Circulating neuronal proteins are elevated in severe autism and associated with patient-specific neuronal dysfunction and behavioral deficits

Abstract

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition with a multifactorial etiology involving both genetic and environmental factors. Although previous studies have identified altered levels of circulating proteins in individuals with ASD, the functional significance of these peripheral biomarkers remains largely unclear. In this study, we recruited a well-characterized cohort of young Arab children with ASD (n = 100) and matched controls (n = 60), aged 2–4 years, in Qatar. We used a multimodal approach, integrating data from proteomics, transcriptomics, induced pluripotent stem cell (iPSC)-derived cortical neurons, cellular imaging, functional assays using microelectrode recordings, and behavioral assays in Drosophila models. High-throughput screening of plasma samples revealed considerable heterogeneity in circulating proteomic profiles across different Arab subpopulations; therefore, we focused our subsequent analyses on Qatari children. We identified elevated levels of several neuronal proteins in the peripheral circulation of individuals with severe, but not mild, ASD symptoms. Furthermore, we found that neurons derived from the severe ASD group showed increased MAP2⁺/DCX⁺ cell counts, upregulated expression of genes associated with neuronal activity, and increased neuronal hyperexcitability in functional assays. Moreover, targeted overexpression of these proteins specifically in the blood of Drosophila models was sufficient to induce ASD-relevant behaviors. Together, our findings reveal key links between peripheral protein dysregulation and neuronal function and behavior, offering new insights into systemic contributions to ASD pathophysiology and highlighting potential therapeutic targets for mitigating symptom severity.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by persistent difficulties in social communication and interaction, alongside restricted and repetitive patterns of behavior. The global prevalence of ASD is estimated to be 1 in 100 children, with a median male-to-female ratio of 4.2 [1]. Symptoms typically emerge within the first three years of life, although the severity of these symptoms varies widely among affected individuals [2]. Currently, there is no cure for ASD; however, studies have demonstrated that early intervention programs can effectively alleviate some of the severity of symptoms [2]. Although the underlying causes of ASD are not fully understood, current evidence suggests that its etiology is multifactorial, involving a combination of genetic predispositions, environmental influences, and the complex interplay between these factors [3–5].

Numerous studies have reported altered expression levels of circulating proteins, such as cytokines, growth factors, and autoantibodies, in individuals with ASD [6–9]. Notably, dysregulated expression of several circulating proteins has been associated with increased severity of ASD symptoms [9–11], suggesting a functional role of these proteins in the disorder. For example, abnormal levels of several cytokines, such as interleukin (IL)-6, tumor necrosis factor (TNF)-α, and IL-1β, have been reported in the blood of individuals with ASD and correlated with worse behavioral outcomes [12–16]. Importantly, peripheral blood cytokines can cross the blood–brain barrier and signal to the brain, directly affecting brain function and behavior [17–20]. These observations highlight the possibility that systemic changes may contribute to neurodevelopmental abnormalities characteristic of ASD.

Despite the abundance of reports identifying differentially expressed circulating proteins in ASD, the functional significance of these changes remains largely unclear. It is not yet well understood whether these altered protein levels are causally involved in the development or progression of ASD, represent compensatory mechanisms, or are simply non-specific correlates of broader systemic dysregulation [10, 21]. This gap in knowledge limits our ability to interpret the biological relevance of circulating protein profiles in ASD and to leverage them for diagnostic or therapeutic purposes. In particular, it is not yet established whether specific circulating proteins represent neurodevelopmental changes relevant to ASD, nor whether their levels correlate with core clinical features of the disorder.

In this study, we aimed to investigate the functional involvement of altered circulating proteins in ASD pathophysiology. We recruited a well-defined study cohort and employed an integrative approach combining proteomics, transcriptomics, induced pluripotent stem cell (iPSC)-derived neurons, cellular imaging, functional assays using microelectrode recordings, and behavioral assays in Drosophila models. We performed high-throughput proteomic profiling of plasma samples from individuals with ASD and controls, followed by characterization of the neuronal features derived from the same individuals exhibiting dysregulated expression of circulating proteins. Additionally, we generated novel Drosophila models of ASD with targeted gene edits specifically in the hemocytes (analogous to human blood cells), followed by assessment of ASD-relevant behaviors. Our findings provide novel insights into the potential mechanistic relevance of peripheral protein alterations in ASD and suggest new avenues for investigating systemic contributions to neurodevelopmental disorders.

Materials and methods

Study population

A total of 160 children were enrolled into this study, including 100 children with ASD and 60 control subjects. Participants with ASD were recruited from the Child Development Center in Rumailah Hospital of Hamad Medical Corporation (HMC; Doha, Qatar), and their matched control participants were recruited from Al-Wajbah Health Center of Primary Health Care Corporation (PHCC; Doha, Qatar). This study was reviewed and approved by the Institutional Review Board (IRB) Ethics Committee of HMC (approval number: MRC-02-18-116) for the recruitment of subjects with ASD and the PHCC (approval number: 2020/06/064) for the recruitment of control subjects. The study also received exempt IRB approval from the Qatar Biomedical Research Institute (QBRI, Doha, Qatar; approval number: 2019-003). This study was conducted in accordance with the guidelines of the Declaration of Helsinki. Written informed consent for study participation was obtained from the families of all subjects.

Subject screening and enrollment in the study

All participants were initially screened using a self-reported questionnaire to confirm their adherence to our study’s eligibility criteria as previously described [7, 9]. The subjects enrolled in this study were of Arabic ethnicity and nationality and were between 2 and 4 years of age. Additionally, all subjects fulfilled the following criteria: 1) being born in Qatar; 2) having their mothers reside in Qatar for most of their pregnancy; 3) primarily living in Qatar since birth; 4) absence of immune conditions like autoimmune diseases, asthma, allergies, and eczema; 5) no neurological conditions such as epilepsy; 6) no suspected difficulties with vision, hearing, or mobility; 7) no other medical complications such as heart, lung, or kidney diseases; and 8) not taking medications or having recent infections or vaccinations at the time of study enrollment. Control subjects were unrelated to ASD individuals, had no family history of ASD, and were assessed using the Social Communication Questionnaire (lifetime version) with a cutoff score < 12 to exclude ASD risk. Those diagnosed with ASD had received a clinical confirmation by qualified professionals according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) and Autism Diagnostic Observation Schedule-second edition (ADOS-2). Subjects with ADOS-2 scores < 7 were considered to have mild-to-moderate ASD while subjects with ADOS-2 scores ≥ 7 were considered to have severe ASD.

Sample collection and processing

Blood samples were extracted from the enrolled subjects into EDTA-containing anti-coagulant tubes (cat #3366643, BD Diagnostics) at HMC or PHCC sites, transported, and processed in the research facility at QBRI within two hours of sample collection. Blood samples were slowly layered over Histopaque-1077 (cat #10771, Sigma-Aldrich) at an equal ratio and centrifuged at 400 x g for 30 min with an acceleration of 3 and zero braking at standard room temperature of 20–23 °C. After centrifugation, plasma samples were collected from the upper layer into a new tube while the peripheral blood mononuclear cells (PBMCs) were carefully collected from the middle fraction into another tube without mixing with the plasma collected from the upper layer. Cells were washed twice with phosphate buffered saline (PBS; cat #10010015, Thermo Fisher Scientific), centrifuged at 400 x g for 10 min, then cryopreserved at −150 °C in liquid nitrogen for later use. Plasma samples were centrifuged at ~ 1800 × g for 15 min to remove cell debris, and plasma aliquots of 200 μl were stored at –80 °C until further analysis.

