Macrophages boost muscle spindles with glutamate for better movement

Mice

Animal work was conducted according to UK Home Office licence legislation under the Animals (Scientific Procedures) Act 1986, with Local Ethical Review by the Imperial College London Animal Welfare and Ethical Review Body Standing Committee or, under the EU and Danish legislation to ensure the animals’ well-being. Mice were maintained under standard housing conditions on a 12 h light/dark cycle with food and water provided, at a constant room temperature and humidity (21–24 °C and 45–65%, respectively). Details of the wild-type and transgenic mouse lines are described:

C57BL6/J mice (Charles River Laboratories); B6.129P2(Cg)-Cx3cr1 tm1Litt/J (Jackson laboratories) known as CX3CR1^GFP mice were provided by M. Malcangio; B6J.B6N(Cg)-Cx3cr1tm1.1(cre)Jung/J), referred to as CX3CR1^Cre (code 025524, Jackson Laboratories); B6;129S-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J), referred to as ChR2^YFP (code 012569, Jackson Laboratories); B6;129S-Gt(ROSA)26Sortm39(CAG-hop/EYFP)Hze/J), known as NpHR3^YFP (code 014539, Jackson Laboratories) and B6;129P2-Pvalbtm1(cre)Arbr/J), known as Pv^Cre (code 008069, Jackson Laboratories).

All male and female mice were used between 8 and 12 weeks of age were used for all experiments. Transgenic lines were screened by PCR analysis for the presence of the transgenes in genomic DNA purified from ear biopsies. Anaesthesia by isoflurane inhalation or intraperitoneal injection of ketamine (100 mg kg⁻¹)/xylazine (10 mg kg⁻¹) was used to kill the mice to minimize suffering. Cervical dislocation was performed following anaesthesia to ensure death. Experimenters were blinded to experimental groups during scoring and quantifications.

Immunofluorescence

Mice were anaesthetized by receiving an intraperitoneal injection of ketamine (100 mg kg⁻¹) and xylazine (10 mg kg⁻¹). Once unconscious, the thoracic cavity was opened to expose the heart and a perfusion needle was directly inserted into the lower ventricle. The mouse was perfused with 10 ml of Dulbecco’s phosphate-buffered saline (DPBS, pH 7.4, Sigma) followed by 20 ml of 4% paraformaldehyde (PFA). Mouse hindlimb skeletal muscles including extensor digitorum longus (EDL), soleus, tibialis anterior and gastrocnemius (Gastro) muscles, sciatic nerves and DRGs were fixed in 4% PFA overnight at 4 °C, then transferred to 30% sucrose for 3 days. Fresh human muscle biopsy was collected from deltoid muscles and snap frozen by immersion in isopentane precooled in liquid nitrogen. Mouse or human tissues were embedded in Tissue-Tek O.C.T. and sliced into 20–50 μm sections on a cryostat and collected on Superfrost plus slides (Fisherbrand). Human muscle sections were prefixed with Acetone at 4 °C for 30 min. Tissue sections were dried for 30 min at room temperature and rinsed with washing buffer (PBS with Tween (PBST), 0.1% Tween in DPBS) three times (5 min each time) before incubation in blocking solution (10% normal donkey serum (NDS) + 0.3 or 1% Triton X-100 in PBST) for 1 h at room temperature to reduce non-specific binding. Tissue sections were then incubated in primary antibody solution (10% NDS + 0.3 or 1% Triton X-100 in PBST, mixed primary antibodies) at room temperature or 4 °C overnight. The next day, after washing with PBST three times, tissue sections were incubated in secondary antibody solution (10% NDS + 1% Triton X-100 in PBST + mixed secondary antibodies) for 1.5–2 h at room temperature. Tissue sections were washed three times and mounted with Prolong Glass Antifade mounting solution (Life Technologies) and coverslips. A list of the antibodies used in this study is provided in Supplementary Table 7.

Microscopy and image analysis

Images were acquired on SP8 confocal microscope (Lecia) at ×20 or ×40 magnification and processed with the LAS-AM Lecia software (Lecia) and ImageJ. A minimum of four MS, NMS and NMJ areas were imaged per biological replicate with a minimum of three biological replicates. The number of macrophages and the area of MS, NMS and NMJ were counted and measured in ImageJ. Three-dimensional (3D) imaging was conducted in ImageJ using the 3D Viewer and 3D Project plugins.

