Biomarker Discovery in Childhood Asthma: A Pilot Study of Serum Metabolite Analysis for IgE-Dependent Allergy

Introduction

Asthma is the most common chronic respiratory disease in children [1,2]. It is characterized by a cough, especially at night, or a cough caused by other factors such as exercise, wheezing, and breathlessness [3]. Asthma in children <6 years of age is diagnosed based on clinical criteria, including characteristic symptoms, severity of symptoms, and response to treatment, and family history of asthma [3]. For diagnosing children under the age of 5 years, it is crucial to base the diagnosis on the guidelines developed by the Global Initiative for Asthma (GINA) [4]. In addition, it is necessary to obtain information on family and patient history of allergic diseases, which can occur with the first asthma symptoms. If the diagnosis is made accurately and early, the patient not only achieves faster symptom control but also has a lower risk of irreversible airway obstruction. This is particularly important in the youngest patients when the disease begins at a stage of intensive child development. In children over the age of 5 years and in adults, respiratory function tests such as spirometry are used to confirm the diagnosis [3]. A spirometry test measures airflow through the airways and its variability by 2 parameters: forced expiratory volume (FEV1) and forced vital capacity (FVC) [5]. However, it is impossible for children under 5 years of age to perform a spirometry test due to lack of patient cooperation [6]. Therefore, there is a need for new diagnostic methods that can be used regardless of age to facilitate early diagnosis of asthma.

A further challenge in diagnosing asthma is that it frequently co-occurs with allergic conditions, such as allergic rhinitis, atopic dermatitis, and food allergy [4,7]. This makes asthma a complex, heterogeneous disease with various endotypes and phenotypes. Also, depending on geographical location, climate, or age group, the severity of allergic diseases can vary [8]. Allergic asthma is the most common asthma phenotype in children and is mainly associated with inhalant/environmental allergens [9,10]. The frequent occurrence of asthma and allergy together suggests that their mechanisms are related to Th2 cell responses [11]. Looking at phenotypic features seems insufficient to accurately diagnose childhood asthma. Looking upstream (at metabolites and proteins) may be informative and facilitate a better understanding of the mechanisms underlying different asthma phenotypes and endotypes.

Due to the heterogeneity of asthma and allergy diseases, their phenotypes and endotypes are being investigated using different omics disciplines such as metabolomics, proteomics, transcriptomics, and genetics [12,13]. Metabolomic analysis involving the most advanced analytical techniques can contribute to differentiation between the patient groups studied to characterize specific disease endotypes. Metabolomic methods allows the analysis of low-molecular-weight compounds, making it possible to compare metabolite groups such as amino acids, acylcarnitines, biogenic amines, phosphatidylcholines (PC), or sphingomyelins (SM). In addition, this discipline can be used as a tool to gain knowledge and understanding of the underlying mechanisms responsible for the metabolic variations in asthma and related diseases. Metabolomics studies have focused on different targets and the search for potential asthma disease biomarkers in various biological samples [14–17]. In addition, previous metabolomic studies looked for differences in the metabolomic profile associated with response to pharmacological treatment [18,19] and disease severity [20,21] in children diagnosed with asthma.

The selection of clinical material is one of the most important steps in planning a metabolomic experiment. A variety of biospecimens can be used in studying asthma, including plasma [22,23], serum [24], urine [23,25], saliva, induced sputum [26], exhaled-breath condensate (EBC) [27,28], nasal lavage fluid, bronchoalveolar lavage fluid (BALF) [29], and stool samples [30]. The factors considered during matrix selection are relevance and suitability to the research question and ease and invasiveness of sample collection procedures [30]. We decided to use serum samples in our experiment. Although this biospecimen is not specific to airway physiology, it contains molecules secreted, excreted, and discarded by tissues all over the body. Therefore, serum provides an integrative view of the metabolic status of the organism and potentially offers a large number of metabolites that can be determined [30]. The serum’s protein, lipid, and metabolite composition is well-described, which helps to develop the best sample preparation procedure to minimize the matrix effect. Furthermore, serum is easy to collect from blood. Blood collection is minimally invasive, and its protocol is standardized, which ensures good repeatability and reliability of the results [31]. Most omics studies conducted in childhood asthma were based on analysis of EBC samples [27,32–34] or urine samples [25,35,36]. Therefore, our study involving the serum of preschool children provided new insights into metabolic perturbations occurring in asthma.

