Abstract
Hypoxia–ischemia is a common cause of neurological impairments in newborns, but little is known about how neuroinflammation contributes to the long-term outcome after a perinatal brain injury. In this study, we investigated the role of the fractalkine receptor chemokine CX3C motif receptor 1 (CX3CR1) and of toll-like receptor (TLR) signaling after a neonatal hypoxic–ischemic brain injury. Mice deficient in the TLR adaptor proteins Toll/interleukin-1 receptor-domain-containing adaptor protein inducing interferon β (TRIF) or myeloid differentiation factor-88 (MyD88) and CX3CR1 knock-out (KO) mice were subjected to hypoxia–ischemia at postnatal day 3. In situ hybridization was used to evaluate the expression of TLRs during brain development and after hypoxic–ischemic insults. Behavioral deficits, hippocampal damage, reactive microgliosis, and subplate injury were compared among the groups. Although MyD88 KO mice exhibited no differences from wild-type animals in long-term structural and functional outcomes, TRIF KO mice presented a worse outcome, as evidenced by increased hippocampal CA3 atrophy in males and by the development of learning and motor deficits in females. CX3CR1-deficient female mice showed a marked increase in brain damage and long-lasting learning deficits, whereas CX3CR1 KO male animals did not exhibit more brain injury than wild-type mice. These data reveal a novel, gender-specific protective role of TRIF and CX3CR1 signaling in a mouse model of neonatal hypoxic–ischemic brain injury. These findings suggest that future studies seeking immunomodulatory therapies for preterm infants should consider gender as a critical variable and should be cautious not to abrogate the protective role of neuroinflammation.
Introduction
Neonatal hypoxic–ischemic encephalopathy is a major cause of long-term neurodevelopmental disability in term infants (Volpe, 2012). Despite the difficulties of adapting the diagnosis criteria of perinatal asphyxia to preterm newborns, it is accepted that neonatal hypoxic–ischemic encephalopathy can also occur in this population, leading to significant mortality and neurologic morbidity (Logitharajah et al., 2009; Chalak et al., 2012). Hypoxia–ischemia is also one of the main upstream mechanisms associated with the encephalopathy of prematurity, characterized by a combination of cerebral gray and white matter damage in preterm infants (Kinney, 2009).
Recent evidence has shown that systemic infection/inflammation can contribute to the pathogenesis of the encephalopathy of prematurity by disrupting white matter and cortical development (Dean et al., 2011; Favrais et al., 2011), as well as by exacerbating hypoxic–ischemic neuronal injury (Lehnardt et al., 2003, Wang et al., 2009). However, little is known about the role of the neuroinflammatory response after a sterile hypoxic–ischemic brain injury in the immature brain (Vexler and Yenari, 2009). Therefore, it is still necessary to identify the role of key signaling pathways involved in the regulation of this response, such as the activation of toll-like receptors (TLRs) by damage-associated molecular patterns (DAMPs) and the engagement of chemokine receptors by their ligands.
TLRs are transmembrane receptors that recognize pathogen-associated molecular patterns and DAMPs. Up to now, five different adaptor proteins have been implicated in the signaling of TLRs: myeloid differentiation factor-88 (MyD88), MyD88-adaptor-like (MAL/TIRAP), Toll/interleukin-1 receptor (TIR)-domain-containing adaptor protein inducing interferon β (TRIF), TRIF-related adaptor molecule (TRAM) and sterile α- and HEAT/armadillo-motif-containing protein (SARM). Each adaptor is used by specific TLR complexes, leading to the activation of transcription factors that control the inflammatory response. For example, although MyD88 is involved in the signaling of all TLRs (except TLR3), TRIF is used by TLR4 (through TRAM) and TLR3. The only exception is SARM, which negatively regulates TRIF signaling (O'Neill and Bowie, 2007; Mallard, 2012).
In this study, we investigated the mRNA expression of several TLRs during telencephalic development and after cerebral hypoxia–ischemia in postnatal day 3 (P3) mice, a model of perinatal brain injury in very preterm infants (Vannucci et al., 1999; Sizonenko et al., 2003). We have also used this model to assess the functional role of MyD88 and TRIF signaling in the development of hippocampal atrophy and long-term cognitive/motor deficits. Finally, we investigated the role of the chemokine CX3C motif receptor 1 (CX3CR1) signaling, an important regulator of microglial activity (Chapman et al., 2000; Cardona et al., 2006), in the progression of learning deficits, subplate neuronal injury, and hippocampal damage in this model.
Our results reveal a novel, gender-specific protective role of TRIF and CX3CR1 signaling, suggesting that the disruption of protective mechanisms of the neuroinflammatory response could aggravate a hypoxic–ischemic brain injury in preterm infants, especially in females.
Materials and Methods
Animals.
CX3CR1−/− (B6.129P-Cx3cr1tm1Litt/J) mice were purchased from The Jackson Laboratory. MyD88−/− mice were kindly provided by S. Akira (Osaka University, Osaka, Japan). TRIF−/− mice were kindly provided by Dr. T.J. Lin (Dalhousie University, Halifax, Canada). All mouse strains were maintained in a C57BL/6J background. Mice were housed two per cage and acclimated to standard laboratory conditions (12 h light/dark cycle) with ad libitum access to mouse chow and water. All protocols were conducted according to the Canadian Council on Animal Care guidelines as administered by the Laval University Animal Welfare Committee. Mice of both genders were used in the experiments. The number of animals used for each experiment is shown in Table 1 and Table 2.
Neonatal hypoxia–ischemia and sham surgery.
The Rice-Vannucci model of neonatal hypoxia–ischemia was used to induce brain damage in P3 mice (P0 = day of birth; Vannucci et al., 1999). Mice were subjected to either sham surgery or to right common carotid artery occlusion under isoflurane anesthesia (2.5–3% for induction and 1.5–2% for maintenance). Briefly, a midline cervical incision was made and the right common carotid artery was exposed and double ligated with 6-0 silk suture thread. The sham surgery consisted of a midline cervical incision followed by the exposition of the right common carotid artery. After a 2 h period in which the pups were allowed to recover with the mother, the occluded mice were subjected to a hypoxic episode (8% O2 balanced with 92% N2) for 40 min. After that, the animals were placed on a temperature-controlled blanket for 20 min and then returned to their dams.
Tissue collection.
