Background
Leukemia inhibitory factor (LIF) is a soluble glycoprotein that belongs to the family of interleukin (IL)-6-type cytokines. Other members of this family include IL-6, IL-11, ciliary neurotrophic factor (CNTF), oncostatin M (OSM), cardiotrophin-1 (CT-1) and novel neurotrophin-1 (NNT-1) [
1], which display pronounced trophic as well as protective properties during pathophysiology of the central nervous system (CNS) and are hence referred to as neuropoietic cytokines or neurokines [
2]. Specific functions of LIF in the nervous system include induction of cholinergic differentiation of sympathetic neurons, induction of neuropeptide and choline acetyltransferase (ChAT) gene expression [
3], regulation of polyneuronal innervation of neuromuscular junction [
4,
5] and regulation of the HPA axis [
6,
7]. Furthermore, LIF signaling is crucial for development of the nervous system, including development of sensory and motor neurons [
8,
9] and glial cells [
10]. Consistently, reduced numbers of astrocytes and oligodendrocytes are found in LIF knock-out mice [
11]. During inflammation, LIF has been suggested to be both pro- and anti-inflammatory and appears to play a key role in neural injury and regeneration. We and others have previously demonstrated the neuroprotective properties of LIF against damages caused by excitotoxicity, light,
et cetera[
12‐
14]. Moreover, promotion of axonal regeneration and oligodendrocyte growth and survival by LIF suggests its potential for reducing damage associated with central inflammatory demyelinating diseases such as multiple sclerosis [
15‐
17].
In the CNS, astrocytes are considered to be the major source for LIF [
18,
19], and its expression in the brain is significant during pathological conditions including ischemia [
20,
21], multiple sclerosis [
22], Alzheimer’s and Parkinson’s diseases [
23] and brain injury [
24]. The factors responsible for elevated LIF induction during CNS pathology are largely unknown. One of the candidates identified recently to induce LIF expression in astrocytes is ATP [
18,
25], levels of which also rise during conditions like high-frequency neuronal activity, seizure, ischemia and hypoxia [
26,
27]. However, extracellular ATP is rapidly hydrolyzed by a cascade of ectonucleotidases resulting in an enhanced level of adenosine [
26,
28]. Correspondingly, excitotoxic conditions such as ischemia, hypoxia, seizure and head injury are known to induce a rapid increase in extracellular adenosine concentrations, up to 100 times that of the resting concentration [
29‐
33]. There is abundant evidence for immune regulation by adenosine [
34] including expression and release of growth factors and cytokines such as nerve growth factor (NGF), S100beta, IL-6 and CCL2 in glial cells [
35‐
39]. However, it is not known whether adenosine can induce LIF expression in astrocytes.
In the present study, we investigated the potential influence of adenosine receptor activity on LIF release from cultured astrocytes.
Methods
Chemicals and reagents
Neurobasal media, Hank’s balanced salt solution (HBSS), phosphate-buffered saline (PBS), sodium pyruvate, L-glutamine, penicillin-streptomycin, hydroxyethyl piperazineethanesulfonic acid (HEPES), glutaMAX-1 and B27 supplement were obtained from Gibco (Breda, The Netherlands). Dulbecco’s modified Eagle’s medium (DMEM) and fetal calf serum (FCS) were obtained from PAA Laboratories (Cölbe, Germany). Trypsin was obtained from Life Technologies (Breda, The Netherlands). L-leucine methyl ester (LME) and the remaining cell medium components were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands). Recombinant mouse LIF (rmLIF: LIF2005) was obtained from Millipore (Amsterdam, The Netherlands). Brefeldin A (BFA), caffeine, L-glutamate, adenosine A2B receptor antagonist (MRS 1754), protein kinase A (PKA) inhibitor (KT 5720), protein kinase C (PKC) inhibitor (Ro 31–8220), p38 mitogen-activated protein kinases (MAPK) inhibitor (SB 203580), and adenosine analog (5′-N-Ethylcarboxamide or NECA) were obtained from Sigma-Aldrich (Zwijndrecht, The Netherlands). Non-hydrolysable ATP (2MeSATP), adenosine A2A receptor antagonist (ZM 241385), adenosine A2A receptor agonist (CGS 21680) and MEK1/2 inhibitor (U 0126) were obtained from Tocris Bioscience (Bristol, UK). NF-kB inhibitor (BAY 11–7082) and c-Jun N-terminal kinase (JNK) inhibitor (SP 600125) were obtained from Calbiochem (Darmstadt, Germany). Reagents used in immunoblotting experiments were purchased from Bio-Rad Laboratories (Veenendaal, The Netherlands) with the exception of the polyvinylidene fluoride (PVDF) membranes that were obtained from Millipore (Bedford, MA).
