Integrative analysis of γδT cells and dietary factors reveals predictive values for autism spectrum disorder in children

BACKGROUND
Autism spectrum disorder (ASD) includes a range of multifactorial neurodevelopmental disabilities characterized by a variable set of neuropsychiatric symptoms. Immunological abnormalities have been considered to play important roles in the pathogenesis of ASD, but it is still unknown which abnormalities are more prominent.


METHODS
A total of 105 children with ASD and 105 age and gender-matched typically developing (TD) children were recruited. An eating and mealtime behavior questionnaire, dietary habits, and the Bristol Stool Scale were investigated. The immune cell profiles in peripheral blood were analyzed by flow cytometry, and cytokines (IFN-γ, IL-8, IL-10, IL-17A, and TNF-α) in plasma were examined by Luminex assay. The obtained results were further validated using an external validation cohort including 82 children with ASD and 51 TD children.


RESULTS
Compared to TD children, children with ASD had significant eating and mealtime behavioral changes and gastrointestinal symptoms characterized by increased food fussiness and emotional eating, decreased fruit and vegetable consumption, and increased stool astriction. The proportion of γδT cells was significantly higher in children with ASD than TD children (β: 0.156; 95% CI: 0.888∼2.135, p < 0.001) even after adjusting for gender, eating and mealtime behaviors, and dietary habits. In addition, the increased γδT cells were evident in all age groups (age < 48 months: β: 0.288; 95% CI: 0.420∼4.899, p = 0.020; age ≥ 48 months: β: 0.458; 95% CI: 0.694∼9.352, p = 0.024), as well as in boys (β: 0.174; 95% CI: 0.834∼2.625, p < 0.001) but not in girls. These findings were also confirmed by an external validation cohort. Furthermore, IL-17, but not IFN-γ, secretion by the circulating γδT cells was increased in ASD children. Machine learning revealed that the area under the curve in nomogram plots for increased γδT cells combined with eating behavior/dietary factors was 0.905, which held true in both boys and girls and in all the age groups of ASD children. The decision curves showed that children can receive significantly higher diagnostic benefit within the threshold probability range from 0 to 1.0 in the nomogram model.


CONCLUSIONS
Children with ASD present with divergent eating and mealtime behaviors and dietary habits as well as gastrointestinal symptoms. In peripheral blood, γδT cells but not αβT cells are associated with ASD. The increased γδT cells combined with eating and mealtime behavior/dietary factors have a high value for assisting in the diagnosis of ASD.


