Microglia have immune memory properties in vitro

Studies have reported that microglia acquired different inflammatory phenotypes upon repeated LPS stimulation in vitro [34, 39, 40]. To compare the different inflammatory responses following LPS preconditioning, we stimulated microglial BV2 cells with 10 ng/ml LPS for different durations and evaluated the transcriptional changes of inflammatory cytokines using RT-qPCR. The transcription levels of pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) peaked at 4 h and then decreased. In contrast, the levels of anti-inflammatory cytokines (Il-4 and Il-10) exhibited different dynamic changes. After exposure to a low-dose of LPS, Il-4 levels decreased, while Il-10 levels showed a moderate increase from 0 to 8 h and a sharp increase at 24 h (Fig. 1A). Based on these findings, we designed an experimental schedule to observe whether the inflammatory states of BV2 cells following a low-dose LPS pretreatment for different time (first stimulus) influenced their responses to a subsequent high-dose LPS challenge (second stimulus). There were four experimental groups: Control (Ctrl), Inflammation (Inflammation, IF), innate immune training (Training, TR) and innate immune tolerance (Tolerance, TL) (Fig. 1B). We found that preconditioning BV2 cells to the peak of pro-inflammatory state exacerbated the subsequent high-dose LPS response even with a resting interval of 6 h, as indicated by dramatically higher expression of Il-1β, Il-6 and Tnf-α compared to BV2 cells treated solely with LPS (IF group). However, BV2 cells preconditioned to an anti-inflammatory state somewhat maintained the anti-inflammatory properties, implied by increased transcriptional levels of Il-10 (Fig. 1C). These results suggested that BV2 microglial cells exhibited two forms of innate immune memory (IIM), of which the pro-inflammatory state induced-memory was termed innate immune training (TR) and the anti-inflammatory state induced-memory was termed innate immune tolerance (TL).

Fig. 1

The Expression of inflammatory cytokines in LPS-treated BV2 microglial cells and the impact of their conditioned medium on the survival of SH-SY5Y cells. A Transcriptional changes of pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) and anti-inflammatory cytokines (Il-4 and Il-10) in BV2 cells exposed to 10 ng/ml LPS for 0, 4, 6, 8 and 24 h. n = 3–4/group. B The experimental schedule diagram showing establishment of innate immune memory models in BV2 microglial cells. C Transcription levels of pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) and anti-inflammatory cytokines (Il-4 and Il-10) in the BV2 microglial cells detected by RT-qPCR. n = 4–5/group. Ctrl, Control; IF, Inflammation; TR, Training; TL, Tolerance. D Cell viability of SH-SY5Y cells. The experimental procedure for LPS-pretreatment and the resting time interval were the same as B, conditioned medium (CM) was collected from BV2 cells after a 24 h-incubation with 100 ng/mL LPS. Cell viability was determined 48 h after incubated with CM collected from BV2 cells. n = 4/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001 vs control groups; #p < 0.05, ##p < 0.01, ###p < 0.001. The difference shown by “+” was analyzed by unpaired t test. +p < 0.05

In addition, we assessed the potential impact of conditioned medium (CM) from BV2 cell models of IIM on the viability of dopaminergic SH-SY5Y cells. The findings revealed that the exposure to CM from IF group led to a decrease in cell survival. However, the viability of SH-SY5Y cells treated with CM from TL group was comparable to that of the control group. Interestingly, we observed higher cell viability after the treatment with TL CM compared to TR CM (Fig. 1D). These findings suggested that the CM from BV2 cell models of IIM exhibited diverse effects on the viability of SH-SY5Y cells; the IF group and TR group exerted detrimental effects and the TL group exhibited no harmful effect.

