Overexpression of ErbB2 in the exocrine pancreas induces an inflammatory response but not increased proliferation



Tyrosine kinase receptors such as members of the epidermal growth factor receptor family and their respective ligands are frequently overexpressed in pancreatic cancer as well as in chronic pancreatitis. In this study, the role of ErbB2 in the exocrine pancreas was examined by ectopic overexpression under the control of the proximal rat elastase promoter. Three independent transgenic mouse lines overexpressing ErbB2 were established by pronuclear injection. Pancreatic mRNA and protein levels were analyzed by real time PCR, immunohistochemistry and immunoblot analysis, RAS activity by using a specific immunoprecipitation assay and various kinase activities by phosphospecific antibodies. Overexpression of ErbB2 in the exocrine pancreas resulted in increased RAS activity and downstream activation of ERK1/2, but not in transgenic increased proliferation of acinar and ductal cells. At later timepoints, some mice showed focal areas of acinar cell damage with upregulated mRNA levels for Cyclin D1 and p16INK4a. Despite the increased mRNA level, cyclin D1 protein levels were downregulated. We observed areas of focal infiltrations with inflammatory cells interspersed in the exocrine pancreas. NF-κB activity was induced in transgenic acinar cells compared with controls contributing to high levels of chemokine and cytokine gene expression such as CCR-1 and CCL3. These data suggest that overexpression of ErbB2 in acinar cells leads to increased RAS activity without cell cycle progression and mediates inflammation via NF-κB. We conclude that the biological response of ErbB2/RAS signaling depends highly on cellular context. © 2007 Wiley-Liss, Inc.

Pancreatic cancer is one of the most lethal human cancers. Despite considerable progress in radiation therapy, chemotherapy and surgical treatment options, the 5-year survival is still <20%. This is mostly due to the fact of first time diagnosis at late stages and its overall resistance to available therapies. Even patients undergoing surgical therapy at early stages often suffer from disease recurrence or metastatic disease. Developing strategies to detect preinvasive pancreatic neoplasms and establishing more effective treatment regimes requires a better understanding of molecular events initiating or proceeding this fatal disease.1 Analyses in early pancreatic lesions, so called pancreatic intraepithelial neoplasias (PanINs), revealed loss of heterozygosity at a number of gene loci and alterations in a number of genes and proteins including KRAS, HER-2/neu (ErbB2), p16INK4a, Tp53, DPC4(SMAD4) and BRCA2. Overexpression of ErbB2 together with mutations in the KRAS gene have been detected in PanIN-1 lesions, while the inactivation of Tp53 and p16INK4a as well as of DPC4 (SMAD4) has been observed in progressive changes of the duct epithelium.2, 3, 4

The growth factor receptor ErbB2 belongs to a family of receptor tyrosine kinases, termed epidermal growth factor receptor family. This family comprises 4 closely related Type 1 receptor tyrosine kinases (RTK), EGF-receptor (EGFR, ErbB1), ErbB2 (neu, HER-2), ErbB3 (HER-3) and ErbB4 (HER-4).5, 6 These receptors are involved in mitogenic and antiapoptotic processes as shown mainly in breast cancer tissue.7, 8 Their ligands, e.g. EGF and TGF-α induce after binding to the receptor the formation of hetero- or homodimers, which then recruit adaptor proteins and thereby activate the RAS signaling pathway.9, 10 This pathway is important for cyclin D1 activation via assembly with cyclin-dependent kinases (CDKs). This leads to phosphorylation and hence inactivation of the growth-suppressive retionoblastoma protein (Rb) during G1 phase of the cell cycle thus contributing to unregulated cell proliferation.11

In addition, EGFR activation leads to activation of the transcription factor NF-1κB, which is activated in a variety of cancers. Several reports suggest that nuclear translocation of NF-1κB results in triggering antiapoptotic signaling pathways due to activation of PI3K/AKT.12, 13 Furthermore, NF-κB/Rel has been shown to be involved in inflammatory and immune cell responses, cell cycle regulation, and differentiation.14 Under normal conditions, NF-κB/Rel heterodimers are sequestered in the cytoplasm by binding inhibitory proteins, so called IκB's. Upon stimulation, a kinase complex, termed IκB kinase (IKK), phosphorylates IκB's, which then undergo degradation by a proteasome dependent mechanism. This enables the nuclear translocation of NF-κB/Rel heterodimers and the subsequent binding to specific DNA-motifs.

