Nanoparticle-Mediated Nose-to-Brain Delivery for Ischemic Stroke Therapy: Preclinical Insights

The immediate aftermath of an ischemic stroke is characterized by the rapid death of brain cells through two primary mechanisms: oxidative stress and excitotoxicity.

Following the acute phase, neuroinflammation contributes to further damage and cell death. Cell death in the ischemic core releases pro-inflammatory damage-associated molecular patterns (DAMPs), initiating a sterile inflammation response [6,27]. This activates resident immune cells like microglia and recruits peripheral immune cells (neutrophils, monocytes) to the brain. Microglia are the first to respond in the central nervous system, phagocytizing of cellular debris and releasing cytokines and chemokines [2]. Neutrophils are among the first peripheral cells to enter the brain, contributing to blood–brain barrier (BBB) damage and tissue injury by releasing neutrophil extracellular traps (NETs) and inflammatory mediators. Monocyte-derived macrophages (MDMs) also infiltrate the brain early on and differentiate into pro-inflammatory types [6]. T and B lymphocytes enter the brain within days to weeks. Pro-inflammatory T-cell subtypes worsen stroke outcomes, and natural killer (NK) cells also contribute to the damage [6,27]. Activation of the complement system further fuels inflammation and tissue destruction [6]. In the subacute phase, reactive astrocytes form a glial scar around the lesion [28,29,30]. Late-stage microglia can shift to a pro-inflammatory phenotype, inducing an inflammatory response in astrocytes, which then secrete neurotoxic substances that hinder recovery. Chronic inflammation is neurotoxic and contributes to ongoing damage.

Despite the initial damage, the brain possesses remarkable capacity for repair and regeneration, a process that is also heavily influenced by neuroinflammation and reactive gliosis. During the later stage of ischemic injury, microglia can shift to an anti-inflammatory phenotype, facilitating inflammation resolution by clearing dead cells and secreting nourishing factors such as IGF-1 [6,31]. Regulatory T cells (Tregs) are beneficial T-cell subtypes that promote neurogenesis and white matter repair through factors like amphiregulin and osteopontin [32,33]. B lymphocytes can also play a protective role by producing anti-inflammatory cytokines like IL-10 [6,34].

Moving beyond the inflammatory response, the glial scar, formed mainly by reactive astrocytes and other neuroglial cells, initially acts as a protective barrier, isolating the damaged tissue and compartmentalizing the immune response [28,35]. However, it can also inhibit neural regeneration due to the release of growth-inhibiting factors like chondroitin sulfate proteoglycans [36,37]. On the other hand, matrix metalloproteinases (MMPs) in the recovery phase can mediate angiogenesis, neurogenesis, and synaptic remodeling [36]. Reactive astrocytes can also acquire neural stem cell-like characteristics and can be reprogrammed into neurons [38,39].

A critical repair process is angiogenesis, which helps restore blood flow and partially repair the blood–brain barrier (BBB) [40]. Vascular endothelial growth factor (VEGF) is a key promoter of angiogenesis, but in the early phase, it can also increase BBB permeability and the risk of bleeding. Pericytes, cells that maintain BBB integrity, can differentiate into neural and vascular cells and play a crucial role in re-establishing tight junctions to restore the BBB [41]. Remyelination of axons is another crucial recovery process in degenerated white matter. Tregs and astrocytes can promote the formation of new oligodendrocytes via osteopontin and Brain-Derived Neurotrophic Factor (BDNF), aiding in white matter repair and functional recovery [32,36,42].

Finally, neuroplasticity is the brain’s ability to rewire and reorganize neural circuits in response to injury. This process is most active in the first few months after a stroke [40]. It involves the sprouting of new axons, the formation of new synapses, and the remodeling of existing connections. Axon growth inhibitory factors are downregulated, and pro-growth factors like growth differentiation factor 10 (GDF10) are upregulated [43]. In addition, synaptic and dendritic spines are initially lost but regenerate during the recovery phase in the peri-infarct area [44]. The overall recovery of function involves the expansion of dendritic branches, the formation of new connections, and the sprouting of axons from the undamaged side of the brain to support the damaged area.

Focal ischemia models are the most clinically relevant and widely utilized category in stroke research, as they mimic the occlusion of a specific cerebral artery, which accounts for the majority of human strokes. Among these, the Middle Cerebral Artery Occlusion (MCAO) model stands as the “gold standard,” a status reflected by its predominant use in the preclinical studies summarized throughout this review. The MCAO model’s strength lies in its ability to generate a large, reproducible infarct affecting both the cortex and the underlying striatum, closely recapitulating the vascular territory of the human middle cerebral artery [8,46,47]. The most common method for inducing MCAO is the intraluminal filament technique [48]. This procedure involves the insertion of a silicone-coated monofilament through the external carotid artery. The filament is then advanced into the internal carotid artery to physically block the origin of the MCA, thereby interrupting blood flow to a significant portion of the cerebral hemisphere. This interruption of blood flow leads to the formation of lesions in the territory supplied by the middle cerebral artery. This territory includes deep subcortical structures (like the basal ganglia, internal capsule, and thalamus), and the overlying frontotemporal and parietal cortices [8,47].

