MYD88, NFKB1, and IL6 transcripts overexpression are associated with poor outcomes and short survival in neonatal sepsis

Toll-like receptor (TLR) family signature has been implicated in sepsis etiopathology. We aimed to evaluate the genetic profile of TLR pathway-related key genes; the myeloid differentiation protein 88 (MYD88), IL1 receptor-associated kinase 1 (IRAK1), the nuclear factor kappa-B1 (NFKB1), and interleukin 6 (IL6) in the blood of neonates with sepsis at the time of admission and post-treatment for the available paired-samples. This case–control study included 124 infants with sepsis admitted to the neonatal intensive care unit and 17 controls. The relative gene expressions were quantified by TaqMan Real-Time qPCR and correlated to the clinic-laboratory data. MYD88, NFKB1, and IL6 relative expressions were significantly higher in sepsis cases than controls. Higher levels of MYD88 and IL6 were found in male neonates and contributed to the sex-based separation of the cases by the principal component analysis. ROC analysis revealed MYD88 and NFKB1 transcripts to be good biomarkers for sepsis. Furthermore, patients with high circulatory MYD88 levels were associated with poor survival, as revealed by Kaplan–Meier curves analysis. MYD88, NFKB1, and IL6 transcripts showed association with different poor-outcome manifestations. Clustering analysis split the patient cohort into three distinct groups according to their transcriptomic signature and CRP levels. In conclusion, the study TLR pathway-related transcripts have a gender-specific signature, diagnostic, and prognostic clinical utility in neonatal sepsis.


Results
Baseline characteristics of the study population. The study included 124 sepsis infants (81 males and 43 females) and 17 healthy controls (12 males and 5 females) (p = 0.78). Both study groups were matched for baseline demographic parameters such as age at sampling (2 ± 1.1 days versus 3.3 ± 3.9, p = 0.17), gestational age (37.1 ± 1.9 weeks versus 36 ± 3.6 weeks, p = 0.22), and birth weight (2.6 × 10-3 ± 0.4 × 10-3 g versus 2.4 × 10-3 ± 0.7 × 10-3 g, p = 0.16) in cases and controls, respectively. Of the sepsis cases, 93 were early-onset (onset of sepsis features within 72 h of life), while 31 neonates presented with late-onset sepsis (onset of sepsis features after 72 h of life). Comparison between male and female patients is presented in Table 1. A higher frequency of premature rupture of membranes was observed in male newborns (42% vs. 23.3%, p = 0.049). In male Figure 1. Toll-like receptor (TLR) signaling pathway. On exposure to pathogens secreting pathogen-associated molecular patterns (PAMPs), their recognition is initiated via pattern recognition receptors, including several Toll-like receptors (TLRs). For example, lipopolysaccharides on gram-negative bacteria activate TLR4, lipoteichoic acid of gram-positive bacteria activates TLR2, while TLR3 recognizes viral PAMPs, generating innate immune responses via the MyD88-dependent pathway that leads to the production of pro-inflammatory cytokines with activation of nuclear factor kappa B (NFKB) and the downstream gene targets 9,10 . The MyD88independent pathway associated with the induction of type I interferon (IFN) and IFN-inducible genes [Data source: KEGG pathway, hsa04620 and created by Biorender.com]. Bioinformatic selection of Toll-like receptor pathway. The transcriptomic analysis of 324 differentially expressed genes (DEGs), including 305 upregulated and 19 downregulated genes in preterm infants during late-onset sepsis, is shown in Figure S1. Functional enrichment analysis of DEGs identified TLR signaling pathway (hsa04620 | hits = 13 out of 104 | FDR = 7.93e−5) and NFKB signaling pathway (hsa04064 | hits = 12 out of 100 | FDR = 1.83e−4) to be the top deregulated pathways (Table S1). A total of 73 gene hits involved in pathways is depicted in the gene heat map ( Figure S2). For the Toll-like receptor pathway, CCL3L3, CCL4, CXCL8, IL1B, IRAK4, IRF7, MAP2K6, MYD88, NFKBIA, TLR1, TLR2, TLR4, and TLR5 genes were significantly enriched. As shown in Fig. 1, MYD88, IRAK4, and NFKBIA genes master a cascade of signaling transduction which leads to the expression of many inflammatory genes for cytokines, chemokines, endothelial adhesion molecules, and costimulatory molecules. NFKBIA gene was iterated in 9 out of the 10 top pathways, MYD88 gene was enriched in 3 pathways, and IRAK4 was the only gene located on chromosome X. Their diagnostic accuracy was compared to IL6, which is used in our hospital protocol as a gold standard test for inflammatory reaction in sepsis.
