CD93/AA4.1: A Novel Regulator of Inflammation in Murine Focal Cerebral Ischemia

This information is current as Denise Harhausen, Vincent Prinz, Gina Ziegler, Karen of October 4, 2021. Gertz, Matthias Endres, Hans Lehrach, Philippe Gasque, Marina Botto, Philip F. Stahel, Ulrich Dirnagl, Wilfried Nietfeld and George Trendelenburg J Immunol 2010; 184:6407-6417; Prepublished online 3 May 2010; doi: 10.4049/jimmunol.0902342 Downloaded from http://www.jimmunol.org/content/184/11/6407

References This article cites 58 articles, 22 of which you can access for free at: http://www.jimmunol.org/ http://www.jimmunol.org/content/184/11/6407.full#ref-list-1

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists by guest on October 4, 2021

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2010 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

CD93/AA4.1: A Novel Regulator of Inflammation in Murine Focal Cerebral Ischemia

Denise Harhausen,* Vincent Prinz,* Gina Ziegler,† Karen Gertz,* Matthias Endres,‡,x Hans Lehrach,† Philippe Gasque,{ Marina Botto,‖ Philip F. Stahel,# Ulrich Dirnagl,* Wilfried Nietfeld,† and George Trendelenburg*

The stem-cell marker CD93 (AA4.1/C1qRp) has been described as a potential complement C1q-receptor. Its exact molecular func- tion, however, remains unknown. By using global expression profiling we showed that CD93-mRNA is highly induced after transient focal cerebral ischemia. CD93 is upregulated in endothelial cells, but also in selected and . To elu- cidate the potential functional role of CD93 in postischemic brain damage, we used mice with a targeted deletion of the CD93 . After 30 min of occlusion of the middle cerebral artery and 3 d of reperfusion these mice displayed increased leukocyte infiltration 6 6 3 into the brain, increased edema, and significantly larger infarct volumes (60.8 52.2 versus 23.9 16.6 mm ) when compared with Downloaded from wild-type (WT) mice. When the MCA was occluded for 60 min, after 2 d of reperfusion the CD93 knockout mice still showed more leukocytes in the brain, but the infarct volumes were not different from those seen in WT animals. To further explore CD93- dependent signaling pathways, we determined global transcription profiles and compared CD93-deficient and WT mice at various time points after induction of focal cerebral ischemia. We found a highly significant upregulation of the chemokine CCL21/ Exodus-2 in untreated and treated CD93-deficient mice at all time points. Induction of CCL21 mRNA and protein was confirmed

by PCR and immunohistochemistry. CCL21, which was formerly shown to be released by damaged neurons and to activate http://www.jimmunol.org/ microglia, contributes to neurodegeneration. Thus, we speculate that CD93-neuroprotection is mediated via suppression of the neuroinflammatory response through downregulation of CCL21. The Journal of Immunology, 2010, 184: 6407–6417.

rain tissue damage after cerebral ischemia is exacerbated by flammation, some chemokines, including chemokine ligands CCL19 a complex pattern of pathophysiologic mechanisms including and CCL21, are constitutively expressed and control cell movement B excitotoxicity, peri-infarct depolarizations, inflammation, and during homeostasis (9). (1–4). Besides proinflammatory cytokines, such as IL-1b Whereas a plethora of linked to the pathophysiology and TNF-a, TLRs and several chemokines have been described to of cerebral ischemia has already been identified using candidate by guest on October 4, 2021 mediate the inflammatory response in the CNS (2, 5, 6). Chemokines approaches, screening techniques have proved to be valuable in are a family of that direct leukocyte migration and activation to identifying novel neuroprotective signaling pathways in the post- inflammatory stimuli. Depending on the position of conserved cysteine ischemic brain (10–15). Analysis of expression profiles of .39,000 residues, chemokines have been subdivided into the families C, CC, mRNA transcripts in brain hemispheres of C57BL/6 wild-type CXC, and CX3C (5, 7). Chemokines have been characterized as (WT) mice at various time points after induction of cerebral is- playing a role in balancing the central and peripheral immune re- chemia using Affymetrix GeneChip arrays revealed nearly 2000 sponses between lymphoid and nonlymphoid tissues (8). Although the genes, which were significantly upregulated at least 2-fold, with p , expression of most chemokines is induced during infection and in- 0.05 (G. Trendelenburg and W. Nietfeld, unpublished observations). Several of the identified genes have been associated already with focal cerebral ischemia in candidate approaches or previous x *Experimentelle Neurologie, ‡Klinik fu¨r Neurologie, and Center for Stroke Re- screening experiments, but many represent new candidates. Several search, Charite´-Universita¨tsmedizin; †Max Planck Institute for Molecular Genetics, of these induced genes are related to inflammation or apoptosis Berlin, Germany; {Laboratoire Biochimie et Ge´ne´tique Mole´culaire, Universite´ de la ‖ (KEGG Pathway Database [www.genome.jp/kegg/pathway.html]; Re´union, Ile de la Re´union, Cedex 9, France; Molecular Genetics and Rheumatology Section, Imperial College, London, United Kingdom; and #Department of Orthopae- BioCarta Charting Pathways of Life [www.biocarta.com/genes/ dic Surgery, Denver Health Medical Center, University of Colorado School of Med- index.asp]) (15–17). This may also be the case for the CD93 gene icine, Denver, CO 80204 (1419589_at). Received for publication July 21, 2009. Accepted for publication March 23, 2010. The 126-kD CD93 protein was initially characterized as one of This work was supported by Grant DFG No. TR742/1-1, 1-2 from the German Re- several putative complement C1q receptors and was given the name search Society (to G.T., P.F.S., and U.D.), the SFB TR43 project A3 (to M.E. and U.D.), the Lichtenberg Professorship program (founded by the Volkswagen founda- C1qRp (18–21). However, subsequent analysis revealed that CD93 tion), the Center for Stroke Research (founded by the Federal Ministry of Education interacts with C1q only under nonphysiologic conditions (18, 22). and Research), and the Hermann and Lilly Schilling Foundation. CD93/C1qRp was also characterized as a Address correspondence and reprint requests to Dr. George Trendelenburg, Exper- marker Ag and named AA4.1 (23). A subset of CD93-positive bone imentelle Neurologie, Charite´-Universita¨tsmedizin Berlin, CCM, Charite´platz 1, D- 10117 Berlin, Germany. E-mail address: [email protected] marrow progenitor cells express neural stem cell markers and thus Abbreviations used in this paper: Affy, Affymetrix GeneChip analysis; C, contralateral; may differentiate into neural cells (24). In addition, surface CD93 fc, fold change; I, ipsilateral; Illum, Illumina hybridization assay; LTB, lymphotoxin b; expression has proved to be inducible by inflammatory stimuli (25, LTBR, lymphotoxin-b receptor; MCAO, middle cerebral artery occlusion; MT, metal- 26). The molecular function of CD93, however, is still an enigma. lothionein; Pdpn, podoplanin; TM, TaqMan PCR; U, untreated; WT, wild-type. CD93 expression has also been found in microglia, as well as in Copyright Ó 2010 by The American Association of Immunologists, Inc. 0022-1767/10/$16.00 a subset of pyramidal neurons, but the vascular has www.jimmunol.org/cgi/doi/10.4049/jimmunol.0902342 6408 CD93/AA.1 IN EXPERIMENTAL STROKE been described as the predominant site of CD93 expression (18, 27, derived from the ipsilateral brain hemisphere of MCAO-treated mice. 28), particularly during the remodeling of the vascular tree (29). Commercially available Affymetrix software and Beadstudio (Illumina, San It was hypothesized that CD93 is involved in angiogenesis and Diego, CA) were applied for data analysis of Affymetrix GeneChips and Illumina Sentrix array, respectively. CD93 expression was assessed by the endothelial , and possibly also in leukocyte extrava- probe 1419589_at (Affymetrix-ID) and scl18542.4.1_65-S (Illumina-ID), sation at postcapillary venules (18, 22, 30, 31). Nevertheless, CD93- respectively. CCL21-expression was assessed by probes scl0065956.1_31-S, deficient mice (CD932/2) are viable and showed no gross abnor- scl0018829.1_65-S, and scl0020298.1_98-S (Illumina-ID). 2/2 malities in their vascular development. However, CD93 mice do Real-time PCR have a defect in the phagocytic removal of apoptotic cells in vivo Isolated RNA (Trizol, Invitrogen) was transcribed into cDNA (SuperScript (18), demonstrating that CD93 contributes to the in vivo clearance of II, Invitrogen); expression of mRNA was ascertained by using custom dying cells. We therefore hypothesized that CD93 participates in configured TaqMan low-density arrays and normalization as recommended postischemic brain injury, where much cellular debris has to be by the manufacturer (Applied Biosystems, Foster City, CA). CD93 ex- removed. In this study, we explored whether CD93 mediates the pression was monitored using assay ID Mm00440239_g1 (Applied Bio- 9 neuroinflammatory changes observed in postischemic brain injury. systems) with primers NM_010740S8 and NM_010740R8 (5 -GGA CGA AGA GAC CTG TTG T-39 and 59-AGA TGA GTG TTC GGA CGC-39, respectively) (www.ncbi.nlm.nih.gov/nuccore/144922670). Materials and Methods To verify CCL21 expression results obtained by Illumina Sentrix arrays, Animals CCL21-mRNA levels were analyzed by real-time PCR using LightCycler FastStart DNA Master SYBR Green I (Roche Applied Science, Mannheim, 2/2 The CD93 mice, generated as previously described (18), had been Germany) with the following primers: forward 59-AAT CCT GTT CTTACC backcrossed onto the C57BL/6 background for 10 generations. Genotype CCG GAA GCA CTC TAA -39, and reverse 59-CTC TTG AGG GCT GTG 2/2 was determined by PCR as described below. Adult 10–12-wk-old CD93 TCT GTT CAG TTC TCT-39, according to National Center for Bio- Downloaded from mice and age-matched C57BL/6 control mice (BfR, Berlin, Germany) were technology Information accession number NM_011124.4 (www.ncbi.nlm. used. All animal handling and surgery was performed in accordance with nih.gov/nuccore/118130240). For quantification purposes, the cycle at the Guidelines for the Use of Animals in Neuroscience Research (Society which the threshold for detection of the fluorescence signal was crossed was for Neuroscience, www.sfn.org/index.cfm?pagename=guidelinesPolicies_ determined (ct), and used for calculation of relative expression (fold change UseOfAnimalsandHumans). All experiments were approved by the in- [fc]). The relative expression (fc) of CCL21-mRNA in CD93-deficient mice stitutional Animal Care Committee, (LAGeSo No. G0382/05). The mice was calculated using the following formula: fc 5 22DDct , where DDct 5 were bred in a selective pathogen-free environment and under standardized DctCD93 2 Dctwt, where DctCD93 5 ctCCL21 in CD932∕2 mice 2 ctb‐actin in CD932∕2 conditions of temperature (21˚C), humidity (60%), and and dark cycles mice, and where Dctwt 5 ctCCL21 in wt mice 2 ctb‐actin in wt mice. http://www.jimmunol.org/ (12 and 12 h each), with food and water provided ad libitum. Western blotting Genotyping Cerebral hemispheres of mice (n = 3 for each time point) were dissected at Genotyping was performed using the REDExtract-N-Amp Tissue PCR different time points after the ischemic insult and homogenized in lysis (Sigma-Aldrich, St. Louis, MO) with DNA extracted from mice tails and the buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Na- + 9 following primers: CD93/W (5 -AGG GAT CCC AGC GAG GAA GGG DodSO4 (SDS), 1% Triton-X-100, 1% deoxycholate acid and 103 pro- CAA CTG-39), neomycin 39-(59-GGG ATC GGC AATAAA AAG AC-39), 2 tease inhibitor mixture, using a sonicator (Bandelin, Sonoplus HD70). The and CD93/U (59-GTC CTG GCA CTC ATC TAT ATC-39) (18). Ampli- lysate was incubated for 10 min on ice and centrifuged, and the resulting fication of the WT gene resulted in an amplicon of ∼721 bp, whereas the supernatant was taken for determination of protein concentration according targeted gene corresponded to an ∼480-bp fragment.