The MCAO model is further refined into distinct variants that correspond to different clinical scenarios. The transient MCAO (tMCAO) model involves withdrawing the filament after a defined period of occlusion, typically ranging from 60 to 90 min, to allow for reperfusion [49,50]. This variant is of paramount importance because it simulates the clinical sequence of successful recanalization therapy, such as intravenous thrombolysis with tissue plasminogen activator (tPA) or mechanical thrombectomy. As discussed in the previous section, the restoration of blood flow, while essential, paradoxically triggers a secondary wave of injury driven by a massive surge in reactive oxygen species (ROS), inflammation, and blood–brain barrier disruption [22]. Consequently, the tMCAO model is an essential experimental paradigm for evaluating therapies designed to mitigate this ischemia–reperfusion injury. In contrast, the permanent MCAO (pMCAO) model, where the filament is left in place, simulates clinical situations where reperfusion is not achieved. This results in a more severe and consistently larger infarct, providing a platform to study the maximal extent of ischemic damage and the brain’s endogenous responses without the confounding variable of reperfusion injury. A third variant, the distal MCAO (dMCAO), involves a craniotomy to directly ligate or cauterize a distal branch of the MCA. dMCAO offers a more invasive but highly precise method for producing a purely cortical infarct [50,51].

An alternative and increasingly utilized method for inducing focal ischemia is the photothrombotic (PT) model. This technique operates on a different principle than mechanical vessel occlusion. It involves systemic administration of a photosensitive dye called Rose Bengal. Subsequent illumination of a targeted cortical region through the intact skull or a thinned-cranial window activates the dye [52,53]. This activation generates singlet oxygen from the dye, which then induces endothelial cell damage, platelet aggregation, thrombus formation, and the occlusion of microvessels within the illuminated area. The resulting infarct is highly circumscribed, reproducible in size and location, and largely confined to the cortex, with minimal primary damage to the underlying white matter and a very limited penumbra [47,52]. While the MCAO model is superior for assessing acute neuroprotective agents aimed at salvaging the penumbra, the PT model’s precision allows researchers to meticulously track long-term structural and functional changes in the peri-infarct cortex. The predictability of the lesion size and location facilitates studies on axonal sprouting, synaptic remodeling, and the reorganization of cortical maps over weeks to months [47,54]. Therefore, the selection of the PT model often signifies a shift in research focus from acute neuroprotection to subacute and chronic neuroregeneration.

Despite their indispensability, it is crucial to acknowledge the inherent limitations of these preclinical models and the challenges they pose for clinical translation. The failure of many promising neuroprotective agents in clinical trials is partly attributed to the “translational chasm” between animal models and the human clinical condition. Preclinical stroke research is predominantly conducted in young, healthy, male rodents under highly controlled experimental conditions [8]. In contrast, the typical human stroke patient is elderly and presents with comorbidities, such as hypertension, diabetes, atherosclerosis, and hyperlipidemia, all of which profoundly alter the brain’s vulnerability to ischemia and its response to therapy [55]. For example, a study comparing spontaneously hypertensive rats (SHR) with normotensive rats (NTR) revealed marked differences in infarct patterns and severity [56]. Furthermore, the controlled nature of experimental stroke induction—with a fixed occlusion time and often immediate therapeutic intervention—does not reflect the clinical variability in stroke onset, severity, and time to treatment [8]. In addition, there are structural differences in the brain. The rodent brain contains less than 10% white matter, while the brains of humans, non-human primates (NHPs), and pigs contain over 60%. This disparity makes it challenging to replicate subcortical white matter strokes in rodents, which constitute approximately 25% of all human strokes [8]. Nevertheless, the strategic selection of models allows researchers to ask specific and relevant questions. The MCAO model remains the most appropriate choice for assessing acute neuroprotectants and therapies targeting the multifaceted injury cascades associated with reperfusion. The PT model, in turn, provides an unparalleled system. It is ideal for dissecting the molecular and cellular mechanisms of long-term cortical reorganization and repair. Despite their limitations, these rodent models provide the essential and standardized platforms upon which the proof-of-concept and efficacy of the novel intranasal therapeutics discussed in the subsequent sections are built and rigorously evaluated.

