Journal of Neurology Research, ISSN 1923-2845 print, 1923-2853 online, Open Access
Article copyright, the authors; Journal compilation copyright, J Neurol Res and Elmer Press Inc
Journal website https://jnr.elmerpub.com

Review

Volume 16, Number 4, December 2026, pages 210-237

Neuroimmune Mechanisms in Autism Spectrum Disorder: Current Evidence, Pathophysiological Insights, and Therapeutic Implications

Figure

↓  Figure 1. Integrated, evidence-graded model of candidate neuroimmune pathways in autism spectrum disorder (ASD). The schematic summarizes the review from upstream maternal, environmental and host factors (top tier), through immune perturbations, neuroimmune and central nervous system (CNS) processes, and neurodevelopmental and circuit effects, to behavioral and clinical phenotypes (bottom tier). Arrow style indicates the strength of support for each link in the author’s appraisal (Section “Review Approach and Evidence Appraisal”): solid arrows, comparatively stronger support combining human data with convergent mechanistic evidence; dashed arrows, moderate or indirect support, such as a human association without an established mechanism or a mechanism demonstrated only in animal models; dotted arrows, preliminary or hypothesized links. Links are drawn as parallel, partially overlapping candidate pathways rather than a single causal sequence; any pathway may operate in some individuals and not in others, and none has been shown to be necessary or sufficient for ASD in humans. A note beneath the arrow key restates that the pathways are partially overlapping candidates. The figure contains no patient data and is drawn with editable Word shapes. 4EPS: 4-ethylphenylsulfate; BBB: blood–brain barrier; CSF: cerebrospinal fluid; FRAA: folate receptor alpha autoantibody; IL: interleukin; MAR: maternal autoantibody-related; MIA: maternal immune activation; Th17: T helper 17 cell; TSPO: translocator protein.
Figure 1.

Tables

↓  Table 1. Author-Derived Framework for Interpreting Evidence in ASD Neuroimmunology Informed by the Studies Reviewed in This Article
 
Evidence classRepresentative studiesPrincipal strengthPrincipal limitationCausal inference supported
All entries represent the author’s appraisal rather than findings reported in the cited studies.
Postmortem neuropathology[1–4, 41, 42, 44]Direct CNS tissue evidenceSmall samples; agonal and medication confounds; late-stage samplingPresence of a process; not its timing or cause
Bulk tissue transcriptomics[13–15, 33]Unbiased, genome-wideTissue-level, not cell-type resolved; cannot distinguish cause from responseTissue-level association
Single-cell and single-nucleus genomics[12]Cell-type-resolvedFew donors; postmortem; cannot distinguish cause from responseCellular localization of the signal
In vivo TSPO-PET[45–48]Living brain; repeatableTSPO not cell-specific; small samples; differing radioligands and quantification; directionally heterogeneousAltered TSPO signal; not its cellular source or cause
Peripheral biomarker studies[23–26, 38, 62–64]Scalable; accessibleCorrespondence with CNS not established; assay heterogeneity; frequent high risk of biasAssociation only
CSF studies[1, 65–67]Proximal to CNSSmall samples; invasive; matched controls often absent; few paired serum–CSF designsAssociation; within-person compartment comparison
Epidemiological cohorts[28–31, 68–71]Population scale; temporal orderingResidual confounding not excluded by designRisk association; exposure timing
Rodent MIA and autoantibody models[17–19, 49, 72–75, 81]Experimental control; mechanismConstruct validity limits; species differencesSufficiency of a pathway in the model
Non-human primate models[76–78]Neurodevelopmental proximityVery small samples; costTranslational plausibility
Randomized controlled trials[22, 111, 116, 118, 119, 121, 122, 124, 127]Controlled test of intervention effectsMany studies are small, short, or adjunctive; behavioral endpoints predominateIntervention effect if adequately powered; mechanism requires demonstrated target engagement

 