Animals
Wild-type C57BL/6 J (1 to 2 days postnatal) mice were obtained from Central Laboratory Animal Facility (University of Groningen, The Netherlands). Adenosine A2B receptor knock-out (A2B KO) mice (1 to 2 days postnatal) with the same genetic background were kindly provided by Professor Marco Idzko (University of Freiburg, Germany). Wild-type C57BL/6 J (14 to 15 days embryonic) mice were obtained from Harlan (Horst, The Netherlands). All procedures were in accordance with the regulation of the Ethical Committee for the use of experimental animals of the University of Groningen, The Netherlands (License number DEC 4623A and DEC 5913A). Animals were housed in standard Makrolon™TM (Bayer AG, Leverkusen, Germany) cages and maintained on a 12 hour light/dark cycle. They received food and water ad libitum.
Primary neuronal culture
Primary culture of cortical neurons from mouse embryo (~E
15) was established as described previously [
13]. Briefly, cortices from embryonic brains were dissected in ice-cold HBSS supplemented with 30% glucose. Meninges were removed, and the tissues were treated with trypsin before they were gently dissociated by trituration in neuronal culture media (neurobasal medium supplemented with 2% B27, 1 mM sodium pyruvate, 2 mM L-glutamine and 50 U/mL penicillin-streptomycin). The cell suspension was filtered using cell strainer (70 μm) (BD Falcon, Franklin Lakes, NJ, USA) before centrifugation (800 rpm for 10 minutes). Cells were then seeded on poly-D-lysine-coated six-well plates (1.5 x 10
6 cells/well) and maintained in neuronal culture media in a humidified atmosphere with 5% CO
2 at 37°C. The culture medium was refreshed the next day to get rid of debris. The neuronal purity as determined by Microtubule-associated protein 2 (MAP2)-staining was around 98% (data not shown) [
13]. Cultures were used after 5 days
in vitro.
Induction of excitotoxicity
Cortical neuron cultures were subjected to an excitotoxic challenge with glutamate (50 μM, for 1 hour), after which cultures were refreshed with fresh media and were incubated at 37°C. Supernatants from neuron cultures (untreated and glutamate-treated) were collected 18 hours after glutamate challenge and were applied to the primary astrocyte cultures.
Primary astrocyte cultures
Primary astrocyte cultures were established from cerebral cortices of postnatal (1 to 2 days) C57BL/6 J and A
2B KO mice according to a previously described procedure [
40], which was modified to reduce microglial contamination [
41]. Microglial cells were separated from the astrocytic monolayer by 1-hour shake-off at 150 rpm. This procedure was repeated two times with an interval of 4 days
in vitro between each shake off, followed by an overnight shake-off at 240 rpm to remove oligodendrocyte precursor cells. Purified astrocytes were washed with HBSS buffer containing 1 mM ethylenediaminetetraacetic acid (EDTA) and further detached using HBSS with 0.1% trypsin. Cells were reseeded with fresh astrocyte culture medium (DMEM supplemented with 5% FCS, 2 mM L-glutamine, 1 mM sodium pyruvate and 50 U/mL penicillin-streptomycin) in multi-well plates (5 x 10
4 cells/cm
2) and maintained in culture to confluency. To further reduce microglial contamination, confluent astrocyte cultures were treated with 5 mM LME, a lysosomotropic agent [
42], for 4 to 5 hours. Astrocytes were ready for experiments after 1 to 2 days. Our cell preparations had a high percentage of astrocytes (≥95%), which was confirmed by immunostaining against GFAP (astrocyte specific marker) and CD11b (microglial specific marker) (data not shown).
