Different doses of systemic LPS induce different degrees of polarization of microglia and astrocytes

Microglia and astrocytes are activated in different phenotypes to exert opposite effects. The recently reported intraperitoneal injection of 5 mg/kg lipopolysaccharides (LPS) to promote A1 astrocytes by activating M1 microglia was found to cause high mortality. Furthermore, reported doses of systemic LPS used to induce M1 microglia vary widely (0.1 ~ 5 mg/kg). We aimed to study microglia and astrocytes polarization induced by various LPS doses in the central nervous system, and assess whether downregulation of C3a receptor (C3aR) in astrocytes contributes to an increased A2/A1 ratio. Methods Seventy-two genes for A1, A2, A-pan, M1, and M2 markers were detected by real-time polymerase chain reaction 24 hours after LPS treatment in the cerebral cortex, hippocampus, and spinal cord. C3aR in astrocytes was knocked down by intrathecal injection of AAV-C3aR-GFAP 21 days before LPS administration. Co-immunouorescence of C3aR with microglia, astrocytes, and neuron markers were performed to verify the specicity of C3aR knockdown in astrocytes. Changes in the 72 genes in the spinal cord were detected again 24 hours after LPS injection.

Activated microglia and astrocytes play a crucial role in neuroin ammation, which is known to be involved in neurodegenerative diseases, neuropsychiatric disorders, brain trauma, and pain [1][2][3] . Microglia are sensitive to pathogens or damage; they can be stimulated by lipopolysaccharides (LPS) or interferon (IFN)-γ to an M1 phenotype for the expression of pro-in ammatory cytokines or by interleukin (IL-4) or IL-13 to an M2 phenotype to resolve in ammation and repair tissue [4][5] . Upon injury, astrocytes are also activated and play dual roles. Neuroin ammation induced by LPS promotes neurotoxic astrocytic polarization, termed the A1 phenotype, while ischemia induced by middle cerebral artery occlusion (MCAO) leads to a neuroprotective astrocyte polarization, termed the A2 phenotype [6] . Systemic LPSinduced neuroin ammation has been evidenced by increased microglia and astrocytes activation [7,8] .
Numerous scienti c studies have demonstrated that systemic LPS can activate microglia to an M1 phenotype and astrocytes to the A1 phenotype. Furthermore [9,10] , A1 astrocytes induced by LPS were found to be activated through microglial secretion of the neuroin ammatory cytokines tumor necrosis factor (TNF)-α, IL-1α, and the complement component C1q [11] . However, unlike the commonly used low dose of LPS used to induce M1 activation [12][13][14] , the dosage used to activate A1 astrocytes has been reported to be as high as 5 mg/kg [11,15] , a similar dose to induce liver injury or sepsis [16,17] . In another ongoing study on the mechanism of A1 astrocytes in a rat chronic postoperative pain model, we employed 5 mg/kg systemic LPS to induce A1 astrocytes. However, a high mortality rate was observed 24 hours after LPS injection, and the survival rate could not be improved by increasing the weight of rats from 250 g to 350 g. Since this LPS dose was also used to create a liver injury and sepsis model, which could lead to death [16,17] , we wondered whether decreasing the LPS dose could still successfully induce A1 polarization. LPS has been found to induce A1 polarization not directly but through M1 activation, and the reported LPS dose-activating M1 phenotype was much lower than 5 mg/kg, although it varied [11][12][13][14][15] .
Each phenotype has multiple markers, and the selected markers often vary [7,18,19] . In a study performed by Xu et al., TNF-α and IL-1β in the cerebral cortex were used as M1 markers, and CD206, Ym1/2, and Arg1 were used as M2 markers [18] . In Wang's research, CD16, TNF-α, inducible nitric oxide synthase (iNOS), and monocyte chemoattractant protein (MCP)-1 in the hippocampus were used as M1 markers and CD206, transforming growth factor (TGF)-β, Arg1, and Ym1 as M2 characteristic markers [19] . It remains unclear whether any of these phenotype markers are appropriate classical markers of LPS in any central nervous tissue. Zamanian et al. showed that reactive astrocytes induced by LPS or ischemic stroke upregulated over 1,000 genes [7] . Some of the upregulated genes are unique to the LPS subtype (A1) or the MCAO subtype (A2); for example, H2-D1 and Serping1 are markers for A1 astrocytes. It remains unclear whether the upregulated genes induced by LPS present a dose-dependent response to LPS.