Animals were deeply anesthetized by hypothermia (for P3-P6 mice pups) or via an intraperitoneal injection of a mixture of ketamine hydrochloride and xylazine (100/10 mg/kg; for older animals) and then transcardially perfused with ice-cold 0.9% saline, followed by 4% paraformaldehyde (PFA) in a PBS, pH 7.4. Brains were rapidly removed from skulls, postfixed in PFA for 1 d at 4°C, and cryoprotected in a PFA solution containing 20% (w/v) sucrose overnight. The frozen brains were then sectioned into 30-μm-thick coronal sections using a microtome. Slices were collected in a cold cryoprotectant solution (0.05 m sodium phosphate buffer, pH 7.3, 30% ethylene glycol, 20% glycerol) and stored at −20°C. The brains used for qRT-PCR experiments were collected after transcardial perfusion with ice-cold sterile, 0.9% saline. The cerebellum, brainstem, and olfactory bulbs were removed and the two hemispheres were separately stored at −80°C.
qRT-PCR.
Tissue samples were homogenized and extracted in TRIzol reagent (Sigma-Aldrich). RNA concentration was measured using a NanoDrop spectrophotometer (Thermo Scientific). First-strand cDNA synthesis was accomplished using 6–10 μg of isolated RNA in a reaction containing 200 U of Superscript II, 11 μg of oligo-dT12-19, 50 mm Tris-HCl, pH 8.3, 75 mm KCl, 3 mm MgCl2, 500 μm deoxynucleotides triphosphate, 2 U RNAsine ribonuclease inhibitor, and 10 mm dithiothreitol (Invitrogen) in a final volume of 21 μl. The reaction was performed at 42°C for 50 min, followed by 70°C for 15 min, and then treated with 0.9 U of E. coli RNaseH (Invitrogen) for 20 min at 37°C. The resulting products were purified with the GenElute PCR Clean-Up Kit (Sigma-Aldrich).
Oligoprimer pairs were designed by GeneTool 2.0 software (Biotools) and their specificity was verified by blast in the GenBank database. The synthesis was performed by Integrated DNA Technology (Table 3). A quantity corresponding to 20 ng of of cDNA was used to perform fluorescent-based RT-PCR quantification using the LightCycler 480 (Roche Diagnostics). Reagent LightCycler 480 SYBRGreen I Master (Roche Diagnostics) was used as described by the manufacturer. The conditions for PCRs were as follows: 45 cycles, denaturation at 95°C for 10 s, annealing at 60°C for 10 s, elongation at 72°C for 14 s, and then 74°C for 5 s (reading). A melting curve was performed to assess nonspecific signal. Calculation of the number of copies of each mRNA was performed according to Luu-The et al. (2005) using the second derivative method and a standard curve of Cp versus logarithm of the quantity. The standard curve was established using known amounts of purified PCR products (10, 102, 103, 104, 105, and 106 copies) and a LightCycler 480 v1.5 program provided by the manufacturer (Roche Diagnostics). PCR amplification efficiency was verified. Normalization was performed using reference genes shown to have stable expression levels from embryonic life through adulthood in various tissues (Warrington et al., 2000): hypoxanthine guanine phosphoribosyl transferase 1 (HPRT1) and 18S ribosomal RNA (18S). qRT-PCR measurements were performed by the Centre Hospitalier Universitaire de Québec Research Center Gene Expression Platform (Québec, Canada).
Fluoro-Jade B staining.
Every eighth section of brain slices, starting from the end of the olfactory bulb to the end of the cerebral cortex, was mounted on Colorfrost/Plus microscope slides (Fisher Scientific). The Fluoro-Jade B staining procedure was used to reveal neuronal death as described previously (Pimentel-Coelho et al., 2013). Briefly, dried mounted brain sections were dehydrated and rehydrated through graded concentrations of alcohol (50, 70, 100, 70, and 50% EtOH, 1 min each) and rinsed for 1 min in distilled water. Slides were then treated with potassium permanganate 0.06% for 10 min and rinsed for 1 min in distilled water, followed by incubation for 20 min in a solution containing 0.004% Fluoro-Jade B (Histochem), 0.1% acetic acid, and 0.0002% 4,6-diamidino-2-phenylindole (DAPI; Invitrogen). Slides were thereafter rinsed in distilled water (3 × 1 min), dried overnight at 37°C, dipped in xylene (3 × 2 min), and then coverslipped with distrene plasticizer xylene (DPX) mounting medium (Electron Microscopy Sciences).
Immunohistochemistry.
Free-floating sections were washed with potassium PBS (KPBS; 3 × 10 min) and then incubated for 30 min in a permeabilization/blocking solution containing 4% goat serum, 1% bovine serum albumin, and 0.4% Triton X-100 (Sigma-Aldrich) in KPBS. Sections were incubated overnight with the primary antibody (diluted in the same permeabilization/blocking solution) at 4°C. The following primary antibodies were used: mouse anti-NeuN monoclonal antibody (EMD; Millipore), mouse anti-glial fibrillary acidic protein (GFAP) monoclonal antibody (EMD; Millipore), rabbit anti-high-mobility group protein B1 (HMGB1) polyclonal antibody (Abcam), rat anti-CD68 monoclonal antibody (AbD Serotec), and rabbit anti-ionized calcium binding adaptor molecule 1 (Iba1) polyclonal antibody (Wako Chemicals). The sections were then rinsed in KPBS (3 × 10 min), followed by a 90 min incubation with one of the following secondary antibodies: Cy3-conjugated goat anti-rabbit IgG (H+L) antibody (Jackson Immunoresearch Laboratories), Cy3-conjugated goat anti-rat IgG (H+L) antibody (Jackson Immunoresearch Laboratories), or Alexa Fluor 488-conjugated goat anti-rabbit IgG (H+L) antibody (Invitrogen). Sections were then rinsed in KPBS (2 × 10 min), stained with 0.0002% DAPI for 10 min, rinsed again in KPBS (2 × 10 min), mounted on Colorfrost/Plus slides (Fisher Scientific), and coverslipped with antifade medium composed of 96 mm Tris-HCl, pH 8.0, 24% glycerol, 9.6% polyvinyl alcohol, and 2.5% diazabicyclooctane (Sigma-Aldrich). Alternatively, for NeuN, GFAP, and Iba1 immunohistochemistry, a biotinylated goat anti-mouse IgG (H+L) secondary antibody (Vector Laboratories) and a biotinylated goat anti-rabbit IgG (H+L) secondary antibody (Vector Laboratories) were used. Binding was visualized using the peroxidase-based Vectastain ABC kit (Vector Laboratories) and 3,3′-diaminobenzidine. Tissues were thereafter counterstained with thionin (0.25%), dehydrated through graded concentrations of alcohol, cleared in xylene, and coverslipped with DPX mounting medium (Electron Microscopy Sciences). Bright-field and epifluorescence images were taken using a Nikon C80i microscope equipped with both a motorized stage (Ludl) and a Microfire CCD color camera (Optronics). Confocal laser scanning microscopy was performed with a BX-61 microscope equipped with the Fluoview SV500 imaging software 4.3 (Olympus).
cRNA probes and in situ hybridization.