Intravenous (IV) administration is the most common and minimally invasive route for systemic drug delivery. It allows for widespread distribution throughout the body, but this becomes a major drawback for CNS disorders. The BBB severely restricts the passage of most therapeutics, especially large biologics like proteins, nucleic acids, and cell-based therapies, resulting in suboptimal drug concentrations in the brain [7,57]. This low bioavailability necessitates higher systemic doses, which in turn increases the risk of off-target side effects [58]. For instance, systemically delivered mesenchymal stem cells (MSCs) often suffer from poor homing to the brain and reduced viability [57].

Intra-arterial (IA) administration offers a more targeted approach by delivering drugs directly into the cerebral circulation, bypassing the first-pass metabolism and achieving a higher local concentration in the brain compared to IV [57]. Preclinical studies have shown that IA delivery of stem cells can lead to superior recovery in terms of infarct volume reduction compared to other routes [59]. However, this method is highly invasive, requiring arterial catheterization by a skilled interventionalist. It also carries significant risks, including arterial dissection, hemorrhage, and distal emboli [57]. Moreover, even with IA delivery, therapeutics must still cross the BBB at the capillary level to reach the brain parenchyma.

To circumvent the BBB entirely, direct injection into the central nervous system has been investigated. Intracerebroventricular (ICV) or intrathecal (IT) administration involves injecting therapeutics into the cerebrospinal fluid (CSF), allowing for broad distribution throughout the CNS [7,57,58]. This route has proven effective for delivering cell-based therapeutics in preclinical models [40,60]. However, it remains a highly invasive procedure requiring surgery or a lumbar puncture, posing risks of infection and procedural complications. Intraparenchymal or intracerebral (IC) injection into the brain parenchyma is the most direct method, ensuring the highest possible concentration at a specific target site [40,57]. While it completely bypasses the BBB, it is also the most invasive approach, carrying a substantial risk of tissue damage, inflammation, and hemorrhage at the injection site [57]. Furthermore, drug distribution is severely limited by slow diffusion from the point of injection, making it unsuitable for treating the widespread and diffuse damage characteristic of an ischemic stroke [61].

The nasal cavity anatomy crucial for drug delivery is divided into three main regions: the vestibular region, the respiratory region, and the olfactory region. The vestibular region is situated at the nasal entrance and is covered by non-keratinized stratified squamous epithelium. It contains nasal hairs, primarily filtering inhaled particles [12,62]. However, its small surface area suggests it contributes minimally to drug absorption [11]. The respiratory region is the most expansive area within the nasal cavity, constituting up to 90% of the human nasal cavity. It is lined by ciliated pseudostratified columnar epithelium with microvilli [11,12]. This region is rich in blood vessels, contains mucus-secreting goblet cells, and is innervated by the facial nerve and trigeminal nerve [62,63]. The olfactory region is located in the superior aspect of the nasal cavity. It covers the underside of the cribriform plate of the ethmoidal bone and is covered by pseudostratified columnar epithelium. In humans, this region accounts for about 10% of the total nasal surface area, while 50% in rodents [12,64]. The olfactory epithelium consists of Olfactory Sensory Neurons (OSNs), basal cells (acting as progenitor cells), and supporting cells [12,64].

Key structures within the nasal cavity that facilitate drug transfer to the CNS include the olfactory nerve, the trigeminal nerve, and the vascular and lymphatic systems. The OSN or olfactory nerve has axons that project to the olfactory bulb. These axons are encapsulated by layers of olfactory ensheathing cells (OECs) and olfactory neural fibroblasts (ONFs), which together form a perineural sheath [9]. This perineural space is physically continuous with the meninges of the brain. The trigeminal nerve is responsible for general sensations like pain and temperature and is distributed across both the respiratory and olfactory mucosa [11,13,64]. Its branches, specifically the V1 ophthalmic nerve and V2 maxillary nerve, are distributed in the nasal epithelium, with nerve fibers connecting directly from the nasal mucosa to the brainstem (pons). The lamina propria beneath the nasal mucosa is highly vascularized, promoting systemic absorption, and also contains lymphatic vessels. The olfactory/nasal lymphatic system provides a pathway for cerebrospinal fluid (CSF) to drain to the deep cervical lymph nodes, which can be exploited for drug delivery [9,62,65,66,67]. It has been reported that dural lymphatic vessels near various cranial nerve outlets at the skull base connect to deep cervical lymph nodes through nasopharyngeal lymphatics or via nasal and hard palate lymphatics [65,66]. Notably, the nasal mucosa is known to be composed of structures rich in lymphatic vessels and venous sinusoids [67].

Drugs administered intranasally access the CNS through two principal pathways, both allowing them to bypass the BBB.

