Background

Introduction

Autism spectrum disorder (ASD) is a complex neurodevelopmental disability with a range of clinical presentations and associated etiologies. ASD is characterized by persistent deficits in social interactions across multiple contexts combined with the presence of restricted, repetitive patterns of behavior, interests, or activities [1]. According to the Autism and Developmental Disabilities Monitoring (ADDM) Network, in 2020, 1 in 36 children were estimated to be identified with ASD. There is a 4:1 male to female ratio and it is seen across all racial, ethnic, and socioeconomic groups [2].

Rates have been steadily increasing over the past few decades, partially due to improved screening leading to increased disease recognition. Yet a complete explanation for increased prevalence has remained elusive. Additionally, despite decades of research, no substantial pharmacologic option for targeting the core symptoms of ASD has been identified. It is therefore becoming imperative to uncover the underlying pathophysiologic processes that contribute to the development of ASD. To date, there have been a multitude of clinical characteristics associated with the development of ASD, encompassing prenatal, perinatal, and postnatal environmental exposures as well as genetic risk factors [3, 4]. What remains unknown is how the spectrum of these causative factors converge upon a common neurodevelopmental phenotype [5].

Increased inflammatory biomarkers and impaired oxytocin signaling have been reported in ASD patients compared to controls. However, a direct pathophysiological mechanism has yet to be established. The Receptor for Advanced Glycosylation End-products (RAGE) system is a multi-ligand receptor system that is part of the immunoglobulin (Ig) superfamily and is implicated in pro-inflammatory gene signaling involving NF-kB, IL-1, IL-6, and TNFα. Differences in RAGE axis signaling have been implicated in patients with ASD. Based on emerging evidence, we propose an ASD pathophysiology based on interactions between the oxytocin, inflammatory, and gut microbiome systems that is mediated through the RAGE system in autism subpopulations.

Inflammation in ASD

Human epidemiological and animal pathophysiological evidence suggests that prenatal and postnatal inflammation are associated with the development of ASD. Maternal immune activation (MIA) has been epidemiologically implicated in the development of ASD whether triggered by autoimmune / immune disease (AID) or infectious causes. A longstanding association between first-trimester neurotropic viral infections (specifically rubella, measles, mumps, varicella, herpes simplex virus, and most commonly cytomegalovirus) and the subsequent development of ASD is well known [6]. A 2020 French systematic review and meta-analysis including 180,000 children with NDD and 14 million control children found AID in mothers to be associated with increased risk for ASD (Adjusted OR 1.27 [95% CI 1.03; 1.57] p = 0.02). This increased ASD risk was also seen to a lesser extent in fathers with AID (adjusted OR 1.18 [95% CI 1.07; 1.30] p = 0.01) [7].

The underlying pathophysiology for this increased prevalence of ASD in offspring of parents with AID is unknown. However, immune signaling of hematopoietic stem cells, t-cell population differences, transplacental cytokine signaling, and maternal antibodies to fetal brain tissue have been implicated. Evidence in a murine model suggests that maternal immune activation stimulates changes in fetal hematopoietic stem cells that persist well beyond delivery [8]. Maternal immunologic factors that have been shown to contribute to alterations in neurodevelopment include maternal antibodies targeting fetal tissue and dysregulation of the developing immune system. T-cell population changes and specific cytokine profiles may provide biomarkers of this proposed pathophysiological mechanism.

Epidemiologically, there is substantial evidence for a correlation between prenatal inflammation during gestation and development of neuropsychiatric illnesses, in particular ASD, OCD, and schizophrenia. Several major studies on this topic come from Denmark, including a 1.6 million child cohort study that found admission to the hospital due to viral or bacterial infection during the first two trimesters of pregnancy to be associated with an increased risk of ASD development [9]. An even larger Dutch systematic review involving 2.2 million singleton infants followed over 38 years revealed that 2.26% of children (n = 50,863) were born to mothers with MIA during the gestational period, and that these children had a slightly increased risk for intellectual disability (HR, 1.19; 95% CI, 1.07–1.34) and childhood autism (HR, 1.21; 95% CI, 1.08–1.36), (see Table 1) [10].

Table 1 “Association of Offspring’s risk of specific mental disorders with prenatal exposure to any autoimmune disease.” This table was adapted from a 2022 Dutch systematic review, which found a 31% increased risk of schizophrenia and 42% increased risk of OCD in children born to parents with MIA. Table adapted from He, 2022
Full size table

There is evidence from fMRI data that schizophrenia, OCD, and ASD have shared features including alterations to fronto-striatal pathways. In terms of schizophrenia, a fMRI study of 22 unmedicated schizophrenic patients and 20 healthy controls (HCs) showed reduced functional activity in the anterior cingulate cortex, striatum, and midbrain of schizophrenic patients compared to HC [11]. A similar finding has also been observed in OCD patients and is reproduced in multiple studies in both conditions. One fMRI study of 44 OCD patients and 43 HCs demonstrated reduced functional connectivity between the caudate and the ventrolateral prefrontal cortex that was selectively associated with reduced cognitive flexibility as defined by attentional set-shifting [12]. In ASD, there are several fMRI studies that demonstrate reduced activation in frontal cortex and ventral striatum, suggesting a mechanism for the behavioral inflexibility seen in ASD [13, 14]. Schizophrenia, OCD, and ASD prevalence is increased in mothers with immune activation during gestation, and there is evidence for altered fronto-striatal pathway signaling in all these conditions.

Several cytokines are implicated in the pathophysiology of neuroinflammation and ASD. In humans, one study detected evidence for HLA-DR4 activation being associated with NDD with an estimated odds ratio of 4.67 (95% CI: 1.34–16.24) [15]. In the mouse model, interleukin 6 (IL-6), and interleukin 17 (IL-17) secretion during pregnancy mediate the occurrence of ASD-like behavior in pups [16, 17]. Furthermore, a more recent study has shown that maternal autoantibody-related (MAR+) mothers have a higher level of proinflammatory cytokine interferon-gamma and pathogen exposure, resulting in a consistent increase in multiple proinflammatory cytokines and chemokines [18].

Intriguingly, maternal antibodies to fetal neural tissue are implicated in some ASD pathogenesis. Cord blood from the Childhood Autism Risk from Genes and the Environment Study demonstrated IgG reactivity to fetal brain protein found at 37 and 73 kDa in 7 of 61 (11.5%) mothers of ASD children that were not found in mothers of neurotypical children. This increase in fetal exposure to maternal autoantibodies has been replicated in multiple other studies with concurrent rates around 10%–12% [19]. A large 2013 study using plasma from the Simons Simplex collection assessed the plasma of 2,431 pregnant mothers of children who would later be diagnosed with ASD and compared it with plasma from 654 mothers of neurotypical children. It found that mothers of children with ASD were four times more likely to have circulating brain-reactive IgG [20]. This data suggests that maternal IgG plays a role in ASD pathogenesis, where a subset of ASD patients may have more neural permeability to maternal autoantibodies potentially resulting in synaptic dysfunction.