The expression of TLRs, fractalkine, and connective tissue growth factor (CTGF) were analyzed using cRNA probes against their respective mRNAs (Table 4). Plasmids were linearized and the sense and antisense 35S-labeled riboprobes were synthesized with the appropriate RNA polymerase, as described previously (Laflamme et al., 2001; for details, see Table 4). Standard in situ hybridization was performed on every eighth coronal section, starting from the end of the olfactory bulb to the end of the cerebral cortex, using 35S-labeled cRNA probes, as described previously (Laflamme et al., 2001). The sections were then exposed at 4°C to x-ray films (Biomax; Kodak) for 1–2 d. The slides were thereafter defatted in xylene, dipped into NTB-2 nuclear emulsion (diluted 1:1 with distilled water; Kodak), and exposed for 14–21 d. The slides were then developed in D19 developer (Kodak) for 3.5 min at 14–15°C, washed for 15 s in water, and fixed in rapid fixer (Kodak) for 5 min. Tissues were thereafter rinsed in running distilled water for 1 h, counterstained with thionin (0.25%), dehydrated through graded concentrations of alcohol, cleared in xylene, and coverslipped with DPX mounting medium (Electron Microscopy Sciences). All images were captured using a Nikon Eclipse 80i microscope equipped with a digital camera (QImaging) and processed to enhance contrast and brightness using Adobe Photoshop CS6 (Version 13.0; Adobe Systems).
Cell quantification and analysis of brain damage.
For the quantification of microglial cells, 30-μm-thick coronal brain sections were immunostained for Iba1 or CD68 and nuclear staining was obtained with DAPI. Cell counts were performed by a blinded experimenter on two sections per animal. Briefly, the contours of the whole hippocampal regions CA1/CA2 and CA3 were traced as virtual overlay on real-time images (1600 × 1200 pixels) obtained using a Nikon C80i microscope equipped with both a motorized stage (Ludl) and a Microfire CCD color camera (Optronics). Such an apparatus was operated using Stereo Investigator software (Version 9.10.6) designed by Microbrightfield Bioscience. The number of microglial cells (3 d after the injury) was determined using the exhaustive count method (for the whole hippocampal CA1/CA2 and CA3 areas in each section). Areas of the analyzed hippocampus were also calculated. Similarly, the number of degenerating neurons stained with Fluoro-Jade B (3 d after the injury) or the area occupied by Fluoro-Jade B staining (24 h after the injury) in the hippocampal CA1/2 and CA3 regions were quantified using the exhaustive count method with the aid of this apparatus. Hippocampal damage was determined by measuring the hippocampal area (CA1/CA2 and CA3 regions) in both hemispheres 14 weeks after the surgery using the same apparatus. The ratio of ipsilateral to contralateral NeuN-positive hippocampal areas was calculated. Subplate neuronal damage was evaluated using the following damage scoring: 0 = no apparent damage; 1 = lack of CTGF mRNA expression in a single area of the subplate; 2 = lack of CTGF mRNA expression in multiple, noncoalescent areas of the subplate; and 3 = lack of CTGF mRNA expression in multiple, coalescent areas of the subplate. The area of fractalkine mRNA expression in the hippocampus was quantified in dark-field photomicrographs of in situ hybridization stained sections using Stereo Investigator software (Version 9.10.6; Microbrightfield Bioscience). All quantifications were performed by a blinded observer.
Behavioral analysis.
Behavioral tests were performed during the “lights-off” phase of the day, when the animals are more active, as described by Guo et al. (2011). The behavioral experimenter was blinded to the genetic and treatment status of the animals.
T-water maze.
The T-water maze task was used to assess hippocampal-dependent spatial learning and memory 12 weeks after the surgery, as described previously (Pimentel-Coelho et al., 2013). In this paradigm, the mouse's ability to remember the spatial location of a submerged platform was evaluated. The T-maze apparatus (length of stem, 64 cm; length of arms, 30 cm; width, 12 cm; height of walls, 16 cm) was made of clear fiberglass and filled with water (23 ± 1°C) at a height of 12 cm. An escape platform (11 × 11 cm) was placed at the end of the target arm and was submerged 1 cm below the surface. The position of the platform was chosen randomly for each animal before testing. In the learning phase, which allows the evaluation of left–right spatial learning, the mice were placed in the stem of the T-maze and swam freely until they found the submerged platform (located either in the right or in the left arm of the T-maze apparatus) and escaped to it. If the animals did not find the platform within 60 s, they were gently guided onto it. After reaching the platform, the mice remained on it for 20 s. During this learning paradigm, the escape latency (time to reach the platform) was recorded in each trial for a total of eight trials.
Rotarod.
The rotarod performance test was performed 5 weeks after the surgery. The animals were placed in a neutral position and the rod was set to accelerate at 0.2 rpm/s. Mice were subjected to three trials per session, with an interval of 10 min between each trial, and the time spent on the rotarod (i.e., the latency to fall) was recorded automatically in each trial. The best trial (i.e., the longest time spent on the rotarod) was chosen for each animal.
Statistical analysis.
Behavioral tests were analyzed using two-way ANOVA followed by Bonferroni's post hoc test. Subplate damage score, reactive microgliosis, hippocampal damage, and qRT-PCR results were analyzed using Kruskal–Wallis test followed by Dunn's test. The Mann–Whitney U test was used for the Fluoro-Jade B analysis and for the analysis of qRT-PCR results (when only two experimental groups were analyzed). A p value <0.05 was considered statistically significant. All analyses were performed using GraphPad Prism version 5.01 software for Windows.
