Genome-Wide Transcriptome Profiling of Region-SpecificVulnerability to Oxidative Stress in the Hippocampus

Xinkun Wanga,b, Ranu Pala,b, Xue-wen Chenc, Keshava N. Kumara,b, Ok-Jin Kimb, and EliasK. Michaelisa,b,*

a Higuchi Biosciences Center, 2099 Constant Avenue, The University of Kansas, Lawrence, KS 66047, USA

b Department of Pharmacology and Toxicology, The University of Kansas, Lawrence, KS 66045, USA

c Department of Electrical Engineering and Computer Science, The University of Kansas, Lawrence, KS66045, USA

AbstractNeurons in the hippocampal CA1 region are particularly sensitive to oxidative stress (OS), whereasthose in CA3 are resistant. To uncover mechanisms for selective CA1 vulnerability to OS, we treatedorganotypic hippocampal slices with duroquinone and compared transcriptional profiles of CA1vs. CA3 cells at various intervals. Gene Ontology and biological pathway analyses of differentiallyexpressed genes showed that at all time points, CA1 had higher transcriptional activity of stress/inflammatory response, transition metal transport, ferroxidase, and pre-synaptic signaling activity,while CA3 had higher GABA-signaling, postsynaptic, and calcium and potassium channel activity.Real-time PCR and immunoblots confirmed the transcriptome data and the induction of OS byduroquinone in both hippocampal regions. Our functional genomics approach has identified in CA1cells molecular pathways as well as unique genes, such as, guanosine deaminase, lipocalin2,synaptotagmin 4, and latrophilin 2, whose time-dependent induction following the initiation of OSmay represent attempts at neurite outgrowth, synaptic recovery, and resistance against OS.

KeywordsHippocampus; CA1; CA3; oxidative stress; inflammation; neurite outgrowth; gene expressionprofiling

INTRODUCTIONOxidative stress (OS), resulting from an imbalance in the generation and scavenging of reactiveoxygen (ROS) and nitrogen species, can lead to cell damage during aging and to age-associatedneurodegenerative diseases [1–3]. Neurons in the central nervous system demonstrate differentlevels of sensitivity to OS, with some neuronal populations being very sensitive and othersquite resistant. A good example of differential neuronal vulnerability is found in thehippocampus, a brain structure involved in learning and memory processes. Neurons in thehippocampal CA1 region are selectively vulnerable to OS, while those in the adjacent CA3region are resistant [4,5]. CA1 neurons are also vulnerable to other adverse conditions,

*Corresponding Author: Dr. Elias K. Michaelis, Department of Pharmacology and Toxicology, University of Kansas, Lawrence, KS66045, USA, Tel: (785) 864 4001, Fax: (785) 864 5219, E-mail: emichaelis@ku.edu.Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customerswe are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resultingproof before it is published in its final citable form. Please note that during the production process errors may be discovered which couldaffect the content, and all legal disclaimers that apply to the journal pertain.

NIH Public AccessAuthor ManuscriptGenomics. Author manuscript; available in PMC 2008 August 1.

Published in final edited form as:Genomics. 2007 August ; 90(2): 201–212.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

including epileptic seizures [6] and hypoxia- or ischemia-induced neuronal damage [7,8]. SinceOS is associated with these adverse conditions, it is quite possible that the sensitivity of CA1neurons to these seemingly different conditions is due to a shared commonality, i.e., sensitivityto OS. There is little information on the underlying mechanisms of selective neuronalvulnerability. Therefore, the study of molecular mechanisms that may determine selectivevulnerability to OS may provide important information on the molecular determinants ofneuronal sensitivity to various types of brain insult.

There have been few attempts to elucidate the molecular underpinnings of the differentialsensitivity of CA1 neurons to OS. In a previous study, we probed the expression of known pro-oxidant and anti-oxidant genes in CA1 and CA3 regions of un-stimulated, non-stressedhippocampal slices maintained in culture [4]. Based on these analyses, we concluded that CA1neurons exhibit high intrinsic oxidative activity even in the absence of exogenous stimuli andwe proposed that this property of CA1 neurons may be responsible for their selectivevulnerability to OS [4]. Other investigators have also attempted to determine differences intranscriptional patterns among the various regions of the hippocampus under normal conditions[9–13]. However, none of these studies was designed to probe the dynamic, time-dependentresponses of neurons in CA1 and CA3 to acute exposure to OS.

In an attempt to shed more light on the molecular mechanisms of CA1 vulnerability to OS, weconducted dynamic genome-wide transcriptional analyses on the response of CA1 and CA3neurons to in vitro-applied OS. Four time points representing different periods of exposure toOS and of post-OS recovery were selected in order to detect differences in dynamic changesin the induction of gene expression between the two hippocampal regions. Our hypothesis wasthat differential transcriptional responses to OS in CA1 vs. CA3 underlies (at least partially)their opposite fates following such stress.

RESULTS AND DISCUSSIONDifferential vulnerability of CA1 and CA3 neurons to OS

We used organotypic hippocampal slice cultures [14] to study the differential vulnerability ofCA1 and CA3 neurons to increased OS. Treatment of the cultures with duroquinone (DQ) wasused to generate ROS in cells. DQ elevates superoxide (O2•−) levels in cells [15]. Followinga relatively quiescent period between 0 and 3h of treatment with DQ in terms of cell death,there was substantial cell loss in CA1, not in CA3, after removal of DQ and continuedincubation in culture medium (Fig. 1A). Exposure to OS for 3 h caused delayed cell death inCA1 that occurred at 9 and 21 h post termination of treatment with DQ (Fig. 1A). At 21 hfollowing termination of DQ treatment, 64% of CA1 neurons were dead whereas only 7% ofCA3 neurons died during the same period. Neuronal death reached a plateau in both regionsafter this time post-treatment with DQ (data not shown). The pattern of neuronal death shownin figure 1 was in agreement with that seen in previous studies [5].

Based on the results shown in figure 1A, we selected four time points for the conduct of studiesof dynamic changes in gene expression following induction of OS. DQ treatment of slices wasfrom 0 to 3 h (Fig. 1B). The time points selected for the conduct of mRNA extraction and genearray analyses were: 0 h, the basal condition of neurons just before initiation of DQ treatment;0.5 h, the early phase of neuronal response to DQ treatment; 3 h, the end of the OS treatment;and 12 h, a time point that was 9 h post-termination of treatment with DQ. The time point of0.5 h from initiation of treatment was selected in order to detect changes in expression of earlystress-response genes. DQ treatment for 0.5 and 3 h did not produce significant cell death ineither CA1 or CA3 regions (Fig. 1A) yet was likely to have caused changes in expression ofearly stress-response genes. There was, however, loss of approximately 40% of CA1 neuronsat 12 h (9 h post-treatment with DQ). Because of this cell loss, the genomic analyses performed

Wang et al. Page 2

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

on slices at 12 h were likely to underestimate the magnitude of changes in gene expression ofCA1 neurons. In order to avoid further confounding of the gene expression studies by thepresence of excessive cell death in CA1, we did not include the 24 h time point in our genearray analyses.

Reproducibility of gene array data obtained from treated hippocampal slicesThe microarray data obtained using the treatment protocol shown in figure 1B and AffymetrixGeneChip RAE230A arrays, showed high correlation between replicate samples (n = 4) and amean correlation coefficient of 0.94 (ranging from 0.91 to 0.97). These results indicated highreproducibility for each hippocampal region at each time point and across slices obtained fromdifferent animals. The high reproducibility achieved in this analysis was, at least in part, areflection of the homogeneity of the dissected CA1 and CA3 samples used for the arrayhybridization. The manual microscopic cell procurement used in the present study was close,in terms of reproducibility of the generated microarray data, to the laser capture microdissectionmethod (LCM). In a separate series of experiments, we used fresh, coronal brain sections usingLCM to collect CA1 and CA3 neurons for microarray analyses (identical GeneChip used withLCM and hand microdissection). The mean correlation coefficient between replicates usingLCM was 0.96 (range: 0.91–0.99), i.e., very similar to the results of hand microdissection.

Time course of common transcriptional responses to OS in both CA1 and CA3 neuronsIn order to identify genes showing significant time-dependent alterations in expression duringand after DQ treatment, we used a BioConductor package called “timecourse”, based onmultivariate empirical Bayes statistics [16]. The package ranked all “present” (i.e., expressed)genes by statistical significance associated with their time-dependent dynamic changes.Analyses based on multivariate empirical Bayes statistics have some advantages overtraditional F-statistics, such as the incorporation of correlations among time-course samples,integration of replicate variances, and the integration of information across genes. Using thistype of analysis, probesets showing significant responses to DQ treatment in both CA1 andCA3 were identified. Seventy seven probesets were identified whose response patterns in CA1and CA3 converged (Pearson’s correlation coefficient > 0.75) to the DQ treatment. The 77probesets represented 0.49% of the total 15,866 probesets on the array. A hierarchical clusteringof all 77 probesets is shown in figure 2A. There were concordant response patterns of the samegenes in the two hippocampal regions, an indication that the CA1 and CA3 shared a certaindegree of commonality in terms of the response to DQ treatment. The currently known genesamong the 77 probesets are listed in table 1.

