Introduction

Astrocytes, a type of glial cells, are distributed throughout the central nervous system and are known to play diverse roles in maintaining homeostasis [1]. For example, astrocytes modulate extracellular pH and ion and neurotransmitter concentrations through specific channels and transporters that affect neuronal functions in the brain and spinal cord [2]. Accumulating evidence indicates that astrocytes exhibit intra- and inter-regional heterogeneity in terms of molecular, morphological, and functional characteristics [3]. In particular, recent advanced technologies have shed light on astrocyte heterogeneity in individual brain regions (e.g., the hippocampus [4, 5], cortex [6, 7], hypothalamus [8], and amygdala [9]) and have categorized astrocytes into clusters based on their gene expression profiles. In the spinal cord, astrocytes are distributed in tile form in the white and gray matter (WM and GM, respectively), and their morphology [10] and gene or protein expression profiles [11] are quite different from those of brain astrocytes. Recently, we identified a population of spinal cord astrocytes defined by the expression of hairy and enhancer of split 5 (Hes5). Hes5+ astrocytes are selectively localized to the superficial laminae (laminae I–IIIo) of the spinal dorsal horn (SDH) and play a crucial role in modulating the processing of mechanical information from the periphery [12]. Furthermore, another astrocyte defined by the expression of GFP under the control of a lunatic fringe promoter was recently found in laminae III–IV [13]. Despite recent progress, there have been no reports on the comprehensive characterization of astrocytic heterogeneity in the spinal cord at the single-cell level, for example, using single-cell RNA-sequencing (scRNA-seq). One reason for this may be the lack of efficient methods for isolating spinal astrocytes from adult mice. Previously, spinal astrocytes have been isolated using papain-based tissue digestion. However, the yield of isolated astrocytes was much lower than that of brain astrocytes [14]. Furthermore, the conventional cell dissociation method using a protease requires heating, which has been reported to artificially induce or change the expression of genes, such as immediate early genes, in several brain cells including astrocytes [9].

To overcome these limitations, we used a cold-active protease from Bacillus licheniformis [15] with an astrocyte cell surface antigen-2 (ACSA-2) antibody [16] and efficiently isolated astrocytes from the mouse spinal cord in high yields using fluorescence-activated cell sorting (FACS). We also found that the cells in the ACSA-2high fraction were GM astrocytes. Thus, our established method enables the selective and high-yield isolation of GM astrocytes from the spinal cord of adult mice and may be useful for further transcriptomic and epigenetic analyses at the bulk and single-cell levels to advance our understanding of astrocyte heterogeneity under physiological and pathological conditions.

Results

Selective immunolabeling of GM astrocytes in the spinal cord by ACSA-2 antibody

Aldh1l1-EGFP mice, in which almost all astrocytes express EGFP reporter proteins [17], and an ACSA-2 antibody have been frequently used to isolate adult mouse brain astrocytes using FACS [4, 16, 18, 19]. To explore the best method of FACS isolation of spinal astrocytes in adult mice, we first compared the distribution patterns of EGFP+ and ACSA2+ cells using Aldh1l1-EGFP mice. EGFP+ cells in the lumbar spinal cord of Aldh1l1-EGFP mice were ubiquitously observed in both GM and WM (Additional file 1: Supplementary Fig. 1a). The fluorescence intensity of EGFP in individual cells was slightly higher in GM astrocytes than in WM astrocytes, and was not detected in wild-type (WT) mice (Additional file 1: Supplementary Fig. 1b). EGFP+ cells in the GM were co-labelled with the astrocytic marker SRY-related high-mobility group box 9 (SOX9) but not with SRY-related high-mobility group box 10 (SOX10, for oligodendrocyte lineage cells), neuronal nuclear antigen (NeuN, for neurons), or ionized calcium-binding adapter molecule 1 (IBA1, for microglia) (Additional file 1: Supplementary Fig. 1c). In the WM of the spinal cord, some EGFP+ cells were co-stained with SOX9 and others with SOX10 (Additional file 1: Supplementary Fig. 1d). In contrast, ACSA-2 immunofluorescence in the spinal cord was selectively observed in the GM (Fig. 1a). However, by increasing the gain of image acquisition (at a saturated level in the GM), weak immunofluorescence signals of ACSA-2 were detected in the WM, which was higher than that in the IgG control (Fig. 1b). These results suggest that the ACSA-2 antibody is a useful tool for selectively isolating GM astrocytes from the spinal cord of adult mice.

