Chemogenetic manipulation of CX3CR1⁺ cells transiently induces hypolocomotion independent of microglia

Abstract

Chemogenetic approaches using Designer Receptors Exclusively Activated by Designer Drugs (DREADD, a family of engineered GPCRs) were recently employed in microglia. Here, we used Cx3cr1^CreER/+:R26^hM4Di/+ mice to express Gi-DREADD (hM4Di) on CX3CR1⁺ cells, comprising microglia and some peripheral immune cells, and found that activation of hM4Di on long-lived CX3CR1⁺ cells induced hypolocomotion. Unexpectedly, Gi-DREADD-induced hypolocomotion was preserved when microglia were depleted. Consistently, specific activation of microglial hM4Di cannot induce hypolocomotion in Tmem119^CreER/+:R26^hM4Di/+ mice. Flow cytometric and histological analysis showed hM4Di expression in peripheral immune cells, which may be responsible for the hypolocomotion. Nevertheless, depletion of splenic macrophages, hepatic macrophages, or CD4⁺ T cells did not affect Gi-DREADD-induced hypolocomotion. Our study demonstrates that rigorous data analysis and interpretation are needed when using Cx3cr1^CreER/+ mouse line to manipulate microglia.

Introduction

Microglia regulate neuronal functions in the healthy brain and respond to central nervous system (CNS) damage ^1–3. G protein coupled receptors (GPCRs) help microglia sense and respond to CNS environment ^{4, 5}. Microglia express multiple Gi protein coupled receptors (Gi-GPCR), including purinergic receptor P2Y12 (P2Y12), CX3C chemokine receptor 1 (CX3CR1), and C3a anaphylatoxin chemotactic receptor (C3aR) ⁶. These microglial Gi-GPCRs play multi-facets functions in normal and diseased brain ^{6, 7}. Consistently, inhibition of microglia Gi-GPCRs using genetically encoded pertussis toxin (PTX) alters microglia morphology, reduces microglia surveillance, and induces neuronal hyperexcitability ⁸. Thus, it is important to further understand Gi-GPCRs mediated signaling pathways in microglial functions in vivo.

To this end, a novel way to study Gi-GPCR signaling is through Designer Receptors Exclusively Activated by Designer Drugs (DREADD), a family of engineered GPCRs activated solely by inert exogenous compounds ⁹. It is a powerful tool for studying neuronal circuit mechanisms in various behaviors ¹⁰. Recently Gi-DREADD approaches were applied to regulate microglia functions in pain. The activation of microglial Gi-DREADD inhibits microglia proliferation, neuroinflammation, synaptic potentiation, and consequently attenuated neuropathic pain ^11–15.

In this study, we used Cx3cr1^CreER/+:R26^hM4Di/+ mice to express Gi-DREADD (hM4Di) on CX3CR1⁺ cells, comprising microglia and peripheral immune cells. We observed that Cx3cr1^CreER/+:R26^hM4Di/+ mice showed hypolocomotion immediately after activating hM4Di receptors. Surprisingly, we found that hypolocomotor phenotype persists with microglial depletion. The unexpected results highlight the need for caution in data interpretation when using the Cx3cr1^CreER/+ mouse line to manipulate microglia.

Results

Activation of Gi-DREADD in long-lived CX3CR1⁺ cells induces acute hypolocomotion.

We intended to utilize the Gi-DREADD (hM4Di) ⁹ to study the functions of microglial Gi protein coupled receptors in vivo. To induce hM4Di expression on long-lived CX3CR1 expressing cells like microglia, we used Cx3cr1^CreER/+:R26^hM4Di/+ mice and performed experiments at least 4 weeks after tamoxifen injection (Fig. 1A). As expected, hemagglutinin (HA) tagged hM4Di was found on the membrane of microglia, but not in neurons nor astrocytes in the brain (Fig. 1B and C). In addition, using ex vivo 2-photon imaging in acute brain slices, we found that hM4Di activation induced the chemotaxis of microglial processes toward Clozapine N-oxide (CNO, 5mM), a hM4Di agonist (Fig. 1D and E). This mimics the function of microglial Gi-GPCRs, such as P2Y12 and C3aR, which are known to mediate the movement of microglia processes towards ATP or C3a gradients ^{16, 17}. These results indicate the functional expression of hM4Di in microglia in vivo.

Figure 1. Activation of Gi-DREADD in CX3CR1⁺ cells induces hypolocomotion in mice.

(A) The genetic strategy of Cx3cr1^CreER/+:R26^hM4Di/+ mice to express Gi-DREADD on CX3CR1⁺ cells, including brain microglia.

(B and C) Immunostaining (B) and quantification (C) showing that hM4Di (red, identified by fused HA-tag) localized on the membrane of Iba1⁺ microglia (green) but not on neurons (NeuN⁺, green) or astrocytes (GFAP⁺, green). Thus, hM4Di is mainly expressed by microglia in the brain in Cx3cr1^CreER/+:R26^hM4Di/+ mice (n = 3 mice in each group). Scale bars: 50 μm (B, upper) and 10 μm (B, lower).

(D) Representative images of microglial process chemotaxis (green) towards CNO source (red) at 0, 5, and 10 minutes. Scale bar: 20 μm.

(E) Quantification demonstrating that activation of hM4Di induced the extension of microglial process chemotaxis within 15 minutes (n = 3 mice in each group).

(F) Timeline of inducing Gi-DREADD expression and OFT for Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after saline injection.

(G) Representative mouse movement trajectories of the saline control groups in the OFT.

(H and I) Quantification of total moving distance (H) and central time (I) of the saline control groups, suggesting the overexpression of hM4Di does not influence mouse behaviors (n = 7 mice in each group).

(J) Timeline of inducing Gi-DREADD expression and OFT or EPM for Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection.

(K) Representative mouse movement trajectories of Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection in the OFT.

(L and M) Quantification revealing that Cx3cr1^CreER/+:R26^hM4Di/+ mice traveled less distance (L) and spent less time in the central area (M) than Cx3cr1^CreER/+ mice after CNO injection in OFT (n = 12 mice, Cx3cr1^CreER/+; n = 11 mice, Cx3cr1^CreER/+:R26^hM4Di/+).

(N) Representative heatmaps (a color gradient ranging from blue to red reflects the time spent in location) of Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection in the EPM.

(O-Q) Quantification showing that Cx3cr1^CreER/+:R26^hM4Di/+ mice had less moving distance (O), spent less time in the open arm (P), and entered less to the open arm (Q) than Cx3cr1^CreER/+ mice after CNO injection in EPM (n = 13 mice, Cx3cr1^CreER/+; n = 15 mice, Cx3cr1^CreER/+:R26^hM4Di/+).

Data are represented as mean ± SEM. *p < 0.05 and ***p < 0.001. Two-way ANOVA (E) or Student’s t-test (H, I, L, M, O, P, and Q).

Loss of microglial Gi-GPCRs, such as P2Y12, CX3CR1, and C3aR, causes increased innate fear and anxiety-like behaviors in mice ^18–21. We performed OFT (open field test) and EPM (elevated plus maze) to test the general function of microglial Gi signaling in vivo. In the OFT, mice did not have behavioral differences after saline injection (Fig. 1F–I). However, 30 minutes after 2.5 mg/kg CNO administration, Cx3cr1^CreER/+:R26^hM4Di/+ mice but not Cx3cr1^CreER/+ mice showed a significant decrease in travel distance and time spent in the central area after CNO (Fig. 1J–M). Likewise, in the EPM, Cx3cr1^CreER/+:R26^hM4Di/+ mice given CNO showed less locomotion, time spent in the open arm, and entries into the open arm than the controls (Fig. 1N–Q).