Olink proteomic assays

Plasma samples were analyzed using 11 commercially available Olink Target 96 Panels–Neurology, Neuro Exploratory, Inflammation, Immune Response, Cell Regulation, Development, Metabolism, Organ Damage, Cardiovascular II, Cardiovascular III, and Cardiometabolic–provided by Olink (Uppsala, Sweden). Each panel contains a total of 92 protein targets; however, a few targets are common across panels, resulting in a total of 984 unique proteins being analyzed. Samples were analyzed according to Olink’s standard protocol. Briefly, the plasma samples were randomized and aliquoted into 96-well plates. Each sample was incubated with a pool of oligonucleotide-labeled antibody pairs that bind to specific target proteins. In this configuration, proximity-dependent DNA polymerization occurs, and the resulting amplicons were quantified using a microfluidic chip-based qPCR system on a Fluidigm Biomark instrument. The signal from each unique DNA amplicon is proportional to the protein concentration and is expressed as a normalized protein expression (NPX) value. Various internal controls were included in each run to minimize inter- and intra-assay variability. The runs were conducted at QBRI in an Olink-certified proteomics core facility. All samples passed QC assessment and were used for downstream analysis. Differential expression analysis of Olink data was performed using the iDEP platform, version 2.0 [22]. Olink NPX values were uploaded into the iDEP software for analysis. Statistical significance was determined using a false discovery rate (FDR) of 0.05 and a fold change cutoff of 1.25.

Sample selection for iPSC generation

Due to the technical challenges of generating iPSC lines from every sample in the cohort, we employed stratified random sampling to select n = 3 samples from each study group. Participants in each group (i.e., controls, mild ASD, severe ASD) were first stratified into subgroups based on sex and nationality criteria, specifically including males of Qatari nationality, to allow for more accurate analyses and reduce the influence of demographic variations on the results. Within each stratified subgroup, individuals were manually selected at random from the full list of eligible Qatari male participants. This stratified random selection was performed manually, without the use of computerized tools or systematic ordering, ensuring unbiased representation across subgroups.

Generation of iPSCs from PBMC samples

We generated iPSC lines from a total of nine selected samples, as described earlier. These iPSC lines were derived from PBMC samples of n = 3 controls, n = 3 mild ASD (ADOS-2 scores 4–5), and n = 3 severe ASD (ADOS-2 scores 8–10). All these iPSC lines were fully characterized and validated, as previously described [23]. Briefly, PBMCs were reprogrammed into iPSCs using CytoTune-iPS 2.0 Sendai Reprogramming Kit (cat #A16517, Thermo Fisher Scientific) according to the manufacturer’s instructions. Pluripotency was confirmed through flow cytometry and spontaneous differentiation into three germ layers. Conventional karyotyping analysis was also performed to verify the normal karyotype of the generated iPSC lines.

Differentiation of iPSCs into cortical neurons

iPSCs were differentiated into cortical neurons as previously described [24], with minor modifications. Single-cell suspensions of iPSCs were seeded onto 6-well plates coated with Matrigel (cat #354277, Corning) and maintained in mTeSR1 plus medium (cat #100-0276, Stem Cell Technologies) supplemented with 10 μM Y-27632 (cat #72305, Stem Cell Technologies). Upon reaching 90–100% confluence, neuronal differentiation was initiated by replacing the mTeSR1 plus medium with neuronal medium consisting of a 1:1 mixture of DMEM/F12 (cat #11320-074) and neurobasal medium (cat #21103-049) supplemented with B-27 (cat #17504044), N2 supplement (cat #17502-048), L-glutamine (cat #25030081), non-essential amino acids (cat #11140050), 50 μM 2-mercaptoethanol (cat #21985-023), and penicillin-streptomycin (cat #15140-122) from Thermo Fisher Scientific. For dual SMAD inhibition, 10 μM SB431542 (cat #1614, Tocris) and 2 μM dorsomorphin dihydrochloride (cat #3093, Tocris) were added to the neuronal medium for 12 days.

Cells were dissociated on days 8 and 14 using TrypLE (cat #12604-013, Thermo Fisher Scientific) and replated onto fresh Matrigel-coated 6-well plates in neuronal medium containing 10 μM Y-27632. During the neural progenitor expansion phase (days 14–18), cultures were supplemented with 20 ng/mL bFGF (cat #233-FB, R&D Systems). On day 20, neural progenitor cells (NPCs) were either cryopreserved or plated for terminal differentiation on Matrigel-coated 12-well plates in neuronal medium further supplemented with 10 ng/mL BDNF (cat #248-BDB, R&D Systems), 10 ng/mL GDNF (cat #212-GD, R&D Systems), 2 μg/mL insulin (cat #4512-01, Merck), 20 μM dibutyryl cAMP (cat #1141, Tocris), and 200 μM L-ascorbic acid (cat #A4544, Sigma). On day 28, cells were transferred onto surfaces coated with 100 μg/mL poly-L-ornithine (cat #3436-100-01, R&D Systems) and 20 μg/mL laminin (cat #23017015, Thermo Fisher Scientific). For RNA experiments, 175,000 cells were seeded per well in 12-well plates, whereas for immunostaining, 30,000 cells were seeded per well in 8-well chamber slides (cat #154534PK, Thermo Fisher Scientific). Cultures were maintained for 4–8 weeks with half medium changes every 2–3 days.

RNA extraction from neurons, sequencing and analysis

RNA samples were extracted from cultured cortical neurons at week 4 or week 8 using Direct-zol RNA Miniprep (cat #R2051, Zymo Research), according to the manufacturer’s instructions. A NanoDrop 2000/2000c Spectrophotometer (cat #ND-2000, Thermo Fisher Scientific) was used to perform RNA quality checks, which showed a 260/280 ratio of approximately 1.7–2.0. RNA samples were stored at −80 °C until further analysis. We then performed library preparation, QC, and total RNA sequencing (150 bp, PE, ~50 M reads/sample). Sequencing data were processed using the DRAGEN RNA Pipeline (Illumina). Read counts were then exported and analyzed in iDEP platform, version 2.01 [22]. Statistical significance was determined using an FDR of 0.05 and a fold change cutoff of 1.50.

Immunostaining and quantification of cortical neurons

Cells were fixed in 4% paraformaldehyde for 20 min, permeabilized with 0.3% Triton X‑100 in PBS (PBST) for 30 min, and blocked in 6% bovine serum albumin (BSA) in PBST for 2 h at room temperature. Cells were then incubated overnight at 4 °C with the following primary antibodies diluted in 3% BSA in PBST: DCX (rabbit, 1:800; cat #48-1200, Thermo Fisher Scientific), Ki67 (mouse, 1:100; cat #ab238020, Abcam), and MAP2 (chicken, 1:500; cat #ab5392, Abcam). After washing, cells were incubated for 1 h at room temperature with Alexa Fluor‑conjugated secondary antibodies (all from Thermo Fisher Scientific), each diluted 1:1000 in 3% BSA in PBST: donkey anti-rabbit (Alexa Fluor 488; cat #A21206), donkey anti-mouse (Alexa Fluor 568; cat #A10037), and donkey anti-chicken (Alexa Fluor 647; cat #A78952). Finally, nuclei were counterstained with Hoechst 33258 (cat #H3569, Thermo Fisher Scientific) at 1:5000 in PBS for 5 min, then washed and embedded in Permount mounting medium (cat #SP15-100, Thermo Fisher Scientific), and coverslipped.

