IS occurs when cerebral blood vessels supplying the brain become obstructed. Occlusion caused by locally formed clots within cerebral vessels is termed cerebral thrombosis, which is closely associated with lipid deposits in atherosclerotic arteries. Alternatively, clots originating from extracranial sites, termed cerebral embolism, may travel through the bloodstream to cerebral vessels, with atrial fibrillation being the primary etiology (42). Due to the structural characteristics of cerebral vasculature and its hierarchical perfusion network, vascular occlusion typically manifests as focal neurological deficits confined to specific brain regions supplied by the affected vessel. The central core region experiences near-complete cessation of cerebral blood flow (CBF), resulting in rapid cellular necrosis within minutes. Surrounding this core lies a transitional zone termed the “penumbra,” where reduced perfusion permits temporary cellular survival. This salvageable tissue constitutes the primary target for neuroprotective interventions (43). Following temporary CBF deprivation in the penumbra, reperfusion restores oxygen and nutrient supply but paradoxically induces significant secondary tissue damage through ischemia-reperfusion (I/R) injury (42, 44). This dual-phase injury mechanism is not unique to cerebral ischemia but has been well-documented in other organ systems, including myocardial infarction and acute kidney injury (45–48).

Reperfusion injury constitutes a multifactorial process involving sustained neuroinflammatory responses, oxidative stress, excitotoxicity, calcium overload, and their intricate interconnections, collectively exacerbating post-ischemic cerebral damage (16, 49, 50). The neuroinflammatory cascade in IS strictly fulfills the four diagnostic criteria of neuroinflammation while orchestrating a sophisticated interplay between tissue destruction and repair mechanisms. As summarized in Figure 1, this interplay is intricately regulated by a network of non-coding RNAs, particularly circRNAs and lncRNAs, which function through diverse mechanisms, including sponging microRNAs and directly interacting with signaling proteins, to precisely modulate key inflammatory pathways and cell death processes such as pyroptosis. Notably, the term “neuroinflammation” suffers from conceptual overgeneralization in practical applications. Its ambiguous definitional boundaries not only risk misinterpretation of pathological mechanisms but may also induce cognitive biases among novice researchers. To address this conceptual ambiguity, previous studies have systematically defined four diagnostic hallmarks based on peripheral inflammation parallels: microglial activation, brain-specific infiltration of peripheral immune cells, significant elevation of inflammatory mediators, and secondary neurodegeneration (51–53). These pathological processes may induce either necrosis or programmed apoptosis, with the specific mode determined by injury severity and neuronal metabolic status. In post-ischemic inflammatory responses, damaged and dying cells release damage-associated molecular patterns (DAMPs) that initiate immune activation and exert critical regulatory functions (54). These DAMPs precisely engage pattern recognition receptor clusters, including Toll-like receptor (TLR) families and scavenger receptors, on microglia and cerebrovascular endothelial cells. Functioning as biological radars within the sterile microenvironment devoid of exogenous pathogens, they breach the molecular silence of the BBB to activate inflammatory responses across neurovascular unit components: microglia initiate phagocytic programs, astrocytes release chemokines, and endothelial cells deploy adhesion molecules to establish molecular pathways for subsequent leukocyte transmigration (55–57).

In summary, inflammation is a recognized core regulatory axis in stroke pathology, functioning as an integral component of I/R-induced cascades. It persists throughout the entire pathological process from onset to post-injury repair, exhibiting dual regulatory characteristics of damage-repair homeostasis across different post-stroke phases (58, 59). This process can be divided into three stages:

While glial scarring is essential for reconstructing vascular networks and sealing lesion areas to restore tissue integrity, it concurrently induces axonal regeneration barriers that significantly impair central nervous system functionality. Clinical studies demonstrate that elevated inflammatory levels correlate with poorer therapeutic outcomes in IS, establishing inflammation as a critical prognostic factor. For instance, early measurements of C-reactive protein (CRP) effectively predict functional disability and mortality in acute IS patients (61). The detrimental effects of inflammation extend beyond these observations. In focal and global cerebral I/R injury, intracranial hemorrhage, or mechanical brain trauma, inflammatory responses consistently exacerbate secondary damage despite incomplete mechanistic understanding (57). Current evidence suggests that targeted modulation of neuroinflammation may provide protective effects in IS, opening new therapeutic avenues. Rational immunomodulation could suppress pathological damage while enhancing reparative functions, though potential side effects require caution. Future research should clarify molecular mechanisms and signaling pathways underlying inflammatory regulation, precisely distinguish beneficial versus harmful targets, and develop safer therapeutic strategies to improve patient prognosis.