Based on the hierarchical clustering shown in figure 2A, the genes were further subdividedinto two groups marked as “I” and “II”. The total number of genes in Group I was 23. Thecurrently known genes in this group are listed in table 1 and represent genes whose expressionwas significantly down-regulated in both CA1 and CA3 following DQ treatment. Down-regulation of genes such as CD36 (thrombospondin receptor, a class B scavenger receptor), isassumed to offer protection against DQ-elicited neuronal damage, because the over-expressionof this gene is known to increase OS [17]. The CD36 protein is involved in scavenging by-products of OS in cells [18] and its expression is increased during β-amyloid deposition [19],a stimulus of OS in Alzheimer’s disease.

Of the 54 genes in Group II, the currently known ones are shown in table 1. The expression ofsome genes in this group, such as Hspa1b (heat shock 70 kDa protein 1B), Atf3 (activatingtranscription factor 3), and Bdnf (brain derived neurotrophic factor), was rapidly increased,i.e., within 0.5 h following the application of OS (Fig. 2A). The rapid up-regulation of thesegenes was, probably, a reflection of cellular attempts at controlling damage. HSP70 is one ofthe early responders to stress and is reported to be protective against OS [20–22]. Atf3 is another

Wang et al. Page 3

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

immediate early gene responding to a wide variety of stress conditions, including OS [23].Bdnf has also been found to increase quickly after induction of cell stress and to limit thedamaging effects of stress in brain [24]. It is important to point out also that BDNF release inthe vicinity of neurons leads to the induction of Nrn1 (neuritin) [25], a gene whose productpromotes neurite outgrowth in hippocampal neurons. Nrn1 expression was significantlyincreased in both CA1 and CA3 following induction of OS (Table 1) and the increasedexpression of Nrn1 fit with the observed up-regulation of Bdnf expression. Other stress-inducible genes whose expression was increased in both CA1 and CA3 were Gadd45b andGadd45g (Table 1). Both genes are induced by various environmental stresses [26].

The increases in expression of some of the genes in Group II occurred more slowly. This wasthe case for Nrgn (neurogranin), Ncdn (neurochondrin), Camk2b (calcium/calmodulin-dependent protein kinase II, beta subunit), and Nptxr (neuronal pentraxin receptor). Many ofthese genes encode proteins related to calcium-mediated signaling, such as regulation of theCa2+- and Ca2+/calmodulin-dependent signaling pathway [27]. The increases in the expressionof these genes may point to the importance of Ca2+ regulation and of pathways controlled byintracellular Ca2+ in the defense of cells against stimuli initiated by OS.

Another gene whose expression was induced during the period of OS was Slc17a6 (vesicularglutamate transporter 2 or Vglut2) (Table 1). The up-regulation of Slc17a6 expression underconditions of OS may be a sign of increases in progenitor cell differentiation to mature neurons[28] rather than increases in excitability within the hippocampus. Finally, some of the GroupII genes showed more complex patterns of up-regulation in both CA1 and CA3. Expression ofthese genes was elevated within 0.5 h after the initiation of exposure to DQ, followed by adecrease between 0.5 and 3 h, and a subsequent second wave of increase. Rgs4 (regulator ofG-protein signaling 4), Rgs7, and Drd1ip (dopamine receptor D1 interacting protein) had suchpatterns of expression. Rgs4 and Rgs7 are predominantly expressed in brain [29,30], are capableof binding to Gα subunits, can deactivate Gα, and terminate downstream signals [31,32], thusprotecting neurons from ROS-induced activation of Gα [33,34]. Non-monotonic expressionprofiles for genes such as Egr-1 (early growth response-1) have been observed under otherstress conditions, e.g., in hypoxia [35]. We speculate that genes showing this expression patternmay play dual roles, as genes that respond to early stress and as genes involved in cell repairat a later stage (i.e., 12h).

In order to ascertain the cellular or biological functions of the genes that had common responsepatterns in neurons of the CA1 and CA3, Gene Ontology (GO) analyses were performed. Byusing a controlled vocabulary to describe gene and protein attributes, the GO project assignsgenes and their products into relevant GO categories in terms of their molecular functions(MF), involvement in biological processes (BP), and associated cellular components (CC). GOanalysis of identified genes is an important approach to uncovering functional profiles ofensembles of genes [36]. We used a statistical test based on the hypergeometric distributionimplemented in Onto-Express [37,38] to detect functional categories significantly altered byOS. Based on the presence or absence of the identified genes within each GO term, a p valuewas first calculated for each GO term. The p values were further processed by multiple-testcorrection [39] in order to effectively control the false discovery rate and, as a result, generatea closer estimation of statistical significance for each GO term. Table 2 presents the GO termsthat had a corrected p value less than 0.05 and contained at least two of the identified genes.

The results of the GO analyses revealed several functional categories of genes whoseexpression was altered in a similar manner in CA1 and CA3 following exposure to OS. Thesecategories were: (i) signal transduction, (ii) stress response, (iii) apoptosis, and (iv)neurogenesis. A large number of genes were associated with signal transduction, includingboth intercellular and intracellular signaling. Some signal transduction genes went up, while

Wang et al. Page 4

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

others went down during the period of OS. Expression of genes in the categories of synaptictransmission, calmodulin binding, and GPCR signaling was up-regulated during OS treatment.Heightened OS led to increases in expression of genes related to cell stress responses in neuronsof both CA1 and CA3. The GO categories “response to stress” and “DNA repair” were enrichedwith genes whose expression was increased during OS. On the other hand, the expression ofsome genes related to immune-responses was decreased in both CA1 and CA3. For example,expression of chemokine genes Ccl6 (chemokine C-C motif ligand 6) and Cxcl4 (chemokineC-X-C motif ligand 4 or platelet factor 4) was decreased during most of the period of OS. Thechanges in expression of these genes differed from those of other chemokine and cytokinegenes that were strongly up-regulated across time in CA1 (Table 3).

Although CA1 neurons died to a much greater extent than CA3 neurons following exposureto DQ, up-regulation of genes related to apoptosis and anti-apoptosis occurred in both CA1and CA3 cells. These results were an indication that pathways other than those sub-served bythis group of genes might have been responsible for the differential cell death of CA1 neurons.The last GO category enriched with genes up-regulated in both CA1 and CA3 was that ofneurogenesis. The enrichment of this category of genes may be a reflection of attempts by CA1and CA3 neurons to activate repair processes, replace lost neurons, or re-grow neurites.

Biological pathway analysis was also conducted to uncover functional profiles of the DQ-responsive genes whose expression is up- or down-regulated in both CA1 and CA3. Thebiological pathway approach assembles genes (or corresponding proteins) with differentfunctions into a biologically meaningful framework needed for a certain biological task orprocess. The pathway databases on which we based our analyses were those covered by KyotoEncyclopedia of Genes and Genomes (KEGG) [40], BioCarta and GenMAPP [41]. Weidentified the calcium signaling pathway as one that was significantly enriched with both up-regulated (p = 0.031, after multiple test correction) and down-regulated genes (p = 0.027) inboth CA1 and CA3. This may indicate that neurons in both regions were attempting to controlsignal transduction mediated through changes in intracellular Ca2+ concentration. Anotherpathway involved in signal transduction, the MAPK signaling pathway, was also found to besignificantly enriched with genes whose expression was up-regulated in both CA1 and CA3neurons (p = 0.040). These results were in agreement with those of the GO analysis describedabove (Table 2).

Differences in gene transcription patterns between CA1 and CA3 neurons during OSUsing the multivariate empirical Bayes model in the “timecourse” package, we identified 97genes (0.61% of the 15,866 probesets) whose expression patterns diverged significantlybetween CA1 and CA3 upon DQ treatment. The divergence in expression of genes betweenCA1 and CA3 was determined using the MB and T̃2 statistics-based ranking and visualizationoffered by the “timecourse” package. The expression patterns were classified by hierarchicalclustering (Fig. 2B). Based on the clustering results, we grouped these genes into threecategories (Fig. 2B). The 27 genes in Category I generally showed higher expression in CA1than CA3 during DQ treatment. The currently known genes in Category I are listed in figure2B, C and table 3. The 31 genes in Category III were expressed more highly in CA3 than CA1during DQ treatment. The known genes in Category III are also listed in table 3. The remaindergenes belonged to Category II, showing different dynamic changes in response to DQ, but noconsistent pattern in one region vs. the other (Fig. 2B).