All images were captured at 20X magnification using 405, 488, 561, and 633 nm lasers on a confocal microscope (Zeiss LSM 780) equipped with ZEISS ZEN software (version 3.12). For each sample, three images were acquired from both week 4 and week 8 neurons (i.e., total 54 images across all study groups). Confocal images were analyzed using CellProfiler software (version 4.2.8) to ensure robust and unbiased cell quantification across groups. The analysis pipeline included the IdentifyPrimaryObjects module with a global Otsu thresholding method to detect four primary objects. For nuclear objects (Hoechst and Ki67⁺), a diameter range of 10–40 pixels was applied to exclude debris and merged nuclei, whereas cytoplasmic objects (DCX⁺ and MAP2⁺) were identified using a broader 20–80 pixel range to account for their cytoplasmic distribution. The RelateObjects module was then applied to quantify total co-localization events between total nuclei (Hoechst) and each of the other markers (Ki67, DCX, and MAP2).

Microelectrode array (MEA) analysis

Extracellular recordings were conducted using Axion’s Maestro Pro platform in combination with Axion 6-well CytoView MEA plates, as previously described [25, 26]. Each well contained an 8×8 (64-channel) electrode array with seven reference electrodes. Extracellular voltage recordings were collected at a sampling rate of 12.5 kHz per channel. A band-pass filter (200 Hz to 3 kHz cut-off frequencies) was applied, and a variable threshold spike detector was set at ±5.5 standard deviations of the root mean square of the background noise. Recordings were performed before media change or two days after media change. During recording, the MEA plate was maintained at 37 °C and 5% CO₂. MEA data for 5 min of recording showing active neural responses were selected for analysis. MEA data analysis was performed using the Axion Biosystems Neural Metric Tool (Axion Biosystems). An electrode was considered active with a threshold of 5 spikes/min. The weighted mean firing rate was defined as the average firing rate of only the active electrodes. A burst was defined as at least 5 consecutive spikes from a single active electrode with inter-spike intervals (ISI) of less than 100 ms. Network bursts were defined as at least 5 consecutive spikes across multiple electrodes with ISIs of less than 100 ms and a minimum involvement of 10% of the electrodes.

Fly husbandry and strains

All Drosophila strains were maintained at 25 °C, 70% humidity, with 12-hour light/dark cycles in Darwin Chambers (IN084-AA-LT-DA-MP). For genotypic controls, we used transgenic F1 progeny from crosses between the driver line (HML-Gal4, BDSC #6395) and responder lines (UAS-cDNA). Several UAS lines were obtained as gifts: UAS-Pent [27] from Dr. Marcos Gonzalez Gaitan, UAS-CDase [28] from Dr. Jairaj K. Acharya, and UAS-Contactin [29] from Dr. Manzoor A. Bhat. Additional lines were sourced from the Bloomington Drosophila Stock Center: UAS-EGFP (BDSC #5431), UAS-Robo2 (BDSC #66886), and UAS-Pvf1 (BDSC #58425). Transgenic F1 flies obtained from crosses of HML-Gal4 and responder lines were used in both hemolymph extractions and behavioral assays. All experimental flies were cultured on Nutri-Fly MF medium (cat #66-116, Genesee Scientific). The food medium was prepared by dissolving 177 g of NutriFly powder in 1 L of water, bringing the mixture to a boil, and then cooling it to 70–75 °C. To prevent microbial growth, preservatives were added to the cooled medium: 4.9 mL propionic acid (cat #8006052500, Merck-Sigma) and 10 mL of 10% Nipagene (Tegosept, cat #20-259, Genesee Scientific), prepared in absolute ethanol. The medium was then immediately dispensed into plastic vials and bottles.

Hemolymph extraction and RT-qPCR analysis

Hemolymph samples were extracted from adult transgenic Drosophila as previously described [30], with minor modifications. Briefly, 3–4 holes were punctured in the bottom of a 0.5 ml microcentrifuge tube, which was then placed inside a 1.5 ml microcentrifuge tube. Adult flies (3–4 days old) were anesthetized, and each fly’s thorax was punctured bilaterally using a 33-gauge tungsten needle to facilitate hemolymph release. Forty punctured flies were transferred to the perforated 0.5 ml tube, kept on ice, and centrifuged at 2000 rpm for 1 min at 4 °C. Following centrifugation, the 0.5 ml tube containing fly carcasses was discarded, and the collected hemolymph pellet was immediately stored at −20 °C until further analysis. A total of three independent biological replicates were collected for each experimental condition.

For gene expression analysis, hemolymph samples were processed using the Cells-to-CT 1-Step TaqMan Kit (cat #A25602, Thermo Fisher Scientific) according to the manufacturer’s instructions. A total of 1 × 10⁵ cells were lysed in Cells-to-CT lysis buffer supplemented with DNase I, and the cell lysates were analyzed using TaqMan Gene Expression Assays from Thermo Fisher Scientific: Contactin (Assay ID: Dm02136688_g1), Pvf1 (Dm01813949_m1), CDase (Dm02146507_g1), Pent/Magu (Dm01812540_m1), Robo2 (Dm01843954_m1), Gpdh1 (Dm01841185_m1), Rpl32 (Dm02151827_g1), Rps17 (Dm01822503_g1), and Act5c (Dm02361909_s1). All reactions were performed in technical duplicates using the QuantStudio 7 Flex Real-Time PCR System under standard cycling conditions. Gene expression levels were normalized to the geometric mean of four endogenous housekeeping genes (Gpdh1, Rpl32, Rps17, and Act5c), and the relative expression of target genes in the overexpression fly lines was calculated relative to EGFP control lines using the comparative Ct (ΔΔCt) method [31]. Analysis was performed on three independent biological replicates for each line.

Social interaction

We evaluated social interaction behavior by placing four Drosophila male flies (2–3 days old) into a 65 mm diameter circular arena, as previously described [32]. Briefly, flies were habituated at 25 °C for approximately 15 min prior to recording. Flies were recorded for 20 min using a FLIR camera with 30 fps at 25 °C. The locomotor trajectories of each of the four flies were quantified using the IowaFLI Tracker [33], which produced social interaction characteristics such as interaction duration and number of interactions.

Open field behavioral assay

Locomotor and exploratory behaviors were analyzed using an open field assay consisting of multiple 15 mm circular arenas. Flies were loaded individually into each arena after short cold anesthesia and acclimatized for at least 15 min at 25 °C prior to recording. Video-based locomotion tracking and analysis was performed using DART software [34]. Flies’ trajectories were captured and tracked at 15 fps over a 10-min duration. Flies were tracked, and speed and percent time in the periphery was analyzed in DART, as previously described [35, 36].