Targeted inhibition of pro-in ammatory phenotypes or promotion of anti-in ammatory phenotypes is a promising technique for alleviation of neuroin ammation. Downregulation of A1 astrocytes has been shown to protect against white matter injury in the chronic hypoperfusion model mice [20] . Furthermore, promoting astrocytes polarized into A2 astrocytes by silencing miR-21 was shown to improve the formation of synapses and nerve growth after acute ischemic spinal cord injury [21] . Complement C3 was recently found to be a classical marker of A1 astrocytes [11,15] , and its interaction with the C3a receptor (C3aR) plays a crucial role in neuroin ammation regulation [22,23] . C3 and C3aR expression have been reported to be positively correlated with cognitive decline. Deletion of C3aR in PS19 mice reportedly attenuates neuroin ammation, synaptic de cits, and neurodegeneration [22] . Interestingly, in in vitro astrocytes, C3 was found to interact with itself C3aR, and downregulation of C3aR in astrocytes not only decreased C3 expression but also many transcription factors of A1 astrocytes [24] . To test whether downregulation of C3aR in astrocytes could protect astrocytes from polarization into the detrimental A1 phenotype induced by LPS in vivo, we knocked down C3aR in astrocytes by intrathecal injection with AAV2/9-r-C3ar1 shRNA-glial brillary acidic protein (GFAP) before LPS administration.
Although targeting of the anti-in ammatory phenotypes shows much promise, the doses of systemic LPS adopted to induce microglia and astrocytes polarization vary considerably. Therefore, we conducted this study to explore the differences in various LPS dosages by covering multiple phenotype markers and tissue variances. By comparing the differences, we expect to provide more speci c information about a proper marker. Furthermore, the involvement of C3aR in astrocytes in LPS-induced pro-in ammatory phenotypes was also studied.

Animals
Adult male Sprague-Dawley rats weighting 120~150 g and 250~300 g used in this study were provided by the Experimental Animal Center of the Chinese Academy of Medical Sciences (Beijing, China). Rats were housed at 23 ± 1 ℃ under a 12 hours light-dark cycle with food and water ad libitum. Experimental protocols were approved by the Institutional Animal Care and Use Committee in the Chinese Academy of Medical Sciences (Beijing, China), and were performed in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals. For the LPS dose groups, rats weighting 250~300 g were randomly divided into 6 groups (0, 0.1, 0.33, 1, 3, and 5 mg/kg LPS), n = 8. To verify the e ency and speci city of AAV, rats weighting 120~150 g were randomly divided into 2 groups (AAV2/9-r-C3ar1 shRNA-GFAP, AAV2/9-GFAP), n = 3. For the invovelment of C3aR in LPS induced neuroin ammation, rats weighting 120~150 g were randomly divided into 2 groups (AAV2/9-r-C3ar1 shRNA-GFAP + LPS , AAV2/9-GFAP + LPS), n = 4.

LPS treatment
LPS from Escherichia coli 055:B5 (dissolved in endotoxin-free PBS) was puri ed by phenol extraction (L2880, Sigma-Aldrich, St. Louis, MO). The animals were injected with 0.1, 0.33, 1, 3 and 5 mg/kg LPS intraperitoneally (i.p.), respectively. The injection volume was diluted to 500 μl with endotoxin-free PBS for each injection. In the control group, 500 μl endotoxin-free PBS were injected i.p.. Tissues including the cerebral cortex, hippocampus and spinal cord were extracted 24 hours after LPS treatment.

