NF‐κB Activity in Perinatal Brain During Infectious and Hypoxic‐Ischemic Insults Revealed by a Reporter Mouse

Abstract

Infants suffering from infection or hypoxia–ischemia around the time of birth can develop brain damage resulting in life‐long impairment such as cerebral palsy, epilepsy and cognitive disability. Inflammation appears to be an important contributor irrespective of whether the primary event is infection or hypoxia–ischemia. Activation of the transcription factor NF‐κB is a hallmark of inflammation. To study perinatal brain inflammation, we developed a transgenic reporter mouse for imaging NF‐κB activity in live animals and tissue samples. The reporter genes firefly luciferase and a destabilized version of enhanced GFP (dEGFP) were regulated by common NF‐κB sites using a bidirectional promoter. Luciferase activity was imaged in vivo, while dEGFP was detected at cellular level in tissue sections. In newborn mice subjected to experimental models of infections or hypoxia–ischemia; luciferase signal increased in brains of live animals. In brain sections dEGFP expression, revealing NF‐κB activation was observed in the endothelial cells of the blood–brain barrier in all disease models. In meningitis and hypoxia–ischemia expression of dEGFP was also induced in perivascular astrocytes. In conclusion, by using this transgenic reporter mouse in experimental models of perinatal complications, we could assess NF‐κB activity in vivo and subsequently determine the cellular origin in the tissues.

INTRODUCTION

In the perinatal period of human development, insults can lead to death or life‐long encephalopathy such as cerebral palsy, epilepsy and vision, language and behavioral disorders 16, 39, 41, 75. Neonatal encephalopathy, which is defined as the clinical manifestation of disordered brain function of full‐term neonates, occurs in about 3 per 1000 live births (42). The long‐term outcome of these children beyond cerebral palsy is not well described and cognitive disabilities are probably much more common than earlier assumed (15). In the more vulnerable condition of preterm neonates with low birth weight (<1.5 kg), 10% develop cerebral palsy and about 50% exhibit cognitive and behavioral deficits (76). The financial and human costs to infants affected, their parents and the society are considered to be immense 25, 42. The etiology of the encephalopathy is not fully understood but it is probably multifactorial with maternal or fetal infections and neonatal hypoxia–ischemia among the most common causes, but metabolic disorders, maternal stress, exposure to medication and drug abuse are also suggested to be involved 12, 17, 18, 45, 74. The underlying pathogenetic mechanisms seem very complex and are to a large extent unknown; however, inflammatory processes appear to have a central role both in infection and hypoxia–ischemia [for review see 17, 18, 22, 73].

Fundamental in regulation of inflammation is the family of transcription factors called nuclear factor‐κB (NF‐κB), which is a set of dimeric transcription factors present in virtually every cell in the body 51, 65. Activation of NF‐κB is triggered by a range of different stimuli typically arising during tissue injury and infection, and initiates numerous processes involved in defense and repair of tissue damage. In the brain, the role and function of NF‐κB depends on the cellular context. In neurons, where NF‐κB is suggested to be constitutively active (36) although debated (49), it appears to be involved in memory and learning as well as in neuroprotection 19, 24, 53, 54, 59. In astrocytes, NF‐κB is highly inducible and seems crucial for initiating inflammatory responses such as recruitment of macrophages and microglia to the site of injury. This is beneficial during the first stages of an injury, whereas sustained activation of NF‐κB can aggravate inflammatory responses and consequently worsen the outcome in adult mice 5, 20. Less is known for neonates; however, pharmacological inhibition of NF‐κB significantly reduced brain injury after hypoxic‐ischemic insults manifested as improvement in long‐term motor and cognitive functions (72), and it would be of significance to determine both the dynamics and cell‐specific activation of NF‐κB during injuries and potential curative treatments.

To assess NF‐κB during perinatal brain injury, we have developed an NF‐κB reporter mouse, which expresses the two reporter genes luciferase and destabilized enhanced green fluorescent protein (dEGFP) (46). Exploiting these two reporter genes allow assessment of NF‐κB activity noninvasively in living animals by luciferase and in single cells of tissue samples by dEGFP. Conventional methods for assessments of NF‐κB such as electrophoretic shift assay and antibody staining are both tedious and not very sensitive, thus NF‐κB reporter mice has emerged as particularly useful tools to easily and rapidly assess NF‐κB activity (38). The present NF‐κB reporter mice are unique because two reporter genes are transcribed in opposite directions by a bidirectional promoter implementing regulation of the reporter genes by the same NF‐κB response elements, which thus likely promote co‐regulation.