Visual startle and habituation assay

To examine non-associative learning, we followed the previously published light-switch on protocol [32]. Briefly, we shined a blue light at maximum intensity for 1 s in complete darkness to measure visual startle response, and repeated the blue light stimulus 5 times at 1-min intervals to measure fly habituation to repeated light stimuli. Flies were loaded into a 65 mm × 5 mm × 2 mm rectangular arena and tracked using DART. Flies were acclimatized in the dark for 15 min at 25 °C prior to behavioral recording. Each fly’s baseline activity was measured using locomotion traces as an active average speed during the first 40 s. Their startle response to blue light was assessed using an active average speed of 10 s before and after the first pulse and their habituation as the locomotor reactivity to the fifth pulse.

Sleep assay

Assessment of sleep behavior was performed by housing individual male Drosophila flies (2-3 days old) in Trikinetics monitor tubes (∅ 5 mm; length 65 mm) sealed with food on one end and an air-permeable cotton cap on the other. This system consisted of two separate arenas holding 20 tubes each, which were placed in an incubator for sleep assessment. Flies were allowed to recover from ice and habituated to the monitor tubes for approximately 15 h prior to starting the first light cycle. Video recordings of flies’ activity were recorded in avi format at 5 fps using DART software [35]. The activity of individual fly was monitored and quantified over a period of 72 h (3 days and 3 nights at 25 °C with 12:12 h of light: dark cycles), starting with the light cycle. Data of dead or non-moving flies were excluded from the analysis. For each fly, inactivity or immobility lasting more than five min was considered an instance of sleep [37]. Total sleep duration was calculated by averaging the 72-hour activity data into one hour activity. Flies’ activity during the light period was averaged into hourly sleep metrics, referred to as daytime sleep. Conversely, nighttime sleep was derived from activity during the dark cycle.

Statistical analysis

Data were tested for normality using the Shapiro–Wilk test. Demographic data were analyzed using the Mann–Whitney U test for continuous variables, and Pearson’s chi-square test or Fisher’s exact test, as appropriate, for categorical variables in SPSS (version 26.0). Correlation analyses were performed using Spearman’s correlation coefficients for continuous variables. CellProfiler quantification data were analyzed using two-way ANOVA followed by Tukey’s multiple comparisons test. MEA recording data were analyzed using the non-parametric Kruskal-Wallis test with Dunn’s post hoc test for multiple comparisons. Hemolymph RT-qPCR data were analyzed using unpaired t-test on ΔCt values. All these analyses were conducted in GraphPad Prism (version 10). Drosophila behavioral data were analyzed using a two-sided permutation t-test. Cliff’s Δ was used as the effect size measure for data with a nonparametric distribution. Effect size analyses and significance tests were conducted using estimationstats.com [38]. Data were considered statistically significant when the p-value was < 0.05 for two-group comparisons and the adjusted p-value was < 0.05 for multiple comparisons.

Results

Study design and demographic characteristics of the study population

This study aimed to investigate the correlation between circulating biomarkers in ASD and neuronal and behavioral phenotypes relevant to the disorder as illustrated in Fig. 1. Therefore, we investigated how the identified neuronal proteins in the circulating blood in ASD are reflected with altered neuronal phenotypes from the same subjects (Fig. 1). Additionally, we explored whether overexpression of these identified biomarkers in the blood alone was sufficient to induce ASD-relevant behaviours in Drosophila (Fig. 1).

Fig. 1. Schematic diagram summarizing the experimental design of the study.

We recruited a well-defined cohort of young children with autism spectrum disorder (ASD) and matched control subjects. Blood samples were collected and separated into plasma and PBMCs. Plasma samples were used for biomarker discovery using the Olink platform, while PBMC samples from the same selected subjects were reprogrammed into iPSCs and differentiated into cortical neurons. This approach enabled us to investigate how altered circulating biomarkers in ASD are reflected in neuronal phenotypes in vitro, as well as in ASD-related behaviors in Drosophila.

In this study, we focused on Arab children aged 2–4 years who were sex-matched between the study groups and met the eligibility criteria detailed in the Methods section. This approach was designed to minimize potential confounding factors, such as differences in age, ethnicity, and other comorbid conditions, which might affect the results. We also ensured that all enrolled subjects were born in Qatar, that their mothers had resided in Qatar for most of their pregnancies, and that the subjects continued to live primarily in Qatar up to the time of study participation. These criteria were established to control for environmental influences during both prenatal and postnatal development.

The study population comprised of children with ASD (n = 100) and control (n = 60) subjects. The study population showed matching demographic characteristics, in terms of age, sex, and nationality. The age of children with ASD [median (interquartile range, IQR)] was 3.37 (2.99–3.70) years, which was comparable to the age of the control subjects 3.56 (2.98–3.81) years (p = 0.376). Additionally, the two study groups were sex-matched at an almost 4:1 male-to-female ratio, 79/21 in ASD cases vs. 48/12 in the control group (p = 0.880). Furthermore, the nationality distribution was not significantly different between the two study groups, with almost comparable percentages of Qatari (27% vs. 28.3%), Egyptian (16% vs. 20%), Syrian (13% vs. 20%), Yemeni (17% vs. 10%), Sudanese (12% vs. 13.3%), Jordanian (3% vs. 3.3%), and other Arab citizens (12% vs. 5%) in the ASD vs. control groups (p = 0.563). Further details about the clinical characteristics of the study population and ADOS-2 severity scores of the subjects with ASD are described in our previous publications [7, 9].

High-throughput screening reveals proteomic heterogeneity within the Arab population

We analyzed the expression of 984 protein targets in circulating plasma samples from individuals with ASD compared to the control group (FDR < 0.05). A total of 22 differentially expressed proteins (DEPs) were identified in the ASD group vs. controls across the entire study population (Fig. 2A, B). We then examined whether nationality and ASD severity classification were confounding factors influencing the results in the whole cohort. After adjusting for these variables, the total number of DEPs increased from 22 to 33 in the ASD vs. control group (Fig. 2B, C). The full list of these DEPs before and after correction is included in Table S1. These results suggest that both nationality and severity classification are likely confounding factors affecting the results and should be considered in future ASD studies.

Fig. 2. High-throughput proteomic screening reveals heterogeneity in the protein profiles of the ASD and control groups within the Arab population.

(A) Pie charts showing matched nationality distributions between the ASD and control (CTRL) groups. (B) Volcano plots demonstrating a higher number of differentially expressed proteins (DEPs) in the ASD vs. CTRL group after correction for nationality and ASD severity classifications in the whole study cohort. (C) Total numbers of DEPs were 22 before correction and 33 after correction for the nationality and ASD severity classifications. (D) Nationality-specific comparisons revealed heterogeneity in DEPs within the study cohort. (E) Bar plot demonstrating a total of 109 downregulated and 448 upregulated proteins in the Qatari ASD vs. Qatari CTRL group. (F) Venn diagram summarizing the number of unique and common markers in the ASD vs. CTRL group between the Qatari samples and the whole cohort.

Next, we performed separate analyses of each nationality within each study group in comparison to the rest of the cohort from the same group. Remarkably, we identified a total of 524 DEPs in the Qatari control group compared to other control samples (Fig. 2D and Table S1). Additionally, we found a total of 16 DEPs in the Syrian control group vs. the rest of control group, and 82 DEPs in the Sudanese ASD vs. rest of ASD cohort (Fig. 2D and Table S1). These findings indicate proteomic heterogeneity within the Arab population. Accordingly, we focused subsequent analyses on the Qatari nationality, as it exhibited the most profound differences compared to other nationalities. Within the Qatari group, we found a total of 557 DEPs in ASD vs. control group, including 537 unique and 20 common markers when compared to the rest of the cohort (Fig. 2E, F and Table S1).