to the BCA assay protocol (Pierce, Rockford, IL). Western blotting was by guest on October 4, 2021 Induction of focal cerebral ischemia performed according to the protocol of Laemmli et al. (33). Protein was denatured in nonreducing loading buffer containing 125 mM Tris-HCl (pH Middle cerebral artery occlusion (MCAO) was induced by inserting a 6.8), 1% SDS, 20% glycerol, and 0.008% bromphenol blue, and was silicone-coated 8/0 nylon monofilament (Xantopren M Mucosa and Acti- loaded, 70 mg per lane, on 7.5% SDS-polyacrylamide minigels, followed vator NF Optosil Xantopren, Heraeus Kulzer, Wehrheim, Germany) via the by electrophoresis and semidry blotting onto nitrocellulose membrane internal carotid artery as described by Hara et al. (32). As a sham control, (Serva, Heidelberg, Germany). Incubation with a polyclonal rabbit Ab the operation was performed without advancing the filament. Sufficiency raised against mouse CD93 (30) was performed with a titer of 1:1000 after of occlusion and reperfusion of the middle cerebral artery was monitored the membrane was blocked with 5% BSA. For chemiluminescence stain- by Laser Doppler flowmetry (Peri Flux 4001 Master, Perimed, Stockholm, ing, a secondary anti-rabbit HRP-linked Ab (Cell Signaling Technology, Sweden). Mice were anesthetized with 2% isoflurane for induction and Danvers, MA) in combination with Luminol detection reagent (Santa Cruz maintained with 1.5% isoflurane, 70% N2O, and 30% O2 via a face mask. Biotechnology, Santa Cruz, CA) was used. A broad range, prestained SDS- The period of anesthesia did not exceed 10 min. After 30 min or 1 h of PAGE protein standard (Invitrogen) was used to determine protein sizes. ischemia, the animals were re-anesthetized and the filament was removed Assessment of infarct volume to permit reperfusion. During surgery and ischemia, body temperature was measured and maintained between 37.0 and 37.5˚C with a heating pad. Two or 3 d after induction of ischemia for 60 min or 30 min, respectively, There was no significant difference in the mean body weight between the mice were deeply anesthetized and sacrificed. The brains were removed different groups. rapidly from the skull and snap-frozen in 2-methylbutane on dry ice. Brains were sectioned (12 mm) on a microtome, dried overnight, and stained with Microarray analysis of gene expression hematoxylin (Merck, Darmstadt, Germany). The sections were digitized, the At specified time points of reperfusion, mice were anesthetized deeply and area of infarction was quantified by using Sigma Scan Pro Version 5.0.0 decapitated. The brains were removed rapidly from the skull. RNA derived Software (Jandel Scientific, San Rafael, CA), and infarct volumes were from ipsilateral (ischemic) hemispheres, pooled from six WT C57BL/6 mice calculated. A correction for edema (brain swelling) was applied by calcu- (after MCAO, respectively sham treatment) at each time point, was used for lating the indirect infarct volume as the volume of the contralateral hemi- Affymetrix GeneChip experiments using Affymetrix GeneChip Mouse sphere minus the noninfarcted volume of the ipsilateral hemisphere. The Expression Set 430 (Affymetrix, Santa Clara, CA). For Illumina Sentrix difference between direct and indirect infarct volumes represents brain arrays(Illumina,SanDiego, CA),ipsilateraland contralateralhemispheres of swelling. Relative infarct size was calculated as a percentage of the size of the CD93-deficient and C57BL/6 WT micewere pooled from three male animals contralateral hemisphere. Different volumes were measured to control for in each group and time point. The RNA was isolated with Trizol Reagent a hypothetical (inverse) influence of edema on the infarct size. It has been (Invitrogen,Karlsruhe, Germany) and transcribedinto cDNA (SuperScript II, shown that direct measurement of infarct volume during the development of Invitrogen). To determine the induction of specific mRNAs in the ipsilateral brain edema is associated with an overestimation of infarct volume, and this hemispheres of MCAO-treated animals, expression values were compared artifact can be reduced with the procedure described above (34). with those derived from the uninjured (contralateral) hemisphere of these Immunohistochemistry animals, as well as to those derived from sham-operated animals. Thus, mRNAs of the contralateral hemispheres of MCAO-treated mice, as well as Staining was performed on fresh-frozen (CD93), respectively PFA-perfused mRNAs from the ipsilateral brain hemispheres of sham-treated WT mice (CCL21), tissue harvested at different times of reperfusion. PFA-perfused (C57BL/6),were used as a reference for determining the inductionof mRNAs brains were fixed in 4% PFA and dehydrated before they were embedded The Journal of Immunology 6409 into paraffin and then sectioned in 5-mm slides. From the fresh-frozen tissue, 12-mm coronar cryosections at interaural positions 6.6, 5.3, 3.9, 1.9, and 0 mm were thaw-mounted onto glass slides. Adjacent sections were used to determine stroke volume (see above). Slides were air-dried for 30 min and fixed for 10 min in 220˚C methanol and acetone (1:1). The sections were incubated in blocking solution containing 3% normal goat serum and 0.3% Triton X-100 (Sigma-Aldrich) in PBS for 60 min. The slides were incubated for 2 h at room temperature with a polyclonal rabbit anti-CD93 Ab (30) with a dilution of 1:200, or with a rat-anti CD93/ C1qRp Ab (eBioscience, San Diego, CA). For the detection of primary Ab, slides were incubated with Alexa488 conjugated goat anti-rabbit IgG (Invitrogen) diluted 1:300 for 1 h at room temperature. To determine the cell-type of CD93-positive cells, additional staining of endothelial cells was performed using rat anti-CD31 (550274; BD Pharmingen, San Jose, CA) and polyclonal goat anti-CD34 (1:25; sc-1407; Santa Cruz Bio- technology), whereas anti-CNP (Santa Cruz Biotechnology) was used to stain oligodendrocytes. Further cell type-specific double staining was performed using the following Abs: astrocytes, Guinea pig anti-GFAP (Advanced ImmunoChemical, Long Beach, CA); granulocytes, rabbit anti- myeloperoxidase (MPO; Chemicon, Temecula, CA; 1:50); microglia and macrophages, goat anti-Iba1 (ab5076; Abcam, Cambridge, U.K.); neurons, mouse anti-NeuN (Chemicon, Temecula, CA). The primary Abs were detected in the brain sections by goat anti-mouse, goat anti-rabbit, goat- anti rat, goat anti-guinea pig, or donkey-anti goat Abs conjugated, re- Downloaded from spectively, with Texas Red or Cy3 (Molecular Probes, Leiden, The Netherlands). Omission of the primary Ab was used to ensure specificity of FIGURE 1. Up-regulation of CD93-mRNA and protein in postischemic the secondary Abs; specificity of the CD93-Ab was verified by control mouse brain tissue. A, Fold change (FC) expression after various reper- 2/2 staining of sections derived from CD93 mice. A proprietary blocking fusion times in an MCAO model with 1 h of occlusion time determined by agent (M.O.M. kit, Vector) was used to minimize background signal from Affymetrix GeneChip analysis (Affy), real time TaqMan PCR (TM), and endogenous mouse IgG in the NeuN stains. Confocal detection of double Illumina hybridization assays (Illum). For each time point, RNA was

