Characterization of EVs from MSCs and of its transport to hippocampus
Negative staining of the isolated vesicles and visualization by transmission electron microscopy confirmed the presence of small concave-shaped extracellular vesicles (< 200 nm) in the samples (Fig. 2A). Mode size diameter of EVs was 126 ± 8 nm measured by nanoparticle tracking analysis in three replicates, with a concentration of 1.14 ± 0.09 × 1011 particles/mL. Representative size profile distribution is shown in Fig. 2B. Western blot analysis shows that EVs isolated from MSCs contain the EV markers Alix, Hsp70, Flotillin-2 and CD9, as well as TGFβ (Fig. 2C). These data confirm that the samples contain true EVs. We also evaluated the presence/absence of positive and negative EV markers in EVs and non-EV fractions (whole cell lysates, supernatant discarded after the last ultracentrifugation step (SP) and cell culture medium (CCM). Isolated EVs were enriched in EV markers such as Alix, Flotillin-2 and CD9, while they lack non-EV markers, such as calnexin, lamin or histones, present in cells (Fig. 2D). As expected, these markers were not detected in the discarded supernatant or the cell culture medium.
Injected EVs reached the hippocampus (Fig. 2E). Dil-labeled EVs (red signal) were detected mainly in microglia (Fig. 2E-I) and pyramidal layer neurons (Fig. 2E-II). The red fluorescence signal of Dil-labeled EVs co-localized with Alix (green fluorescence), a marker of EVs, confirming that it corresponds to injected labeled EVs (Fig. 2E-III). The immunofluorescence against Alix also stained in green some EVs that are not stained in red, indicating that it also stains endogenous EVs. We did not observe a clear co-localization with astrocytes (Fig. 2E-IV). These results are consistent with previous studies conducted by Li et al. [52] and Otero-Ortega et al. [63]. In the study conducted by Li et al. [52] fluorescent EVs were intravenously injected to mice and they observed that the EVs reached the brain and were taken up mainly by microglia (86.8%) and neurons (12.6%) and only in a small percentage (0.8%) by astrocytes. While the level of astrocytic EVs uptake was low, the observation of astrocytic activation was clear, so that the authors speculate that the astrocytic activation was a secondary even of the activated microglia, as reported by Liddelow et al. [53]. Otero-Ortega et al. [63] obtained similar results after intravenous injection of Dil-labeled MSC-EVs from adipose tissue, finding co-localization of the EVs with microglia and neurons in the brain.
In vivo administration of EVs from MSCs reverse microglial and astrocytic activation in hippocampus of hyperammonemic rats and normalizes TNFα and IL-1β content
Hyperammonemic rats show neuroinflammation, with activation of microglia and astrocytes in hippocampus. Activated microglial cells acquire an amoeboid shape and reduce their processes, thus presenting a reduction in their area. The area of microglial cells was reduced in hippocampus of hyperammonemic rats (290 ± 15 μm2 versus 428 ± 28 μm2 in control rats, p < 0.05) and the injection of MSC-EVs reversed this effect (436 ± 21 μm2 in comparison with 290 ± 15 μm2, p < 0.05) (Fig. 3A and E). The percentage of area stained with GFAP increased in hippocampus of hyperammonemic rats (157 ± 15% of control group, p < 0.01), reflecting an increase in astrocyte activation (Fig. 3B and F), which was also reversed by injection of MSC-EVs (113 ± 8% of control group, p < 0.05). The content of the pro-inflammatory cytokines TNFα and IL-1β were increased in neurons of the CA1 region of hippocampus of hyperammonemic rats (131 ± 5%, p < 0.01; and 115 ± 3%, p < 0.05, respectively), as shown in Fig. 3C, D, G and H. Injection of MSC-EVs normalized the amount of both cytokines (90 ± 5%, p < 0.001; and 92 ± 5%, p < 0.001, respectively).
None of these parameters was significantly altered in control rats injected with MSC-EVs compared to control rats injected with PBS (area of microglia: 492 ± 51 μm2; percentage of area stained with GFAP in comparison with control group: 101 ± 11%; TNFα content in CA1: 113 ± 7% and IL-1β content in CA1: 103 ± 1%).