To elucidate the role of ErbB2 in the exocrine pancreas, we have generated transgenic mouse lines overexpressing human ErbB2 under the control of the elastase promoter. Although most human pancreatic neoplasms show a ductal phenotype and new targeting strategies specific for the pancreatic ductal epithelium have been established, there are several reports suggesting that injured or transformed acinar cells can acquire a ductal phenotype.15, 16 Therefore, it is conceivable that under stress conditions certain forms of pancreatic neoplasia might originate from acinar cells. Acinar cells overexpressing ErbB2 have been associated with pancreatic head enlargement in patients with chronic pancreatitis (CP),17 an accepted risk factor for pancreatic cancer in humans.18, 19 These observations have encouraged us to target human ErbB2 using the well-characterized acinar cell-specific Ela enhancer/promoter to analyze its role in pancreatic oncogenesis and inflammation.20, 21 Here we report that overexpression of ErbB2 in pancreatic acinar cells does not result in tumor formation, but provokes an inflammatory phenotype suggesting an important role of ErbB2 in chronic processes of the pancreas.

Material and methods

Generation of transgenic mice

The human ErbB2 complementary DNA (cDNA) was subcloned into the E0.2-human growth hormone (hGH) plasmid (generously provided by R. D. Palmiter) carrying the rat elastase promoter (−205 to +5) and a genomic fragment of the hGH. A 3.5 kb expression cassette was excised using AfIII and PvuI, and transgenic mice were generated by pronuclear injection of purified fragment into C57Bl/6J-DBA/2J F2 hybrid zygotes by standard procedures. Offspring were genotyped by polymerase chain reaction (PCR) of tail DNA using a primer pair specific for the hGH sequence (5′-GGC TTT TTG ACA ACG CTA TG and 5′-GCG CGG AGC ATA GGG TTG TC) and a second primer pair spanning part of the rat elastase promoter and the human ErbB2 cDNA (5′-AAG AGC CGT ATA AAG AGG GT and 5′-TAT CCT GTA ATT TTA TCT TCA GTT).

Animals were kept under specific pathogen-free conditions and analyzed together with littermate controls after 2 backcrosses to the C57Bl/6J background. The survival of 20 ErbB2-transgenic mice was followed for at least 720 days to evaluate possible tumor induction or chronic inflammatory processes. All experiments were performed according to the guidelines of the local Animal Use and Care Committees.

Preparation of total cell lysates and nuclear extracts

Whole cell lysates were prepared by incubating pelleted isolated acini or fresh pancreatic tissues from transgenic mice and wild-type littermate controls for 10 min at 4°C in lysis buffer (50 mM HEPES, 150 mM NaCl, 1mM EDTA, 0.5% NP-40, 10% glycerol, 1 mM dithiothreitol, 1 mM phenylmethylsulfonylflourid and 5 mM NaF). Insoluble material was removed by centrifugation and lysates were aliquoted and stored at −80°C. Nuclear extracts were prepared according to the method of Schreiber et al.22 Extracts were normalized for protein content using a colorimetric method (Bradford Biorad Assay, Bio-Rad).

Microscopic analysis and immunohistochemistry

To detect proliferating cells transgenic mice and wildtype littermates were labeled with 200 mg of BrdU/kg body weight by i.p. injection and sacrificed 2 h later. Pancreatic tissues were fixed in 4% paraformaldehyde and paraffin embedded. Sections of 5 μm were mounted on adhesive-coated slides and stained with H&E. Immunohistochemical analysis were performed as described.23 Sections were stained using a monoclonal anti-BrdU antibody (1:400, Boehringer Mannheim), anti-CCR-1 (1:300, Abcam), anti-ErbB2 (1:100 Santa Cruz Biotechnology, Dako Hercep Test), p16INK4a (BD Pharmingen, 1:100) and anti-CD 19 (1:200, BD Pharmingen). Fresh tissue samples of the pancreas were frozen in OCT compound (Ames, Elkhart, IN) and stored at −80°C until use. Frozen tissue sections (10 μm) were air-dried for 2 h, fixed in acetone for 10 min, air-dried again for 20 min and stained by the avidin-biotin complex method as described using purified anti-CD4 and anti-CD8 antibodies (BD Pharmingen).24 Signals were developed by the avidin-biotin peroxidase method using the 3,3′-diaminobenzidine substrate kit (Vector Laboratories). The specificity of the staining was confirmed for each antibody by control reactions in which the primary antibody was omitted.