We here describe the development of this NF‐κB reporter mouse and its utility to study inflammation in the brain of newborn mice caused by infectious agent and hypoxia–ischemia. In vivo imaging showed clear increase in luciferase‐mediated signal from brain. Furthermore, increased levels of dEGFP were detected in endothelial cells and astrocytes, but with variation between different disease models. Thus, this reporter mouse, prove, for the first time, cell‐specific activation of NF‐κB during brain injury in the perinatal mouse brain.

METHODS

Development of an NF‐κB‐responsive transgene

The transgenic DNA construct is based on the pBI‐L vector, which contains the firefly luciferase gene (Clontech, Mountain View, CA, USA). The tetracycline‐controlled transcriptional activation (TET)‐responsive promoter was replaced with a novel bidirectional promoter responsive to NF‐κB. The dEGFP gene was inserted at the side of the promoter opposite to the luciferase gene (Figure 1A). First, four oligos: two complementary oligos for each of the two core promoters were annealed and inserted in the PstI/MluI sites of pBI‐L. The two core promoter units were ligated together by an AvrII site and thus oriented in opposite directions of each other. The two promoters contain identical initiation site, TATA‐box and BRE (transcription factor IIB recognition element). Next, dEGFP from pd2EGFP‐1 vector (Clontech) was inserted between the AgeI and NotI sites. Finally, two complementary oligos containing one NF‐κB site and AvrII overhangs were annealed and ligated into the AvrII site between the core promoters. This will give a cocktail of bidirectional promoters with various number of NF‐κB sites depending on number of oligos inserted. Clones were amplified in SURE2 bacteria (Stratagene/Agilent, Santa Clara, CA, USA) and number of NF‐κB sites was estimated based on length of polymerase chain reaction (PCR) fragments spanning the bidirectional promoter. Clones that contained from zero to nine NF‐κB sites were confirmed by sequencing and transfected, using Lipofectamine reagents (Invitrogen, Carlsbad, CA, USA) in the hepatoma cell line HepG2 for evaluation of luciferase and dEGFP expression.

Figure 1.

A bidirectional NF‐κB‐regulated promoter co‐expressing luciferase and dEGFP. A. Schematic representation of the bidirectional transgene. The two core promoters contain a TATA‐box, a BRE and a transcriptional initiation site (inr) for efficient transcription. Six NF‐κB sites are placed between the core promoters. The dEGFP contains a PEST sequence to accelerate proteasome‐dependent degradation. Thus, dEGFP achieve a similar half‐life as luciferase of approximately 3 h. B. In vivo imaging of luciferase‐mediated luminescence from an untreated and LPS‐treated mouse. C. Ex vivo image of dissected organs from the same mice. Warmer colors indicate more luminescence.

Oligos

Core promoter directed toward luciferase

• 1 ^P‐GTCCAGGCGATCTGACGGTTCACTAAACGAGCTCTGCTTATATAGGCGCCGC
• 2ACGTCAGGTCCGCTAGACTGCCAAGTGATTTGCTCGAGACGAATATATCCGCGGCGGATC^‐P

Core promoter directed toward dEGFP

• 1 ^P‐CTAGGCGGCGCCTATATAAGCAGAGCTCTTTAGTGAACCTCAGATCGCCTGGAACCGGTA
• 2CGCCGCGGATATATTCGTCTCGAGAAATCACTTGGAGTCTAGCGGACCTTGGCCATGCGC^‐P

NF‐κB binding site (NF‐κB response element is underlined)

• 1 ^P‐CTAGGGGACTTTCC
• 2 CCCTGAAAGGGATC^‐P

PCR Primers for determining the number of NF‐κB binding sites

• 1Forward: CTTATGCAGTTGCTCTCC
• 2Reverse: ACCCTGAAGTTCTCAGTC

Generation of transgenic mice

The transgene containing the NF‐κB‐responsive bidirectional promoter and reporter genes were isolated as a ∼4.4 kb Spe1/Cla1 fragment and microinjected into the pronucleus of B6CBA zygotes, which subsequently were implanted into pseudopregnant CD‐1 recipient mice. This was performed at the Norwegian Transgenic Centre at the University of Oslo. Transgenic animals were identified by PCR of sequences in the luciferase and dEGFP genes.

