Engineered Source-Specific Extracellular Vesicles for Neurodegenerative Diseases: From Biological Barriers to Targeted Delivery Strategies

Abstract

With the rapidly aging global population, neurodegenerative diseases (NDDs) present an escalating challenge to public health. The complex, self-perpetuating pathological network of NDDs, encompassing aberrant protein aggregation, chronic neuroinflammation, synaptic impairment, and mitochondrial dysfunction, significantly impedes the efficacy of traditional single-target pharmacological interventions. Extracellular vesicles (EVs), which encapsulate a diverse array of functional cargoes, including proteins, lipids, and nucleic acids, mediate both horizontal intercellular communication and cross-kingdom signaling. Their distinctive biological properties—characterized by low immunogenicity, favorable biocompatibility, and an intrinsic ability to cross the blood-brain barrier (BBB)—position EVs as a promising multi-target pharmacological platform for treatment of NDDs. By integrating genetic, chemical, and physical engineering strategies, these vesicles have been redefined as programmable nanotherapeutic tools. To address the cost and scalability constraints associated with mammalian cell-derived EVs, plant-derived EV-like nanoparticles (PDENs) represent a promising “green” platform. In this review, we systematically delineate the pathological mechanisms of NDDs, the diverse sources and characteristics of EVs, their functionalization via advanced engineering, and their multifaceted applications in NDDs therapy. Finally, we discuss the translational potential of these platforms, particularly their capacity to address the challenges associated with BBB penetration, and outline a framework to guide the development and clinical evaluation of EV-based therapies for NDDs.

Introduction

Neurodegenerative diseases (NDDs), including Alzheimer’s disease (AD), Parkinson’s disease (PD), Huntington’s disease (HD), and amyotrophic lateral sclerosis (ALS), are chronic disorders characterized by the progressive dysfunction and loss of specific neuronal populations, leading to gradual neurological impairment.¹ According to the Global Burden of Disease 2021 report, neurological conditions affected approximately 3.4 billion individuals in 2021 and collectively ranked as the leading cause of disability-adjusted life years globally.² This burden is further exacerbated by rapid population aging. A recent modeling study projected that the number of people living with PD worldwide would reach 25.2 million by 2050, representing a 112% increase from 2021, highlighting the growing pressure on health systems worldwide.^3,4

At the cellular molecular levels, NDDs arise from an interconnected network of pathological processes. Key mechanisms include aberrant protein aggregation, chronic neuroinflammation, synaptic dysfunction, and mitochondrial impairment.^5,6 Genomic instability, axonal transport deficits, and disrupted energy metabolism further exacerbate neuronal injury.^7–9 Rather than acting independently, these processes interact through feedforward loops that heighten selective neuronal vulnerability and promote neuronal loss, ultimately leading to progressive functional decline.¹⁰ This mechanistic complexity poses a major therapeutic challenge, as targeting a single pathway may be insufficient to interrupt disease progression.

In recent years, advances in understanding NDD pathogenesis, together with developments in biotechnology, have expanded the therapeutic landscape, including novel small-molecule drugs, monoclonal antibodies (mAbs), and gene- and stem cell-based therapies.¹¹ However, their clinical translation remains constrained by three major challenges: limited brain delivery, the multifactorial nature of NDDs, and safety concerns.¹² First, the blood-brain barrier (BBB) restricts the central nervous system (CNS) penetration of many systemically administered agents, particularly biologics. Achieving effective concentrations in the brain may therefore require higher systemic doses, increasing the risk of off-target toxicity.¹³ Second, the involvement of multiple interacting pathological pathways can limit the efficacy of single-target therapies. Current symptomatic treatments for AD, such as donepezil and memantine, provide clinical benefits but do not halt disease progression.¹⁴ Finally, the biosafety and long-term tolerability of emerging therapies remain important concerns. Stem cell-based approaches carry risks such as tumorigenicity, while anti-Aβ antibodies like lecanemab require monitoring for amyloid-related imaging abnormalities (ARIA), including cerebral edema and hemorrhage.¹⁵ Together, these limitations underscore the need for biocompatible, multifunctional therapeutic platforms that combine efficient BBB penetration, multi-target activity, and low immunogenicity.¹⁶

Against this backdrop, extracellular vesicles (EVs) have emerged as promising therapeutic and drug-delivery platforms (Figure 1).¹⁷ EVs are a heterogeneous population of lipid-bilayer-delimited particles ranging from tens of nanometers to several micrometers in diameter, with small EVs generally defined operationally as particles smaller than 200 nm.¹⁸ EVs mediate intercellular communication by transporting proteins, lipids, and nucleic acids, while their lipid bilayer protects these cargoes from extracellular degradation.¹⁹ Notably, EV-associated microRNAs (miRNAs) can regulate gene expression post-transcriptionally and may contribute to neuroprotective effects.^20,21 The molecular composition and biological activity of EVs depend strongly on their cellular or tissue source.²² Native membrane lipids and surface proteins facilitate interactions with recipient cells and may promote BBB transport through mechanisms such as transcytosis. Surface modification and cargo loading can further enhance targeting and therapeutic delivery, supporting the development of engineered EV-based systems.²³

Figure 1.

Engineered EVs for brain delivery in neurodegenerative diseases. (A) Intravenous administration. (B) Proposed receptor-mediated transport across the BBB, with an inset showing TfR-targeting ligands and CD47 on the EV membrane. (C) Intended delivery to affected neural cells in the brain parenchyma. EVs are shown in green; vessels and endothelial cells in pink; pericytes in orange; brain parenchyma in grey; lesions in red; neural cells in brown; and TfR-targeting ligands in yellow. Solid arrows indicate the proposed delivery sequence.

Abbreviations: CD47, cluster of differentiation 47; RMT, receptor-mediated transcytosis.

In this review, we summarize recent advances in EV-based therapies for NDDs. We first outline the pathological complexity of NDDs and the limitations of current treatments, and then compare stem cell-derived EVs, neural cell-derived EVs, and plant-derived EV-like nanoparticles (PDENs), with emphasis on their source-specific properties and therapeutic potential. We next summarize engineering strategies to improve CNS targeting and therapeutic efficacy, and conclude with an assessment of clinical translation, remaining barriers, and future research priorities. Recent reviews have generally focused either on comparing EVs from different sources or on summarizing engineering strategies, but have provided limited integration of these aspects with the disease-specific pathological mechanisms of NDDs. By linking NDD pathophysiology with source-specific EV selection and bioengineering strategies, this review provides an integrated framework encompassing both mammalian and PDENs.

This narrative review was informed by a structured search of PubMed, Web of Science, Embase, Scopus, and Google Scholar for articles published from January 2000 to June 2026. The search terms covered three domains: neurodegenerative diseases (“neurodegenerative disease”, “Alzheimer’s disease”, “Parkinson’s disease”, “Huntington’s disease”, and “amyotrophic lateral sclerosis”), extracellular vesicles (“extracellular vesicle”, “small extracellular vesicle”, “exosome”, and “plant-derived EV-like nanoparticle”), and engineering or therapeutic applications (“surface modification”, “cargo loading”, “brain targeting”, “blood-brain barrier”, and “therapy”). Source-specific terms, including “mesenchymal stem cell”, “neural cell”, and “medicinal plant”, were added where appropriate. After duplicates were removed, titles and abstracts were screened for relevance, and potentially eligible articles were assessed in full text. Reference lists of relevant studies and reviews were also manually screened to identify additional publications. Only peer-reviewed publications within the predefined scope were considered; non-peer-reviewed and out-of-scope records were excluded. Final selection was based on scientific relevance and contribution to the mechanistic and translational themes of this narrative review.

Pathological Mechanisms of Neurodegenerative Diseases

NDDs arise from a complex network of interacting pathological modules rather than a single causal lesion. Across AD, PD, ALS and HD, several convergent mechanisms drive disease progression: proteostasis failure (linked to pathogenic protein aggregation), chronic neuroinflammation, mitochondrial impairment, and synaptic dysfunction.^24,25 These modules are mutually reinforcing, exacerbating selective neuronal vulnerability that leads to functional decline (Figure 2). In this section, we delineate a modular “pathology map” to establish a mechanistic foundation for subsequent therapeutic strategies—most notably, EVs-mediated interventions.

Figure 2.

Interconnected pathological mechanisms in NDDs. (A) Protein aggregation and impaired clearance. (B) Neuroinflammation involving microglia and reactive astrocytes. (C) Synaptic dysfunction and impaired axonal transport. (D) Mitochondrial dysfunction, impaired mitophagy, oxidative stress, and reduced energy availability. Solid arrows indicate interactions among these processes; black and red downward arrows (↓) indicate decreased levels, while black upward arrows (↑) indicate increased levels.

Abbreviations: Aβ, amyloid-beta; ALP, autophagy–lysosome pathway; ATP, adenosine triphosphate; ETC, electron transport chain; GDNF, glial cell line-derived neurotrophic factor; IL-1β, interleukin-1β; IL-6, interleukin-6; IL-18, interleukin-18; LTD, long-term depression; LTP, long-term potentiation; mHTT, mutant huntingtin; ROS, reactive oxygen species; TDP-43, transactive response DNA-binding protein 43; TNF-α, tumor necrosis factor-α; UPS, ubiquitin–proteasome system.

