Interacting Chemokine Signals Regulate Dendritic Cells in Acute Brain Injury

Brain trauma is known to activate inflammatory cells via various chemokine signals although their interactions remain to be characterized. Mice deficient in Ccl3, Ccr2 or Cxcl10 were compared with wildtype mice after controlled cortical impact injury. Expression of Ccl3 in wildtypes was rapidly upregulated in resident, regularly spaced reactive microglia. Ccl3-deficiency enhanced endothelial expression of platelet selectin and invasion of peripheral inflammatory cells. Appearance of Ccr2 transcripts, encoding the Ccl2 receptor, reflected invasion of lysozyme 2-expressing phagocytes and classical antigen-presenting dendritic cells expressing major histocompatibility complex class II. Ccr2 also directed clustered plasmacytoid dendritic cells positive for the T-cell attracting chemokine Cxcl10. A reduction in Ccr2 and dendritic cells was found in injured wildtype cortex after cyclophosphamide treatment resembling effects of Ccr2-deficiency. The findings demonstrate the feasibility to control inflammation in the injured brain by regulating chemokine-dependent pathways.


Introduction
Traumatic brain injury (TBI) in mice results in distinct upregulation of hundreds of gene transcripts including many linked to inflammation [1,2]. The inflammatory cascade activated by TBI lacks specific pharmacological treatments and is elicited by mechanisms including reactive oxygen species (ROS), hemorrhage and release of nucleotides sensed by upregulated pyrimidinergic receptors such as P2ry6 [1]. This leads to activation of resident microglia offering an instantaneous response [3]. In addition, invading classical (cDCs) and plasmacytoid dendritic cells (pDCs) [4] become engaged in the inflamed central nervous system. The interactions and signals among these inflammatory cells are not fully understood but one possibility includes a balance among chemokine networks.
Chemokines and chemokine receptors are known to orchestrate inflammatory responses via chemotaxis in injured tissues [5,6]. We have previously described the appearance of clustered inflammatory cells expressing chemokine Cxcl10 scattered in grey and white matter of injured wildtype (wt) brains using a controlled cortical impact (CCI) injury model [1,2,7,8]. At the molecular level, strong transcriptional activation after traumatic brain injury can also be seen in chemokine pathways including Ccl3 (with cognate receptors Ccr1 and Ccr5) and Ccr2 (with strongly upregulated ligands Ccl2 and Ccl12) as listed in Table 1. Although this suggests postinjury chemokine functions in brain tissue the interactions among different chemokines, cellular sources of signals and targets for signaling remain unknown. In order to establish a possible interplay among chemokine pathways, we studied knock-out Ccl32/2, Ccr22/2 and Cxcl102/2 mice subjected to TBI. Based on the findings, a model for hierarchical chemokine activities in the injured cerebral cortex is currently presented.
In a previous report [2] we demonstrated general similarities among inflammatory responses after TBI and neurodegenerative disorders in mouse models. Our results distinguish the brain injury response in resident microglia cells in the brain parenchyma from the contribution of invading immune cells providing phagocytes, antigen-presenting dendritic cells as well as clustered Cxcl10producing cells. We also investigated posttreatment in injured mice using the immune-suppressing agent cyclophosphamide, clinically administered to patients with severe forms of systemic lupus erythematosus, in order to consider a pharmacological compound for treatment of TBI. Similar to the outcome of TBI in Ccr2-deficient mice, cyclophosphamide dampened the increases of dendritic cells and limited markers of antigen presentation in the injured brain. The study design is presented in Table 2.

