Inhibition of the Nuclear Factor-κB Pathway Prevents Beta Cell Failure and Diet Induced Diabetes in Psammomys obesus

Background High doses of anti-inflammatory drugs, such as aspirin and salicylates, improve glucose metabolism in insulin resistant and type 2 diabetic patients. It has also been shown that the glucose lowering effect is related to the unspecific ability of these drugs to inhibit inhibitor kinaseβ (IKKβ). In this study we have investigated the effect of a selective IKKβ-inhibitor on beta cell survival and the prevention of diet induced type 2 diabetes in the gerbil Psammomys obesus (P. obesus). Methodology/Principal Findings P. obesus were fed a diabetes inducing high energy diet for one month in the absence or presence of the IKKβ-inhibitor. Body mass, blood glucose, HbA1C, insulin production and pancreatic insulin stores were measured. The effects on beta cell survival were also studied in INS-1 cells and primary islets. The cells were exposed to IL-1β and subsequently reactive oxygen species, insulin release and cell death were measured in the absence or presence of the IKKβ-inhibitor. In primary islets and beta cells, IL-1β induced the production of reactive oxygen species, reduced insulin production and increased beta cell death, which were all reversed by pre-treatment with the IKKβ-inhibitor. In P. obesus the IKKβ-inhibitor prevented the development of hyperglycaemia and hyperinsulinaemia, and maintained pancreatic insulin stores with no effect on body weight. Conclusions/Significance Inhibition of IKKβ activity prevents diet-induced diabetes in P. obesus and inhibits IL-1β induced reactive oxygen species, loss of insulin production and beta cell death in vitro.


Introduction
Obesity is characterised by a state of low-grade inflammation [1][2][3], which may contribute to the development of insulin resistance and type 2 diabetes [3,4]. Type 2 diabetes is characterized by insulin resistance, pancreatic beta cell dysfunction and reduced beta cell mass [5]. Although interventions that reduce inflammatory parameters have been associated with improved insulin and glucose metabolism during insulin resistance and type 2 diabetes [2,6], the mechanisms connecting inflammation and diabetes are still unclear.
Inflammatory processes are characterised by the migration of proliferating white blood cells from the circulation to the tissues and activated macrophages and cytokines released from the adipose tissue have been associated with insulin resistance [7,8].
Whereas TNFa has been associated with peripheral insulin resistance [3], islet inflammation and increased beta cell death have in particular been associated with the cytokine IL-1b [9,10]. IL-1b may be produced by beta cells themselves or by homing macrophages within the pancreas. The involvement and therapeutic potential of IL-1b in the treatment of type 2 diabetes was recently demonstrated by treatment with the IL-1b receptor antagonist Anakinra, which resulted in improved beta cell secretory function and glycaemia control in human type 2 diabetic patients [11].
The transcription factor nuclear factor-kappa-B (NFkB) is central in the regulation of inflammatory responses and control of the innate immune system [12,13]. Both TNF-a and IL-1b stimulate NFkB activation. NFkB is activated by a protein complex, where interaction between the inhibitor kinase (IKK) proteins -a, -b, -c leads to phosphorylation of the inhibitor kB (IkB) protein. IkB is subsequently ubiquitinated and targeted for degradation by the proteasome, allowing release and translocation of NFkB to the nucleus and subsequent gene transcription [14,15].
In animal models, the lack of IKKb in immune cells protects the animals from insulin resistance during High fat feeding [16]. It has also been shown that the anti-inflammatory drug salicylate improves insulin resistance in diabetic rodents as well as glycaemia in obese and diabetic patients, at least in part, through the inhibition of the NFkB pathway [17,18].
In most cell-types regulation of inflammatory pathways is associated with the induction of oxygen-derived free radicals, often referred to as reactive oxygen species (ROS). Increased ROS production has been demonstrated to be a unifying mechanism in endothelial cell dysfunction and late diabetic complications [19]. Additionally, ROS seems to be associated with insulin resistance in adipocytes, as normalisation of ROS levels reverses the inflammatory response and increases insulin sensitivity of these cells [20]. Observations in beta cells have shown that IL-1b mediated toxicity is in part mediated through induced production of reactive oxygen species in particular nitric oxide (NO) [21,22].
We have utilized the gerbil Psammomys obesus (P. obesus) in our studies, as it is a rodent model of nutritionally induced type 2 diabetes. When transferred from their normal low-calorie saltbush diet to a more energy rich diet (regular rodent laboratory chow) the majority of the animals develop diabetes within a few weeks [23,24]. This animal model is characterized by a fast depletion of insulin stores as a diabetic phenotype develops [24]. Similar to humans, these animals display increased pancreatic inflammation and associated beta cell death [9,25].
The aim of the present study was to evaluate if inhibition of the IKKb and NFkB pathways could prevent 1) IL-1b mediated beta cell toxicity in vitro and 2) diet induced diabetes in vivo. This was done using a selective IKKb-inhibitor originally developed against steroid-resistance asthma [26].