The study analyzed serum samples from patients under 5 years of age who were clinically diagnosed with asthma with IgE-dependent allergy and children without asthma and IgE-dependent allergy. The objective of this study was to find differences in metabolomic profiles between these groups. A high-throughput targeted metabolomic methodology was employed to gain a broad overview of metabolic abnormalities in the serum of young patients with asthma and IgE-dependent allergy. The heterogeneous nature of asthma and allergy frequently becomes apparent at an early age, which can complicate the diagnostic process. Children under 5 years of age are a large group of patients who frequently visit doctors. Therefore, a standardized diagnosis at an early stage of life is essential. By studying the metabolic changes associated with asthma and allergy phenotypes, it may be possible to discover specific markers that could enable early diagnosis, even before the onset of clinical symptoms. Such biomarkers would not only help distinguish asthma and allergy from other respiratory diseases, but also provide insights into the underlying pathophysiology of the diseases. Therefore, this study compared metabolite concentrations in serum samples from 13 children with asthma and IgE-dependent allergy and 10 children without asthma and IgE-dependent allergy, aged 6 months to 5 years, using the AbsoluteIDQ p180 kit, and analyzed by triple quadrupole tandem mass spectrometry coupled to high-performance liquid chromatography.

Material and Methods

CHEMICALS AND REAGENTS:

The targeted metabolomic analysis was performed using an AbsoluteIDQ p180 kit (Biocrates Life Sciences AG, Innsbruck, Austria). To prepare the samples and phases, the following solvents were required: water from a Simplicity UV water purifying system (Merck Millipore, Darmstadt, Germany), liquid chromatography-mass spectrometry (LC-MS)-grade methanol, acetonitrile, isopropanol, formic acid, phenylisothiocyanate (PITC), pyridine, ammonium acetate, and phosphate-buffered saline (PBS) (Merck KGaA, Darmstadt, Germany).

STUDY SUBJECTS:

The study was approved by the Bioethics Committee of Poznań Medical University, Poland (decision no. 960/22). The research was conducted in accordance with the Declaration of Helsinki [37]. All parents or guardians of study participants provided written informed consent before sample collection. The study was conducted on serum samples from 2 groups of subjects aged between 6 months and 5 years: the study group of children with asthma and IgE-dependent allergy (n=13) and the control group of healthy children without asthma and IgE-dependent allergy (n=10) (Table 1). The asthma diagnosis was made by a pediatric allergy specialist. The main criterion for inclusion in the study group was the presence of at least 3 episodes of bronchial obstruction (characterized by expiratory wheezing, coughing, shortness of breath, and difficulty breathing) with documented improvement following the administration of short-acting beta-agonists (SABA) or even a single but severe episode requiring systemic steroids and/or hospitalization, followed by improvement after 2–3 months of inhaled glucocorticosteroid therapy, and this was sufficient for inclusion. In children who were able to cooperate effectively, bronchial obstruction was additionally confirmed using forced oscillometry. This test can be performed in children over the age of 3 years and is one of the few available diagnostic tools for younger children, in whom standard methods such as spirometry are not feasible due to limited cooperation. Additional diagnostic criteria included: wheezing, shortness of breath, and coughing triggered by physical activity, play, laughter, or crying; exclusion of other causes of wheezing; as well as the presence of atopic diseases and/or asthma in the family, particularly in parents. To confirm or exclude the diagnosis of asthma, each child was evaluated by a specialist 2 to 3 times over a 6-month observation period. The study group included patients with symptoms indicative of bronchial asthma, while the control group consisted of children with recurrent respiratory infections but no clinical signs of bronchial asthma. In patients’ serum blood smear, determination of total immunoglobulin E (IgE), immunoglobulin M (IgM), immunoglobulin G (IgG), and immunoglobulin A (IgA) concentrations, and determination of IgE antibodies to selected 30 allergens using the Atopic Polycheck 30-I panel (Biocheck, Münster, Germany) were performed. Participants were recruited from the Department of Pediatric Pneumonology, Allergology and Clinical Immunology, K. Jonscher Clinical Hospital, Poznań University of Medical Sciences, and the medical practice of Joanna Matysiak, Kalisz.

