Sepsis is regarded as arising from an unusual systemic response to infection but the physiopathology of sepsis remains elusive. At present, sepsis is still a fatal condition with delayed diagnosis and a poor outcome. Many biomarkers have been reported in clinical application for patients with sepsis, and claimed to improve the diagnosis and treatment. Because of the difficulty in the interpreting of clinical features of sepsis, some biomarkers do not show high sensitivity and specificity. MicroRNAs (miRNAs) are small noncoding RNAs which pair the sites in mRNAs to regulate gene expression in eukaryotes. They play a key role in inflammatory response, and have been validated to be potential sepsis biomarker recently. In the present work, we apply a miRNA regulatory network based method to identify novel microRNA biomarkers associated with the early diagnosis of sepsis. By analyzing the miRNA expression profiles and the miRNA regulatory network, we obtained novel miRNAs associated with sepsis. Pathways analysis, disease ontology analysis, and protein-protein interaction network (PIN) analysis, as well as ROC curve, were exploited to testify the reliability of the predicted miRNAs. We finally identified 8 novel miRNAs which have the potential to be sepsis biomarkers.
Sepsis is among the common causes of death in the intensive care units’ patients [
Over the past decade, sepsis has been considered as a hidden public health disaster [
Until now, there are many works reported to identify putative microRNA biomarkers [
The schematic workflow in our study for identifying miRNAs as potential sepsis biomarkers.
We conducted exhaustive search in Medline database with the key words “sepsis or severe sepsis or septic shock,” “miRNA or microRNA,” and “biomarker or marker or indicator.” Publication date (before October 31, 2013) and human studies were used as filters. We then extracted from each paper the relevant information of biomarkers, for example, microRNA name, accession number in miRBase [
The miRNA expression profiles were retrieved from EBI ArrayExpress (
To identify miRNAs of interest, we adopted student
We created union miRNA-mRNA interactions for human, which combine experimentally validated targeting data and computational prediction data. The experimentally validated data were extracted from miRecords [
Herein, we mapped the genes uniquely regulated by candidate miRNAs to GeneGo database for analysis of enriched signaling pathway and disease ontology [
Text mining in NCBI PubMed was used to identify miRNAs as sepsis biomarker. By setting the specific key words, we collated 10 miRNAs that were already proven to be helpful for diagnosis or prognosis of sepsis. To analyze common characteristics of 10 known biomarkers, the number of genes targeted exclusively by a specific microRNA in union miRNA-mRNA interactions database was conducted and we termed it as a novel out degree (NOD) to indicate the independent regulation power for an individual miRNA [
The details of sepsis miRNA biomarkers extracted from the literature.
MicroRNA name (Hsa-) | Accession number (MIMAT) | Biomarker type | Detection technology | Study design | Expression in sepsis patients | PMID | Reference |
---|---|---|---|---|---|---|---|
miR-15a | 0000068 | Diagnosis | qRT-PCR | Serum | Up | 22868808 | [ |
miR-16 | 0000069 | Diagnosis | qRT-PCR | Serum | Up | 22868808 | [ |
miR-122 | 0000421 | Diagnosis | qRT-PCR | Serum | Down | 23026916 | [ |
miR-146a | 0000449 | Diagnosis | qRT-PCR | Serum | Down | 20188071 | [ |
miR-223 | 0000280 | Diagnosis | qRT-PCR | Serum | Down | 20188071 | [ |
miR-483-5p | 0004761 | Prognosis | qRT-PCR | Serum | Downregulated in survivors | 22719975 | [ |
miR-499-5p | 0002870 | Diagnosis | qRT-PCR | Serum | Down | 23026916 | [ |
miR-574-5p | 0004795 | Prognosis | qRT-PCR | Serum | Upregulated in survivors | 22344312 | [ |
miR-150 | 0000451 | Diagnosis | qRT-PCR | Serum | Down | 19823581 | [ |
miR-193b* | 0004767 | Prognosis | qRT-PCR | Serum | Downregulated in survivors | 22719975 | [ |
The distribution of NOD value was compared between known miRNA biomarkers and all miRNAs in database. Though we constructed miRNA-mRNA interactions network, the number of genes targeted exclusively by a specific microRNA can be computed. So each miRNA has a NOD value. Kolmogorov-Smirnov test (K-S test) was used to test whether two underlying one-dimensional probability distributions differ. The above boxplot really highlights the difference between two samples. The
With the result above, we exploited miRNA expression profiles to predict disease biomarker. As described in Methods, we identified 10 significantly and differentially expressed miRNAs to be candidate sepsis miRNA biomarkers from our selected miRNA expression dataset. Among these miRNAs, miR-16 [
The diagnostic potential of candidate miRNAs was evaluated by ROC curve analysis and the discriminatory accuracy was presented by AUC values. We found that the minimum of AUC is 0.81, the maximum is 0.97, and the average of 5 miRNAs’ AUC is above 0.90. Because the property of ROC is measured as area under the curve (AUC), the ROC curve comparing sepsis patients and healthy controls provides a graphical demonstration of the superiority of candidate miRNA as sepsis marker. Finally, we plot the false positive rate (1−specificity) versus true positive rate (sensitivity) of a test (see Figure
Candidate miRNAs with outlier activity in sepsis.
