Distinct protein profiles in cord blood plasma of children with autism spectrum disorder Authors: Lital Kalich Philosoph, Chen Yoffe, Bat El Biri, Polina Baryakh, Efrat Glick Saar, Margarita Mayorov, Omer Bar Yosef Published: 2025 ABSTRACT AbstractAutism spectrum disorder (ASD) has a genetic origin in approximately 30% of cases, while the remaining causes are primarily linked to epigenetic and non-genetic factors. This study aims to identify potential non-genetic causes contributing to ASD by investigating the underlying cellular mechanisms through proteomic analysis. Extracellular vesicles (EVs) mediate cellular communication and are linked to brain development disorders. Here, we utilize mass-spectrometry‐based proteomic analysis of EVs derived from umbilical cord blood plasma collected from 30 children diagnosed with non-syndromic ASD and 30 neurotypical controls. The analysis identified 565 proteins with significantly different expression levels, most of which were more abundant in the ASD group. Notably, the protein expression variances were markedly lower in the ASD group, suggesting a similar cellular activity. Differentially expressed proteins were clustered using String software into 3 groups: mitochondria, endoplasmic reticulum (ER), and a mixture of immune and cytoskeletal proteins. Further statistical analysis identified 11 strongly predictive ASD proteins, while the Human Protein Atlas recognized an additional 13 brain-specific proteins. A portion of these 24 proteins was associated with synaptogenesis and myelination. These findings suggest that, despite diverse etiologies, ASD may converge on a common final pathway involving mitochondrial and ER dysfunction, resulting in abnormal synaptogenesis. This study presents the first unbiased proteomic analysis of exosomal proteins aimed at determining whether neurodevelopmental disruptions linked to ASD originate prenatally or postnatally.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-29420-5.Keywords: Autism spectrum disorder, Extracellular vesicles, Umbilical cord blood, Protein expression, Neurogenesis, Synaptogenesis, Microglia, Neuro-inflammationSubject terms: Cell biology, Developmental biology, Neuroscience, Biomarkers, Neurology Introduction Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by deficits in social communication, repetitive behaviors, restricted interests, and abnormal sensory processing1. ASD is highly heritable, with 60–90% heritability in monozygotic twins2. Approximately 2% of ASD is associated with known genetic syndromes, with many of these variants involved in encoding neuronal and transcriptional regulation3,4. The genetic variants could be either inherited or de novo4–6. Inherited variants are divided by their occurrence rate (polygenic load of common variants and rare variants) and by size, ranging from single-nucleotide to larger deletions or duplications, known as copy number variations (CNVs)7. Approximately 1500 genes are associated with ASD etiology3. Nevertheless, only 17.5% of the variability observed in ASD susceptibility can be attributed to identifiable genetic influences3,4,7. A useful approach to investigating ASD etiology is to focus on the critical stages of brain development and to examine biomarkers during these periods. Fetal brain development is particularly important, as the essential processes of neural maturation are predominantly completed by the end of gestation. Consequently, any harmful exposure to the brain during this timeframe may result in neurodevelopmental deficits, including ASD. Moreover, genes associated with transcriptional regulation during pregnancy constitute the largest share of rare de novo variants associated with ASD4. Based on the findings that most brain imaging studies do not present brain malformation, and post-mortem studies of individuals with ASD have revealed abnormalities in brain microstructure8, suggest that the period of synaptogenesis and myelination towards the end of the pregnancy is particularly vulnerable8. Testing for biomarkers of abnormal brain development during this period would enhance the likelihood of finding such elements. Studies examining late-gestational development are mainly based on cord blood sampling9. Epigenetic studies of DNA methylation have been conducted on postmortem brain tissue and peripheral blood samples, revealing sex-specific patterns in ASD10–12. One study examined sex-specific epigenetic signatures in cord blood and found distinct differential methylation patterns in males and females with ASD compared to controls. Seven specific loci distinguish males and 31 females with ASD. At the level of cis-regulatory regions, 537 DNA methylated regions (DMRs) were identified in males and 1762 in females. The genes likely modify transcription factors associated with brain development, suggesting that sex-specific epigenetic dysregulation occurs during early pregnancy, rather than at the sampling time13. To further investigate the functional consequences of these epigenetic changes, transcriptome analysis of cord blood from the same cohorts was performed. While no differentially expressed transcripts were identified using an FDR q-value < 0.0513, a more permissive statistical threshold revealed 172 transcripts with altered expression. These transcripts were involved in the response to toxic substances and epigenetic regulation, potentially reflecting the impact of prenatal exposure to endocrine-disrupting chemicals during critical windows of transcriptional regulation14. Further analysis of gene expression found that expression of superoxide dismutase 2 (SOD2), retinoic acid–related orphan receptor alpha (RORA), and G protein–coupled estrogen receptor-1 (GPER) was significantly reduced, while progesterone receptor (PGR) expression was elevated15. These findings suggest that early dysregulation of oxidative stress responses, nuclear receptor activity, and hormone signaling pathways is involved in ASD pathogenesis. However, further investigation is required to connect these transcriptomic changes to specific physiological mechanisms associated with neurodevelopment. One challenge in interpreting the effect of the