A comparative proteomic analysis for non‐invasive early prediction of hypoxic‐ischemic injury in asphyxiated neonates

Hypoxic Ischemic Encephalopathy (HIE) is one of the principal causes of neonatal mortality and long‐term morbidity worldwide. The neonatal signs of mild cerebral injury are subtle, making an early precise diagnosis difficult. Delayed detection, poor prognosis, and lack of specific biomarkers for the disease are increasing mortality rates. In this study, we intended to identify specific biomarkers using comparative proteomic analysis to predict the severity of perinatal asphyxia so that its outcome can also be prevented.


BACKGROUND
Birth asphyxia or perinatal asphyxia is a clinical condition that arises due to the impaired gas exchange in neonates.Some of the factors which are responsible for impaired gas exchange are alveolar-capillary membrane changes, decreased oxygen-carrying capacity of blood, abnormal RBC structure and its reduced life span, increased blood viscosity, predisposition to bacterial pneumonia/pulmonary infarcts, and so on [1][2][3].It further leads to progressive hypoxia, hypercarbia, and acidosis, depending upon the extent and duration of the interruption.The interruption of oxygen supply causes energy failure and triggers a biochemical cascade, leading to cell dysfunction and death [4].The following criteria are considered for the diagnosis of asphyxia: (i) profound metabolic or mixed acidemia (pH <7.00) in an umbilical artery blood sample, if obtained, (ii) persistence of an APGAR (Appearance, Pulse, Grimace, Activity, and Respiration) (score of 0−3 for longer than 5 min, (iii) neonatal neurologic sequelae (e.g., seizures, coma, hypotonia), and (iv) multiple organ involvement (e.g., kidney, lungs, liver, heart, intestines) [5].Globally 2.4 million newborn deaths occur annually, contributing to ∼47% of the under-5 child mortality.
Hypoxia of the newborn accounts for 20% and is the third most important cause of neonatal deaths, which is 900,000 neonates every year worldwide [6].The frequency of perinatal asphyxia is approximately 3-5 in 1000 live births in developed countries with advanced obstetric and neonatal care.Each year, approximately a quarter of global neonatal deaths occur in India, with neonatal mortality rate of 28 per 1000 live births and an infant mortality rate is 40-49 per 1000 live births [7,8].Birth asphyxia can arise from several antepartum and intrapartum risk factors [9,10].The severity and duration of the hypoxia are the most important factors determining the degree and the extent of organ and tissue damage associated with perinatal asphyxia.
One of the outcomes of asphyxia is Hypoxic-ischemic encephalopathy (HIE).Due to a lack of oxygen and blood supply to the brain, the neurons are injured, and some even die.After assessing the severity of encephalopathy or insult by Sarnat and Sarnat staging criteria, HIE is categorized into three stages: HIE-1 (Mild), HIE-2 (Moderate), and HIE-3 (Severe).Hypoxia also has other adverse impacts such as respiratory distress syndrome, disseminated intravascular coagulation, subcutaneous fat necrosis, myocardial ischemia, adrenal hemorrhage, metabolic disorders, or acute tubular necrosis [11,12].
The current measures used to determine birth asphyxia include intrapartum electronic fetal monitoring, fetal or umbilical cord pH measurement, APGAR score, EEG, MRI, HIE, and major organ disorders [13].Early prediction of HIE is vital for selecting new-born infants who could benefit from neuroprotective treatment such as hypothermia [14,15].Unfortunately, even when combining clinical signs and the current techniques, early identification of neonates, who may subsequently have a poor neurodevelopmental outcome, can still be challenging [16].This necessitates an urgent need to identify biomarkers that may aid in the early diagnosis of perinatal asphyxia and promptly identify neonates who require immediate attention for neuroprotection.Though the research in this field has been active in the last few decades, one of the biggest challenges lies with the prediction,