↓  Table 2. Glial Findings in ASD: Convergence and Divergence
 
FindingModalityDirectionRegionsLimitation/author appraisal
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
Astroglial and microglial reactivity; elevated MCP-1, TGF-β1; no lymphocytic infiltrate [1]Postmortem + CSF↑Cerebellum, ACC, midfrontalInnate, not adaptive; n small
Increased microglial density; activated morphology [2]Postmortem↑DLPFC, esp. white matterRegion-specific
Abnormal microglia–neuron spatial organization [3]PostmortemAlteredDLPFCSpatial relationship rather than cell number
Increased microglial density [41]Postmortem↑Fronto-insular and visual cortexRegions differ from DLPFC studies
Elevated GFAP [42]Postmortem↑Frontal, parietal, cerebellarMarker-dependent
Region-dependent glial marker expression [4]Postmortem (transcript)↑/mixedPFC (astrocytic and microglial ↑); cerebellum (astrocytic ↑ only)Partial discordance; transcript level
NF-κB p65 upregulation in neurons and glia [43]Postmortem↑Orbitofrontal cortexMost pronounced in microglia; n = 9 per group
Higher [11C]-(R)-PK11195 binding [45]PET, in vivo↑Multiple; cerebellar predominanceMen; first-generation ligand
Lower [11C]PBR28 signal [46]PET, in vivo↓Multiple cortical and subcortical regionsMen; relative uptake measure
No overall difference in [18F]FEPPA binding [47]PET, in vivoNo difference (↓ post hoc)Prefrontal, temporal, cerebellar, anterior cingulateMixed sex; post hoc exclusion
Higher [11C]PBR28 signal in women [48]PET, in vivo↑ (regional)Midcingulate cortex; spleniumWomen; pilot sample
Immune-glial module upregulation [13–15]Bulk transcriptomics↑CortexTissue-level; cannot separate cause/response
Cell-type-specific glial signatures [12]Single-cell genomics↑CortexSupports cellular localization

 

↓  Table 3. Cytokine Abnormalities in ASD: Synthesis Across Evidence Levels
 
MediatorPeripheral directionConsistencyCentral or tissue-level evidenceAuthor appraisal
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”). Cross-diagnostic group differences do not constitute validated individual-level diagnostic accuracy.
IL-1β↑Replicated across meta-analyses [23, 24, 26]Not establishedStrongest pooled peripheral signal; non-specific innate marker
IL-6↑Replicated [23, 24, 26]; attenuated after excluding high risk-of-bias studies [24]Frontal cortex tissue [44]Central in murine MIA [49]
IL-8/CXCL8↑Replicated [23, 24, 26]Frontal cortex tissue [44]; ↑ CSF in one adult study [67]Non-specific
MCP-1/CCL2↑ [23]; not significant in largest pooling [24]Inconsistent; very large heterogeneity [24]↑ brain and CSF [1]; ↓ CSF in adults [67]Directionally inconsistent across compartments
TNF-α↑ [24]; not ↑ in earlier pooling [23]Sensitive to study quality [24]Frontal cortex tissue [44]Higher in ASD than ADHD [26]
IL-2Not consistently ↑ versus controlsHigher in ASD than ADHD [26]NoneCross-diagnostic contrast only
IL-17A↑ in subgroups [57] and one cohort [62]Not significant on pooling [24]↓ IL-17 mRNA in frontal cortex [89]Central in murine MIA [18]; human support weak
IL-23↓ [55]; no difference [24, 62]Low↓ IL-23 mRNA in frontal cortex [89]Not aligned with the murine IL-17A axis
IFN-γ↑Replicated [23, 24, 62]Frontal cortex tissue [44]Co-elevated with Th2- and Th17-associated mediators [62]
IL-4/IL-10IL-4 ↑ [24, 26, 62]; IL-10 ↓ [25], ↑ [62], no difference [24]InconsistentNoneDirection varies by synthesis
IL-34/CSF-1↑ [62]Single cohortNone; measured in serum onlyNot evidence of microglial activation
TGF-β1↓ [23]; no difference [24, 25, 62]Low↑ in brain tissue [1]Earlier pooled reduction not reproduced
CXCL1/MIF↑ [24]Few studies; not robust to quality restriction [24]NoneRequires replication
IL-1RA↓ [25]; ↑ only in quality-restricted analysis [24]InconsistentNoneDirection depends on study selection

 

↓  Table 4. Maternal Immune Activation: Evidence Hierarchy and Translational Status
 