Isolation of acini

Pancreas was dissected from transgenic mice and littermate controls, and dispersed pancreatic acini were isolated by a standard collagenase digestion method as described previously.25 Briefly, pancreata of mice were injected with collagenase (100 U/ml, Worthington Biochemical Corporation, NJ) and incubated at 37°C for 30 min with shaking. Acini were then dispersed by filtration through a nylon mesh with 150-μm pores (Hartenstein, Würzburg, Germany). Acini were purified by centrifugation through a solution containing 4% BSA and were resuspended in HEPES-buffered Ringer's solution supplemented with 0.2 M glucose, Eagle′s minimal essential amino acids, 2 mM glutamine, 0.1 mg/ml SBTI and 0.5% BSA. Acinar cell viability was assessed by trypan blue exclusion. More than 80% of the isolated acini were intact. Acini were incubated at 37°C in HEPES buffered Ringer solution and subsequently separated into aliquots.

RAS activity assay and western blotting

To determine RAS activity, fresh pancreatic tissue from transgenic animals and wildtype littermate controls was immediately transferred to Mg2+ lysis buffer (Upstate Biotechnology) with protease inhibitors (Boehringer Mannheim), homogenized and finally centrifuged at 4°C for 20 min to remove insoluble material. RAS activity was detected according to the manufacturer's protocol (Upstate Biotechnology). Briefly, pancreas of mice was lysed in ice-cold MLB buffer (125 mM HEPES, pH 7.5, 750 mM NaCl, 5% Igepal CA-630, 50 mM MgCl2, 5 mM EDTA, 10% glycerol). Lysates were cleared of insoluble cell debris by centrifugation (5 min, 14,000g, 4°C). Equal amounts of protein were incubated with Raf-1 RBD agarose beads (10–20 μg) for 45 min at 4°C with gentle agitation. After centrifugation supernatant was discarded, the beads were washed 3 times with MLB and resuspended in Laemmli reducing sample buffer. After boiling supernatants were subjected to western-blot analysis. Proteins were transferred to nitrocellulose. Blotted membrane was incubated with anit-RAS (clone RAS 10) antibody (Upstate Biotechnology).

For western blot analysis lysates of fresh pancreas were separated on SDS-gel, transferred and blotted with β-actin (Sigma, Deissenhofen, Germany), anti-phospho-ERK1/ERK2 E10, anti-phospho-Akt, anti-phospho SAPK/JNK, ERK1/ERK2, Akt, SAPK/JNK (all of them from Cell signaling, Beverly, MA), IκBα, IκBβ, p65, Cyclin D1, Cyclin E, p16, p19, p21, lamin A/C, ErbB2/neu or EGFR (purchased from Santa Cruz, San Diego, CA); CCR-1 was purchased from Abcam. Immunodetection of the signals was performed using HRP-conjugated corresponding secondary antibodies.

Microarray analysis of gene expression

Total pancreatic RNA (8 μg) was labeled and hybridized to Affymetrix MOE430A GeneChips according to the manufacturer's instructions. Two biological replicates per mouse line were analyzed. Hierarchical clustering was performed using the program Genesis (release 1.6.0 Beta 1).

Quantitative reverse-transcription PCR

Expression of mRNA was evaluated by using quantitative real-time (RT)-PCR (TaqMan, PE Applied Biosystems). Reverse transcription of total RNA was performed in duplicate and analyzed independently. The PCR reaction (denaturation at 95°C for 2 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min; SybrGreen PCR Mastermix, PE Applied Biosystems) was performed in triplicates and normalized to endogenous cyclophillin mRNA levels for each reaction. Quantification was done using the ΔΔ CT method.26 The primers were used were as follows:


Generation of transgenic mice overexpressing ErbB2 in the exocrine pancreas

To analyze the effect of ErbB2 overexpression in acinar cells, we generated several ErbB2 transgenic mouse lines. The human ErbB2 cDNA was placed under control of an elastase promoter followed by a genomic fragment of the hGH. Three transgenic mouse lines I, III, XVI were established, in which the transgene was expressed over several generations. To verify expression of the transgene ErbB2, we analyzed mRNA expression levels performing RT-PCR (see Table). Using hGH-specific primers, we detected a 140 bp mRNA product in the tested transgene mouse lines. In all transgenic lines, ErbB2 was selectively overexpressed in the pancreas as other tested organs were negative for the spliced transgene (Fig. 1a, founder line III).