Further characterization of Gi-DREADD-induced behaviors found that 12 hours after CNO injection, Cx3cr1^CreER/+:R26^hM4Di/+ mice traveled similar distance and spent comparable time to Cx3cr1^CreER/+ controls in the central area (Fig. S1A–D), indicating these behaviors are dependent on the transient hM4Di activation. Notably, a lower dose of CNO (1.0 mg/kg) was also able to induce similar behaviors in the OFT (Fig. S1E–H). To exclude the effect of anxiety on locomotion behaviors, we also measured the mouse activities in their homecage 30 minutes after CNO injection. Our results again showed the hypolocomotion in Cx3cr1^CreER/+:R26^hM4Di/+ mice compared with Cx3cr1^CreER/+ controls (Fig. S1I–K), suggesting Gi-DREADD induced hypolocomotion is not secondary to anxiety. Since mice showed hypolocomotion in behavior tests, we next examined their motor coordination by rotarod. We found that CNO did not affect the motor coordination in Cx3cr1^CreER/+:R26^hM4Di/+ mice (Fig. S1L and M). Additionally, sex differences were not observed in behavioral tests, that is, both male and female mice showed similar hypolocomotion in OFT, EPM or home cage after Gi-DREADD activation (Fig. S2). Together, these results suggest that systemically activating Gi signaling in long-lived CX3CR1⁺ cells induces acute hypolocomotion without affecting ataxia in mice.

Gi-DREADD-induced hypolocomotion does not require microglia.

The Cx3cr1^CreER/+ mouse line is commonly used to genetically manipulate microglia when peripheral CX3CR1-expressing immune cells are turned over 4 weeks after tamoxifen injection ^22–24. To determine if Gi-DREADD-induced hypolocomotion was mediated by microglia, we used PLX3397 to deplete microglia in both Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice (Fig. 2A). PLX3397 is a colony stimulating factor 1 receptor (CSF1R) inhibitor that ablates microglia and border associated macrophages in the CNS (but also affects peripheral immune cell compartments) (Fig. S3) ^{25, 26}. As expected, most microglia were depleted after 2 weeks of PLX chow treatment (Fig. 2B and C). However, we found that Cx3cr1^CreER/+:R26^hM4Di/+ mice with microglial ablation still moved less in the total field and in the central area compared with Cx3cr1^CreER/+ control mice after CNO injection (Fig. 2D–F). Thus, these results indicate that microglial Gi signaling is not responsible for the Gi-DREADD-induced hypolocomotion in Cx3cr1^CreER/+:R26^hM4Di/+ mice.

Figure 2. Gi-DREADD-induced hypolocomotion is independent of microglia.

(A) Timeline of inducing Gi-DREADD expression, microglia ablation, and OFT for Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection.

(B and C) Representative images (B) and quantitative analysis (C) showing PLX treatment efficiently ablated Iba1⁺ cells (green) in the brain (n = 3 mice in each group). Scale bar: 200 μm (B).

(D) Representative mouse movement trajectories of Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice with microglial ablation after 2.5 mg/kg CNO injection in the OFT.

(E and F) Quantification revealing that, after microglia ablation, Cx3cr1^CreER/+:R26^hM4Di/+ mice had less distance moved (E) and less time spent in the central area (F) than Cx3cr1^CreER/+ mice after CNO injection in OFT (n = 7 mice, Cx3cr1^CreER/+; n = 10 mice, Cx3cr1^CreER/+:R26^hM4Di/+).

(G) The genetic strategy of generating Tmem119^CreER/+:R26^hM4Di/+ mice to express Gi-DREADD (hM4Di) selectively in microglia.

(H and I) Immunostaining (H) and quantification (I) showing that hM4Di (red, identified by fused HA-tag) localized on the membrane of Iba1⁺ microglia (green) but not on astrocytes (GFAP⁺, green) or neurons (NeuN⁺, green) in Tmem119^CreER/+:R26^hM4Di/+ mice (n = 3 mice in each group). Scale bars: 50 μm (H, upper) and 10 μm (H, lower).

(J) Timeline of inducing Gi-DREADD expression and OFT for Tmem119^CreER/+ and Tmem119^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection.

(K) Representative mouse movement trajectories of Tmem119^CreER/+ and Tmem119^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection in the OFT.

(L and M) Quantification showing that Tmem119^CreER/+:R26^hM4Di/+ mice had similar travel distance (L) and the central time (M) with Tmem119^CreER/+ group (n = 7 mice in each group).

(N and O) Immunostaining (N) and quantification (O) showing that hM4Di (red, identified by fused HA-tag) localized on the membrane of Iba1⁺ microglia (green) 2 weeks after PLX ablation in Tmem119^CreER/+:R26^hM4Di/+ mice. Scale bar: 50 μm (N).

(P and Q) UMAP plot (P) and quantification (Q) demonstrating the majority of Lgals3⁺ microglia expresses Tmem119 (a color gradient ranging from grey to red reflects the relative Tmem119 level).

Data are represented as mean ± SEM. *p < 0.05 and ***p < 0.001. One-way ANOVA followed by Tukey’s post hoc test (C) or Student’s t-test (E, F, L, and M).

To further test whether hypolocomotor phenotype in Cx3cr1^CreER/+:R26^hM4Di/+ mice is microglia independent, we repeated the OFT in Tmem119^CreER/+:R26^hM4Di/+ mice (Fig. 2G). Tmem119 is a microglial signature gene, and the Tmem119^CreER/CreER mouse line was recently developed to specifically target microglia ^{27, 28}. Indeed, in the Tmem119^CreER/+:R26^hM4Di/+ mice, hM4Di was expressed on the membrane of more than 90% of microglia, but not neurons or astrocytes 4 week after tamoxifen injection (Fig. 2H and I). However, we found that Tmem119^CreER/+:R26^hM4Di/+ mice showed a similar travel distance and the central time to those in Tmem119^CreER/+ group 30 minutes after CNO administration (Fig. 2J–M). Importantly, when we treated Tmem119^CreER/+:R26^hM4Di/+ mice with PLX chow, we found microglia resistant to CSF1R inhibition also expressed Gi-DREADD (Fig. 2N and O). Our staining results are consistent with the previous single-cell RNA sequencing data showing Lgals3⁺ CSF1R inhibition resistant microglia still express Tmem119 (Fig. 2P and Q) ²⁹. These results demonstrate that Tmem119^CreER/+:R26^hM4Di/+ mice allow us to activate Gi-DREADD on both CSF1R inhibition sensitive and resistant microglia. Given that Tmem119^CreER/+:R26^hM4Di/+ mice did not show hypolocomotion, it is unlikely that PLX-resistant microglia attributed to the hypolocomotion. Together, these results indicate that activation of Gi signaling in CX3CR1⁺ cells induced hypolocomotion, which is not dependent on brain microglia.

To test whether microglia activation caused Gi-DREADD induced hypolocomotion, we performed daily intracerebroventricular (i.c.v.) minocycline (Mino) administration for 4 days prior to CNO injection and OFT (Fig. S4A). Our results showed that microglia inhibition did not rescue the Gi-DREADD induced hypolocomotion (Fig. S4B–D). Next, we examined the neuronal activity in response to Gi-DREADD activation. Neuronal activity changes during OFT were assessed by immunostaining of c-fos (Fig. S5A). Two hours after OFT (which is 2.5 hours after CNO injection), we harvested brains from Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice. Immunostaining for c-fos on these brain slices showed that there was no difference in the number of c-fos⁺ neurons in primary motor cortex (MOp), prefrontal cortex (PFC), and ventral hippocampus (vHIP) between two groups (Fig. S5B–G).

CX3CR1 is expressed within the myeloid lineage.

