A Combined Proteomics and Metabolomics Profiling to Investigate the Genetic Heterogeneity of Autistic Children

Autism spectrum disorder (ASD) has become one of the most common neurological developmental disorders in children. However, the study of ASD diagnostic markers faces significant challenges due to the existence of heterogeneity. In this study, genetic testing was performed on children who were clinically diagnosed with ASD. Children with ASD susceptibility genes and healthy controls were studied. The proteomics of plasma and peripheral blood mononuclear cells (PBMCs) as well as plasma metabolomics were carried out. The results showed that although there was genetic heterogeneity in children with ASD, the differentially expressed proteins (DEPs) in plasma, peripheral blood mononuclear cells, and differential metabolites in plasma could still effectively distinguish autistic children from controls. The mechanism associated with them focuses on several common and previously reported mechanisms of ASD. The biomarkers for ASD diagnosis could be found by taking differentially expressed proteins and differential metabolites into consideration. Integrating omics data, glycerophospholipid metabolism and N-glycan biosynthesis might play a critical role in the pathogenesis of ASD.


Introduction
Autism spectrum disorder (ASD) is a group of developmental neurological disorders characterized by early onset of abnormal social communication and restricted repetitive behaviors and interests. In recent years, its incidence has gradually increased. ASD is about four times more common among boys than girls [1]. Although its pathogenesis has not been clarified, studies have shown that it may be a multifactorial disorder. Twin studies suggest genes play a key role in the pathogenesis of ASD [2]. It is estimated that ASD may involve thousands of genes [1]. However, ASD is highly heterogeneous, with each gene individually accounting for less than 1% of cases [3]. ASD may also be caused by the environmental factors. Therefore, it may be caused by interaction between genes and environmental factors [1].
Currently, there is no specific treatment for ASD. Studies show that early intervention can significantly improve ASD symptoms of diagnosed children and help them transition into society [4,5]. Early diagnosis is crucial for early detection and early intervention for autistic children. However, at present, there is a lack of effective biomarkers for clinical diagnosis [1,[6][7][8]. The diagnosis of ASD is based on the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-V) [9]. This standard brings a certain level of subjectivity into diagnosis, and is not entirely conducive to early detection of the disease.
Yuxi Zhao and Xueshan Cao contributed equally to this paper. The plasma contains abundant substances that are convenient for sampling. Thus, it becomes a suitable candidate when screening diseases. In the search for reliable biomarkers for ASD, some studies on blood protein-or metabolitesbased biomarkers have been carried out [1,[6][7][8]. Some other studies suggest that peripheral blood cells such as peripheral blood mononuclear cells (PBMCs) [10,11] and lymphoblastoid cell lines (LCLs) [12] may become useful tools to investigate systemic neurochemical changes in ASD. However, due to the existence of heterogeneity, finding peripheral blood or cell markers for the diagnosis of ASD can be challenging. Heterogeneity in etiology, phenotype, and outcome are hallmarks of ASD [13]. This may be due to the fact that ASD is a multifactorial disease. Heterogeneity complicates the quest for personalized medicine in ASD. Combining data from autistic children with different etiologies may lead to confusing results. Factors such as genetic variation, comorbidities, and gender contribute to the heterogeneity of ASD. Among them, genetic variation is the main contributor.
To explore the influence of gene heterogeneity on the search for diagnostic markers of ASD, we performed genetic testing in children clinically diagnosed with ASD in the present study. Some children were found to be carrying ASD susceptibility genes. The plasma and PBMCs of these children and their gender-and age-matched healthy controls were collected, plasma and PBMCs proteomics and plasma metabolomics analysis were carried out. Based on protein and metabolite profiles collected, the diagnostic markers and the pathogenesis of ASD were investigated.

