Bone marrow-derived mesenchymal stem cells for ischemic stroke: mechanisms, delivery strategies, clinical evidence, and translational challenges

Abstract

Ischemic stroke remains a major global health burden because of its high incidence, disability, and mortality. Current reperfusion therapies are constrained by narrow therapeutic windows and have limited capacity to reverse established neuronal injury or promote long-term functional recovery, highlighting the need for strategies that support neural repair and neurovascular remodeling. Bone marrow-derived mesenchymal stem cells (BMSCs) have been widely investigated as candidate therapeutic products for ischemic stroke because of their paracrine, immunomodulatory, trophic, angiogenic, and microenvironment-regulating properties. Current evidence suggests that their therapeutic effects are more likely mediated by regulation of the post-ischemic microenvironment than by direct neuronal replacement. This narrative review discusses BMSC-based strategies for ischemic stroke by integrating evidence from preclinical models, early clinical studies, and translational research. We summarize major proposed mechanisms, including regulation of neuroinflammation, suppression of apoptosis and oxidative stress, blood-brain barrier protection, angiogenesis, neurotrophic support, axonal remodeling, and myelin repair. Emerging mechanisms, such as autophagy regulation, mitochondrial transfer, and extracellular vesicle-mediated effects, are also discussed according to their current level of evidence. In addition, this review evaluates delivery routes, functional enhancement strategies, BMSC-derived extracellular vesicles, combination approaches, clinical evidence, and key translational barriers. Particular attention is given to the distinction between experimentally supported mechanisms, early clinical safety and feasibility signals, and unresolved issues related to optimal dose, administration timing, route-dependent biodistribution, product heterogeneity, mechanism-linked potency assays, and Good Manufacturing Practice-compatible manufacturing. This review provides a balanced framework for understanding the therapeutic potential, current limitations, and future development requirements of BMSC-based therapies for ischemic stroke.

Graphical Abstract

Introduction

Stroke remains one of the leading causes of death and long-term disability worldwide, with ischemic stroke accounting for more than 80% of all stroke cases [1]. According to the Global Burden of Disease study, the lifetime risk of stroke in China is estimated to be 39.9%, ranking among the highest worldwide [1, 2]. Following cerebral ischemia, interruption of blood flow initiates a complex cascade of pathological events, including energy failure, blood-brain barrier (BBB) disruption, oxidative stress, inflammatory activation, and apoptosis, ultimately leading to neuronal death and neurological deficits [3].

Current acute treatments for ischemic stroke mainly include intravenous thrombolysis with recombinant tissue plasminogen activator and endovascular thrombectomy. However, these reperfusion strategies are restricted by narrow therapeutic windows, strict eligibility criteria, and the risk of hemorrhagic complications [4, 5]. Moreover, reperfusion therapy is primarily designed to restore blood flow and has limited capacity to reverse established neuronal injury or promote long-term neurovascular repair. Two major pathological challenges remain after ischemic stroke: irreversible neuronal loss in the ischemic core and secondary injury in the ischemic penumbra, which is driven by BBB disruption, vasogenic edema, inflammatory infiltration, oxidative stress, and glial activation [6, 7]. Therefore, therapeutic strategies capable of modulating the post-ischemic microenvironment, reducing secondary injury, and supporting neurovascular remodeling are urgently needed.

Mesenchymal stem cells (MSCs) are multipotent stromal cells with paracrine, immunomodulatory, trophic, angiogenic, and microenvironment-regulating properties, and have been widely investigated as candidate therapeutic products or strategies for ischemic stroke [8, 9]. Among different MSC sources, bone marrow-derived mesenchymal stem cells (BMSCs) remain one of the earliest, most extensively characterized, and most frequently studied cell populations in experimental and translational stroke research [10–13]. Their biological properties, including immunomodulatory capacity, secretion of neurotrophic and angiogenic factors, interaction with the injured neurovascular unit, and release of extracellular vesicles (EVs), have been investigated across multiple ischemic stroke-related models [14–16]. In addition, BMSCs have accumulated early clinical safety and feasibility data, as well as relatively well-documented culture, phenotypic, paracrine, and safety profiles [10–13, 17, 18]. These features make BMSCs an important reference cell population for evaluating MSC-based therapeutic strategies in ischemic stroke.

BMSCs were selected as the focus of this review because they represent a well-characterized MSC population with a substantial body of preclinical, translational, and early clinical evidence in ischemic stroke. As a narrative review, this article synthesizes representative evidence from ischemic stroke-related experimental models, including middle cerebral artery occlusion, cerebral ischemia-reperfusion injury, and oxygen-glucose deprivation/reoxygenation systems, together with translational and early clinical studies of BMSC-based therapy.

To improve transparency in literature selection, relevant studies were identified through searches of PubMed, Web of Science, and Google Scholar, supplemented by manual screening of references from key articles. The main search terms included “ischemic stroke,” “cerebral ischemia,” “middle cerebral artery occlusion,” “cerebral ischemia-reperfusion injury,” “oxygen-glucose deprivation/reoxygenation,” “bone marrow-derived mesenchymal stem cells,” “bone marrow-derived mesenchymal stromal cells,” “BMSCs,” “extracellular vesicles,” “exosomes,” “neuroinflammation,” “BBB,” “angiogenesis,” “autophagy,” “mitochondrial transfer,” and “clinical trial.” Priority was given to studies directly related to BMSC-mediated neuroprotection, post-ischemic microenvironmental regulation, delivery strategies, extracellular vesicle-based approaches, clinical translation, and manufacturing considerations in ischemic stroke. Evidence from non-ischemic neurological diseases or non-BMSC sources was included only when it provided necessary background for source comparison, terminology clarification, or translational context. The following sections discuss relatively well-supported preclinical mechanisms, emerging mechanisms, administration routes, functional enhancement strategies, extracellular vesicle-based approaches, clinical evidence, and manufacturing and regulatory considerations.

Preclinical mechanisms of BMSCs in ischemic stroke

The therapeutic effects of BMSCs in ischemic stroke have been investigated mainly in preclinical models, including transient or permanent middle cerebral artery occlusion (tMCAO/MCAO), cerebral ischemia-reperfusion injury (CIRI), and oxygen-glucose deprivation/reoxygenation (OGD/R)-based cellular systems. These studies suggest that BMSCs can influence multiple pathological processes after ischemic injury, including neuroinflammation, apoptosis, oxidative stress, BBB disruption, angiogenesis, axonal remodeling, autophagy, mitochondrial dysfunction, and brain-organ immune-metabolic communication. This section first discusses mechanisms that have been repeatedly reported in experimental stroke models and are therefore relatively well supported at the preclinical level. Pyroptosis, autophagy regulation, mitochondrial transfer, and brain-organ axis regulation are then discussed as emerging mechanisms that require further validation. The mechanistic framework is summarized in six schematic figures, which illustrate the major and emerging pathways through which BMSCs may regulate the post-ischemic microenvironment and support neurovascular repair (Figs. 1, 2, 3, 4, 5 and 6).

Fig. 1.

Proposed preclinical mechanisms by which BMSCs attenuate ischemic brain injury through modulation of neuroinflammation, oxidative stress, apoptosis, and pyroptosis. BMSCs may regulate multiple pathological processes after ischemic stroke, including suppression of inflammatory signaling, enhancement of antioxidant defenses, inhibition of apoptosis-related pathways, and attenuation of NLRP3 inflammasome-mediated pyroptosis. This schematic is based primarily on findings from experimental studies, including OGD/R systems and animal stroke models, and should not be interpreted as established clinical mechanisms. Neuroinflammation, oxidative stress, and apoptosis are supported by relatively extensive preclinical evidence, whereas pyroptosis remains an emerging mechanism requiring further validation. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; Bcl-2, B-cell lymphoma 2; Bax, Bcl-2-associated X protein; Cyt-c, cytochrome c; GCLC, glutamate-cysteine ligase catalytic subunit; GCLM, glutamate-cysteine ligase modifier subunit; GCL, glutamate-cysteine ligase; GGCT, γ-glutamyl cyclotransferase; GSDMD, gasdermin D; GSH, glutathione; HO-1, heme oxygenase-1; IκB, inhibitor of kappa B; IL, interleukin; JAK/STAT, Janus kinase/signal transducer and activator of transcription; JNK, c-Jun N-terminal kinase; c-Jun, transcription factor jun; MDA, malondialdehyde; MOMP, mitochondrial outer membrane permeabilization; NF-κB, nuclear factor-kappa B; p50, nuclear translocation of NF-κB p50; p65, nuclear translocation of NF-κB p65; NLRP3, nod-like receptor family pyrin domain-containing 3; Nrf2, nuclear factor erythroid 2-related factor 2; RIP1, receptor-interacting protein kinase 1; RIP3, receptor-interacting protein kinase 3; ROS, reactive oxygen species; SIRT1, sirtuin 1; SOD, superoxide dismutase; CAT, catalase; TGF-β, transforming growth factor-β; Th17, T helper 17 cells; TLR4, toll-like receptor 4; TNF-α, tumor necrosis factor-α; TRAF6, TNF receptor-associated factor 6; Treg, regulatory T cells; TSG-6, TNF-α-stimulated gene-6; C3, complement component 3

Fig. 2.

Proposed preclinical mechanisms by which BMSCs protect the blood-brain barrier after ischemic stroke. BMSCs may preserve blood-brain barrier integrity by maintaining junction-associated proteins, reducing MMP-9-mediated matrix and tight junction degradation, downregulating ICAM-1 and VCAM-1, inhibiting the PKCδ/ERK1/2/Ets-1/MMP-9 signaling axis, decreasing AQP4-associated vasogenic edema, and supporting pericyte-like vascular remodeling. This schematic is based primarily on findings from in vitro BBB-related models and animal studies of ischemic stroke. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; BMECs, brain microvascular endothelial cells; AQP4, aquaporin-4; ERK1/2, extracellular signal-regulated kinase 1/2; Ets-1, E26 transformation-specific sequence-1; ICAM-1, intercellular adhesion molecule-1; JAM-A, junctional adhesion molecule-A; MMP-9, matrix metalloproteinase-9; PDGFR-β, platelet-derived growth factor receptor-β; PKCδ, protein kinase Cδ; N-cadherin, neural cadherin‌; SDF-1α, stromal cell-derived factor-1α; VCAM-1, vascular cell adhesion molecule-1; VEGF, vascular endothelial growth factor; ZO-1, zonula occludens-1; α-SMA, alpha-smooth muscle actin

Fig. 3.

