5-Fluorouracil Induces Enteric Neuron Death and Glial Activation During Intestinal Mucositis via a S100B-RAGE-NFκB-Dependent Pathway

5-Fluorouracil (5-FU) is an anticancer agent whose main side effects include intestinal mucositis associated with intestinal motility alterations maybe due to an effect on the enteric nervous system (ENS), but the underlying mechanism remains unclear. In this report, we used an animal model to investigate the participation of the S100B/RAGE/NFκB pathway in intestinal mucositis and enteric neurotoxicity caused by 5-FU (450 mg/kg, IP, single dose). 5-FU induced intestinal damage observed by shortened villi, loss of crypt architecture and intense inflammatory cell infiltrate as well as increased GFAP and S100B co-expression and decreased HuC/D protein expression in the small intestine. Furthermore, 5-FU increased RAGE and NFκB NLS immunostaining in enteric neurons, associated with a significant increase in the nitrite/nitrate, IL-6 and TNF-α levels, iNOS expression and MDA accumulation in the small intestine. We provide evidence that 5-FU induces reactive gliosis and reduction of enteric neurons in a S100B/RAGE/NFκB-dependent manner, since pentamidine, a S100B inhibitor, prevented 5-FU-induced neuronal loss, enteric glia activation, intestinal inflammation, oxidative stress and histological injury.


5-FU increases S100B protein in GFAP-expressing cells.
We found that 5-FU enhanced (P < 0.01) S100B protein expression in the jejunum, ileum and colon compared to the control group ( Fig. 1A and S1). S100B protein was expressed in GFAP-positive cells (Fig. 1C). The mean fluorescence intensity of GFAP and S100B was significantly (P < 0.05) increased in the 5-FU group compared to the control group (Fig. 1D).
Taken together, these data suggest that 5-FU activates EGCs since GFAP and S100B are glial-cell markers.