Results
Ccl3 and Ccr2 transcripts are differentially regulated in the injured cerebral cortex TBI rapidly results in local expression of chemokine Ccl3 in the mouse neocortex [1]. By in situ hybridization, responding cells were seen regularly spaced in ipsilateral neocortex, hippocampus and subcortical structures including mesencephalon in a pattern indicating Ccl3 expression in reactive resident microglia (Fig-ure 1A). Quantitative RT-PCR (qRT-PCR) show increases in Ccl3 transcript in neocortex one hour after injury with a peak after four hours ( Figure 1B), expression remaining modestly increased one to three days. Also, the Ccl2 transcript is upregulated within one hour after TBI [1]. The transcript of Ccr2, encoding a cognate receptor for Ccl2, became detectable four hours after injury with a peak at three days ( Figure 1B). This temporal pattern may reflect delayed invasion of inflammatory cells from the periphery. Levels of both Ccl3 and Ccr2 transcripts remained well above background three weeks and three months after the injury.
To examine injury responses in the cerebral neocortex deficient in Ccl3 signaling, we studied TBI in Ccl32/2 mice [9]. Surprisingly, the Ccr2 transcript was upregulated even further after three days in neocortex compared to wt mice ( Figure 1C). In contrast, Gfap expression by reactive astrocytes was increased to the same extent in wt and Ccl3-knockout mice implying similar severity of the inflicted injury.
Considering the stronger increase of Ccr2 transcripts in the injured cerebral cortex of Ccl3 knockout mice, transcripts representing inflammatory functions were examined by unbiased microarray analysis. Comparison of wt and Ccl32/2 mice identified over 30 genes that fulfilled the following criteria: increased more than three-fold three days post-injury comparing injured with uninjured wt mice, upregulated at least 50% further in the injured Ccl3-deficient mice compared to wt mice. Identified transcripts involved in inflammatory responses included C3ar1, Ccr2, Csf2rb1, Cybb, Dab2, H2-Aa, Il2rg, Msr2, Selp, Tgfbi, Thbs1 and Tlr1. In contrast to the increases in Ccr2, the cognate Ccr2 ligands Ccl2 ( Figure 1D, left) and Ccl12, possibly produced by reactive astrocytes, were equally enhanced in injured wt and Ccl32/2 brains as was Gfap expression. Thus, augmented ligand levels are not likely to account for the further increase of Ccr2 in the injured Ccl32/2 mice. Rather, the strong increase in platelet selectin (Selp) in the injured Ccl3-deficient mice ( Figure 1D, right) indicates endothelial involvement [1,6] in recruitment of Ccr2positive leukocytes attracted to the injured brain.
Ccr2 deficiency results in reduced Lyz2 level and smaller cavity volume We also examined the outcome of TBI in wt mice compared to homozygous Ccr2-deficient mice [10] three days postinjury. Lysozyme 2 (Lyz2) previously identified by microarray analysis as injury-induced in neocortex 1 was among the transcripts found reduced in the injured Ccr22/2 cortex. From in situ hybridization, Lyz2 expression was obvious in large phagocyte-like cells only partially overlapping with the activated microglia cell-surface marker isolectin B4 (IB4) [1] in injured neocortex and hippocampus ( Figure 1E). This pattern contrasts with the clustered appearance of Bst2 (bone marrow stromal cell antigen 2, encoding PDCA-1, plasmacytoid dendritic cell antigen 1) expressed by pDCs [11] ( Figure 1E, insert left). qRT-PCR confirmed that Lyz2 injury-increased expression was lower in Ccr22/2 compared to wt cortices. In contrast, Lyz2 was further upregulated in injured Ccl32/2 brains compared to wt brains and thus in parallel with Ccr2 expression levels ( Figure 1F).
The injury-induced cortical cavity volume seen in wt brains was significantly reduced in the Ccr22/2 mice seven days after injury ( Figure 1G). This is in line with the reduced upregulation of Lyz2 in the injured Ccr22/2 cortex and suggests that fewer tissueeliminating phagocytic cells invaded the injured brain when the Ccl2/Ccl12 attraction mediated by Ccr2 signaling was interrupted.
Deletion of Ccr2 does not affect injury-evoked Ccl3 but reduces Cxcl10 expression The Ccl3 expression increased in concert in wt and Ccr22/2 mice ( Figure 2A) whereas Cxcl10 injury-induced expression [1,2] was markedly dampened in brains lacking Ccr2 ( Figure 2B). In contrast, qRT-PCR showed that Cxcl10 transcript was elevated above wt levels in the injured Ccl3-deficient brain, similar to shifts in Ccr2 and Selp transcripts ( Figures 1C and D). The increased Cxcl10 expression appears in clustered inflammatory cells [1] and we earlier showed [2] that injured Cxcl102/2 mice [12] lacked this spotted staining. The clusters of positive cells appear with higher intensity in the injured Ccl32/2 cortex compared to wt ( Figure 2C). In injured Ccr22/2 mice, expression of Cxcl10 was evident in cell clusters but at reduced labeling intensity in line with the qRT-PCR data ( Figure 2B). Cxcl10 expression did not affect injury-induced upregulation of Ccl3 and Ccr2 transcripts three days post-injury in traumatized Cxcl102/2 mice ( Figure 2D). A model of chemokine interactions during the initial three days following injury is presented in Figure 2E. The injury resulted in independent increases in cortical levels of Ccl3 and Ccr2 transcripts. Since, Ccr2 affected Cxcl10 levels ( Figure 2B), but not the reverse ( Figure 2D), Cxcl10 is positioned downstream and positively regulated by Ccr2. Furthermore, Ccl3-defiency increased the injury-evoked Ccr2 levels ( Figure 1C) indicating that Ccl3 normally exerts a suppressive effect on Ccr2 and thus indirectly limit Cxcl10 expression in the injured brain.