Real-time polymerase chain reaction
Total RNA of primary astrocytes was extracted, purified and transcribed into cDNA as described previously [
13]. Quality of the cDNA was examined using the following housekeeping gene
Glyceraldehyde-3-phosphate dehydrogenase (
GAPDH) primer pairs: Fw 5′-CATCCTGCACCACCAACTGCTTAG-3′ and Rev 5′-GCCTGCTTCACCACCTTCTTGATG-3′ [Accession number: NM-008084]. The effect of neuronal supernatants and NECA on LIF mRNA expression in cultured astrocytes was analyzed by real-time PCR (qPCR) using the iCycler and iQ™ SYBR™ Green supermix (Bio-Rad, Veenendaal, The Netherlands ). Mouse
Hypoxanthine phosphoribosyltransferase 1 (
HPRT1) and
GAPDH primers were used for normalization to housekeeping genes (data normalized to
HPRT1 are not shown), and these genes showed no variations in response to the experimental treatments. The primer pairs used for qPCR were: LIF (Fw 5′-ATGTGCGCCTAACATGACAG-3′ and Rev 5′-TATGCGACCATCCGATACAG-3′) [Accession number: NM-008501];
GAPDH (Fw 5′-ATGGCCTTCCGTGTTCCTAC-3′ and Rev 5′-GCCTGCTTCACCACCTTCTT-3′) [Accession number: AF106860] and
HPRT1 (Fw 5′-GACTTGCTCGAGATGTCA-3′ and Rev 5′-TGTAATCCAGCAGGTCAG-3′) [Accession number: NM-013556]. The comparative C
t method (amount of target amplicon X in Sample S, normalized to a reference R and related to a control sample C), was calculated by:
and was used to determine the relative gene expression levels [
43].
Western blot
Western blotting on cultured cortical astrocytes was performed as previously described [
13]. Equal amounts of protein (30 μg) were loaded to 12.5 or 15% sodium dodecyl sulfate-polyacrylamide gels and subsequently transferred to PVDF membranes. The membranes were blocked using Odyssey™ Blocking Buffer (OBB; LI-COR Biosciences, Cambridge, UK; diluted 1:1 in PBS) for 1 hour and incubated overnight at 4°C with different combinations of primary antibodies (diluted in 1:1 OBB and PBS + 0.1% Tween 20 (PBS-T)): mouse monoclonal anti-β-actin (1:8000, Abcam, Cambridge, UK); rabbit monoclonal anti-phospho-NF-κB p65 (Ser536) (1:1000, Cell Signaling Technology, Leiden, The Netherlands); and rat monoclonal anti-LIF (MAB449; 1 μg/mL, R&D Systems, Oxford, UK). The next day, membranes were washed in PBS-T (four times for 5 minutes each time) and incubated for 1 hour at room temperature with appropriate fluorescence conjugated secondary antibodies (diluted in PBS-T): donkey anti-mouse IR Dye 680 (1:10000, LI-COR Biosciences, Cambridge, UK); goat anti-rat IR Dye 680 (1:10000, LI-COR); and donkey anti-rabbit IR Dye 800CW (1:10000, LI-COR). Membranes were washed again in PBS-T (four times for 5 minutes each time) and the fluorescent bands were detected using LI-COR’s Odyssey™ infrared imaging system.
Leukemia inhibitory factor ELISA
A total of 1 mL of supernatant was collected from each well of the six-well plates of primary mouse astrocyte cultures, and these samples were stored at −20°C. ELISA plates (96-well, Costar, Corning Life Sciences, Amsterdam, The Netherlands) were coated overnight at room temperature with 100 μl/well of primary antibody goat anti-LIF (AF449; 0.5 μg/mL, R&D Systems, Oxford, UK) diluted in 0.01 M PBS (pH 7.4). The following day, the plates were washed six times with wash buffer (0.25 M Tris–HCl pH 8, 0.15 M NaCl, 0.05% Tween-20) using an automated microplate washer and air dried (this step is repeated after each incubation step). Plates were subsequently incubated for 1 hour at room temperature with 200 μl/well of blocking buffer (0.01 M PBS, 2% BSA). After blocking, the plates were incubated with supernatants from astrocyte cultures (100 μl/well) for 2 hours at room temperature. Two dilutions (1:2 and 1:4) of each sample, diluted in incubation buffer (0.01 M PBS, 0.2% gelatin, 0.05% Tween-20), were made in triplicates. The plates were then incubated for 1 hour at room temperature with 100 μl/well of the detection antibody, biotinylated goat anti-LIF (BAF449; 0.05 μg/mL; R&D Systems, Oxford, UK) diluted in incubation buffer, followed by an incubation for 30 minutes at room temperature with 100 μl/well of Streptavidin-horseradish peroxidase (HRP) conjugate (1:8000, Sanquin Reagents, Amsterdam, The Netherlands). The plates were then incubated for 15 to 20 minutes at room temperature with 100 μl/well of TMB detection buffer (0.1 M acetate buffer, 0.1 M sodium-acetate, pH adjusted with 1 M citric acid (0.21 g/mL; dissolve 2 tablets of 3, 3′, 5, 5′-tetramethyl benzidinedihydrochloride in 11 mL of TMB buffer and add 2 μl of 30% H2O2)). Upon stable color formation the reactions were stopped by adding 100 μl/well of 1 M H2SO4. Absorbance of the samples was measured using VersaMax, a spectrophotometric ELISA plate reader, and SoftMax Pro software (Molecular Devices, CA, USA) at 450 nm, with a background correction at 575 nm. Recombinant mouse LIF (15 to 2000 pg/mL) was used to plot the standard curve.