Proteostasis Failure: The Core Pathological Axis

Proteostasis is a sophisticated dynamic equilibrium governing the synthesis, folding, trafficking, and degradation of the cellular proteome.^26,27 To maintain this balance, neurons rely on robust quality control network comprising molecular chaperones, the ubiquitin-proteasome system (UPS), and the autophagy-lysosome pathway (ALP).²⁷ Proteostasis failure and the subsequent aggregation of pathogenic proteins represent a fundamental pathological axis and a critical upstream event in NDD progression.^28,29 Since the landmark identification of “plaques and tangles” in 1906,⁵ molecular research has elucidated their biochemical substrates: β-amyloid (Aβ), hyperphosphorylated Tau, α-synuclein, mutant huntingtin (mHTT), and TDP-43.^6,30,31 Under oxidative and inflammatory stress, this proteostatic capacity progressively declines, allowing misfolded proteins to escape clearance. These misfolded species trigger early neurotoxicity, driving synaptic impairment and chronic cellular stress.^32–35

This pathological axis is also influenced by genomic instability.³⁶ Hereditary mutations and impaired DNA damage repair (DDR) directly shape specific pathological phenotypes.³⁷ For instance, increased amyloid beta precursor protein (APP) gene copy numbers in early-onset AD drive Aβ overproduction, initiating the cascade of abnormal aggregation and cerebral amyloid angiopathy (CAA).³⁸ In addition, neurons are exceptionally vulnerable to DNA damage accumulation during aging and chronic stress. Robust DDR systems—including single-strand/double-strand break (DSB) repair and base excision repair—are essential for maintaining this genomic stability.^39,40 In AD, pathological Aβ-associated signaling impairs DSB repair by affecting key DDR factors. For instance, activated extrasynaptic NMDA receptors induce the proteasomal degradation of BRCA1, a crucial DNA repair protein, ultimately accelerating synaptic dysfunction and cognitive decline.⁴¹

Moreover, the initial proteostasis failure triggers a self-reinforcing vicious cycle. Pathogenic protein aggregates do not merely accumulate as inert debris; they actively compromise the cellular clearance systems. For instance, in HD models, mHTT aggregates sequester essential chaperones—such as Hsp70 and Hdj1/2, which are required to deliver these aggregates to the degradation machinery.⁴² Similarly, in ALS models, poly-GA aggregates can physically obstruct the 26S proteasome, directly leading to systemic UPS dysfunction.⁴³ This creates a lethal paradox: as the burden of aggregates grows, the cell’s clearance capacity is increasingly impeded, further fueling proteostatic collapse. Furthermore, recent evidence suggests that these aggregates can exhibit “seeding” activity, potentially propagating this proteostatic failure to neighboring neurons through the release and internalization of toxic aggregates.^44,45 This feedback loop between aggregation and degradation failure, synchronized with intercellular spreading, serves as a cornerstone of NDD chronicity. Therefore, interrupting this cycle—primarily by accelerating the endogenous clearance of toxic aggregates—represents a strategic entry point for therapeutic interventions.

Chronic Neuroinflammation

Protein aggregation not only damages neurons directly but also fuels a self-sustaining inflammatory cascade by recruiting the CNS innate immune system. This chronic neuroinflammation is primarily driven by the phenotypic transition of microglia and astrocytes from homeostatic, neuroprotective states to reactive, neurotoxic profiles.^46,47 Rather than acting in isolation, these glial populations engage in a complex pathological crosstalk. As the initial responders to proteostatic collapse, microglia recognize and internalize protein aggregates via pattern-recognition receptors, triggering their amoeboid transformation and release of pro-inflammatory mediators, including IL-1β, IL-18, and TNF-α.³² These microglial-derived factors serve as molecular triggers that induce a reactive astrocyte phenotype—often termed A1—characterized by a gain of neurotoxic properties and the loss of essential homeostatic functions.⁴⁸

This synergistic interaction creates a self-amplifying cycle of neurotoxicity across various NDD models. For instance, in AD and PD, activated microglia release apoptosis-associated speck-like protein containing a CARD (ASC) speck via the NLRP3 inflammasome; these specks act as extracellular scaffolds that cross-activate neighboring astrocytes, exacerbating their pro-inflammatory output.^49,50 Furthermore, reactive astrocytes, once primed by microglial signals, can release factors like lipocalin-2 (LCN2) or complement C3, which feed back to sustain microglial activation and impair their phagocytic clearance of protein aggregates.⁵¹ Ultimately, this reciprocal signaling establishes a neurotoxic feedback loop: microglial activation drives astrocyte reactivity, which in turn compromises the BBB and recruits peripheral immune cells, fostering an environment hostile to neuronal survival. Therefore, intercepting this glial crosstalk is a cornerstone for immunomodulatory interventions.

Integrated omics analyses, together with patient samples and experimental models, are helping to clarify the cellular mechanisms of chronic neuroinflammation. In primary open-angle glaucoma, combined single-cell RNA-seq and ATAC-seq analyses implicated Bcl3/NF-κB and Edn2/Hippo signaling, with Bcl3 and Edn2 expression associated with immune-cell infiltration. Increased expression was corroborated in patient samples, glutamate-injured retinal cells, and experimental models, while pharmacological inhibition of the NF-κB or Hippo pathway reduced retinal ganglion-cell injury.⁵² In multiple sclerosis, MRI-informed single-nucleus RNA sequencing and patient brain tissue immunostaining identified lymphocyte-microglia-astrocyte interactions at chronic active lesion edges and implicated C1q in microglial activation. Microglia-specific C1qa deletion attenuated reactive microgliosis at the onset of experimental autoimmune encephalomyelitis, while C1q blockade reduced microglial inflammatory markers during its chronic phase.⁵³

Mitochondrial Impairment and Energy Crisis

Mitochondria are the primary organelles sustaining ATP supply and redox homeostasis in neurons. Given their extensive axonal architecture and the high metabolic cost of synaptic transmission, neurons are uniquely vulnerable to defects in mitochondrial quality control and energy supply. Extensive evidence identifies mitochondrial dysfunction and dysregulated energy metabolism as fundamental drivers of NDD pathogenesis.^19,54–56 In PD models, for instance, α-synuclein aggregation influences key mitochondria-associated proteins—including adenylate kinase 2 (AK2), ATP synthase, and DJ-1—resulting in collapsed membrane potential and bioenergetic failure, which directly impair electron transport efficiency and trigger a surge in reactive oxygen species (ROS).⁵⁷ These events further disrupt mitochondrial fusion/fission dynamics and mitophagy, compromising axonal transport and exacerbating protein aggregation.

Beyond intracellular deficits, impaired intercellular metabolism further precipitates the neuronal “energy crisis”. In AD, soluble Aβ oligomers and associated vascular pathologies restrict cerebral blood flow and glucose utilization; notably, a 30–35% reduction in glucose uptake directly correlates with accelerated cognitive decline.^58,59 Similarly, in early-stage HD, mutant huntingtin (mHTT) accumulates within striatal astrocytes, inhibiting glycolysis and reducing the regional metabolic rate by approximately 20%.⁶⁰ This persistent metabolic shortage impairs the ability of astrocytes to clear glutamate, triggering excitotoxicity and progressive death of spiny neurons. Collectively, mitochondrial impairment and metabolic failure serve as shared pathological cornerstones that synergize with protein aggregation and neuroinflammation to drive NDD chronicity.

Synaptic Dysfunction and Axonal Transport Deficits

Synapses serve as the fundamental structural and functional units of neural circuitry. Their dysfunction—manifesting as impaired plasticity and compromised transmission efficiency—typically precedes neuronal loss and acts as the primary determinant of early clinical symptoms, such as cognitive and motor deficits.^61–64 Therefore, synaptic dysfunction represents a pathological convergence point for proteostatic collapse, neuroinflammation, mitochondrial impairment and energy crisis. Phenotypically, this is characterized by the downregulation of synapse-associated proteins, the loss of dendritic spines, and a profound shift in synaptic plasticity from long-term potentiation (LTP) toward long-term depression (LTD).^65–68 In NDDs, pathological protein aggregates could reduce the expression of key synaptic protein (eg, PSD-95, Rab3A, SNAP-25), diminish spine density and neurotransmitter release, directly contributing to spatial memory loss and motor impairment.^69–71 This intrinsic damage is further exacerbated by the “extrinsic” microenvironment: chronic neuroinflammation promotes synaptic loss via cytokine release, while restricted ATP supply from dysfunctional mitochondria impairs neurotransmitter release and ion pump activity.⁷²

Beyond local insults, the collapse of axonal transport including microtubules and actin filaments serves as a primary driver of synaptic dysfunction.⁷³ Structural and functional microtubule defects paralyze long-distance transport, such as dynamitin (p50)-mediated dynein disruption in ALS, survival of motor neuron 1 (SMN1)-related aberrant α-tubulin tyrosination in SMA, and mHTT-induced β-tubulin degradation via the immunoproteasome LMP2 in HD, which starves distal synapses of essential mitochondria and signaling complexes.^74–77 Complementing this, dysregulated actin dynamics disrupt the “last mile” of synaptic delivery and structural plasticity. In AD, Aβ-induced “cofilin-actin rods” physically obstruct axonal trafficking while destabilizing dendritic spines.⁷⁸ Similarly, in SMA, SMN1-Plastin 3 (PLS3) deficiency disrupts F-actin dynamics and prevents the proper surface localization of tropomyosin receptor kinase B (TrkB) receptors, hindering synapse formation and maturation.⁷⁹ Ultimately, these factors merged into a self-amplifying synaptic imbalance circuit. Given its role as an early indicator of disease progression, synaptic dysfunction provides critical targets for early diagnosis and precision interventions.

Neuronal Loss: The Terminal Phase of Neurodegeneration

Neuronal loss represents the final structural foundation of clinical progression in NDDs, serving as the “execution stage” where multiple pathological mechanisms converge. Under the synergistic interplay of proteostasis failure, chronic neuroinflammation, mitochondrial crisis, and synaptic disconnection, neurons are forced into diverse regulated cell death (RCD) programs, transforming chronic functional impairment into irreversible structural attrition.^80,81 Proteostatic failure serves as a primary initiator; for instance, Aβ aggregates stabilize c-Fos, which upregulates the pro-apoptotic gene Bim, directly accelerating neuronal loss.⁸² This intrinsic vulnerability is further exploited by the chronic neuroinflammatory microenvironment, where pro-inflammatory cytokines and ROS overactivate poly (ADP-ribose) polymerase (PARP) and the RIPK1/RIPK3/MLKL axis, triggering necroptosis and irreversible membrane damage.^83,84 Simultaneously, mitochondrial dysfunction acts as a critical metabolic executioner. Oxidative mediators such as nitric oxide (NO) induce the S-nitrosylation of the microtubule-associated proteins, which inhibits the mitochondrial ubiquitin ligase (MITOL). This molecular blockade prevents the effective clearance of damaged mitochondria, leading to their pathological aggregation and driving apoptosis cascade.⁸⁵ This process is further exacerbated by the collapse of axonal transport and synaptic integrity, which deprives neurons of essential neurotrophic support.⁸⁶

Moreover, neuronal loss exhibits profound regional and subtype specificity, providing the pathological basis for the diverse clinical phenotypes observed in NDDs.⁸⁷ Recent advancements in snRNA-seq and spatial profiling (eg, Slide-seqV2) have illustrated these patterns with unprecedented resolution. In AD, CBLN²⁺/LINC00507⁺ excitatory neurons in layers 2/3 of the inferior temporal gyrus are selectively depleted as tau burden increases.⁸⁸ In HD, there is marked reduction in Enk⁺ medium spiny neurons projecting to the external globus pallidus (GPe), whereas SP⁺ projection to the internal globus pallidus (GPi) or substantia nigra pars compacta (SNc) remain largely intact.⁸⁹ In PD, the SOX6⁺/AGTR1⁺ dopaminergic (DA) subpopulation in the ventral tier of the substantia nigra is selectively lost, whereas the CALB1⁺/GEM⁺ subpopulation in the dorsal tier is relatively resilient.⁹⁰ The death threshold of specific subpopulations is determined by a complex interplay of intracellular proteostatic capacity, metabolic load, axonal transport demands, and the local immune microenvironment. These insights imply that future therapeutic strategies must shift from “pan-brain interventions” toward “precision protection” targeted at specific neuronal populations.