Proposed preclinical mechanisms by which BMSCs promote neurotrophic support, angiogenesis, and neurovascular remodeling after ischemic stroke. A BMSCs may enhance neurotrophic support by increasing BDNF-related signaling, supporting DCX-positive neural precursor responses, and promoting endogenous neural repair. B BMSCs may facilitate angiogenesis and neurovascular remodeling through VEGF-related signaling, Notch1/Hes1 activation, and Rabep2-mediated VEGFR2 regulation. These mechanisms are based primarily on experimental stroke models and should not be interpreted as established clinical mechanisms. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; BDNF, brain-derived neurotrophic factor; DCX, doublecortin; Hes1, hairy and enhancer of split-1; Notch1, notch receptor 1; Rabep2, RAB GTPase-binding effector protein 2; VEGF, vascular endothelial growth factor; VEGFR2, vascular endothelial growth factor receptor 2

Fig. 4.

Proposed preclinical mechanisms by which BMSCs support axonal remodeling, synaptic reconstruction, and myelin repair after ischemic stroke. BMSCs may enhance structural plasticity by increasing MAP-2, GAP-43, SYP, MBP, and NF-200, reducing GSK3β/CDK5-associated CRMP2 phosphorylation, and downregulating axon growth-inhibitory molecules such as Nogo-A, MAG, and OMgp. These effects may contribute to axonal remodeling, synaptic reconstruction, and myelin repair in experimental stroke models. This schematic is based primarily on preclinical evidence and should be interpreted as BMSC-associated enhancement of structural repair rather than direct neuronal replacement or established clinical neural circuit integration. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; CDK5, cyclin-dependent kinase 5; CRMP2, collapsin response mediator protein 2; p-CRMP2, phosphorylated CRMP2; GAP-43, growth-associated protein 43; GSK3β, glycogen synthase kinase-3β; MAG, myelin-associated glycoprotein; MAP-2, microtubule-associated protein-2; SYP, synaptophysin; MBP, myelin basic protein; NF-200, neurofilament-200; Nogo-A, neurite outgrowth inhibitor-A; OMgp, oligodendrocyte myelin glycoprotein

Fig. 5.

Proposed preclinical mechanisms by which BMSCs regulate autophagy and mitochondrial transfer after ischemic stroke. A BMSCs may modulate autophagic flux through AMPK/mTOR, PI3K/AKT, SYNPO2-related signaling, and autophagy-associated proteins, including LC3-II/LC3-I, Beclin-1, p62, ATG5, and ATG12. These changes are interpreted as regulation of autophagic balance rather than simple activation or inhibition of autophagy. B BMSCs may provide mitochondrial support to injured neurons through Miro1-associated mitochondrial transfer, thereby improving mitochondrial function and cellular recovery. These mechanisms are based primarily on in vitro and animal studies.The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; AKT, protein kinase B; AMPK, AMP-activated protein kinase; ATG12, autophagy-related protein 12; ATG5, autophagy-related protein 5; LC3-I, microtubule-associated protein 1 light chain 3-I; LC3-II, microtubule-associated protein 1 light chain 3-II; Miro1, mitochondrial Rho GTPase 1; mTOR, mechanistic target of rapamycin; PI3K, phosphatidylinositol 3-kinase; p62, sequestosome 1; SYNPO2, synaptopodin-2

Fig. 6.

Proposed preclinical brain-organ axis mechanisms of BMSCs after ischemic stroke. BMSCs may modulate post-stroke brain-organ communication through gut-brain and gut-liver-brain axis regulation. In the gut-brain axis, BMSCs may improve gut microbiota composition, increase beneficial bacteria such as Lactobacillus, enhance short-chain fatty acid-related signaling, reduce microbial translocation, and attenuate T cell-associated systemic inflammation and neuroinflammation. In the gut-liver-brain axis, BMSCs may influence hepatic GC metabolism involving HSD11B1, HSD11B2, and SRD5A1, thereby supporting intestinal mucus barrier homeostasis and reducing secondary brain injury. This schematic is based primarily on animal studies and represents an emerging preclinical mechanism. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; GC, glucocorticoid; GPCR, G protein-coupled receptor; HSD11B1, 11β-hydroxysteroid dehydrogenase type 1; HSD11B2, 11β-hydroxysteroid dehydrogenase type 2; SCFAs, short-chain fatty acids; SRD5A1, steroid 5alpha-reductase 1; PPARγ, peroxisome proliferator-activated receptor γ; T cell, T lymphocyte

Regulation of neuroinflammation

Neuroinflammation is a major contributor to secondary brain injury after ischemic stroke. Preclinical studies suggest that BMSCs attenuate post-ischemic inflammatory injury mainly by regulating innate and adaptive immune responses and reshaping the inflammatory microenvironment [19–23]. In experimental models of ischemic stroke, BMSC treatment has been reported to inhibit toll-like receptor 4 (TLR4)/nuclear factor-kappa B (NF-κB) signaling by downregulating TLR4 expression, reducing nuclear translocation of NF-κB p65/p50, and increasing inhibitory kappa B (IκB) expression [24]. These changes are associated with reduced production of pro-inflammatory mediators and suppression of inflammatory amplification within the ischemic brain.

BMSCs may also regulate the balance between T helper 17 (Th17) cells and regulatory T (Treg) cells, an immune axis that involved in post-stroke inflammatory injury [25, 26]. Experimental evidence indicates that BMSCs can inhibit Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling, reduce interleukin-17 (IL-17)-mediated inflammatory responses, and enhance immunoregulatory activity [25]. In addition, BMSCs have been reported to increase the expression of tumor necrosis factor-alpha-stimulated gene-6 (TSG-6), which may contribute to the attenuation of complement-associated neuronal injury [27].

Regulation of microglial activation represents another important component of BMSC-mediated immunomodulation. In preclinical stroke models, BMSC treatment has been associated with a phenotypic shift from a pro-inflammatory M1-like state toward an anti-inflammatory or reparative M2-like state. This shift is reflected by changes in markers such as ionized calcium-binding adaptor molecule 1 (Iba1) and cluster of differentiation 206 (CD206), together with increased production of anti-inflammatory mediators, including interleukin-10 (IL-10) and transforming growth factor-beta (TGF-β) [28, 29]. BMSCs may further suppress neuroinflammatory responses by enhancing sirtuin 1 (SIRT1) expression and inhibiting NF-κB activity through deacetylation-dependent mechanisms [30].

Inhibition of apoptosis

Apoptosis is an important mechanism of delayed neuronal loss in the ischemic penumbra. After cerebral ischemia, an increased B-cell lymphoma 2-associated X protein (Bax)/B-cell lymphoma 2 (Bcl-2) ratio promotes mitochondrial outer membrane permeabilization, followed by the release of cytochrome c (Cyt c) and apoptosis-inducing factor into the cytosol. These events activate the caspase cascade, particularly caspase-3, and ultimately contribute to apoptotic cell death [31–34]. Receptor-interacting protein kinases 1 and 3 (RIP1 and RIP3) may further amplify cell death signaling and exacerbate ischemic injury in a caspase-mediated manner [35].

BMSCs may attenuate apoptosis by regulating mitochondrial and death receptor-associated signaling pathways [30, 31, 36, 37]. Reported effects include increased expression of neuronal nuclei (NeuN), fewer TUNEL-positive cells, and reduced infarct volume [38]. At the molecular level, BMSC-mediated anti-apoptotic effects have been linked to downregulation of Bax, caspase-3, cleaved caspase-3, RIP1, and RIP3, together with upregulation of Bcl-2 [35, 39]. BMSCs have also been reported to increase survivin expression, suppress tumor necrosis factor receptor-associated factor 6 (TRAF6)-related apoptotic signaling, reduce mitochondrial Cyt c release, and inhibit c-Jun N-terminal kinase (JNK) phosphorylation within the mitogen-activated protein kinase (MAPK) pathway [40, 41]. These findings suggest that BMSCs may limit post-ischemic neuronal apoptosis by preserving mitochondrial stability, suppressing caspase activation, and modulating death-associated stress signaling.

Reduction of oxidative stress

Oxidative stress is a major contributor to ischemic brain injury and is characterized by excessive production of reactive oxygen species (ROS), lipid peroxidation, protein oxidation, and nucleic acid damage [42, 43]. BMSCs may strengthen antioxidant defenses by activating the nuclear factor erythroid 2-related factor 2 (Nrf2)/heme oxygenase-1 (HO-1) pathway, regulating glutathione (GSH) metabolism, enhancing antioxidant enzyme activity, and reducing the malondialdehyde (MDA) levels [44–46].

BMSC treatment has been linked to increased expression of antioxidant enzymes such as superoxide dismutase and catalase, reduced MDA levels, decreased inducible nitric oxide synthase expression, and lower peroxynitrite generation [45, 47].

At the metabolic level, BMSCs may promote GSH biosynthesis by upregulating γ-glutamylcyclotransferase‌ (GGCT), catalytic (GCLC) and modifier (GCLM) subunits of glutamate-cysteine ligase (GCL) and increasing related metabolic intermediates such as 5-oxoproline [48–51]. These changes are associated with reduced ROS accumulation and improved antioxidant capacity [22, 52, 53]. Together, these findings indicate that BMSCs may alleviate oxidative injury through coordinated regulation of antioxidant signaling, nitric oxide-related oxidative damage, and glutathione-dependent redox homeostasis.

Protection of the blood-brain barrier

BBB disruption is a key pathological event after ischemic stroke and contributes to vasogenic edema, inflammatory cell infiltration, and secondary neuronal injury. The BBB is composed of brain microvascular endothelial cells, pericytes, astrocytic endfeet, basement membrane components, and intercellular junctional proteins, including occludin, claudin family members such as claudin-1 and claudin-5, zonula occludens-1 (ZO-1), junctional adhesion molecule-A (JAM-A), and neural cadherin (N-cadherin) [54, 55]. After ischemia, hypoxia and inflammatory cytokines, including interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α), can upregulate matrix metalloproteinase-9 (MMP-9), intercellular adhesion molecule-1 (ICAM-1), and vascular cell adhesion molecule-1 (VCAM-1), thereby promoting tight junction degradation, leukocyte adhesion, and increased barrier permeability [56].

BMSCs may protect BBB integrity by preserving tight junction proteins, reducing MMP-9 activity, and downregulating ICAM-1 and VCAM-1 [57, 58]. Mechanistically, BMSCs have been linked to modulation of protein kinase Cδ (PKCδ)-related signaling, inhibition of extracellular signal-regulated kinase 1/2 (ERK1/2) phosphorylation, and reduced E26 transformation-specific sequence-1 (Ets-1)-mediated transcriptional activation, thereby suppressing MMP-9 transcription and matrix degradation [47, 59, 60]. BMSCs have also been reported to modulate aquaporin-4 (AQP4) and claudin-1 expression, which may contribute to the attenuation of vasogenic edema in specific experimental settings [47, 48, 59]. In addition, their pericyte-like properties may contribute to microvascular remodeling through platelet-derived growth factor receptor-β (PDGFR-β), vascular endothelial growth factor (VEGF), alpha-smooth muscle actin (α-SMA), stromal cell-derived factor 1alpha (SDF-1α) and N-cadherin-related pathways [61]. Together, these findings suggest that BMSCs may help maintain BBB integrity by regulating endothelial junction stability, inflammatory adhesion, matrix degradation, edema formation, and microvascular support.