5-FU increases NFκB p65 protein expression in the intestine.
Because S100B is involved in NFκB activation by binding to RAGE, as previously shown in the CNS 18 , we assessed whether 5-FU treatment increases NFκB p65 protein expression in the jejunum, ileum and colon. Consistent with previous studies 24 , we found that 5-FU augmented (P < 0.05) NFκB p65 protein expression in the jejunum, ileum and colon compared to the control group (Fig. 1B and S2). and S100B (green) and their co-localization (Merge, yellow). Nuclei were stained with DAPI. Scale bar, 50 µm.
(D) Graph represents the mean ± SEM of the immunofluorescence intensity of GFAP, S100B and their colocalization (GFAP and S100B) in colon tissue from 5 fields per mouse from 4 mice in each group. Intensity of fluorescence were quantified using ImageJ. # P < 0.05 versus control group, Student's t test.
SCIeNtIFIC RepoRts | (2019) 9:665 | DOI: 10.1038/s41598-018-36878-z The S100B inhibitor Pentamidine reduces 5-FU-related loss of body weight and histopathological injury. Because, S100B is involved in neuron loss and glial activation in CNS through its interaction with RAGE and NFκB activation [18][19][20][21][22][23][24][25] , we hypothesized that S100B is involved in 5-FU-induced glial-cell activation, neuron loss and intestinal mucositis. To confirm this hypothesis, we inhibited S100B by intraperitoneal administration of pentamidine, a S100B inhibitor. We first investigated whether pentamidine affected 5-FU-induced weight loss and histological alterations. 5-FU caused a pronounced loss of body weight ( Fig. 2A), which was partially rescued by a low dose of pentamidine (0.8 mg/kg). Following animal euthanasia, we collected intestinal segments for histopathological analysis (Fig. 2B). We found that 5-FU induced intestinal damage in all intestinal segments investigated, and this damage involved shortened villi, loss of crypt architecture, submucosal edema and a pronounced inflammatory cell infiltrate in the lamina propria (Fig. 2C) compared to the control group. We also observed hypertrophy of myenteric ganglions in 5-FU-injected mice (Fig. 2C). The deleterious effects of 5-FU were also visible in the morphometry of the small intestine, which showed a significant reduction in villus height (Fig. 2D, P < 0.01) and crypt depth (Fig. 2E, P < 0.01) in all sections evaluated compared to the control group. Pentamidine treatment significantly prevented intestinal injury ( Fig. 2C; Fig. 2D, P < 0.01; Fig. 2E, P < 0.01). Intestinal damage was also measured by blind semi-quantitative analysis ( Table 1). All intestinal segments showed moderate to severe intestinal damage in 5-FU-administered mice (Table 1), and pentamidine significantly attenuated this injury in the duodenum and jejunum (P < 0.01 vs 5-FU). S100B inhibition reduces 5-FU-induced glial activation, downregulates GFAP and S100B expression in the jejunum and prevents neuron loss. We next evaluated whether pentamidine decreased 5-FU-induced glial activation in the intestine. We found that 4 mg/kg pentamidine reduced (P < 0.01) the percentage of GFAP-immunopositive and S100B-immunopositive area in the small intestine of mice with 5-FU-induced intestinal mucositis compared to the 5-FU group (Fig. 3A). Qualitative analysis showed that 4 mg/kg pentamidine decreased (P < 0.01) GFAP and S100B immunostaining in the mucosal, submucosal and myenteric plexuses of the jejunum compared to the 5-FU group (Fig. 3A). To confirm these results, we measured GFAP and S100B mRNA levels by qPCR. GFAP expression was 19-fold higher in 5-FU-treated mice than in controls (P < 0.01), whereas pentamidine induced a 10.5-fold decrease (P < 0.01) in GFAP mRNA levels in the jejunum compared to the 5-FU group (Fig. 3D). S100B mRNA levels were 6.8-fold higher in 5-FU-treated mice than in the controls (P < 0.01), whereas pentamidine induced a 3.8-fold decrease (P < 0.01) in S100B mRNA levels in the jejunum compared to the 5-FU group (Fig. 3E). In addition, S100B protein expression was 0.9-fold lower in pentamidine-treated mice with 5-FU-induced intestinal mucositis than in the 5-FU group (P < 0.01, Fig. 3F and S3). The treatment with pentamidine alone, caused no changes on S100B expression compared with control group (Fig. S4), Taken together, these results show that S100B inhibition with pentamidine diminish glial activation in 5-FU-induced intestinal mucositis due to decreased levels of S100B and GFAP in the jejunum.
We also found that 5-FU decreased (P < 0.01) the percentage of HuC/D-immunopositive area in the small intestine ( Fig. 3A) compared to control animals. Next, we investigated whether S100B inhibition could reduce 5-FU-induced neuronal loss. Immunostaining results showed that 4 mg/kg pentamidine increased (P < 0.01) the percentage of HuC/D-immunopositive area in the small intestine of 5-FU-treated mice compared to mice treated only with 5-FU (Fig. 3A). Furthermore, 4 mg/kg pentamidine prevented the loss of neurons in both the submucosal and myenteric plexuses in the jejunum (Fig. 3B, right). To confirm these findings, we performed Western Blot analysis for HuC/D on the jejunum. We found that HuC/D protein expression was 1.13-fold lower in 5-FU-treated mice compared to the control group (P < 0.01), whereas HuC/D expression was 1.5-fold higher in pentamidine-treated mice in 5-FU-induced intestinal mucositis than in the 5-FU group (P < 0.01, Fig. 3G).
Due to the ability of the S100B inhibitor to prevent 5-FU-induced decreased expression of the HuC/D protein, we evaluated cell death in the intestine by TUNEL assay. We found that 5-FU induced cell death in the intestinal crypts, smooth muscle layer and myenteric plexus (primarily neurons) (Fig. 3C). Quantitative analysis demonstrated that 5-FU enhanced the number of TUNEL-positive cells in the jejunum compared to the control group (P < 0.01), whereas 4 mg/kg pentamidine reduced the number of TUNEL-positive cells in 5-FU-induced intestinal mucositis compared to the 5-FU group (P < 0.01) (Fig. 3H).