Whether fetal brain inflammation originates from maternal antibodies crossing into the brain, direct infection with a neurotropic virus, or secondary interactions from other cytokine systems, preliminary evidence suggests that epigenetic modification occurs during gestational immune activation that predispose the development of ASD. When pregnant mice are injected with Poly I: C, an immunostimulant, there are notable methylation changes in the offspring. The prefrontal cortex glutamic acid decarboxylase 1 (GAD1) and glutamic acid decarboxylase 2 (GAD2) genes become hypermethylated, which is in turn associated with increased MecP2 binding and lower gene expression [21]. This mouse model provides a potential link between a downregulated GABAergic system, intrauterine neuroinflammation, and the development of an autistic phenotype.

After the child is born, epidemiological evidence suggests ongoing inflammatory hyperactivity in some ASD patients, as demonstrated by an increased prevalence of asthma, atopic dermatitis, allergic rhinitis, and irritable bowel syndrome (IBS). A 2015 Japanese metanalysis of including 2,234 children with ASD and 8,146 neurotypical control children found ASD children were 69% more likely to have asthma, and 66% more likely to have allergic rhinitis than age-matched controls [22]. Insight into the magnitude of the difference comes from a slightly larger 2016 Chinese cross-sectional metanalysis which found the prevalence of asthma in ASD to be 20.4%, and only 15.4% in age-matched controls (P < 0.001) [23]. Preliminary evidence suggests that atopy (the triad of allergic rhinitis, eczema, and asthma) is associated with the relative severity of ASD core symptoms, with a 2017 Australian study of 45 atopic and 95 non-atopic children, which found that on ADOS-2 testing, atopic children were 2.3 times more likely to experience ADOS-2 highest level severity symptoms and 2.9 times more likely to show social difficulties when compared to non-atopic ASD children [24]. Finally, a 2017 Johns Hopkins University literature review of 84 studies with mostly questionnaire-driven prevalence data of GI symptoms in ASD found that 46.8% of ASD patients report GI symptoms with the most reported symptoms being constipation followed by diarrhea, followed by abdominal pain [24]. In summary, epidemiological evidence suggests an increased prevalence of immune activation in ASD children as demonstrated by an increased prevalence of the atopic triad and IBS symptomology seen in ASD.

Some of these issues may be due to decreases in immune system regulating T-reg cell populations along with increases in immune system activating TH17 populations, which has been noted in several studies and is disturbed to a greater magnitude in ASD children with co-morbid GI symptoms [25]. A recent metanalysis of T-cell characterization studies found a statistically significant decreased ratio of T-reg / TH17 populations in ASD patients (n = 114) compared to controls (n = 107) [26]. These T-cell populations assist in antibody presentation and the adaptive immune response and may trigger microglia to disrupt appropriate neurodevelopment. In fact, estimates suggest that up to 20% of ASD cases may have maternal anti-brain reactive antibodies, and therefore are potential targets for immunomodulation during gestation to reduce prevalence of autism in at-risk subgroups [27]. Figure 1 below gives a generalized overview of a proposed mechanism of immune cell activation contributing to the pathogenesis of ASD:

Fig. 1
Fig. 1
Full size image

Graphic depicting how immune system activation results in differential synaptic pruning and subsequent NDD. (A) Immune dysregulation occurs through autoantibody production and altered complement/cytokines profiles. (B) Autoantibodies and complement enter the central nervous system and influence synaptic pruning. (C) The NLRP3 inflammasome pathway within macrophages and microglia leads to the propagation of the inflammatory response through cytokine release. Cytokines then effect neuronal and oligodendrocyte proliferation, survival vs. death. The net result of these immunologic factors is altered neurodevelopment and behavior. Images adapted from Van der Water, et al. 2017 and Blevins, 2022. [28, 29] Created with BioRender.com

One proposed model of ASD pathogenesis advanced by Naviaux and colleagues implicates the cell danger response (CDR). The CDR is an evolutionarily conserved, adaptive response to threats such as infection, tissue injury, hypoxia, or toxic exposure that reprograms cellular metabolism, mitochondrial function, and inflammatory signaling toward survival and defense. In this state, extracellular ATP–purinergic signaling increases, biosynthetic and growth-related pathways are downregulated, and cellular resources are redirected away from repair and development. Although this is adaptive after acute insult, failure to resolve the CDR has been proposed as a mechanism contributing to chronic disease states [30]. In Naviaux’s 3-hit model of ASD, the first hit is a genetic or epigenetic predisposition that lowers the threshold for CDR activation or impairs its resolution; the second hit is an early-life exposure that initiates the CDR; and the third hit is recurrent or persistent exposure to CDR-activating triggers for at least 3–6 months during critical neurodevelopmental windows from late gestation through the first 18–36 months of life [31]. This framework provides a mechanistic explanation for the frequent observation of proinflammatory signaling in ASD, whereby persistent, unresolved CDR sustains immune activation. Variability in the timing and type of insult differentially impacts the development of neural circuit formation, therefore restoring CDR to normal may be insufficient to rescue the phenotype in subpopulations.

In summary, evidence from MIA prevalence studies, maternal immunoglobulin found in fetal brains with ASD, and altered T-cell populations in ASD children suggests that prenatal and postnatal inflammation correlate strongly with ASD prevalence. Additionally, early and chronic activation of the CDR has been hypothesized as a central mediator of ASD pathogenesis strongly implicating inflammation in the process. The increased atopy and GI symptoms seen in ASD patients may be propagated by a dysregulated immune system. The next section will discuss the inflammatory biomarkers previously implicated in ASD neurodevelopment.

Inflammatory biomarkers in ASD

Several inflammatory biomarkers have been implicated in the pathogenesis of ASD, including TNF-α, NF-kB, IL-1β, IL6, and IL-17a. While these biomarkers are frequently found in other systemic inflammatory conditions, the significance of their elevation in ASD patients has yet to be elucidated. The research discussed below will delve into the differences in brain tissue and serum levels of these molecules in ASD children versus age matched controls. It is important to understand that these cytokines represent different inflammatory signaling pathways that culminate in the formation of multiprotein oligomers called inflammasomes which activate innate immune responses. Whether triggered by TNF-alpha, interleukins, nuclear factors, or toll-like receptors, there is formation of inflammasomes that activate pattern recognition receptors (PRRs) through pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs), thereby leading to prolonged systemic inflammation [32]. Excessive inflammasome formation then activates t-cells and microglia resulting in a prolonged systemic inflammatory state (summarized in Fig. 1).

Tumor Necrosis Factor-alpha

Tumor Necrosis Factor-alpha (TNF-α) is both a cytokine implicated in immune system signaling and an adipokine implicated in the insulin resistance associated with obesity-induced type 2 diabetes. Abnormal levels of TNF-α have been implicated in many conditions including Alzheimer’s, major depression, psoriasis, inflammatory bowel disease, and more recently in Autism. A 2017 Chinese study of 11 cytokines measured in children (n = 50) found significantly increased TNF-α, IL-1β, and IL-17a concentrations in ASD children compared to typically developing controls, but only TNF-α concentrations were positively correlated with severity of ASD symptoms on all 5 different ABC sub-scales, as well as being predictive of an ASD phenotype [33]. A 2018 Iranian study (n = 71) demonstrated upregulation of TNF- α, IL-6, and IL-17 in age matched controls, as well as a down regulation of IL-2 in ASD children [34]. In a mouse model, addition of TNF-α to the brain tissue demonstrated regional specificity in the frontal lobes and temporal region, in a process involving the NF-κB signaling pathway [35].