Ccr2-deficient cortex
We further investigated the role of Ccr2 signaling in appearance of inflammatory cells in the injured neocortex. Flow cytometry in fractionated dissociates from cortex three days after injury demonstrated the presence of both cDCs and pDCs in the injured neocortex ( Figures 3A and B). Cd11c encoded by Itgax (alpha-X integrin) was used as a marker for cDCs [13]. pDCs [11] were characterized by PDCA-1 whereas CD45 (encoded by Ptprc) was used as a pan-leukocyte marker [1,2]. Optimal gating for the two classes of dendritic cells ( Figure 3A, left) identified double-positive cells ( Figure 3A, right). Isotype controls gave low background signals whereas Cd11c-and PDCA1-positive cells increased less in the Ccr2-deficient compared to wt brains ( Figure 3B). Uninjured brains showed only few cells double-positive for these markers. The data support the notion of impaired infiltration of peripheral inflammatory cells in mice lacking the Ccr2 chemokine receptor. in Ccl32/2 mice as compared to wt mice three days postinjury. Gfap expression was also strongly upregulated but with no differences between the strains. (d) The Ccl2 transcript, encoding a Ccr2 ligand, was upregulated after injury with no differences between wt and Ccl32/2 mice as shown by microarray analysis. In contrast, Selp transcript expressed by endothelial cells was more upregulated in Ccl32/2 compared to wt brains. (e) Lyz2 expression and the microglial surface marker isolectin B4 (IB4) showed only partial overlap three days post-injury in wt brains (in situ hybridization and histochemistry). The Bst2, characterizing pDCs, exhibited a different expression pattern (insert). (f) At three days postinjury, the Lyz2 increase shifted in opposite directions in the Ccl32/2 and Ccr22/2 compared to wt brains. (g) Cavity volume was reduced in the Ccr22/2 compared to wt mice seven days after injury in accordance with downregulation of inflammatory Lyz2-response. doi:10.1371/journal.pone.0104754.g001 Temporal shifts in inflammatory transcripts in the injured Ccr22/2 cortex Unbiased microarray analysis identified genes less upregulated in Ccr2-deficient cortex compared to wt three days postinjury. Data revealed over 50 genes being upregulated more than two-fold in injured wt brains while reaching less than half of this level in Ccr2-deficient mice. In addition to Cxcl10 and Lyz2, inflammatory transcripts included Arg1, Chi3l3, Ccr1, Ccr5, Cybb, H2-Aa, Ifi204, Ifi205, Ms4a4c, Stat1, Tgfbi and Tlr1. Microarray data were confirmed by qRT-PCR in a selection of these genes in Ccr22/2 mice and revealed persisting transcriptional differences one and two weeks postinjury ( Figure 3C). Uninjured brains showed only trace expression of the examined transcripts. A reduction in injury-increases in the Ccr22/2 mice was seen for Itgax previously shown to be expressed in immune cells in the vicinity of Cxcl10-positive cell clusters [1,2]. Cortical Itgax levels were reduced in the Ccr22/2 compared to wt brains and Ccr22/2 at three and seven days after injury. However, two weeks post-injury levels had increased equally in wt and mutant brains, indicating a delayed recruitment of antigen-presenting cDCs when Ccr2 is missing. The H2-Aa transcript encoding an alpha chain in the antigen-presenting major histocompatibility complex class II (MHC II) was upregulated at three, seven days and two weeks postinjury in wt mice. The lower expression persisted in injured Ccr22/2 mice.
The Bst2 transcript was distinctly upregulated in the injured wt cortex three days after injury but to a more modest degree in the Ccr22/2 mice ( Figure 3C). At seven days after injury, levels of Bst2 in the mutants reached the same levels as in wt mice. A slight reduction from these levels was seen two weeks after injury in both wt and Ccr2-deficient mice. The lack of a Cxcl10 peak three days post-injury in the Ccr22/2 brains resembles the delayed expression of Bst2. However, at seven days postinjury the Ccr22/2 mutant had caught up with wt Cxcl10 expression. The difference in Lyz2 upregulation between wt and Ccr22/2 seen at three days did not persist at peak levels seven days after injury. In contrast, the heme oxygenase (decycling) 1 (Hmox1) transcript located juxta-positioned to the focal injury [1], was similarly upregulated in injured wt and mutant brains with a peak after three days ( Figure 3C).