Since the hypolocomotion does not stem from the microglia, we then hypothesized that peripheral long-lived CX3CR1⁺ cells might play key roles in those behaviors. To identify cells expressing Cx3cr1 mRNA, we queried mouse Cx3cr1 expression patterns in 20 different organs at the single-cell level using the Tabula Muris database ³⁰. We examined 53,760 cells from the database to map Cx3cr1 transcription by organ (Fig. 3A). Among all the tissues, Cx3cr1 is mainly expressed in brain myeloid cells consisting of microglia. In addition, Cx3cr1 expression is also prevalently detected in other peripheral organs (Fig. 3B and C). At the single cell level, Cx3cr1 is mainly found within the myeloid lineage, including brain microglia/macrophages, renal macrophages, hepatic macrophages, pulmonary monocytes, and monocytes of the bone marrow (Fig. 3D). Of note, there are also progenitor cells that express Cx3cr1. For example, some granulocyte monocyte progenitor cells express Cx3cr1 at least transiently (Fig. 3D). Thus, once the gene is recombined after tamoxifen induction in these progenitor cells, Gi-DREADD could also be expressed in their progenies in Cx3cr1^CreER/+:R26^hM4Di/+ mice. The query suggests that Cx3cr1 controlled CreER recombinase could induce gene rearrangement in various cell populations in addition to brain microglia.

Figure 3. CX3CR1 is expressed in peripheral cells and Cx3cr1-controlled tdTomato (tdT) reporter labels circulating leukocytes.

(A) Query of Tabula Muris database showing cells in 20 mouse organs by t-SNE plot.

(B) t-SNE plot of all cells from the database demonstrating the Cx3cr1 was densely expressed in brain myeloid cells and prevalently detected in other peripheral organs (a color gradient ranging from grey to red reflects the relative Cx3cr1 level).

(C) UMAP plot of all Cx3cr1 expressing cells colored by the organ.

(D) Detailed cell ontology analysis revealing that Cx3cr1 was mainly found in myeloid lineage in various organs.

(E) The genetic strategy of generating Cx3cr1^CreER/+:R26^tdT/+ mice to trace the CX3CR1⁺ lineage.

(F) Timeline of inducing tdT expression and blood collection for flow cytometric analysis.

(G) Representative flow cytometry plots examining the constitutive expression of EYFP and inducible expression of tdT in circulating CD45⁺ leukocytes in Cx3cr1^CreER/+, Cx3cr1^CreER/+:R26^tdT/+, and R26^tdT/tdT mice 4 weeks after tamoxifen injection.

(H) Quantitative analysis showing that tdT⁺ EYFP⁺ leukocytes were only found in Cx3cr1^CreER/+:R26^tdT/+ mice.

(I) Histogram and quantification of the percentage of tdT⁺ cells (peak between dash lines) in EYFP⁺ cells.

(J) Quantitative analysis demonstrating that tdT⁺ EYFP^neg leukocytes were only found in Cx3cr1^CreER/+:R26^tdT/+ mice.

(K) Pie chart illustrating that B220⁺ B cells, CD4⁺ and CD8α⁺ T cells were 3 major populations in the tdT⁺ EYFP^neg cell population.

(L-Q) Histograms and quantification of the percentage of tdT⁺ EYFP^neg cells (peak between dash lines) in B220⁺ B cells (L), CD4⁺ T cells (M), CD8α⁺ T cells (N), Ly6C⁺ myeloid cells (O), Ly6C^neg myeloid cells (P), and Ly6G⁺ neutrophils (Q).

Data are represented as median ± quartiles (D). Data are represented as mean ± SEM (H, J, and L-Q). n = 4 mice, Cx3cr1^CreER/+; n = 3 mice, Cx3cr1^CreER/+:R26^tdT/+ and R26^tdT/tdT. ***p < 0.001. One-way ANOVA followed by Tukey’s post hoc test (D and F).

Cx3cr1-controlled gene expression.

We next explored which of these CX3CR1⁺ cells are long-lived that might be responsible for Gi-DREADD-induced hypolocomotion in Cx3cr1^CreER/+:R26^hM4Di/+ mice. To this end, a Cx3cr1^CreER/+:R26^tdT/+ reporter mouse line was used. In this mouse line, CX3CR1⁺ cells express EYFP constitutively but only induce tdTomato (tdT) expression after gene recombination (Fig. 3E). Four weeks after tamoxifen injection in Cx3cr1^CreER/+:R26^tdT/+ reporter mice, we collected blood, spleen, liver, lung, and kidney to identify tdT expressing cells by flow cytometry or immunofluorescence.

In the blood, we observed a small portion of circulating leukocytes expressing tdT in Cx3cr1^CreER/+:R26^tdT/+ mice 4 weeks after tamoxifen (Fig. 3F and G, and Fig. S6). Among them, 0.36 % of blood CD45⁺ cells expressed both tdT and EYFP (Fig. 3H) and tdT⁺ cells composed 11.87 % of EYFP⁺ cells (Fig. 3I). In addition, 2.00 % of CD45⁺ cells were EYFP negative but expressed tdT (tdT⁺ EYFP^neg; Fig. 3J). The tdT⁺ EYFP^neg population mainly consisted of B220⁺ B cells, CD4⁺ and CD8α⁺ T cells (Fig. 3K). Regarding the percentage of tdT⁺ cells in each cell type, tdT was expressed in 2.08 % of B cells (Fig. 3L), 1.62 % of CD4⁺ T cells (Fig. 3M), 3.44 % of CD8α⁺ T cells (Fig. 3N), 4.09 % of Ly6C⁺ myeloid cells (Fig. 3O), 7.15 % of Ly6C^neg myeloid cells (Fig. 3P), and 0.15 % of Ly6G⁺ neutrophils (Fig. 3Q). The results indicate that the gene rearrangement could be found in CX3CR1⁺ cells and potentially the CX3CR1^neg progeny of CX3CR1⁺ cells in the blood 4 weeks after the activation of CreER recombinase.

Similar to the blood, splenic tdT⁺ EYFP⁺ and tdT⁺ EYFP^neg cells were observed in Cx3cr1^CreER/+:R26^tdT/+ mice 4 weeks after tamoxifen injection (Fig. 4A–D, and Fig. S7). The tdT⁺ cells constituted 18.70 % of EYFP⁺ cells (Fig. 4E), and the tdT⁺ EYFP^neg cell population was mainly made up of B220⁺ B cells, CD4⁺ and CD8α⁺ T cells (Fig. 4F and G). In each cell population we tested, tdT labeled 8.24 % of F4/80⁺ macrophages (Fig. 4H), 0.69 % of B220⁺ B cells (Fig. 4I), 0.97 % of CD4⁺ T cells (Fig. 4J), 2.28 % of CD8α⁺ T cells (Fig. 4K), 2.77 % of Ly6C⁺ myeloid cells (Fig. 4L), 3.84 % of Ly6C^neg myeloid cells (Fig. 4M), and 0.01 % of Ly6G⁺ neutrophils (Fig. 4N). These results demonstrate that Cx3cr1-controlled tdT reporter is expressed in splenic immune cells 4 weeks after the induction of gene recombination. In addition, using immunostaining, we observed tdT⁺ Iba1⁺ cells with elongated cytoplasmic processes and round shape tdT⁺ Iba1^neg cells in the liver, lung, and kidney (Fig. 4O). tdT⁺ Iba1⁺ cells made up 18.54 % of total Iba1⁺ macrophages in the liver, 45.99 % in the lung, and 64.63 % in the kidney (Fig. 4P). Together, the comprehensive analysis of tdT reporter mice suggests that a decent amount of cells with Cx3cr1-controlled gene rearrangement persist in blood and peripheral organs.

Figure 4. Cx3cr1-controlled tdTomato (tdT) reporter is expressed in peripheral organs.

(A) Timeline of inducing tdT expression and peripheral organ collection for flow cytometric analysis or immunostaining.

(B) Representative flow plots showing the constitutive expression of EYFP and inducible expression of tdT in splenic CD45⁺ leukocytes in Cx3cr1^CreER/+, Cx3cr1^CreER/+:R26^tdT/+, and R26^tdT/tdT mice 4 weeks after tamoxifen injection.

(C) Quantitative analysis revealing that splenic tdT⁺ EYFP⁺ leukocytes were only observed in Cx3cr1^CreER/+:R26^tdT/+ mice.