Significance Statement
Despite advances in monitoring technology and knowledge of fetal and perinatal medicine, the improvements in long-term neurological outcomes of neonates with hypoxicischemic encephalopathy (HIE) remain modest.The precise biomarkers for the early identification of neonates with subtle brain injury remain elusive.Using shotgun proteomics followed by ELISA, we validated a panel of four urinary proteins for predicting hypoxic injury in asphyxiated neonates within the first few hours of life.The cumulative ROC curve produced for AGT and FABP1, and APP and FN1 was significant for all three stages of HIE.This makes it a highly sensitive prediction model with high accuracy, sensitivity, and specificity.After validation on a larger cohort, this combination of urinary biomarkers can be developed into a diagnostic kit for early identification of brain injury in asphyxiated neonates.It can also help in understanding the disease progression.The prompt treatment of neonates may also reduce mortality and neurodevelopmental impairment.
detection, and grading of neonatal HIE, as this grading impacts the therapeutic intervention.The emergence of MS-based proteomic platforms as prominent bio-analytical tools in clinical applications has enhanced the identification of protein biomarkers in urine [17,18].Several studies have highlighted the specific role of a few promising biomarkers in the urine of the neonatal population [19][20][21][22], however, none are routinely used in the clinic.Since urine sampling is minimally invasive compared to sampling of blood, urinary biomarkers have an important diagnostic approach.
In the present study, we employed a quantitative proteomics approach to identify urinary biomarkers to detect early brain injury in asphyxiated neonates with different stages of HIE-Mild (HIE-1), Moderate (HIE-2), and Severe (HIE-3).Our analysis revealed early biomarkers like Uromodulin (UMOD), Apolipoprotein-A2 (APOA2) for HIE-1 and also differential or severity biomarkers like Fatty Acid Binding Protein 1 (FABP1), Apolipoprotein-M (APOM), Parkinsonism Associated Deglycase (PARK7), Osteopontin (SPP1) for all the three HIE stages.The differentially expressed proteins were associated with neurodegenerative diseases, oxidative stress, amyloid fiber formation, and programmed cell death.This study may pave a path for identifying clinically relevant biomarkers specific to brain injury or its severity during birth asphyxia, which will further help to improve the therapeutic outcomes.

Study design
This case-control study was conducted in a tertiary care hospital in South India to identify candidate markers in the urine of neonates to predict birth asphyxia.The study included infants admitted and treated for birth asphyxia and its outcomes.The urine samples of both asphyxiated and healthy neonates were collected after obtaining approval from Yenepoya Ethics Committee-1 (Study protocol no. YEC-1/2017/183) and institutional ethics guidelines were followed.
A written informed consent form was obtained from the parents of all the neonates and.The asphyxiated and healthy (control) neonate's urine sample was collected in the pediatric uro bag (Romsons) and transferred to a sterile container.The first spot urine sample of the asphyxiated and healthy neonates was collected within 24 h of life, and the second urine sample was taken only for asphyxiated neonates within 72 h of life.
For the diagnosis of birth asphyxia, the following criteria were considered for the selection of neonates: APGAR score should be less than or equal to 5 at 5 min for birth asphyxia, evidence of encephalopathy using Sarnat and Sarnat staging, either evidence of fetal distress or assisted ventilation for at least 10 min after birth or evidence of any organ dysfunction, and ABG abnormality should be present at the time of birth for birth asphyxia.The samples with all birth weight categories and gestational age in neonates were included.The study excluded neonates with major congenital renal anomalies and maternal renal failure (File S1).

Clinical details
We collected the data on demographic details, risk factors for birth asphyxia, details of type of delivery with need for resuscitation and APGAR scoring.Postnatally the babies were assessed for the evidence of HIE using Sarnat and Sarnat staging and multiorgan dysfunction.
Simultaneously need for ionotrophs and assisted ventilation were also recorded.Neonate's condition was monitored until discharge from NICU or death.The biochemical parameters which are indicated in birth asphyxia were measured in blood as per institutional treatment protocol.These include Arterial blood gas (ABG), Renal function test (RFT), Serum electrolytes, and Septic screening parameters.The normal values of these parameters are given in File S2.The identified biomarkers were correlated with the clinical severity of the disease.

Proteomics sample preparation for the discovery phase
The urine was centrifuged at 3000 rpm for 15 min at 4 • C, and the supernatant of the urine was passed through a 0.22 um PES filter.
Then the urine was concentrated using an Amicon 3 kDa filter unit (Amicon Ultra-15 Centrifugal Filter Unit, 15 mL, Millipore, Catalog # UFC900396) and multiples washes using 100 mM TEABC were carried out to remove excess of salts from the urine.The concentrated urine, the retentate, was transferred in cryovials and stored immediately at −80 • C. The estimated sample size for the discovery phase is five neonates in each condition (Healthy, HIE 1, HIE 2, and HIE 3).Protein concentration was estimated using the Bicinchoninic acid (BCA) assay, and the same was confirmed visually resolving on a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel.Protein normalization was performed across all stages and all the four pooled conditions were split into two halves serving as technical replicates.
Based on the protein concentrations, 250 µg of protein from each condition, that is, 256, 416, 684.5, and 737.6 µL of concentrated urine from Healthy, HIE-1, HIE-2, and HIE-3 respectively was taken.Later, these samples were subjected to reduction of protein cysteine bridges by treating with 5 mM dithiothreitol for 45 min at