Results
Inhibition of the NFkB pathway inhibits IL-1b induced beta cell failure in INS-1 cells We first tested our hypothesis that inhibition of the NFkB pathway with the IKKb-inhibitor would rescue beta cells against IL-1b-induced toxicity. This was done in INS-1 cells by measuring viability and apoptosis during IL-1b exposure. Forty-eight hours of IL-1b exposure reduced mitochondrial activity around two-fold ( fig. 1A) in parallel with an increased caspase 3/7 activity ( fig. 1B). Co-incubation with the IKKb-inhibitor completely restored mitochondria activity and significantly reduced caspase 3/7 activity.
Since previous observations in beta cells have shown that IL-1b mediated beta cell destruction is in part mediated through nitric oxide (NO) production [27], we also measured expression levels of inducible nitric oxide synthase (iNOS). As expected iNOS transcription was massively induced in response to 24 hours IL-1b exposure, which was fully prevented by co-incubation with the IKKb-inhibitor (Fig. 1C). To examine if the IKKb-inhibitor also protected against IL-1b induced loss of insulin secretion, accumulated and GSIS were measured. This was done after 48 hours incubation in the absence or presence of IL-1b and with or without the IKKb-inhibitor. As seen from fig. 1D accumulated insulin production was halted by IL-1b exposure, but completely unaffected when the IKKbinhibitor was also present. Analysis of GSIS in control cells demonstrated a more than three-fold induction (8,2 vs. 26,1 ng/ml insulin) (Panel E). Cells exposed to IL-1b for 2 days displayed a diminished insulin production at both basal glucose (Panel E white columns, although non-significant) and followed stimulation at high glucose (Panel E. black columns). Co-incubation with the IKKbinhibitor significantly protected the cells against the IL-1b-induced decrease in insulin secretion (16,1 vs. 6,2 ng/ml insulin).

IL-1b induced radical formation prevented by the IKKb-inhibitor
To further characterize the effects of IL-1b on INS-1 cells, the production of radical formation was measured by H 2 DCFDA. H 2 DCFDA is normally referred to as an indicator of ROS despite its affinity for both reactive oxygen and nitrogen species (Table 1) [28]. We confirmed and extended previous observations demonstrating that that IL-1b exposure caused the level of ROS to rise in a time and concentration dependent manner with maximal production after 24 hours ( Fig. 2A). Co-incubation with the IKKb-inhibitor reduced the ROS production in a dose-dependent manner (Fig. 2B). 0,5 mM of the IKKb-inhibitor attenuated the effect of 50 pg/ml IL-1b whereas a double dose, 1 mM was required to inhibit the effect of the 150 pg/ml IL-1b treatment. A macroscopic examination revealed that not only a subpopulation of dividing cells, but all cells were affected by the treatments (Fig. 2C).

Effect of IL-1b and the IKKb-inhibitor using primary newborn rat islets
To evaluate if the effects observed in the INS-1 beta cell line could be observed in primary beta cells, we isolated newborn rat islets and exposed them to IL-1b in the absence or presence of the IKKb-inhibitor. We found that IL-1 b increased the production of ROS by more than two-fold and this effect was attenuated by by the IKKb-inhibitor ( fig. 3A). In parallel, iNOS expression was significantly up-regulated in IL-1b, which was prevented by the NFkB pathway inhibitor (Fig. 3B).

