The involvement of regulatory non-coding RNAs in sepsis: a systematic review

We searched articles indexed in PubMed (MEDLINE), EMBASE and Europe PubMed Central databases using Medical Subject Heading (MeSH) or Title/Abstract words (“microRNA or miRNA or lncRNA or circRNA” and “sepsis or septic shock”) from inception up to 30 Sep 2016. There were no limitations imposed on language or type of study. We included original research articles in which the role of host-derived regulatory RNAs (miRNA, lncRNA, or circRNA) in sepsis was examined in relation to disease pathogenesis, diagnosis, prognosis, and treatment. Investigation on exogenous regulatory RNAs or non-original research articles such as review articles, conference proceedings, editorials, and book chapters were excluded. Titles and abstracts were independently screened for relevancy by two authors. Disagreement was resolved by consensus or consultation with senior authors.

Data were extracted in terms of the role of non-coding RNAs (i.e., non-coding RNA species investigated, laboratory detection methods, and cellular pathways) and their use as biomarkers or therapeutic agents. The following data were abstracted: (1) first author and year of publication; (2) type of study; (3) non-coding RNA species investigated; (4) methods used to detect the corresponding non-coding RNA; (5) number of replicates/specimens (in vitro and in vivo studies) or patients (clinical studies); (6) cellular pathways involved; and (7) major conclusions. Two researchers independently performed the data extraction and evaluated the quality of the included studies using a modified guideline from the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE), a tool based on the Cochrane Collaboration Risk of Bias tool [31]. One item concerning random housing of animals was removed in the modified version. This tool contains nine items assessing selection bias, performance bias, detection bias, attrition bias, and reporting bias. These factors are common amongst in vitro, in vivo, and human studies. The higher the SYRCLE score was, the better the quality of the study would be. The maximum achievable score is 9.

This literature review indicates that the pattern of non-coding RNA expression differs substantially upon encountering various microbial moieties [38, 41]. Analysis of peripheral blood by quantitative RT-PCR and miRNA microarrays has been widely used for expression profiling of miRNA in septic patients [15, 16, 49, 50]. Notably, several miRNA species, including miR-126, miR-21, miR-16, and miR-27a, increased more than 30-fold in sepsis [15]. Single-candidate miRNA studies have established the association of miR-146a, miR-25, and miR-15a/16 with sepsis [17, 51‐54]. Further stratification of the systemic inflammatory response syndrome (SIRS) from sepsis revealed differential miRNA deregulation [16]. These results suggest that miRNA expression may be pathogen-specific and that its pattern could be used as biomarkers or therapeutic targets.

While regulatory RNAs have been recognized for more than a decade, their use as biomarkers for sepsis diagnosis and prognostication has not been thoroughly investigated until recently (Table 1). Microarray analyses, next-generation sequencing, and quantitative RT-PCR are important tools in developing biomarkers [15‐21, 50, 51, 53, 57, 71, 97, 107, 119‐131]. To date, candidate regulatory RNAs are limited to miRNAs. No study has evaluated the feasibility of using lncRNAs or circRNAs as biomarkers in sepsis.

Investigations revealed that the expression level of miR-25, miR-143, miR-146a, miR-15a, miR-16, miR-126, miR-150, miR-223, and 472-5p-iso could differentiate SIRS from sepsis [16‐21, 48‐51, 107, 132]. However, an independent research group could not detect any difference in miR-223 expression between septic patients and healthy controls [110]. Recently, massive screening using Solexa sequencing has identified nine novel miRNAs which are correlated with sepsis mortality (AUC = 0.681–0.863) [50]. Of note, selective reporting of sensitivity, specificity, and the associated AUC value was common in several included studies. In addition, none of these studies used an external cohort to validate the biomarkers investigated. This reporting bias complicates the analysis of results between studies.

Other miRNAs have been investigated to predict complications associated with sepsis. For instance, miR-122 predicts the development of liver injury in septic patients [70, 126]. miR-574-5p and miR-155 may predict the development of septic shock, immunosuppression and respiratory failure [71, 127, 133, 134].

Clearly, genome-wide profiling of miRNA expression distinguished septic from nonseptic patients. However, prediction of the likelihood of a nonseptic patient developing sepsis may be more clinically relevant to reduce mortality and morbidity in critical care settings.