genetic, epigenetic, and transcriptomic variations lies in their position at the upstream of complex biological cascades that ultimately influence proteins and biochemical pathways. These interrelated cascades are highly responsive to environmental factors. Thus, a different approach to studying these network effects is by analyzing metabolites, which represent the final outcome of such processes. Metabolite studies in cord blood demonstrated changes in the lipid and amino acid pathways. One study found changes in the metabolites of polyunsaturated fatty acids, which are associated with an inflammatory effect16. Another study reported elevated levels of homocysteine and a reduced level of L-Isoleucine, suggesting disruptions in methylation and amino acid metabolism17. The functional consequences of the metabolic alterations still require further investigation. An additional process that was assessed in cord blood studies was immune activation. Maternal immune activation (MIA) is described as the inflammatory activation of microglia during pregnancy, which adversely affects synaptogenesis and myelination18–21. A large cohort study of cord blood plasma and maternal mid-gestational blood found a specific cytokine profile distinguishing ASD and typically developing controls22. Similar findings were also reported in other cohorts23,24. The variety of studies examining different biomarkers in cord blood suggests two processes are involved with the development of ASD: early inflammatory changes, characterized by variations in cytokines, and alterations in transcriptional regulation associated with brain development. Although these studies investigate the cord blood characteristics from different aspects, the downstream tissue-level effects of these changes remain unclear. To address this gap, we quantify the protein content of extracellular vesicles (EVs), which may reflect molecular processes occurring in fetal tissues. To evaluate this effect, we characterize the protein content of EVs in umbilical cord blood plasma (UCB-PL) from children later diagnosed with ASD25. EVs include three subtypes routinely excreted from cells: microvesicles, exosomes, and apoptotic bodies. These vesicles contain proteins, lipids, nucleic acids, and metabolites, serving essential functions in cell-to-cell communication and the clearance of cellular debris26. EVs primarily originate from the Golgi apparatus and endoplasmic reticulum (ER) systems, with a portion released directly through exocytosis at the cell membrane27. There is scarce literature on EVs derived from the plasma of children with ASD28. Existing studies report that EVs extracted from the plasma of children with ASD and neurotypical (NT) controls across various ages exhibit similar size and morphology. The top five downregulated proteins were WW domain-containing protein 2 (WWP2), Heat shock protein 27 (HSP27), C-type lectin domain family 1 member B (CLEC1B), Cluster of differentiation 40 (CD40), and folate receptor alpha (FRalpha). These proteins are mainly involved with immune modulation, and FRalpha and CLEC1B are brain-specific28. EVs derived from the peripheral blood mononuclear cells (PBMCs) have been shown to exhibit differences in 41 proteins associated with mitochondrial function, ER activity, and inflammatory processes29. Our study evaluates EVs’ protein content in UCB-PL, exploring potential markers linked to ASD. The protein content of EVs reflects cellular processes occurring in various body tissues, including the brain and the immune system. Deviations in the protein profile of children with ASD may serve as a sign for the condition and may provide insights into the mechanisms underlying its development and potentially aid in diagnosis. Cord blood proteomics presents a unique opportunity to measure the prenatal protein profile, capturing biological processes in utero before any postpartum environmental influences occur. This study aims to compare the protein content of EVs in UCB-PL between children later diagnosed with ASD and NT controls. We hypothesize that EVs from children with ASD will exhibit a distinct protein profile indicative of altered molecular processes that interfere with neurodevelopment and ultimately contribute to ASD beginning prenatally or emerging postnatally. Results Demographic profiles in ASD and NT groups All samples were collected at a tertiary medical center between 2021 and 2024, using an identical protocol. The study analyzed 60 cord blood plasma samples from 30 children with ASD (26 males) and 30 NT children (15 males). The ASD group had a higher proportion of males, following the expected sex differences in ASD. These differences did not significantly affect the proteomic difference between the groups, as explained below. Table 1 presents the cohort characteristics and data collected from 27/30 children in the NT group. The data of three children could not be retrieved as their cord blood units were collected anonymously for quality control. Maternal age, gestational age, birth weight, and Apgar scores at birth were compared between ASD and NT children and were found to be similar. ASD severity levels were categorized into three levels: 11 children required low support, 11 required moderate support, and 8 children required marked support. Cognition assessments conducted using the Leiter brief IQ yielded a mean score of 83.3 (SD = 33.6), with 36% (11/30) having an estimated IQ below 70, aligning with proportions reported in large cohort studies30. Table 1Delivery and development assessment of 30 diagnosed children with ASD and 30 NT children. ASD severity is calculated based on the DSM-5 score.NT (n)NTASD (n)ASDP valueMale/Female3016/143026/40.001Maternal age at delivery (year) (Mean, std)2732.4 ± 3.731.6 ± 5.60.4Gestational age (week) (Mean, std)39.7 ± 1.138.7 ± 1.30.4Weight (gr) (Mean, std)3336 ± 4603298 ± 3190.3Apgar 5 (Median IQR)10 (10, 10)10 (10, 10)0.03ASDAge of diagnosis (month) (Mean, std)3038.8 ± 17.3Brief IQ Leiter (Mean, std)83 ± 33Brief IQ Leiter < 7011Severity (DSM-5 Criteria)11Mild11Moderate8Severe Delivery and development assessment of 30 diagnosed children with ASD and 30 NT children. ASD severity is calculated