labeling studies was performed by confocal laser-scanning microscopy http://www.jimmunol.org/ pooled from ipsilateral postischemic hemispheres of C57BL/6 WT mice using Leica DM LFSA (Leica, Solms, Germany). For CCL21-staining, sections were deparaffinized three times in xylene, (n = 4 for Illum and TM; n = 6 for Affy). Expression values of ipsilateral followed by ascending alcohol series, and unmasked with trypsin (Abcam) hemispheres were compared with values of sham-treated control animals for 20 min at 37˚C. Nonspecific binding was blocked with 15% normal (Affy) or contralateral hemispheres (TM; Illum). CD93-gene specific donkey serum. Primary goat-anti CCL21 Ab (1:200; R&D Systems, primers were used (identifier numbers Mm00440239_g1 [FC TM], Wiesbaden, Germany) was incubated over night at 4˚C followed by scl18542.4.1 65-S [FC Illum], respectively 1419589_at [FC Affy]). n.d., staining using donkey anti-goat Cy3 (1:200). Slides were also counter- not determined. B, C57BL/6 WT mice underwent 1 h MCAO or were left stained with Hoechst 33258, which stains DNA (Invitrogen). untreated (control). After 12 or 48 h of reperfusion, protein was extracted Leukocyte quantification from the contralateral (C) and ipsilateral (I) hemispheres. Extracts were immunoblotted with a polyclonal rabbit anti-CD93 specific Ab. The arrow by guest on October 4, 2021 WT and CD93 knockout mice brain sections (see above) at different time denotes the 130-kDa band of the CD93 isoform; 70 mg total protein extract points of reperfusion were stained with CD11b Ab (see above for pro- (determined by BCA protein assay) per lane were used. Results are rep- cedure). CD11b-positive cells in the whole ipsilateral (ischemic) hemi- resentative of three separate experiments. spheres at the following positions were counted using a stereo investigator 7 (MicroBrightField Bioscience, Williston) stance to bregma 5.3 mm, 3.9 mm, and 1.9 mm (n = 3–4 animals). to investigate early mRNA expression patterns over time, gene Statistical analysis expression was also compared between ipsilateral and contralat- Students t test for independent samples was applied to determine statistical eral hemispheres (expression patterns over space). Gene expres- significance between the mean values of two study groups if not stated sion was independently determined from three pooled ipsilateral otherwise. Values of p , 0.05 were considered statistically significant. For and contralateral brain hemispheres, respectively, for each time comparison of infarct volumes between different groups Mann-Whitney U point (3, 12, 24, and 48 h) to compare gene-specific expression in test was used, and significance was accepted when the calculated U value the ischemic hemisphere not only with the sham-operated mice as was below the critical U value. reference (as done for the initial Affymetrix screening experi- Results ment), but also to allow comparison between ipsilateral and contralateral hemispheres. Upregulation of CD93 mRNA (Target CD93 mRNA is induced after focal cerebral ischemia in the ID: scl18542.4.1_65-S; Illumina Sentix arrays) in the ipsilateral mouse hemisphere was detected from 3 to 48 h (Fig. 1A). Upregulation of Expression values for .39,000 mRNA transcripts from six pooled CD93 mRNA in the ipsilateral, ischemic brain hemisphere was ipsilateral WT brain hemispheres were analyzed at 1, 3, 6, 12, and further confirmed by real-time PCR (TaqMan LowDensity Arrays, 24 h of reperfusion after 1 h of MCAO and compared with sham- probe ID Mm00440239_g1) in combination with an independent treated mice by using Affymetrix GeneChip arrays (Set 430). The murine tissue preparation (Fig. 1A). Affymetrix GeneChip results revealed a significant upregulation of CD93 mRNA during the first day of reperfusion, with a maxi- Expression of CD93 protein in ischemic mouse brain mal induction at 12 h after MCAO when compared with sham- Induction of CD93 protein in the ischemic hemisphere was con- treated mice (Fig. 1A). Moreover, mRNA of control genes, such as firmed by Western blotting. An ∼130-kDa band appeared in protein metallothionein-I/-II and , which interacts with CD93 (34), extract derived from both the ipsilateral (ischemic) and contralat- was shown to be induced as described by us and other groups eral hemispheres of postischemic mice at 12 and 48 h of reperfusion before (11, 14), whereas the expression of the CD93-homologous (Fig. 1B). This ∼130-kDa band corresponds to the 130-kDa band protein (35) did not exceed a 1.5-fold change as described by Dean et al. (31), shown to be expressed in mouse measured by Affymetrix GeneChip and RT-PCR (data not shown). spleen and brain. The ∼130-kDa CD93 protein was most promi- Apart from Affymetrix GeneChip analysis, which was performed nently induced at 48 h of reperfusion in the ischemic hemisphere 6410 CD93/AA.1 IN EXPERIMENTAL STROKE of WT animals, compared with the expression in contralateral and microglia, which mainly occurred at later time points of hemispheres, the expression at shorter reperfusion times, or the reperfusion. Some macrophages and microglia were stained expression in brain tissue of untreated animals (Fig. 1B). with anti-CD93 Ab in the penumbra at 72 h of reperfusion. In- frequently, CD93 colocalized with neurons of the peri-infarct Immunohistochemistry region (data not shown). In the infarct core, CD93 signal was The induction of CD93 protein in postischemic brain tissue of confined to selected macrophages and microglia, but at 48 h of WT mice was also observed by the use of immunohistochemistry reperfusion anti-CD93 Ab almost always stained vascular CD31- at various time points after MCAO, but not in the brains of CD93- positive structures with high intensity (Fig. 2). Selected CD93- deficient mice. A strong upregulation of CD93 was found at positive cells were located immunohistochemically in the direct various times of reperfusion in the ipsilateral hemisphere, with the vicinity of GFAP-positive cells and were found exclusively at the highest expression in the peri-infarct region and in the infarct border of the infarct core. No colocalization was observed with core. Anti-CD93 Ab stained vascular structures of the ischemic the oligodendrocyte marker CNP (data not shown). Confocal hemisphere intensively, whereas vascular structures of the non- laser-scanning microscopy confirmed CD93 expression in small ischemic contralateral hemisphere were stained only with low vessels (CD34-positive cells), but did not reveal a colocalization intensity. Double labeling with anti-CD31, anti-GFAP and anti- of CD93 and GFAP (Fig. 2B), despite an expression of GFAP and Iba1 (Fig. 2) identified the majority of CD93-immunopositive CD93 in direct vicinity (Fig. 2B,2D–F), indicating that as- cells as endothelial cells and selected infiltrating macrophages trocytes do not express CD93. Downloaded from

FIGURE 2. CD93 expression in ischemic brain tissue. A, Double staining of ischemic mouse brain tissue in the infarct border zone at 48 h after 1 h MCAO in WT mice

(a–f, k–m), at 48 h after 1 h MCAO of CD93- http://www.jimmunol.org/ deficient mice (o), at 72 h after 30 min MCAO in WT mice (g–j), and brain tissue of untreated WT (CD93+/+) mice (n). Immunohistochemical staining of CD93 and cell specific markers, respectively. Right column (c, f, j, m), Merge of the first (CD93-specific staining [a, d, g, k]) and second column (cellular markers [b, e, h, l]) is dis- played. Intensely stained CD93-positive cells in the is- chemic hemisphere (a, d, k) were found to be mostly endothelial cells (l and m, rat anti-CD31 [BD Pharmin- gen, San Jose, CA]). Some microglial cells (e, f, h,andj, by guest on October 4, 2021 goat anti-Iba1 [Abcam, Cambridge, U.K.]; i and j,visu- alization of cell nuclei using Hoechst 33258 DNA stain) also displayed weak CD93-positive staining at 48 h after 1 h MCAO (arrow in f), whereas at 72 h after 30 min various Iba1-positive cells were observed with intense CD93-staining (g–j). Most astrocytes do not colocalize with CD93-positive cells (b, c, guinea pig anti-GFAP [Advanced ImmunoChemical, Long Beach, CA]). In contrast to ischemic tissue of WT mice, staining of nonischemic mouse brain (n)revealedonlyaweak CD93-staining in endothelial cells. o, control staining using CD93-specific Ab in ischemic mouse brain tissue of CD93-knockout [CD932/2]miceat48hafter1h MCAO. Original magnification 340 (a–f, k–o)and3100 (g–j). Scale bar, 50 mm. B, Confocal laser-scanning mi- croscopy (Leica DM LFSA, Leica, Solms, Germany) of ischemic brain tissue of WT mice at 48 h after 1 h MCAO using rat anti-CD93 Ab (a, d) in combination with anti-CD34 Ab (b), respectively. Anti-GFAP Ab (e), which demonstrates CD93 expression in CD34- expressing (endothelial) cells, but argues against a co- localization of CD93 in GFAP-expressing cells (f). Double-labeling studies were performed using rat anti- mouse CD93/C1qRp Ab (eBioscience) in combination with goat anti-rat secondary Ab (Alexa 488; Invitrogen), and goat anti-CD34 (Santa Cruz Biotechnology) in combination with Cy3 donkey anti-goat Ab (Molecular Probes). Original magnification 3200 (a–f). The Journal of Immunology 6411