Intrathecal administration
Under anesthesia, rats were put in a prone position with a manual bolster under the lower abdomen to fully expose the L4-L5 intervertebral space. Incisions were made in the back skin of the rats, and the lumbar fascia was exposed. The L5 acantha was held with a clamp by the left hand, and a 25 μl microsyringe needle was vertically inserted into the L4-L5 intervertebral space through the ligamenta ava with the right hand. The microsyringe was stopped when a tail ick or a paw retraction was observed, as these responses demonstrated correct intrathecal needle placement. Then, 20 μl AAV2/9-r-C3ar1 shRNA-GFAP or AAV2/9-GFAP was injected slowly through the needle and held a few seconds before removal. A tail ick could been observed when the needle was removed, and then the wound was sutured. Rats that showed any movement disorder on the next day were excluded from the experiments. Twenty-one days after AAV administration, rats received a single dose of LPS injection determined by the above LPS dosage response experiment. Chemogenetical AAV (titer, 2.5 × 10 12 vg/mL) was purchased from Hanbio Technology (Shanghai, China).

Real-time polymerase chain reaction (RT-PCR)
Rats were sacri ced with overdose anesthetics 24 hours after LPS treatment. The cerebral cortex, hippocampus and L4-L5 spinal cord from the same group were mixed respectively, and total RNA was extracted with Trizol regent (15596026, Invitrogen, Carlsbad, CA) by using ultrasonic cracking, and reverse transcribed with Prime Script RT Master Mix (RR036, Takara, Otsu, Japan) according to the manufacturer's instructions. Quantitative reverse transcriptase (RT-PCR) was performed using SYBR Premix Ex Taq (RR820, Takara) and a Step One Real Time PCR System (Applied Biosystems, Foster City, CA). Three repeated spottings were set, and if necessary, another three more spottings were repeated when the three spottings showed inconsistent result. The sequences of primers used were presented in Supplementary Table 1.
Immunohistochemistry Rats (n = 3) from the 3 mg/kg LPS group and the control group were sacri ced with excessive anesthetics 24 hours after LPS treatment, and sacri ced 21 days after AAV intervention. Rats were then transcardially perfused with PBS followed by fresh 4% paraformaldehyde. The whole brain and L4-L5 spinal cord were extracted and post xed in 4% paraformaldehyde for 4 hours, and then dehydrated in 30% sucrose overnight until the tissues settled to the bottom of the EP pipe at 4 °C. Tissues were embedded in OCT (4583, Tissue-Tek, SAKURA, USA) and frozen at -80 °C, and then sectioned at 12 μm thick using a cryostat (Leica 2000, Germany). The sections were permeated in 0.3% Triton-X-100 in PBS for 30 min and incubated with 10% bovine serum albumin in PBS for 60 mins, following an overnight incubation with primary antibodies at 4 °C. Then sections were incubated with corresponding secondary antibodies for 1 hour at room temperature (the primary and secondary antibodies used for immunohistochemistry were listed in Supplementary Table 2), and visualized using a CCd spot camera (Olympus DP71).

Statistical analysis
Results were presented as means ± SEM, and statistical analyses were performed using GraphPad Prism 8.0 software package (version 8 for Windows, San Diego, CA).

Results
Systemic LPS induced polarization of microglia and astrocytes As a classical reagent to induce M1 and A1 activation, systemic LPS is expected to predominantly polarize the pro-in ammatory M1 microglia and/or A1 astrocytes. However, in addition to the expected reactive M1 and A1, we observed obvious M2 and A2, especially A-pan activation in the cerebral cortex, hippocampus, and spinal cord ( Figure 1). The majority of markers of A1 and A-pan astrocytes showed an increase when exposed to LPS, even at a low dose in the cerebral cortex, and a large LPS dose induced more activation. Markers of M1 and A2 showed an inconsistent trend with increasing LPS, with a fair number of markers showing no change, such as IL-12a, IL-6, Ptgs2 of M1 genes, and Clcf1, Emp1, and Sphk1 of A2 genes, while some others showed an obvious increase with LPS. Many M2 markers, such as IL-10, Jak3, IL-1ra, Ym1, and Clec7a, were also activated by LPS ( Figure 1A). Reactive microglia induced by LPS in the hippocampus were similar to those in the cortex, while much fewer markers of A2 and Apan were activated in the hippocampus compared to the cortex ( Figure 1B). The majority of markers in A1 and A-pan, except Ugt1a1 and Amigo2, showed an increase with LPS in the spinal cord, and the increased consistency of A2 markers was better than that in the cortex and hippocampus. Changes in the M1 and M2 markers in the spinal cord were similar to those in the cortex ( Figure 1C). To further con rm the mRNA data, we performed co-immuno uorescence of reactive astrocytes.