Figure 1.

(a) Specific expression of the transgene in exocrine pancreas was tested by analyzing mRNA of different organs using quantitative PCR specific for the spliced transgene construct. Amplification of cyclophilin (101-base pair PCR product) served as internal control. DNA was visualized using ethidium bromide staining. (b) Western blot analyses showed upregulation of ErbB2 in whole pancreatic extracts of 3 different lines of ErbB2 transgenic mice (Ela-ErbB2) compared with littermate control (wt). Isolated acini from founder line III were tested for expression of ErbB2 in western blot. Representative samples of three different animals each were used. (c, d) Immunostaining of ErbB2 in littermate controls (c) and transgenic mice (d) using the ABC technique (50×, brown is positive staining).

Table  . 

Protein expression levels of ErbB2 were determined by immunoblot techniques with ErbB2 specific antibodies. In all 3-tested transgenic mouse lines high amounts of ErbB2 compared with wild type were detected in the pancreas or isolated acini (founder line I, III and XVI, Fig. 1b). This was confirmed by immunohistochemical analyses, which revealed a strong staining of ErbB2 in the pancreas of transgenic mice compared with wild type littermates (Fig. 1c, founder line III).

ErbB2 overexpression induces RAS activation, but not cell proliferation

To test functionality of the transgene in vivo, activation status of downstream effectors were tested in freshly prepared pancreatic tissues from transgenic animals and wild type littermates. Immunoprecipitation and immunoblot analysis demonstrated RAS activity in transgenic mice, predominantly in lines I and III (Fig. 2). Total RAS protein levels were comparable in whole pancreatic extracts. Concomitant phosphorylation of the downstream effector ERK1/2 was detected in founder lines I and III, with weak activation in line XVI. Although the PI3K/Akt signaling pathway is known to be downstream of RAS, differences in Akt activation were not observed (Fig. 2).

Figure 2.

Analyses of downstream signaling pathways in pancreas of ErbB2 overexpressing mice. While the founder lines III and I revealed strong RAS activation, the founder line XVI displayed weak RAS induction. Concomitant ERK1/2 (p44/p42) phosphorylation could be detected in founder line III and I, and to a lesser extent in line XVI. Overexpression of ErbB2 in pancreas did not mediate the PI-3-Kinase dependent Akt phosphorylation in any founder lines compared with wildtype control. Representative samples of 3 different animals each were used.

Posttranscriptional downregulation of cyclin D1 and induction of p16INK4a in transgenic mice

To test whether ErbB2 mediates proliferation via activated RAS and its downstream effector ERK1/2, we analyzed BrdU incorporation in all transgenic founder lines and wild type littermates, where we detected very little BrdU positive cells. When counting BrdU-positive cells per high power fields, the proliferation index was not significantly different in transgenic mice compared with littermate controls (data not shown). Transgenic mice did not display signs of malignant transformation even after 2 years of observation, although RAS and its downstream effector ERK1/2 were induced. Therefore, we performed immunoblot analysis for cell cycle proteins. Transgenic mice had significantly lower protein levels of cyclin D1, whereas the expression level of cyclin E remained unaltered (Fig. 3a). Protein expression levels of p16INK4a were upregulated and p21 and p27 levels were not affected in transgenic mice compared with wild type littermates. Nuclei of acini in transgenic pancreatic tissue displayed p16INK4a expression, whereas nuclei in comparable sections in wild type mice were negative (Fig. 3b).

Figure 3.

(a) Western blot analyses of cell cycle regulating proteins. Decreased cyclin D1 expression and upregulation of p16 levels were detected in transgenic mouse lines (XVI, III, I) compared with wildtype littermates. Cyclin E, p21 and p27 were not altered in their expression level. (b) Tissue sections of 3 months old nontransgenic and transgenic mice were stained with an antibody detecting p16INK4a (arrowheads indicating positive nuclei). Mononuclear cells in the ileum displayed p16INK4a expression and served as positive controls. (c) Real-Time-PCR-analyses of wildtype and transgenic mice using cyclin D1, cyclin E, p16 and p27 specific primer (n ≥ 4). The values were normalized to endogenous cyclophilin as described in Materials and Methods.

