After traumatic brain injury, memory dysfunction is due in part to damage to the hippocampus. To study the molecular mechanisms of this selective vulnerability, the authors used laser capture microdissection of neurons stained with Fluoro-Jade to directly compare gene expression in injured (Fluoro-Jade-positive) and adjacent uninjured (Fluoro-Jade-negative) rat hippocampal neurons after traumatic brain injury and traumatic brain injury plus hemorrhagic hypotension.
Twelve isoflurane-anesthetized Sprague-Dawley rats underwent moderate (2.0 atm) fluid percussion traumatic brain injury followed by either normotension or hemorrhagic hypotension. Animals were killed 24 h after injury. Frozen brain sections were double stained with 1% cresyl violet and 0.001% Fluoro-Jade. RNA from 10 Fluoro-Jade-positive neurons and 10 Fluoro-Jade-negative neurons, obtained from the hippocampal CA1, CA3, and dentate gyrus subfields using laser capture microdissection, was linearly amplified and analyzed by quantitative ribonuclease protection assay for nine neuroprotective and apoptosis-related genes.
In injured CA3 neurons, expression of the neuroprotective genes glutathione peroxidase 1, heme oxygenase 1, and brain-derived neurotrophic factor was significantly decreased compared with that of adjacent uninjured neurons. Superimposition of hemorrhagic hypotension was associated with down-regulation of neuroprotective genes in both injured and uninjured neurons of all subregions. Expression of apoptosis-related genes did not vary between injured and uninjured neurons, with or without superimposed hemorrhage.
The authors show, in the first direct comparison of messenger RNA levels in injured and uninjured hippocampal neurons, that injured neurons express lower levels of neuroprotective genes than adjacent uninjured neurons.
PARASAGITTAL, fluid percussion traumatic brain injury (TBI) in rats produces a combination of focal cortical contusion and diffuse subcortical injury1–4that replicates several clinically relevant features of human TBI. In rats, fluid percussion TBI causes memory dysfunction and scattered loss of neurons in the hippocampus, most prominently in the CA3 subfield.4–6In patients surviving TBI, memory dysfunction, likely due to hippocampal damage, is a frequent, disabling complication. After clinical TBI, secondary injury due to hypotension, which frequently accompanies head injury, markedly reduces the likelihood of favorable neurologic outcome.7
We previously reported that enhanced expression of neuroprotective genes was associated with the resistance of hippocampal subfields to injury in an in vivo model of TBI in rats8and that chelation of neurotoxic zinc both increased expression of neuroprotective genes and reduced neuronal cell death in rat brains after TBI.9Despite extensive study of the molecular mechanisms of neuronal cell death in animal models of TBI, no neuroprotective treatment has improved outcome in clinical trials of human brain injury.10Because many of these treatments involved inhibition of signaling components involved in cell death and degeneration, these failures suggest that alternative therapeutic strategies are needed.
Recent studies of the mechanisms of cancer resistance have suggested that a viable therapeutic strategy would be to boost inhibitors of angiogenesis that occur naturally as part of the endogenous defense against tumor formation.11Similarly, a viable therapeutic strategy for patients with TBI could be to augment endogenous neuroprotective responses elicited by brain injury. To that end, in this study, we sought to pursue the identification of the molecular factors (genes) that potentially contribute to injury resistance and survival after TBI. We hypothesized that this could be addressed by examining the effects of TBI on a homogeneous population of brain cells that received an identical injury (same injury level, same time), provided that one could distinguish an injured, degenerating cell from an uninjured cell.