Circulatory levels of Toll-like receptor signaling pathway genes in neonatal sepsis compared with healthy newborns. Fold changes of the four tested genes (MYD88, IRAK1, NFKB1, and IL6) in 81 males, and 43 females are shown in Fig. 2. Stratification analyses by sex and time of onset of the disease and comparing paired samples at admission with the second blood sample three days after treatment are shown. As demonstrated in Fig. 2, across all neonates, relative expression levels of MYD88 (p < 0.001), NFKB1 (p < 0.001), and IL6 (p = 0.027) were significantly higher in sepsis cases compared to controls. Higher levels of MYD88 (p < 0.001) and IL6 (p < 0.001) were found in male infants compared to females. None of the gene levels were significantly altered after receiving treatment. Clustering of patients into distinct phenotypes. Sepsis patients exhibited a different pattern of gene expression combinations. Gene co-expression analysis across the 124 sepsis cohorts showed that IL6 gene expression was directly correlated to MYD88 (r = 0.26, p = 0.004) and NFKB1 (r = 0.19, p = 0.039) transcript levels ( Fig. 3A). Clustering analysis split cohorts into three distinct groups according to their transcriptomic signature and CRP levels; cluster 1 included 67 infants with low IRAK1 gene expression, cluster 2 including 49 neonates Association of gene expression levels with clinical characteristics of the patients. As shown in Table 3, MYD88 expression was upregulated in patients presented with necrotizing enterocolitis (p = 0.043),  www.nature.com/scientificreports/ transient tachypnea (p = 0.043), poor feeding (p = 0.037), congenital pneumonia (p = 0.021), acquired pneumonia (p = 0.003), and jejunal atresia (p = 0.029). Elevated circulatory levels of the MYD88 were associated with mortality (p = 0.007). NFKB1 overexpression was associated late-onset sepsis (p = 0.002), cohorts with hemodynamic instability (p = 0.044), poor feeding (p = 0.032), and oliguria (p = 0.026). Moreover, IL6 was significantly upregulated in males (p < 0.001), low-weight birth babies (p = 0.032), and infants of triple I mothers (p = 0.006). Cluster 1 was associated with temperature instability (p = 0.048) and poor feeding (p = 0.018). All cluster 3 were females, while males were more representative in cluster 1 (p < 0.001). Spearman's correlation analysis of the gestational age with gene expression of the studied four genes revealed a very weak correlation coefficient ranging from 0.02 to 0.08 ( Figure S3).  www.nature.com/scientificreports/ In those neonates with positive blood culture, cohorts were categorized into gram-positive and gram-negative groups. Gene expression analysis revealed no significant difference in both groups ( Figure S4). In addition, the type of organism in positive blood culture patients did not show a significant impact on mortality risk as

Functional enrichment analysis of Toll-like receptor signaling pathway genes. Pathway analysis
for the four studied genes showed them to be involved in multiple inflammatory-related pathways and host immune responses (Fig. 6).

Discussion
Although the sepsis incidence continues to increase and impacts a wide range of ages, our understanding of human response to sepsis across different age groups remains inadequate and limits our ability to modify outcomes 6 . Given the essential roles TLR and NFKB-pathways play in sepsis as identified previously 11,12 and through our in silico analysis, this work aimed to explore the transcriptomic signature and the clinical utility of MYD88, IRAK1, NFKB, and IL6 in a sample of neonatal sepsis and to correlate the gene signature with the clinic-laboratory data.