In summary, the pathogenesis of neurodegeneration is a systemic and self-amplifying cascade, where initial genomic and proteostatic defects are exacerbated by neuroinflammation and metabolic crises, ultimately leading to synaptic dysfunction and neuronal loss. The pathological mechanism clarification advanced the development of multitarget therapeutic platforms to halt disease progression.

Current Therapeutic Strategies and Challenges

Current Therapeutic Strategies for NDDs

In recent years, an in-depth understanding of the pathogenesis of NDDs, coupled with rapid advancements in biomedical technology, has catalyzed the development of diversified therapeutic strategies. These include novel small-molecule drugs, monoclonal antibodies (mAbs) and biologics, and gene and stem cell therapies.¹²

Small-Molecule Drugs

Small-molecule drugs remain the primary clinical strategy for NDDs due to their BBB permeability and well-established pharmacokinetics. These agents are broadly categorized into neurotransmitter modulators for symptomatic relief and emerging disease-modifying therapies (DMTs). In the context of symptomatic management, therapeutic efforts focus on restoring the neurotransmitter equilibrium disrupted by neuronal loss. For AD, FDA-approved cholinesterase inhibitors (ChEIs)—such as donepezil, rivastigmine, and galantamine—elevate synaptic acetylcholine (ACh) concentrations to reinforce cholinergic signaling and ameliorate cognitive decline.⁹¹ These are frequently combined with cholinergic precursors like choline alphoscerate to achieve synergistic neuroprotection and attenuate regional brain atrophy.⁹² Similarly, for PD, levodopa remains the “gold standard” for replenishing dopamine levels, often supplemented by dopamine receptor agonists (eg, pramipexole, ropinirole) and monoamine oxidase B (MAO-B) inhibitors (eg, selegiline, rasagiline) to promote motor control and delay complications.^93,94

Beyond neurotransmitter modulation, the focus of small-molecule research has pivoted toward DMTs designed to intervene in the molecular cascades of neurodegeneration. The representative interventions include agents targeting proteostatic failure and RNA metabolism. For instance, arimoclomol facilitates the repair or degradation of misfolded proteins by enhancing heat shock protein activity, which has shown clinical efficacy in delaying progression for lysosomal storage disorders like Niemann-Pick disease type C.⁹⁵ On the transcriptomic level, novel inhibitors such as buntanetap suppress the translation of multiple neurotoxic mRNAs, while splicing modulators like branaplam lower mutant huntingtin levels by inducing aberrant splicing and premature termination.^96–98 Furthermore, endogenous anti-inflammatory molecules, such as palmitoylethanolamide, are being explored in Phase II trials for their ability to scavenge ROS and mitigate the chronic neuroinflammation that fuels disease chronicity.^99,100 Despite these advancements, the limited half-life and off-target effects of systemic small-molecule delivery continue to necessitate more sophisticated, brain-targeted delivery platforms.

Monoclonal Antibodies & Biologics

Monoclonal antibodies (mAbs) and biologics represent an innovative shift toward pathology-specific clearance and precision immunomodulation in clinical strategies. Their core advantage lies in high-affinity antigen recognition, enabling the neutralization of toxic aggregates, receptor blockade, or replenishment of deficient proteins. Notably, immune clearance of pathological aggregates has achieved landmark progress. Anti-Aβ antibodies, such as lecanemab and donanemab, have demonstrated the ability to significantly reduce cerebral amyloid burdens and modestly delay cognitive decline in Phase III trials.^101,102 Similarly, prasinezumab targets α-synuclein with ultra-high affinity (KD≈0.05 nM) to reduce peripheral free aggregates in Parkinson’s research.¹⁰³ Beyond direct protein clearance, mAbs are increasingly utilized for precision neuro-immunomodulation. Temelimab targets the pro-inflammatory HERV-W Env protein to mitigate glial-mediated brain atrophy in MS,¹⁰⁴ while AL002 modulates the TREM2 receptor to activate protective microglial functions, thereby enhancing the brain’s endogenous clearance capacity.¹⁰⁵

Complementary biologics, including recombinant proteins, enzyme replacement therapies (ERT), and peptide, offer targeted solutions for specific NDD subtypes. In MS treatment, peginterferon β-1a utilizes PEGylation to extend drug half-life and enhance long-term immunomodulatory efficacy.^106,107 For lysosomal storage disorders like CLN2, intracerebroventricular administration of recombinant human TPP1 provides essential enzyme replacement, effectively delaying neurological deterioration.¹⁰⁸ Furthermore, peptide-based strategies such as OCS-05 (BN201)—an SGK2 activator—promote neuronal survival by modulating the SGK2/FOXO3 signaling pathway, offering a novel neuroprotective approach for acute optic neuritis and MS.¹⁰⁹ Despite high specificity, biologics and monoclonal antibodies are severely restricted by negligible BBB permeability and potential immunogenicity during chronic administration.

Gene Editing & RNA Interference

Gene therapy represents an etiology-based intervention by precisely modulating pathogenic gene expression at the DNA or RNA level. Currently, antisense oligonucleotides (ASOs) and RNA interference (RNAi) offer a reversible and finely-tuned approach to suppress neurotoxic products. For instance, the ASO tofersen, targeting the SOD1 gene, has significantly improved pathological biomarkers in ALS.¹¹⁰ While nusinersen, has revolutionized prognosis in spinal muscular atrophy (SMA) by modulating SMN2 splicing.¹¹¹ In Parkinson’s research, intranasal delivery of siRNA nanoparticles targeting synuclein alpha (SNCA) mRNA successfully reduced α-synuclein aggregation and motor deficits in preclinical models.¹¹²

Despite their safety and reversibility, RNA-based therapies require repeated administration, thereby driving the development of gene editing strategies with more enduring curative potential. AMT-130, utilizing an adeno-associated virus (AAV5) vector for stable miRNA-mediated silencing of mutant HTT, has shown significant reductions in cerebrospinal fluid (CSF) neurofilament light chain and dose-dependent clinical improvements over 24 months.¹¹³ Advancements in CRISPR-Cas9 technology have further enabled the precise repair of NDDs-associated mutations. Utilizing homology-directed repair in induced pluripotent stem cell (iPSC) models, researchers successfully restored γ-secretase activity and Aβ 42/40 ratio by correcting PSEN1 point mutations.¹¹⁴ Furthermore, combining focused ultrasound (FUS)-mediated blood–brain barrier opening with systemic delivery of AAV-CRISPR-Cas9 vectors enabled localized APOE4 knockdown in the mouse hippocampus, accompanied by reduced apolipoprotein E4 (apoE4) protein levels and markers of glial activation.¹¹⁵ These innovations highlight a transition toward “one-time” permanent interventions for both monogenic and complex NDDs.

Stem Cell Therapy

Stem cell therapy treats NDDs through two primary ways: direct cell replacement and indirect paracrine support. In cell replacement strategies, iPSCs or neural stem cells (NSCs) are transplanted to replenish specific lost neuron populations. For instance, dopaminergic precursor products such as bemdaneprocel and Raguneprocel^® have demonstrated functional integration and significant improvements in dopamine metabolism and motor scores in early-phase clinical trials.^116,117 Concurrently, mesenchymal stem cells (MSCs) leverage their potent immunomodulatory properties to exert paracrine effects. By secreting neurotrophic and anti-inflammatory factors, MSCs optimize the neural microenvironment and mitigate neuroinflammation. Clinical evidence highlights the efficacy of these approaches: spinal injection of autologous hypoxia-preconditioned olfactory mucosa-derived MSCs (hOM-MSCs) led to a 30% reduction in Unified Parkinson’s Disease Rating Scale (UPDRS) scores in PD patients,¹¹⁸ while intrathecal MSC delivery improved nerve conduction in SMA.¹¹⁹

Current interventions are evolving toward standardized, “off-the-shelf” allogeneic products. NestaCell^®, derived from deciduous tooth pulp, has shown promise in reducing white matter loss in HD.¹²⁰ Similarly, the intravenous allogeneic product Lomecel-B demonstrated promising safety in mild AD, correlating with elevated anti-inflammatory factors (VEGF, IL-4, IL-10) and a transient increase in left hippocampal volume.¹²¹ These multi-mechanistic benefits underscore the potential of stem cells not only as a structural replacement but also as a biological factory for neuroprotection.