Basic pH reverse-phase liquid chromatography (bRPLC)
The pooled sample consisting of labeled peptides was reconstituted using bRPLC solvent A (10 mM TEABC in water; pH 8.5) and loaded on the Xbridge C18 column (4.6 × 250 mm, 2.5 µm; Waters, Milford, MA) and fractionated by basic pH reversed-phase liquid chromatography using a high-performance liquid chromatography (HPLC) system (Hitachi HPLC system-LC2400, Elite, LaChrom).In gradient mode, the peptides were separated by passing solvent B (10 mM TEABC in 90% acetonitrile, pH 8.5) at a 1 mL/min flow rate.The solvent B was increased from 3% to 50% over 120 min.The peptides were fractionated into 96 fractions and concatenated to obtain six fractions.These fractions were dried and stored at −20 • C until LC-MS/MS analysis.

Database search for peptide and protein identification
The raw data files were processed using Proteome Discoverer v2.2 (Thermo Fisher Scientific, Bremen, Germany).The data were searched against the human proteome (v109, RefSeq, NCBI) and common contaminant proteins databases using SequestHT and MASCOT search algorithms.The search was performed with trypsin and semi-trypsin as proteolytic enzymes separately.The remaining search parameters used were: a minimum peptide length of seven amino acids with one maximum missed cleavage.Precursor and fragment level mass tolerances were set at 10 ppm and 0.02 Da, respectively.Carbamidomethylation of cysteine (C) and TMT modification at the peptide N-terminus were set as fixed modifications.At the same time, TMT modification at the lysine residues, oxidation of methionine (M), and acetylation of protein N-termini were set as dynamic modifications.False discovery rate (FDR) cut-off of 1% (q-value <0.01) was applied at peptide spectrum match (PSM), peptide and protein level to remove false-positive identifications.Data normalization was carried out on the total peptide amount using Proteome Discoverer.

Validation by ELISA
We selected Amyloid-beta precursor protein (APP), fibronectin 1(FN1), fatty acid-binding protein (FABP1), and Angiotensinogen (AGT) from the significantly altered proteins identified in the discovery phase for validation.These biomarkers were monitored across different time points (24 and 72 h) and were further validated in a larger cohort (n = 38) using sandwich ELISA (FN1, FABP1, APP from Wuhan Fine Biotech and AGT from Krishgen Biosystems).Due to the COVID-19 pandemic, we could collect n = 8 samples for HIE stage 3.
The ELISA test was carried out as per the manufacturer's instructions and in duplicates.A semi-log regression model curve fit was applied to generate a standard curve.GraphPad Prism-5 was used for statistical analysis.One-way ANOVA with Tukey's post-hoc test was used to compare groups and different time points, and a p-value <0.05 was considered significant.The candidate proteins were evaluated for their capacity to distinguish between healthy and asphyxiated neonates using receiver operating characteristic (ROC) curves based on the ELISA scores, which were calculated and plotted using an R script.

Data analysis
The normalized protein abundances were imported into the Perseus software and analyzed for differential regulation and statistics.The abundance values were transformed to their log10 and tested for statistical significance with the ANOVA test with post-hoc Tukey's testing to determine the p-value and the significant pairs.The FDR corrected p-value (q-value) was also calculated using the Permutations-based FDR.The abundance values were used for the fold change calculations with respect to the healthy group, of which the proteins having an FC cut-off >1.5 and <0.67 were considered biologically significant.

Baseline characteristics of the asphyxiated neonates
A total of 10 healthy neonates and 28 asphyxiated neonates were recruited for this study.In our study, out of 28 asphyxiated babies 10 were females (35.7%).Also 23 out of 28 asphyxiated neonates were term (between 37 and 41 weeks) (82.1%) and 21 out of 28 asphyxiated neonates were normal-weight babies (2.5-3.5 kg) (75%).The sample collection details, characteristics of the neonates, along with their antenatal details, are presented in (Table S1).Among the various risk factors associated with birth asphyxia, meconium-stained liquor was found statistically significant (p-value = 0.002).Low APGAR score can predict birth asphyxia.In our study, APGAR scoring at 1 min was below 3 for all the HIE-3 neonates and was statistically signifi-cant (p-value ≤ 0.001).Table 1 outlines the risk factors and neonatal characteristics of our study.
The details of biochemical parameters including Arterial Blood Gas (ABG), Serum electrolytes used for the classification of birth asphyxia provided in Table 2. Most of the neonates in our study had metabolic acidosis, which is a common phenomenon in asphyxiated neonates.
19 (67%) neonates were having suspected sepsis, although only 4 had positive blood cultures.In our study, 14 (50%) and 10 (35%) had abnormal urea and potassium values with a significant p-value of 0.02 and 0.001, respectively.However, the levels of blood Creatinine, Calcium, and Sodium were not found to be significant for HIE staging.The Chisquare test and One-way ANOVA were performed to determine the p-value of these clinical parameters.Hence in our study, there is a statistically significant correlation between few risk factors of birth asphyxia and clinical and biochemical parameters which are used to classify birth asphyxia.