(D) Immunostaining showing that tdT (red) is expressed in Iba1⁺ (green; arrow) and round shape Iba1^neg (arrowhead) cells in the spleen. Scale bars: 20 μm.

(E) Histogram and quantification of the percentage of tdT⁺ cells (peak between dash lines) in EYFP⁺ cells.

(F) Quantitative analysis showing that splenic tdT⁺ EYFP^neg leukocytes were only observed in Cx3cr1^CreER/+:R26^tdT/+ mice.

(G) Pie chart showing that the tdT⁺ EYFP^neg cells mainly consisted of B220⁺ B cells, CD4⁺, and CD8α⁺ T cells.

(H-N) Histograms and quantification of the percentage of tdT⁺ EYFP^neg cells (peak between dash lines) in F4/80⁺ macrophages (H), B220⁺ B cells (I), CD4⁺ T cells (J), CD8α⁺ T cells (K), Ly6C⁺ myeloid cells (L), Ly6C^neg myeloid cells (M), and Ly6G⁺ neutrophils (N).

(O) Immunostaining showing that tdT (red) is expressed in Iba1⁺ (green; arrow) and round shape Iba1^neg (arrowhead) cells in the liver, lung, and kidney. Scale bar: 50 μm.

(P) Quantification of the percentage of tdT⁺ cells in Iba1⁺ cells in liver, lung, and kidney.

Data are represented as mean ± SEM. n = 3 mice in each group. **p < 0.01 and ***p < 0.001. One-way ANOVA followed by Tukey’s post hoc test (C and F).

Gi-DREADD is expressed in splenic leukocytes and tissue resident macrophages.

After understanding Cx3cr1-controlled gene rearrangement, we next assessed the Gi-DREADD expression in the circulating immune cells and peripheral organs in Cx3cr1^CreER/+:R26^hM4Di/+ mice. The expression of Gi-DREADD was designed to be identified by the mCitrine reporter and the fused HA tag. We first intended to use mCitrine to surrogate Gi-DREADD expression, but unexpectedly found that the mCitrine expression is independent of the activation of Cre recombinase. Specifically, approximately 40% of CD45⁺ leukocytes in the blood express mCitrine in the R26^hM4Di/hM4Di mice, which is comparable to the Cx3cr1^CreER/+:R26^hM4Di/+ (Fig. 5A–E). Additionally, t-SNE and quantitative analysis found that CX3CR1^neg mCitrine⁺ population includes 61% of CD4⁺ T cells, 12% of Ly6C⁺ myeloid cells, 12% of Ly6G⁺ neutrophils, and 8% of CD8α⁺ T cells, respectively (Fig. S8A–G). Thus, mCitrine is not directly associated with the expression of hM4Di. For example, prevalent mCitrine⁺ splenic cells were observed in both Cx3cr1^CreER/+:R26^hM4Di/+ and R26^hM4Di/hM4Di mice. However, HA is only expressed on the membrane of splenic cells in Cx3cr1^CreER/+:R26^hM4Di/+ mice but not R26^hM4Di/hM4Di mice after tamoxifen induction (Fig. S8H).

Figure 5. Gi-DREADD is expressed in peripheral immune cells.

(A) The genetic strategy of generating Cx3cr1^CreER/+:R26^hM4Di/+ mice to express Gi-DREADD (hM4Di) and mCitrine on CX3CR1⁺ cells.

(B) Timeline of tamoxifen induction of Gi-DREADD expression and blood collection for flow cytometric analysis.

(C) Representative flow cytometry plots examining CX3CR1⁺ EYFP/mCitrine⁺ (double positive) and CX3CR1^neg EYFP/mCitrine⁺ (EYFP/mCitrine⁺ only) cells in circulating CD45⁺ leukocytes in wild type (WT), Cx3cr1^CreER/+, Cx3cr1^CreER/+:R26^hM4Di/+, and R26^hM4Di/hM4Di mice 4 weeks after tamoxifen injection.

(D) Quantitative analysis showing that most of the EYFP⁺ cells were CX3CR1⁺ in the Cx3cr1^CreER/+ mice. A similar double positive population (CX3CR1⁺EYFP⁺) is found in Cx3cr1^CreER/+:R26^hM4Di/+. Surprisingly, R26^hM4Di/hM4Di mice also contain the double positive population (CX3CR1⁺mCitrine⁺) (n = 3 mice, WT; n = 3 mice, Cx3cr1^CreER/+; n = 4 mice, Cx3cr1^CreER/+:R26^hM4Di/+; n = 4 mice, R26^hM4Di/hM4Di).

(E) Quantification demonstrating that a CX3CR1^neg EYFP/mCitrine⁺ population was observed in Cx3cr1^CreER/+:R26^hM4Di/+ mice. Surprisingly again, R26^hM4Di/hM4Di mice have CX3CR1^neg EYFP/mCitrine⁺ population similar to those from Cx3cr1^CreER/+:R26^hM4Di/+ mice. These results suggest the mCitrine expression is leaky prior to the introduction of Cre recombinase (n = 3 mice, WT; n = 3 mice, Cx3cr1^CreER/+; n = 4 mice, Cx3cr1^CreER/+:R26^hM4Di/+; n = 4 mice, R26^hM4Di/hM4Di).

(F) Timeline of tamoxifen induction of Gi-DREADD expression and OFT for Cx3cr1^CreER/+, Cx3cr1^CreER/+:R26^hM4Di/+, and R26^hM4Di/hM4Di mice after 2.5 mg/kg CNO injection.

(G) Representative mouse movement trajectories of Cx3cr1^CreER/+, Cx3cr1^CreER/+:R26^hM4Di/+, and R26^hM4Di/hM4Di mice after 2.5 mg/kg CNO injection in the OFT.

(H and I) Quantification revealing that Cx3cr1^CreER/+:R26^hM4Di/+ mice had less move distance (H) and spent a shorter time in the central area (I) than Cx3cr1^CreER/+ or R26^hM4Di/hM4Di mice (n = 8 mice in each group).

(J-R) Representative images and quantitative analysis of spleen slices demonstrating that HA fused hM4Di (red) was colocalized with Iba1⁺ cells, CD4⁺ T cells (J-L; arrow), CD8α⁺ T cells, Ly6C⁺ myeloid cells (M-O; arrow), Ly6G⁺ neutrophils, and B220⁺ B cells (P-R; arrow) (n = 3 mice in each group). Scale bars: 50 μm (J, upper) and 10 μm (J, lower).

(S) Representative immunostaining of liver, lung, and kidney sections showing that some Iba1⁺ (green) cells in these organs expressed hM4Di (red, identified by fused HA-tag; arrow). Scale bar: 50 μm.

(T) Quantification of the percentage of HA⁺ cells in Iba1⁺ cells in liver, lung, and kidney (n = 4 mice in each group).

Data are represented as mean ± SEM. **p < 0.01, and ***p < 0.001. One-way ANOVA followed by Tukey’s post hoc test (D, E, H, and I).

Considering the potential leakage of mCitrine in R26^hM4Di/hM4Di mice, we further tested their behaviors in the OFT. While Cx3cr1^CreER/+:R26^hM4Di/+ mice exhibited hypolocomotion in OFT after CNO, R26^hM4Di/hM4Di mice have similar normal behaviors as control Cx3cr1^CreER/+ mice (Fig. 5F–I). These results indicate a minimal expression of hM4Di, although there is leakage of mCitrine expression. Therefore, we further used HA to map the expression patterns of hM4Di in the spleen, liver, lung, and kidney. In the spleen, hM4Di was colocalized with 48.71 % Iba1⁺ cells, 16.08 % CD4⁺, 3.58 % CD8α⁺ T cells, 3.87 % Ly6C⁺ myeloid cells, 1.29 % Ly6G⁺ neutrophils and 0.63 % B220⁺ B cells (Fig. 5J–R). Moreover, Iba1⁺ HA⁺ cells comprised 9.51 % of Iba1⁺ cells in the liver, 47.49 % in the lung, and 14.85 % in the kidney (Fig. 5S and T). Together, these results indicate that the Gi-DREADD, which is surrogated by HA tag but not mCitrine, is specifically expressed in splenic leukocytes and tissue resident macrophages.