In this study, significant upregulation of MYD88 was identified in neonates with sepsis relative to uninfected controls. Interestingly, MYD88 mRNA showed a sex-specific signature being higher in male neonates than females, contributing to the clustering of both groups by the principal component analysis. The findings also highlight the potential clinical utility of this transcript as a diagnostic/prognostic molecular biomarker in terms of having a high area under the ROC curve and associating with unfavorable clinical phenotypes (e.g. necrotizing enterocolitis, transient tachypnea, poor feeding, congenital and acquired pneumonia, jejunal atresia) and poor survival.
Given the central role MyD88 plays in TLR signaling, TLR-induced death, and the innate immune response activation 13 , among others (as shown in Fig. 6), it is not surprising to find MYD88 overexpression in PBMC of the sepsis cohort in line with the findings of previous clinical studies and experimental septic models [13][14][15][16] . For    www.nature.com/scientificreports/ patients, which were persisted across the different stages of sepsis 13,14 . Adib-Conquya et al. also reported a significant upregulation of MyD88s in monocytes of septic patients compared to either resuscitated patients after cardiac arrest or healthy controls. Their transfection experiments identified that the short form of "MyD88" could negatively regulate "TLR2-dependent NF-B response" 15 . Furthermore, Khailova et al. confirmed the significant increase of MyD88 with other assessed TLR pathway-related molecules (TLR-2 and NFΚ) in the lungs of septic mice model of cecal ligation and puncture peritonitis relative to healthy shams 16 . This overexpression was ameliorated on "Lactobacillus rhamnosus GG" or "Bifidobacterium longum" probiotic treatment. The present finding of MYD88 upregulation showed sex disparity with significantly higher levels in male neonates than females, contributing to the previous observations in which females exhibited a less exaggerated immune response than males 17,18 . While the exact mechanism underlying this observation is not conclusive, hormonal and epigenetic factors were proposed to contribute to this phenomenon by impacting the pathogenspecific inflammatory responses (particularly lipopolysaccharides) and the immunological differences [18][19][20] . Naugler et al. have found that "MYD88-dependent activation of IL6 production" could be negatively regulated by estrogen in their model of hepatocellular carcinoma induced by liver inflammation 21 . Estrogen has displayed a suppressive effect on inflammation, mainly via downregulating the NFKB1 transcriptional activity with a subsequent decrease in pro-inflammatory cytokine/chemokine production, including IL6 [22][23][24] . El Sabeh et al., in their recent preprint, revealed that MYD88/estrogen receptor-α interaction during the inflammatory signaling might contribute to the gender-dependent bias in the inflammatory response 25 . In addition, Crisostomo et al. identified that the stimulated female-derived murine mesenchymal stem cells have demonstrated less inflammatory response, including IL6 and tumor necrosis factor-α production, compared with the male-derived ones 26 . All these observations, including ours, highlight the importance of running "sex-based therapeutic interventions," which will improve neonatal outcomes, particularly in a neonatal intensive care unit (NICU) 27 .
Signals transduced by MYD88 should continue to the nucleus to upregulate the pro-inflammatory gene expressions, including the IL6 gene, through the transcriptional factor NFKB (shown in Fig. 1), which orchestrates the immune response through multiple downstream targets leading to inflammatory reactions and ultimately severe phenotype 28 . This supports in part the observed upregulation of IL6 and NFKB1 in septic neonates and association with some features of poor outcomes, including low-weight birth babies and infants of triple I mothers (in case of IL6), and hemodynamic instability, poor feeding, and oliguria (in case of NFKB1).
The pro-inflammatory cytokine IL6 is expressed by immune cells (macrophages, dendritic cells, B cells, and epithelial cells). It has been found to "exhibit a more prolonged response to the pathogenic challenge", even in the period of temperature stabilization, in comparison to other cytokines 20 . Such type of pleiotropic interleukin mediates its molecular signaling through JAK/STAT and MAPK pathways and is involved in many biological processes, including cell survival/apoptosis, T cell maturation, T helper-1/2/17 differentiation, and inflammation 29 . It showed higher levels in neonatal males versus females in the present study and adult males with sepsis versus females in earlier studies [30][31][32] and after administering lipopolysaccharides in experimental work 19 . High serum levels of IL6 showed a correlation with the low circulating lymphocyte counts and correlated with a high risk of developing acute respiratory distress syndrome 33 . These findings could support the predictive role that such cytokine plays in septic episode severity as proposed by Wang and colleagues 34 and highlighting the IL6 as a promising target for the current monoclonal antibody therapy in the clinics 35 .
