Genome-wide sequencing of cellular microRNAs identifies a combinatorial expression signature diagnostic of sepsis

Abstract

Rationale: Sepsis is a common cause of death in the intensive care unit with mortality up to 70% when accompanied by multiple organ dysfunction. Rapid diagnosis and the institution of appropriate antibiotic therapy and pressor support are therefore critical for survival. MicroRNAs are small non-coding RNAs that play an important role in the regulation of numerous cellular processes, including inflammation and immunity.

Objectives: We hypothesized changes in expression of microRNAs during sepsis may be of diagnostic value in the intensive care unit (ICU).

Methods: Massively parallel sequencing of microRNAs was utilised for screening microRNA candidates. Putative microRNAs were validated using quantitative real-time PCR (qRT-PCR). This study includes data from both a training cohort (UK) and an independent validation cohort (Sweden). A linear discriminant statistical model was employed to construct a diagnostic microRNA signature.

Results: A panel of known and novel microRNAs were detectable in the blood of patients with sepsis. After qRT-PCR validation, microRNA miR-150 and miR-4772-5p-iso were able to discriminate between patients who have systemic inflammatory response syndrome and patients with sepsis. This finding was also validated in independent cohort with an average diagnostic accuracy of 86%. Fractionating the cellular components of blood reveals miR-4772-5p-iso is expressed differentially in monocytes. Functional experiments using primary human monocytes demonstrate that it expressed in response to TLR ligation.

Conclusions: Taken together, these data provide a novel microRNA signature of sepsis that should allow rapid point-of-care diagnostic assessment of patients on ICU and also provide greater insight into the pathobiology of this severe disease.

Figure 1. Fold Changes of microRNA sequencing data and candidate microRNA expression validated by qRT-PCR.

(A) Each bar represents fold increase or decrease in the number of reads for one microRNA candidate comparing sepsis group with SIRS group; grey bar represents fold change data from septic patients with low CD64 expression whereas black bar represents patients with high CD64 expression. A pool of 4 whole blood samples for each group was used for sequencing experiments; (B) Box and whisker plot shows microRNA expression in raw Ct values (Y axis) of 7 candidate microRNAs (X axis); n = 61; (C) to (I) microRNA levels of 7 candidates and (J) mRNA level of IL-18RAP (normalized to GUSB) in samples from healthy donors (n = 17), sepsis patients (n = 22) and SIRS patients (n = 22) detected by qRT-PCR. Kruskal-Wallis ANOVA test was applied, * significant at p<0.05, ** significant at p<0.01, *** significant at p<0.001.

Figure 2. Simplified diagram illustrates the genomic location of miR-4772 family on Chromosome q12.1 within intron 5 of IL-18 RAP.

Figure 3. Construction and Validation of a Linear Discriminant Model for diagnosing sepsis from SIRS.

(A) LDA score was achieved using linear discriminant analysis (LDA) based on results of miR-4772-5p-iso and miR-150 in UK cohort; Mann-Whitney U test were applied. *** significant at p<0.001; (B) Prediction plot based on LDA score (X axis) and re probability of sepsis (Y axis), Red dots: Sepsis, Black dots: SIRS; ROC Curves demonstrate the diagnostic capacities of (C) miR-150 alone, (D) miR-4772-5p-iso alone and (E) LDA score; (F) miR-150 expression and (G) miR-4772-5p-iso expression of 2 patient groups from both UK and Sweden, Kruskal-Wallis ANOVA test was applied; ROC curves show the diagnostic power of (H) miR-150 and (I) miR-4772-5p-iso alone; (J) LDA score from Swedish cohort.

Figure 4. Expression of miR-4772-5p-iso in CD14+ monocytes between sepsis and SIRS compared with CD14 depleted PBMCs, and expression levels in healthy monocytes stimulated by TLR ligands.

(A) Small RNA sequencing of CD14+ monocytes vs CD14 depleted PBMCs for miR-4772-5p-iso expression in Healthy, Sepsis, and SIRS samples (experiments were performed using RNA from a pool of samples in each group, see Table S1.). (B) Fold changes of sequencing results based on the number of reads/million in Sepsis vs SIRS group; (C) qRT-PCR results of microRNA miR-4772-5p-iso expression in CD14+ monocytes from healthy donors stimulated with different TLR ligands for 24 hours; does and source of TLR ligands listed in methods section. qRT-PCR was done using triplicates. Fold changes were calculated based on RPMI medium alone (see method) then normalised to percentage of maximum effect in each experiment, error bars represents Standard Error of the Mean (SEM); n = 2.