Genetic background does not account for the differential effects of chemokine signaling
We tested for possible contribution from the genetic background to the strikingly opposite effects of Ccl3and Ccr2-deficiencies ( Figure 4). Crossing the inbred knockout strains resulted in expected Mendelian ratios of homozygous Ccl32/2 and Ccr22/2 mice in male F2 offspring, in line with location of these genes on chromosome 11 and 9, respectively. Cortical injury increased Lyz2 levels in F2 Ccl32/2 mice whereas being reduced in F2 Ccr22/2 mice. The injury increases in Itgax and Cxcl10 transcripts did not reach above wt levels in the F2 Ccl32/2 mice in contrast to Bst2. In the F2 Ccr22/2 cortices, Lyz2, Itgax, Cxcl10 and Bst2 were robustly reduced confirming results from the parental strains even in a mixed genetic background. It is of note that our genotyping of Ccl3-deficient mice does not distinguish Ccr2+/+;Ccl32/2 from Ccr2+/2;Ccl32/2 and this heterogeneity may contribute to the lack of upregulation of Itgax and Cxcl10 in F2 Ccl32/2 mice. Irrespective of genotype and genetic background, Gfap levels were similar in all injured cortices.
Antigen presentation marker is limited to classical dendritic cells Next, we isolated dendritic cell populations from wt neocortex three days after injury using anti-Cd11c and anti-PDCA-1 antibodies coupled to magnetic beads. Subsequent RNA analysis of the inflammatory cells sorted on anti-Cd11c showed enrichment of Itgax and H2-Aa transcripts ( Figure 5A). Cells sorted on anti-PDCA-1 beads showed only weak expression of these two transcripts while highly expressing Bst2. In order to examine correlation between pDCs and Cxcl10 expression, antibodies directed against PDCA-1 were administered to TBI mice and cells fractionated from the brain for RNA isolation. qRT-PCR showed that Itgax and H2-Aa transcripts were induced to the same extent in the injured neocortex whether the injected antibodies were directed against PDCA-1 or represented control IgG immunoglobulins ( Figure 5B, left). In contrast, Bst2 and Cxcl10 transcripts were reduced in mice receiving anti-PDCA-1 antibodies compared to mice given equal doses of normal rat IgG ( Figure 5B, right). These data support pDCs as the source of Cxcl10 in injured neocortex.

Cyclophosphamide treatment resembles Ccr2 deletion by limiting injury-induced transcripts
We finally turned to intraperitoneal administration of the chemotherapeutic and anti-inflammatory agent cyclophosphamide to examine possible effects on injury-induced inflammatory cells in wt cerebral cortex three days postinjury. Microarray data from the cyclophosphamide-treated mice were compared to corresponding data from injured Ccr22/2 and wt mouse neocortex. Analysis of these three groups identified 20 genes that fulfilled the criteria of being upregulated more than two-fold in injured wt neocortex and reaching levels less than half of this in both Ccr22/2 mice and cyclophosphamide-treated mice ( Table 3). Several of these transcripts are involved in inflammatory processes e.g. Ccr2, Chi3l3, H2-Aa, Ifi204, Ms4a4c, Plac8, Stat1 and Tgfbi. The Ccr2 reduction by cyclophosphamide in concert with transcripts supporting antigen presentation indicates a major effect on cDCs (H2-Aa, Itgax; Figures 6A and B). Lyz2 and Cxcl10 transcripts (qRT-PCR; n.s.) were not affected whereas a slight reduction of Bst2 (qRT-PCR; P,0.05) suggests marginal influence of cyclophosphamide on pDCs but not phagocytes/macrophages. The upregulation of Cx3cr1 encoding the fraktalkine receptor and Ccl3 ( Figures 6A and B), both characteristic of resident microglia, as well as Gfap and Hmox1 (qRT-PCR; n.s.), were not affected by the cyclophosphamide treatment.  Flow cytometry three days after injury treated with cyclophosphamide showed a robust drop in number of inflammatory CD45 (Ptprc) and Cd11c (Itgax) double-positive cells isolated from neocortex ( Figure 6C) supporting RNA data at the protein/cell levels. Double isotype labeling in the area gated for cDCs yielded low background. The reduced Cd11c expression after cyclophosphamide treatment was confirmed by repeated experiments ( Figure 6C, bottom right). Overall, these results demonstrate that it is feasible to interact pharmacologically with selected inflammatory cells after brain trauma as summarized in Table 4.