RT-PCR results showed a 7.9-fold in upregulation of C3, which marks A1, a 2.3-fold increase in S100a10, which marks A2, and a 2.2-fold increase in GFAP, which marks A-pan in the 3 mg/kg LPS group in the spinal cord. Consistent with these results, our immuno uorescence data showed an obvious increase in the expression of C3, S100a10, and GFAP. The co-immuno uorescence of C3 with GFAP ( Figure 2A) and S100a10 with GFAP ( Figure 2B) were also upregulated in the 3 mg/kg LPS group compared to the control group.
Number of phenotypic markers upregulated more than 2-fold by LPS Interestingly, in the LPS-induced activation of M1 and A1, some classical markers did not show a large increase, but most showed a 2-to 10-fold upregulation, such as CD86 (2-to 4-fold), TNF-α (2-to 9-fold), and C1qa (2-to 4-fold). For further quantitative analysis of the complex heat map results, we compared the number of markers with an upregulation of more than 2-fold under LPS stimulation. When exposed to a low LPS dose, such as 0.1 or 0.33 mg/kg, most markers of A1 and A-pan in the cortex and spinal cord were activated, while fewer markers of M1, M2, and A2 were reactive. With increasing LPS dose, markers of M1 and M2 presented a similar upregulation in both the brain and spinal cord. The majority of A1 and A-pan markers were activated, especially in the cortex and spinal cord, and those of A2 markers presented a relatively lower activation at large LPS doses. Although largely increased, there seems to be no difference in the activated numbers between 3 and 5 mg/kg LPS doses ( Figure 3).
Large LPS dose induced more microglia activation, but not astrocytes Considering that 5 mg/kg LPS led to higher mortality, and similar numbers (≥ 2-fold) of reactive gliocytes induced by 3 and 5 mg/kg LPS (Figure 3), we further analyzed whether 3 mg/kg LPS could replace 5 mg/kg LPS to induce microglial and astrocytes polarization. We compared the number of markers reaching the maximum in all LPS doses at 3 mg/kg or 5 mg/kg LPS.
Results showed that all the phenotypes except the A1 subtype in the cortex reached a maximum at 5 mg/kg LPS. More markers of M1, M2 and A-pan in the hippocampus peaked at 5 mg/kg compared to 3 mg/kg LPS. However, only more markers of M2 in the spinal cord peaked at 5 mg/kg compared to 3 mg/kg LPS. From the phenotype perspective, more M1, M2, and A-pan in the cranial tissues were activated at 5 mg/kg LPS than at 3 mg/kg. For A1 and A2 astrocytes, there was no obvious difference between the two LPS doses (Figure 4).

Phenotypic markers presenting a good dose response with LPS
More than ten markers were classi ed as special genes of each phenotype, but only a few of the 72 markers showed a good dose response with LPS in any nervous tissue; they were Marco marked M1, Ym1 marked M2, and C3 marked A1 ( Figure 5).