based on the DSM-5 score. Differential expression and reduced variability in ASD proteomic profiles reveal quantitative differences We conducted a mass spectrometry‐based proteomic analysis of EVs from UCB-PLs, identifying 5,478 proteins that were significantly expressed (Fig. 1). The mean expression level (MEL) of EV markers (CD, cluster of differentiation), CD9, CD81, and CD63 was detected at similar levels across all samples, confirming consistent identification of EVs in each group with no significant difference (see SM Fig. S1). Workflow for analyzing and identifying the most informative proteins distinguishing ASD and NT groups. A volcano plot (Fig. 2a) illustrates the results of the FDR test (two-sample S0 test 0.5, FDR 0.01) comparing the ASD and NT groups. This analysis revealed significant differences in the expression of 565 proteins, referred to as the Dif group. ASD proteins had higher mean expression levels and significantly lower variance. (a) Volcano plot depicting differential expression analysis between NT and ASD groups. Multiple two-sample t-tests were conducted with significance thresholds set at an S0 value of 0.5 and an FDR of 0.01. The x-axis represents the differences in t-test values between the NT and ASD groups. The y-axis shows the negative log10-transformed p-values from the two-sample t-test. Red markers indicate proteins with significantly different MEL between NT and ASD groups (NT-ASD), meeting both the S0 threshold and the FDR criterion. (b) Scatter plot depicting the MEL of proteins from the ASD and NT groups. The black dashed line represents the ASD-NT equality line. The blue line indicates the trend line, and the light blue shading represents the variation. (c) Histogram showing the distribution of MEL of all proteins in the NT (blue) and ASD (orange) groups. (d) Histogram showing the variance distribution of the expression levels of all proteins in the NT (blue) and ASD (orange) groups. (e) Histogram representing the distribution of MEL of Dif proteins from the NT (blue) and ASD (orange) groups. f. Histogram showing the distribution of MEL of all proteins from the NT (blue) and ASD (orange) groups. Among the 2592 proteins significantly expressed in at least 21 children in each group, 565 proteins exhibited differential expression. Of these, the majority (532 proteins) exhibited elevated MEL in the ASD group, indicating primarily quantitative differences between the groups. The remaining group of proteins, which shows no significant differences, is categorized as NDif. A scatter plot comparing protein expressions in the ASD and NT groups depicts a strong correlation (R = 0.94, p < 0.001), demonstrating that all proteins detected in the ASD group were also present in the NT group (Fig. 2b). Figure 2c–f provides insights into group-level differences in protein expression. For Dif proteins, the MEL was significantly elevated in the ASD group compared to the NT group (Fig. 2c, p < 0.001). This was accompanied by a marked reduction in expression variance in the ASD group (Fig. 2d, p < 0.001). In contrast, for NDif proteins, the MEL did not differ significantly between groups (Fig. 2e, p > 0.05), yet the variance remained significantly lower in the ASD group (Fig. 2f, p < 0.001). Together, these findings indicate not only increased expression of Dif proteins in ASD but also a consistent reduction in proteomic variability across both Dif and NDif proteins. Differentially expressed proteins in ASD cluster into mitochondrial, ER, and immune system pathways Using STRING software, we conducted a protein interaction network analysis for both Dif and NDif proteins. The analysis revealed three main clusters: mitochondrial, ER, and a mixed cluster containing proteins related to the immune system, inflammation, and cytoskeleton (Fig. 3a). Notably, 295 out of 565 proteins in the Dif proteins are predicted to originate from the brain and central nervous system (Fig. 3b). Mitochondrial proteins were exclusively present in the Dif group, with the largest cluster comprising 254 out of 565 proteins (45%) associated with mitochondrial function. In contrast, STRING analysis of the NDif group showed that mitochondrial proteins accounted for only a small portion (1.5%) of the total (Fig. 3c). Differentially expressed proteins in ASD are enriched in CNS, mitochondrial, and endoplasmic reticulum compartments. (a) K-means clustering of Dif proteins revealing three distinct clusters. The red color is mainly mitochondrial proteins involved in aerobic respiration, the green color is mainly ER proteins, and the blue color is a mixture of proteins mainly associated with neutrophil degradation. Clustering was performed using the K-means (absolute) algorithm in STRING software (Legacy version). (b) Tissue expression analysis (TISSUES) from STRING software presents proteins potentially originating from the CNS in purple. A total of 296 out of 565 proteins are predicted to be expressed in the CNS. (c) Subcellular localization analysis (COMPARTMENTS) was performed using STRING software for enrichment analysis against a whole-genome statistical background. The x-axis represents the sub-cellular compartment source, while the y-axis represents the percentage of proteins from each group associated with each compartment. A total of 160 compartments were significantly enriched with Dif proteins (green), while 311 compartments were significantly enriched with NDif proteins (gray). According to STRING cluster analysis and Gene Ontology (Biological Process)31, the two primary processes within the mitochondrial cluster are aerobic respiration and the formation of protein complexes in the inner mitochondrial membrane. Based on the KEGG database32–34, mitochondrial proteins are mainly involved in oxidative phosphorylation (61 out of 254, strength 1.57). Sixteen proteins (strength 1.65) are related to the Tricarboxylic acid (TCA) cycle, and they were upregulated (Table S2). Disease-gene association analysis (DISEASES) indicates that 21% of proteins in this cluster (53 out of 254, strength 1.33) are related to mitochondrial metabolism. The second largest cluster consists of ER-related proteins, comprising 33% (187 