SAMPLE COLLECTION:

Blood from preschool-age patients was collected during medical examination into 7.5 mL tubes with a clotting activator (S-Monovette^® Serum Gel CAT, Sarstedt, Nümbrecht, Germany). After the collection to obtain serum, blood was centrifugated at 2500×g for 10 min at 20°C. Serum samples were stored in the freezer at −80°C between collection and analysis.

SAMPLE PREPARATION:

The serum samples were analyzed using the AbsoluteIDQ p180 kit (Biocrates Life Sciences AG, Innsbruck, Austria). This methodology enables the simultaneous quantitation of 188 analytes, including 21 amino acids, 21 biogenic amines, 40 acylcarnitines, 90 glycerophospholipids, 15 sphingolipids, and the sum of hexoses. The assay is based on a two-layer, 96-deep-well plate with selected isotope-labeled internal standards partially integrated into the filter layer of the plate. The first step of the analytical procedure was to pipette 10 μL of a mixture containing the rest of the internal standards and 10 μL of serum on the filter layer of the plate. The plate was dried for 30 minutes under nitrogen flow. In the next step, 50 μL of derivatization solution (5% phenylisothiocyanate) was added to each well. After 25 minutes of incubation at room temperature, the plate was dried for 60 minutes under nitrogen flow. Then, 300 μL of 5 mM ammonium acetate in methanol serving as extraction solvent was put into each well. Afterward, the plate was shaken at room temperature for 30 minutes using an orbital plate shaker at 450 rpm. The sample extracts were passed through the filter layer to the lower capture plate using a positive-pressure manifold 96 processor (Agilent Technologies, Santa Clara, CA, USA). After the extraction, 150 μL of the final extract was transferred to a second 96-deep-well plate. The samples intended for liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis were diluted with water, whereas the flow injection analysis (FIA) mobile phase was used to dilute the ones intended for flow injection analysis-tandem mass spectrometry (FIA-MS/MS) analysis. Each plate was sealed with a silicone mat and placed into an autosampler until analysis.

LC-MS/MS ANALYSIS:

The quantification was carried out using a liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) system comprising 1260 Infinity high-performance liquid chromatography (HPLC) (Agilent Technologies, Santa Clara, CA, USA) coupled with 4000 QTRAP mass spectrometer (Sciex, Framingham, MA, USA). The separation of amino acids and biogenic amines was performed on a ZORBAX Eclipse XDB-C18 (3.0×100 mm, 3.5 μm) column (Agilent Technologies, Santa Clara, CA, USA) with a pre-column (C18, 4.0×3.0 mm) SecurityGuard (Phenomenex, Torrance, CA, USA). The injection volume of the sample in LC-MS/MS analysis was 10 μl. A 0.2% aqueous formic acid solution was used as mobile phase A, and 0.2% formic acid in acetonitrile served as mobile phase B, at a flow rate of 0.5 mL/min. The gradient was started with 100% of mobile phase A, decreased to 5% phase A between 0.5 and 5.5 min, and then held at 5% of phase A from 5.5 to 6.5 min. Finally, the gradient returned to 100% of phase A (6.5–7 min) and was maintained at these conditions from 7 to 9.5 min. The column temperature was set at 50°C. The parameters of the mass spectrometer are shown in Table 2.

FIA-MS/MS ANALYSIS:

Lipids, acylcarnitines, and hexoses were analyzed using an FIA-MS/MS. The injection volume of the sample in this analysis was 20 μl. We used 290 ml of methanol with 1 ampoule of FIA mobile phase additive as the FIA solvent, with the flow rate varying in time from 0.03 mL/min (0.0–1.6 min) to 0.20 mL/min (2.4–2.8 min) and 0.03 mL/min again (3.0 min). The instrument parameters were set according to the manufacturer’s protocol for the AbsoluteIDQ p180 kit (Table 2) and were provided in our previous paper [38].

INTERNAL STANDARDS AND QUALITY CONTROL SAMPLE PREPARATION:

To assess the quality of the assay and calculate metabolite concentrations, internal standards and quality control samples from the AbsoluteIDQ p180 kit were prepared. The internal standard mix was reconstituted in 1200 μL of water and shaken for 15 minutes at 1200 rpm prior to addition to the kit plate. The quality control samples were reconstituted in 100 μL of water, shaken for 15 minutes at 1200 rpm, and centrifuged at 4°C for 5 minutes at 2750×g. Then, they were managed in the same way as the experimental samples.