Introduction
Autism spectrum disorder (ASD) is a neurodevelopmental disease marked by social and communication impairments, as well as restricted interests and repetitive behaviors (Regier et al., 2013). The incidence is around 1-3% globally and is showing a continued increasing trend (Arinda et al., 2021;Christensen et al., 2018;Christensen et al., 2019;Gentil-Gutiérrez et al., 2021;Málaga et al., 2019;Rahaman et al., 2021;Sabbagh et al., 2021). The etiology and pathogenesis of ASD remain largely unclear, but abnormal immune function has been frequently identified as one of the associated factors Ahmad et al., 2017a;Ahmad et al., 2017b). Maternal immune activation during pregnancy has been implicated as a potential cause of abnormal synaptic pruning and microglia-mediated neurogenesis in children with ASD (Bergdolt and Dunaevsky, 2019;Estes and McAllister, 2016;Faust et al., 2021), and maternal infection during pregnancy may lead to an increased risk of ASD (Jiang et al., 2016). Proinflammatory cytokines such as IL-6, IL-17A (Brimberg et al., 2013;Singer et al., 2008) and circulating anti-lymphocyte antigen antibodies (Meltzer and Van de Water, 2017;Warren et al., 1990) and anti-fetal brain antibodies are increased in the serum of children with ASD's mothers, and these are associated with the increased susceptibility to ASD and delayed brain development after birth. Importantly, in children with ASD increased cytokines such as IL-17A, IL-8, IL-1β, IL-6 in the peripheral blood are closely related to impairments in behavior (Hoogenraad and Riol-Blanco, 2020), memory, learning, cognition, and adaptability and may play important roles in ASD development (Ashwood et al., 2011;Goines and Ashwood, 2013;Hughes et al., 2022;Nadeem et al., 2022;Nadeem et al., 2020a;Nadeem et al., 2020b;Patel et al., 2020). Moreover, dysregulated immune cells, such as decreased CD4T cells, especially regulatory T cells, and increased Th17/Treg cells of the CD4 subset, are seen in children with ASD (Ahmad et al., 2017b;Akbari et al., 2022;De Giacomo et al., 2021;Ellul et al., 2021). However, contradictory results suggest that the general CD4T cell proportion is not significantly associated with ASD-related traits based on the blood-derived methylation analysis (Yap et al., 2021).
Recent studies also suggest that children with ASD are commonly affected by eating disorders and have a high incidence of gastrointestinal co-morbidities including chronic constipation and diarrhea (Baraskewich et al., 2021;Ding et al., 2020;Sanctuary et al., 2018). Clearly, the metabolic fate of dietary metabolites, such as fat, protein, and carbohydrates, is important for immune function (Lee and Dixit, 2020). For example, consumption of cranberry polyphenols is found to enhance human γδT cell proliferation and reduce the symptoms associated with colds and influenza (Nantz et al., 2013), and ingestion of concentrated fruit and vegetable juices increases the number of circulating γδT cells (Nantz and Rowe, 2006). Clinical studies have also demonstrated that oral administration of L-theanine, a tea component, can enhance γδT cell proliferation and interferon-γ secretion to strengthen its effect in immune surveillance against malignant cells (Percival et al., 2008). Thus, there is a close relationship between γδT cells and diet. Dietary factors and the gut microbiota (Ding et al., 2020) appear to be closely related to ASD and may play an important role in ASD development; however, the association between dietary factors, immune dysfunction, and ASD remains poorly understood. To test whether ASD children have altered peripheral immune cell profiles, particularly γδT cells, and the potential association with dietary factors, we conducted a cohort study including a total of 105 children with ASD and 105 typically developing (TD) controls, all of whom came from one single center in order to guarantee that the criteria were consistent for all of the assessments. The results were further validated using an external validation cohort of an additional 82 ASD and 51 TD children.

Study design and participants
Children with a definitive diagnosis of ASD and aged 3 to 12 years were included unless they met the following exclusion criteria: (a) eating a special diet or having eaten a special diet within the past 3 months; (b) ASD combined with other neurological diseases; (c) complications with inherited metabolic diseases or autoimmune/inflammatory disease; (d) experiencing a fever or other infectious symptoms; or (e) suffering from Crohn's disease or other digestive system diseases. All diagnoses were made by two professional child psychologists through DSM-5 testing. Informed consent was collected after explaining the study in detail to the participants and their guardians, and the participants were informed that they could withdraw from the project if they had any hesitation or reluctance to participate during the experiment (Fig. 1A). All of the children with ASD came from the Department of Child Developmental Behavior of the Third Affiliated Hospital of Zhengzhou University, Henan Province, China, and were enrolled from October 2019 to June 2021 (n = 105). The TD children matched as controls were from a kindergarten in the same region (n = 105). The internal validation cohort set was selected randomly from the primary cohort. The external validation set was recruited from the Department of Child Developmental Behavior of the Third Affiliated Hospital of Zhengzhou University, Henan Province, China, and were enrolled from October 2021 to October 2022 and included 51 children with ASD matched for age and gender with TD children from a kindergarten in the same region at the same time.
In this case-control study, we detected CD3T, γδT, αβT, and Vγ9δ2T immune cells in the peripheral blood of the participants by flow cytometry. Their dietary behavior and habits and Bristol Stool Scale scores were investigated in parallel. A deep analysis to explore potential information was performed through machine learning (Fig. 1B).