Results
TLR mRNA expression is developmentally regulated in the mouse telencephalon
Recent studies using qRT-PCR and immunoblots have observed that TLR1–9 mRNA (Kaul et al., 2012) and TLR2–3 protein levels (Lathia et al., 2008) change during mouse brain development. TLR2 immunostaining has been observed in the white matter and in the hypothalamic paraventricular nucleus of P10 mice (Stridh et al., 2011) and TLR7, TLR8, and TLR9 mRNA and protein expression were demonstrated in the developing neocortex (Ma et al., 2006; Kaul et al., 2012). To further elucidate the spatial and temporal pattern of TLR mRNA expression during telencephalic development, we performed in situ hybridization using specific probes for TLR2, TLR3, TLR6, TLR7, and TLR9 in coronal sections of embryonic and early postnatal mouse telencephalon.
We observed that TLR2 expression could be observed in the choroid plexus and in the subcallosal region at the midline at E18 (Fig. 1A). At P7, there was an increase in the expression of TLR2, particularly in the white matter (below the apex of the cingulum; Fig. 1B), in the region immediately below the external capsule (Fig. 1B), and in the callosal/subcallosal regions ranging from the dorsolateral aspect of the lateral ventricles to the midline (Fig. 1C). By the end of the second postnatal week (P14), TLR2 expression was restricted to the choroid plexus (Fig. 1D). Indeed, at all the studied ages, TLR2 expression could be detected in the choroid plexus, which is one of the few regions expressing the mRNA of this receptor in the adult brain (Laflamme et al., 2001), indicating that this region may act as an immune sentinel in the brain throughout life (Nguyen et al., 2002). TLR3 presented a different pattern of expression, being distributed throughout the cortex and hippocampus from E18 to P22 (Fig. 1E–H). At E14, a few scattered cells expressed TLR7 in the cortical plate and there was a strong expression of this receptor mRNA in the hippocampus (Fig. 1M). TLR7 expression could also be observed in scattered cells in the cerebral cortex and hippocampus during the first postnatal week (Fig. 1O,P). We also observed that TLR6, TLR7, and TLR9 mRNA were mainly detected in the subcallosal region at midline at E18 (Fig. 1I,N,R). During early postnatal life, the expression of TLR6, TLR7, and TLR9 resembled TLR2 distribution at P3 (Fig. 1J,O,S) and P7 (Fig. 1K,P). Such common pattern suggests that complex functional interactions could occur, given that TLR2 heterodimerization with TLR6 extends the spectrum of ligand recognition (Farhat et al., 2008) and that TLR9 can inhibits TLR7 signaling (Wang et al., 2006). Nevertheless, although TLR6 and 7 expression decreased later on, not being detected at P22 (Fig. 1L,Q), TLR9 expression could be observed throughout the cerebral cortex and hippocampus at P14 (Fig. 1T).
We also used qRT-PCR to measure the expression of TLR2, TLR3, TLR6, TLR7, and TLR9 mRNA during the first postnatal week in male and female mice. No significant differences were observed in the expression of such TLR mRNA between males and females at P3 and P7 (Fig. 2A). These experiments also revealed that TLR3 and TLR7 mRNA expression increased in the brains of male mice from P3 to P7 (Fig. 2A; p < 0.05, Kruskal–Wallis test with Dunn's multiple-comparisons test). A similar trend was also observed in the brains of female mice (Fig. 2A), although it did not reach statistical significance.
Unfortunately, our experience and those from others indicate that most of the commercially available antibodies against TLRs are nonspecific (Mallard, 2012) and, for this reason, we were not able to precisely identify the phenotype of the TLR-expressing cells. However, we observed that the common pattern of TLR2, TLR6, TLR7, and TLR9 mRNA expression at E18, P3, and P7 was mainly localized in regions with a transient high density of activated microglial cells. At E18, the expression of TLR2, TLR6, TLR7, and TLR9 was mainly observed below the corpus callosum (Fig. 1A,I,N,R), where we observed a chain of Iba1+ microglia between the lateral ventricles (Fig. 2C), as well as a population of GFAP+ astrocytes (Fig. 2B). These microglial cells exhibited an amoeboid morphology and expressed the activation marker CD68 (Fig. 2D–F). Similarly, TLR2, TLR6, TLR7, and TLR9 mRNA were present in the dorsal periventricular region at P7 (Fig. 1C,K,P and data not shown, respectively), where we found a high density of astrocytes and microglia at this age (Fig. 2G,H, respectively). These microglial cells exhibited an activated phenotype, as indicated by their morphology and by the strong coexpression of CD68 (Fig. 2I–K). Interestingly, it has been shown recently that these microglia express the growth factors macrophage colony-stimulating factor and insulin-like growth factor 1, contributing to the survival of layer V subcerebral and callosal projection neurons during early postnatal development (Hristova et al., 2010; Ueno et al., 2013). Although the population of astrocytes was still present in this periventricular region at P22 (Fig. 2L), the high density of round-shaped activated microglia was not observed at this age (Fig. 2M), coinciding with the absence of TLR2, TLR6, TLR7, and TLR9 mRNA expression in this region at P14 and P22 (data not shown). Indeed, periventricular microglial cells had a ramified morphology and expressed lower levels of CD68 at P14 (N-P). Similar observations could be made in the white matter below the apex of the cingulum, where there was a strong expression of TLR2, TLR6, TLR7, and TLR9 mRNA during the first postnatal week (Fig. 1B,J,O,S). At P7, this region was populated by astrocytes (Fig. 2Q) and activated microglia (Fig. 2R–U). Whereas astrocytes were still observed at P22 (Fig. 2V), the population of microglia acquired a ramified morphology at P14 and P22 (Fig. 2W–Z), when the expression of TLR2, TLR6, and TLR7 mRNA could no longer be visualized in this particular region (Fig. 1D,L,Q). These results suggest that the spatiotemporal pattern of microglial activation/deactivation coincides with the pattern of TLR2, TLR6, TLR7, and TLR9 mRNA expression in the early postnatal brain.
Neonatal hypoxic–ischemic brain injury in P3 mice
Our previous results indicated that the expression of TLR2, TLR3, TLR6, and TLR7 in the mouse brain peaked during the first postnatal week, which corresponds to the brains of very preterm infants, who are particularly vulnerable to hypoxic–ischemic insults during the perinatal period (Volpe et al., 2011). For this reason, we decided to investigate the role of TLR signaling in the setting of an acute sterile brain injury in P3 mice using the Rice-Vannucci model of cerebral hypoxia–ischemia (Vannucci et al., 1999).