The core contradiction in acute IS treatment lies in the spatiotemporal limitations of recanalization and associated therapeutic risks. Current strategies primarily include IVT and EVT.

Antithrombotic therapy is a cornerstone of primary and secondary IS prevention, utilizing antiplatelet agents (e.g., aspirin, clopidogrel) or anticoagulants (e.g., warfarin, non-vitamin K antagonist oral anticoagulants [NOACs]) to suppress thrombogenesis (12). Treatment strategies are tailored to thrombotic etiology: antiplatelet regimens are favored for non-cardioembolic IS subtypes, such as atherosclerotic and small vessel disease (73, 74), while anticoagulation is preferred for cardioembolic origins, such as atrial fibrillation, or hypercoagulable states (75, 76). However, all antithrombotic approaches require a careful balance between ischemic protection and hemorrhagic risk, particularly gastrointestinal and intracranial hemorrhage. Effective clinical decision-making involves: Personalized bleeding risk assessment using validated scoring systems; Continuous monitoring of hemostatic parameters; and Prompt regimen adjustments based on the recurrence of thrombotic or hemorrhagic events. This delicate risk-benefit balance highlights the critical need for precision medicine in IS prophylaxis.

Current therapeutic strategies for IS emphasize rapid reperfusion through IVT and EVT, paired with etiology-guided antithrombotic management using antiplatelet or anticoagulant therapy. However, their clinical efficacy is limited by strict time windows, eligibility criteria, and an inability to mitigate neuroinflammation-driven secondary injury following reperfusion (77, 78). Increasing evidence positions inflammation as a central regulator in I/R cascades and a key mediator of risk factors. Notably, inflammatory processes compromise neurovascular unit (NVU) integrity prior to IS onset and continuously worsen secondary damage post-stroke (57, 79, 80). After reperfusion, aberrant neuroimmune responses initiate uncontrolled inflammatory cascades via mechanisms like inflammasome activation and mitochondrial dysfunction, resulting in BBB disruption, reactive oxygen species (ROS) bursts, and subsequent neuronal apoptosis and glial activation (81–83). Recent research identifies ncRNAs as critical regulators of post-reperfusion neuroinflammation, modulating neuroimmune interactions, suppressing inflammatory signaling pathways (e.g., NF-κB), and preserving NVU homeostasis (84). Their dynamic expression profiles, such as miR-124 and lncRNA MALAT1, not only act as early biomarkers of injury but also provide innovative therapeutic opportunities to overcome time window constraints through targeted delivery systems, such as exosomal vectors, facilitating precise anti-inflammatory interventions (32, 85–87). In this regard, stem cell-derived exosomes have emerged as highly effective ncRNA carriers, owing to their BBB permeability and biocompatibility. Experimental studies demonstrate their ability to deliver functional ncRNAs to ischemic lesions efficiently, maintaining bioactivity and coordinating molecular regulation in affected regions (88–90) (Figure 2). This advancement lays a theoretical groundwork for multidimensional strategies that integrate reperfusion, anti-inflammation, and neuroprotection, potentially overcoming the fundamental limitations of conventional IS therapies.

Although members of the ncRNA family share common anti-neuroinflammatory functions, their mechanisms of action differ significantly (Table 1). As a key member of this family, circRNAs are covalently closed, single-stranded circular molecules formed through back-splicing of precursor RNAs (91). Based on their biogenesis mechanisms and structural features, circRNAs can be classified into four major subtypes: intron-derived ciRNAs (circular intronic RNAs), exon-intron chimeric EIciRNAs (exon-intron circRNAs), exon-cyclized ecircRNAs (exonic circRNAs), and tRNA-processing-derived tricRNAs (tRNA intronic circRNAs) (92, 93). Current studies reveal that circRNAs exhibit diverse functions, including acting as miRNA sponges, interacting with RNA-binding proteins (RBPs), regulating alternative splicing, transcription, and translation, generating pseudogenes, and participating in cellular transport and communication (93). In cerebral I/R injury, circRNAs primarily function as dynamic regulators of neuroinflammation. As evidenced by mechanistic studies, they not only directly modulate inflammatory responses but also critically sequester miRNAs through sponge activity to fine-tune inflammation levels, thereby steering neuroinflammatory progression (94, 95). The following representative examples illustrate this regulatory duality:

CDR1as/ciRS-7 exemplifies a unique circRNA-mediated neuroprotective mechanism. Enriched with miR-7 binding sites, it stabilizes miR-7 via incomplete complementary pairing, extending its biological half-life. This interaction suppresses α-synuclein (α-Syn), a key regulator of the TLR4/NF-κB pathway, ultimately attenuating neuroinflammation, reducing neuronal apoptosis, and enhancing motor function recovery in ischemic models (96). circ_0000831 represents another pivotal anti-inflammatory circRNA, characterized by its downregulation in stroke patients. By serving as a molecular sponge for miR-16-5p, it activates adiponectin receptor 2 (AdipoR2) and the PPARγ signaling pathway, significantly alleviating neuroinflammation, vertigo, neurological deficits, and apoptosis in MCAO mice. Dual-luciferase assays confirm its direct binding to miR-16-5p, solidifying its mechanistic role (97).

Conversely, a distinct class of pro-inflammatory circRNAs exacerbates ischemic injury. For example, circ_0129657 competitively binds miR-194-5p, relieving its suppression of glia maturation factor beta (GMFB). This derepression promotes endothelial apoptosis and inflammatory cytokine secretion, while its silencing reduces cerebral infarct volume and ameliorates neurological dysfunction (98). circZfp609 mediates damage through a dual regulatory network: it serves as a sponge for miR-145a-5p both to suppress NF-κB inhibition and upregulate BACH1, thereby promoting apoptosis and pro-inflammatory factor release (e.g., IL-1β, TNF-α). Genetic knockout of circZfp609 alleviates cerebral ischemia/reperfusion injury and inhibits OGD/R-induced cellular damage by modulating astrocyte apoptosis and inflammatory responses (99, 100). Other pro-inflammatory circRNAs—including circ_0007290 (via miR-496/PDCD4), circ_0000566 (targeting miR-18a-5p), circ_0008146 (miR-709/CX3CR1 axis), and circ_0000647 (miR-126-5p/TRAF3)—likewise function as miRNA sponges to amplify ROS bursts, cytokine secretion, and endothelial injury (101–104).

In summary, circRNAs critically orchestrate neuroinflammatory responses post-stroke, primarily through sponge-mediated miRNA regulation. While anti-inflammatory members (e.g., CDR1as, circ_0000831) stabilize protective miRNAs to suppress inflammatory cascades, their pro-inflammatory counterparts (e.g., circ_0129657, circZfp609) sequester repressive miRNAs to activate apoptotic and inflammatory pathways. This functional duality positions circRNAs as promising therapeutic targets for modulating post-stroke neuroinflammation.

LncRNAs are non-coding RNA molecules exceeding 200 nucleotides (nt) in length. Thousands of lncRNAs have been identified in the human genome, regulating gene expression through chromatin modification and post-transcriptional regulatory mechanisms (105–107). LncRNA dysregulation is closely associated with pathological processes in various diseases. The central nervous system harbors the most extensive repertoire of lncRNA transcripts, which participate in multidimensional regulation of neural development and homeostasis maintenance. Studies using microarray and high-throughput RNA sequencing technologies have identified hundreds of differentially expressed lncRNAs in IS patients and animal models. These expression profile abnormalities are strongly correlated with pathological processes such as BBB disruption and neuroinflammation (108–110).

At the neuroimmune interface, lncRNA Maclpil exhibits unique properties by blocking monocyte-derived macrophage (MoDM) activation and migration through inhibition of LCP1-mediated actin dynamics. Crucially, LCP1 knockout fully replicates Maclpil’s neuroprotective effects, confirming LCP1 as its key effector. Consistent with this, animal studies demonstrate that this axis reduces infarct volume and improves neurological recovery (111). Similarly, KLF3-AS1 stands out for its dual regulatory mechanisms. In cerebral I/R models, its downregulation correlates with elevated pro-inflammatory factors (TNF-α, IL-6, IL-1β). Mechanistically, KLF3-AS1 suppresses neuroinflammation via the Sphk1/NF-κB pathway (112). Furthermore, when delivered by BMSC-derived exosomes, it competitively binds miR-206 to upregulate USP22, promoting Sirt1 deubiquitination and extending its half-life to mitigate inflammation and apoptosis (113). Other protective lncRNAs (e.g., OIP5-AS1, SERPINB9P1, SNHG16) similarly modulate ceRNA networks, chaperone activity, and oxidative pathways to mitigate injury (114–116).