Discussion
The present findings confirm that interactions among chemokines in the injured brain set the stage for inflammatory cell activation. While resident microglia and astrocytes become engaged, a flow of invading immune cells from the periphery after trauma is also evident.
In particular, the current data demonstrate a strong activation of dendritic cells in the injured brain depending on immigration of Ccr2 positive cells from the peripheral circulation [14]. Our data support a pivotal role of infiltrating cells of monocyte lineage in disease progression as demonstrated in previous reports using parabiotic mice and genetic deletion of the Ccr2 receptor [15]. We show that this diapedesis is valid for cDCs [4], pDCs [16] and large phagocytes expressing Lyz2. The activation of dendritic cells in the injured cortex is distinct from the inflammatory response among resident, locally renewing microglia [17][18][19] characterized by Cx3cr1 expression [20] and found to exert important surveillance functions in the brain [3].
cDCs express integrin alpha X (Itgax/Cd11c) and major histocompatibility class II complex (MHC II, e.g. H2-Aa) [21][22][23]. These cells coordinate a host of immune responses by capturing and processing proteins to peptides that are presented to T cells on MHC II [4]. Our data link H2-Aa with Itgax-expressing DCs serving as professional antigen presenting cells. These cells have been demonstrated to sufficiently trigger auto-reactive encephalitogenic T cells and to initiate CNS inflammation [24]. A distinct focal co-localization of cDCs and T cells in the inflamed brain, highly reminiscent of the current injury-induced inflammatory cell clusters, was shown and suggested to be indicative of CNS acting as a neolymphoid tissue [24]. The present results also connect pDCs, characterized by the Bst2 transcript [11,25] and its encoded protein PDCA-1, with the expression of Cxcl10 in injured mouse cortex, in agreement with reports on CXCL10 production in activated human pDCs [26,27].
Injury-induced activation of the chemokine Ccl2/Ccr2 axis is demonstrated by the current data. Previously, monocytes have been identified to express Ccr2 as well as high levels of Ly-6C and found to be crucial for autoimmune inflammation in the CNS [28], offering a target for treatment strategies [29]. An important role for the Ccr2 receptor in Ccl2 attraction of monocytes has been demonstrated [10]. Moreover, using a selective Ccr2 antagonist reduced apoptosis while improving behavioral performance in a rat TBI model [30]. Currently, a drastic increase in Ccr2-linked transcripts was observed in the injured Ccl3-deficient brains. This is not attributable to increased Ccl2 levels attracting inflammatory Ccr2 positive cells. A plausible explanation is offered by enhanced upregulation of platelet selectin, encoded by Selp, found in the vasculature of the injured Ccl32/2 brains. The Ccl32/2 mice have been shown to exhibit normal numbers of hematopoietic progenitor cells in peripheral blood and bone marrow and of T cells in lymph nodes and spleen [9]. The pathway by which Ccl3 is regulating Selp expression remains unknown but may result in curtailed endothelial expression of Selp and position Ccl3 as a suppressor of Ccr2-expressing cells in the injured neocortex.
The present data also demonstrate that large phagocytic cells expressing Lyz2 depend on invading Ccr2-positive cells and do not overlap with activated endogenous microglial cells. The injuryinduced cortical cavity formation was reduced in our Ccr22/2 mice after seven days. This is in parallel with the reduction of Lyz2 and adds to the idea that Ccr2 is a key receptor in recruitment of different invading peripheral cells. A knock-in strategy to express Cre recombinase from the Lyz2 locus in mice [31] showed high expression of Lyz2 in macrophages and only low expression in Cd11c+ splenic dendritic cells. Also, absence of Ccr2 reduced infarct sizes in a mouse model of cerebral ischemia/reperfusion injury [32], correlating with the current demonstration of reduced cortical cavity after TBI in Ccr2-deficient mice.
The alkylating agent cyclophosphamide affects DNA synthesis e.g. in immune cells in mice [33], acting as an immunosuppressant with clinical applications in autoimmune diseases such as systemic lupus erythematosus and rheumatoid arthritis. Presently, an impaired invasion of inflammatory cells in injured wt brain Table 3. Transcripts increased at least two-fold three days postinjury in wt cerebral cortex and with upregulation reduced by at least 50% in Ccr22/2 mice as well as in wt mice treated with cyclophosphamide. receiving cyclophosphamide administration was shown by restriction of injury-evoked Ccr2, Itgax and H2-Aa expression linked to a reduction of inflammatory Cd11c/Itgax positive cDCs. In contrast, cyclophosphamide did not affect the injury-elicited upregulation of Ccl3 or Cx3cr1 in resident microglia, had only marginal effects on pDCs and showed no manifest influence on phagocytes.
To what extent do the current findings from the injured mouse brain apply to human clinical conditions? A poor resemblance of inflammatory responses in humans and mice has been suggested from large scale genomic studies of systemic inflammation [37]. One particular case is the Selp promoter that is strongly upregulated in the mouse by various insults but unresponsive to inflammation in humans [38]. However, Selp may have other functional counterparts in the endothelia of patients. Nevertheless, in general the present findings are in line with observations of both cDCs and pDCs in patients with neurological conditions involving inflammation [39]. Also, several of the currently identified inflammatory transcripts are associated with human injuries. Thus, CCL2 and CCL3 transcripts were both increased in patients with posttraumatic brain contusion [40]. CXCL10, CCL3 and CCL2 were upregulated in plasma and microdialysis perfusates in TBI patients [41,42] and CCL2 found elevated in cerebrospinal fluid [43]. Moreover, increased levels of the CCL2 protein in serum of both civilian TBI patients and military blast-induced mild TBI cases have been demonstrated and found to be a potential risk factor for subsequent dementia [44]. Finally, the human orthologue of mouse H2-Aa, known as HLA-DQA1, encodes one of the HLA class II alpha chains and has been associated with autoimmune conditions including asthma, myasthenia gravis and celiac disease [45][46][47]. Taken together, our results suggest similarities in the studied systems between man and mouse.
The present data demonstrate separate repertoires among invading inflammatory cells of monocytic lineage distinct from those of activated resident cells. Moreover, a potential of dampening specific inflammatory cells invading the injured brain by pharmacological means is revealed. Thus, our findings suggest a time-window for therapeutic interference of invading Ccr2positive, antigen-presenting cells after traumatic brain injury.