Fig. 1
figure 1

Specific ACSA-2-immunostaining in the gray matter (GM) of spinal cord in adult mice. a Immunolabeling with an ACSA-2 antibody in the fourth lumbar (L4) spinal cord of Aldh1l1-EGFP mice. The dashed line indicates the boundary between GM and white matter (WM). b Representative immunohistochemical images of ACSA-2 and matched IgG control antibodies in the spinal dorsal horn (SDH) of Aldh1l1-EGFP mice. The immunofluorescence intensities of ACSA-2 and IgG was enhanced to a saturated level in the GM by increasing the gain of image acquisition. Scale bars, 500 μm (a) and 100 μm (b)

Full size image

Isolation of spinal astrocytes by FACS using ACSA-2 antibody

Papain is the most frequently used cysteine protease to dissociate brain and spinal cord tissues to isolate astrocytes [13, 18, 19]. Heating tissue samples is needed to achieve their highest enzymatic activity, but this has been shown to allow for artificial changes in gene expression in several cell types, including astrocytes [9]. Thus, we selected a cold-active serine protease purified from the psychrophilic soil bacterium Bacillus licheniformis [15]. The lumbar spinal cord tissues (the third and fourth lumbar (L3/4) segments) were treated with this enzyme at 4 °C for 20 min, myelin debris and oligodendrocytes were removed from the suspension using myelin removal beads (20 µL per one L3/4 spinal cord tissue), and ACSA-2+ cells were isolated by FACS (Fig. 2a-c). After doublet exclusion, 7-aminoactinomycin D (7-AAD)-negative viable cells immunolabeled with the ACSA-2 antibody, but not the IgG control, were selected as ACSA-2+ cells (Fig. 2b). As ACSA-2 recognizes the cell surface protein ATPase Na+/K+-transporting subunit beta 2 (ATP1B2), the fraction of ACSA-2+ cells may have included astrocytic process fragments. To purify the cell bodies of astrocytes, we further treated the ACSA-2 antibody-containing cell suspension with DRAQ5, a cell-permeable far-red fluorescent DNA dye [20] and found that ACSA-2high and ACSA-2low cells were clearly identifiable (Fig. 2c). Consistent with previous data [16], quantitative polymerase chain reaction (qPCR) analysis of the sample extracted from ACSA-2+ cells showed that astrocyte (Sox9), but not other cell-types markers (Sox10 for oligodendrocyte lineage cells, Mog for mature oligodendrocytes, Itgam for microglia, Rbfox3 for neurons and Pecam1 for endothelial cells except Cspg4 for perivascular cells and oligodendrocyte precursor cells) were enriched in ACSA-2+ population compared with the L3/4 spinal cord tissues (Fig. 2d). These results indicate that this protocol enables the isolation and purification of spinal astrocytes from adult mice.

Fig. 2
figure 2

FACS isolation of ACSA-2+ cells from the lumbar spinal cord. a, b, c Gating strategy for collecting ACSA-2+ cells from total cells in the L3/4 spinal cord of adult mice stained with IgG-PE (a), ACSA2-PE (b), or ACSA-2-PE with DRAQ5 (c). ACSA-2+ cells were gated using singlets of 7-AADneg (live), DRAQ5+ (cell body), and ACSA-2+ . d Quantitative polymerase chain reaction (qPCR) analysis of cell-type specific markers in isolated ACSA-2+ and ACSA-2neg cells (n = 5 mice). Values represent the relative ratio of mRNA (normalized to Gapdh mRNA) to the mean mRNA expression level of total cells in the L3/4 spinal cord sample. Data show mean ± SEM.

Full size image

Highly efficient isolation of the GM astrocytes and ependymal cells from the mouse spinal cord