Neurotrophic support and angiogenesis

Neurotrophic and angiogenic factors are essential for neuronal survival, synaptic plasticity, endogenous neurogenesis, vascular remodeling, and neurovascular unit repair after ischemic injury [62]. BMSCs may improve the post-ischemic microenvironment by increasing the expression of neurotrophic factors such as brain-derived neurotrophic factor (BDNF) and by promoting doublecortin (DCX)-positive neural precursor cells in the cortex or hippocampus [62–64]. These effects suggest that BMSCs may support endogenous repair programs rather than acting solely through acute neuroprotection.

Angiogenesis is another important reparative process after ischemic stroke. In experimental stroke models, BMSC treatment has been associated with increased VEGF expression, greater microvessel density, and improved local perfusion within ischemic areas [65–69]. VEGF and BDNF may act synergistically to promote endothelial cell survival, migration, and neurovascular remodeling [70, 71]. At the signaling level, BMSCs may regulate the notch receptor 1 (Notch1)/hairy and enhancer of split-1 (Hes1) pathway or enhance RAB GTPase-binding effector protein 2 (Rabep2)-mediated VEGF receptor 2 (VEGFR2) signaling, thereby promoting endothelial proliferation and collateral circulation remodeling [72, 73]. These findings indicate that BMSCs may facilitate post-ischemic repair by coordinating neurotrophic support, angiogenic signaling, and neurovascular reconstruction.

Axonal remodeling and myelin repair

Axonal remodeling, synaptic plasticity, and myelin repair provide structural support for functional recovery after ischemic stroke [74, 75]. BMSC treatment has been associated with increased expression of microtubule-associated protein-2 (MAP-2) and growth-associated protein-43 (GAP-43), suggesting enhanced neuronal structural plasticity and axonal sprouting in the ischemic penumbra [76, 77]. Increased levels of myelin basic protein (MBP), high-molecular-weight neurofilament protein (HMW-NF), and neurofilament-200 (NF-200) further indicate improved axonal integrity and myelin preservation [34, 78, 79].

At the molecular level, He et al. reported that BMSCs may reduce collapsin response mediator protein 2 (CRMP2) phosphorylation by modulating glycogen synthase kinase-3β (GSK-3β) and cyclin-dependent kinase 5 (CDK5) signaling, thereby facilitating microtubule assembly and axonal extension [80]. BMSCs may also increase synaptophysin (SYP) expression while suppressing axon growth-inhibitory molecules, including neurite outgrowth inhibitor-A (Nogo-A), myelin-associated glycoprotein (MAG), and oligodendrocyte myelin glycoprotein (OMgp) [80, 81]. Together, these findings suggest that BMSCs may promote axonal remodeling, synaptic reconstruction, and myelin preservation. However, these effects should be interpreted as enhanced structural plasticity within the post-ischemic microenvironment rather than as direct evidence of neuronal replacement or durable neuronal integration.

Modulation of pyroptosis

Pyroptosis is an inflammatory form of programmed cell death driven mainly by activation of the NOD-like receptor family pyrin domain-containing 3 (NLRP3) inflammasome, caspase-1, and gasdermin D (GSDMD) [82, 83]. Once activated, caspase-1 promotes the maturation of interleukin-1β (IL-1β) and interleukin-18 (IL-18), whereas cleaved GSDMD forms membrane pores that amplify inflammatory injury [82, 84]. BMSCs have been reported to suppress key components of this pathway, including NLRP3, caspase-1, and GSDMD, thereby reducing pyroptosis-associated inflammatory injury [85].

Compared with inflammation, apoptosis, oxidative stress, and BBB protection, evidence for BMSC-mediated regulation of pyroptosis in ischemic stroke remains at an earlier stage. Whether inhibition of pyroptosis is a major contributor to functional recovery or mainly a secondary consequence of broader anti-inflammatory regulation requires further clarification.

Regulation of autophagy

Autophagy has a context-dependent role in cerebral ischemia-reperfusion injury. Moderate autophagy may help remove damaged organelles and maintain cellular homeostasis, whereas excessive, insufficient, or dysregulated autophagy can aggravate neuronal injury [86]. BMSCs may regulate autophagic activity through AMP-activated protein kinase (AMPK)/mechanistic target of rapamycin (mTOR) signaling and phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) signaling [37, 87]. Changes in the phosphorylation status of AMPK and mTOR, together with alterations in autophagy-related proteins such as microtubule-associated protein 1 light chain 3-II (LC3-II), LC3-I, Beclin-1, and sequestosome 1 (p62), suggest that BMSCs may influence autophagic flux after ischemic injury [37, 87].

BMSCs have also been linked to downregulation of synaptopodin-2 (SYNPO2) and modulation of autophagy-associated molecules, including p62, Beclin-1, and the LC3-II/LC3-I ratio [88, 89]. These findings indicate that BMSCs may protect neurons by restoring autophagic balance rather than simply increasing or suppressing autophagy. Because the functional consequences of autophagy depend on timing, cell type, ischemic severity, and reperfusion status, BMSC-mediated autophagy regulation should be interpreted as a context-dependent and still emerging mechanism.

Mitochondrial transfer

Mitochondrial transfer has been proposed as a direct intercellular support mechanism through which BMSCs deliver functional mitochondria to injured cells. This process may occur through tunneling nanotubes or extracellular vesicle-associated pathways and may improve energy metabolism, reduce oxidative stress, and attenuate apoptosis and inflammation [90, 91]. BMSCs can upregulate the mitochondrial Rho GTPase 1 (Miro1), thereby enhancing mitochondrial transfer to injured neurons and improving mitochondrial function [92].

Despite its biological appeal, the contribution of mitochondrial transfer to BMSC-mediated protection in ischemic stroke remains incompletely defined. Key questions include whether mitochondrial transfer occurs efficiently in vivo after BMSC administration, whether it reaches a sufficient magnitude to influence tissue repair, and how much it contributes to functional recovery relative to paracrine and immunomodulatory mechanisms.

Brain-organ axis mechanisms

The brain-gut axis is a bidirectional communication network involving the central nervous system, autonomic and neuroendocrine signaling, immune responses, intestinal barrier integrity, and gut microbiota [93, 94]. After ischemic stroke, hypothalamic-pituitary-adrenal (HPA) axis activation, autonomic dysregulation, increased intestinal permeability, microbial dysbiosis, and bacterial translocation may amplify systemic inflammation and aggravate neurological injury [95–98]. Reduced abundance of beneficial bacteria and decreased production of short-chain fatty acids (SCFAs) may further impair intestinal barrier function and immune homeostasis [97, 99, 100]. BMSCs have been reported to modulate gut microbiota composition, strengthen intestinal barrier integrity, reduce bacterial translocation, and attenuate systemic inflammation and neuroinflammation in stroke models, thereby contributing to neural recovery [101].

The gut-liver-brain axis has also emerged as a potential inter-organ pathway in ischemic stroke. BMSCs may preferentially migrate to the liver, where they regulate glucocorticoid metabolism through peroxisome proliferator-activated receptor-gamma (PPARγ)-mediated modulation of 11β-hydroxysteroid dehydrogenase type 1 (HSD11B1), 11β-hydroxysteroid dehydrogenase type 2 (HSD11B2), and steroid 5alpha-reductase 1 (SRD5A1) to support intestinal mucus barrier homeostasis [102, 103]. This hepatic metabolic regulation may indirectly improve intestinal mucus barrier function, limit gut-derived bacterial translocation and reduce secondary brain injury.

Although these findings broaden the mechanistic framework of BMSC therapy beyond the brain itself, brain-organ axis regulation remains an emerging area. Its reproducibility, dose dependence, source specificity, and relevance to human ischemic stroke require further clarification.

Administration routes and route-dependent brain exposure of BMSCs

Several administration routes have been explored for BMSC-based therapy in ischemic stroke, including intravenous, intra-arterial, intrathecal, intraventricular, intranasal, and local stereotactic delivery. These approaches differ in technical feasibility, safety profile, biodistribution, cell retention, and actual exposure of the ischemic brain to administered cells. Therefore, the administered cell dose should not be equated with the biologically effective dose reaching the target tissue. A route-by-route comparison of delivery approach, main advantages, limitations, translational considerations, and supporting references is provided in Table 1.

Table 1.

Comparison of administration routes for BMSC-based therapy in ischemic stroke

Administration route	Delivery approach	Main advantages	Main limitations	Translational considerations	References
Intravenous (IV)	Systemic delivery through peripheral veins	Minimally invasive; repeatable; technically simple; compatible with large-scale clinical use	Prominent pulmonary first-pass retention; low and variable cerebral exposure; therapeutic effects may depend partly on systemic immunomodulation rather than direct brain engraftment	Most clinically practical route; supported by early autologous BMSC stroke studies and MSC biodistribution studies	[10, 11, 104–109, 118]
Intra-arterial (IA)	Delivery via arteries supplying the ischemic territory	May increase local cerebral exposure and reduce initial pulmonary sequestration	Requires interventional expertise; exposure depends on vascular patency, lesion location, cell size, suspension quality, dose, and infusion rate; risk of microvascular obstruction or embolic events	Potentially useful in selected settings but requires strict safety optimization	[105, 107, 108, 110, 111, 118]
Intrathecal (IT, lumbar puncture)	Infusion into the CSF via lumbar puncture	Partially bypasses systemic first-pass effects; repeatable; less invasive than intracerebral injection	CSF exposure does not guarantee delivery to ischemic parenchyma; distribution depends on CSF dynamics	Moderately feasible but operator-dependent; clinical utility remains insufficiently established	[112, 113]
Intracerebroventricular (ICV)	Injection into the ventricular system with CSF-mediated distribution	Partially bypasses the BBB and increases exposure within the CSF compartment	Invasive; requires stereotactic or imaging guidance; risk of infection, hemorrhage, inflammation, or hydrocephalus	Mainly experimental or trial-oriented; limited routine applicability	[107, 119]
Intranasal (IN)	Noninvasive delivery through olfactory and trigeminal-associated pathways	Avoids surgical access; repeatable; may support brain-directed delivery	Actual delivered dose, lesion exposure, and cell persistence are difficult to quantify; high inter-individual variability; protocols remain insufficiently standardized	Promising but still largely experimental for BMSC-based stroke therapy	[115–117]
Stereotactic intracerebral or peri-infarct implantation	Direct local delivery into or around the infarct region	High local exposure; suitable for chronic-stage local repair strategies; enables lesion-targeted delivery and imaging follow-up	Highly invasive; less suitable for acute stroke; requires neurosurgical expertise; local tissue responses need careful interpretation	Clinically explored in selected chronic or subacute stroke studies, including modified allogeneic BMSC-derived products and labeled autologous BMSCs	[18, 114]

This table summarizes route-specific feasibility, biodistribution, exposure, and safety considerations based on representative BMSC or MSC studies in ischemic stroke and related biodistribution research. Numerical values that are not consistently supported by direct route-specific evidence were not retained in the main table. Reported cell dose, infusion rate, labeling method, stroke phase, and follow-up duration vary substantially across studies and should not be directly compared across routes without considering product type, viable cell dose, delivery timing, lesion characteristics, and route-dependent exposure. BBB, blood-brain barrier; BMSCs, bone marrow-derived mesenchymal stem cells; CSF, cerebrospinal fluid; MSCs, mesenchymal stem cells

Intravenous administration is the most commonly used route in clinical studies because it is technically simple, repeatable, and compatible with large-scale clinical application [10, 104, 105]. However, brain delivery after intravenous infusion is limited and variable. A large proportion of infused cells may be retained in the pulmonary circulation and subsequently redistributed to the liver, spleen, and other organs enriched in the mononuclear phagocyte system, thereby limiting direct cerebral exposure [105–109]. Accordingly, the therapeutic effects of intravenously administered BMSCs may depend not only on cell entry into the injured brain but also on systemic immunomodulation, paracrine signaling, and interactions with peripheral immune organs.