Materials
The materials were purchased from the following companies: Dithiothreitol from Biosharp (Biosharp, Hefei, China); iodoacetamide from Sigma Aldrich (Sigma-Aldrich, St. Louis, USA). NH 4 HCO 3 , methanol and acetonitrile were acquired from Macklin (Macklin, Shanghai, China). The ultrafiltration filter (10 kDa cut-off) from Millipore (Millipore, Billerica, USA). Isobaric tags for relative and absolute quantification (iTRAQ) reagents were obtained from AB Sciex (AB Sciex, Foster City, USA). The glass injection flask from CNW Technologies (Anpel, Shanghai, China). The enzyme linked immunosorbent assay (ELISA) kit was purchased from Cusabio Biotech (Cusabio Biotech, WuHan, China). Pierce bicinchoninic acid (BCA) Protein Assay kit and formic acid from Thermo Fisher Scientific (Thermo Fisher Scientific, MA, USA). RIPA lysis buffer from Beyotime (Beyotime, Shanghai, China). All other reagents, except otherwise noted, were obtained from Sigma-Aldrich. Formic acid and trypsin were mass spectrometry grade. Acetonitrile and methanol were of chromatographic purity, and all of the other reagents were analytically pure.

A New Generation of Sequencing Technology for Targeted Sequencing of Autistic Candidate Genes
An overview of the workflow used in this study is shown in Figure 1. One hundred twenty-two autistic children (2-6 years old) were recruited from the Maternal and Child Health Hospital of Baoan in Shenzhen, China. The diagnosis of ASD was based on the criteria of autistic disorders as defined in the DSM-V by a child neuropsychiatrist.
EDTA (ethylenediaminetetraacetic acid) anticoagulation tubes were used to collect 4 mL of peripheral fasting blood from children in the morning. Genomic DNA was extracted from fresh blood samples using the phenol-chloroform extraction method. DNA samples were stored in anhydrous ethanol dissolved in TE buffer and diluted to 50ng/μL for subsequent experiments. The molecular inversion probe (MIP) was used to design probes for autism risk genes as previously described [14]. The target genes were captured, digested, and amplified by PCR (Polymerase chain reaction). After enrichment and purification of the product, highthroughput sequencing was carried out on a hiseq 3000 platform (Illumina, USA). The original sequencing data were then processed by quality control, comparison, and filtration. The mutation sites were annotated and analyzed for function.

Protein Sample Preparation
Subsequently, we performed plasma and PBMCs proteomics studies on the 5 children with de novo mutations (ASH1L, SCN2A, GIGYF2, NAA15, and DDX3X), along with gender-and age-matched healthy controls (Table S1). Four milliliters of fasting blood samples was collected in EDTAcoated plastic tubes. Plasma was separated by centrifugation at 800×g for 10 min at 4 °C, and the supernatant was collected and centrifuged at 10,000×g for 30 min at 4 °C, and then divided into aliquots and stored immediately afterwards at -80 °C. PBMCs were separated by density gradient centrifugation using Ficoll-Hypaque (Sigma-Aldrich, St. Louis, MO, USA) [11].
For proteomics analysis, the plasma samples were processed using Multiple Affinity Removal Columnas (Hu6; MARC; 4.6×50 mm, Agilent, Palo Alto, USA) previously described [15]. PBMCs were lysed in RIPA lysis buffer, sonicated 10 times for 5s with 10s pause interval in an ice-water bath, and then centrifuged at 10,000×g at 4 °C for 60 min [11]. The supernatant was taken away and stored at -80°C until use. The protein concentrations were quantified using a Pierce BCA Protein Assay kit [16].

iTRAQ Labeling and High-pH RPLC Fractionation
Protein digestion was conducted using the filter-aided sample preparation procedure [11,15,17]. For each sample, 100μg of proteins were placed on an ultrafiltration filter that consisted of 200μL of urea buffer (8 M urea, 150 mM Tris-HCl, pH 8.0), centrifuged at 14,000rcf for 30min, and then washed using 200μL of urea buffer. One hundred microliters of 10mM dithiothreitol was added to the sample and then kept in 30 °C for 2h. Approximately 100μL of 50mM iodoacetamide was then added to the filter in order to block any reduced cysteine residues. The samples were then kept at room temperature for 15 min in the dark, which was followed by centrifugation at a speed of 12,000rcf for 30min. The filters were then washed twice with urea buffer (200μL), followed by centrifugation at 12,000rcf for 20 min after every wash. Next, approximately 200μL of a NH 4 HCO 3 solution was placed on the filter, and then centrifuged at 12,000rcf for 20 min, and then repeated twice. Then, the protein suspensions were subjected to enzyme digestion using 2μL of trypsin buffer (2μg trypsin in 20μL of dissolution buffer) for 16-18 h at 37°C. Last, the filter unit was then transferred to a new tube and spun at 12,000rcf for 30min. The collected peptides were collected in the form of a filtrate and lyophilized. Then these samples were labeled with iTRAQ reagents. The information of iTRAQ tags for plasma and cells is shown in Table S2. iTRAQ analysis included set A and set B. After labeling, these were incubated at room temperature for 2h, and then mixed and lyophilized. The dried samples were reconstituted in 100μL double-distilled water (ddH 2 O) and injected into the Agilent HPLC (highperformance liquid chromatography; Agilent Technologies, USA) with a high pH RP (reverse phase) column (Durashell, C18, 250mm×4.6mm, 5μm; Bonna-Agela Technologies Inc., USA). Peptides were eluted and combined into 16 groups and lyophilized.

