1 INTRODUCTION

Excitotoxicity, related to ischaemic stroke, traumatic brain injury and seizures, occurs through over-activation of excitatory glutamate receptors. The glutamatergic receptors involved are those selectively activated by either N-methyl-D-aspartate (NMDA) or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA), and their excessive activation causes neuropathology.¹ Interestingly, glutamatergic activity can induce repair signalling. For instance, excitotoxic stimulation of NMDA receptors induces neuronal impairment and death, but in the developing brain these receptors play a role in neuronal survival.² In addition, related signalling avenues have been suggested to promote cell survival in different in vitro and in vivo models.^3-5 Thus, identifying competing genetic programmes is important to fully understand the influence of opposing pathways underlying the brain's response to injury.

The brain's response to an excitotoxic injury involves pathogenic and protective pathways, and clinical outcome is likely determined by the temporal relationships between such counteracting gene responses. Excitatory over-activation, in particular, activates such opposing pathways, with excitotoxicity causing transcriptional induction of both cell death- and cell survival-related genes.^{6, 7} Examples of differential gene expression during cerebral ischaemia and models of excitotoxic events were demonstrated by measuring brain RNA expression profiles.^7-10 Notably, temporal expression changes were identified in the transient middle cerebral artery occlusion (MCAO) stroke model by employing neurobiology-focussed arrays.⁷ The MCAO study provides strong evidence of distinct acute versus delayed gene expression profiles comprised of diverse functional categories.

The present study investigates the excitotoxic injury response in the hippocampus using the cultured hippocampal slice, a physiologically relevant three-dimensional tissue model with native pathogenic responsiveness with regard to neurodegenerative insults.^{3, 4, 11-13} Different excitotoxic insults in the explants indeed resulted in the expected delayed neuronal damage along with associated compensatory signals found involved in governing cellular damage.^{3, 4, 6} The hippocampus is particularly important to study as it is a higher brain centre involved in memory encoding and behavioural responses, as well as being distinctly vulnerable to excitotoxicity. Here, excitotoxicity was induced through activation of NMDA receptors and the associated transcriptional responses were assessed one hour (early) and 24 h (delayed) later with Affymetrix focussed arrays containing 1300 neurobiologically relevant sequences. The utilization of categorical filtering was an additional informatics step to identify transcriptional regulation events with high propensity of being involved in cellular degeneration versus cellular protection. Networks of co-expressed neurobiology-relevant genes were computationally generated and found to be in congruence with the temporal patterns observed using the experimental hippocampal slice cultures. Additionally, Gene Ontology annotations demonstrated unique functional networks involving the early and delayed expression profiles.

2 METHODS

3 RESULTS

For the study of transcriptional events elicited specifically in the hippocampus, we prepared rat hippocampal slices and maintained them in culture on Millipore inserts (Figure 1A). The explants displayed long-term maintenance of the major neuronal subfields of the hippocampus (Figure 1B) and their dense processes populated with evident staining of synaptic terminals (Figure 1C–E). Former studies reported that the slice cultures exhibit connectivity, plasticity and pathogenic responsiveness as found in vivo.^{11, 13, 50-52} Here, we used an NMDA infusion period that led to pathogenic changes as previously described.^{6, 51} The protocol for the excitotoxic stimulation and subsequent measuring of gene expression profiles is shown in Figure 1F, in which the explants were infused with NMDA for 20 min followed by rapid washout in the presence of glutamatergic antagonists for precise insult duration due to quenching any further excitatory activation.

To test for early versus delayed expression profiles after the defined excitotoxic insult, RNA was isolated from groups of slice cultures, cDNA was generated and used to synthesize biotinylated cRNA, which was followed by the chip hybridization protocol employing Affymetrix Rat Neurobiology U34 arrays. Across the approximately 1300 neurobiologically relevant sequences in the arrays, the scatter plot in Figure 1F indicates a close correlation of hybridization signal intensities between non-treated control samples. The linear distribution of expression intensities confirms reproducible measures of transcription levels.