Nuclear factor kappa B

Nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) is a nearly ubiquitously expressed immune regulating protein that establishes a cytokine feedback mechanism that can produce chronic or excessive inflammation and has been implicated in the pathogenesis of ASD. Several genes related to this NF-κB pathway are associated with Autism according to the Simons Foundation Autism Research Initiative (SFARI) database— most notably NFIB, but also NFIA, and NFE2L3 which have known interactions with NF- κB. The Child and Adolescent Twin Study in Sweden (CATSS), studied 12,319 children aged 9–12 year old and found two SNPs in the NF-kappa-B inhibitor-like protein 1 gene (NFKBIL1) that were significantly associated with autistic traits in males [36]. These data indicate that genetic perturbation of the NF- κB system is implicated in ASD pathogenesis.

Some evidence suggests that NF- κB expression may be elevated in certain brain regions of ASD patients and may represent a common pathophysiology by which stress inflammatory signaling modulates behavior [37]. NF-kB expression in the brain tissue of autistic children from the London Brain Bank (ASD: n = 3, controls: n = 4) and Harvard Brain Bank (ASD: n = 6, controls: n = 5) found NF-κB was aberrantly expressed in orbitofrontal cortex in patients with ASD [38]. The altered NF-kB levels in human brain samples does not appear to be a diffuse process, and frontal lobes of post-mortem ASD children did not show the elevated NF-kB in a different pathology review of 7 ASD children with 7 age matched controls from the NICHD Brain and Tissue Bank for Developmental Disorder [39]. Dysregulation of the orbitofrontal cortex is associated with inappropriate behavioral reinforcement/extinction, and according to fMRI studies of ASD patients, this region along with the medial-lateral fusiform gyrus and amygdala are hypoactive in the majority of ASD children when they are exposed to faces [40].

Inflammasome activation

The elevated TNF-α, interleukins, nuclear factors, and toll-like receptor levels seen in some subgroups of ASD children may reflect a chronic neuroinflammatory process through inflammasome activation of T-cells and microglia. Several subtypes of inflammasomes exist in the brain: NLRP1 inflammasomes (NLR Family Pyrin Domain Containing 1) have been found in human neurons and NLRP1 inflammasome activation is associated with age-related neuron damage, learning, and memory in rodents [41]. Human microglia contain NLRP3 and NLRC4 inflammasomes which both activate CASP1 to convert pro-IL-1β to active, secreted IL-1β which may create positive feedback for continued inflammation [42]. This process of NLRP3 inflammasome formation is depicted in Fig. 3 and demonstrates a common pathway for the previously discussed inflammatory biomarkers.

In this 2016 Italian study assessing mRNA expression by RT-PCR of white blood cells of 25 ASD patients aged 3–11 years, (n = 23 siblings and 30 controls), they found elevated NLRP3, IL1, and IL18 along with more LPS reactive leukocytes in ASD patients when compared to siblings and age matched controls. This study demonstrated that some ASD patients have serum leukocytes that are more likely than controls to form NLRP3 inflammasomes, and it correlates with magnitude of social impairment. The biomarkers TNF-α, NF-kB, IL-1β, IL6, and IL-17a are elevated in some subgroups of ASD patients, and may result in inflammasome formation in neurons, leukocytes, and microglia and may provide a molecular mechanism by which stress and inflammation during neurodevelopment predisposes the development of ASD [43]. Though increased inflammatory signaling is noted across many studies of ASD children, research rarely includes subjects younger than 2 years of age. Therefore, it is difficult to ascertain whether these inflammatory biomarkers are also perturbed during early neurodevelopment. Furthermore, the process by which inflammation contributes to ASD pathogenesis remains incompletely understood. The receptor for advanced glycation end products (RAGE) system is a potential mechanism with the promise of linking the inflammatory phenotype with the core symptoms of ASD.

Main text

The RAGE system

The RAGE system is a multi-ligand receptor within the immunoglobulin (Ig) superfamily that plays a role in inflammatory gene signaling. The gene for RAGE is encoded on the major histocompatibility complex locus class III region of chromosome 6p21.3 and is uniquely expressed in mammals [44, 45]. RAGE has multiple expressed isoforms and can be found as a membrane bound (mRAGE), soluble (sRAGE), and endogenously secreted (esRAGE) forms. Ligand binding to mRAGE results in multiple intracellular signaling cascades leading to proinflammatory NF-kB gene expression that culminates in increased IL-1, IL-6, and TNFα release [46]. Refer to Fig. 2 for a diagram depicting the RAGE system. The extracellular metalloproteinases ADAM10 and MMP can proteolytically convert mRAGE into sRAGE, while esRAGE is an alternatively spliced product of the RAGE gene [47, 48]. In contrast to mRAGE, sRAGE and esRAGE work by binding RAGE ligands and preventing them from interacting with mRAGE, thereby decreasing inflammatory gene expression.

Fig. 2
Fig. 2
Full size image

(A) RAGE mediated signaling on the intestinal endothelium leads to increased permeability in response to dietary AGEs. RAGE also serves as a transport mechanism for oxytocin across the intestinal-blood barrier. (B) RAGE is expressed on the endothelial surface of blood vessels interfacing at the blood-brain barrier. (C) Depiction of mRAGE multiligand binding that leads to an intracellular signaling cascade culminating in gene expression through the NF-kB, EGr-1, AP-1, and STAT3 pathways. Additionally, mRAGE serves as an oxytocin transport mechanism, allowing oxytocin transport across the blood-brain and blood-intestinal barriers. (D) The sRAGE/esRAGE is depicted as a “scavenger” for other mRAGE ligands, thereby preventing mRAGE signaling and allowing for a greater oxytocin transport capacity. (E) Systemic circulation containing endogenously secreted oxytocin from the posterior pituitary gland, sRAGE/esRAGE, and various RAGE ligands interacting and influencing RAGE signaling at the blood-brain barrier

Ligands that have been shown to bind to the RAGE protein include advanced glycation end products (AGEs), HMGB1, various S100 proteins (S100A4, S100A7, S100A8/A9, S100A12, S100B, S100P), oxytocin, amyloid β, C1q, Mac-1 and phosphatidylserine (PS) [46, 49]. Each of these ligands serves a role as either increasing inflammatory gene expression, moderating transport of biomolecules, or mediating immune cell migration and phagocytosis.

AGEs, HMGB1, and S100 proteins bind RAGE to stimulate inflammatory gene expression. AGEs are stable end products formed by the reaction of the dicarbonyl compounds methylglyoxal (MGO) and glyoxal (GO) with amino acids in proteins formed during glycolysis or lipid peroxidation [50]. In this way, the RAGE system serves as an intermediate inflammatory signaling process to excessive metabolic throughput. Evidence suggestive of this claim include the observation that chronic hyperglycemic states found in diabetic patients lead to AGE accumulation, RAGE activation, and an amplified inflammatory state that mediates the progression of diabetic complications [51]. The accumulation of AGEs has also been shown to induce peripheral inflammation during high fat-feeding in mice via a RAGE-dependent process [52].