Challenges and Bottlenecks in NDD Therapeutics

Despite the advantages of current therapeutic strategies, clinical intervention for NDDs faces a lot of challenges and bottlenecks. The progress of clinical translation is primarily inhibited by three convergent structural challenges. The high selectivity of the BBB remains the most significant physiological checkpoint. Many biologics, including monoclonal antibodies and recombinant proteins, show extremely low intracranial uptake in preclinical models following intravenous administration relative to plasma exposure, as determined by quantitative biodistribution analyses.^122,123 Similarly, the high polarity and instability of nucleic acid therapeutics (eg, ASOs and siRNAs) prevent them from autonomously traversing the BBB. This often mandates highly invasive delivery routes, such as intrathecal injections that lead to poor patient compliance and potential localized trauma.¹²⁴

The insufficient target specificity within the complex CNS microenvironment leads to significant off-target effects and biosafety concerns. Small-molecule drugs, while manufacturing-scalable, frequently trigger systemic adverse events, such as the motor fluctuations and dyskinesia associated with chronic levodopa therapy.¹²⁵ Concurrently, live-cell transplantation faces the dual challenge of low engraftment rates in hostile lesion areas and potential risks of tumorigenicity or immune rejection.¹²⁶ Even for precision tools like CRISPR-Cas9, the risk of off-target genomic integration necessitates a safer, more biocompatible delivery platform capable of providing cell-type-specific protection and controlled release.

Furthermore, NDDs are characterized by a multifaceted interplay of proteostasis failure, chronic neuroinflammation, mitochondrial and synaptic dysfunction. Traditional single-target interventions—whether they be antibodies neutralizing a single protein or small molecules modulating one neurotransmitter—are often inadequate to disrupt these interconnected pathological cascades. This “single-target paradox” explains why many promising DMTs fail to achieve promising clinical outcomes.

Therefore, to overcome these systemic bottlenecks, new creative platforms are needed to be developed. Notably, EVs are promising strategies, which possess innate BBB-penetrating capabilities, low immunogenicity, and a unique capacity for multi-target regulation via their complex cargo of proteins and RNAs. Emerging evidence suggests that the neuroreparative and immunomodulatory benefits of stem cells are largely mediated by their secreted EVs—rather than the cells themselves.⁷⁰ By leveraging these endogenous properties, EVs are emerging as the versatile delivery platform to bypass the limitations of traditional modalities and usher in a new era of precision therapy for NDDs.

EVs as a Multidimensional Therapeutic Platform

EVs: Composition and Biological Properties

EVs are heterogeneous particles released by cells, delimited by lipid-bilayer, and unable to replicate independently. They range from tens of nanometers to several micrometers in diameter, with small EVs generally defined operationally as particles smaller than 200 nm.¹⁸ Because the biogenesis of isolated vesicles is rarely demonstrated directly, the generic term “EV” is used throughout this review unless an endosomal origin has been established.

EVs exhibit substantial morphological and molecular heterogeneity. Conventional transmission electron microscopy (TEM) often gives isolated EVs a cup-shaped appearance because of dehydration and vesicle collapse during sample preparation, whereas cryo-electron microscopy (Cryo-EM) preserves their native, predominantly rounded morphology.¹²⁷ Their molecular composition reflects the source and physiological state of the parent cell.¹²⁸ Commonly assessed EV-associated proteins include tetraspanins such as CD9, CD63, and CD81, integrins, and cytosolic proteins such as Alix (PDCD6IP) and TSG101.¹²⁹ However, their abundance varies by source, and no single marker is universal to all EVs. EVs carry diverse bioactive molecules, including proteins, lipids, metabolites, and nucleic acids such as messenger RNA, microRNA (miRNA), and long non-coding RNA. EV-associated miRNAs can regulate gene expression post-transcriptionally after delivery to recipient cells and may contribute to neuroprotective signaling.^20,21

Compared with conventional delivery platforms, EVs offer several potential biological advantages for NDD therapy. Their lipid bilayer and surface proteins may facilitate interactions with brain endothelial cells and transport across the BBB, a major obstacle to CNS drug delivery.^130–132 For example, intravenously administered MSC-derived EVs have been detected in damaged brain regions and associated with therapeutic effects in preclinical models.¹³³ However, BBB transport varies with EV source, administration route, dose, and disease state. The membrane composition of EVs may confer favorable biocompatibility and relatively low immunogenicity, while their lipid bilayer protects encapsulated cargo from enzymatic degradation in complex biofluids.^134,135 Source-specific surface proteins may also confer tissue tropism and promote accumulation at sites of neural injury or pathology.¹³⁶ Moreover, their ability to carry diverse bioactive cargoes may enable EVs to modulate interconnected pathological pathways, supporting a multi-target therapeutic approach. However, cargo diversity alone does not establish synergistic or disease-modifying activity. Most reported effects have been attributed to specific cargo molecules, whereas direct evidence for cargo synergy remains limited and nonspecific contributions have rarely been systematically evaluated. Collectively, these properties support the development of EVs as promising cell-free therapeutic and drug-delivery platforms for NDDs.

Classification of EVs for NDD Therapy

Stem Cell-Derived EVs

Stem cell-derived EVs, particularly those from MSCs, represent the most extensively studied “cell-free” therapeutic platform in neuroregeneration. These vesicles serve as concentrated reservoirs of the parent cells’ regenerative ability, sequestering a diverse array of neurotrophic factors (eg, BDNF, GDNF), immunomodulatory proteins, and regulatory non-coding RNAs.^137,138 By delivering these bioactive cargos, stem cell-derived EVs orchestrate a multi-pronged therapeutic response across several critical pathological axes (Figure 3) (Table 1).

Figure 3.

Representative sources, molecular components of EVs and plant-derived EV-like nanoparticles. (A) Stem cell-derived EVs, illustrated by mesenchymal stem cells and bioreactor-based production. (B) Neural cell-derived EVs from neural stem cells, neurons, astrocytes, and microglia. (C) Plant-derived EV-like nanoparticles, with representative lipid, RNA, and metabolite components. The components shown vary with the biological source and preparation method.

Abbreviations: MSCs, mesenchymal stem cells; NSCs, neural stem cells; CD, cluster of differentiation; BDNF, brain-derived neurotrophic factor; GDNF, glial cell line-derived neurotrophic factor; IL-10, interleukin-10; miR-21, microRNA-21; miR-124, microRNA-124; TGF-β, transforming growth factor-β; PA, phosphatidic acid.

Table 1.