Compared with the expression levels measured in vehicle-treated control slices, the NMDA insult led to 106 differentially expressed genes across the two post-insult times. The small set from the 1300 neurobiologically relevant genes met the criteria of at least a 50% increase or decrease in expression and represented a wide range of functionality. In Table 1, functional information about the 106 genes was used to classify them under several broad titles, consisting of cell cycle proteins, cytokines and inflammatory agents, transcription factors, chaperons and heat shock proteins, receptors and channel proteins, synaptic proteins, kinases and signal transduction proteins, growth hormones, transporters and neurotransmitters, adhesion molecules, and calcium homeostasis. Clones that could not be placed under any of the headings were designated as other. The differentially expressed genes evident 1-h vs. 24-h post-insult produced distinct heat maps (Figure 1G). The distinctions between the early and delayed transcriptional regulation events were further distinguished with dendrograms formulated through linkage clustering. To identify possible functional differences in the differentially expressed gene clusters, counts of functional categories in each cluster were completed and depicted as pie charts in Figure 1G. The pie charts express the per cent of regulated genes in each of the functional categories. The early transcriptional regulation events are distinct in that they are clustered in the cytokine and transcription factor gene groups, whereas the delayed events were mostly found outside those two classifications. The largest up-regulation events included 11- to 50-fold increases in four cytokines at 1-h post-insult, as well as early 13- to 24-fold changes to two members of the transcription factor group (NGFI-B and fos-related antigen Fra-1). The largest delayed increase in expression was a 4.9-fold change to the chemokine monocyte chemoattractant protein (MCP).

Further profiles were constructed in order to test for distinct patterns of differentially expressed genes in response to the excitotoxic insult, as well as for those specific genes exhibiting unique transcriptional changes. Up-regulation events at 1-h post-insult consisted of a distinct set of 26 transcripts (Figure 2A, red bars on left), and among these, only five exhibited statistically significant changes in expression 24 h after the NMDA exposure (right bars). Note that using the same order of genes listed for the 1-h data, the 24-h post-insult responses consist mostly of grey bars determined to be outside statistical significance, along with two biphasic genes exhibiting delayed down-regulation (green bars) and three genes with continued up-regulation (red bars). The latter three exhibiting early and delayed up-regulation were MCP, up-regulated 5.81 ± 0.36- and 4.92 ± 2.46-fold (mean ± SEM), the transcription factor C/EBP up-regulated 2.98 ± 0.25- and 2.08 ± 0.23-fold and haem oxygenase-1 (HO-1 or HSP32) up-regulated 2.66 ± 1.16- and 2.23 ± 0.30-fold, respectively.

At 24-h post-insult, a new pattern of 20 up-regulated genes was found (Figure 2B), and this delayed pattern shares only MCP, C/EBP and HO-1 with the profile of early up-regulated genes. The remaining 17 genes in the delayed pattern were found to be statistically unchanged with regard to their early expression alterations. Thus, two distinct groups of genes with very little overlap exhibit increased expression levels 1 h vs. 24 h after the excitotoxic insult. The 24-h post-insult time was further distinctive by having 63 genes down-regulated exclusively in a delayed manner (Figure 2C). The behaviour of these 63 genes at the early time found that two of the genes, early response factor 1 (Egr-1) and the neuron-derived orphan receptor-1 (NOR-1)—both transcription factors—displayed biphasic responses exhibited by early induction followed by delayed suppression. Egr-1 displayed dramatic induction 1 h after the excitotoxic insult (eightfold), but that regulated level of expression was suppressed 95% in hippocampal slices harvested 24 h after NMDA treatment, reaching a level 58% below the baseline of vehicle-treated slice cultures (0.42 ± 0.12 of control expression). NOR-1’s early induction in the excitotoxic slice model was 4.42 ± 0.82-fold after 1 h, followed by a 93% reduction in expression by 24 h (0.30 ± 0.05 of control). To verify these unique responses, qRT-PCR analyses were conducted resulting in confirmation of biphasic transcriptional responses by Egr-1 (early vs. delayed Log2 of 2.0 vs. −1.3) and NOR-1 (1.9 vs. −0.82).