Bioluminescence imaging

Luciferase expression in cell cultures and in mice was recorded as bioluminescence with an IVIS 100 imaging system (Caliper, Alameda, CA, USA). Cells cultured in 96 or 12 well plates were incubated with the luciferase substrate d‐luciferin (200 µg/mL; Biosynth, Staad, Switzerland) for 8 minutes prior to imaging. For in vivo imaging of luciferase, d‐luciferin (75 mg/kg) was injected intraperitoneally in adult and subcutaneously in newborn mice. Adult mice were immobilized during the imaging session with 2.5% isoflurane anesthesia, while newborn mice were carefully fixed with tape across the back or placed in small containers. In vivo images were taken 8 minutes after administration of d‐luciferin with exposure time between 5 and 30 s. Ex vivo imaging of dissected organs was conducted 15 to 20 minutes after administration of d‐luciferin intraperitoneally. This was done on whole organs, except for liver and lung, where one specific lobe was imaged and for mesenteric fat where a portion of approximately 0.5 cm³ was dissected and imaged. When luciferase activity was compared between individual organs, the time from luciferin injection to ex vivo imaging was always the same. The images were acquired and analyzed with the Living Image software (Caliper).

Western blot analyses

Proteins were extracted from cells and organs by homogenization in lysis buffer from Promega (Madison, WI, USA) supplemented with protease inhibitor (Complete EDTA‐free, Roche, Mannheim, Germany) and quantified using Bradford solution (Bio‐Rad, Hercules, CA, USA). Proteins were separated by 12.5% sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) and blotted on polyvinylidene fluoride (PVDF) membranes. The blots were incubated in blocking buffer for 1 h; primary antibodies overnight at 4°C; and secondary antibodies for 1 h at room temperature. The blocking and primary antibody solutions were both tris‐buffered saline (TBS) with 5% nonfat milk and 0.1% Tween‐20. The secondary antibody solution was identical except without milk. Rabbit anti‐GFP antibodies were used with a dilution of 1:1000 (Invitrogen, Paisley, UK) and secondary antibody with 1:8000 dilution (Southern Biotechnology, Birmingham, AL, USA). The secondary antibodies were conjugated with horseradish peroxidase (HRP), which was detected by chemiluminescence (Supersignal West Pico) as recommended by the manufacturer (Pierce/Thermo Scientific, Rockford, IL, USA).

Primary astrocyte culture

Astrocytes were isolated from 4‐day‐old mice according to a modified protocol from McCarthy and deVellis (52). After decapitation and removal of the meninges, cerebral cortices from three brains were passed through a 70‐µm cell strainer and plated in poly‐L‐lysine‐coated 75‐cm² flasks (Corning, Lowell, MA, USA). The cells were incubated at 37°C/5% CO₂ and culture media were changed every second day for 1 week. In order to detach microglia and precursor cells seeded on top of protoplasmic astrocytes cell layer, the flasks were gently shaken for 2 minutes during medium change. Immunostaining for glial fibrillary acidic protein (GFAP) and the appearance of flat polygonal morphological phenotype indicated presence of more than 95% astrocytes. Culture medium was in the entire procedures Dulbecco's Modified Eagle Medium (DMEM)‐glutamax with 10% fetal bovine serum (FBS), 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco/Invitrogen).

Fixation of cell cultures and brain sections

Astrocyte primary cultures was fixed in 4% formaldehyde in phosphate‐buffered saline (PBS) for 30 minutes and rinsed in PBS. Neonatal mice were fixed 6 h after treatment by transcardial perfusion using 4% formaldehyde in PBS after washing out blood cells with 2% dextran in phosphate buffer. Brains were dissected, postfixed overnight at 4°C and cryoprotected successively in 10%, 20% and 30% sucrose. The brains were frozen in optimal cutting temperature (OCT) compound (Sakura Finetek, Alpena an den Rijn, the Netherlands) and cut in sagittal and coronal sections of 14 µm.