Our results showed that C3aR was widely co-expressed with GFAP in the spinal white matter ( Figure 6A) and Neun in the spinal gray matter ( Figure 6B), and decreased in the AAV2/9-r-C3ar1 shRNA-GFAP group (AAV). Compared with the AAV2/9-GFAP group (Control), C3aR was mostly downregulated in astrocytes in the AAV group ( Figure 6A), with almost no changes in microglia ( Figure 6C) and neurons ( Figure 6B). In addition, the expression of C3aR mRNA in the spinal cord of the AAV group was signi cantly downregulated ( Supplementary Fig. 1), indicating the e ciency of AAV2/9-r-C3ar1 shRNA-GFAP. Rats were then injected with 3 mg/kg LPS (i.p.). The dose was selected because no activation advantage of A1 and A2 astrocytes was found at 5 mg/kg LPS compared with 3 mg/kg LPS ( Figure 5), while the latter contributed to a higher survival rate ( Supplementary Fig. 2).
treated mice with an even lower LPS dose (0.1 mg/kg) for 24 hours to induce M1 activation, and they detected upregulation of TNF-α, IL-1β, IL-6, CD86, CD16, and CD32 in the brain [12] . In our experiments, we showed that systemic LPS with doses ranging from 0.1 to 5 mg/kg induced M1 activation in both the brain and spinal cord. As the LPS dose increased, the number of activated M1 markers (≥ 2-fold) increased slowly, with 13 of 22 activated at 5 mg/kg LPS. In particular, M2 markers presented a similar increasing trend and degree as M1. Thus, we infer that systemic LPS lasting 24 hours with a dosage ranging from 0.1 to 5 mg/kg may not be a proper choice to induce M1 polarization. On the one hand, it did not make microglia polarize into M1 more than M2; on the other hand, only half of the M1 genes were activated, with less than 25% upregulated at the most reported dose of 0.1 mg/kg. In a study comparing 4 hours and 24 hours of LPS treatment, a weaker activation of M1 markers CD86, iNOS, and IL-1β was detected at 24 hours compared to the 4-hour group [13] . Whether a shortened stimulation of LPS contributes more to M1 polarization than M2 polarization is not clear. Moreover, in Wang and Zhao's study [19,26] , small LPS doses (0.75 and 0.25 mg/kg, respectively) were administered i.p. for 7 days to induce cognitive de cits by activating M1 microglia.
LPS has been recently reported to induce A1 activation. Studies by Liddelow [11] , Zamanian [7] , and Kano [27] all used systemic LPS 5 mg/kg to induce A1 polarization, and Zhang [28] selected 0.83 mg/kg LPS to activate A1 astrocytes to create murine depression-like behavior. Our results showed that even a small intraperitoneal LPS dose (0.1 mg/kg) could activate more than half of the A1 markers. As LPS increased, more than 80% of A1 markers were activated in both the brain and spinal cord. Furthermore, fewer A2 markers were activated, especially in the brain, indicating that LPS contributes to superior A1 polarization, thus supporting it as a good inducer to activate A1 astrocytes.
Besides the upregulated M1 and A1 genes induced by LPS, we also found that M2 and A2 markers were signi cantly increased, as demonstrated by our co-immuno uorescence of elevated S100a10 with GFAP certi cates. The dichotomy of M1 and M2, or A1 and A2, is an oversimpli ed conceptual framework. M2, which is often induced by IL-4 can be further characterized into several main states [29] : M2a with alternate activation and involvement in repair and regeneration; M2b with an immunoregulatory phenotype; and M2c with an acquired-deactivating phenotype. Chhor et al. found that the majority of M1 markers (10/13) and M2b markers (8/13) were maximally expressed after 12 hours of exposure to LPS. Moreover, M2c genes (IL-10, Il-4Rα, and SOCS3) were also increased in enriched microglia 4 hours after LPS injection [30] . Similarly, we found that more than half of M2 genes, especially M2a genes, were activated when the dose of LPS injection was larger than 0.33 mg/kg, and most reached a maximum at LPS 5 mg/kg. Identically, although less sensitive to LPS compared with A1, A2 was also activated, particularly in the spinal cord. A2 is often induced by MCAO and upregulates many bene cial in ammatory factors, such as hypoxia-induced factors, IL-6 and IL-10 [11,31] . Markers of A2 could also be activated by LPS, although they were more sensitive to MCAO. For example, as A1 markers, H2-D1 was induced 30-fold by LPS, but only 3-fold by MCAO, and Serping1 was induced 6.5-fold after MCAO and 34fold after LPS [11] . Our results showed that Serping1 was induced 34-fold in the cortex, 111-fold in the hippocampus, and 55-fold in the spinal cord under LPS 5 mg/kg, indicating a difference in tissues induced by LPS. Overall, LPS exposure could induce not only a pro-in ammatory response, but also antiin ammatory action. However, the question remains whether the transient M2 or A2 polarization is a selflimiting reaction that is bene cial for regeneration or just a recovery failure.