MicroRNA name (Hsa-) | Accession number (MIMAT) |
|
Fold change ( |
NOD value |
|
AUC value (95% CI) |
---|---|---|---|---|---|---|
let-7b | 0000063 | 0.020 | 85.93 | 53 |
|
0.81 |
miR-16 | 0000069 | 0.030 | 55.79 | 35 |
|
0.84 |
miR-15b | 0000417 | 0.001 | 192.07 | 33 |
|
0.95 |
miR-146a | 0000449 | 0.002 | −6.89 | 20 |
|
0.90 |
miR-210 | 0000267 | 0.023 | 1.64 | 15 | 0.0006 | 0.97 |
miR-340 | 0004692 | 0.021 | −1.18 | 11 | 0.0021 | 0.88 |
miR-145 | 0000437 | 0.021 | 13.03 | 11 | 0.0021 | 0.83 |
miR-484 | 0002174 | 0.002 | 3.74 | 11 | 0.0021 | 0.92 |
miR-324-3p | 0000762 | 0.021 | 2.45 | 10 | 0.0041 | 0.84 |
miR-486-5p | 0002177 | 0.019 | 102.49 | 8 | 0.0151 | 0.97 |
Receiver operating characteristic (ROC) curves of the 10 candidate miRNAs for their performance of diagnosis of sepsis.
Previous researches have revealed that microRNAs emerged as key gene regulators in diverse biological pathways [
For pathway analysis, we retrieved 29 significantly enriched pathways (
Pathway enrichment analysis for the target genes of the 10 candidate sepsis miRNA biomarkers. The uniquely regulated and targeted genes of the candidate sepsis miRNA biomarkers from our method were retrieved and annotated with analysis of pathway enrichment in GeneGo database. In total, 207 genes are uniquely regulated and targeted by the 10 candidate miRNA biomarkers. The statistical significance level
Disease ontology is created based on the classification in medical subject headings (MeSH). Each disease in disease ontology has its corresponding biomarker gene or set of genes. After mapping the uniquely regulated and targeted genes of candidate miRNA biomarkers, we noted that the most significant disease is septic shock. Septic shock is severe sepsis plus a state of acute circulatory failure characterized by persistent arterial hypotension unexplained by other causes despite adequate volume resuscitation [
Disease ontology analysis for uniquely regulated and targeted genes of the 10 candidate sepsis miRNA biomarkers. The uniquely regulated and targeted genes of the candidate sepsis miRNA biomarkers from our method were retrieved and annotated with disease ontology analysis. In total, 207 genes are uniquely regulated and targeted by the 10 candidate miRNA biomarkers. The statistical significance level (
MicroRNAs implement their function by regulating their target genes, thereby directly affecting expression of their target genes at the posttranscriptional level and the related protein-protein interaction network [
We constructed candidate miRNAs regulatory networks, containing miRNAs, genes exclusively targeted by them, and the genes directly connected to the targets. The extended network nodes were obtained by appending known interactions form the PINA database. Protein interaction network analysis (PINA) platform integrated protein-protein interaction data from six public curated databases containing 108477 binary interactions [
Summary of constructed 10 miRNA regulated PINs. N0: gene was included in PINA database; N1: the extended subnetwork of N0 gene directly connected to N0 gene; N2: the total genes of miRNA regulated subnetwork.