Conversely, some lncRNAs drive pathology. lncRNA H19 demonstrates significant clinical relevance, with its expression in neutrophils of IS patients positively correlating with leukocyte activation. Mechanistically, H19 acts as a ceRNA for miR-29b, relieving its inhibition of C1QTNF6 and driving pro-inflammatory factor release (IL-1β, TNF-α), thereby exacerbating BBB disruption and ischemic damage (117). Similarly, lncRNA XIST exhibits striking cell-type specificity in neurons. It sponges miR-25-3p to elevate TRAF3 levels, activating NF-κB-mediated inflammatory responses. Supporting its detrimental role, XIST silencing inhibits neuronal apoptosis and cytokine release, while exogenous TRAF3 overexpression reverses this protection (118).

In summary, as lncRNAs bridge epigenetic regulation and immune dynamics, they represent not just therapeutic targets but potential “molecular switches” for reprogramming neuroinflammation, offering a promising platform for developing lncRNA-mediated precision medicine strategies in stroke recovery.

MiRNAs are small non-coding RNA molecules approximately 20–24 nucleotides (nt) in length. They exert critical post-transcriptional regulatory roles by mediating target mRNA degradation or translational repression through incomplete base complementarity. Notably, a single miRNA can regulate the translation of hundreds of mRNAs, demonstrating its broad functional regulatory potential (119–122). Given their ability to directly target key inflammatory factor mRNAs, miRNAs have become a research focus in I/R injury in recent years. For example, multiple studies confirm that targeting specific miRNAs effectively modulates I/R-associated neuroinflammatory cascades, offering novel therapeutic avenues for IS.

Specific examples highlight this therapeutic potential. miR-193a-5p, for example, orchestrates neutrophil immune polarization. In acute IS patients, its significant downregulation in circulating neutrophils disrupts PPARγ-mediated anti-inflammatory signaling. Mechanistically, miR-193a-5p targets UBE2V2 to inhibit PPARγ ubiquitination, promoting N2 phenotype polarization and reducing pro-inflammatory cytokine release. Importantly, rescue experiments confirm that UBE2V2 knockdown reverses neutrophil hyperactivation in miR-193a-5p-deficient models (123). Recent advances in exosome delivery strategies have opened new avenues for miRNA-based therapies. For instance, exosomes derived from recombinant human growth differentiation factor 7 (rhGDF7)-preconditioned bone marrow mesenchymal stem cells (BMSCs) are enriched with miR-369-3p. This miRNA targets the 3’-untranslated region (3’-UTR) of the Pde4d gene, suppressing PDE4D protein expression and activating the AMPK signaling pathway, thereby significantly inhibiting I/R-induced neuronal inflammation and oxidative stress damage (124). Similarly, miR-6328 exerts critical anti-inflammatory effects in I/R injury by targeting the IKKβ/NF-κB signaling axis. Mechanistically, this miRNA binds to IKKβ to inhibit its expression, thereby blocking NF-κB pathway activation. Notably, sequencing at 72 hours post-I/R revealed elevated miR-6328 levels accompanied by reduced IKKβ, and reverse functional experiments further validated the specificity of its anti-inflammatory effects (125). In addition to these, other protective miRNAs (e.g., miR-124-5p, miR-30c-5p, miR-182-5p, miR-149) orchestrate anti-inflammatory responses primarily by suppressing neuronal apoptosis and reinforcing BBB integrity (126–129).