Fluoro-Jade B staining revealed the pattern of neurodegeneration in this model, showing clusters of degenerating neurons in the ipsilateral hippocampus (Fig. 3A,B) and cerebral cortex (Fig. 3A,D) 24 h after the insult. Degenerating neurons were observed in the CA1/CA2 and CA3 hippocampal regions of the ipsilateral hemisphere (Fig. 2B), but not in the contralateral hemisphere (Fig. 3C).
Given that male sex is associated with an increased risk of developing cognitive impairments (Helderman et al., 2012) and cerebral palsy (Beaino et al., 2010) in extremely low gestational age newborns and very preterm infants, respectively, we compared the extent of neuronal death between male and female mice subjected to neonatal hypoxia–ischemia. Although there was no difference in the area occupied by Fluoro-Jade B staining in the hippocampal CA1/2 and CA3 regions 24 h after the injury (data not shown), significantly higher numbers of degenerating neurons were found in the hippocampal CA3 region of male mice (Fig. 3E) compared with female mice (Fig. 3F,G; p < 0.05; Mann–Whitney U test) 3 d after the insult. A similar trend was observed in the CA1/2 region, although this did not reach statistical significance (Fig. 3G). Next, we performed in situ hybridization to qualitatively evaluate the expression of fractalkine mRNA after the insult. This chemokine is constitutively expressed by neurons (Harrison et al., 1998), regulating microglia function through the interaction with the CX3CR1 receptor. We observed that the expression of fractalkine mRNA was drastically decreased in the damaged regions of the ipsilateral hippocampus compared with the contralateral hippocampus 24 h (Fig. 3I,J) and 3 d (Fig. 3K,L) after the injury. Quantification of the area of fractalkine mRNA expression in the hippocampal CA1/2 and CA3 regions revealed a similar degree of decrease in male and female mice 24 h (data not shown) and 3 d (Fig. 3H) after the injury. Interestingly, a lack of fractalkine mRNA expression could be detected up to 5 weeks after injury in the damaged regions of the ipsilateral hippocampus (Fig. 3M,N). These regions, particularly the CA3 area, exhibited a decreased density of thionin-stained cells, which were replaced by a glial scar composed mainly of GFAP+ astrocytes (Fig. 3O), but not of Iba1+ microglia (Fig. 3P).
TLR mRNA expression is induced after neonatal hypoxia–ischemia
We also investigated whether the induction of the expression of TLR2, TLR3, TLR6, and TLR7 mRNA occurred after the hypoxic–ischemic insult. qRT-PCR revealed the upregulation of TLR2, TLR3, TLR6, TLR7, and TLR9 mRNA expression (normalized to Hprt1 expression) in the ipsilateral hemisphere (Fig. 3Q), at 3 d after injury. No significant differences were observed in the expression of such TLR mRNA between males and females at this time point (Fig. 3Q).
We also performed in situ hybridization to evaluate qualitatively the spatial distribution of TLR2, TLR3, TLR6, TLR7, and TLR9 mRNA after neonatal hypoxia–ischemia. We observed an increased expression of TLR2 mRNA in the ipsilateral hippocampus (Fig. 4A) as early as 8 h after the injury. The upregulation of TLR2 mRNA in the ipsilateral cerebral cortex and hippocampus became more evident at 24 h (Fig. 4B,C), persisting for at least 3 d after the injury (Fig. 4D,E). TLR2 mRNA expression could not be detected at 7 d after the injury despite the hippocampal atrophy observed in most of the animals (Fig. 4F), indicating that the expression of this molecule is tightly controlled after neonatal hypoxia–ischemia. However, in very rare cases, when a porencephalic cyst was formed in the cerebral cortex of severely affected animals, TLR2 expression persisted around the walls of the cyst for up to 7 d (Fig. 4G). We could not detect major changes in TLR3 expression after the injury (data not shown), probably due to the strong basal expression of this TLR during the first postnatal week. Conversely, there was a strong upregulation of TLR6 mRNA in the ipsilateral cerebral cortex and hippocampus, particularly at 3 d after the injury (Fig. 4H–K). A marked increase of TLR7 expression was also noticed in the damaged hippocampus and cerebral cortex at 24 h (Fig. 4L–O) and 3 d (Fig. 4P,Q) after the injury compared with the contralateral hemisphere. Similarly, TLR9 expression increased in the ipsilateral cerebral cortex and hippocampus, especially at 3 d after the insult (Fig. 4R–W). These results indicate an induction of TLR2, TLR6, TLR7, and TLR9 mRNA expression in the acute/subacute phase of the hypoxic–ischemic insult. Finally, we used immunohistochemistry to evaluate the expression of the endogenous TLR ligand HMGB1, a key mediator linking neuronal death and neuroinflammation (Kim et al., 2006; Maroso et al., 2010). We observed that HMGB1 is constitutively expressed in the developing brain and that this expression was drastically decreased in the damaged hippocampal CA3 region and in the ipsilateral cerebral cortex at 24 h (Fig. 4X,Y) and 3 d (Fig. 4Z–Zc) after the insult. This result is in accordance with previous findings showing that HMGB1 could be released by dying neurons into the extracellular space (Muhammad et al., 2008), acting as a DAMP and triggering microglial activation through the receptor for advanced glycation end products (RAGE), TLR2, TLR4, and Mac1 receptor (Kim et al., 2006; Maroso et al., 2010; Fang et al., 2012).
Disruption of TRIF signaling impairs spatial learning and motor function in hypoxic–ischemic female mice
Given that several TLRs are already expressed in the developing brain and that hypoxia–ischemia further induces the upregulation of these receptors' mRNA, we used TRIF knock-out (KO) and MyD88 KO mice to investigate the role of TLR signaling on the development of long-term neurological deficits after neonatal hypoxia–ischemia. The T-water maze task was used to assess spatial learning 12 weeks after the surgery. Surprisingly, there was no difference in the latency to find the platform in any of the trials, between hypoxic–ischemic wild-type (WT) males (HI WT male) and sham-operated WT males (Fig. 5A), nor between hypoxic–ischemic WT female mice (HI WT female) and sham-operated WT female mice (Fig. 5B). There was also no difference between hypoxic–ischemic MyD88 KO male mice (HI MyD88 KO male) and sham-operated MyD88 KO male (Fig. 5A), nor between hypoxic–ischemic MyD88 KO female mice (HI MyD88 KO female) and sham-operated MyD88 KO female animals (Fig. 5B).
