Traumatic brain injury induces macrophage subsets in the brain

Macrophages are recruited to the lesion site in large numbers early post-TBI

To assess the immune response following TBI, we used an adult murine controlled cortical impact model. Histological analysis of brain sections following TBI confirmed cortical injury, which extended into the hippocampus (Fig. 1A). Hematoxylin and eosin (H&E) staining revealed increased cellular recruitment to cortical tissues adjacent to the lesion (Fig. 1A). Immunohistochemical staining for F4/80 showed that macrophages/microglia are widely present at the pericontusional site (e.g. in areas of the cortex adjacent to the lesion) (Fig. 1B).

To assess leukocyte subset frequencies present in the brain following TBI, single-cell suspensions from TBI and sham-injured brain contralateral and ipsilateral hemisphere tissues were analyzed by flow cytometry. Flow cytometry permitted discrimination of macrophages from microglia based on levels of CD45 expression; both microglia and macrophages express CD11b, but macrophages express a higher level of CD45 30, 31. In our analyses of macrophages and microglia, neutrophils (which also express CD45 and CD11b) were consistently excluded by using an antibody against Ly6G (Clone 1A8). Blood leukocytes were excluded by perfusing the brain prior to cell recovery.

Flow cytometry plots of cell preparations from brain tissues 4 days following TBI of WT mice showed that macrophages are a major part of the inflammatory response to TBI primarily on the side of injury (Fig. 1C); macrophages comprised 40 ± 2% of all CD45⁺ leukocytes in the ipsilateral TBI hemisphere compared with 5.7 ± 1.5% of CD45⁺ cells in sham control tissues (p < 0.001).

Quantification of the kinetics of macrophage numbers that accumulate in brain hemispheres after TBI revealed that macrophage infiltration in ipsilateral hemispheres of TBI mice increased by 21-fold on day 1 (mean ± SEM, 22 115 ± 1732), and by 77-fold on day 4 (46 968 ± 5918) compared with sham controls (1081±151 and 613± 205, respectively) (Fig. 1D). On day 7, WT ipsilateral TBI macrophage numbers declined but were still 25-fold higher than levels in sham controls, and on day 14 macrophage numbers were fourfold higher (Fig. 1D).

On the first day following TBI, there was also a substantial increase in neutrophils (CD45^hiCD11b⁺Ly6G⁺) in the brain (41 520 ± 4533 compared with 1419 ± 94 in sham controls), with a decline thereafter (Fig. 1D). These findings are similar to the recent findings of Jin et al. 32, although our results add quantification of absolute cell numbers as well as proportions, and we find that macrophage levels are higher on day 4 than on day 1.

Early macrophage response to TBI includes Arg1⁺ and Arg1⁻ subsets

To examine macrophage polarization post-TBI, we first sought to trace the genetic expression of Arg1, which is highly expressed during M2 polarization, or of Il12b, the gene for IL-12p40, a signature of M1 polarization. To do this, we took advantage of two reporter mouse strains, YARG (YFP-Arginase-1) and Yet40 (YFP-enhanced transcript for IL-12p40) 28, 33. TBI was performed in YARG and Yet40 mice, and YFP expression in brain and peripheral blood leukocytes was compared by flow cytometry to WT animals, which lack YFP expression.

One day after TBI, 21 ± 1.5% (mean ± SEM, n = 6) of ipsilateral hemisphere brain macrophages in YARG mice expressed YFP (Fig. 2A), but brain macrophages in the contralateral hemisphere and from either hemisphere of sham animals uniformly lacked YFP (data not shown). YFP expression in YARG brain macrophages peaked on day 1 after TBI, fell to 4–7% of the macrophage population by day 4, and was undetectable on days 7 and 14 (data not shown). YFP expression could not be detected in microglia following TBI at any time point. F4/80⁺ blood monocytes isolated from the same injured YARG animals also lacked expression of YFP (Fig. 2A), suggesting that TBI induces macrophage differentiation after localization in the tissue. Brain macrophages and blood monocytes from TBI animals differed markedly not only in YFP expression but also in their gene expression profiles as assessed by microarray (Fig. 4 and Supporting Information Fig. 1), confirming that macrophages isolated from brains were not significantly contaminated by blood monocytes. Yet40 mice subjected to TBI had little or no upregulation of YFP in macrophages or microglia on days 1, 4, 7, and 14 (day 1 is shown), and this was subsequently confirmed for macrophages by microarray analysis for IL-12p40 on day 1 where all comparison ratios were close to 1, indicating no change in expression in comparison to blood monocytes or between brain macrophage subsets. Thus, TBI rapidly induces a macrophage response that is characterized at early time points by at least two major subsets of cells that differ in Arg1 expression, and these are hereafter called Arg1⁺ and Arg1⁻ cells.