NanoLC-Mass Spectrometry (MS)/MS Analysis
An Ultra 2D Plus nanoflow HPLC (Eksigent Inc., Dublin, CA, USA) coupled with Triple TOF 6600 quadrupole time-of-flight mass spectrometer (AB Sciex) was used for analytical separation of peptides [11,15]. A ChromXP C18 (5μm, 0.3 × 10 mm, 120 Å , Agilent Technologies, Santa Clara, USA) trap cartridge was utilized for online trapping and desalting with 100% solvent A (water/acetonitrile/ formic acid (A, 98/2/0.1%, v/v/v; B, 2/98/0.1%, v/v/v),) at 4μL/min for 5 min, while microfluidic analytical columns packed with ChromXP C18 (3μm, 0.3 × 150 mm, 120 Å, Eksigent, Dublin, USA) were used in analytical separation using an elution gradient of 8-38% solvent B within 40 min at 4μL/min. Data was obtained by using 2.4kV ion injection voltage, 35PSI curtain gas, 12PSI sprayer gas, and 150°C interface heater. The reverse phase microLC eluent was subjected to positive ion microflow electrospray analysis in an information dependent acquisition mode (IDA). For IDA mode, the investigation scan was obtained within 250 ms. If the threshold of 260 CPS was exceeded and the charged ions was 2-4. The maximum of 40 product ion scans  Figure 1 (50ms) were collected. The rolling collision energy setting was applied to all precursor ions for collision induced dissociation. Dynamic exclusion was set to 16 s.

Database Search, iTRAQ Quantification, and Bioinformatics Analysis
Protein identification and quantification were performed using ProteinPilot v5.0 (AB Sciex). The quantitative data were loaded to OMICSBEAN website [18]. Data were normalized and t-test was performed. The cutoff value for upregulation was 1.2-fold change, for down-regulation was 0.83-fold change, and FDR (false discovery rate) p-value < 0.05 was established for significantly differentially expressed proteins (DEPs) between autistic children and controls [19]. Principal component analysis (PCA) was performed by using MetaboAnalyst 4.0 [19,20]. Partial least squares discriminant analysis (PLS-DA) were performed using SIMCA-P 14.1 software package (V14.1, Sartorius Stedim Data Analytics AB, Umea, Sweden). The model was validated using a 7-fold cross validation method and tested with 200 random permutations. The DEPs were loaded to String [21] database for biological process (BP), cellular component (CC), KEGG and REACTOME pathways analysis. Protein-protein interaction (PPI) networks were analyzed using Cytoscape 3.8.2 and STRING.

ELISA Validation
We chose the complement protein, C1S, which was found to altered in ASD plasma for the first time in this study and could be purchased with a commercial ELISA kit for verification. Twenty-nine (29) aqueous crude plasma samples from children with ASD and age-and sex-matched healthy subjects were detected according to the manufacturer's instructions. Receiver operating characteristic (ROC) curve was constructed to assess the diagnostic value by using SPSS package (SPSS 24.0, Chicago, IL, USA). The statistical analysis was performed and scatter plot was drawn by using Graphpad software (GraphPad Software, Inc., San Diego, CA, USA). Two-tailed t-test was used for statistical analysis, and p-value < 0.05 were considered statistically significant.