Although recent work suggests that a different gene expression signature might be present in the case of different bacteria (gram-positive vs. gram-negative) 36 , we found no significant difference between neonates with positive culture results of gram-positive and gram-negative. Also, despite some in vitro stimulation experiments (to imitate sepsis-like conditions), revealed that conventional antibiotic therapy might be associated with immunomodulatory properties associated with the change in transcriptomic signature for some TLRs and related cytokines 37,38 , none of the studied gene transcript levels in the present study were significantly altered at the second sampling time point (after three days of receiving antibiotics). It is worth noting that what was found in in vivo and in vitro studies does not always translate to the clinic. Also, it does not discount the possibility that the gene signatures could be normalized with a longer treatment time. Furthermore, differences in the study design, mammalian cells investigated, bacterial species, class of antibiotics tested, and sepsis stage, among others, should be considered.
This study is limited by the relatively small sample size and the study design (a case-control study), hence large-scale and follow-up studies are recommended. Also, it was difficult to find healthy preterm neonates during sample collection in our hospital and obtain parental consent for sample withdraw in such cases (if any); hence, this study included full-term neonates as a control group. This issue (matched case-control groups regards the gestational age) should be considered in future studies. Furthermore, because of the small volume of blood collected in this work, the authors could only investigate the expression of the studied genes at the level of mRNA. It would be useful to demonstrate the protein levels of the studied genes in future studies as the mRNA levels may not always reflect the corresponding protein expression due to the posttranscriptional processing. Indeed, the role of other genetic/epigenetic and environmental factors should be considered.
In light of the present findings, the development and validation of new approaches for neonatal sepsis treatment are recommended, such as the potential use of TLR-pathway antagonists/inhibitors to limit the TLRs association with MyD88 preferentially and reduces the NFKB1 activity and/or IL6 8,[39][40][41] . Furthermore, as mentioned earlier, clinically relevant sex-based therapeutic strategies should be implicated as adjunct approaches to the ordinary ones to increase the potentially favorable outcomes in male neonates.  42 and expected to be diagnosed clinically as neonatal sepsis were included. Neonates with a gestational age below 28 weeks, a history of perinatal hypoxia, hypoxic-ischemic encephalopathy, or gross congenital anomalies and genetic syndromes were excluded. If mothers had positive hepatitis C or B infection or were known to have a history of misuse drug intake during pregnancy, the neonates were also excluded. The controls included 17 samples of healthy full-term neonates collected in the same period during the routine screening of serum bilirubin. Because of the difficulty of blood sampling in some neonates, the drop-out sample (due to non-survivors), and the mother's non-approval for including a second sampling of their neonates, only 43 paired samples were available for the comparison.
Clinical assessment. Each participant was subjected to history taking from the mother to detect any sign of sepsis, full maternal history, including maternal age, gravidity, and parity, medical history, details of labor with an emphasis on any prenatal hazards (as pre-eclampsia, premature rupture of membranes, antepartum hemorrhage, or intrapartum fever), detailed perinatal history of neonates, including gestational age, mode of delivery, early postnatal cyanosis, jaundice or convulsions, and full birth record, including methods and duration of resuscitation and mode of lactation (if present). The clinical data included (1) gestational age assessment, weight, and sex of the full-term neonates, (2) general and systemic examination, including (a) the respiratory system: tachypnea, apnea, increased ventilator support, and oxygen desaturation, (b) the cardiovascular system: bradycardia, pallor, hypotension, and decreased perfusion (c) metabolic changes: hypothermia, hyperthermia, feeding intolerance, glucose instability, metabolic acidosis, and (d) neurologic changes: lethargy, hypotonia, and decreased activity.