Effects of IL-1b and IKKb-inhibition on islets of Langerhans isolated from P. obesus
To evaluate the in vivo potential of the IKKb-inhibitor in the prevention of type 2 diabetes, the animal model P. obesus was selected. First we established the effects of the mouse IL-1b and the IKKb-inhibitor on islets isolated from healthy adult animals. We measured the production of ROS and insulin following IL-1b exposure with or without co-incubation with the IKKb-inhibitor. Exposure to IL-1b for 24 hours increased the production of ROS, which was attenuated by the inhibitor ( fig. 4A). At basal glucose insulin secretion was not different between control islets and islets exposed to IL-1b or the IKKb-inhibitor ( fig. 4B white columns). However, at high glucose ( fig. 4B black columns) IL-1b significantly Table 1. Fluorescence response of H 2 DCFDA to various reactive oxygen and reactive nitrogen species [28].  reduced insulin secretion (from 49,0 to 7,2 ng/ml), an effect which was significantly attenuated in the presence of the IKKb-inhibitor (19,5 vs. 7,2 ng/ml insulin).

IKKb-inhibition prevents diet induced diabetes in P. obesus
We next examined if inhibition of the NFkB pathway could prevent diet induced diabetes. We used the diabetes prone P. obesus on a high energy diet treated with vehicle or 60 mg/kg/day IKKb-inhibitor for 28 days. As seen in fig. 5A the inhibitor treatment had no effect on body weight gain. Vehicle treated animals developed diet induced diabetes during the 28 days on a high-energy diet ( fig. 5B). Within one week the animals had a significant increase in non-fasting BG levels (from 3,4 mM at day one to above 10 mM at day 7 (p,0,001)), which was further increased to more than 15 mM by day 28 5D) were significantly lower than the vehicle treated control animals and did not show a significant increase from day 0 to day 28.
To further evaluate the effects of treatment on beta cell function and capacity, we stained for insulin and measured total insulin content in acid ethanol extracts of pancreas from animals treated for 28 days with vehicle or the IKKb-inhibitor in addition to age matched non-diabetic animals. As seen in figure 6A we found an inverse correlation between pancreatic insulin content and blood glucose. Furthermore, the treatment had a significant effect on pancreatic insulin supply, as reflected by a more than three-fold increase of pancreatic insulin content in animals treated with the IKKb-inhibitor compared to the vehicle treated animals (day 28: IKKb-inh. 2667,2 mg vs. vehicle 761,5 mg; p,0,05). After 28 days on high energy diet insulin stainings also showed that most beta cells barely expressed insulin ( fig. 6B). However, animals treated with the IKKb-inhibitor showed clear and dense insulin staining of most beta cells ( fig. 6C), which were comparable with insulin staining of beta cells in non-diabetic lean controls animals ( fig. 6D).

