Analysis of potential biomarkers and immune infiltration in autism based on bioinformatics analysis

Abstract

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder caused by both environmental and genetic factors. However, its etiology and pathogenesis remain unclear. The purpose of this study was to establish an immune-related diagnostic model for ASD using bioinformatics methods and to identify ASD biomarkers. Two ASD datasets, GSE18123 and GSE29691, were integrated into the gene expression Database to eliminate batch effects. 41 differentially expressed genes were identified by microarray data linear model (limma package). Based on the results of the immune infiltration analysis, we speculated that neutrophils, B cells naive, CD8+ T cells, and Tregs are potential core immune cells in ASD and participate in the occurrence of ASD. Finally, the differential genes and immune infiltration in ASD and non-ASD patients were compared, and the most relevant genes were selected to construct the first immune correlation prediction model of ASD. After the calculation, the model exhibited better accuracy. The calculations show that the model has good accuracy.

1. Introduction

Autism spectrum disorder (ASD) is a complex and heterogeneous neurodevelopmental disorder caused by both environmental and genetic factors. It is a comprehensive problem in modern psychology, neurology, and related sciences, and is being widely discussed. Its clinical manifestations are heterogeneous and include attention deficit hyperactivity disorder, growth retardation, motor abnormalities, mental retardation, and epilepsy. Recent studies have shown that the overall prevalence rate of ASD at home and abroad shows a gradual upward trend^[1,2] and mainly in males (about 4.2:1).^[3] With the continuous increase in the number of ASD, the demand for lifelong support in education, healthcare, life, family, and other fields has also increased, making the disease a major international public health problem.^[4] Finding an effective treatment for autism is of immediate importance, and its potential molecular and cellular mechanisms are one of the hotspots in the field of children’s mental psychology.

Although extensive studies have been conducted, the etiology and pathogenesis of ASD remain unclear. At present, there are many theories in the research: genetic hypothesis, non-neurogenic hypothesis, neurochemical hypothesis, mitochondrial dysfunction and oxidative stress hypothesis, immune hypothesis, opioid hypothesis, local translation disorder hypothesis, mTOR signaling pathway hypothesis and so on.^[5,6] The immune hypothesis is considered the main factor leading to the pathogenesis of autism, and it also explains the differences in clinical phenotypes and complications that affect the progression and severity of the disease.^[7] Research shows that the cytokine level IL-6, IL-8, IL-1 β Changes in cytokine levels can affect the stereotyped behavior of ASD,^[8] and the reduction of TGF-b1 level is related to the reduction of adaptive behavior and the deterioration of behavioral symptoms.^[9] While mononuclear cells, myeloid dendritic cells and the size of amygdala are related to and the abnormal behavior of ASD,^[10,11] even more studies believe that the severity of ASD can be determined according to the proportion of “Th17” or “Treg” cells.^[12,13] Although many studies have recognized that the change of immune response is the main etiological component of ASD, and a large number of studies have been done on it, the specific functional changes of immune response are not very clear.^[14]

Intervention for ASD after early recognition and diagnosis at 12 to 18 months is beneficial for improve patient prognosis of patients.^[1] However, due to the heterogeneity of the clinical manifestations of autism, ASD patients can only be identified in children aged 2–3 years.^[15] Therefore, it is of great significance to find and improve early screening methods, help identify ASD populations in the early stages, accelerate the clinical intervention process, and improve the prognosis of ASD patients.

Therefore, this study screened the risk genes related to ASD using bioinformatics methods and built them into a risk prediction model for ASD.^[16,17] A nomogram can be used as a diagnostic and prognostic prediction model that is widely used in the tumor and non-tumor fields.^[18] This may become a promising method to identify the risk of autism, assist in the early clinical diagnosis of ASD, promote early intervention and treatment of ASD, and improve clinical prognosis. At the same time, we discuss the mechanism of ASD through the formula of ASD immune infiltration analysis and further promote the study of potential ASD biomarkers.^[19]

2. Materials and methods

2.1. Dataset selection

From the NCBI gene expression omnibus database (GEO; https://www.ncbi.nlm.nih.gov/geo/)^[20] searches for ASD-related sequencing results were performed and ASD sequencing data sets (GSE18123, GSE29691) that met the requirements were downloaded. GEO databases belong to public databases. The patients involved in the database have obtained ethical approval. Users can download relevant data for free for research and publish relevant articles. Our study is based on open source data, so there are no ethical issues and other conflicts of interest. GSE18123 included 115 non-autistic and 170 autistic samples. GSE29691 included 13 non-autistic and 2 autistic samples. GSE18123 uses GPL6244, [HuGene-1_0-st] Affimatrix Human Gene 1.0 ST Array [transcript (gene) version] and GPL570, [HG-U133_Plus_2] Affimatrix Human Genome U133 Plus 2.0 Array for platform detection. GSE29691 uses GPL570, [HG-U133_Plus_2] Affimatrix Human Genome U133 Plus 2.0, and Array for platform detection; both datasets are gene expression arrays.