Genes that are induced by both neuroin ammation and ischemia were classi ed as pan-reactive (A-pan) genes [15] , different from the speci c genes (A1) induced by neuroin ammation. Our research showed that, with increasing LPS dose, A-pan markers presented a similar trend to A1, especially in the cerebral cortex and spinal cord. It should be noted that the dichotomy of A1, A2, and A-pan is an oversimpli ed conceptual framework, and the status of astrocytes may include a battery of different, but overlapping, functional phenotypes. Although LPS is often used as a classical reagent to induce A1 astrocytes, it could activate other astrocyte phenotypes like A-pan, which may show an overlapping function with A1. We infer that selective blocking of A1 formation could be discounted by the function-similar A-pan, and this requires further study.
In addition to the various LPS doses, gene changes in different tissues were also detected. The cortex and hippocampus were preferred when studying LPS-induced impairments in spatial learning and memory. In Wang et al.'s study, mice were administered i.p. with LPS (250 μg/kg) once daily for 7 days, a 2-fold increase in TNF-α was found in the cortex and hippocampus by ELISA, and increases were nearly 3-fold for IL-1β in the cortex and 2.5-fold in the hippocampus [19] . Consistent with this, our results showed that when exposed to 0.33 mg/kg LPS, TNF-α mRNA was increased 2-fold both in the cortex and hippocampus, while IL-1β was raised by 19-fold in the hippocampus and 36-fold in the cortex, which may correlate with the increased LPS dose-induced pro-in ammatory state. In particular, more markers of microglia and fewer astrocytes were activated in the hippocampus than in the cortex and spinal cord.
These data remind us that activated genes induced by LPS vary not only with LPS dose, but also tissue type and stimulation time. The complex signals in the lesion microenvironment determine the polarization of glial cells.
Our results showed that several markers, such as Marco, Ym1, and C3, presented a signi cant doseresponse relationship with LPS increase in any tissue. TLR4 is an important receptor of LPS, and the interaction of TLR4 with adaptor MyD88 leads to the activation of downstream NF-κB and subsequent production of pro-in ammatory cytokines [32] . With increasing LPS, more pro-in ammatory cytokines were released. Whether these dose-dependent markers are involved in in ammation production or could re ect the degree of in ammation or glial cell polarization remains to be elucidated.
Lectin Ym1, a well-established marker of murine M2, is a secretory protein strongly induced by IL-4 and IL-13. The protective action of Ym1 expression after stroke has been demonstrated in several studies. Fumagalli et al. found increased Ym1 levels with reduced infarct sizes 24 h after transient MCAO in CX3CR1-de cient mice [33] . Barbera-Cremades showed the expression of Ym1 was associated with the inhibition of macrophage proliferation and is induced both by anti-in ammatory signal adenosine and ATP, and pro-in ammatory signal LPS [34] . This is consistent with Lee's study, which showed that LPS signi cantly increased Ym1 in the anterior cortex, hippocampus, and entorhinal cortex [35] . This raises questions regarding how the upregulation of Ym1 and other M2 or A2 markers induced by LPS is regulated by pro-in ammatory stimuli, and why mRNA levels of Ym1 are increased in AD patients and animal models of amyloid deposition, which are typically considered to be associated with a proin ammatory (M1 or A1) cytokine environment. These results also suggest a more complex set of microglia and astrocyte phenotypes than the dipolar M1/M2 or A1/A2 characterization, and indicate that LPS-induced neuroin ammation may be mediated by distinct activation subtypes. Meanwhile, the classical markers of M1, such as CD86 and M2 marker CD206, increased with LPS in the cortex and hippocampus, but not in the spinal cord, suggesting that the particular markers of phenotypes should vary with tissues.
