Learn more: PMC Disclaimer | PMC Copyright Notice
Stem/progenitor cells from bone marrow decrease neuronal death in global ischemia by modulation of inflammatory/immune responses
Associated Data
Abstract
Human mesenchymal stromal cells (hMSCs) were injected into the hippocampus of adult mice 1 day after transient global ischemia. The hMSCs both improved neurologic function and markedly decreased neuronal cell death of the hippocampus. Microarray assays indicated that ischemia up-regulated 586 mouse genes. The hMSCs persisted for <7 days, but they down-regulated >10% of the ischemia-induced genes, most of which were involved in inflammatory and immune responses. The hMSCs also up-regulated three mouse genes, including the neuroprotective gene Ym1 that is expressed by activated microglia/macrophages. In addition, the transcriptomes of the hMSC changed with up-regulation of 170 human genes and down-regulation of 54 human genes. Protein assays of the hippocampus demonstrated increased expression in microglia/macrophages of Ym1, the cell survival factor insulin-like growth factor 1, galectin-3, cytokines reflective of a type 2 T cell immune bias, and the major histocompatibility complex II. The observed beneficial effects of hMSCs were largely explained by their modulation of inflammatory and immune responses, apparently by alternative activation of microglia and/or macrophages.
Observations in rodent and primate models suggest that a potential therapy for ischemia of the central nervous system is the administration of the adult stem/progenitor cells from bone marrow referred to as mesenchymal stem cells or multipotent mesenchymal stromal cells (MSCs) (1–3). Administration of MSCs also produced beneficial effects in animal models for neurodegenerative diseases, such as Parkinson's disease, experimental autoimmune encephalomyelitis, and amyotrophic lateral sclerosis (2–5). MSCs initially attracted interest for their ability to differentiate into multiple cellular phenotypes in culture and in vivo (1–7). However, recent observations indicate that only small numbers of the cells engraft into most injured tissues, and they disappear quickly (2–5, 8–10). When human MSCs (hMSCs) were injected into the dentate gyrus (DG) of the hippocampus in adult immunodeficient (ID) mice, most of the cells disappeared within 1 week, but they enhanced proliferation, migration, and neural differentiation of the endogenous neural stem cells (8). These and related observations have focused attention on the paracrine effects of MSCs (2, 3, 11). However, it has not been established whether the beneficial effects of MSCs in ischemic models of brain injury are explained by enhanced neurogenesis (8) or by neuroprotection.
Experiments here were performed in a mouse model of global ischemia to assess the neuroprotective effects of hMSCs. Administration of hMSCs 1 day after transient common carotid artery occlusion (tCCAO) improved neurologic function and decreased the delayed neuronal cell death of the hippocampus. Surveys with microarrays indicated that the hMSCs decreased expression of many of the mouse genes that were induced by ischemia and that were involved in inflammatory and immune responses. In addition, the transcriptome of the hMSCs changed in response to the ischemic environment. The results were confirmed by assays for immune-related cytokines (12–17).
Results
Intracranial Implantation of hMSCs.
In initial experiments, 105 hMSCs were implanted into each DG of either immunocompetent (IC) mice (C57/BL6) or ID mice (C57/BL6/scid), and the brains were assayed with quantitative real-time PCR for the highly repetitive human Alu sequences (hAlu). Within 5 min after the injection, 52% ± 13% (mean ± SE) of the injected hAlu were recovered in the hippocampus, 16% ± 15% in the cerebral cortex, 9.2% ± 8.1% in the striatum, and 1.6% ± 0.7% in the cerebellum (Fig. 1A). The recovery of human DNA stayed constant or increased slightly 1 day after implantation and then decreased rapidly over the next 3–7 days (Fig. 1B). The recovery of hAlu in the ID mice was slightly but not significantly greater than in the IC mice. Similar experiments were performed in IC mice in which global ischemia was produced by tCCAO 1 day before administration of hMSCs [supporting information (SI) Fig. S1] Again the recovery of human DNA increased slightly for 1 day and then decreased over the next several days (Fig. 1B). The low engraftment of the hMSCs was confirmed by immunohistochemistry with anti-human nuclei antigen (HuNu) (Fig. 1 C and D).
