Anti–miR-93-5p therapy prolongs sepsis survival by restoring the peripheral immune response

Sepsis remains a leading cause of death for humans and currently has no pathogenesis-specific therapy. Hampered progress is partly due to a lack of insight into deep mechanistic processes. In the past decade, deciphering the functions of small noncoding miRNAs in sepsis pathogenesis became a dynamic research topic. To screen for new miRNA targets for sepsis therapeutics, we used samples for miRNA array analysis of PBMCs from patients with sepsis and control individuals, blood samples from 2 cohorts of patients with sepsis, and multiple animal models: mouse cecum ligation puncture–induced (CLP-induced) sepsis, mouse viral miRNA challenge, and baboon Gram+ and Gram– sepsis models. miR-93-5p met the criteria for a therapeutic target, as it was overexpressed in baboons that died early after induction of sepsis, was downregulated in patients who survived after sepsis, and correlated with negative clinical prognosticators for sepsis. Therapeutically, inhibition of miR-93-5p prolonged the overall survival of mice with CLP-induced sepsis, with a stronger effect in older mice. Mechanistically, anti–miR-93-5p therapy reduced inflammatory monocytes and increased circulating effector memory T cells, especially the CD4+ subset. AGO2 IP in miR-93–KO T cells identified important regulatory receptors, such as CD28, as direct miR-93-5p target genes. In conclusion, miR-93-5p is a potential therapeutic target in sepsis through the regulation of both innate and adaptive immunity, with possibly a greater benefit for elderly patients than for young patients.


Isolation and activation of primary human T cells
Standard Ficoll-mediated isolation of PBMCs was followed by PAN-T cell isolation using the human Pan T cell Isolation Kit (Miltenyi Biotec, Cat# 130-096-535) according to the manufacturer's instructions. T cells were cultured and activated as described previously (15) in RPMI 1640 plus 10% human AB serum at 37°C with 5% CO2, together with 6.25L/1 million cells Dynabeads™ Human T-Activator CD3/CD28 for T Cell Expansion and Activation (Thermo Fisher, Cat# 11161D) and 50IU/mL human IL2 (R&D, Cat# 202-IL). IL2 supplemented RPMI 1640 plus 10% human AB serum medium was refreshed every 48h. Medium was refreshed, and cell density was assessed using hemacytometermediated counting every 48h.

RNA extraction from plasma, peripheral blood mononuclear cells, and whole blood
Total plasma, peripheral blood mononuclear cells (PBMC), white blood cells (WBCs), and whole blood RNA were extracted and reverse transcribed as previously described (16). RNA was obtained from 100 μL of plasma, PBMCs, or whole blood using the total RNA purification kit (NorgenBiotek, Cat. #37500) according to the manufacturer's instructions. At the end of the extraction process, RNA was eluted in 50 μL elution solution, and RNA concentrations and quality were determined using NanoDrop-1100. For the normalization of plasma sample-to-sample variation in the RNA isolation step, the C. elegans, cel-miR-39-3p and cel-miR-54-3p, (ThermoFisher SCIENTIFIC, Cat # A25576, and Cat #A25576), 25 fmol of each in a total volume of 1 μL, were used. The geometric means (17) of the Ct values of cel-miR-39-3p and cel-miR-54-3p were used for normalization. For the normalization of sample-to-sample variation of RNA extracted from PBMCs and whole blood, U6 was used as an endogenous normalizer.

RNA extraction from solid tissues (liver, kidney, heart)
Solid organs were homogenized in liquid nitrogen before RNA isolation in Trizol (Life Technologies Corporation, Carlsbad, CA, USA). Next, the RNA isolation included prolonged precipitation and centrifugation steps to preserve the small RNA fractions. Phase separation was made by incubation on ice for 30 min, followed by centrifugation at 12,000 g for 20 min at 4 °C. Total RNA samples were precipitated overnight on ice and centrifuged at full speed (21,000 g) for 30 min at 4 °C.
For the normalization of sample-to-sample variation of RNA extracted from organs, U6 was used as an endogenous normalizer. Genes: The High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Cat# 4368814) was used according to the manufacturer's instructions using 1μg RNA input in a 20μL reaction. Reverse transcription was performed using the following program: 25°C for 10min, 37°C for 2h, 85°C for 5min, and then 4°C on hold. cDNA was stored at −20°C until analysis.