Laser-capture microdissection

To reduce RNase-induced RNA degradation, RNaseZap (Thermo Fisher) was used to clean the hood, microscope and all dissection tools. Mice were anaesthetized by intraperitoneally injecting ketamine (100 mg kg⁻¹) and xylazine (10 mg kg⁻¹) in filtered saline and then perfused with 20 ml of sterilized cold DPBS. Muscles were dissected and directly snap frozen in precooled 2-methylbutane for 15 s (for EDL and soleus muscles) or 35 s (for tibialis anterior and Gastro muscles) and embedded in Tissue-Tek O.C.T. Muscles were cut into 30 μm sections with CryoJane cryostat and attached to Lecia PEN-membrane slides (pretreated with RNaseZap and washed with nuclease-free water, exposed to ultraviolet light for 1 h). Slides were immediately fixed with cold 75% ethanol for 1 min and processed for staining protocol (washing with 75% ethanol: 30 s three times, staining with 4% Cresyl Violet: 1 min, washing and dehydrating with 75% ethanol: 30 s twice, 95% ethanol: 30 s, 100% ethanol: 30 s and Xylene: 5 min). Stained muscle sections were placed on Laser Microdissection with Zeiss Palm Microbeam microscopy. MS and NMS areas were selectively captured and collected into laser-capture microdissection collection tubes (with 150 μl of RLT lysis buffer, Qiagen). Collection tubes were then frozen with dry ice and kept at −80 °C for RNA extraction.

FACS of macrophages

To sort cells in vivo, mice were anaesthetized with ketamine (80 mg kg⁻¹) and xylazine (10 mg kg⁻¹). Muscle, heart and lung were dissected and kept in cold DPBS on ice after cardiac perfusion (20 ml DPBS). For muscle macrophages, to reduce the contamination of NMS macrophages, fat, perimysium, nerves and tendons were carefully removed from dissected muscles. Muscle (EDL, soleus, tibialis anterior and gastro), heart and lung were transferred to digestion buffers and cut into small cubes. Digestion buffer contained Collagenase B (2 mg ml⁻¹), Dispase II (0.83 mg ml⁻¹), DNase I (0.25 mg ml⁻¹) and DNase I buffer (10 nM Tris Base, 2.5 mM MgCl₂, 0.1 mM CaCl₂), dissolved in RNase-free DPBS. Tissue suspensions were incubated in 15 ml of digestion buffer in a 37 °C water bath for 50 min. Tubes were shaken every 5 min. Then 10 ml of ice-cold MojoSort buffer (BioLegend) was added to stop digestion. Cell suspension was filtered through 70 and 40 μm cell strainers in series. Then, cells were washed with 5 ml of cold MojoSort buffer twice and resuspended with 800 μl of cold MojoSort buffer for FACS staining. Cell suspension was incubated with mixed antibodies at 4 °C for 20 min in the dark. The following antibodies were used: PE-conjugated anti-CD45 (Biolegend, catalogue no. 103106, 1:100), APC/Cyaine7-conjugated anti-CD11b (Biolegend, catalogue no. 101226, 1:100), APC-conjugated anti-F4/80 (Biolegend, catalogue no. 123115, 1:100), Brilliant Violet-conjugated anti-CX3CR1 (Biolegend, catalogue no. 149027, 1:100). LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific, catalogue no. L34966, 1:200) was used to identify live or dead cells. Stained cells were washed twice with 1 ml of cold MojoSort, resuspended with 600 μl of MojoSort buffer (with RNase inhibitor, 1:500), filtered through a 40 μm cell strainer and kept on ice until sorting. Becton Dickinson FACS Aria Fusion Flow Cytometer with a 100 μm nozzle was set to 4 °C to protect RNA from degradation. The gating was set to FSC-A/SSC-A-Singlet/Doublet-Live/Dead⁻-CD45⁺-CD11b⁺-F4/80⁺-CX3CR1⁺. Single stain, Fluorescence Minus One and negative controls were used for the gating boundaries. Target cells were directly collected into the cold collection buffer (DPBS with RNase inhibitor, 1:500). Around 80,000 cells were collected from each sample. Cell suspension was centrifuged at 500g for 5 min, resuspended with RLT lysis buffer and processed for RNA extraction for bulk RNA-seq, or resuspended with cold DPBS for single-cell RNA-seq or for cell culture. FlowJo was used for data analysis.