Analysis of markers for cell activation and for antigen presentation on macrophages from YARG mice revealed that both Arg1⁺ and Arg1⁻ populations upregulated the activation marker CD86 compared with sham control macrophages (Fig. 2B). Few Arg1⁺ macrophages, however, expressed MHC class II antigens (MHCII; Fig. 2C), a marker that has been described on both M1 and M2 cells 17, 34. In contrast, 25–30% of Arg1⁻ macrophages expressed MHCII (Fig. 2C). This is similar to the proportion of macrophages that express MHCII in sham brains (Fig. 2C), and it suggests that the Arg1⁻ cells include at least two subpopulations, one lacking and the other expressing MHCII.

Although microglia from TBI brains did not express detectable MHCII (Fig. 2C), virtually all microglia upregulated CD86 following TBI (Fig. 2B). This finding is consistent with previous observations that TBI induces widespread activation of microglia 35, 36.

To examine the spatial localization of YFP⁺ cells in YARG mice post-TBI, we performed immunofluorescent colabeling for YFP and F4/80 in brain sections 2.2 days post-TBI, when macrophage infiltration of the brain peaks. F4/80⁺ macrophages/microglia localized in and around the area of injury (Fig. 3, second row). F4/80 expression was below level of detection by immunofluorescence in sham-injured tissues (data not shown). The Arg1⁺ cells were scattered among the F4/80⁺ cells in TBI mice (Fig. 3, third row) and were not detectable in the contralateral hemisphere or in sham-treated mice. The majority of the Arg1⁺ cells costained with F4/80. As suggested from our flow cytometry data in which only a subset of macrophages expresses YFP, the majority of F4/80⁺ cells were Arg1⁻ (Fig. 3).

Arg1⁺ and Arg1⁻ brain wound macrophages represent at least two distinct subsets

To further characterize the nature of brain macrophages following TBI, we sorted the Arg1⁺ and Arg1⁻ cell populations from the ipsilateral hemisphere of YARG mice 1 day after TBI (when the proportion of Arg1⁺ cells peaked) and performed gene expression analysis of both cell populations. We also examined gene expression by peripheral blood monocytes from injured animals to assess the expression state of monocytes prior to their infiltration into the brain and differentiation into macrophages. As a control, peripheral blood monocytes from uninjured animals were also analyzed. It was not technically feasible to perform arrays on brain macrophages from sham animals, because there were insufficient cells to generate adequate amounts of RNA. Pairwise analyses of differentially expressed genes showed that Arg1⁺ and Arg1⁻ brain macrophages differed in the expression of 1360 genes, and both populations showed even greater differences from TBI monocytes (11 799 genes differed between Arg1⁺ macrophages and TBI monocytes; 9932 genes differed between Arg1⁻ macrophages) (Fig. 4A). TBI monocytes displayed few differences compared with normal monocytes (15 genes) (Fig. 4A). Principal component analysis (PCA), an analytical technique that uses dimensionality reduction to identify dominant patterns within highly multivariate data, was performed. PCA confirmed that distinctions separating macrophages from monocytes were the largest source of variance in the dataset (principal component (PC) 1), and that the monocyte populations had fewer differences that were not represented in either of the top two PCs (Fig. 4B). PCA also confirmed that Arg1⁺ and Arg1⁻ brain macrophages represented two distinct populations, representing the second most significant PC (PC2) (Fig. 4B).

TBI-induced macrophages exhibit transcriptional responses distinct from known macrophage subsets