As noted above, experimental TBI causes a distinctive pattern of scattered hippocampal neuronal death that preferentially affects CA3 and the dentate gyrus.12,13Because neuronal injury is aggravated by post-TBI hemorrhagic hypotension,14,15we used Fluoro-Jade, a polyanionic fluorescein derivative that sensitively binds to and stains irreversibly injured, degenerating neurons,16,17to distinguish injured from uninjured hippocampal neurons after TBI or TBI combined with severe hemorrhagic hypotension (TBI/HH). We then used laser capture microdissection (LCM) to obtain 10 injured and 10 uninjured neurons, identified by Nissl staining with cresyl violet, from the CA1, CA3, and dentate gyrus (DG) subregions of each rat hippocampus. We hypothesized that, in morphologically similar hippocampal neurons, those that were uninjured after TBI would express higher levels of neuroprotective gene expression than adjacent, injured neurons. Although one previous study of Fluoro-Jade–positive degenerating neurons in the rat hippocampus examined expression of the nuclear factor κB p50 subunit in Fluoro-Jade–positive CA1 pyramidal neurons by immunostaining,18ours is the first study to examine gene expression in Fluoro-Jade–positive, laser-captured hippocampal neurons. For the current study, we selected a representative group of genes whose expressions we had previously found to be altered by TBI9or were known from previous studies to be altered by TBI.19,20
Materials and Methods
All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch, Galveston, Texas. Moderate parasagittal fluid percussion TBI was performed in 12 isoflurane-anesthetized rats as previously described.21Rats were then randomly assigned to a normotensive group or a hemorrhagic hypotension group, in which was hypotension induced by withdrawing sufficient blood to reduce mean arterial blood pressure to 40 mmHg for 45 min, after which shed blood was returned and animals were permitted to recover from anesthesia. Because in a previous microarray study9we observed the greatest number of changes in gene expression 24 h after injury, we chose that time interval for the current study. Therefore, 24 h after TBI, rats were anesthetized with 4% isoflurane and decapitated, and their brains were rapidly removed and frozen on dry ice. Brains were stored at −80°C until they were embedded in OCT (optimal cutting temperature) medium for sectioning on a cryostat.
LCM and Linear mRNA Amplification
Frozen 10-μm coronal brain sections containing both the ipsilateral and the contralateral hippocampus (3.6–4.5 mm posterior to bregma) were collected on clean uncoated ribonuclease (RNase)-free slides (Fisher Scientific International, Hampton, NH) and kept at −20°C until they were stained. Slides were thawed for 30 s, fixed in 75% ethanol (1 min), rinsed in RNase-free water (1 min), stained with 1% cresyl violet (15 s), rinsed twice with RNase-free water (30 s each), stained for 4 min with 0.001% Fluoro-Jade (diluted 1:10 in 0.1% acetic acid from 0.01% stock solution made with RNase-free water) in a light-tight container, rinsed three times in RNase-free water (1 min each), dehydrated in 95% and 100% ethanol (30 s each) and twice in xylene (3 min each), air-dried in a hood for 10–15 min, and stored with desiccant in a closed slide box until LCM was performed. Neurons were identified by staining with cresyl violet, a Nissl stain that stains only neurons and not glia and that is compatible with preservation of RNA integrity in laser microdissected cells. Fluoro-Jade has a green iridescence with an excitation peak at 480 nm and an emission peak at 525 nm and is visualized using a fluorescein/fluorescein isothiocyanate filter. Adjacent Fluoro-Jade positive and Fluoro-Jade negative cells were only obtained from the ipsilateral (below fluid percussion injury site) hippocampus because there were virtually no injured cells in the contralateral hippocampus of any of the 12 injured rats. Ten Fluoro-Jade–positive neurons and 10 Fluoro-Jade–negative neurons each from each hippocampal subregion (CA1, CA3, and DG) were captured on separate CapSure HS caps (Arcturus Engineering, Mountain View, CA) by LCM using a PixCell II laser capture microscope (Arcturus Engineering). Using an RNAqueous Micro Kit and MessageAmp Kits (both from Ambion, Austin, TX), we performed total RNA isolation and linear amplification of messenger RNA (mRNA) (as previously described8) with T7 RNA polymerase, which reproducibly amplifies limited quantities of mRNA while maintaining proportional representation of gene expression.22