RNA-seq preparation and analyses

Laser-capture micro-dissected muscle tissues or macrophages sorted from muscles, hearts and lungs were processed for RNA extraction. RNA was extracted by using RNeasy micro-kit and DNase on-column digestion (Qiagen) following the manufacturer’s guidelines. RNA purity and concentration were verified by Agilent 2100 Bioanalyzer. RNA with an RNA integrity number above 8 was used for library preparation. Complementary DNA (cDNA) libraries for each sample were generated at the Source Bioscience Laboratory (Cambridge, UK) by using NEBNext Single Cell/Low Input Library Prep Kit (Ilumina). Sequencing was performed using Illumina HiSeq 4000 75 base pairs (bp) paired-end sequencing to obtain 60 million reads per sample. SNMP RNA-seq datasets (Fastq) were extracted from a published dataset GSE144708 and were processed with the same pipeline as the FACS-sorted macrophages from muscles, heart and lung. Sequences were demultiplexed and adapters trimmed with bcl2fastq-v.2.20 (Illumina). Read quality controls were carried out using FastQC-v.0.11.9, trimming reads less than Phred33 quality score 20 and removing remaining adapters with Trim-Galore (v.0.6.6). RNA-seq analysis was run using the COMBINE laboratory’s Salmon-DESeq2 pipeline. Salmon-v.1.6.0 was used in mapping-based mode to quantify read sets. Reads were mapped to the M25 GENCODE reference mouse genome. Salmon’s output files were imported and converted by Tximeta (v.1.12.4) for DESeq2. Differential expression analysis was performed using DESeq2 (v.1.34.0) in R (v.4.1.2). Normalized counts were generated using the median of ratios method in DESeq2. Differentially expressed genes were filtered to include genes with a fold change greater than 1.5. GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were carried out using the latest available DAVID (Database for Annotation, Visualization and Integrated Discovery) for all differentially expressed genes. GO and KEGG terms were filtered by P 

Loss of function experiments

Cre-dependent perturbation

We used a Cre-dependent viral strategy to selectively silence vesicle release from a CX3CR1⁺ cell of the gastrocnemius muscles. We injected 25 μl of AAV9/2-DJ/2-FLEX-mCherry-2A-FLAG-TeTxLC (Viral Vector Facility, Universität Zürich) in both gastrocnemius muscles of CX3CR1^Cre::ChR2^YFP mice to prevent the release of SNARE vesicles. We injected AAV9/2-DJ/2-flex-mCherry (Viral Vector Facility, Universität Zürich) in another group of animals (the control group). Similarly, we also used a Cre-dependent depletion strategy as an alternative by injecting in the same muscles (gastrocnemius) AAV9/2-DJ/2-Flex-(pro)taCasp3 (Viral Vector Facility, Universität Zürich) in a group of mice defined as Cas3. In this way, the Cre⁺ cells of the gastrocnemius muscle were depleted as previously shown for other types of cell including neurons and microglia^50,51. Five weeks later, a cohort of mice from each group were euthanized and perfused. Muscles were dissected, fixed and processed for immunostaining. The control group, the TeLC group and the Cas3 group underwent behavioural testing 5 weeks after injection. Furthermore, some of the animals from the control and TeLC groups were tested with optical stimulation of MSMP in the muscle while recording tibialis and gastrocnemius.

Systemic depletion of macrophages

CSF1 inhibitor, BLZ945 (4 mg per mouse per day, S7725, Selleckchem) dissolved in sterile water with 20% 2-hydroxypropyl-β-cyclodextrin (H107, Sigma) or control with 20% 2-hydroxypropyl-β-cyclodextrin were delivered to the mice for four consecutive days by oral gavage. Clodronate liposomes or control liposomes (Liposoma) were injected into mice by intraperitoneal injection (200 μl per mouse per day) on days 1 and 3. Depleted or control mice were euthanized on day 4 (4 h after the last BLZ945 gavage) or day 11. Muscles together with spinal cords and sciatic nerves were dissected and freshly processed for FACS or fixed with 4% PFA and dehydrated with 30% sucrose for cryostat sectioning followed by immunostaining. Mice without or with macrophage depletion (oral gavage of control or BLZ945 in combination with control or clodronate liposomes) were used for behavioural assessments. Mouse hindlimbs were bilaterally assessed on day 11 after depletion. Animals were randomly and equally grouped before depletion treatments (15 biological replicates for each condition). All investigators were blind to the groups and treatments.