Higher concentration of S100B but not 5-FU, stimulates enteric neuronal cell death in vitro.
Because pentamidine decreased the number of TUNEL-positive cells in 5-FU-induced intestinal mucositis, we investigated whether S100B and 5-FU could directly promote enteric neuronal cell death. According to neuronal morphology features showed in Fig. 4A, we incubated enteric neurons with S100B or 5-FU in two different conditions in vitro (Days in vitro -DIV0 and DIV4). We found that lower concentrations of S100B (0.05 or 0.5 µM) decreased the percentage of TUNEL-positive cells compared to the control group (p < 0.01). However, the treatment with 5-FU did not promote enteric neuronal cell death (Fig. 4B). Given that higher concentrations of S100B have been reported to stimulate neuronal cell death in the CNS, we increased the dosage of S100B in vitro to mimic the effect reported in CNS. We found that higher concentration of S100B (500 µM) increased TUNEL-positive cells compared to the controls (p < 0.01) (Fig. 4C).
Similar results were found when investigated the effect of S100B and 5-FU on enteric neurons death (incubating them on day 4) (Fig. 4D,E). However, 50 µM S100B also stimulated (p < 0.01) enteric neurons death compared with control group (Fig. 4D). S100B inhibition reduces 5-FU-related RAGE and NFκB expression. Because most of the effects of S100B are mediated via RAGE activation, we evaluated the distribution of RAGE within the jejunum and investigated the expression of RAGE by enteric neurons during 5-FU-induced intestinal mucositis. We found that RAGE  To confirm these findings, we next examined RAGE protein expression by Western Blot. RAGE protein expression was 0.25-fold higher in the jejunum of 5-FU-treated mice than in controls (P < 0.01), whereas its expression was 0.24-fold lower in pentamidine-treated mice with 5-FU-induced intestinal mucositis compared to the 5-FU group (P < 0.01, Fig. 5B).
To evaluate whether S100B was required for NFκB activation in enteric neurons during 5-FU-induced intestinal mucositis, we assessed NFκB NLS expression, which is a marker of activated NFκB because the NLS (nuclear localization sequences) appears when NFκB is uncoupled from IκB. NFκB NLS immunostaining was highly increased in 5-FU-treated mice compared to the control group (Fig. 5C). We observed the increase of NFκB NLS immunostaining in enteric neurons of the submucosal and myenteric plexuses, as shown by HuC/D and NFκB NLS co-localization. Furthermore, 5-FU increased NFκB NLS immunostaining in cells from the intestinal epithelium and lamina propria. In contrast, pentamidine decreased NFκB NLS immunostaining in the jejunum of mice with 5-FU-induced intestinal mucositis (Fig. 5C).
We next assessed NFκB p65 protein expression by Western Blot. NFκB p65 protein expression was 4.95-fold higher in 5-FU-treated mice compared to the control group (P < 0.01). Inhibition of S100B with 4 mg/kg pentamidine significantly decreased NFκB p65 protein expression in the same proportion compared to mice treated only with 5-FU (P < 0.01) ( Fig. 5D and S5).

S100B inhibition decreases the expression of pro-inflammatory cytokines and oxidative stress markers.
Inducible nitric-oxide synthase levels were markedly increased in the jejunum of 5-FU-treated mice relative to the control group detected by immunofluorescence (Fig. 6A), qPCR (P < 0.01, Fig. 5B), Western Blot (P < 0.01, Fig. 6C). Griess reaction also detected an increase in nitrite/nitrate levels on this group (P < 0.05, Fig. 6D). We identified an increased iNOS immunostaining in the intestinal epithelium, lamina propria and myenteric plexus (Fig. 6A). Furthermore, pentamidine treatment completely abrogated the expression of this pro-inflammatory enzyme compared to the 5-FU-injected group ( Fig. 6A-D).
Moreover, 5-FU also induced a significant increase in the levels of TNF-α (P < 0.01, Fig. 6E), IL-6 (P < 0.01, Fig. 6F), and MDA (P < 0.01, Fig. 6G) in the jejunum. However, pentamidine significantly decreased the level of these proinflammatory markers compared to the 5-FU-treatment ( Fig. 6E-G). However, the injection of pentamidine alone did not cause any changes on the levels of nitrite/nitrate, TNF-α and MDA, but decreased the levels of IL-6 (p < 0.01) compared to control group (Fig. S6).