Effects of tCCAO with and without injection of hMSCs into DG of hippocampus of mice. (A) Within 5 min after cell injection [hMSCs (+)] into the DG of IC mice without tCCAO the cells were found in the hippocampus (Hip), neocortex (Ctx), striatum (Str), and cerebellum (Ce) (n = 6 hippocampi from three mice). (B) Assays by real-time PCR for hAlu sequences after injection of hMSCs into ID or IC mice with or without previous tCCAO (n = 6 hippocampi). The survival of hMSCs in ID mice was not statistically different from that in IC mice. (C) Typical labeling for human cells (anti-HuNu) in the hippocampus 4 days after tCCAO (arrows). (D) Higher-magnification images of C indicate the HuNu label in nuclei (DAPI; blue). (E) Neurologic deficit (ND) scores were evaluated by open-field behavior test on day 1 (pretreatment) and on days 2, 3, and 4 after tCCAO. (F) Typical images of FJB staining for degenerating neurons in the hippocampus CA1 region (arrows). (G) Quantitation of FJB-stained cells (n = 6 mice; P < 0.05) indicated decreased death of neurons in mice treated with hMSCs 4 days after tCCAO.
Injection of hMSCs Suppressed Neurologic Deficits and Neuronal Death After tCCAO.
Previous reports demonstrated that the tCCAO model caused delayed neuronal cell death in the hippocampal CA1 region in that low levels of neuronal death were detected on day 1 after tCCAO and the levels increased on days 2–4 (18, 19). Neurologic deficit (ND) scores by open-field behavior indicated that administration of the hMSCs produced a significant decrease in the ND scores 4 days after tCCAO (Fig. 1E). The tCCAO mice that received hMSCs continued to improve in that the ND score was 2.21 ± 0.29 SE (n = 6) on day 8. In addition, there was a decrease in the number of degenerative neurons detected in the hippocampal CA1 region (228 ± 86 cells/mm2 vs. 1112 ± 362 cells/mm2 in the Hanks' balanced salt solution [HBSS] control, P < 0.05) (Fig. 1 F and G). There were no differences in the total area assayed for Fluoro-Jade B (FJB) staining, mortality, and body weight (data not shown) between the hMSC- and HBSS-injected groups.
Effects of Ischemia on the Mouse Transcriptome.
To survey changes in the mouse transcriptome in response to tCCAO, mRNA was isolated from the hippocampi 2 days after tCCAO and assayed with mouse microarrays. Comparison of data from mice subjected to tCCAO (ischemia/HBSS) with data from control mice (uninjured/HBSS) indicated that the ischemia up-regulated 586 mouse genes (463 nonredundant) twofold or more (Fig. 2 A and B). Analysis of gene ontologies (GOs) indicated that many of the genes were in relevant categories, such as cytokine activity, immune system process, and response to ischemic stress (Table S1). The microarray data also indicated that 41 mouse genes (all nonredundant) were down-regulated in response to ischemia (Fig. 2 A and C and Table S2).
Microarray assays of hippocampi 2 days after tCCAO and 1 day after injection of HBSS or hMSCs. (A) Effects of ischemia on the mouse transcriptome. Signal intensities (SI) were compared between samples from uninjured/HBSS and ischemia/HBSS mice. (B) Venn diagram of 586 mouse genes up-regulated by ischemia, eight genes further up-regulated by injection of hMSCs, and 80 genes down-regulated by the hMSCs. (C) Venn diagram of 41 mouse genes down-regulated by ischemia and one of the down-regulated genes up-regulated by injection of hMSCs. (D) Effects of hMSCs on the transcriptome of mice after tCCAO. The hMSCs produced up-regulation of only three mouse genes (Lyz, lysozyme; Ym1, also known as Chi313; and Tgbi, TGFβ induced protein). The hMSCs produced down-regulation of 80 mouse genes. The most common GO terms are indicated. (E) Venn diagram of 189 human genes (170 nonredundant) up-regulated in hMSCs by the ischemic environment. (F) Venn diagram of 57 human genes (54 non-cross-hybridizing) down-regulated by the ischemic environment. The genes were filtered for cross-hybridization as indicated in SI Methods. (G) Effects of the ischemic microenvironment on the hMSC transcriptome.