Hypolocomotion does not depend on Kupffer cells, splenic resident macrophages, or CD4⁺ cells.

We aimed to narrow down the CX3CR1⁺ cell populations responsible for the Gi-DREADD-induced hypolocomotion. Considering that tissue resident macrophages express hM4Di throughout organs, we first depleted resident macrophages by i.p. injection of clodronate (Clod) encapsulated liposomes. PBS encapsulated liposomes were used as the control (Fig. 6A). After Clod treatment, we observed sparse Iba1⁺ Kupffer cells (Fig. 6B and C) and F4/80⁺ splenic macrophages (Fig. 6D–F), suggesting the efficient elimination of macrophages in the liver and spleen. In addition, the i.p. administration of Clod eliminated some renal macrophages but did not affect the number of microglia or pulmonary macrophages (Fig. S8I–N). However, Clod ablation of tissue resident macrophages did not prevent Gi-DREADD-induced hypolocomotion in Cx3cr1^CreER/+:R26^hM4Di/+ mice. Compared with Cx3cr1^CreER/+ mice with macrophage ablation, Cx3cr1^CreER/+:R26^hM4Di/+ mice treated with Clod or PBS moved less and spent a shorter time in the central area in OFT (Fig. 6G–I). Therefore, these results revealed that hM4Di on Kupffer cells and splenic resident macrophages did not mediate Gi-DREADD-induced hypolocomotion.

Figure 6. Resident macrophages or CD4⁺ T cells are not required for Gi-DREADD-induced hypolocomotion.

(A) Timeline of tamoxifen induction of Gi-DREADD expression, peripheral macrophage elimination by clodronate, and OFT for Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection.

(B and C) Immunofluorescent images (B) and quantification (C) showing Iba1⁺ cells are ablated in the liver after Clod treatment (n = 3 mice, PBS; n = 6 mice, Clod). Scale bar: 50 μm (B).

(D-F) Flow plots (D) and quantitative analysis (E and F) demonstrating that F4/80⁺ splenic macrophages were largely ablated by Clod (n = 3 mice, PBS; n = 6 mice, Clod).

(G) Representative mouse movement trajectories of Clod treated Cx3cr1^CreER/+, PBS treated Cx3cr1^CreER/+:R26^hM4Di/+, and Clod treated Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection in the OFT.

(H and I) Quantification revealing that Cx3cr1^CreER/+:R26^hM4Di/+ mice treated with Clod or PBS had less move distance (H) and spent a shorter time in the central area (I) than Cx3cr1^CreER/+ mice had macrophage depletion (n = 3 mice in each group).

(J) Timeline of tamoxifen induction of Gi-DREADD expression, CD4⁺ T cell elimination by anti-CD4, and OFT for Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection.

(K and L) Flow plots (K) and quantitative analysis (L) showing that anti-CD4 efficiently ablated CD4⁺ T cells in the blood (n = 3 mice in each group).

(M) Representative mouse movement trajectories of anti-CD4 treated Cx3cr1^CreER/+, IgG treated Cx3cr1^CreER/+:R26^hM4Di/+, and anti-CD4 treated Cx3cr1^CreER/+:R26^hM4Di/+ mice after 2.5 mg/kg CNO injection in the OFT.

(N and O) Quantification demonstrating that Cx3cr1^CreER/+:R26^hM4Di/+ mice treated with anti-CD4 or IgG traveled less distance (N) and spent less time in the central area (O) than Cx3cr1^CreER/+ mice with anti-CD4 administration (n = 3 mice in each group).

Data are represented as mean ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001. Student’s t-test (C, E, F, and L) or one-way ANOVA followed by Tukey’s post hoc test (H, I, N, and O).

CD4⁺ T cell is another major population expressing hM4Di 4 weeks after tamoxifen injection. To test the dependence between the hypolocomotion and CD4⁺ T cells, we injected CD4 depletion antibody (anti-CD4, i.p.) into Cx3cr1^CreER/+:R26^hM4Di/+ mice and Cx3cr1^CreER/+ mice (Fig. 6J–L). However, Cx3cr1^CreER/+:R26^hM4Di/+ mice still showed robust Gi-DREADD-induced hypolocomotion compared with Cx3cr1^CreER/+ mice in the OFT after CD4⁺ T cell depletion (Fig. 6M–O). These results suggest that the CD4⁺ cells did not mediate this hypolocomotion. Altogether, though various cells have Cx3cr1-controlled hM4Di expression, we rule out that Kupffer cells, splenic resident macrophages, or CD4⁺ cells were responsible for Gi-DREADD-induced hypolocomotor phenotype.

Peripheral immune cells are capable to infiltrate the brain in neurological diseases, such as epilepsy ³¹, stroke ³², and brain injury ³³. To test whether Gi-DREADD activation promotes the brain infiltration of long-lived CX3CR1⁺ cells, we performed flow cytometry to examine the CD45^high Cd11b^high infiltrating immune cells 30 minutes in Cx3cr1^CreER/+:R26^hM4Di/+ mice and Cx3cr1^CreER/+ mice after CNO administration (Fig. S9). However, we did not observe significant alterations of CD4⁺ T cells, CD8α⁺ T cells, B cells, Ly6C⁺ myeloid cells, Ly6C^neg myeloid cells, or Ly6G⁺ neutrophils between the two groups (Fig. S10).

Discussion

In this study, we initially planned to study the function of microglia Gi signaling in vivo by chemogenetic approaches using Cx3cr1^CreER/+:R26^hM4Di/+ mice. Upon Gi-DREADD activation in CX3CR1⁺ cells, Cx3cr1^CreER/+:R26^hM4Di/+ mice dramatically decreased movement and central time in OFT, and decreased travel distance and open arm activity in EPM. Unexpectedly, this hypolocomotion persisted when microglia were depleted. Although the Cx3cr1^CreER/+ mouse line is commonly used for the genetic manipulation of microglia, we should be cautious about the data interpretation, considering the presence of peripheral long-lived CX3CR1⁺ cells.

Peripheral long-lived CX3CR1⁺ cells in Gi-DREADD-induced hypolocomotion.

We demonstrated that Gi-DREADD-induced hypolocomotion is microglia independent and provided a comprehensive analysis of gene recombination in peripheral cells 4 weeks after the tamoxifen injection. Interestingly, it was briefly mentioned in a previous report that the activation of hM4Di could elicit unfavorable sedative behaviors during acute periods when using the Cx3cr1^Cre/+:R26^hM4Di/+ strain to constitutively express hM4Di in all CX3CR1⁺ cells ³⁴. To narrow down the causative cell population, we tested the role of tissue resident macrophages and CD4⁺ T cells with higher portions of hM4Di⁺ cells in hypolocomotor phenotype. However, the Gi-DREADD-induced hypolocomotion still persisted when we ablated these cells by clodronate liposomes or the CD4 depletion antibody.

Of note, i.p. injection of clodronate liposomes could efficiently ablate splenic macrophages, Kupffer cells, macrophages in the peritoneal cavity, and some lymph nodes ³⁵. But clodronate is unable to eliminate brain microglia, renal macrophages, and pulmonary macrophages efficiently. Besides, a small portion of circulating monocytes, CD8⁺ T cells, and B cells express hM4Di and might be clodronate insensitive. Thus, activation of Gi-DREADD in resident macrophages in the kidney, lung, or some circulating CX3CR1⁺ cells could be responsible for the hypolocomotion. Additionally, intramuscular CX3CR1⁺ macrophages play an important role in regulating muscle functions ^{36, 37}. It is possible that intramuscular CX3CR1⁺ resident macrophages attribute to the hypolocomotion by regulating muscle fiber functions. Since our results on rotarod test are similar between the two groups, the muscle function may not be significantly affected by intramuscular macrophage Gi-DREADD activation. It would be interesting for researchers to tease out the causality between these cell populations and hypolocomotion in the future. Nevertheless, our results suggest the hitherto unappreciated importance of peripheral long-lived CX3CR1⁺ cells in locotomtion behaviors.