Traumatic brain injury (TBI)
Male C57BL/6 (B6) mice with a body weight of 25-35 g were used for the traumatic brain injury (TBI). The mice were anaesthetized with 3.5% isoflurane (in combination of 70% nitrous oxide and 30% oxygen) before being transferred to a stereotactic frame. Anaesthesia was continued with 1-2% isoflurane, using mask ventilation and spontaneous breathing for the rest of the procedure. Bupivacaine (0.5 mg) was injected subcutaneously in the neck providing local analgesia. Body temperature (37uC) was controlled rectally throughout surgery. A craniotomy (approximately 3 mm diameter) was made over the right parietal cortex between midline, bregma and lambda. The mice were subjected to controlled cortical impact (CCI) injury [7,8] by a pneumatic impact device (model AMS 201, AmScien Instruments, Richmond, Virginia, USA). The compression depth was 0.5 mm, compression duration 100 ms and the velocity 3.1 m/s, resulting in a severe, focal injury. Control tissues were from uninjured mice without craniotomy. All efforts were made to minimize animal suffering in compliance with ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines. Uppsala Ethical Committee on Animal Experiments evaluated and authorized all experimental protocols.

In situ hybridization
Mice were deeply anaesthetized and perfused by a cardiac infusion of sodium chloride followed by 4% paraformaldehyde. Brains were postfixed in formaldehyde overnight. A vibratome was used to cut 60 mm coronal slices collected in phosphate-buffered saline, dehydrated in methanol and stored at 220uC over night. Sections were rehydrated in methanol and saline containing 0.1% Tween-20, bleached in 6% H 2 O 2 , treated with 0.5% Triton X-100 and permeabilized with proteinase K and postfixed in formaldehyde. Antisense riboprobes encoding Ccl3, Cxcl10, Lyz2 and Bst2 were synthesized from linearized IMAGE clones using SP6, T3 or T7 polymerase and DIG RNA Labeling Kit (Roche, Mannheim, Germany). Analysis was based on several sections respresenting at least two injured brains subjected to repeated hybridizations giving similar results. Sections were prehybridized two hours at 55uC. Probes (1 mg/ml) were denatured at 80uC and added to the sections for incubation at 55uC over night. Anti-DIG antibody was diluted 1:5,000 in blocking solution for an overnight incubation at 4uC. Levimasole was used to inhibit endogenous phosphatase. Finally, sections were developed with BM-purple alkaline phosphatase substrate (Roche) at 37uC before being washed in saline, mounted onto microscope slides and photographed.

