Microglia actively and continuously monitor the cerebral microenvironment by extending and retracting their processes in adult brain (40). The main role of microglia in a healthy brain is a spectrum of surveilling functions, including normal phagocytosis and synaptic pruning. When stimulated by cell injury, microglia rapidly activate and shift to pro- or anti-inflammatory phenotype. Both spatial and temporal dimensions are very important in microglia dynamics because microglia present with pro- or anti-inflammatory phenotype depending on specific pathophysiological stages and brain regions after stroke (41). Pro-inflammatory microglia (M1-like) secrete detrimental cytokines and molecules such as interleukin (IL)-6, IL-1β, NO, TNF-α, IL-12, and IL-23 and express surface markers such as CD16 and CD32, which aggravate brain injury. However, anti-inflammatory microglia (M2-like) secrete anti-inflammatory cytokines such as IL-4, IL-10, transforming growth factor-β (TGF-β), IL-13, and growth factors such as vascular endothelial growth factor (VEGF) and brain-derived neurotrophic factor (BDNF) and express markers such as CD206,arginase1 (Arg1), TGF-β, insulin growth factor 1 (IGF-1), and IL-10, which promote neurological recovery (40, 42, 43). The M2-like microglia are further categorized as M2a, M2b, and M2c depending on their cellular function (5, 44, 45). M2a microglia, induced by IL-13 or IL-4, are involved in tissue regeneration and repair. M2a microglia also produce significant amounts of Arg1, Ym-1, CD206, and Fizz1 (5, 45). Microglia of M2b phenotype, activated by immune complexes such as Fcγ receptors, Toll-like receptors (TLRs), or IL-1R, exert immunoregulatory effects (5, 44, 46). The M2c phenotype microglia, deactivated by IL-10, TGF-β, and glucocorticoids, promote tissue regeneration when the inflammation reaction recedes (5, 47, 48). The function of microglia is also different between sexes. Most genes associated with inflammatory processes, including regulation of cell migration and cytokine production, are more expressed in male mice microglia compared with females (49). Male microglia are more prone to inflammatory reactions than female microglia due to transcriptionally activated NF-κB (49). However, female microglia have much more expression of genes related to cell plasticity, inflammatory response control, and repair than male microglia (49).

In a transient ischemic stroke model, activated Iba1⁺microglia appeared within 24 h and reached the peak within 4–7 days in the infarct core area. However, in the peri-infarct region, microglia emerged much earlier within ~3.5 h and peaked at 7 days (50). The temporal kinetics of microglia/macrophage polarization was described in a mouse model of middle cerebral artery occlusion (MCAO) (10). Microglia/macrophages initially attracted to the ischemic area exhibit mainly the M2 phenotype (CD206⁺Iba⁺ double-positive cells) only for a short period of time within 7 days post ischemia. The early recruitment of M2-like microglia/macrophages may act to clean ischemic tissue and restrict brain damage. M1-like microglia/macrophages (CD16/32⁺Iba⁺ double-positive cells) with decreased phagocytosis and enhanced secretion of pro-inflammatory mediators increasingly occurred from day 3 and remained high level until day 14 after ischemia. These cells may exacerbate neuronal demise and impair axon regrowth. Hence, the “helpful” M2-like microglia were first induced in response to ischemic injury during the early, acute phase of ischemic stroke, of which the function is protecting the neurons in the infarct area, phagocytizing cellular debris, and helping to restrict the area of damage. Then, the “harmful” M1-like microglia become more abundant in the next several days, and the neuronal damage caused by ischemia is increased. The shift of M2- to M1-like microglia in the process of chronic inflammation after stroke exacerbate the neuronal injury, leading to reduced neuronal recovery.

Microglia may respond differently between sexes under ischemic stroke. When subjected to pMCAO, male mice exhibited larger ischemic lesions than female mice (49). The progression of damage was significantly lower in male mice after transplantation of female microglia, indicative of the protective effects of female microglia (49). This phenomenon can be probably interpreted by the fact that Ym1 (a marker for microglia anti-inflammatory activation) immunoreactivity in Iba1-positive cells surrounding the ischemic lesion was higher in male mice transplanted with female microglia than in control (49). Hence, sex-specific microglia phenotype is intrinsically independent of the hormonal environment.