Intranasal (IN) drug delivery offers several advantages over conventional administration routes such as oral or intravenous injection. (1) Because IN delivery is non-invasive, patient compliance is improved. (2) This route provides a direct pathway that bypasses the blood–brain barrier (BBB) while also avoiding hepatic first-pass metabolism, thereby enhancing drug bioavailability in the brain and reducing the required dosage [73,78]. (3) Systemic side effects can be minimized; for instance, intranasal administration of deferoxamine (DFO) after stroke resulted in significantly higher concentrations in brain tissue while maintaining lower levels in blood and peripheral organs compared to intravenous delivery [79]. Consistently, recent studies employing circular RNA demonstrated greater accumulation in the peri-infarct region and reduced nonspecific distribution in other organs when administered intranasally rather than intravenously [80]. Furthermore, intranasal delivery of anti-Nogo-A antibodies achieved comparable concentrations in CNS tissue to those observed after intrathecal injection, highlighting its ability to reach therapeutic levels in the brain without invasive procedures [81]. (4) Finally, IN delivery provides rapid access to the CNS, with drugs detected in brain tissue within 5 min of administration and reaching more distant brain regions within 30 min [11,73], collectively underscoring its promise as an efficient and clinically valuable route for CNS drug delivery.

While nose-to-brain drug delivery presents distinct advantages over conventional administration routes, significant translation hurdles impede its clinical implication. The most significant and systemically overlooked anatomical discrepancy lies in the olfactory epithelium, the primary portal for the direct neuronal nose-to-brain (N2B) pathway. In rodents (rats and mice), this specialized region is expansive, constituting as much as 50% of the total nasal surface area [64]. In contrast, the human olfactory region is a small, remote patch located in the superior aspect of the nasal cavity, accounting for only approximately 10% of the total nasal surface area. This 5-fold difference in relative surface area implies a fundamental, qualitative difference in the dominant transport mechanism between the species. In a rodent model, a drug administered into the nasal cavity has a high probability of interacting with the massive olfactory epithelium, leading to efficient direct brain transport via the olfactory nerve pathway. In a human, however, 90% of the nasal cavity is lined with respiratory epithelium. Consequently, for any therapeutic administered via a standard nasal spray, the dominant outcome is not direct brain delivery. Instead, the drug is either rapidly cleared by mucociliary action or, more likely, absorbed into the systemic circulation via the highly vascularized respiratory mucosa, nullifying the primary BBB-bypassing advantage [12]. Therefore, what is observed as a high-efficiency, primary delivery route in rodent models is, in humans, very likely a low-efficiency, auxiliary pathway. Preclinical studies are thus systemically biased to overestimate the potential for direct brain transport.

In addition, significant differences exist in the structure of nasopharyngeal immune tissues. Rodents possess well-characterized, organized nasopharynx-associated lymphoid tissue (NALT), whereas humans have a more diffuse Waldeyer’s ring (comprising the lingual, pharyngeal, palatine, and tubal tonsils) [82]. This difference in immune architecture has profound implications for the translatability of N2B biologics, proteins, and cell therapies. These treatments may elicit vastly different immune responses and trafficking patterns in human versus rodents. Furthermore, there are tremendous differences in the lengths, volumes, and surface areas between the nasal cavities of adult humans and rodents. In addition, species-specific variations in mucosal permeability and metabolic processes (e.g., enzymatic degradation) complicate direct translation [83].

Beyond small molecules and protein-based therapies, recent advances in nucleic acid medicine have enabled specific modulation of the complex gene networks involved in stroke. For instance, gene expressions can be modulated using tools like siRNA and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR). A study by Boutej et al. demonstrated that intranasal delivery of an siRNA targeting the RNA-binding protein SRSF3, a key regulator of the innate immune response in the brain [145], efficiently reduced SRSF3 protein levels in the brain. This reprogramming of innate immune response led to a significant reduction in the infarct volume. This was achieved by restoring the synthesis of critical immune proteins like Leukocyte Immunoglobulin-Like Receptor subfamily B member 4a (LILRB4a), Heme Oxygenase 1 (HMOX1), and Tissue Inhibitor of Metalloproteinases 1 (TIMP1). Similarly, Ryu et al. developed a protein-based CRISPR/dCas9 system to upregulate the expression of the neuroprotective gene sirtuin1 (Sirt1), a member of NAD+-dependent protein deacetylase [107,146]. Sirt1 knockout mice subjected to stroke displayed larger infarct volume [146]. Ryu et al. successfully enhanced Sirt1 expression through targeted intranasal delivery of dCas9-VP64 protein and its corresponding single-guide RNA (sgRNA) to the brain via the MCT1 transporter. This, in turn, mitigated secondary brain damage, reduced brain edema, and improved survival rates in mice [107].