Behavioural tests

Grid walk

Mice were trained to walk through the metal grid (50 × 5 cm) plastic grid (1 × 1 cm) placed between two vertical 40 cm high wood blocks. Three runs per session were carried out and measured. For each hind paw, the total number of steps and missteps per run were analysed by a blinded investigator.

Treadmill

Mice were trained to run on a motorized treadmill at two different speeds (20 and 40 cm s⁻¹). Mice were placed on the treadmill and their paw prints were captured by a video camera with 147 frames per second) mounted inside the treadmill chassis beneath the transparent belt (DigiGait lmaging system, Mouse Specifics Inc.). Each mouse, running at two different speeds, was recorded for a minimum of four consecutive strides. After processing all video frames, gait signals for each paw were filtered and 39 gait features were calculated to quantify the paw pattern alternation in mice. For each gait feature (variable)’s dataset, normality and equality of variance were assessed. Variables conforming to a normal distribution were analysed with unpaired Student’s t-tests, and those from non-normally distributed datasets were analysed with the Mann–Whitney test. A P value less than 0.05 was considered statistically significant (IBM SPSS Statistics v.26.0). The Cohen’s effect size quantified the size of the mean difference between two groups by comparing it to the variability of the dataset. A Cohen’s value below 0.2 indicated a weak effect, a value between 0.2 and 1 indicated a moderate effect and a value greater than 1 signified a strong effect⁵².

Swimming

The swimming ability of mice was evaluated using a transparent Plexiglas chamber (90 × 14.5 × 30 cm) filled with tap water, maintained at a constant temperature of 23 °C. Platforms were positioned at both ends of the chamber to allow the animals to rest before and after each trial. A single run spanned a unidirectional course from the start point to the endpoint, roughly 60 cm in length. A run was deemed complete if the animal moved from one platform to the other in less than 5 s. Runs not completed within this timeframe were stopped and recorded as incomplete. Each animal completed a minimum of three valid runs, with a 5 min rest period between runs. After testing, the animals were manually dried with tissue, then placed in a single cage with a heating pad at 37.2 °C for 20 min to ensure complete drying. The animals’ movements were recorded at 200 FPS with two cameras (model ACA640-90UC, Basler) located outside the chamber, providing lateral and ventral views.

Overground walking

Locomotor performance was tested in a transparent corridor (120 cm long and 10 cm wide) with 20 cm high lateral walls. Animal movements were captured at 200 FPS using two cameras (model ACA640-90UC, Basler) to obtain ventral and lateral views, as previously demonstrated⁵³. Each animal traversed the corridor from one end to the other, constituting a run. Each animal was required to complete a minimum of three runs without stops. To prevent fatigue and maintain consistency, particularly in speed, the same experimenter handled all animals throughout the testing process.

Ladder

The evaluation of skilled walking in mice was conducted using a horizontal ladder, 120 cm long and 10 cm wide, with evenly spaced rungs set 2 cm apart. This setup challenged the mice to execute precise and coordinated movements. The animals’ movements were recorded from ventral and side views by two cameras. Each animal repeated the ladder walk three times, totalling a maximum of 180 steps.

Locomotor analysis

The swimming and walking abilities were assessed through video recordings, subsequently tracked using marker-less pose estimation technology based on deep-learning and convolutional neural networks⁵⁴. This process involved identifying major key points on the mouse’s anatomy, specifically two on the head, five on the trunk and four on the tail. Furthermore, kinematics of the limbs was detailed using four key points for each forelimb (shoulder, elbow, wrist, paw) and five for each hindlimb (hip, knee, ankle, paw, toes). These points on the limbs formed interconnected chains of segments. From these key points, kinematic analysis of the hindlimb and forelimb generated a comprehensive set of 37 gait parameters using custom MATLAB scripts, which parsed the pose estimation data. To discern differences across experimental conditions and identify relevant parameters explaining these differences, a multifactorial analysis was used. For each experimental dataset, PCA commenced by constructing the covariance matrix X from all gait parameters, which were normalized to avoid bias towards parameters with larger values by subtracting their mean and dividing by the standard deviation. The eigenvectors and eigenvalues of X were then calculated, representing the principal components and the variance each component explained, respectively. The principal components were sorted by decreasing order of variance, with the first few typically capturing the most substantial portion of the variance in the dataset. The significant principal components served as the dependent variables in subsequent analyses to assess differences between experimental conditions, using analysis of variance (ANOVA) and post hoc tests where appropriate.