Ribonuclease Protection Assay, Northern Blots, and Real-time PCR Analyses
To compare mRNA expressions quantitatively, we used ribonuclease protection assay (RPA) analysis, a simple and linear assay sensitive enough to detect even low-abundance transcripts. RPA analysis was performed with 100 ng linearly amplified mRNA samples as previously described8using the HybSpeed RPA kit (Ambion). Probes for RPA and Northern blots were cloned using reverse-transcription polymerase chain reaction (PCR) as previously described.8Probes for RPA included consensus promoter sequences for T3 (sense strand) and T7 (antisense strand) RNA polymerase. Using reverse-transcription PCR, we cloned nine genes that are neuroprotective (glutathione peroxidase 1 [GPX-1], heme oxygenase 1 [HO-1], brain-derived neurotrophic factor [BDNF], heat shock protein 70 [HSP70]) or proapoptotic or antiapoptotic (Bcl-2, caspase 3, caspase 9, FAS, interleukin 1β[IL-1β]) and prepared two multiprobe sets of sense riboprobes from these templates to hybridize to the antisense amplified mRNA (aRNA) from the neuronal cell samples. We previously reported that one round of linear amplification with T7 RNA polymerase yielded sufficient aRNA (typically 400–1,500 ng) from 600–800 neurons to conduct several RPA experiments.8Subsequently, we demonstrated equivalent detection of a moderately expressed gene BDNF using 100 ng amplified mRNA from 5, 10, or 25 hippocampal neurons in an RPA analysis with a sense BDNF riboprobe. Consequently, we chose to study 10 injured neurons and 10 uninjured neurons from the three subfields of each rat hippocampus. In previous experiments in which we examined the differences in gene expression between young and aged rat brains with and without TBI, we found that normalizing mRNA levels to β-actin or glyceraldehyde-3-phosphate dehydrogenase (G3PDH) or both gave the same results.8β-Actin has been shown to be unchanged after injury in the rat retina,23after fluid percussion TBI in rat brain,24and after cerebral ischemia in mice.25Because these two commonly accepted “housekeeping genes” are completely different and serve different functions, we concluded that using G3PDH as an internal control was justified in our experimental paradigm. Further, a recent study of the use of housekeeping genes in quantitative gene expression studies26showed that genes such as G3PDH are good single housekeeping genes for any given tissue type. Therefore, in the current study, we used G3PDH to normalize mRNA expression because we were comparing mRNA levels in adjacent neurons of homogeneous populations of hippocampal cells and the size of the protected fragment fit more appropriately with the multiprobe gene sets.
The first sense, multiprobe, riboprobe set was composed of GPX-1 (GenBank acc# ×12367), BDNF (GenBank acc# NM_012513), HO-1 (GenBank acc# J02722), IL-1β (GenBank acc# NM_031512), and G3PDH (GenBank acc# NM_017008); the second set consisted of caspase 3 (GenBank acc# NM_012922), caspase 9 (GenBank acc# NM_031632), FAS ligand (GenBank acc# U03470), Bcl-2 (GenBank acc# L14680), HSP70 (GenBank acc# Z27118), and G3PDH. Quantitative real-time PCR with rat BDNF primers27was performed with the Applied Biosystems 5700 thermal cycler (Foster City, CA) and Applied Biosystems kits and reagents, according to the manufacturer’s protocols, using pooled uninjured CA3 aRNA and pooled injured CA3 aRNA from all six TBI rats. Reaction conditions for first strand synthesis were 25°C for 10 min, 48°C for 30 min, and 95°C for 5 min. The 50-cycle, two-step PCR conditions were 95°C for 10 min followed by 50 cycles of 15 s at 94°C and 1 min at 60°C. BDNF expression was normalized to protein gene product (PGP9.5). The protein gene product 9.5 primer sequences were forward 5′-CCTGGGCTCAGTGCCA-TCT-3′ and reverse 5′-CCTCAATTTGTTTTTTCCTGAA-GTTT-3′, and the probe was 5′-ATTCTCTGCATCCGTCCTCACCGG-3′. Northern blot analysis of injured and uninjured CA3 aRNA samples (pooled samples from six TBI rats) was performed according to standard protocols and as previously described8using a random primer [α-32P]deoxycytidine triphosphate–labeled GPX-1 probe. RPA and Northern blots were exposed to phosphoimaging screens, scanned on a Cyclone PhosphorImager (Perkin Elmer Life Sciences, Downers Grove IL), analyzed with OptiQuant software (Perkin Elmer), and imported into an Excel (Microsoft, Redmond, WA) spreadsheet for further analysis.