Although cell and animal models have demonstrated the use of miRNA modulators in combating sepsis, considerable challenges have to be overcome in order to successfully translate these into clinical use. Apparently, the expression of miRNAs is tissue-dependent, questioning the appropriateness of systemic delivery of antagomir or miRNA mimics, as has been commonly performed in animal models. Although targeted drug delivery may be an alternative, this is further complicated by the technology of the delivery and the heterogeneity of clinical manifestations among septic patients.

Recent animal studies have recognized that miRNAs are associated with medical interventions and septic complications [125‐138]. In SPRET/Ei mice, glucocorticoid induced miR-511 upregulation, inhibiting the TNF receptor TNFR1 and, hence, reducing their sensitivity to TNF-α [135]. Similarly, administration of dexamethasone in a LPS-induced murine sepsis model downregulated the expression of miR-155 in the liver and alleviated proinflammatory cytokine production [116, 138]. Interestingly, miRNAs are also involved in cortisol nonresponsiveness, which may occur during the therapy. This resistance phenotype is partly related to altered expression of one of the cortisol receptor isoforms, glucocorticoid receptor α. In sepsis patients, this isoform is significantly downregulated by miR-124, which is increased three-fold upon exposure to glucocorticoid [137].

After exposure to LPS for a while, our body switches to tolerance mode, which avoids prolonged proinflammatory response. Silencing transcription and translation of acute inflammatory genes in vitro during LPS tolerance is mediated by various miRNAs [139]. Regulation of this tolerance status can be significantly disrupted by overexpression of miR-146a in cellular models of sepsis [140]. While sepsis-induced differential miRNA expression involves a diversity of miRNA species in multiple organs, modulation of miRNAs in endotoxin tolerance is seen predominantly in macrophages mediated by miR-146a and miR-155 [141, 142]. As in sepsis, in vitro studies suggested that the regulation of miRNA during tolerance involves the TLR-NFkB-cytokine pathways [143‐145].

Experimental evidence suggests that alternative therapies in treating sepsis involve modulation of miRNAs. Predominantly, these miRNA species are linked to some, if not all, pathways in innate immunity in cellular [146, 147] and in animal models [148, 149]. One of the well-known anti-inflammatory dietary components is flavonoids. In vitro investigation of a flavonoid, apigenin, revealed that it suppresses LPS-induced miR-155 expression in macrophages, leading to upregulation of the anti-inflammatory regulators forkhead box O3a and MAD-related protein 2 [146]. The protective effect was further elaborated by a murine sepsis model in which an apigenin-rich diet considerably reduced the expression of miR-155 and TNF-α in the lungs [146]. Recent innovations in septic treatment include stem cell therapy. In this regard, in vivo mechanistic studies revealed that mesenchymal stem cells improved survival of CLP-induced septic mice by downregulating miR-143 [147, 150]. Microarray analyses revealed more than 1.5-fold differential expression of 77 miRNAs in septic mice treated with noncultured-derived mesenchymal cells [149]. This was accompanied by a reduced inflammatory response and apoptosis [149]. Elucidating the mechanisms using animal models of sepsis in relation to a 20-HETE analog, N-(20-hydroxyeicosa-5Z,14Z-dienoyl glycine, revealed the involvement of miR-150, miR-223, and miR-297 [148]. Further in vitro studies revealed that these miRNAs were downregulated leading to suppression of the MyB88/NF-κB pathway [151].

Evaluation of the feasibility of miRNA as septic therapy predominantly employs two approaches: use of antagomir or miRNA mimics. Among all miRNAs, miR-146a is the most comprehensively studied candidate. By targeting IRAK1 and TRAF6, miR-146a attenuates cardiac dysfunction in septic mice [152]. Delivering miR-146a agomir by in vivo jetPEITM instillation into airways of septic mice inhibited proinflammatory cytokine production and alleviated lung tissue injury [153]. An independent group of researchers revealed that this miRNA additionally interferes in vitro with Th1 cell differentiation of human CD4⁺ T lymphocytes via PRKC [154]. Other miRNAs, including miR-124, miR-142-3p, and miR-195, have also been demonstrated to be useful in preventing hyperinflammation, apoptosis, and multiple organ injury in murine sepsis models [8, 140, 155, 156]. Similarly, indirect induction of miR-126 expression in vitro by CTEC-0214, a stromal cell-derived factor 1 alpha analog, preserved endothelial cell barrier integrity and attenuated pulmonary vascular leak [157].