Many of the Category II genes showed tri-phasic dynamic patterns of expression in CA1neurons, with increases in expression during 0 to 0.5 h, decreases between 0.5 and 3 h, andsubsequent increases during the period 3 to 12 h. In CA3, the same genes showed eitherincreased expression only in the last period of treatment, or were consistently increasedthroughout the period of exposure to DQ. Several Category II genes were related to synaptic

Wang et al. Page 5

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

vesicle trafficking and docking, processes involved in neurotransmitter release. These genesencode SNARE proteins (soluble N-ethylmaleimide-sensitive factor attachment proteinreceptors), such as SNAP25 (synaptosomal-associated protein, 25kD) and VAMP1 (vesicle-associated membrane protein 1, or synaptobrevin 1), and synaptic vesicle-associated proteins,such as CPLX1 (complexin 1), SYN2 (synapsin II), and SYNGR1 (synaptogyrin 1). SNAP25,is highly sensitive to ROS [42] and is a potential pre-synaptic ROS sensor that mediates ROS-induced impairment in neurotransmitter release. Up-regulation of SNAP25 transcription mayreflect efforts of neurons to compensate for the loss of its activity. The up-regulation of severalof these genes was more rapid in CA1 than CA3 neurons (Fig. 2B). The more rapid up-regulation of these genes in CA1 might indicate that the presynaptic neurotransmitter releasingmachinery was affected to a greater degree in this vulnerable region.

The patterns of regional expression of some of the identified genes, e.g., Pou3f1, Lphn2 (bothhigher in CA1, Fig. 2C and Table 3), and Nfl (higher in CA3, Table 3), were in agreement withpreviously reported mRNA and protein expression levels in hippocampal regions determinedby either in situ hybridization or immuno-staining. POU3F1 protein is expressed in CA1 athigher levels than in CA3 [43]. This general pattern was supported by our GeneChip data. Theexpression of the mRNA for the receptor for α-latrotoxin (latrophilin 2), Lphn2, is confined toonly the CA1 region of the hippocampus [44]. By GeneChip analysis, we found that this genewas expressed in CA1 at a significantly higher level than in CA3. On the other hand,neurofilament light polypeptide, Nfl, is more abundant in CA3 than in CA1 [45], which wasalso found to be the case in the present study. The similar patterns of expression in CA1 andCA3 detected in previous studies using various experimental techniques and in the presentstudy by GeneChip analysis provided partial confirmation of the validity of the results obtainedby the microarray analysis.

Among the genes showing overall higher expression in CA1 than CA3 were several immuneresponse genes, Cxcl1 and Cxcl2, and Ifitm3 (Table 3 and Fig. 2C). Cxcl1 (also known asGro1) is a gene that stimulates growth of some tumor cells and has an important function ininflammation and wound healing [46]. Another gene whose expression was differentiallyincreased in CA1 over CA3 neurons was Tfpi2 (tissue factor pathway inhibitor 2). Expressionof Tfpi2 is increased during acute inflammatory responses [47], therefore, this gene may beinvolved in anti-inflammatory responses in cells of the CA1. Added to the list of genes whoseexpression is increased during inflammation and whose function is to suppress inflammatoryresponses, is the gene Serpina3n (serine peptidase inhibitor, clade A, member 3N, or α-1-antichymotrypsin) [48]. Serpina3n expression was higher in CA1 than CA3 neurons beforeand during DQ treatment. SERPINA3 forms part of the inflammatory response inneurodegenerative diseases [48]. Increases in expression of these genes in CA1 might beindicative of higher activation of wound healing or recovery from trauma in cells of this brainregion. This would be consistent with the differential increases in expression of other genesinvolved in anti-oxidant and anti-inflammatory defenses in cells of the CA1 region describedbelow.

The higher levels of expression in CA1 over CA3 neurons of Tf (transferrin) and Cp(ceruloplasmin) before and during the period of OS, fit well with our previous observation thatCA1 neurons, in the absence of any exogenously-induced OS, express higher levels of thesetwo genes [4]. Both proteins are involved in metal transport and are part of either pro- or anti-oxidant processes in cells [49]. Ceruloplasmin is a ferroxidase [50] and functions as anantioxidant protein during inflammation or hyperoxia [51]. Another gene that codes for aprotein that sequesters iron is Lcn2 (lipocalin2) [52]. Expression of this gene was also higherin CA1 than CA3 cells during OS. This gene has not been previously linked to neuronal stressor damage. Thus, transport, sequestration, and oxidation of iron and copper appeared to be

Wang et al. Page 6

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

primary functions of genes which were consistently higher in CA1 than CA3 before and duringOS.

Hspb1 (heat shock protein 27 or Hsp27) also had higher levels of expression in CA1 over CA3cells before, during, and after induction of OS. HSP27 is a microtubule-interacting, heat shockprotein that protects a variety of cells, including neuronal cells, from stress stimuli that lead toapoptosis [53,54]. HSP27 decreases ROS production in cells that over-express the mutatedHuntington’s disease protein and protects these cells from OS-induced cell death [55]. Thehigh levels of expression of Hspb1 by CA1 neurons might be viewed as part of a coordinatedattempt to overcome the damaging effects of OS. Other notable genes whose expression washigher in CA1 as compared with CA3 neurons were Arc, Calb1, Gda, Thbs4, and Syt4 (Table3). Arc (activity-related cytoskeletal-associated protein) is a neuronal immediate early genewhose expression is rapidly and transiently increased in the hippocampus CA1 region uponsynaptic activation of pyramidal neurons [56] or following electroconvulsive stimulation ofthe cortex [57]. In this study, expression of Arc in the CA1 and CA3 regions increased rapidly(at 0.5 h), and continued to increase to a significantly higher level in CA1 than CA3 at 12 h.

A gene whose expression is highly correlated with resistance of hippocampus and dentate gyrusneurons to various types of stresses, including ischemia and epileptic seizures, is Calb1(calbindin-D28K). The differential regulation between CA1 and CA3 of Gda (guanosinedeaminase or Cypin or p51-nedasin), the gene for a protein that increases dendrite branchingin neurons [58], might be another indicator of attempts by neurons of the CA1 region toovercome the damaging effects of OS and re-establish normal synaptic activity. Thisinterpretation is supported by the selective induction in CA1 of the genes for two other proteinsthat promote neurite outgrowth and local synapse differentiation, Thbs4 (thrombospondin 4)[59] and Syt4 (synaptotagmin 4) [60] (Fig. 2C and Table 3).

It is also important to note that proteins such as GDA may act in the formation of synapses,especially the post-synaptic structures. GDA interacts with PDZ domain-containing proteinsat postsynaptic sites of excitatory synapses [61]. Another protein that interacts with PDZdomain-containing proteins is that encoded by the gene Lphn2, a protein that is almostexclusively expressed in CA1 [44] and whose expression is differentially increased in CA1 byDQ treatment (Table 3).

Only one gene that falls in the general functional category of a stress-response gene wasexpressed at higher levels in CA3 than CA1 cells during OS. This gene was Il16 (interleukin16). There was, however, higher expression in CA3 than in CA1 of Gabra5 (GABA receptoralpha 5). Increases in the activity of the inhibitory neurotransmitter γ-aminobutyric acid(GABA) would be expected to decrease neuronal excitation and thus may lead to some levelof protection of neurons in the CA3 region against any excitatory effects induced by OS.

The results of GO analyses of the genes in Categories I and III are summarized in table 4. TheGO analysis indicated a higher level of up-regulation of transcription in CA1 than in CA3neurons of genes related to stress responses, including chemokine, cytokine, and inflammatoryresponse genes. This finding is further supported by our analysis of biological pathways thatidentified the cytokine-cytokine receptor interaction pathway as the only one having reachedstatistical significance (corrected p = 0.0019). Many of the genes shown in table 3 and discussedabove were included in the GO category and biological pathway described as stress-responsegenes. The GO analyses also identified the categories of genes related to synaptic vesicletransport and neurotransmitter release and of G-protein coupled receptors (GPCR) as beingmore highly expressed following OS in CA1 as compared with CA3 (Table 4). Althoughexpression of genes in the GO categories under synaptic vesicle transport was significantly

Wang et al. Page 7

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

higher in CA1 than CA3, there was only one gene in each of these categories whose expressionwas higher.

The CA3 region had statistically higher expression than CA1 in GO categories related to post-synaptic activity, GABA activity, and voltage-gated calcium and potassium transport (Table4). Although expression of genes in the various GO categories under GABA activity, voltage-gated calcium and potassium transport was significantly higher in CA3 than CA1, there wasonly one gene in each of these categories whose expression was higher.