Sample size calculation
The sample size was calculated using Epitools (https://epitools.au svet.com.au/onemean). The expected difference between the ASD and TD groups was based on our previous study on ASD . The sample size calculation was conditioned on a standard deviation value of 8.32, 95% power (α = 0.05; two-sided) and 1.7% accuracy, and 93 children were required for each group. Based on our previous clinical study experience, we presumed a drop-out rate of 11% through infection, efficacy, compliance, or other unforeseeable factors, implying a minimum of 105 children for each group.

The dietary assessment and eating and mealtime behaviors survey
An indirect method to assess gastrointestinal health is through a dietary assessment. The questionnaire was based on our previously published protocol with modification to better suit the dietary characteristics of Chinese children . The dietary assessment questionnaire included diet flavor preferences, nutrition supplements, and the consumption of vegetables, fruits, eggs, etc., (Supplementary Table 1).
The Children's Eating Behavior Questionnaire (CEBQ) (Wardle et al., 2001) was administered to investigate the children's eating and mealtime behaviors (Supplementary Table 2). The questionnaire includes eight subscalessatiety responsiveness, speed of eating, food fussiness, food response, food preferences, eagerness for beverages, emotional undereating, and emotional overeating. Higher scores indicate more extreme eating behavior.

Assessment of gastrointestinal symptoms
Gastrointestinal symptoms were assessed by the Bristol Stool Scale that describes the shapes and types of stools . The chart classifies the stool into seven levels, namely type I: a dry nut-like ball-shaped stool that is difficult to discharge; type II: sausage-like but very hard; type III: sausage-like with surface cracks; type IV: sausagelike or snake-like and smooth and soft; type V: soft lumps with clearcut edge (and easy to discharge); type VI: loose fragments, broken edges, or mushy stool; and type VII: watery stool with no solid parts. Type IV is an ideal healthy state.

Flow cytometry assay with cytokine intracellular staining
To determine the cellular source of plasma cytokines, 200 µl of fresh blood samples from 14 ASD and 13 gender and age-matched TD children were stimulated by Leukocyte Activation Cocktail (BD Bioscience Gol-giPlug™, Cat.

Statistical analyses
Statistical analysis was performed using GraphPad Prism 8, R 4.2.0, and IBM SPSS 26.0. Scatter plots and columnar statistics were plotted in GraphPad Prism 8. In the univariable analysis, normality testing of the data was performed by the Kolmogorov-Smirnov test. For parametric data, unpaired Student's t-test or one-way ANOVA with Tukey's posthoc test was used for comparisons of two or three groups, respectively. For non-parametric data, the Mann-Whitney U test or Kruskal-Wallis test with Dunn's post-hoc test was used as appropriate. All immune profile assays were performed in a blinded manner, and the data were subsequently unblinded and analyzed independently by a statistician.

Statistics used in the machine learning analysis
In the multivariate analysis, variables with p ≤ 0.05 were recruited in the least absolute shrinkage and selection operator (LASSO) binary logistic regression model (R packages "psych" and "glmnet"). Tuning parameter (λ) selection in the LASSO model used 10-fold crossvalidation via minimum criteria. Variables at the optimal values, as determined by using the minimum criteria and the 1 standard error of the minimum criteria (the 1-SE criteria), were added to further multivariate binary logistic regressions in a stepwise manner. The variables screened by multivariate logistic regression analysis (R package "plyr") were introduced to construct nomograms (R packages "survival", "survminer", and "regplot"). The criterion for the result was having statistical significance at p ≤ 0.05. Samples were then randomly split into a 7:3 ratio (R packages "caret", "ggplot2 ′′ ), with the 70% portion being used as the internal validation set. Receiver operating characteristic (ROC) curves were plotted (R package "pROC"), and the area under the curve (AUC) was calculated to test the performance of our nomograms in the training and validation groups. Calibration curves (R packages "calibrate" and "rms") were used to assess the degree of agreement between the actual results and the nomogram predictions in the training and validation groups. The bootstrap method (R package "riskRegression") was used to repeat the sampling 1000 times, and the relative corrected consistency index (C-index) was calculated to verify the prediction performance of the nomogram. If the C-index or AUC was > 0.75, the nomogram we established was considered to have good predictive performance, while it was considered acceptable at 0.5 ~ 0.75. Finally, we calculated the net benefit, drew the decision curve (R packages "caret" and "ggplot2 ′′ ), and used the decision curve analysis method to judge the clinical utility of the nomogram.