RNA preparation
Neocortex from the injured side of the brain was dissected and stored in the RNAlater reagent (Qiagen Inc., Valencia, California, USA). The tissue was homogenized using a Polytron homogenizer and total RNA isolated by RNeasy Mini kit (Qiagen) with absorbance determined at 260 and 280 nm. RNA was prepared from uninjured wt mice (n = 10) and at the following time points after injury: one hour (n = 3), four hours (n = 4), twenty-two hours (n = 4), three days (n = 13), seven days (n = 11), two weeks (n = 12), three weeks (n = 9) and three months (n = 9). Moreover, the Cxcl102/2 RNA consisted of uninjured mice (n = 2) and injured mice at three days (n = 7) and seven days (n = 9) postinjury. RNA from Ccl32/2 brains were represented by uninjured mice (n = 2) and injured at three days (n = 9) and seven days postinjury (n = 5). The Ccr22/2 RNA comprised of two uninjured brains and postinjury at three days (n = 9), seven days (n = 11), two weeks (n = 11) and three weeks (n = 6). The RNA from the injured F2 hybrids consisted of five brain samples of Ccl32/2 mice and six samples of Ccr22/2 mice that were compared with five injured wt mice.

Quantitative reverse-transcriptase PCR (qRT-PCR)
Total RNA (10 ng) was analyzed and measurements were repeated at least twice using duplicated microwells (25 ml reaction volume). Injury-induced transcriptional changes were studied using the following primer pairs (reference sequence number and upper as well as lower primer stated): Ccl3 (NM_011337, 59-GCC TGC TGC TTC TCC TAC AG-39 and 59-TCT GCC  GGT TTC TCT TAG TC- Step RT-PCR Kit with SYBR Green (Bio-Rad Laboratories, Inc., Hercules, California, USA) was used. Reverse transcription was run for 10 min at 50uC. Thereafter, qRT-PCR was initiated by a hot-start at 95uC during 5 min followed by 36 cycles (95uC for 10 s, 60uC for 30 s) using a CFX96 thermal cycler (Bio-Rad). Melting curves were obtained by temperature increments of 0.5uC from 55.0uC to 94.5uC. Expression of 28S rRNA (X00525, 59-GGG AGA GGG TGT AAA TCT CGC-39 and 59-CTG TTC ACC TTG GAG ACC TGC-39) was used as a reference. Threshold cycle differences (DDCt) were transformed to linear fold changes with reference to equal amounts of total RNA in uninjured cortex.

Genetically modified mice
TBI was performed in three strains of chemokine signaling deficient mice. Ccl3 knockout B6.129P2-Ccl3 tm1Unc /J (JAX #002687) [9] and Ccr2 knockout B6.129S4-Ccr2 tm1Ifc /J (JAX #004999) [10] were obtained as homozygous females crossed with our local B6 mice. Cxcl10 knockout B6.129S4-Cxcl10 tm1Adl /J (JAX strain #006087 Jackson Laboratories, Bar Harbor, Maine, USA) [12] was obtained as homozygous breeding pairs as described previously [2]. The strains were then maintained by crossing heterozygotes or homozygotes to produce the required number of homozygous males for TBI. All experiments used knockout mice backcrossed for at least nine generations with B6 wt mice. The three chemokine-defect strains of mice exhibit no overt phenotypic deficiencies, impaired development or fertility, nor did the targeted alleles deviate from expected Mendelian ratios. Control wt B6 males were from our local breeding colony or obtained commercially. In order to test for any contribution from the genetic background of the knockout mice, hybrid F1 Ccl3+/2 ;Ccr2+/2 mice were created to serve as parents for the F2 hybrid generation. For genotyping, tail-tips from three-week-old pups were taken and the pups marked by ear tags. The tail-tips were put in 0.3 ml lysis buffer (0.5 ml 0.1 M Tris-HCl buffer pH 8.5 with 5 mM EDTA, 0.2% SDS and 0.2 M NaCl with proteinase K freshly added to 0.1 mg/ml final concentration) and incubated overnight at 60uC. For DNA-precipitation, equal volume of isopropanol was added. Genotyping by PCR was performed with the primers specified below using AmpliTaq Gold polymerase (Applied Biosystems, Foster City, California, USA) using the following steps: 35 (NM_011337, 59-GCC TGC TGC TTC TCC TAC AG-39 and 59-TCT GCC  GGT TTC TCT TAG TC-39) gave a fragment length of 350 base pairs. In addition, to detect the neomycin resistance cassette we used primers (U43612, 59-CTT GGG TGG AGA GGC TAT  TC-39 and 59-AGG TGA GAT GAC AGG AGA TC-39) yielding an amplicon length of 280 base pairs. Fragment size was evaluated by gel electrophoresis (using 1.4-2.0% agarose gels with 0.005% ethidium bromide) followed by examination by UV light in a ChemiDoc imager (Bio-Rad, Hercules, California, USA).