Intra-arterial administration can theoretically increase local brain exposure by delivering cells closer to the ischemic vascular territory. However, target-site delivery depends on vascular patency, lesion location, collateral circulation, cell size, suspension quality, infused dose, and infusion rate [108, 110]. This route also carries safety concerns, including microvascular obstruction, embolic events, and procedure-related complications [107, 111]. Thus, intra-arterial delivery may be useful in selected experimental or clinical settings, but it requires careful optimization and should not be considered universally superior to other routes.

Intrathecal and intraventricular administration can partially bypass the BBB and increase exposure within the cerebrospinal fluid (CSF) compartment. Nevertheless, these approaches are invasive and may be associated with infection, hemorrhage, CSF flow disturbance, and other procedure-related risks [112–114]. Moreover, cerebrospinal fluid exposure does not necessarily ensure efficient delivery to the ischemic parenchyma. Intranasal administration is less invasive and may support brain-directed delivery through olfactory and trigeminal nerve-associated pathways. However, the actual delivered dose, lesion-specific exposure, and persistence of BMSCs after intranasal delivery remain difficult to quantify, and standardized protocols are still lacking [115–117].

Overall, the optimal administration route for BMSC therapy in ischemic stroke remains unresolved. Route selection should be guided by stroke phase, BBB permeability, lesion location, inflammatory status, safety considerations, clinical feasibility, and the intended mechanism of action. Future studies should report route-specific biodistribution, cell persistence, administration parameters, target-site exposure, pharmacodynamic biomarkers, and functional outcomes, rather than relying only on the administered cell number.

Preclinical functional enhancement strategies for BMSCs

BMSCs have been widely investigated as cell-based therapeutic candidates for ischemic stroke. However, their therapeutic efficacy may be limited by inefficient homing to ischemic tissue, poor survival in the post-ischemic microenvironment, donor- and passage-related heterogeneity, and variability in paracrine activity. To improve their reparative potential, several functional enhancement strategies have been explored, including genetic modification, hypoxic or pharmacological preconditioning, extracellular vesicle-based strategies, combination approaches, and engineered delivery systems.

These strategies differ in their mechanistic rationale, technical complexity, and translational maturity. Genetic modification and preconditioning have relatively consistent support in ischemic stroke-related experimental models, particularly in improving BMSC survival, migration, stress resistance, and paracrine function. BMSC-derived EVs provide a cell-free alternative that may retain part of the biological activity of parental BMSCs, but their characterization, dose definition, storage, and potency assessment remain insufficiently standardized. Combination strategies involving traditional Chinese medicine (TCM), chemical compounds, physical stimulation, or advanced delivery systems are promising but most remain largely exploratory or at an early translational stage. The following sections summarize representative strategies and their reported effects on BMSC function and post-ischemic repair. Figure 7 provides a conceptual overview of the proposed, partly overlapping mechanisms underlying these effects, based primarily on evidence from preclinical studies.

Fig. 7.

Functional enhancement strategies for BMSCs and their proposed mechanisms in ischemic stroke. This schematic presents a conceptual overview of strategies designed to enhance BMSC function in ischemic stroke, including genetic modification, preconditioning, BMSC-derived EVs, combination approaches, and engineered delivery systems. These strategies have been reported or proposed to improve BMSC survival, migration, paracrine activity, and lesion-directed repair in experimental ischemic stroke models. The mechanisms summarized in the figure include attenuation of neuroinflammation, oxidative stress, apoptosis, and pyroptosis, together with BBB protection, trophic support, angiogenesis, axonal remodeling, autophagy regulation, mitochondrial transfer, gut-brain axis regulation, gut-liver-brain axis regulation, and neural repair. These mechanisms should be interpreted as proposed, partly overlapping, and strategy-dependent pathways derived mainly from preclinical studies, rather than as clinically established relationships or as evidence that each enhancement strategy acts through all listed mechanisms. Autophagy regulation, mitochondrial transfer, gut-brain axis regulation, and gut-liver-brain axis regulation remain emerging preclinical mechanisms that require further validation. The illustration was independently developed by the authors using the Figdraw platform (www.figdraw.com). BMSCs, bone marrow-derived mesenchymal stem cells; EVs, extracellular vesicles

Genetic modification

Genetic modification has been used to enhance selected biological properties of BMSCs, including survival, migration, antioxidant capacity, trophic factor secretion, angiogenic activity, and resistance to the ischemic microenvironment. Chen et al. reported that BMSCs overexpressing extracellular superoxide dismutase reduced oxidative injury, apoptosis, and infarct volume after transplantation [120]. BMSCs modified with fibronectin type III domain-containing protein 5 (FNDC5) may regulate autophagy-related signaling by downregulating specificity protein 1 (Sp1) and its downstream Unc-51-like autophagy activating kinase 2 (ULK2), thereby improving cell survival and neuroprotective activity [121].

Other genetic modifications have focused on strengthening the paracrine and reparative functions of BMSCs. Overexpression of neurotrophic or angiogenic factors, such as nerve growth factor (NGF), hepatocyte growth factor (HGF), and VEGF, has been associated with enhanced trophic support, vascular remodeling, and tissue repair in ischemic injury models [46, 63, 122, 123]. Modulation of migration- and survival-related pathways, including extracellular signal-regulated kinase 1/2 (ERK1/2), C-C chemokine receptor type 2 (CCR2), and Notch1, may further improve BMSC homing, resistance to apoptosis, inflammatory regulation, BBB protection, and interactions with the injured neurovascular unit [124, 125]. In summary, genetic modification provides a targeted approach to enhancing specific BMSC functions, particularly antioxidative defense, trophic factor secretion, migration, and survival.

Preconditioning of BMSCs

Preconditioning aims to prepare BMSCs for the hostile post-ischemic microenvironment before administration. Unlike genetic modification, this approach uses environmental or pharmacological stimuli to enhance stress tolerance, migration, and paracrine activity without permanently altering the cellular genome.

Hypoxic preconditioning is one of the most extensively studied approaches. Non-lethal hypoxia can activate adaptive responses in BMSCs and increase the expression of chemokine receptors and migration-related molecules [126, 127]. Hypoxia-preconditioned BMSCs show enhanced migration toward ischemic regions, partly through upregulation of C-X-C chemokine receptor 4 (CXCR4) and activation of SDF-1α/CXCR4 axis [128]. Hypoxia may also strengthen the vascular and trophic functions of BMSCs. For example, hypoxia-preconditioned BMSCs have been linked to perivascular aggregation, collateral remodeling, and blood flow recovery through pathways involving VEGF and RAB GTPase-binding effector protein 2 (Rabep2) [68]. Reported effects include reduced infarct size, improved neurological outcomes, enhanced angiogenesis, increased neurotrophic factor release, and attenuation of inflammatory or cell death-related responses [128–130]. These findings suggest that hypoxia induces a BMSC phenotype with improved migratory capacity and stress resistance.

Stroke-serum preconditioning represents a disease-relevant strategy that exposes BMSCs to humoral factors generated after ischemic injury. Post-stroke serum, compared with healthy serum, has been reported to increase the expression of VEGF, BDNF, MMP-9, C-C chemokine receptor type 7 (CCR7), NGF, and HGF in BMSCs [131, 132]. These molecules are involved in cell migration, neurotrophic support, angiogenesis, and tissue repair. Administration of stroke-serum-preconditioned BMSCs has been associated with reduced infarct volume, increased angiogenesis- and neurogenesis-related markers, decreased apoptosis, and modulation of inflammatory responses [131].

Pharmacological preconditioning uses defined compounds to activate reparative properties of BMSCs before administration. This strategy may enhance migration and homing by regulating receptors such as CXCR4 and may improve cell survival under ischemia-like stress conditions [133, 134]. Pharmacologically preconditioned BMSCs have also been associated with reduced infarct volume, decreased edema, and improved functional recovery in selected experimental settings [134, 135]. Compared with more complex conditioning systems, pharmacological preconditioning may offer a more controllable way to modulate BMSC function, provided that the active compound, concentration, exposure duration, washout procedure, and release criteria are clearly defined.

BMSC-derived extracellular vesicles

EVs are heterogeneous lipid bilayer-enclosed particles released by cells. According to their biogenesis, EVs are commonly described as including exosomes, microvesicles, and apoptotic bodies [136, 137]. As key mediators of paracrine communication, BMSC-derived EVs have attracted increasing interest as a cell-free strategy for ischemic stroke. They can transfer proteins, lipids, mRNAs, microRNAs, and other regulatory cargo from parental BMSCs to recipient cells, thereby modulating the post-ischemic microenvironment [138–140]. In ischemic stroke-related models, BMSC-derived EVs have been reported to regulate neuroinflammation, apoptosis, pyroptosis, autophagy, angiogenesis, BBB integrity, and glial responses [139, 141–144].

Liu et al. reported that BMSC-derived exosomes attenuated cerebral ischemia-reperfusion injury by suppressing NLRP3 inflammasome-mediated neuroinflammation and pyroptosis and by modulating microglial M1/M2 polarization [145]. Consistently, BMSC-derived EVs have been associated with reduced expression of pro-inflammatory microglia/macrophage markers, including inducible nitric oxide synthase (iNOS), cluster of differentiation 68 (CD86), and TNF-α, together with increased expression of anti-inflammatory markers such as arginase-1 (Arg-1), TGF-β, and IL-10. Beyond inflammatory regulation, BMSC-derived EVs may also influence autophagic activity and BBB stability [127, 133, 146]. Reported mechanisms include AMPK-dependent autophagic flux, ETS variant transcription factor 4 (ETV4)/SIRT1-mediated regulation of autophagic flux, suppression of caveolin-1-dependent endocytosis of ZO-1 and claudin-5, and modulation of the caveolin-1/CD147/VEGFR2/MMP-related signaling axis [143, 144, 147–149].