Summary of Representative EVs Derived for NDDs

Source	Isolation Method	Key Markers/Cargo	Representative Disease Models	Experimental Model	Administration Routes	Therapeutic Mechanisms	Results	Ref.
Rat BMSC-EVs	Differential centrifugation + ultracentrifugation	miR-214-3p	AD	Aβ1-42-induced AD rats	Intravenous injection (i.v).	Targets and suppresses CD151	↑antioxidant capacity; ↓CA3 neuronal apoptosis and oxidative stress; improved learning and memory	[139]
Human ADSC-EVs	Exo-Quick precipitation	p-CREB/PGC-1α–related cargo	HD	R6/2 mouse-derived neuronal cells (in vitro)	— (in vitro)	Up-regulates the p-CREB/PGC-1α pathway	↓mHtt aggregates; ↑mitochondrial function and cell viability; ↓apoptosis	[140]
Human ADSC-EVs	Exo-Quick precipitation	p-CREB/CREB, PGC-1α	ALS	G93A-SOD1 mouse NSC-derived neurons (in vitro)	— (in vitro)	Restores p-CREB/CREB ratio and PGC-1α expression	↓mutant SOD1 aggregation; ↓mitochondrial protein abnormalities	[141]
Mouse BM-MSC-EVs	Filtration + ultrafiltration + exosome isolation reagent	SphK1/S1P, NEP	AD	APP/PS1 transgenic mice	i.v.	Activates SphK/S1P signaling; down-regulates BACE1/PS1; up-regulates NEP	↓Aβ production; ↑Aβ clearance; ↓amyloid plaques; ↑NeuN; improved spatial learning and memory	[142]
Rhesus monkey BM-MSC-EVs	Differential ultracentrifugation	Anti-inflammatory cargo	MS	EAE mice + cuprizone demyelination model	i.v.	Inhibits TLR2/IRAK1/NF-κB pathway; drives microglial M1→M2 polarization	↑OPC differentiation and remyelination; ↑oligodendrocytes and MBP; ↓neuroinflammation; improved neurological and cognitive outcomes	[143]
Human BMSC-EVs	ExoQuick-TC precipitation	GDF-15	AD	Aβ42-induced SH-SY5Y cells (in vitro)	— (in vitro)	Activates the AKT/GSK-3β/β-catenin pathway via GDF-15; up-regulates NEP/IDE	↑Aβ42 degradation; ↑cell viability; ↓apoptosis and inflammatory factors	[144]
Human ADSC-EVs	Serial ultracentrifugation	Anti-inflammatory cargo	PD	MitoPark transgenic PD mice	i.v.	Suppresses microglial activation	↓midbrain neuroinflammation; improved motor function and memory	[145]
Mouse BMSC-EVs	Filtration + ultracentrifugation	Immunomodulatory cargo (IL-10/TGF-β)	MS	EAE mice	i.v.	Increases Treg frequency; down-regulates TNF-α/IL-1β/IL-6; up-regulates IL-10/TGF-β	Efficacy comparable to parental BMSCs; ↓immune imbalance in EAE	[146]
hucMSC-EVs	Differential centrifugation + ultracentrifugation	miR-223	AD	Aβ1-40-induced SH-SY5Y cells (in vitro)	— (in vitro)	Targets PTEN via miR-223; activates PI3K/Akt pathway	↓neuronal apoptosis; ↑cell migration	[147]
hucMSC-EVs	Differential ultracentrifugation	miRNA profile	PD	6-OHDA-induced PD rats	i.v.	Inhibits caspase-3 expression	↓apoptosis; EVs crossed BBB and reached substantia nigra; ↓asymmetric rotation; ↓dopaminergic neuron loss; ↑striatal dopamine	[148]
Rat BMSC-EVs	Differential centrifugation + ultracentrifugation	miR-34b/circRNA.2837	PD	Rotenone-induced PD rats	i.v.	Modulates miR-34b/circRNA.2837 axis; down-regulates α-synuclein; restores DJ-1/PARKIN	↑motor function; improved histopathology; superior to L-Dopa	[149]
OM-MSC-EVs	Ultracentrifugation	lncRNA A2M-AS1	PD	MPP+-treated HT22 cells + MPTP PD mice	Intracerebroventricular injection (ICV)	Regulates the A2M-AS1–IGF2BP1–TP53INP1 axis; induces mitophagy	↓oxidative stress; improved PD symptoms in HT22 cells and MPTP mice	[150]
Human MSC-EVs	Differential ultracentrifugation	ICAM1	PD	MPP+-treated HBMECs + MPTP PD mice	Intraperitoneal injection (i.p).	Activates SMAD3/P38MAPK pathway	↑HBMEC angiogenesis; maintained neurovascular unit; ↓MPP+ injury; contributed to PD recovery	[151]
Human iPSC-NSC-EVs	Ultracentrifugation	BBB tight junction–repairing cargo	AD	5×FAD mouse endothelial BBB model (in vitro)	— (in vitro)	Repairs BBB tight junctions	↓BBB leakage and endothelial dysfunction (5×FAD-derived endothelial cells)	[152]
Mouse NSC-EVs	Size-exclusion chromatography	Catalase	PD	α-synuclein-overexpressing/6-OHDA-treated SH-SY5Y cells (in vitro)	— (in vitro)	Delivers catalase to scavenge ROS	↓ROS accumulation and apoptosis; ↑dopaminergic neuron survival	[153]
Hypoxia-pretreated mouse NSC-EVs (H-NSC-Exos)	Filtration + ultrafiltration	CDC42	PD	MPP+-treated HCMECs + MPTP PD mice	i.v.	Binds and inhibits ACSL4 via CDC42 (ferroptosis pathway)	↓ferroptosis; ↑angiogenesis; ↓vascular injury and PD progression	[154]
Human NSC (F3)-EVs	TFF + Total EV Isolation Kit	miR-182-5p, miR-183-5p, miR-9, let-7	PD	6-OHDA-treated SH-SY5Y cells + 6-OHDA PD mice	Intracerebral injection (ic; substantia nigra)	Suppresses apoptotic pathways; down-regulates pro-inflammatory signaling	↓intracellular ROS; ↓reactive glial activation; ↓dopaminergic neuronal loss; ↓inflammatory cytokines	[155]
Human NSC-EVs	ExoQuick-TC + ultracentrifugation	Down-regulates BACE1/PSEN1; up-regulates ADAM10	AD	SH-SY5Y cells + HMC3 glial cells (in vitro)	— (in vitro)	Inhibits β/γ-secretases (BACE1/PSEN1); activates α-secretase (ADAM10); down-regulates NF-κB/ERK/JNK signaling	↓Aβ and p-tau; ↓acetylcholinesterase; ↓inflammatory mediators; ↑cell viability	[156]
Microglia (BV2)-EVs	Gradient centrifugation	circZNRF1	PD	Paraquat-treated BV2 + MN9D co-culture (in vitro)	— (in vitro)	Sponges miR-17-5p via circZNRF1; up-regulates Bcl2	↓apoptosis; reversed neuronal viability loss; ↑Bcl2/Bax ratio	[157]
M2 microglia-EVs	Differential centrifugation + ultracentrifugation	miR-223	AD	APP/PS1 mice + in vitro cell models	i.v.	Mediates miR-223 sorting into EVs via YB-1	↓neuroinflammation; ↓nerve damage; improved cognition; ↑M2 microglia proportion	[158]
Astrocyte-EVs (A-Exo)	Filtration + size-exclusion chromatography	HepaCAM	ALS	SOD1G93A mice + in vitro excitotoxicity models	Intraspinal (stereotactic)	Protects axons against excitotoxicity via HepaCAM	IL-1α/TNF-α/C1q ↓A-Exo secretion and abolished neuroprotection; no EV toxicity	[159]
Astrocyte-EVs (ADEXs)	Ultracentrifugation	miR-200a-3p	PD	MPP+-treated SH-SY5Y/dopaminergic neurons (in vitro)	— (in vitro)	Targets MKK4 (Map2k4) via miR-200a-3p; inhibits JNK cell death pathway	↓MPP+-induced cell death	[160]
Lycium ruthenicum ELNs (LRM-ELNs)	PEG precipitation + freeze-drying	Plant miRNAs, proteins/lipids	AD	Aβ-induced PC12 cells (in vitro)	— (in vitro)	Acts via MAPK and PI3K/AKT signaling pathways	↓Aβ-induced PC12 apoptosis	[161]
Citrus lemon ELNs (EXO-CLs)	Differential ultracentrifugation	Antioxidant components	AD	SH-SY5Y oxidative stress model (in vitro)	— (in vitro)	Exerts antioxidant activity (comparable to ascorbic acid)	Crossed the BBB; non-toxic (>80% cell viability)	[162]
Salvia hairy root-EVs	Differential ultracentrifugation + SEC	Triterpenoids	PD	6-OHDA-treated SH-SY5Y cells (in vitro)	— (in vitro)	Inhibits 6-OHDA autoxidation and accumulation of toxic oxidative products	Preserved cellular metabolic homeostasis; ↓apoptosis; ↓oxidative stress	[163]
Gardenia-EVs (GDEVs)	Sucrose density-gradient ultracentrifugation	p38 MAPK/p53-modulating cargo	PD	Rotenone-induced PC12 cells + C. elegans PD model	— (in vitro); worm exposure	Decreases p38 MAPK/p53 phosphorylation; increases Bcl-2/Bax ratio	↑mitochondrial function; ↓cytochrome C release and apoptosis; ↓α-synuclein; ↑dopamine release and motility	[164]
Dihuang Yinzi plasma EVs (DHD-Exo)	Commercial plasma EV extraction kit	BDNF	AD	Scopolamine-induced cognitive impairment mice	i.v.	Delivers BDNF via EVs	↑cholinergic markers (ChAT); ↑synaptic plasticity markers (SYN-1, PSD95, M1 mAChR); improved learning and memory; ↓hippocampal neuronal loss	[165]
Ganoderma lucidum ELNs (GLENVs)	Differential ultracentrifugation	Ganoderic acids	AD	5×FAD mice	Intranasal	Inhibits JAK2/STAT3 signaling pathway	↓neuroinflammation; improved learning and memory; ↓Aβ deposition, overactivated microglia, reactive astrocytes and pro-inflammatory factors; no hepatic/renal toxicity	[166]
Treg-EVs	Ultracentrifugation + Exo-Quick	TLR4/NF-κB-inhibiting cargo	PD	MPP+-treated BV-2 microglia (in vitro)	— (in vitro)	Inhibits TLR4/NF-κB signaling	↓inflammatory response and oxidative stress; ↓ROS/MDA; ↑SOD activity; ↓IL-1β/IL-6/TNF-α	[167]
Ex vivo expanded Treg-EVs	PEG precipitation or TFF	Treg-associated markers	ALS	LPS-induced inflammation mice + mSOD1 ALS mice	i.v./Intranasal	Suppresses pro-inflammatory transcripts; up-regulates anti-inflammatory transcripts; modulates myeloid/Treg immune balance	↓peripheral inflammation and neuroinflammation; slowed disease progression; ↑survival; ↓spinal cord inflammation	[168]
HBMVEC-EVs	Sequential ultracentrifugation	P-glycoprotein (P-gp)	AD	Aβ42-injected mice + in vitro BBB model	i.v.	Captures Aβ via P-gp and promotes its efflux out of the brain	↑cerebral Aβ clearance; ↓cognitive dysfunction and hippocampal neuronal injury	[169]
Human umbilical cord blood-EVs (UCB-Exos)	Filtration + ultracentrifugation	HspB1/Ppef2-regulating cargo	PD	MPTP PD mice + MPP+-treated MN9D/SH-SY5Y cells	i.v.	Inhibits MAPK p38/ERK1/2 hyperphosphorylation; regulates HspB1/Ppef2 transcription	↓neuronal oxidative damage, senescence and energy metabolism disorders; improved motor function and cognition; ↓nigral pathology	[170]

Note: Upward arrows (↑) and downward arrows (↓) indicate increased and decreased levels, respectively.

Stem cell-derived EVs counteract proteinopathy by both suppressing the production of toxic aggregates and accelerating their degradation. In AD models, MSC-EVs reduce Aβ production at the source by activating the SphK/S1P signaling axis to inhibit the abnormal amyloidogenic cleavage.^142,171 In HD and ALS models, MSC-EVs interfere with the translational efficiency of mutant HTT (mHtt) and SOD1 mRNA, effectively lowering toxic protein synthesis.^140,141 Concurrently, they activate intracellular ALP. For instance, BMSC-EVs enriched with Growth Differentiation Factor 15 (GDF-15) activate the AKT/GSK-3β/β-catenin pathway to promote Aβ clearance,¹⁴⁴ while others facilitate the removal of insoluble SOD1 and mHtt inclusions via LC3-II-mediated mitophagy.^140,141

A defining attribute of stem cell-derived EVs is their ability to reprogram the hostile neuroinflammatory microenvironment and preserve neuronal viability. In PD and demyelination models, MSC-EVs effectively shift microglia from a pro-inflammatory state toward an anti-inflammatory, neuroprotective phenotype. This is achieved by inhibiting the TLR2/IRAK1/NF-κB pathway and reducing the release of neurotoxic cytokines such as IL-1β.^143,145 In Experimental Autoimmune Encephalomyelitis (EAE) models, bone marrow-derived MSC-EVs regulate miRNA profiles to inhibit the release of pro-inflammatory cytokines (TNF-α, IL-1β, IL-17) while promoting anti-inflammatory IL-10 and TGF-β expression and regulatory T-cell (Treg) differentiation.¹⁴⁶, By delivering specific miRNAs, such as miR-223, miR-134-5p and miR-1908-5p, these EVs target the PTEN/PI3K/Akt axis to inhibit neuronal apoptosis.¹⁴⁷ In 6-OHDA-induced PD models, MSC-EVs significantly reduce dopaminergic cell death by downregulating pro-apoptotic genes like BAD, CYCS, and TRAF2.¹⁴⁸ In male Sprague–Dawley rats with Aβ1-42-induced AD-like pathology, intravenously administered BMSC-derived EVs reduced ROS and MDA levels and TUNEL positivity in the hippocampal CA3 region. Mechanistic experiments implicated the miR-214-3p/CD151 axis in these effects.¹³⁹

Beyond direct neuroprotection, stem cell-derived EVs could also stabilize cellular energetics and the vascular scaffold. In PD models, MSC-EVs deliver miR-34b to upregulate Parkin expression, enhancing the mitophagy-mediated clearance of damaged mitochondria.^149,150 They also upregulate PGC-1α, restoring mitochondrial membrane potential and reversing metabolic dysfunction in HD mice.¹⁴⁰ NSC-derived EVs (NSC-EVs) and adipose-derived mesenchymal stem cell-derived EVs (ADSC-EVs) demonstrate a unique capacity to protect neurovascular unit (NVU). They reduce the permeability of the brain microvascular endothelial barrier and promote angiogenesis via SMAD3/p38 MAPK signaling.^151,152 Furthermore, hypoxia-preconditioned NSC-EVs have been shown to mitigate ACSL4-mediated ferroptosis in vascular cells, preserving the structural integrity of the BBB.^153,154

The therapeutic efficacy of MSC-EVs is currently being validated in human subjects (Table 2). For ALS, clinical studies (eg, IRCM-2021-296, NCT06598202) demonstrated the safety of both intravenous and intranasal administration.^172,173 An open-label, single-center, three-arm phase I/II dose-escalation trial of intranasal allogeneic adipose-derived MSC-EVs in AD (NCT04388982; n = 9) reported no adverse events during treatment and exploratory reductions in ADAS-Cog scores in the medium-dose group.¹⁷⁴ However, the study lacked a placebo control, and changes in amyloid or tau deposition did not differ significantly among the dose groups.