out of 565) of the total Dif proteins. STRING cluster analysis and Gene Ontology identified ER stress proteins (30 out of 187, strength 1.15) as a key process. According to KEGG32–34, these proteins are involved in protein processing (30 out of 187, strength 1.29), and N-glycan biosynthesis (9 out of 187, strength 1.29). DISEASES links this three-protein cluster to major histocompatibility complex (MHC) class I deficiency (strength 2.02) and to congenital disorders of glycosylation (10 out of 187, strength 1.25). The final cluster includes a diverse group of proteins primarily associated with the immune system, with a total of 125 proteins. KEGG analysis indicates involvement in pathways such as lysosome (10 out of 125, strength 1.1)32–34. STRING cluster analysis and Gene Ontology identified only small groups (2–7 proteins) with a strength analysis larger than one. These groups consist of immune system response proteins of both the innate immune system and cytokine production. Additionally, DISEASES analysis identifies four proteins associated with chronic granulomatous disease (strength 1.85). The large number of proteins in the Dif group suggests that many of them belong to a common protein ensemble. To identify the most informative proteins distinguishing between the ASD and NT groups, we applied a combination of three analytical methods (Random Forest, Boruta algorithm, and two-sample t-tests) as presented in Fig. 1. This analysis identified eleven proteins with unique value in differentiating ASD from NT groups (Fig. 4a). These proteins represent each identified cluster, and their function and roles in brain development are summarized in Table S1. Expression levels of key proteins distinguishing NT and ASD groups. (a) Box plot illustrating the expression levels of proteins that most effectively distinguish between NT and ASD children. Statistically significant differences in MEL between the groups are indicated by double asterisks (**), with all comparisons yielding p < 0.01 (t-test, FDR). The color of the proteins on the x-axis is linked to their cluster origin. (b) Box plot illustrating the expression levels of proteins predominantly expressed in the brain, as identified by HPA. Statistically significant differences in MEL between NT and ASD groups are indicated by double asterisks (**), with all comparisons yielding p < 0.01 (t-test, FDR). The color of the proteins on the x-axis is linked to their cluster origin. The next step involved identifying proteins within the Dif group that are dominantly found in the brain. Using the Human Protein Atlas (HPA)35, 13 proteins were identified, representing each cluster (Fig. 4b). Their function and role in brain development are summarized in Table S2. The most informative proteins (Table S1) were further analyzed. While none of these proteins are brain-specific, they are distributed between the clusters, with a significant proportion involved in ER processes. Two proteins have been previously associated with ASD, according to the SFARI gene database36. One of these is TM9SF4, which is linked to microduplication of 20q11.21 observed in several family members associated with ASD or bipolar disorder. The second protein, MYH9, encodes a cellular myosin involved in cytokinesis. The MYH9 gene, located on chromosome 22, was associated with 11 out of 47 family members diagnosed with ASD37. Additionally, three proteins are involved in redox homeostasis (PRDX4, TM9SF4, ATP2C1), and HYOU1 has a role in regulating ER activity, specifically during ER stress13. Furthermore, DAGLβ is part of the immune response, regulating macrophages and dendritic cells38 (Table S1). Using the HPA database (Table S2)35, 13 brain-dominant proteins were recognized. These proteins are likely to originate from brain-derived EVs. The brain-dominant proteins were found in all three clusters (Fig. 4b). Four proteins (CLEC2L, SIRPα, TREML1, and CMTM5) are involved in immune system regulation, specifically microglia and the innate immune system. Additional proteins are involved in brain myelination, axonal projection, and the extracellular matrix (including RELN, TMEM63A, and DMTN). Five other proteins are involved in various ER processes and are known to be modulated during stress (DNAJC5, PSCK6, STRIP2, RTN3, and TMCC2). Lastly, GLUD1, a mitochondrial protein, is involved in the conversion of L-glutamate into alpha-ketoglutarate. Notably, both RELN and GLUD1 are described in the SFARI database as ASD-associated genes, classified under categories 1 and 2, respectively. (Table S2). Association of proteomic analysis and epidemiological factors We examined potential correlations between the identified proteins and the variables of sex and ASD severity. No significant associations were observed between protein expression clusters and ASD severity levels. Discussion Despite substantial progress, ASD presents a significant challenge due to its heterogeneous nature and complex etiology. Evidence suggests that ASD arises from a multifaceted, unbalanced interplay between genetic, environmental, and immune factors. This study sought to investigate how these connections may already be reflected at birth in cases of non-syndromic ASD, under the assumption that they may converge toward abnormal brain microstructure and function8,39. The aim of this study was to find the protein profile contained in EVs, reflecting this diversion. To evaluate this, we analyzed protein expression profiles in EVs derived from UCB-PL samples from children with ASD and NT controls. Differences in protein expression between the groups may serve as an early indicator of susceptibility to ASD and provide insight into the underlying abnormal mechanism that precedes its onset. Although sex differences are known to influence protein expression40, ROC analysis incorporating both sex and the most informative proteins revealed that protein expression levels were a significantly stronger predictor of ASD development. Proteomic analysis of EVs from UCB-PL revealed elevated expression of 565 proteins in individuals with ASD, accompanied by lower variability in protein expression, particularly within the Dif group. This implies