Interestingly, it should be pointed out that the excitotoxic hippocampal slice model was without any gene expression changes that met the minimal criteria of ≥50% decline at 1-h post-insult. Among the delayed down-regulation events, 9 of the 63 genes were markedly reduced in expression by 89%–95%. The nine genes represent mostly synaptic proteins or proteins linked to synaptic mechanisms. Among the synaptic proteins, GluR-K3 expression was reduced by 94% and synapsin 2 was reduced by 89% (see Table 1). SNAP 25B expression was reduced by 90% at 24-h post-insult and by 49% at the 1-h time point, the early regulation being just shy of the filtering criteria. It should be noted that the related protein SNAP-25A exhibited a delayed reduction in expression of 80%. In addition, Ca²⁺/calmodulin-dependent protein kinase IIβ (CaMKIIβ) was reduced by 95%, neurogranin (RC3) by 90% and brain-derived neurotrophic factor (BDNF) by 91% at 24-h post-insult.

To identify networks of differentially expressed genes involved in early versus delayed responses, a gene coexpression network was generated using GeneMania. To test for differentially expressed gene networks involved in the early vs. delayed response to excitotoxicity, we created a single-gene coexpression network using GeneMania,¹⁵ where each of the 103 differentially expressed genes is represented by a node. Edges denote evidence of gene coexpression found in previous studies in NCBI Gene Expression Omnibus via GeneMania. In this way, we informed our neuro-specific gene expression network with prior gene expression studies but the network was not biased by cell or tissue type. As shown in Figure 3A, the generated network clustered into two distinct sets based on the gene coexpression data from the two post-insult assessments. We then overlaid our early and late gene expression data onto our coexpression network. Interestingly, we found that our temporal gene profiles mirrored the coexpression network clustering, suggesting our insult-induced neuro-specific temporal gene expression pattern may be part of a more basic stress response pattern, which has been observed across a variety of gene expression studies. Distinct functional networks associated with the temporal profiles were also evident when mapping the genes to biological function using Gene Ontology databases. For example, as compared to early changes in biological function, delayed changes included cytoskeleton organization and neurotransmitter and ion transport, suggesting repair and energy-dependent processes were taking place.

Both pathogenic and repair responses have been linked to excitotoxicity in the brain; therefore, categorical filtering was applied accordingly to the 106 differentially expressed genes. Over 40 of the 106 NMDA-induced genes are commonly categorized as either primarily involved in pathology or primarily in cellular repair/survival. By assessing such categorical distinction across genes influenced at the two post-insult times, discrete patterns of degenerative vs. repair events were identified (Table 2). The early degenerative signalling consists of 15 up-regulated pathogenic genes, 9 of them being increased 3- to 27-fold 1 h after the NMDA insult (MCP, C/EBP, NOR-1, MIP-1α, Fra-1, IL-1β, NGFI-B, TNF-α and MIP-2) while CREM, cJun, IRF-I, PDE4, ICAM-1 and tPA were up-regulated to a smaller degree (1.58- to 1.74-fold). Of these 15 genes, only MCP and C/EBP exhibited enhancement of expression at the 24-h post-insult time as well, being among a small group of up-regulated pathogenic genes that also included caspase 1, calpain II and bax. Also illustrated in Table 2, early degenerative changes appear to consist only of increased pathogenic genes, whereas delayed degenerative changes included increased pathogenic genes as well as down-regulated expression of a putative set of survival genes negatively influenced by the excitotoxic stimulation (Egr-1, neuro-D4, trkB, PKCβ, neuronatin α, PKCγ, BDNF, IP3K and ILK). The qRT-PCR analyses confirmed many transcriptional responses identified in the microarray data including the early, marked induction of NOR-1, IL-1β and Egr-1, and the delayed down-regulation of trkB and neuronatin α using specific primers.

The categorical filtering also found evidence for opposing pathways. Early responses to the NMDA insult consisted of degenerative increases in the 15 primarily pathogenic genes listed above as well as protective increases in eight primarily pro-survival genes (see Table 2). The 24-h post-insult findings were more complicated but still involved opposing pathways. The delayed responses to the excitotoxic episode involved both degenerative changes (up-regulated pathogenic genes and down-regulated survival genes) and protective changes (up-regulation of 6 survival genes and down-regulation of three pathogenic genes). Among the 14 up-regulated survival genes, only HO-1 exhibited sustained enhancement, having increased expression at the early and delayed time points. Regarding the two biphasic responding genes, Egr-1 expressed an early protective response followed by a delayed degenerative event of down-regulation. In contrast, the pathogenic gene NOR-1 expressed an early degenerative response followed by delayed down-regulation.