Table 2.

Summary of EV-Based Clinical Studies in NDDs

Study/Registry Number	Disease	Source	Administration Routes	Clinical Phase and Study Design	N	Sponsor (Country)	Key Outcomes (Safety/Efficacy)	Ref.
NCT04388982	AD (mild to moderate)	Allogeneic adipose MSC-derived exosomes (ahaMSCs-Exos)	Intranasal drip, twice weekly × 12 weeks; 5–20 μg protein per dose (low/medium/high-dose arms)	Phase I/II; open-label, single-center, non-randomized, sequential dose-escalation (3+3)	9	Ruijin Hospital (China)	No adverse events; medium-dose arm showed ADAS-Cog reductions of 2.33 (12 wk) and 3.98 (36 wk) from baseline; no significant amyloid/tau changes; less hippocampal volume loss in the medium-dose arm	[174]
Crose 2024 pilot study (IRB approval No.IRCM-2021-296; no trial registration number)	ALS	Human bone marrow MSC-derived EVs (ExoFlo™, Direct Biologics; ~0.6–0.8 trillion particles per 10 mL)	Intravenous infusion, 10 mL × 2 doses, 1 month apart; 3-month follow-up	Prospective, open-label, dual-site, single-investigator pilot safety study (uncontrolled)	10	Capitis Research Institute/Direct Biologics (USA)	No serious adverse events and no IP-related adverse events (4 subjects reported transient fatigue/cough/lightheadedness/paresthesia resolving within 48 h, deemed unrelated); 3/10 (30%) showed no ALSFRS-R decline over 3 months; group mean decline ~1 point/month (consistent with natural ALS progression)	[172]
NCT06598202	ALS	Human umbilical cord blood MSC-derived exosomes/small EVs (nasal drops)	Intranasal drip (dose-finding)	Phase I/II; randomized, quadruple-blind, placebo-controlled	38	Shengqi Medical Technology (Guangzhou) Co., Ltd.	Uncompleted	[173]
NCT01668849	OM induced by chemoradiation in HNC (supportive oncology care)	Grape-derived EVs (oral grape powder formulation)	Oral, once daily × 35 days (during chemoradiation)	Phase I; randomized, parallel, open-label; control = standard pain medication/mouthwash	60	University of Louisville (USA)	Completed, no results posted	ClinicalTrials.gov record (no published results)
NCT01294072	CRC (plant exosome-mediated curcumin delivery study)	Plant exosomes conjugated with curcumin (oral tablets)	Oral, 3.6 g/day × 7 days	Phase I (registered as N/A); randomized, three-arm (curcumin alone/curcumin + plant exosomes/no intervention), open-label	35 (planned)	University of Louisville (USA)	Recruiting/Active not recruiting	ClinicalTrials.gov record (no published results)

Neural Cell-Derived EVs

Compared to exogenous cell sources, neural cell-derived EVs (NC-EVs) offer a distinct therapeutic advantage through a “homologous compensation” mechanism.¹⁷⁵ Due to their native origin within the CNS, these vesicles possess an inherent affinity for neural cells, allowing them to be more readily internalized and to directly participate in the restoration of metabolic homeostasis.

EVs from NCs also demonstrate significant potential in modulating the pathological landscape of NDDs. NSC-EVs are significantly enriched with miRNAs (eg, miR-124, miR-137) and essential neurotrophic factors (eg, BDNF, GDNF). These molecules are pivotal effectors in modulating synaptic plasticity and driving neuroregeneration in damaged brain regions.^155,176 Beyond regeneration, NSC-EVs exert potent anti-inflammatory effects by specifically inhibiting NF-κB and extracellular signal-regulated kinase (ERK) signaling pathways within microglia, thereby preventing the secondary cascade of neurotoxicity.¹⁵⁶ EVs from mature CNS glial cells often act in a context-specific manner. Under oxidative stress, microglial EVs enriched with circZNRF1 can specifically promote mitophagy to clear damaged organelles, highlighting their role as “stress-responders”.¹⁵⁷ The functional profile of microglial EVs is highly dependent on the polarization state of the parent cell. For instance, anti-inflammatory microglial EVs carry miR-223, which has been shown to reduce Aβ load and neuroinflammation by inhibiting neuronal PTEN.¹⁵⁸ Astrocytes contribute to the neuroprotective shield by secreting EVs laden with protective signals such as hepatic and glial cell adhesion molecule (HepaCAM) or miR-200a-3p.^159,160 These payloads are critical for counteracting glutamate-induced excitotoxicity and suppressing apoptotic pathways in neighboring neurons.

Despite the high biological relevance and efficacy of animal-derived EVs, their path to large-scale clinical application is obstructed by several critical hurdles. These include limited source availability, the high cost and complexity of standardized manufacturing, and potential long-term immunogenicity or safety concerns.¹⁷⁷ Based on this, PDENs have emerged as a more promising direction. Their broad availability, cost-effectiveness, and naturally low immunogenicity position them as a promising “green” platform for the next generation of NDD therapeutics.¹⁷⁸

Plant-Derived EV-Like Nanoparticles (PDENs)

PDENs, first observed in carrot cell walls in 1996,¹⁷⁹ have emerged as highly complex, natural delivery systems shaped by long-term evolution. These nanovesicles (30–1000 nm in diameter) are enriched with plant-specific miRNAs, bioactive lipids, antioxidant proteins, and diverse secondary metabolites.^180,181 Crucially, PDENs exhibit the unique biological characteristic of “cross-kingdom regulation”, allowing them to be internalized by mammalian cells and precisely modulate key signaling pathways in NDDs.^182,183

PDENs function as efficient natural delivery platforms that surpass the efficacy of crude extracts by protecting their cargo from degradation and significantly enhancing cellular uptake.¹⁶¹ EVs derived from citrus and lemon exhibit exceptional antioxidant capacity, significantly alleviating neuronal oxidative stress and enhancing survival rates in stressed cell models.¹⁶² EVs from Lycium ruthenicum deliver plant-derived ata-miR156c-3p, which downregulates the pro-apoptotic protein Bax while upregulating the anti-apoptotic protein Bcl-2. This targeted delivery effectively shields neurons from Aβ-induced apoptosis, demonstrating superior performance compared to traditional phytochemical delivery.¹⁶¹

EVs derived from medicinal plant, or “medicinal PDENs”, also inherit the pharmacologically active substances of their parent herbs, representing a paradigm shift in the modernization of traditional herbal medicine (THM).¹⁸⁴ Therefore, these vesicles naturally integrate the THM therapeutic logic of “multi-component and multi-target” action at the nanoscale. EVs isolated from Salvia sclarea and S. dominica are enriched with triterpenoids such as asiatic acid, ursolic acid, and oleanolic acid, as well as antioxidant enzymes. These vesicles downregulate PINK1/PARK2 expression, alleviating mitochondrial damage and maintaining metabolic homeostasis in PD models.¹⁶³ Gardenia-derived EV-like nanovesicles (GDEVs) suppress the p38 MAPK/p53 signaling pathway and increase the Bcl-2/Bax ratio, reducing dopaminergic apoptosis and enhancing dopamine release.¹⁶⁴ Similarly, Dihuang Yinzi-derived EVs have been confirmed to carry bioactive BDNF protein, activating neurotrophic signaling.¹⁶⁵ Furthermore, Ganoderma lucidum-derived EVs (GLENVs) promote the clearance of Aβ by enhancing autophagic flux.¹⁶⁶

Beyond their biological potency, PDENs possess significant advantages for industrialization. The wide availability of plant raw materials markedly reduces production costs and ensures superior batch consistency compared to mammalian cell-based manufacturing. EVs from sources like grapes and ginger have already entered early-stage clinical trials for oncology (NCT01668849, NCT01294072, NCT03493984). These trials have preliminarily validated the safety and low immunogenicity of PDENs, laying a solid foundation for expanding their therapeutic indications to NDDs.

The diversity of EVs sources—ranging from stem cells and neural cells to medicinal plants—provides a versatile toolkit for intervening in neurodegenerative pathologies. Each source offers unique strategic advantages. Stem cell-derived EVs excel in immunomodulation and broad-spectrum repair. NC-EVs offer superior CNS-homing and “homologous compensation”. PDENs provide a scalable, low-immunogenic platform for the delivery of natural bioactive compounds. Beyond these, EVs from immune cells,^167,168 adipocytes,¹⁸⁵ vascular endothelial cells,^169,186 and blood¹⁷⁰ are also garnering attention for their source-specific functions. By virtue of their excellent biocompatibility, low immunogenicity, and inherent ability to cross the BBB, EVs are rapidly transitioning from basic research toward individualized clinical applications. They are poised to play a significant role in the precision treatment of NDDs.^187,188

Limitations of the Current Evidence

The evidence supporting source-specific EVs for NDD treatment remains predominantly derived from cell cultures and rodent models (Table 1). Cell culture studies help identify bioactive cargoes and molecular mechanisms but cannot reproduce the integrated effects of the BBB, systemic immunity, and whole-body pharmacokinetics. Rodent studies provide in vivo evidence of biological activity and functional improvement, although species differences and incomplete representation of human disease limit clinical extrapolation. Large-animal studies, where available, can provide additional information on biodistribution, safety, and therapeutic activity; however, findings from healthy animals or acute CNS injury models do not establish efficacy in chronic NDDs.