In Real-Time-PCR-analyses, we confirmed microarray data showing equal levels of Cyclin E and p27, but increased mRNA levels of cyclin D1 and p16 in ErbB2 transgenic mice compared with controls (Fig. 3c and Table). These results suggest posttranscriptional downregulation of cyclin D1 in transgenic mice, which in conjunction with p16INK4a induction may explain the lack of proliferation increase despite significant RAS activation in ErbB2 overexpressing mice.

ErbB2 overexpression in the exocrine pancreas induces an inflammatory response

Next we analyzed the phenotype of all transgenic lines compared with age-matched littermates. Mice were observed over a period of >2 years. Starting at Day 30, we detected lymphocytic infiltration dispersed in the pancreas in Erb2 transgenic mice in contrast to age-matched wild type littermates (Fig. 4a–4c asterisk). The extent of infiltrations was not markedly changed during the course of their lives. All of the detected infiltrations seemed to harbor mostly CD4+ but not CD8+ lymphocytes (Fig. 4d and data not shown). Age-matched wild type littermates were negative for CD4 and CD8 staining.

Figure 4.

(a) Tissue sections of wildtype littermates were stained with H&E. (b, c) Pancreas from age matched transgenic mice were analyzed. Mononuclear cells accumulated as conglomerates next to normal acinar cells or ducts (*). (d) Immunostaining of CD4 demasked infiltrates as predominantly CD4+ lymphocytes. (e, f) Focal destruction of acinar cells with appearance of tubular complexes in the pancreas (founder line I, age 375 days). (g, h) Founder line III (age 380 days) displays acinar cell damage next to normal unaffected acinar compartment.

At late ages (over 1 year) founder lines I and III displayed focal areas of chronic inflammation with acinar cell destruction and tubular complexes, while pancreatic weight was not different between wildtype littermates and mice from founder lines I and III (0.183 ± 0.002 g vs. 0.173 ± 0.004 g). These areas were very rare and did not increase in size or affect large areas of the pancreas (Fig. 4e–4h).

Microarray analysis of pancreatic tissue from one month old transgenic mice revealed increased expression of CCR-1 (2.46 over control, Fig. 5a) and CCL3 (Mip 1α) (1.24 over control, data not shown). The increase was confirmed with real time PCR analysis (Fig. 5b and Table) and immunoblot analysis (Fig. 5c). Therefore, these data suggest that the CCR1/CCL3 receptor/ligand interaction might at least in part explain this inflammatory phenotype.

Figure 5.

(a) Pancreata from nontransgenic and transgenic (founder line III) mice were removed. Total pancreatic RNA (8 μg) (n = 2) was labeled and hybridized to Affymetrix MOE430A GeneChips. Inflammation-associated genes being differentially expressed (1.5-fold over control) were compiled in a list. (b) Real-Time-PCR analyses revealed not only overexpression of CCR-1mRNA, but also upregulation of its ligand CCL3 (Mip 1α) (n ≥ 4). (c) In Western Blot analyses CCR-1 expression has been found to be high in 2 founder lines (I, III). (d) Nuclear localization of p65 in isolated acinar cells has been detected in at least 2 founder lines. Cytosolic extracts were further analyzed for IκBα and IκBβ expression levels. Lamin A/C and β-actin were used as loading controls for nuclear and cytosolic extracts, respectively. Representative samples of 3 different animals each were used.

The transcription factor NF-κB/Rel is known to be involved in ErbB2/RAS signaling pathways and concomitant inflammatory responses regulating cytokine and chemokine synthesis. To test whether NF-κB/Rel was induced in pancreatic acinar cells of transgenic mice we isolated them to rule out any effects from infiltrating cells. Founder lines I and III displayed enhanced nuclear accumulation of p65, the transactivating subunit of NF-κB, compared with wild type littermates. The inhibitor protein IκBα is less expressed in founder lines I and III, likely as a result of continuous degradation or nuclear accumulation due to NF-κB/Rel activation. Protein levels of the inhibitor IκBβ were not affected (Fig. 5d).