Based on the differences in ACSA-2 immunofluorescence in the lumbar spinal cord (Fig. 1a, b), we speculated that the ACSA-2high and ACSA-2low populations (Fig. 3a) could be GM and WM spinal astrocytes, respectively. To characterize the populations of ACSA-2high and ACSA-2low cells, mRNA was extracted from these populations separately. The expression of alanine–serine–cysteine transporter-1 (ASC-1; encoded by Slc7a10), an Na+-independent amino acid transporter, is highly specific to GM astrocytes in the spinal cord [21, 22]. As expected, ACSA-2high but not the ACSA-2low population expressed Slc7a10 (Fig. 3b). Galectin-3 (encoded by Lgals3) which is known to be expressed in WM astrocytes [23] was specifically detected in the ACSA-2low population (Fig. 3b). However, in contrast to the expression of Sox9, Aqp4 which is a marker of astrocytes, was much lower in ACSA-2low cells than in ACSA-2high cells (Fig. 3b). Immunohistochemical analysis showed that galectin-3 and SOX9 were expressed in WM astrocytes (Fig. 3c) and ependymal cells (Fig. 3d). Consistent with this result, ACSA-2low but not ACSA-2high cells highly expressed Dynlrb2, a marker of ependymal cells [24, 25] (Fig. 3e). By increasing the gain of image acquisition, weak immunofluorescence signals of ACSA-2 were also detected in the ependymal cells, the intensity of which was higher than that in IgG control (Additional file 1: Supplementary Fig. 2). These results suggest that the ACSA-2high and ACSA-2low populations are GM spinal astrocytes and ependymal cells, respectively.

Fig. 3
figure 3

Identification of ACSA-2high and ACSA-2low cells as GM astrocytes and ependymal cells, respectively. a Representative scatter plots of ACSA-2high and ACSA-2low cells. b qPCR analysis of Slc7a10, Lgals3, Sox9 and Aqp4 mRNAs in fluorescence-activated cell sorting (FACS)-sorted ACSA-2high and ACSA-2low cells (n = 5 mice). c, d Immunostaining of galectin-3 (green) and SOX9 (magenta) in the SDH (c) and the central canal (d) of wild-type mice. e qPCR analysis of Dynlrb2 mRNA in FACS-sorted ACSA-2high and ACSA-2low cells (n = 5 mice). Scale bars, 100 μm (c) and 20 μm (d). Data show mean ± SEM

Full size image

Optimized isolation procedure enhances the yield of GM spinal astrocytes

To enhance the yield of ACSA-2high population, we optimized the processes of myelin removal and enzymatic reactions. As the treatment with myelin removal beads has been reported to affect the yield of astrocytes [26], we tested different volumes of myelin removal beads (5, 7.5, or 10 µL per one hemi-section of the L3/4 spinal cord). A low volume of myelin removal beads slightly increased the number of ACSA-2high cells (Fig. 4a). Furthermore, as a longer duration of enzymatic treatment allowed for effective dissociation at a lower temperature [9], we compared the effect of three different dissociation periods (20, 40, or 60 min) and found that enzymatic dissociation for 60 min enabled the isolation of the highest number of ACSA-2high cells (Fig. 4b). In contrast to the ACSA-2high cells, these procedures had non-significant effects on the yield of ACSA-2low cells (Additional file 1: Supplementary Fig. 3a, b). Under this procedure (myelin removal beads 5 µL, and enzymatic dissociation for 60 min), we confirmed that FACS-purified ACSA-2high cells (7079 ± 269 cells per L3/4 spinal cord tissue, n = 5) highly expressed astrocytic markers (Sox9 and Aqp4) and exhibited low expression of Sox10 and Cspg4 (Fig. 4c, d). Furthermore, the mRNAs of Slc7a10 and Hes5 (a marker for our identified subpopulation of astrocytes in the SDH) were enriched in the ACSA-2high population (Fig. 4d). In contrast, Sox9, Lgals3, and Dynlrb2 were enriched in the ACSA-2low population (1355 ± 56 cells per the L3/4 spinal cord, n = 5) (Additional file 1: Supplementary Fig. 3c). These results indicate that our established method enables the highly effective and selective isolation of GM astrocytes (and ependymal cells) from the spinal cord of adult mice.

Fig. 4
figure 4

Optimization of tissue processing for enhancing the yield of spinal GM astrocytes.  a, b Effects of myelin removal (a) and enzymatic reaction (b) on the yield of isolated ACSA-2high cells from the hemisection of the L3/4 spinal cord (n = 5 mice). c Representative FACS plots of cells stained with the ACSA-2 antibody under the optimized myelin removal and tissue dissection protocol. d qPCR analysis of cell type-specific markers in FACS-sorted ACSA-2high cells (n = 5 mice). Data show mean ± SEM