DATA ANALYSIS:

The metabolomic data were obtained using Analyst software version 1.6.3 (Sciex, Framingham, MA, USA) and analyzed using Biocrates WebIDQ software. The compatible MetaboINDICATOR software (Biocrates Life Sciences AG, Innsbruck, Austria) was used to obtain sums and ratios of metabolites. After LC-MS/MS and FIA-MS/MS analyses and obtaining concentration values, the data were reviewed, and any variables with more than 50% missing data were then excluded. Missing values in the remaining data were filled in with the group’s average. A statistical analysis of the data included was performed. All statistical analyses were conducted using PQStat software 2024 (PQStat v.1.8.6.122., Poznań, Poland). A value of P<0.05 was considered statistically significant.

The normality of the data distribution was first assessed using the Shapiro-Wilk test because the study group consisted of less than 100 patients. Then, the t test or Mann-Whitney U test was performed. These 2 statistical methods were used to search for statistically significant differences between the groups. When the data had a normal distribution, the t test was used to compare the means of the 2 groups, but when the data did not have a normal distribution, the Mann-Whitney test was used for non-parametric variables to compare the ranks of the 2 groups (Figure 1).

In addition to statistically significant results, single-factor receiver operating characteristic (ROC) curves were plotted. The ROC curve is a graphical tool used to evaluate the performance of classification models. The curve helps visualize how effectively a model distinguishes between positive (sensitivity) and negative (specificity) classes across different thresholds. Sensitivity is the probability of predicting that a diseased individual has the disease, and specificity is the probability of predicting that a non-diseased individual does not have the disease [39]. Sensitivity is the ratio of true positives (TP) to the sum of true positives and false negatives (FN), and specificity is the ratio of true negatives (TN) to the sum of true negatives and false positives (FP). The true positive rate (TPR) is plotted on the y-axis, and the false positive rate (FPR) is plotted on the x-axis to create the ROC curve. The area under the curve (AUC) provides a summary measure of the ability of a test to discriminate between the positive and negative classes. AUC values higher than 0.5 and closer to 1.0 indicate better performance and distinguishing between classes [39]. AUC and confidence interval (CI) were calculated using the DeLong method. A cut-off point determines a threshold value below or above which values are classified as positive or negative. This study’s cut-off point was calculated using the Youden index cut-off point method.

Results

Out of a total of 188 metabolites covered by the applied method, 131 metabolites were detected and quantified in the samples (Table 3). Internal standards and quality control samples prepared for assay quality assessment were within acceptable ranges. Table 4 provides information on the number and groups of determined metabolites. The most abundant group of metabolites found was the phosphatidylcholines, which accounted for 62% of the total determined metabolites. Additionally, of all 136 sums and ratios of different metabolites calculated with the MetaboINDICATOR tool, 48 had less than 50% missing data (Table 5). The 48 sums and ratios of metabolites were also included in the statistical analysis.

Depending on the normality of data distribution, t tests and Mann-Whitney U tests were performed to find statistically significant differences in levels of metabolites between the analyzed groups. The statistically significant differences were found for the concentrations of 6 metabolites from the glycerophospholipids and sphingolipids class of metabolites and 1 metabolite ratio - serotonin synthesis (Table 6, Figure 1). These 6 metabolites are lipid compounds, 4 from the glycerophospholipid class and 2 from the sphingolipid class. Concentrations of all metabolites were decreased in the study group compared to the control group PC 36:0 (1.06±0.42 μM), PC 38:0 (2.32±0.77 μM), PC 28:1 (2.20±0.38 μM), SM 44:2 (0.36±0.08 μM), SM 35:1 (2.60±0.42μM), and serotonin synthesis (0.014±0.006 μM), except for the metabolite PC 40:4 (2.97±0.53 μM), which was higher in the study group. Values, including median, average, P value, and statistical tests, are summarized in Table 6 to illustrate the differences between groups. Of the acylcarnitines, amino acids, and biogenic amines groups, none of the determined levels of metabolites were identified as significantly different between groups. This may be due to the large dispersion of results within the analyzed groups (Table 3).