In vivo injection of EVs induce a shift in hippocampus of hyperammonemic rats from a pro-inflammatory to an anti-inflammatory state
The content of pro-inflammatory cytokines IL-6 and IL-1β was increased (131 ± 10%, p < 0.05; and 126 ± 6%, p < 0.05, respectively) in hippocampi of hyperammonemic rats compared to control rats (Fig. 4A and B). The injection of MSC-EVs normalized the levels of both cytokines (98 ± 8%, p < 0.01; and 78 ± 7%, p < 0.001, respectively).
In contrast, the amount of anti-inflammatory cytokines IL-4 and IL-10 was reduced (73 ± 5%, p < 0.05; and 81 ± 5%, p < 0.05) in hyperammonemic rats (Fig. 4C and D) and the injection of MSC-EVs reversed this effect, normalizing IL-4 and IL-10 content (111 ± 9%, p < 0.01; and 102 ± 5%, p < 0.05, respectively). The content of arginase 1, a marker of anti-inflammatory microglia, was reduced (67 ± 6%, p < 0.05) in hippocampi of hyperammonemic rats (Fig. 4E) and was also normalized by injection of MSC-EVs (98 ± 10%, p < 0.05 compared to hyperammonemic rats).
None of these parameters was significantly altered in control rats injected with MSC-EVs compared to control rats injected with PBS (IL-6: 108 ± 10%; IL-1β: 105 ± 8%; IL-4: 98 ± 8%; IL-10: 99 ± 6%; Arg1: 88 ± 9%).
EVs from MSCs restore memory and learning in hyperammonemic rats
Hyperammonemic rats showed impaired cognitive function, with reduced discrimination ratio both in object location (0.07 ± 0.04 versus 0.22 ± 0.04 in control rats, p < 0.05) and object recognition memory (0.42 ± 0.03 versus 0.56 ± 0.04 in control rats, p < 0.05) tests (Fig. 5A and B). Injection of MSC-EVs to hyperammonemic rats reversed this impairment (0.23 ± 0.03 in comparison to 0.07 ± 0.04 in HA rats injected with PBS, p < 0.05; and 0.71 ± 0.03 in comparison to 0.42 ± 0.03 in HA rats injected with PBS, p < 0.0001, respectively). Hyperammonemic rats injected with EVs showed a discrimination ratio in the OLM similar to control rats and even better than control rats in ORM (Fig. 5A and B). Control rats injected with MSC-EVs showed discrimination ratios similar to control rats injected with PBS in both tests (0.26 ± 0.05 versus 0.22 ± 0.04 in the OLM and 0.65 ± 0.03 versus 0.56 ± 0.04 in the ORM).
Hyperammonemic rats also showed impaired short-term memory as measured by discrimination ratio in Y-maze (0.47 ± 0.02 in comparison to 0.72 ± 0.06 in control rats, p < 0.01), which was also reversed by injection of EVs (0.67 ± 0.03 in comparison to 0.47 ± 0.02 in HA rats injected with PBS, p < 0.05) (Fig. 5C). Injection of MSC-EVs to control rats does not induce significant differences in this parameter (0.71 ± 0.06 versus 0.72 ± 0.06 in control rats injected with PBS).
Learning, reference memory and working memory were assessed in the 8-arm radial maze. Learning index was significantly lower (7.6 ± 0.7 versus 11 ± 0.6, p < 0.01) in hyperammonemic than in control rats at day 4 of the test and was normalized by injection of EVs (10 ± 0.7, p < 0.05) (Fig. 5D). Hyperammonemic rats showed impaired reference memory, with increased (13 ± 0.6 versus 10 ± 0.7 in the control group, p < 0.05) reference memory errors at day 4 of the test (Fig. 5E and F) and total number of reference memory errors (56 ± 1 versus 49 ± 2 in control rats, p < 0.01) (Fig. 5G). Both parameters were normalized by injection of MSC-EVs (10 ± 0.7 in comparison to 13 ± 0.6, p < 0.05; and 51 ± 1 in comparison to 56 ± 1, p < 0.05). No significant differences were found in working memory errors among experimental groups (Fig. 5H and I), although a tendency towards an increased total number of working memory errors in hyperammonemic rats (24 ± 2 versus 19 ± 1 in control rats) and a certain reduction by the injection of MSC-EVs (22 ± 2 versus 24 ± 2 in HA rats injected with PBS) can be observed in Fig. 5I.
No significant effects were observed in the control rats injected with MSC-EVs in any of the afore-mentioned parameters (learning index at day 4: 9.4 ± 0.6; reference memory errors at day 4: 11 ± 0.7; total reference memory errors: 50 ± 1; and total working memory errors: 22 ± 3).