Another RAGE ligand that stimulates inflammatory gene expression is High Mobility Group Box 1 protein (HMGB1). HMGB1 is a nuclear protein that binds DNA to serve as a non-histone architectural protein essential for DNA transcription, replication/repair, as well as a secretory product from immune cells (macrophages, monocytes, dendritic cells). When secreted, HMGB1 acts as a damage-associated molecular pattern (DAMP) molecule and helps to regulate inflammatory response [53].

The third major RAGE ligand group implicated in inflammatory gene expression are the S100 proteins. S100 proteins are a family of calcium binding cytosolic proteins that have a broad range of intracellular and extracellular functions. Intracellular S100 proteins functions include intracellular receptor signaling, membrane protein recruitment/transportation, transcriptional regulation, and DNA repair. Extracellularly, S100 proteins serve as alarmins and are crucial for regulating immune homeostasis, post-traumatic repair responses, and inflammation [54]. This RAGE ligand inflammatory system is depicted in Fig. 2.

In addition to the RAGE mediated expression of inflammation, mRAGE has been shown to be a transporter of certain molecules. For example, amyloid β peptide was found to be transported across the BBB via mRAGE and results in accumulation in the brain [55]. Oxytocin has also been shown to be transported by RAGE, the details of which will be expounded upon in a section to follow [56].

Another key aspect of the RAGE system is its chemotactic role implicated in immune cell migration and phagocytosis. More specifically, the RAGE system has been shown to interact with C1q and Mac-1 in a process that leads to recruitment of leukocytes and aids in C1q mediated phagocytosis [46, 57]. Similarly, RAGE has been demonstrated to enhance apoptotic cell clearance by binding with PS to induce Rac1 activation in macrophages [58].

Given RAGE is associated with such a wide range of physiologic processes, it comes as no surprise that this system is implicated in a host of disease processes. This diverse array of pathologies includes Alzheimer’s disease, multiple sclerosis, amyotrophic lateral sclerosis, diabetes, obesity, asthma, and inflammatory bowel disease. Mounting evidence suggests that ASD pathogenesis is in part due to aberrations in the RAGE axis occurring during the early prenatal and/or postnatal developmental periods. This fits mechanistically well with the aforementioned 3-hit model with RAGE signaling serving as an amplifier of the CDR.

Influence of prenatal RAGE on neurodevelopment

There is a paucity of direct mechanistic data linking prenatal RAGE dysregulation and ASD, however there is clear overlap between several known prenatal ASD risk factors implicating RAGE. Gestational diabetes mellitus (GDM), maternal obesity, and preeclampsia are three prenatal conditions with known RAGE-related pathophysiology that each correlate with an increased incidence of children later developing ASD. Fundamentally these three disorders predispose a CDR stress signaling disrupting early neurodevelopment. Though the directionality of the association between these conditions and ASD is fairly consistent across observational studies, the magnitude of the effect is somewhat inconsistent, likely due to confounding of overlapping risk factors such as maternal BMI, GDM, preeclampsia, and socioeconomic factors.

Metanalysis based on hundreds of international observational studies provides clear consensus that both gestational diabetes and pre-gestational diabetes are associated with an increased incidence of developmental delays, ASD, ADHD, and ID. A 2025 Chinese metanalysis of 202 observational studies including more than 56 million mother–child pairs from around the world also showed an increased incidence of ASD in children exposed to maternal diabetes (OR 1.25) [59]. A 2025 Mexican metanalysis compiling studies from all over the world showed increased risk of autism in the offspring exposed in utero to T2D (OR = 1.48; n = 3,141,255), T1D (OR = 1.73; n = 2,791,607), and GDM (OR = 1.31; n = 3,259,557) [60]. Severity of fetal exposure to hyperglycemia correlates with degree of neurodevelopmental impairment, with pre-gestational insulin-dependent T2DM mothers typically showing the largest ASD risk while non-insulin dependent GDM mothers show only a minor increase to ASD risk.

We propose that chronic hyperglycemia activates CDR and dysregulated RAGE further propagates this signaling, predisposing the increased risk of ASD. Advanced Glycation Endproducts (AGEs) formation is accelerated in hyperglycemia. Soluble RAGE (sRAGE) acts as a decoy receptor, binding AGEs and preventing them from interacting with membrane bound RAGE, thereby potentially mitigating the harmful oxidative stress effects associated with AGEs. An 18-year prospective study found that people with low sRAGE had an increased risk for later developing diabetes (hazard ratio 1.64 [95% CI 1.10–2.44]), coronary heart disease (1.82 [1.17–2.84]), and mortality (1.72 [1.11–2.64]) [61]. sRAGE is lower in those vulnerable to developing diabetes but once a patient develops chronic diabetes, a compensatory effect occurs resulting in increased sRAGE. A 2024 metanalysis compiling 32 studies (n = 4948) assessing sRAGE levels in humans with diabetes found that sRAGE levels are elevated in T1DM patients (SMD 0.45, CI: 0.16–0.73, P = 0.002) and T2DM patients with complications (SMD 1.59, CI: 0.77–2.41, P = 0.0001), while a decrease is observed in newly diagnosed T2DM patients (SMD-0.40, CI: -0.71- -0.09, P = 0.01) [62]. The increase in sRAGE levels is suggested to be a compensatory response to the chronic hyperglycemia, inflammation, and oxidative stress associated with diabetes. Low sRAGE is associated with vulnerability to excessive metabolic throughput, and high sRAGE is associated with a chronic compensatory response.

Maternal obesity appears to have a modest association with increased ASD risk though very few studies appropriately control against co-occurring GDM, T2DM, or preeclampsia. One 2016 meta-analysis comprising 8,403 ASD cases and 509,167 controls showed that children born to overweight and obese mothers have a 28% (RR 1.28) and 36% (RR 1.36) higher relative risk of developing ASD compared with children whose mothers were of normal weight [63]. This metanalysis, like most related metanalyses, did not control for the confounding variables of gestational hyperglycemia or preeclampsia. There is one 2020 Scandinavian study of 649,043 live born infants that successfully controlled for maternal diabetes status while assessing maternal obesity. Higher maternal prepregnancy BMI (moderate and severe obesity) was associated with an increased risk of children having any neuropsychiatric illness (HR 1.30), though non-diabetic severely obese mothers showed no statistically significant increased risk of having a child with ASD when compared to normal weight mothers. The association between maternal BMI and ASD risk appears to be modified by diabetes status: among severely obese mothers (BMI > 35), there is no significant increase in ASD risk without diabetes (HR 1.21, 95% CI 0.88–1.66) or with gestational diabetes (HR 1.29, 95% CI 0.95–1.74), whereas risk is significantly elevated in the presence of T2DM (HR 2.28, 95% CI 1.18–4.41) and is highest with insulin-treated pregestational diabetes (HR 5.93, 95% CI 2.81–12.52) [64]. In summary, the increased risk of having a child with ASD is null-to-modest in obese non-diabetic and obese GDM mothers. However, as hyperglycemia increases, an obesity-related additive ASD risk emerges whereby insulin-treated normal weight mothers have no increased risk of having a child with ASD but insulin-treated severely obese mothers show a nearly 6-fold increased risk of having a child with ASD.