RT-qPCR profiling
miRNAs: The diluted cDNA (3 μL) was used as a template in a quantitative PCR (qPCR) reaction with a total final volume of 5 μL. DNA amplification was performed using TaqMan primers specific for

Histological assessment of tissue injury
For quantitative analysis of tissue damage, we analyzed the lung, heart, kidney, liver, and spleen of 16-month-old CLP-mice treated with anti-miR-93-5p or scramble miRNA form the survival experiment. For each mouse, all organs were formalin-fixed and paraffin-embedded in a single block.
The heart tissue damage was scored using a modified method from the one described by Kishimoto et al. (18). Briefly, three different parameters were analyzed: myocardial necrosis, fibrosis, and immune infiltrates. Each of the three parameters was scored from 0 to 4 (0 = absent, 1 < 25% of the tissue, 2 = 25-50%, 3 = 50-75%, and 4 > 75%) on an ordinal scale with a step size of 0.5. The tissue damage score represents the sum of the three parameters.
The lung tissue pathological assessment was performed as previously described (19). Briefly, we scored from 0 to 3 (0 = absent, 1 = mild, 2 = moderate, 3 = severe) using an ordinal scale with a step size of 0.5 for the presence of exudates, hyperemia/congestion, immune infiltrates, intraalveolar hemorrhage/debris, and cellular hyperplasia. The tissue damage score represents the sum of all five parameters.
The kidney damage pathological scoring was performed similarly to the one used by Yu et. al. (20). Using a 0 to 4 ordinal scale with a step size of 1 scoring system (0 = none, 1 = 0-20%, 2 = 20%-50%, 3 = 50%-70%, 4 > 70%), we looked for tubules that displayed necrosis, loss of brush border, interstitial edema, vacuolization, tubule dilatation, and immune infiltrates. For the analysis of the tubules and glomeruli, the PAS and Jones' silver stain were essential. The tissue damage score represents the sum of the six parameters.
The liver damage was analyzed using a modified method from the one described previously by Martin et al. (21). Four parameters -immune infiltration, hypoperfiusion to necrosis, central vein congestion, and sinusoidal congestion were scored from 0 to 3 (0 = none, 1 = mild, 2 = moderate, 3 = severe) on an ordinal scale with a step size of 0.5 scale. The tissue damage score represents the sum of the four parameters.
For spleen tissue damage analysis, we used a modified version of the protocol reported by Karamese et al. (22). We analyzed five different parameters: increased immune infiltrate, apoptotic cells, hemosierin deposits, hemorrhage, and thrombi, which were scored from 0 to 4 (none -0, mild -1, moderate -2, severe -3, and more severe -4) on an ordinal scale with a step size of 0.5. The tissue damage score represents the sum of the four parameters.
Additionally, we noted an increasing number of PAS-positive foamy immune cells in the lungs and spleens of mice, and these were scored separately. We counted PAS-positive immune cells from five high-power fields with the most abundant PAS-positive cell infiltrate (40x magnification), and averaged the five data points, and reported as the number of PAS-positive macrophages/HPF. The slides were evaluated using an Olympus BX46 (Olympus Europe). Histological images were acquired with the PANNORAMIC 1000 digital slide scanner (3DHISTECH).