Full size image

Discussion

Increasing evidence indicates molecular and functional heterogeneity of astrocytes in the brain under physiological and pathological conditions [3]. However, the extent to which spinal cord astrocytes are heterogeneous remains to be fully understood because there is a lack of efficient methods to sort spinal cord astrocytes, which hinders a comprehensive analysis of astrocytic heterogeneity, such as scRNA-seq. Here, we established an optimal method for the efficient and high-yield isolation of GM astrocytes from the spinal cord of adult mice. The key to this achievement was to employ cold-active protease-based cell dissociation, cell body isolation using the DNA dye DRAQ5, and cell labeling using an ACSA-2 antibody. Using DRAQ5, the indistinct ACSA-2+ cell population was divided into two distinguished populations: ACSA-2high and ACSA-2low. Thus, the ACSA-2+ fraction may include process fragments of astrocytes that were produced by the tissue dissociation procedure. Furthermore, by analyzing gene expression, we demonstrated for the first time that ACSA-2high cells in the spinal cord of adult mice are GM astrocytes, enabling GM astrocyte-selective analyses. Moreover, our method achieved a high-yield isolation of GM astrocytes by optimizing the protocols for enzymatic tissue dissociation and myelin removal from cell suspensions. The optimized protocol yielded approximately 7,000 cells per L3/4 spinal cord from one mouse, which is much higher than the yield in a previous report using the cysteine protease papain (approximately 9,000 cells per whole spinal cord tissue from one mouse). The reason for this difference remains unclear. However, the serine protease used in our study may be suitable for cell dissociation to form a single-cell suspension from the spinal cord tissues of adult mice.

The process in which oligodendrocytes are removed from cell suspension by myelin removal beads is also important for efficient astrocyte purification using ACSA-2 antibody because, in addition to astrocytes, Atp1b2 mRNA has been reported to be expressed in Cspg4+ or O1+ cells, which are considered oligodendrocyte precursors and mature oligodendrocytes, respectively [4, 16]. Compared to Sox10 mRNA whose expression level was low in the ACSA-2+ population (Fig. 2d), these cells expressed Cspg4 mRNA, a marker of oligodendrocyte precursor cells and perivascular cells. However, considering our further data showing that ACSA-2high and ACSA-2low cells had extremely low or low levels, respectively, of Cspg4 mRNA (Fig. 4d), Cspg4 expression detected in the ACSA-2+ population could be due to the inclusion of a few pericytes. In fact, scRNA-seq has revealed that brain cells isolated using the ACSA-2 antibody include mural cells [4] and astrocytes and pericytes in the spinal cord express Atp1b2 [27].

An unexpected finding in this study was that cells in the ACSA-2low fraction were ependymal cells, but not astrocytes (Fig. 3 and Additional file 1: Supplementary Figs. 2, 3). In fact, ACSA-2low cells did not express the astrocytic gene Aqp4 but did Dynlrb2 which is a marker of ependymal cells. Other astrocytic genes, Sox9 and Lgals3 were detected in ACSA-2low cells, but SOX9 and galectin-3 were also found in ependymal cells of the central canal, which is consistent with previous data. Considering our data, ACSA-2+ cells isolated using only ACSA-2 magnetic beads could be a mixture of astrocytes and ependymal cells; thus, caution should be exercised when interpreting data obtained from non-fractionated ACSA-2+ cells. However, because, to our knowledge, there is currently no efficient isolation method for ependymal cells by antibody-based FACS in non-transgenic WT mice, our method could also be useful for ependymal cell isolation, although there is one previous report that isolated ependymal cells and investigated their heterogeneity using a reporter protein-expressing transgenic mouse line [25].

A limitation of this study was that WM astrocytes were not isolated from the spinal cord using our method. A similar difficulty was also reported in the optic nerve, which may be related to the dense myelination of these tissues [28]. Fluorescence-activated nuclei sorting [29] or Ribo Tag immunoprecipitation [30] may be used to overcome this limitation. Another limitation was that this study examined only adult mice because the expression of ATP1B2 in the brain is detectable from embryonic day 18.5 (E18.5) [16]; however, its expression level and cell-type specificity in the spinal cord remain to be determined. Further investigations are needed to clarify whether the current protocol applies to pre- and postnatal or aged mice.

An advantage of our method is to apply for isolating spinal astrocytes from genetically modified mice (e.g., cell-type-specific conditional knockout mice or genetic animal models of disease) because there is no need to cross additional transgenic mice, such as Aldh1l1-EGFP mice. Recent single-cell RNA-seq analysis revealed that Atp1b2 is predominantly expressed in astrocytes of the marmoset [31], developing [32] and adult human spinal cords [33]. Therefore, our established method can be adapted to other species. Among more important challenges are to determine whether there is heterogeneity within the ACSA-2high population in the spinal cord by scRNA-seq analysis of isolated spinal astrocytes using our established method, since a recent scRNA-seq analysis of ACSA-2+ brain astrocytes has revealed that these cells are divided into several populations [4]. Furthermore, previous papers have shown that cultured astrocytes acutely isolated from mice and human have abilities to secrete molecules, promote neuron survival and respond to various stimuli [34,35,36]. Thus, the functional analysis of isolated spinal astrocytes using our established method is also an important subject. Thus, our technique will be useful for transcriptomic and epigenetic analyses at bulk and single-cell levels, deepening our understanding of spinal astrocyte heterogeneity under physiological and pathological conditions.