Quantitative proteomics highlights differentially altered proteins across different stages of HIE
We performed a global proteomic profiling of the urine samples of healthy and asphyxiated neonates categorized based on the severity of HIE (Stage 1-mild, Stage 2moderate, and Stage 3-severe).For the discovery phase, five urine samples each were collected from neonates within 24 h of their life for all the four groups.The schematic of the quantitative proteomic analysis followed by validation is provided in F I G U R E 1 Graphical summary of the comparative urinary proteomics analysis by shotgun proteomics followed by validation using ELISA of healthy and asphyxiated neonates.

Clustering pattern and classification of differentially expressed proteins
K-means clustering analysis grouped the differentially expressed proteins into six clusters.To understand the significance of these dysregulated proteins in the context of HIE, pathway enrichment analysis was carried out using the Enrichr and Reactome tool for each cluster [32].The significant pathways were identified (p < 0.05) and four clusters were selected for further analysis based on their relevance to the disease.The clusters 2, 3, 5, and 6 were narrowed down for pathway enrichment analysis and some crucial pathways enriched in these clusters included disease of programmed cell death, neurodegenerative diseases, amyloid fiber formation, nervous system development, integrin cell surface interactions, and neutrpohil degranulation.In addition to this, collagen degradation and response to metal ions was also enriched, illustrated in Figure S1C.We selected a few of the enriched pathways related to our study (Figure 2D) and proteins involved in them (File S6A).
Further interactome analysis of the differentially regulated proteins was carried out to identify the proteins which are interacting and various pathways associated with them.Nodes with different color gradients were used to visualize the network and identify the highly expressed proteins (based on fold change) in urine.Some of the proteins with high abundance in urine were Fibronectin 1 (FN1), Alpha-hemoglobin-stabilizing protein (AHSP), and Beta-2microglobulin (B2M).The three distinctive clusters were involved in pathways like neurogenerative diseases, amyloid fiber formation, and so on.This led us to understand which differentially expressed proteins were involved in significant pathways (Figure 2E).
As hypoxia and oxidative stress are two leading biochemical pathways that occur during birth asphyxia, we compared the altered proteins of our study against their known proteins.We observed that 32 and 40 altered proteins from our study were common in hypoxia and oxidative stress specific proteins (Figures S2A and S2B), respectively.

Functional annotation reveals enrichment of dysregulated proteins
To better understand the functional relevance of the dysregulated proteins found in our study, GO enrichment analysis was performed.
Analysis of the subcellular localization of proteins showed that intracellular organelle lumen (20.98%), collagen-containing extracellular matrix (17.84%) and secretory granule lumen (11.76%) were the most enriched terms.Similarly, extracellular matrix organization (9.6%), platelet degranulation (4.9%), and regulation of the apoptotic process (8.82%) were among the most enriched terms based on the biological process.Likewise, Proteins with endopeptidase inhibitor activity (4.5%), actin-binding (3.33%), and hormone activity (1%) were most enriched in molecular function category.The association between the top enriched GO terms and these differentially expressed proteins is illustrated by a chord plot in Figure 3 (Files S6B, S6C, S6D).
Biological process classification revealed that the dysregulated proteins like APP, AGT, and S100A8 were involved in the positive regulation of response to external stimulus.Furthermore, the proteins like APP, B2M, and APOA1 were involved in amyloid fibril formation.The molecular functional analysis showed that proteins like APP, AGT, SER-PINA10, SPOCK1, and FN1 were involved in endopeptidase inhibitor activity, actin-binding, growth factor activity, and so on.

Classification of commonly altered proteins
To obtain more insights into the dysregulated proteins, a comparison between differentially expressed proteins of the different HIE stages was performed.Here, we observed 113 proteins commonly altered across all three conditions (Figure S3A).Moreover, we also found 53, 36, and 209 dysregulated proteins exclusive to HIE1, HIE2, and HIE3 stages respectively and can potentially serve as stage-specific biomarkers.For instance, LTF, UMOD, and SPP1specific to HIE1 can serve as early markers.A few examples along with their increasing expression trend across the three stages of HIE has been presented in Table S2.A heat map was generated to explore the expression profile of these common and differentially expressed proteins (Figure S3B).Unsupervised clustering of the commonly altered proteins demonstrated global differences in protein expression patterns in the different stages of HIE compared to healthy.In contrast, technical replicates of each group clustered together.To deduce the various functions of these proteins, we carried out Gene Ontology (GO)-based classification using DAVID (Version 6.8).We found that most of the enriched terms in biological process, subcellular localization and molecular function were similar to the terms found in the GO annotation analysis of differentially regulated proteins (Figure S3C).