Discussion
Here, we provide the first evidence of the deleterious effect of S100B on 5-FU-induced intestinal mucositis pathogenesis. We also showed that S100B is produced by EGCs and may be involved in enteric neuron death and reactive gliosis during mucositis as well as in the histological alterations, animal weight loss, inflammatory response and oxidative stress through a RAGE/NFκB-dependent mechanism. Moreover, we showed that high concentrations of S100B, but not 5-FU, causes enteric neuronal cell death in vitro, suggesting that S100B mediates the effect of 5-FU on neuronal loss during intestinal mucositis.
5-FU-induced intestinal mucositis animal model is widely used since it reproduces findings in the clinical disease, including intestinal dysmotility, loss of body mass and intestinal inflammation [4][5][6][7][8][9]26,27 , as observed in colorectal cancer patients undergoing chemotherapy 28 . In this present study, 5-FU caused body weight loss, shortened villi with vacuolated cells and intense inflammatory cell infiltration along the intestinal wall in mice with mucositis relative to the control group. We also demonstrated that 5-FU upregulates S100B protein expression in GFAP-positive cells during mucositis, suggesting that EGCs are an important source of S100B on this model. In Duodenum the brain, GFAP and S100B upregulation by astrocytes are an indication of reactive gliosis 29 . However, this study is the first to report these findings for 5-FU-induced intestinal mucositis. Previous studies reported that GFAP expression is increased in the inflamed intestinal mucosa of ulcerative colitis patients 11 , although this cell marker is reduced during necrotizing enterocolitis and in the non-inflamed mucosa of Crohn's disease patients 30 . In the CNS, GFAP is the major intermediate filament protein, and it has been found to control cell shape and movement in mature astrocytes 31 . in each group. # P < 0.01 (GFAP) or # P < 0.01 (S100B) versus control group, *P < 0.01 (GFAP) or # P < 0.01 (S100B) versus 5-FU group. (F) S100B and (G) HuC/D protein expression were evaluated by Western Blot.
To test the role of S100B, we treated mice with the S100B inhibitor pentamidine 36 . Interestingly, pentamidine prevented 5-FU-induced neuronal loss, EGCs activation, intestinal inflammation, oxidative stress and histological injury. The protective inhibitory effect of pentamidine was previously reported in animal models of Alzheimer's disease, where it prevented neuroinflammation and RAGE activation 37 . In the gut, pentamidine also protected animals against experimental DSS-induced colitis 38 . A previous study showed that inhibition of the metabolic activity of GFAP-positive cells through a gliotoxin (fluorocitrate) reduced motility and transit in the small intestine, indicating that enteric glia modulates enteric neuronal activity 39,40 .
The effect of S100B inhibition on reducing 5-FU-induced weight loss might be associated with increased villus height and consequently increasing the absorptive area at least in the duodenum (0.8 mg/kg). It is known that pentamidine decreases the appetite when administered in major doses, which could explain the reduced effect We hypothesize that S100B is a key regulatory molecule since its inhibition led to the recovery of intestinal architecture. Previous studies suggest that S100B impairs intestinal epithelial cell proliferation, possibly by binding to and activating RAGE 19 .
Our present findings suggest that 5-FU increases RAGE expression in enteric HuC/D-positive neurons. Several studies have used HuC/D as a pan-neuronal marker of enteric neurons [41][42][43] . We demonstrated that 5-FU decreased HuC/D expression within the gut wall and induced neuronal death. Neuronal loss may be one factor that contributes to 5-FU-induced altered intestinal contractility, which persists even after active inflammation has been resolved 9 . In accordance with these findings, a pilot study from colonic biopsies of patients after 5-FU treatment showed a decrease in enteric neurons of the myenteric plexus, including morphological and electrophysiological changes in these neurons 44 . McQuade et al. 45 demonstrated that 5-FU decreases the number of nitric oxide synthase-immunoreactive neurons during the inflammatory phase. In ulcerative colitis, Crohn's disease and necrotizing enterocolitis, enteric neurons cell death has also been observed 30 and it is likely to contribute to changes in the intestinal motility. We provide evidence that 5-FU-induced neuronal death occurs in a S100B/RAGE/NFκB-dependent manner since pentamidine-treated animals expressed increased HuC/D levels and decreased RAGE and NFκB levels in neurons. In accordance to this hypothesis, we showed that 5-FU was unable to directly cause enteric neuronal death in vitro. In contrast, S100B have a dual effect, since higher S100B concentrations caused a significant neuronal death, while lower S100B concentrations showed a protective action.
Notably, pentamidine treatment reduced iNOS expression, IL-6 and TNF-α levels and MDA accumulation in the small intestine. Cirillo et al. 21 showed that the use of antibodies against S100B or RAGE reduces  Bars represent mean ± SEM of 6 mice in each group. # P < 0.01 versus control group, *P < 0.01 versus 5-FU group. One-way ANOVA followed by Bonferroni. (H) Proposed model of the role of S100B in 5-FU-induced glial cell activation, neuronal loss and intestinal mucositis. 5-FU disrupts the intestinal epithelial barrier leading to EGCs activation by mediators released by macrophages (TNFα, IL-6 and nitric oxide) and bacteria. This process results in increased expression of GFAP and S100B and further release of S100B into the extracellular environment. Additional to that, it also stimulates neuronal death and activates macrophages through the S100B/RAGE/NFκB pathway to release NO, which can also participate in neuronal death.
SCIeNtIFIC RepoRts | (2019) 9:665 | DOI:10.1038/s41598-018-36878-z lipopolysaccharide (LPS)-or INF-γ-induced iNOS expression in primary culture of EGCs. Additionally, S100B stimulated iNOS gene expression, production of nitric oxide and TNF-α protein expression in macrophages isolated from the peritoneal cavity in a concentration-dependent manner 46 . Brown et al. 41 suggested that EGC mediate neuronal cell death in colitis through an oxidative stress reaction. Therefore, inhibition of S100B signaling can improve intestinal mucositis, partially by modulating enteric ganglia and reducing pro-inflammatory mediators and oxidative stress.
In conclusion, this study demonstrates an important role for enteric glia in 5-FU-induced intestinal mucositis pathogenesis. EGCs release S100B, which in turn activates NFκB signaling in a RAGE-dependent manner in neurons, leading to neuronal cell death. An autocrine pathway might also induce glial cell expression of pro-inflammatory cytokines, iNOS-derived nitric oxide release and oxidative stress (Fig. 5H). Taken together, these results show that the S100B/RAGE/NFκB pathway may be critical in the pathogenesis of 5-FU-induced intestinal mucositis. Therefore, the S100B signaling pathway appears to be a promising pharmacological target to prevent 5-FU-induced intestinal mucositis and ENS injury.