Moderate LPS-induced inflammatory stimulation in the periphery regulates brain innate immune response in vivo

Previous reports have demonstrated that microglia are main innate immune cells in brain and can retain innate immune memory (IIM) [41,42,43]. To explore the induction efficiency of innate immune memory of microglia in vivo, LPS extracted from three different kinds of Gram-negative bacteria was intraperitoneally injected daily to mice. LPS-1 (from Salmonella enterica serotype typhimurium), LPS-2 (from Escherichia coli O111:B4) and LPS-3 (from Escherichia coli O26:B6) were used in this study. First of all, we established mouse models of acute innate immune memory (AIIM) including Training (TR) and Tolerance (TL). The control group received 4 injections of normal saline (NS); The 2xLPS (1xLPS/1xLPS) group received LPS injections for two consecutive days, in which the first LPS injection serving as the first stimulus, and the second as the second stimulus; the 4xLPS group (3xLPS/1xLPS) received LPS injections for four consecutive days, in which the first three injections were regarded as the first stimulus and the fourth injection as the second stimulus (Fig. 2A). The changes in body weight during LPS injections were surveyed. The innate immune response of brain induced by the peripheral inflammatory stimuli was investigated by examining the transcriptional levels of pro- and anti-inflammatory cytokines in mouse striatum by RT–qPCR. We found that the immune response somewhat depended on the sources of LPS, despite the comparable trend in body weight losses (Fig. 2B, C). In the TR group, the levels of all the three pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) were significantly increased, whereas the expression of anti-inflammatory cytokine Il-4 was decreased. The expression of Il-10 transcript did not change compared to the control group. On the other hand, in the TL group, the transcriptional levels of pro-inflammatory cytokines remained low, and the expression of Il-4 did not change. However, the expression of Il-10 transcript was significantly elevated in this group (Fig. 2C). Among the three kinds of representative LPS, peripheral injection of LPS-1, but not LPS-2 and LPS-3, induced the most explicit AIIM in mouse brain. Moreover, the protein levels of IL-1β and IL-4 in mouse striatum were measured. We found that LPS-1 stimulation elicited a significant increase in IL-1β in TR group compared to the Control and TL groups. Conversely, there was a significant decrease in the level of IL-4 in TL group compared to the Control group, with no statistical difference observed between TR and TL groups (Fig. 2D). This means we confirmed the microglial IIM in vivo. By immunofluorescence staining of microglial marker IBA1 and cell counting, we found that the numbers of microglia were increased, and the area of soma, the numbers of endpoints, and the summed process length of microglia were significantly elevated as well, in the striatum of both TR and TL groups (Fig. 2E, F). Western blot results showed that the protein levels of IBA1 and GFAP in TL group were higher than those in the Control and TR groups, while TH protein levels remained unchanged (Fig. 2G). These results indicated that the patterns of brain IIM could be regulated by the systemic stimuli of different paradigms of LPS.

Fig. 2

The innate immune responses in the brain of mice systemically challenged with different low doses of LPS. A The experimental schedule diagram showing establishment of innate immune memory models in mice. LPS-1: Lipopolysaccharides from Salmonella enterica serotype typhimurium; LPS-2: Lipopolysaccharides from Escherichia coli O111:B4; LPS-3: Lipopolysaccharides from Escherichia coli O26:B6. B The changes in body weight during different paradigms of LPS stimulation. n = 3–10 /group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001. C Transcription of pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) and anti-inflammatory cytokines (Il-4 and Il-10) in mouse striatum detected by RT-qPCR. n = 3–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01. D Concentration of IL-1β and IL-4 proteins in the striatum of LPS-1 treated mice measured by ELISA. n = 3–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, vs control groups. #p < 0.05. E Representative images of IBA1 immunostaining in the striatum of mice injected with LPS-1. Scale bar, 25 µm. Microglial morphology was shown by zoomed representative images. Scale bar, 5 µm. F Quantification and morphology analysis of IBA1+ microglia in mouse striatum. n = 5–6 mice for each experimental group; N = 20–30 microglia for different groups. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests(quantification) or Kruskal–Wallis test (morphology analysis). **p < 0.01, ***p < 0.001, vs control groups. ###p < 0.001. G Protein levels of TH, GFAP and IBA1 in mouse striatum detected by Western Blot. Quantification of relative TH, GFAP and IBA1 protein expression is shown in the right panel. n = 6–7/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, ***p < 0.001 vs control groups. #p < 0.05, ###p < 0.001