Plasma Pretreatment and Mass Spectrometry Analysis
Internal standard (L-2-chlorophenylalanine, 0.3mg/mL, methanol) was pre-mixed with enough methanol-acetonitrile (2:1, v/v) as the metabolite extract, which was pre-cooled at -20°C. For each sample, 100μL plasma was mixed with 300μL methanol-acetonitrile (2:1, v/v), sonicated in an ice water bath for 10 min, stood at -20°C for 30 min, and then centrifugated at 4 °C at 13,000×g for 15 min. Additionally, 100μL supernatant was filled into a glass injection flask, and mixed with 10μL supernatant from each sample as quality control (QC) sample.
A Waters Acquity UPLC system coupled with a Q-TOF Synapt G2 high-definition mass spectrometer (Waters, USA) was used for metabolomics analysis. Chromatographic separation was performed on an ACQUITY UPLCBEH C18 column (2.1×100 mm, 1.7μm) at 45°C. Mobile phase A was an aqueous solution containing 0.1% formic acid, and mobile phase B was an acetonitrile solution containing 0.1% formic acid. The flow rate of mobile phase A and B was 0.4mL/min. The gradient elution is shown in Table S3. The injection volume was 2μL. The MS analysis was carried out by the ESI ion source, and the positive and negative ion scanning modes were used to collect the signal respectively. The parameters of MS analysis are shown in Table S4.

Processing of the Results of Metabolomics Mass Spectrometry
The raw spectrum was processed by Progenesis QI 2.0 software (Nonlinear Dynamics, Newcastle, UK) for peak selection, comparison, standardization, and identification. R package metaX was used to perform further statistical analysis on the normalized peak intensity. We also used public spectral library MassBank, HMDB, LipidBlast, and METLIN for secondary identification. After characterization, duplicate qualitative results were eliminated based on the score.

Metabolomics Data Analysis
The profile of metabolomics data was analyzed using multivariate statistical methods. PCA and PLS-DA analysis were carried out. VIP (variable importance in projection) score of PLS-DA was obtained. The metabolites with VIP scores ≥ 1, fold change ≥ 1.2 or ≤ 0.83 (ASD versus control) as the threshold, and p < 0.05 were identified as differential metabolites. Pathway analysis were performed using Meta-boAnalyst 4.0.

Integrated Analysis of DEPs and Differential Metabolites Identified in This Study
The DEPs identified by proteomics analysis between the cases and controls, including the DEPs of plasma,PBMCs, and the differential metabolites between the two groups identified by plasma metabolomics were loaded to MetaScape v3.1.3 in Cytoscape 3.4.0 for integrated analysis [22].

The Results of Gene Testing
By genetic testing, among the 122 autistic patients, 5 children were found to be associated with de novo mutations in autism risk genes, including histone-lysine N-methyltransferase ASH1L (ASH1L), ATP-dependent RNA helicase DDX3X (DDX3X), N-alpha-acetyltransferase 15, NatA auxiliary subunit (NAA15), GRB10-interacting GYF protein 2 (GIGYF2), and sodium channel protein type 2 subunit alpha (SCN2A), respectively. Another child was found to be associated with the mutations of the MECP2 (methyl-CpGbinding protein 2) gene. Since MECP2 is associated with Rett syndrome, which has been removed from ASD based on DSM-V diagnostic criteria, so we have only further studied the other five children.

The Results of Plasma iTRAQ Quantitative Proteomics Analysis
By proteomics analysis, 300 and 350 plasma proteins were identified in the set A experiment and set B experiment, respectively. Among these, 280 proteins were identified as common. The cluster and PCA analysis showed that the expression pattern of plasma proteins was heterogeneous. The characteristics of total plasma protein failed to distinguish ASD individuals from healthy controls (Figure 2A and B). Out of the total proteins, 13 proteins were identified as DEPs between children with ASD and healthy controls ( Figure 2C, Table S5). Among them, 5 proteins were upregulated and 8 proteins were downregulated. The cluster analysis showed that these DEPs could distinguish the ASD group from the control group. However, the five controls were not completely clustered together, and the expression characteristics of the controls with similar age were more similar ( Figure 2D).

The Results of PBMCs Quantitative Proteomics Analysis
By proteomics analysis, 2800 proteins were identified as common in the set A experiment and set B experiment. Cluster and PCA analysis showed that the expression characteristics of total proteins failed to distinguish the ASD group from the control group ( Figure 3A and B). By comparative analysis, 111 proteins were identified as DEPs between children with ASD and controls ( Figure 3C, Table S6). Compared with the controls, 88 proteins were up-regulated and 22 proteins were downregulated in the ASD individuals. Cluster analysis showed that the ASD group and control group could be well distinguished by these DEPs (Figure 3D). Of note, among the DEPs identified in PBMCs, the genes encoding 4 DEPs (SLC9A9, CSDE1, CCT4, and DDX3X) are included in the SFARI (Simons Foundation Autism Research Initiative) gene database (https:// gene. sfari. org/). Interestingly, the DDX3X gene has been detected in gene testing as described above.