Methods

Animals

Male C57BL/6 mice (CLEA Japan, Tokyo, Japan) were used. Male and female Aldh1l1-EGFP mice (B6;FVB-Tg(Aldh1l1-EGFP/Rpl10a)JD130Htz/J) [17] were purchased from The Jackson laboratory. All mice used were 8–10 weeks of age at the start of each experiment and were housed at 22 ± 1 °C with a 12-h light-dark cycle. All animals were fed food and water ad libitum. All animals were housed in standard polycarbonate cages in groups of same-sex littermates. All animal experiments were conducted according to relevant national and international guidelines contained in the ‘Act on Welfare and Management of Animals’ (Ministry of Environment of Japan) and ‘Regulation of Laboratory Animals’ (Kyushu University) and under the protocols approved by the Institutional Animal Care and Use committee review panels at Kyushu University.

Immunohistochemistry

Mice were deeply anesthetized with intraperitoneal (i.p.) injection of pentobarbital and transcardially perfused with phosphate-buffered saline (PBS) followed by ice-cold 4% paraformaldehyde (PFA)/PBS. The transverse fourth lumbar (L4) segments of spinal cord was removed, postfixed in the same fixative for 3 h at 4°C. According to a method of our previous paper [12], transverse sections of L4 spinal cord (30 µm) were incubated for 48 h at 4°C with primary antibodies for ACSA-2-PE (1:50, 130-116-244, Miltenyi Biotec, Bergisch-Gladbach, Germany), IgG-PE (1:50, 130-113-450, Miltenyi Biotec, Bergisch-Gladbach, Germany), polyclonal goat anti-SOX9 (1:2000, R&D Systems, MN, USA), polyclonal goat anti-SOX10 (1:500, AF2864, R&D Systems, MN, USA), monoclonal rabbit anti-NeuN (1:2000, ab17748, abcam, Cambridg, UK), polyclonal guinea pig anti-IBA1 (1:2000, 234 00, Synaptic Systems, Goettingen, Germany) and monoclonal rat anti-galectin-3 (1:500, CL8942AP, Cedarlane Labs, Burlington, Canada). Tissue sections were incubated with secondary antibodies conjugated to Alexa Fluor 488 (1:1000, A-21208, Thermo Fisher), Alexa Fluor 546 (1:1000, A-11056, Thermo Fisher) or DyLight 405 (1:1000, 711-475-152 or 706-475-148, Jackson ImmunoResearch) and mounted with or without 4’,6-diamidino-2-phenylindole (DAPI; Vector Laboratories, USA). Three sections from one tissue were randomly selected and images were taken using an LSM700 Imaging System (Zen 2012, Carl Zeiss) and BZ-X800 fluorescence microscope (Keyence Corporation, Osaka, Japan).

Fluorescence activated cell sorting (FACS)