that, despite the diverse origins of non-syndromic ASD, affected individuals present a remarkably similar EV protein profile at term gestational age. This convergence of pathways near the time of delivery may reflect a similar cellular activity associated with ASD development. Given that the low variance was observed exclusively in the ASD group, we suggest that this uniform activity is associated with the future emergence of ASD. Identifying inflammation, mitochondrial, and ER clusters as the primary groups within the Dif proteins emphasizes the main pathways potentially involved in brain development in ASD. Moreover, several specific proteins are known to be involved in neuroinflammation, mitochondrial dysfunction, and ER dysregulation (Table S2). These findings, while circumstantial, provide evidence for the role of these pathways in the development of autism (Fig. 5). Schematic illustration of the convergence hypothesis. An illustration of the suggestive convergence of different etiologies, including genetic, epigenetic (e.g., prematurity, environmental pollution), and inflammation, that impact mitochondria and ER function. This dysregulation may disrupt brain development, particularly synaptogenesis. MIA during pregnancy is often triggered by infections or inflammatory conditions, and it is associated with ASD21,41,42. In these cases, elevated levels of cytokines and antibodies can cross the placenta and enter the fetal circulation, activating the fetal microglia within the developing nervous system21,43,44. Once activated, microglia produce additional pro-inflammatory cytokines, disseminating the state of neuroinflammation. The neuroinflammation ultimately dysregulates gene expression and RNA editing related to neurodevelopment45 and specifically to ASD, such as FMR1, CHD8, mTOR, and Eif4e46,47. Five proteins from the most informative and brain-dominant lists (Table S2) are known to be associated with neuroinflammation. Among them, SIRPα, together with CD47, is localized to the neuronal membrane. At the same time, DAGLβ and TREML1 are associated with the modulation of microglial activity. The proteins SIRPα and DAGLβ reduce inflammation and phagocytosis, whereas TREML1 has a pro-inflammatory effect. CLECL2, a protein found on platelets and immune cells, also modulates microglial activity and influences the integrity of the blood–brain barrier. The activation of microglia by these proteins eventually triggers phagocytosis, excretion of cytokines, and release of reactive oxygen species (ROS)48. These processes influence synaptic development by modifying synaptogenesis, maturation, and pruning at various stages of neural growth49. A recent study of plasma EVs in children diagnosed with ASD identified two downregulated proteins related to inflammation, HSP27 and CLEC1B. This finding suggests that inflammatory processes continue to play a key role in ASD even beyond early development. Such inflammatory activity also alters the function of sub-cellular organelles such as mitochondria or ER, both of which are independently essential for proper brain development28. Mitochondria have a major role in several key cellular processes, including aerobic and anaerobic energy production, calcium homeostasis, production of free radicals (involving oxygen and nitrogen species), and programmed cell death. Mitochondrial diseases may result from genetic variants in mtDNA or nucDNA. The resulting phenotype depends on the severity of the genetic variants, the distribution of heteroplasmy in different brain regions, and external triggers to mitochondria (i.e., infection, ischemia). The prevalence of mitochondrial disease in individuals with ASD is ~ 5.0%, markedly higher than the estimated ~ 0.01% in the general population50–52. Moreover, the prevalence of biomarkers indicating mitochondrial dysfunctions is even higher. These perturbations can significantly affect brain development and contribute to the onset of ASD50,53,54. Although most cases of mitochondrial dysregulation are mainly secondary, primary genetic causes have also been described55,56. Recent studies demonstrated that neural maturation is associated with a transition of the mitochondria from a fusion state to a fission state. The rate of change between the mitochondrial states and the mitochondrial metabolic rate is linked to the rate of brain maturation in different species57. Moreover, mitochondria localized in the dendritic tree are necessary for the development of synapses at the end of pregnancy58. Myelination, another crucial brain development process that intensifies around term birth, relies profoundly on mitochondrial energy production59. Mitochondrial metabolism also plays a critical role in neurotransmitter production and synaptic transmission by stabilizing intracellular calcium concentrations, which enables the excretion of neurotransmitter vesicles. Consequently, normal mitochondrial function is essential for proper brain function, and any divergence, primary or secondary, would damage basic brain activity and development60. The mitochondrial proteins described in Table S1 (GLUD1, PRDX4, TM9SF4, and ATP2C1) are involved in redox homeostasis and glutamate metabolism61. Moreover, fourteen proteins of the TCA cycle had higher MEL (Table S3). These processes are known to be associated with ASD development50,53,54,62. Their increased presence may be a marker for elevated mitochondrial activity, potentially associated with increased synaptogenesis and myelination during late gestation57–59. Another explanation is that EVs may contain vesicles of mitochondrial or apoptotic vesicles26,27. Thus, their increased presence may be a sign of mitochondrial injury secondary to oxidative, toxic, or inflammatory processes53,54. Modulation of these proteins may be a result of genetic or epigenetic origin or secondary to mitochondrial stress, which eventually leads to abnormal synaptogenesis and impaired myelination63. The abundance of MIA and mitochondrial proteins in the EVs content could be interpreted as a response to a variety of harmful processes. This response was previously conceptualized as the cell danger response (CDR)64,65. An evolutionary protective