Bioinformatics Analysis of DEPs in Plasma and PBMCs
For the DEPs of plasma, the BP and KEGG pathway are shown in the Table S7 and S8. These DEPs were involved in complement activation, immune response, and complement and coagulation cascades ( Figure 4A and B). Interestingly, some complement proteins were related to synapse punting (C1QA and C1QB) and neuronal development and plasticity (C1QA, C1QB, C1S, C2, and COLEC10).
For the DEPs of PBMCs, the BP, KEGG and Reactome pathways are shown in Table S9-S11. These DEPs mainly belong to immunity and metabolism related protein. KEGG pathway showed that they were associated with proteasome, ubiquitin mediated proteolysis, protein processing in endoplasmic reticulum, necroptosis, and nervous system diseases include Parkinson's disease (PD), Alzheimer's disease (AD), amyotrophic lateral sclerosis, and Huntington disease (Figure 4C-G). Based on STRING database, we also found that 12 DEPs belonging to mitochondria-related proteins,

The Results of ELISA Verification
As shown in Figure 5, consistent with the iTRAQ analysis, the expression of C1S protein in plasma of children with ASD was significantly higher than that in the control group (p<0.05). The area under the ROC curve (AUC) value was 0.8157 (95% confidence interval, 0.6989-0.9325, p <0.0001). The sensitivity and specificity were 83% and 76%, respectively.

The Results of Plasma Metabolomics Analysis
Subsequently, we conducted plasma metabolomics studies with the 5 children with ASD and their corresponding controls. In the positive ion mode of MS analysis, 9281 mass spectrum peaks were identified. The PCA analysis established that the cases and controls could not well be distinguished, although except for a gene, other cases and controls have aggregation trend ( Figure 6A). PLS-DA analysis revealed that the cases and controls could be distinguished ( Figure 6B). However, this model was a bit overfitting ( Figure 6C). It may be due to individual heterogeneity or small sample size. Likewise, 286 metabolites were identified as differential metabolites between the cases and controls ( Figure 6D, Table S12). Compared with the controls, 157 of them were increased and 129 were decreased. They could well distinguish the case group from the controls (Figure 6E), and were mainly related to niacin and niacinamide metabolism, vitamin B6 metabolism, arginine biosynthesis, and sphingolipid metabolism ( Figure 6F). The pathways associated with the differential metabolites are shown in Table S13.
In the negative ion mode, 6989 mass spectrum peaks were identified. PCA analysis exhibited that the case group was not well distinguished from the controls. The case group showed a tendency to be more concentrated, while the control group was more dispersed among different individuals ( Figure 7A). PLS-DA analysis separated the two groups effectively, however, this model was also a bit overfitting ( Figure 7B and C). Meanwhile, 190 metabolites were identified as differential metabolites between the cases and the controls. Compared with the controls, 98 were increased and 92 were decreased for controls ( Figure 7D, Table S14). They could effectively discern the cases from the control group ( Figure 7E), and were mainly involved in alanine, aspartic acid and glutamate metabolism, D-glutamine and D-glutamate metabolism, arginine biosynthesis, glycerophospholipid metabolism, and nitrogen metabolism ( Figure 7F, Table S15).

The Integration Analysis of DEPs and Differential Metabolites
By integration analysis, two pathways associated with DEPs and differential metabolites identified in this study were enriched, including glycerophospholipid metabolism (Figure 8A) and N-Glycan biosynthesis ( Figure 8B).