It should be noted that the beneficial effects of EVs on hyperammonemic rats is not due to reduction of hyperammonemia. Injection of EVs did not affect blood ammonia levels, which were similar in hyperammonemic rats injected (68 ± 6 μM) or not (73 ± 8 μM) with EVs. These levels were higher (p < 0.01) than in control rats injected (38 ± 5 μM) or not (36 ± 3 μM) with EVs.
The above results show that i.v. injection of EVs from MSCs reduces neuroinflammation in hippocampus and restores cognitive function in hyperammonemic rats. To advance in the understanding of the mechanisms involved in the beneficial effects of EVs from MSCs, we used an ex vivo system allowing to analyze in detail the mechanisms involved. Freshly isolated hippocampal slices from hyperammonemic rats were treated ex vivo with EVs from MSCs. We first assessed if this ex vivo system reproduces the effects on neuroinflammation found in vivo.
Ex vivo administration of EVs from MSCs reverses microglial and astrocytic activation in hippocampus of hyperammonemic rats and normalizes TNFα and IL-1β content
The area of microglial cells was reduced in hippocampal slices of hyperammonemic rats (174 ± 6 μm2 in comparison to 236 ± 12 μm2 in control slices, p < 0.01) and treatment with MSC-EVs reversed this effect (235 ± 10 μm2, p < 0.01) (Fig. 6A and C). The area stained with GFAP increased in hippocampal slices from hyperammonemic rats (128 ± 3%, p < 0.01), reflecting an astrocytes activation (Fig. 6B and D). This was reversed ex vivo by MSC-EVs (98 ± 7%, p < 0.01).
The content of TNFα and IL-1β was increased in neurons of the CA1 region of hippocampus of hyperammonemic rats (158 ± 11%, p < 0.01; and 125 ± 2%, p < 0.01, respectively). Treatment with MSC-EVs normalized the amount of TNFα and IL-1β (107 ± 8%, p < 0.05; and 95 ± 5%, p < 0.05, respectively) (Fig. 7A–D). Therefore, the ex vivo system reproduces the effects of EVs from MSCs on neuroinflammation found in vivo.
Ex vivo treatment with EVs induces a shift in hippocampus of hyperammonemic rats from a pro-inflammatory to an anti-inflammatory state
The content of IL-6, IL-1β and TNFα were increased in hippocampal slices of hyperammonemic rats compared to control rats (132 ± 5%, p < 0.0001; 145 ± 6%, p < 0.0001; and 129 ± 5%, p < 0.001, respectively) as analyzed by Western blot (Fig. 8A–C). Treatment with MSC-EVs normalized the levels of these pro-inflammatory cytokines: IL-6 (97 ± 3%, p < 0.0001), IL-1β (100 ± 5%, p < 0.0001) and TNFα (100 ± 4%, p < 0.01). Similar results were obtained when the levels of IL-1β and TNFα were analyzed by ELISA. Hyperammonemia increases (p < 0.01) the content of IL-1β to 244 ± 36 pg/mg protein compared to 126 ± 26 pg/mg protein in control rats. Treatment with MSC-EVs normalized IL-1β levels to 132 ± 29 pg/mg protein (Fig. 8G). Hyperammonemia increases (p < 0.01) the content of TNFα to 470 ± 38 pg/mg protein compared to 273 ± 28 pg/mg protein in control rats. Treatment with MSC-EVs normalized TNFα levels to 300 ± 25 pg/mg protein (Fig. 8H).
The contents of IL-4 and IL-10 and arginase 1 were reduced in hippocampal slices from hyperammonemic rats (75 ± 4%, p < 0.001; 67 ± 4%, p < 0.0001; and 66 ± 6%, p < 0.0001) (Fig. 8D–F) and addition of MSC-EVs normalized them (100 ± 2%, p < 0.0001; 101 ± 2%, p < 0.0001; and 100 ± 6%, p < 0.01, respectively) These results show that the ex vivo system reproduces the effects on microglia polarization found in vivo in hyperammonemic rats injected with EVs from MSCs and it is therefore adequate to analyze the underlying mechanisms.