We postulate that this additive risk effect is mainly due to RAGE pathway activation of the CDR during neurodevelopment. RAGE knock out mice remained significantly leaner than their wild-type counterparts when fed a western diet, exhibiting reduced body weight gain and smaller adipocyte size [65]. Obese humans show lower sRAGE levels than normal weight humans. sRAGE serum concentration negatively correlates with BMI, blood glucose, IL-6, leptin, and HDL [66]. The highest negative correlation coefficients (r) were estimated between sRAGE and BMI (r= − 0.7537) and proinflammatory IL-6 (r= − 0.7129). Interestingly, blood glucose showed the lowest correlational significance (r= − 0.3217) and serum insulin did not correlate with sRAGE levels at all. From these data we hypothesize that the well-documented pro-inflammatory state seen in obesity is related to dysfunctional RAGE signaling.

Hypertension during pregnancy can disrupt nutrient flow to the developing body and brain resulting in preterm birth, low birth weight, and a similarly increased predisposition to develop ASD. A 2018 metanalysis of 10 preeclampsia case-control and cohort studies (n = 1,166,307 cases and 397,504 controls) found the risk of ASD was 32% higher in offspring who had intrauterine exposure to pre-eclampsia compared with those not exposed (RR = 1.32, 95% CI 1.20–1.45) [67]. The effect of preeclampsia on ASD risk is duration-dependent, with earlier onset of preeclampsia increasing ASD risk. Of all three isolated conditions discussed in this section, preeclampsia most credibly shows an ASD association after controlling for the other comorbid maternal metabolic conditions.

Preeclampsia limits oxygen and nutritional flow to the fetus, increasing oxidative stress, hypoxia signaling, and exacerbating the RAGE-mediated CDR. Adults with chronic essential hypertension show lower sRAGE levels than normotensive controls [68]. Normal human physiological pregnancy demonstrates a change in the composition of sRAGE throughout pregnancy, where the ratio of esRAGE compared to all isoforms of sRAGE decreases throughout pregnancy. A 2011 prospective study of 87 healthy pregnant women and 28 pregnant women with preeclampsia found the esRAGE/sRAGE ratio normally decreases between 1st and 3rd trimester (P = 0.007 and P = 0.003). However mothers with preeclampsia showed significantly increased esRAGE/sRAGE ratios compared to controls (P = 0.005 and P < 0.001) [69]. A 2025 study of RAGE pathway ligands, sRAGE, HMGB1, S100B comparing placental samples from mothers with preeclampsia and with normal pregnancy found significant upregulation of these pathway molecules in preeclamptic placentas compared to control [70].

In summary, fetal exposure to gestational diabetes mellitus (GDM), maternal obesity, and preeclampsia result in an additive risk for developing ASD and other neurodevelopmental disabilities. Prolonged CDR through proinflammatory RAGE-signaling provides a unifying and potentially modifiable pathophysiological mechanism that could be modulated to reduce severity and prevalence of ASD and neurodevelopmental disabilities in appropriately selected subgroups.

Postnatal evidence of proinflammatory RAGE signaling in ASD

The previously discussed proinflammatory gene expression ligands AGEs, HMGB1, S100 family have preliminary evidence of being elevated in ASD [71,72,73,74]. Advanced glycation end-products Nε-carboxymethyl-lysine and Nω-carboxymethyl-arginine were found to be significantly increased in ASD patients (n = 38) compared to normal developing controls (n = 31) in one study [71].

In a study of 18 ASD patients, significantly reduced levels of esRAGE (P = 0.0023) and elevated concentrations of S100A9 (P = 0.0012) were demonstrated when compared to age matched healthy controls [72]. Increased concentrations of HMGB1 have been found in patients with ASD when compared to normal developing controls (1.25 ± 0.84 ng/mL versus 1.13 ± 0.79 ng/mL, respectively, p = 0.039) [73]. Additionally, a metanalysis of 822 participants showed that there was a statistically significant increase in S100B peripheral blood levels [74].

A proteomic study of ASD children aged 4–6 (n = 69) compared to typically developing children demonstrated a significant increase in serum C1q in the ASD cohort [75]. This is taken to signify the systemic proinflammatory state in ASD. Conversely, C1q has been shown to have a significantly decreased expression in the postmortem brains of ASD individuals [76]. C1q has been demonstrated to play a pivotal role in synaptic pruning and thus decreased C1q brain expression would impair this process [77]. The exact way C1q interacts with the RAGE system to give rise to these findings remains unknown.

Finally, another study demonstrated a significant increase in total secreted amyloid precursor protein (APP) in autism patients with a regressive phenotype [78]. RAGE has also been demonstrated to cleave APP to amyloid β peptide by modulating β- and γ- secretase [79]. It is therefore postulated that APP increases in autism patients due to decreased available RAGE to participate in APP metabolization.

In conclusion, there is evidence of dysfunctional RAGE system components and alterations in RAGE ligands in patients with ASD. As detailed above, ASD patients demonstrate several differences in RAGE signaling; elevations in inflammatory gene expression ligands (AGEs, HMGB1, S100 family), decreases in esRAGE, regionally altered C1q, and impaired APP metabolism. Regardless, these changes alone do not adequately explain RAGE system contributions to the pathogenesis of the ASD phenotype. For a further understanding of this process, attention must be turned to the oxytocinergic system.

The oxytocinergic system in ASD

Oxytocin is a biomolecule with substantial evidence for its contribution to ASD. Oxytocin is a nonapeptide hormone and neuropeptide that plays a critical role in various physiological and behavioral functions, including lactation, uterine contraction, and social bonding. Oxytocin, vasopressin, and their g-coupled receptor receptors appear to have evolved from an ancestral gene duplication event over 600 million years ago. Many invertebrates, and all vertebrates utilize the system for reproductive behaviors and memory potentiation. Mammals additionally utilize oxytocin in gestation, milk production, and pair bonding [80]. Produced primarily in the hypothalamus, oxytocin is released into the bloodstream via the posterior pituitary gland. Beyond its peripheral roles, in humans there is additional evidence to support oxytocin’s role in social development and even severity of ASD core symptoms. This process may occur through reduced oxytocin receptor density in the ventral pallidum mesolimbic dopamine reward pathway of ASD patients, resulting in a reduced experience of OXT-mediated social reward in ASD [81].

Although oxytocin receptor expression occurs prenatally, more critical evidence for oxytocin’s role in social neurodevelopment comes from the perinatal and early postnatal periods [82]. Parturition is a major developmental transition during which oxytocin not only facilitates labor, but also exerts neuroprotective, analgesic, and immunomodulatory effects. Importantly, oxytocin signaling during delivery has been shown to promote the transient shift of GABA signaling toward inhibition by lowering neuronal intracellular chloride, a process linked to regulation of the chloride transporters NKCC1 and KCC2 [83]. Converging evidence further suggests that ASD is associated with altered chloride homeostasis and impaired maturation of GABAergic signaling through the imbalance of NKCC1/KCC2 [84]. Taken together, these findings support the hypothesis that disrupted oxytocin signaling could contribute to failure or delay of the normal GABAergic inhibitory switch, thereby promoting aberrant circuit formation and excitatory/inhibitory imbalance relevant to ASD pathophysiology.