Discussion
The transcription factor NFkB is a ''master switch'' in inflammatory processes. NFkB is found in almost all cell types and is involved in cellular responses to stimuli such as stress, cytokines, free radicals, ultraviolet irradiation, oxidized LDL, and bacterial or viral antigens. A connection between inflammation, islet function and type 2 diabetes has been found. Non-specific inhibitors of the NFkB pathway, including aspirin and aspirinrelated drugs, have shown positive therapeutic effects in diabetic patients, indicating the importance of NFkB activation in the development of type 2 diabetes [2,6,17,18,29].
As beta cell dysfunction and death in type 2 diabetes has been associated with cytokine-induced radical formation [25], we incubated INS-1 cells, rat islets and P. obesus islets with IL-1b and measured free radical formation, beta cell function and beta cell death. In isolated beta cells and islets, IL-1b increased the production of free radicals significantly, which was reflected in a loss of insulin production and beta cell death. In beta cells, IL-1binduced free radical formation was shown in a time and dose dependent manner, with maximum production after 24 hours. IL-1b induced beta cell death has been associated with increased  activity of NFkB. One key NFkB-regulated gene that is associated with beta cell death, is iNOS. iNOS is primarily responsible for the production of NO radicals [21,22], and is normally silent, however IL-1b treatment induces a significant induction of the iNOS mRNA expression. We found that IL-1b up-regulated the expression of iNOS by greater than 500-fold in both beta cells and islets. IL-1b induced iNOS transcription, radical formation, loss of insulin production and beta cell death were all reversed by the IKKb-inhibitor, indicating a link between iNOS activity, radical formation, beta cell function and death. A tight redox balance is important for the beta cell since it is weakly protected against oxidative damage, as a result of low levels of the defense proteins superoxide dismutase and glutathione peroxidise [30,31]. There are many reactive species, each with various harmful effects. One example is the superoxide radical (O 2 *-), which is produced continuously as a by-product of oxidative phosphorylation, but is not very reactive. Similar to superoxide, the NO radical reacts slowly, however, when the two are combined together they form the highly toxic peroxynitrite radical (ONOO 2 ) [32]. The probe used in our study to quantify radical formation has a low affinity for both NO and superoxide compared to peroxynitrite, and since peroxynitrite effectively oxidizes proteins, lipids and DNA, it is likely that IL-1b induced radicals caused beta cell failure and ultimately beta cell death.
P. obesus is a well established model of diet induced type 2 diabetes. When these animals are switched from low caloric density diet to a high caloric density diet they rapidly develop obesity, hyperglycaemia and hyperinsulinaemia [24,33]. In this study, a high-energy diet resulted in a rapid rise in insulin levels in the vehicle treated animals. However, the higher level of insulin was not sufficient to maintain normal blood glucose control as both BG and subsequently HbA 1C continued to rise. After three weeks of a highenergy diet there was a decrease in insulin levels and a further deterioration in BG control. Loss of insulin production is expected Figure 5. IKKb inhibition prevents diet induced diabetes in P. obesus. Diabetes-prone male and female P. obesus, age of 13-14 weeks, were transferred from ad libitum low energy diet to ad libitum high energy diet. At the same day (day 0) treatment was initiated with vehicle (open circles n = 12) or with the IKKb-inhibitor (black circles n = 10). A: Body weight was measured daily. B: Non fasting blood glucose was measured twice weekly. C: HbA1c was measured once weekly. D: Insulin was measured at day 0, 11 and then twice a week through the rest of the study. All data are shown as the mean 6 S.E.M. (*p,0,05; **p,0,01; ***p,0,001). doi:10.1371/journal.pone.0013341.g005 to coincide with the loss of insulin production and beta cell destruction reported during the late stage of diabetes development in this animal model [24,25,33]. In animals treated with the IKKbinhibitor, the high-calorie diet only resulted in a transient increase in BG. However, a small counter-regulatory increase in insulin production was sufficient to prevent an incipient hyperglycaemia. The protective effect against hyperglycaemia by the treatment with the IKKb-inhibitor was also reflected in HbA 1C levels as these were unchanged through the study.
It is not resolved as to whether the inhibitor treatment resulted in a short-term effect on beta cell function as a result of a lower demand for insulin due to increased insulin sensitivity, or if it has a permanent effect on beta cell survival. However, during the 4 week study, the animals treated with the IKKb-inhibitor showed no sign of insulin depletion or beta cell failure. An early sign of beta cell failure in P. obesus is decreased pancreatic insulin content, which leads to decreased insulin production and reduced plasma insulin levels [24,33]. When we analyzed pancreatic insulin content we found an inverse relationship between pancreatic insulin stores and blood glucose. We also found that animals treated with the IKKb-inhibitor had significantly more pancreatic insulin content, as demonstrated by extraction and immunohistochemistry.
The presented data demonstrates that the inhibition of the NFkB pathway in vitro protects against beta cell failure and in vivo protects against hyperglycaemia and decreased pancreatic insulin content. Still, the in vivo effect on beta cell function could be a secondary effect of increased insulin sensitivity which our experiments did not investigate. However, previous studies have demonstrated that inhibition of IKKb activity protects against lowered insulin sensitivity. For instance, knock out of IKKb in hepatocytes and myeloid cells have shown to protect high fat fed mice against insulin resistance in both muscle, liver and fat [16,34]. Furthermore, heterozygous deletion of IKKb also protects against the development of insulin resistance both during high-fat feeding and in obese Lep ob/ob mice [18]. Non-steroidal antiinflammatory drugs such as aspirin and salicylate also have a glucose lowering effect in both rodents and humans [2,18,29]. These drugs are cyclooxygenase but also function as non-selective inhibitors of NFkB activity, which has been ascribed as the mechanism connected to increased insulin sensitivity [17].
In summary, we have shown that the development of dietinduced diabetes in P. obesus was prevented by treatment with a selective IKKb-inhibitor. The IKKb-inhibitor also prevented cytokine induced loss of insulin production and beta cell death in primary islets and beta cells, possibly by inhibition of ROS production. This was also reflected in vivo since treated animals had much larger insulin reserves compared to the exhausted pancreata in vehicle treated animals. In conclusion, these data support the

IKKb-inhibitor
The IKKb-inhibitor was synthesized at Novo Nordisk A/S. The structure and properties have been published by Ziegelbauer and colleague [26]. In all in vitro experiments a pure form of the IKKbinhibitor was used whereas an enantiomeric mixture of the compound was used in the in vivo assay. Ziegelbauer and colleague [26] referred to the pure and the enantiomeric mixture of the IKKb-inhibitor as COMPOUND A and COMPOUND B, respectively.