In parallel, the direct delivery of therapeutic nucleic acids offers a distinct approach to compensate for the genetic dysregulation following brain injury. Circular RNAs (circRNAs), which are highly stable endogenous RNAs characterized by back-splicing and a closed-loop structure, and abundant in the CNS, have been studied for their restorative function in ischemic stroke. Among them, circular RNA form of Scm Polycomb Group Protein Homolog 1 (SCMH1) gene (circSCMH1), whose expression levels are decreased in the plasma of patients with acute ischemic stroke [147], was extensively studied in preclinical settings (Figure 3). Yang et al. (2020) reported that in rodents and nonhuman primates, intravenous (IV) delivery of circSCMH1 in extracellular vesicles (EV) promoted neuroplasticity and suppressed glial activation [147]. A following study by Li et al. demonstrated that IV delivery of circSCMH1 also enhanced vascular repair [148]. Building on these findings, intranasal delivery of circSCMH1 effectively bypasses the blood–brain barrier (BBB). The circRNA accumulated well within the peri-infarct region while minimizing off-target accumulation (Figure 3) [80]. The treatment significantly improved sensorimotor and cognitive function and reversed key pathological gene expression profiles.

Mesenchymal stem cell (MSC) therapy has emerged as a promising approach for treating ischemic stroke, with a shared focus on the cells’ powerful paracrine effects rather than direct cellular replacement [40]. To overcome limitations of conventional systemic delivery, such as unstable in vivo homing and reduced cell viability, researchers have increasingly explored intranasal delivery [149]. This approach has moved beyond simple cell administration, with studies attempting various methods to augment therapeutic efficacy and delivery efficiency to the target site.

The preclinical studies summarized above demonstrate the potential of various intranasally delivered agents across different phases of ischemic stroke. Small molecules targeting acute mechanisms like oxidative stress (e.g., Edaravone-IL) or excitotoxicity, along with early-acting anti-inflammatory agents (e.g., NaB, possibly (+)-Naloxone depending on timing), are primarily relevant for acute neuroprotection. These small molecules are ideally administered within hours of stroke onset. This timeframe aligns with the existing window for reperfusion therapies, suggesting potential combination strategies. However, this requires rapid diagnosis and administration capabilities in clinical settings. Repurposed drugs like naloxone may have an accelerated path to clinical trials due to existing safety data.

In contrast, agents promoting neurogenesis, angiogenesis, white matter repair, and neuroplasticity, such as growth factors (BDNF, IGF-1, Wnt3a), certain cytokines (IL-4, IL-13), complement peptides (C3a), repurposed tPA, anti-Nogo-A antibodies, and nucleic acid therapies (circSCMH1), are generally administered in the subacute to chronic phases (days to weeks post-stroke) in preclinical models. These target the brain’s endogenous repair mechanisms and aim to improve long-term functional recovery. Their delayed administration window is clinically more feasible for many patients who miss the hyperacute treatment window.

However, translating these findings faces significant hurdles. Biologics (proteins, antibodies) often have poor stability, complex manufacturing processes, and potential immunogenicity, although intranasal delivery aims to minimize systemic exposure. Nucleic acid therapies face similar challenges regarding stability, delivery efficiency into target cells, and potential off-target effects or immunogenicity [172]. Cell therapies (MSCs) present unique challenges including manufacturing scale-up under GMP conditions, ensuring batch consistency, cell viability post-administration, potential immunogenicity (for allogeneic cells), and long-term safety (e.g., tumorigenicity, although low for MSCs) [82]. While the PASSIoN trial demonstrated feasibility in neonates, translating this to the adult population requires addressing anatomical differences and potentially different dosing or delivery strategies. The success of these subacute/chronic therapies hinges not only on demonstrating efficacy in robust preclinical models (including those with comorbidities) but also on developing scalable manufacturing processes and establishing long-term safety profiles.

Hyaluronidase, an enzyme of approximately 55–61 kDa with a diameter of ~3 nm, has been used as a pretreatment in ischemic stroke models to loosen the extracellular matrix by degrading hyaluronic acid (HA). By reducing structural barriers, hyaluronidase enables more efficient delivery of the drugs and MSCs to target regions of the brain [173]. Intranasal administration of hyaluronidase can increase nasal mucosal permeability and facilitate the entry of therapeutic agents, such as stem cells or nanoparticles, into the CNS while bypassing the BBB [174]. However, dosing must be carefully controlled: while a dosing range of approximately 100 U per mouse is recommended, excessive amounts can cause acute inflammation and anaphylaxis.