2.2. Data processing

The 2 datasets were processed by background calibration, normalization, log2 transformation, etc. When a gene pairs multiple probes, the average value is taken as its expression value, and the SVA package is used to remove the batch effect between datasets.

2.3. Screening differentially expressed genes

The limma package was used to screen the differentially expressed genes with log2 fold change (FC) |>0.5 and P value < .05.

2.4. Immune cell infiltration

We used cell type identification by estimating relative subsets of transcripts to analyze and calculate the infiltration of immune cells in ASD,^[21] calculate the percentage of immune cells, and show the results in a bar chart. We drew an immune cell heat map and a related map.^[22,23]

2.5. Model for predicting the risk of ASD

By comparing the differential genes and performing immune infiltration analysis between ASD and non-ASD patients, the most relevant genes were selected to construct the risk prediction model, and a calibration curve was constructed to determine the calibration of the model. A receiver operating characteristic curve was drawn, the area under the curve was calculated, and the accuracy of the prediction model was evaluated.

3. Results

3.1. Contrasting differential genes

Comparing the differential genes of the GSE18123 and GSE29691 datasets, 41 differential genes were found, including 20 upregulated and 21 downregulated genes. The 5 upregulated genes were EIF1AY, DDX3Y, CYP4F11, RPS4Y1, and USP9Y. The first 5 downregulated genes were DMRTB1, ACY1, COL24A1, ZNRF2P1, and EPYC (Fig. 1).

Figure 1.

The heat map and volcano map of the different genes. (A) each row of the heat map represents a DEG, and each column represents a sample, either normal or ASD. Red and blue represent up-regulated and down-regulated DEG, respectively. (B) Red dots indicate up-regulated DEG and blue dots indicate down-regulated DEG. ASD = autism spectrum disorder, DEGs = differentially expressed genes.

3.2. Analysis of immune cell infiltration

We used cell type identification by estimating relative subsets of transcripts to predict immune cell infiltration in patients with ASD and controls. The percentage of the 22 immune cells in each sample is shown in the histogram and heatmap (Fig. 2). As shown in the figure, the proportion of immune cells varied significantly among the different samples and groups. The histogram of the difference in immune cell infiltration showed that, in patients with ASD, the proportion of neutrophils and resting mast cells increased, the degree of infiltration was higher, B cells naive, CD8+ T cells, and Treg cells decreased, and the degree of infiltration was lower (Fig. 3A). In addition, we analyzed the correlation between the infiltrating immune cells in ASD and found synergistic or competitive effects among different immune cells. Among them, T cells regulatory (Tregs) and T cells CD8 have strong synergistic effect (r = 0.57), gamma delta T cells and T cells follicular helper have strong synergistic effect (r = 0.58), neutrophils and T cells CD8 have strong competitive effect (r = −0.55) (Fig. 3B).

Figure 2.

The immune infiltration between ASD and normal controls. (A) the relative percentage of 22 kinds of immune cells. (B) 22 kinds of immunized heat map cells. ASD = autism spectrum disorder.

Figure 3.

Distribution and visualization of immune cell infiltration. (A) 22 kinds of immune cell subtypes were compared between ASD group and control group. Blue and red represent normal and ASD samples, respectively. (B) the composition of all 22 correlated matrix immune cell subtypes. Both horizontal and vertical axes demonstrate immune cell subtypes. Composition of immune cell subtypes (higher, lower and the same correlation levels are shown in red, blue and white, respectively). ASD = autism spectrum disorder. (C) The correlation between significantly differentially expressed genes (up regulated top five and down regulated top five genes) and immune function.

3.3. Building a forecasting model

Based on differential gene expression in ASD, we selected the most relevant genes were selected to construct the risk prediction model (Fig. 4A). The calculated C-index value of the model C index is 0.767. We calculated the model C index to be 0.767, and drew a model calibration chart (Fig. 4B) and receiver opeating characteristic curve (Fig. 4C). Based on these results, we believe that the proposed model has good discrimination and prediction capabilities.

Figure 4.

(A) Nomograms of ASD. (B) Calibration diagram of ASD prediction model. (C) ROC curve. The area under the curve is 0.843, and the model has good discrimination and prediction ability. ASD = autism spectrum disorder, ROC = receiver operator characteristic.