Effects of hMSCs on the Mouse Transcriptome.
To survey the effects of administration of hMSCs, a comparison was made of signal intensities on the mouse microarray between samples from mice with tCCAO that received hMSCs and those that received vehicle (ischemia/MSC and ischemia/HBSS mice in Fig. 2D). Of the 586 total mouse genes up-regulated by ischemia (Fig. 2 A and B), hMSC treatment further up-regulated only three nonredundant genes (total of eight) (Fig. 2 B and D). Two of the three nonredundant genes were neuroprotective (Tables S3 and S4): the gene for the antibacterial enzyme lysozyme (Lyz) and the gene for chitinase 3-like 3 (Chi3l3; also known as Ym1) that is characteristically expressed by alternatively activated macrophages, by resting microglia, and at higher levels by alternatively activated microglia (17). The third gene, TGFβ-induced protein (Tgfbi), may also protect against tissue injury because it was induced after a cerebral stab wound in the rat (20).
Administration of hMSCs had more extensive effects on down-regulation of the ischemia-up-regulated mouse genes in that they down-regulated 80 mouse genes (65 nonredundant) that were up-regulated by ischemia (Fig. 2 B and D). Of the 65 nonredundant genes, 19 were involved in immune response, 21 in response to stimuli, 4 in defense response, and 3 in cell death and apoptosis as classified by GO (Fig. 2D and Tables S5 and S6). Of special interest was that 12 IFN-responsive genes were down-regulated (Table S6). Administration of hMSCs up-regulated only 1 of 41 mouse genes that were down-regulated by ischemia (Fig. 2C), a hypothetical gene of unknown function.
Effects of Ischemia on the hMSCs Transcriptome.
Because hMSCs were used in the experiments, it was possible to search for changes in the transcriptome of the human cells by assaying the same samples of hippocampal RNA on human microarrays. To correct for cross-hybridization with mouse mRNAs, the data from the human microarrays were filtered with data from three controls: uninjured/MSC, uninjured/HBSS, and ischemia/HBSS (see SI Methods). After filtering, comparison of the signal intensities on the human microarray of ischemia/MSC and uninjured/MSC indicated that 189 genes (170 nonredundant) in the hMSCs were up-regulated by ischemia (Fig. 2 E and G). An unbiased examination of the GOs indicated that the 170 nonredundant up-regulated genes were in 27 major ontologies (P < 0.01) (Table S7). A more selective examination of the data indicated that 45 of the 170 up-regulated genes were ischemia-related genes assigned to 15 GO terms (Table S8), with most assigned to the categories of (i) immune system process/immune response and antigen processing and presentation/MHC receptor activity (14 genes); (ii) inflammatory response/IκB kinase/NFκB cascade (4 genes); and (iii) cell growth/growth factor/binding TGFβ receptor signaling pathway (14 genes) (Fig. 2G). Therefore, the results indicated that the ischemia or cross-talk with ischemic mouse cells produced major changes in the hMSC transcriptome.
The filtered data indicated that 57 human genes (54 nonredundant) were down-regulated twofold or greater (Fig. 2 F and G). Unbiased analysis of the data indicated that 37 of the 54 nonredundant genes were in 15 of the GO categories (P < 0.01) (Table S9). Human genes that were down-regulated (Fig. 2G) included three immune response genes: cathepsin E (CTSE), platelet factor 4 variant 1 (PF4V1), and transcription factor 12 (TCF12). Again, the results indicated that the ischemic microenvironment or cross-talk with the ischemic mouse cells altered the hMSCs transcriptome.
hMSCs Modified Expression of Mouse Cytokines in Ischemic Hippocampus.
