The hippocampus takes centre stage in latest autism research
A comprehensive review published in Neurobiology of Disease this month synthesises recent evidence positioning the hippocampus as central to autism spectrum disorder pathology. Not peripheral. Not secondary. Central. And it’s precisely this type of language that requires me to be covering it here.
The hippocampus (“The Encoder” for its role in encoding experiences into retrievable memory) is essential for memory consolidation, spatial navigation, and social cognition. It exhibits structural disruptions, synaptic dysfunction, altered neurogenesis, and inflammatory activation across multiple ASD models and human studies. The review, authored by researchers at Khalifa University and fertility centres across the UAE, documents how genetic mutations, epigenetic alterations, and environmental factors converge on hippocampal circuits to produce core ASD features.
Structural MRI studies show altered hippocampal volumes, subfield-specific anomalies in CA1 and dentate gyrus regions, and reduced functional connectivity with large-scale brain networks including the Default Mode Network (“The Muse”), central Executive Network (“The Will”), and the Salience Network (“The Shadow”). Genetic disruptions in Shank3, Syngap1, Fmr1, Nlgn3, and other ASD-associated genes impair synaptic plasticity specifically within hippocampal circuits. Neuroinflammation — microglial activation, cytokine release, blood-brain barrier dysfunction — exacerbates hippocampal pathology across developmental stages.
But here’s what the comprehensive synthesis can’t definitively answer: Is hippocampal dysfunction driving autism, or is it just along for the ride?
Nevertheless, the authors conclude that hippocampal dysfunction represents a “central contributor to cognitive, emotional, and social difficulties characteristic of ASD” and propose it as both a biomarker and therapeutic target.
The case for hippocampal centrality
Before examining whether hippocampal dysfunction is primary or secondary, the scope of evidence linking it to autism needs establishing. The convergence isn’t limited to one type of study or one level of analysis — it spans structural neuroimaging, genetic models, synaptic mechanisms, inflammatory processes, and epigenetic regulation.
This breadth matters because it rules out the possibility that hippocampal involvement is an artefact of a particular research methodology or model system. The patterns replicate across human post-mortem tissue, live neuroimaging, mouse genetic knockouts, environmental exposure models, and computational network analyses.
So the question isn’t whether the hippocampus is involved in autism. The question is what that involvement means.
Convergence on a single brain region
The review documents extensive evidence linking hippocampal abnormalities to ASD across multiple levels of analysis.
Structurally, neuroimaging reveals heterogeneous hippocampal volume changes — some studies report enlargement, particularly in individuals with overall brain overgrowth, while others show reductions in specific subfields. The CA1 region, critical for integrating processed inputs and supporting temporal sequencing, shows impaired functional connectivity. The dentate gyrus, responsible for pattern separation (discriminating between similar experiences), exhibits reduced neurogenesis and disorganised cellular architecture in animal models.
Post-mortem tissue from children with ASD reveals downregulation of synaptic proteins PSD-95 and SYN1, alongside upregulation of MMP9, an enzyme involved in synaptic remodeling and neuroinflammation. Dendritic alterations include increased branching in CA1 pyramidal neurons in male mice and hypoarborization in dentate gyrus granule cells in females, with both sexes showing elevated dendritic spine density.
Genetically, the convergence is striking. Shank3 knockout mice show reduced social memory neuron activity and disrupted neuronal ensemble dynamics specifically in ventral CA1. When Shank3 is deleted from enough vCA1 neurons, social memory deficits emerge — demonstrating a threshold effect where hippocampal function collapses only when sufficient neurons lack the protein.
Syngap1 haploinsufficiency impairs critical periods of pyramidal neuron maturation, resulting in synaptic instability and less dendritic branching in newly born neurons. The protein regulates both synaptic function at the postsynaptic density and cytoskeletal remodeling during neuronal differentiation.
Fmr1 knockout (the Fragile X model) shows reduced expression of GABA receptor subunits in CA1, decreased inhibitory transmission, and elevated parvalbumin-positive interneuron density in dentate gyrus alongside reduced excitatory input to CA3. The result: excitatory/inhibitory imbalance across hippocampal circuits.