Discussion
In this study, we identified five de novo mutational genes associated with five ASD children among 122 clinically diagnosed autistic children. This is similar to results from previous studies of coding-sequence mutations making up 5-10% of ASD patients [3]. These genes include ASH1L, SCN2A, GIGYF2, NAA15, and DDX3X. They has been reported to be associated with ASD in the previous studies [23][24][25][26][27][28][29]. We then carried out plasma, PBMCs proteomics, and plasma metabolomics studies in these ASD children and healthy controls. The results showed that total plasma and PBMCs proteins, total plasma metabolites, were not well distinguished between the two groups, but DEPs and differential metabolites between these two groups were able to distinguish the two groups. Here, we focus on these DEPs, differential metabolites, and their related mechanisms.
By plasma proteomics analysis, 13 DEPs were identified. Consistent with the previous studies, these DEPs identified in the blood were mainly involved in complement and coagulation cascades, inflammatory and immune responses. Altered complement proteins in the blood from patients with ASD have been broadly reported, including changes in protein expression levels [15,[30][31][32][33] and post-translational modifications [34,35]. Complement C3 was also found to be up-regulated in the PBMCs of children with ASD in our previous study [11]. Here, C1QA, C1QB, C1S, and C2 were significantly upregulated in the plasma of children with ASD, whereas COLEC10 was downregulated. To the best of our knowledge, C1S, C2, and COLEC10 were first Blue and red dots: hits with p < 0.05 and |log 2 FC|> 0.26. (E) Cluster analysis the differential metabolites between children with ASD and healthy controls. (F) Metabolic pathways related to differential metabolites between children with ASD and healthy controls reported to be associated with ASD. The result of C1S protein ELISA verification was consistent with the proteomics result with a higher AUC value, sensitivity and specificity, showing its potential as a diagnostic marker. The expression of most complement proteins is upregulated in the periphery of autistic patients, suggesting that the complement pathway may be activated in the periphery of autistic patients. However, the expression trends of complement proteins in the brain of autistic patient are not completely consistent [36,37]. Beyond its involvement with innate immune responses, complement proteins has been increasingly implicated in playing an important role in neurodevelopment, including neurogenesis, neuronal migration, and synaptic remodeling [36][37][38]. It is unclear whether changes in the levels of complement molecules in the central nervous system correlate with peripheral changes [39]. Nevertheless, decreased levels of complement proteins in the ASD brain may lead to reduced complement-mediated synaptic pruning, contributing to the cortical hyperconnectivity and behavioral phenotypes of ASD [36,39,40]. A recent study revealed that complement C4 levels were lower in induced pluripotent stem cell (iPSC)-derived astrocytes from ASD patients (D) Volcano plot analysis of differential metabolites between children with ASD and healthy controls. The log 2 fold change (FC) is plotted versus the -log 10 of the p value. Blue and red dots: hits with p < 0.05 and |log 2 FC|> 0.26. (E) Cluster analysis the differential metabolites between children with ASD and healthy controls. (F) Metabolic pathways related to differential metabolites between children with ASD and healthy controls [41]. Besides, it is known that immune dysregulation plays a role in the neurodegenerative/psychiatric disorders [42,43]. The role of complement system in other neuropsychiatric diseases has recently received a lot of attention, such as schizophrenia and AD [42,44,45]. In the brain of schizophrenic patients, the high expression of complement protein C4A may cause faulty synapse elimination, thereby resulting in the decrease of synaptic density [46,47]. The results of changes in complement protein concentration in peripheral blood of patients with AD were inconsistent [48]. Pathological complement-mediated synapse loss has been reported to be associated with AD [46,49,50]. Overall, the changes in Fig. 8 Integrated analysis of proteomics and metabolomics results. (A) Glycerophospholipid metabolism was associated with the DEPs of proteomics and the differential metabolites of metabolomics. (B) N-Glycan biosynthesis was associated with the DEPs of proteomics and the differential metabolites of metabolomics the complement system in peripheral blood, PBMCs, and brain of ASD patients highlights that this system may play a key role in the pathogenesis of ASD and is worthy of further study.
Accumulating evidence suggests a potential role of the immune system in the pathophysiology of ASD during the pre, neo-, and postnatal periods [13,51]. In this study, 7 plasma DEPs and 36 PBMCs DEPs were involved in the immune system, further supporting the association between immunity and ASD pathogenesis. Ten DEPs of PBMCs were associated with interleukin-1 (IL-1), together with our previous study (IL-12) [11] and other studies [52,53], suggesting that IL-1 and IL-12 might be key inflammatory cytokines in peripheral blood and blood mononuclear cells.
Mitochondrial dysfunction has been implicated in immune dysregulation in ASD [54]. Mitochondrial dysfunction also affects oxidative stress in ASD. Oxidative stress plays a role in immune regulation, and some of the metabolites found may act upstream or downstream of specific pathways. The role of mitochondrial dysfunction, oxidative stress, immune dysregulation/inflammation in ASD as well as their relationship, have recently received extensive attention [55][56][57]. In this study, the expression of 11 mitochondria-related proteins was up-regulated in PBMCs of autistic children, including 3 proteins involved in energy metabolism (i.e., NDUFA7, NDUFA11, and NDUFA13). The results were similar to previous observations in PBMCs [11] and LCL cells [58], different tissues and organs [59], T cell, NK cell, and monocyte [60] in children with autism, and in a recent study in the cerebral cortex of valproic acidinduced rat model of autism [61]. Activation and proliferation of microglia and astrocytes have also been observed in the brains of ASD subjects [62]. Interestingly, transcriptome analysis between autistic brains and normal brains identified discrete modules by gene co-expression network analysis: a neuronal module and a module enriched for immune genes and glial markers [63]. This may explain the uniqueness of mitochondrial dysfunction in ASD and how autistic children may have nontraditional mitochondrial diseases. After all, mitochondrial dysfunction has been reported by postmortem brain tissue examinations on ASD subjects and reduction of Electron transport chain (ETC) complexes has also been observed in different brain regions of autistic children [64][65][66].