With the development of recent technology such as single-cell RNA sequencing and cytometry by time-of-flight mass spectrometry (CyTOF), spatial, temporal, and functional diversity of microglia during development, homeostasis, and disease in mice and humans has been unveiled (51). Varying degrees of pro- and anti-inflammatory markers may coexist on the same microglial cell. Hence, microglia may exhibit multiple phenotypical subtypes in vivo rather than two individual states based on the clustering of transcriptomic data. For example, 14 microglia sub-clusters were discovered in the cortex penumbra at the early stage of ischemic stroke without complete expression of M1 or M2-type marker genes (52). Five distinct microglial subtypes were identified in the mouse model of MCAO, which did not have a complete overlap with classic M1 or M2 subsets at the single-cell level (53). At least six transcriptionally distinct microglial subsets were uncovered in the stroke-aged brain (54). Further research on temporal and regional heterogeneity of activated microglia based on transcriptomic studies will be helpful in better understanding the pathophysiological mechanism of ischemic stroke.

MSCs, a population of non-hematopoietic cells, were identified for the first time in the bone marrow (55). Subsequently, MSCs have been successfully isolated and cultured from various tissues in mammals, including circulating blood, umbilical cord blood (UCB), menstrual blood, the placenta, the heart, the adipose tissue, the skeletal muscle, the pancreas, and the dental pulp (56, 57). MSC surface markers include CD105, CD73, and CD90, and more than 95% of the culture do not express CD34, CD45, and CD14 or CD11b, CD79a or CD19, and HLA-DR (58).

MSCs of different origins have been demonstrated to exert immunomodulatory effects on microglia. One study investigated the effects of adipose tissue-derived mesenchymal stem cells (AMSCs) on microglia morphology and involved intracellular molecules in vitro (59). The results showed that AMSCs were capable of inducing a round, flat microglia phenotype (inflammatory) into a ramifying, anti-inflammatory one. Ramified microglia induced by AMSC have decreased secretion of the pro-inflammatory cytokines such as TNF-a and IL-6, increased phagocytic activity, and the upregulation of neurotrophic factors and of the Arg-1 (a marker for M2-like regulatory microglia) (59). AMSC-secreted CSF-1 binds to its receptor, resulting in phosphorylation of PI3K and ERK1/2, followed by activation of small RhoGTPases, actin polymerization, and cell morphology change and migration (59). The ratio of ramified to hypertrophic microglia was significantly increased in the ischemic brain at 12 weeks when MSC-EVs were IV administered in monkeys 24 h and 14 days after ischemic injury (60). Moreover, microglia in the perilesional gray (PG) of MSC-EV-treated monkeys showed significantly greater branching complexity, process intersections, and length compared with control (60). Therefore, MSC-EV treatment could shift microglial morphology from hypertrophic to ramified microglia, which present with stronger surveillance capacity and homeostatic functions (60).