Because the microarray surveys suggested that the hMSCs down-regulated expression of many inflammatory and immune response genes, attempts were made to verify the data with immunoassays for cytokines previously associated with pro-inflammatory and anti-inflammatory responses. Mouse-specific ELISAs demonstrated that the administration of hMSCs after tCCAO significantly increased hippocampal levels of insulin-like growth factor 1 (IGF-1) (Fig. 3A), a cytokine that is coexpressed with anti-inflammatory cytokines and that is neuroprotective after ischemia (14, 15). There were also small but not statistically significant changes in IL-4, TNFα, and IFNγ (Fig. S2 a–d). However, the immune balance significantly shifted to anti-inflammatory (P < 0.05) when expressed as either the ratio of IL-4/IFNγ or IL-4/TNFα to reflect Th2/Th1 bias (Fig. 3 B and C). Immunohistochemistry of brain sections indicated that the IGF-1 colocalized with CD11b, suggesting that IGF-1 was expressed in microglia and/or macrophages (Fig. 3D).
Injection of hMSCs into hippocampus after tCCAO modified the immune response to a Th2 immune bias. Mouse-specific ELISAs were performed (n = 4–6 mice from each group). (A) Increase in the level of IGF-1 in the hippocampus (n = 5–6; P < 0.05). (B and C) Changes in the IL-4/IFNγ and IL-4/TNFα ratios in the hippocampus indicated a Th2/Th1 response (P < 0.05). (D) IGF-1 (red) coexpressed (arrows) with the microglial and macrophages marker (CD11b, green) in the hippocampus 4 days after tCCAO and 3 days after the injection of hMSCs.
Immunoassays of serum for IGF-1, IL-4, TNFα, and IFNγ did not detect any differences between hMSC-treated and nontreated mice (data not shown). Therefore, the data suggested that the hMSCs modified local immune bias in the ischemic mouse, apparently by their local rather than systemic effects on microglia and/or macrophages.
hMSCs Activated Microglia and/or Macrophages.
To test further the hypothesis that the effects of hMSCs were mediated by microglia and/or macrophages, mouse-specific ELISAs were performed for galectin-3 (Gal-3, also known as Mac-2), a galactosidase-binding lectin that is characteristically secreted by microglia and macrophages (15). Administration of hMSCs after tCCAO increased the Gal-3 levels in both the hippocampus and the cerebrospinal fluid (CSF) (Fig. 4 A and B). Gal-3 (+) cells colocalized with microglial and macrophage markers (F4/80 and CD11b in Fig. 4C) but not with markers for astrocytes (GFAP), neurons (MAP2), endothelial cells (CD31) of microvessels, or oligodendrocytes (CNPase) (Fig. S3). Double staining for MSCs with anti-HuNu demonstrated that Gal-3 (+) cells were clustered around hMSCs (Fig. 4D) but were not colocalized in the same cells (Fig. 4E).
Injection of hMSCs increased the levels of Gal-3 and numbers of Gal-3 (+) microglia/macrophages 4 days after tCCAO. (A) Mouse-specific Gal-3 ELISA indicated increased levels in hippocampus of the ischemia/hMSCs mice (n = 5 to 6; **, P < 0.01). (B) Increased levels of Gal-3 in the CSF (n = 8–10; *, P < 0.05). (C) Colabeling (arrows) in the hippocampus of Gal-3 (+) cells (green) with markers for microglia/macrophages (F4/80 and CD11b, red). (D) Gal-3 (+) cells (green) in close association with hMSCs (HuNu, red). (E) Higher magnification to demonstrate that Gal-3 (+) cells (arrowheads) were not colabeled with HuNu (arrows).
hMSCs Induced Microglia and/or Macrophages to Express Phenotype of Antigen-Presenting Cells and Ym1.
The hippocampus was also assayed for cells expressing major histocompatibility complex class II (MHC II), a marker of antigen-presenting cells (APC). After tCCAO and administration of hMSCs, numerous cells in the hippocampus were immunoreactive for MHC II (Fig. 5A). The MHC II (+) cells were clustered in regions containing HuNu (+) cells, but the two labels were not colocalized (Fig. 5A). The expression of MHC II in microglia and/or macrophages was confirmed by colabeling with Gal-3 (Fig. 5B). Comparisons of hippocampi from ischemia/HBSS and ischemia/MSC mice demonstrated that administration of hMSCs increased the number of MHC II (+)/Gal-3 (+) cells (Fig. 5B). As expected, the same cells expressing MHC II and Gal-3 also expressed IGF-1 (Fig. 5C).