DRG neuronal nuclei sorting for single-nucleus RNA-seq

L4-6 DRG from control or macrophage-depleted mice were euthanized, dissected and snap frozen in liquid nitrogen on day 4. On the day of sorting, DRG were placed into homogenization buffer (0.25 M sucrose, 25 mM KCl, 5 mM MgCl₂, 20 mM tricine-KOH pH 7.8, 5 μg ml⁻¹ actinomycin, 1% BSA, 0.15 mM spermine, 0.5 mM spermidine, EDTA-free protease inhibitor, phosphatase inhibitor, RNase inhibitor). DRGs were homogenized by pelleting with a plastic pestle for 15 s. Trition-X-100 (Sigma) was added to the homogenization mixture to reach the final concentration of 0.1%. Another 15 s stroke was added to further homogenize the DRGs. Finally, DRG homogenization suspension was filtered through 70 μm followed by 40 μm cell strainers and incubated with NeuN-Alexa Fluor488 antibody (MAB377X Clone A60, 1:100) for 1 h in cold room with gentle rotation. After washing with washing buffer (homogenization buffer + 0.1% Triton X-100 + 2% BSA) three times, DRG nuclei were resuspended in 2 ml of washing buffer and incubated with DAPI (D5942, Sigma) for 10 min. DAPI⁺/NeuN⁺ DRG neuronal nuclei were sorted by Aria III sorter and collected in washing buffer for 10X single-nucleus RNA-seq according to 10X Chromium Next GEM Single Cell 3′ Reagent Kits v.3.1 (Dual Index) protocol.

Optogenetic stimulation of neurons (Pv⁺) or macrophages (CX3CR1⁺)

To specifically activate ChR2 in either proprioceptive neurons or macrophages, we used mice expressing Channel rhodopsin (Ai32, Jackson Laboratories) under the parvalbumin (Pv^Cre, Jackson Laboratories) or the macrophage (Cx3cr1^Cre, Jackson Laboratories) promoters. Before surgery, the animals were anaesthetized with isoflurane (induction 4% and maintenance 2%) and an eye ointment was applied to prevent eyes dehydration. The animal was placed on a heating pad (Stoelting Warming Systems) and the anaesthesia’s depth was verified through the absence of reflexes and tail pinch responses, ensuring the animals were in a controlled and unresponsive state throughout the whole procedure. A sizable incision was made to expose the sciatic nerve at the thigh level (for stimulation of sensory neurons) and to reveal the gastrocnemius and tibialis muscles (for MSMP optical stimulation). On exposure, an optical fibre (0.63 NA, two for muscle stimulation) was positioned roughly 1 mm away from the tissue surface of the sciatic nerve (or muscles) on left side of the mouse body. The right-side sciatic nerve (or muscles) was exposed without light treatment for contralateral control. This configuration enabled optical stimulation of the sciatic nerve (or muscle) for 1 h, using an optical fibre (core diameter of 200 µm) with 0.63 NA, Prizmatix connected to a dual LED optical source (Prizmatix) triggered by a pulse stimulator (Master 9). Two fibres were used for each muscle in the case of MSMP stimulation. Stimulation at a wavelength of 450–460 nm, with a pulse duration of 2 ms at a frequency of 1 Hz with a sufficient power (maximum power greater than 80 mW) to result in myogenic activity. In all case, the hindlimbs were securely immobilized to minimize movement and the exposed tissue was continually infused with warm saline to maintain its functional viability and avoid dehydration.

MSMP sorting for single-cell RNA-seq

After 1 h of activation of Pv neurons, a cardiac perfusion was performed using 20 ml of cold DPBS. MSMP were dissected and FACS-sorted from both light-treated and control side of Pv-ChR2 mice with the previously described gating strategy. Sorted MSMP were collected in cold DPBS and processed for single-cell RNA-seq according to 10X Chromium Next GEM Single Cell 3′ Reagent Kits v.3.1 (Dual Index) protocol.