4. Discussion

ASD is a complex and heterogeneous neurodevelopmental disorder whose pathogenesis remains unclear. In the past, research on gene expression was limited to a single method or small samples. In this study, we synthesized 2 ASD sequencing datasets and used limma and weighted gene correlation network analysis methods to identify genes. First, 41 differentially expressed genes, including 20 upregulated and 21 downregulated genes, were identified using the limma package to compare gene expression between the ASD and non-ASD groups. The first 3 upregulated genes were EIF1AY, DDX3Y, and CYP4F11, and the first 3 downregulated genes were DMRTB1, ACY1, and COL24A1. The first 3 upregulated genes were Y chromosome-related genes. Therefore, we believe that an imbalance in sexual dimorphism on the human Y chromosome may be the cause of autism and other nervous system diseases.^[24,25]

Based on differential gene expression in patients with ASD, we selected the most significant genes to construct an immune-related risk prediction model for ASD. The results indicate that the model is accurate.

In addition, an increasing number of studies are focusing on the role of immunity in the pathogenesis of ASD.^[12,26–28] However, analyses of immune infiltration in ASD are scarce. Therefore, we analyzed immune cell infiltration in ASD and non-ASD patients. The number of Tregs in the ASD group was lower than that in the non-ASD group. Tregs are a subgroup of CD4+ helper T cells (CD4+ Th). CD4+ Th cells are divided into Th1, Th2, Th3, and other effector and regulatory cell subsets. Th1 cells participate in cell-mediated immune responses to clear intracellular pathogens.^[29,30] Th2 cells participate in humoral immune response, clear extracellular organisms and produce immunoglobulins.^[31] Th17 cells play an important role in inducing autoimmune diseases and preventing inflammatory reactions such as bacterial and fungal infections.^[32] Tregs are a regulatory subset of CD4+ T helper (Th) cells. It is characterized by the transcription factor Forkhead box P3 (FoxP3) and is a key component of immune regulation. It plays a key role in maintaining immune homeostasis and preventing autoimmune diseases.^[33,34] Activated Tregs (CD4 + CD25 + Foxp3+) largely reverse the proinflammatory immune spectrum and behavioral abnormalities induced by maternal immune activation.^[35] Defects in Treg cells are related to behavioral regulation and core characteristics of autism.^[13] Treg cells can inhibit nitrification by microglia α-Synuclein and other stimulating reactions to protect nerves.^[36]

In addition, we found that the proportions of neutrophils and resting mast cells in patients with ASD. On the one hand, neutrophils can produce a pro-inflammatory cytokine IL-17^[37,38], which is related to neurodevelopmental disorders. However, an increase in neutrophil oxidative stress leads to a disorder of the enzyme antioxidant network in peripheral innate immune cells. Oxidative and pro-inflammatory cytokines produced by neutrophils are involved in the pathogenesis of ASD.^[39]

Mast cells are located around blood vessels close to neurons and microglia, mainly in the soft brain, thalamus, and hypothalamus, and especially in the median eminence.^[40] Mast cells can selectively release mediators such as IL-6,^[41] change the permeability of the blood-brain barrier,^[42] and allow neurotoxic molecules to enter the brain, causing local inflammation and promoting the onset of ASD.^[40,43–45] The combination of B cells and other immune cells may cause IL-6 and TNF in the systemic circulation of ASD subjects-α With the increase of B cells^[46] and other immune cells have surface receptors of cytokine/chemokine signals, which can be activated by inflammatory cytokines produced by T cells/monocytes of ASD subjects, thus changing the immune response of B cells, which can promote the inflammation of ASD.^[47–50] Some studies have shown that BTBR mice have a high blood CD4/CD8 T cell ratio; however, the specific role of CD8 T cells in ASD has not been reported.^[51]

5. Conclusion

In this study, we explored the mechanism of ASD by analyzing immune infiltration in ASD. We speculate that neutrophils, resting mast cells, B cells naive, CD8+ T cells, and Treg cells are potential core immune cells in ASD and participate in the occurrence of ASD. In addition, this study established the first immune correlation prediction model of ASD. It may be helpful for potential diagnosis and treatment targets, and the construction of risk prediction models will help identify patients with autism at an early stage, provide early intervention, and improve the prognosis of patients with ASD.

Acknowledgments

We thank all our colleagues in the Neonatal Intensive Care Unit of the First Affiliated Hospital of Zhengzhou University for their support of our project.

Author contributions

Conceptualization: Mengyuan Lei, Xinru Cheng, Zanyang Shi, Fengxia Mao, Qian Zhang.

Data curation: Wenjun Cao, Chenghan Luo, Zhaohan Fan, Mengyuan Lei, Xinru Cheng, Zanyang Shi, Qianya Xu, Zhaoqin Fu, Qian Zhang.

Formal analysis: Zhaohan Fan.

Methodology: Xinru Cheng, Zhaoqin Fu.

Validation: Zanyang Shi.

Visualization: Wenjun Cao, Zhaohan Fan, Zanyang Shi.

Writing – original draft: Wenjun Cao.

Writing – review & editing: Wenjun Cao, Chenghan Luo, Mengyuan Lei, Zanyang Shi, Fengxia Mao, Qianya Xu, Zhaoqin Fu, Qian Zhang.