TGFβ mediates the beneficial effects of MSC-EVs observed ex vivo
It has been proposed that MSCs modulates microglia activation via TGFβ secretion and also that EVs from MSCs contain TGFβ on their surface which mediates some beneficial effects of these EVs [62, 74, 86, 93, 94]. On the basis of these reports we hypothesized that the beneficial effects of EVs from MSCs on neuroinflammation in hyperammonemic rats would be mediated by TGFβ present in their membranes. To assess this possibility, we tested if the beneficial effects of EVs from MSCs in the ex vivo system are prevented by blocking TGFβ action by co-incubating with anti-TGFβ or by adding an antagonist of TGFβ receptor along with the EVs.
We also prepared MSCs lacking TGFβ (see “Materials and methods” section) and we assessed if the EVs from these MSCs lacking TGFβ loss their beneficial effects. Finally, we also assessed if the direct addition of recombinant TGFβ to the hippocampal slices from hyperammonemic rats reproduces the beneficial effects of EVs from MSCs.
As shown in Fig. 6, the capacity of EVs from MSCs to reverse microglia and astrocytes activation ex vivo was eliminated when the EVs were co-incubated with anti-TGFβ (area of microglia: 180 ± 17 μm2 versus 235 ± 10 μm2 in the HA + EVs group, p < 0.01; area stained with GFAP: 124 ± 8% versus 99 ± 7% in the HA + EVs group, p < 0.01) or when the EVs lacked TGFβ (area of microglia: 170 ± 9 μm2 versus 235 ± 10 μm2 in the HA + EVs group, p < 0.001; area stained with GFAP: 136 ± 6% versus 99 ± 7% in the HA + EVs group, p < 0.0001), thus supporting that TGFβ in the surface of the EVs is inducing these effects. This is further supported by the fact that addition of recombinant TGFβ to the hippocampal slices from hyperammonemic rats also reduced microglia (213 ± 10 μm2 versus 174 ± 6 μm2 in HA slices, p < 0.05) and astrocytes activation (103 ± 3% versus 128 ± 3% in HA slices, p < 0.05) similarly to MSC-EVs (Fig. 6). Slices from hyperammonemic rats incubated with anti-TGFβ showed microglia activation (180 ± 2 μm2 versus 236 ± 12 μm2 in control slices, p < 0.01) and astrocytes activation (121 ± 4% in comparison with control slices, p < 0.05) similar to slices of HA rats, indicating that the addition of anti-TGFβ to the slices did not have an effect on these parameters and that it was not responsible of the improvement observed in the slices treated with MSC-EVs previously incubated with anti-TGFβ (Fig. 6).
Similar results were obtained for inflammatory markers. Incubation of the MSC-EVs with anti-TGFβ prevented the reduction by EVs of IL-1β (161 ± 13% versus 107 ± 8% in the HA + EVs group, p < 0.05) and TNFα (123 ± 5% versus 95 ± 5% in the HA + EVs group, p < 0.01) in hippocampal neurons of CA1 region, as assessed by immunohistochemistry (Fig. 7). Incubation of slices from hyperammonemic rats with anti-TGFβ did not affect the levels of IL-1β (153 ± 11% of control, p < 0.05) or TNFα (122 ± 2% of control, p < 0.05), which were increased in comparison to slices from control rats.
TGFβ is also responsible for the shift from pro-inflammatory to anti-inflammatory induced by EVs from MSCs in hippocampus of hyperammonemic rats (Fig. 8). Co-incubation with anti-TGFβ prevented the reduction by EVs of the levels of pro-inflammatory cytokines IL-6 (116 ± 5%, p < 0.01) (Fig. 8A), IL-1β (130 ± 7%, p < 0.05) (Fig. 8B) and TNFα (134 ± 9%, p < 0.01) (Fig. 8C) as well as the increase of the anti-inflammatory IL-4 (85 ± 5%, p < 0.001) (Fig. 8D), IL-10 (84 ± 3%, p < 0.05) (Fig. 8E) and arginase 1 (70 ± 7%, p < 0.001) (Fig. 8F). Depletion of TGFβ from the MSCs also prevented the effects of MSC-EVs on these pro-inflammatory (IL-6: 112 ± 3%, p < 0.05; IL-1β: 127 ± 5%, p < 0.05; TNFα: 127 ± 7%, p < 0.05) and anti-inflammatory factors (IL-4: 74 ± 6%, p < 0.001; IL-10: 87 ± 1%, p < 0.05; Arginase 1: 70 ± 3%, p < 0.01) as assessed by Western blot (Fig. 8A-F). Similar results were obtained when IL-1β and TNFα were analyzed by ELISA. MSCs-EVs reduced the levels of IL-1β in hyperammonemic rats from 244 ± 36 to 132 ± 29 pg/mg protein; however, treatment with TGFβ-depleted EVs did not reduce IL-1β, maintaining it at 253 ± 22 pg/mg protein (Fig. 8G). EVs reduced the levels of TNFα in hyperammonemic rats from 470 ± 38 to 300 ± 25 pg/mg protein; however, treatment with TGFβ-depleted EVs maintained TNFα at 492 ± 58 pg/mg protein (Fig. 8H).