MTT assay
Survival of cultured embryonic cortical neurons or cultured neonatal astrocytes against various experimental treatments was measured by the colorimetric MTT (3-(4,5-dimethylthiazol-2-yl-) 2,5-diphenyltetrazolium bromide) assay, as described previously [
44]. MTT solution (0.5 mg/mL final concentration) was added to cultured cells and incubated for 4 hours, after which, cells were lysed and MTT-formazan solubilized in dimethyl sulfoxide (DMSO) on an orbital shaker for 15 minutes. Optical density measure of each sample was determined using an automated ELISA reader - the Varioskan Flash spectral scanning multimode reader (Thermo Scientific, FL, USA) at 570 nm, with a background correction at 630 nm.
Immunocytochemistry and confocal microscopy
Astrocytes cultured on glass cover slips were fixed for 15 minutes in 4% paraformaldehyde. After several washes in PBS, the cells were blocked for 45 minutes with 5% normal goat serum (Vector Laboratories, Burlingame, CA, USA) in PBS containing 0.1% TritonX (Sigma, Zwijndrecht, The Netherlands). The cover slips were then incubated overnight at 4°C with rat anti-LIF primary antibody (5 μg/mL, R&D Systems, Oxford, UK) in combination with one of the following primary antibodies: rabbit anti-Rab11 (1:400, Zymed, San Francisco, CA, USA); rabbit anti-chromogranin A & B (1:100, Novus Biologicals, Cambridge, UK); mouse anti-clathrin (1:1000, Abcam, Cambridge, UK); rabbit anti-pHogrin C-terminal (1:100, kind gift of Professor J.C. Hutton (Denver, USA)) and rabbit anti-giantin (1:1000, Covance, Princeton, NJ, USA). The following day, cells were rinsed three times with PBS and incubated for 1 hour with the appropriate secondary antibodies: donkey anti-rat CY3 (1:500, Jackson ImmunoResearch Laboratories, Uden, The Netherlands); donkey anti-rabbit Alexa Fluor 488 (1:500, Molecular Probes, Leiden, The Netherlands)and donkey anti-mouse Alexa Fluor 488 (1:500, Molecular Probes). The cover slips were then rinsed with PBS and mounted on microscopic slide with Mowiol (Sigma, Zwijndrecht, The Netherlands) and analyzed with a Leica SP2 AOBS system (Leica Microsystems, Rijswijk, The Netherlands). Pictures were deconvoluted using the software Huygens Pro (SVI, Hilversum, The Netherlands). Primary antibody omission served as the control.
Statistical data analysis
The absolute data values were normalized to the control in order to allow multiple comparisons. Statistical analyses were performed by one-way analysis of variance (ANOVA) followed by Bonferroni post-hoc test, using the Statistical Package for the Social Sciences (SPSS, Chicago, IL, USA). In all cases, P values < 0.05 were considered statistically significant.
Discussion
We have previously shown that recombinant LIF protects neurons against glutamate-induced excitotoxicity [
13]. In this study, we investigated the mechanism by which astrocytes produce and release LIF. Here we show that glutamate-induced neuronal excitotoxicity leads to adenosine receptor-mediated increase in LIF mRNA expression in cultured cortical astrocytes. We demonstrate that the upregulation of LIF mRNA and protein is adenosine A
2B receptor-dependent, and is mediated through G
q/11-PLC-PKC-MAPK-NF-κB signaling pathways. We furthermore show that LIF is transiting through the Golgi and is found in recycling endosomes rather than in LDCV. Finally, LIF produced by astrocytes can protect neurons against excitotoxicity.
