRNA targets were enriched (Fig. 4). Specifically, miRNA targets associated with HIV-1 belonged to multiple pathways such as pathways in cancer, viral carcinogenesis, adherens junctions, TGF-β, fatty acid biosynthesis, p53 signaling, cell cycle, pathways regulating pluripotency of stem cells and proteoglycans in cancer (Fig. 4). Next, we performed GO analysis to identify the biological processes associated with the miRNAs. A total of 30 GO biological processes were observed (Fig. 5). The highest enrichment GO terms targeted by these miRNAs included biosynthetic process followed by viral process, catabolic process, cell death, ion binding, membrane organization, mitotic cell cycle, RNA metabolic process, poly (A) RNA binding and neurotropin TRK receptor signaling pathway (Fig. 5).

miRNA-gene interaction network.
To understand the association of differentially expressed miRNAs in HIV-1 infected HM and their target proteins, miRNA-gene interaction network was generated using miRNet tool. The 19 differentially expressed miRNAs were uploaded into miRNet platform and miRNA-gene interactions were observed which generated 4190 target nodes and 6042 edges. Shortest path filter with "all but miRNA www.nature.com/scientificreports/ nodes" generated 124 nodes with 105 targets and 393 edges (Fig. 6). The top cluster hubs included miR-16-5p followed by miR-497, miR-93-5p, miR-30a-5p and miR-23a-5p. The biological functions were determined within the "reactome database" using the "hypergeometric test" algorithm and P-value < 0.05. Results showed that prenotch transcription and translation (TP53, E2F3, AGO2, CCND1) was the top group followed by mitotic G1/S phase (Wee1, CDK6 (Fig. 7). Furthermore, when miR-630 and miR-378g were combined, it yielded ROC AUC of 0.86 (95% CI 0.72 to 0.86) (Fig. 7) suggesting that miR-630 and miR-378g could serve as biomarkers to distinguish HIV-1 infected HM from non-HIV-1 infected HM.  13 . Using the search term HIV, we identified that miR-378g has one target site in 3′ UTR of TARBP2 (ENST 00000552857.1) from 382 to 388 (Fig. 8A). TARBP2 is known to promote HIV-1 LTR expression and viral production whereas its siRNA-mediated knockdown inhibits HIV-1 LTR expression and viral production 55 . A schematic of the hypothetical layout is shown in Fig. 8B, where we speculate that miR-378g mediated RNA interference would lower HIV-1 expression and viral production essentially as previously described 56 .