Infarct volumes of CD932/2 mice and WT controls at 2 d of inflammatory response (36, 37). In contrast to the findings after 60 min reperfusion after induction of transient focal cerebral ischemia MCAO (Fig. 3), direct infarct volumes in CD93-deficient mice of To test whether the CD93 protein is of functional relevance in the mixed gender were significantly increased compared with WT mice at standard murine stroke model, infarct volumes of male and female 72 h of reperfusion after 30 min MCAO (difference of direct infarct 6 2/2 6 CD93-deficient mice and matching WT mice after 1 hof MCAO and 48 volume in pooled gender as given by mean SD: CD93 : 60.8 52.2 mm3;CD93+/+:23.96 16.6 mm3; p =0.022(median[25th h of reperfusion were compared. Although therewas a tendency toward 2/2 +/+ larger infarctvolumes inthe CD93knockoutmicewhencomparedwith percentile; 75th percentile]: CD93 : 33.2 [18.3; 104.5], CD93 : WT mice, no significant difference was detected in the mixed-gender 18.5 [12.7; 24.3]) (Fig. 4). The difference proved not to be gender- groups (given by mean 6 SD; CD932/2:137.56 31.0 mm3;CD93+/+: specific, but was only statistically significant in male mice and mixed 6 3 gender. Direct infarct volumes in male mice as given by mean 6 SD: 121.7 36.2 mm ; p = 0.194; median [25th percentile; 75th percen- 3 2/2 3 +/+ 2/2 +/+ 74.0 6 54.0 mm /CD93 versus 27.5 6 19.2 mm /CD93 ; p = tile]; CD93 : 130.7 [116.3; 154.8], CD93 : 116.7 [93.7; 142.7]), 2/2 2/2 6 3 +/+ 0.020 (median [25th percentile; 75th percentile]: CD93 :59.2[27.1; in the female mice (CD93 :145.6 32.7 mm versus CD93 : +/+ 6 3 131.6], CD93 : 22.8 [13.1; 28.9]); direct infarct volumes in female 127.3 43.0 mm ; p = 0.316; median [25th percentile; 75th percen- 3 2/2 3 +/+ 2/2 +/+ mice: 31.7 6 37.7 mm /CD93 versus 18.0 6 4.7 mm /CD93 tile]: CD93 : 145.1 [124.0; 162.0], CD93 : 124.0 [95.1; 143.3]), or 2/2 2/2 6 3 +/+ (median [25th percentile; 75th percentile]: CD93 : 13.7 [13.2; in the male mice (CD93 : 125.3 26.1 mm versus CD93 :114.8 +/+ 6 26.6 mm3; p = 0.473; median [25th percentile; 75th percentile]: 20.3], CD93 : 14.0 [13.2; 20.3]). Differences of calculated indirect CD932/2: 120.0 [110.1; 138.4], CD93+/+: 111.3 [94.1; 124.8]) (Fig. 3). infarct volumes did not reach significance, but brain swelling (edema) significantly increases in CD932/2 mice when compared with WT 2/2 CD93 mice of both genders have an increased infarct mice (Fig. 4). Thus, brain swelling mainly contributes to increased volume when compared with WT-mice at 72 h after induction of direct infarct volumes (Fig. 4). Downloaded from transient focal cerebral ischemia BecauseoftheinvolvementascribedtoCD93inphagocytoticprocesses CD93 deficiency leads to increased inflammatory cell and angiogenesis, we analyzed infarct volumes in a second experi- infiltration in the postischemic brain tissue mentalstroke model, which does notcause infarction (i.e.,pannecrosis) Because of the role attributed to CD93 protein in leukocyte extrav- but rather selective neuronal cell death with a profound neuro- asation, its endothelial expression, and its structural homology to http://www.jimmunol.org/ by guest on October 4, 2021

FIGURE 3. Comparison of CD93-deficient and WT mice (C57BL/6) at 48 h of reperfusion and 60 min of transient MCAO. Direct infarct volumes (A, D, G), indirect infarct volumes (B, E, H), and volume of the edema (C, F, I) given as box-and-whisker plots at 48 h of reperfusion after induction of ischemia are visualized for both CD932/2 and WT mice for both genders (A–C)andmale(D–F)andfemale(G–I) mice separately. Statistical analysis was performed using the Mann-Whitney U test. The indirect infarct volume was calculated as the volume of the contralateral hemisphere minus the noninfarcted volume of the ipsilateral hemisphere. There is no significant difference between infarct sizes of CD932/2 and WT mice, albeit there is a small but insignificant tendency in both genders toward larger direct infarct volumes and edema in the CD932/2 mice. CD93 knockout total, CD93-knockout mice of both genders (n = 15); WT total, WT mice of both genders (n =18);CD93 knockout female, female CD93-deficient mice (n = 9); WT female, female WT mice (n = 10); CD93 knockout male, male CD93-deficient mice (n =6);WTmale,male WT mice (n = 8). In all box plots, the top of the box represents the 75th percentile, the bottom of the box represents the 25th percentile, and the line in the middle represents the 50th percentile. The whiskers (the lines that extend from the top and bottom of the box) represent the highest and lowest values that are not outliersorextremevalues. 6412 CD93/AA.1 IN EXPERIMENTAL STROKE other adhesion molecules (such as ) (30, 36), we asked CD93 gene. Forty-two genes were found to be induced $1.5-fold whether CD932/2 mice have increased leukocyte infiltration into the in untreated CD93-deficient mice compared with untreated WT postischemic brain. Indeed, CD11b staining, which stains invading mice, and only nine genes were found to be induced in untreated macrophages, , and activated microglia, demonstrated CD93-deficient animals at least 2-fold when compared with un- a significant increase of CD11b-positive cells in the ischemic treated WT mice (Table I). At 48 h of reperfusion after induction hemisphere of CD932/2 mice when compared with WT mice at 48 h of cerebral ischemia for 60 min, 18 genes were differentially (Fig. 5A) and 72 h (Fig. 5B) of reperfusion. The increased in- expressed at least 2-fold, and 85 genes at least 1.5-fold between flammatory cell counts in CD93-deficient mouse brains do not ap- the ipsilateral (ischemic) hemispheres of CD93-deficient and WT pear to be secondary to different infarct sizes, because CD11b- mice. Comparison of genes in CD932/2 mice with WT mice in the positive cell counts at 48 h differed significantly between the two ischemic hemisphere at 72 h reperfusion after induction of cere- strains (CD932/2 versus WT) despite similar infarct sizes (Fig. 5C). bral ischemia for 30 min revealed 58 with at least 2-fold induction (data not shown). Gene expression analysis of genes differentially expressed 2 2 In contrast, comparison of the number of genes induced in the between CD93 / mice and WT mice ischemic hemisphere at 48 h after 60 min MCAO versus the non- To further decipher CD93-dependent pathways, we compared ischemic (contralateral) hemisphere revealed 289 genes induced at global expression profiles of ischemic (ipsilateral) and nonischemic least 2-fold in CD932/2 mice, 211 of which are also induced in WT (contralateral) hemispheres of CD93-deficient mice with expres- mice (Fig. 6). Counting the number of genes induced at least 2-fold sion profiles of WT mice; 46,374 probes were investigated at each in the ischemic hemisphere at 48 h after 1 h MCAO, when compared condition by the use of Illumina Sentrix array technology (detailed with the expression level in the non-treated hemisphere, revealed information is accessible via GEO accession number GSE21002 or 439 genes in CD93-deficient mice and 360 genes in WT mice, of Downloaded from www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21002). After which 309 are induced in both strains (Fig. 6). 72 h reperfusion and induction of transient focal cerebral ischemia for 30 min (induction for 60 min and 48 h of reperfusion, re- CD93 deficiency leads to induction of CCL21 spectively), transcripts extracted from the ischemic hemisphere of (Exodus2)-mRNA expression in untreated knockout mice and in CD93-deficient mice were compared with those derived from the postischemic mice ischemic hemisphere of WT control mice. As shown in Table I, Interestingly, the mRNA of a single gene, Exodus-2 (also named http://www.jimmunol.org/ gene expression analysis identified several genes that were dif- CCL21), which is represented in three different probe sets, was ferentially regulated depending on the presence or absence of the found to be upregulated almost 10-fold in untreated CD93-deficient by guest on October 4, 2021