The next step in the statistical analysis of the data was using ROC curve analysis, which enabled the calculation of sensitivity, specificity, AUC, and cut-off points to validate the diagnostic utility of the differentiating metabolites (Table 7). The greater the AUC, the better the model classifies the samples into groups. The 5 metabolites – PC 40:4, PC 36:0, PC 38:0, SM 44:2, and SM 35:1 – and serotonin synthesis ratio had AUCs of 0.70 and higher. Nevertheless, only serotonin synthesis and PC 40:4 were found to have an AUC of 0.80, which suggests their potential diagnostic utility in diagnosing childhood asthma with IgE-dependent allergy [40]. An essential indicator for verifying true negatives and positives is specificity and sensitivity, which affect the appropriateness of the choice of diagnostic marker. The ratio of serotonin synthesis and PC 40:4 had the highest specificity of 100% and 90%, respectively. Sensitivity was highest for SM 35:1 (100%), PC 36:0, PC 44:2, and PC 28:1 (92.3%), possibly indicating their diagnostic utility due to reduced false-negative results. Most metabolites were de-stimulants, so the increased levels of this parameter indicate the reduced chance of the occurrence of the disease. In the performed ROC curves analysis, only serotonin synthesis and PC 40:4 had a P value below 0.05, both of which were 0.01 (Table 7). A P value of less than 0.05 indicates that the null hypothesis has been rejected. The null hypothesis states that the classifier’s ability to discriminate between classes is no better than random guessing.

We also looked for a multi-metabolite marker that can be useful in differentiating the 2 groups of patients enrolled in the study. Multivariate ROC curve analysis was used to test the diagnostic utility of 2 variables (serotonin synthesis and PC 40:4) together. These variables were selected for the multivariate ROC curve analysis because they exhibited the best performance in univariate ROC curve analysis, and their P values were 0.01 (Table 7). This statistical analysis showed a result of AUC (0.89), cut-off line (0.668), sensitivity (76.9%), specificity (100%), and P value of 0.001 (Table 7, Figure 2). A graphical presentation of the AUC measured by the DeLong method is presented in Figure 2. An AUC of 0.89 means that the combination of selected parameters can predict disease occurrence with a very high degree of accuracy. In addition, the AUC value was higher than in the case of the individual variables, indicating a higher probability of correct classification of patients between groups by the proposed multiclassifier.

Discussion

Due to the increasing prevalence of childhood asthma, rapid and precise diagnostic methods are needed. Childhood asthma often coexists with allergies and atopic diseases; however, it can also occur without allergies. Many patients under 5 years old who visit the doctors have asthma and allergy. Until now, limited is known about differences at the metabolic level between children with asthma and IgE-dependent allergy and those without asthma and IgE-dependent allergy.

This study searched for serum metabolomic markers in preschool children with asthma and IgE-dependent allergy compared to preschool children without asthma and IgE-dependent allergy. The study found statistically significant differences between patient groups in concentrations of 6 metabolites – PC 40:4, PC 36:0, PC 38:0, PC 28:1, SM 44:2, and SM 35:1 – and serotonin synthesis ratio (Table 6). Most previously conducted omics studies of asthma involved adults, young people, and children aged 5 years and over, for which a wider range of diagnostic tools can be used [22,27,41]. There are significant differences between childhood and adulthood asthma regarding immunology, histopathology, and clinical symptoms. In this study, we narrowed down the age of the patients to 5 years of age and selected a group of patients with asthma and IgE-dependent allergy as the most common disease phenotype of asthma [9]. For the study, we used a targeted metabolomic approach, allowing the determination of a panel of metabolites from 6 different classes of acylcarnitines, amino acids, biogenic amines, phosphatidylcholines, sphingomyelins, and the sum of hexoses.

The lipid metabolites had the greatest differences between the study and control groups. Lipid compounds have essential bodily functions, such as energy storage, transport of substances, and cell development and regulation [42]. The lipids are fatty acyls, glycerolipids, glycerophospholipids, sphingolipids, sterol lipids, prenol lipids, saccharolipids, and polyketides [43]. The methodology allowed us to find the most significant concentration differences between groups among glycerophospholipids and sphingolipids (Table 6). Similarly, other metabolomic studies in children with asthma indicated different levels of lipid group metabolites, suggesting that these merit further investigation [42,44].