Methods
Seventy-six male Swiss mice, weighing 25-30 g (6-8 weeks of age), were housed in temperature-controlled rooms under 12 h light-dark cycles. The animals received water and food ad libitum. Surgical procedures and animal treatments were conducted in accordance with the Guidelines for Institutional and Animal Care and Use of the Federal University of Ceará, Brazil. All procedures involving animals were approved by the Federal University of Ceará Committee on the Ethical Treatment of Research Animals (Protocol No. 80/2015).
After deparaffinization, antigens were recovered by incubating the slides in citrate buffer (pH 6.0) for 20 min at 95 °C. Endogenous peroxidase was blocked with 3% H 2 O 2 for 10 min to reduce nonspecific binding. Sections were then incubated with GFAP (IS524, DAKO), S100B (IR504, DAKO) or HuC/D (A21271, Invitrogen, 1 µg/ mL) antibodies for 1 h. Sections were then incubated for 30 min with polymer (K4061, DAKO). Antibody binding sites were visualized by incubating the samples with diaminobenzidine-H 2 O 2 (DAB, DAKO) solution. Sections incubated with antibody diluent without a primary antibody were used as negative controls. Antibody specificity was evaluated using positive controls for GFAP, S100B and HuC/D in the mouse cerebral cortex or for RAGE in the rat liver (data not shown). The amounts of DAB products from immunostaining were estimated from digital images of at least ten different areas of each section (from 4 specimens per group) at 400x (GFAP and S100B) and 1000 × (HuC/D) magnification using Adobe Photoshop software. The percentage of immunopositive area was calculated by dividing the DAB-positive staining (immunostaining-positive pixels) by the number of pixels per total tissue image multiplied by 100, as described previously 49 .

Terminal Deoxynucleotidyl Transferase-Mediated Deoxyuridine Triphosphate Nick-End
Labeling (TUNEL) Assay. Cell death was investigated using a TdT-mediated dUTP nick end-labeling (TUNEL) assay according to the manufacturer's protocol (ApopTag, S7101, Merck, Millipore). Four micron-thick sections were prepared from paraffin-embedded jejunum segments. After deparaffinization and rehydration, antigens were recovered with 20 μg/mL proteinase K for 15 min at room temperature. Endogenous peroxidase was blocked with 3% H 2 O 2 for 10 min to reduce nonspecific binding. After washing, sections were incubated in a humidified chamber at 37 °C for 1 h with TdT buffer containing TdT enzyme and reaction buffer. Specimens were incubated for 10 min at room temperature with a stop/wash buffer and then incubated in a humidified chamber for 30 min with anti-digoxigenin peroxidase conjugate at room temperature. After a series of PBS washes, slides were covered with peroxidase substrate to develop color and then washed in three changes of dH 2 O and counterstained in 0.5% methyl green for 10 min at room temperature. TUNEL-positive cells were counted from digital images of at least ten different areas of each section (from 4 specimens per group) at 1000x using ImageJ software (NIH, Bethesda, MD, USA).