Immunofluorescence

Astrocyte cultures and brain sections were treated equally. Permeabilization was done with 0.1% Triton X‐100 in PBS for 10 minutes before blocking in 2% bovine serum albumin or benzenesulfonic acid (BSA) in PBS for 2 h and incubation with primary antibodies in the same solution overnight (4°C). The primary antibodies were used at the following dilutions: anti‐rat‐CD31 1:800 (BD Biosciences, Franklin Lakes, NJ, USA); anti‐chicken‐GFAP 1:1000 (Nordic BioSite, Oslo, Norway), anti‐mouse‐NeuN 1:1000 (Chemicon/Millipore, Billirica, MA, USA), anti‐rabbit‐GFP 1:1000 (Invitrogen), anti‐rabbit‐Oligo2 1:200 (Chemicon), anti‐goat‐ionized calcium‐binding adaptor molecule (Iba) 1:200 (Abcam, Cambridge, UK) and anti‐rabbit‐aquaporin 4 (AQP4) 1:1000 (Chemicon). After washing in PBS, the specimens were incubated for 1 h with secondary donkey antibodies diluted 1:1000 in PBS: Cy2 anti‐chicken, Cy5 anti‐rat and Cy3 anti‐rabbit (Jackson Immunoresearch Laboratories, West Grove, PA, USA). Mounting were done in Prolong Gold with DAPI (Molecular Probes/Invitrogen) and analyzed using a Zeiss confocal laser scanning microscope with a 40× oil objective (Carl Zeiss, LSM 510 Meta, Thornwood, NY, USA).

Animal models of perinatal brain insults

Animal experiments were in accordance with the national guidelines and with the approval of the Animal Care Committee in Norway. Mice of both genders at age 4 or 5 days were used. Each experiment contained mice from two litters. Brain inflammation was induced by three methods. To mimic meningitis, 2 µL lipopolysaccharide (LPS, 0.05 µg/µL, serotype 055:B5, Sigma‐Aldrich, St. Louis, MO, USA) solution was injected just intracranial at the midpoint between bregma and lambda 1 mm laterally to the midline (44). Systemic infection was induced by subcutaneous injection of LPS (0.15 µg/kg) in the neck of the animals. Experimental brain hypoxia–ischemia was induced by ligating the left carotid artery succeeded by the following breathing conditions: first in 8% oxygen for 1 h then in 100% oxygen for 45 minutes (40). Surgery and intracranial injection were done under isoflurane anesthesia.

RESULTS

An NF‐κB reporter mouse expressing luciferase and dEGFP by a bidirectional promoter

We cloned a bidirectional expression cassette comprised of two almost identical core promoters oriented in opposite directions. The luciferase and dEGFP reporter genes were inserted downstream on each side of the promoter (Figure 1A). Between the two core promoters, we inserted from zero to nine NF‐κB response elements in order to evaluate the significance of the number of cis‐elements on reporter gene expression. HepG2 cells, transfected with the different clones and stimulated with 5 ng/mL tumor necrosis factor‐α (TNF‐α), were assessed for luciferase activity and dEGFP signal. The promoter containing six NF‐κB sites showed most advantageous properties with respect to luminescence strength and relative increase (induction ratio) in signal from luciferase and dEGFP (data not shown) and were accordingly used to produce transgenic mice. Five PCR‐positive founders, after pronuclear injections, were evaluated for luciferase dependent signal in response to LPS stimulation, both in terms of amount of luminescence and induction ratio. LPS (1 mg/kg) was injected intraperitoneally and mice were imaged in vivo and dissected organs were imaged ex vivo (Figure 1B,C). In response to the 5‐h LPS treatment, the in vivo luciferase signal of the different founders ranged from 4.3 × 10⁴ to 4.7 × 10⁶ photons/s/cm²/sr. The induction ratio differed from 12‐ to 25‐fold. Additionally, as expected, the ex vivo images of organs revealed substantial differences between the founders. As we intended to study NF‐κB response in the brain, we identified the founder, which best met the criteria of high signal strength and induction ratio in the brain, and used it in the rest of the experiments.

The presence of dEGFP was verified with western blots of proteins isolated from liver and brain from mice stimulated with LPS (1 mg/kg) intraperitoneally (Figure 2A). As shown in Figure 2B the dEGFP expression reached maximum after approximately 6 h and then declined, whereas the luciferase signal reached maximum after 4 h. This is in accordance with previous studies of these reporter genes 8, 46.

Figure 2.

Expression of dEGFP in mice. A. Western blots with an antibody against dEGFP on tissue extracts from control (C) and LPS‐stimulated mice. Proteins were isolated from liver and brain of both wild‐type (WT) and transgenic reporter mice. B. Graph comparing the time dynamics of dEGFP expression, quantified by western blots, with luciferase expression quantified by in vivo imaging (n = 5, error bars are SEM).

