Study reveals gene expression changes at birth that could allow early detection and treatment of neonatal sepsis.
Study: Prognostic gene expression signature diagnoses neonatal sepsis before clinical presentation. Image credit: Iryna Inshyna/Shutterstock.com
In a recent study published in eBioMedicine, A team of researchers has identified gene expression biomarkers that allow early prediction of neonatal sepsis (a serious bacterial infection in infants under 28 days old, leading to systemic inflammation and serious complications) at birth, before clinical symptoms, facilitating early intervention .
Background
Sepsis leads to life-threatening organ failure due to dysregulated host response to infection and occurs uniquely in all age groups.
Neonatal sepsis, which affects infants in their first 28 days, occurs in 2-3 per 100 live births worldwide, particularly affecting premature and low-birth-weight infants. Mortality rates can be as high as 17.6%, especially in low/middle income countries, with sepsis being the leading cause of neonatal death.
Nonspecific clinical signs often delay diagnosis, making early treatment critical. Current diagnostic methods, including blood cultures, are insufficient, highlighting the urgent need for specific biomarkers for the early and accurate identification of sepsis. Further research is necessary to develop predictive biomarkers for early intervention.
About the study
The Systems Biology to Identify Biomarkers of Neonatal Vaccine Immunogenicity study enrolled 720 healthy full-term neonates (≥37 weeks’ gestation) in The Gambia from 2017 to 2019. Whole blood was collected from these neonates at two time points: the day of life (DOL ) 0 and a second sample taken randomly on DOL 1, 3, or 7.
Among this cohort, 33 neonates were hospitalized within 28 days for clinical signs consistent with sepsis. Of these, 21 were diagnosed with sepsis based on either blood culture results or clinical diagnosis, while the remaining 12 had localized infections without systemic signs. Septic neonates were categorized into early-onset sepsis (EOS) and late-onset sepsis (LOS).
Blood samples were processed for ribonucleic acid (RNA) sequencing and bioinformatic analyzes were performed using R. Differentially expressed genes (DEGs) were identified using Differential Gene Expression Analysis based on the Negative Binomial Distribution (DESeq2), comparing healthy controls, cases of localized infection and septic neonates.
Predictive gene biomarkers for sepsis were developed through machine learning methods, including Sparse Partial Least-Squares Discriminant Analysis (sPLS-DA) and Regression Least Absolute Shrinkage and Selection Operator (LASSO). The study provided rigorous statistical methods for evaluating DEGs, with findings validated against external datasets.
Study results
The cohort consisted of full-term neonates, all born between 37 and 42 weeks’ gestation, with a median gestational age of 40 weeks. At birth, neonates had Apgar scores ranging from 8 to 10, with a median score of 10.
Demographic and clinical parameters of healthy neonates and those hospitalized for sepsis were comparable, with no statistically significant differences, which may be attributable to the limited cohort size. During hospitalization, septic neonates had significantly higher neutrophil counts, received antibiotics more frequently, and had a longer hospital stay compared to those with local infections only.
Transcriptional differences were evident in septic neonates even before clinical symptoms appeared. Among neonates with DOL 0 samples who later developed sepsis, a total of 469 DEGs were identified compared to healthy neonates and 476 DEGs relative to those with localized infections.
Further analysis revealed that neonates with EOS exhibited 1,067 DEGs compared to those with LOS, 984 DEGs compared to cases of localized infection, and 1,086 DEGs compared to healthy controls.
Pathway analysis highlighted the dysregulation of several processes, including regulation of the heat shock response and various cell cycle pathways, indicating significant changes in the host response prior to clinical presentation.
Machine learning algorithms were used to identify predictive biomarkers for EOS at birth. sPLS-DA identified heat shock protein family H member 1 (HSPH1) and heat shock protein family member DnaJ B1 (DNAJB1) as major contributors in distinguishing neonates with EOS from others.
These heat shock proteins were upregulated in EOS cases and correlated with the observed enriched pathways. In addition, the LASSO regression model identified a 4-gene signature (HSPH1, DNA primase subunit 1 (PRIM1), BORA aurora kinase activator A (BORA), and non-SMC G2 condensate complex II subunit (NCAPG2)) that demonstrated excellent predictive ability for EOS with area under the curve (AUC) 0.94, sensitivity 0.93 and specificity 0.92.
The study also included follow-up samples from the first week of life to assess how neonatal sepsis may affect developmental trajectories. Analyzing DOL7 samples from EOS neonates and matched healthy controls, the researchers found 3,931 DEGs in EOS neonates compared to 2,456 in healthy neonates.
While many pathways were disrupted in EOS neonates, several metabolic and immune pathways remained preserved, highlighting the complexity of immune and metabolic adaptations during early development and the potential of early biomarkers to predict sepsis outcomes.
conclusions
In summary, this study successfully identified gene expression biomarkers that predict neonatal sepsis at birth despite challenges in developing such predictors. HSPH1 was a significant predictor for EOS, being part of a 4-gene signature that distinguished EOS from LOS, localized infection, and healthy controls.
Furthermore, gene expression changes during the first week of life showed that sepsis disrupts immune and metabolic development while many ontogeny processes remain preserved.
These findings highlight the need for improved diagnostics in neonatal sepsis, particularly in low- and middle-income countries, and highlight the potential for wider application in different populations.