HIF-1α plays a role in the regulation of brain innate immune memory

To assess whether the mouse striatum in TR and TL groups was characterized by a distinct gene expression profile, RNA-sequencing (RNA-seq) analysis of striatum from TR and TL mice collected at 3 h after LPS treatment was conducted. 137 genes were upregulated and 211 genes were downregulated in TL group compared to TR group (Fig. 3A). Gene ontology (GO) analysis showed the differentially expressed genes between the two groups were mainly concentrated in the cellular component, such as the cytoplasm and plasma membrane, and were involved in biological processes mainly related to defense response, innate immune response, inflammatory response, and cellular response to lipopolysaccharide (Fig. 3B, Additional file 1: Figure S1A). KEGG enrichment analysis showed that the enrichment pathways mainly focused on human diseases, environmental information processing and cellular processes including JAK-STAT signaling pathway (Additional file 1: Figure S1B). These results indicated that mouse brains acquired two types of IIM after the moderate LPS-induced inflammatory stimulation in the periphery. The findings from RNA-seq were further confirmed by RT-qPCR assays. On the whole, the transcription of M1-associated genes, including Il-1β, Tnf-α and CD86, was increased in TR group compared to Control group. However, the transcription of M2-associated genes (Tgf-β, Ym1/2) was increased both in TR group and TL groups, whereas CD206, Arg-1 and Il-13 transcripts were comparable among Control, TR and TL groups. The differences in the transcription of inflammatory genes including Hif-1α, Ifn-β, Hdac6, Myd88 and Nlrp3 between TR and TL groups further demonstrated that the signal pathways involved were distinct (Fig. 3C; Additional file 1: Figure S2). Several studies have reported that HIF-1α played a crucial role in immune inflammatory response [25, 28, 44]. By analyzing the protein level of HIF-1α in LPS-treated BV2 cells, we found that HIF-1α responded to inflammatory stimulation in microglia (Fig. 3D). Subsequent double-immunofluorescence staining for HIF-1α and IBA1 further confirmed that HIF-1α was expressed in activated microglia in mice (Fig. 3E). In view of the above results, CD16/32 was chosen as an indicator of pro-inflammatory phenotypic marker to explore the involvement of HIF-1α in microglial activation state. As depicted in Fig. 3F, all IBA1-immunopositive cells were classified into four clusters: HIF-1α− CD16/32− (white arrow, expressing neither HIF-1α nor CD16/32), HIF-1α+ CD16/32+ (yellow arrow, expressing both HIF-1α and CD16/32), HIF-1α+ CD16/32− (magenta arrow, expressing HIF-1α but not CD16/32), HIF-1α− CD16/32+ (red arrow, expressing CD16/32 but not HIF-1α).