RNA extraction and reverse transcription.
Total RNA was isolated using an RNA isolation protocol (PROMEGA). RNA was quantified by NanoDrop (Thermo Fisher Scientific), and RNA quality was determined by examining the 260/280 ratio >1.8. A total of 1 µg RNA was then reverse transcribed using a High-capacity cDNA reverse transcription Kit (Applied Biosystems) according to the manufacturer's protocol.
TaqMan qPCR. mRNA expression was analyzed by TaqMan quantitative PCR (qPCR) according to the manufacturer's instructions using pre-made probes (Applied Biosystems, USA). Sequences or pre-made probe ID numbers are listed in Table 2. To compare gene expression between conditions, the expression (normalized by GAPDH endogenous control) was quantified relative to the control condition. A total of 40 ng cDNA (diluted 5x) was added to the 20x TaqMan gene expression assay (1 µL of S100B, GFAP, iNOS or Gapdh), 2x TaqMan gene expression master mix (10 µL) and RNA-free water (5 µL) to a final volume of 20 µL. Then, 20 µL of PCR reaction mix was transferred to each well. The plate was loaded into the instrument, sealed and centrifuged. The thermocycler parameters were 50 °C for 2 min and 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. All fold changes were calculated by the ΔΔC t method.

ID assay Mm99999915_g1
TaqMan probe GGTGTGAACGGATTTGGCCGTATTG Amplicon length: 109 Jejunal nitrite level detection. Jejunal segments were removed for nitrite/nitrate (NO X ) quantification using the Griess method 50 . First, nitrate was reduced to nitrite by incubation with nitrate reductase and nicotinamide adenine dinucleotide overnight at room temperature. Griess reagent (1% sulfanilamide and 0.1% naphthyl ethylenediamine dihydrochloride in 5% phosphoric acid) was then added and the total nitrite concentration was determined by its 540 nm absorbance value, assigned to the Griess reaction. A calibration curve was obtained by incubating sodium nitrite (10 to 200 µmol/L) with Griess reagent. Nitrate/nitrite levels were expressed as micromoles per milligram of intestinal tissue.
Jejunal TNF-α and IL-6 levels. Jejunal samples were collected and homogenized in 1% PBS to evaluate TNF-α and IL-6 levels by using a commercial ELISA kit (R&D Systems) according to the manufacturer's instructions. TNF-α and IL-6 levels in the jejunum were measured as picograms per milliliter of tissue.

MDA.
Lipid peroxidation was measured as malondialdehyde (MDA) production through the thiobarbituric acid-reactive substances (TBARS) 51 in jejunal segments of mice. Briefly, 250 µL of 10% homogenate of intestinal tissue was mixed with 1.5 mL of 1% H 3 PO 4 and 0.5 mL 0.6% thiobarbituric acid aqueous solution, and the mixture was stirred and heated in boiling water for 45 min. After cooling, 2 mL of n-butanol was added, and the mixture was homogenized. The butanol layer was separated, and the difference between the optical densities at 535 and 520 nm was used to calculate MDA concentrations, which were expressed as micromoles of MDA per milligram intestinal tissue.
TUNEL assay for in-vitro experiments. Cell death was analyzed by TUNEL assay following the protocol from the manufacturer (ApopTag, S7101, Merck-Millipore or In situ cell death detection kit, Fluoroscein, ROCHE). Apoptotic cells were counted from Confocal (Leica SP5) of at least three-eight distinct fields of each well per group from 2-5 different experiments. Cells were counted using ImageJ software (NIH, Bethesda, MD, USA).

Statistical analysis.
Data are presented as the mean ± standard error of the mean (SEM) or as medians when appropriate. Student's t test, one or two-way Analysis of Variance (ANOVA) followed by the Bonferroni test was used to compare means, and the Kruskal-Wallis and Dunn tests were used to compare medians. P < 0.05 was considered significant 43 .