Cell culture
A rat insulinoma cell line, INS-1 cells (a gift from Prof. Claes Wollheim, Geneva [35]) was maintained in RPMI 1640 medium with 11 mM glucose supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 mg/ml streptomycin and 50 mM bmercaptoethanol (all from Gibco-BRL). For preparation of cell culture for MTT, Caspase and insulin measurement, 50.000 cells/ well were seeded in 96-well poly-L-lysine coated (to insure adherence) culture dishes (Cyto-well 96F. Nunc, Roskilde, Denmark). Two days after plating, the cells were pre-incubated for one hour with the IKKb-inhibitor (0,5 or 0,1 mM) before IL-1b (mouse IL-1b, Sigma) (50 or 150 pg/ml) were added. Following exposure for the period of time indicated in the figures, insulin secretion, viability (measured as MTT activity), caspase 3/7 activity and production of ROS were quantified. For RT-PCR, 1,5 mil cells/well were seeded in 6-well culture dishes (Nunc, Roskilde, Denmark). Two days after plating, the cells were preincubated for one hour with the IKKb-inhibitor (0,1 mM) before IL-1b (150 pg/ml) was added. Cells were exposed for 24 hours before RNA was extracted.
In vitro islet studies in isolated islets of Langerhans from newborn Wistar rats Pancreata from 3-to 5-day-old Wistar rats (Taconic, Ry, Denmark) were collagenase-digested and the islets were isolated by Percoll gradient purification (GE Healthcare, Denmark) as previously described [36]. Islets were maintained in RPMI 1640 supplemented with 1% penicillin/streptomycin, 10% newborn calf serum, 11 mM glucose (Gibco-BRL) and incubated in a 5% CO 2 humidified atmosphere at 37uC. For the measurement of ROS, about fifteen islets/well were cultured in 96-well poly-L-lysine coated dishes (Cyto-well 96F), whereas 1000 islets/well were cultured in 6-well dishes (Nunc, Roskilde, Denmark) for RT-PCR. One day after plating, islets were pre-incubated for one hour with the IKKb-inhibitor (0,1 mM) before incubation with IL-1b (150 pg/ml) for another 24 hours.

Insulin release
Before measurement of GSIS the culture media were collected and frozen for quantification of accumulated insulin secretion. To determine GSIS, the cells or islets were washed in Krebs-Ringer buffer (115 mM NaCl, 4,7 mM KCl, 2,6 mM CaCl 2 , 1,2 mM KH 2 PO 4 , 1,2 mM MgSO 4 , 5 mM NaHCO 3 , 10 mM HEPES, 0,2% BSA, pH 7.4) containing 3.7 mM glucose, and preincubated for half a hour in the same buffer. The buffer was replaced with a fresh buffer containing 3.7 mM glucose for one hour before the buffer were collected (basal) and replaced with a buffer containing 16.7 mM glucose for an additional one hour (glucose stimulated). Supernatants were collected and frozen for subsequent insulin determination; using a rat-specific luminescence oxygen channeling immuno assay.

Measurement of intracellular ROS production
5-(and-6)-chloromethyl-29,79-dichloro-dihydrofluorescein diacetate (H 2 DCFDA) (Molecular probes) is a cell-permeable dye that within the cells is cleaved by esterases to H 2 DCF. By the action of different ROS (see Table 1) H 2 DCF is subsequently oxidized to its fluorescent form [28]. At the indicated times cells or islets were washed with DPBS (Gibco-BRL) and incubated in the dark with 10 mM fluorescent probe H 2 DCFDA in DPBS. Fluorescence was measured on a NovoStar Microplate Reader (Ramcon, Denmark). Results are shown as delta fluorescence (t90 min -t45 min).

Measurement of gene expression by real-time PCR
Cells were lysed in RLT buffer and islets in Qiazol Lysis Reagent before RNA extraction according RNeasy protocol (Qiagen). cDNA was transcribed using iScript cDNA Synthesis Kit (Bio-Rad laboratories). Samples were assayed on an Mx3000 (Stratagene) using probs and primers designed by Probe Library (Roche Diagnostics). The expression level of the gene of interest was normalized to the housekeeping gene 36B4 and analyzed by the DDCt method [38].