The biological activity of BMSC-derived EVs can be further shaped by hypoxic preconditioning or pharmacological treatment of parental BMSCs. These approaches may enrich selected cargo profiles or enhance specific functional effects, but they also introduce additional product variability. Therefore, BMSC-derived EVs provide a bridge between cell therapy and cell-free therapy, whereas their translational development requires standardized characterization, dose definition, storage stability, and mechanism-linked potency evaluation.

Combination strategies

Combination strategies aim to enhance BMSC-mediated repair by improving BMSC survival, migration, and paracrine activity while also modifying the injured microenvironment. TCM-derived compounds, extracts and formulas have been explored in this context because of their multi-target regulatory properties. Ligustrazine, puerarin, icariin, Dendrobium officinale-related preparations, Angelica gigas extract, ginsenoside Rg1, and other herbal-derived agents have been reported to enhance BMSC homing, angiogenesis, neurotrophic factor secretion, anti-apoptotic activity, inflammatory regulation, and BBB protection in ischemic stroke-related models. These effects are often associated with SDF-1α/CXCR4, PI3K/AKT, VEGF, BDNF, and inflammation-related signaling pathways. Representative studies are summarized in Table 2.

Table 2.

Representative TCM-derived compounds, formulas, and related pharmacological strategies combined with BMSCs in experimental ischemic stroke

Agent or formula	Administration or experimental approach	Reported molecular changes	Reported biological effects	References
Icariin	Caudal vein injection	↑ VEGF, BDNF, Bcl-2, p-ERK1/2, PI3K pathway	Inhibited neuronal apoptosis, promoted angiogenesis and neurogenesis	[64]
Hydroxysafflor yellow A	Caudal vein injection	↑TGF-β, IL-10 ↓TNF-α, IL-1β, MyD88	Inhibited the inflammatory response	[150]
Panax notoginseng saponins	Caudal vein injection	↑NGF	Enhanced neural growth-related responses, inhibited apoptosis	[151]
Mangiferin	1. Co-culture with microglia and astrocytes 2. Caudal vein injection	↑SDF-1α, CXCR4, Cyclin D, BDNF, NGF, IL-10 ↓ TNF-α	Promoted the proliferation and migration of BMSCs and inhibited the inflammatory response	[152]
Ginsenoside Rg1	Femoral vein injection	↑Bcl-2 ↓Bax	Reduced apoptosis	[153]
Angelica gigas extract	External carotid artery administration	↑Ang-1, Tie-2, VEGF-A, ZO-1, occludin	Promoted angiogenesis and reduced BBB permeability	[154]
Sodium ferulate	1. External carotid artery administration 2. Co-culture with human brain microvascular endothelial cells	↑VEGF, TGF-β, Ang-1, TSP4, CD31, BDNF PI3K/AKT/mTOR pathway ↓TNF-α, IL-1β, ROS	Inhibited inflammatory response, apoptosis	[155, 156]
Salvia miltiorrhiza extract	External carotid artery administration	↑ Bcl-2 caspase-3, p-ERK, p-AKT ↓Bax	Modulated autophagy-related signaling and promoted the survival of BMSCs	[157]
Dendrobium officinale polysaccharide	Lateral ventricle injection	↑Wnt-3a, Wnt pathway, β-tubulin III, Thy-1, Enolase ↓Jagged1, TNF-α, IL-1β, IL-6, Notch pathway	Increased neuronal differentiation-related marker expression and inhibited inflammatory response	[158, 159]
Salidroside	Lateral ventricle injection	↑NeuN ↓GFAP	Enhanced neurogenesis and inhibited neuroinflammation in the hippocampus	[160]
Tenuigenin	Lateral ventricle injection	↑SOX2, NeuN ↓DCX, GFAP	Promoted neural lineage marker expression in BMSCs	[161]
Puerarin	Intraperitoneal administration	↓Beclin-1, LC3-II/LC3-I	Modulated autophagy-related signaling	[37]
Tanshinone II	Intraperitoneal administration	↑BDNF ↓Nogo-A, NgR	Promoted axonal remodeling	[162]
N-butylphthalide	In vitro BMSC experiment	↑ Bcl-2 ↓TNF-α, IL-1β, IL-6, p65, caspase-3, Bax	Inhibited the inflammatory response and apoptosis	[163]
Astragaloside IV combined with Panax notoginseng saponins	Caudal vein injection	↑Ang-1, TGF-β1, VEGF	Enhanced angiogenesis-related responses	[164]
Shexiang Huangqi Compound dripping pills-containing/ Shexiang Huangqi Compound Dripping pill	Caudal vein injection	↑claudin-5, ZO-1 ↓ PKC, NF-κB, IL-1β, IL-6	Regulated BBB permeability and inhibited inflammatory response	[165]
		↑OATP1a4, OATP1 ↓ AQP4, P-gp	Supported BBB homeostasis	[166]
	Co-culture with astrocytes	↑GFAP	Promoted BMSC proliferation, migration, and astrocyte-like differentiation-related marker expression	[165]
Buyang Huanwu Decoction	Caudal vein injection	↑occludin, claudin-5, JAM-A, ZO-1	Preserved BBB tight junction integrity	[56]
	Lateral ventricle injection	↑VEGF, VEGFR2, p-VEGFR2, p-AKT, PI3K/AKT pathway	Induced endothelial cell proliferation and migration, promoted angiogenesis	[167]
		↑αVβ3, FAK, p-FAK,	Promoted angiogenesis and vascular stability	[168]
Zhali Nusi Formula	internal carotid artery administration	↑occludin, claudin-5	Preserved BBB tight junction integrity	[169]

Ang-1, angiopoietin-1; AKT, protein kinase B; AQP4, aquaporin-4; Bax, Bcl-2-associated X protein; Bcl-2, B-cell lymphoma 2; caspase-3, cysteine-dependent Aspartate-specific Protease-3; BDNF, brain-derived neurotrophic factor; CD31, cluster of differentiation 31; NF-κB, nuclear factor-kappa B; CXCR4, C-X-C chemokine receptor type 4; DCX, doublecortin; ERK, extracellular signal-regulated kinase; FAK, focal adhesion kinase; GFAP, glial fibrillary acidic protein; IL, interleukin; JAM-A, junctional adhesion molecule-A; LC3, microtubule-associated protein 1 light chain 3; p65, nuclear translocation of NF-κB p65; MyD88, myeloid differentiation primary response 88; NF-κB, nuclear factor-kappa B; NGF, nerve growth factor; NgR, Nogo receptor; NeuN, neuronal nuclei; OATP, organic anion-transporting polypeptide; P-gp, P-glycoprotein; PI3K, phosphatidylinositol 3-kinase; PKC, protein kinase C; ROS, reactive oxygen species; SDF-1α, stromal cell-derived factor-1 alpha; SOX2, SRY-box transcription factor 2; TGF-β, transforming growth factor-beta; Thy-1, thymocyte differentiation antigen-1; Tie-2, tunica interna endothelial cell kinase 2; TNF-α, tumor necrosis factor-alpha; TSP4, thrombospondin-4; VEGF, vascular endothelial growth factor; VEGF-A, vascular endothelial growth factor A; VEGFR2, vascular endothelial growth factor receptor 2; Wnt-3a, Wnt family member 3 A; ZO-1, zonula occludens-1; αVβ3, integrin αVβ3; β-tubulin III, beta-tubulin III; Nogo-A, neurite outgrowth inhibitor-A; Jagged1, jagged canonical Notch ligand 1‌; mTOR, mechanistic target of rapamycin; p-, phosphorylated; BMSCs, bone marrow-derived mesenchymal stem cells; BBB, blood-brain barrier. ↑, upregulated or increased; ↓, downregulated or decreased

Pharmacological agents have also been used to improve the host microenvironment or potentiate BMSC-based interventions. Stiripentol combined with BMSCs has been reported to reduce microglial activation and apoptosis and improve neurological function [170]. Pioglitazone combined with BMSCs has been associated with attenuated oxidative stress and inflammatory injury, reduced infarct volume, and improved tissue repair in ischemia-reperfusion models [53]. Oxiracetam combined with BMSCs may influence calpain activity and modulate transient receptor potential canonical 6 (TRPC6)/cAMP response element-binding protein (CREB)-related neuronal survival signaling [171]. These findings suggest that pharmacological modulation of the ischemic microenvironment may strengthen the paracrine and reparative effects of BMSCs.

Physical stimulation represents another combination approach. Electroacupuncture combined with BMSC administration has been associated with increased neural progenitor cell proliferation and elevated levels of BDNF and other neurotrophic factors in MCAO models [172, 173]. Ultrasound-targeted microbubble destruction may transiently increase BBB permeability, thereby facilitating the accumulation of intravenously administered BMSCs in ischemic brain tissue and enhancing anti-apoptotic or regenerative responses [174]. Pulsed electromagnetic fields and electrical stimulation have also been reported to influence BMSC proliferation, differentiation-related marker expression, migration, and inflammatory regulation [175, 176]. However, these approaches differ substantially in stimulation parameters, timing, safety profile, and compatibility with cell delivery. Their therapeutic value therefore depends on careful optimization rather than simple combination with BMSC administration.

Advanced delivery and engineering systems

Advanced delivery and engineering systems have been developed to improve BMSC targeting, cargo delivery, and persistence in ischemic tissue. Surface modification is one representative strategy. By incorporating targeting peptides, chemokine-responsive molecules, or membrane-associated functional groups, engineered BMSCs or BMSC-derived EVs may interact more efficiently with the ischemic microenvironment. Dual-modified BMSC-derived EVs carrying targeting peptides and therapeutic nucleic acid cargo have been investigated to improve BBB penetration and lesion-directed delivery in MCAO models [177]. Similarly, mangiferin-modified BMSCs may enhance CXCR4/SDF-1α signaling, increase migration-related activity, and promote neurotrophic factor secretion in ischemic injury models [152].

Biomimetic carriers and composite systems provide another strategy for combining targeting capacity with gene or cargo delivery. Peptide-modified BMSC membrane systems and non-viral vectors carrying regulatory microRNAs have been used to improve lesion localization and neuroprotective signaling. Huang et al. reported that Pep-133b-modified BMSCs integrate targeting peptide-mediated localization, miR-133b-associated neuroprotection, and the intrinsic reparative properties of BMSCs [178]. These systems illustrate how cell-based, EV-based, and gene delivery approaches can be integrated to enhance lesion targeting and therapeutic signaling.