VALIDATION OF A SELECTED PANEL OF PROTEINS BY ELISA
Based on the expression pattern, biological and statistical significance, a panel of four proteins was selected for validation by sandwich ELISA.Of these four proteins, Angiotensinogen (AGT) and Fatty acid-binding protein (FABP1), were found to be significantly upregulated, and Amyloid-beta precursor protein (APP) and Fibronectin 1 (FN1), were found significantly downregulated.The FABP1 was upregulated across all the HIE stages within 24 h while as AGT was upregulated in HIE-3.While the APP was downregulared in HIE-3 and FN1 for HIE-2 and HIE-3 (Table S3).The expression of these four proteins was analyzed at24-hour time point for the healthy and asphyxiated neonates.The expression of this panel of proteins was validated by sandwich-based ELISA using urine from 10 healthy neonates and 28 asphyxiated neonates (File S7).Similar expression trend is observed between ELISA and mass spectrometry data (Table 3).All the four markers show a trend in their fold change in both the time points across the three stages of HIE.So, they might serve as differential biomarkers to distinguish between HIE stages.
AGT and FABP1 were significantly upregulated in both the time points in the urine of neonates with HIE than the healthy neonates (Figure 4A,C) and in agreement with our mass spectrometry findings (Figure 4B,D).
On the other hand, APP and FN1 were significantly downregulated in both the time points in the urine of neonates who were having HIE compared to the healthy neonates (Figure 5A,C) and in agreement with our mass spectrometry findings (Figure 5B,D).The maximum standard deviation for each protein in 24 hours was 21.8 for AGT, 22.45 for FABP1, 0.30 for APP, and 3.35 for FN1.The statistical summary of these four proteins is given in Table S4.

Receiver Operative Characteristic (ROC) analysis revealed a biomarker panel for early diagnosis of HIE
The ROC curve was plotted to observe the sensitivity and specificity of these four significantly dysregulated proteins, that is, AGT (Figure 6A), FABP1 (Figure 6B), APP (Figure 6C), and FN1 (Figure 6D) in each stage of HIE.HIE cases and healthy control were selected as the dependent variable and protein levels as the independent variable.The ROC curves of these selected proteins were significant, with a p-value of <0.0001 for all the stages of HIE.The summary of the ROC analysis is given in Table 4.The cumulative ROC curve was plotted for the upregulated proteins to comprehend their sensitivity, specificity, and accuracy as biomarkers.The ROC produced for AGT, and FABP1 was significant with a p-value <0.0001 for all the three stages and area under the curve of 82.3%, 98.6%, and 100% for HIE stages 1, 2, and 3, respectively (Figure 7A,C,E).A similar analysis was carried out for downregulated proteins APP and FN1, which revealed that the ROC curve for this model was also significant, with a p-value <0.0001 for HIE stage 3, while the p-value for HIE stage 1 and HIE stage 2 was <0.01 and <0.001, respectively.This model's area under the curve was 73.7%, 78.7%, and 84.6% % for HIE stages 1, 2, and 3, respectively (Figure 7B,D,F).The specificity of the model AGT and FABP1 was 95%, 95%, and 100% and sensitivity was 58.9%, 97.3%, and 100% for HIE stage 1, 2, and 3, respectively.On the other hand, for FN1 and APP the specificity was 100%, 100%, and 80% along with the sensitivity of 46.1%, 55.8%, and 76.9% for HIE stage 1, 2, and 3, respectively.Hence, the analysis revealed that AGT and FABP1 together formed a better model for all the stages of HIE.The characteristics of the ROC curve for these four proteins are exhibited in Table S5.
F I G U R E 3 GO analysis of the differentially expressed (DEX) proteins.Gene ontology (GO) enrichment analysis of the biological process, cellular component, and molecular function of the DEX proteins.The chord plot relationship between a few proteins and the top enriched GO terms.

TA B L E 3
The differences in the mean concentration of early urine biomarkers in all the three stages of HIE in neonates from sandwich ELISA were compared with the mean concentration of healthy neonates and the same have been shown in fold change differences.

(A) (B) (D) (C)
F I G U R E 4 Boxplot (A) and (C) depicts the expression of AGT and FABP1 proteins in asphyxiated neonates at 24 and 72 h compared to healthy neonates from ELISA.PSM (B) and (D) with respective reporter ion abundances depicts the quality of identification and expression of AGT and FABP1, respectively from proteomics.