Consistent with our previous study [11], the main pathways associated with DEPs of PBMCs have also involved in proteasome, ubiquitin mediated proteolysis, and protein processing in the endoplasmic reticulum, which strongly suggest that endoplasmic reticulum stress (ER) being linked to the PBMCs of children with ASD. ER stress occurs when the amount of unfolded proteins in the ER reach an unmanageable level, triggering the unfolded protein response (UPR) [67]. Under excess or chronic ER stress, cell apoptosis is induced to eliminate unhealthy cells [68]. Here, 6 DEPs were associated with necroptosis. Similarly, LCLs from children with autism were shown to be more sensitive to necrosis than their non-autistic siblings [69]. Thus, necroptosis may be associated with the ER stress in the PBMCs of ASD subjects. Moreover, increase in ER stress has been observed in the brains of autistic children [70,71] and those of ASD model mice [72]. The mutation of ASD related genes and the aggregation of their encoded proteins [11,70], along with oxidative stress may be responsible for ER stress in children with ASD [70]. On the other hand, the proteostasis network machinery plays a role in the establishment, maintenance, and plasticity of stable and dynamic dendritic arbors. Ubiquitin-proteasome system is required for developmental dendritic pruning [73,74].
Interestingly, among the DEPs of PBMCs, three other genes have also been linked to ASD in addition to DDX3X, including SLC9A9 [75], CSDE1 [76], and CCT4 [29]. Among them, the SLC9A9 gene encodes Na + /H + transpteron-9 endometrium protein (NHE9), which has been shown to be related to endocytosis, protein ubiquitination, and phagosome [75]. CSDE1 encodes RNA-binding proteins that may be involved in translation-coupled mRNA conversion and is associated with neurodevelopment and neuropsychiatric disorders [76]. The CCT4 gene encodes a molecular chaperone that assists in protein folding during ATP hydrolysis [77]. These results further support that protein folding and ER stress may be associated with the pathogenesis of ASD, and that at least some children with ASD carry more than two risk genes in this study.
By metabolomics analysis, the differential metabolites in the plasma between cases and controls were mainly involved in amino acid (alanine, aspartate and glutamate, arginine biosynthesis, and D-glutamine and D-glutamate), vitamin (nicotinate and nicotinamide, and vitamin B6), and lipid (glycerophospholipid and sphingolipid) metabolism. Nitrogen metabolism, N-Glycan biosynthesis, and neomycin, kanamycin, and gentamicin biosynthesis were also involved.
In the present study, among the altered amino acids, L-Glutamate was down-regulated in plasma of controls while L-Glutamine was increased in plasma of autistic children. Although previous studies showed that the altered levels of glutamate and glutamine in the blood of autistic patients were inconsistent [78][79][80][81][82][83], these data support the current view that excitatory/inhibitory imbalance, especially the abnormality of the excitatory neurotransmitter glutamate, is one of the pathogenesis of ASD [82,84]. Indeed, it has been reported that glutamate signals in the anterior cingulate cortex and cerebellum of ASD patients were significantly decreased [85]. More recently, the gut metabolites involved in alanine, aspartate, and glutamate metabolic pathways were reported to be significantly lower in children with ASD, which was associated with differences in the abundance of gut microbiota related to D-Glutamine and D-glutamate metabolism, suggesting that the gut microbiota might contribute to abnormal glutamate metabolism in autistic children [86]. Besides, similar with the present study, aspartate has been observed to be decreased in the fecal ASD subjects [87], while ornithine was observed to be increased in the blood of children with ASD [81].
Our results showed that three differential metabolites were also involved in the nicotinate and nicotinamide metabolism pathway. Niacin (NA), also known as Vitamin B3 and nicotinic acid, can be biosynthetically converted into nicotinamide adenine dinucleotide (NAD). NAD has a variety of biological functions and plays a central role in redox reactions [88]. Nicotinate and nicotinamide metabolism have been reported to be altered in the prefrontal cortex [89], urine [90], and blood [91] of ASD individuals, and associated with microbiota transfer therapy (MTT) of autistic children [91]. Nicotinamide is derived from tryptophan, while abnormal tryptophan metabolism has been observed in children with ASD [90,[92][93][94]. In addition, three differential metabolites (O-phospho-4-hydroxy-Lthreonine, pyridoxamine, and 4-pyridoxate) involved in vitamin B6 metabolism were found to be decreased in autistic children. Reduced levels of Vitamin B6 have been observed in the urine of autistic children [95]. It is the main cofactor of biological reactions and is important for the synthesis of neurotransmitters and trans-sulfuration. Its deficiency is related to oxidative stress, high blood homocysteine and hypomethylation in children with AD [95]. In general, the lack of vitamin B group in children with ASD may be caused by nutritional deficiency, poor absorption or alteration in gut microbiota [91,96,97], contributing to the pathogenesis of ASD. The brain is particularly enriched in lipids with a diverse lipid composition compared to other tissues [98]. Glycerophospholipids are critical components of neuronal membranes and myelin, and principal regulators of synaptic function [98]. Sphingolipid is involved in neuronal differentiation, synaptic transmission in neuronal-glial connections, and myelin stability. Disturbance of their metabolism has been linked to various neuropsychiatric diseases include autism [99], Rett syndrome [100], and ASD [101].
Finally, integrating omics data, glycerophospholipid, and N-linked glycosylation metabolisms were associated with the DEPs and differential metabolites. N-linked glycosylation is important in brain structure and function. The extracellular glycans and glycoconjugates may contribute to the etiology and pathogenesis of pervasive neurodevelopmental disorders include idiopathic ASDs. Glycobiology related genes were implicated in ASD. Mutations in glycogenes associated with ASD affect the downstream steps of N-glycan biosynthesis [102].