To demonstrate dEGFP expression in the brain, we injected ∼5 ng TNF‐α directly into the brain. After 6 h, the luminescent signal from the head region was 100 times stronger in comparison with control animals (Figure 3A). However, when trying to identify fluorescent cells of cryosections from fixed brains with confocal laser scanning microscopy, we found the dEGFP fluorescence to be relatively weak, which further made it difficult to reliably distinguish stimulated from unstimulated cells (Figure 3B). In order to enhance the signal, we stained brain sections with an antibody against dEGFP. This amplification step clearly improved the signal (Figure 3C). Double staining with the antibodies against dEGFP, together with antibodies against various cell‐specific markers, showed that dEGFP was mainly expressed in endothelial cells. We additionally observed some labeling in cellular processes, which appear to be astrocytes. We could not observe dEGFP in neurons, microglia or oligodendrocytes (Figure 3D). To verify the astrocyte expression, we isolated and cultured these cells from brains of 4‐day‐old mice and stimulated them with TNF‐α (5 ng/mL). Luciferase signal was increased in the stimulated cell cultures and western blots of protein extracts showed expression of dEGFP (Figure 4A,B). Immunohistochemistry with an antibody against dEGFP showed strong labeling in the cells, which was confirmed to be astrocytes with antibodies against GFAP (Figure 4C).

Figure 3.

Expression of luciferase and dEGFP in perinatal mouse brain. A. In vivo imaging of luciferase‐mediated luminescence 6 h after intracranial injection of TNF‐α. B. Confocal microscopy of fluorescence from dEGFP in cortical brain sections with and without TNF‐α stimulation. Nuclei are DAPI stained. C. Immunofluorescence in cortical brain sections co‐stained with antibodies against dEGFP and cell‐specific markers for astrocytes (GFAP) and endothelial cells (CD31). The dEGFP signal is present in pia, arachnoidea and subpial astrocytic endfeet. dEGFP expression was increased in astrocytes (arrow), astrocytes perivascular endfeet (double arrows) and endothelial cells (arrowhead), with expression also in some of the vessels in the subarachnoid space. D. Double staining against dEGFP and cellular markers for neurons (NeuN), microglia (Iba1) or oligodendrocytes (Oligo2) show no co‐localizations. Scale bars are 50 µm. Asterisks show the subarachnoid space.

Figure 4.

Expression of luciferase and dEGFP in astrocyte cultures. A. Induction of luciferase‐mediated luminescence in astrocyte cultures after 6 h of TNF‐α stimulation (n = 6). Inset is image of cell wells with and without stimulation. Error bars are SEM. B. Western blot with cell culture extracts using an antibody against dEGFP. C. Fluorescent images of astrocytes stained with antibodies against dEGFP (green) and the astrocyte marker GFAP (red). Scale bars are 50 µm.

NF‐κB activation is induced in astrocytes and endothelial cells in a model of meningitis

To mimic meningitis in perinatal mice, we injected LPS intracranially at postnatal day four. Six hours after treatment, a strong luminescent signal from the head region was present (n = 5, Figure 5A). Additionally, a strong signal came from the abdominal area with the strongest signal from the liver region. This latter activation is possibly caused by LPS entering the blood stream; however, increased immune activity in the brain is suggested to induce NF‐κB activation in the liver through neural signaling (7). The dEGFP‐mediated fluorescent signal is increased in sections from the LPS‐stimulated brain (Figure 5B). However, as described above, immunohistochemical staining significantly improved the signal to noise ratio (Figure 5C). After co‐staining with antibodies against dEGFP and cell‐specific markers, we observed distinct labeling in astrocytes and endothelial cells of the LPS‐stimulated animals. Labeling was also present in typical perivascular endfeet of astrocytes. We saw the same pattern in both white and gray matter in both neocortex and deeper ventricular layers. No dEGFP labeling was revealed in neurons, oligodendrocytes or microglia (Figure 5D).

Figure 5.