FIGURE 4. Comparison of infarct volumes of CD93-deficient and WT mice (C57BL/6) at 72 h after 30 min of transient MCAO. Direct infarct volumes (A, D, G), indirect infarct volumes (B, E, H), and volume of the edema (C, F, I) given as box-and-whisker-plots are visualized. Statistical analysis was performed with the Mann-Whitney U test. pp # 0.05. The indirect infarct volume was calculated as the volume of the contralateral hemisphere minus the noninfarcted volume of the ipsilateral hemisphere. In contrast to 48 h of reperfusion time (Fig. 3), after 72 h of reperfusion there is a significant difference of infarct volumes of WTand CD93- deficient mice in mixed gender and male mice: infarct volumes are larger in both genders in the knockout mice, when compared with age- and gender-matchedWT control mice. CD93 knockout total, CD93 knockout mice of both genders (n = 16); WT total, WT mice of both genders (n = 21); CD93 knockout female, female CD93-deficient mice (n = 5); WT female, female WT mice (n = 7); CD93 knockout male, male CD93-deficient mice (n = 11); WT male, male WT mice (n = 14). In all box plots, the top of the box represents the 75th percentile, the bottom of the box represents the 25th percentile, and the line in the middle represents the 50th percentile. The whiskers (the lines that extend from the top and bottom of the box) represent the highest and lowest values that are not outliers or extreme values. The Journal of Immunology 6413 Downloaded from

FIGURE 5. CD11b-positive inflammatory cells in postischemic mouse brain after 48 h, respectively 72 h reperfusion time in postischemic hemispheres of CD93-deficient or WT mice. Absolute counts of CD11b (which stains macrophages/monocytes and microglia)-positive cells in whole ischemic brain hemispheres of WT and CD93-deficient (CD93) mice, determined in different brain sections. A, CD11b-positive cell counts at 48 h of reperfusion after induction of cerebral ischemia for 60 min. B, Cell counts at 72 h of reperfusion after induction of cerebral ischemia for 30 min. Distance to bregma: a,

5.3 mm; b, 3.9 mm; and c, 1.8mm. pp , 0.05, sd, unpaired t test. nWT/48h =4;nCD93/48h =3;nWT/48h =4;nCD93/72h = 4. To evaluate inflammatory cell counts http://www.jimmunol.org/ in correlation to infarct volume sizes, absolute CD11b-positive cell counts (summarized value of the values derived from the three different brain sections of each animal) were plotted against the indirect infarct volumes of each WT, respectively CD93-deficient (CD93) animal. C, Cell counts at 48 h of re- perfusion after induction of cerebral ischemia for 60 min. D, Cell counts at 72 h of reperfusion after induction of cerebral ischemia for 30 min. mice compared with WT control mice. This dramatic difference in Exodus2 (CCL21) protein is induced in cerebral brain tissue of CCL21 expression in brain tissue of naive CD93 knockout mice was untreated and in postischemic brain tissue of CD93-deficient also found after MCAO. To evaluate results obtained with micro- mice arrays, expression of CCL21 mRNA was confirmed by real-time To analyze whether induction of CCL21-transcripts in CD93- by guest on October 4, 2021 PCR in separate experiments. Quantitative PCR confirmed expres- deficient mice provokes a similar induction of CCL21-protein in the sion data derived from Illumina microarrays: CCL21 mRNA is 2/2 2/2 mutant (CD93 ) mice, brain and spleen tissue were stained with highly induced in the CD93 brain tissue of untreated mice and in CCL21-specific Ab. Specific staining was observed in the spleens of postischemic brain tissue at all time points after induction of cere- WT mice (Fig. 7A), but an abundant and much more intense CCL21- bral ischemia, compared with WT brain tissue (data not shown). positive staining in the spleens of CD93-knockout mice (Fig. 7D). In However, in comparison with microarray expression data, PCR normal brain tissue, weak CCL21 staining was observed in individual results demonstrate an even higher induction ratio of CCL21 mRNA 2/2 cells of CD93-deficient mice, but a CCL21-specific signal was almost expression in the ischemic hemisphere of CD93 mice when absent in the brain tissue of WT mice (data not shown). In contrast to compared with WT animals; an almost 100-fold increase of the the weak cerebral staining pattern observed in untreated mice, CCL21 expression in CD93 knockout mice was detected by PCR CCL21-positive cell number increased after induction of cerebral when compared with WT mice at 48h after 1h MCAO. Moreover, in ischemia for 60 min and after 48 h of reperfusion in WT mice as well the normal brain of male untreated CD93-deficient mice, CCL21 as in CD93-deficient animals (Fig. 7). However, CCL-21-positive was found to be induced 127-fold compared with the CCL21- cells, which were mainly located at the infarct border zone in both expression in normal WT brains. As observed previously by other animal strains, revealed a highly intense CCL21-specific staining in groups, mRNA quantification by PCR revealed higher induction CD93 knockout mice, but only a low- to medium-intensity CCL21- ratios than those measured by hybridization technologies (e.g., 127- signal in WT mice. Moreover, CCL21-positive cells were more fold versus 8-fold induction in untreated animals). In contrast to abundant in CD93 knockout mice than in WT mice. the dramatic induction of CCL21 expression in knockout animals, we observed only a maximum 2.2-fold increase of the CCL21- mRNA expression in the ischemic hemisphere at 72 h after 30 min Discussion MCAO as measured by PCR and compared with the nonischemic By using a recent Affymetrix microarray platform, we identified hemisphere (data not shown). Furthermore, to detect a potential CD93 as a significantly induced gene whose role in cerebral is- compensatory regulation of the CCL21 receptors (CXCR3 and chemia was unknown until now. Both CD93 mRNA and CD93 CCR7) and any other chemokine signaling system that could con- protein were significantly induced after cerebral ischemia. CD93 tribute to neuroinflammation (CCL2/CCR2), expression of CCL21- was further shown to be neuroprotective in specific conditions: associated genes (e.g., CD30 and its ligand) (39), and CXCR3, CD93-deficient mice displayed significantly larger infarct volumes CCR7, CCL2 and its receptor CCR2 were studied. This revealed no in an experimental stroke model with a more selected neuronal significant differential expression between CD932/2 and WT mice death in combination with a profound neuroinflammatory response for either CCL2, its receptor CCR2, or the CCL21-receptors CCR7 (37). Interestingly, there was only a minor, nonsignificant tendency and CXCR3 (Table II). toward increased infarct volumes in CD93-deficient mice when 6414 CD93/AA.1 IN EXPERIMENTAL STROKE

Table I. mRNA expression of genes most significantly upregulated in brain tissue of CD93-deficient mice when compared with WT mice

Gene Illumina Number Ratio U Ratio 48 h Ratio 72 h CCL21B scl0065956.1_31-S 9.659 14.286 12.251 (LOC100041504) scl0018829.1_65-S 8.040 9.744 10.609 scl0020298.1_98-S 7.875 10.414 10.536 LOC100041516 ri|D130052N13|PX00184G04|AK051496|2048-S 5.449 7.153 5.955 LOC674214 GI_38078272-S 4.366 4.896 4.588 RPL23 scl39638.4.1_92-S 4.264 1.260 0.797 PLA2G4E scl18791.2_626-S 3.343 3.905 3.677 STK4 scl19945.14_178-S 2.955 3.332 3.875 MT-ND5 scl0017721.1_131-S 2.274 1.222 1.744 PPP1R16B scl19990.13_167-S 2.160 1.791 2.484 ZC3H13 scl067302.8_2-S 2.008 1.008 1.383 CNTFR scl0002829.1_12-S 1.993 2.370 1.730 CASC4 scl0003123.1_3-S 1.921 1.689 1.796 2310001H12RIK GI_21704131-S 1.905 1.939 1.718 INMT scl28971.3_6-S 1.821 1.464 1.329 MTF2 scl0017765.2_254-S 1.776 1.040 0.900 NR1D2 scl00353187.1_48-S 1.745 1.153 1.327 ATF4 scl47732.2_4-S 1.729 1.153 1.159 UBR1 scl0022222.2_5-S 1.727 3.130 2.194

PRSS18 scl019144.7_7-S 1.675 1.348 1.364 Downloaded from LOC100048622 scl056398.1_30-S 1.669 0.988 0.898 BOK scl17622.1.190_1-S 1.649 1.097 1.393 EPB4.1L1 scl0003233.1_11-S 1.649 1.187 1.379 D19ERTD721E scl53400.8.1_51-S 1.626 1.278 1.165 ZFP106 scl0003002.1_1949-S 1.626 0.950 0.976 TPH2 scl37491.11.1_193-S 1.624 0.751 0.910 LRRC59 scl39705.9.188_7-S 1.616 1.054 1.308

B930097J01RIK ri|B930097J01|PX00166N21|AK047604|3109-S 1.613 1.726 1.496 http://www.jimmunol.org/ LOC100047369 scl000093.1_57-S 1.605 1.048 1.039 Shown are the differentially regulated genes, ranked by their degree of induction in brain tissue derived from CD93-knockout mice when compared with WT brain tissue. Second column: gene identifiers according to the Illumina denotation (Illumina, San Diego, CA). Third column: relative expression, given by the expression level of each gene in CD93-knockout normal male brain tissue as determined by IlluminaBeadChip technology, divided by the expression level in normal (untreated) WT brain tissue. Fourth column (ratio 48 h): relative expression in CD93-deficient brain tissue at 48 h after 1 h MCAO when compared with the expression level in WT brain tissue at 48 h after 1 h MCAO. Last column (ratio 72 h): relative expression in postischemic hemispheres of CD93 knockout mice at 72 h after 30 min MCAO when compared with WT mice. mRNA was pooled from three male mice per group. MT, metallothionein.