The degenerative and protective gene differences were subsequently used to assess networks involved in the distinct pathways. Starting from the gene coexpression networks generated in Figure 3A, all nodes not categorized as pathogenic and/or protective based on literature searches were removed. Next, the remaining nodes were then annotated based on four functional categories (from Table 1): (i) transcription factors, (ii) cytokines, (iii) receptors and channels, and (iv) kinases and transduction, or as ‘other.’ As a result, diagrammed in Figure 3B, gene expression data for pathogenic and protective genes at one and 24 h post-insult were overlaid onto the coexpression networks to create four networks of differential gene responses. At 1 h post-insult, both protective and pathogenic genes were up-regulated and pathogenic gene expression clustering matched that of the coexpression network. In the protective early gene expression network, clustering was less apparent and a wide variety of gene functional categories were observed. At 24 h post-insult, protective and pathogenic genes were both up- and down-regulated and temporal gene expression data mirrored coexpression clustering.

To further understand how the temporal changes at a gene expression level might impact system behaviour, functional networks involving the early versus delayed differentially expressed genes were generated by mapping the gene name to biological function using Gene Ontology (GO). Networks were subsequently created using Cytoscape, where each node represents a GO biological function category (Figure 4). Node sizes represent the number of genes within a given GO category while edge sizes denote the number of genes shared between the category nodes.

To identify a candidate signalling avenue activated by NMDA and involved in opposing gene responses, we tested the role of NF-κB which also served to test the value of the signalling profiles established in the present study. NF-κB was selected for testing due to being activated by synaptic stimulation,^{6, 53} and this transcription factor has been implicated in both cell survival and cell death.^{6, 54} A set of 22 opposing genes were first assessed for whether they possess an NF-κB-binding site in their promoter region. In the second tier of assessment, the excitotoxic NMDA treatment was applied to hippocampal slices that were also treated with an inhibitor of NF-κB activity (MG-132) before, during, and after the NMDA infusion to identify those genes linked to NF-κB. MG-132 was confirmed to block the distinct early and delayed phases of NMDA-induced NF-κB activation (see Figure 5A). The resulting data listed in Table 3 suggest that NF-κB is responsible for a subset of opposing genes involved in the brain's response to an excitotoxic episode. The early inductions of IL-1β, IRF-I and TNF-α expression were blocked by the NF-κB inhibitor, and these pathogenic genes indeed possess a regulatory site for the transcription factor (see rows in bold in Table 3A).

Besides the degenerative role of NF-κB regulating the three aforementioned pathogenic genes, NF-κB also has an apparent regulatory role with an opposing pro-survival gene, HES-1, early after the NMDA insult (Table 3B). Note that three other survival genes possess an NF-κB binding site (Egr-1, IL-6 and SOCS-3), however, the MG-132 inhibitor had no effect on their NMDA-induced expression. Of the pathogenic genes induced at 24 h post-insult, MCP, caspase 1 and cJun have an NF-κB-binding site but blocking NF-κB activity only disrupted the enhanced expression of caspase 1 (Table 3C). Also at 24 h, two NMDA-induced survival genes with NF-κB-binding sites were blocked by the NF-κB inhibitor (JAK2 and IGFII), whereas three others with binding sites (HSP27, HSP10 and IGFII BP3) did not have their regulated expression blocked by MG-132 (Table 3D). These data provide an example of determining induced transcriptional responses in order to build an understanding of how a single regulator can influence opposing pathways. Illustrated in Figure 5B are the potential roles NF-κB has on protective and pathogenic responses in the excitotoxic hippocampus. Note that, based on promoter binding sites, NF-κB’s regulation of other transcription factors may connect NF-κB indirectly to the early induction of NOR-1 (via HES-1 and/or IRF-I), Egr-1 (via HES-1) and IL-1β (via IRF-I) — the IL-1β induction also has a direct connection to NF-κB based on the MG-132 results. Interestingly, results with the MG-132 inhibitor also suggest indirect effects that may contribute to the delayed down-regulation of BDNF and Egr-1, with the latter having the potential to influence NOR-1 and IP3K as well.