Building on the preclinical evidence for MSC-based therapies described in Section 6, recent efforts have focused on nanotechnology and pretreatment strategies to enhance the efficiency of intranasal delivery. A report by Chau et al. showed that intranasal administration of BMSCs, when combined with hyaluronidase pretreatment 30 min prior to delivery, markedly improved therapeutic outcomes in ischemic stroke [167]. In this study, hypoxic-preconditioned BMSCs (HP-BMSCs) were employed, which exhibited enhanced survival potential and robust migration to the peri-infarct cortex. The rats showed significant increase in angiogenesis, neurogenesis and overall functional recovery. Mechanistically, expression ratios of HIF-1α/18s and Bcl-xL/18s were elevated 3–5 fold compared with controls. With a similar concept, another study demonstrated that hyaluronidase pretreatment also enhanced the intranasal delivery efficiency of BMSCs when co-administered with IGF-1 [109]. More recently, Yamamoto et al. extended this approach to multilineage-differentiating stress-enduring (Muse) cells, isolated from human bone marrow using stage-specific embryonic antigen-3 (SSEA-3) antibody [175]. Following hyaluronidase pretreatment, mitochondria-positive Muse cells demonstrated engraftment near the infarct core and showed improved sensorimotor function, whereas only a marginal number of BMSCs were detected in control animals.

Extracellular vesicles (EVs) are nano-sized, lipid bilayer-enclosed particles released by nearly all cell types into the extracellular space. They facilitate intercellular communication by transferring active biomolecules, including proteins, lipids, and nucleic acids. EVs can be derived from the MSCs and other autologous cells. After multiple homogenization, purification, and filtering steps, EVs can be prepared as particles a few hundred nanometers in diameter. EVs contain lipids and proteins of cellular origin, which may lower their immunogenicity.

However, EV-based therapies face significant translational challenges. First, large-scale production remains a major bottleneck. The number of MSCs and autologous cells is often insufficient to produce sufficient quantities of EVs for therapeutic use. While other types of allogenic cells have been used to solve this, this approach induces the possibility of immune responses. Furthermore, the dose-dependent safety of EV-based therapeutics must also be studied, given their cargo of endogenous proteins or loaded drugs.

Second, the instability of EVs is another critical factor to consider for intranasal therapeutics. When administered intranasally, intact EVs are vulnerable to physical clearance (mucociliary clearance and drainage), enzymatic degradation (by proteases and nucleases), and the impermeability of nasal mucosa. These factors can lead to the rapid disintegration of the vesicles into their individual components [176].

As discussed in Section 6, BDNF has emerged as a potential neurotrophic factor for ischemic stroke but faces major translational hurdles due to its large molecular size and instability in systemic circulation. Recent nanotechnology-based strategies have sought to address these issues by developing optimized carriers for intranasal delivery. One approach involves packaging BDNF into small EVs derived from genetically engineered MSCs [177] (Figure 4). Using lentivirus packaging, BDNF was overexpressed on iPSC-derived MSCs, and small EVs (BDNF-sEVs) were purified by anion chromatography. In a transient middle cerebral artery occlusion (tMCAO) model, BDNF-sEVs (50–150 nm) were delivered via the intranasal route following hyaluronidase pretreatment. BDNF-sEVS promoted sustained activation of BDNF/TrkB signaling, and ultimately improved functional recovery. Nonetheless, the mass production of BDNF-sEVs remains a significant challenge for future clinical applications.

Similarly, BMSC-derived EVs (BMSC-EVs) were tested in a tMCAO mouse model without hyaluronidase pretreatment [165]. Intranasal delivery of 150 nm BMSC-EVs resulted in modest functional recovery. These results highlight that both cellular origin of EVs and the delivery route critically influence therapeutic efficacy in ischemic stroke. In another study, ProtheraCytes, human CD34+ stem cells, were treated in an MCAO mouse model via three different routes: intracerebral (IC), intranasal (IN), and intra-arterial (IA) administration [178]. IC and IA routes produced superior recovery compared with IN in terms of infarct volume and sensorimotor functions. However, neurogenesis and angiogenesis markers were expressed at comparable levels across all delivery routes.

Liposomes are artificial vesicles composed of synthetic lipid bilayers, while extracellular vesicles (EVs) are naturally secreted by cells. This fundamental difference leads to key distinctions between liposomes and EVs. Liposomes are easier to scale up and allow higher drug-loading capacity. However, EVs are more biocompatible, naturally target tissues, and exhibit lower immunogenicity. Due to these complementary strengths, researchers are developing hybrid vesicles that combine the advantages of both systems [179]. In most cases, liposomes demonstrate longer circulation times than EVs because of their structural stability.

Basic fibroblast growth factor (bFGF) has been demonstrated as neuroprotective proteins in ischemic stroke mouse models when delivered via intracerebroventricular infusion. However, as a 16.5 kDa protein, bFGF does not cross the BBB efficiently, making systemic administration (intravenous, intraperitoneal, or subcutaneous) largely ineffective [180]. To improve delivery, the authors synthesized 100~130 nm bFGF-loaded liposomes (bFGF-NL) and injected them via the intranasal route. The bFB-NL treatment significantly improved sensorimotor function and reduced the infarct volumes in ischemic stroke mouse models [180].