Confirmation of the microarray gene expression patterns using real-time PCR (qPCR)For the qPCR studies, we selected five genes that are known to be involved in anti-oxidantresponses in cells: Junb, Mt1a (metallothionein 1A), Nqo1 (NAD(P)H:quinone oxidoreductase1), Nrf2 (NF-E2-related factor-2) and Tf. The selection of these genes was based also on theirdifferential expression in CA1 vs. CA3. Although the expression patterns of these genes (exceptTf) did not reach statistical significance of differential expressed in CA1 vs. CA3 according tothe “timecourse” package, all five genes showed higher expression in CA1 than CA3 for alltime points. There was a significant correlation between the microarray data and qPCR resultsfor all five genes (Fig. 3). The estimated Pearson product-moment correlation coefficient washighly significant (Fig. 3). The standard error of both methods was comparable: the medianfor GeneChip was 0.55 (range: 0.03 – 2.17), while the median for qPCR was 0.69 (range: 0.04– 3.63). These results, along with the high reproducibility of the GeneChip data, confirmed thevalidity of the data gathered by the GeneChip microarray method.

Molecular confirmation of the induction of OS based on the levels for JUNB and SQSTM1Two proteins whose levels in neurons are known to be rapidly altered by a variety of stresses,especially OS, are JUNB and SQSTM1 (sequestosome 1). These two proteins were chosen toconfirm at the molecular level the induction of OS in hippocampus. JunB mRNA and proteinlevels are increased following traumatic brain injury or chemically-induced OS [62,63].Sqstm1 expression increases in response to a variety of stress conditions, including OS [64,65]. In the GeneChip analyses, we found that the transcripts of Junb, in both CA1 and CA3,increased rapidly (within 0.5 h) after exposure of the slices to DQ and then gradually returnedtoward baseline levels at 12 h. The protein level of JUNB, as detected by immunoblots, showeda continuous increase throughout the period of OS in both CA1 and CA3, with the correlationcoefficients (r) between JUNB level and elapsed time being 0.89 in CA1 and 0.95 in CA3.Also, there was a high degree of correlation between the JUNB levels in CA1 and those in CA3across all time points (r = 0.97; p = 0.034).

For Sqstm1, mRNA levels also showed a rapid increase, in both CA1 and CA3, within 0.5 hafter DQ exposure, followed by slow but continuous further increases in Sqstm1. Estimatedcoefficient for inter-region (CA1 vs. CA3) correlation was: r = 0.99 (p = 0.003). The proteinlevel of SQSTM1 also increased quickly after initiation of OS and continued to rise up to 3 h.The levels of SQSTM1 were highly correlated between CA1 and CA3 (r = 0.98, p = 0.022).The GeneChip and Western blot results for JunB and Sqstm1 expression described above, werea strong indication that OS was indeed established in hippocampus slices exposed to DQ.Completely identical patterns for mRNA and protein changes across time of exposure to DQwere not expected, as changes in mRNA levels would normally precede those in protein levels.

Concluding remarksTo begin to identify molecular mechanisms for the selective vulnerability of CA1 neurons toOS, a multiple-time-point dynamic transcriptome analysis was employed in the assessment oftime-dependent changes in neuronal injury induced by OS. Functional genomic analyses wereused to compare genome-wide gene expression responses to increased OS in a vulnerable,

Wang et al. Page 8

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

CA1, and a resistant, CA3, region of the hippocampus. The clearest difference between thesetwo regions in terms of their OS responses was the much higher level of induction of genesrelated to stress and anti-inflammation in CA1 as compared with CA3. Such differentialinduction might be an indication that CA1 neurons are endogenously, and possiblycontinuously, under higher stress than CA3 neurons. A state of high endogenous stress in CA1neurons, we hypothesize, may predispose them to cell damage and loss of viability whenconfronted with further increases in OS. Several genes that either have not been previouslylinked to neuronal responses to stress, or have not been considered as part of the gene expressionrepertoire in response to OS, such as Gda, Lcn2, Lphn2, and Syt4, were identified as ones thatwere differentially expressed in CA1 following the induction of OS. The genes identified inthe present study offer potential new targets for future experimental manipulations to assesshow their expression alters the sensitivity of neurons to OS. In addition, the GO categoriesidentified in this study provide potential new molecular and cellular pathways that can beprobed for the causes of differential neuronal vulnerability to OS or other forms of stress.

Experimental MethodsEstablishment of organotypic hippocampus cultures

Organotypic hippocampal slice cultures were established from male Sprague-Dawley rat pups[4,14]. All animal procedures were performed in accordance with guidelines of the Universityof Kansas IACUC. Ten-day old male rat pups were anesthetized, the hippocampi dissectedout, and slices (300 μm) prepared using a McIlwain tissue chopper. The slices, maintainedbriefly in ice-cold artificial cerebrospinal fluid (in mM: 124 NaCl, 2.8 KCl, 2 CaCl2, 2MgSO4, 26 NaHCO3, 1.25 NaH2PO4, 10 glucose, pH 7.4), were transferred onto BD™ FalconCell Culture Inserts (0.4 μm). The inserts were positioned on the surface of cell culture mediumconsisting of 50% minimum essential medium (Sigma-Aldrich, St. Louis, MO, USA), 25%Hanks balanced salt solution, and 25% heat-inactivated horse serum (Invitrogen, Carlsbad,CA, USA) in a solution containing 4.5 mg/ml glucose, 1 mM glutamine, and 20 ml/L penicillin-streptomycin. The medium was changed every third day.

Induction of OS in culturesTo induce OS, the slices were treated with DQ [5]. The culture medium was changed to serum-free medium 24 h prior to the initiation of DQ treatment. An aliquot of DQ (100 mM stocksolution) was added to the medium (final concentration 100 μM). The duration of DQ treatmentwas 3 h. At the end of the treatment period, the medium was removed and replaced with serum-free medium. At 0.5, 3, 12 and 24 h after the initiation of DQ treatment, the slices were rinsedthree times with the same medium and stained with propidium iodide (PI, 5 μg/ml) to determinecell death [5].

Microdissection of CA1 and CA3 pyramidal cell layer and isolation of total RNASlice cultures not stained with PI along with the polyethylene terephthalate matrix at the bottomof the inserts were detached from the plastic frame, fixed and stained using HistoGene™ LCMStaining Kit (Arcturus, Mountain View, CA, USA). Dissection of the pyramidal cell layers ofCA1 and CA3 was performed under a dissection microscope with tungsten needles, mini-scalpels and micro-forceps. The dissections were performed at 0 h (no-treatment with DQ),0.5, 3 and 12 h post-exposure to DQ. The microdissected layers were immersed in 350 μl RLTlysis buffer in Rneasy® Mini Kit (Qiagen, Valencia, CA, USA), incubated for 30 min at 42°C, centrifuged at 5,400 g for 2 min, and total RNA extracted from the supernatant accordingto the protocol of Qiagen RNeasy® Mini Kit.

Wang et al. Page 9

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Oligonucleotide microarray hybridization, data collection and analysisFour cultures from 4 rat littermates under each condition of treatment with DQ were used. The4 replicates of mRNA were subjected to Affymetrix GeneChip (Affymetrix, Santa Clara, CA,USA) analyses. The amount of total RNA was in the range of 10–100 ng and the GeneChipTwo-Cycle cDNA Synthesis Kit (Affymetrix) was used for target preparation. The volume ofthe isolated RNA was first concentrated to 7 μl in an Eppendorf® Vacufuge Concentrator(Brinkmann Instruments, Westbury, NY, USA) and 3.5 μl was used for two cycles of reverseand in vitro transcription. Affymetrix Rat Expression 230A (RAE230A) oligonucleotide arrayscontaining 15,866 probesets and interrogating 14,280 Unigene clusters, were used forhybridization (45 °C for 16 h) with a mixture of 15 μg of fragmented cRNA. Three RNAsamples extracted from slices of two animals failed to produce enough quantity and quality ofcRNA after the amplification process. GeneChip data generated from these were not subjectedto further analyses. Subsequent washing and staining steps were performed on a GeneChipFluidics Station 400 and the chips were scanned on a GeneArray Scanner GA2500(Affymetrix). Instrument control and data collection were carried out using GeneChipOperating Software (GCOS, ver 1.1.1). In order to minimize experimental variability, all stepsin slice culture preparation, DQ treatment, microdissection, RNA isolation and microarrayoperation were performed by a single investigator. The quality and quantity of the RNAsamples and of the cRNA probes were determined with a Bioanalyzer 2100 using the RNA6000 Pico LabChip (Agilent Technologies, Palo Alto, CA, USA), and by spectrophotometricmeasurements at 260 and 280 nm on GeneQuant (Amersham Biosciences, Piscataway, NJ,USA). The microarray data from all chips met the quality control criteria set by Affymetrix,including low background and noise, positive detection of QC probesets such as bioB, percentof genes called present in normal range (between 40–60%), similar scaling factors across allchips, and 3′/5′ ratio.