Glycerophospholipids include phosphatidylcholine, phosphatidylethanolamine, phosphatidylserine, phosphatidylinositol, cardiolipin, and phosphatidylglycerol. Phosphatidylcholine (PC) is the most abundant phospholipid in all mammalian cells, including lung cells, and has a role in cell signalling [45,46]. Therefore, phosphatidylcholine may have a role in signalling pathways that are involved in inflammatory processes. PC can be converted into metabolites such as lysophosphatidylcholine (LPC) as a result of phospholipase A2 (PLA2) [47]. Also, PC deficiency can lead to cellular dysfunction, and PC metabolism contributes to other lipids’ cellular homeostasis [48]. In this study, the metabolites with the greatest statistically significant differences in the comparison of children with asthma and IgE-dependent allergy and the control group were PC 40:4, PC 36:0, PC 38:0, and PC 28:1, from the 81 determined (Tables 3, 6). Few studies have compared lipid compounds as possible disease biomarkers in studies of childhood asthma up to the age of 5 years. The study by Jiang et al [42] used plasma samples from patients with asthma and healthy controls aged 18 to 72 years. The authors showed that phosphatidylethanolamine (PE), SM, triglycerides (TG), phosphatidylinositol (PI), phosphatidylglycerol (PG), ceramide (Cer) and LPC may be associated with asthma. Furthermore, elevated levels of PE (20:0/18:1) and TG (16:0/16:0/18:1) were observed in patients with asthma and IgE-dependent allergy [42]. This highlights the crucial role of activated mast cells by IgE in the pathogenesis of asthma, particularly in allergic phenotypes. Based on sputum samples from patients with asthma compared with patients with chronic obstructive pulmonary disease (COPD), Correnti et al [49] found 22 metabolites with statistically significant differences. Among these, increased lipids levels such as PC (33:0), PC (32:2), and PC (38:3) were observed. In our study, the increased PC 40:4 level in patients with asthma was noticed. Our observations on the association of PC and asthma are consistent with a study by Ried et al [50], which showed that PC is critical in differentiating in serum patients with asthma. In contrast to the present study, they compared cohorts of adult patients with asthma only with healthy controls. From the analysis of 151 metabolites quantified by directed mass spectrometry, 2 metabolites from the PC group demonstrated statistically significant differences. They attempted to create a combination of omics studies that would link results between the most significant lipids and also of phosphatidylcholines levels in asthma, which can be associated with asthma risk alleles from the PDED3 and MED24 genes based on the measurement of single-nucleotide polymorphisms SNP. In research similar to the present study, Chiu [23] included patients aged 3–5 years old with asthma and allergies. Their results showed significant differences in metabolite levels, especially in the amino acids, in contrast to our work, where we found significant differences in lipid metabolites. They used 2 biofluids – plasma, and urine – which can provide a broader view of metabolomic disturbances occurring in asthma. A more extensive study was carried out by Qu et al [51] on a cohort of 1195 patients aged 6–21 years, with and without asthma. Their methods, based on the nuclear magnetic resonance (NMR) platform, allowed the determination of 249 selected metabolites. Qu et al [51] also found differences in serum lipid metabolites, mainly lipoprotein subclasses, as well as amino acids and ketone bodies, in patients with asthma. Metabolomic studies of childhood asthma conducted so far suggest that lipids may have a role in mechanisms associated with childhood asthma. In addition, attention should be paid to their role in the immune system and inflammation, which may be related to allergic reactions in asthma.

The synthesis of the serotonin ratio was another statistically significant difference observed in our study. The calculation of this variable was based on dividing serotonin by tryptophan, determining the rate of tryptophan conversion to serotonin in 2 steps by tryptophan hydroxylase and aromatic L-amino acid decarboxylase activity. Serotonin synthesis is the biochemical process by which this neurotransmitter is formed from tryptophan by tryptophan hydroxylase (TPH) and aromatic amino acid decarboxylase [52]. During an allergic reaction, IgE antibodies attach to mast cells. Next, calcium uptake leads to mast cell degranulation and the release of proinflammatory mediators, including histamine, tryptase, leukotrienes, prostaglandins, and serotonin [53,54]. In our study, serotonin synthesis levels were lower in the study group compared to the control group. Other studies of asthma showed differences in serotonin levels. A study investigating differences between symptomatic and asymptomatic patients with asthma showed a statistically significant difference in plasma serotonin [55]. Lechin et al [55] involved patients aged 6–43 years, but our study focused on a different age range – from 6 months to 5 years. In that study, increased serotonin levels were observed in patients with symptomatic asthma. In our study, we observed decreased serotonin levels in patients with asthma and IgE-dependent allergy. One study used mice as a model of allergic airway inflammation, sensitized according to the protocol to ovalbumin and house dust mites [56], and they also performed a segmented allergen provocation in patients over the age of 20 years with and without asthma. Dürk et al [56] reported that serotonin accumulates in the airways in allergic airway inflammation, and they focused on the possibility of using serotonin cells in the anti-serotonergic treatment of pathophysiological features associated with asthma. In addition to metabolomic studies, studies on serotonin transporter gene polymorphism and asthma have been performed [54]. These studies used PCR-based methods in patients aged 5–18 years. However, despite the growing interest in serotonin in asthma research, little is known about its effect on changes in the metabolome of patients with asthma. Most studies focused on its genetic influence on the mechanisms underlying the disease. For this reason, further research is needed to gain a better understanding of the actual effect of serotonin on asthma, particularly in children under the age of 5 years. Nevertheless, given the involvement of serotonin in inflammatory processes during allergic reactions, it may be a promising diagnostic indicator for asthma coexisting with IgE-dependent allergy.