4 DISCUSSION

Over-activation of glutamatergic synapses recapitulates many aspects of brain disorders linked to excitotoxicity. The present study utilized a defined over-activity period for NMDA-type glutamatergic receptors and identified early and delayed expression signatures in a vulnerable brain region, the hippocampus. The distinct signatures act together to make up the brain's response to cellular responses that occur early after an excitotoxic insult or in a delayed manner. Informatics methods were combined with the cultured hippocampal tissue model, thereby eliminating systemic variables and allowing the study to focus on the types of cellular pathways activated. In addition, the use focussed gene array analyses resulted in transcriptional responses representing neurobiology-relevant pathways. The expression profiles appear to consist of competing genetic programmes, and the distinct programmes can be used to evaluate opposing pathways that make up the brain's response to injury.

The pattern of transcriptional responses found early after the excitotoxic stimulation was distinct as compared to delayed responses. Most dramatic changes one hour post-insult were twofold-fivefold up-regulated levels of cytokines and transcription factors involved in inflammatory responses, cell proliferation and apoptosis. Many of the induced signalling components indicate that competing pathways take part in the brain tissue's response to injury.

Adding to the many comprehensive lists of regulated genes from models of excitotoxic insults, this report will help advance understanding of the brain's response to hypoxic/ischaemic episodes that are linked to excitotoxicity. The identified responses to excitotoxic stimulation involved genes with a wide range of functionality. The differentially expressed genes and their classified functional networks vary widely between the early activated gene regulation events and the delayed groups. As a result, distinct patterns of cellular responses were found early after the brief NMDA insult as compared to responses identified 24 hours later. Dramatic changes one hour post-insult included twofold to fivefold up-regulated message levels for cytokines and transcription factors involved in inflammation, cell proliferation and apoptosis, suggesting there are competing pathways at the early post-insult period and/or the presence of different cell types having opposing response signatures. Further, twofold to fourfold changes 24 hours later also exhibited opposing directions and included the down-regulation of kinases and transduction molecules.

The delayed down-regulation of cell signalling processes is clearly indicated in the NMDA insult model used, perhaps as part of compensatory pathways for protection/repair that require sufficient time for a cascade of cellular events and crosstalk.^7-9 Gene ontology analysis also showed neurotransmitter transport and other terms appearing only in the delayed networks. As compared to the early changes in biological function nodes, delayed changes exhibited the inclusion of nodes for ion transport, cytoskeleton organization and neurotransmitter transport, thus suggesting repair/compensatory responses and energy-dependent processes were activated. Interestingly, in addition to the wide range of functional categories among NMDA-induced up- and down-regulated genes, long non-coding RNAs (lncRNAs) assessed in a young rat model of lithium/pilocarpine-induced status epilepticus were found to exhibit up- and down-regulation for wide functional influence.⁵⁵ The modified lncRNAs identified were found to influence the regulation of genes involved in cell proliferation, inflammation, angiogenesis and autophagy.

Findings from the hippocampal explants highlight the brain tissue's complex responses to an excitotoxic insult, which involve opposing pathways and networks. The degree of pathology produced by over-activated neurons is determined by both the induced degenerative pathways and the induced compensatory signalling as part of repair mechanisms. With the combination of informatics methods and categorical filtering, the current study found evidence for opposing profiles, and the counteracting pathways may provide distinguishable features between different stages of excitotoxic damage. Early changes in gene expression show that opposing pathways occur soon after an excitotoxic insult, in which degenerative activation of pathogenic genes occurs at the same time protective activation ensues regarding opposing survival genes. Of course, opposing pathways may stem from distinct areas of the hippocampal circuitry with distinctly responsive connectivity. While some circuits may respond to intense neuronal activity by initiating degenerative signals, other over-activated circuits may trigger repair signalling. The present study supports the idea that the brain is not a passive recipient of pathogenic insults and the induction of degenerative processes, but rather, the brain can trigger pathways of cellular repair that counter pathological responses to injury.