The oxytocin (OXT) and oxytocin receptor (OXTR) genes are associated with the ASD phenotype according to the SFARI autism gene database. There is also some evidence that single nucleotide polymorphisms (SNPs) of OXTR may be implicated in ASD. A 2015 metanalysis of 3,941 ASD patients found 16 OXTR SNPs, with 4 SNPs that were significantly associated with ASD [85]. A 2023 Polish study assessed 58 ASD boys with IQ > 90 and found 2 OXTR SNPs and 2 AVP1A SNPs that were not associated with ADOS-2 performance, though the rs5372 genotype of OXTR was associated with increased severity of clinical picture of social cognition disorders in reading mind in the eyes test (RMiE) and empathy quotient (EQ) studies [86]. Overall, polymorphisms to OXTR may be associated with some social cognition deficits, but these do not appear to be specific only to the pathogenesis of Autism.

A 2021 metanalysis of 28 studies (n = 726) of intranasal oxytocin administration to adults with ASD noted moderate improvements in questionnaire assessments of social functioning, eye gaze, emotion recognition, and prosocial behaviors, while there was no evidence for improvements to non-social core ASD criteria [87].

A 2021 Swiss systematic review of 31 studies of oxytocin levels in children with ASD vs. controls found that non-ASD children had significantly higher serum levels of OXT than ASD children [88]. Interestingly, this discrepancy does not seem to persist into adulthood, which raises questions about the temporal dynamics of oxytocin levels and their relevance to the ASD phenotype.

Lower oxytocin levels in younger children with ASD have been correlated with more severe social deficits, but similar social deficits are also seen in neurotypical children with low serum OXT. A 2014 study of children (ages 3–12) with ASD, their unaffected siblings, and unrelated neurotypical controls (n = 193) found no evidence to support that OXT deficit directly causes ASD, but they did find a strong positive correlation between serum OXT and predicted theory of mind and social communication performance in ASD and neurotypical children [89].

In terms of the involved neuroanatomy, there is a notable 2018 study of AVP1A and OXTR binding in 44 pediatric and adult human brain samples (half ASD and half neurotypical) that showed significantly lower ventral pallidum oxytocin receptor density in ASD brains (see Fig. 3). The ventral pallidum is part of the mesolimbic dopamine reward pathway and is important in processing reinforcing stimuli. The authors speculated that reduced OXTR levels in the ventral pallidum may be related to a reduced experience of OXT-mediated social reward in ASD [81].

Fig. 3
Fig. 3
Full size image

Effect of autism spectrum disorder on OXTR binding in five regions of the human brain sampled. OXTR binding is significantly decreased in the Ventral Pallidum (VP) and significantly increased in the Nucleus Basalis of Meynart (NBM) in ASD patients compared to control. Adapted from data from Freeman, 2018. Created using Biorender.com

This same study also found OXTR density changes with age, with the greatest difference between TD and ASD OXTR density noted in the 2–5 year old period, with this difference in oxytocin binding density variably conserved across patient lifetime. This could be taken to mean that OXTR density is more influential during infancy and early toddlerhood, which begs the question whether oxytocin administration could overcome the OXTR density differences seen in ASD.

In one 2021 Chinese systematic review of 28 studies of intranasal oxytocin in adults with ASD (n = 726 ASD patients), the researchers found that OXT therapy did not reliably improve nonsocial features of ASD but may have modest beneficial effects on social functioning [87].

In children with ASD, a 2023 Chinese metanalysis reviewed 5 major trials of intranasal OXT in ASD children with a total of 486 ASD children receiving OXT (mean age of 8.97 ± 3.98 years). The systematic review found a modest improvement to social deficits with intranasal OXT administration only after excluding nearly half (n = 139) of the ASD children in the NEMJ 2021 Sikich, et al. trial, because it was not double blinded [90]. The Sikich, et al. trial was randomized, placebo controlled, and spanned multiple highly reputable teaching hospitals to include 290 participants, and this study showed no improvement of ASD core criteria with OXT administration over 24 weeks [91]. In children and adults with ASD, there may be a modest to null improvement to social functioning with intranasal OXT administration.

The modest to null effect of intranasal OXT in the treatment of children with ASD may in part be due to the late age at which the therapy is initiated. The rodent models demonstrating complete rescue from ASD symptoms with OXT administration often initiate therapy in the newborn pup period, which has yet to be studied in pediatric participants. In human children, the greatest difference seen in oxytocin receptor binding density seen in brain tissue is observed in the 2–5 year old cohort, which implies that more robust response to oxytocin would occur in infants and young toddlers rather than school aged children.

From the same nonameric peptide ancestor as oxytocin, vasopressin is also produced in the hypothalamus, and has three main functions: (1) increasing solute-free water reabsorption from the kidney, (2) constricting arterioles to raise blood pressure, (3) social behaviors such as sexual motivation, pair bonding, and maternal responses to stress. Vasopressin receptors AVP1A and AVP1B mutations are implicated in the development of ASD according to a 2023 search of the SFARI Database. Serum vasopressin levels may predict theory of mind scoring in some ASD children, though other studies have had difficulty corroborating this [92]. A randomized placebo-controlled pilot study from Stanford University in 2019 provides preliminary evidence that vasopressin nasal spray reduces social impairment in some children with ASD [93].

In summary, there is evidence for serum oxytocin levels in young ASD children being lower than controls, and that serum OXT levels correlate with social communication testing across all groups of children. Moreover, there is human autopsy evidence that the ventral pallidum has reduced OXTR levels in ASD brains, and that this difference is greatest in ASD children between 2 and 5 years old, which is incidentally the most common age to diagnose ASD and for ASD regression to occur. OXT and vasopressin intranasal supplementation may have a modest effect on ameliorating core symptoms of ASD, but more studies with younger participants are needed to confirm this.

Implicated microbiota in ASD

Several studies have noted the differences in gut microbiome profiles between patients with ASD and neurotypical individuals. In one study, Bacteroides fragilis was decreased in autism spectrum disorder patients [94]. Due to this bacterium having the enzymatic capability of converting the serotonin precursor tryptophan into other indoles, it raises the question whether this difference plays a pathogenic role in ASD. ASD patients have also been shown to have less gut microbiome diversity with a sparsity in Prevotella, Coprococcus, and unclassified Veillonellacea species, when compared to neurotypical individuals [95]. Increased gastrointestinal Candida albicans has also been demonstrated in patients with autism [96]. Some of these microbiome differences have been associated with altered inflammatory and even oxytocin signaling.

As previously discussed, inflammation during neurodevelopment and elevated inflammatory biomarkers are associated with ASD pathogenesis and this inflammatory process has been linked to microbiome profiles in some ASD patients. In one study of children and teenagers with ASD, there was a reduction in beneficial bacterial species Bacteroides and Lachnospiraceae in the setting of significantly higher plasma levels of IL-2, IL-4, IL-5, IL-6, IL-10, TNF-α, TNF-β and INF-γ. Furthermore, it was found that relative abundances of Lachnospiraceae taxa were negatively correlated with IL-6 and INF- γ levels and associated with patients with severe ASD as defined by CARS [97]. There is preliminary evidence that gut dysbiosis contributes to the systemic inflammation seen in ASD patients.

