**3. Results**

#### *3.1. Data Analysis for the Personal and Clinical Characteristics of School Children*

The study population was well-balanced between the doctor-diagnosed asthmatic children (52%) and healthy children (48%), in which 60% of them were from urban schools. The majority of asthmatic children tested positive for at least one of the allergens, with 74.0% of them sensitised towards house dust mites (Derp1 and Derf1), followed by cat dander (30.0%) (Supplementary Table S1). The FeNO levels were statistically higher among the doctor-diagnosed asthmatic children than healthy children (*p* < 0.001). The researchers observed that the total percentage of eosinophil count in the sputum samples was slightly higher in the doctor-diagnosed asthmatic children but not statistically different between the two groups (*p* > 0.05) (Table 1).


**Table 1.** Students' characteristic and inflammation status for the doctor-diagnosed asthma and control group.

Values are the median (IQR) for clinical characteristics. IQR = Interquartile range. \* *p* < 0.05; \*\* *p* < 0.001.

Both the eosinophils and neutrophils expressed an activated phenotype in both groups of induced sputum samples. The expression of CD11b was significantly upregulated in both their cell surfaces among the doctor-diagnosed asthmatic children (*p* < 0.05). The expression of CD63 on the neutrophils was also significantly higher in the sputum samples of the doctor-diagnosed asthmatic children as compared to healthy children (*p* < 0.001). Samples from the doctor-diagnosed asthmatic children with a high expression of CD11b (tertiary) and CD63 (azurophilic/crystalloid) on the eosinophil and neutrophil surfaces indicated a moderately degranulated state of sputum granulocytes (Figure 2).

**Figure 2.** Expression profile of sputum granulocytes compared between the doctor-diagnosed asthmatic and healthy school children. Expression of degranulation and activation markers on eosinophil (**A**,**B**) and neutrophil (**C**,**D**). The expression of CD11b as the degranulation and classical activation marker is displayed twice in this graph. \* *p* < 0.05, \*\* *p* < 0.001.

#### *3.2. Levels of Indoor Pollutants and Building Inspection Data*

All the classrooms were designed with natural ventilation, which is equipped with glass windowpanes on both sides of the wall. The classrooms were equipped with an average of three ceiling fans in each classroom. The schools were painted, and the floor surface was furnished with concrete. There were bookshelves, whiteboard, and soft boards in every classroom. Some of the classrooms had window curtains fixed on both sides of the class. The statistical analysis showed a significant difference in all of the indoor environmental parameters between the urban and suburban schools (Table 2).

**Table 2.** Comparison of the environmental parameters between schools located in urban and suburban areas.


*n* = 32; IQR = Interquartile range, Min = Minimum, Max = Maximum. \*\* *p* < 0.001.

#### *3.3. Chemometrics Analysis of the Biomarkers and Indoor Air Pollutants*

A PCA was performed to explain the variance observed between the biomarkers and indoor pollutants in a more efficient way. Prior to the PCA, Bartlett's sphericity and Kaiser–Meyer–Olkin (KMO) tests were conducted to determine the correlation difference and sampling adequacy, with both measurements achieving the required levels. The PCA was applied with a normalization procedure and the coefficient factor loadings produced also expressed the correlation between the variables [36]. The factor loadings of the four factors with eigenvalues >1 were extracted from the eosinophil and neutrophil expression markers. A total of 39.0% of the variation for both Factor 1 and Factor 2 was observed in the expression of markers on eosinophils tested with indoor pollutants. Moderate factor loadings were identified for expression of CD11b and CD35, together with a strong factor loading for the concentrations of NO2 and formaldehyde in Factor 1. In other words, the upregulation of CD11b and CD35 expression in the eosinophils were associated with exposure to NO2 and formaldehyde; in turn, Factor 2 had moderate factor loadings of FeNO levels, PM10, and PM2.5. Moderate factor loadings were observed for FeNO levels, CD66b, and PM10 in Factor 3, with a total variation of 14.9%. High factor loading of CD63 was observed in Factor 4, with a total variation of 13.0%. For neutrophils, a strong factor loading of CD66b expression together with moderate factor loadings for FeNO levels, CD11b, and CD63 were recorded in Factor 1, with a variation of 24.2%. Factor 2 showed a strong factor loading for the concentration of formaldehyde together with a moderate factor loading for the concentrations of CO2, PM10, and PM2.5, with 19.3% of the total variation. There were moderate loading factors for expression of CD35 and the concentrations of NO2 and PM10 in Factor 3, with 14.3% of the total variation. Factor 4 showed moderate factor loadings of CD11b expression and concentration of PM2.5, with 13.2% of the total variation (Table 3).


**Table 3.** Factor loadings using PCA for eosinophils and eosinophils. The moderate (0.5–0.75) and strong (>0.75) factor loadings are highlighted in bold.