Meningitis activated NF‐κB in astrocytes and brain endothelial cells. A. Experimental meningitis induced by delivery of LPS intracranially. In vivo imaging of luciferase‐mediated luminescence 6 h after LPS injection. B. Confocal microscopy images of fluorescence signal from dEGFP in brain sections. The dEGFP signal (green) was strongest in stimulated brain. Nuclei are DAPI stained (blue). C. Immunofluorescence of cortical brain sections co‐stained with antibodies against dEGFP and cellular markers for astrocytes (GFAP) and endothelial cells (CD31). dEGFP expression was increased in astrocytes (single arrows), astrocyte endfeet (double arrow) and endothelial cells (arrow heads). D. Immunofluorescent images of co‐staining with antibodies against dEGFP and cellular markers for neurons (NeuN), microglia (Iba1) and oligodendrocytes (Oligo2). No co‐staining was observed here. Asterisks indicate the subarachnoid space. Scale bars are 50 µm.

Systemic infection induced activation of NF‐κB in brain endothelial cells

Maternal infection can cause inflammation in the fetus (12). To model this situation, we induced systemic inflammation in perinatal mice through subcutaneous injection of LPS in 4‐day‐old mice. In vivo imaging of luciferase activity showed a time‐dependent increase, which peaked around 4 h (Figure 6A). The signal increased in the whole body with particularly strong intensity from the abdomen. From the head region, the increase was around 20‐fold in comparison with basal levels. Ex vivo images of the brains showed 10‐fold stronger luminescence from LPS‐treated animals compared to controls (n = 5, Figure 6B). Immunohistochemical staining of brain sections with antibodies against dEGFP and cell‐specific markers showed clear expression of dEGFP in endothelial cells (Figure 6C), but not in astrocytes, oligodendrocytes, microglia or neurons (Figure 6C,D). The pattern of dEGFP staining was also here found in various layers of the brain including gray and white matter of neocortex and ventricular zones. Additionally, we stained with antibody against AQP4 because upregulation of this brain‐specific water channel is previously observed in adult mice after activation of NF‐κB (33). AQP4 staining was present in astrocytes (Figure 6E); however, we observed no clear change in signal strength after LPS treatment.

Figure 6.

Systemic inflammation induced NF‐κB in endothelial cells of the brain. A. In vivo imaging of luciferase‐mediated luminescence at different time points after subcutaneous injection of LPS. B. Ex vivo imaging of dissected brains 6 h after LPS injection. C. Immunofluorescence of anti‐GFP co‐stained with antibodies against cell‐specific markers. Increased dEGFP expression was observed in endothelial cells (CD31, arrowheads), but not in astrocytes (GFAP). D. Double staining against dEGFP and cellular markers for neurons (NeuN), microglia (Iba1) or oligodendrocytes (Oligo2) showed no co‐localizations. E. Antibodies against AQP4 show expression in astrocytes. Asterisks indicate the subarachnoid space. Scale bars are 50 µm.

Experimental cerebral hypoxia–ischemia induced NF‐κB activation in endothelial cells and astrocytes

Cerebral hypoxia–ischemia was induced in 5‐day‐old mice by ligation of the left carotid artery followed by exposure to 8% oxygen for 1 h and 100% oxygen for 45 minutes. The animals were imaged in vivo 6 h after surgery, and in comparison with control mice, a clear increase in luminescent signal was observed from the left side of the head (n = 5, Figure 7A). Immunohistochemistry of brain sections showed significantly more labeling in insulted brains contra controls. Double labeling with antibodies against dEGFP and markers for endothelial cells and astrocytes verified increased expression of dEGFP in these cells (Figure 7B). We find the same pattern of EGFP staining in both gray and white matter of neocortex and ventricular zones (7D). No labeling for dEGFP was found in oligodendrocytes, microglia and neurons (Figure 7E). We here also stained for AQP4, and as for systemic inflammation, we observed expression in astrocytes but without significant differences between hypoxic‐ischemic and normal brains (Figure 7F).

Figure 7.

Hypoxia–ischemia activated NF‐κB in astrocytes and endothelial cells. A. In vivo recording of luciferase‐mediated luminescence 6 h after occlusion of the left carotid artery. B. Sagittal section of whole brain to indicate the regions of staining against dEGFP. C. Confocal immunofluorescent image of untreated neocortical brain sections stained with antibodies against dEGFP. D. Double staining against dEGFP (green) and cellular markers for astrocytes (GFAP; red) and endothelial cells (CD31; blue) in neocortex, subcortical white matter and periventricular zone. dEGFP expression is increased in astrocytes (arrows) and endothelial cells (arrow heads). E. Double staining against dEGFP and cellular markers for neurons (NeuN), microglia (Iba1) or oligodendrocytes (Oligo2) showed no co‐localizations. F. Antibodies against AQP4 show expression in astrocytes. Asterisks indicate the subarachnoid space. Scale bars are 50 µm.