Mice were deeply anesthetized by i.p. injection of pentobarbital and perfused transcardially with ice-cold PBS. The L3/4 spinal cord was rapidly removed from the vertebral column and placed into ice-cold Hanks’ balanced salt solution (HBSS) without Ca2+, Mg2+ (HBSS (-)). The dura and pia matter and dorsal/ventral roots were carefully removed from the spinal cord, and spinal cord tissue was cut into 8 pieces. The spinal tissue pieces were treated with ice-cold 1 mL enzymatic solution [10 mg/mL of protease from Bacillus licheniformis enzyme (Sigma-Aldrich, P5380) with 125 units/mL of DNaseI (Worthington Biochemical Corporation) and 5 mM CaCl2] in HBSS (-) for 10 or 30 min at 4 °C. The tissues were homogenized by passing through a 23G needle attached with 1 mL syringe and were further incubated for 10 or 30 min at 4 °C. After that, the tissues were homogenized by passing through a 26G needle, and the enzymatic reaction was stopped by HBSS with Ca2+, Mg2+. After centrifugation (1000 × g, 5 min, 4 °C), the cells were resuspended in HBSS (-) with 0.5% BSA, 2mM EDTA and 10 or 20 µL Myelin Removal Beads II (Milteny Biotec) for 15 min at 4 °C to remove myelin debris and oligodendrocytes from the suspension. In the case of the hemi-section of the L3/4 spinal cord, the tissue pieces were treated with ice-cold 0.5 mL enzymatic solution for 20, 40 or 60 min at 4 °C and the cell suspension was incubated with 5, 7.5 or 10 µL Myelin Removal Beads II. After centrifugation (300 × g, 10 min, 4 °C), the cells were resuspended in HBSS (-) with 0.5% BSA and 2mM EDTA, and passed through a magnetic cell sorting (MACS) LS column (Milteny Biotec) according to the manufacturer’s protocol at 4 °C. After centrifugation (300 × g, 10 min, 4 °C), the cell suspension was blocked by incubating with Fc Block (1:200, 553142, BD Biosciences) for 5 min and immunostained with IgG-PE (1:200, 130-113-450, Miltenyi Biotec) or ACSA2-PE (1:200, 130-116-244, Miltenyi Biotec) for 10 min at 4 °C in the dark. After washing, the cell suspension was treated with 7-aminoactinomycin D (7-AAD; 1:40, 00-6993-50, eBioscience) and DRAQ5 (1:666, 130-117-343, Milteny Biotec) and incubated for at least 15 min on ice for viability and targeting cells with nucleus staining prior to cell sorting. 7-AADneg DRAQ5+ ACSA-2+ cells were gated as live cells and sorted depending on ACSA-2 fluorescent intensity using cytoFLEX SRT (Beckman coulter, CA, USA).

Quantitative polymerase chain reaction (qPCR)

The sorted cells were subjected total RNA extraction using the Quick-RNA Micro-Prep kit according to the manufacturer’s protocol (Zymo Research, CA, USA). The extracted RNA from the sorted cells was transferred to reverse transcriptional reaction with Prime Script reverse transcriptase (Takara, Japan). qPCR was performed with Premix Ex Taq (Takara, Japan) using QuantStudio3 (Applied Biosystems, MA, USA). Expression levels were normalized to the value for Gapdh and values below detection level were regarded as zero. The following primers and probes were obtained from Integrated DNA Technologies (IA, USA): Sox9 (NM_011448; SRY (sex determining region Y)-box 9), Sox10 (NM_011437; SRY (sex determining region Y)-box 10), Mog (NM_010814; myelin oligodendrocyte glycoprotein), Itgam (NM_001082960; integrin subunit alpha M), Rbfox3 (NM_001039168; RNA binding protein, fox-1 homolog (C. elegans) 3), Pecam1 (NM_001032378; platelet/endothelial cell adhesion molecule 1), Cspg4 (NM_139001; chondroitin sulfate proteoglycan 4), Aqp4 (NM_009700; aquaporin 4), Slc7a10 (NM_017394; solute carrier family 7 (cationic amino acid transporter, y+ system), member 10), Hes5 (NM_010419; hairy and enhancer of split 5) and Dynlrb2 (NM_029297; dynein light chain roadblock-type 2). The sequences of TaqMan primer pairs and probe were described below.

Gapdh: 5′-FAM-ACCACCAACTGCTTAGCCCCCCTG-TAMRA-3′ (probe), 5′-TGCCCCCATGTTTGTGATG-3′ (forward primer), 5′-GGCATGGACTGTGGTCATGA-3′ (reverse primer).

Lgals3: 5’-FAM-CCCGCTGGACCACTGACGGTG-TAMRA-3’ (probe), 5’-TCCTGCTGCTGGCCCTTAT-3’ (forward primer), 5’-TTCACTGTGCCCATGATTGTG-3’ (reverse primer).

Statistical analysis

Quantitative data were expressed as the means ± SEM. Data were analyzed by one-way analysis of variance (ANOVA) with the post hoc Tukey’s multiple comparisons test (Fig. 4a, b and Additional file 1: Supplementary Fig. 3a), Kruskal–Wallis test with the post hoc Dunn’s multiple comparisons test (Additional file 1: Supplementary Fig. 3b), as appropriate, after determining the normality (Shapiro–Wilk test) and variance (F-test or Brown–Forsythe test) of the experimental data. Statistical analyses were performed using Prism 7 (GraphPad, CA, USA). Values were considered significantly different at P < 0.05.