It has been known for more than a decade that astrocytes are the major source for LIF in the CNS [
18,
19,
59,
60]. However, the factors responsible for the regulation of LIF expression in these cells are still largely unknown. It is well known that stressed neurons release nucleotides such as ATP and adenosine [
30,
61]. Recently, it was demonstrated that astrocytes increase LIF production and release in response to ATP receptor stimulation [
18]
. In this study, the authors demonstrate that neurons during action potentials can secrete ATP, which triggers LIF production in astrocytes. This ATP-dependent upregulation of LIF by astrocytes is responsible for the promotion of oligodendrocyte-mediated myelination around neuronal axons. ATP is also known to be secreted by neurons during stressful conditions such as seizure, ischemia and hypoxia [
26,
27]. However, when we blocked adenosine receptors with the non-selective antagonist caffeine, or with specific A
2A/A
2B receptor antagonists, the effect of glutamate-stressed neuronal supernatants on LIF expression in astrocytes was absent, suggesting that adenosine, but not ATP, is responsible for astrocytic LIF production during glutamate-induced neuronal stress. Thus, it might be hypothesized that depending on the CNS status, astrocytic LIF expression and secretion is differentially regulated; during normal neuronal activity and development ATP is involved whereas during neuronal insults, adenosine might enhance LIF secretion by astrocytes.
Several studies have demonstrated the involvement of adenosine A
2B receptors in the regulation of IL-6 expression in various cell types
in vitro[
38,
47,
48,
62,
63] as well as
in vivo[
64], suggesting that A
2B receptors might also be essential in the regulation of other IL-6-type cytokines. Our results show that adenosine-dependent LIF regulation is mediated through the A
2B receptor, since no increase in LIF expression was found in cultured astrocytes from A
2B receptor deficient mice. Instead NECA caused a down-regulation of LIF mRNA after 8 and 24 hours in these cells, indicating that knocking out A
2B receptors may have unmasked an inhibitory effect on LIF mRNA expression of an unidentified adenosine receptor. Whether or not this might explain the very short-lived effect of NECA on LIF mRNA expression in wild-type astrocytes is at the moment unclear and a subject of future investigations. We furthermore demonstrated that A
2B-mediated LIF expression is dependent on the PKC, but not the PKA pathway. These data are in line with the study of Aloisi and colleagues, which demonstrated that LIF modulation by pro-inflammatory cytokines in human astrocytes was mediated through PKC activation [
59]. Moreover, PKC has also been shown to be essential in IL-6 regulation [
47,
48,
62,
65], revealing a prominent role for PKC in the signaling pathway controlling LIF gene expression.
MAPKs have been reported to be involved in adenosine A
2B receptor-mediated regulation of IL-6 gene expression in astrocytoma cells [
48]. In our experiments, both basal as well as NECA-induced LIF gene expression and release in cultured astrocytes were inhibited by specific inhibitors of p38 and ERK1/2, but not JNK-MAPKs. In line with our findings, it has been shown that LIF expression in Schwann cells is mediated through PKC pathway-induced ERK1/2 activation [
49]. Furthermore, we show here that adenosine-dependent LIF expression in astrocytes is regulated through the NF-κB transcription factor. This observation is in line with several studies showing an NF-κB-dependent regulation of IL-6 gene by this transcription factor in several cell types [
38,
50,
51,
65,
66]. It has been shown that NECA-induced NF-κB activation and the resultant IL-6 gene expression was abolished by inhibitors of MAPK pathways [
65]. In our study, preliminary observations indicate that NECA-induced activation of the NF-κB pathway is reduced by selective inhibitors of p38 and ERK1/2 pathways (data not shown), suggesting that these pathways might play as upstream mediators in NF-κB-dependent LIF expression in astrocytes.