DISCUSSION
The extensive inception of therapeutic hypothermia as a standard of care for HIE has increased the pressure on clinicians to make an early and precise assessment of the degree of hypoxic injury (HI) that has occurred and the severity of the encephalopathy that will ensue [33].
No single urine-based marker at present is robust enough to detect significant brain/hypoxic injury or predict its outcome.Combining clinical and biochemical data with "omics" technology is currently the most likely path toward improved detection and prediction of outcomes in neonatal HIE [34][35][36].We performed this case-control cohort study intending to identify potential biomarker(s) for detecting early HI brain injury in asphyxiated neonates.We generated quantitative proteomics data to investigate diagnostic biomarkers of neonatal HIE.The differentially expressed urinary proteins between HIE neonates and healthy controls were screened at different time points.This led in identification of an early prognostic biomarker panel that may predict brain injury and disease progression in asphyxiated neonates, though further validation is required.
The clinical assessment of the asphyxiated neonates indicated that in our study, the predisposing factors like meconium stain liquor and male sex were statistically significant for developing birth asphyxia in concordance with previous literature [37,38].The evaluation of renal function test (RFT), Arterial blood gas (ABG), and serum electrolytes are normal in healthy neonates.Their abnormality is one of the clinical indications of birth asphyxia.Previous literature shows the evidence of renal abnormalities and their correlation with the degree of asphyxia [39,40].The degree of serum electrolytes, that is, sodium, potassium, and calcium, is related to the severity of birth asphyxia [41,42].This corresponds with our findings where urea and potassium were significant, with a p-value of 0.02 and 0.001, respectively.The abnormal ABG values revealed that most neonates were predisposed to metabolic aci-dosis [43].One of the outcomes of our study was probable sepsis which corresponds with the reported literature [44,45].
The pathway analysis identified pathways like the disease of programmed cell death, neurodegenerative diseases [46], innate immune system [47], and amyloid fiber formation [48], has been associated with hypoxic injury and poor neurodevelopmental outcomes.Similarly, enriched gene ontology terms like amyloid fibril formation, positive regulation of response to external stimulus, and hormone and growth factor activity might have a role in brain injury and poor neurodevelopmental outcomes.
Interestingly, we identified proteins in correlation with the disease pathways, and a panel of four proteins was shortlisted for validation.
These four proteins were APP, FN1, FABP1, and AGT.We evaluated this panel of proteins as potential biomarkers for the early diagnosis of hypoxic injury.APP and FN1 were significantly downregulated, while AGT and FABP1 were upregulated in our data.As discussed below, the role of these proteins is well established in neonatal diseases but remains unexplored or markedly less in birth asphyxia and HIE.
FN1 is an extracellular matrix protein that plays a critical role in cytoskeletal organization, cell cycle progression, growth, and cell survival and differentiation [49].In our study, it is related to the pathways like post-translational protein modifications and degradation of the   [50].FN1 is overexpressed in developing embryos and may also be involved in early blastocyst formation [51].
The role of FN1 is not much explored in HIE, but the knowledge of fetal FN results may reduce preterm birth before 37 weeks and its upregu-lation has been associated with preeclampsia [52,53].Prematurity and preeclampsia are the risk factors for birth asphyxia [54].So, FN1 might have a role in birth asphyxia.In our study, it is showing a decreasing trend across all the stages of the HIE in 24-and 72-h time points as observed by mass spectrometry and validated by ELISA.
AGT plays a crucial role in the pathophysiological modulation of cardiovascular functions.It is the primary trigger for generating reactive oxygen species (ROS) in various tissues [55].Public database such as The lack of AGT has also been associated with failure of BBB repair after an injury in mice [56].Also, the brain-specific renin-angiotensin system (RAS) plays essential role in brain homeostasis.One of the receptor, that is, AT 2 R expression was found in developing fetal tissues, which reduces after birth and maintains a relatively low level during adulthood [57].Although there is less literature available showing the association of AGT with HIE [58], due to its role in BBB repair and brain homeostasis it might have a role to play in brain injury as well.Our val-idation experiment shows the increase in expression pattern at both time points with the severity of HIE.
APP is a transmembrane protein expressed mainly in the brain.It is responsible for neurodegenerative diseases like Alzheimer's [59] and has a significant role in the migration of nerve cells [60].APP accumulation in the brain is an early marker of brain injury, and the low plasma APP levels have been previously correlated with HIE progression [61,62].An animal study conducted by Benterud et al. exhibited a similar drop in APP levels after neonatal asphyxia [63].It was downregulated across all conditions at both time points and was associated with pathways like neurodegenerative disease and amyloid fiber formation.Hence, our findings for this protein resonates with the previous reported literature.
FABP1 is a cytoplasmic protein that participates in lipid metabolism [64].Elevated FABP disrupts the blood-brain barrier and causes cerebral ischemic injury in mice [65].Higher levels of FABP have been reported in the serum of neonates having a hypoxic injury [66].
According to the Human Developmental Biology Resource (HDBR), the RNA-seq data of prenatal human brain development shows FABP1 expression in the cerebellum, medulla oblongata, and spinal cord regions of neonatal brain [67].Therefore, FABP1 is present in the brain of neonates and brain injury can release them to the circulatory system.
However, a detailed study on this with negative controls (neonates with other diseases) can further substantiate this claim.In our study, the fold change of this upregulated protein increases with the severity of HIE, indicating the disease progression.The pathways associated with this protein in our study are cellular response to stress and innate immune system.
Birth asphyxia is a multifactorial disease; we compared and examined the collective effect of these proteins as opposed to their individual impacts.A multivariable biomarker panel approach is preferable as it offers accuracy, sensitivity, and specificity [68,69].The cumulative ROC curve provides better prediction of the disease than individual ROC curve.Measuring a panel of proteins for efficient categorization of disease conditions from the healthy is more effective than measuring single proteins [70,71].In our study, the ROC curve produced for AGT and FABP1 and APP and FN1 was significant for all three stages of HIE.
This makes it a highly sensitive prediction model with high accuracy, sensitivity, and specificity.
Our findings suggest the involvement and possible role of FN1, AGT, APP, and FABP1 in HIE, in a stage-dependent manner.These proteins can be developed as potential biomarkers for asphyxiated neonates with hypoxic injury.A rapid test detection kit when used in a hospital setting could aid in providing accurate disease diagnosis by complementing clinical examination and expertise of the clinician.It will also help to assess the treatment efficiency and prognosis of the disease.As this is a preliminary study, further studies are necessary to characterize and validate the functional role of these proteins in disease progression and brain injury.