Moreover, addition of recombinant TGFβ was also able to induce the shift to the anti-inflammatory state, reducing IL-6 (115 ± 2%, p < 0.05), IL-1β (103 ± 5%, p < 0.001) and TNFα (100 ± 10%, p < 0.05) and increasing IL-4 (99 ± 2%, p < 0.05), IL-10 (81 ± 5%, p < 0.01) and arginase 1 (96 ± 2%, p < 0.05) (Fig. 8A–F). These data indicate that TGFβ in the surface of the EVs is responsible for the reduction of glial activation and neuroinflammation induced by EVs from MSCs.
Ex vivo administration of MSC-EVs reverses the alterations in membrane expression of AMPA and NMDA receptors in hippocampal slices from hyperammonemic rats
Hernandez-Rabaza et al. [40], Taoro-Gonzalez et al. [78, 79] and Balzano et al. [9] have shown that neuroinflammation induces alterations in the membrane expression of AMPA (GluA1 and GluA2) and NMDA (NR2B) receptor subunits in hippocampus, which are responsible for the impairment of spatial learning in hyperammonemic rats and that treatments that normalize membrane expression of these subunits restore cognitive function.
We therefore assessed using the cross-linker BS3 if addition of EVs from MSCs to hippocampal slices from hyperammonemic rats normalizes membrane expression of AMPA and NMDA receptors subunits. Hyperammonemia increased membrane expression of the NR2B subunit of NMDA receptors (151 ± 8%, p < 0.0001) (Fig. 9A) and of the GluA2 subunit of AMPA receptors (150 ± 10%, p < 0.001) (Fig. 9C) and reduced membrane expression of the GluA1 subunit of AMPA receptors (67 ± 5%, p < 0.0001) (Fig. 9B) in the hippocampal slices. Treatment with EVs from MSCs normalized the membrane expression of NR2B (96 ± 5%, p < 0.0001), GluA1 (100 ± 4%, p < 0.0001) and GluA2 (96 ± 7%, p < 0.001) subunits (Fig. 9). This normalization of membrane expression of AMPA and NMDA receptor subunits would mediate the restoration of cognitive function. The normalization of membrane expression of NR2B, GluA1, and GluA2 did not occur in the presence of anti-TGFβ (136 ± 15%, p < 0.01; 71 ± 6, p < 0.01; and 138 ± 8%, p < 0.05, respectively) or when EVs lacking TGFβ were used (140 ± 8%, p < 0.05; 70 ± 9%, p < 0.05; and 136 ± 4%, p < 0.05, respectively). Conversely, the normalization induced by MSC-EVs was mimicked by addition of recombinant TGFβ (NR2B: 104 ± 7%, p < 0.01); GluA1: 108 ± 6%, p < 0.0001; and GluA2: 97 ± 5%, p < 0.01) (Fig. 9).
This indicates that TGFβ in the surface of MSCs-EVs is responsible for the normalization of membrane expression of NMDA and AMPA receptor subunits, which in turn would be responsible for restoration of learning and memory in hyperammonemic rats.
Incubation with MSC-EVs reduces NF-κB activation in hippocampal slices from hyperammonemic rats through the TGFβ–TGFβR2–Smad7–IkBα pathway
To further advance in the understanding of the mechanisms by which EVs from MSCs reduce neuroinflammation in hippocampus of hyperammonemic rats, we assessed whether nuclear translocation of NF-κB is increased in hippocampal slices from hyperammonemic rats and if this is reversed by MSC-EVs. Dadsetan et al. [22] reported that in rats with porta-cava shunts, another model of MHE, the increased levels of IL-1β and TNFα in hippocampus are a consequence of increased activation and nuclear translocation of NF-κB.