Three-dimensional (3D) culture systems have also been investigated to improve the functional properties of BMSCs before administration. Compared with conventional two-dimensional (2D) adherent culture, 3D culture may better preserve cell-cell interactions and modulate cell size, migration-related receptor expression, and paracrine activity [179]. In BMSC-related experimental studies, 3D culture has been associated with preservation of multipotency, increased secretion of tissue-repair and immunomodulatory factors, enhanced neurotrophic support, and reduced pulmonary entrapment [180, 181]. Specifically, 3D BMSC culture has been reported to increase the secretion of interleukin-1 receptor antagonist (IL-1Ra), VEGF, and granulocyte colony-stimulating factor (G-CSF), while reducing microglial TNF-α production in vitro [182]. BMSC spheroids may also support neuronal repair through neurotrophic and immunomodulatory mechanisms [183]. These engineering and culture-based strategies may improve the biological activity of BMSC-based products, but their translational application requires reproducible manufacturing, product comparability assessment, potency testing, and safety evaluation.

Clinical research

Clinical studies of BMSC-based therapy for ischemic stroke have mainly focused on safety, feasibility, and preliminary signals of neurological recovery. Compared with the extensive preclinical literature, the available clinical evidence remains limited and heterogeneous. Considerable variation exists across studies with respect to cell source, product type, administration route, dose, treatment timing, stroke phase, rehabilitation exposure, follow-up duration, and outcome measures. The key characteristics and findings of representative clinical studies are presented in Table 3. Accordingly, the current clinical literature should be interpreted as preliminary translational evidence rather than definitive proof of efficacy.

Table 3.

Representative clinical studies of BMSC-based interventions for ischemic stroke

Cell type	Source	Route of administration	Dose	Timing	Stroke phase	Sample size	Study design	Main outcomes	Major limitations	References
Culture-expanded MSCs	Autologous bone marrow	Intravenous infusion	1 × 10^8 cells in total, given as two infusions of 5 × 10^7 cells	Approximately 4–5 and 7–9 weeks after stroke onset	Subacute severe ischemic stroke, mainly MCA territory infarction	30 patients; 5 treated and 25 controls	Prospective randomized controlled exploratory study	Feasible and well tolerated; preliminary functional improvement signals during follow-up	Limited treated sample size; limited blinding; no dose-response analysis; difficult to exclude spontaneous recovery or rehabilitation effects	[10]
Culture-expanded MSCs	Autologous bone marrow	Intravenous infusion	Divided infusions, approximately 1 × 10^8 cells in total	Subacute administration after cell expansion, followed by long-term observation	Severe ischemic stroke	52 patients included in the final analysis; 16 treated and 36 controls	Long-term follow-up study of intravenous autologous MSC transplantation	No major long-term cell-related safety signal; lower mRS scores and a higher proportion of patients with mRS 0–3 were reported during follow-up	Attrition from the initial cohort; nonblinded treatment allocation; limited number of treated patients; not a large multicenter phase III trial	[11]
Autologous serum-expanded MSCs	Autologous bone marrow	Intravenous infusion	0.6–1.6 × 10^8 cells per patient	36–133 days after stroke onset	Subacute to early chronic ischemic stroke	12 patients	Phase I/IIa, open-label, single-arm study	Feasible cell preparation and infusion; no major treatment-related adverse events; neurological improvement signals and imaging-based changes were reported	No randomized control group; limited treated sample size; heterogeneous timing and lesion characteristics; efficacy cannot be confirmed	[184]
Autologous MSCs	Autologous bone marrow	Intravenous infusion	1–3 × 10^8 cells	After ex vivo expansion during the subacute period; patients were enrolled within 2 weeks after stroke onset	Subacute moderate-to-severe ischemic carotid stroke	31 patients; 16 treated and 15 controls	Single-center, open-label randomized clinical trial with 2-year follow-up	Feasible and generally safe ; motor recovery signals were reported in motor-NIHSS, motor Fugl-Meyer assessment, and motor cortex fMRI activity	Small single-center open-label trial; autologous manufacturing feasibility was incomplete; global outcomes such as NIHSS, Barthel Index, and mRS were not uniformly improved	[186]
Culture-expanded MSCs	Autologous bone marrow	Intravenous infusion	2 × 10^6 cells/kg	Within 2 months after severe MCA infarction	Subacute MCA infarction	17 patients; 9 treated and 8 controls	Phase II randomized controlled trial with blinded outcome assessment	Safe and well tolerated; no significant between-group difference in NIHSS, mRS, or Barthel Index at 12 months; infarct volume change showed a possible signal	Limited number of treated patients; baseline infarct volume imbalance; underpowered for efficacy; no definitive functional benefit	[104]
Preconditioned autologous MSCs	Autologous bone marrow	Intravenous infusion	1 × 10^8 cells	5–89 days after stroke onset, within approximately 3 months	Subacute to early chronic major ischemic stroke with persistent deficits	54 patients; 39 treated and 15 controls	Prospective randomized open-label controlled trial with blinded outcome evaluation	Feasible and safe; no clear improvement in the primary global functional outcome; motor and neuroimaging analyses suggested possible recovery-related signals	Open-label treatment; limited sample size; short primary endpoint window; exploratory neuroimaging findings; clinical efficacy not established	[187]
Ischemia-tolerant MSCs	Allogeneic adult bone marrow	Intravenous infusion	Dose escalation of 0.5, 1.0, and 1.5 × 10^6 cells/kg; expanded cohort at 1.5 × 10^6 cells/kg	More than 6 months after ischemic stroke	Chronic ischemic stroke with persistent deficits	36 patients	Phase I/II open-label dose-escalation and expanded safety study	Generally safe and well tolerated; behavioral improvement signals were reported	No placebo control; open-label design; chronic selected population; product-specific allogeneic and hypoxia-conditioned MSC preparation	[188]
SB623, modified MSCs product	Allogeneic bone marrow	Stereotactic intracerebral or peri-infarct implantation	2.5, 5, or 10 × 10^6 cells	6–60 months after stroke onset	Chronic ischemic stroke with stable motor deficits	18 patients	Phase I/IIa open-label dose-escalation study with long-term follow-up	No dose-limiting toxicity; motor improvement signals were reported in some patients; peri-lesional MRI changes suggested local tissue responses	Invasive delivery; small uncontrolled cohort; selected chronic-stage population; genetically modified product; limited generalizability to unmodified BMSCs or acute/subacute stroke	[185]
MRI-trackable SPIO-labeled BMSCs	Autologous bone marrow	Stereotactic intracerebral or peri-infarct transplantation	2 or 5 × 10^7 cells	Subacute period after ischemic stroke	Subacute ischemic stroke	7 patients	Phase I/II first-in-human open-label safety and imaging-based study	Feasible transplantation and MRI-based cell tracking; no major safety signal reported	Very small cohort; no control group; imaging signal reflects exposure rather than validated efficacy; limited long-term efficacy data	[13]

BMSCs, bone marrow-derived mesenchymal stem cells; fMRI, functional magnetic resonance imaging; MCA, middle cerebral artery; MRI, magnetic resonance imaging; mRS, modified Rankin Scale; MSCs, mesenchymal stem cells; NIHSS, National Institutes of Health Stroke Scale; SPIO, superparamagnetic iron oxide

Early clinical studies primarily investigated intravenous infusion of autologous BMSCs in patients with severe ischemic stroke. These studies reported acceptable tolerability and suggested possible functional improvement signals during follow-up [10, 11, 184]. However, most cohorts were small, and several studies were open-label, early-phase, or only partially controlled. These design features limit the ability to distinguish cell-related effects from spontaneous recovery, rehabilitation-associated improvement, baseline heterogeneity, or placebo-related effects. Despite these limitations, these studies provided important preliminary evidence that BMSC administration is clinically feasible and has not raised major short-term safety concerns in selected patients.

Subsequent studies evaluated different BMSC-based products and delivery strategies. SB623, a modified allogeneic BMSC product, has been stereotactically implanted into peri-infarct regions in patients with chronic ischemic stroke and persistent motor deficits [12, 185, 186]. These studies reported no dose-limiting toxicity and suggested sustained motor improvement in some patients. Transient peri-lesional T2-fluid-attenuated inversion recovery (T2-FLAIR) changes were also observed and may reflect local tissue responses rather than durable cell engraftment. In another imaging-focused study, superparamagnetic iron oxide-labeled autologous BMSCs were stereotactically implanted in patients with subacute ischemic stroke, enabling magnetic resonance imaging (MRI)-based tracking of transplanted cells near the infarct region [13]. These approaches provided information on local delivery, lesion exposure, and cell tracking that cannot be obtained from systemic infusion studies alone.

Overall, current clinical evidence suggests that BMSC-based interventions are technically feasible and generally well tolerated in selected patients with ischemic stroke. However, the evidence remains preliminary and supports safety, feasibility, and biological plausibility more strongly than clinical efficacy. Larger randomized, blinded, adequately powered trials with long-term follow-up are needed to determine whether BMSC-based therapies can produce reproducible functional benefits beyond spontaneous recovery, standard rehabilitation, and supportive care.

Source-related advantages and translational positioning of BMSCs

MSCs derived from bone marrow, umbilical cord-associated tissues, adipose tissue, and other sources share core MSC features, but they differ in their translational profiles. Tissue source can influence accessibility, donor variability, expansion capacity, senescence, secretome composition, EV cargo, manufacturing scalability, and safety monitoring [14, 189–191]. Therefore, the rationale for focusing on BMSCs should be based not on ease of acquisition alone, but on their disease-specific evidence base, biological characterization, and translational positioning in ischemic stroke.

Comparative studies indicate that perinatal or umbilical cord-associated MSCs, including umbilical cord blood- and Wharton’s jelly-derived MSCs, often show greater expansion potential and fewer senescence-related features than adult tissue-derived MSCs. For example, umbilical cord blood-derived MSCs have been reported to exhibit higher proliferation and clonogenicity, together with lower expression of senescence-associated markers such as tumor protein p53 (p53), cyclin-dependent kinase inhibitor 1 (p21), and cyclin-dependent kinase inhibitor 2 A (p16), compared with bone marrow- and adipose tissue-derived MSCs [192, 193]. Wharton’s jelly-derived MSCs also have practical advantages, including non-invasive collection, lower donor-age-related variability, and suitability for scalable allogeneic banking [194, 195]. Adipose tissue-derived MSCs are abundant and relatively accessible, and their higher tissue yield may facilitate cell preparation [196]. By contrast, BMSCs require bone marrow aspiration and may be affected by donor age, vascular risk factors, and comorbidities.

Functional differences among MSC sources are also reflected in their secretomes. In a systematic comparison of human adipose tissue-, bone marrow-, and Wharton’s jelly-derived MSC secretomes under xenogeneic-free culture conditions, Wharton’s jelly-derived MSC secretomes showed the most comprehensive angiogenesis-related protein network, followed by BMSC secretomes. In the same experimental setting, adipose tissue-derived MSC secretomes showed lower expression of angiogenesis-related protein expression and weaker angiogenic functional responses [191]. Another proteomic study comparing secretomes from adipose tissue-, bone marrow-, placenta-, and Wharton’s jelly-derived MSCs showed source-dependent clustering, with perinatal tissue-derived MSCs, including placenta- and Wharton’s jelly-derived MSCs, displaying more diverse protein profiles than adult tissue-derived MSCs, including adipose tissue- and bone marrow-derived MSCs [190]. These findings indicate that the translational value of BMSCs should be interpreted in relation to their characterized biological profile and disease-specific evidence, rather than as a general advantage in expansion capacity or secretome diversity.