Circular RNAs (circRNAs), already shown in preclinical models to promote neuroplasticity and vascular repair (see Section 6 and Figure 3), has been further optimized through encapsulation in lipid nanoparticles (LNPs) [80]. By incorporating PEGylated lipids, ionizable lipids, DSPC, and cholesterol, researchers generated a circSCMH1@LNP formulation suitable for intranasal administration [80]. When combined with hyaluronidase pretreatment, this system achieved efficient accumulation in the peri-infarct region while minimizing peripheral distribution. Beyond enhancing delivery efficiency, the liposomal strategy facilitated multiple reparative processes—including angiogenesis, synaptic remodeling, BBB stabilization, and myelin regeneration. These processes ultimately resulted in significant functional recovery in ischemic stroke models.

Recently, intranasal delivery of interleukin-4 (IL-4)-loaded lecithin-based liposomes was shown to enhance white matter integrity and attenuate long-term sensorimotor and cognitive deficits in MCAO models [97]. IL-4 promoted the structural and functional robustness of myelinated fibers, stimulated the differentiation of oligodendrocyte progenitor cells (OPCs) into myelinating oligodendrocytes and markedly improved long-term neurological outcomes.

The long-term circulation of the macrophage-mimetic liposomes can be achieved through the camouflage effect of the lipid components derived from the endogenous macrophages. A recent study tested liposomes co-loaded with panax notoginseng saponins (PNS) and ginsenoside Rg3 (Rg3), incorporated into macrophage-mimetic core–shell nanocomposites (MM-Lip-Rg3/PNS) [181]. These 150~180 nm nanocomposites preferentially targeted ischemic regions following intranasal administration, where they significantly reduced infarct volume, improved neurological function, and decreased pro-inflammatory cytokines (TNF-α, IL-6, and IL-1β). However, the precise mechanisms underlying the transport of most of the liposomal systems during intranasal delivery remain unclear.

Intranasal drug delivery using inorganic nanoparticles has been demonstrated in several preclinical studies. With appropriate surface modifications and synthetic methods, gold nanoparticles (GNPs) can load antigen, drugs, and enzymes at the surface of GNPs and be delivered to the ischemic brain in mouse models. In one study, an ApoE-mimetic peptide (CS15) was covalently conjugated carboxylate thiol-PEG ligands on the surface of GNPs (CS15-GNPs) [182]. Following intranasal administration, 40 nm CS15-GNPs with a near-non-charged surface promoted recovery in the infarcted regions. Treatment with CS15-GNPs significantly suppressed neuronal apoptosis and cerebral inflammation, while no evidence of toxicity or accumulation of CS15-GNPs was observed.

Platelet-mimetic membrane-coated Mn/Co₃O₄ nanoparticles were also delivered intranasally [183] (Figure 5). These 300 nm particles effectively scavenged reactive oxygen species (ROS) and induced M2 polarization of macrophages. This shift in M1/M2 balance resulted in decreased expression of pro-inflammatory cytokines (TNF-α, IL-6, and IL-1β) and increased expression of the anti-inflammatory cytokine IL-10.

Calcium phosphate (CaP) NPs have been widely studied as vehicles for drug and gene delivery due to their biocompatibility, biodegradability, and chemical similarity to the human bone. Researchers synthesized CaP NPs loaded with nucleotides (AMP, ADP, ATP and GTP) by mixing calcium solutions with each nucleotide (NT) solution [184]. To enhance intracellular uptake, branched polyethyleneimine (bPEI, 1.8 kDa) was coated onto the surface of nucleotide-loaded CaP NPs. Because sufficient nucleotide supply is required for recovery of ischemic stroke, intranasal delivery of NT-loaded CaP NPs reduced infarct volume and improved neurological deficits. Although the negative charge of ATP reduced its delivery efficiency, bPEI-tethered ATP loaded CaP NPs facilitated intracellular delivery.

In addition, Cas9 protein-loaded CaP NPs were intranasally delivered and shown to upregulate SIRT1 expression without acute cytotoxicity [107]. Even without hyaluronidase pretreatment, Cas9 protein successfully reached the infarcted brain region. The greatest increase in Sirt1 mRNA expression was achieved by tethering PEGs and PEIs onto the surface of Cas9-loaded CaP NPs.