We further analysed the biomarkers and indoor pollutants to categorize them based on their homogeneity levels using agglomerative hierarchical clustering (AHC). Our data showed the cluster of markers and indoor pollutants from the PCA and AHC analyses were relatively identical. Three clusters were generated for eosinophils, which consisted of CD66b, FeNO levels, and concentrations of CO2 in Cluster 1, with 98.4% of variance within-class; whereas Cluster 2 presented 56.5% of the variation within-class for CD11b expression and concentrations of NO2 and formaldehyde. High (82.9%) variance within-class was observed for CD35 and CD63 expression and concentrations of PM10 and PM2.5 in Cluster 3; three clusters were also generated for neutrophils, which consisted of CD35, CD63, CD66b, and concentrations of CO2, PM10, and PM2.5, with 95.7% of the variation within-class for Cluster 1. The concentrations of NO2 and formaldehyde were grouped in Cluster 2 with 26.0% of the variation within-class. Cluster 3 showed 56.7% of the variation within-class for CD11b expression and FeNO levels. The clusters generated through this process confirmed the contributing factors in the PCA earlier. This showed that exposure to CO2, PM10, and PM2.5 have resulted in degranulation of both eosinophils and neutrophils, with upregulation of the markers for tertiary (CD11b), specific (CD66b), and azurophilic/crystalloid (CD63). In this study, exposure to NO2 and formaldehyde also triggered the activation of tertiary eosinophil surface markers (Figure 3A).

#### *3.4. Binary Logistic Regression (LR)*

Finally, models were built to predict the toxicodynamic effects of indoor pollutants towards marker expression on the eosinophil and neutrophil in the sputum samples among doctor diagnosed asthmatic children. For this objective, the researchers next used the binary logistic regression to model the prediction with the potential confounders of gender, atopy, parental asthma/allergy status, and area of schools. Cluster 1 showed an overall accuracy of 76.0% in predicting asthmatic children by using CD66b expression markers on the eosinophil and FeNO levels in relation to CO2 exposure. Meanwhile, the model generated for Cluster 2 showed a 76.0% accuracy and 68.0% sensitivity in predicting asthmatic children by using the upregulation of CD11b expression on eosinophils in relation to the NO2 exposure. In the model generated for Cluster 3, the upregulation of CD63 expression on eosinophils and PM2.5 concentration was significantly associated (*p* < 0.05), with a 71.9% accuracy and 60.0% sensitivity. Similarly, the upregulation of the neutrophil expression markers, CD63 and CD66b, was significant (*p* < 0.05), and could be predicted from the PM2.5 exposure with a 72.9% accuracy and 72.0% sensitivity. Overall, children with a status of atopy, parental asthma/allergy, and from urban school were more likely to develop asthma (*p* < 0.05) (Table 4 and Figure 4).

**Figure 3.** Agglomerative hierarchical clustering (AHC) analysis using the Ward linkage method and using Euclidean distances to generate the clustering of degranulation and the activation of markers and environmental pollutants measured inside classrooms for (**A**) eosinophils and (**B**) neutrophils. The dotted line represents the pruning level to generate distinct clusters. § Activation marker for eosinophils; # Activation marker for neutrophils.


**Table 4.** Summary of the binary logistic regression models based on the clusters generated in the AHC analysis.


**Table 4.** *Cont.*

<sup>\*</sup> *p* < 0.05, \*\* *p* < 0.001.

**Figure 4.** Receiver operating curve (ROC) for the models predicting an asthmatic or healthy child classification based on the clusters generated in the AHC analysis. The corresponding predictors for each curve are presented in Table 4. For eosinophils: (**A**) ROC for Cluster 1, (**B**) Cluster 2, and (**C**) for Cluster 3. For neutrophils: (**D**) ROC for Cluster 1 and (**E**) for Cluster 3.

## **4. Discussion**

The role of biomarkers in airways is complex and specific, which is helpful in evaluating the aetiology, characterisation of phenotyping, and treatment of allergy and lung inflammation [37]. In this study, the FeNO levels were significantly higher among asthmatic school children, which are similar with the studies conducted in China [38], Terengganu, Malaysia [39], and Penang, Malaysia [40]. The result showed that there was inflammation in the airways and the average value was above the threshold of 50 ppb, which could reflect a high degree of inflammation. Liu et al. [41] and Carlsen et al. [42] reported that there was a significantly positive relationship between the FeNO levels and almost all pollutants, namely PM10, PM2.5, SO2, NO2, CO, and VOCs. This advocates a relationship between the high levels of all pollutants measured inside the classroom of urban schools and the high levels of FeNO among school children in this study. Some researchers estimated that FeNO is positively correlated up to five-fold and two-fold when exposed to NO2 [43] and finer particles, such as PM2.5 [44], respectively, which could be modulated by DNA methylation in the arginase–nitric oxide synthase pathway [45,46].