Eating and mealtime behavior changes, dietary habits, and gastrointestinal symptoms in children with ASD
As judged by the CEBQ (Table 1), children with ASD showed a significant disparity in eating and mealtime behavior scores for food fussiness, food response, food preferences, eagerness for beverages, and emotional eating compared with TD children. While taste preferences and consumption of fruits, vegetables, dairy, and eggs showed great disparity between the two groups ( Fig. 2), there were no significant differences regarding satiety responsiveness or speed of eating between the children with ASD and TD children. Regarding stool types, the proportions of each stool type showed a large discrepancy between children with ASD and TD children. Among them, Type IV stool only accounted for 5.7% of children with ASD compared with 74.3% in the TD children, while Type III stool was seen in 68.6% of the children with ASD (p < 0.001 compared to controls). These results suggest that there were significant differences in eating and mealtime behavior and stool types between children with ASD and TD children.

The influence of age and gender on eating and mealtime behavior, dietary habits, and gestational symptoms in children with ASD
We compared the eating and mealtime behavior in those < 48 months of age versus those ≥ 48 months of age, and we found that there were clearly different eating and mealtime behaviors between children with ASD and TD children, despite the differences in each parameter not always being significant for both age groups (Supplementary Table 2). In both children younger and older than 48 months, there were significant differences for Bristol Stool Scale, food fussiness, emotional eating, and consumption of vegetables, dairy, and eggs between children with ASD and TD children.
When analyzing the association between gender and eating and mealtime behavior and gestational symptoms, we found that Bristol Stool Scale, food preferences, emotional eating, salty preference, and consumption of vegetables, dairy, and eggs differed greatly in boys with ASD compared with TD boys but not in girls (Supplementary Table 3).

Alterations in peripheral blood immune cells in children with ASD
Flow cytometry analysis of CD3T, αβT, and γδT cells in the peripheral blood showed distinct cluster groups ( Fig. 3B-3C). There was no significant difference in CD3T cells between children with ASD and TD children, but the proportion of αβT cells was significantly lower in children with ASD compared to TD children. In contrast, the proportion of γδT and Vγ9δ2T cells and the γδT/αβT ratio in the ASD group were significantly higher than in the TD group (Fig. 3D, 3E).
To exclude the potential influence of confounderssuch as demographics, eating and mealtime behaviors, dietary habits, and gastrointestinal symptomson the peripheral immune cell profile changes observed in children with ASD, we conducted further analysis using linear multivariable regression models (Table 2)

The plasma cytokine changes in children with ASD
To further explore the immune profile changes in children with ASD, cytokines in the plasma in all participants were analyzed by Luminex assay. As shown in Fig. 4A, among all cytokines analyzed, including IFNγ, IL-8, IL-10, IL-17A, TNF-α, there were significant increases in IFN-γ and IL-17A and significant decreases in IL-8 in children with ASD compared with the TD group. Additional FACS intracellular staining showed increased γδT-IL17A and decreased γδT-1L-8 in children with ASD (Fig. 4B), suggesting that altered γδT cells may contribute to the plasma cytokine profile changes in ASD.