Clinical evidence remains limited. Safety and tolerability findings from healthy volunteers, where available, should be distinguished from evidence of therapeutic efficacy in patients with NDDs. In patients, the open-label study of intranasal allogeneic adipose-derived MSC-EVs in AD (NCT04388982) reported preliminary tolerability and exploratory cognitive findings, but included only nine participants and lacked a placebo control.¹⁷⁴ These findings are insufficient to establish efficacy or disease modification.

Cross-study comparisons are also limited by differences in EV isolation and characterization, dose definitions, administration routes, and outcome measures. Safety and activity reported for one preparation cannot be assumed for other EV sources or engineered variants, particularly with repeated administration. Findings should therefore be interpreted as mechanistic evidence, preclinical outcomes, or preliminary clinical observations according to the experimental model and study design. Adequately controlled clinical trials with sufficient follow-up and clinically meaningful endpoints are needed to establish therapeutic efficacy and long-term safety. The associated manufacturing, pharmacokinetic, and immunological challenges are discussed in the Translational Barriers and Clinical Considerations section.

Engineering Strategies to Improve EV Targeting and Therapeutic Delivery

While native EVs demonstrate promising biocompatibility and the innate potential to traverse the BBB, their clinical translation for NDDs is often constrained by insufficient targeting specificity, limited payload capacity, and unpredictable in vivo release kinetics. To overcome these bottlenecks, advanced engineering strategies—encompassing genetic, chemical, and physical approaches—have emerged to transform EVs into programmable neurotherapeutic platforms without compromising their intrinsic biological advantages (Figure 4) (Table 3).

Figure 4.

Surface engineering and cargo loading of EVs. (A) Surface functionalization through donor-cell genetic engineering, post-isolation covalent conjugation, or noncovalent lipid insertion. The schematic is divided into two color-coded modules. (B) Cargo loading through donor-cell-based approaches before EV release or post-isolation methods, including electroporation, sonication, and saponin-assisted membrane permeabilization. Arrows indicate the engineering workflow.

Abbreviations: CuAAC, copper(I)-catalyzed azide–alkyne cycloaddition; DSPE–PEG, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-polyethylene glycol; Lamp2b, lysosome-associated membrane protein 2b; miRNAs, microRNAs; MSCs, mesenchymal stem cells; RVG, rabies virus glycoprotein; siRNAs, small interfering RNAs.

Table 3.

Summary of Engineered EVs Strategies for CNS Targeting

Engineered EVs (Parental Source)	Engineering Approach	Specific Modification	Isolation Method	Key Markers/Cargo	Representative Disease Models	Experimental Models	Administration Routes	Therapeutic Mechanisms	Results	Ref.
RVG-EXO-CD10dm (HEK293)	Surface – genetic	RVG–hLAMP2b fusion + CD10dm (mutant neprilysin) co-expression	Differential ultracentrifugation	RVG; CD10dm	AD	Aβ-producing N2a cells; 20-mo aged mice	Intravenous injection (i.v).	Targets α7-nAChR-expressing neurons via RVG; degrades Aβ via CD10dm	α7-nAChR-dependent uptake; ↓Aβ40 in N2a cells; ↑hippocampal targeting in vivo; ↓IL-1α/TNF-α/NF-κB and ↑IL-10 in brain	[189]
Fe65-EXO-Cory-B (HT22 neurons)	Surface – genetic	Fe65 overexpression (pCI-Fe65); Cory-B loaded by sonication	Total Exosome Isolation reagent (polymer precipitation)	Fe65; corynoxine-B	AD	APP-HT22/N2a cells; 5xFAD mice	i.v.	Blocks APP–Fe65 signaling via Fe65–AICD binding; induces autophagy via Cory-B to clear Aβ/tau	↑BBB crossing and brain accumulation; ↑autophagy; ↓Aβ/amyloid plaques and neuroinflammation; ↑synapse formation and cognitive function	[190]
Dopa-EVs (ADSC)	Surface – chemical (covalent)	Dopamine grafted via EDC/sulfo-NHS amidation	Tangential flow filtration	Dopamine; native ADSC-EV	PD	α-Syn preformed fibrils neurons; 6-OHDA and preformed fibrils mice	i.v.	Mediates neuronal uptake via D1R/D2R; induces autophagy	↑brain accumulation; ↑dopaminergic neuroprotection; ↑motor function in both models	[191]
Rho-BR-EVs (rat brain EVs)	Surface – chemical (click)	Alkyne tagging (EDC/NHS) + CuAAC click of rhodamine-type mitochondria-targeting ligand	Ultracentrifugation	Mitochondria-targeting ligand (rhodamine-derived)	NDD/CVD (proof of concept)	Primary neurons; healthy rats	Intracerebroventricular injection (ICV)	Targets mitochondrial membranes via cationic mitochondria-targeting ligand	Mitochondrial colocalization in primary neurons; distribution in hippocampus and cortex; no detectable neuroinflammation	[192]
MSCEXO/PMA (HUMSC)	Surface – non-covalent	Zwitterionic PMA nanoparticles assembled on EXO (NO-driven nanomotor)	Ultracentrifugation	PMA artificial module; CD63/CD81+	PD	SH-SY5Y/BV2/bEnd.3 cells; MPTP mice	i.v.	Drives iNOS/ROS-triggered NO-mediated chemotaxis; targets lesions through a multi-step process; scavenges ROS	~72% in vitro BBB penetration; ↑PD-brain/SN targeting; ↓ROS, neuroinflammation and α-syn aggregates; ↑TH/GAP-43 and striatal dopamine; ↑motor function and spatial memory	[193]
Exo-VU01 (HEK293T)	Cargo – physical	VU0155069 (PLD1 inhibitor) loaded by electroporation/sonication	Differential ultracentrifugation + tangential flow filtration	VU0155069	AD/ADRD	Healthy CD-1 mice (PK study)	i.v.	Crosses the BBB as an EV “stealth” carrier; inhibits PLD1	↑brain Cmax (235.75→6822.98 pg/mL); ↑brain AUC (~24-fold vs free drug)	[194]
MSC-EVs-anta (rat BMSC)	Cargo – physical	miR-206-3p antagomir loaded by electroporation	Differential ultracentrifugation	miR-206-3p antagomir	AD	Primary cortical/hippocampal neurons; 5×FAD mice	Intranasal	Inhibits miR-206-3p and activates BDNF/TrkB signaling	↑neurite outgrowth, hippocampal neurogenesis and synaptic plasticity; ↓Aβ deposition; ↑learning and memory; superior to MSC-EVs or antagomir alone	[195]
Tom40-exosomes (HEK293)	Cargo – genetic	Tom40 packaged via XPack lentiviral system	PEG-NaCl precipitation	Tom40 protein	NDD (AD/PD)	H2O2-stressed HEK293 cells (in vitro only)	— (in vitro)	Restores mitochondrial protein import channels via Tom40	↑cell viability under 250–500 μM H2O2; ↑mitochondrial bioenergetics/transcription genes (ATP5B, NRF1, TFAM); ↑antioxidant genes (SOD1/SOD2)	[196]
RVG-CRISPRi-Exo (HEK293T)	Cargo – genetic (+ surface)	RVG-Lamp2b + dCas9-DNMT3A/SNCA-sgRNA co-loading; FUS-assisted	Differential ultracentrifugation	RVG; dCas9-DNMT3A + sgRNA	PD	MPP+-Neuro-2a cells; MPTP mice	i.v. + focused ultrasound (FUS)	Opens the BBB via FUS; targets neurons via RVG; methylates SNCA CpGs via dCas9-DNMT3A to ↓SNCA transcription	↑SNCA-specific methylation; ↓SNCA/α-syn and neuronal apoptosis; ↑motor performance, balance and neurosensitivity; ↑nigrostriatal synaptic function	[197]
C3/TPP-EXO(-CUR) (plasma EVs)	Hybridization (lipid anchor)	DSPE-PEG2000-C3/TPP insertion by incubation; curcumin encapsulated	Ultracentrifugation	C3 + TPP; curcumin	AD	Okadaic acid-induced HT22 cells; Tau P301S mice	i.v.	Targets neurons via C3→NCAM; targets mitochondria via TPP; exerts anti-tau and antioxidant effects via CUR	↓tau phosphorylation and neuronal apoptosis; ↑mitochondrial and synaptic protection; ↑cognitive function	[198]
H/Exos (MSC × neutrophil HCs)	Hybridization (cell fusion)	PEG1500-mediated MSC–neutrophil fusion→hybrid-cell-derived EVs	Ultracentrifugation	Sca-1/CD44 (MSC) + LFA-1 (neutrophil)	AD	APP/PS1 mice; Aβ1-42-injected mice	i.v.	Targets inflamed endothelium via LFA-1/ICAM-1; promotes microglial phagocytic polarization	↑AD-lesion homing and microglial Aβ clearance; ↓neuroinflammation; ↑neural repair and cognitive function (Morris water maze)	[199]

Note: Upward arrows (↑) and downward arrows (↓) indicate increased and decreased levels, respectively.