response that initially signals the near physiological environment by excreting ATP and TCA metabolites to allocate energy to neutralize the threat and to protect the undamaged tissue. The second step is recovery, regeneration, and healing of the damaged tissue. Mitochondria and the immune system are the major key players in both steps of the CDR switching from neutralizing mode to healing mode during the process. CDR is one of the etiologies for chronic neurological deficits, including ASD64,65. ER is another important factor of the CDR, characterized by a unique response to cellular threat. ER-related proteins are also the second most abundantly represented group of organelle-derived proteins in our samples. Notably, brain-dominant ER proteins are frequently associated with ER stress18,29. This process is triggered by the accumulation of unfolded proteins in the ER, arising from various etiologies. In response, the unfolded protein response (UPR) is activated to restore cellular balance. This involves enhancing chaperone activity to assist in proper protein folding. It also promotes the degradation of unfolded proteins and temporarily slows overall protein production66. The presence of proteins associated with UPR implies that the ER was under stress at the time of birth. Mitochondria actively communicate with the ER through multiple pathways, and this crosstalk plays a central role in regulating neurogenesis, synaptogenesis, and myelination. Both organelles are key factors in regulating cellular calcium homeostasis, a necessary process for synaptic transmission, and the determination of cell fate54,64,65,67. Similar to mitochondria, the ER is influenced by microglia activation in similar processes. These disruptions have been linked to abnormal neurodevelopment in both mouse models and cellular studies65,68. Among the Dif proteins, we identified at least six highly expressed proteins associated with ER stress (e.g., DNAJC5, PSCK6, STRIP2, TMCC2, RTN3, and HYOU1). These proteins may indicate the presence of CDRs and could influence the specific nature of proteins observed in our cohort. The key findings of this study reveal that 565 proteins differ between the ASD and NT groups, originating from mitochondrial, ER, and inflammatory pathways, and exhibiting very low variation. Although these findings may stem from genetic causes, the abundance of mitochondrial proteins more strongly suggests a cellular response to stress, consistent with the concept of the CDR50,54,64,65. This suggests that mitochondrial and ER dysregulation, both known to impair brain development, may represent a common pathway among individuals in our cohort. MIA could be either the cause of this dysregulation or act as an additional factor influencing brain development, as suggested by CDR. Clearly, this assumption requires validation in future studies utilizing induced pluripotent stem cells (iPSCs) from the children in these cohorts. These iPSCs differentiated into cortical organoids containing glial cells. Exposing these organoids to physiological and stress-inducing conditions will enable us to study cellular changes, intercellular interactions, and their role in synaptogenesis and myelination. We predict that neural cells derived from ASD participants will have a distinct CDR response compared to those from NT participants, eventually leading to abnormal regenerative and healing processes. Limitations Sex differences appear to have minimal impact on proteomic predictors of ASD. As shown in Table 1, the NT group includes a higher proportion of females compared to the ASD group, which can potentially impact the proteomic results. To assess the impact of sex on the results, we conducted ROC analysis to distinguish between NT and ASD samples, using the most informative proteins as predictive factors. This analysis was done initially using protein data alone and then repeated with sex included as a variable. As demonstrated in Table 1, the inclusion of sex only had a marginal effect on the diagnostic accuracy. This study has several limitations. First, the cohort includes 30 children with ASD, and the results are valid only for this specific sample. However, the uniformity in sample collection timing and procedures reduces potential variability, supporting the validity of our results. Second, the NT group has a higher proportion of females compared to the ASD group, which may influence proteomic results. Although ROC analysis showed that including sex as a variable had only a marginal effect on diagnostic accuracy (Table 2), future studies with balanced sex distributions are needed. Table 2The ROC-AUC (receiver operating characteristic—area under the curve) scores for the predictive performance of each protein individually and in combination with sex as a supplementary factor.ROC-AUC ScoreSingle ProteinSex & Single ProteinESYT20.970.97HYOU10.990.99MYH90.930.95PRDX40.960.97RER10.890.92SPCS20.990.98TM9SF40.970.97TOMM510.98ATP2C10.970.98NDUFS80.910.91DAGLB0.910.92ROC-AUC values closer to 1.0 indicate higher predictive accuracy. The ROC-AUC (receiver operating characteristic—area under the curve) scores for the predictive performance of each protein individually and in combination with sex as a supplementary factor. ROC-AUC values closer to 1.0 indicate higher predictive accuracy. Although canonical EV markers (CD9, CD63, CD81) were consistently detected, we did not systematically assess the presence of exclusion markers for non-EV compartments such as the ER, Golgi, or nucleus. As a result, we cannot fully exclude the possibility of minor contamination, which should be addressed in future studies using dedicated exclusion marker panels. Third, our findings require validation through cellular and molecular studies using iPSCs from both cohorts. Differentiation into cortical organoids will help explore cellular changes, synaptogenesis, and mitochondrial function, particularly in ASD-derived neural cells. Future work will also examine brain-derived EVs (BDEVs) during early development, correlating molecular changes with developmental assessments. Moreover, analyzing maternal plasma BDEVs using iPSCs and proteomics may provide insights