Interestingly, although hypoxic–ischemic TRIF KO males (HI TRIF KO males) and sham-operated TRIF KO males performed this task similarly (Fig. 5C), TRIF KO female mice subjected to hypoxia–ischemia (HI TRIF KO females) exhibited a prolonged latency to reach the platform in the second trial of this learning test compared with the sham-operated TRIF KO female group (Fig. 5D; p < 0.001, two-way ANOVA with Bonferroni post hoc test).
In a similar way, HI TRIF KO females were the only group to exhibit a decreased time to fall from the rotarod compared with their corresponding sham-operated control group (Fig. 5E,F; p < 0.01, two-way ANOVA with Bonferroni post hoc test) 5 weeks after the insult. In addition, there was a significant difference in the latency to fall from the rotarod between HI TRIF KO females and HI WT females (Fig. 5F; p < 0.001, two-way ANOVA with Bonferroni post hoc test). These results indicate that TRIF signaling presents an important role in protecting female mice from long-term cognitive and motor impairments after neonatal hypoxia–ischemia.
TRIF KO male mice have increased hypoxic–ischemic hippocampal atrophy
Hippocampal atrophy and developmental amnesia can occur as a consequence of hypoxic–ischemic brain injury early in life, even in the absence of other major neurological deficits (Gadian et al., 2000). To determine whether the absence of MyD88 or TRIF signaling could affect the development of hippocampal atrophy, CA1/CA2 and CA3 areas were measured in both hemispheres in NeuN-immunostained coronal sections 14 weeks after the insult. Hippocampal damage was therefore expressed as the ratio between ipsilateral/contralateral areas for each region. Similar ratios were observed between HI WT males and HI MyD88 KO males (Fig. 6A,C), as well as between HI WT females and HI MyD88 KO females (Fig. 6B,D), in both regions (Fig. 6G,H). Surprisingly, although there was no difference between HI WT females and HI TRIF KO females (Fig. 6B,F,G,H), HI TRIF KO males exhibited a pronounced atrophy of CA3 (but not of CA1/CA2) compared with HI WT males (Fig. 6A,E,G,H; p < 0.05, Kruskal–Wallis test followed by Dunn's test).
Effects of MyD88 or TRIF deficiency on the neuroinflammatory response after neonatal hypoxia–ischemia
The hippocampal damage elicited an acute/subacute inflammatory response, as indicated by an increase in the number of Iba1+ microglial cells in the ipsilateral hippocampus compared with the contralateral hippocampus (Fig. 6I) 3 d after the injury. These activated microglial cells of the ipsilateral hippocampus presented an amoeboid morphology and strongly expressed CD68, in sharp contrast with the ramified microglia of the contralateral hippocampus, which expressed low levels of CD68 (Fig. 6I). Quantification of the number of Iba1+ microglia revealed increased ratios of microglial cell numbers between the ipsilateral hippocampus (CA1/2 and CA3 regions) and the corresponding contralateral regions in HI WT males, HI WT females, HI TRIF KO males, and HI TRIF KO females 3 d after the injury (Fig. 6P,R). In addition, we observed significant differences between the ipsilateral/contralateral ratios of microglial cell numbers in the CA1/CA2 and CA3 regions when comparing HI WT males and HI MyD88 KO males (Fig. 6P,R; p < 0.01 for each region, Kruskal–Wallis test with Dunn's multiple-comparisons test). Similar results were obtained when the number of CD68+-activated microglia was quantified (Fig. 6J–O,Q,S). Therefore, these data revealed that MyD88 signaling is important for the recruitment of microglial cells to the hippocampus after neonatal hypoxia–ischemia, particularly in males.
We also performed qRT-PCR to evaluate the expression of the cytokines Interferon-β (IFN-β) and Interleukin-1 β (IL-1β) 3 d after the injury. IFN-β mRNA expression was below the detection levels in the contralateral hemisphere, but could be detected in the ipsilateral hemisphere of 5/7 HI WT males and 2/7 HI WT females (Fig. 6T). Although it is known that the TRIF pathway is used by TLR3 and TLR4 to induce IFN-β expression, we detected IFN-β mRNA expression in the ipsilateral hemisphere of 2/7 HI TRIF KO males and 3/7 HI TRIF KO females (Fig. 6T). These results suggest that the transcription of this cytokine could at least be partially induced by TRIF-independent pathways after neonatal hypoxia–ischemia, similarly to what has been shown recently in models of infection (Reim et al., 2011; Aubry et al., 2012). It was also interesting that HI TRIF KO female mice exhibited an increased ratio of IL-1β mRNA levels between the ipsilateral and contralateral hemispheres compared with HI WT male mice 3 d after the injury (Fig. 6U; p < 0.05, Kruskal–Wallis test with Dunn's multiple-comparisons test). A similar trend was observed when comparing HI TRIF KO female and HI WT female mice (Fig. 6U), indicating an increased subacute proinflammatory response in HI TRIF KO female mice.
Disruption of CX3CR1 signaling impairs spatial learning in hypoxic–ischemic female mice
To further investigate the role of the microglial response after the injury, we used mice lacking the fractalkine receptor CXRCR1, given that we observed a drastic reduction in the expression of fractalkine mRNA in the damaged hippocampus, as described above. CX3CR1 is exclusively expressed by microglia/macrophages in the developing and adult brain (Mizutani et al., 2012) and fractalkine-CX3CR1 interaction is a crucial signaling pathway by which neurons modulate microglial function (Harrison et al., 1998; Chapman et al., 2000). For example, it is known that membrane-bound fractalkine-CX3CR1 interaction constitutively inhibits microglial activation, keeping these cells in a surveillance phenotype. Therefore, a decrease in fractalkine-CX3CR1 interaction could result in microglia hyperactivation under pathological conditions (Wolf et al., 2013). In addition, it is known that excitotoxic insults can lead to the cleavage and release of neuronal fractalkine, which then acts as a chemokine (Chapman et al., 2000). To elucidate the impact of CX3CR1 deficiency on the functional outcome after the hypoxic–ischemic insult, the T-water maze task was used to assess spatial learning 12 weeks after the injury. Although there was no difference between hypoxic–ischemic CX3CR1 KO males (HI CX3CR1 KO males) and sham-operated CX3CR1 KO males (Fig. 7A), hypoxic–ischemic CX3CR1 KO female mice (HI CX3CR1 KO female) exhibited a prolonged latency to reach the platform in the second and third trials compared with the sham-operated CX3CR1 KO females (Fig. 7B; p < 0.05 and p < 0.01 in each trial, respectively, two-way ANOVA with Bonferroni post hoc test).