Correlation of several circulating proteins with ASD severity

We then sought to identify circulating proteins that were significantly associated with the clinical severity of ASD symptoms. We first adjusted for nationality and severity classification as covariates and identified 299 DEPs between the ASD and control groups across the whole cohort (FDR < 0.05; Table S2). No fold-change filter was applied, in order to include the maximum number of proteins in the correlation analysis with ASD severity. Interestingly, we found 20 DEPs that were significantly correlated with ASD severity (p-value < 0.05; Table S2). This included two proteins that were negatively correlated–ILKAP and WASF3–and 18 proteins that were positively correlated: AMIGO2, CDH2, CNTNAP2, CDH15, TAFA5, SEZ6L2, CNTN5, GFRA3, ITGA11, SORCS2, OMG, CNTN2, ENTPD6, HPGDS, ULBP2, PLXDC1, ENAH, and TACSTD2 (Table S2).

Upregulation of circulating neuronal proteins in young children with severe ASD

In order to investigate how changes in circulating proteomic profiles in ASD are associated with neuronal and behavioral abnormalities, we randomly selected three Qatari male children from each study group to identify the DEPs between them. This was followed by iPSC generation and neuronal characterization from the same selected subjects (Fig. 3A). We identified 14 DEPs that were significantly upregulated in the severe ASD group compared to the control group, with the mild ASD group showing intermediate expression levels, as illustrated in the heatmap and principal component analysis (Fig. 3B, C). Among these DEPs in the Qatari ASD group, three proteins–SMOC2, CDH15, and PCDH17–and two proteins–CDH15 and CNTN5–remained consistent with the results from the whole cohort in ASD vs. controls and ASD severity correlation, respectively, as listed in Tables S1 and S2. Notably, CNTN5 and ROBO2 are considered strong candidate genes for ASD, while CDON shows suggestive evidence, and SLITRK2 is classified as a syndromic gene, according to the SFARI Gene database.

Fig. 3. Neuronal proteins are upregulated in the circulating blood of young children with severe ASD.

(A) Schematic diagram of the overall study design, with a focus here on biomarker discovery, in which we randomly selected Qatari male children with and without ASD (n = 3 per group). (B) Heatmap demonstrating 14 differentially expressed proteins (DEPs), particularly in the severe ASD vs. control (CTRL) group. (C) Principal component (PC) analysis shows a clear separation between the severe ASD vs. CTRL group. (D) Network analysis reveals protein-protein interactions between several DEPs. (E) Pathway enrichment analysis demonstrates that the DEPs in the circulating blood in ASD are involved in neuronal functions based on the gene ontology (GO) database for biological process.

We then performed protein-protein interaction analysis of the 14 DEPs using the iDEP platform based on STRING database. This analysis revealed significant network interactions among several of these DEPs (enrichment p-value = 5.1e-05; Fig. 3D). We also conducted pathway enrichment analysis using ShinyGO (v. 0.80) to explore the functional roles of the DEPs. We analyzed the data using three gene ontology (GO) databases–Biological Process, Cellular Component, and Molecular functions–each ranked by fold enrichment (Table S3). Interestingly, these 14 DEPs are involved in key neuronal functions, including synapse assembly, cell adhesion, and neurodevelopment, according to the GO Biological Process database (Fig. 3E and Table S3). Additional enrichment was observed in pathways related to GABAergic synapse function, axolemma, and calcium ion binding (Table S3). These findings highlight the neuronal functions of these circulating proteins and suggest their potential relevance to ASD pathophysiology.

Developmental defect in neuronal maturation in ASD

We then sought to determine the neuronal characteristics associated with subjects exhibiting elevated levels of neuronal proteins in circulating blood. Hence, we generated iPSC lines from the PBMC samples of the same individuals included in the previously described proteomic analysis, followed by directed differentiation into cortical neurons (Fig. 4A). RNA sequencing analysis was then performed on both immature (week 4) and mature (week 8) neurons to assess neurodevelopmental differences between the study groups. Unexpectedly, we identified 1863 differentially expressed genes (DEGs) in the control group during neuronal maturation, whereas only 8 and 13 DEGs were detected in the mild and severe ASD groups, respectively (FDR < 0.05; Fig. 4B, C). The full list of DEGs across the three groups is included in Table S4. We then performed pathway enrichment analysis to determine the functions of these DEGs. Notably, the control group exhibited significant downregulation of pathways related to mitosis and cell cycle processes–changes that were not observed in the ASD groups (Fig. 4D and Table S5). These results suggest a normal maturation process in control neurons from week 4 to week 8 that appears to be disrupted in ASD neurons. In contrast, the severe ASD group showed significant upregulation of pathways involved in axon extension, nervous system processes, and nerve impulse transmission, indicating a potential presence of aberrant neuronal activity in this group (Fig. 4E and Table S5).

Fig. 4. Developmental defect in neuronal maturation in ASD.

(A) Schematic diagram of the overall study design, with a focus here on neuronal characterization (B) RNA-seq analysis reveals hundreds of differentially expressed genes (DEGs) in mature (week 8) vs. immature (week 4) neurons derived from control (CTRL), but not ASD samples. (C) Heatmaps illustrating a higher number of DEGs in week 8 (W8) compared to week 4 (W4) neurons generated from the CTRL group, as opposed to the severe ASD group. (D) Pathway enrichment analysis shows normal downregulation of cell cycle processes in maturing CTRL neurons at W8 compared to W4. (E) Mature neurons in the severe ASD group exhibit several dysregulated pathways, particularly those involved in neuronal functions.

Remarkably, we further validated and extended these observations by quantifying Ki67⁺, DCX⁺, and MAP2⁺ cells across the three groups. While Ki67 labels actively proliferating cells, both DCX and MAP2 are neuron-specific markers that reflect distinct stages of neuronal maturation and might be indirectly linked to neuronal excitability. Our findings demonstrated that the control group showed a significant reduction in Ki67⁺ cells from week 4 to week 8 (p < 0.001; Fig. 5A), while DCX⁺ and MAP2⁺ counts remained unchanged over time (Fig. 5B, C). Meanwhile, mild ASD group exhibited significantly fewer Ki67⁺ cells in week 4 compared to controls (p < 0.01; Fig. 5A), along with a significant decline in DCX⁺ and MAP2⁺ cells by week 8 (p < 0.05; Fig. 5B, C). In contrast, severe ASD neurons displayed a significant increase in DCX⁺ and MAP2⁺ cells at week 8 compared to week 4 (p < 0.05; Fig. 5B, C), and significantly higher numbers of DCX⁺ (p < 0.001) and MAP2⁺ (p < 0.01) cells at week 8 relative to both control and mild ASD groups (Fig. 5B, C). Taken together, these findings reveal distinct neurodevelopmental trajectories in ASD subtypes and suggest abnormal neuronal maturation and activity, particularly in the severe ASD group, which also exhibited elevated levels of neuronal proteins in the blood.

Fig. 5. Distinct patterns of cell proliferation and neuronal characteristics across ASD subtypes.