Injection of hMSCs after tCCAO increased APC-like features of Gal-3 (+) cells and caused alternative activation of the microglia and/or macrophages in hippocampus. (A) Cells labeled with the APC marker MHC II (green) and with hMSCs (HuNu; red) were seen near the site of injection (Inj.), but the cells were not colabeled. The images were from immediately beneath the DG (above the double arrows). (B) Injection of tCCAO mice with hMSCs increased cells colabeled with MHC II (green) and Gal-3 (red). (C) In tCCAO mice injected with hMSCs, some Gal-3 (+) cells (purple) were colabeled (arrows) with both IGF-1 (green) and MHC II (red). (D) Some MHC II (+) cells (red) near the injection site (Inj.) colabeled (arrows) with Ym1 (green). (E) Injection of hMSCs into tCCAO mice increased number of cells labeled with Ym1 (green) (arrows). (F) Western blots confirmed that injection of hMSCs increased levels of Ym1 in the hippocampi 4 days after tCCAO. The pretreatment control was taken 1 day after tCCAO but before injection of hMSCs. (G) Densitometry of the Western blots in F.
In addition, the hippocampi were also assayed for expression of Ym1 (also known as Chi3l3), one of the neuroprotective mouse genes up-regulated by administration of hMSCs in microarray analysis (Fig. 2D). As expected, Ym1 was coexpressed in the same hippocampal cells that expressed MHC II (Fig. 5D). Comparisons of hippocampi from ischemia/HBSS and ischemia/MSC mice indicated that the number of Ym1 (+) cells was greater after treatment with hMSCs (Fig. 5E). The observations were confirmed by Western blots. The levels of Ym1 were increased 2 days after tCCAO in both control and hMSC-treated mice. However, 4 days after tCCAO, the levels were significantly higher in the hMSC-treated mice (Fig. 5 F and G).
Discussion
When hMSCs were injected into the DG of adult mice 1 day after transient global ischemia, assays for hAlu sequences and by immunohistochemistry demonstrated that the hMSCs disappeared over the next 7 days. However, the hMSCs both improved neurologic function and markedly decreased the neuronal cell death that begins 1 day after tCCAO and increases for 2–4 days thereafter (18, 19). Surveys with microarrays indicated that ischemia induced up-regulation of hundreds of mouse genes, and administration of hMSCs 1 day after the ischemia down-regulated more than 10% of the ischemia-induced genes. The effects of the hMSCs were accompanied by changes in their own transcriptome, suggesting that they were activated in response to the ischemic environment or to cross-talk with ischemic mouse cells, as was seen in other systems (21, 22). Most of the mouse genes down-regulated by the hMSCs were involved in inflammatory and immune responses. Therefore, the data suggested that the beneficial effects of hMSCs were probably explained by their modulation of inflammatory and immune reactions to ischemia. The suggestion was supported by protein assays that demonstrated increased expression by microglia/macrophages in the hippocampus of the neuroprotective factor Ym1, the cell survival cytokine IGF-1, cytokines reflecting a Th2 bias, Gal-3 that is characteristically expressed in activated macrophages and microglia, and MHC II. Therefore, all of the data supported the conclusion that a major effect of hMSCs was to modulate inflammatory and immune reactions to ischemia, at least in part by alternatively activating microglia and/or macrophages, cell phenotypes that were not distinguishable under the conditions used here.
The results are consistent with previous reports that injection of MSCs into rodents protected against neuronal injury and reduced the neuronal death after permanent or transient focal ischemia (2–5, 23). The conclusion that MSCs produce anti-immune or immune-suppressive responses is consistent with numerous reports that the cells have similar effects in culture, in animal disease models, and in patients with graft-vs.-host disease (24–27). The anti-inflammatory effects are consistent with anti-inflammatory effects of MSCs in two mouse models for lung injury (28, 29). The observations presented here are also consistent with recent demonstrations that mammalian responses to tissue injury frequently invoke excessive inflammatory and immune reactions that exacerbate the injury and that require specific cellular responses to modulate them (30).