CX3CR1 KO female mice have an increased hypoxic–ischemic brain injury
It has been shown that subplate neurons of P1/P2 rats are particularly vulnerable to hypoxic–ischemic insults (McQuillen et al., 2003) and that subplate damage can be observed in preterm infants with periventricular leukomalacia (Kinney et al., 2012). These neurons are transiently present in the developing brain, playing an important role in the maturation of thalamocortical synapses, in thalamocortical and corticofugal axon guidance, in developmental plasticity, and in the formation of the columnar neocortical architecture (Kanold and Luhmann, 2010). We performed in situ hybridization using a specific probe to CTGF, a molecular marker of subplate neurons in the early postnatal mouse brain (Hoerder-Suabedissen and Molnár, 2012), to evaluate whether these neurons are affected by neonatal hypoxia–ischemia in P3 mice. We observed that CTGF mRNA expression in the subplate was decreased in a subpopulation of HI WT male (7/22 mice) and HI WT female mice (4/26 mice) 3 d after the injury. Because these changes varied from the lack of CTGF expression in a single area to multiple coalescent areas lacking CTGF in the subplate region, we used a score to compare these changes among the experimental groups (Fig. 7M). Interestingly, HI CX3CR1 KO female mice exhibited an increased damage score compared with HI WT females (Fig. 7G–J,M; p < 0.05, Kruskal–Wallis test with Dunn's multiple-comparisons test), whereas there was no difference between HI WT males and HI CX3CR1 KO males (Fig. 7C–F,M). It was also interesting that, in very rare cases with a more severe brain injury, CTGF expression was induced in the damaged cerebral cortex, whereas it was still decreased in the subplate (Fig. 7K,L).
Furthermore, HI CX3CR1 KO females had a more pronounced atrophy of the hippocampal CA1/2 and CA3 regions compared with HI WT females (Fig. 7O,Q–S; p < 0.05, Kruskal–Wallis test followed by Dunn's test), whereas no significant differences were found between HI WT males and HI CX3CR1 KO males (Fig. 7N,P,R,S) as assessed by the ratio between ipsilateral/contralateral NeuN-stained areas 14 weeks after the insult. These data clearly indicate that CX3CR1 signaling protects female mice from a hypoxic–ischemic brain injury.
We also observed that HI CX3CR1 KO males and HI WT males presented similar ratios of microglial cell numbers between the ipsilateral hippocampus (in both CA1/2 and CA3 regions) and the corresponding regions of the contralateral hemisphere (Fig. 8E,G). Conversely, HI CX3CR1 female mice exhibited a trend to have higher ratios, particularly in the CA3 region, compared with HI WT female mice (Fig. 8G). Similar results were obtained when the number of CD68+-activated microglia was quantified (Fig. 8A–D,F,H). In addition, no significant differences were observed in the ratio of IL-1β mRNA levels between the ipsilateral and contralateral hemispheres of HI CX3CR1 KO males, HI WT males, HI CX3CR1 KO females, and HI WT females, despite the trend of CX3CR1 KO mice to have higher ratios at 3 d after the insult (Fig. 8I).
Discussion
TLRs have been involved in the regulation of several developmental mechanisms, such as neurogenesis, neurite outgrowth, axonal growth, and neuronal apoptosis (Ma et al., 2006; Cameron et al., 2007, Lathia et al., 2008; Shechter et al., 2008). In this study, we observed that the expression of TLR2, TLR6, TLR7, and TLR9 mRNA in the brain increased after birth, in accordance with previous findings (Lathia et al., 2008; Kaul et al., 2012). In the late embryonic period, there was a strong expression of TLR2, TLR6, TLR7, and TLR9 in the subcallosal region in the midline. This region comprises the glial wedge and the subcallosal sling, which have an important role in the formation of the corpus callosum (Paul et al., 2007), suggesting that further studies are needed to determine whether TLR signaling affects the development of this commissural structure. These results confirm that TLR mRNA expression is developmentally regulated in the mouse telencephalon and that the first postnatal week represents a critical period when all of the studied TLRs are highly expressed. The mouse brain in this period corresponds to the brain of very preterm infants, which is particularly vulnerable to hypoxic–ischemic insults (Volpe et al., 2011). For this reason, we used a model of cerebral hypoxia–ischemia in P3 mice to evaluate the role of TLR signaling in the development of cognitive/motor deficits and hippocampal damage.
We first observed that neonatal hypoxia–ischemia induces the transcriptional activation of TLR2, TLR6, TLR7, and TLR9. The spatial pattern of expression of these receptors corresponded to areas where neuronal damage had occurred and the peak of expression occurred between 24 h and 3 d after the insult. A recent study using a PCR array system found upregulation of TLR1, TLR2, and TLR7; downregulation of TLR5; and no changes in the levels of TLR3, TLR4, TLR6, TLR8, and TLR9 24 h after hypoxia–ischemia in P9 mice. In addition, the investigators observed a smaller infarct size in TLR2 KO mice but not in TLR1 KO mice (Stridh et al., 2011).
In the present study, we also observed the downregulation of HMGB1 in areas of neuronal injury, which is probably caused by the secretion of this molecule into the extracellular space by dying neurons (Muhammad et al., 2008). It is known that HMGB1 can act as a DAMP and that it is possible to achieve neuroprotection by the immunoblockade of HMGB1 after stroke in adult mice (Muhammad et al., 2008). However, the role of HMGB1 signaling in animal models of perinatal brain damage remains to be determined.
Given that a reduction in hippocampal area was the most reproducible long-term outcome in this model of neonatal hypoxia–ischemia and that a reduction in the hippocampal volume is one of the possible alterations found in the brains of preterm/very-low-birthweight children (de Kieviet et al., 2012), this parameter was used for the comparison of the structural outcome among the groups 14 weeks after injury. For measurement of functional outcome, we chose a cognitive test (T-water maze task) and a motor test (rotarod performance test). Surprisingly, there was no difference between hypoxic–ischemic WT animals and sham-operated controls in both functional tests, probably due to the increased neuroplasticity of the developing brain (Quairiaux et al., 2010). Conversely, HI TRIF KO female mice presented greater learning and motor impairments compared with sham-operated controls. In addition, HI TRIF KO male mice exhibited a pronounced atrophy of the hippocampal CA3 region compared with WT mice, indicating that TRIF signaling exerts a protective role after neonatal hypoxia–ischemia and that the mechanisms underlying this effect are gender specific. Among the possible mechanisms involved in this effect, we found a more robust expression of the proinflammatory cytokine IL-1β mRNA in the subacute phase of the insult in HI TRIF KO female mice.