(A) Cell counts of proliferating Ki67⁺ cells are significantly reduced in the control (CTRL) group at week 8 (W8) compared to W4 and are also reduced in the mild ASD group vs. CTRL group at W4. (B–C) Severe ASD neurons show a significant increase in DCX⁺ and MAP2⁺ cells at W8 relative to W4, as well as higher numbers at W8 compared to both CTRL and mild ASD groups. In contrast, mild ASD neurons exhibit a significant decrease in DCX⁺ and MAP2⁺ cells by W8. Representative immunofluorescent images are shown for each marker and group at both W4 and W8. Data are presented as individual values with mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001. Scale bar: 50 µm.

Abnormal synaptic function and neuronal hyperexcitability in severe ASD

We performed RNA analysis to investigate changes in transcriptomic profiles in cortical neurons across the different study groups. We detected a total of 54 DEGs in the severe ASD compared to the control group, whereas only 5 DEGs were found between the mild ASD and control group, and 9 DEGs in the severe vs. mild ASD group (FDR < 0.05; Fig. 6A, B; Table S6). Interestingly, two markers (FKBP5 and CTSZ) showed consistent dysregulation in Qatari individuals across two datasets: plasma protein expression in ASD compared to controls (Table S1) and cortical neuron RNA expression in severe ASD vs. controls and severe vs. mild ASD groups, respectively (Table S6). Enrichment analysis of the total DEGs in the severe ASD vs. control group revealed several upregulated pathways, including those involved in cilium movement and axoneme assembly, which play critical roles in neuronal excitability and function (Fig. 6C). Next, we used MEA recordings to assess neuronal firing activity and network formation in cortical neurons from the three study groups. Remarkably, the severe ASD group exhibited higher neuronal activity compared to both the mild ASD and control groups, as shown in the raster plots (Fig. 6D). In particular, the severe ASD group showed significantly increased mean firing rate and network burst frequency in comparison to the other two groups (FDR < 0.05; Fig. 6E, F). These results support the presence of neuronal hyperexcitability in cortical neurons derived from individuals with severe ASD, which is also reflected in a functionally related gene expression signature in these neurons and in upregulated levels of circulating neuronal proteins in these individuals.

Fig. 6. Abnormal synaptic function and neuronal hyperexcitability in cortical neurons derived from individuals with severe ASD.

(A) RNA-seq analysis shows a greater number of differentially expressed genes (DEGs) in severe ASD compared to the control (CTRL) group. (B) Heatmap illustrates DEGs between severe ASD and CTRL. (C) Enrichment analysis of DEGs in severe ASD vs. CTRL identifies multiple upregulated pathways associated with neuronal activity and excitability. (D) Assessment of neuronal activity and network dynamics using microelectrode arrays (MEA). Representative raster plots of mature cortical neurons derived from CTRL, mild, and severe ASD. Colored bars represent detected network bursts. (E) Quantification of weighted mean firing rate (Hz) and (F) network burst frequency (Hz). Cortical neurons derived from individuals with severe ASD show significantly increased firing rates and network burst frequencies compared to neurons derived from the mild ASD and CTRL groups. Data are presented as individual values with mean ± SEM; *P < 0.05, ***P < 0.0001.

Selective overexpression of ASD-associated neuronal proteins in Drosophila hemocytes induces ASD-relevant behaviors

We investigated whether upregulation of neuronal proteins in the blood alone could induce ASD-relevant behaviors in Drosophila. From the list of 14 DEPs identified in the severe ASD group, we selected five candidates based on DIOPT scores > 0.2 and the availability of transgenic lines: CNTN5 (Drosophila ortholog: Contactin), ROBO2 (Robo2), ASAH2 (CDase), SMOC2 (Pent), and VEGFD (Pvf1), as listed in Fig. S1. These genes were constitutively overexpressed in Drosophila hemocytes using the HML-GAL4/UAS-cDNA system (Fig. S1). We validated overexpression of the control EGFP in the hemocytes of third instar Drosophila larvae, with expression persisting into adulthood (Fig. S1). In addition, RT-qPCR analysis confirmed successful overexpression of the genes of interest in hemocytes from adult Drosophila overexpression lines compared to the EGFP control line (Fig. S1).

We then assayed for ASD-associated behaviors in these transgenic Drosophila models, including social interaction, locomotion, anxiety, habituation learning, and sleep (Fig. 7A, B, E; Fig. 8A). Flies overexpressing Contactin and Robo2 displayed reduced social interaction time and frequency (Fig. 7C, D). In contrast, CDase and Pent overexpression resulted in increased interaction time, while Pvf1 uniquely exhibited reduced interaction frequency (Fig. 7C, D). Meanwhile, Contactin-, CDase-, Pent-, and Pvf1-overexpressing flies demonstrated reduced open-field locomotion (Fig. 7F). Additionally, Contactin, Robo2, CDase, and Pent flies exhibited increased wall-following activity, an anxiety-like behavior, whereas Pvf1 flies showed decreased activity (Fig. 7G). Although Contactin- and Robo2-overexpressing flies exhibited elevated baseline locomotion in darkness, CDase, Pent, and Pvf1 lines showed reduced activity (Fig. 8B). Startle responses to visual stimuli were normal across all models, indicating intact baseline sensory systems (Fig. 8C). However, Contactin and Robo2 flies displayed increased post-stimulus activity (Fig. 8D). Meanwhile, Pvf1 exhibited reduced post-stimulus activity (Fig. 8D). Finally, sleep pattern analyses revealed reduced total and daytime sleep in Contactin and Pvf1 flies, whereas Robo2, CDase, and Pent lines showed increased sleep duration (Fig. 8E–G). Although nighttime sleep followed similar trends, differences in the Robo2 and CDase models were not statistically significant (Fig. 8H). A summary of these results is provided in Table S7. Taken together, our findings demonstrate that overexpression of ASD-related genes specifically in Drosophila hemocytes is sufficient to elicit a range of behavioral phenotypes relevant to ASD.

Fig. 7. Transgenic flies with hemocyte-specific overexpression of ASD candidate genes exhibit impaired social interactions and anxiety-like behaviors.

(A) Schematic diagram of the overall study design, with a focus here on behavioral assessment. (B) Social interaction assay in male flies. (C) Interaction time was reduced in Contactin and Robo2 lines, whereas CDase and Pent lines showed increased interaction time compared to the control EGFP line. (D) Frequency of interactions was reduced in Contactin, Robo2, and Pvf1 flies, whereas Pent exhibited increased frequency of interactions. (E) Schematic diagram of the open-field assay. (F) Average active speed in four transgenic lines, namely Contactin, CDase, Pent, and Pvf1, was reduced compared to control line. (G) Anxiety-like behavior was increased in Contactin, Robo2, CDase, and Pent flies as indicated by increased thigmotaxis, while Pvf1 flies showed decreased thigmotaxis relative to the control EGFP line.

Fig. 8. Drosophila models demonstrated habituation learning deficits and disrupted sleep patterns.

(A) Schematic diagram of the habituation assay protocol, which included a 10-min dark acclimation followed by five 1-second blue light pulses (1-min intervals). (B) Baseline locomotion in darkness showed that Contactin and Robo2 flies had elevated activity, while CDase, Pent, and Pvf1 flies exhibited reduced activity vs. control EGFP line. (C) Startle response was unchanged across genotypes. (D) Post-habituation locomotion (after fifth pulse) revealed that Contactin and Robo2 lines displayed reduced habituation (increased speed), while Pvf1 showed enhanced habituation. Both CDase and Pent flies matched control rates. (E) Representative sleep profiles over three days (white) and nights (grey). (F–H) Total sleep during both daytime and nighttime was reduced in Contactin and Pvf1 lines, while Robo2, CDase, and Pent lines exhibited increased sleep compared to the control EGFP line.