Fig. 3

The expression of HIF-1α in microglia in innate immune memory models. A, B RNA-seq analysis of the striatum between mice treated with 2xLPS or 4xLPS. TR, Training; TL, tolerance. A The volcano plot of differentially expressed genes (DEGs). B The gene ontology (GO) enrichment analysis of DEGs. C The summary heatmap of RT-qPCR results based on M1/M2-associated genes and inflammatory genes in mouse striatum. n = 4–5/group. D Protein levels of HIF-1α in BV2 cells treated with LPS for 12 h, detected by Western Blot. n = 7–10/group. Differences were analyzed by unpaired t-test. **p < 0.01. E A representative 3D reconstruction image of activated microglia showing the colocalization of IBA1 (green) and HIF-1α (Magenta). Scale bar, 5 µm. F Co-localization of HIF-1α, CD16/32 and IBA1 in striatal microglia. Representative images from confocal microscopy revealed the colocalization of HIF-1α (Magenta) and CD16/32(red) in IBA1+ (green) microglia. White arrow: HIF-1α− CD16/32− cell; yellow arrow: HIF-1α+ CD16/32+ cell; magenta arrow: HIF-1α+ CD16/32− cell; red arrow: HIF-1α− CD16/32+ cell. Scale bar, 20 µm. G Representative images of IBA1/HIF-1α/CD16/32 triple immunostaining in mouse striatum. Scale bar, 50 µm. Small plots in the upper left quadrant showed the zoomed details. Scale bar, 20 µm. The experimental schedule was described as in Fig. 2A. (Ctrl, Control; TR, Training; TL, Tolerance). H Quantification of the numbers of HIF-1α+, CD16/32+ and HIF-1α+CD16/32+ cells among all the IBA1-immunolabeled cells. n = 3–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001, vs control groups. I The percentage of HIF-1α+, CD16/32+ and HIF-1α+ CD16/32+ cells within the total IBA1+ cells in mouse striatum. n = 3–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. **p < 0.01, ***p < 0.001, vs control groups. #p < 0.05. J The relative proportion of sub-populations of IBA1-immunolabeled cells in the acute innate immune memory mouse model. The difference of HIF-1α−CD16/32− microglia between groups was marked in the figure. n = 3–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. ***p < 0.001 vs control groups. K The relative proportion of IBA1-immunolabeled HIF-1α+ and HIF-1α− cells in acute innate immune memory model. The difference in HIF-1α− microglia between groups was marked in the figure. n = 4–5/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01 vs control groups. #p < 0.05. L Representative images of IBA1/HIF-1α/CD16/32 triple immunostaining in the striatum of mice treated with NS, 1xLPS, or 4xLPS for two or five weeks. Scale bar, 50 µm. Small plots in the upper left quadrant showed the zoomed details. Scale bar, 20 µm. M Quantification of the numbers of HIF-1α+, CD16/32+ and HIF-1α+CD16/32+ cells in all IBA1-immunolabeled cells. n = 3–4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001, vs control groups. #p < 0.05, ##p < 0.01. N The percentages of HIF-1α+, CD16/32+ and HIF-1α+CD16/32+ in the total IBA1+ cells in mouse striatum. n = 3–4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001, vs control groups. ##p < 0.01. O The relative proportions of HIF-1α− CD16/32− (white), HIF-1α+ CD16/32+ (yellow), HIF-1α+ CD16/32− (green), and HIF-1α− CD16/32+ (blue) in all IBA1-immunolabeled cells in acute innate immune memory mouse models. The difference of HIF-1α− CD16/32− microglia in 2w and 5w groups, compared to Control groups, was marked by asterisks (*) inside the figure, and difference of HIF-1α− CD16/32− microglia between 2 and 5w groups was indicated by pound sign (#). n = 3–4/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. **p < 0.01, ***p < 0.001, vs control groups. #p < 0.05

Firstly, the changes of microglial phenotype in AIIM mouse models, established according to the method in Fig. 2A, were examined. Triple immunofluorescence staining of IBA1, HIF-1α and CD16/32 in mouse striatum, followed by cell counting, was performed (Fig. 3G). The results showed that the numbers of each cell type, including HIF-1α+ IBA1+ cells, CD16/32+ IBA1+ cells and HIF-1α+ CD16/32+ cells, were markedly increased in both the TR and TL groups, although no difference was observed between the two groups (Fig. 3H). Notably, the percentage of HIF-1α+ microglia in TL group was decreased compared with TR group (Fig. 3I). The different proportion of four types of microglia (HIF-1α− CD16/32− microglia, HIF-1α+ CD16/32+ microglia, HIF-1α+ CD16/32− microglia and HIF-1α− CD16/32+ microglia) showed the change of microglial molecular signatures in LPS-injected mice (Fig. 3J). Given the sharp increase in the number of microglia in TL group compared to TR group (Fig. 2E and data not shown), we further investigated the relative percentage of IBA1-immunolabeled HIF-1α+ and HIF-1α− cells in AIIM model and found a significant differences in the proportion of HIF-1α− microglia between these groups (Fig. 3K). The results indicated that HIF-1α− microglia constituted the majority of the increased cell population in TL group.