Considering CX3CR1 is mainly expressed in immune cells and the robust hypolocomotive phenotype is similar to the behavior outcomes of systemic LPS treatment ³⁸, the hypolocomotion may be one aspect of the sickness behaviors. Indeed, it was reported that activation of Gi-GPCR signaling could induce proinflammatory cytokine production through Gβγ subunits in peripheral immune cells ³⁹. Activation of Gi-DREADD on peripheral CX3CR1⁺ cells may promote the release of proinflammatory cytokines and cause hypolocomotion. It would be important for future studies to discern whether the behavioral phenotype is consistent with sickness behaviors by including additional criteria such as body temperature, food and water intake, and cytokine levels in the blood ⁴⁰.

Utilizing chemogenetics to study microglial GPCR-mediated signaling pathways.

The DREADD platform is a powerful tool for understanding and manipulating microglial GPCR-induced signaling pathways ¹⁴. GPCR superfamily consists of 4 subfamilies based on the α subunits of the heterotrimeric G proteins, Gi, Gq, Gs, and G12/13 ⁴¹. Microglia express a plethora of GPCRs, including P2Y12, CX3CR1, C3aR1, P2Y6, β2 adrenergic receptor, and G protein-coupled receptor 56 (GPR56) ⁶. Microglial chemogenetics provides a better method to temporally and specifically manipulate microglial GPCR signaling. Moreover, using viral transfection and stereotaxic microinjection, researchers could utilize DREADD to investigate microglia GPCR functions in a particular region ^{11, 12, 42}.

DREADD could mimic the known function of endogenous microglia GPCRs and help us explore the unknown microglial GPCR-activated signaling pathways. Pharmacological activation of microglial Gi-GPCRs collapses filopodia but triggers large processes extension toward the agonist source ^{16, 17, 43}. Similarly, our study shows that activating Gi-DREADD-induced microglia processes follow CNO gradients and extend toward the CNO source, suggesting that hM4Di could recapitulate the general function of microglia endogenous Gi-GPCRs. In addition, hM4Di could inhibit microglial activation, reduce the release of pro-inflammatory cytokines, and relieve neuropathic pain ^{11, 15, 34, 42}. Collectively, although microglial Gi-DREADD did not mediate CNO-induced hypolocomotion, multiple reports have proven that the tool is useful for interrogating microglial functions. In addition, future studies are needed to address how Gi-DREADD activation in peripheral CX3CR1⁺ cells induces hypolocomotion.

Gq-DREADD (hM3Dq) has also been applied to microglial research. A recent study found that hM3Dq activation increased microglia phagocytotic ability, promoted inflammatory cytokine production, elicited pain, and induced a negative affective state ^{12, 13}. Chronic activation of hM3Dq could also induce microglial immune memory formation ⁴⁴. Microglia reprogram their epigenetics for at least six months after receiving the peripherally inflammatory stimuli ⁴⁵. Thus, DREADD manipulations may help us accurately train microglia through GPCR-mediated pathways and modify microglial innate memory.

Limitations of using Cx3cr1^CreER/+:R26^hM4Di/+ mouse line to manipulate microglia Gi signaling.

Gene rearrangement in circulating, splenic macrophages, and Kupffer cells was studied in Cx3cr1^CreER/+:R26^DsRed/+ or Cx3cr1^CreER/+:R26^yfp/+ reporter lines 22, 23. Although peripheral CX3CR1⁺ cells initially expressed fluorescent proteins, they had a higher turnover rate in the circulation and thus are replaced by cells without gene recombination 4 weeks after tamoxifen injection. In contrast, we observed the prevalent gene recombination throughout peripheral organs in Cx3cr1^CreER/+:R26^tdT/+ mice. These results indicate the presence of long-lived peripheral CX3CR1⁺ cells in addition to brain microglia. Similarly, the expression of Gi-DREADD was observed in peripheral cells in Cx3cr1^CreER/+:R26^hM4Di/+ mice. The expression of transgenes in peripheral cells is also affected by the floxed mice. Two recent studies supported this notion and reported the leakiness of the Cx3cr1^CreER/+ mouse line. Indeed, some Cx3cr1^CreER/+ related reporter strains express fluorescent proteins in the absence of tamoxifen administration ^{46, 47}. Moreover, the leakiness is associated with the length of the floxed STOP cassette. A 0.9 kilobases (kb) DNA fragment is more prone to have tamoxifen independent rearrangement than a longer fragment. In our study, STOP codons for both R26^tdT/+ and R26^hM4Di/+ mouse lines are 0.9 kb. Thus, while peripheral CX3CR1⁺ cells with tamoxifen dependent gene recombination could turnover within 4 weeks, replenished peripheral CX3CR1⁺ cells may still carry transgenes due to the spontaneous CreER activation. Thus, the spontaneous CreER activation and peripheral CX3CR1 expression affect the selectivity of iDTR mediated depletion. Similarly, the ablation approach involved CSF1R inhibiton is known to affect peripheral immune populations ²⁶. Further studies should carefully examine whether peripheral immune cells are involved when using depletion methods to understand microglia functions.

Another limitation of Cx3cr1^CreER/+ mouse line is the induction of gene rearrangement in border associated macrophages (BAMs). BAMs are macrophages residing in the dura mater, subdural meninges, choroid plexus, and perivascular spaces in the CNS ⁴⁸. Unlike microglia, BAMs consist of distinct subsets with specific transcriptional signatures and respond uniquely to CNS damages ⁴⁹. To overcome these limitations, several new mouse lines were recently generated. For example, the Tmem119^{CreER/+ 27}, Hexb^{CreER/+ 50}, P2ry12^{CreER/+ 51}, and Cx3cr1^cCre/+:Sall1^nCre/+ mouse lines ⁴⁹ have minimal non-microglia activity. These efforts to produce more microglia specific tools by targeting other microglia signature genes would help us understand microglial functions with better selectivity.

Lastly, we found Cre independent expression of mCitrine in peripheral immune cells in R26^hM4Di/hM4Di mice and the expression of mCitrine was not correlated with Gi-DREADD expression. Thus, the mCitrine is not appropriate for identifying Gi-DREADD expressing cells. This issue may limit some applications, like fluorescence-activated cell sorting or electrophysiology recording, that need the mCitrine reporter. Interestingly, HA tag can be used to surrogate Gi-DREADD expression. Together, our study suggests that rigorous data analysis and interpretation are needed when using Cx3cr1^CreER/+:R26^hM4Di/+ mouse line to manipulate microglia.

Methods

Animals

Cx3cr1^CreER/CreER (B6.129P2(Cg)-Cx3cr1^{tm2.1(cre/ERT2)Litt}/WganJ, 021160), Tmem119^CreER/CreER (C57BL/6-Tmem119^{em1(cre/ERT2)Gfng}/J, 031820), R26^tdT/tdT (B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J, 007914), and R26^hM4Di/hM4Di (B6.129-Gt(ROSA)26Sor^{tm1(CAG-CHRM4*,-mCitrine)Ute}/J, 026219) mouse lines were obtained from the Jackson Laboratory, and then bred at Mayo Clinic. All adult mice received tamoxifen (Sigma, 150 mg/kg, i.p.) 4 times every 2 days to induce Cre-Lox recombination or serve as control. Both sexes were used in the studies. Mice were group housed in a 12-hour light/dark cycle and climate-controlled environment. Food and water were given ad libitum. All experimental procedures were approved by the Mayo Clinic’s Institutional Animal Care and Use Committee (IACUC).