In our study, PC 40:4 and serotonin synthesis yielded the best performance in univariate ROC curve analysis, indicating these variables as the most promising classifiers. Thus, we created a multiclassifier combining PC 40:4 and serotonin synthesis as a candidate diagnostic marker for childhood asthma with IgE-dependent allergy. As a result, we obtained a P value of less than 0.05 and an AUC of 0.89, suggesting the possible diagnostic utility of this multiclassifier. The combination of multiple compounds can be used to create a multiclassifier that would show promising potential in diagnostic applications. Popular methods for creating multiclassifiers are those based on machine learning, that is, data analysis that automates the systematic construction of a model [57]. By collecting data and analyzing it to uncover trends, clinical data, symptoms, and outcomes can be combined to select crucial patient features [58]. Features are scanned with different model algorithms to distinguish diseases like asthma from allergic asthma with high accuracy and precision. Another direction is using multi-omic approaches combining genetic, metabolomic, or transcriptomic studies [59,60]. McGeachie et al [61] combined metabolomic results with genome-wide genotype data. The study involved patients with an average age of 13.3 years and aimed to understand the cellular response to treatment and the effect of lipids on airway inflammation. Also, Rago et al [62] used a combination of metabolomic and genetic studies results. They identified the sphingolipid pathway as potentially associated with asthma risk at an early age, which may be related to genetic variants 17q21 and expression of the serine palmitoyltransferase (SPT) enzyme. Few studies have developed and used multiclassifiers to diagnose asthma with IgE-dependent allergy in children. Nevertheless, metabolomic multiclassifiers could play a key role in expanding and accelerating the diagnosis of asthma with IgE-dependent allergy in children, focusing on metabolome studies that would find the most distinguishing compounds in diseased patients. As asthma is a heterogeneous disease, another attractive direction is to develop a classifier that combines metabolites with other influencing factors from the omic sciences.

Some limitations of the present study should be acknowledged. The applied methodology (AbsoluteIDQ p180 kit) has certain limitations, most importantly a limited number of metabolites from 6 major classes (amino acids, biogenic amines, acylcarnitines, glycerophospholipids, sphingolipids, and the sum of hexoses). To better understand metabolic alterations in childhood asthma, the application of a non-targeted metabolomic analysis followed by a targeted analysis would be useful. In addition, the use of more than 1 biological fluid, such as serum, urine, EBC, or saliva, collected from patients will provide a more comprehensive view of the metabolic changes associated with asthma and its related diseases. Another limitation of our study is the size of the study population. Serum samples from a larger study group of patients should be analyzed to validate the result of this pilot research.

Conclusions

The study provided a snapshot of the metabolic disturbances occurring in the serum of childhood asthma and IgE-dependent allergy patients aged 5 months to 6 years. It supplied data that may help to determine further directions for pediatric research and expanded our knowledge of the mechanisms of these co-occurring diseases. Statistically significant differences were found between individual parameters of sphingolipids and glycerophospholipids, which will broaden the future directions in the search for a potential biomarker. In addition, differences in serotonin synthesis ratio shed new light on the mechanisms underlying the development of childhood asthma. A multiclassifier based on the parameters with the most statistically significant differences (serotonin synthesis and PC 40:4) improves the ability to classify by a diagnostic parameter compared to individual parameters. Creating such multiclassifiers can contribute to a more precise and effective diagnosis, allowing early detection, disease monitoring, and personalized treatment of children with asthma and IgE-dependent allergies.