However, miRNA function is not context-independent. A prominent example underscoring the critical, yet often overlooked, role of age and sex factors in RNA biology is miR-15a/16-1. Significantly upregulated post-stroke and correlating with neuronal injury severity, studies in tMCAO models reveal a striking age-sex interaction: young female mice (3-month-old) exhibit the strongest neuroprotective phenotype, characterized by reduced miR-15a/16–1 induction. Combined antagonism with estradiol synergistically downregulates genes driving neurovascular death, inflammation, and oxidative stress. miR-15a/16–1 target genes further display marked age- and sex-dependent expression, solidifying their biomarker potential for tailored stroke therapy (130). Crucially, while most preclinical stroke studies utilize healthy rodent models, translating findings to human patients requires attention to human-specific factors. Recent investigations focusing on patient-derived biomarkers reveal critical mechanisms specific to human pathology. For instance, circulating miR-30c-2-3p is uniquely upregulated in plasma exosomes of acute ischemic stroke (AIS) patients. It establishes a pro-inflammatory network by targeting SMAD2, a core anti-inflammatory component of the TGF-β pathway, suggesting exosome-mediated miR-30c-2-3p delivery as a key driver of post-stroke neuroinflammation (131, 132). Notably, research delving into specific stroke subtypes uncovers further complexity. Research on the large artery atherosclerotic (LAA) stroke subtype identifies plasma exosome-derived miR-Novel-3 as a dual pathogenic factor. Specifically overexpressed in LAA-AIS patients, it concurrently drives neuroinflammation and ferroptosis by targeting STRN protein in peri-infarct microglia/macrophages, thereby suppressing PI3K-AKT-mTOR signaling to trigger inflammatory mediator release and ferroptosis-related molecular dysfunction, ultimately exacerbating ischemic neuronal injury (133).These findings collectively highlight the necessity of incorporating patient-specific complications (e.g., atherosclerosis) into stroke models to uncover clinically relevant pathological cascades. Additional detrimental miRNAs (e.g., miR-9-5p, miR-188-5p) exacerbate neuroinflammation through dysregulated TLR signaling, driving maladaptive pro-inflammatory cascades (134, 135).

Microglia originate from erythromyeloid progenitors (EMPs) in the early embryonic yolk sac. Their development occurs independently of the hematopoietic stem cell-mediated system, culminating in their differentiation into tissue macrophages resident in the brain parenchyma (136–138). As the central nervous system’s primary immune effector cells, microglia constitute approximately 10% of total neural cells and dynamically maintain neuroimmune homeostasis to support brain function (139–141).

In IS, microglia display two distinct polarization states:

In stroke, excessive DAMP stimulation skews the M1/M2 balance toward the pro-inflammatory M1 phenotype. Encouraging M2 polarization can mitigate inflammation and enhance tissue repair, improving post-stroke outcomes (53, 148). Consequently, modulating microglial polarization—particularly toward the M2 state—represents a key strategy for reducing I/R injury. Among these approaches, ncRNA-mediated regulation stands out for its precision and capacity to target multiple pathways simultaneously (Table 2).

Pyroptosis, an inflammatory form of programmed cell death mediated by Gasdermin proteins, is characterized by plasma membrane rupture and the release of pro-inflammatory factors (161, 162). Unlike apoptosis, which maintains membrane integrity, pyroptosis induces the efflux of cytoplasmic inflammatory contents through membrane pores, culminating in secondary necrosis—a pathological state of altered membrane permeability following complete cell death (161, 163–165).

The Gasdermin protein family comprises six members (GSDMA, GSDMB, GSDMC, GSDMD, GSDME, and GSDMF/PJVK/DFNB59), serving as core executors of inflammatory responses and disease progression (166).

Pyroptosis is closely linked to the initiation and progression of stroke. Post-stroke, Gasdermin-mediated pyroptosis displays cascading amplification effects within the neurovascular unit (167, 168). When pyroptosis is activated in neuroglia, neurons, and brain microvascular endothelial cells (BMECs), membrane perforation leads to the explosive release of inflammatory mediators like IL-1β and IL-18. This release generates self-amplifying cytokine storms (169, 170). This pathology exacerbates brain injury through two primary mechanisms. First, pyroptosis disrupts the physical barrier by dismantling the tight junction structures of the BBB, leading to plasma protein leakage and the infiltration of toxic substances into the brain parenchyma. Second, it causes molecular signaling dysregulation, where extracellular DAMPs activate microglia via pattern recognition receptors such as TLRs. This activation drives the spread of neuroinflammation to the ischemic penumbra, accelerating the conversion of reversible injury zones into infarct cores (171–174).

Recent advances in understanding the role of pyroptosis in stroke pathophysiology have shown that ncRNAs regulate pyroptosis signaling pathways at multiple levels by targeting critical components such as NLRP3 inflammasome activation and Caspase-1 processing (Table 3). This complex epigenetic regulation provides new insights into disrupting the vicious cycle of pyroptosis, neuroinflammation, and tissue.