into prenatal neurodevelopmental influences. Conclusions This unique study uses the possibility to test protein expressions within EVs from NT and ASD children at birth, which can serve as early sign for ASD. The analysis revealed marked differences in protein profiles between the groups, with ASD children consistently exhibiting higher expression levels. Importantly, the proteins that differentiated ASD from NT children originated mainly from the mitochondria, ER, and immune system pathways, suggesting that these organelles and processes play a central role in ASD pathophysiology. Notably, the low variability observed within the ASD group shows a potential convergent mechanism underlying ASD development despite its heterogeneous nature. These findings reinforce the hypothesis that mitochondrial and ER dysfunction contribute to ASD. By identifying the important molecular pathways at birth, this study lays the groundwork for future research aimed at understanding the cellular origins of ASD and developing targeted interventions during critical periods of neurodevelopment. Methods Selection and description of participants Children with ASD were recruited as part of a study to test the efficacy of autologous cord blood for autism. Enrollment took place between 2019 and 2023. All participants received a formal ASD diagnosis and severity level through a comprehensive assessment conducted by neurologists and psychologists according to DSM-V criteria69. The psychological evaluation included the Autism Diagnostic Observation Schedule (ADOS)70, and cognitive levels were assessed using the Leiter International Performance Scale, Third Edition (Leiter-3)71. Maternal, pregnancy, and delivery data were collected from electronic medical records. Chromosomal microarray (CGH) and Fragile X testing were performed, but not whole-exome sequencing. The NT children were chosen from a list of mothers who had consented to cord blood collection through the public cord blood bank at Tertiary Medical Center. As some of the mothers preferred to be anonymous, we were able to collect the epidemiological data only from 27/30 children. These children were not formally evaluated for the study. However, based on the estimate that the prevalence of ASD in certain populations is approximately 1.5%, the likelihood of having more than 2 children with autism in this cohort is very low and approximated to be 1%. Thus, the chance of having a child with ASD in the NT group will only minimally increase the noise of comparisons between the groups. Data collection and measurements Cord blood collection and plasma excretion Cord blood units from both ASD and NT children were collected and cryopreserved following an identical protocol approved by the Israel Ministry of Health. (National Cord Blood Center, Sheba Medical Center, Ramat Gan, 5262100, Israel). The plasma was separated from the collected cord blood before cryopreservation and stored at − 70 °C. This procedure was applied to all cord blood units. https://www.factglobal.org/about/]https://www.aabb.org/standards-accreditation/standards/cellular-therapy-services Isolation of plasma microvesicles Microvesicle isolation was performed as previously described72. Briefly, 400 μl plasma samples were collected and stored at − 80 °C. Samples were thawed on ice and centrifuged at 1000 × g for 20 min to remove cell debris. Microvesicles were subsequently precipitated by centrifugation at 20,000 × g for 60 min at 4 °C. Pellets were washed twice with 1 ml ice-cold phosphate-buffered saline (PBS) and centrifuged again at 20,000 × g for 60 min at 4 °C. In-solution digestion and liquid chromatography-mass spectrometry (LC–MS)/MS analysis Microvesicle pellets were solubilized in 6 M urea and 2 M thiourea in 100 mM Tris–HCl (pH 8.5). The samples were reduced with 1 mM dithiothreitol (DTT) at room temperature (RT) for 30 min and alkylated with 5 mM iodoacetamide (IAA) for 30 min in the dark. The lysates were diluted fourfold with 50 mM ammonium bicarbonate, followed by overnight digestion with Trypsin/Lys-C mix (MS grade Promega, 1:100 enzyme to protein ratio) and sequencing grade modified trypsin (Promega, 1:50 enzyme to protein ratio). The resulting peptides were acidified with trifluoroacetic acid (TFA), purified on C18 StageTips (Pierce), and vacuum-dried. The dried peptides were resuspended in 2% acetonitrile and 0.1% TFA prior to LC–MS/MS analysis. Peptides were analyzed by mass-spectrometry at the Wohl Institute for Translational Medicine Proteomics Unit, Tertiary Medical Center. Purified peptides were separated using an EASYnLC1200 nano-HPLC on EasySpray columns (Pep Map 50 cm, 0.75 µm i.d.) packed with 2 μm C18 material with a 100Å pore size. Separation was performed with a water-acetonitrile gradient, and peptides were injected into the Q-Exactive HF mass spectrometer (Thermo Scientific) via the EasySpray ionization source. The peptides were loaded with Buffer A (0.1% formic acid) and eluted with a gradient of 5–30% Buffer B (80% acetonitrile and 0.1% formic acid) at a flow rate of 300 nl/min over a gradient of 100 min. Mass spectrometric acquisition MS acquisition was performed in data-independent acquisition (DIA) mode using positive-ion detection. The DIA method included a survey scan at a resolution of 60,000 across the 400–1000 m/z range, with an automatic gain control target of 3E6 and 100 ms injection time. Following the scan, 25 DIA windows, spanning 400–975 m/z range, were acquired at a resolution of 30,000. For each window, the automatic gain control target was set to 1E6, with an injection time of 50 ms. The normalized collision energy was set to 27. Mass spectrometric raw data analysis The raw MS files were processed using DIA-NN 1.8.1 in high-precision mode with global cross-run normalization enabled. The output was at 1% precursor and protein group levels. Protein quantification was performed using the MaxLFQ algorithm, which was implemented in the diann R package. Database search was performed using the Human UniProt database in library-free mode. Statistics Protein identification and quantification were imported into Perseus73 for multiple-sample tests. The statistical workflow