The nuclear content of the p50 subunit of NF-κB was increased in hippocampal neurons of CA1 region in slices from hyperammonemic rats (126 ± 6% of control, p < 0.01) (Fig. 10A and B) and the number of microglial cells expressing NF-κB was also increased (31 ± 2 cells/mm2 versus 15 ± 2 cells/mm2 in control slices, p < 0.0001) (Fig. 10E and F) and these increases were reversed by EVs from MSCs (nuclear/cytoplasmic content of p50: 93 ± 4%, p < 0.001; and microglia expressing p50: 14 ± 1 cells/mm2, p < 0.0001). The normalization of nuclear NF-κB did not occur if EVs from MSCs were added in the presence of anti-TGFβ (nuclear/cytoplasmic content of p50: 122 ± 4%, p < 0.01; and microglia expressing p50: 25 ± 1 cells/mm2, p < 0.01) or if the EVs were depleted of TGFβ (nuclear/cytoplasmic content of p50: 123 ± 3%, p < 0.01; and microglia expressing p50: 27 ± 1 cells/mm2, p < 0.001) (Fig. 10A, B). Incubation of the hippocampal slices from hyperammonemic rats with anti-TGFβ did not prevent the increase in p50 nuclear content (133 ± 8%, p < 0.001 compared to control slices) or in the microglia expressing p50 (29 ± 2 cells/mm2, p < 0.0001 compared to control slices). Moreover, addition of recombinant TGFβ reproduced the effects of EVs (nuclear/cytoplasmic content of p50: 104 ± 5%, p < 0.05; and microglia expressing p50: 19 ± 1 cells/mm2, p < 0.001), indicating that TGFβ in the EVs is responsible for this effect. Figure 10C shows axial projections of z-stack taken to confirm that p50 staining was localized in the nuclei.
To corroborate the effects on NF-κB activation, p65 NF-κB transcriptional activity was measured in nuclear extracts using a commercial kit. The results show that hyperammonemia increases p65 activity in nuclear extracts from hippocampal slices (124 ± 7% of control, p < 0.05) and treatment with MSC-EVs reverses this activation (102 ± 3%, p < 0.05). In contrast, MSC-EVs depleted of TGFβ did not reduce p65 activity (128 ± 7%, p < 0.05) (Fig. 10D).
We then tried to understand how TGFβ reduces NF-κB signaling. Noh et al. [62] reported that MSC-secreted TGFβ inhibits the NF-κB pathway in LPS-activated microglia by modulating Smad2/3 phosphorylation through the TGFβ1 receptor. We therefore tested if the Smad2/3 pathway could be mediating the effects of EVs TGFβ on NF-κB signaling in hippocampal slices of hyperammonemic rats. We did not find any change in the phosphorylation of Smad2 or Smad3 in hippocampal slices from hyperammonemic rats. Moreover, treatment with EVs from MSCs did not affect either Smad2 or Smad 3 phosphorylation (not shown). This indicates that the TGFβ–Smad2/3 pathway is not involved in the beneficial effects of EVs from MSCs.
It has been shown that TGFβ may also inhibit NF-κB signaling by inducing Smad7, which enhances the transcription of IkB, a key inhibitor of NF-κB signaling pathway. Smad7 may also disrupt the TRAF–TAK1–TAB2/3 complex, thus inhibiting NF-κB signaling [91]. We therefore assessed if the Smad7–IkB pathway could be mediating the effects of EVs TGFβ on NF-κB signaling in hippocampal slices of hyperammonemic rats. We found that hyperammonemia reduced Smad7 content in hippocampus (79 ± 5% of control, p < 0.05) (Fig. 11A) and this is associated with a parallel reduction of the IkB content (81 ± 2% of control, p < 0.001) (Fig. 11B). Moreover, hyperammonemia also increased the phosphorylation of IkB (140 ± 8% of control, p < 0.0001) (Fig. 11C). All these factors would contribute to enhanced nuclear translocation of NF-κB and activation of NF-κB signaling, including transcription of IL-1β and TNFα.