LevelKey findings (reference)Strength (author appraisal)Unresolved issue (author appraisal)
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
Human registry [68, 69]Infection hospitalization associated with ASD; non-specific across pathogens/trimestersPopulation scale; temporal orderingModest effect; residual confounding not excluded by these designs
Human prospective [70]Prenatal fever, dose-dependent, second-trimester predominanceProxy for cytokine response, not pathogenFever ascertainment; antipyretic confounding
Human negative [71]No robust influenza infection or vaccination associationLarge, well-adjusted; reduces specificity claimsDoes not exclude non-influenza inflammation
Rodent mechanism [49, 72–74]IL-6 necessary and sufficient in the model; fetal brain transcriptional response; persistent offspring immune changeExperimental causality within the modelExperimental stimulus and fixed gestational timing limit construct validity
Rodent pathway [18, 19, 75]Maternal Th17/IL-17A required; microbiota-dependent; reversible via S1DZ circuitMost fully specified experimental pathway reviewed hereNot demonstrated in humans
Non-human primate [76–78]Behavioral, brain growth, cognitive and immune alterationsClosest neurodevelopmental proximity of the models reviewedVery small samples; fixed-timepoint exposure

 

↓  Table 5. Autoantibody Findings in ASD: Mechanistic Maturity
 
FindingHuman evidence (reference)Animal corroborationPhenotypic anchorMaturity (author appraisal)
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
MAR autoantibodies to fetal brain proteins [16, 79]Defined antigens; reported specificity; risk-assessment estimatesFetal intraventricular IgG injection [81]; endogenous-antigen model [17]Abnormal brain enlargement [80]Comparatively well-characterized candidate subgroup; human causality not established
Folate receptor alpha autoantibodies (FRAAs) [87, 88]Substantial prevalence reported in studied samples; assay characteristics described [88]Not applicable (proposed mechanism is transport blockade)Proposed impairment of cerebral folate transportModerate; therapeutically testable
Child brain-reactive antibodies [83, 84]Detected; heterogeneous targetsNoneNone establishedLow
Anti-neural antigen antibodies with mimicry claims [85, 86]Immunoassay cross-reactivityNoneNone establishedLow—methodologically fragile

 

↓  Table 6. Gut–Immune–Brain Axis: What Each Evidence Tier Can and Cannot Establish
 
TierRepresentative evidence (reference)EstablishesDoes not establish (author appraisal)
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
Human descriptive [93–101]Compositional and metabolomic differences; mucosal immune correlatesAssociationCausal direction; diet and medication confounding
Barrier markers [91, 92]Mixed brain tight-junction expression; intestinal tight-junction changes; elevated zonulinMolecular plausibilityFunctional BBB permeability; intestinal markers do not index the BBB
Rodent mechanism [21, 102–104]Metabolite-driven behavioral and oligodendroglial effectsSufficiency of specific metabolitesHuman relevance of concentrations achieved
Rodent reconstitution [105, 106]Single-species rescue via vagal–oxytocin pathwayA manipulable pathway in the experimental modelsApplicability to idiopathic human ASD
Human-to-germ-free transfer [20]Donor-dependent behavioral and metabolic phenotypesDonor-dependent phenotypes in germ-free miceThat it operates in human development
Uncontrolled human studies [107, 108]Symptom improvement reportedAssociation within an uncontrolled designEffect distinguishable from expectancy or developmental change
Controlled trials [22, 109, 111]Oral FMT RCT [22]; small double-blind, placebo-controlled probiotic–prebiotic study [109]; kefir-derived probiotic RCT [111]Controlled estimates of intervention effectsConsistent clinical benefit or microbiome mediation; the FMT trial is limited by sample size and possible functional unblinding [22], and the probiotic RCT showed no between-group clinical effect [111]
Systematic review [110]Evidence base judged small and low qualityBasis for current cautionNothing further; no pooled effect estimate

 