TUNEL assay
Because we morphologically observed apoptotic cells (shrinking and fragmentation of cells) in the spleens and lungs of mice, we used the TUNAL assay to further analyze apoptosis. For this purpose, we built two tissue microarrays (TMA), one including the spleens and a second one including the lungs of 8-month-old mice from the survival study. Representative areas with immune cells showing morphological features of apoptosis were identified on H&E slides and transferred to tissue microarrays (TMAs) with one core per sample. Each core measured 1 mm in diameter. Spleen and lung TMA sections of 4μm were used for the TUNEL assay by means of the HRP-DAB TUNEL staining kit (ab206386), and the slides were counterstained by methyl blue following the protocol provided by the TUNEL staining kit. For the quantification TUNEL assay, tissues were analyzed by quantifying at least 3-5 images at 40x magnification (HPF) per core. The data were reported as number of positive cells per HPF. The slides were evaluated using an Olympus BX46 (Olympus Europe). Histological images were acquired with the PANNORAMIC 1000 digital slide scanner (3DHISTECH).

CRISPR/Cas9-mediated miR-93 knockout
The method described by Rosenlund et al. (23) was used for generating miR-93 knockout clones from JURKAT cells and NB4 using two sgRNAs. Electroporation was performed using the Amaxa Cell Line Nucleofector Kit V (Lonza, Cat# VCA-1003) according to the manufacturer's instructions using program X-001. For knockout validation, DNA from single cell-derived clones was isolated using the Genomic DNA Mini Kit (IBI Scientific, Cat# IB47202). PCR was performed using GoTaq Green Master Mix (Promega, Cat# M712) using the following program: 95°C for 2min, 40 cycles of (95°C for 30sec, 60°C for 30sec, 72°C for 2min), 72°C for 5min, and then 4°C on hold, and PCR products were assessed on 3% agarose gel. Sanger sequencing was performed by the MD Anderson DNA sequencing core facility using both forward and reverse primers. Sequences were analyzed using ChromasPro V2.6.6.
Sequences of sgRNAs and PCR primer sequences for validation by PCR and sanger sequencing are provided in Supplementary Table S6.

AGO2-RIP-Chip
AGO2-RIP was performed as described previously (24). In brief, EZview™ Red Protein G Affinity Gel Beads (Sigma, Cat# E3403) were blocked with 5% BSA and 500 g/mL salmon sperm DNA (Sigma, D7656) in NT2 buffer (50mM Tris, pH7.4; 150mM NaCl; 1mM MgCl2, 0.05% Nonidet P-40) for 1h at 4°C. Then, 10 g of anti-AGO2 antibody or normal mouse IgG (Supplementary Table S6) was coupled to the affinity gel beads rotating overnight at 4°C. Beads were washed in NT2 buffer and resuspended in IP buffer (1xNT2 buffer, 40 U/L RNAseOUT, 5 L vanadyl ribonucleoside complex, 0.1 M DTT, 0.5 M EDTA). Per JURKAT clone and replicate, 30 million cells were lysed on ice in polysome lysis buffer (5 mM MgCl2, 100 mM KCl, 10 mM Hepes, pH7, 0.5% Nonidet P-40, 1mM DTT, 40 U/L RNAseOUT, and 1x Protease Inhibitor cocktail); the lysate was added to the beads in IP buffer. After an input sample was taken, the mix rotated overnight at 4°C before flow through and IP fractions were collected for RNA and protein extraction. RNA fractions were dissolved in Qiazol (Qiagen) and stored at -80°C until isolation; protein fractions were dissolved in Laemmli buffer (BioRad) substituted with 2mercaptoethanol, boiled at 100 °C for 10 min and stored at -20 °C until analysis. RNA was extracted using the miRNeasy mini (input fraction) or micro (IP fraction) kit (Qiagen, Cat# 217004 and 217084, respectively) according to the manufacturer's instructions. RNA quality was assessed using the Agilent Bioanalyzer 2100 using the Eukaryote Total RNA Nano Assay version 2.6. sequencing were performed by the Clinical Genomics Core at the OMRF. RNA-seq data processing followed the guidelines and standard practices of the ENCODE and modENCODE consortia. 75 bp single end raw sequencing reads, in a FASTQ format, were trimmed of residual adaptor sequences using Scythe software. Low-quality bases at the beginning or the end of sequencing reads were removed using sickle software, and then the quality of the remaining reads was confirmed with FastQC.
Trimmed quality reads were aligned to the Papio anubis (olive baboon) genome reference (PapAnu2.0) using STAR v2.4.0h, and transcript-level read counting was performed with HTSeq v0.5.3p9. Readcount normalization and differentially expressed analyses were performed using the edgeR package from Bioconductor. Expression values normalized with the variance modeling at the observational level (voom) function were analyzed for differential expression using the standard functions of the limma package. The group option of the voom function was used to accommodate the serial data collection design, and time series analysis methods, implemented in R package MasigPro, were used to characterize the deregulation of genes related to sepsis over time. Moderate t test p-values were adjusted for multiple testing using the false discovery rate (FDR) method, and FDR < 0.05 was used to filter significant differentially expressed transcripts between conditions and across time. Annotations for baboon transcripts and their human homologs were retrieved from the ensemble database using Biomart.