Recent evidence indicates that, depending on the cell type, different secretory pathways are employed for cytokine release [
67]. For example, T cells use two different release mechanisms: IL-2 and IFN-γ are secreted at the immunological synapse whereas CCL3 and TNF-α are secreted multidirectionally, suggesting different secretory pathways [
68]. In neurons or neuron-like cells, secretory granules called LDCVs are the organelles used for the selective secretion of IL-6, TGF-β2 and CCL21 [
53,
54,
69]. The same organelles are also used in immune cells such as mast cells and neutrophils [
67]. Here we show that LIF protein is transported through Golgi but its secretion by astrocytes is not mediated by secretory granules. Instead, LIF co-localizes with Rab11, a known marker of recycling endosomes [
57,
58]. Moreover, we observed a partial co-localization of LIF with clathrin, which also associates with recycling endosomes where it is implicated in protein sorting [
56]. Recycling endosomes have now been shown to be responsible for cytokine secretion in several cell types. For example, IFN-γ and TNF-α secretion from natural killer cells require Rab11 [
70]. Recycling endosomes are also responsible for the constitutive secretion of IL-6 and TNF-α in macrophages [
71]. Further studies will be needed to better understand LIF sorting, trafficking and release by these vesicles.
Interestingly, our data indicate that LIF is constitutively released from astrocytes. Indeed constant levels of LIF were present in the supernatants of untreated astrocytes when measured by ELISA. Similar data were observed in human astrocyte cultures [
59]. Whether this observation is representative of the physiological behavior of astrocytes
in vivo or is due to the culture conditions remains to be determined. We further show that by blocking the early secretory pathway with BFA, the LIF concentration in the culture supernatant was not increased upon NECA stimulation. The inhibitory effect of BFA indicates that LIF passes through the Golgi prior to its secretion, and thus does not follow non-conventional secretory pathways that by-pass the Golgi and is typically insensitive to BFA, which has recently been reported to be used by other cytokines [
72]. Importantly, the inhibitory effect of BFA suggests that NECA-stimulated release of LIF by astrocytes requires
de novo LIF synthesis, and does not involve a ready-releasable post-Golgi pool of LIF.
It is now clear that one of the major roles of LIF is directed toward cell protection. Indeed, it has been shown that LIF is up regulated in astrocytes and neurons after cerebral ischemia [
21] as well as in astrocytes after cortical brain injury [
24], suggesting a role of LIF in neuronal repair or protection. In line with these data, treatment of rat with LIF prevented loss of motoneurons after peripheral nerve injury [
73,
74] and protection of retinal ganglia cells was compromised in LIF knock-out mice after lens injury [
12]. Finally, LIF was shown to limit demyelination in an experimental autoimmune encephalomyelitis mouse model [
16] and has become a prominent therapeutic candidate for multiple sclerosis [
17]. We have previously shown that LIF can protect cortical as well as hippocampal neurons against glutamate-induced excitotoxicity [
13]. Here we show that LIF coming from the supernatant of NECA-treated astrocytes has the same protective effect. Indeed, astrocytes produce several other cytokines and neurotrophic factors including IL-6, NGF, brain-derived neurotrophic factor, neurotrophin-3, S-100β protein and TGFβ [
75], that might help neurons to cope with excitotoxic stress. Accordingly, conditioned media from astrocyte cultures protected cortical neurons against glutamate (data not shown). In order to confine the neuroprotective effect of astrocytic factors that are released in response to NECA treatment, we had to refresh astrocyte culture medium prior to NECA treatment and testing supernatant on glutamate-stressed neurons. This, together with LIF neutralization, indicates that LIF produced by astrocytes after adenosine receptor stimulation is necessary to witness neuronal protection. Our results provide further evidence for a role of adenosine in neuronal protection. Indeed, it has been shown that adenosine can protect neurons during hypoxia [
76,
77], ischemia [
78,
79] and excessive neuronal activity [
80,
81]. This adenosine protection is often mediated through the A
1 receptor subtype [
82‐
84], but here we show that an indirect protection of adenosine through the stimulation of A
2B receptor on astrocytes leading to LIF upregulation exists. This A
2B receptor activation might be related to an anti-inflammatory process as observed previously by others [
85‐
87].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SM performed the majority of experiments including cell culturing, ELISA and Western blotting. JV did the immunofluorescent part of the study as well as all the microscopy and statistical analysis. Both SM and JV were involved in the redaction of the manuscript. EW did the qPCR experiments. JB took part in ELISA experiments. CHS performed the toxicity and the A2B KO experiments. SCDI and PB were both involved in the secretion part of the study. They gave precious antibodies, were essential in the design of the experiments and participated actively in writing the manuscript. HWGMB and KPHB were both involved in the conception and design of the study as well as in the manuscript redaction. All authors read and approved the final manuscript.