Discussion
HIV-1 is known to cause dramatic changes in cellular miRNA expression profiles 57-60 , however its effect on HM derived exosomal miRNAs remains unknown. Here, we characterized miRNA expression profiles of HM exosomes derived from HIV-1 infected HM and showed that HIV-1 infection significantly altered the expression levels of exosomal miRNAs. Analysis of differentially expressed miRNAs by a gene ontology and KEGG pathway-based approach revealed several biological processes are affected by HIV-1 infection. Furthermore, we identified two dysregulated miRNAs that can potentially discriminate HIV-1 positive HM from uninfected HM, with good predictive power. Collectively, these data provide, for the first time, comprehensive insight into HM exosomal miRNA profiles involved during HIV-1 infection.  20,59 . Importantly, exosomes may act at different levels of HIV-1 pathogenesis by modulating immune responses, infectivity or possibly activating latent viral reservoirs 9 . Indeed, the impact of exosomes on HIV-1 has been suggested as a potential strategy to cure HIV-1 infection and/or therapeutic 61 . Currently, only a few reports exist that demonstrate the immunemodulatory functions of HM derived exosomes 62 which may partly be due to the methodological limitations in their isolation and purification 8 . In this study, we have provided a successful HM exosome isolation method that will potentially aid future studies related to HM derived exosome characterization and mechanism.
To gain more insight into HM derived exosomal miRNAs modulation by HIV-1, we performed NanoString miRNA profiling and showed that HIV-1 perturbed the expression levels of 19 miRNAs (FC > 1.3; P < 0.05; Table 2). Further, we identified 31 KEGG pathways potentially regulated by miRNAs including pathways in cancer, TGF-β, fatty acid biosynthesis, FoxO signaling, p53 signaling, cell cycle, pathways regulating pluripotency of stem cells and proteoglycans in cancer. miRNA-mRNA network showed differentially expressed miRNAs are linked to each other via their target genes. Furthermore, miRNA-mRNA network analysis, in addition to cell cycle, apoptosis identified the involvement of NOTCH and EGFR pathways.
A combination of two miRNAs (miR-630 and miR-378g) had 86% accuracy rate in predicting HIV-1 infection which may serve as a biomarker for segregating HIV-1 positive HM from uninfected HM. The stability of HM exosomes at room temperature also raises the possibility of their utility in initial screening processes prior to HIV-1 specific blood testing in low-to-middle income countries. Mothers living with HIV-1 for 3 and 4-15 years showed no significant fold change differences in miRNA expression levels (Suppl Tables S2, S3). Furthermore, the majority of women who participated in the study were carrying HIV-1 for greater than 5 years (Table 1), thus it is suggested these two miRNAs could be used to monitor AIDS progression in HIV-1 infected women. Indeed, our finding is in agreement with a previous study in which miR-630 was reported as a biomarker in chronic progressors of HIV-1 63 . Interestingly, our HM exosomal miRNA data from the South Africa cohort where HIV-1 infected women were carrying HIV-1 load < 1 year (data not shown) showed downregulation of these miRNAs identified in the current study and correlated with a recent report where it was shown that acute HIV-1 leads  65 . Some of the miRNAs we have identified have previously been implicated in HIV-1 infection including miR-630, miR-4516, miR-16-5p, miR-378, miR-93, miR-23, miR-30a 57,63,66-69 thus strongly suggesting the reliability of the NanoString data obtained in the current study. Further, it has been described that NanoString can perform miRNA profiling with digital precision and the results do not require further validation by another method 70 . miRNA-630 causes apoptosis by targeting BCL2, BCL2L2 and IGF-1R 71 or maintains the apoptotic balance by targeting multiple modulators 72 . www.nature.com/scientificreports/ miR-15a/b, miR-16, miR-20a, miR-93, miR-106b have been shown to bind Pur-α and repress its expression 68 . Pur-α is a cellular partner for Tat regulatory protein of HIV-1 and facilitates its transcriptional activity 73,74 and is required for HIV-1 infection in macrophages 68,75 . We found that Pur-α is a target of HM exosomal miR-93-5p and miR-497-5p. Whether HM exosomal miR-93-5p and miR-497-5p lower the R5-tropic HIV-1 infection of macrophages is not known but will be intriguing to investigate in future studies. HM derived immunomodulatory factors are transferred from mothers to infants via breastmilk which include immunoglobulins, cytokines, chemokines, growth factors, hormones, lactoferrins and Toll-like receptors 26,27,[76][77][78][79][80] . TGF-β is a major cytokine in HM that favors preferential MTCT of R5-tropic HIV-1 81,82 . miR-378g, miR-16-5p and miR-497-5p were found to target SMAD2 which has previously been shown to mediate TGF-β and regulate multiple pathways such as cell proliferation, apoptosis and cellular differentiation 81,83 . HIV-1 positive human skim milk fraction after heat inactivation and proteolytic digestion retains HIV-1 inhibitory activity and was shown to significantly inhibit oral HIV-1 transmission in-vivo 3 . Since, HM exosomal miRNAs are known to reach the systemic circulation 62 , these data suggest the HM derived exosome containing miRNAs reported herein may reach the fetal systemic circulation via breastmilk and play an important role in lowering MTCT of HIV-1 7 . Further future studies are required to elucidate the functional role of these HM exosomal miRNAs.
Using a consensus scoring approach, it has been shown that miR-378 targets HIV-1 envelope gene 67 . mir-378 family consist of 11 mature miRNA members according to miRbase database (www.mirba se.org) comprising of miR-378a-5p, miR-378a-3p, miR-378b, miR-378c, miR-378d, miR-378e, miR-378f, miR-378g, miR-378h, miR-378i and miR-378j; however, this study is the first to indicate the involvement of miR-378g in HIV-1 infection. Furthermore, prediction analysis suggested miRNA-378g targets a site located in 3′ UTR of TARBP2 from 382 to 388 nucleotides (Fig. 8), which is required for HIV-1 expression and virion production 55,84 . Interestingly, astrocytes are shown to be resistant to HIV-1 infection due to low endogenous levels of TARBP2 85 . TARBP2 was originally identified as a protein that binds to the 59 nucleotides conserved TAR element found at the 5′ and 3′ ends of all HIV-1 transcripts and enhances its translation and replication 56,84,86 . In addition, miR-378g was also found to target 4 confidently annotated sites in 3′ UTR of human HIV-1 enhancer binding protein 3 (HIVEP3) (ENST00000372583.1) located at 582-588, 2026-2032, 2208-2214 and 2580-2586 nucleotides (data not shown). Although, the role of HIVEP3 in HIV-1 replication is not clear, targeting HIVEP2 by miRNAs is known to reduce HIV-1 replication 87 . The effect of miR-378g on HIV-1 replication and MTCT must be investigated in future studies.
HIV-1 associated neurologic disease (HAND) occurs in more than 25% of HIV-1 infected patients who develop AIDS 88 . Previously, miR-4516 has been shown to be a biomarker of HAND in HIV-1 infected patients 66 . Our data showed that miR-4516 is upregulated in 4-15 years HIV-1 infected HM which may suggest HIV-1 infected women were HAND-asymptomatic or already had developed HAND. Although, our clinical data did not collect any neurologic symptoms in these infected women, it will be interesting to monitor the immune status as well as the temporal expression pattern of miR-4516 in future studies.
In conclusion, these data are the first to characterize the expression of HM exosomal miRNAs in HIV-1 infected HM. Given the use of current ART in HIV-1 positive mothers does not completely mitigate MTCT of HIV-1, interventions including the use of exosomal miRNA in addition to available ART may be required to prevent new infections of HIV-1 in infants. In this context, our data detailing HM exosome miRNAs could potentially be exploited to lower MTCT HIV-1 transmission. Moreover, the miRNAs reported herein may serve as potential biomarkers of HIV-1 infected HM.

Data availability
The datasets reported and analyzed in the current study are available in NCBI Gene Expression Omnibus (GEO) repository and are accessible through GEO series accession number GSE143039.