DISCUSSION

Here, we have generated a novel transgenic reporter mouse and used it to study activity of NF‐κB in experimental models of perinatal complication. The transgenic construct contained the reporter genes luciferase and dEGFP for in vivo and cell‐specific examination, respectively.

We could easily image in a noninvasive fashion increased luciferase signals from the brain of live neonates in models of meningitis, sepsis and hypoxia–ischemia. Previous studies have demonstrated NF‐κB activation in the brain of neonates; however, little is known about the cell‐specific activation during encephalopathy 6, 63. In a histological study of traumatic brain injury in 9‐day‐old rats, increased staining against NF‐κB was observed in neurons and astrocytes; however, only in astrocytes this increased staining was confined to the nucleus (62). A caveat against interpretation of such histological results is the uncertainty whether nuclear staining represents actual NF‐κB activity, as transcriptional regulation by NF‐κB also depends on its various binding partners and posttranslational modifications 26, 57, 61. To determine cell‐specific activity, we assessed NF‐κB‐regulated dEGFP expression. The change in NF‐κB activity showed some variability between the different disease models we tested. We found strong activation in both endothelial cells and astrocytes in the models of meningitis and hypoxia–ischemia, while in systemic inflammation, only the endothelial cells displayed increased NF‐κB activity. However, we found no activation in neurons, oligodendrocytes or microglia. The difference between astrocytes and endothelial cells can be caused by the distinct initiation sites of injury, which for meningitis and hypoxia–ischemia are in the brain, while in systemic inflammation the proinflammatory mediators reach the brain via circulation (48).

The endothelial cells of the brain are an important constituent of the blood–brain barrier (BBB), and it has been shown that inflammatory processes can open the BBB's tight junction and contribute to brain edema 14, 23, 31. Indeed, the anti‐inflammatory drug dexamethasone is used to treat brain edema (35), and inhibition of NF‐κB is shown to reduce BBB permeability in brain endothelial cell cultures and in adult rats (70). Additionally, NF‐κB can reduce the amount of tight junctions between epithelial cells 47, 68. The increased NF‐κB activity we found in astrocytes was localized to the perivascular area. These astrocytes are intimately associated with endothelial cells of the BBB and take part in regulating its properties. Indeed, activation of NF‐κB in astrocytes, induces release of interleukin‐6 that can target endothelial cells with a subsequent increase in BBB permeability (64). In conclusion, it is pertinent to suggest that NF‐κB plays a central role in astrocytes to determine the integrity of the BBB through interaction with endothelial cells. In neonates, many of the same adverse changes have been observed during neuroinflammation, although a direct role of NF‐κB has not yet been demonstrated. However, as many typical target genes of NF‐κB are clearly involved in compromising the BBB in neonates, it is likely that NF‐κB plays an important role for BBB function both in adult and neonatal brains (66).

The lack of NF‐κB induction in neurons, oligodendrocytes and microglia may have various reasons. A methodological explanation could be that the genomic insertion site of the transgene is in a region with suppressed transcription caused by, for example, chromatin folding, epigenetic factors or endogenous cis‐elements. This would be in agreement with the large tissue variations in the activation levels of NF‐κB we observed between different founders. Furthermore, a transgene can unpredictably be influenced by other transcription factors with differential regulation between cell types (50). However, the great complexity of the NF‐κB signaling also enables more biologically relevant explanations 26, 57, 61. NF‐κB in an active form is a dimer composed from five different subunits in various combinations. The different combinations of NF‐κB dimers bind to a range of NF‐κB cis‐elements, with different affinity (57). We can not rule out activation of other types of NF‐κB dimers with low affinities for our reporter construct. Furthermore, NF‐κB is regulated by numerous intracellular signaling cascades, which show cell‐specific variations. Also, different cells are exposed to and responsive to various environmental factors. Together, these will cause cell specificity with respect to the types of NF‐κB isoforms activated 29, 34, 37, 55. The cis‐element used in our reporter construct binds strongly with the RelA/p50 heterodimer, but not with for example, RelA homodimers (30). It is commonly assumed that RelA/p50 is activated in neurons of the central nervous system (CNS) 24, 54, but this has been questioned by others (49). Mao and coworkers showed that after stimulating neurons with glutamate or LPS, only RelA homodimers were reliably detected in the nucleus (49). Thus, our failure to detect NF‐κB in neurons can be explained by the poor interaction between our chosen cis‐element and putative activated RelA homodimers. It has also been suggested that absence of NF‐κB activation in neurons has a protective function. Selective inhibition of NF‐κB in neurons of adults reduces lesion size after infarction 28, 77. Therefore, restricted NF‐κB activation in neurons could work as a defense mechanism against hypoxic‐ischemic lesions in the neonatal brain, which is known to be relatively protected from hypoxic‐ischemic insults (3).