The influence of age on immune profiles in children with ASD
We conducted further analysis to explore whether or not immune profile changes in ASD are affected by age. According to the developmental characteristics of immune cells (Bunders et al., 2005;Moosmann et al., 2022;Simon et al., 2015;Song et al., 2022), we analyzed the immune profiles of children with ASD using 48 months as the cut-off point. In children < 48 months, there were no significant changes in CD3T cells but significantly decreased αβT cells and increased γδT cells, Vγ9δ2T cells, and γδT/αβT ratio in the children with ASD compared to TD children ( Supplementary Fig. 2A). The plasma cytokine analysis showed increased IFN-γ and IL-17A but decreased IL-8 in children with ASD compared to TD children.
In children ≥ 48 months ( Supplementary Fig. 2B) there were significantly increased percentages of γδT and Vγ9δ2T cells and increased γδT/αβT ratio in children with ASD compared to TD children. The plasma cytokine IL-8 was decreased while IL-10 was increased in children with ASD versus TD children.

The influence of gender on immune profiles in children with ASD
We further analyzed the immune profiles according to gender (Supplementary Fig. 3). In boys, the percentages of CD3T, γδT, and Vγ9δ2T cells and the γδT/αβT ratio were increased, while αβT cells were decreased in children with ASD compared to TD controls, and the plasma level of IL-17A was increased and IL-8 was decreased in boys with ASD compared to TD controls. In girls, the percentages of γδT and Vγ9δ2T cells and the γδT/αβT ratio were increased, and the plasma cytokine levels of IFN-γ, IL-10, and IL-17A were higher while IL-8 was lower in the ASD group compared to TD controls.

γδT was independently associated with ASD after adjusting for confounders
To exclude the potential influence of confounders (such as eating and mealtime behavior and gastrointestinal symptoms) on immune profile changes in children with ASD, we conducted multivariable regression analysis. After adjusting for gender and for significant influencing parameters (Table 3) determined by multivariable regression of eating and mealtime behaviors (for age < 48 months these were Bristol Stool Scale, food fussiness, and total score, and for age ≥ 48 months this was emotional eating) and for dietary habits, γδT cells were the only cell type independently associated with ASD in both age groups (age < 48 months: β: 0.288; 95% CI: 0.420 ~ 4.889, p = 0.02; age ≥ 48 months: β: 0.485; 95% CI: 0.694 ~ 9.352, p = 0.024).
Similarly, we adjusted for factors identified by the multivariate regression model to determine the true relationship between immune cells and gender in the development of ASD (Table 4). The corrected results after the adjustment showed that CD3T cells (β: 0.227; 95% CI: 0.181-13.499, p = 0.044), αβT cells (β: 0.206; 95% CI: 0.093-8.048, p = 0.045), and γδT cells (β: 0.174; 95% CI: 0.834-2.625, p < 0.001) were statistically significant in each model in the boys with ASD, while the immune cell profiles were affected by dietary behavior, dietary habits, and other immune cells in different models in girls with ASD. Taken together, these results suggest that γδT cells were independently associated with ASD in both younger (<48 months) and older (≥48 months) boys with ASD but not in girls with ASD.

The peripheral immune cell changes in children with ASD were confirmed in an independent validation cohort
To confirm the immune profiles observed in the initial analysis cohort, we recruited participants in the external validation cohort, including 82 children with ASD (69 boys and 13 girls) and 51 TD children (42 boys and 9 girls). As expected, flow cytometry analysis showed decreased percentages of αβT cells and increased proportions of γδT cells and Vγ9δ2T cells in children with ASD compared to TD children (Fig. 5A, 5B). In addition, increased CD3T, γδT, and Vγ9δ2T cells and decreased αβT cells were found in boys with ASD, consistent with the results obtained in the previous analysis (Fig. 5C). When dividing the groups by age (<48 months: 35 children with ASD versus 18 TD children; >48 months: 47 children with ASD versus 33 TD children), similar trends in immune cell alterations were seen as described above (Fig. 5D).