MicroRNA name (Hsa-) | Accession number (MIMAT) | NOD count | N0 count | N1 count | N2 count |
---|---|---|---|---|---|
let-7b | 0000063 | 53 | 42 | 424 | 466 |
miR-15b | 0000417 | 33 | 26 | 201 | 227 |
miR-16 | 0000069 | 35 | 28 | 384 | 412 |
miR-145 | 0000437 | 11 | 8 | 256 | 264 |
miR-146a | 0000449 | 20 | 13 | 202 | 215 |
miR-210 | 0000267 | 15 | 10 | 39 | 49 |
miR-324-3p | 0000762 | 10 | 10 | 121 | 131 |
miR-340 | 0004692 | 11 | 9 | 124 | 133 |
miR-484 | 0002174 | 11 | 11 | 246 | 257 |
miR-486-5p | 0002177 | 8 | 6 | 26 | 32 |
GO analysis results of miR-15b regulated PIN. The common GO terms for miR-15b were listed.
MIMAT0000417 (Hsa-miR-15b) | ||
---|---|---|
GO term | Genes |
|
GO:0006916~antiapoptosis | BFAR, HSP90B1, GSK3B, BCL2, HIPK3, TGFBR1, NPM1, UBC, SERPINB2, FAIM3, BCL2L1, HSPA5 |
|
|
||
GO:0009891~positive regulation |
DVL3, HRAS, THRB, GRIP1, PCBD1, RXRB, RXRA, TGFBR1, PPARG, DDX5, CALR, POT1, SREBF2, ATXN1, MAPK1, MEIS2, PSMC5, NCOA2, HNF4A, ATXN7, NPM1, UBC, YAP1 |
|
|
||
GO:0010557~positive regulation |
DVL3, HRAS, THRB, GRIP1, PCBD1, RXRB, RXRA, TGFBR1, PPARG, DDX5, CALR, POT1, SREBF2, ATXN1, MAPK1, MEIS2, PSMC5, NCOA2, HNF4A, ATXN7, UBC, YAP1 |
|
|
||
GO:0010604~positive regulation |
HRAS, THRB, GRIP1, PPARG, PSMD1, PSMD2, PSMD3, H2AFX, PSMD4, YAP1, PSMD6, PSMD7, PRKCA, PCBD1, RXRB, RXRA, PSMA2, UBE2N, MAPK1, NCOA2, HNF4A, PSMA6, PSMA3, UBC, MDM2, CALR, POT1, PIN1, PSMB5, MEIS2, BCL2, UBE2D1, DVL3, TGFBR1, DDX5, FURIN, SREBF2, ATXN1, PSMC6, PSMD14, PSMD13, PSMC5, PSMD12, PSMC4, PSMC3, PSMD11, PSMD10, ATXN7, PSMC2, PSMC1 |
|
|
||
GO:0010605~negative regulation |
THRB, TSG101, PPARG, BCL2L1, TERF2IP, CALR, POT1, PSMB5, MEIS2, NPM1, PSMD1, PSMD2, PSMD3, PSMD4, UBE2D1, PSMD6, PSMD7, PRKCA, RXRA, ZNF24, UBE2I, FURIN, CDK5, SIRT3, PSMA2, ATXN1, PSMD14, PSMC6, PSMD13, NCOA2, PSMC5, PSMA6, HNF4A, PSMD12, PSMC4, PSMC3, PSMD11, PSMD10, PSMC2, PSMA3, PSMC1, UBC, BUB1B, MDM2, FABP4, SMURF2 |
|
|
||
GO:0010628~positive regulation |
DVL3, THRB, GRIP1, RXRB, PCBD1, RXRA, TGFBR1, PPARG, DDX5, SREBF2, ATXN1, MAPK1, MEIS2, PSMC5, NCOA2, HNF4A, ATXN7, UBC, YAP1 | 0.0031 |
|
||
GO:0010941~regulation of cell death | HRAS, BCAR1, BCL2L1, CALR, ITSN1, DYNLL1, BCL2, SOS1, CASP8, RAC1, NPM1, POU4F1, HSPA5, PRKCA, VAV3, TP53BP2, TGFBR1, TMBIM6, RXRA, ACTN1, ACTN2, FURIN, VAV1, CDK5, CASP10, MAPK1, BFAR, HSP90B1, PSMC5, GSK3B, HIPK3, UBC, SERPINB2, ERN1, FAIM3, MAPK8, CACNA1A |
|
|
||
GO:0031328~positive regulation |
DVL3, HRAS, THRB, GRIP1, PCBD1, RXRB, RXRA, TGFBR1, PPARG, DDX5, CALR, POT1, SREBF2, ATXN1, MAPK1, MEIS2, PSMC5, NCOA2, HNF4A, ATXN7, NPM1, UBC, YAP1 |
|
|
||
GO:0042981~regulation of apoptosis | HRAS, BCAR1, BCL2L1, CALR, ITSN1, DYNLL1, BCL2, SOS1, CASP8, RAC1, NPM1, POU4F1, HSPA5, PRKCA, VAV3, TP53BP2, TGFBR1, TMBIM6, RXRA, ACTN1, ACTN2, FURIN, VAV1, CDK5, CASP10, MAPK1, BFAR, HSP90B1, GSK3B, HIPK3, UBC, SERPINB2, ERN1, FAIM3, MAPK8, CACNA1A |