As a result of heterogeneous variances, levels of mRNA were logarithmically transformed (original normalized data and log-transformed data are shown in tables 1 and 2). The transformed data were analyzed using analysis of variance for a split-plot design with a main unit factor of treatment (TBI or TBI/HH) and subunit factors of region (CA1, CA3, and DG) and injury (uninjured and injured) (PROC MIXED in SAS®, release 8.2; SAS Institute Inc., Cary NC).28Main effects and interactions were assessed at the 0.05 level of significance. The Fisher least significant difference procedure with Bonferroni adjustment was used for multiple comparisons.
Coronal sections containing the ipsilateral and contralateral hippocampi of all experimental rat brains were all double-stained with Fluoro-Jade and cresyl violet. We first examined the sections under bright field illumination, which clearly showed the cresyl violet–stained neurons, and after we found the hippocampal neurons, we switched to the fluorescein isothiocyanate filter to find and visualize the Fluoro-Jade–positive neurons and perform the laser capture. We examined the ipsilateral and contralateral hippocampal regions of all sections of all experimentally injured rat brains after double staining with Fluoro-Jade and cresyl violet, and we did not find any injured, Fluoro-Jade–positive neurons in the contralateral hippocampus of any of the injured rats (fig. 1A). In the ipsilateral hippocampus, 24 h after TBI, we found a consistent and distinctive distribution of Fluoro-Jade–positive neurons in the CA1 and CA3 subregions (fig. 1B), as well as scattered positive neurons in the DG and dentate hilus (fig. 1C). These results confirm previous reports of Fluoro-Jade staining in TBI animals.17Fluoro-Jade staining has not been used previously to assess the influence of adding a secondary injury, such as hemorrhagic hypotension, after TBI. In TBI/HH rats, we consistently observed more Fluoro-Jade–positive neurons in all three subregions (fig. 1D), especially the CA1 subregion, which supports previous observations that the CA1 subfield is vulnerable to ischemic injury.29Because our study focused on differences in gene expression between adjacent injured and uninjured neurons, we did not quantify differences in the numbers of Fluoro-Jade–positive neurons.
Using LCM, we obtained 10 injured neurons and 10 immediately adjacent uninjured neurons from each subregion (CA1, CA3, and DG) for comparative molecular analysis (figs. 1E and F). In preliminary experiments using a sense BDNF riboprobe, we determined that the amplified RNA signals from 5, 10, and 25 neurons were equivalent (fig. 2A). After TBI and TBI/HH, we found prominent regional differences in gene expression in injured and uninjured neurons. In hippocampal subregion CA3 of the TBI group, the mean mRNA levels of the neurotrophin BDNF and the antioxidants GPX-1 and HO-1 in uninjured cells were significantly higher than in adjacent injured cells (P < 0.009; figs. 2B[confirmed by Northern Blot and real-time PCR analysis, Figs. 2C,D] and 3A–C). Although there were no differences between injured and uninjured neurons of the CA1 and DG subfields in TBI rats, the mean mRNA levels of BDNF, GPX-1, and HO-1 in the TBI/HH rat neurons (all subregions) were significantly lower than in TBI rats alone (P ≤ 0.0005). Expression of IL-1β, HSP70, caspase 3, and Bcl-2 was not significantly different in injured versus uninjured neurons (all subregions), but mean mRNA levels of all four genes in the TBI/HH group were significantly lower than in the TBI group (P ≤ 0.0001; figs. 3D–G). FAS ligand and caspase 9 were similarly expressed in injured and uninjured neurons in all hippocampal regions in both experimental groups (figs. 3H–I).
To confirm the results obtained using RPA, we performed a limited set of experiments using Northern Blot analysis and reverse-transcription PCR. Using Northern Blot analysis of pooled aRNA from uninjured and injured neurons (amplified from 60 CA3 pyramidal neurons from six TBI rats), we confirmed substantially higher expression of GPX-1 in uninjured neurons (fig. 2C). Real-time PCR analysis of BDNF expression in injured and uninjured CA3 neurons confirmed higher levels (4.4-fold) of BDNF in the uninjured cells (fig. 2D).
Although two previous studies compared gene expression in injured neurons (identified by staining with terminal deoxynucleotidyl transferase–mediated biotinylated deoxyuridine triphosphate nick end labeling) from rats after TBI to uninjured neurons from control animals that had not undergone TBI,30ours is the first study to directly compare mRNA levels in injured and adjacent uninjured neurons from the same animals. Ours is also the first study to use samples obtained using LCM (which is a rapid, precise technique for isolating single cells or populations of cells31) to directly compare gene expression in injured and uninjured neurons after TBI.