Although robust Arg1 expression is often used as a marker for alternative activation of macrophages, we observed that Arg1⁺ and Arg1⁻ brain macrophages after TBI did not represent clear M2 and M1 macrophages, respectively, but instead each subset expressed markers of both M1 and M2 cells. Comparison of gene expression between Arg1⁺ and Arg1⁻ macrophages confirmed that the former expressed much higher levels of Arg1 (eightfold) as well as higher levels of Mrc1 (2.4-fold), which encodes the mannose receptor/CD206 17 (Fig. 5). Increased expression of these two genes is a feature of M2 cells. The expression of other genes, however, indicated that Arg1⁺ macrophages were not identical to M2 cells. For example, Arg1⁺ macrophages preferentially expressed Nos2 (2.1-fold), an M1-associated gene 17 (Fig. 5). Similarly, although Arg1⁻ macrophages had increased expression of Il1b (IL-1β) (2.4-fold), they also preferentially expressed signature M2 markers, notably Retnla (resistin-like α) (2.1-fold) and Clec10a (C-type lectin domain family 10, member A)/CD301 (2.9-fold) 17, 37 (Fig. 5). The relative increases in expression levels of Arg1, Mrc1, Nos2, and Il1b were confirmed by real-time PCR, demonstrating that relative to GAPDH, these genes were indeed transcriptionally active (Fig. 6). In accordance with flow cytometry data (Fig. 2C), gene expression analysis of MHCII, a molecule thought to be on both M1 and M2 cells, revealed that the Arg1⁻ macrophage population as a whole expressed much higher levels of MHCII transcripts (not shown) and higher levels of Ciita (class II, MHC, transactivator) than the Arg1⁺ macrophages (Fig. 5). The MHCII⁺ Arg1⁻ macrophages may thus have increased capacity to present antigen to CD4⁺ T cells.

Taken together, we conclude that Arg1⁺ and Arg1⁻ macrophages each have mixed expression of M2 and M1 properties, and under the conditions of TBI Arg1 cannot be used as a marker for conventional M2 cells. To further compare Arg1⁺ and Arg1⁻ TBI brain macrophages with M1 and M2 macrophages, we performed a meta-analysis of genes differentially expressed between Arg1⁺ and Arg1⁻ TBI brain macrophages compared with genes differentially expressed between IFN-γ- or IL-4-stimulated bone marrow derived macrophages (BMDMs) stimulated in vitro with IFN-γ or with IL‑4, representing M1 and M2 cells, respectively 38. Arg1⁺ and Arg1⁻ macrophages each upregulated a variety of genes that were also expressed by BMDMs in response to either IFN-γ or IL-4 (Fig. 7). Thus, Arg1⁺ and Arg1⁻ TBI brain macrophage subsets have features of both M1 and M2 phenotypes (Fig. 7). There are at least two explanations for these findings, not mutually exclusive: (i) individual brain macrophages may have features of both M1 and M2 cells (including cells that are incompletely polarized or are in transition from between different states of polarization and (ii) there may be subsets of cells within the Arg1⁺ and Arg1⁻ cells that have different expression of M1 and M2 markers. Regardless, the gene expression profiles demonstrate that Arg1⁺ and Arg1⁻ macrophages differ by many genes other than just Arg1.

The most striking and novel differences between Arg1⁺ and Arg1⁻ macrophages were in their unique chemokine profiles. Arg1⁺ macrophages preferentially expressed a chemokine repertoire that included Ccl24 (which is also secreted by M2 cells; 6.2-fold), Cxcl7 (5.4-fold), Cxcl4 (2.4-fold), Cxcl3 (4.5-fold), Cxcl1 (3.6−fold), Cxcl14 (2.4-fold), and Ccl8 (2.3-fold) (Fig. 5). Arg1⁻ macrophages, in contrast, preferentially upregulated Ccl17 (6.8-fold), Ccl5 (4.4-fold), Ccl22 (3.7-fold), and Ccr7 (tenfold) (Fig. 5).

Although the gene profile of the Arg1⁺ macrophages suggests that they are not typical or homogeneously polarized M2 cells, they may have a role in promoting wound healing and in suppressing inflammation. Thus, Arg1⁺ macrophages preferentially expressed Spry2 (sprouty2; 2.4-fold), Cd9 (2.2-fold), Cd38, and Mt2 (metallothionein-2; 4.2-fold, Fig. 5). Sprouty2 and CD9 have protective roles in wound healing in skin injury models 39, 40. Mt2 and Cd38 have been implicated in neuroprotection during brain injury 41, 42. Arg1⁺ brain macrophages also preferentially expressed several other genes that are associated with protection against tissue injury, including Cd36 (3.8-fold), Hmox1 (heme oxygenase 1; 3.4-fold), Folr2 (folate receptor-2; 2.6-fold), Prdx6 (periredoxin-6; 2.5-fold), and Spsb4 (SPRY domain and SOCS box containing protein 4; 2.5-fold) (Fig. 5) 43-49. If Arg1⁺ cells do have the potential for neuroprotection following TBI, this may be overwhelmed by Arg1⁻ cells, which are greater in number and are less transient.