For data analysis, the GCOS .CEL data files were first preprocessed with the RMA method byemploying the rma function in BioConductor package affy (version 1.8.1). This pre-processingstep included 1) model-based probe-specific correction of PM probes, 2) quantilenormalization of the corrected PM probes, and 3) median-polish-based calculation ofexpression measure. The preprocessed data were used to calculate correlation coefficientsbetween replicate samples to evaluate array data reproducibility. For this calculation, thepreprocessed data of all probesets were imported to Microsoft Excel and Pearson correlationcoefficients were calculated using the CORREL function.

To identify genes showing convergent and divergent expression patterns during OS treatmentof the two hippocampal regions, the preprocessed data for all probesets were imported into theBioConductor package timecourse (version 1.2.1). This package offers a comprehensivelibrary of functions for the analysis of microarray time course data. With this package, geneswere ranked based on MB and T̃2 statistics calculated from their overall dynamic expressionprofiles [16]. Since no p values were provided by this package, genes of interest were identifiedby the ranking and visualization of corresponding dynamic expression profiles. Genes that hada fold-difference cutoff of 2.0 were called up- or down-regulated in one region vs. the other.

Hierarchical clustering (or gene tree) of both convergent and divergent genes was conductedin GeneSpring GX (version 7.3). All gene RMA logarithmic values were transformed to linearscale and mean-centered to 1.0 after importing to GeneSpring. The similarity measure usedwas Pearson Correlation.

All gene expression data were deposited in NCBI’s Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/). They are accessible through GEO Series accession numberGSE6693.

Wang et al. Page 10

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Gene Ontology and Biological Pathway AnalysesThe 77 convergent and 97 divergent genes between CA1 and CA3, identified by the abovesteps, were used for GO and Biological Pathway analyses. Convergent genes were furtherseparated into two groups, i.e., up- and down-regulated genes (genes marked “I” and “II” Fig.2A). Divergent or differentially expressed genes formed Group “I” and “III” of figure 2B. Thecomplete gene lists are presented in the online supplement. For GO analysis, individual listsof genes were fed into Onto-Express (http://vortex.cs.wayne.edu/projects.htm#Onto-Express)[37,38] and hypergeometric distribution and false discovery rate corrections were set. Forpathway analysis, the same gene lists were introduced into Pathway-Express (http://vortex.cs.wayne.edu/projects.htm#Pathway-Express) [66], hypergeometric distribution andfalse discovery rate corrections were used, and other parameters were set at default values. TheGO and Biological Pathway databases used for the analyses were those of July 2006. A p valueof less than 0.05 after correction was considered significant.

Real-time qPCRThe remainder of the concentrated total RNA extracts from CA1 and CA3 not used formicroarray analyses, were used for qPCR (SmartCycler, Cepheid, Sunnyvale, CA, USA). TheRNA was reverse-transcribed to cDNA using oligo(dT) primers and Taqman Gene ExpressionAssay reagents (Applied Biosystems, Foster City, CA, USA) used for qPCR assays. Thethermal cycling conditions were: 2 min at 50 °C, 10 min at 95°C, and 45 cycles of 95 °C for15 sec and 60 °C for 1 min. Quantification of target genes was based on the relative standardcurve [67].

SDS-PAGE and immunoblot studies of JUNB and SQSTM1Protein samples extracted from CA1 and CA3 of slice cultures were subjected to SDS-PAGEand immunoblot analyses [68] using Pierce SuperSignal West Femto kit for proteinimmunolabeling. Polyclonal antibodies for JUNB and SQSTM1 were from Santa CruzBiotechnologies, that for neuron-specific enolase from Chemicon. The amount of JUNB andSQSTM1 in each extract was normalized against the densitometric value of the immunolabeledband of neuron-specific enolase. Enolase gene expression was not altered following exposureof slices to OS, thus this protein was used as the normalization standard.

Supplementary MaterialRefer to Web version on PubMed Central for supplementary material.

Acknowledgements

This research was supported by grants from NIA, AG12993, and NICHD, HD02528, a Kansas Technology EnterpriseCorporation (KTEC) grant to the Higuchi Biosciences Center at the University of Kansas, and the Miller-Hedwig-Wilbur Fund.

References1. Emerit J, Edeas M, Bricaire F. Neurodegenerative diseases and oxidative stress. Biomed Pharmacother

2004;58:39–46. [PubMed: 14739060]2. Serrano F, Klann E. Reactive oxygen species and synaptic plasticity in the aging hippocampus. Ageing

Res Rev 2004;3:431–443. [PubMed: 15541710]3. Zecca L, Youdim MB, Riederer P, Connor JR, Crichton RR. Iron, brain ageing and neurodegenerative

disorders. Nat Rev Neurosci 2004;5:863–873. [PubMed: 15496864]4. Wang X, Pal R, Chen XW, Limpeanchob N, Kumar KN, Michaelis EK. High intrinsic oxidative stress

may underlie selective vulnerability of the hippocampal CA1 region. Brain Res Mol Brain Res2005;140:120–126. [PubMed: 16137784]

Wang et al. Page 11

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

5. Wilde GJ, Pringle AK, Wright P, Iannotti F. Differential vulnerability of the CA1 and CA3 subfieldsof the hippocampus to superoxide and hydroxyl radicals in vitro. J Neurochem 1997;69:883–886.[PubMed: 9231752]

6. Mathern GW, Kuhlman PA, Mendoza D, Pretorius JK. Human fascia dentata anatomy and hippocampalneuron densities differ depending on the epileptic syndrome and age at first seizure. J NeuropatholExp Neurol 1997;56:199–212. [PubMed: 9034374]

7. Schmidt-Kastner R, Freund TF. Selective vulnerability of the hippocampus in brain ischemia.Neuroscience 1991;40:599–636. [PubMed: 1676492]

8. Olsson T, Wieloch T, Smith ML. Brain damage in a mouse model of global cerebral ischemia. Effectof NMDA receptor blockade. Brain Res 2003;982:260–269. [PubMed: 12915260]

9. Ginsberg SD, Che S. Expression profile analysis within the human hippocampus: comparison of CA1and CA3 pyramidal neurons. J Comp Neurol 2005;487:107–118. [PubMed: 15861457]

10. Bonaventure P, Guo H, Tian B, Liu X, Bittner A, Roland B, Salunga R, Ma XJ, Kamme F, MeurersB, Bakker M, Jurzak M, Leysen JE, Erlander MG. Nuclei and subnuclei gene expression profilingin mammalian brain. Brain Res 2002;943:38–47. [PubMed: 12088837]

11. Torres-Munoz JE, Van Waveren C, Keegan MG, Bookman RJ, Petito CK. Gene expression profilesin microdissected neurons from human hippocampal subregions. Brain Res Mol Brain Res2004;127:105–114. [PubMed: 15306126]

12. Lein ES, Zhao X, Gage FH. Defining a molecular atlas of the hippocampus using DNA microarraysand high-throughput in situ hybridization. J Neurosci 2004;24:3879–3889. [PubMed: 15084669]

13. Zhao X, Lein ES, He A, Smith SC, Aston C, Gage FH. Transcriptional profiling reveals strictboundaries between hippocampal subregions. J Comp Neurol 2001;441:187–196. [PubMed:11745644]

14. Stoppini L, Buchs PA, Muller D. A simple method for organotypic cultures of nervous tissue. JNeurosci Methods 1991;37:173–182. [PubMed: 1715499]

15. Lagrange P, Livertoux MH, Grassiot MC, Minn A. Superoxide anion production duringmonoelectronic reduction of xenobiotics by preparations of rat brain cortex, microvessels, andchoroid plexus. Free Radic Biol Med 1994;17:355–359. [PubMed: 8001839]

16. Tai YC, Speed TP. A multivariate empirical Bayes statistic for replicated microarray time coursedata. Annals of Statistics. 2006in press

17. Farhangkhoee H, Khan ZA, Barbin Y, Chakrabarti S. Glucose-induced up-regulation of CD36mediates oxidative stress and microvascular endothelial cell dysfunction. Diabetologia2005;48:1401–1410. [PubMed: 15915335]

18. Kuniyasu A, Ohgami N, Hayashi S, Miyazaki A, Horiuchi S, Nakayama H. CD36-mediated endocyticuptake of advanced glycation end products (AGE) in mouse 3T3-L1 and human subcutaneousadipocytes. FEBS Lett 2003;537:85–90. [PubMed: 12606036]