CONCLUSION
Our quantitative study aimed to explore the urinary proteomic expression profile of the asphyxiated neonates using a high-throughput proteomics approach.We have developed an extensive catalog of differentially expressed proteins in HIE.The findings of this study enabled us to gain better insights into the disease pathology by relating the role of dysregulated proteins with their associated pathways.Our analysis and validation present a sensitive biomarker panel of four proteins that have the potential to be developed into a multiparameter rapid testing kit to explore its prospective in clinical settings.The biomarker panel developed in this study exhibits better diagnostic accuracy than the individual proteins.Also, as urine is non-invasive and easy to collect in neonates a urinary diagnostic kit has an additional advantage over other body fluid kits.The development of a rapid detection kit can be helpful in the early screening of brain injury, which is a need for the hour.

Limitations
As this is a preliminary study, we have used a modest sample size (n = 38) for the validation of biomarker panel.Cross-validation of these potential biomarker panels in the clinical setting using a larger cohort is required to explore their potential as a biomarker for HIE detection.This will further confirm their accuracy, precision, sensitivity, specificity, and positive and negative predictive values for a probable biomarker.The median CV% for the technical replicates of the total identified proteins was 22.9% and 22.8% for Healthy, 21.8% and 23.4% for HIE-1, 21.8% and 22.2% for HIE-2, 23.7% and 23.4% for HIE-3.
Another limitation is the absence of samples after the hypothermia treatment.

DATA AVAILABILITY STATEMENT
The MS raw data and the Proteome Discoverer-searched data have been publicly made available by submitting to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org)via the PRIDE repository with the data set identifier PXD031986 (Username: reviewer_pxd031986@ebi.ac.uk,Password: KWvdEmKe).