Other genes implicated include SCN2A (reduced hippocampal excitability and impaired long-term potentiation), KCNH5 (downregulated in valproic acid models, contributing to AKT/mTOR pathway overactivation), GluK2 (kainate receptor knockout producing anxiety and impaired pattern recognition), and RPH3A (gain-of-function variants disrupting NMDA receptor localisation and synaptic transmission).
At the synaptic level, hippocampal circuits show impaired long-term potentiation, enhanced metabotropic glutamate receptor-dependent long-term depression, disrupted NMDA receptor function, and altered AMPA receptor trafficking. These aren’t abstract molecular changes — they translate directly into deficits in spatial learning, contextual memory, and social recognition.
Neuroinflammation adds another layer. Microglial activation, elevated pro-inflammatory cytokines (TNF-α, IL-1β, IL-6, IFN-γ), and disrupted fractalkine signalling (CX3CL1/CX3CR1) compromise hippocampal neurogenesis. Maternal immune activation models show reduced BrdU-positive and DCX-expressing immature neurons in dentate gyrus, with particularly severe effects during early childhood when ASD symptoms typically emerge.
This all makes the evidence for hippocampal involvement extensive, multi-level, and reproducible across human studies, genetic mouse models, and environmental exposure paradigms.
But convergence isn’t causation.
Causation versus correlation and the unresolved mechanism
The review acknowledges a critical limitation: “It is still unclear whether hippocampal abnormalities are a consistent hallmark across all ASD subtypes or restricted to specific genetic or behavioural phenotypes.”
Translation: We don’t know if hippocampal dysfunction causes autism’s core features or if it’s a consequence of broader neurodevelopmental disruption that also affects other brain regions.
The directionality problem appears in multiple forms.
Does Shank3 deficiency impair hippocampal social memory circuits, which then produces social deficits? Or does Shank3 deficiency disrupt multiple brain regions simultaneously — prefrontal cortex, amygdala, striatum, cerebellum — with hippocampal dysfunction being just one component of a distributed network failure?
The paper notes that Shank3 deletion specifically in ventral CA1 reproduces social memory deficits, suggesting the hippocampus plays a causal role. But Shank3 is expressed throughout the brain. Deleting it everywhere produces a different phenotype than deleting it from one region. The localised manipulation proves sufficiency (hippocampal Shank3 is sufficient for social memory) but doesn’t establish necessity in the context of whole-organism development.
Similarly, does Fmr1 loss cause hippocampal synaptic dysfunction, which then impairs cognition? Or does Fmr1 loss affect cortical development, cerebellar circuits, and sensory processing simultaneously, with hippocampal changes being secondary to altered inputs from these other regions?
The inflammation evidence faces the same interpretive challenge. Does hippocampal microglial activation cause reduced neurogenesis, which then produces cognitive and social deficits? Or does systemic immune dysregulation affect multiple organs and brain regions, with hippocampal inflammation being just one manifestation?
The review attempts to position hippocampal dysfunction as primary by arguing that memory, spatial cognition, and social behaviour all route through hippocampal circuits. But this confuses involvement with primacy. The hippocampus is involved in these functions. That doesn’t mean its dysfunction is the root cause of their impairment in autism.
Consider the alternative interpretation: autism arises from disruptions in early cortical development, subcortical connectivity, or cerebellar-cortical loops. These primary disruptions alter the inputs the hippocampus receives, the computational demands placed on it, and the feedback it gets from other regions. Hippocampal dysfunction then emerges as a consequence of operating within an altered network architecture.
Under this model, hippocampal abnormalities would be universal in autism (because the hippocampus is downstream of multiple disrupted systems) without being causal. Every autism-related genetic or environmental perturbation that affects broader brain development would secondarily affect the hippocampus. The convergence on hippocampal pathology wouldn’t indicate driver status — it would indicate hub vulnerability.
The paper can’t distinguish between these possibilities because the evidence is primarily correlational.