Certain gut microbiome profiles have also been implicated in oxytocin signaling and ASD pathogenesis. It has been demonstrated that oxytocin is secreted from enterocytes in response to Limosilactobacillus reuteri [98]. In a cohort of 39 ASD patients with 44 controls, there was a significantly lower serum oxytocin level in ASD patients that correlated with increased concentrations of pathogenic gut microbiota (Enterobacter, Citrobacter, Escherichia, Shigella) and lower levels of butyrate producing bacteria [99].

Unfortunately, the clinical significance of these differences remains unclear, as other variables such as diet influence microbiome profiles, making control of these ASD microbiome studies difficult. In one cohort of 247 patients, differences in gut microbiome composition were found to be linked with autism-related dietary preferences (i.e. less diverse diet) and stool consistency rather than the diagnosis of autism itself [100]. Despite these limitations, there are differences in the gut microbiomes of ASD patients, and further studies are needed to ascertain the extent to which the microbiome influences ASD pathogenesis. To better understand how the gut microbiome affects neurodevelopment, one must consider the mechanisms by which microbial metabolites influence microglial function.

The gut metabolome in ASD

Prior sections provide evidence for altered inflammation, RAGE axis, and oxytocin signaling in ASD compared to controls. There is emerging evidence that these signaling mechanisms alter microbial metabolites in a way that predisposes autistic neurodevelopment [101]. Gut microbiome metabolism of short chain fatty acids, neurotransmitter precursors, and lipopolysaccharides is increasingly linked with ASD.

Short chain fatty acids

In one Chinese study of 30 ASD and 30 normal controls, lower levels of fecal acetic acid and butyrate, along with higher levels of fecal valaric acid, were found in ASD subjects. This finding corresponded to decreased abundances of certain butyrate producing colonies (Ruminococcaceae, Eubacterium, Lachnospiracae, and Erysipelotrichaceae) [102]. Dysregulated gut butyrate in ASD could compromise gut permeability, increasing systemic bacterial metabolic exposure and altering neurodevelopment.

Propionic acid is a fermentation product of certain bacteria, in particular Clostridia, Desulfovibrio, and Bacteroidetes, that has been reported to be higher in concentration in stool samples from patients with ASD compared to controls [103]. Furthermore, intraventricular injection of propionic acid was shown to lead to an autistic phenotype in mice [104]. This preliminary data suggests that increased propionic acid metabolite translocation occurs in some patients with ASD, and if translocated to the brain, can cause an autistic phenotype per mouse model evidence.

Wang et al. reported that all fecal SCFAs levels except caproic acid were elevated in patients with autism [105]. In conclusion, there is preliminary evidence that altered butyrate and elevated propionic acid levels are seen in ASD patients, and there is some mouse model evidence that these changes can result in an autistic phenotype.

Indoles, p-crestol, GABA

Another form of bacterial metabolism implicated in ASD pathogenesis are the metabolites formed by bacterial fermentation of proteins: indoles, GABA and p-crestol.

The indole synthesis pathway tryptophan and the neurotransmitter serotonin has been implicated in the pathogenesis of autism. Additionally, altered indole gut metabolism is seen in patients with autism. In one Italian urine metabolomic study, tryptophan was preferentially converted to the toxic metabolites xanthurenic and quinolinic acid via the kynurenine pathway which is theorized to contribute to cognitive dysfunction in ASD (depicted in Fig. 4) [106]. Elevated levels of serotonin have been found in the serum of ASD patients, with one study reporting as much as a 25% increase in systemic serotonin concentrations compared to normal developing controls [107]. Abnormal indole metabolism, including potentially toxic buildup of non-serotonergic metabolites, and excessive levels of serum serotonin has been seen in ASD patients, and may help drive the autistic phenotype.

Fig. 4
Fig. 4
Full size image

Depiction of the kynurenine pathway. (A) Tryptophan may be converted to serotonin or kynurenine within dendritic cells by indoleamine 2,3-dioxygenase (IDO). (B) Kynurenine can be further metabolized in microglia or leukocytes by kynurenine 3-monoxygenase (KMO) to form the neurotoxic quinolinic acid. (C) Alternatively, kynurenine can be converted by kynurenine acetyltransferase (KAT) in astrocytes to the neuroprotective kynurenic acid [108]. Created using Biorender.com

There are reports of GABA alterations in patients with autism as well. Several studies report lower levels of serum GABA in ASD patients [109]. Bifidobacterium is involved in the synthesis of GABA, and it is proposed that alterations in this microbiota contribute to the alterations in serum GABA levels seen in ASD [110, 111]. Lactobacillus reuteri, brevis, and rhamnosus have also been shown to affect central GABA signaling mediated through the vagus nerve [111,112,113,114]. In a murine model, L. rhamnosus reduced GABAAα2 mRNA expression in the prefrontal cortex and amygdala, but increased GABAAα2 in the hippocampus, while vagotomized mice did not demonstrate these changes [111]. L. reuteri was found to modify GABA receptor expression, as well as oxytocin expression in Shank3 KO mice, and its relative abundance modified social and repetitive behaviors [114].

P-crestol is another bacterial metabolite produced in the gut and has been reported to be elevated in the urine and feces of ASD patients [115]. The effects of p-crestol are thought to be due to its impact on CNS dopamine levels, thereby promoting an autistic phenotype. P-crestol is produced in the gut and requires increased gut permeability to cross into the systemic circulation. P-crestol has been shown to have an inhibitory effect on the enzyme dopamine-β-hydroxylase, thereby altering the neurotransmitter synthesis in the brain. In one BTBR mouse model, P-crestol was shown to exacerbate autism-like behaviors and demonstrated associated brain dopamine metabolism changes [116].

Lipopolysaccharide (LPS)

Lipopolysaccharide (LPS) levels have been reported as elevated in patients with autism compared with controls. LPS is found on the outer membrane of gram-negative bacteria and produces an acute inflammatory response. LPS has been shown to activate microglial cells leading to GABAergic signaling defects [117].

In another mouse model, maternal LPS exposure led to activation of the TLR4 signaling pathway and resulted in microglial activation. The wildtype offspring of LPS exposed mothers showed lower social and self-exploration behavior, and greater anxiety and repetitive behaviors, however the TLR4 -/- did not [118]. This shows that LPS exposure during neurodevelopment promotes the autistic phenotype as mediated through the TLR4 pathway.

Altogether, SCFAs, bacterial protein metabolites (Indoles, p-crestol, GABA), and LPS provide three examples of how gut dysbiosis contributes to impaired gut-blood permeability and is contributory to the promotion of the autistic phenotype.

