Study on identification of a three-microRNA panel in serum for diagnosing neonatal early onset sepsis

Abstract Background Comparing with other diseases, early onset sepsis (EOS) is a global health concern in neonatal period for its high morbidity and mortality rates. In recent years, many studies have contributed to the figure out the expression patterns of circulating micro-RNAs (miRNAs) in different diseases and progressions, which could function as diagnostic biomarkers for EOS. The purpose of this study was to analyze the expression patterns of selected miRNAs and evaluate their diagnostic value for early detection and treatment. Methods This was a prospective cross-sectional study conducted from 1 July 2021 to 30 June 2022. We collected surplus peripheral blood and demographic statistics of septic neonates and non-infected neonates during the first 24 h after delivery and obtained 11 candidate miRNAs by literature screening. First, we extracted the candidate miRNAs from the serum of selected neonates and analyzed their expression levels, and then the receiver operating characteristic (ROC) curve was used to select the differentially expressed miRNAs. We analyzed their sensitivity and specificity and obtained the best diagnostic panel. Finally, with the help of differentially expressed miRNAs, we performed gene ontology (GO) enrichment and protein–protein interaction (PPI) analyses by their target genes. Results In patients with EOS, three miRNAs (mir-223-3p, mir-15a-5p, and mir-17-5p) in serum were significantly downregulated, and mir-146a-5p, mir-1-3p, and mir-16-5p were upregulated. The diagnostic value of these miRNAs (miR-15a-5p, AUC = 0.67; miR-223-3p, AUC = 0.72; miR-16-5p, AUC = 0.68; miR‐17‐5p, AUC = 0.70; miR-1-3p, AUC = 0.69; miR‐146a‐5p, AUC = 0.72) was moderate, and the diagnostic panel constructed by miR-15a-5p, miR-223-3p, and miR-16-5p possessed a comparatively higher diagnostic value (AUC = 0.85, sensitivity: 74.6%, specificity: 86%), indicating that their combined application may be a promising biomarker for clinical diagnosis of EOS. According to GO enrichment analysis, most proteins encoded by target genes were located in the cytosol as for cellular component (CC), for molecular function (MF), most proteins acted as regulators in protein binding, and for biological process (BP). Most genes function in positive or negative regulation of transcription from RNA polymerase II promoter, and the top 10 hub genes were CDKN1A, YAP1, CCNE1, CCND1, CKK6, ERBB4, CHEK1, DICER1, VEGFA, and APP by rank degree after PPI construction. Conclusions The three-miRNA panels (miR-15a-5p, miR-223-3p, and miR-16-5p) may be a novel noninvasive biological marker for EOS screening.


Introduction
Early onset of neonatal sepsis is defined as an invasive infection that occurs within 72 h of birth.One study reported that the population-level estimate for neonatal sepsis was 2202 per 100,000 livebirths, and the incidence of mortality varied from 11% to 19% [1].The reproduction of pathogens could be found deriving from many different ways, such as intra-uterine infection, access to the vaginal flora, invasive venous catheterization, or long-term empirically administrating to broad-spectrum antibiotics.Among them, group B streptococcus (GBS) is the most common pathogen, but Escherichia coli is the one that lead to most common cause of death.The clinical manifestations range from ignoring jaundice and tachypnea to seizures, shock, anuria, apnea, and even sudden death, leading to hard precognition or high mortality [2], and the combination of some specific manifestations and index changes may be potential indicator of the onset of neonatal sepsis, like sepsis prediction score [3].Moreover, clinical biomarkers such as C-reactive protein (CRP), interleukins (ILs), procalcitonin, and white blood cell (WBC) count have been used for the prognosis of sepsis in early-stage, but many studies have revealed that they may have high sensitivity but poor specificity, which cannot alarm a timely treatment or mislead pediatricians in the diagnosis of early onset sepsis (EOS) to long-term antibiotic treatment, and the isolation of a pathogen from sterile body fluid needs 3-5 days for culturing and require -4 mL blood to meet the maximum positive rate.To avoid aggravation of clinically suspected sepsis, many neonates are treated with broad-spectrum antibiotics at the early disease stage, leading to increased risks of secondary infections, antimicrobial resistance, and disturbed intestinal flora.Therefore, novel biomarkers with higher specificity are needed for the early detection and accurate treatment of EOS.
By binding to various regions of transcripts, micro-RNAs (miRNAs), the non-coding RNA molecules with 20-23 nucleotides, could induce either degradation or suppression of translation of their target mRNAs.In recent years, a single miRNA has been found to target hundreds of mRNAs to regulate the expression of downstream or upstream genes, which may function in pathogenesis of many diseases such as allergy, infection, and carcinoma [4].Transported as exosomes or vesicles form and secrete into extracellular fluids, miRNAs could bind to the target cells or genes, which are extremely stable in different biospecimen of body even though in very small volumes.This feature of them could help their quantification and further analysis.Accordingly, the application of miRNAs as novel biomarkers to identify optimal combinations that could maximize the ability for timely diagnosing, treating, and outcome-predicting, which may be promising for clinical management, such as the early diagnosis of sepsis [5].In recent years, many researchers have attempted to correlate the expression of miRNAs with the diagnosis and prognosis of EOS using different methods, such as DNA microarray, quantitative reverse transcription (PCR), or RNA sequencing, and elaborate the mechanisms using cellular or animal tests [6][7][8].