Histochemistry
To detect activated microglia, peroxidase-labeled isolectin B4 from Bandeiraea simplicifolia (L5391, Sigma, Saint Louis, Missouri, USA) was applied to sections, previously hybridized with the Lyz2 probe. Binding of isolectin B4 was examined in duplicate sections by adding DAB as a peroxidase substrate before being mounted.

Injury-induced cavity volume analysis
Mice were perfused by 4% paraformaldehyde one week after injury. The brains were removed, cryoprotected in 30% sucrose, snap-frozen in ice-cold isopentane and stored at 280uC. Brains were embedded in Tissue-Tek (Histolab, Gothenburg, Sweden) and cut in 20 mm coronal sections with a cryostat, beginning at bregma. Every 25th section was saved on glass slides (SuperFrost Plus, Menzel-Glä ser, Braunschweig, Germany). After nuclear staining in Harris hematoxylin (Histolab), slides were dehydrated and mounted. Images of sections of the injured brains were stored in a digitized format for area measurements in order to reconstruct the cavity volume. Ipsilateral area was subtracted from the corresponding contralateral and area difference was transformed into a circular area to calculate the radius. The total cavity volume was approximated as the sum of the volume of each truncated cone, stacked in the injury. Measurements were performed in six Ccr22/2 mice and five wt mice.

Magnetic cell sorting
Inflammatory cells fractionated on Percoll gradients, were incubated with magnetic microbeads coated with anti-Cd11c or anti-PDCA-1 antibodies (Miltenyi Biotech, Bergisch Gladbach, Germany). Four independent experiments were performed, each analysis based on cells pooled from neocortex from two up to six injured mice. The magnetic cell sorting was accomplished using MS Columns and a MiniMACS separation unit (Miltenyi Biotech) for subsequent recovery of RNA. RNA was analyzed in duplicate with qRT-PCR using Itgax, H2-Aa, Bst2 and Cxcl10 probes and normalized against 28S ribosomal RNA.

Depletion of plasmacytoid dendritic cells
Wt mice (n = 6) were injected intraperitoneally 30 minutes after injury with 500 mg rat antibodies directed against mouse PDCA-1 (# 130-091-978, Miltenyi Biotech, Bergisch Gladbach, Germany) to achieve depletion of pDCs. As reference, the same amount of normal rat IgG (I4131, Sigma Aldrich Sweden AB) were administered to wt mice (n = 8). Brains were collected for RNA preparation three days after injury.

Cyclophosphamide treatment
Wt mice were given phosphate buffered saline (n = 6) or cyclophosphamide monohydrate (n = 8), 200 mg/kg; C0768, Sigma Aldrich) intraperitoneally 30 minutes after the TBI and tissue collected three days later for RNA preparation for qRT-PCR and microarray analysis. In addition, injured mice given saline (PBS) or cyclophosphamide were analyzed by flow cytometry.

Statistical analysis
The SigmaStat version 3.1 software (SPSS, Inc., Richmond, CA, USA) was used to perform One Way Analysis of Variance (ANOVA) or Student's t-test for pair-wise comparisons of groups. Non-parametric data were analyzed using Kruskal-Wallis One Way Analysis of Variance on Ranks or Mann-Whitney U test as advised by the program. P = 0.05 or lower was considered to represent statistically significant differences. Data in the text and graphs are presented as mean values 6 standard error of the mean (SEM).