We also observed that MyD88 deficiency does not alter the functional outcome or the hippocampal damage in either gender despite the reduced microgliosis in the ipsilateral hippocampus of HI MyD88 KO male mice, in agreement with a previous study (Wang et al., 2009). However, gender-specific effects were not analyzed and functional tests were not performed in that study. Moreover, two recent studies found no differences in neuronal survival when comparing TRIF KO, MyD88 KO, and WT adult male mice subjected to cerebral ischemia (Hua et al., 2009; Famakin et al., 2011). However, because these studies did not assess long-term outcome, it is not possible to affirm that the protective role of TRIF signaling observed here is specific to the neonatal period. Furthermore, the synthetic TLR3 ligand poly I:C, which signals via a TRIF-dependent pathway, exerts a neuroprotective effect in mixed cortical cultures subjected to oxygen-glucose deprivation (Marsh et al., 2009), and TRIF signaling is crucial for the clearance of axonal debris by microglia, facilitating axonal outgrowth after dorsal root axotomy in adult mice (Hosmane et al., 2012).
In the second part of the present study, we investigated the role of neuron-microglia crosstalk through the fractalkine-CX3CR1 pathway after neonatal hypoxia–ischemia. Although fractalkine is expressed by neurons (Harrison et al., 1998), CX3CR1, the only known fractalkine receptor, is expressed solely by microglia in the brain, as well as by circulating monocytes, dendritic cells, NK cells, and T cells (Mizutani et al., 2012). We observed that the constitutive expression of fractalkine mRNA in the hippocampus was drastically reduced after neonatal hypoxia–ischemia, in agreement with previous findings in adult animals (Harrison et al., 1998; Pimentel-Coelho et al., 2013). CX3CR1 KO adult mice have an increased neuronal loss in several animal models of neurological disease (Cardona et al., 2006; Blomster et al., 2011), although they exhibit decreased neuronal damage after traumatic spinal cord injury (Donnelly et al., 2011) and stroke (Dénes et al., 2008). This conflicting evidence suggests that the role of CX3CR1 signaling in the injured adult brain could be context dependent. We observed that HI CX3CR1 KO female mice presented a pronounced injury in the hippocampus and in the subplate, as well as increased learning deficits, compared with HI WT female mice, whereas there was no difference between HI CX3CR1 KO males and HI WT males. However, despite the critical inhibitory role of fractalkine-CX3CR1 signaling in microglia activity, we cannot discard a possible contribution of CX3CR1 deficiency in monocytes to the outcome of HI CXCR1 female mice. In addition, recent studies have reported several CX3CR1-mediated microglia functions during postnatal brain development (Paolicelli et al., 2011; Hoshiko et al., 2012) and, therefore, it is possible that changes in fractalkine expression and/or in microglial function after the injury could result in secondary alterations in brain development.
Regarding the gender-specific effects of neuroinflammation, Kentner et al. (2010) observed that the systemic administration of lipopolysaccharide in P14 rats increases the expression of cyclooxygenase-2 in the hypothalamus in male, but not in female, rats in adulthood. Gender-specific differences in mitochondrial dysfunction (Weis et al., 2012), in the activation of cell death pathways (Zhu et al., 2006), and in the efficacy of certain neuroprotective drugs (Comi et al., 2006; Wen et al., 2006; Nijboer et al., 2007; Fan et al., 2011; Fleiss et al., 2012) have also been demonstrated in animal models of neonatal hypoxia–ischemia and neonatal stroke. However, it is still not clear how much of these differences could be attributed to sex chromosome effects or to the action of sex hormones (Hill and Fitch, 2012) and how they contribute to the neurological outcome (Arteni et al., 2010; Hill et al., 2011; Sanches et al., 2013). It is known that there are elevated circulating levels of dihydrotestosterone in males during the late embryonic period, which persist through the first year of life in humans (Christine Knickmeyer and Baron-Cohen, 2006), and that the developing brain has the capacity to convert testicularly derived androgens into estrogens (McCarthy, 2009). Naugler et al. (2007) showed that estrogens inhibit the inflammatory response (MyD88-dependent IL-6 production) triggered by hepatocyte necrosis in adult female mice. Conversely, Saijo et al. (2011) demonstrated that 5-androsten-3β,17β-diol, acts as a ligand of estrogen receptor β, suppressing the inflammatory response of microglia and astrocytes, and that 17β-estradiol antagonizes this anti-inflammatory activity. Therefore, it remains to be elucidated whether sex hormones could contribute to the neuroinflammatory response after neonatal hypoxia–ischemia.
In conclusion, the present study provides the first evidence of a gender-specific protective role of neuroinflammation after a sterile hypoxic–ischemic injury in a mouse model of the preterm brain. These findings may have important clinical implications, because there is a growing interest in the development of new therapeutic approaches based on the modulation of neuroinflammation in infants with perinatal brain damage. Our results support the concept that one of the main challenges for the development of novel immunomodulatory therapies is to selectively antagonize the deleterious consequences of neuroinflammation (Volpe et al., 2011; Kaindl et al., 2012) without abrogating its beneficial effects (Covey et al., 2011; Faustino et al., 2011). Furthermore, our findings indicate that gender should be considered as a critical variable when investigating the role of inflammation in the developing brain.
Footnotes
This work was supported by the Canadian Institutes of Health Research, the Canadian Stroke Network, and the Canadian Stroke Network and European Stroke Network Collaborative Research Initiative. P.M.P.-C. was supported by a doctoral scholarship from the Conselho Nacional de Desenvolvimento Científico e Tecnológico in 2009. J.-P.M. is supported by a doctoral scholarship from the Canadian Institutes of Health Research. We thank Marie-Michèle Plante, Paul Préfontaine, and Nataly Laflamme for excellent technical assistance.
The authors declare no competing financial interests.
- Correspondence should be addressed to Serge Rivest, Neuroscience Laboratory, Centre Hospitalier Universitaire de Québec Research Center and Department of Molecular Medicine, Faculty of Medicine, Laval University, 2705 Boulevard Laurier, QC G1V 4G2, Canada. Serge.Rivest{at}crchul.ulaval.ca