Participants and ethics statement
During 72 h after birth, the septic group was diagnosed with EOS based on clinical tests and symptom evaluation, and the diagnostic criteria of EOS were as follows: (1) isolating pathogens from patients' sterile body fluids, such as blood or cerebrospinal fluid; (2) neonates have a clinical presentation, and cerebrospinal fluid analysis suggests central nervous system involvement; (3) the causative organism DNA or antigen is identified by polymerase chain reaction and other new technologies; (4) at least two inflammatory markers are abnormally high or low (WBC count, immature neutrophils, platelet count, CRP, and procalcitonin) [9], and the selected patients conformed to at least one of the above criteria.From 1 July 2021 to 30 June 2022, Peking University Shenzhen Hospital recruited 102 participants, including 43 septic group neonates and 59 control group neonates.Control patients were neonates without any risk of infection during delivery or in the uterus, and they were excluded from sepsis diagnosis based on symptoms and tests such as WBC, IL-6, CPR, and PCT.The inclusion criteria for the control group were like those: (1) admitted to the neonatology department of Peking University Shenzhen Hospital within 24 h after delivery; (2) length of patient stay was >3 days, and (3) exclusion from any type of infection.The inclusion criteria for the septic group were as follows: (1) age �3 days; (2) diagnosed sepsis with at least two positive screening tests.The exclusion criteria for both groups were as follows: neonates who died within 24 h after delivery, or their parents yielded to rescue or refused to sign the consent.This study was approved and reviewed by the Ethics Committee of Peking University Shenzhen Hospital, and the acquisition process of serum from neonates followed relative regulations established by the ethics committee.

Research design
First, by retrieving the PubMed and Gene Expression Omnibus (GEO) databases, a total of 11 miRNAs (hsa-mir-106b-5p, hsa-mir-122-3p, hsa-mir-125b-5p, hsa-mir-1-3p, hsa-mir-141-3p, hsa-mir-146a-5p, hsamir-150-5p, hsa-mir-15a-5p, hsa-mir-16-5p, hsa-mir-17-5p, and hsa-mir-223-3p) were selected from a large amount of literature, and peripheral blood samples were collected from 43 neonates with EOS and 59 controls.First, qRT-PCR and spiked-in normalization methods were used to clarify the expression pattern of candidate miRNAs.Then for the timely detection and intervention of EOS, we evaluated the diagnostics of each miRNAs as novel biomarkers through statistical analysis and built a diagnostic panel using a binary logistic regression model.Finally, we performed bioinformatics analysis, including target gene prediction and gene ontology (GO) enrichment analysis.Based on the above procedure, an accurate process flowchart for this study is shown in Figure 1.

Sample collection and RNA extraction
After obtaining informed consent from the parents, peripheral blood (0.5 mL) was collected from each neonate and separated by centrifugation at 3000 rpm for 10 min at 4 � C. The supernatant (the upper liquid layer of serum) from each sample was collected for subsequent experiments.To account for the variability generated during extraction and purification, we added 2 mL of 10 nmol/L synthetic Caenorhabditis elegans cel-miR-54-5p (RiboBio, Guangzhou, China) to each serum sample prior to the procedure.Total RNA (messenger RNA, miRNA, and other RNAs) was extracted from serum by TRIzol LS isolation kit (RiboBio, Guangzhou, China) following the manufacturer's protocol, resuspended in 30 mL RNase-free water, and stored at −80 � C for subsequent experiment, and the concentration of the all the RNA was measured by NanoDrop 2000 spectrophotometer (Thermo Fisher, Waltham, MA).

qRT-PCR
miRNAs were amplified using specific primers for reverse transcription from the Bulge-Loop miRNA qRT-PCR Primer Set (RiboBio, Guangzhou, China) and then performed real-time polymerase chain reaction using a SYBR Green qPCR kit (SYBR Pre-mix Ex Taq II, TaKaRa, Kusatsu, Japan) in 384-well plates at 95 � C for 30 s, followed by 35 cycles of 95 � C for 10 s, 60 � C for 20 s, and then 70 � C for 10 s.Finally, the relative expression change of miRNAs was calculated through the comparative 2 −DDCt method, and exogenous cel-miR-54-5p mimic was used as a spike-in control for further normalization.