19. Ricciarelli R, D’Abramo C, Zingg JM, Giliberto L, Markesbery W, Azzi A, Marinari UM, PronzatoMA, Tabaton M. CD36 overexpression in human brain correlates with beta-amyloid deposition butnot with Alzheimer’s disease. Free Radic Biol Med 2004;36:1018–1024. [PubMed: 15059642]

20. Bellmann K, Jaattela M, Wissing D, Burkart V, Kolb H. Heat shock protein hsp70 overexpressionconfers resistance against nitric oxide. FEBS Lett 1996;391:185–188. [PubMed: 8706913]

21. Dedeoglu A, Ferrante RJ, Andreassen OA, Dillmann WH, Beal MF. Mice overexpressing 70-kDaheat shock protein show increased resistance to malonate and 3-nitropropionic acid. Exp Neurol2002;176:262–265. [PubMed: 12093104]

22. Yenari MA, Giffard RG, Sapolsky RM, Steinberg GK. The neuroprotective potential of heat shockprotein 70 (HSP70). Mol Med Today 1999;5:525–531. [PubMed: 10562718]

23. Hai T, Wolfgang CD, Marsee DK, Allen AE, Sivaprasad U. ATF3 and stress responses. Gene Expr1999;7:321–335. [PubMed: 10440233]

24. Rage F, Givalois L, Marmigere F, Tapia-Arancibia L, Arancibia S. Immobilization stress rapidlymodulates BDNF mRNA expression in the hypothalamus of adult male rats. Neuroscience2002;112:309–318. [PubMed: 12044449]

25. Naeve GS, Ramakrishnan M, Kramer R, Hevroni D, Citri Y, Theill LE. Neuritin: a gene induced byneural activity and neurotrophins that promotes neuritogenesis. Proc Natl Acad Sci U S A1997;94:2648–2653. [PubMed: 9122250]

Wang et al. Page 12

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

26. Vairapandi M, Balliet AG, Hoffman B, Liebermann DA. GADD45b and GADD45g are cdc2/cyclinB1 kinase inhibitors with a role in S and G2/M cell cycle checkpoints induced by genotoxicstress. J Cell Physiol 2002;192:327–338. [PubMed: 12124778]

27. Wu J, Huang KP, Huang FL. Participation of NMDA-mediated phosphorylation and oxidation ofneurogranin in the regulation of Ca2+- and Ca2+/calmodulin-dependent neuronal signaling in thehippocampus. J Neurochem 2003;86:1524–1533. [PubMed: 12950461]

28. Takamori S, Rhee JS, Rosenmund C, Jahn R. Identification of a vesicular glutamate transporter thatdefines a glutamatergic phenotype in neurons. Nature 2000;407:189–194. [PubMed: 11001057]

29. Druey KM, Blumer KJ, Kang VH, Kehrl JH. Inhibition of G-protein-mediated MAP kinase activationby a new mammalian gene family. Nature 1996;379:742–746. [PubMed: 8602223]

30. Koelle MR, Horvitz HR. EGL-10 regulates G protein signaling in the C. elegans nervous system andshares a conserved domain with many mammalian proteins. Cell 1996;84:115–125. [PubMed:8548815]

31. Hollinger S, Hepler JR. Cellular regulation of RGS proteins: modulators and integrators of G proteinsignaling. Pharmacol Rev 2002;54:527–559. [PubMed: 12223533]

32. Witherow DS, Tovey SC, Wang Q, Willars GB, Slepak VZ. G beta 5.RGS7 inhibits G alpha q-mediated signaling via a direct protein-protein interaction. J Biol Chem 2003;278:21307–21313.[PubMed: 12670932]

33. Nishida M, Maruyama Y, Tanaka R, Kontani K, Nagao T, Kurose H. G alpha(i) and G alpha(o) aretarget proteins of reactive oxygen species. Nature 2000;408:492–495. [PubMed: 11100733]

34. Nishida M, Schey KL, Takagahara S, Kontani K, Katada T, Urano Y, Nagano T, Nagao T, KuroseH. Activation mechanism of Gi and Go by reactive oxygen species. J Biol Chem 2002;277:9036–9042. [PubMed: 11781308]

35. Yan SF, Lu J, Xu L, Zou YS, Tongers J, Kisiel W, Mackman N, Pinsky DJ, Stern DM. Pulmonaryexpression of early growth response-1: biphasic time course and effect of oxygen concentration. JAppl Physiol 2000;88:2303–2309. [PubMed: 10846049]

36. Gene-Ontology-Consortium. The Gene Ontology (GO) project in 2006. Nucleic Acids Res2006;34:D322–326. [PubMed: 16381878]

37. Draghici S, Khatri P, Bhavsar P, Shah A, Krawetz SA, Tainsky MA. Onto-Tools, the toolkit of themodern biologist: Onto-Express, Onto-Compare, Onto-Design and Onto-Translate. Nucleic AcidsRes 2003;31:3775–3781. [PubMed: 12824416]

38. Khatri P, Draghici S, Ostermeier GC, Krawetz SA. Profiling gene expression using onto-express.Genomics 2002;79:266–270. [PubMed: 11829497]

39. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate - a Practical and Powerful Approachto Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological 1995;57:289–300.

40. Kanehisa M, Goto S, Kawashima S, Okuno Y, Hattori M. The KEGG resource for deciphering thegenome. Nucleic Acids Res 2004;32:D277–280. [PubMed: 14681412]

41. Dahlquist KD, Salomonis N, Vranizan K, Lawlor SC, Conklin BR. GenMAPP, a new tool for viewingand analyzing microarray data on biological pathways. Nat Genet 2002;31:19–20. [PubMed:11984561]

42. Giniatullin AR, Darios F, Shakirzyanova A, Davletov B, Giniatullin R. SNAP25 is a pre-synaptictarget for the depressant action of reactive oxygen species on transmitter release. J Neurochem2006;98:1789–1797. [PubMed: 16945102]

43. Frantz GD, Bohner AP, Akers RM, McConnell SK. Regulation of the POU domain gene SCIP duringcerebral cortical development. J Neurosci 1994;14:472–485. [PubMed: 7905511]

44. Kreienkamp HJ, Zitzer H, Gundelfinger ED, Richter D, Bockers TM. The calcium-independentreceptor for alpha-latrotoxin from human and rodent brains interacts with members of the ProSAP/SSTRIP/Shank family of multidomain proteins. J Biol Chem 2000;275:32387–32390. [PubMed:10964907]

45. Hof PR, Rosenthal RE, Fiskum G. Distribution of neurofilament protein and calcium-binding proteinsparvalbumin, calbindin, and calretinin in the canine hippocampus. J Chem Neuroanat 1996;11:1–12.[PubMed: 8841885]

Wang et al. Page 13

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

46. Nirodi C, Hart J, Dhawan P, Moon NS, Nepveu A, Richmond A. The role of CDP in the negativeregulation of CXCL1 gene expression. J Biol Chem 2001;276:26122–26131. [PubMed: 11371564]

47. Hisaka T, Lardeux B, Lamireau T, Wuestefeld T, Lalor PF, Neaud V, Maurel P, Desmouliere A,Kisiel W, Trautwein C, Rosenbaum J. Expression of tissue factor pathway inhibitor-2 in murine andhuman liver regulation during inflammation. Thromb Haemost 2004;91:569–575. [PubMed:14983234]

48. Aisen PS. Inflammation and Alzheimer disease. Mol Chem Neuropathol 1996;28:83–88. [PubMed:8871945]

49. Kang JH, Kim KS, Choi SY, Kwon HY, Won MH, Kang TC. Carnosine and related dipeptides protecthuman ceruloplasmin against peroxyl radical-mediated modification. Mol Cells 2002;13:498–502.[PubMed: 12132593]

50. Ryden, L. Ceruloplasmin. CRC; Boca Raton: 1984.51. Fleming RE, Whitman IP, Gitlin JD. Induction of ceruloplasmin gene expression in rat lung during

inflammation and hyperoxia. Am J Physiol 1991;260:L68–74. [PubMed: 1996664]52. Flower DR. The lipocalin protein family: structure and function. Biochem J 1996;318( Pt 1):1–14.