Microglia in ASD

Microglia, the resident immune cells in the brain, respond to pathogens, clear neuronal and non-neuronal debris, and secrete cytokine and chemokine specific soluble factors to modulate neuroinflammation [119,120,121,122,123]. In children with ASD, microglia mediated neuroinflammatory response coupled with a failure of microglia to regulate synapses seems to play a crucial role in shaping neuronal connectivity [124]. These differences at the cell level culminate in differences in gross neuroanatomy, including sustained brain overgrowth, aberrant neuronal hyperconnectivity, altered synaptic pruning, and neural circuit abnormalities seen in ASD children [119, 125,126,127,128]. Recent studies have demonstrated the role of microglia in inducing neuroinflammation leading to behavioral exacerbations associated with ASD [129, 130].

As was previously discussed, LPS is produced by the gut microbiome, can be translocated into serum in the setting of a leaky blood-gut barrier, and serum LPS levels are known to be increased in ASD patients [131]. LPS activates the RAGE system to increase systemic inflammatory response. According to a Chinese 2016 mouse study, oxytocin pretreatment followed by LPS–stimulation reduced microglial TNF-α and IL-1β production when compared to control microglia not exposed to oxytocin pre-treatment [132]. A 2023 Korean study found that LPS-stimulated microglia induce GABAergic synaptic deficits in mouse hippocampus which were accompanied by cognitive impairment [117]. Altogether, these studies implicate oxytocin in the attenuation of CNS inflammatory response through microglia modulation.

Another proposed mechanism for gut microbiota modulating microglial functioning arises from SCFA signaling. In vitro cultured human THP-1 monocytic cells and differentiated human HL-60 myelomonocytic cells serving as a microglia model treated with multiple SCFAs (acetate, propionate, butyrate, formate, and valerate) were found to have decreased secretion of interleukin (IL)-1β, monocyte chemoattractant protein (MCP)-1, tumor necrosis factor (TNF)-α, and cytotoxins by immune-stimulated THP-1 cells [133]. In another model, mice fed inulin were shown to have increases in systemic SCFAs, and when injected with LPS, isolated microglia from these mice demonstrated decreased ex vivo secretion of TNF-α when compared to microglia from mice not fed inulin [134]. Taken together, these results suggest SCFAs serve a role to decrease microglia mediated neuroinflammation.

In conclusion, differential microglial activation mediated by oxytocin, LPS, and SCFAs provide a mechanism through which excessive neuroinflammation contributes to the ASD phenotype. Furthermore, identification of additional biomarkers, such as the related RAGE system, could lead to novel therapies targeted to shift or alter the infant microbiota potentially resulting in microbiome-immune homeostasis to reduce ASD prevalence.

RAGE-oxytocin dynamics in the brain and gut

The mRAGE signaling system is important in the transport of oxytocin across both the blood brain barrier and the intestinal endothelial surface [56, 135, 136]. In a mouse model, vascular RAGE has been demonstrated to transport oxytocin across the blood brain barrier to facilitate maternal pair bonding. In mice, when exogenously administered oxytocin was given, it was not transported into the brain in RAGE-/- mice, which resulted in dams demonstrating impairment in maternal-infant bonding [56]. Another study demonstrated that intranasal oxytocin administration led to increased oxytocin in the extracellular space of the medial prefrontal cortex of wild type mice but was not observed in RAGE-/- mice. In the same study, inclusion of esRAGE did not alter the transport of oxytocin, implicating mRAGE in oxytocin transport independent of esRAGE levels. Additionally, male mice overexpressing esRAGE had slightly higher oxytocin transport in the brain, albeit not statistically significant [135]. Similarly, oxytocin was found to also be transported across the intestinal epithelium into the blood in a RAGE dependent process in the mouse model [136]. Refer to Fig. 2 for a graphical depiction of the different roles of the RAGE system. It is still unclear whether this same process occurs in humans therefore further research is needed to determine if this mechanism is conserved and to what extent the RAGE system plays in the development of social behavior.

Gut dysbiosis in children with ASD may either be implicated in ASD pathogenesis or the severity of the autistic phenotype. RAGE signaling has a bidirectional relationship with gut microbiota via LPS signaling. Evidence for this relationship includes a study that demonstrated LPS directly binds to RAGE with a subsequent NF-kB dependent inflammatory response in a murine model of septic shock. The injection of soluble RAGE significantly reduced the LPS-induced proinflammatory response and tissue damage [137]. RAGE signaling does influence gut microbiota and is seen with differential species survival related to AGEs intake. Increased AGEs intake has been associated with relatively increased Lawsonia, Parabacteroides, and Ruminococcus and relatively reduced populations of Lactobacillus, Prevotella, Anaerostipes, and Candidatus arthromitus [138]. This suggests that amplified intestinal RAGE signaling mediated by AGEs intake leads to differential gut microbiome populations, some of which have been implicated in ASD patients (i.e. Lactobacillus, Prevotella) [95, 98].

Interestingly, there are reports that suboptimal breastfeeding such as late initiation breastfeeding, non-intake of colostrum, prelacteal feeding and bottle feeding increase the risk for the development of autism [139]. Moreover, the risk of autism decreases in a dose dependent fashion with periods of exclusive and/or continued breastfeeding. There is even evidence that consumption of breast milk can decrease the risk of autism in Fragile-X patients by 1.7-fold [140]. The mechanism by which breastfeeding reduces ASD prevalence is unknown. However, certain glycation end products in breastmilk, such as the milk proteins β-lactoglobulin and casein, significantly increased fermentability of proteins by Lactobacillus and Bifidobacterium, thereby promoting their growth [94]. This process has potential to explain some of breastfeeding’s neuroprotective effects. Further studies are needed to determine the role of milk AGEs on RAGE signaling in humans and how this contributes to formation and maintenance of the gut microbiome.

Conclusions

ASD is increasing in prevalence and there is more urgency than ever before to understand its underlying pathophysiology. Both prenatal and postnatal inflammation have been associated with the development of ASD, specifically alterations in TNF-α, NF-kB, IL-1β, IL6, and IL-17a. This inflammation has been implicated as part of a more global CDR with RAGE signaling serving as an amplifier. A complete understanding of the biochemical processes that influence RAGE activation and ASD pathogenesis remain unknown, necessitating further study.

There are additional challenges impeding the translational progress of this autism research. Innate heterogeneity in oxytocin responsiveness, context-dependent RAGE activation, and a paucity of reliable biomarkers reflecting the neuroimmune status are factors that make a comprehensive mechanistic understanding difficult to garner. Additionally, these processes are dynamic over the course of different developmental periods. Prenatal network disruption predisposes an increased vulnerability to ASD in the postnatal period. There is a significant temporal component to the efficacy of RAGE modulation, because subpopulation vulnerability to develop ASD often begins during early gestation.

It is imperative future work prioritizes longitudinal design to capture critical developmental periods implicated in ASD. Defining when and how the neuroendocrine and inflammatory systems converge to contribute to ASD development will be critical for informing targeted and developmentally appropriate interventions.

In summary, we propose a common pathophysiology of ASD based on interactions between the oxytocin, inflammatory, and gut microbiome systems that is mediated through the RAGE axis. Future research is paramount to investigate how the RAGE system interfaces with the CDR, oxytocin, and the gut microbiome to explain the pathophysiology of ASD in humans. Unraveling the impact of RAGE signaling on neurodevelopment has the promise of creating precision interventions for ASD subpopulations.