The CO2 concentration in both school areas was below the recommended limit of 1000 ppm [47]. Similarly, the PM10 and PM2.5 concentrations were below the 24 h mean of theWorld Health Organisation (WHO) guideline (PM10 = 50 μg/m3, PM2.5 = 25 μg/m3), the National Ambient Air Quality Standard by USEPA (PM10 = 150 μg/m3, PM2.5 = 35 μg/m3), and the new Malaysian Ambient Air Quality Standard 2018 Interim Target-2 (PM10 = 120 μg/m3, PM2.5 = 50 μg/m3) [48]. A few classrooms recorded a concentration of PM2.5 that exceeded the value of 25 μg/m3, especially in the urban areas (37.5%) compared to the suburban (12.5%) areas. The median level of NO2 for the urban and suburban areas was also below the WHO guideline of 40 μg/m<sup>3</sup> (annual mean), with only 18.8% and 25.0% of the classrooms in the urban and suburban areas, respectively, exceeding the limit. Overall, the levels of indoor air pollutants were below the guideline limits. This was due to the su fficient natural ventilation system and a wider window design on both sides of the classroom, together with the adequately equipped ceiling fans. The classroom design has a well-balanced ventilation that suits the temperature of the equatorial region and is able to reduce the particles, NO2, and CO2 concentrations [49]. Additionally, Silvestre et al. [50] reported that an opening of 56% of the classroom windows under natural ventilation conditions was able to keep the CO2 concentration below 1000 ppm.

The schools, classrooms, and children were randomly selected from all secondary schools in the Hulu Langat area, Malaysia. Thus, we concluded that this study was not seriously influenced by selection bias. Moreover, Malaysia has a similar climate all year around; therefore, with the natural ventilation flowing through the windows in the classrooms, the indoor and outdoor levels of pollutants would be expected to be constant throughout the year. This is supported by several studies that have determined equal indoor to outdoor (I/O) ratios for PM10, PM2.5, NO2, CO, and VOCs measured in schools across Peninsular Malaysia [51–53]. Be that as it may, the cross-sectional study design utilized here preludes making conclusions on causality.

In contrast to earlier findings, the total percentage of eosinophil and neutrophil counts in the sputum sample was not statistically di fferent between the doctor-diagnosed asthmatic children and healthy children groups. Previous studies reported that the percentage of eosinophils and neutrophils for asthmatic children was significantly di fferent and in the range of 2.5–13.0% and 15–47%, respectively [54–56]. Meanwhile, for healthy subjects, the percentages were in the range of 0.5–4.0% and 24.1–37.0%, respectively [57,58].

There have been few recent studies on activation and degranulation marker expression in the sputum samples of asthmatic children. The finding of this study confirmed that the sputum granulocytes of the asthmatic children increased the expression of the classical activation markers, CD11 and CD63, in both eosinophil and neutrophil cells, as reported by Tak et al. [28]. The upregulation of these tertiary and azurophilic/crystalloid granules is associated with the circulating cytokines that occur sequentially in response to the stimulus [59]. The mitogen-activated protein kinase (MAPK) pathway is believed to be central to the degranulation process [60]. The present study failed to show the upregulation of CD35 expression on eosinophils. In accordance with the study by Berends et al. [61], the downregulation of CD35 in the sputum of asthmatic children could be partly explained by the absence of intracellular stores for CD35 on eosinophils and neutrophils. Another possible explanation for this discrepancy was that CD35 is highly expressed on blood eosinophils or circulating granulocytes and was only directly associated with antigen inhalation [62] or the lower threshold stimulus required for cell activation [63].

The CD66b (CEACAM8) is a single-chain GPI-anchored glycoprotein and was recognised as an exclusive degranulation marker for neutrophils [64]. CD66b is upregulated when neutrophils are activated. The researchers of this study observed that the expression of CD66b was slightly upregulated on the surface of eosinophils and neutrophils collected from the airways of asthmatic children as compared to healthy children. It represents a normal activation pattern of neutrophils in relation to the migration from the circulating blood in vessels [65]. The late-phase response of the neutrophils could also possibly increase the CD11b, CD11b/18, CD35, CD64, and CD66b expressions [66]. The researchers found one study that identified that CD11b, CD16, and CD66b were consistently expressed on the neutrophils surface and were independent of their location and level of activation [67]. This could be contributed by the increased levels of intracellular cyclic GMP that yielded upregulation in the CD63 and CD66b expression on neutrophils [68].