[PubMed: 8761444]53. Rane MJ, Pan Y, Singh S, Powell DW, Wu R, Cummins T, Chen Q, McLeish KR, Klein JB. Heat

shock protein 27 controls apoptosis by regulating Akt activation. J Biol Chem 2003;278:27828–27835. [PubMed: 12740362]

54. Tezel G, Wax MB. The mechanisms of hsp27 antibody-mediated apoptosis in retinal neuronal cells.J Neurosci 2000;20:3552–3562. [PubMed: 10804196]

55. Wyttenbach A, Sauvageot O, Carmichael J, Diaz-Latoud C, Arrigo AP, Rubinsztein DC. Heat shockprotein 27 prevents cellular polyglutamine toxicity and suppresses the increase of reactive oxygenspecies caused by huntingtin. Hum Mol Genet 2002;11:1137–1151. [PubMed: 11978772]

56. Guzowski JF, Setlow B, Wagner EK, McGaugh JL. Experience-dependent gene expression in the rathippocampus after spatial learning: a comparison of the immediate-early genes Arc, c-fos, and zif268.J Neurosci 2001;21:5089–5098. [PubMed: 11438584]

57. Larsen MH, Olesen M, Woldbye DP, Hay-Schmidt A, Hansen HH, Ronn LC, Mikkelsen JD.Regulation of activity-regulated cytoskeleton protein (Arc) mRNA after acute and chronicelectroconvulsive stimulation in the rat. Brain Res 2005;1064:161–165. [PubMed: 16309632]

58. Akum BF, Chen M, Gunderson SI, Riefler GM, Scerri-Hansen MM, Firestein BL. Cypin regulatesdendrite patterning in hippocampal neurons by promoting microtubule assembly. Nat Neurosci2004;7:145–152. [PubMed: 14730308]

59. Arber S, Caroni P. Thrombospondin-4, an extracellular matrix protein expressed in the developingand adult nervous system promotes neurite outgrowth. J Cell Biol 1995;131:1083–1094. [PubMed:7490284]

60. Yoshihara M, Adolfsen B, Galle KT, Littleton JT. Retrograde signaling by Syt 4 induces presynapticrelease and synapse-specific growth. Science 2005;310:858–863. [PubMed: 16272123]

61. Kuwahara H, Araki N, Makino K, Masuko N, Honda S, Kaibuchi K, Fukunaga K, Miyamoto E,Ogawa M, Saya H. A novel NE-dlg/SAP102-associated protein, p51-nedasin, related to theamidohydrolase superfamily, interferes with the association between NE-dlg/SAP102 and N-methyl-D-aspartate receptor. J Biol Chem 1999;274:32204–32214. [PubMed: 10542258]

62. Przybyla-Zawislak BD, Kim CS, Ali SF, Slikker W Jr, Binienda ZK. The differential JunB responsesto inhibition of succinate dehydrogenase in rat hippocampus and liver. Neurosci Lett 2005;381:354–357. [PubMed: 15896499]

63. Abrous DN, Rodriguez J, le Moal M, Moser PC, Barneoud P. Effects of mild traumatic brain injuryon immunoreactivity for the inducible transcription factors c-Fos, c-Jun, JunB, and Krox-24 incerebral regions associated with conditioned fear responding. Brain Res 1999;826:181–192.[PubMed: 10224295]

64. Lee YH, Ko J, Joung I, Kim JH, Shin J. Immediate early response of the p62 gene encoding a non-proteasomal multiubiquitin chain binding protein. FEBS Lett 1998;438:297–300. [PubMed:9827565]

65. Stumptner C, Fuchsbichler A, Heid H, Zatloukal K, Denk H. Mallory body--a disease-associated typeof sequestosome. Hepatology 2002;35:1053–1062. [PubMed: 11981755]

Wang et al. Page 14

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

66. Khatri P, Sellamuthu S, Malhotra P, Amin K, Done A, Draghici S. Recent additions and improvementsto the Onto-Tools. Nucleic Acids Res 2005;33:W762–765. [PubMed: 15980579]

67. Applied Biosystems, User Bulletin #2: Relative quantitation of gene expression, (PN: 4303859B)(2001).

68. Martinet W, Abbeloos V, Van Acker N, De Meyer GR, Herman AG, Kockx MM. Western blotanalysis of a limited number of cells: a valuable adjunct to proteome analysis of paraffin wax-embedded, alcohol-fixed tissue after laser capture microdissection. J Pathol 2004;202:382–388.[PubMed: 14991905]

AbbreviationsBP

biological process

CC cellular component

DQ duroquinone (2,3,5,6-tetramethyl-p-benzoquinone)

GO gene ontology

LCM laser capture microdissection

MF molecular function

OS oxidative stress

qPCR quantitative real-time PCR

RNS reactive nitrogen species

ROS reactive oxygen species

Wang et al. Page 15

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig 1.Differential, time-dependent cell death in CA1 and CA3 neurons following duroquinone (DQ)treatment (A) and the outline of the treatment procedure selected for the comparative mRNAand protein studies (B). The cell death data shown in (A) were obtained using the PI assay asdescribed under “Experimental Methods”. The data represent the mean (± SEM) percentageof lost cells from the examination of 6 brain slices per time interval of exposure to DQ.

Wang et al. Page 16

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig 2.Hierarchical clustering of genes with similar (A) or divergent (B, C) transcriptional responsesto DQ-induced OS in CA1 and CA3. Relative expression levels represent averages of allreplicate samples across four time periods and are depicted by different colors (color barshown). The mean of all probe sets on the array was centered to 1.0. (A) Genes with similarpatterns of response in CA1 and CA3 and whose expression decreased during and followingDQ treatment were grouped as type “I”, those that increased as type “II”. (B) Clustering ofgroup “I”, “II” and “III” genes that exhibited differential expression patterns between CA1 andCA3. Genes consistently expressed more highly in CA1 than CA3 were labeled as type “I”,those more highly in CA3 than CA1 as type “III”, and those without a consistent pattern ofdifferential expression as type “II”. (C) Enlarged view of type I genes demonstrating twosomewhat different patterns of expression in CA1: low baseline expression followed byincreased expression during and after DQ (from top: Dcn to Itih3, and the last 2 genes at bottom)vs. high baseline expression that remained high (most of the remaining genes). *EST,Expressed Sequence Tag (function unknown); **Gda, the same gene as one above butidentified by a different probeset.

Wang et al. Page 17

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig 3.Correlation of GeneChip and real-time quantitative PCR (qPCR) data. The expression ratiosbetween CA1 and CA3 for five genes, Junb, Mt1a, Nqo1, Nrf2 and Tf, estimated by qPCRwere plotted against those estimated by gene chip microarray analyses. The points representthe ratios estimated at four time points. Three qPCR replicate measurements were made foreach gene at each of the four time points, i.e., each data point represents the mean of threemeasurements. The expression ratios shown represent the means of the ratios from the replicatedeterminations. The median standard error was 0.55 for GeneChip and 0.69 for qPCR. ThePearson correlation coefficient (r) between GeneChip and qPCR data and the estimated p valueare shown at the bottom. The best-fit linear regression line (solid line) and the estimated 95%confidence limits (interrupted lines) are also shown.

Wang et al. Page 18

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Wang et al. Page 19

Table 1Currently known genes showing similar expression patterns in CA1 and CA3 in response to increased OS

Affx_ID Gene Symbol DescriptionGenes whose expression was down-regulated following DQ treatment1368270_at Apobec1 apolipoprotein B editing complex 11373025_at C1qg complement component 1, q subcomponent, gamma polypeptide1389123_at Ccl6 chemokine (C-C motif) ligand 61386901_at Cd36 CD36 antigen1370891_at Cd48 CD48 antigen1368518_at Cd53 CD53 antigen1367776_at Cdc2a cell division cycle 2 homolog A1371250_at Cxcl4 chemokine (C-X-C motif) ligand 41398246_s_at Fcgr3 Fc receptor, IgG, low affinity III1367628_at Lgals1 lectin, galactose binding, soluble 11390510_at Ms4a6b membrane-spanning 4-domains, subfamily A, member 6B1388340_at Ns5atp9 NS5A (hepatitis C virus) transactivated protein 91398540_at Rgs1 regulator of G-protein signaling 11390412_at Slc40a1 solute carrier family 39 (iron-regulated transporter), member 11388809_at Smpdl3a sphingomyelin phosphodiesterase, acid-like 3AGenes whose expression was up-regulated following DQ treatment1369268_at Atf3 activating transcription factor 31370869_at Bcat1 branched chain aminotransferase 1, cytosolic1368677_at Bdnf brain derived neurotrophic factor1398251_a_at Camk2b calcium/calmodulin-dependent protein kinase II beta subunit1390566_a_at Ckmt1 creatine kinase, mitochondrial 1, ubiquitous1368381_at Crtac1 cartilage acidic protein 11370053_at Dlgap1 guanylate kinase associated protein1376345_at Drd1ip dopamine receptor D1 interacting protein (calcyon)1372016_at Gadd45b growth arrest and DNA-damage-inducible 45 beta1388792_at Gadd45g growth arrest and DNA-damage-inducible 45 gamma1387906_a_at Gnas GNAS complex locus1373202_at Gng3 guanine nucleotide binding protein (G protein), gamma 31370997_at Homer1 homer, neuronal immediate early gene, 11370912_at Hspa1b heat shock 70kD protein 1B1389514_at Lrrn6a leucine rich repeat neuronal 6A1367956_at Ncdn neurochondrin1370229_at Ndrg4 N-myc downstream regulated gene 41376362_at Nptxr neuronal pentraxin receptor1370211_at Nrgn neurogranin1386969_at Nrn1 neuritin1368956_at Pcdh8 protocadherin 81367835_at Pcsk1n proprotein convertase subtilisin/kexin type 1 inhibitor1368106_at Plk2 polo-like kinase 2 (Drosophila)1387929_at Pmf31 PMF32 protein1389463_at Prkar1b protein kinase, cAMP dependent regulatory, type I, beta1368505_at Rgs4 regulator of G-protein signaling 41368373_at Rgs7 regulator of G-protein signaling 71368564_at Slc17a6 solute carrier family 17 (sodium-dependent inorganic phosphate co-transporter), member 61371568_at Sncb synuclein, beta1389008_at Spred2 Sprouty-related, EVH1 domain containing 21370042_at Stmn2 stathmin-like 21368562_at Sult4a1 sulfotransferase family 4A, member 11369977_at Uchl1 ubiquitin carboxy-terminal hydrolase L1