MSC transplantation is neuroprotective after stroke at least in part via modulating microglia-mediated neuroinflammation. MSC exhibited inhibition of the pro-inflammatory cytokine expression, such as TNF-α, IL-1β, IL-6, and IFNγ, pro-inflammatory enzyme expression, such as iNOS and COX-2, and chemokine expression, such as MCP1 and GRO-α, when co-cultured with LPS, IL-β, or IFNγ-primed HMO6 or BV2 microglia (61–63) (Table 1). However, MSC-produced paracrine factors may account for its therapeutic effect, considering that >99% of transplanted MSC has been entrapped in the lungs. A study on MSC secretome showed that EVs/exosome was the only component that could successfully reproduce most of the beneficial effects exerted by parent cells (71). In vitro, oxygen- and glucose-deprived (OGD) model can activate microglial-mediated inflammatory response and simulate the ischemic/reperfusion microenvironment in vivo. Several in vitro studies demonstrated that exosomes from MSC could promote M1 to M2 microglia shift in OGD conditions (Table 1). Notably, hypoxia preconditioning could increase the production of exosomes by bone marrow MSC (BMSC), whereas inhibiting the secretion of exosomes to a great extent eliminated its effects on M1 to M2 microglia shift in OGD (66). Therefore, exosomes play pivotal roles in the beneficial effects of MSC on microglia. As a marker of MSC-Exos, miRNAs have been considered the most important component and nearly can reproduce most of the exosomal effects on recipient cells (72). MicroRNA (miR) are small, endogenous, non-coding RNA molecules that can selectively hybridize to the 3′ untranslated region (UTR) poly(A) tail of targeted mRNAs, leading to the prevention of their transcription into proteins or promotion of their degradation (73). When co-cultured with primary microglia cells under OGD conditions, exosomes from adipose-derived stem cells (ADSCs) overexpressing miR-30d-5p have a greater effect in inhibiting inflammatory factor expression and promoting M2 microglial shift by inhibiting autophagy (64). MiR-126⁺ exosomes can also inhibit microglial activation and the expression of inflammatory factors such as TNF-α and IL-1β in the OGD model (65). Human umbilical cord mesenchymal stem cells (hUCMSC)- derived exosomal miR-26b-5p could inhibit M1 polarization of microglia via targeting CH25H to suppress the TLR pathway (68). Rat BMSC-derived exosomal miR-223-3p induced conversion of inflammatory M1 microglia toward M2 microglia via downregulating CysLT2R transcription and expression (67). hUCMSC-exosomal miR-145-5p was demonstrated to attenuate OGD-induced microglial pro-inflammatory activity, evidenced by lowered expression of IL-6, TNF-α, and IL-1β, via suppression of the IRAK1/TRAF6 signaling pathway (70). Compared with EVs derived from other MSC sources, EVs from the umbilical cord have the ability to exert stronger therapeutic immunomodulation and protective effects (74).

A pile of evidence demonstrated the immunomodulatory effects of MSC or MSC-EVs on microglia in cerebral ischemia model (Table 2). The most commonly used species include rat and mice, whereas monkey was used in one study (60). MSC or MSC-EVs were administrated mainly via IV route, and others include intra-arterial, intracerebral, and intranasal routes. When stem cells were infused in stroke models, most of the animal studies have chosen the acute (within 48 h) and subacute (within 7 days after ischemia) phases as the transplantation time. In one study, hUCMSC was administered IV 2 weeks before cerebral ischemia (87). The number of M1-like microglia (Iba⁺CD16/32⁺cells) decreased and M2-like microglia (Iba⁺CD206⁺cells) increased 7 days after photothrombotic stroke (87). These results indicated that MSC could also be developed as a prophylactic therapy for patients post-stroke besides therapeutic treatment. Most preclinical studies showed that the number of activated microglia was significantly decreased by MSC or EV administration (15, 61, 63–65, 75, 77–82, 84, 85, 88). Notably, three-dimensional (3D) spheroid cultured MSCs have been demonstrated to exhibit decreased cell size; therefore, these cells are not likely to be entrapped in lung tissue after IV infusion (39). Moreover, the increased expression of chemokine receptor CXCR4 after 3D culture enhanced the homing ability of the MSC to the ischemic brain area. The number of amoeboid microglia decreased significantly 5 days after MCAO. The mRNA expression of microglial markers Iba1 was reduced. 3D-cultured MSC downregulate pro-inflammatory mediators, including IL-12β, Hmgb1, Cxcl10, Ccl2, and IL-1β and increase expression of anti-inflammatory mediators STC1, HGF, and TSG-6 probably via decrease of Mincle expression in microglia cells (39). 3D-cultured MSC has been shown to have enhanced anti-inflammatory effects on microglia, which provide a novel and efficient method for the treatment of immune-mediated disorders. Exosomes derived from MSC of different origins have been demonstrated to exert enhancing effects on microglia polarization from the M1 to M2 phenotype. For example, ADSC-derived exosomes can reduce the number of M1-like microglia (Iba1⁺/iNOS⁺cells) and increase the number of M2-like microglia (Iba1⁺/CD206⁺cells) at 3 days after MCAO (64). hUCMSCs-exos can also attenuate I/R-induced M1 polarization (68, 70). The number of M1-like microglia (Iba1⁺/CD32⁺ cells) was statistically decreased in MCAO rats treated with BMSC-Exos, while the number of M2-like (Iba1⁺/CD206⁺cells) microglia was significantly increased (69). Hence, IV-administered BMSC-Exos can shift microglia toward neuroprotective M2 phenotype in the subacute phase of ischemia stroke (69).