Even the genetic manipulation studies that delete specific genes from hippocampal neurons can’t fully resolve causation, because they’re creating artificial conditions (region-specific knockout) that don’t reflect how these genes function during natural development when expressed across multiple brain regions simultaneously.
This isn’t a criticism of the research quality. It’s an acknowledgment that establishing causation in complex neurodevelopmental conditions requires evidence the field doesn’t yet have. Longitudinal imaging from infancy, combined with genetic risk profiling and detailed behavioural phenotyping, might eventually clarify whether hippocampal changes precede, coincide with, or follow the emergence of autism symptoms. Current evidence documents robust association but can’t definitively place hippocampal dysfunction in the driver’s seat.
Targeting biology while ignoring environment
The review proposes multiple therapeutic strategies targeting hippocampal pathology: mTOR inhibitors to restore synaptic balance, BDNF-modulating agents to enhance plasticity and neurogenesis, anti-inflammatory interventions to reduce microglial activation, antioxidant-rich diets to protect against oxidative damage, and epigenetic modulators like histone deacetylase inhibitors.
Preclinical evidence supports several approaches. Rapamycin (mTOR inhibitor) restores synaptic balance in ASD-related models. Calcitriol activates PI3K/AKT signalling to modulate hippocampal axon guidance. Taurine supplementation enhances hippocampal neurogenesis through PTEN/mTOR/AKT pathway activation. Fullerenol treatment restores impaired neurogenesis in BTBR mice.
These interventions target real biological mechanisms. mTOR dysregulation does contribute to synaptic dysfunction. Neuroinflammation does impair neurogenesis. Oxidative stress does damage hippocampal neurons.
But there’s one thing that none of these interventions explicitly address.
Structural incompatibility between how hippocampal circuits process information in autism and what standardised environments demand.
The hippocampus doesn’t function in isolation. It operates within environmental contexts that place specific computational demands on its circuits. Standard educational environments demand rapid verbal memory encoding, efficient retrieval under time pressure, and flexible switching between spatial and conceptual frameworks. Standard social environments require quick pattern matching of facial expressions, real-time integration of verbal and non-verbal cues, and immediate contextual memory access during conversations.
If hippocampal circuits in autism process information differently — slower pattern separation in dentate gyrus, altered CA3 recurrent network dynamics, different CA1 temporal integration windows, reduced efficiency in memory consolidation during sleep — then the “dysfunction” is only dysfunction relative to environments designed for neurotypical hippocampal processing speeds and patterns.
Restoring neurogenesis or synaptic plasticity to neurotypical parameters doesn’t solve this. It assumes the goal is making autistic hippocampal function look more like non-autistic hippocampal function, rather than questioning whether environments should accommodate different hippocampal processing characteristics.
many autistic individuals don’t need their hippocampus “fixed” — they need environments that don’t penalise different hippocampal processing patterns, proven by the experiences of those individuals when in environments that are not doing that.
The valproic acid model demonstrates this clearly. VPA-exposed rats show disorganised spatial organization of CA3 pyramidal neurons and altered dentate gyrus granule cell layers. Maternal bumetanide treatment restores “normal” neuronal chloride concentration and GABA function, reversing the CA3 volume increase and improving behaviour.
So what if the disorganised spatial architecture isn’t fundamentally pathological? What if it represents a different computational strategy that functions effectively under certain conditions but fails under the specific demands of standard laboratory behavioural tests?
The same question applies to Fmr1 knockout mice showing enhanced mGluR-dependent long-term depression. This differs from typical activity-dependent synaptic modulation — but “differs from typical” isn’t the same as “dysfunctional.” It might represent an alternative synaptic plasticity mechanism that works well for certain types of learning while performing poorly on the specific memory tasks researchers use to assess “normal” hippocampal function.
The therapeutic focus on normalisation — restoring synaptic protein levels to control values, increasing neurogenesis to match neurotypical rates, reducing inflammation to baseline — treats difference as deficit without interrogating whether the deficit is intrinsic to the biological variation or imposed by environmental mismatch.