↓  Table 7. Biomarker Candidates: Current Standing
 
CandidateCompartmentConsistency (author appraisal)CNS anchorStanding (author appraisal)
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
MAR autoantibody panel [16, 79]Maternal serumHigh specificity reported [16]Brain enlargement [80]; animal model [17, 81]Candidate risk or subtype marker; prospective validation required
Folate receptor α autoantibodies (FRAAs) [87, 88]SerumSubstantial prevalence reported in studied samples [87, 128]Proposed interference with choroid plexus folate transport [87, 88]Candidate predictive marker for folinic acid response; not prospectively validated [127, 128]
Redox markers (GSH, damage markers) [113, 114]BloodAmong the more consistently reported pooled peripheral findings reviewed hereTissue-level findings [112]No paired peripheral–central validation; one brain-spectroscopy study found no group difference [115]
Pro-inflammatory cytokine panels [23, 26, 38]Plasma/serumSmall–moderate pooled differences; quality-sensitive [24]Low serum–CSF correspondence within ASD [65]Not diagnostic; stratification role unvalidated
TGF-β1 [23, 25]Plasma↓ in early pooling [23] onlyOpposite direction reported in brain tissue [1]Not reproduced in later pooling [24, 25]
Immune endophenotypes [39]Stimulated PBMCSubgroup clustering reported in one study [39]NonePotential stratification approach; not diagnostic
Myeloid dendritic cells [59]BloodSingle cohortAmygdala volume correlationHypothesis-generating
Serum zonulin [92]SerumSingle cohortIntestinal, not cerebrovascularLimited; indirect proxy
CSF immune mediators [1, 65–67]CSFDirectionally inconsistent; few controlled studies [66]DirectNot a validated marker; controlled paired serum–CSF data needed

 

↓  Table 8. Immunomodulatory and Mechanism-Directed Interventions in ASD
 
InterventionDesignOutcome (reference)Mechanistic confirmationAuthor appraisal
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (see Section “Review Approach and Evidence Appraisal”).
Celecoxib adjunct [121]RCT, adjunctive, single centerImproved irritability, lethargy, stereotypyNot assessedAdjunctive design; effect cannot be attributed to anti-inflammatory action
Pentoxifylline adjunct [122]RCT, adjunctiveImprovement across ABC subscalesNot assessedSame limitation; overlapping methodology
Prednisolone adjunct, regressive ASD [124]RCT, adjunctiveBehavioral improvementNot assessedCandidate subgroup; mechanism not confirmed
Minocycline [123]Open-label pilotNo clinical improvementPeripheral and CSF biomarker changes only; central target engagement not measuredBiomarker change without detectable clinical improvement
IVIG [125, 126]Open-label pilot; systematic review and meta-analysisImprovement reported; few controlled studies [126]Not assessedNot established outside immunological indications or clinical trials
N-acetylcysteine [116, 117]RCTs; meta-analysisPooled improvement in irritability, hyperactivity, and social awareness (2–4 small trials per outcome) [117]Not assessedShort, small trials; larger replication needed
Sulforaphane [118, 119]Two RCTsImprovement; reversal on withdrawalProposed Nrf2-related mechanism; target engagement not measuredPositive trials; engagement unmeasured; replication needed
Folinic acid, FRAA-associated [127, 128]RCT with biomarker-associated analysis (not stratified randomization)Improved verbal communication; larger response reported in FRAA-positive participantsTransport mechanism not directly measuredPreliminary; prospective biomarker-stratified trial needed
Microbiota-directed interventions [22, 107–111]Open-label; double-blind placebo-controlled; RCTsImprovement in uncontrolled studies; selected differences in one FMT RCT [22]; none in one probiotic RCT [111]Not assessedMixed controlled results; microbiome mediation not shown

 

↓  Table 9. Candidate Neuroimmune Subgroups in ASD: Demonstrated Versus Hypothesized
 