Bioinformatic analysis
Mirwalk (http://mirwalk.umm.uni-heidelberg.de/)was used to identify miRNA target genes, including the predicted and validated genes.The cutoff criteria were as follows: (1) the predicted genes should be from all the databases of TargetScan, MiRDB, and miRTarBase; (2) the corrected p value <. 05.GO analysis of the selected genes was conducted using the DAVID database (https://david.ncifcrf.gov/)and visualized online (https://www.bioinformatics.com.cn/).Protein-protein interaction (PPI) networks were constructed using the STRING (https://cn.string-db.org)database, Cytoscape software (University of California, San Francisco, CA) was used to grade all genes, and the top 10 genes were visualized by degree.

Data analysis
All statistical analysis was performed using SPSS 21.0 (SPSS Inc., Chicago, IL) and GraphPad Prism 7.0 (GraphPad Software, Inc., La Jolla, CA).The normality of demographic and clinical characteristics was analyzed using the Shapiro-Wilk test.All the normally distributed data were exhibited as mean ± SD (by independent-sample t-test), and the abnormal distribution data were represented as first quartile-third quartile (by Mann-Whitney's U-test).The experimental differences between groups were estimated using one-way ANOVA.The diagnostic ability assessment of miRNAs was performed using receiver operating characteristic (ROC) curve and Kaplan-Meier's method.The cutoff points and their optimal sensitivity and specificity were identified using the Youden index (sensitivity þ specificity -1) and the diagnostic value of the panel was built using a binary logistic regression model and the Hosmer-Lemeshow test.

Participants' clinical and demographic characteristics
This experiment involved 43 neonates with sepsis and 59 controls, with a total of 102 patients.The demographic characteristics and clinical detail of the two groups are shown in Table 1.WBC, CRP, PCT, and IL-6 levels in the septic neonates were generally higher than those in the control neonates (p < .05),while there were no statistically different differences in weight, sex, and other parameters between two populations (p > .05).Gestational weeks may be a factor that increases the incidence of sepsis in term neonates, showing that longer gestation seems to be more likely to induce EOS (p < .05).

Differentially expressed miRNAs between two groups
In order to confirm whether the expression of candidate miRNAs was statistically significant or not in the circulating blood of all participants, the expression levels in the serum samples of all neonates were analyzed by qRT-PCR.As illustrated in Figure 2, the expression of serum mir-16-5p, mir-1-3p, and mir-146a-5p in septic patients was significantly higher than that in controls (p < .05).In contrast, mir-17-5p, mir-15a-5p, and mir-223-3p were significantly downregulated.The differential expression of these miRNAs suggests that they might be involved in the regulation of EOS.
Since the combined diagnostic capability of multiple miRNAs may be more accurate than that of a single one, we then constructed diagnostic panels for every three miRNAs and chose the panel with the highest AUC.The highest AUC value was 0.85 (95% CI: 0.77-0.92)for three miRNAs panel in Figure 4, the included miRNAs are follows: mir-15a-5p, mir-223-3p, and mir-16-5p.After the Hosmer-Lemeshow test, we estimated the goodness of fit and obtained a p value of .074(>.05), indicating a proper fit; the accurate OR value of each miRNA is shown in Figure 5, which shows that mir-16-5p may be the one that mostly distinguishes between sepsis and non-sepsis states from those miRNAs.We used the Youden index to calculate the best cutoff value and list the best specificity and sensitivity of these miRNAs for diagnosing EOS (Table 2), finding that mir-15a-5p, mir-1-3p, and mir-146a-5p had higher sensitivity than other miRNAs, ranging from 81.4% to 97.4%, but all six miRNAs had comparatively low specificity, ranging from 50.8% to 69.8%; in contrast, the diagnostic panel could not only provide a modest sensitivity of 74.6%, but a higher specificity of 86%.