However, there were some conflicting results. Compared with 7 days post-ischemia, administration of either allogeneic (rMSCs) or xenogeneic (hMSCs) stem cells at 1 day resulted in a significantly greater recovery of motor behavior after MCAO that was possibly related to an increase in activated microglia (CD11) in the infarct region (76). Transplantation of BMSC reduces the proportion of Iba-1⁺microglia cells expressing TGF-β in the cerebral infarction area 48 h after ischemia (88). IV allogenic BMSC significantly increased the inflammation, evidenced by an increase of TNFαand IL-1βand the number of Iba+ microglia/macrophages in the ischemic core cortex at day 2 after MCAO (83). Hence, MSC delivered at 2 days after MCAO can stimulate but not suppress the immune response (83, 88). It is reasonable that the immunosuppressive capacity of MSC requires a certain amount of time to be induced under inflammatory conditions. The inflammatory reaction will be induced by MSC without being fully activated, followed by exacerbated disease outcome (83). Therefore, the immunosuppressive and anti-inflammatory effects of MSC may be dependent on time. The neurotrophic effects of microglia, evidenced by an increased number of IGF-1⁺CD68⁺ and BDNF⁺Iba-1⁺ double-positive cells, are most likely responsible for the therapeutic effects of MSC in the acute phase (83). However, in another study, ADMSCs delivered IV at 2 or 7 days post-cerebral ischemia do not have a profound influence on the number or phenotype of Iba+ microglia in the perilesional cortex at 44 days after pMCAO (89). This may be the reason that subpopulations of microglia with different phenotypes were recruited to the perilesion and were involved in tissue repair in different activation stages (89, 91). There are some factors that probably account for these controversial effects of transplanted cells on the number and activation of microglia, including the route, timing, and dose of transplantation, follow-up time, stroke model, and techniques of staining and counting (89).

As mentioned above, most of the in vivo studies did not demonstrate the real causal effects of MSC or MSC-EVs on microglia in stroke. It was possible that morphological and phenotype transformation of microglia in vivo may be secondary. Moreover, neuroprotective function of MSC on ischemic stroke is comprehensive. The effect of MSC on microglia cannot completely account for the amelioration of functional outcome in ischemic stroke. This is because MSC can also promote macrophage polarization toward anti-inflammatory phenotype (92–94). Besides, infiltration of immune cells such as monocytes and neutrophils to inflammation sites can be prevented by MSC (93, 95). Selective elimination of microglia subtype may help better understand the role of MSC in stroke. Most in vivo studies just observed morphological and phenotype changes of microglia after transplantation of MSC or MSC-EVs in stroke model. Accompanied functional alteration has not been explored in detail. Considering various dynamic states of microglia subtype may be present, determination of microglia by M1 or M2 phenotype based on surface markers is oversimplified and inappropriate. The function of numerous microglia subtype in vivo will be future research focus. Indeed, most of the in vivo studies even cannot reliably distinguish microglia from infiltrating macrophages just by detection of biomarkers. The two kinds of immune cells may be distinguished based on CCR2 expression (96). Combined with the extent of CD45 and CD11b expression, the resident microglia can be characterized as CD11bhi/CD45lo/CCR2⁻, and infiltrating macrophage as CD11blo/CD45hi/CCR2⁺ (97). Further research is needed for a precise identification of microglia and macrophage in vivo. Considering that EVs contain proteins, lipids, and nucleic acids, molecular profiles of these EVs will be the focus to determine which components may exert beneficial roles on microglia in ischemic stroke. Besides, the underlying specific molecular mechanisms of MSC in modulating microglia in vivo need to be fully elucidated.