The immune profile in children with ASD was not associated with eating and mealtime behavior, dietary habits, or gastrointestinal symptoms
To further determine if there were any potential relationships between the peripheral immune cells and eating and mealtime behavior, dietary factors, and gastrointestinal symptoms in children with ASD, we conducted an analysis using the R package "cor". The associations between immune cells and eating and mealtime behavior, dietary habits, and Bristol Stool Scale scores are shown in Fig. 6. There were no significant correlations between immune cells and dietary behavior, dietary habits, or Bristol Stool Scale scores, although the Bristol Stool Scale score was negatively correlated with food fussiness and emotional eating.

Machine learning analysis showed that immune profile alterations in ASD have potential clinical applications
To determine the potential predictive value of the combination of immune cell profile with eating and mealtime behavior in children with ASD, machine learning nomograms, which use a points-based system in which a patient accumulates points based on levels of risk factors, were used to predict medical outcomes.
First, we used univariable analysis and selected the variables with p ≤ 0.05 among all variables, including both eating and mealtime behaviors and immune profile markers (Fig. 7A). Using these variables for subsequent LASSO regression analysis, we found 17 correlated variables at log λ minimum and 4 variables at log λ1_SE (Fig. 7B). We selected the 4 variables for further analysis, which included food fussiness, γδT cells, emotional undereating, and salty taste preference.
The ROC analysis was used to determine the discriminatory ability of  (Fig. 8A). Machine learning was performed separately by age and by gender . The discrimination of the models is shown in Fig. 8B (age < 48 months: model 1 AUC = 0.982; age ≥ 48 months: model 2 AUC = 0.944; boys: model 3 AUC = 0.981; girls: model 4 AUC = 1.00), and the decision curves (Fig. 8C) showed that children can receive obtain higher diagnostic benefits in the nomogram models (age < 48 months: model 1; age ≥ 48 months: model 2; boys: model 3; girls: model 4) suggesting a good predictive value for the diagnosis of ASD.