Sample collection. Five milliliters of peripheral blood were withdrawn on admission for 124 neonates under aseptic conditions for routine chemistry and immunological assessment (2 ml on plain tubes), complete blood count (1 ml on EDTA tubes), blood culture (1 ml), and genetic analysis (1 ml on EDTA tubes). The latter test tubes were transferred immediately to the genetic lab within 20 min to be centrifuged with separation of the buffy coat in sterile Eppendorf for the subsequent genetic analysis. Sampling was then repeated after three days of treatment (as post-treatment samples) for the available 43 neonates.
Blood culture. The collected venous blood was inoculated directly into blood culture medium vials and send to the clinical microbiology laboratory for cultivation. In brief, the blood cultures were incubated aerobically at 37 °C and observed daily for the first three days to identify any visible microbial growth. Simultaneously, subcultures were made during three consecutive days on enriched and selective media, including chocolate, MacConkey, blood, and mannitol salt agar plates. The subcultures were examined for growth after incubation for 24-48 h. The protocol was repeated until the 7th day before blood culture was considered sterile 43 .
In silico selection of candidate pathway. Gene Expression Omnibus (GEO) database, a public functional genomics data repository, was screened for RNAseq experiments on neonatal sepsis. Transcriptomic signature of GSE138712 experiment was utilized for analysis, for which whole blood specimens of nine preterm neonates (< 30 weeks gestational age) with late-onset sepsis were compared to another nine age-matched preterm infants without sepsis. Data was downloaded from GEO RNA-seq Experiments Interactive Navigator (GREIN (ilincs.org). Transcriptomic analysis of the GEO dataset was performed using NetworkAnalyst (www. netwo rkana lyst. ca). Differentially expressed genes (DEGs) were identified using Limma R package (false discovery rate "FDR" < 0.05 and |FC|> 1.0). Pathway enrichment analysis was performed in String database version 11.0 (string-db.org).
Gene signature analysis. RNA  www.nature.com/scientificreports/ (2 ×) (TaqMan, Applied Biosystems) on StepOne™ Real-Time PCR System (Applied Biosystems). All reactions were run in duplicate, and a "No-template" and a "No-RT" controls were included in each run. Each 96-well plate run initially at 95 °C for 5 min, followed by 40 cycles of denaturation at 95 °C (15 s), annealing at 60 °C (1 min), and elongation at 72 °C (1 min). Real-time PCR was run following the "Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines" 45 . The fold change of the transcriptomic signature of the four genes in the neonates with sepsis relative to the mean value of the controls was calculated using the LIVAK method based on the quantitative cycle (Cq) values with the equation (2 −ΔΔCq ) 46 .
Statistical analysis. Data analysis was performed using SPSS version 27.0, GraphPad Prism version, and RStudio 1.3.1056. Using the G*Power 3.1.9.2. with the specified study design (gene expression), alpha error = 0.05, an effect size = 0.74, and a total sample size of 138 can give 81% power of the study. http:// www. gpower. hhu. de/A two-sided Chi-Square test was employed for categorical variables, and Mann-Whitney or Kruskal Wallis was applied for quantitative data. Univariate logistic regression analysis was performed, and results were reported as odds ratio (OR) and 95% Confidence intervals (CI). Spearman's correlation analysis was used. The statistical significance cutoff level was set at P value < 0.05. Hierarchical clustering analysis and K-means clustering were generated and visualized using BioVinci (Bioturing, San Diego, CA, USA) through the following parameters: Ward's minimum variance clustering method, Euclidian distance, dendrogram, scaling and centering. Gene expression was categorized at the median cutoff values into high and low expression. To compare the survival in high and low expressor groups, "Kaplan-Meier" curves were plotted. Cox Hazards Proportional Regression analysis was carried out to identify independent risk factors for mortality, and results are reported as hazards ratio (HR) and 95%CI. Multivariate analysis was executed for data exploration in the principal component analysis (PCA) using "Psych, Factoextra, FactoMineR, ggplot2, ggpubr, and magrittr" packages.
Institutional review board statement. Author's agreement of originality and statement of copyright transfer. This is an original article that has not been published in any other publication.
Informed consent. Informed consent was obtained from all included neonate caring relatives in the study.

Data availability
All data generated or analyzed during this study are included in this submitted article and Supplementary Materials.