To further investigate the persistence of the phenotype in microglia, we conducted triple immunofluorescence staining for IBA1, HIF-1α and CD16/32 in the striatum of mice received training- and tolerance-inducing stimuli for two or five weeks (Fig. 3L). The results revealed that the numbers of HIF-1α+, CD16/32+ and HIF-1α+ CD16/32+ microglia were higher in the training-induced group (1xLPS) compared than those in the Control group over time. Similarly, in the tolerance-induced group (4xLPS), numbers of these cell types were higher than those of the Control group at 2 weeks but decreased over time to reach similar levels as the Control group. Notably, 2 weeks post-LPS injection, there was a significant difference in the numbers of HIF-1α+ and HIF-1α+ CD16/32+ cells between the 1xLPS and 4xLPS groups (Fig. 3M). Furthermore, the dynamic trends in the proportions of HIF-1α+, CD16/32+ and HIF-1α+ CD16/32+ microglia were consistent with the observed quantity trends (Fig. 3N). The differences in the proportion of HIF-1α− CD16/32− microglia between different groups were remarkable (Fig. 3O). The relative proportions of each type of cell revealed that the phenotypic transformation of microglia persisted for weeks, even as the density of microglia returned to control levels.

Innate immune memory can last for one month and influence MPTP-induced immune response

The state of microglia in the brain may influence the pathology of CNS disorders [45, 46]. Studies have shown that the pathological process of AD can be regulated by stimuli applied months before [12]. To further investigate whether the sustained microglial IIM had an impact on PD-like pathology, we established the MPTP-induced mouse model of PD based on the IIM models. As indicated in the experimental schedule diagram, there was a one-month interval between the training- and tolerance-inducing stimuli (1x or 4xLPS, first stimulus) and the MPTP administration (second stimulus) (Fig. 4A). Because the interval between the first stimulus and the second stimulus in this model was one month, we called it chronic innate immune memory model (CIIM). Among them, mice in 1xLPS/MPTP group were PD mice with innate immune training (short for TR PD mice) and mice in 4xLPS/MPTP group were PD mice with innate immune tolerance (short for TL PD mice). The injection of LPS caused a momentary weight loss followed by a recovery before MPTP administration (Additional file 1: Figure S3A). In addition, the Pole test results showed that low dose(s) of LPS did not cause motor dysfunction (Additional file 1: Figure S3B). After MPTP injection, mice in 4xLPS/MPTP group exhibited the smallest weight changes compared with mice in 1xLPS/MPTP and NS/MPTP groups (Fig. 4B). MPTP administration alone triggered an inflammatory response [47,48,49,50]. Luminex assay was conducted to determine the protein levels of cytokines and chemokine in the striatum of mice at 3 days after MPTP treatment. Compared to the NS/NS group, the levels of IL-1β, IL-6, TNF-α and MCP-1 in the striatum of NS/MPTP group were significantly increased, but the levels of IFN-γ and IL-4 were not significantly different from those of NS/NS group. Compared with 1xLPS/NS group, the expression of IL-1β and IL-4 proteins in the striatum of 1xLPS/MPTP group increased, while the levels of IL-6, IFN-γ and MCP-1 proteins decreased. Only the level of IL-6 protein in the striatum of 4xLPS/MPTP group was higher than that of 4xLPS/NS group. The levels of IL-1β, TNF-α, IFN-γ, IL-4 and MCP-1 proteins in 4xLPS/MPTP group were similar to those in 4xLPS/NS group, while the expression of IL-1β, TNF-α and MCP-1 proteins was significantly lower than that of NS/MPTP group. In addition, there was no significant difference in IL-10 protein expression among all experimental groups (Fig. 4C, Additional file 1: Figure S3C). On the whole, the results suggested that the expression levels of inflammatory factors were altered dependent on the first stimulus one month ago, and compared to the NS/MPTP and 1xLPS/MPTP group, mice in 4xLPS/MPTP group exhibited the smallest changes in cytokine protein levels in the striatum. Furthermore, immunostaining for IBA1 in the striatum and SNpc of mice showed that MPTP-induced increase in the number of IBA1+ cells was reversed due to the IIM provoked by the first stimulus (Fig. 4D, E).