Discussion
In the current study, we conducted a careful analysis in a large cohort of children with ASD and found that these children had significantly altered eating and mealtime behaviors and obvious gastrointestinal symptoms compared to TD children. Children with ASD showed altered immune profiles characterized by increased γδT cells and IL-17A and decreased IL-8 in the peripheral blood. Importantly, the immune profile in children with ASD was not associated with eating behaviors, dietary habits, or gastrointestinal symptoms. However, the increased numbers of γδT cells in the peripheral blood in combination with different eating and mealtime behaviors and dietary factors had a high predictive value for assisting in the diagnosis of ASD, especially for children older than 48 months.
The γδT cells, which are ubiquitous on mucosal and epithelial surfaces (Howie et al., 1998), recognize antigens without major histocompatibility complex restriction (Frascoli et al., 2012), and they are critical in the defense against bacterial and fungal pathogens (Zhang et al., 2017). In the peripheral blood, γδT cells make up a relatively smaller proportion among all leucocytes, among which Vδ2 TCR and Vγ9 TCR are the two main subsets (Roden et al., 2008). In this cohort, we found increased percentages of γδT cells in children with ASD compared with TD children, and this trend persisted even after age division and gender stratification, except in the case of girls. Importantly, γδT cells were independently associated with ASD in boys. Although the gender bias in γδT cells has not been reported previously, it is well recognized that ASD has a dramatic gender-biased prevalence and is gender-risked (Hirota and King, 2023;Loomes et al., 2017). Gender is also a biological variable that impacts the functions of the immune system (Klein and Flanagan, 2016;Takahashi and Iwasaki, 2021;Wilkinson et al., 2022). Studies have shown that boys with ASD have higher proportions of CD2 + , CD4 + , and CD8 + immune cells , while IL-8, IL-1β, and IL-15 are shown to be gender-biased (Masi et al., 2017). Additionally, genes belonging to glial/immune modules (Gupta et al., 2014;Voineagu et al., 2011) and astrocyte and microglial marker genes in prenatal cortex (Cahoy et al., 2008;Kissel and Werling, 2022;Walsh et al., 2022;Werling et al., 2016) are found to be male-biased in ASD. Conversely, the ASD-downregulated neuronal/synaptic modules (M12, Mod1) are female-biased (Kissel and Werling, 2022). Furthermore, we found that γδT cells were vital for the ASD diagnosis in both age groups and in boys (but not in girls) by LASSO analysis. In all, the increased γδT cell proportion is associated with ASD and may serve as one of the predictive factors for children with ASD.
The γδT cells are activated by signals such as cellular antigens or antibodies, and they differentiate into subsets expressing different cytokines in a manner similar to CD4T cells, including γδT1, γδT2, γδT17, γδTfh, and γδTreg (Frascoli et al., 2012;McVay et al., 1998). These subsets release their respective factors to exert immunosuppressive or stimulatory functions. Here we found increased levels of IL-17A in the plasma and in γδT cells, in children with ASD, and this was consistent with previous studies (Al-Ayadhi and Mostafa, 2012;Basheer et al., 2018;Moaaz et al., 2019;Nadeem et al., 2019). IL-17A is a cytokine produced by Th17 cells and γδT cells (Albertsson et al., 2018;McVay et al., 1998), and blocking IL-17A/IL-17 receptor signaling may be beneficial for children with ASD (Nadeem et al., 2018). IL-17A is a potent pro-inflammatory cytokine that has been shown to inhibit the proliferation of neural progenitor cells (Ribot et al., 2014). Furthermore, in newborn mice meningeal γδT cells have Th17-like features (Zelco et al., 2021), and this promotes abnormal cortical and ASD-like behaviors (Choi et al., 2016). However, IL-17A can induce two opposing behavioral outcomes depending on when its upregulation occurs, again suggesting immune dysfunction in ASD. It can regulate hippocampal neurogenesis to improve spatial learning in mice (Choi et al., 2022) and can ameliorate anxiety-like behaviors (Alves de Lima et al., 2020;Reed et al., 2020) as well as promote neurite outgrowth (Chisholm et al., 2012;Choi et al., 2016). In addition, studies have found that the permeability of the blood-brain barrier changes in children with ASD (Fiorentino et al., 2016;Sweeney et al., 2018), and thus peripheral immune abnormalities might affect the central immune system. Also, increased IFN-γ was seen in children with ASD, although only in girls and in children with ASD younger than 48 months. Indeed, neurons are found to respond to IFN-γ secreted by meningeal T cells in animal models (Choi et al., 2022), which in turn enhances GABAergic inhibition (Filiano et al., 2016). Stimulating microglia by IFN-γ impairs neurogenesis in the hippocampus and leads to depression-like behavioral and cognitive deficits .
IL-8, also known as chemokine CXCL8, is a cytokine secreted by macrophages, epithelial cells, and γδTh1 cells. IL-8 can be synthesized and released by the above cells through the activity of IL-1, TNF-α, PHA, LPS, and other inducers (Baggiolini and Clark-Lewis, 1992), and IL-8 induces neutrophils to express surface adhesion molecules, release storage enzymes, promote respiratory bursts, generate reactive oxygen species metabolites, and activate a series of inflammation reactions. In human newborns, IL-8 is found to have the potential to activate antimicrobial neutrophils and γδT cells (Gibbons et al., 2014). The results of studies of IL-8 in ASD have been ambiguous (Masi et al., 2015;Saghazadeh et al., 2019), but studies have found that IL-8 has significant gender differences and is negatively correlated with severity in the female ASD population (Masi et al., 2017). IL-8 is also associated with the levels of parental cytokines and the subdomains of social awareness, cognition, and motivations in children with ASD (Shen et al., 2021). Intriguingly, in the current study, which to our knowledge represents the largest sample size for cytokine analysis (n = 83/group), we found highly significant decreases in IL-8 in all groups of children with ASD, and this had a high diagnostic value for distinguishing children with ASD from TD children when combined with measurements of γδT cells (Fig. 8A, AUC: 0.802). Taken together, these results provide evidence of an altered cytokine profile and inflammatory state, thus confirming the immune dysfunction in ASD.
Many studies have reported that children with ASD have cooccurring eating and mealtime behavior changes. Gastrointestinal disturbances, particularly constipation, are often reported to afflict children with ASD (Ferguson et al., 2019;Holingue et al., 2018;Restrepo et al., 2020), and these disturbances are related to the disease severity. Thus, we investigated these factors in the same cohort and analyzed the possible association between these factors and γδT immune cells. We found that the dietary structure of children with ASD differed from that of TD children, not only in eating and mealtime behaviors such as food fussiness, food response, food preference, and emotional eating, but also in dietary habits such as nutritional supplementation, consumption of fruits, vegetables, and dairy products, and gastrointestinal symptoms, which is consistent with previous findings (Leader et al., 2020). These findings suggest that dietary factors and gastrointestinal symptoms are often co-existing and non-negligible in ASD. Individuals with ASD have specific sensory patterns and can be particularly sensitive to olfactory, gustatory, or other food characteristics, which makes them selective for food and limits their diet, leading to gastrointestinal problems (Ristori et al., 2019). However, no correlation between immune cells and diet or gastrointestinal symptoms was found, which agrees with previous findings showing that dietary preferences are associated with alpha and beta diversity in the gut microbiota but are not significantly associated with CD4 T cell proportions in children with ASD (Yap et al., 2021).
There are some limitations of this study. Firstly, although to our knowledge this is so far the largest cohort study on circulating γδT profiles in children with ASD, sampling bias may still exist, particularly when subgroup analyses were performed, and larger sample size studies thus need to be conducted to overcome this shortcoming. Secondly, there is a lack of systematic observational studies on ASD diet factors, so the conclusions drawn from the current study are more appropriately used as a reference for systematic dietary research. Thirdly, a wider range and more complete spectrum of cytokine profiles and their cellular source were not included in the current study, which would have provided a more complete picture of immune profile alterations in children with ASD.
In summary, children with ASD present with significant disparities in eating and mealtime behaviors as well as gastrointestinal symptoms compared to TD children. In peripheral blood, γδT cells but not αβT cells are associated with ASD, but this association is not related to eating and mealtime behavior changes (Fig. 9). To our knowledge, this is the first study revealing abnormalities in the proportion of γδT cells in the peripheral blood in children with ASD, and the increased γδT cells together with eating and mealtime behavior changes and dietary factors might assist in the diagnosis of ASD.