ORCID
Sneha M. Pinto https://orcid.org/0000-0001-8541-8761 Arun AB https://orcid.org/0000-0003-3140-9266 60 • C followed by alkylation for 60 min with 10 mM iodoacetamide at room temperature under dark condition.The proteins were then precipitated with 6× ice-cold acetone overnight at −20 • C. Precipitated proteins were separated by centrifugation at 4 • C with 12,000 rpm for 20 min and reconstituted in 50 mM of TEABC and digested with TPCK-trypsin overnight at a ratio of 1:20 of enzyme/protein at 37 • C. Digestion efficiency was evaluated by resolving the samples on 10% SDS-PAGE and by checking the percentage of peptides detected with 1 maximum missed cleaves and it resulted in the identification of 15.99% and 17.33% of peptides with 1 missed cleaves.The peptides were dried overnight in SpeedVac and stored at −20 • C until labeling.
All the six fractions were subjected to C18-based desalting, where each fraction was cleaned separately using C18 StageTip.The C18 bed was activated and equilibrated by passing 100% acetonitrile (ACN) and 0.1% formic acid (FA) at a slow rate.Peptides were reconstituted with 0.1% FA and loaded onto the C18 StageTip.Unbound contents were removed by a wash using 0.1% FA, and bound peptides were made to elute by passing elution buffer (0.1% FA in 40% ACN).The eluted peptides were dried and reconstituted in Mobile Phase A (0.1% FA) before the LC-MS/MS acquisition.Thermo Scientific Orbitrap Fusion Tribrid mass spectrometer (Thermo Fischer Scientific, Bremen, Germany) connected to Easy-nLC1200 nanoflow liquid chromatography system (Thermo Scientific) was used for data acquisition.Each fraction was reconstituted with Mobile Phase Aandloaded onto a trap column (Acclaim PepMap™ 100, 75 µm × 2 cm, nanoViper, C18, 3 µm, 100Å).Peptides bound to the trap column were separated and eluted by passing Mobile Phase B (0.1% FA in 80% ACN) through an analytical column (nanoViper, 75 µm silica capillary, 2 µm C18 Aq) at a flow rate of 300 nL/min.The percentage of Mobile Phase B was increased gradually from 5% at 0 min to 40% in 75 min.This was further increased to 70% in 23 min, followed by 100% in another 7 min, and kept at the same level for 15 min.The total run time for each fraction was 120 min.The analytical column temperature was set to 45 • C throughout the sample acquisition process.The raw data was acquired in data-dependent acquisition (DDA) mode with a Top speed of 3 s.The Orbitrap mass analyzer acquired peptide precursors within m/z of 400-1600 mass range at a mass resolution of 120K (200 m/z).Automatic gain control (AGC) target and maximum injection time (Max.IT) for precursor scan were set to 2e 5 and 5 ms, respectively.Precursors with charge (z) state 2-8 and minimum intensity of 5e 4 were isolated with a 1.6 m/z isolation window and fragmented in HCD with 34 ± 3%.The AGC target and Max.IT for the isolated precursors was set to 1e 5 and 200 ms, respectively.The resulting fragments between 110 and 2000 m/z were acquired in Orbitrap mass analyzer at a mass resolution of 60K (200 m/z).The dynamic exclusion rate was set to the 30 s. Data acquisition was carried out in technical triplicates for both the technical replicates.

Figure 1 .
Figure 1.From the tryptic and semi-tryptic database searches, 1094 and 1099 proteins were identified respectively.Out of these 656, 389, 41 were high, medium and low confidence proteins from tryptic database search.And 824, 193, 74 were high, medium and low confidence proteins from semi-tryptic database search (File S3).The comparative proteomic analysis resulted in identification of 8654 nonredundant peptides corresponding to 1201 proteins, of which 1095 were quantifiable (Figure S1A).The fold change (FC) of these 1201 proteins and their respective p-values and q-values are given in (File S4).A correlation matrix plot was generated to observe the pattern in which these four groups correlate with each other, which is shown in Figure S1B.The clustering of the groups is based on the biological replicates and technical triplicates.The Pearson's correlation coefficient was equal to 1, that is, in the case of similar stages and their replicates.On the contrary, negative co-relation or minimal positive co-relation was observed when the three stages of HIE were related to healthy, indicating variance.The inter individual variability of proteins can help in differentially diagnosing between the different stages of HIE.

2
Clustering pattern of differentially expressed proteins.Volcano plots showing the differentially expressed proteins in HIE1 (A), HIE2 (B) and HIE3 (C) stages when compared with healthy neonates.(D) Sankey diagram showing the involvement of DEX proteins in various biological pathways (E) Interactome analysis depicting expression of DEX proteins in urine.

5
Boxplot (A) and (C) depicts the expression of APP and FN1 proteins in asphyxiated neonates at 24 and 72 h compared to healthy neonates from ELISA.PSM (B) and (D) with respective reporter ion abundances depicts the quality of identification and expression of APP and FN1, respectively from proteomics.

F I G U R E 7
Cumulative ROC curves (A), (C), and (E) depicts the biomarker model for upregulated proteins AGT and FABP1 in HIE1, 2 and 3, respectively.Cumulative ROC curves (B), (D), and (F) depict the biomarker model for downregulated proteins APP and FN1 in HIE1, 2 and 3, respectively.BrainSpan Atlas of the Developing Human Brain shows AGT expression in 37 weeks after post conception and early postnatal (1-3 years).
Risk factors and neonatal characteristics in the study.

TA B L E 2
Clinical parameters of the study.
Summary of Receiver Operative Characteristic (ROC) curve features for urine biomarkers.
in amygdala, cerebellum, dorsolateral prefrontal cortex, hippocampus, medial dorsal nucleus of thalamus, primary visual cortex, and striatum regions of neonatal brain