A comparative summary of these intricate mechanisms, detailing the functions and targets of key ncRNAs in preclinical neuroinflammation models, is provided in Table 4. Notwithstanding these promising mechanistic insights, the therapeutic promise of exosome-based ncRNA delivery hinges on overcoming substantial technical challenges in production, storage, and delivery. The therapeutic promise of exosome-based ncRNA delivery hinges on overcoming substantial technical challenges in production, storage, and delivery. Exosome isolation exemplifies these hurdles, as current techniques struggle to balance efficiency and integrity. Conventional methods, such as ultracentrifugation (UC) and ultrafiltration (UF), yield low recovery rates—typically 5–25% for UC—while risking membrane damage and vesicle deformation. Emerging alternatives, including microfluidic chips, offer rapid separation in under five minutes but demand sophisticated instrumentation, limiting scalability. Immunoaffinity purification achieves exceptional purity (>90%) yet relies on costly, specific antibodies (182). To address these limitations, multi-technique integration strategies, such as sequential centrifugation combined with UF and UC, have become essential for producing high-purity, structurally intact exosomes (183). Success in these approaches requires careful optimization of six critical parameters: recovery rate, sample compatibility, molecular integrity, separation efficiency, cost-effectiveness, and operational simplicity (184). However, unlike synthetic drugs, exosome production is fraught with industrialization risks stemming from biological variability, which complicates batch-to-batch consistency. Each phase—from cell culture to harvest—profoundly influences yield and quality, necessitating stringent oversight to meet clinical standards (185).

Storage poses another formidable challenge, as exosomes are highly susceptible to degradation. Short-term storage at 4 °C suffices for less than 48 hours, but long-term preservation demands deep-freezing at -80 °C. Even under these conditions, exosomes suffer biofunctional decay, reduced concentration, vesicle fusion, and structural fragmentation. Prolonged storage or repeated freeze-thaw cycles exacerbate these issues, triggering aggregation that increases particle size and diminishes biological activity. Such degradation not only compromises therapeutic potential but also heightens immunogenicity risks through membrane rupture and cargo leakage (186). To mitigate these effects, protective strategies are indispensable. Storing exosomes at -80 °C in biocompatible cryoprotectants, such as 5% sucrose solution, outperforms PBS by better preserving size distribution, concentration, and membrane protein integrity. Complementing this, programmed gradient cooling at 1 °C/min suppresses ice crystal formation, reducing vesicle breakage and enhancing stability during extended storage (187).

Even with optimized production and storage, delivering ncRNA therapeutics via exosomes remains a complex endeavor. Nucleic acids face inherent vulnerabilities—nuclease degradation, poor stability, and limited transmembrane transport due to their negative charge—which impede efficient targeting and therapeutic impact (188). Achieving precise tissue- or cell-specific delivery is vital to limit off-target effects and systemic toxicity, presenting substantial challenges for delivery system design (189). Platforms like PEGylated nanoparticles improve circulation time and exploit the EPR effect (190), while focused ultrasound (FUS) enhances BBB penetration (191, 192). However, challenges such as immunogenicity persist.

Exosomes, as endogenous extracellular vesicles, offer a compelling solution, leveraging their natural nucleic acid-loading capacity, low immunogenicity, and tissue tropism. In preclinical studies of IS, intravenous injection predominates as the delivery route. Yet, natural exosomes encounter rapid clearance by the MPS, driven by plasma-derived opsonins binding to their hydrophobic surface domains. Consequently, over 90% of injected exosomes accumulate in peripheral organs, primarily the liver and spleen, leaving less than 1% to reach the ischemic brain parenchyma (193, 194). This nonspecific biodistribution undermines neuroprotective, anti-inflammatory, and pro-angiogenic effects while raising concerns about dose-dependent toxicity. Attempts to elevate brain concentrations through higher doses result in off-target accumulation in hepatic macrophages and splenic antigen-presenting cells, potentially triggering hepatotoxicity and systemic inflammation.

To overcome these delivery barriers, innovative engineering strategies have emerged. Conjugating “don’t-eat-me” signals, such as CD47 or CD55, reduces phagocytic clearance (195), while functionalizing exosomes with targeting ligands like RGD peptides enhances site-specific accumulation (196). Exploring alternative administration routes, such as intranasal delivery via the olfactory pathway, improves BBB penetration and brain enrichment. Advances in drug loading, exemplified by magnetic extrusion-derived exosome-mimetics using the ammonium sulfate gradient method, integrate enhanced payload capacity with precise targeting. Additionally, selecting specific cellular sources, such as neuron-derived exosomes bearing neural adhesion molecules, capitalizes on inherent BBB-crossing capabilities (197).These modifications collectively aim to refine delivery precision and efficacy.