Discussion

Previous studies have identified several altered biomarkers in the circulating blood of individuals with ASD; however, the functional contributions of these proteins to ASD pathophysiology remain unclear. In this study, we screened for hundreds of proteins in plasma samples from young Arab children with ASD compared to controls and observed heterogeneity in proteomic profiles across different Arab nationalities. Accordingly, we focused our subsequent analyses on Qatari male children and randomly selected samples from the control, mild, and severe ASD stratified study groups. Interestingly, we identified 14 DEPs in the circulating blood that are involved in neuronal function and correlate with the clinical severity of ASD symptoms. To investigate their functional relevance, we generated iPSC-derived cortical neurons from the same individuals in the three study groups. Remarkably, we found that elevated levels of circulating proteins in the severe ASD group were reflected with increased neuronal cell counts and hyperexcitability in their derived neurons. Additionally, we found that overexpression of these proteins in the blood alone was sufficient to induce ASD-relevant behaviors in Drosophila models. To the best of our knowledge, this is the first study to demonstrate that circulating neuronal biomarkers elevated in severe ASD may contribute to ASD pathophysiology and, therefore, could represent potential therapeutic targets for alleviating ASD symptoms, which should be further investigated in future studies.

In this study, we focused on young Arab children with ASD residing in Qatar to minimize potential effects of ethnic variation on the results. Notably, we identified heterogeneity in the proteomic profiles across different Arab subpopulations, particularly among the Qatari, Syrian, and Sudanese groups. Although we stratified each Arab subpopulation based on nationality, multiple population‑scale genomic studies across the Arab region have demonstrated that each nationality often represents a distinct genetic structure with shared regional ancestry, as supported by previous studies on Qatari, Egyptian, Sudanese, and Levant nationals [39–43]. These observations suggest that the nationality‑based proteomic differences observed in our study may be influenced by underlying population‑specific genetic variation, which warrants further validation in future studies. Furthermore, prior research has revealed novel candidate genes associated with ASD in Qatari and other Middle Eastern populations [44–46]. These findings highlight the importance of accounting for inherent proteomic differences among Arab populations in ASD studies. Therefore, we focused our analyses on Qatari subjects and identified upregulated levels of 14 neuronal proteins in the circulating blood of individuals with severe ASD. Remarkably, four of these proteins–CNTN5, ROBO2, CDON, and SLITRK2–are associated with ASD according to the SFARI Gene database. However, the relationship between these genetic variants and protein expression, as well as the functional implications of altered circulating neuronal proteins in ASD, remains unclear. We propose that elevated levels of circulating neuronal proteins may reflect increased neuronal activity in the brains of individuals with severe ASD. Collectively, our findings reveal proteomic heterogeneity within Arab populations and suggest a potential role for altered circulating neuronal proteins in the pathophysiology of ASD.

We then explored whether elevated levels of neuronal proteins in the circulating blood of individuals with severe ASD might serve as a window into the abnormal neuronal processes occurring in the brain. Importantly, patient-specific iPSCs retain some epigenetic memory unique to the individual, originally shaped by their in vivo environment, including exposure to systemic factors [47, 48]. Therefore, iPSC differentiation in vitro may reflect patient-specific phenotypes influenced by circulating molecules present in vivo [47, 48]. To investigate this notion, we generated iPSC-derived cortical neurons from individuals with severe ASD who exhibited elevated levels of neuronal proteins in their blood, as well as from individuals with mild ASD and control subjects. We identified hundreds of DEGs during the maturation of the derived neurons from control subjects, but not from those with ASD. The majority of these DEGs in control neurons were associated with downregulated cell cycle processes, reflecting normal neuronal maturation, consistent with previous reports [49–51]. In contrast, neurons derived from individuals with ASD showed a limited number of DEGs during maturation, further supporting the notion of a neurodevelopmental defect in ASD [52, 53]. We further validated these findings through cell quantitative analyses, which revealed divergent neurodevelopmental trajectories across ASD subtypes. A significant increase in MAP2⁺ cells, along with a persistent increase in DCX⁺ cells in severe ASD neurons at week 8, may reflect prolonged neurogenesis or delayed maturation. This is consistent with previous reports of sustained progenitor proliferation and delayed neuronal differentiation in ASD iPSC models, particularly those associated with macrocephaly [54–56]. Conversely, the reduced Ki67⁺ counts and declining DCX⁺/MAP2⁺ populations observed in mild ASD neurons may indicate premature cell-cycle exit or limited neurogenic potential, aligning with previous observations of early maturation and reduced neuronal output in certain ASD subtypes [57, 58]. Together, these findings support a typical maturation trajectory in control neurons and highlight altered neurodevelopmental phenotypes in ASD-derived neurons.

Furthermore, we analyzed the transcriptomic profiles and functional activity of neurons derived from control subjects and ASD subgroups. We found that neurons from the severe ASD group exhibited upregulated expression of several genes involved in neuronal activity. These included genes related to cilium movement and axoneme assembly, which play critical roles in regulating synaptic signalling, neuronal excitability, and connectivity across neural networks [59–61]. Recent evidence supports a strong link between ASD and altered expression of cilium and axoneme related genes and function [54, 62–64]. Additionally, we observed dysregulated expression of two markers (FKBP5 and CTSZ) across both plasma proteomics and neuronal RNA data, and these markers have been previously associated with ASD [65, 66]. In particular, altered FKBP5 expression affects neuronal excitability and synaptic plasticity, whereas CTSZ modulates neuronal function indirectly by regulating neuroinflammatory pathways [67–69]. To further characterize ASD neurons, we performed MEA recordings to simultaneously assess neuronal activity and network formation. Remarkably, we observed increased neuronal hyperexcitability specifically in the severe ASD group, suggesting enhanced synaptic transmission and network activity, as previously reported [25, 70]. Overall, these findings suggest that elevated levels of circulating neuronal proteins may represent functionally relevant biomarkers that reflect elevated neuronal activity in the brains of individuals with severe ASD.

We then sought to further demonstrate the functional relevance of altered circulating proteins to ASD by investigating whether elevated levels of neuronal proteins in the blood alone could induce ASD-relevant behaviors. We used Drosophila, a well-established genetic model for studying ASD phenotypes [71, 72]. Five selected candidates were overexpressed in Drosophila hemocytes, and ASD-related behaviors were subsequently assessed. Both Contactin- and Robo2-overexpressing flies displayed reduced social interactions, mirroring the core ASD symptom of social withdrawal [2]. Unexpectedly, CDase and Pent models showed increased social interactions, which might reflect inappropriate social judgement observed in some ASD subtypes [73, 74]. Meanwhile, flies overexpressing Contactin, CDase, Pent, and Pvf1 exhibited reduced open-field locomotion, resembling hypoactivity and depressive-like behaviors seen in some individuals with ASD, as well as in mouse models [2, 75–77]. Anxiety-like behavior was also elevated in Contactin, Robo2, CDase, and Pent models, aligning with the increased prevalence of anxiety in ASD [78, 79]. Furthermore, both Contactin and Robo2 models exhibited impaired habituation learning, suggesting learning difficulties or heightened sensory sensitivity commonly associated with ASD [2]. Finally, all models showed disrupted daytime sleep patterns, consistent with the sleep disturbances frequently observed in ASD [2, 80, 81]. Together, these findings demonstrate that overexpressing neuronal proteins in Drosophila hemocytes alone is sufficient to elicit ASD-related behaviors. They also highlight the behavioral heterogeneity observed across different genetic perturbations, reflecting the clinical diversity of symptom presentation in individuals with ASD.