Note that opposing proapoptotic and antiapoptotic genes were reported to be induced in the hippocampus of rats subjected to global cerebral ischaemia⁵⁶ as well as in hippocampal explants exposed to a defined NMDA exposure⁶ as used in the present study. Excitotoxicity through the activation of NMDA receptors shares many of the cellular cascades associated with a variety of brain injuries. The utilization of categorical filtering was an additional informatics step used to identify transcriptional regulation events with a high propensity of being involved in cellular degeneration vs. cellular protection. Further analyses in the current study found that the potent transcription factor NF-κB appears to play a role, at least in part, in both of the pathogenic and protective expression signatures identified in the excitotoxic hippocampus. Such also supports the assertion of improved profile effectiveness by including a categorical informatics step to identify candidate signalling elements, in this case one potentially responsible for opposing gene responses.

NF-κB has often been linked to neuronal survival, but with many studies also linking it to excitotoxic pathology.¹ In addition to regulating responses for a wide array of events including cellular proliferation, inflammation and tissue remodelling, NF-κB signalling facilitates both protection avenues in surviving cells and cell death pathways for the clearance of dying cells. Such a dynamic and dual role is likely important, due to the complexity of cellular decisions, to allow crosstalk between divergent pathways for cells to adapt to stress, repair injuries and foster tissue health. NF-κB is activated by synaptic over-stimulation and it represents signalling reported to underlie complex and often opposing responses,^{6, 54, 57, 58} with links to the MAPK pathway during biphasic gene regulation in surviving tissue.⁸ The transcription factor also frequently responds to excitotoxic insults that have an early phase of cytoskeletal and synaptic compromise.^{50, 51, 59} Interestingly, NMDA exposure in the current study activated pathogenic as well as protective genes in correspondence with putative transcription programmes linked to biphasic NF-κB signalling. As previously noted,^{50, 60} excitotoxic NMDA receptor stimulation is a pathogenic event shared by different neurological disorders with wide ranging aetiologies, including, strokes, ischaemia and brain injuries. In addition, alterations to the receptors may underlie the early degenerative processes in Alzheimer's disease and the memory impairment in ageing. Many of the genes identified that converge on NF-κB have been associated with the variety of brain disorders, and perhaps the biphasic action of this transcription regulator underlies influential expression signatures stemming from neuronal over-stimulation. NMDA-mediated bidirectional regulation of cellular responses may also involve the distinct functions of synaptic vs. extrasynaptic NMDA receptors, an issue being addressed at the level of receptor subunit composition in the two neuronal locations.⁶¹ Thus, the indication that opposing pathways converge on a single response element may involve (i) different patterns of temporal regulation, (ii) distinct subcellular locals of the activated pathways or (iii) distinct receptor subtypes triggering the transcriptional induction. Our study identifies transcription profiles and functional networks to begin to understand the combinations of the above events that lead to dual influences on cellular survival and death. The usefulness of the informatics data includes constructing putative signalling networks underlying the opposing directions of cellular fate in order to find avenues to offset clinical outcomes.

CONFLICT OF INTEREST

Dr. Bahr has patents related to neuroprotection and protease regulation (US Patents 8,163,953 and 10,702,571), but they do not involve information in the present study. The remaining authors have no conflicts of interest to declare.

AUTHOR CONTRIBUTIONS

Ebru Caba: Conceptualization (equal); Data curation (equal); Formal analysis (supporting); Investigation (equal); Methodology (equal); Writing-original draft (equal). Marcus D. Sherman: Data curation (equal); Formal analysis (equal); Methodology (equal). Karen L. G. Farizatto: Data curation (equal); Formal analysis (equal); Methodology (equal). Britney Alcira: Data curation (supporting); Investigation (supporting). Hsin-wei Wang: Data curation (equal); Formal analysis (equal). Charles Giardina: Methodology (equal); Supervision (equal). Dong-Guk Shin: Conceptualization (equal); Investigation (equal); Methodology (equal); Supervision (equal). Conner I. Sandefur: Data curation (equal); Formal analysis (equal); Methodology (equal); Supervision (equal). Ben A. Bahr: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing-review & editing (equal).