Bioinformatics analysis of candidate miRNAs
Mirwalk was used to predict the possible target genes of the six miRNAs.According to the criteria above, a total of 442 genes were predicted and hsamir-15a-5p, hsa-mir-16-5p, and hsa-mir-17-5p could target hundreds of genes, but hsa-mir-223-3p, hsamir-1-3p, and hsa-mir-146a-5p target genes number ranging from 3 to 5 and miR-17-5p possessed the highest number of targeted genes, indicating that miR-17-5p may be a key regulator of sepsis-related gene expression.According to the GO enrichment analysis, in terms of cellular component (CC), most proteins encoded by targeted genes were located in the cytosol, nucleoplasm, cytoplasm, nucleus, and centrosome, as shown in Figure 6.As for cellular function (molecular function, MF), most of the genes acted as regulators of protein binding (approximately 159 genes) and biological processes (BPs), and most genes function in positive or negative regulation of transcription from RNA polymerase II promoter, others in the Wnt signaling pathway, cellular response to DNA damage stimulus, and other gene expressions.After PPI construction, the top 10 hub  genes were CDKN1A, YAP1, CCNE1, CCND1, CKK6, ERBB4, CHEK1, DICER1, VEGFA, and APP ranked by degree, as shown in Figure 7.

Discussion
Neonatal sepsis is a life-threatening condition characterized by interplay between the host immune system and invading microbial pathogens, resulting in severe systemic inflammation, which leads to multisystem failure [10].EOS is generally diagnosed based on a combination of clinical presentation and laboratory diagnosis, including sepsis prediction score, microbiological cultures, and infection markers, such as CRP, PCT, and WBC.However, the diagnosis of EOS can be delayed because of nonspecific clinical symptoms and infection markers [11].Owing to the limitations of culture-independent diagnostics and clinical manifestations, more diagnostics are needed for the early detection of EOS to reduce the burden of neonatal sepsis sequelae.Recent studies have found that miRNAs are short noncoding RNAs that are involved in posttranscriptional gene regulation, such as cell growth, development, and activity [8], indicating a massive potential for miRNAs as prognostic and diagnostic biomarkers in sepsis.In clinics, an operational biomarker should not only provide a proper accuracy  in recognizing the presence of neonatal sepsis in time, but in neonatal period, it is also urgently requiring the minimum consuming of blood and least invasive operation in neonates especially for the preterm and low weight newborns, accordingly miRNAs could satisfy those above requirements and reduce the incidence of anemia and slow increase of weight.For example, a study once pointed out that miR-34a exhibited high specificity of 97% and 0.94 of AUC in diagnosing neonatal sepsis [12].Although these results cannot be directly extrapolated to neonatal patients with sepsis owing to the widely different conditions, age, disease stage, and overall state of the organism struck by neonatal sepsis [7], the exploration of novel biomarkers and potentially functioning genes is still necessary for the early diagnosis of EOS.
In this panel, the expression level of mir-15a-5p (AUC ¼ 0.67; 95% CI: 0.57-0.78;sensitivity ¼ 81.4%, specificity ¼ 53.5%) was downregulated in patients with sepsis.It has been reported to be involved in the inflammatory process during sepsis by activating the NF-jB pathway and targeting TNIP2, suggesting that the miR-15a-5p inhibitor may be a novel anti-inflammatory agent and therapeutic strategy for sepsis [13].Similarly, in LPS-treated lung cells, circFADS2 overexpression can reduce miR-15a-5p overexpressioninduced apoptosis and increase circFADS2 levels [14].It could also be related to the pathogenesis of sepsisinduced AKI through the miR-15a-5p-XIST-CUL3 regulatory axis, which indicates the potential value of prognosis in evaluating organ dysfunction.Similar to the above finding, miR-17-5p level also declined in the serum of patients with septic AKI, which means that both miRNAs may be considered as biomarkers for kidney injury, except for creatinine and urea nitrogen [7,15].CircTLK1 sponges miR-17-5p to aggravate mtDNA oxidative damage, mitochondrial dysfunction, and cardiomyocyte apoptosis by activating the PARP1/ HMGB1 axis during sepsis, indicating that circTLK1 may be a putative therapeutic target [16].A thesis get the similar conclusion that miR-146a-5p level was elevated in the serum of septic patients, overexpression of it promoted proliferation and inhibited apoptosis of LPS-induced cardiomyocyte [17], miR-146a-5p deletion alleviated the activation of T cells and attenuated the imbalance of Th17/Treg, and an animal experiment revealed that miR-146a-5p downregulation alleviated T cell activation, inflammation, lactate production, and glucose uptake in sepsis mice [18].Another study also revealed that in septic patients, plasma miR-146a-5p concentrations are closely associated with sepsis outcomes, blood lactate, and coagulopathy, and miR-146a knockout in mice offers protection against sepsis with attenuated interleukin-6 (IL-6) storm and organ injury, improved cardiac function, and better survival [19].