Downregulation of C3aR in astrocytes affected the status of activated markers A1 and A2 induced by LPS. Only Ggta1 (4-fold) and Srgn (2-fold) of A1 were upregulated, while more markers of A2, including S100a10, Clcf1, Emp1, Sphk1, Slc10a6, and especially Ptx3 (10-fold) and CD14 (7.4-fold), were elevated, indicating that A2 polarization dominated the reactive astrocytes when C3aR in astrocytes was knocked down. Interestingly, C3aR downregulation in astrocytes under LPS stimulation also affected microglial activation, such as decreasing IL-1β, Stat1, and TNF-α in M1 and increasing CD206, IL-1ra, and Clec7a in M2. The signi cant upregulation of IFN-γ and downregulation of TNF-α re ect a dominant antiin ammatory state, consistent with the superior activation of M2 and A2. We infer that the superior M2 activation may be caused by the A2 polarization. During development, astrocytes can sense subtle changes in neurons to induce the production of C1q in neuronal synapses, which interacts with the microglial C3aR to prune the neuronal synapses through the classic cascade complement pathway [23] . In the context of Alzheimer's disease pathology, overproduction of C3 from astrocytes can simultaneously communicate with microglial C3aR and neuronal C3aR to dynamically regulate microglial phagocytosis and impair dendritic morphology, as well as synaptic function, subsequently resulting in deterioration of cognitive function [22] . Meanwhile, reactive A2 astrocytes could also stimulate more active M2 microglia to produce an anti-in ammatory microenvironment.

Limitations
There were several limitations to our study. Our focus was on the dose relationship between LPS and gliocyte phenotypes, so we mainly used RT-PCR to quantify this. Morphology data on gliocyte polarization were partly shown, which could provide more information about polarization. We employed three variables-dose, phenotypes, and tissues-in our study, and did not include time variants.
Considering that astrocytes are often activated following microglia, polarization changes with time need to be detected in future studies. Though we did not show the classical one-way ANOVA or Student's t test statistical results which we have done, the presented gures could help better interpretate the multipledimensional RT-PCR results.
In conclusion, our results showed that systemic LPS activates not only M1 and A1, but also M2, A2, and A-pan. Only a few markers of the 72 genes showed a signi cant dose response to LPS, which may correlate with the degree of in ammation caused by LPS. No more A1 astrocytes were polarized at 5 mg/kg LPS than at 3 mg/kg LPS in the brain and spinal cord. In addition, the downregulation of C3aR in astrocytes contributed to more A2 and M2 polarization in the spinal cord when exposed to LPS.

Declarations
Ethics approval: All animal procedures performed in this study were reviewed and approved by the Institutional Animal Care and Use Committee in the Chinese Academy of Medical Sciences (Beijing, China), and were performed in accordance with the National Institutes of Health guide for the care and use of laboratory animals.
Consent for publication: Not applicable.
Availability of data and materials: There are no data, software, databases, and application/tool available apart from the reported in the present study. All data are provided in the manuscript and supplementary data.
Competing interests: The authors declare that they have no any competing interests.   Systemic LPS increased expressions of both A1 marker C3 and A2 marker S100a10. Compared with the control group (equal dose of PBS), injection of 3 mg/kg LPS intraperitoneally increased the colocalization of A1 marker C3 with astrocytic marker GFAP (a), and also A2 marker S100a10 with GFAP (b) in the spinal cord. 20*, 40* presenting magni cation. n = 3 per group.  Number of markers reaching maximum of all the LPS doses are compared between 3 and 5 mg/kg LPS. a-b More microglia markers reaching maximum at 5 than 3 mg/kg LPS. c More A1 astrocytes were reactive at 3 than 5 mg/kg LPS. d-e 5 mg/kg LPS activated more A2 and A-pan markers in the brain, while no advantage in the spinal cord compared with 3 mg/kg LPS.