With the select group of genes examined in this study, there were significant differences in neuroprotective gene expression between injured and uninjured neurons, particularly in the CA3 region, which has been shown to be selectively vulnerable to fluid percussion injury.6On the other hand, we found a striking lack of detectable differences in proapoptotic gene expression in either group. We interpret this unexpected result with caution because we examined only two members of the proapoptotic gene family and there is no reason, a priori , to assume that the expression of all apoptosis-related genes would be altered by TBI. Furthermore, a recent study showed that some apoptosis-related genes are also critical for cell survival; e.g. , in an in vitro model of ischemic preconditioning, neuroprotection from a severe injury seemed to depend on sublethal activation of caspase 3 expression and subsequent up-regulation of downstream neuroprotective pathways.32Future studies involving comprehensive global gene chip analysis ongoing in our laboratory are expected to shed light on TBI-induced changes in apoptotic gene expression. Nonetheless, the results of our study suggest that increases in neuroprotective gene expression are associated with neuronal survival in the CA3 subregion after TBI rather than down-regulation of proapoptotic gene expression. Furthermore, after TBI in rats, severe hemorrhagic hypotension dramatically suppressed expression of both neuroprotective and potentially harmful genes (seven of the nine genes studied) in both injured and uninjured neurons. Although unexpected, this may be the result of the overall suppression of transcription after TBI and hemorrhage, due to, for example, injury-induced suppression of transcription factors such as activating transcription factor 2.33The lack of differences in neuroprotective gene expression in the injured and surviving neurons of rats subjected to TBI/HH suggests the involvement of additional factors critical to survival of individual neurons, at least in the first 24 h after TBI. Future studies in our laboratory will address these issues.
In a previous study, we showed that higher endogenous expression of other neuroprotective and survival genes in the CA1 subfield correlated highly with resistance to injury.8In that study, we laser dissected intact groups of cells containing several hundred neurons and glia from each region of the hippocampus. Although it is known that more than 95% of the cells in the CA1 region, for example, are neurons,34the laser-captured cells in that study were, nevertheless, a mixed population of neurons, interneurons, and a few glia. In the current study, individual Nissl-stained neurons were captured singly by LCM; therefore, the gene expression data obtained from these cells are representative of a much finer resolution than previously achieved.
We propose several possible explanations for the differences between injured and uninjured neurons in the CA3 subfield and the dramatic suppression of gene expression by hemorrhagic hypotension after TBI. The mammalian brain has 2,500–5,000 specific neuronal cell types,35and morphologically similar but functionally distinct neurons may differentially express certain genes. Even within apparently homogeneous populations of cells, there is substantial heterogeneity,36so perhaps uninjured neurons were differentially capable of increasing neuroprotective gene expression after TBI, whereas adjacent, apparently similar neurons were not. Another possible explanation is that differences between injured and uninjured neurons are the result of stochastic, or randomly varying, gene expression37(i.e. , at the time of injury, expression of neuroprotective genes was randomly increased in some neurons, which then were better able to survive an injury, but not in others). Finally, the results of our study could suggest that injury-induced signals may actively suppress neuroprotective gene expression in some neurons, which would render them vulnerable to traumatic injuries. Further comparative gene expression studies of injured and uninjured neurons will help to identify other injury-induced genes that may play a key role in the transcriptional regulation of neuroprotective gene expression.
