Plasma MicroRNA Profiling of Plasmodium falciparum Biomass and Association with Severity of Malaria Disease

Severe malaria (SM) is a major public health problem in malaria-endemic countries. Sequestration of Plasmodium falciparum–infected erythrocytes in vital organs and the associated inflammation leads to organ dysfunction. MicroRNAs (miRNAs), which are rapidly released from damaged tissues into the host fluids, constitute a promising biomarker for the prognosis of SM. We applied next-generation sequencing to evaluate the differential expression of miRNAs in SM and in uncomplicated malaria (UM. Six miRNAs were associated with in vitro P. falciparum cytoadhesion, severity in children, and P. falciparum biomass. Relative expression of hsa-miR-4497 quantified by TaqMan-quantitative reverse transcription PCR was higher in plasma of children with SM than those with UM (p<0.048) and again correlated with P. falciparum biomass (p = 0.033). These findings suggest that different physiopathological processes in SM and UM lead to differential expression of miRNAs and pave the way for future studies to assess their prognostic value in malaria.

assess adhesion by light microscopy. After washing with water, cells were stained with 10% Giemsa. iE and niE bound were counted in 6 different wells/assay in >500 nuclei cell/well.
Results were presented as the number of adhered iE per 500 nuclei of cells. Estimation of Pf adhesion to purified receptors (CD36, CD54, and g1CqR) as well as platelet-mediated (PM)agglutination and rosetting was performed as described elsewhere (2,3). Cytoadherence was defined as positive only if the number of iEs bound per mm 2 > the mean binding +2 SD to Duffy-Fc coated petri dishes. Pf isolates were considered positive for PM clumping if the frequency of clumps was higher in the presence of platelets than in buffer-control and for rosetting if the frequency of rosettes was >2% (2,4).

Small RNA Sequencing
Before RNA extraction, the level of hemolysis in plasma samples was assessed by spectrophotometry (EPOCH, BioTek) at a wavelength of 414nm (absorbance peak of free hemoglobin). Samples were classified as nonhemolysed if the optical density at 414nm <0.2 (5).
RNA was extracted from cell-conditioned media (3 mL) using the miRNeasy tissues/cells kit and plasma samples (1 mL) using miRNeasy plasma/serum kit (both QIAGEN, https://www.qiagen.com), with the use of 5µg UltraPure glycogen/sample (Invitrogen, Thermo Fisher). Given that the plasma samples were conserved in heparin, RNA was precipitated with lithium chloride (LiCl) as described elsewhere (6). Purified RNA quality and quantity were determined using the Bioanalyzer (Agilent Technologies, https://www.agilent.com) followed by preparation of libraries using NEBNext Small RNA Library Prep Set for Illumina (New England Biolabs, https://www.neb.com), then separation of libraries in polyacrylamide gels (Novex; Invitrogen). The Bioanalyzer was again used to quantify and assess the size of the libraries.
Further, libraries were pooled at the same equimolar concentrations and no more than 18 libraries were sequenced in the same lane using a HiSeq 2000 (Illumina) platform following the protocol for small RNAs (7).
A previously published pipeline was used to assess the sequencing quality, identification, and quantification of small RNAs and normalization (7). First, a quality control (QC) was conducted using FASTX-Toolkit and FastQ Screen. After adaptor removing, reads with the following features were removed: reads <18 nt; mean PHRED scores <30; and low complexity reads based on the mean score of the read. Good quality reads were then annotated to main RNA categories (tRNA, rRNA, and miRNAs), and miRNA complexity was estimated as the number of distinct miRNAs observed in each sample. Finally, contamination with RNA from other species was evaluated by mapping reads to clade-specific mature miRNA sequences extracted from miRBase v21 (8). The tested species categories include animal sponges, nematodes, insects, lophotrochozoan, echinoderms, fish, birds, reptiles, rodents, and primates.
Sequences that passed the QC were subjected to the seqBuster/seqCluster tool that retrieves miRNA and isomiRs counts (9,10). To detect miRNAs and isomiRs, reads were mapped to the precursors and annotated to miRNAs or isomiRs using miRBase version 21 with the miraligner (9). DESeq2 R package v.1.10.1 (R version 3.3.2) (11) was used to perform an internal normalization in which the counts for a miRNA in each sample were divided by the median of the ratios of observed counts to the geometric mean of each corresponding miRNAs over all samples.

Reverse Transcription Quantitative PCR
We used 50 µL of plasma with no hemolysis from the children recruited in 2014 for RNA extraction as described above. A synthetic RNA mimicking ath-miR-159a (Arabidopsis thaliana; Metabion, http://www.metabion.com) was added after lysis reaction at a final concentration of 1.5 pM. cDNA synthesis and RT-qPCR (ABI PRISM 7500 HT Real-Time System; Applied Biosystems) were performed using the TaqMan Advanced miRNA assays. A standard curve of 5 serially diluted points was prepared with cDNA of 6 randomly selected samples and run in triplicate for each miRNA. Results were normalized using a combination of endogenous controls (ECs). The selection of ECs was based on the following criteria: a) reported in scientific literature as previously used as ECs (12,13), b) coefficient of variance (CV) of normalized counts across all samples <5%, c) basemean ≥3000, d) SD <1, and e) log2fold change between SM and UM patients <1. Finally, the best 2 ECs tested as housekeepings using the NormFinder (14) were used for normalization of RT-qPCR data. miRNA relative expression levels (RELs) were calculated with the 2 −ΔCt method, where ∆Ct = [Ct (miRNA) -Mean Ct (ECs)], considering efficiencies of 100% for all the miRNAs and ECs (12).

In silico Analysis
The selected miRNAs were screened through 4 different gene target prediction programs: DIANA-microT-CDS (15), MiRDIP (16), MirGate (17), and TargetScan (http://www.targetscan.org/vert_71). Identified gene targets of each program were compared using an online tool, Venny2.1.0 (http://bioinfogp.cnb.csic.es/tools/venny). The gene targets that occurred in more than one database were selected and screened through the miRTarBase (18) online program to check if these genes have been experimentally validated previously. These gene targets were anticipated to be true positive targets present at detectable levels in field samples. The identified gene targets were further analyzed by DAVID 6.8 using Homo sapiens as the reference species. Genes were clustered to Gene Ontology terms and KEGG pathways (fold enrichment >1.5, p<0.05).

Statistical Analysis
Differential expression of miRNAs and isomiRs was assessed using DESEq2 and IsomiRs packages in R (9,10), which use negative binomial generalized linear models adjusted for multiple testing with the false discovery rate (FDR) by the Benjamini-Hochberg method (19).
Those with an FDR <5% were selected for posterior analysis. Analysis of the modification in the bases of the seed region was carried using isomiR package to determine a possible change in the target messenger RNAs. We performed Mann-Whitney U test to compare continuous data and χ 2 tests to compare categorical data. Spearman correlation analysis was performed to assess the correlation of miRNA RELs (log transformed) with log transformed HRP2 levels. A two-sided p < 0.05 was considered statistically significant. All statistical analyses were performed using R 3.3.2 in Linux-based system and graphs were prepared with GraphPad.
Appendix Table 1. Characteristics of microRNAs detected in cell-conditioned media of human brain endothelial cells exposed to Plasmodium falciparum−infected and noninfected erythrocytes*

Sample
Total reads