|
|
||
GO:0043066~negative regulation |
HRAS, TMBIM6, TGFBR1, BCL2L1, ITSN1, FURIN, BFAR, HSP90B1, GSK3B, HIPK3, BCL2, NPM1, UBC, SERPINB2, FAIM3, MAPK8, HSPA5, CACNA1A |
|
|
||
GO:0043067~regulation |
HRAS, BCAR1, BCL2L1, CALR, ITSN1, DYNLL1, BCL2, SOS1, CASP8, RAC1, NPM1, POU4F1, HSPA5, PRKCA, VAV3, TP53BP2, TGFBR1, TMBIM6, RXRA, ACTN1, ACTN2, FURIN, VAV1, CDK5, CASP10, MAPK1, BFAR, HSP90B1, PSMC5, GSK3B, HIPK3, UBC, SERPINB2, ERN1, FAIM3, MAPK8, CACNA1A |
|
|
||
GO:0043069~negative regulation |
HRAS, TMBIM6, TGFBR1, BCL2L1, ITSN1, FURIN, BFAR, HSP90B1, PSMC5, GSK3B, HIPK3, BCL2, NPM1, UBC, SERPINB2, FAIM3, MAPK8, HSPA5, CACNA1A |
|
|
||
GO:0045941~positive regulation |
DVL3, THRB, GRIP1, RXRB, PCBD1, RXRA, TGFBR1, PPARG, DDX5, SREBF2, ATXN1, MAPK1, MEIS2, PSMC5, NCOA2, HNF4A, ATXN7, UBC, YAP1 | 0.0022 |
|
||
GO:0060548~negative regulation |
HRAS, TMBIM6, TGFBR1, BCL2L1, ITSN1, FURIN, BFAR, HSP90B1, PSMC5, GSK3B, HIPK3, BCL2, NPM1, UBC, SERPINB2, FAIM3, MAPK8, HSPA5, CACNA1A |
|
The miRNA-210 regulated protein-protein interaction network (PPIN). In this network, red node denotes the miRNA, yellow nodes denote miRNA directly targeted genes, and green nodes denote genes connected with target genes. The red lines represent a negative regulatory relationship initiated by miRNAs. The black lines represent interactions between protein and protein.
Further studies are needed to confirm the relationship between 14 GO terms and sepsis. The 12 of 14 terms could be divided into two processes: one is cell death and the other is macromolecule biosynthetic process. As shown in Figure
The ancestor chart for common GO terms obtained from the GO analysis of the 10 candidate miRNAs. The grey circle represents GO term related to cell death process. The black rectangle represents GO term related to macromolecule biosynthetic process. All marked GO terms are included in the common GO terms.
In this study, we applied an integrative approach to identify microRNAs as sepsis biomarkers from miRNA expression profiles. Comparing with the work by Vasilescu et al., we identified 10 novel and reliable miRNA biomarkers for sepsis, supported by our pathways analysis, disease ontology analysis, and protein-protein interaction network analysis, as well as ROC curve comparison. These putative miRNA biomarkers could hopefully promote the precision diagnosis of sepsis.
The authors declare that there is no conflict of interests regarding the publication of this paper.
Jie Huang and Zhandong Sun contribute equally to this work.
This work was supported by Grants from Key Medical Subjects of Jiangsu Province (XK201120), Innovative Team of Jiangsu Province (LJ201114), Special Clinical Medical Science and Technology of Jiangsu Province (BL2012050 and BL2013014), Key Laboratory of Suzhou (SZS201108, SZS201307), and National Natural Science Foundation (81100371, 81370627, 81300423, and 81272143).