Recently it has been shown that ectopic expression of mutant KRAS in acinar cells using the rat elastase promoter led to preinvasive pancreatic neoplasia.27 In addition to mutated KRAS, ErbB2 is also overexpressed in PanIN1A lesions. To analyze the role of ErbB2 in a similar context, we established 3 transgenic mouse lines using the rat elastase promoter driving human ErbB2 expression in pancreatic acinar cells. We confirmed expression of the transgene in all founder lines. In 2 of 3 mouse lines downstream signaling of ErbB2 resulted in RAS activation and ERK1/2 phosphorylation.

Ela-ErbB2 transgenic mice revealed an unexpected finding characterized by lymphocytic infiltration and chronic inflammation in the pancreas. Although somewhat similar to a recent published transgenic mouse line overexpressing mutant KRAS under the control of the CK-19 promoter,28 there are differences in the morphological changes between the two models. We found that the infiltrates in the ErbB2 transgenic mice also harbor CD4+ lymphocytes suggesting that ErbB2 activates signaling pathways involved in the initiation of an inflammatory response. In contrast to the model of Brembeck model we do observe in addition areas of chronic inflammation. The involvement of ErbB2 in chronic inflammation in the pancreas has been highlighted in a study by Friess et al.17 They were able to show that ErbB2 expression is increased in tissue samples from CP patients, in particular in a subgroup of patients showing pancreatic head enlargement. Although being a rather rare event, at late age transgenic founder lines I and III displayed focal areas of chronic inflammation. Our transgenic mouse model therefore recapitulates characteristics observed in the mouse model presented by Brembeck et al.28 and clinical findings described by Friess et al.17

An important mediator in this scenario might be NF-κB/Rel. This transcription factor is a key factor in initiating an inflammatory responses via pro-inflammatory cytokines and chemokines, such as the CC chemokine CCL3.24, 29 In line with this hypothesis we detected higher levels of CCL3 in the pancreas of transgenic mice, which may contributes to the recruitment of the also detected CCR-1 positive T-cells.

Although ErbB2 transgenic mice display RAS activation and phosphorylation of downstream effectors such as ERK1/2, we could not detect an increased proliferation rate in the pancreas of transgenic mice compared with nontransgenic littermates. This was unexpected since ectopic expression of ErbB2 in mammary glands has been shown to induce a strong proliferative response.37 It is known that the expression of cyclin D1 and its assembly into a complex with CDK4 and CDK6 requires both RAS activation and signals via the RAF-MEK-ERK/MAPK pathway in early G1-phase.30 Although microarray analysis revealed higher expression levels of cyclin D1 mRNA, protein levels were lower in ErbB2 transgenic mice compared with littermate controls. This shows that RAS induced transcriptional activation of cyclin D1 did not result in higher protein levels in our model. Another possible mechanism by which ErbB2 mediated RAS activation fails to induce proliferation is the induction of the tumor suppressor protein p16INK4a. The protein p16INK4a, encoded by the INK4A gene, inhibits CDK 4/6-mediated phosphorylation of Rb and thereby blocks entry into S-phase of the cell cycle. RAS mediated p16INK4a induction has been shown to be important for premature senescence, a potential defense mechanism against oncogene activation.31, 32 The relevance of this mechanism has been demonstrated in in vivo models in which oncogenic KRAS has been expressed in the pancreas.33 Therefore, it is conceivable that acinar cells overexpressing ErbB2 and activate RAS undergo cellular senescence via a p16INK4a dependent mechanism.2, 34, 35, 36 Cyclin D1 instability and upregulation of p16INK4a might be responsible for differences observed between our transgenic mouse model and the one described by Grippo et al.27

In summary we have generated a transgenic mouse model overexpressing human ErbB2 under the control of the rat elastase promoter to elucidate its role in pancreatic carcinogenesis and chronic inflammation. Although ErbB2 has been shown to be upregulated in early pancreatic lesions, transgenic mice did not develop tumors or preneoplastic lesions but prominent lymphocytic infiltrations and focal areas of chronic inflammation. This mouse model might give insights into the role of ErbB2 in inflammatory processes of the pancreas.


We thank Mrs. Rosi Rittelmann, Mrs. Janet Köhler and Mrs. Beate Knobel for excellent technical assistance and Mrs. Andrea Lohner and Mrs. Karen Dlubatz for support in mouse experiments. We further thank Dr. Roger Vogelmann and Dr. Matthias Treiber for critical reading of the manuscript.