Brain slice preparation

Freshly isolated cortical slices were prepared from 5–10 weeks old mice. Briefly, mice were anesthetized and swiftly decapitated. Then, brain was carefully placed in ice-cold oxygenated (95 % O₂ and 5 % CO₂) sucrose cutting solution (185 mM sucrose, 2.5 mM KCl, 1.2 mM NaH₂PO₄, 25 mM NaHCO₃, 25 mM glucose, 10 mM MgSO4, 0.5 mM CaCl₂; 290–300 mOsM, 7.3–7.4 pH). Coronal slices (400 μm) were subsequently transferred to a recovery chamber for 30 minutes with oxygenated artificial cerebrospinal fluid (ACSF; 126 mM NaCl, 2.5 mM KCl, 1 mM NaH₂PO₄, 26 mM NaHCO₃, 10.5 mM glucose, 1.3 mM MgSO₄, 2 mM CaCl₂; 300–305 mOsM, 7.35–7.4 pH) at 37 °C and cooled down to room temperature before 2-photon imaging.

Two-photon imaging

Experiments were performed at room temperature, and slices were maintained in oxygenated ACSF at a flow rate of 1–3 mL/minute. A pipette containing 5 mM CNO or aCSF was moved beside microglia using a micromanipulator (Scientifica). A CNO gradient was created by passive diffusion of CNO molecules without puffing. The pipette was visualized by 0.05 mg/mL Alexa Fluor 594 carboxylic acid (Invitrogen). EYFP labeled microglia were imaged by a multiphoton microscope (Scientifica) equipped with a Mai-Tai DeepSee laser (Spectra Physics) tuned to 920 nm with a 16x water immersion lens (NA: 0.8, Nikon). EYFP signal was passed through a 525/50 filter, and Alexa Fluor 594 signal was passed through a 630/75 filter (Chroma). The laser power was maintained at 6 mW or below. Images were collected at 1 Hz frame rate at 512 × 512 pixel resolution for 15 minutes. To quantify the microglial responses to CNO gradients, we measured the number of microglial processes entering from the outer area ($R_o(\mathit{0})$, 70 mm in diameter) into the inner area ($R_i(t)$, 35 mm in diameter) surrounding the tip of the pipette over time ⁵². The microglial response at any time point is given by $R(t)=(R_i(t)-R_i(\mathit{0}))/R_o(\mathit{0})$.

Behavioral tests

Mice were transferred to the test room at least 1 hour before the behavioral measurements started. Behavior tests were performed in the light phase. Clozapine N-oxide (CNO, Cayman chemical, 1.0 mg/kg or 2.5 mg/kg, i.p.) was administrated to all groups 30 minutes before the following tests unless otherwise noted.

Open Field Test (OFT).

The ENV-510 test chambers (27.3 cm × 27.3 cm × 20.3 cm, Med Associates) equipped with infrared photo beams were used to evaluate spontaneous mouse locomotor activity. Mice were placed in the center of the arena to begin the 10-minute exploration. Mouse movement was tracked and analyzed by the activity monitoring software (Med Associates) for the total travel distance and the time spent in the central zone (13 cm × 13 cm).

Elevated plus maze (EPM).

The maze, consisting of two oppositely positioned open arms (35 cm × 6 cm), two oppositely positioned closed arms (35 cm × 6 cm × 22 cm), and a connecting central zone (6 cm × 6 cm), was elevated 74 cm above the floor (Med Associates). Mice were initially placed at the junction of the 4 arms, facing an open arm. Over the course of a 5-minute free exploration, mouse movement was recorded by an overhead camera, and the total travel distance, the duration, and entries in the open arms were analyzed by the monitor software (Ethovision-XT, Noldus).

Rotarod.

Mouse locomotor function was tested using a 5-lane rotarod treadmill (Ugo Basile, Stoelting). Mice were placed on the rotating drum at a speed of 4 rounds per minute, and the speed uniformly accelerated to 40 rounds per minute over a 5-minute testing period. The latency to fall off the drum was measured.

Homecage locomotor activity.

The homecage environment was chosen to minimize the anxious stimuli during the locomotion behavioral test. Mouse movement was recorded by an overhead camera. Total moving trajectory was manually tracked and moving distance was calculated by ImageJ (National Institutes of Health).

Immunofluorescence staining

Isoflurane anesthetized mice were perfused transcardially with 40 mL 0.1 M phosphate-buffered saline (PBS) followed by 40 mL of cold 4% paraformaldehyde (PFA) in PBS. The whole brain, spleen, lung, kidney, and liver were harvested, kept in 4% PFA overnight at 4 °C, and cryoprotected in 30% sucrose in PBS for at least 48 hours before cryosection. Mounted tissue sections (15 μm) and free-floating sections (30 μm) were obtained by a cryostat (Leica). For immunostaining, sections were washed 3 times in the tris-buffered saline (TBS), blocked with TBS buffer containing 5% goat or donkey serum and 0.4% Triton X-100 at room temperature for 1 hour, and then incubated overnight at 4 °C with primary antibodies or conjugated antibodies: rabbit anti-c-fos (1:500, Cell Signaling Technology, 2250S), rabbit anti-HA (1:200, Cell Signaling Technology, 3724S), rabbit anti-Iba1 (1:1000, Abcam, Ab178847), goat anti-Iba1 (1:500, Wako, 011–27991), mouse anti-NeuN (1:500, Abcam, Ab104224), mouse anti-GFAP (1:500, Cell Signaling Technology, 3670S), chicken anti-GFP (1:1000, Aves Labs, GFP-1010), PE rat anti-CD4 (1:200, BioLegend, 100408 ), PerCP rat anti-Ly6C (1:100, BioLegend, 128028), APC-Cy7 rat anti-CD8α (1:200, Tonbo Bioscience, 25–0081-U100), or APC rat anti-Ly6G (1:200, BioLegend, 127614). The following day, sections were washed with TBS 3 times and incubated with goat or donkey secondary antibodies (1:500, Alexa-Fluor 488/555/594/647 anti-rabbit, anti-mouse, or anti-goat, Invitrogen, A11008, A11037, A21428, A21246, A11029, A21447, A21207, and A21206 or 1:500, Alexa-Fluor 488 anti-chicken, Jackson ImmunoResearch, 703–545-155) in blocking buffer for 1200 min at room temperature. After that, sections were washed and mounted with DAPI Fluoromount-G mounting medium (SouthernBiotech). Fluorescent images were obtained with an inverted fluorescence microscope (BZ-X, Keyence) or a laser scanning confocal microscope (LSM 980, Zeiss). Cell counting was manually quantified, and brightness and contrast were adjusted by ImageJ (National Institutes of Health).

Intracerebroventricular (i.c.v.) injections

Mice were anesthetized with isoflurane. A 26 Ga cannula was placed dorsal to the lateral ventricle (from the bregma: −0.2 mm anteroposterior, +1.0 mm mediolateral, −2.0 dorsoventrial). Four days after cannula implantation, mice received either 1 μL minocycline (2μg/μL in sterile PBS; Sigma, M9511) or sterile PBS daily for next 4 days.

Cell ablation

Microglia ablation.

Chow containing colony-stimulating factor 1 receptor (CSF1R) inhibitor, PLX3397 (600 mg/kg, Chemgood), was given ad libitum at least for 2 weeks to fully deplete microglia. Microglial ablation efficiency was assessed by immunostaining brain sections.

CD4⁺ T cell depletion.