Animals

C57BL/6 WT males (age 10–16 weeks) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). YARG and Yet40 knockin mice were generated from C57BL/6 mice as previously described 28, 33 and bred in the AALAC-approved transgenic animal facility of the San Francisco VA Medical Center. YARG mice express enhanced YFP from an internal ribosome entry site (IRES) inserted at the 3′ end of the Arg1 gene, leaving the gene and regulatory regions intact, and Yet40 mice express enhanced YFP from an IRES inserted at the 3′ end of the IL-12p40 promoter. Where indicated, mice were administered LPS at 10 mg/kg i.p. and euthanized 4 days later.

Surgery

Controlled cortical impact surgery or sham surgery was performed on anesthetized animals under a protocol approved by the San Francisco VA Medical Center Animal Care Committee. Briefly, bupivacaine was administered subcutaneously above the skull, and an incision was made followed by a 2.5 mm circular craniectomy. TBI was inflicted by a 2 mm circular, flat pneumatic piston traveling at 3 m/s, penetrating 1.5 mm, for 150 ms (Amscien Instruments, Richmond, VA, USA with extensive modifications by H&R Machine, Capay, CA, USA). Target brain coordinates for the center of injury were 1.5 mm lateral, 2.3 mm posterior to the bregma point. After minor bleeding had ceased, the skin was clipped together and animals were monitored for recovery. Sham animals received all surgical procedures without piston impact. As needed, animals were given rehydration therapy for the first 3 days.

Brain and blood leukocyte isolation

Brain leukocytes were harvested according to previously published methods 30. Briefly, following perfusion brain tissues were obtained and mechanically disassociated through a 100 μm cell strainer. Washed cells were treated with 400 U/mL DNase I (Sigma-Aldrich) and 0.5 mg/mL collagenase type I (Worthington) at 37°C for 30 min. Leukocytes were isolated by separation on a Percoll gradient (Amersham Biosciences). For PBL isolation, mononuclear cells were separated from peripheral blood using ficoll-hypaque (GE Healthcare).

Flow cytometry and antibodies

Fc receptors were blocked with 10% rat serum (Sigma) and cells were stained with fluorescent antibodies. Leukocyte analysis used a combination of the following antibodies: anti-CD45 (clone Ly5) allophycocyanin (eBioscience), anti-CD11b (clone M1/70) PE (Invitrogen) or PE-Cy5 (eBioscience), anti-Ly6G (clone 1A8) PE-Cy7 (BD Biosciences), F4/80 (clone BM8) FITC or PE-Cy5 (eBioscience), MHCII (clone M5/114.15.2) PE (eBioscience), CD86 (clone GL1) PE (eBioscience). SYTOX Blue (Invitrogen) was used to gate out dead cells. Cells were sorted on a FACSAria (BD Biosciences) and data were analyzed using FlowJo Software (Treestar). All data represent mean ± SEM.

Histology

Brains were perfused with saline followed by 3.7% formaldehyde. After a 2-h fixation, brains were incubated in 30% sucrose overnight and frozen in tissue-freezing medium (Sakura, Inc.). For H&E staining, brains were sectioned 10 μm thick onto glass slides, heat-dried, and stained (at least three animals per group were analyzed, five sections per animal). For F4/80 staining, 5 μm sections that were quenched for endogenous peroxidases and blocked with streptavidin and biotin (VectorLabs) were immunostained with an anti-F480 antibody (Clone BM8, eBioscience), followed by goat anti-rabbit biotinylated antibody and visualized using a Vectastain ABC elite kit (VectorLabs) (three animals per group and at least five sections per animal were analyzed).

For immunofluorescent labeling of YFP and F4/80, a biotinylated goat anti-YFP antibody (Abcam) and streptavidin-HRP (Perkin Elmer) were used and amplified by fluoresceinated tyramide (Perkin Elmer). After an additional round of quenching and blocking, mounted sections were further stained with a biotin-conjugated anti-F4/80 (Clone BM8, eBioscience) antibody followed by streptavidin-HRP and Alexa fluor-555 conjugated tyramide (Invitrogen). DAPI (Invitrogen) was used at 300 nM to identify cellular nuclei. Sections were mounted by using Fluorogel (Electron Microscopy Services). All sections were imaged using either a Nikon Eclipse 80i microscope or an Olympus BX-51 microscope. Three TBI animals were analyzed and at least five sections per animal were analyzed.