Surface Functionalization: Optimizing BBB Permeability and Homing

Surface engineering aims to endow EVs with “navigation systems” capable of bypassing the BBB and precisely recognizing pathological lesions in the CNS. Genetic engineering is a “design-at-the-source” strategy that utilizes the biosynthetic machinery of donor cells to present targeting ligands on the EV surface, typically by fusing targeting peptides with EV transmembrane proteins such as Lamp2b or CD63. The fusion of Rabies Virus Glycoprotein (RVG) with Lamp2b yields EVs that specifically recognize nicotinic acetylcholine receptors on neurons, significantly enhancing delivery efficiency across the BBB.¹⁸⁹ Similarly, overexpression of the adaptor protein Fe65 in HT22 hippocampal neuronal cells generated Fe65-enriched EVs (Fe65-EXO). These EVs showed preferential uptake by APP-overexpressing neuronal cells in vitro and increased brain accumulation following intravenous administration in 5xFAD mice. When further loaded with the autophagy inducer corynoxine-B by sonication, the resulting formulation (Fe65-EXO-Cory-B) enhanced autophagy, reduced Aβ pathology, and improved cognitive performance in 5xFAD mice.¹⁹⁰

Compared to genetic remodeling, chemical modification offers a more flexible and universal path for post-isolation functionalization, enabling the stable anchoring of small molecules or synthetic probes. Via a covalent grafting strategy, Dopa-EVs were engineered by conjugating dopamine onto the surface of MSC EVs through EDC/Sulfo-NHS amidation. This modification leverages the intrinsic affinity of dopamine for D1/D2 receptors, thereby achieving significant brain enrichment in PD models.¹⁹¹ Utilizing Copper-catalyzed Azide-Alkyne Cycloaddition (CuAAC)—a highly efficient and site-specific form of Click Chemistry—to attach mitochondria-tropic modules, researchers developed Rho-BR-EVs. These vesicles exhibit high mitochondrial localization in AD models, exerting neuroprotective effects without inducing undesirable inflammatory responses.¹⁹²

Non-covalent modifications, on the other hand, achieve reversible integration of functional modules through hydrophobic interactions or electrostatic adsorption. A notable example is the integration of a nitric oxide (NO)-driven nanomotor with MSC-EVs. They were activated by NO release within the high-ROS environment of NDD lesions, enabling autonomous chemotaxis and enhanced BBB penetration to mitigate oxidative stress.¹⁹³

Cargo Loading: Empowering the Therapeutic Payload

Optimizing the encapsulation of therapeutic agents—ranging from lipophilic small molecules to complex gene-regulatory tools—is essential for maximizing therapeutic potency. Physical loading methods employ external energy of mechanical forces to temporarily disrupt the EV membrane, thereby facilitating the incorporation of therapeutic cargo. Electroporation was used to load the PLD1 inhibitor VU0155069 into HEK-293T-derived EVs. In male CD-1 mice, a single intravenous tail-vein dose of this formulation (Exo-VU01; 1 mg/kg VU0155069-equivalent) produced approximately 29-fold higher peak brain concentration (Cmax) and 24-fold greater area under the brain concentration–time curve up to the last quantifiable time point (AUC0–t) than free VU0155069 at the same drug dose. Brain homogenate drug concentrations were quantified by LC–MS/MS over 24 h.¹⁹⁴ Similarly, loading miR-206-3p antagonists via electroporation into MSC-EVs enables the activation of the BDNF/TrkB signaling pathway, promoting neurogenesis and cognitive recovery.¹⁹⁵ The introduction of externally controllable modules, such as superparamagnetic iron oxide nanoparticles (SPIONs), creates “intelligent” systems like Spion-Ex. Under an external magnetic field, these EVs are actively pulled across the BBB and enriched in brain lesion areas, significantly improving post-stroke cognitive recovery.¹³³

Genetic engineering allows for the stable encapsulation of complex proteins or gene-editing systems during vesicle biogenesis. Overexpressing the mitochondrial protein Tom40 in donor cells yields Tom40-EVs, which replenish neuronal mitochondrial entry channels and restore energy metabolism.¹⁹⁶ The “RVG-CRISPRi-Exo” platform successfully delivers the dCas9-DNMT3A complex to neurons. This allows for the epigenetic downregulation of SNCA expression without altering the genomic sequence, significantly mitigating motor deficits in PD models.¹⁹⁷

In addition, hybridization strategies, such as the C3/TPP-EXO platform, synergistically enhance BBB crossing and mitochondrial enrichment with curcumin delivery.¹⁹⁸ Furthermore, EVs–liposome hybrid and cell-cell hybrids (eg, MSC-neutrophil multifunctional EVs (H/EVs)) have demonstrated superior disaggregation of α-syn and enhanced microglial modulation by shifting microglial phenotypes.^199,200

In summary, the evolution of engineered EVs is moving toward multimodal, programmable tools, representing a landmark of precision medicine in NDDs. Although challenges in scalable and standardized manufacturing remain, these engineering techniques provide an experimental basis for developing EV-based therapies for NDDs.

Translational Barriers and Clinical Considerations

Following systemic administration, EVs interact with plasma components and are rapidly sequestered by the mononuclear phagocyte system (MPS; historically termed the reticuloendothelial system), particularly through uptake by hepatic Kupffer cells and splenic macrophages. Surface features such as exposed phosphatidylserine can facilitate macrophage recognition and internalization, followed by intracellular degradation, resulting in rapid blood clearance and preferential accumulation in the liver and spleen. Representative intravenously administered EV preparations in mice have exhibited circulation half-lives of approximately 2–7 min.^201,202 However, these values vary with EV source, surface composition, dose, administration route, species, disease state, and labeling or tracking method. Product-specific PK studies should therefore distinguish intact EVs from released cargo or free labels and characterize circulation, biodistribution, and clearance, whereas PD studies should relate tissue exposure and target engagement to therapeutic responses and toxicity.

Extensive chemical modification or lipid insertion may alter EV surface properties and membrane stability or introduce immunogenic components. These changes may activate complement and promote opsonization and MPS-mediated uptake in the liver and spleen, thereby shortening circulation and reducing the delivery of intact EVs across the BBB, particularly after repeated administration.^203,204 The type and extent of surface modification should therefore be carefully optimized and evaluated for complement activation, immune responses, membrane stability, pharmacokinetics, biodistribution, and BBB delivery.

Before clinical translation, engineered EVs should undergo risk-based immunotoxicity assessment, including cytokine-release assays using human whole blood or peripheral blood mononuclear cells (eg, IL-6, TNF-α, and IFN-γ);²⁰⁵ measurement of complement activation products (eg, C3a, C5a, and sC5b-9);^203,204 and flow-cytometric profiling of major immune-cell subsets and activation states.^206,207 For products intended for repeated administration, studies should also assess systemic cytokine levels, anti-drug antibodies, histopathological changes in immune organs, and changes in pharmacokinetics, including accelerated clearance.

Viral vector-based therapies and mRNA-based nanomedicines are two other rapidly advancing approaches being explored for NDD treatment. AAV vectors can support sustained transgene expression but face anti-vector immunity and redosing constraints, whereas LNP-mRNA formulations offer transient expression and scalable manufacturing but face formulation-dependent inflammatory risks and CNS delivery challenges. EVs may combine source-specific biological activity and multi-component cargo delivery with favorable biocompatibility. However, surface engineering intended to improve EV targeting may introduce immunogenic components or alter membrane properties, potentially promoting antibody responses, complement activation, and MPS-mediated clearance that offset delivery gains, particularly during repeated administration.

Conclusion and Future Perspectives

The interconnected pathological mechanisms of NDDs complicate treatment with single-target therapies.^24,25 Source-specific EVs offer potential for combining biological activity with therapeutic cargo delivery.¹⁷ Their proteins, lipids, and regulatory RNAs may modulate multiple pathological pathways, while surface modification and cargo loading can improve CNS targeting and therapeutic delivery.¹⁸⁹ Despite encouraging preclinical results, clinical translation remains at an early stage.¹⁷⁴

Clinical translation requires standardized isolation and characterization, reproducible potency assays, scalable GMP-compliant manufacturing, and validated storage conditions (Figure 5). To support batch-to-batch consistency and reproducibility, quality control of EV products intended for clinical translation should include standardized assessments of particle-size distribution (eg, by DLS, NTA, or RPS), particle concentration (eg, by NTA or RPS), morphology, and zeta potential as an indicator of surface charge and colloidal stability.²⁰⁸ In addition, product-specific PK/PD studies defining MPS-mediated clearance, circulation half-life, biodistribution, exposure-response relationships, dosing regimens, and safety are essential before clinical translation. Non-native surface components, including synthetic targeting peptides, virus-derived fusion proteins, and bioorthogonal chemical anchors, may introduce immunogenic neoepitopes and elicit binding or neutralizing antibodies.²⁰⁹ This risk may increase with the repeated systemic administration required for long-term treatment of NDDs; repeated-dose studies should therefore assess antibody formation and its effects on EV clearance, BBB delivery, and therapeutic activity.

Figure 5.

Manufacturing considerations and a proposed clinical development pathway for EV-based neurotherapeutics. (A) Source-dependent production and isolation, including mammalian cell expansion in bioreactors and mechanical processing of plant tissues. (B) Quality control and formulation/storage, including assessments of identity, purity, particle size and concentration, membrane integrity, and potency, together with cryopreservation or lyophilization. (C) Preclinical evaluation, including pharmacokinetic and pharmacodynamic profiling, followed by Phase I–III clinical trials and regulatory review.

PDENs represent an emerging alternative source that may help address some cost and scalability constraints of mammalian cell-derived EVs.^210–213 However, yield comparisons require clearly defined metrics and consideration of plant species, tissue source, and extraction and purification methods. Moreover, their suitability for oral or intranasal delivery, including cargo stability and CNS exposure, requires preparation- and route-specific validation.¹⁷⁸ Looking forward, Artificial Intelligence (AI) and machine learning may support the prediction of ligand-receptor interactions and cargo optimization, helping prioritize candidates for experimental testing.²¹⁴ Their value for EV engineering will depend on reliable datasets and experimental validation of loading efficiency, CNS delivery, therapeutic activity, and safety.

In summary, source-specific and engineered EVs offer a highly versatile approach for CNS drug delivery and may enable the modulation of multiple pathological pathways involved in NDDs.¹⁷ However, most supporting evidence remains preclinical, and their comparative efficacy, safety, and clinical benefits have not yet been established. Well-designed preclinical studies and adequately controlled clinical trials are needed to determine their therapeutic value and long-term safety.

Funding Statement

This work was supported by the National Natural Science Foundation of China (22507092, 22377021), the Joint Funds of the Zhejiang Provincial Natural Science Foundation of China (LKLY25H290002), and Startup grants from Wenzhou Institute, University of Chinese Academy of Sciences (WIUCASQD2019002, WIUCASQD2023029).

Data Sharing Statement

No new datasets were generated or analyzed in this review. All information presented is derived from the published studies cited in the manuscript; therefore, data sharing is not applicable.

Disclosure

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.