Conclusions
In this study, our results showed that although there was heterogeneity at the genetic level in children with ASD, the DEPs of plasma and PBMCs and differential metabolites of plasma could still distinguish the cases from controls. The proteomic results highlighted the roles of complement, inflammation and immunity, mitochondrial dysfunction, proteasome, ubiquitin mediated proteolysis, and ER stress in the pathogenesis of ASD. Three complement proteins (i.e., C1S, C2, and COLEC10) were first reported to be altered in the plasma of children with ASD, among them, the expression of the C1S was confirmed by ELISA analysis. Metabolomic results mainly showed the disturbances of amino acid, vitamin, and lipid metabolism in children with ASD ( Figure 9). The mechanisms and pathways associated with the DEPs and differential metabolites have been reported in previous studies. The results agreed with the view that children with ASD might have an important underlying common mechanism. They are not only potential therapeutic targets for ASD but also significant contributors for studying biomarkers for the disorder [1,[6][7][8]13]. This study presents a paradigm for using high-throughput omics to investigate the omics profiling of plasma/PBMC to explore ASD genetic heterogeneity. It needs to be pointed out that the number of samples carrying risk genes used in omics studies is limited, and further research with a large sample size is required. It is also interesting to add a group in which that the children were diagnosed as ASD, but not detected with risk genes in further research. Moreover, further verification of DEPs or differential metabolites is also important and interesting.

Data Availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethics Approval and Consent to Participate The Maternal and Child Health Hospital of Baoan (No.20170801), and the Shenzhen University approved study procedures (No.M20220203). The experiments were performed after obtaining written consent from caretakers of the children under observation according to the guidelines of this hospital.

Competing Interests
The authors declare no competing interests.