Single-cell or single-nucleus RNA-seq data processing

The library of MSMP or DRG was sequenced by Illumina NovaSeq 6000 platform. Samples were demultiplexed into FASTQ reads and then aligned to the mouse GRCm39 genome reference. Sample demultiplexing, sequence alignment, barcode processing and single-cell 3′ unique molecular identifier counting were performed by using Cell Ranger Software Suite (v.7.1.0). Seurat objects for each sample were created by converting the count matrix and performing initial quality control. Cells were filtered on the basis of thresholds for read counts (greater than 500), gene features (greater than 250) and mitochondrial content (less than 15%). Genes expressed in fewer than ten cells were filtered out. Doublets were removed from the dataset by DoubletFinder package (doublet formation rate was set as 7.5%). The filtered datasets were normalized by using SCTransform. The count matrices from different samples were merged into a single Seurat object. The analysis then progressed to perform data integration using Seurat (v.4) integration function. Anchors were identified for dataset integration, and hierarchical clustering is used to perform dimensionality reduction. PCA and uniform manifold approximation and projection (UMAP) were used for dimensionality reduction and visualization. After PCA, 40 principal components were selected for cell clustering. Clusters were defined on the basis of a chosen resolution, and the resulting cell clusters were visualized through UMAP plots. Seurat FindAllMarkers function enabled the identification of marker genes associated with specific cell clusters. For DRG neurons, we annotated proprioceptors by using Pth1r, Esrrg, Pvalb, Etv1. C-LTMRs were annotated by using Tafa4, Cacna1h, Zfp521. Aδ-LTMR were annotated by using Ntrk2, Cacna1h, Cadps2, Kcnq3. Non-peptidergic C-fibre nociceptors (NPs) subsets were annotated using Trpc3, Gfra2, Plxnc1 (NP1); Osmr, Il31ra, Nppb (NP2); Tnr, Gfra1, Mrgpra3 (NP3). C-fibre peptidergic nociceptors (PEPs) subsets were annotated using Tafa1 (PEP1); Gfra3, Kcnmb2 (PEP2); Tafa1, Kcnt2 (PEP3), Ryr2, Ryr3, Cntn5, Gm10754 (PEP4). Cold thermoreceptor subsets were annotated using Trpm8, Foxp2.

Genes that were differentially expressed in each cell cluster between two conditions were identified by using the Seurat FindMarkers function with the MAST test. Differentially expressed genes between each pair of clusters were determined with a false discovery rate (FDR) less than 0.05 and performed GO analysis. GO pathways were categorized by the DAVID Bioinformatics Resource with a P value less than 0.05. GO categories highlighted in red and blue were upregulated or downregulated pathways, respectively.

Electromyography and dorsal root recording with electric stimulation of the ‘H reflex’ or stimulation of stretch reflexes

The animals were deeply anaesthetized with isoflurane as previously mentioned (for optical stimulation of sensory neurons) and a shave at the level of the spinal columns and the two hindlimbs. Iodine and 70% ethanol were then applied to the shaved skin. After 10 min, another incision of the skin exposed specific muscles to allow bipolar recordings and monitor the activity of the tibialis, the gastrocnemius, and the semitendinosus muscles. Two dual core Teflon-coated platinum–iridium wires with a diameter of 125 µm (WPI, code number PTT0110) were threaded through a 30.5 gauge hypodermic needles. The Teflon cover was removed from about 0.5 mm of the tips and the two electrodes were then inserted 1–1.5 mm apart in the belly of each muscle. Successively, another skin incision at the level of the spinal column followed by a laminectomy exposed the thoraco-lumbar segments of the spinal cord with the respective dorsal roots. The dura mater was opened, and the dorsal roots were gently moved apart for easy identification of the belonging segment. Bleeding due to the laminectomy was blocked. A dorsal root between L2 and L6 was chosen and then recorded using a suction electrode. Electric activation of sensory neurons was obtained by stimulating the sciatic nerve using a bipolar stimulating electrode with a hook shape. Electric stimulation (50 ms, 50–150 mA, low threshold stimulation) elicit primarily the typical muscle activity obtained with activation of the proprioceptive afferents, so electromyography recordings that correspond in latencies and threshold to the ‘M wave’ and to the ‘H reflex’ in mice⁵⁵. The M wave represents the direct stimulation of motor axons whereas the H reflex is generated by the monosynaptic activation of motor neurons by Ia afferent fibres in the muscles. The amplitude of the H reflex is proportional to the number of activated motor neurons. Electrical signals were acquired using a band pass filter (100–1,000 kHz) and all data were stored for off-line analysis. An NMDA blocker (AP-5, 5 mM) or AMPA blocker (CNQX, 5 mM) was injected into the muscle. Wash out of the drugs and restoration of responses were obtained by injecting saline.