Fig. 4

MPTP-induced immune responses in mouse brain can be attenuated to a certain degree one month after intraperitoneal injections of low-dose(s) of LPS. A The experimental schedule diagram illustrating the establishment of MPTP-induced mouse model of Parkinson’s disease following the induction of IIM. B The changes in body weight during and after MPTP injections. n = 6–7 /group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. **p < 0.01, ***p < 0.001. C The summary heatmap of protein levels based on Luminex assay of pro-inflammatory cytokines (IL-1β, IL-6, TNF-α and IFN-γ), anti-inflammatory cytokines (IL-4 and IL-10) and chemokine (MCP-1) in the striatum of mice at 3 days post-saline or MPTP treatment. n = 4/group. D, E Assessments of microglial activation in the nigrostriatal pathway. Immunofluorescence staining of IBA1 (red) in the striatum (D) and immunofluorescence double staining of TH (green) and IBA1 (green) in the substantia nigra (E) at 7 days post-saline or MPTP administration. Scale bar, 100 μm. Quantification of IBA1+ cells in the dorsal striatum and the SNpc is shown in the bottom panel. n = 4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. **p < 0.01, ***p < 0.001

Altered MPTP toxicity in the nigrostriatal pathway of mice acquired innate immune memory

Given that mice exposed to MPTP exhibit movement disorders and damage to the nigrostriatal pathway [36, 49, 51], we investigated whether IMM had any effects on PD-like behaviors and pathology. The Pole test results showed that mice treated with MPTP alone took longer to turn around, climb down, and complete the entire process. Mice in 1xLPS/MPTP group also displayed some motor impairment, as evidenced by the increased time to climb down and total time spent. However, mice in 4xLPS/MPTP group showed no motor impairment compared to NS group, and they performed significantly faster than those in NS/MPTP and 1xLPS/MPTP groups. Similarly, the results of the Rearing test showed that there was a significant decrease in the number of rearing counts in NS/MPTP group compared to NS group. However, such a decline was reversed in 4xLPS/MPTP group (Fig. 5A). Western blotting and immunohistochemistry assays were conducted to assess the damages of dopaminergic system in nigrostriatal pathway. Treatment with MPTP in mice caused a decreased in striatal TH protein levels in all the groups regardless of the pre-stimulations. However, TH levels in both NS/MPTP and 1xLPS/MPTP groups were significantly reduced compared to 4xLPS/MPTP group (Fig. 5B). By assessing the optical density of TH+ nerve fibers in the striatum and TH+ neurons in the SNpc, we observed that MPTP treatment led to a depletion of striatal TH+ nerve fibers in mice of NS/MPTP and 1xLPS/MPTP groups. Additionally, MPTP treatment resulted in a reduction of TH+ neurons only in mice of NS/MPTP group. However, pre-stimulation with 4xLPS elicited resistance to MPTP-induced dopaminergic damage both in the striatum and SNpc, as indicated by the elevated TH+ nerve fibers and TH+ neurons in 4xLPS/MPTP group compared to 1xLPS/MPTP and NS/MPTP group, respectively. Moreover, mice in 1xLPS/MPTP group also showed no dramatical loss of TH+ neurons in the SNpc (Fig. 5C, D). These results revealed that 4xLPS alleviated the dopaminergic neuronal loss, while the effect of 1xLPS was limited. Since there exists bi-directional communication between microglia and astrocytes, the astrocyte reactivity was assessed by western blot and immunohistochemical staining. The levels of striatal GFAP proteins and the number of GFAP+ cells in the NS/MPTP and 1xLPS/MPTP groups were increase. Striatal level of GFAP proteins in the 4xLPS group did not restore to the normal level by the time of the second stimulus (MPTP) administration and was significantly increased compared to NS and 1xLPS groups. However, MPTP treatment did not further upregulate the expression of GFAP proteins in 4xLPS/MPTP group (Fig. 5B). Interestingly, the number of GFAP+ cells in the striatum was significantly higher in 4xLPS/MPTP group compared to 4xLPS/NS group, and there were no differences observed among the three groups that received MPTP injections (Additional file 1: Figure S4). These results indicated that the innate immune tolerance memory lasted for at least one month and led to a certain resistance against dopaminergic degeneration in PD mouse model. However, the modified microglia in TR group did not induce aggravated PD-like pathology in mice.