is presented in Fig. 1. The proteomic analysis revealed 5478 proteins. Subsequent statistical analyses were restricted to 2924 proteins with more than 21 valid values (70%) in each group (ASD and NT). A two-sample Student’s t-test was performed with an S0 = 0.1, and a false discovery rate (FDR) < 0.05 was applied to identify statistically significant differences in protein expression between ASD and NT samples. This analysis identified 565 proteins with significantly different expression levels, while 2359 proteins showed no significant difference. To further refine the identification of the most informative proteins that effectively distinguish between ASD and NT groups, feature selection was conducted (Fig. 1). Initially, Bonferroni correction for multiple comparisons, a more stringent approach, was employed to compare the ASD and NT groups in parallel to FDR. This conservative statistical method identified 344 differentially expressed proteins (p-value < 0.05). Then, the Boruta algorithm (Python package Boruta) and the Random Forest feature selection algorithm (Python package sklearn.feature_selection) were applied to each cohort of differentially expressed proteins. The FDR-based feature selection yielded 25 proteins, while the Bonferroni-based selection identified 24 proteins. Notably, 11 proteins were common to both cohorts. These proteins were subjected to additional biological analysis to assess their relevance and functional significance73. The complementary application of FDR and Bonferroni corrections provided a dual layer of statistical rigor, balancing discovery sensitivity with stringent control of false positives. This approach ensured the reliability and robustness of the identified proteins. Additional statistical analyses were performed using the Python package scipy.stats. The normality of protein expression data was assessed using the Shapiro–Wilk test. Based on data distribution, protein expression levels were compared using either the unpaired t-test or the Mann–Whitney U test. Correlation analysis was conducted using the Pearson correlation coefficient, with statistical significance set at p-value < 0.05. Protein clustering and characterization were performed using STRING version 12.0. The STRING database (https://string-db.org/) systematically collects and integrates protein–protein interactions, both physical interactions as well as functional associations31. Pathway and sub-cellular origin characterization were conducted using the Kyoto Encyclopedia of Genes and Genomes (KEGG), which is integrated into STRING32–34. Subgrouping of the proteins will be presented as a portion of a specific group of proteins from the whole group and its enrichment strength (Log10 of Observed/Expected presence of proteins from this group), all of which have an FDR < 0.05. Supplementary Information Supplementary Information. Acknowledgements We would like to express our gratitude to Michaela Smilovici Ofir, PhD, Research Grants and Academic Collaborations Manager at Sheba Medical Center, for her invaluable support. We thank Dr. Noa Burshtein for editing the manuscript. Abbreviations ATPase secretory pathway Ca2+ transporting 1 Autism diagnostic observation schedule Autism spectrum disorder Brain-derived extracellular vesicles Cluster of differentiation Cell danger response Chromodomain helicase DNA binding protein 8 C-type lectin domain family one member B C-type lectin domain family two member L Copy number variations CKLF like MARVEL transmembrane domain containing 5 Dematin DNA methylated regions Diacylglycerol lipase beta Data-independent acquisition Significant differences No significant differences Disease-gene association analysis DnaJ heat shock protein family (Hsp40) member C5 Eukaryotic translation initiation factor 4E Extended synaptotagmin 2 Extracellular vesicles Endoplasmic reticulum Fragile X messenger ribonucleoprotein 1 Glutamate dehydrogenase 1 G protein–coupled estrogen receptor-1 Human protein atlas Heat shock protein 27 Hypoxia up-regulated 1 Induced pluripotent stem cells Kyoto encyclopedia of genes and genomes Leiter international performance scale, third edition Maternal immune activation Mitochondrial DNA Mean expression level Myosin heavy chain 9 Mechanistic target of rapamycin Neurotypical Nuclear DNA NADH: ubiquinone oxidoreductase core subunit S8 Peripheral blood mononuclear cells Proprotein convertase subtilisin/kexin type 6 Peroxiredoxin 4 Progesterone receptor Reelin Retention in endoplasmic reticulum sorting receptor 1 Reticulon-3 Receiver operating characteristic Receiver operating characteristic Area under the curve Retinoic acid—related orphan receptor alpha Reactive oxygen species Superoxide dismutase 2 Signal peptidase complex subunit 2 Signal-regulatory protein alpha Striatin-interacting protein 2 Subcellular localization analysis Transmembrane 9 superfamily member 4 Transmembrane and coiled-coil 2 Transmembrane crotein 63A CSC1-like protein 1 Tissue expression analysis Translocase of outer mitochondrial membrane 5 Triggering receptors expressed on myeloid cells like 1 Tricarboxylic acid Umbilical cord blood plasma Unfolded protein response Author contributions L.K. analyzed the data, prepared Tables S1–S3 and Fig. 5, and contributed to the main manuscript text. O.B.Y. analyzed the data and contributed to writing the main manuscript text. C.Y. analyzed the data and prepared Figs. 2 and 4, as well as Table 2. B.B. collected data and prepared Table 1. M.M. and P.G. collected samples and data. E.G.S. collected data and prepared Fig. S1. All authors reviewed the manuscript. Data availability The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Declarations Conflict of interest The authors declare that they have no conflict of interest. Ethical approval All experiments were performed in accordance with relevant guidelines and regulations. The study was approved by the Sheba Medical Center Ethics Committee (IRB-5105-18-SMC), and informed consent was obtained from all participants and/or their legal guardians. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Information. Data Availability Statement The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.