Author Contributions
LZ analyzed the data and wrote the manuscript. YX designed the study and analyzed the data. SS, LC, DP, ML, HL, YL, TJ, SJ, XS, YY, YW, ZL, CS and GD provided the clinical data of ASD patients. WL, HL, LX, and BL ensured the accuracy and integrity of the results. JWL provided support for data interpretation and contributed to manuscript revision. XW, CZ devised and supervised the study. All authors read and approved the final version of the manuscript.

Funding
This study was supported by the National Natural Science Foundation of China (U21A20347 and 82203969), the Henan Key Laboratory of Population Defects Prevention (ZD202103), the Department of Science and Technology of Henan Province, China (212102310221), the Swedish Research Council (2018-02267, 2022-01019), Swedish Governmental grants to scientists working in health care (ALFGBG-965197), the Brain Foundation (FO2022-0120), the Adlerbert Research Foundation , and Stiftelsen Edit Jacobsons Donationsfond (2021-102).

Data availability
Data are available in the main text or the supplementary materials and also are available from the corresponding author upon reasonable request.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request. Fig. 9. Summary of the major findings in the current study. Serious eating behavior problems in autistic children are more common, such as food fussiness (resistance to fruit and vegetable intake) and emotional eating, as well as disturbed gastrointestinal symptoms. Moreover, γδT cells were increased in blood. The plasma cytokine levels of IFN-γ, IL-10, and IL-17A were higher while IL-8 was lower in children with ASD compared to TD controls.