compared with WT mice in an alternative, more severe stroke Cerebral ischemia leads to a significant inflammatory immune by guest on October 4, 2021 model, which causes extensive pan-necrosis. response. The resulting neuroinflammation is thought to exacerbate To unravel potential CD93-mediated neuroprotective mecha- postischemic brain injury during the initial phase of reperfusion nisms, we performed globalanalysis of expression patterns ofCD93- after cerebral vessel occlusion, although it might sustain restoration deficient and WT mice at different time points before and after of tissue homoeostasis at later time points. Ongoing repair mech- cerebral ischemia. Despite the finding that several genes were dif- anisms require the of large amounts of dead cells. The ferentially expressed, only one single gene—CCL21 (Exodus-2)— protein CD93 was shown recently to contribute to the in vivo was found to be highly induced in CD93 knockout mice under all clearance of dying cells (18). conditions when compared with control animals. Thus, we speculate However, no information on CD93 function in cerebral ischemia a central role of this chemokine in CD93-mediated neuroprotection. has been available until. Recently, van der Net et al. (40) confirmed

FIGURE 6. CD93 deficiency leads to altered gene expression patterns. Venn diagrams demonstrating the number of genes induced at least 2-fold when mea- sured in the ischemic brain hemisphere of WT or CD932/2 mice after 1 h MCAO and 48 h reperfusion (A, B) or 30 min MCAO and 72 h reperfusion (C, D), respectively, when compared with the nonischemic contralateral hemisphere (A, C) or the healthy hemi- sphere of untreated mice (B, D) using Illumina Sentrix arrays. Ratios were calculated using expression level in the ischemic (ipsilateral) hemisphere divided by the expression level in the nonischemic (contralateral) hemisphere (A, C), respectively, in the same hemi- sphere of untreated mice (B, D). Only male mice were used; n = 3 for each condition. The Journal of Immunology 6415

Table II. Expression of selected CCL21-related genes in CD93-deficient mice compared with WT mice, respectively, in the contralateral hemisphere

Ratio U Ratio 48 h Ratio 72 h CD93 Ipsi versus Contra WT Ipsi versus Contra

CD93/U CD93/48h CD93/72h CD93 Ipsi versus CD93 Ipsi versus WT Ipsi versus WT Ipsi versus Gene versus WT/U versus WT/48h versus WT/72h Contra/48h Contra/72 h Contra/48h Contra/72h CCL2 0.91 1.23 1.23 1.85 1.69 1.82 1.31 CXCR3 1.17 0.93 0.85 1.06 1.16 1.13 1.15 CCR7 1.05 1.08 1.01 1.00 1.13 0.76 1.14 CCR2 0.89 1.02 0.82 0.86 0.78 0.96 0.99 CXCL13 0.91 1.13 1.00 1.25 1.06 0.96 1.05 CD30L 0.92 1.07 1.04 1.16 1.06 1.31 0.95 CD30 0.95 1.13 1.03 1.16 1.04 0.95 1.27 Pdpn 1.02 1.02 1.39 2.91 3.00 2.94 2.36 LTB 1.01 1.10 1.05 0.85 0.99 0.92 0.85 LTBR 0.97 0.97 1.27 1.43 1.43 1.43 1.10 Regulation of the CCL21-related genes as determined by Illumina Sentrix arrays. Values are given as relative expression levels, determined as the ratio of the absolute signal intensity in the ipsilateral (column 2; non-treated; columns 3–6; ischemic) CD93-knockout hemisphere divided by the value obtained in the WT hemisphere (columns 2–4) in the two different infarct models (columns 3, 5, 7: 1 h MCAO and 48 h reperfusion; columns 4, 6, 8: 30 min MCAO and 72 h reperfusion). Columns 5–8 demonstrate the values of the ipsilateral (ischemic) hemisphere divided by the value derived from the contralateral (nonischemic) hemisphere (columns 5–8) of the same mice (CD93-knockout mice: columns 5, 6; WT mice: columns 7, 8) as reference. mRNA was pooled from three mice per group. Contra, contralateral; Ipsi, ipsilateral; LTB, lymphotoxin b; LTBR, lymphotoxin-b receptor; Pdpn, podoplanin; U, untreated. Downloaded from a genetic link between CD93 gene polymorphisms and myocardial a prevailing endothelial expression of CD93 and its hypothetical infarction in a human risk population. role in the immune response and remodeling of the vascular tree We found the CD93 protein to be highly induced after focal (18, 29, 28, 41, 42). ischemia in endothelial cells and, to a lesser extend, on selected The CD93 molecule is a heavily O-glycosylated type I trans- invading macrophages/microglia, which is in accordance with membrane protein consisting of a carbohydrate-recognition domain