PCA and AHC are very helpful approaches for dimensionality reduction in proteomics data. In this current study, these analytical approaches depicted similar group factors of biomarkers and air pollutants. The final regression analysis generated relatively moderate prediction models. The researchers noted that the upregulation of CD63 expression on both leukocytes and CD66b expression on neutrophils was related to particle exposure. The likely risks were observed among children under atopic and parental asthmatic/allergic conditions and children from schools located in urban areas. This finding reinforces the previous in vitro study conducted by Jin et al. [69]. They suggested that particulate allergens potentiated the mast cells to modulate the recruitment of eosinophils in the airways by internalising the particulate allergens into the CD63+ intracellular compartments through an endolytic pathway. Another in vitro study reported that CD66b only activated the neutrophils in the peptidoglycan challenge but did not upregulate the surface activation of eosinophils [70]. This is the possible explanation of why CD66b expression is clustered and associated with PM10 and PM2.5 in neutrophils but not with eosinophils in the AHC and regression analyses. Likewise, the study conducted by Banerje et al. [71] using flow cytometry analysis reported that there were increased CD35, CD16, and CD11b/CD18 expression on circulating neutrophils and a high percentage of eosinophils in the sputum of adults who have been exposed to PM10 and PM2.5.

To the researchers' knowledge, this study was the first study that explored the interrelation of CD11b, CD63, CD35, and CD66b marker expression on eosinophils and neutrophils with different parameters of air pollutants. This study revealed that CD11b was not clustered together with PM10 and PM2.5. This result was coherent with the study conducted by Ishii et al. [72] using immunocytochemistry, which showed that the expression of CD11b on alveolar macrophages was unaffected after two hours of stimulation with PM10. They suggested that the adhesive interaction between CD11/CD18 on alveolar macrophages with CD54 on the bronchial epithelial cells contributed to the amplification of cytokine production from the alveolar macrophages. The chemometrics analysis in this current study was also clustered and showed a significant relationship between the CD11b expression on eosinophils with NO2, especially among asthmatic children under atopic and parental asthmatic/allergic conditions. This finding was consistent with the review article by Hiraiwa and Eeden [73] and an in vitro study reported by Hodgkins et al. [13]. They found that dendritic cells expressed an upregulation of CD11b at 48 h during NO2-promoted allergic sensitisation. A study has shown that CD11b is directly involved in cellular adhesion, which is expressed in many leukocytes, including neutrophil, monocytes, natural killer cells, and macrophages. The migration of these leukocytes to the inflammation site will only take place if the CD18 subunit is present [74]. The other possible roles of CD11b were reported by Medoff et al. [75] in their experiment in which the CD11b+ had critical roles in mediating the Th2 cell and eosinophil recruitment in the airways via STAT6-dependent chemokine production.

This study also showed that the FeNO levels were positively correlated with the expression of the activation (CD66b, CD11b) and degranulation (CD66b, CD11b) markers for both leukocytes. This result is consistent with the results in the previous studies conducted by Guo et al. [76] and Kobayashi et al. [77], who also indicated that FeNO levels were reflected by eosinophilic airway inflammation [78]. In fact, the activated neutrophils can recruit the Th17/IL-17 and Th1 cells via chemokine release [79] and cause neutrophil infiltration within the airways [80]. This finding also reinforces that eosinophils and eosinophils are the binary indicators for the phenotyping of asthma. It was found in previous studies that exposure to PM10 was significantly associated with the increased levels of FeNO in healthy children tested on robust multi-pollutant models [41,81–83]. In line with this report, the chemometric analysis results in this study provided further evidence on the positive effects of PM10 on bronchial inflammation and resulted in the increase of FeNO levels among children.

The cluster approach used in this study, which is aimed at improving the interpretability of the data, interestingly revealed that the formaldehyde and NO2 concentrations were in the same factor, with a total variance of 26.0%. This finding confirmed that the formation of formaldehyde was through the photochemical reactivity of NO2 in the air with VOCs generating different aldehydes [84]. Formaldehyde also originates from furniture made out of wood and plastics, plywood, textile, table laminate, and consumer products, which is commonly available in the classroom [85]. As reported by Hua [86], NO2 potentially originates from the process of fossil-fuel combustion, biomass burning, and agricultural activities. Hereafter, all the schools in this current study are located very close to the main road and industrial area, which was considered the probable sources of NO2 in the classrooms. The researchers found that PM10 and PM2.5 were grouped together in the dimensionality analysis of the PCA and AHC, and this indicates that both particles originated from the same source. The primary sources of particles in the urban and suburban areas were industrial emissions, transportation, and traffic emissions [87,88]. It was possible that the indoor particles also originated from the occupant's activities or re-suspension of deposited particles, soil materials from the school children's shoes, skin flakes, furniture fragments, and less frequent cleaning [89]. Children are at risk of day-long exposure to the same indoor pollutants, not only at school, but also at home—which was reflected in the time spent during non-school days. This possibility merits further investigation.