Studies have showed that mir-223-3p is significantly downregulated in patients with sepsis and has been proven by many animal or cellular tests to be involved in inflammation.For MKNK1, one target of miR-223-3p might be involved in sepsis by regulating neutrophil abundance by mediating the expression of inflammatory factors [20], and lncRNA-SNHG14 regulates autophagy by controlling miR-223-3p/Foxo3a as a ceRNA and participating in alveolar type II epithelial cell injury and acute lung injury induced by LPS [21], and a study had pointed out that DLX6-AS1 mediated LPS-mediated cytotoxicity and proptosis in HK-2 via miR-223-3p/NLRP3 axis [22].
In GSE65682 from the GEO database, a total of 682 subjects with various causes of sepsis were included for consensus weighted gene co-expression network analysis (WGCNA), which was performed to identify modules of sepsis, and got that mir-16-5p was one of the top five miRNAs that regulated the highest number of genes [23].Another study on the microarray dataset GSE94717 (including six sepsis-induced AKI samples and three control samples) revealed that mir-16-5p is involved in the mTOR signaling pathway and is enriched in the PI3K-Akt signaling pathway [15].Both studies have implicated that mir-16-5p functions in the progression of inflammation.For example, upregulating exosomal miR-16-5p can inhibit macrophage inflammation and relieve lung injury in septic mice [24] and exacerbate sepsis by upregulating aerobic glycolysis via the SIRT3-SDHA axis [25].
Regarding the mechanism of mir-1-3p in sepsis, there are relatively few studies, one illustrating that upregulated miR-1-3p inhibits cell proliferation, promotes apoptosis and cytoskeleton contraction, increases monolayer endothelial cell permeability, and membrane injury by targeting SERP1, which leads to dysfunction of endothelial cells and weakens the vascular barrier function involved in the development of acute lung injury [26].Another study revealed that in septic mouse models, omega-3 fatty acids elevated the expression of Notch3 by downregulating miR-1-3p and blocking the Smad pathway to alleviate intestinal epithelial inflammation and oxidative stress injury caused by sepsis [27].Both studies showed that mir-1-3p plays a role in the pathological process of sepsis and may be a promising therapeutic candidate for sepsis-induced lung or intestinal injury.
Although the findings are meaningful, there are some limitations for this study regarding method and design.First, for the delivery stress response, the inflammatory indexes are regularly high and descending with time, and reliable blood culture results as the gold standard are usually difficult to obtain due to the high false negative ratio.We cannot ascertain the diagnosis of EOS because the diagnostic standards from the national consensus are inclined to meet empirical judgment because a pediatrician cannot make newborns risk the prognosis of neurological impairment or vital septic shock.Second, although we have demonstrated the clinical value of these three miRNAs' diagnostic power in EOS, the miRNAs' ability to predict the short-term or long-term outcome remains to be experimentally validated and further elucidated.Third, this study was a single-center and total number of cases included in the study was relatively small, larger number of patients and multi-center verification and experiments are required.Additionally, it is not clear whether the miRNAs could predict the incidences of neonatal sepsis in early stage, which consequently happens after necrotizing enterocolitis, pneumonia, or other local infections.
Combined with the results, we drawn the conclusion that mir-15a-5p, mir-223-3p, and mir-16-5p could serve as predictive biomarker of EOS, with relatively higher sensitivity, and are significantly associated with the prediction of organ dysfunction in sepsis, such as acute kidney injury, acute lung injury, and septic cardiomyopathy.Although there are limited reports associated with the top 10 hub genes after retrieving databases, which were mostly studied in the metabolism of cancer and rarely researched in neonatal fields, further studies on EOS should still be carried out to obtain an early and precise diagnosis and ultimately avoid poor prognosis or prolonged antibiotic application in newborns.

Figure 1 .
Figure 1.The flowchart of study design.ROC: receiver operating characteristic.

Figure 3 .
Figure 3. Receiver operating characteristic curve analyses of six differential expressed miRNAs.

Figure 4 .
Figure 4. Receiver operating characteristic curve analyses of three miRNAs panel.

Figure 5 .
Figure 5.The OR value of three miRNAs; CI: confidence interval.

Figure 6 .
Figure 6.GO analysis of related genes.BP: biological process; CC: cellular function; MF: molecular function.

Table 1 .
Clinical data of the study population.

Table 2 .
Outcomes of receiver operating characteristic curves and associated criterion for six candidate miRNAs and the three-miRNA panel.