(called CLECT-thrombomodulin-like domain for C-type -like http://www.jimmunol.org/ domain [CTLD] of the type found in thrombomodulin and endosialin) (41), five epidermal growth factor-like domains, a mucin domain, a single transmembrane domain, a PDZ-binding domain (44), and a short intracellular domain(18,35). The molecular structure of CD93 is reminiscent of thrombomodulin, a receptor for thrombin that is recognized as a major natural anticoagulant mainly expressed on endothelial cells and (31, 36). CD93 and thrombomodulin seem to be derived from a common ancestor (31). Thrombomodulin serves as a physiologic anticoagulant abundantly and has been shown by guest on October 4, 2021 to regulate endothelial function in inflammation. Furthermore, thrombomodulin polymorphisms have been shown to be associated with an increased risk of early onset ischemic stroke (45). Whereas homozygous thrombomodulin knockout mice are not viable, mice lacking an N-terminal part of thrombomodulin are viable but show a range of abnormalities resulting from a latent inflammatory state (46). Despite its membrane-bound form, CD93 might be shed from the cell surface upon cellular activation (26, 47, 48). A similar ecto- domain release has been reported for L-selectin and CD44, as well as for TNF-a and other cytokines (35). McGreal et al. (22) demonstrated specific binding of a soluble recombinant CD93-Fc chimeric protein to vascular endothelial cells in sections of in- flamed human tonsil, indicating the presence of a CD93 ligand at this site (22). Considering these data, we cannot exclude that part of the endothelial CD93-positive signal might derive from soluble CD93 protein rather than locally expressed CD93 membrane pro- tein. It must still be clarified whether CD93 is cleaved in cerebral ischemia, soluble CD93 might have clinical application as an anti- adhesion molecule to dampen local inflammation (36), or phar- FIGURE 7. CD93 deficiency leads to upregulation of CCL21/Exodus-2 macologic modulation of CD93 expression (49) might improve protein. CCL21 protein expression as revealed by CCL21-specific Ab postischemic outcome must still be clarified. 2/2 staining (R&D Systems) in spleens (A–F) of WT mice (A–C) and CD93 The exact molecular function of CD93 has not yet been identified. mice (D–F), and expression in the peri-infarct region of the ischemic brain Our finding of a robust induction of chemokine CCL21-expression hemisphere (G–R) of WT mice (G–I, M–O) and CD93-deficient mice (J–L, in CD93-deficient mice could explain how CD93 can protect the P–R). Samples were taken at 48 h of reperfusion after 60 min MCAO. First column, representative images of the tissue stained with anti-CCL21 Ab; brain after cerebral ischemia; it could also explain other CD93- second column, visualization of nuclei using Hoechst 33258 DNA stain; third related observations. It was recently demonstrated that CD93 is column, overlay of columns one and two. Original magnification 310 (A–F), important for the maintenance of plasma cells in the bone marrow 320 (G–L), and 3100 (M–R). Scale bar, 50 mm. (50). It was further shown that CD93-deficient mice were able to 6416 CD93/AA.1 IN EXPERIMENTAL STROKE respond normally with serum IgG to immunization, but showed expression and link CD93 regulation in ischemic brain tissue to a statistically significant decrease in IgG levels at later time points altered chemokine CCL21 expression and, therefore, to the in- (51). Accordingly, one can speculate that these observations may flammatory response. also be influenced by the differentially regulated expression of CCL21/Exodus-2 in CD932/2 mice. Acknowledgments Chemotactic signals, such as CCL21, were thought to control We thank Catherine Aubel for critical reading of the manuscript, Francisco leukocyte navigation by regulating migration from the blood into Fernandes Klett for help with confocal laser-scanning microscopy, and Mar- tissues and subsequent localization of cells within the tissue mi- kus Albrecht for assistance with handling of chip data. croenvironment (51). CCL21 was postulated to be a part of the neuron-microglia signaling system in cerebral ischemia (5, 7). In Disclosures contrast to most chemokines found in brain tissue, which are ex- The authors have no financial conflicts of interest. pressed by glial cells and by infiltrating leukocytes, CCL21 was found to be expressed exclusively on neurons and not on glial cells References (7). CCL21 was shown to be involved in the signaling between 1. Hossmann, K. A. 2006. Pathophysiology and therapy of experimental stroke. endangered neurons and microglial cells (5, 52). Damaged neurons Cell. Mol. Neurobiol. 26: 1055–1081. rapidly induce and release CCL21 in vitro and in vivo, and they 2. Dirnagl, U., C. Iadecola, and M. A. Moskowitz. 1999. Pathobiology of ischaemic stroke: an integrated view. Trends Neurosci. 22: 391–397. activate microglia via the chemokine receptor CXCR3 (5, 7, 52), 3. Trendelenburg, G., and U. Dirnagl. 2005. Neuroprotective role of astrocytes in whereas basically all T cells in the cerebrospinal fluid of healthy cerebral ischemia: focus on ischemic preconditioning. Glia 50: 307–320. individuals express CCR7, the second receptor for the chemokine 4. Gasque, P. 2004. Complement: a unique innate immune sensor for danger sig- nals. Mol. Immunol. 41: 1089–1098. Downloaded from CCL21 (53). 5. de Haas, A. H., H. R. J. van Weering, E. K. de Jong, H. W. G. M. Boddeke, and CCL21 and its receptor in the ischemic CNS, CXCR3, were pre- K. P. H. Biber. 2007. Neuronal chemokines: versatile messengers in central viously found to be induced in ischemic mouse brain (54–56). nervous system cell interaction. Mol. Neurobiol. 36: 137–151. 6. Trendelenburg, G. 2008. Acute neurodegeneration and the inflammasome: cen- CXCR3-deficiency was associated with reduced microglial activity tral processor for danger signals and the inflammatory response? J. Cereb. Blood and reduced loss of secondary neurons in hippocampal formation in Flow Metab. 28: 867–881. an entorhinal cortex lesion model (57). Furthermore, ectopic CCL21 7. Biber, K., A. Sauter, N. Brouwer, S. C. Copray, and H. W. Boddeke. 2001. Ischemia-induced neuronal expression of the microglia attracting chemokine http://www.jimmunol.org/ expression in oligodendrocytes induced a massive brain inflammation Secondary Lymphoid-tissue Chemokine (SLC). Glia 34: 121–133. that killed the animals within 3 d after onset of the expression (55). 8. Lo, J. C., R. K. Chin, Y. Lee, H. S. Kang, Y. Wang, J. V. Weinstock, T. Banks, C. F. Ware, G. Franzoso, and Y. X. Fu. 2003. Differential regulation of CCL21 in Because CD93-deficiency induces a dramatic increase of CCL21 lymphoid/nonlymphoid tissues for effectively attracting T cells to peripheral expression in the brain (but also in other organs, such as the spleen), tissues. J. Clin. Invest. 112: 1495–1505. we hypothesize that the upregulation of this chemokine interferes 9. Fo¨rster, R., A. C. Davalos-Misslitz, and A. Rot. 2008. CCR7 and its ligands: balancing immunity and tolerance. Nat. Rev. Immunol. 8: 362–371. with leukocyte homing into the brain and microglia activation, 10. Lu, A., Y. Tang, R. Ran, J. F. Clark, B. J. Aronow, and F. R. Sharp. 2003. Ge- thereby promoting postischemic inflammation. This hypothesis fits nomics of the periinfarction cortex after focal cerebral ischemia. J. Cereb. Blood well with our observation of lower leukocyte counts in postischemic Flow Metab. 23: 786–810. 2 2 11. MacManus, J. P., T. Graber, C. Luebbert, E. Preston, I. Rasquinha, B. Smith, and brain tissue of WT animals than in CD93 / animals (Fig. 5). The J. Webster. 2004. Translation-state analysis of gene expression in mouse brain by guest on October 4, 2021 observation that leukocyte infiltration between CD93-deficient and after focal ischemia. J. Cereb. Blood Flow Metab. 24: 657–667. WT mice significantly differs in a model that does not have sig- 12. Meller, R., S. L. Stevens, M. Minami, J. A. Cameron, S. King, H. Rosenzweig, K. Doyle, N. S. Lessov, R. P. Simon, and M. P. Stenzel-Poore. 2005. Neuro- nificant differences in lesion size argues against a simple secondary protection by osteopontin in stroke. J. Cereb. Blood Flow Metab. 25: 217–225. effect because of different infarct volumes. Because other che- 13. Stenzel-Poore, M. P., S. L. Stevens, Z. Xiong, N. S. Lessov, C. A. Harrington, mokines are not induced, or are only marginally induced (e.g., M. Mori, R. Meller, H. L. Rosenzweig, E. Tobar, T. E. Shaw, et al. 2003. Effect of ischaemic preconditioning on genomic response to cerebral ischaemia: sim- CCL2), in CD93-deficient mice, we postulate a major CCL21- ilarity to neuroprotective strategies in hibernation and hypoxia-tolerant states. mediated effect on leukocyte extravasation and microglia activa- Lancet 362: 1028–1037. tion in CD932/2 mice (Table II). 14. Trendelenburg, G., K. Prass, J. Priller, K. Kapinya, A. Polley, C. Muselmann, K. Ruscher, U. Kannbley, A. O. Schmitt, S. Castell, et al. 2002. Serial analysis of This effect resembles the way in which other receptors act on the gene expression identifies metallothionein-II as major neuroprotective gene in initiation of neuroinflammation: following stroke receptors, such as mouse focal cerebral ischemia. J. Neurosci. 22: 5879–5888. the TLRs, upregulate the expression of adhesion molecules that 15. Ziegler, G., D. Harhausen, C. Schepers, O. Hoffmann, C. Ro¨hr,V.Prinz,J.Ko¨nig, H. Lehrach, W. Nietfeld, and G. Trendelenburg. 2007. TLR2 has a detrimental role facilitate the infiltration of lymphocytes into the ischemic brain in mouse transient focal cerebral ischemia. Biochem. Biophys. Res. Commun. 359: region, thereby contributing to neuronal damage (58). However, the 574–579. exact contribution of other signaling pathways (e.g., inflammation 16. Kanehisa, M., S. Goto, M. Hattori, K. F. Aoki-Kinoshita, M. Itoh, S. Kawashima, T. Katayama, M. Araki, and M. Hirakawa. 2006. From genomics to chemical mediated by other chemokines, IL-1b–, or TLR-signaling) (6, 51) genomics: new developments in KEGG. Nucleic Acids Res. 34(Database issue): in CD93-mediated neuroprotection remains to be determined. D354–D357. Furthermore, we cannot exclude additional effects by altered 17. Huang, R., A. Wallqvist, and D. G. Covell. 2006. Comprehensive analysis of pathway or functionally related gene expression in the National Cancer In- signaling networks in the transgenic mice. Our experimental design stitute’s anticancer screen. Genomics 87: 315–328. does not allow a causal relationship between CCL21 induction and 18. Norsworthy,P.J.,L.Fossati-Jimack,J.Cortes-Hernandez,P.R.Taylor,A.E.Bygrave, R. D. Thompson, S. Nourshargh, M. J. Walport, and M. Botto. 2004. Murine CD93 leukocyte invasion in CD93-deficient mice. Future studies need to (C1qRp) contributes to the removal of apoptotic cells in vivo but is not required for C1q- determine the exact signaling cascade by which CD93 inhibits the mediated enhancement of phagocytosis. J. Immunol. 172: 3406–3414. CCL21-expression and to what extend this pathway and those 19. Løvik, G., J. T. Vaage, E. Dissen, C. Szpirer, J. C. Ryan, and B. Rolstad. 2000. Characterization and molecular cloning of rat C1qRp, a receptor on NK cells. linked to the other differentially expressed genes contribute to the Eur. J. Immunol. 30: 3355–3362. neuroprotection observed in male CD93-deficient mice. The future 20. Nepomuceno, R. R., A. H. Henschen-Edman, W. H. Burgess, and A. J. Tenner. use of conditional transgenic mice or double-knockout mice should 1997. cDNA cloning and primary structure analysis of C1qR(P), the human C1q/ MBL/SPA receptor that mediates enhanced phagocytosis in vitro. Immunity 6: allow us to dissect these pathways in more detail. 119–129. We demonstrate a significant upregulation of CD93 mRNA and 21. Kishore, U., and K. B. M. Reid. 2000. C1q: structure, function, and receptors. protein, as well as a significant neuroprotective effect of this gene in Immunopharmacology 49: 159–170. 22. McGreal, E. P., N. Ikewaki, H. Akatsu, B. P. Morgan, and P. Gasque. 2002. specific conditions after cerebral ischemia at 3 d of reperfusion. Human C1qRp is identical with CD93 and the mNI-11 antigen but does not bind Besides its induction, we describe a neuroprotective effect of CD93 C1q. J. Immunol. 168: 5222–5232. The Journal of Immunology 6417