Source-dependent differences also extend to MSC-derived EVs. Proteomic comparisons of EVs from bone marrow-, adipose tissue-, and umbilical cord-derived MSCs have shown that their cargo profiles and predicted functional annotations are not identical. BMSC-derived EVs were predicted to be more closely associated with regeneration-related functions, adipose tissue-derived MSC-EVs with immune regulation, and umbilical cord-derived MSC-EVs with tissue repair-related functions [197]. These findings support the need to distinguish BMSC-derived EVs from broader MSC-EV effects, because EVs derived from different parental MSC sources should not be assumed to be functionally interchangeable.

BMSCs remain a rational focus for ischemic stroke research mainly because of the depth and continuity of their stroke-specific evidence. They were among the earliest MSC sources evaluated in patients with ischemic stroke, and early autologous intravenous studies and follow-up investigations provided foundational evidence for clinical feasibility, short-term safety, and preliminary functional signals [10, 198]. In experimental stroke models, BMSCs have also been repeatedly used to investigate post-ischemic inflammation, neurovascular injury, angiogenesis, BBB disruption, and microenvironmental regulation. This accumulated disease-specific evidence makes BMSCs a well-characterized reference source for evaluating MSC-based therapeutic strategies in ischemic stroke.

Overall, this review focuses on BMSCs because they have a longer stroke-specific research history, established phenotypic characterization, accumulated early clinical experience, and a coherent connection between parental cell studies and BMSC-derived EV investigations. Umbilical cord-associated and adipose tissue-derived MSCs remain important alternative sources, particularly for scalable allogeneic manufacturing and future product development. Standardized head-to-head comparisons among MSCs and MSC-derived EVs from different sources are still needed to clarify source-specific differences in therapeutic potency, safety, manufacturing feasibility, and clinical applicability under ischemic stroke-relevant conditions.

Problems and challenges

Although BMSC-based therapy has shown biological plausibility, reproducible biological activity in experimental ischemic stroke models, and encouraging safety and feasibility signals in early clinical studies, its clinical development remains by several unresolved translational challenges. These challenges are no longer limited to whether BMSCs can exert neuroprotective or reparative effects. Instead, they concern how BMSC products are manufactured, how they are delivered, how effective target-tissue exposure is measured, how product potency is defined, and how emerging combination or delivery-enhancement strategies should be advanced toward clinical testing. The major barriers can be summarized in four interrelated areas: dose, route, timing, and brain exposure; Good Manufacturing Practice (GMP)-compatible manufacturing and product standardization; long-term safety monitoring; and the clinical implementation of combination strategies and advanced delivery systems.

Dose uncertainty, route-dependent exposure, and timing dependence

Dose selection remains unresolved in BMSC-based therapy for ischemic stroke. Reported doses vary widely across experimental and clinical studies, and the administered cell number cannot be interpreted independently of product type, viable cell dose, delivery route, stroke phase, lesion characteristics, and actual biological exposure [108, 199, 200]. Clinical studies illustrate this heterogeneity. Early autologous intravenous studies used culture-expanded autologous BMSCs during the subacute stage, including protocols in which approximately 100 million cells were infused several weeks after stroke onset or administered as repeated intravenous infusions after ex vivo expansion [10, 201]. In contrast, the modified allogeneic bone marrow-derived cell product SB623 was delivered by stereotactic implantation in chronic ischemic stroke, with dose levels such as 2.5, 5, or 10 million cells in early-phase studies and different fixed-dose groups in subsequently sham-controlled studies [18, 185]. These examples show that dose and timing are closely shaped by manufacturing logistics, delivery route, and disease stage, rather than representing universally applicable therapeutic parameters.

Route-dependent biodistribution further complicates dose interpretation. Intravenous infusion is technically simple and minimally invasive, but a substantial proportion of infused cells may be retained in the lungs and later redistributed to the liver, spleen, and other organs enriched in the mononuclear phagocyte system, thereby limiting direct cerebral exposure [202–204]. This does not exclude therapeutic benefit, because systemic immunomodulation and paracrine signaling may still influence post-ischemic inflammation and repair. However, it makes it difficult to determine how much of the administered product reaches, persists in, or functionally affects the ischemic brain. Intra-arterial delivery may increase the opportunity for local cerebral exposure, but target-site exposure depends on vascular patency, lesion location, cell size, suspension quality, dose, and infusion rate [107, 108, 205]. Stereotactic intracerebral or peri-infarct implantation can achieve higher local exposure and may be more compatible with chronic-stage studies, but it is invasive and less suitable for acute stroke scenarios [107, 206, 207].

Timing should also be interpreted in relation to the intended mechanism of action. The post-ischemic brain is a dynamic microenvironment in which BBB permeability, inflammatory activation, oxidative stress, edema, glial responses, and tissue remodeling change over time. Acute or early subacute administration may be more relevant to anti-inflammatory, cytoprotective, immunomodulatory, and BBB-stabilizing effects. By contrast, delayed or focal implantation in chronic stroke may be more closely related to local plasticity, trophic support, and repair-associated signaling [208–210]. Therefore, timing should not be defined by a single post-stroke interval. It should be matched to disease stage, route-specific exposure, product availability, and mechanism-specific therapeutic objectives.

Another important issue is that sustained engraftment within the brain is not required for all observed therapeutic effects. Functional improvement or infarct reduction may occur even when transplanted cells show limited long-term survival in brain tissue [211–213]. This supports the view that transient paracrine signaling, systemic immune modulation, and microenvironmental regulation contribute substantially to BMSC-mediated effects. Future studies should therefore report not only the administered cell dose but also the viable cell dose, pre-administration viability, infusion rate, route-specific biodistribution, retention time, imaging evidence of exposure when feasible, product potency, and pharmacodynamic biomarkers related to inflammation, vascular integrity, angiogenesis, or tissue repair.

Manufacturing, GMP-compatible quality control, and regulatory barriers

BMSCs are living products, and their biological activity is sensitive to donor characteristics, tissue source, isolation procedures, culture conditions, passage number, expansion system, cryopreservation, and release criteria. Therefore, inconsistent clinical outcomes cannot be explained solely by administered dose or delivery route [214]. A key translational challenge is to determine whether variability in outcomes reflects inadequate delivery, insufficient product potency, disease heterogeneity, or limitations in trial design.

Autologous and allogeneic BMSC products require different manufacturing and regulatory control strategies. Autologous BMSCs have the theoretical advantage of immune compatibility and a lower risk of alloimmune rejection, which partly explains their use in early exploratory stroke trials [215, 216]. However, autologous products require patient-specific bone marrow harvesting, individualized ex vivo expansion, quality testing, and release evaluation. This process increases product variability, limits batch comparability, and may delay administration, thereby reducing compatibility with acute or early subacute therapeutic windows. In addition, autologous BMSCs obtained from stroke patients may be influenced by age, vascular risk factors, diabetes, hypertension, systemic inflammation, and other comorbidities, potentially affecting proliferative capacity, senescence status, immunomodulatory activity, and secretome composition [217, 218].

Allogeneic BMSCs may offer advantages in donor selection, scalable manufacturing, pre-release testing, cryopreserved availability, and batch preparation before patient enrollment [14, 216]. However, allogeneic products introduce distinct quality-control and regulatory challenges, including donor-to-donor variability, immunogenicity assessment, repeat dosing feasibility, cryopreservation-induced changes, post-thaw viability, batch comparability, batch-to-batch consistency, and regulatory release requirements. Therefore, autologous and allogeneic BMSCs should not be treated as interchangeable products. Product consistency and clinical interpretability depend on donor screening, manufacturing control, storage conditions, post-thaw viability, release criteria, and potency assays aligned with the intended mechanism of action.

Culture and expansion conditions further contribute to functional heterogeneity. Passage number, seeding density, oxygen tension, culture duration, inflammatory preconditioning, serum-containing versus xeno-free media, and 2D versus 3D expansion may influence cell phenotype, viability, cell size, trophic factor secretion, extracellular vesicle release, and immunomodulatory potency [219, 220]. Although 3D culture or bioreactor-based expansion may improve scalability and preserve certain functional properties, these systems require comparability testing, process control, and reproducibility assessment under GMP conditions.

Current release criteria usually include sterility, endotoxin, mycoplasma testing, viability, cell count, and basic MSC identity markers [219]. MSC identity can be supported by cluster of differentiation 73 (CD73), cluster of differentiation 90 (CD90), and cluster of differentiation 105 (CD105) positivity and absence of hematopoietic or immune lineage markers such as cluster of differentiation 34 (CD34), cluster of differentiation 45 (CD45), cluster of differentiation 14 (CD14) or cluster of differentiation 11b (CD11b), cluster of differentiation 19 (CD19) or cluster of differentiation 79a (CD79a), and human leukocyte antigen-DR (HLA-DR) [14]. However, these identity markers do not directly measure disease-relevant therapeutic potency [221]. For ischemic stroke, potency assays should be aligned with the intended mechanism of action, including suppression of inflammatory cytokine responses, secretion of trophic and angiogenic factors, protection of endothelial tight junctions, regulation of microglia/macrophage activation, or extracellular vesicle-mediated signaling. Future trials should connect product quality attributes with in vivo exposure, imaging biomarkers, inflammatory and vascular biomarkers, rehabilitation exposure, and functional outcomes.

For BMSC-derived EV products, standardization should follow contemporary EV reporting principles. Parental BMSCs should be clearly defined, including donor source, passage number, culture medium, conditioning protocol, viability, and MSC identity markers. EV preparations should be reported according to isolation method, morphology, particle size distribution, protein markers, purity, cargo reproducibility, dose unit, storage stability, and functional potency. Morphology should be assessed using transmission electron microscopy, cryo-electron microscopy, or atomic force microscopy to confirm membrane-bound vesicular structures. Conventional transmission electron microscopy may show round or cup-shaped vesicles, but the cup-shaped appearance can partly reflect dehydration or staining artifacts; therefore, morphology alone should not be used to define exosomes [136].

Particle size distribution should be quantified using methods such as nanoparticle tracking analysis, tunable resistive pulse sensing, dynamic light scattering, or single-particle flow cytometry when appropriate. Small extracellular vesicle preparations are commonly described operationally as particles below approximately 200 nm, whereas larger EVs are generally above this range. However, size alone cannot prove endosomal biogenesis. Therefore, unless the cited study demonstrates exosomal origin, terms such as “BMSC-derived extracellular vesicles” or “BMSC-derived small extracellular vesicles” are preferable to “BMSC-derived exosomes” [222, 223].