Mouse CD4 depletion antibody (rat anti-CD4, clone: GK1.5, BioXCell) or control rat IgG (Sigma, I4131) was i.p. injected on days −2 and −1 relative to behavior tests, 200 μg per animal per day. PerCP rat anti-CD45 antibody (1:1000, clone: 30-F11, BioLegend, 103130), PE-Cy7 rat anti-TCRβ antibody (1:200, clone: H57–597, Tonbo Bioscience, 60–5961-U100), and APC rat anti-CD4 antibody (1:100, clone: RM4–5, BioLegend, 100516) was used to identify the CD4⁺ T cells in the blood. Depletion efficiency was assessed by spectral flow cytometry (Cytek Aurora, Cytek Biosciences).

Macrophage depletion.

Clodronate-containing liposomes or PBS-containing liposomes (Encapsula Nanosciences) were i.p. injected on days −3 and −1 relative to behavioral measurement, 200 μL per animal per injection. PE-CF594 rat anti-CD45 antibody (1:1000, clone: 30-F11, BD Bioscience, 562420), PE-Cy5 rat anti-CD11b antibody (1:500, clone: M1/70, Tonbo Bioscience, 55–0112-U100), APC rat anti-Ly6G antibody (1:100, clone: 1A8, BioLegend, 127614), and BUV395 rat anti-F4/80 antibody (1:100, clone: T45–2342, BD Bioscience, 565614) was used to identify the splenic macrophages. Depletion efficiency was assessed by spectral flow cytometry (Cytek Aurora, Cytek Biosciences). The ablation efficiencies of Kupffer cells, renal macrophages, pulmonary macrophages, and microglia were assessed by immunostaining liver, kidney, lung, and brain sections, respectively. Clodronate liposomes are only able to deplete macrophages when they can be reached. Tissues can form barriers for liposomes. Thus, we observed lower ablation efficiency in the brain, lung, and kidney (Fig. S8I–N) ⁵³.

Isolation of immune cells from blood, spleen and brain

Blood (80 μL) was collected from the tail vein and immediately placed in the flow cytometry tubes containing 1 mL heparin solution (Sigma). Subsequently, cells were spined down by a centrifuge (Thermo Scientific), and red blood cells were lysed by 2 mL ammonium-chloride-potassium (ACK) lysis buffer (8.3 g ammonium chloride, 1 g potassium bicarbonate, and 37.2 mg EDTA per 1 L water, pH 7.2–7.4) for 3 minutes. The reaction was terminated by diluting ACK with PBS, and cells were washed twice at 400 g with PBS. After the final wash, the supernatant was removed, and the pellet was resuspended in 100 μL PBS for further flow cytometric staining.

Before harvesting the spleens and brains, mice were anesthetized by isoflurane and transcardially perfused with 40 mL 0.1 M phosphate-buffered saline (PBS). Then, spleens were dissected, weighed, and dissociated using the rubber end of a syringe in RPMI 1640 (Corning). Brains were processed by dounce homogenization followed by 7840 g centrifugation on a 30% Percoll (Sigma, P1644–1L) for 30 minutes. Subsequently, dissociated cells were washed once at 400 g with RPMI 1640 in 50 mL tubes. Next, pellets were transferred to the flow cytometry tubes, and 0.6 mL of ACK lysis buffer was added to samples for 30 seconds to lyse erythrocytes. Finally, PBS was added to quench the reaction, and samples were washed one more time with PBS prior to staining for flow cytometric analysis. Cells from the spleen or brain were counted in a hemocytometer (Hausser Scientific) using 0.4% trypan blue solution (Gibco).

Flow cytometry

Dissociated cell samples were stained with combinations of surface antibodies together with Fc blocking antibody, rat anti-CD16/CD32 (1:100, clone: 2.4G2, BD Pharmingen, 553142). Spark NIR 685 rat anti-B220 (1:200, clone: RA3–6B2, BioLegend, 103268), BV510 rat anti-CD4 (1:100, clone: GK1.5, BioLegend, 100449), PE rat anti-CD4 (1:500, clone: GK1.5, BioLegend, 100408), BV785 rat anti-CD8α (1:200, clone: 53–6.7, BioLegend, 100750), APC-eFlour 780 rat anti-Cd11b (1:500, clone: M1/70, Invitrogen, 47–0112-82), PE-Cy5 rat anti-CD11b (1:500, clone: M1/70, Tonbo Bioscience, 55–0112-U100), BV650 rat anti-CD11b (1:500, clone: M1/70, BioLegend, 101239), PerCP-Cy5.5 rat anti-CD45 (1:1000, clone: 30-F11, BioLegend, 103132), PE-CF594 rat anti-CD45 (1:1000, clone: 30-F11, BD Bioscience, 562420), Pacific Blue rat anti-CX3CR1 (1:100, clone: SA011F11, BioLegend, 149038), BUV395 rat anti-F4/80 (1:100, clone: T45–2342, BD Bioscience, 565614), BV650 rat anti-Ly6C (1:100, clone: HK1.4, BioLegend, 128049), PerCP rat anti-Ly6C (1:100, clone: HK1.4, BioLegend, 128028), APC rat anti-Ly6G (1:100, clone: 1A8, BioLegend, 127614), or PE-Cy7 rat anti-TCRβ (1:200, clone: H57–597, Tonbo Bioscience, 60–5961-U100) antibodies were used to stain cells from spleen and blood. Ghost dye red 780 (1:1000, Tonbo Bioscience, 13–0865-T100), Zombie UV (1:1000, BioLegend, 423107), or Zombie NIR viability dye (1:1000, BioLegend, 77184) was used to label dead cells. Single stain reference controls were prepared at the same time for spectral deconvolution. Next, cell populations in each sample were assessed by a spectral flow cytometer (Cytek Aurora, Cytek Biosciences) equipped with SpectroFlo software (Cytek Biosciences). Acquired flow cytometry results were analyzed by FlowJo software (BD Life Sciences).

Single-cell RNA sequencing database query

The Tabula Muris database was used to determine the Cx3cr1 expression in 20 mouse organs at the single-cell level ³⁰. All indexed R objects were downloaded from Figshare (https://figshare.com/articles/dataset/Robject_files_for_tissues_processed_by_Seurat/5821263), updated to the newest Seurat v4.0 object ⁵⁴ by UpdateSeuratObject function, and merged into one Seurat object by the merge function. The gene counts were then normalized by NormalizeData function with a scale factor of 1e6, which equaled to the log₂CPM transformation, and then scaled by ScaleData function. Highly variable features were called by FindVariableFeatures function. Dimensionality reduction was performed first by principal component analysis (PCA) followed by t-distributed stochastic neighbor embedding (tSNE) or uniform manifold approximation and projection (UMAP). Cx3cr1 expressing cells, as well as cells of individual organs, were subsetted from the original Seurat object and went through the same pipeline above. Dimplot was used to plot tSNE or UMAP plots. FeaturePlot was used to plot Cx3cr1 expression level, with the max.cutoff parameter set to 1.

The single-cell RNA sequencing database from Zhan et al. was used to examine the expression of Tmem119 in microglia resistant to PLX treatment ²⁹. Raw sequencing data were downloaded from the GEO database under the series number GSE150169. Data were normalized and scaled as described above. Harmony package was used for data integration of different experiment groups ⁵⁵. The MAC2-positive progenitor-like microglial population was subset out by the expression of the MAC2 encoding gene Lgals3. FeaturePlot was used to plot Tmem119 expression level, with the max.cutoff parameter set to 1.

Statistics

The statistical details for each specific experiment, including sample size and statistical methods, are described in the figure legends. The sample size for each experiment is comparable to similar studies with these experimental paradigms ^{12, 18}. The statistical analysis started by testing the normality, and all data followed the Gaussian distribution. Unpaired 2-tailed Student’s t-test was used for statistical comparison of experiments with 2 groups. When comparing more than 2 groups in an experiment, a one-way ANOVA was performed, followed by Tukey’s post hoc analysis for multiple comparisons. Gi-DREADD-induced chemotaxis was analyzed by two-way ANOVA. Results are illustrated as mean ± standard error of the mean (SEM). Statistically significant was determined when p < 0.05. Statistical analyses were conducted using GraphPad Prism 8 software.