Despite such progress, translating engineered exosomes into clinical practice remains hindered by persistent obstacles. Production at GMP scale incurs prohibitively high costs, compounded by suboptimal drug encapsulation efficiency and storage-related degradation. Complex pharmacokinetics—characterized by significant hepatic and splenic sequestration, low brain targeting, and a short half-life—further complicate therapeutic development. Addressing these challenges demands a comprehensive approach, where full-process GMP management plays a pivotal role. By enforcing rigorous raw material screening, terminal product quality control, strict aseptic operations, batch-to-batch consistency monitoring, and stability validation, this strategy ensures quality assurance and bridges critical logistical gaps. Ultimately, the clinical advancement of exosome-based ncRNA therapeutics hinges on systematically resolving these technical bottlenecks through integrated technological innovation and robust quality control.

Biological heterogeneity, encompassing non-modifiable factors such as age, sex, and genetics, profoundly shapes the pathophysiology and progression of IS. These factors critically influence disease susceptibility, clinical presentation, and therapeutic outcomes (198, 199). Epidemiological evidence demonstrates that stroke incidence doubles every decade after age 55 (200). Sex-specific disparities are evident, with women facing a higher lifetime risk and mortality burden from stroke. Women account for 57.1% of global stroke deaths, and stroke ranks as the third leading cause of death in women over 55, compared to fifth in men (201–203). This disparity stems from female-specific risk factors, such as premature menopause and pregnancy complications, as well as amplified effects of shared comorbidities like hypertension and diabetes.

Preclinical studies reveal intricate interactions between age and sex in stroke outcomes. For instance, young female rodents show enhanced responses to neuroprotective treatments, such as reduced infarct volume with miR-15a/16–1 modulation. However, these sex differences diminish with age due to accumulated repair deficits (130). Interestingly, middle-aged female rodents exhibit greater cognitive vulnerability post-stroke but achieve optimal functional recovery with intervention (204)​​. These findings underscore the role of estrogen-mediated neuroprotection and the age-dependent decline in neural plasticity.

Comorbidities, particularly hypertension and diabetes, further complicate the impact of biological heterogeneity on stroke. Hypertension escalates stroke risk in a dose-dependent manner (205), while diabetes independently doubles the risk and hastens stroke onset, especially in younger individuals (205, 206). Importantly, these comorbidities interact with aging to alter therapeutic efficacy. For example, the effectiveness of FosDT siRNA diminishes in aged hypertensive models and necessitates dose adjustments in diabetic contexts. This is attributed to accelerated aging of the neurovascular unit via the NF-κB-FosDT-REST inflammatory axis (207).

Diabetes significantly worsens stroke-induced neurovascular damage through mechanisms such as metabolic dysregulation, BBB disruption, white matter injury, and heightened inflammation. These factors not only impair recovery but also reduce the efficacy of standard neurorestorative therapies, such as exosomes derived from non-diabetic mesenchymal stem cells (208). However, exosomes from diabetic rat mesenchymal stem cells (T2DM-MSC-Exo) have shown unique therapeutic benefits in diabetic stroke models. They restore BBB integrity, reduce hemorrhage, attenuate neuroinflammation, and promote white matter repair through pathways involving miR-9/ABCA1 and IGF1R (209). Additionally, diabetes exacerbates post-stroke cardiac dysfunction via impaired miR-126 signaling, a deficit that exosome therapy can address by targeting multi-organ inflammatory cascades (28). These findings highlight how metabolic dysregulation, microvascular damage, and chronic inflammation compromise recovery in patients with comorbidities (208).

A major barrier to translating stroke therapeutics from bench to bedside is the reliance on preclinical models that overlook biological heterogeneity. While most studies employ young, healthy rodents, stroke primarily affects elderly individuals with multiple comorbidities (210). This mismatch is critical, as systematic reviews indicate that less than 10% of preclinical studies test interventions in comorbid models, where therapeutic efficacy is often significantly reduced (210).