To test the inactivation of MSMP (CX3CR1⁺) on the stretch reflexes, mice were anaesthetized and the hindlimb at the knee joint were immobilized to a horizontal plate. Then, the Achilles tendon of the gastrocnemius was cut and secured to a small diameter bar (diameter of 3 mm) with a suture wire (Mersilk, Ethicon). The bar was controlled by a programmable micromanipulator to (MCL3, WPI) executing movements to elongate the muscle from 0.5 to 4 mm s⁻¹ for a duration of 2 s. The elongations were randomly executed. The inhibitory opsin was activated by yellow laser light (590 nm) during the time of the stretch.

MSMP glutamate assay

MSMP were isolated from mouse hindlimb skeletal muscles by FACS. Sorted macrophages were directly collected in the culture media (B27 (1:50), glucose (17.5 mM), glutamine (31.25 μM) and penicillin–streptomycin (1:100) in DMEM (no glucose, no glutamine, no phenol red)). After washing once with culture media, MSMP were resuspended in culture media with 20 ng ml⁻¹ M-CSF1 and cultured in a sterile 96-well plate (20,000 cells per well). On the next day, the original culture medium was collected and changed into the medium with or without glutamine (50 mM). Glutaminase inhibitor CB-839 (2 μM, Selleckchem) was added to the cultured MSMP 1 h before adding glutamine. Glutamine transporter inhibitors α-(Methylamino) isobutyric acid (MeAIB, 20 mM, Sigma) and V9302 (50 μM, Selleckchem), or calcium chelators BAPTA-AM (10 μM, Thermo Fisher Scientific) and EGTA (10 mM, Sigma) were dissolved in glutamine-free culture medium and added to cultured MSMP for 10 min of preincubation, respectively. After washing with glutamine-free culture medium twice, MSMP were treated with 50 mM glutamine (dissolved in culture medium) together with or without dimethylsulfoxide (DMSO), MeAIB and V9302, or BAPTA-AM and EGTA. Control cells were treated with glutamine-free medium with or without DMSO. After 2 or 8 h of incubation, the MSMP-conditioning medium was collected and centrifuged with high speed to remove debris. Glutamate concentration was measured with the Glutamate-Glo Assay Kit (Promega) according to the manufacturer’s instructions.

MSMP calcium imaging

MSMP were isolated from mouse hindlimb skeletal muscles by FACS. Sorted macrophages were directly collected in the culture media (B27 (1:50), glucose (17.5 mM), glutamine (31.25 μM) and penicillin–streptomycin (1:100) in DMEM (no glucose, no glutamine, no phenol red)). After washing once with culture media, MSMP were resuspended in culture media with 20 ng ml⁻¹ M-CSF1 and cultured in a sterile 96-well plate (20,000 cells per well). On the next day, MSMP were washed with glutamine-free culture medium and incubate with Fluo-4-AM (2 μM, Thermo Fisher Scientific) for 30 min at 37 °C. Cells were washed three times with glutamine-free culture medium and preincubated with glutamine-free medium with or without DMSO, MeAIB (20 mM, Sigma) and V9302 (50 μM, Selleckchem), or BAPTA-AM (10 μM, Thermo Fisher Scientific) and EGTA (10 mM, Sigma) for 10 min. After washing with glutamine-free culture medium twice, MSMP were treated with 50 mM glutamine (dissolved in culture medium) together with or without DMSO, MeAIB and V9302, or BAPTA-AM and EGTA for 5 min. MSMP were immediately washed and fixed by 4% PFA for 30 min at room temperature and processed for immunostaining. Cells were blocked with 5% BSA for 1 h, incubated with F4/80 primary antibody and secondary antibody for 2 and 1 h, respectively.

MSMP real-time qPCR

MSMP were isolated from mouse hindlimb skeletal muscles by FACS after 30 min of freely swimming or resting physiological conditions (control). FACS-sorted macrophages were collected in cold DPBS with RNase inhibitor and directly processed for cell lysis, reverse transcription and preamplification using a SuperScript IV Single Cell/Low Input cDNA PreAmp Kit (Invitrogen) according to the manufacturer’s guidelines. Real-time quantitative PCR (qPCR) was performed on the Quant Studio 3 Flex System with 10 ng cDNA per reaction using KAPA SYBR FAST qPCR Master Mix kit (Sigma) according to the manufacturer’s instructions. A list of the qPCR primers purchased and used in this study is provided in Supplementary Table 7. Expression values for each tested gene relative to a glyceraldehyde 3-phosphate dehydrogenase endogenous control were calculated using the ΔΔCt method.