Our data suggest that higher endogenous expression of neuroprotective genes 24 h after TBI is associated with protection of neurons from irreversible injury. Antioxidants have provided prolonged and robust neuroprotection in animal models of brain injury38and in vitro models of injury,39which are all associated with oxidative stress.40Transgenic mice overexpressing GPX-1 have been shown to be resistant to ischemia–reperfusion heart and brain injury as well as various types of oxidant tissue injury.41Furling et al. 42reported improved recovery of synaptic transmission in CA1 pyramidal cells of GPX-1 transgenic mice after a short period of hypoxia, suggesting that a moderate increase in endogenous antioxidant levels was sufficient to prevent irreversible damage. Increases in antioxidant activity are also positively associated with several biologic phenomena, including lifespan extension43and stress resistance.44We showed previously that up-regulation of HO-1 was significantly associated with the protective effects of zinc chelation after experimental TBI.9Induction of HO-1 was cytoprotective in pathologic states associated with oxidative stress (HO-1 catalyzes the rate-limiting reaction in the degradation of heme, a pro-oxidant that is released from hemoglobin after brain injury).45Increases in expression of neurotrophins such as BDNF have been shown to be associated with neuroprotection in brain injury models.27,46–48In immortalized mouse hippocampal HT22 cells, the neuroprotective effects of BDNF have been shown to be mediated by activation of the extracellular signal–regulated protein kinase and phosphatidylinositol 3-kinase survival pathways.49Finally, a recent report on a new therapeutic approach to amyotrophic lateral sclerosis showed that enhancement of the neuroprotective heat shock response slowed motor neuron degeneration and extends survival in a transgenic mouse model of the disease,50suggesting that selective enhancement of neuroprotective responses is sufficient to decrease neuronal cell death.
Interestingly, the lack of differences in proapoptotic gene expression between injured and uninjured neurons suggests that antiapoptotic strategies might be expected to exert little therapeutic effect after TBI. Even more important, further suppression of proapoptotic genes by subsequent hemorrhage suggests that the worsening of the outcome by a secondary injury is not mediated by proapoptotic factors. However, the lower levels of neuroprotective genes in all neurons after the combination of TBI and hemorrhagic hypotension suggest that enhancing the normal endogenous neuroprotective response may be a rational therapeutic strategy for patients with TBI. We interpret our results with caution because we did not examine the corresponding protein expression levels in the identified injured and uninjured neurons. However, for some of these genes, we and others have shown up-regulation in brain tissue after injury, though not at the single-cell level.9,38,42Future studies in our laboratory will examine the expression of some of these proteins at the single-cell level using immunohistochemical methods. Nonetheless, although our study provides only an association of increased neuroprotective gene expression with survival, we suggest that our results support the value of future studies of neuroprotective molecular mechanisms.
We have also demonstrated in the current study that quantitative mRNA analysis of small numbers of laser capture microdissected neurons is eminently straightforward and practical if current linear mRNA amplification methods are combined with sensitive RPA analysis. In principle, these methods can be readily applied to any neurodegenerative disease (e.g. , Parkinson or Alzheimer disease) in which dying or degenerating cells can be identified histologically. The transcriptional profile of these cells can be directly compared with adjacent cells that are not injured (or not yet injured) at that time point, allowing identification of factors that may contribute to neuronal survival in these neurodegenerative disorders. Furthermore, this approach allows facile comparison of the expression of any gene of interest in differentially identifiable cells in any type of tissue.
Finally, the results of this study illustrate the importance of studying gene expression at the single-cell level. In a previous microarray study using whole brain RNA, we found that HSP27 and HSP70 gene expression increased significantly after TBI and further increased after a neuroprotective zinc chelation treatment.9However, analysis of HSP70 expression in individual hippocampal neurons showed that there were no significant differences in mRNA levels between injured and uninjured neurons. We conclude that, at least in hippocampal neurons, increased HSP expression is not associated with neuronal survival. Further studies of injury-induced gene expression in discrete populations of neurons are necessary for better understanding of the molecular determinants of neuronal survival after TBI.
The authors thank Maria A. Micci, Ph.D. (Assistant Professor, Internal Medicine-Gastroenterology), and Willemijntje Hoogerwerf, M.D. (Assistant Professor, Internal Medicine-Gastroenterology), for their assistance with Fluoro-Jade photomicrography; Darshana Choksi, B.S. (Research Technician, First-Year Graduate Studies/Basic Biomedical Science), for assistance with real-time polymerase chain reaction analysis of brain-derived neurotrophic factor; C. Jordan Kicklighter, B.A. (Director, Editorial Services, Department of Anesthesiology), for assistance with manuscript preparation; and Christy Perry (Editor, Editorial Services, Department of Anesthesiology), for assistance with figures, at the University of Texas Medical Branch in Galveston, Texas.