Bridging this translational gap requires the adoption of more clinically relevant preclinical models. This entails using aged animals, ensuring balanced sex representation, and incorporating validated comorbidities such as hypertension and diabetes (210–212). Moreover, given species-specific physiological differences, advanced models like non-human primates (NHPs) are crucial. For instance, pharmacokinetic studies in pig-tailed macaques show that human exosomes have a circulation half-life of approximately 40 minutes—four times longer than in rodents—and display unique immune interactions, such as rapid binding to B lymphocytes (213). These insights highlight the shortcomings of rodent models for assessing biodistribution and underscore the value of NHPs in evaluating exosome pharmacodynamics to facilitate clinical translation.

To further address the translational challenges, Table 5 provides a systematic comparison of dysregulation patterns and functional consistency of key exosome/ncRNA regulators across preclinical and human contexts. This analysis identifies both promising concordant targets, such as circ_0000831 and miR-193a-5p, and critical discrepancies, like those involving lncRNA GAS5, which necessitate validation in human models.

The choice of exosome source is pivotal in optimizing ncRNA delivery for treating cerebral ischemia, as different cellular origins confer distinct therapeutic advantages and limitations. Both stem cell-derived and differentiated cell-derived exosomes have demonstrated potential in promoting neural plasticity and functional recovery in stroke models, largely through their cargo of ncRNAs and bioactive lipids, which mitigate inflammation and secondary neuronal damage (214).

For instance, mesenchymal stem cell (MSC)-derived and M2 microglia-derived exosomes both deliver miR-124 but exert their effects through divergent mechanisms. MSC-derived exosomes, when engineered with rabies virus glycoprotein (RVG) for enhanced targeting, penetrate the BBB more effectively and accumulate in ischemic lesions. This promotes neurogenesis and angiogenesis while indirectly ameliorating the inflammatory microenvironment (215). In contrast, M2 microglia-derived exosomes directly modulate neuroinflammation by reducing pro-inflammatory cytokines (IL-1β, TNF-α), elevating anti-inflammatory factors (IL-10, TGF-β), and inducing microglial polarization toward the M2 phenotype. However, their therapeutic impact requires repeated administration, indicating limited persistence in vivo (216).

Similarly, brain endothelial cell-derived exosomes (EC-Exo) and adipose-derived stem cell exosomes (ADSC-Exo) exhibit distinct properties in delivering miR-126. EC-Exo, naturally enriched with miR-126, efficiently crosses the damaged BBB and accumulates in endothelial cells and neurons within the ischemic border zone. By enhancing vascular integrity and promoting M2 macrophage polarization, EC-Exo reverses hyperglycemia-induced miR-126 deficiency and neurovascular damage in diabetic stroke models (28). Yet, the challenge of isolating brain endothelial cells constrains large-scale production. Conversely, ADSC-Exo requires genetic modification to boost miR-126 loading and, despite weaker BBB penetration, effectively suppresses microglial activation and pro-inflammatory cytokine secretion, indirectly fostering neurogenesis and angiogenesis (27). The ease of sourcing ADSC-Exo from adipose tissue supports clinical scalability, though repeated administration is necessary to sustain efficacy, suggesting rapid clearance.

In summary, stem cell-derived exosomes, such as those from MSCs and ADSCs, offer versatility through their amenability to engineering and their capacity for multi-dimensional repair—synergistically addressing inflammation, BBB protection, and neural/vascular regeneration. This makes them well-suited for complex, multi-target conditions like stroke. Differentiated cell-derived exosomes, such as those from M2 microglia and brain endothelial cells, provide precision by directly regulating specific pathways (e.g., microglial polarization or vascular reconstruction). However, their clinical translation is hindered by production challenges, source scarcity, and pathology-dependent risks. The optimal exosome source must therefore be selected based on a careful evaluation of targeting needs, therapeutic mechanism breadth, production feasibility, and administration logistics.

While providing a systematic synthesis, this review has the following limitations: (i) It focuses primarily on neuroinflammation, not encompassing all stroke pathologies; (ii) Its conclusions rely solely on English-language studies from PubMed, potentially omitting relevant unpublished data or non-English findings, and it did not perform a meta-analysis, limiting statistical power and formal assessment of heterogeneity and publication bias; (iii) The field’s rapid evolution means new discoveries may supersede included evidence; (iv) Interpretation is constrained by inconsistent reporting of statistical parameters (e.g., power analysis, multiple testing corrections, sample size justification) across studies, variability in study quality, potential publication bias (underrepresentation of negative/null results), and low external/internal validity across diverse preclinical models; (v) Clinical biomarker data, though promising, require extensive validation.