Transcription factors are proteins that can regulate the transcriptional activity of genes by binding to DNA. NF-κB translocates from the cytoplasm into the nucleus in microglia following activation after ischemic stroke. Then, activate microglia were transformed into the M1 phenotype, followed by the production of inflammatory cytokines, resulting in secondary brain injury after stroke (98, 99). Neuroinflammatory responses can be attenuated in ischemic stroke via decreasing hypoxia-induced factor-1α following inhibition of activated NF-κB (p65) upregulation (100). MAPKs are a family of serine/threonine protein kinases, including JNK, ERK, and p38, which also play essential roles in the activation and polarization of microglia (101). MSC-derived soluble factor TSG-6 decreases NF-κB activity and the activation of MAPK signaling in LPS-induced microglia, in which the levels of expression of neuroinflammatory cytokines are reduced (62). B10 transplantation (a human MSC line) specifically decreases the NF-κB protein level in macrophage/microglia in the rat MCAO model (79). It is likely that MSC act on NF-κB signaling by inhibiting TLR2 and CD40 expression at the receptor level, increasing the inhibitory cytokine expression such as IL-4 and IL-10 at the activation level and inhibiting the expression of NF-κB itself (79) (see Figure 1).

Members of the signal transducer and activator of transcription (STAT) family of transcription factors have been demonstrated to play critical roles in cellular proliferation, immunity, and inflammation (102, 103). STAT3 can regulate macrophage/microglia polarization to anti-inflammatory phenotype (104, 105). However, some studies indicated that STAT3 could induce M1 microglia polarization in both an MCAO-induced and a bilateral common carotid arteries stenosis (BcaS)-induced model of ischemic stroke (106, 107). Consistently, hMSC reduced microglial STAT3 gene expression and activation of Y705 phosphorylated STAT3 following transplantation in an ouabain-induced brain ischemia rat model, which resulted in decreased microglia activation (108). The effect of STAT3 on microglia is likely contradictory, and further studies are required.

CX3CL1/CX3CR1 signaling is reported to regulate the microglia activity. It has been demonstrated that CX3CL1/CX3CR1 inhibits LPS-induced microglial activation and decreases expression of inflammatory factors in microglia such as NO, IL-6, and TNF-α via activating the PI3k/Akt pathway, which effectively protect the neuron (109, 110). The infusion of exogenous CX3CL1 reduced cerebral infarct size, neurological deficits, and caspase-3 activation in an MCAO mouse model (111). CX3CL1 expressed by B10 cells (human mesenchymal lines) inhibited cytokine-induced pro-inflammatory gene expression, including iNOS in a human microglia cell line in vitro (61). In addition, when transplanted in rat MCAO model, B10 cell decreased the accumulation of Iba-1⁺ microglia and inhibited pro-inflammatory gene expression in the core and ischemic border zone (IBZ) (61). In our study, BMSC intraventricular injection significantly increased the number of M2-type microglia (Iba⁺/CD206⁺ cells) and decreased the number of M1-type microglia (Iba⁺/TNFα⁺ cells) in the infarct region of MCAO rats, especially at 3 days and 7 days post transplantation (90). The above effects can be reversed via interfering with CX3CL1 administration, indicating that BMSC transplantation could shift microglia from pro-inflammatory type to anti-inflammatory in the infarct region of MCAO rats through triggering the secretion of CX3CL1 (90).

CCL2/CCR2 also plays an important regulatory role in post-stroke inflammation. Overexpressed CCL2 has been shown to increase macrophage infiltration and cerebral infarction area, indicating that CCL2 can promote inflammatory response and impede recovery post-stroke (112, 113). In a study, hUCMSC-derived exosome overexpressing CCR2 promoted M2 microglia/macrophage polarization possibly via competitively binding CCL2 to inhibit the excessive activation and M1 polarization of microglia/macrophages (86).

CD200, a member of the immunoglobulin superfamily, plays a critical role as a membrane glycoprotein in the maintenance of immune homeostasis. IL-4-treated hMSCs significantly reduced microglia secretion of IL-6 and IL-1β when co-cultured with LPS-activated microglia through CD200 expression on hMSCs (108). The immunomodulatory effect of hMSC CD200 expression was effectively inhibited by anti-CD200 antibodies (108). In another study, CD200 was more highly expressed under hypoxic conditions in human placenta amniotic membrane-derived mesenchymal stem cells (AMSCs) (63). When CD200 was silenced, the inhibitory effect of AMSC on pro-inflammatory cytokine expression in co-cultures with LPS-primed BV2 was decreased (63). Moreover, AMSC transplantation significantly decreased microglia activation in the boundary region in MCAO through upregulation of CD200 expression (63).