miRNA expression profiles from public resources
Global miRNA expression profiles (CCLE_miRNA_20180525.gct) were downloaded from the Cancer Cell Line Encyclopedia (CCLE; https://depmap.org/portal/download/). miR-93 expression was pooled by cancer subtype and displayed as normalized read counts.

Microarray analysis
A microarray analysis was performed for each clone (i.e., JURKAT parental, Control, miR-93 KO#1 and KO#2) in duplicate for both input and IP fractions using the Applied Biosystems human Clariom TM S assay (Cat# 902927) according to the manufacturer's instructions. Relative signal intensities were extracted using Affymetrix feature extraction software version 1.1.0.1567. The CEL files, generated from Affymetrix RNA microarray image analysis software, were processed through Transcriptome Analysis Console 4.0 which normalizes (and applies the log2 function to) array signals using a robust multiarray averaging algorithm.
Approximately 16,000 microarray probes that were expressed above background in both duplicates of either the parental, control, or both miR-93 KO samples were selected for further analysis.
Target genes of miR-93 were defined as being enriched in the AGO2-IP vs. the control IgG-IP fraction of the parental or control samples (i.e., miR-93-5p expressed) but not in either of the two generated knockout clones (i.e., miR-93-5p expression absent). Differentially expressed mRNAs in a comparative analysis were further identified by analysis of variance (eBay) with a P-value cutoff of 0.05 and a fold change more than 1.5 or less than -1.5.

MiRNA target prediction and pathway analysis
We obtained experimentally confirmed miRNA-mRNA interactions from miRTarBase

Western blotting
Protein concentration was quantified by the Bradford assay (Bio-Rad). In total, 20 ug of proteins were loaded on 4%-20% acrylamide Criterion TGXTM precast gels (Bio-Rad) and transferred to nitrocellulose membranes by the semidry method. The membranes were incubated overnight with the corresponding primary antibodies for MCM7 and GAPDH (normalizer) (Supplementary Table S6) and then incubated with the appropriate horseradish peroxidase-conjugated secondary antibody.
Immunoreactivity was detected by incubation with ECL SuperSignal West Femto Substrate (ThermoFisher Scientific), and detected by the autoradiographic film.

Statistical analysis
Statistical analysis and graphical representation were performed with GraphPad Prism (version 8.1.2). P values < 0.05 were considered statistically significant. The Shapiro-Wilk normality test was performed for each group to assess whether the data were normally distributed. For normally distributed data, a 2-tailed t test was used to compare mean values between different groups. When one of the groups did not pass the normality test, the Mann-Whitney-Wilcoxon nonparametric test was used to assess statistical differences between the different groups. For samples with matched repeated measures, we performed the ANOVA test with Geisser-Greenhouse correction for normally distributed data and the Friedman test if the data did not pass the normality test. Correlation analyses between expression data and clinical parameters were performed using the Pearson correlation test. We adjusted for multiple testing using the false discovery rate (FDR) (26).