Protein characterization should include EV-associated transmembrane or lipid-bound proteins such as cluster of differentiation 9 (CD9), cluster of differentiation 63 (CD63), and cluster of differentiation 81 (CD81), as well as cytosolic EV-associated proteins such as tumor susceptibility gene 101‌ (TSG101), apoptosis-linked gene 2-interacting protein X (ALIX), or heat shock protein 70 (HSP70). Source-associated markers such as cluster of differentiation 73 (CD73), cluster of differentiation 90 (CD90), cluster of differentiation 105 (CD105), cluster of differentiation 44 (CD44), or cluster of differentiation 29 (CD29) may help support BMSC origin, but they should not be treated as universal EV markers. Negative or contaminant markers should also be assessed, including intracellular organelle proteins such as calnexin or golgi matrix protein 130 (GM130) and, when serum-containing media or plasma-derived materials are used, lipoprotein or serum contaminants such as apolipoprotein A-I (ApoA-I), apolipoprotein B (ApoB), or albumin. Dose definition should not rely only on total protein concentration. Particle number, protein amount, particle-to-protein ratio, parental cell equivalent, and EV recovery after isolation should be reported [223]. Disease-relevant potency assays may include suppression of TNF-α, IL-1β, or IL-6 release in activated microglia/macrophages, preservation of endothelial tight junction proteins such as ZO-1, occludin, and claudin-5, reduction of neuronal apoptosis in OGD/R systems, promotion of endothelial tube formation or migration, and modulation of ischemia-related inflammatory or angiogenic biomarkers.

Long-term safety and tumor tropism

Long-term safety remains insufficiently characterized. Most clinical studies have had limited sample sizes and follow-up durations, which restricts the detection of rare or delayed adverse events. Potential safety concerns include delayed immune reactions, ectopic tissue formation, unwanted differentiation, vascular complications, inflammatory reactions, and risks associated with repeated dosing or invasive delivery methods. These concerns become more important when BMSCs are extensively expanded ex vivo, genetically modified, cryopreserved, combined with biomaterials, or administered together with EVs or other therapeutic agents.

Tumor tropism deserves specific attention. BMSCs can migrate toward inflammatory, hypoxic, and tissue-remodeling microenvironments, and similar chemokine gradients may also exist in neoplastic tissues [224, 225]. This property has been explored for tumor-targeted delivery in oncology, but it raises safety considerations when BMSCs are used for non-cancer indications such as ischemic stroke. Current evidence does not support a simple conclusion that properly prepared BMSCs are inherently tumorigenic, but MSC-tumor interactions appear to be context-dependent and may be influenced by cell source, culture expansion, genetic modification, host immune status, and the tumor microenvironment [226]. Future clinical protocols should exclude patients with active malignancy when appropriate, assess relevant cancer history, monitor unexpected proliferative lesions, and include long-term surveillance, especially for repeatedly administered, engineered, or extensively expanded BMSC-based products.

Combination strategies and advanced delivery systems

Combination strategies and advanced delivery systems are intended to address limitations of BMSC therapy alone, including insufficient homing, poor survival, variable paracrine potency, inadequate lesion exposure, and heterogeneity of the post-ischemic microenvironment. TCM-derived interventions, pharmacological agents, physical stimulation, biomaterials, hydrogels, nanoparticles, engineered EVs, and targeting systems may enhance BMSC survival, migration, paracrine activity, angiogenesis, BBB protection, local retention, or sustained cargo delivery. Their value should be judged by whether they address a defined translational bottleneck rather than simply adding another intervention to BMSC-cell therapy.

TCM-BMSC combination strategies have a distinctive theoretical basis because many TCM-derived bioactive compounds, extracts, and formulas exert multi-target regulatory effects. These interventions may improve the ischemic microenvironment, enhance SDF-1α/CXCR4-mediated homing, promote PI3K/AKT- or VEGF-related survival and angiogenesis, and support neurotrophic factor secretion. This creates a potential model in which TCM-derived interventions modulate the injured host microenvironment, whereas BMSCs provide paracrine, immunomodulatory, angiogenic, and repair-supportive effects. Such a strategy may be particularly relevant in the subacute or recovery stage, when inflammatory resolution, neurovascular remodeling, and rehabilitation-related functional recovery become central therapeutic goals.

However, the clinical translation of combination strategies involving TCM-derived interventions and BMSCs is more complex than the development of either approach alone. TCM formulas contain multiple active components, and their pharmacokinetics, quality markers, dose-response relationships, and interactions with BMSCs are often incompletely defined. It remains unclear whether observed benefits are mainly driven by TCM-mediated microenvironmental regulation, improved BMSC survival and homing, altered BMSC secretome, or additive effects of both interventions. Compatibility between herbal components and cell products should be evaluated, including effects on BMSC viability, phenotype, migration, paracrine activity, immunomodulatory potency, EV release, and safety. For compound formulas, chemical fingerprinting, active component identification, batch consistency, and quality-control standards are required before clinical translation can be considered.

A practical development pathway for TCM-derived interventions should begin with defined clinical scenarios and mechanism-linked product design. First, representative bioactive compounds or standardized formulas should be selected according to specific objectives, such as enhancing homing, improving survival in the ischemic microenvironment, promoting angiogenesis, or regulating inflammation. Second, herb-cell interaction assays should determine how the TCM intervention modifies BMSC viability, migration, secretory profile, EV release, or immunomodulatory function. Third, dose-response and timing-response studies should clarify whether TCM-derived interventions should be administered before BMSC administration, together with BMSC delivery, or during the recovery phase. Finally, efficacy should be evaluated using clinically relevant outcomes, including motor function, infarct evolution, vascular remodeling, inflammatory biomarkers, and rehabilitation-related recovery.

Advanced delivery systems provide another route to improve the feasibility of BMSC-based therapy. Biomaterial scaffolds, hydrogels, nanoparticles, surface-modified cells, engineered EVs, and targeting peptide-based systems may improve local retention, sustained release, lesion-directed delivery, or resistance to the hostile ischemic microenvironment. Biomaterials may be useful when spatial retention and local support are required, whereas engineered EVs or targeting systems may improve cargo delivery and reduce dependence on long-term cell engraftment. However, these products are more complex than unmodified BMSCs and require clear definition of product identity, release criteria, dose unit, degradation behavior, sterilization method, storage stability, immunogenicity, biodistribution, and regulatory classification.

The clinical feasibility of combination and delivery-enhancement strategies should be matched to stroke phase and therapeutic objective. For acute or early subacute stroke, rapidly available and minimally invasive products, such as allogeneic BMSCs or well-characterized BMSC-derived EVs, may be more practical than individualized autologous products. For subacute or recovery-stage stroke, TCM-derived interventions, rehabilitation-integrated protocols, or non-invasive physical stimulation may be compatible with repeated intervention and functional recovery. For chronic stroke, localized delivery or engineered products may be considered when the therapeutic goal is to enhance local plasticity or repair-associated signaling. In all cases, combination strategies should be advanced according to a clear mechanism, standardized product quality, route-specific feasibility, safety monitoring, and clinically meaningful functional endpoints.

Conclusion

BMSC-based therapy for ischemic stroke has evolved from an early emphasis on cell replacement to a broader framework centered on biological regulation, microenvironmental modulation, and repair-supportive signaling. Current evidence suggests that BMSCs act mainly through paracrine signaling, immunomodulation, angiogenesis, anti-inflammatory and anti-apoptotic effects, BBB and neurovascular protection, and regulation of the post-ischemic microenvironment. Although neuronal differentiation-like changes have occasionally been reported in experimental studies, durable neuronal differentiation, synaptic integration, and meaningful replacement of lost neurons in vivo remain controversial and are not generally considered dominant mechanisms underlying BMSC-mediated benefit.

Preclinical studies provide strong biological plausibility for BMSC-based interventions and have reported beneficial effects across multiple pathological processes involved in ischemic stroke. Early clinical studies have generated encouraging safety and feasibility signals. However, current clinical evidence remains heterogeneous and is largely derived from small-scale or early-phase studies; therefore, reproducible clinical efficacy has not yet been established.

Future progress will depend on adequately powered randomized and blinded clinical trials with appropriate controls, standardized BMSC products, potency assays linked to the intended mechanism of action, rigorous quality control procedures, biomarker-guided patient stratification, optimized delivery strategies, and long-term safety monitoring. Continued advances in GMP-compatible manufacturing, EV characterization, optimization of route and dose, and translational study design may improve the clinical applicability of BMSC-based approaches. At present, BMSC-based products, including BMSC-derived EVs, should be regarded as promising investigational therapies rather than clinically established treatments for ischemic stroke.

Abbreviations

BMSCs

Bone marrow-derived mesenchymal stem cells

Bcl-2

B-cell lymphoma 2

Bax

Bcl-2-associated X protein

Cyt-c

Cytochrome c

GCLC

Glutamate-cysteine ligase catalytic subunit

GCLM

Glutamate-cysteine ligase modifier subunit

GCL

Glutamate-cysteine ligase

GGCT

γ-glutamyl cyclotransferase

GSDMD

Gasdermin D

GSH

Glutathione

HO-1

Heme oxygenase-1

IκB

Inhibitor of kappa B

IL

Interleukin

JAK/STAT

Janus kinase/signal transducer and activator of transcription

JNK

c-Jun N-terminal kinase

c-Jun

Transcription factor jun

MDA

Malondialdehyde

MOMP

Mitochondrial outer membrane permeabilization

NF-κB

Nuclear factor-kappa B

p50

Nuclear translocation of NF-κB p50

p65

Nuclear translocation of NF-κB p65

NLRP3

Nod-like receptor family pyrin domain-containing 3

Nrf2

Nuclear factor erythroid 2-related factor 2

RIP1

Receptor-interacting protein kinase 1

RIP3

Receptor-interacting protein kinase 3

ROS

Reactive oxygen species

SIRT1

Sirtuin 1

SOD

Superoxide dismutase

CAT

Catalase

TGF-β

Transforming growth factor-β

Th17

T helper 17 cells

TLR4

Toll-like receptor 4

TNF-α

Tumor necrosis factor-α

TRAF6

TNF receptor-associated factor 6

Treg

Regulatory T cells

TSG-6

TNF-α-stimulated gene-6

C3

Complement component 3

Author contributions

W.J.Z.: wrote the first draft of the manuscript, revised it accordingly and drew the figures. H.Q.X.: revised, and checked the whole manuscript. W.J.Z., H.Q.X., F.W.G., Z.Q.Z.: prepared the graphs. W.L.: revised the manuscript, provided supervision and developed the major research programs. All authors reviewed the manuscript.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82174367). Special Expert Project of Taishan Scholars in Shandong Province (TSTP20250547).

Data availability

Data availability is not applicable to this article as no new data were created or analyzed in this study.

Declarations

Competing interests

The authors declare no competing interests.