In terms of therapeutic applications, administration of MSCs has the advantage over currently used anti-inflammatory and anti-immune agents in that the cells exert most of their effects locally at the site of injury and apparently specifically respond to the nature of the injury. The disadvantages include the difficulty of delivering the cells to some tissues (9, 10) and the possibility that their modulation of the normal responses to some injuries may exacerbate the effects. In addition, MSCs home to and enhance growth of some cancers (31, 32).
The observations suggesting that the effects of hMSCs were mediated through microglia and/or macrophages are consistent with indications that both microglia and macrophages can be alternatively activated to play contrasting roles in response to injury (33–35). The increases in Gal-3 and Ym1 observed here provided a direct link between the effects of the hMSCs and activation of microglia or macrophages that may have been produced either by paracrines secreted by the hMSCs or by direct cell-to-cell contacts, as was observed during immunosuppression of T cells by MSCs (27).
The results presented here and previously by others suggest the following sequence of events, summarized in Fig. 6: (i) the hMSCs were activated by stress signals from the ischemic tissue or cross-talk with ischemic cells; (ii) the activated hMSCs up-regulated expression (Fig. 2G and Table S8) of MHC I and a series of anti-immune, anti-inflammatory, and antiapoptotic-related factors (LTBP2, TGFβ2 and TGFβ3, IGFBP3–5, and TNFAIP6); (iii) the activated hMSCs caused alternative activation of microglia and/or macrophages, either by secretion of paracrine factors or by cell–cell contacts; and (iv) the alternatively activated microglia and/or macrophages increased expression of MHC II, Gal-3, the cell survival factor IGF-1, and the neuroprotective protein Ym1, and they established a Th2 immune bias. The overall effect of hMSCs was to rescue neuronal cells by modulating both the inflammatory/immune responses and decreasing apoptosis.
Materials and Methods
More detailed methods are presented in SI Methods.
Preparation of hMSCs.
Frozen vials of extensively characterized preparations of hMSCs from normal healthy donors were obtained from the Tulane Center for the Preparation and Distribution of Adult Stem Cells under an institutional-review-board-approved protocol. The cells were expanded and suspended in HBSS before administration (8).
Transient Forebrain Ischemia.
Mice (C57/BL6 or C57/BL6/scid) were anesthetized by inhalation of 2.5% sevoflurane and subjected to tCCAO (18, 19). One day after tCCAO, hMSCs were injected into the DG. ND score was evaluated with the open-field behavior test.
hAlu Real-Time PCR Assays and Species-Specific Microarrays.
The mice were anesthetized and the hippocampus, cerebral cortex, striatum, and cerebellum isolated. Genomic DNA was extracted (n = 6), and real-time PCR was performed with 100 ng of target DNA and human Alu-specific primers. Two days after tCCAO and 1 day after the injection, mice were killed and RNA was extracted from the hippocampi. The 7 μg of total RNA was used for assay on either the mouse (MG-430 2.0) or human (HG-U133 Plus 2.0) microarrays (Affymetrix) and analyzed by the dChip program.
Histologic Examinations.
Four days after tCCAO, the brains were obtained and cryosectioned at 50 μm (bregma −0.9 to −2.8 mm). Every 10th serial section was stained with FJB. The FJB-positive neurons were counted in the entire hippocampal CA1 region.
Cytokine ELISAs and Immunohistochemistry.
Under anesthesia, CSF, serum, or hippocampus were collected and assayed by using mouse-specific ELISA kits. For immunohistochemistry, 50-μm sections of brain were immunostained with the free-floating method with the antibodies listed in Table S10.
Western Blot Analysis.
The supernatants from hippocampal homogenates were electrophoresed and transferred to PVDF membranes. The protein bands were detected by chemiluminescence.
Acknowledgments.
This work was supported in part by National Institutes of Health Grant P40 RR 17447 and grants from the W. M. Keck Foundation, the Amon Carter Foundation, and the Louisiana Gene Therapy Research Consortium.
Footnotes
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/cgi/content/full/0803670105/DCSupplemental.