MSMP patch clamp

MSMP were isolated from mouse hindlimb skeletal muscles by FACS. Target cells were collected into culture media (1% sodium pyruvate, 1% penicillin–streptomycin, 10% FBS and 20 ng ml⁻¹ M-CSF1 in RPMI1640 media). After washing once with culture media, cells were seeded onto 8 mm diameter precoated 100 µg ml⁻¹ poly-d-lysine coverslips in sterile 48-well plates at a final density of 30,000 cells per well. Non-adherent cells were removed by changing medium before functional experiments. Patch-clamp experiments were performed not before 72–96 h postisolation. For this, cultures were positioned in a recording chamber on the stage of an upright BX51WI Olympus microscope, equipped with an ×40 water immersion objective and epifluorescence illumination. The chamber was continuously perfused with artificial cerebral spinal fluid (aCSF), composed of 125 mM NaCl, 2.5 mM KCl, 25 mM NaHCO₃, 2.5 mM CaCl₂, 1.3 mM MgCl₂, 1.25 mM NaH₂PO₄ and 25 mM d-glucose. The aCSF was saturated with carbogen (95% O₂, 5% CO₂) and maintained at a flow rate of 1–2 ml min⁻¹.

Patch electrodes were fabricated from borosilicate glass using a P-1000 pipette puller (Sutter Instruments Co.) and showed resistances of 4–7 MΩ. Macrophages expressing either ChR2-YFP⁺ or NpHR3-YFP⁺ were identified under epifluorescence and subjected to whole-cell patch-clamp recordings in current-clamp mode. The intracellular recording solution contained 108 mM K-gluconate, 8 mM Na-gluconate, 2 mM MgCl₂, 8 mM KCl, 1 mM EGTA, 4 mM K2-ATP, 0.3 mM Na-GTP and 10 mM HEPES; pH was adjusted to 7.2 with KOH and osmolarity to roughly 285 mOsm. All recordings were conducted at room temperature.

After achieving whole-cell configuration, cells were allowed to stabilize for at least 5 min before measuring the resting membrane potential. Optogenetic activation of ChR2 and NpHR3 was performed using blue and red LED light, respectively, delivered through the recording objective by means of an X-Cite 120 LED boost lamp (Excelitas). Electrical signals were amplified and filtered using a MultiClamp 700B amplifier, then digitized at a 10 kHz sampling rate with a Digidata 1440A system (Molecular Devices). Cell capacitance and access resistance were monitored continuously; recordings were deemed acceptable if these parameters varied by less than 20%. Data analysis was performed off-line using the pClamp (v.11.2.1) software (Molecular Devices).

Statistics and reproducibility

Statistical analysis was performed using GraphPad Prism v.8 software. For comparisons between two groups with normally distributed data and equal variances, we used a two-tailed unpaired Student’s t-test. When variances were unequal, we used Welch’s t-test or Wilcoxon. For non-normally distributed data comparing two groups, we applied the Mann–Whitney test. One-way ANOVA was used to compare more than two groups with normally distributed data, followed by appropriate post hoc tests to determine group-specific differences. In cases in which some groups of data were normally distributed and others were not, or when some groups showed unequal variances, we consistently applied the most conservative statistical test. Specifically, when assumptions of normality or homogeneity of variances were violated, we used non-parametric tests or tests designed to handle unequal variances (for example, Welch’s t-test or Mann–Whitney test) to ensure robust and reliable results. A significance threshold of P P P P P 

Sample size was determined on the basis of similar, previously established experimental designs or calculated using the Animal Experimentation Ethics Committee’s power calculator (http://www.lasec.cuhk.edu.hk/sample-size-calculation.html), with parameters set to detect a difference of 1.5, with 80–90% power and a 5% significance level.

Each dataset was replicated and examined over two to three independent experiments.

Replication

All attempts at replication were successful.

Blinding

All analyses were conducted with blinding to the experimental groups. Data were obtained from two to three independent experiments.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.