Animals
Male and female P. obesus, aged 7-9 weeks, were obtained from Harlan (Israel). Animals were housed three per cage under controlled conditions with a 12 hours light-12 hours dark cycle, temperature of 25uC and fed low energy chow (LE-2.4 kcal/g; Teklad, Jerusalem, Israel) and water ad libitum. At 10-11 weeks of age animals were challenged with a high energy diet (HE-3.1kcal/g; Purina 5008). Animals that showed sign of hyperglycaemia (BG.10 mM) or no signs (BG,8 mM) during the test period of 10 days were classified as diabetes-prone animals or diabeticresistant animals, respectively. All animals were returned to LE diet. At the age of 14-15 weeks all diabetes-prone animals were normoglycemic again and transferred to an ad libitum high energy diet. Animals were divided into two groups, treated orally with either 2 ml/kg of either vehicle (10% 2-hydroxypropyl betacyclodextrin in saline (HPCD/saline)) or the IKKb inhibitor (60 mg/kg in HPCD/saline) every morning, for 4 weeks. During this period body-weight (BW) was measured daily and morning blood glucose (BG), HbA 1C and insulin was measured one or two times a week. The study was approved by the Animal Experiments Inspectorate, Ministry of Justice, Denmark (Approval number 2001/561-460).

Blood analysis
Blood samples was taken from the tail tip capillary into 10 ml glass capillary tubes and immediately suspended in buffer (500 ml of Biosen analysis buffer) and analysed for glucose on the test day on a Biosen S_Line (EKF-diagnostic, GmbH). For HbA 1C measurement 5 ml was immediately suspended in 250 ml of Hitachi haemolysate buffer and analyzed on a Hitachi 912 (Roche Diagnostics). For insulin measurement blood samples were taken from the tail tip capillary into 80 ml capillary EDTA tubes and immediately centrifugation at 5000 rpm for 5 minutes (at 4uC). Plasma was collected and frozen until measurement of insulin using a rat-specific luminescence oxygen channelling immuno assay.

Pancreatic insulin content
At day 28 animals were anaesthetized by isoflurane before the pancreas was removed as described above. Total pancreas was weighted and sampled using ''the smooth fractionator'' technique as earlier described [39]. Two thirds of the pancreas was homogenized on a tissue-lyzer (Qiagen) at 25 Hz for 2 min in acid-ethanol (30 mM HCl in 75% ethanol) using 1 ml/mg tissue. The samples were placed on a rocker at 4uC overnight before centrifuged at 200006g at 4uC for one hour. Supernatants were collected and the insulin concentration was quantified as described above.

Pancreatic beta cell staning for insulin
One third of the pancreata from Psammomys obesus (sampled as described above) were fixed in 4% paraformaldehyde, followed by paraffin embedding. Sections were deparaffinized and rehydrated. Endogenous peroxidase was blocked by submersion in 0.5% H 2 O 2 for 20 minutes before washed in Tris-buffer saline (TBS) with 0,1% Tween 20 (TBS-T). The remaining steps were carried out in an autostainer (DAKO Corp., Carpinteria, California, USA). Sections were incubated 10 minutes with Avidin blocking solution (DAKO) followed incubation for 10 minutes with Biotin blocking solution (DAKO), with a washing procedure in TBS-T between each step. Sections were incubated for 15 minutes with 10% rabbit serum (X0902, DAKO) in TBS-T before stained for 60 minutes for insulin with guinea pig anti-insulin serum (A0564, DAKO) 1:500 in 7% rabbit serum and 3% rat serum in TBS-T. Sections were washed with TBS-T before incubated for 30 minutes with peroxidise labelled rabbit anti-guinea pig (P0141, DAKO) 1:100 in TBS-T. Peroxidise reactivity were developed for 12 minutes with Nova Red (Vector laboratories, Burlingame, California, USA). Sections were washed in tap water before counterstained for 15 seconds with haematoxylin, dehydrated and mounted with Pertex.

Statistical analysis
Data were analyzed with an unpaired t-test assuming equal variances. If variances were found to be different between groups, Welsh's correction was applied. Results are represented as mean 6 SEM. p-values less than 0.05 are considered significant and are indicated with *, while p-values of p,0.01 and p,0,001 are marked with ** and ***, respectively.