None of this means biological interventions are useless. Severe neuroinflammation that causes neuronal death isn’t “just a difference” — it’s damage that impairs function across any environment. Profound synaptic dysfunction that prevents basic memory formation isn’t “neurodiversity” — it’s pathology requiring treatment.
But the majority of hippocampal alterations described in this review fall into an ambiguous middle ground where they could represent either pathology or variation depending on environmental context. The research doesn’t distinguish between these possibilities because it assumes neurotypical function as the universal standard.
Biomarkers don't redesign incompatible worlds
The review concludes that hippocampal dysfunction “holds translational value in ASD, both as a source of measurable biomarkers and as a target for therapeutic intervention.”
Proposed biomarkers include hippocampal volume measurements, functional connectivity patterns with large-scale networks, and oscillatory signatures. These could serve as endpoints in trials aimed at restoring synaptic plasticity or neurogenesis.
But this framework operates within a medical model positioning hippocampal difference as pathology requiring correction.
Biomarkers are tools for categorisation. They don’t create solutions. Identifying reduced hippocampal connectivity doesn’t clarify whether the problem is the connectivity pattern or environments assuming neurotypical function for success.
The sovereignty alternative: use biomarkers to identify who needs environmental modification rather than intervention. Different CA1-CA3 connectivity might not require synaptic enhancement but work structures accommodating slower memory consolidation or educational approaches providing more processing time.
The same measure guides opposite responses. Medical model: reduced neurogenesis indicates need for BDNF intervention. Sovereignty model: reduced neurogenesis indicates need for reduced cognitive load and environmental predictability.
The paper’s focus on “restoring” function assumes neurotypical hippocampal operation as the template. But if autism represents a different trajectory with different computational trade-offs, restoration might erase adaptive specialisations functioning well under certain conditions.
Hippocampal research holds translational value. The question is what it’s translated toward: making autistic individuals function like non-autistic individuals in existing environments, or building environments where different processing patterns aren’t penalised.
The evidence for hippocampal involvement is extensive. Whether it’s driving autism or along for the ride remains unresolved. Either way, fixing the hippocampus won’t fix environments designed for neurotypical function.
Update — May 2026
The driver-versus-downstream question has not been resolved in the months since publication, but a strand of subsequent research has made the more important point unavoidable: even where biological convergence is documented, normalising the autistic brain toward neurotypical parameters has now been shown to be logically impossible.
The Directory’s coverage of Huang et al.’s GABA-B receptor study and what it proved about the normalisation framework demonstrated that no single intervention can shift all autistic brain networks toward non-autistic measurements simultaneously — improving one network actively destabilises another. That finding extends directly to the hippocampal therapeutic strategies discussed here. Restoring synaptic plasticity, neurogenesis or inflammatory markers to “neurotypical” values is not a clean target; it is a moving one that produces internal contradictions.
Information theory frameworks applied to autistic neural systems push the argument further — describing what autistic brains do without requiring the concept of deficit at all. And the Cambridge findings covered in genetic timing — when autism becomes visible is written in DNA show that autism arises through partly separate genetic routes with distinct developmental rhythms, undermining any single-pathway hippocampal driver model.
The biology is real. The interpretive frame around it is what’s failing.
Citations
Neurobiology of Disease — Review on the role of hippocampus in autism spectrum disorder: Recent insights into neuropathology, genetics, and emerging therapeutic strategies (Bhamidimarri et al., 2026)
Shank3 studies — Tao et al. (2022), Garrido et al. (2022), Kouser et al. (2013)
Syngap1 research — Singh et al. (2025), Clement et al. (2012)
Fmr1 knockout models — Sabanov et al. (2017), Reinhardt et al. (2020)
Neuroinflammation studies — Rexrode et al. (2024), Traetta et al. (2021)
mTOR pathway interventions — Sato & Ikeda (2022), Xiaoyan et al. (2024)
Hippocampal neurogenesis — Gong et al. (2025), Kuipers et al. (2015)