Fig. 5

The pre-stimuli of 1xLPS or 4xLPS change MPTP-induced motor impairments and dopaminergic damage to the nigrostriatal axis in mice. A The results of behavioral tests in mice 2 days post-MPTP administration. Time to turn around, climb down and the total time in the Pole test, and the rearing count over a 3-min period in the Rearing test were shown. n = 6–10/group. The difference shown by “++” was analyzed by unpaired t test. ++p < 0.01. B The protein levels of TH and GFAP in mouse striatum detected by Western Blot. n = 4–8 /group. C Assessments of the optical density of TH+ nerve fibers in the striatum of mice by immunohistochemistry staining. Scale bar, 200 μm. n = 4/group. D Assessments of TH+ neurons in the SNpc of mice by immunohistochemistry staining and stereological counting. Scale bar, 100 μm. n = 4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001. The difference shown by “+” was analyzed by unpaired t test. +p < 0.05. ns, no significant difference

HIF-1α plays a role in the MPTP-induced inflammation in chronic innate immune memory mouse models

According to the above experimental results, the existence of innate immune tolerance memory alleviated the MPTP-induced damage in nigrostriatal pathway. In view of the regulating effect of HIF-1α on IIM, we investigated the activation state of microglia in PD mouse models with IIM. The experimental schedule followed what was described in Fig. 4A. Results of triple immunofluorescent staining and the subsequent quantification analysis revealed an increased density of HIF-1α+ microglia and CD16/32+ microglia in 1xLPS or 4xLPS groups, compared to NS group, one month after the pre-stimulation. MPTP administration increased the number of HIF-1α+ microglia and HIF-1α+ CD16/32+ microglia in all groups, and increased the number of CD16/32+ microglia in NS/MPTP (PD) and 1xLPS/MPTP (TR PD) groups, but not in the 4xLPS/MPTP (TL PD) group. After the second stimulus of MPTP, mice in TL PD group displayed reduced density of CD16/32+ microglia and HIF-1α+ CD16/32+ microglia compared to both the PD and TR PD groups. At the same time, the TR PD mice exhibited an increased number of HIF-1α+ microglia compared to PD and TL PD mice (Fig. 6A, B). After analyzing the relative proportion of microglia subpopulations in MPTP-induced PD mice based on IIM models, we found there were significant variations. The differences in the proportion of HIF-1α− CD16/32− microglia between these groups were remarkable. Compared to NS group, the proportions of HIF-1α− CD16/32− microglia were lower in all the other groups. Administration of MPTP resulted in a decrease in the proportion of HIF-1α− CD16/32− microglia compared to their respective control groups (NS, 1xLPS or 4xLPS). Among the three groups intoxicated with MPTP, 1xLPS/MPTP group exhibited the lowest proportion of HIF-1α− CD16/32− microglia compared to the other two groups (Fig. 6C). Considering all of the quantitative results from the HIF-1α and CD16/32 staining, we conducted a correlation analysis to assess the density of HIF-1α+ and CD16/32+ cells among all the IBA1-immunolabeled microglia. Correlation diagram (scatter plot) revealed a significant positive correlation with an r value of 0.7879, p < 0.001 (Fig. 6D). Altogether, these results suggested the expression of HIF-1α in microglia had correlation with CD16/32 and could serve as a biomarker for assessing microglia phenotype.