Wang et al. further demonstrated theranostic applications using rare-earth element-doped near-infrared (NIR, ~1060 nm) emissive NPs. These elements allowed in vivo imaging of superior tissue penetration depth, while also promoting recovery from ischemic injury [185] (Figure 6). Moreover, a stroke-homing peptide (SHp) was conjugated to polyethylene glycol (PEG) and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine (DSPE) at the terminus of the PEG. The amphiphilic polymer, DSPE-TK-PEG-SHp (TS), was synthesized by employing a ROS-cleavable thioketal (TK) as the intermediate chain to connect DSPE and PEG. The TS surface encapsulated 100 nm sized NaYF₄:Nd@NaLuF₄@MSN@NH₄NO₃ (NMNP-TS), which scavenged hydrogen peroxide, reduced infarct size, and suppressed pro-inflammatory cytokines (TNF-α, and IL-6). However, the precise delivery pathways of inorganic NPs during transportation from the nasal cavity to the brain remain to be fully elucidated.

Organic compounds have been developed for intranasal delivery for decades. FDA-approved polymers and surfactants have been widely used in this context. The use of organic polymers in intranasal drug delivery has been shown to enhance drug absorption and direct pharmaceutical agents to specific targets such as the brain. Commonly used organic polymers include natural polysaccharides such as chitosan and starch, as well as cellulose derivatives like hydroxypropyl methylcellulose (HPMC). Synthetic polymers, including poly(lactic-co-glycolic acid) (PLGA) and polyacrylates, have also been extensively applied due to their mucoadhesive and gel-forming properties.

Joachim et al. reported osteopontin (OPN)-loaded gelatin NPs (OPN-GNPs) could be used as a treatment for ischemic stroke [101]. Following intranasal delivery of OPN-GNPs in a middle cerebral artery occlusion (MCAO)-induced ischemic stroke, a significant reduction in infarct volume was observed. The same research group further reported the use of gelatin nanoparticles for the delivery of inducible nitric oxide synthase (iNOS) siRNA to the post-ischemic brain [186].

Reactive oxygen species (ROS)-responsive nanocarriers (SPNP) that can target mitochondria have been demonstrated to promote neurological functional recovery [187]. Puerarin (PU), an isoflavone from kudzu root, exerts anti-inflammatory, antioxidant, and lipid-lowering effects, and has shown therapeutic potential in metabolic, cardiovascular, hepatic, and neurological disorders. It is clinically approved in China for the treatment of myocardial ischemia. However, its poor solubility limits broader applications. To overcome this limitation, researchers developed a PU-loaded SPNP (chitosan-based polysaccharides NPs) formulation, which enables efficient delivery of small-molecule drugs into the brain. Intranasal delivery of PU-loaded SPNPs significantly reduced oxidative stress and pro-inflammatory cytokine levels and inhibited neuronal apoptosis.

The interaction between small-molecule drugs and cyclodextrin (CD) is characterized by a high degree of host–guest stability, making it well-suited for sustained drug release. In a recent study, Teng et al. reported a temperature-dependent sol–gel transition of drug-loaded NPs, which has significant implications for the treatment of ischemic stroke [188]. The hydrophobic pocket within CDs has been utilized as an inclusion site for borneol, a compound that has demonstrated efficacy in the treatment of inflammation- and oxidation-related diseases. In the experimental setup, poloxamer 407 and edaravone (EDA), a free radical scavenger, were co-loaded with borneol-loaded CD (BP) to form the EDA-BP sol. The solution can be injected into the nasal cavity, where it undergoes a phase transition. Injected solution transforms into a hydrogel upon contact with body temperature. The sol–gel transformation has been shown to enhance drug bioavailability. Intranasal delivery of EDA-BP significantly reduced infarct volume and improved neurological outcomes.

Similar to poloxamer 407, ionic liquids (ILs) have been applied for EDA (edaravone-IL) in ischemic mice [87]. Among these, ethoxy-ILs showed the highest cell viability and radical scavenging activity. Following intranasal administration of edaravone-IL, EDA was detected in the brain, olfactory bulb, nasal mucosa, and plasma. This formulation demonstrated exceptional therapeutic efficacy by reducing infarct volume and improving sensorimotor function.

Polymer-based drug delivery platforms have been developed for the treatment of ischemic stroke. Polylactic acid-hyperbranched polyglycerol (PLA-HPG) has been used as a surfactant to synthesize dexmedetomidine (DEX)-loaded NPs, which were further conjugated with the amines of dendritic polyamidoamine (PAMAM) to form a Schiff base (BNP-PAMAM/DEX) [88]. A Schiff base is an organic compound characterized by an azomethine group (–C=N–), a carbon-nitrogen double bond formed by the condensation of a primary amine with an aldehyde or ketone. This azomethine group can be broken into two different parts under slightly acidic conditions, releasing DEX into the olfactory bulb region. Intranasal administration of BNP-PAMAM/DEX resulted in the passive diffusion of DEX. It reduced neuronal apoptosis and decreased the expression of IL-6 and TNF-α, thereby providing neuroprotection.