Although we overexpressed these genes in Drosophila hemocytes, we predict that this may result in increased expression and release of the corresponding proteins into the circulating hemolymph (analogous to human blood), which may then infiltrate the brain and affect neuronal function and behavior in these Drosophila models. While Drosophila possesses a blood-brain barrier (BBB) that restricts circulating hemocytes from accessing the brain directly, soluble proteins may still cross the BBB via passive diffusion or active transport mechanisms [82]. Future studies are needed to determine whether our proteins of interest are capable of crossing the BBB in Drosophila or other animal models, such as rodents, to further elucidate their mechanisms of action and potential molecular contributions to ASD pathophysiology.

Our study has several strengths. We recruited a well-defined cohort of young children with ASD and matched control subjects with comparable demographic characteristics, including age, sex, nationality, and other factors, as described in our previous publications [7, 9]. Specifically, we included only subjects of Arab ethnicity within a narrow age range of 2–4 years and with matched sex distribution between the study groups to minimize the influence of these variables on the results. Additionally, we recruited only children whose mothers had resided in Qatar for most of their pregnancies and continued to live in Qatar after birth. These criteria aimed to ensure that participants in both study groups were exposed to similar environmental factors, as variations in such factors may affect the findings. Although we focused on Arab children, we observed notable proteomic heterogeneity across Arab subpopulations. In particular, Qatari children exhibited the most profound proteomic differences compared to other Arab subpopulations. Accordingly, we focused our subsequent analyses on Qatari male children to have a more homogeneous study population and minimize the effects of demographic variation on the results. Through this approach, we identified novel mechanisms linking altered proteomic profiles to neuronal and behavioral deficits associated with ASD. These findings underscore the importance of precision medicine approaches in studying ASD mechanisms, which could ultimately lead to identifying population-specific therapeutic targets.

Nevertheless, our study has some limitations. Although we identified a subset of proteins elevated in a small cohort of individuals with ASD, these findings need to be replicated in much larger Arab populations to confirm population-specific differences. While our proteomics, neuronal phenotyping, and Drosophila behavioral assessments were based on the Qatari group, future studies should include analyses of other non-Qatari Arab populations to determine whether there are shared or distinct features across Arab subgroups. We observed limited overlap between plasma proteomics and neuronal RNA data, which likely reflects methodological differences, as RNA-seq captures a broad range of expressed genes while the proteomics panel is limited to predefined targets. Future studies should apply the same analytical methods across sample types to enable direct comparison. Although FKBP5 and CTSZ overlapped between the plasma proteomics and neuronal RNA datasets in ASD, they were not identified during initial study design, which included 14 plasma proteomics candidates for further analysis. They were instead identified after the completion of neuronal characterization and Drosophila behavioral assays. Nevertheless, FKBP5 and CTSZ remain interesting candidates for future ASD-related behavioral studies.

Additionally, it remains unclear whether the observed changes in protein expression levels contribute to ASD pathophysiology or are a consequence of the disorder. ASD often co-occurs with other conditions, including intellectual disability, anxiety, and attention-deficit/hyperactivity disorder. However, these comorbid conditions are reliably diagnosed after the age of four years, which is beyond the age range of our study cohort [83–85]. Future longitudinal studies are needed to capture dynamic changes in proteomic profiles before and after the onset of ASD symptoms. Such approaches may clarify whether these proteomic changes are causative or an outcome of ASD and help investigate their correlation with co-occurring conditions and clinical phenotypes in ASD. Furthermore, the cellular origin of the altered circulating neuronal proteins in ASD remains unclear–whether they are derived from the peripheral or central nervous system, or even from non-neuronal tissues–and needs to be investigated in future studies. Additional factors such as environmental exposures, genetic predispositions, and dietary habits may also influence protein expression and should be explored to better understand their contributions to the observed proteomic profiles in ASD. While we employed correlative analyses, our use of Drosophila models exhibiting ASD-related behaviors suggests a potential role for the identified proteins in ASD pathophysiology. However, these behaviors varied across different genetic models. Therefore, future studies should include genetic analyses in individuals with ASD to establish a more direct link between specific genes, their downstream protein expression, and ASD behaviors.

In conclusion, our study demonstrates the presence of proteomic heterogeneity within the Arab population and underscores the need for precision medicine approaches in identifying biologically relevant markers in ASD. We found elevated levels of 14 neuronal proteins in the circulating blood of Qatari children with severe ASD. Functional studies linked these biomarkers to increased neuronal excitability in patient-derived iPSC cortical neurons. Furthermore, we established multiple novel Drosophila models exhibiting distinct behavioral profiles relevant to ASD as a result of gene overexpression in hemocytes alone. These models not only recapitulate key aspects of ASD-related behaviors but also reveal behavioral diversity linked to specific genes. Overall, our findings indicate that the identified circulating neuronal biomarkers are associated with ASD severity and may provide insights into its underlying pathophysiology. Further studies are needed to evaluate their potential as therapeutic targets for mitigating symptom severity in affected individuals.

Author contributions

A.R.A.-S conceived the study. S.M.L. contributed to the recruitment of study participants and the collection of demographic data. S.M.L. also performed cell culture work, RNA extractions, correlation analysis, immunostaining, confocal imaging, CellProfiler analysis, and RT-qPCR experiments. W.N.-E provided some assistance with cell culture work. K.C.S. performed MEA recordings and data analysis. S.S., S.H., and S.B.M.G. conducted the Drosophila behavioral assessments and data analysis. A.M. performed hemolymph extractions. K.O. performed RNA-seq and initial data analysis. I.B. performed the Olink experiments. A.R.A.-S analyzed the Olink, RNA-seq, and RT-qPCR data. A.R.A.-S, F.M., Y.P., and H.B.A. supervised and managed the work. S.M.L. and A.R.A.-S wrote the draft, and A.R.A.-S finalized the manuscript. All authors have read and approved the manuscript.

Funding

This work was supported by a research grant from Hamad Bin Khalifa University’s Office of the Vice President for Research (project code HBKU-INT-VPR-TG-02-07) awarded to A.R.A.-S and F.M. It was also supported by an ECRA award from the Qatar National Research Fund (project code ECRA01-001-3-001) and a research grant from the Qatar Biomedical Research Institute (project code VR03) to A.R.A.-S. The authors are solely responsible for the findings of this study.

Data availability

The raw datasets from this study will be available upon reasonable request to the corresponding author.

Competing interests

The authors declare no competing interests.

Ethics statement

This study was reviewed and approved by the Institutional Review Boards (IRBs) of the Qatar Biomedical Research Institute, Hamad Medical Corporation (HMC), and Primary Health Care Corporation (PHCC) in Doha, Qatar.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04261-6.