Microarrays

For gene expression profiling of macrophages from YARG mice, Arg1⁺ (YFP⁺ CD45^hi CD11b⁺ Ly6G⁻ SYTOX Blue⁻) and Arg1⁻ macrophages (YFP⁻ CD45^hi CD11b⁺ Ly6G⁻ SYTOX Blue⁻) were isolated by flow cytometry from ipsilateral brain hemispheres at day 1 following TBI (n = 4 for each cell sample). Monocytes (CD11b⁺ F4/80⁺) from peripheral blood were also collected. Sorted cells were immediately lysed in denaturation buffer and frozen. RNA was isolated by using an RNAqueous Micro kit (Ambion). Further sample preparation, labeling, and array hybridizations were performed according to standard protocols from the UCSF Shared Microarray Core Facilities and Agilent Technologies. RNA quality was assessed using a Pico Chip on an Agilent 2100 Bioanalyzer (Agilent Technologies), and RNA was amplified by use of a whole transcriptome amplification kit (Sigma-Aldrich). Subsequent Cy3-CTP labeling was performed by using a NimbleGen one-color labeling kit (Roche-NimbleGen, Inc.). The quality of the amplified products was assessed by using an Agilent 2100 Bioanalyzer and Nanodrop ND-8000 (Nanodrop Technologies, Inc.). The products were hybridized to Agilent whole mouse genome 4×44K microarrays according to the manufacturer's protocol. Arrays were scanned with an Agilent microarray scanner, and raw signal intensities were extracted with Feature Extraction v10.5 software. Data were normalized by using the quantile normalization method 54. No background subtraction was performed, and the median feature pixel intensity was used as the raw signal before normalization. A one-way ANOVA linear model was fitted to the comparison to estimate the false discovery rate for each gene for the comparison of interest, and genes with a false discovery rate < 0.05 were considered significant. Scatter plots compared averaged log₂ gene expression from each group. PCA was performed using the top 15% of genes exhibiting the most variance across all samples, using the PopulationDistances module of GenePattern (PMID: 16642009). For heatmaps, data were log₂ transformed and median centered across genes. Replicates were hierarchically clustered (PMID: 16939791). Heatmaps of genes selected from the top 15% most variable genes that exhibited interesting pairwise comparisons were visualized using Java Treeview (http://sourceforge.net/projects/jtreeview/files/) (PMID: 15180930).

Meta-analysis of transcriptional responses of brain wound macrophages to BMDMs stimulated by either IFN-γ or IL-4 was performed using previously published tables 38. Macrophage genes with significant changes in expression upon IFN-γ or IL-4 stimulation were compared with genes with significant expression differences between Arg1⁺ brain macrophages versus Arg1⁻ brain macrophages.

Accession number

Microarray data were deposited in Gene Expression Omnibus (GEO) under accession number GSE39759.

Semiquantitative real-time PCR

Total RNA was isolated from sorted cell populations, including macrophages from injured brain hemispheres and monocytes from peripheral blood, by using an RNAqueous micro kit (Ambion). RT was performed using oligo dT primers and Superscript II reverse transcriptase (Invitrogen). Amplicons were amplified using SYBR green (New England Biolabs) and the rate of amplification was measured using a 7500 real-time PCR machine (Applied Biosystems). Relative transcript levels for each gene were normalized to GAPDH controls by calculating delta cycle of threshold values. The following primers were used for: Arg1 5′-CTCCAAGCCAAAGTCCTTAGAG-3′, 5′-GGAGCTGTCATTAGGGACATCA-3′; Mrc1 5′-CTCTGTTCAGCTATTGGACGC-3′, 5′-TGGCACTCCCAAACATAATTTGA-3′; Nos2 5′-TGTGGCTGTGCTCCATAGTT-3′, 5′-CCAGGGCTCGATCTGGTAGT-3′; Il1b 5′-GCAACTGTTCCTGAACTCAACT-3′, 5′-ATCTTTTGGGGTCCGTCAACT-3′; Ccl24 5′-TCTTGCTGCACGTCCTTTATT-3′, 5′-CTAACCACTCGGTTTTCTGGAAT-3′; Cxcl4 5′-CCTGGGTTTCCGGACTGGGC-3′, 5′-CCGCAGCGACGCTCATGTCA-3′; Cxcl3 5′-CAGAGCTTGACGGTGACGCCC-3′, 5′-CCAGACACCGTTGGGATGGA-3′; Spp1 5′-ATCTCACCATTCGGATGAGTCT-3′, 5′-CTTGTGTACTAGCAGTGACGG-3′; GAPDH 5′-ATTCAACGGCACAGTCAAGG-3′, 5′-TGGTTCACACCCATCACAAA-3′.