To obtain co‐regulated expression of the reporter genes, luciferase and dEGFP, we chose a bidirectional promoter. There are various other strategies to coordinate gene expression such as internal ribosome entry site (IRES), multiple uses of similar promoters, self‐processing polyproteins and fusion proteins. IRES is inserted between the genes and widely used, but the expression efficiency of the genes on each side of the IRES vary according to the type of transgene and tissue; it is therefore relatively unpredictable 4, 27, 56. Multiple promoters are shown to cause mutual interference, which again is under influence by various biological factors 10, 11. In self‐processing polyproteins, the genes are intermitted by a small peptide, such as the 2A‐like peptide, which is split during translation (67). These proteins as well as fusion proteins require molecular engineering, which introduces sequence changes that can affect the expression, activity, stability and trafficking of the proteins 2, 13. Previously, a bidirectional promoter has, to our knowledge, not been used to express two reporter genes; however, it has been used in a transgenic mouse to co‐express a functional gene together with luciferase (43), and to drive coordinated expression of two genes in mice after transfer by a lentiviral vector (2). Furthermore, 10%–20% of the promoters in the mammalian genome are suggested to be bidirectional 1, 60, 71. When we transiently transfected our bidirectional construct into cell lines, both fluorescence from dEGFP and luciferase‐mediated luminescence were robustly induced by stimulation of NF‐κB. However, we only observed weak fluorescent signal from dEGFP in transgenic animals carrying the same construct. The destabilized version of EGFP we have used has a PEST sequence (rich in proline, glutamic acid, serine and threonine) fused to its C‐terminal end, which provides a proteolytic signal (46). As a result, dEGFP has a half‐life of 2 to 3 h as opposed to the ∼26 h it has without this modification (9). This is of importance in order to properly interpret the dynamic expression pattern of NF‐κB activation and it also correlates well with the 3‐h half‐life of luciferase (69). Additionally, in reporter mice, the promoters usually have some baseline level of expression. This is probably because the transcription factor of interests have activity beyond the experimental conditions examined and because the promoter is influenced by other transcription factors. As a consequence, reporter genes with long half‐life can accumulate inside the cells and potentially mask an induced response. However, a short half‐life also limits the maximum obtainable cytoplasmic concentration of dEGFP, which probably needs a concentration of around 1 µM to distinguish it from autofluorescent background (58). Thus, we believe that a combination of low expression and the rapid turnover of dEGFP is the reason for the difficulty to detect the fluorescent signal. As a consequence, while the luciferase signal was easily detected, the dEGFP signal needed amplification through antibody staining to obtain reliable results. Although direct visualization was the ambition, the model is indeed useful as antibody staining against luciferase in sections has proven difficult. We have ourselves tested several types of luciferase antibodies with unreliable results. This difficulty is also experienced by other researchers, for example, as noted by Nishikawa and Herschman (32).

In conclusion, our results demonstrate for the first time that NF‐κB is highly inducible in astrocytes and endothelial cells in the perinatal brain following complications that can lead to encephalopathy. Currently, hypothermia is the only established treatment option for perinatal hypoxia–ischemia, and it provides only moderate neuroprotection (21). Knowing that hypoxia–ischemia and infections leads to similar pathological outcome of encephalopathy, our data could indicate that NF‐κB is a point of convergence of the etiologies evolving from these two different pathological triggers. Thus, it is plausible that targeted therapeutic strategies against NF‐κB can be considered either alone or in combination with hypothermia treatment or other pharmacological agents. The transgenic reporter mouse to assessing NF‐κB activity will thus be valuable to validate such treatments. Using luciferase and dEGFP as co‐regulated reporters of NF‐κB activity, we can validate inflammation and effect of treatment both in a time‐dependent manner in live animals and in individual cells of tissue preparations.