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Wang et al. Page 20

Table 2GO terms significantly associated with genes that show similar patterns of expression during OS in neurons ofCA1 and CA3

GO Termsa Total Number Observed Number Expected Valueb Corrected pvaluec

GO terms significantly associated with down-regulated genes

Signal transduction and transportCell surface receptor linked signaltransduction (BP)

112 2 0.16 3.52E-04

Signal transduction (BP) 646 2 0.94 0.010Kinase activity (MF) 408 2 0.59 0.013Transport (BP) 832 2 1.21 0.017Receptor activity (MF) 757 2 1.02 0.039Immune responseImmune response (BP) 174 5 0.25 5.35E-07

GO terms significantly associated with up-regulated genes

Signal transductionActivation of MAPKKK (BP) 3 2 0.01 4.05E-07Negative regulation of proteinkinase activity (BP)

20 2 0.07 1.58E-04

Regulation of GPCR proteinsignaling pathway (BP)

25 2 0.09 2.10E-04

Calmodulin binding (MF) 111 3 0.38 0.002Synaptic transmission (BP) 231 3 0.79 0.005Cell-cell signaling (BP) 171 2 0.58 0.009Signal transduction (BP) 646 5 2.21 0.010Heterotrimeric G-protein complex(CC)

38 2 0.13 0.016

Intracellular signaling cascade (BP) 295 2 1.01 0.026Stress responseResponse to stress (BP) 69 2 0.24 0.002DNA repair (BP) 87 2 0.30 0.003ApoptosisAnti-apoptosis (BP) 86 2 0.29 0.003Apoptosis (BP) 213 2 0.73 0.014NeurogenesisNeuron differentiation (BP) 23 2 0.08 2.28E-04Neurogenesis (BP) 259 3 0.89 0.006

aThe Gene Ontology (GO) database used for the analysis was that of July 2006.

bThe Expected Value was calculated as: ((Total number of differentially expressed genes × Total number of genes in each GO term) / Total number of

genes in the whole gene array).

cThe corrected p value is an indicator of statistical significance of the change in each GO term. The values shown were obtained after FDR (False Discovery

Rate) multiple test correction.

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Wang et al. Page 21

Table 3Currently known genes showing differential expression patterns between CA1 and CA3 following OS.

Affx_ID Gene Symbol Description Rank*Genes whose expression was consistently higher in CA11370956_at Dcn decorin 11370228_at Tf Transferrin 31387011_at Lcn2 lipocalin 2 191367973_at Ccl2 chemokine (C-C motif) ligand 2 211387316_at Cxcl1 chemokine (C-X-C motif) ligand 1 281368420_at Cp ceruloplasmin 291370432_at Pou3f1 POU domain, class 3, transcription factor 1 411368760_at Cxcl2 chemokine (C-X-C motif) ligand 2 421370201_at Calb1 calbindin 1 461368224_at Serpina3n Serine protease inhibitor 481369644_at Lphn2 latrophilin 2 501387068_at Arc activity regulated cytoskeletal-associated protein 521387659_at Gda guanine deaminase 541375877_at Syt4 synaptotagmin IV 571387291_at Itih3 Inter-alpha trypsin inhibitor, heavy chain 3 631387995_a_at Ifitm3 interferon induced transmembrane protein 3 651367577_at Hspb1 heat shock 27kDa protein 1 661390112_at Efemp1 epidermal growth factor-containing fibulin-like extracellular matrix protein 1 741377340_at Tfpi2 tissue factor pathway inhibitor 2 871372294_at Perc64 PE responsive protein c64 (peroxidase and cell adhesion domains) 881388138_at Thbs4 thrombospondin 4 97Genes whose expression was consistently higher in CA31367749_at Lum lumican 91368990_at Cyp1b1 cytochrome P450, subfamily 1B, polypeptide 1 151374787_at Kcnf1 potassium voltage-gated channel, subfamily F, member 1 161379747_at Prss35 protease, serine, 35 171368114_at Fgf13 fibroblast growth factor 13 231390358_at Cacna2d3 calcium channel, voltage dependent, alpha2/delta subunit 3 271370058_at1370059_at

Nfl neurofilament, light polypeptide 32

1368861_a_at Mag myelin-associated glycoprotein 341371057_at Gabra5 gamma-aminobutyric acid A receptor, alpha 5 371372953_at Ncald neurocalcin delta 391368523_at Cadps Ca2+-dependent secretion activator 451373646_at Rab15 RAB15, member RAS onocogene family 591368810_a_at Mbp myelin basic protein 641376895_at Il16_mapped interleukin 16 (mapped) 701369670_at Cd200 Cd200 antigen 911387961_at Opcml opioid-binding protein/cell adhesion molecule-like 931386943_at Pllp plasma membrane proteolipid 95*Based on time-course expression difference detected by timecourse R package.

Genomics. Author manuscript; available in PMC 2008 August 1.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Wang et al. Page 22

Table 4Comparison of Gene Ontology (GO) categories associated with genes expressed at higher levels in CA1 overCA3, or the reverse

GO Termsa Total Number Observed Number Expected Valueb Corrected pvaluec

GO terms significantly associated with CA1-higher genes

Stress ResponseChemokine activity (MF) 21 3 0.04 5.31E-07Cytokine activity (MF) 83 3 0.14 6.07E-05Inflammatory response (BP) 115 3 0.20 1.10E-04Response to unfolded protein (BP) 37 1 0.06 0.001Transition Metal ActivityFerroxidase activity (MF) 2 1 <0.01 1.89E-05Iron ion homeostasis (BP) 15 2 0.03 2.68E-05Copper ion homeostasis (BP) 5 1 0.01 8.37E-05Ferric Iron binding (MF) 5 1 0.01 1.26E-04Copper ion transport (BP) 11 1 0.02 2.42E-04Iron ion transport (BP) 14 1 0.02 3.09E-04Synaptic vesicle transport and neurotransmitter secretionSynaptic vesicle transport (BP) 11 1 0.02 2.57E-04Calcium ion-dependent exocytosis (BP) 15 1 0.03 3.40E-04Synaptic vesicle (CC) 67 1 0.11 5.15E-04Neurotransmitter secretion (BP) 33 1 0.06 0.001GPCR activityGPCR protein signaling pathway (BP) 355 2 0.61 0.006GPCR activity (MF) 254 1 0.43 0.044

GO terms significantly associated with CA3-higher genes

GABA ActivityGABA signaling pathway (BP) 22 1 0.04 0.002GABA-A receptor activity (MF) 33 1 0.06 0.006Calcium channel (voltage-gated) activityVoltage-gated calcium channel activity(MF)

32 1 0.06 0.006

Calcium ion transport (BP) 92 1 0.18 0.011Voltage-gated calcium channel complex(CC)

33 1 0.06 0.019

Potassium channel (voltage-gated) activityVoltage-gated potassium channelactivity (MF)

64 1 0.13 0.011

Potassium ion transport (BP) 145 1 0.28 0.020Voltage-gated potassium channelcomplex (CC)

66 1 0.13 0.033

Postsynaptic activityNeurotransmitter receptor activity (MF) 45 1 0.09 0.008Synaptic transmission (BP) 231 2 0.45 0.009Postsynaptic membrane (CC) 69 1 0.14 0.025

aThe Gene Ontology (GO) database used for the analysis was that of July 2006.

bThe Expected Value was calculated as described for Table 2.

cThe corrected p value is defined in Table 2.

Genomics. Author manuscript; available in PMC 2008 August 1.