Fig. 6

The expression of HIF-1α in striatal microglia of chronic innate immune memory model mice treated with NS or MPTP. A Representative images of IBA1/HIF-1α/CD16/32 triple immunostaining in the striatum of MPTP-induced PD mice with innate immune memory. Scale bar, 50 µm. Small plots in the upper left quadrant showed the zoomed details. Scale bar, 20 µm. The experimental schedule was described as in Fig. 3A, B. Quantification of the number of HIF-1α+, CD16/32+ and HIF-1α+ CD16/32+ cells in all IBA1-immunolabeled cells from mouse striatum, 7 days post-MPTP administration. n = 4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001. C The relative proportion of IBA1-immunolabeled cells in MPTP-induced PD mice based on innate immune memory models. The changes in the proportion of HIF-1α− CD16/32− microglia in the experimental groups compared to Control groups was marked by asterisks (*) within the figure, and differences between these groups was indicated by pound sign (#). n = 4/group. Differences were analyzed by one-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001 vs control groups. ###p < 0.001. D Pearson’s correlation analysis showing density of HIF-1α+ and CD16/32+ cells in all of the IBA1-immunolabeled microglia. Pearson's correlation coefficient (r) and p-values are shown above

Specific knockout of Hif-1α in microglia hinders the formation of innate immune memory

To further explore the role of HIF-1α in innate immune memory regulation, we employed a tamoxifen-inducible Hif-1α conditional knockout (cKO) mouse model whereby Hif-1α was selectively deleted in TMEM119+ microglia. Mice were treated with tamoxifen for 5 consecutive days, and the subsequent experiments were conducted 2 weeks later. We performed PCR and immunofluorescence staining to examine the absence of microglial HIF-1α in cKO mice. Notably, HIF-1α was rarely expressed under physiological conditions (Fig. 7A, Additional file 1: Figure S5A-C). Our results confirmed the high efficiency of Hif-1α gene deletion in microglia.

Fig. 7

Acute innate immune memory models cannot be established in microglial Hif-1α conditional knockout mice. A Experimental strategy for tamoxifen-induced deletion of Hif-1α specifically in microglia. B The changes of body weight during LPS injections. n = 5–8/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, ***p < 0.001. The experimental schedule was the same as Fig. 2A. (Ctrl: Cre-negative control littermates; cKO: Hif-1α cKO). C Transcription of pro-inflammatory cytokines (Il-1β, Il-6 and Tnf-α) and anti-inflammatory cytokines (Il-4 and Il-10) in the striatum of Ctrl and cKO mice detected by RT-qPCR. n = 3–5 /group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. **p < 0.01, ***p < 0.001 vs normal saline (NS) control. ##p < 0.05, ###p < 0.001 vs Ctrl mice. ^^p < 0.01, ^^^p < 0.001. D Representative images of IBA1 (red) immunostaining in the striatum of mice 3 h after NS, 2xLPS or 4xLPS injection. Scale bar, 100 µm. Quantity analysis of IBA1+ cells in the dorsal striatum is shown in the right panel. n = 3–4/group. Differences were analyzed by two-way ANOVA followed by LSD multiple comparison tests. *p < 0.05, ***p < 0.001 vs normal saline (NS) control. #p < 0.05, ##p < 0.01 vs Cre-negative control. ^^p < 0.01

To investigate whether lack of HIF-1α in microglia had any effects on IIM, we established AIIM models in Cre-negative control (Ctrl) and Hif-1α cKO mice. The experimental schedule diagram was the same as Fig. 2A. We found that cKO mice exhibited less weight loss compared to controls during LPS injections (Fig. 7B). When compared to Ctrl mice, transcriptional levels of pro-inflammatory cytokines, including Il-1β, Il-6 and Tnf-α, in the striatum of cKO mice treated with 2xLPS were significantly inhibited. On the other hand, the transcription of these cytokines in the striatum of both Ctrl and cKO mice following the stimuli of 4xLPS did not change, in comparison to the NS group, and there was no significant difference between the two 4xLPS-treated groups (Fig. 7C). Notably, the expression of inflammatory cytokines in NS-treated Ctrl and cKO mice showed no difference. By analyzing the above results, we observed that the deficiency of HIF-1α in microglia h