Treatment of the hippocampal slices from hyperammonemic rats with EVs from MSCs normalized the levels of Smad7 (99 ± 2%, p < 0.05) and IkB (100 ± 7%, p < 0.01), as well as the phosphorylation of IkB (100 ± 4%, p < 0.001), which returned to values similar to control rats (Fig. 11A–C). Normalization of these parameters did not occur if EVs were added in the presence of anti-TGFβ (Smad7: 73 ± 7%, p < 0.01; IkB: 84 ± 9%, p = 0.1; and phospho-IkB: 130 ± 7%, p < 0.05) or if TGFβ-depleted EVs were used (Smad7: 78 ± 5%, p < 0.05; IkB: 73 ± 8%, p < 0.01; and phospho-IkB: 160 ± 20%, p < 0.0001), while the addition of anti-TGFβ alone did not alter them (Smad7: 73 ± 7%, p < 0.01; IkB: 84 ± 9%, p = 0.1; and phospho-IkB: 130 ± 7%, p < 0.05). Moreover, the levels of Smad7 and IkB, and phosphorylation of IkB were also normalized if recombinant TGFβ was added to the hippocampal slices from hyperammonemic rats (Smad7: 76 ± 7%, p < 0.05; IkB: 76 ± 4%, p < 0.001; and phospho-IkB: 130 ± 7%, p < 0.0001) (Fig. 11A–C).
These data support the idea that TGFβ in the surface of EVs from MSCs reverses the enhanced NF-κB signaling in hippocampus of hyperammonemic rats by normalizing the levels of Smad7 and IkB.
We then assessed if the reduced levels of Smad7 and IkB in hippocampus of hyperammonemic rats could be due to reduced levels of TGFβ, or to reduced content or membrane expression of its receptors. The content of TGFβ was reduced (64 ± 6%, p < 0.0001) in hippocampal slices from hyperammonemic rats and was restored to normal levels by treatment with EVs from MSCs (105 ± 4%, p < 0.0001) or with recombinant TGFβ (104 ± 7%, p < 0.0001), but not by EVs co-incubated with anti-TGFβ (79 ± 2%, p < 0.01) or EVs lacking TGFβ (69 ± 4%, p < 0.0001) (Fig. 11D). Hyperammonemia also reduced the total content (82 ± 3%, p < 0.05) (Fig. 11E) and membrane expression (54 ± 6%, p < 0.0001) (Fig. 11F) of the TGFβ receptor 2, which were also normalized by treatment with EVs from MSCs (104 ± 6%, p < 0.01; and 119 ± 10%, p < 0.0001, respectively) or with recombinant TGFβ (107 ± 6%, p < 0.01; and 105 ± 2%, p < 0.0001, respectively) but not by EVs in the presence of anti-TGFβ (78 ± 7%, p < 0.01; and 73 ± 6%, p < 0.0001, respectively) or by EVs lacking TGFβ (83 ± 3%, p < 0.05; and 60 ± 7%, p < 0.0001, respectively) (Fig. 11E, F).
To confirm that the beneficial effects of EVs from MSCs are mediated by activation of TGFβ receptors by TGFβ present in the membrane surface of EVs, we assessed if these beneficial effects were prevented by blocking TGFβ receptors 1 and 2 with a selective antagonist. The results show that this is the case. Blocking TGFβ receptors also prevents the beneficial effects of EVs from MSCs on neuroinflammation, preventing the reduction of NF-κB activation in neurons (p50 nuclear content: 125 ± 2%, p < 0.01) (Fig. 10A, B) and microglia (microglia expressing p50: 26 ± 3 cells/mm2, p < 0.01) (Fig. 10E, F), the normalization of Smad7 (77 ± 3%, p < 0.05), IkB (75 ± 4%, p < 0.001) and p-IkB (150 ± 10%, p < 0.0001) (Fig. 11A–C), of TGFβ levels (79 ± 3%, p < 0.01) (Fig. 11D) and of TGFβ receptor 2 amount (83 ± 3%, p < 0.001) and membrane expression (66 ± 2%, p < 0.0001) (Fig. 11E, F). Blocking TGFβ receptors also prevents the reduction of microglial (area of the microglial cells 151 ± 4 μm2 versus 235 ± 10 μm2 in the slices treated with EVs, p < 0.05) and astrocytes activation (135 ± 4%, p < 0.001) (Fig. 6) and the shift from pro- to anti-inflammatory state induced by EVs, preventing the changes in IL-6 (114 ± 2%, p < 0.05), IL-1β (145 ± 14%, p < 0.001), TNFα (129 ± 4%, p < 0.05), IL-4 (75 ± 7%, p < 0.001), IL-10 (84 ± 2%, p < 0.01) and arginase 1 (79 ± 2%, p < 0.05) (Fig. 8).