Candidate subgroup or mechanismHuman evidence and replicationMechanistic evidenceBiomarkerPhenotype and therapeutic relevanceCurrent status (author appraisal)
Status categories represent the author’s appraisal (Section “Review Approach and Evidence Appraisal”), ordered from comparatively well characterized to hypothesized. No candidate meets all criteria for an independently replicated, biologically defined subgroup. FRAA: folate receptor alpha autoantibody; IgG: immunoglobulin G; PBMC: peripheral blood mononuclear cell; PET: positron emission tomography; TSPO: translocator protein.
Maternal autoantibody-related ASDDefined antigen patterns in mothers; high case–control specificity; data largely from one research program [16, 79, 82]Fetal intraventricular IgG and endogenous-antigen mouse models [17, 81]Maternal serum antigen panel [16]Brain enlargement in a subgroup [80]; no targeted therapyComparatively well-characterized candidate; independent replication and human causal evidence lacking
FRAA-positiveReported in a substantial proportion of studied samples; assay-dependent [87, 88, 128]Proposed blockade of choroid plexus folate transport; not directly measured [87]Serum blocking/binding FRAAs [88]Larger folinic acid response in biomarker-associated analysis [127]Replicated association, incomplete biological definition; predictive value untested prospectively
Stimulation-defined immune endophenotypesCytokine-response clusters in one cohort [39]Not establishedStimulated PBMC cytokine profileBehavioral correlates reported in the same cohort [39]; no therapy testedExploratory; requires replication
Regression with elevated cytokinesPlasma cytokines associated with regression [38, 54]Not establishedPlasma cytokinesPrednisolone adjunct improved behavior [124]; minocycline showed no benefit [123]Partially supported stratification; mechanism unconfirmed
Gastrointestinal symptoms with mucosal immune signaturesMucosal and peripheral immune differences in ASD with gastrointestinal symptoms [100, 101]Barrier and microbiota mechanisms in animal models [102]Mucosal profiles; serum zonulin [92]Symptom-specific associations [100]; microbiota trials not stratified [22, 111]Partially supported; ASD-specific definition lacking
Maternal infection or fever exposureModest, non-specific registry and cohort associations [68–70]Rodent and primate MIA models [49, 76–78]Exposure history; no validated biomarkerNot establishedMechanistically plausible; not a definable subgroup in humans
Maternal IL-17A axisNot demonstrated; human IL-17 data inconsistent [24, 55, 57]Defined murine pathway [18, 19, 75]None validatedNoneHypothesis derived from animal models
Complement-related synaptic alterationMixed brain mRNA and plasma protein findings [89, 90]Developmental pruning biology [50–52]NoneDirection of any effect unknownExploratory; direction unresolved
Altered TSPO signalSmall, directionally heterogeneous PET studies [45–48]Signal not cell-type specificTSPO-PETPossible sex difference [46, 48]Insufficiently validated

 

↓  Table 10. Major Neuroimmune Hypotheses and Evidence Domains in ASD: Current Appraisal and Testability
 
MechanismStrongest support (reference)Principal gap or discordant finding (reference)Current status (author appraisal)
Interpretive columns represent the author’s appraisal rather than findings reported by the cited studies (Section “Review Approach and Evidence Appraisal”).
Altered microglial and astroglial statesPostmortem [1–3, 41, 42]; bulk and single-cell transcriptomic [12–15]Region-dependent marker findings [4]; directionally heterogeneous TSPO-PET [45–48]Consistently reported at tissue level; in vivo status unresolved
MIA via IL-6 and IL-17ARodent pathway sufficiency [18, 19, 49, 75]; primate [76–78]Pathway not demonstrated in humans; modest epidemiological associations [68–71]Experimentally well specified; limited human anchoring
Maternal brain-reactive autoantibodiesAntigen definition [79]; reported specificity [16]; animal models [17, 81]; brain enlargement [80]Applies to a subgroup; human causality not establishedComparatively well-characterized candidate subgroup
Complement-mediated over-pruningDevelopmental biology [50–52]Reduced C3 with adverse behavioral effect in rodents [89]; mixed proteomic findings [90]Direction unresolved on the evidence reviewed here
Barrier dysfunctionMixed brain tight-junction and MMP-9 findings [91]; intestinal markers [91, 92]Unreplicated; no functional BBB measurement; low serum–CSF correspondence [65]Among the least well-established components reviewed here
Microbiota-immune-brain signalingMetabolite mechanisms [21, 102, 103]; reconstitution [105, 106]; transfer [20]; controlled clinical trials [22, 111]Human descriptive studies not designed to exclude dietary and medication confounding [93–100]Experimentally supported; controlled clinical findings remain mixed and preliminary
Oxidative stress and redox dysregulationTissue [112]; large meta-analyses [113, 114]; NF-κB findings [43]No group difference in in vivo glutathione [115]; no paired peripheral–central design; no direct bioenergetic dataAmong the more consistent peripheral findings reviewed here; central status unclear
Immune-genetic liabilityHLA and rare variant signals [34–37]Immune modules not genetically enriched [15]Possible modifying role