The receptor proteins IRAK1 and TRAF6 are widely distributed in the cytoplasm and nucleus of various cell types. They are largely involved in Toll-like receptor (TLR)-initiated pathways, which contribute to the expression of pro-inflammatory mediators (70). Overexpression of IRAK1 and TRAF6 can activate NF-κB (114). hUMSC-Exos administration decreased IBA-1⁺CD16⁺ (M1 type) and increased IBA-1⁺CD206⁺ (M2 type) microglia cells at 72 h post-stroke in the MCAO model (70). The immunomodulatory effect on microglia polarization may be mediated by miR-146a-5p, which binds to 3′UTR of the mRNAs encoding IRAK1 and TRAF6 and downregulates the inflammatory response (70).

Cysteinyl leukotrienes (CysLTs) are potent inflammatory mediators derived from the 5-lipoxygenase metabolites of arachidonic acid after cell necrosis (67). The action of CysLTs is mainly mediated by receptors of CysLT1 and CysLT2 (CysLT1R and CysLT2R), which are activated in various types of cells after brain injury (115). During cerebral ischemia, CysLTs receptors (mainly CysLT1R and CysLT2R) are upregulated in activated microglia (116). In a study, BMSC exosome overexpressing MiR-223-3p promoted M2 microglia transformation of M1 microglia when co-cultured with BV2 cells exposed to OGD in vitro through inhibiting CysLT2R (67). Concomitantly, pro-inflammatory factors such as IL-6 and IL-1β decreased and anti-inflammatory factors such as IL-10 increased (67).

Toll-like receptors (TLR) are signaling receptors in the innate immune system that can promote microglial activation and polarization (113, 117). Microglial activation was induced by TLR4 recognition of LPS on the cell surface after brain injury, followed by the production of pro-inflammatory factors via inhibiting the NF-κB pathway (118, 119). Targeted downregulation of TLR4 expression after ischemia-reperfusion injury can inhibit microglial activation and M1-type polarization, and reduce the release of pro-inflammatory factors (117). hUCMSCs-exos attenuated I/R-induced M1 microglia polarization and inflammatory response. The underlying mechanism maybe that exosomal miR-26b-5p could inhibit CH25H to inactivate the TLR pathway, resulting in the repression of M1 polarization of microglia (68).

MicroRNAs (miRNAs) can regulate gene expression post-transcriptionally by combining with the 3'-UTR, coding sequence, or 5'UTR of target genes (120). They play key roles in the remodeling process under stroke conditions (121). MSC-derived exosomes are capable of transferring miRNAs to recipient cells and subsequently contribute to microglia polarization after stroke. Several studies elucidated the role of specific miRNAs in microglia polarization. ADSC-derived exosomes mediated the delivery of miR-30d-5p to the microglia and reversed OGD-induced and autophagy-mediated microglial polarization to M1 (64). The in vivo study also demonstrated that ADSC exosomal miR-30d-5p promoted M2 microglia/macrophage in the MCAO model by suppressing autophagy (64). IV administration of miR-126⁺ exosomes from ADSC post-stroke also inhibited microglial activation and the expression of inflammatory factors in vivo (65). hUCMSCs exosomal miR-26b-5p could repress M1 polarization of microglia by targeting CH25H to inactivate the TLR pathway after cerebral I/R (68). MiR-223-3p from BMSC exosome could promote M1 transformation to M2 microglia when co-cultured with BV2 cells exposed to OGD in vitro (67). As mentioned earlier, miR-146a-5p from hUMSC-Exos could promote the microglial shift from M1 to M2 after administration in the MCAO model via downregulation of IRAK1 and TRAF6 (70).

The rate of transcription of genes can be influenced by histone post-translational modifications (PTMs) at gene promoters and distal regulatory elements. Histone PTMs can exert negative and positive effects on transcription, including methylation (122). H3K4me1 increased dramatically in mice photothrombotic stroke models receiving LPS, in which CD16/32-positive microglia elevated and CD206-positive microglia decreased surrounding parainfarct region (87). When hUCMSC was administrated, microglial H3K4me1 modification was robustly alleviated, accompanied by a microglial shift from M1 toward M2 phenotype (87). Therefore, inhibiting H3K4 methylation maybe another putative mechanism for MSC to regulate microglia polarization.