23. Danet, G. H., J. L. Luongo, G. Butler, M. M. Lu, A. J. Tenner, M. C. Simon, and 42. Sumanas, S., T. Jorniak, and S. Lin. 2005. Identification of novel vascular en- D. A. Bonnet. 2002. C1qRp defines a new human stem cell population with dothelial-specific genes by the microarray analysis of the zebrafish cloche mu- hematopoietic and hepatic potential. Proc. Natl. Acad. Sci. USA 99: 10441– tants. Blood 106: 534–541. 10445. 43. Marchler-Bauer, A., J. B. Anderson, M. K. Derbyshire, C. DeWeese-Scott, 24. Goolsby, J., M. C. Marty, D. Heletz, J. Chiappelli, G. Tashko, D. Yarnell, N. R. Gonzales, M. Gwadz, L. Hao, S. He, D. I. Hurwitz, J. D. Jackson, et al. P. S. Fishman, S. Dhib-Jalbut, C. T. Bever, Jr., B. Pessac, and D. Trisler. 2003. 2007. CDD: a conserved domain database for interactive domain family analysis. Hematopoietic progenitors express neural genes. Proc. Natl. Acad. Sci. USA 100: Nucleic Acids Res. 35(Database issue): D237–D240. 14926–14931. 44. Bohlson, S. S., M. Zhang, C. E. Ortiz, and A. J. Tenner. 2005. CD93 interacts 25. Jack, R. M., B. A. Lowenstein, and A. Nicholson-Weller. 1994. Regulation of with the PDZ domain-containing adaptor protein GIPC: implications in the C1q receptor expression on human polymorphonuclear leukocytes. J. Immunol. modulation of phagocytosis. J. Leukoc. Biol. 77: 80–89. 153: 262–269. 45. Cole, J. W., S. C. Roberts, M. Gallagher, W. H. Giles, B. D. Mitchell, 26. Bohlson, S. S., R. Silva, M. I. Fonseca, and A. J. Tenner. 2005. CD93 is rapidly K. K. Steinberg, M. A. Wozniak, R. F. Macko, L. J. Reinhart, and S. J. Kittner; shed from the surface of human myeloid cells and the soluble form is detected in Stroke Prevention in Young Women Study. 2004. Thrombomodulin Ala455Val human plasma. J. Immunol. 175: 1239–1247. Polymorphism and the risk of cerebral infarction in a biracial population: the 27. Fonseca, M. I., P. M. Carpenter, M. Park, G. Palmarini, E. L. Nelson, and Stroke Prevention in Young Women Study. BMC Neurol. 4: 21. A. J. Tenner. 2001. C1qR(P), a myeloid cell receptor in blood, is predominantly 46. Weiler, H. 2004. Mouse models of thrombosis: thrombomodulin. Thromb. expressed on endothelial cells in human tissue. J. Leukoc. Biol. 70: 793–800. Haemost. 92: 467–477. 28. Nepomuceno, R. R., and A. J. Tenner. 1998. C1qRP, the C1q receptor that en- 47. Park, M., and A. J. Tenner. 2003. Cell surface expression of C1qRP/CD93 is hances phagocytosis, is detected specifically in human cells of myeloid lineage, stabilized by O-glycosylation. J. Cell. Physiol. 196: 512–522. endothelial cells, and platelets. J. Immunol. 160: 1929–1935. 48. Ikewaki, N., J. K. Kulski, and H. Inoko. 2006. Regulation of CD93 cell surface 29. Petrenko, O., A. Beavis, M. Klaine, R. Kittappa, I. Godin, and I. R. Lemischka. expression by protein kinase C isoenzymes. Microbiol. Immunol. 50: 93–103. 1999. The molecular characterization of the fetal stem cell marker AA4. Im- 49. Woltman, A. M., N. Schlagwein, S. W. van der Kooij, and C. van Kooten. 2004. munity 10: 691–700. The novel cyclophilin-binding drug sanglifehrin A specifically affects antigen 30. Jordan, C. T., C. M. Astle, J. Zawadzki, K. Mackarehtschian, I. R. Lemischka, uptake receptor expression and endocytic capacity of human dendritic cells. and D. E. Harrison. 1995. Long-term repopulating abilities of enriched fetal liver J. Immunol. 172: 6482–6489. stem cells measured by competitive repopulation. Exp. Hematol. 23: 1011–1015. 50. Chevrier, S., C. Genton, A. Kallies, A. Karnowski, L. A. Otten, B. Malissen, 31. Dean, Y. D., E. P. McGreal, and P. Gasque. 2001. Endothelial cells, mega- M. Malissen, M. Botto, L. M. Corcoran, S. L. Nutt, and H. Acha-Orbea. 2009. Downloaded from karyoblasts, platelets and alveolar epithelial cells express abundant levels of the CD93 is required for maintenance of antibody secretion and persistence of mouse AA4 antigen, a C-type lectin-like receptor involved in homing activities plasma cells in the bone marrow niche. Proc. Natl. Acad. Sci. USA 106: 3895– and innate immune host defense. Eur. J. Immunol. 31: 1370–1381. 3900. 32. Hara, H., P. L. Huang, N. Panahian, M. C. Fishman, and M. A. Moskowitz. 1996. 51. Babcock, A. A., W. A. Kuziel, S. Rivest, and T. Owens. 2003. Chemokine ex- Reduced brain edema and infarction volume in mice lacking the neuronal iso- pression by glial cells directs leukocytes to sites of axonal injury in the CNS. form of nitric oxide synthase after transient MCA occlusion. J. Cereb. Blood J. Neurosci. 23: 7922–7930. Flow Metab. 16: 605–611. 52. de Jong, E. K., I. M. Dijkstra, M. Hensens, N. Brouwer, M. van Amerongen,

33. Laemmli, U. K. 1970. Cleavage of structural proteins during the assembly of the R. S. Liem, H. W. Boddeke, and K. Biber. 2005. Vesicle-mediated transport and http://www.jimmunol.org/ head of bacteriophage T4. Nature 227: 680–685. release of CCL21 in endangered neurons: a possible explanation for microglia 34. Lin, T. N., Y. Y. He, G. Wu, M. Khan, and C. Y. Hsu. 1993. Effect of brain edema activation remote from a primary lesion. J. Neurosci. 25: 7548–7557. on infarct volume in a focal cerebral ischemia model in rats. Stroke 24: 117–121. 53. Kivisa¨kk, P., D. J. Mahad, M. K. Callahan, C. Trebst, B. Tucky, T. Wei, L. Wu, 35. Zhang, M., S. S. Bohlson, M. Dy, and A. J. Tenner. 2005. Modulated interaction E. S. Baekkevold, H. Lassmann, S. M. Staugaitis, et al. 2003. Human cerebro- of the ERM protein, moesin, with CD93. Immunology 115: 63–73. spinal fluid central memory CD4+ T cells: evidence for trafficking through 36. Dean, Y. D., E. P. McGreal, H. Akatsu, and P. Gasque. 2000. Molecular and choroid plexus and meninges via P-selectin. Proc. Natl. Acad. Sci. USA 100: cellular properties of the rat AA4 antigen, a C-type lectin-like receptor with 8389–8394. structural homology to thrombomodulin. J. Biol. Chem. 275: 34382–34392. 54. Rappert, A., K. Biber, C. Nolte, M. Lipp, A. Schubel, B. Lu, N. P. Gerard, 37. Katchanov, J., C. Waeber, K. Gertz, A. Gietz, B. Winter, W. Bru¨ck, U. Dirnagl, C. Gerard, H. W. Boddeke, and H. Kettenmann. 2002. Secondary lymphoid R. W. Veh, and M. Endres. 2003. Selective neuronal vulnerability following mild tissue chemokine (CCL21) activates CXCR3 to trigger a Cl- current and che- focal brain ischemia in the mouse. Brain Pathol. 13: 452–464. motaxis in murine microglia. J. Immunol. 168: 3221–3226.

38. Endres, M., A. Meisel, D. Biniszkiewicz, S. Namura, K. Prass, K. Ruscher, 55. Chen, S. C., M. W. Leach, Y. Chen, X. Y. Cai, L. Sullivan, M. Wiekowski, by guest on October 4, 2021 A. Lipski, R. Jaenisch, M. A. Moskowitz, and U. Dirnagl. 2000. DNA methyl- B. J. Dovey-Hartman, A. Zlotnik, and S. A. Lira. 2002. Central nervous system transferase contributes to delayed ischemic brain injury. J. Neurosci. 20: 3175– inflammation and neurological disease in transgenic mice expressing the CC 3181. chemokine CCL21 in oligodendrocytes. J. Immunol. 168: 1009–1017. 39. Bekiaris, V., F. Gaspal, M. Y. Kim, D. R. Withers, F. M. McConnell, G. Anderson, 56. Utans-Schneitz, U., H. Lorez, W. E. Klinkert, J. da Silva, and W. Lesslauer. and P. J. Lane. 2009. CD30 is required for CCL21 expression and CD4 T cell 1998. A novel rat CC chemokine, identified by targeted differential display, is recruitment in the absence of lymphotoxin signals. J. Immunol. 182: 4771–4775. upregulated in brain inflammation. J. Neuroimmunol. 92: 179–190. 40. vanderNet,J.B.,D.M.Oosterveer,J.Versmissen,J.C.Defesche,M.Yazdanpanah, 57. Rappert, A., I. Bechmann, T. Pivneva, J. Mahlo, K. Biber, C. Nolte, A. D. Kovac, B. E. Aouizerat, E. W. Steyerberg, M. J. Malloy, C. R. Pullinger, J. J. Kastelein, et al. C. Gerard, H. W. Boddeke, R. Nitsch, and H. Kettenmann. 2004. CXCR3- 2008. Replication study of 10 genetic polymorphisms associated with coronary heart dependent microglial recruitment is essential for dendrite loss after brain lesion. disease in a specific high-risk population with familial hypercholesterolemia. Eur. J. Neurosci. 24: 8500–8509. Heart J. 29: 2195–2201. 58. Tang, S. C., T. V. Arumugam, X. Xu, A. Cheng, M. R. Mughal, D. G. Jo, 41. Brady, J., J. Neal, N. Sadakar, and P. Gasque. 2004. Human endosialin (tumor J. D. Lathia, D. A. Siler, S. Chigurupati, X. Ouyang, et al. 2007. Pivotal role for endothelial marker 1) is abundantly expressed in highly malignant and invasive neuronal Toll-like receptors in ischemic brain injury and functional deficits. brain tumors. J. Neuropathol. Exp. Neurol. 63: 1274–1283. Proc. Natl. Acad. Sci. USA 104: 13798–13803.