The brain expression of genes involved in inflammatory response, the ribosome, and learning and memory is altered by centrally injected lipopolysaccharide in mice

Abstract

Neuroinflammation plays a role in the progression of several neurodegenerative disorders. We used a lipopolysaccharide (LPS) model of neuroinflammation to characterize the gene expression changes underlying the inflammatory and behavioral effects of neuroinflammation. A single intracerebroventricular injection of LPS (5 µg) was administered into the lateral ventricle of mice and, 24 hours later, we examined gene expression in the cerebral cortex and hippocampus using microarray technology. Gene Ontology (GO) terms for inflammation and the ribosome were significantly enriched by LPS, whereas GO terms associated with learning and memory had decreased expression. We detected 224 changed transcripts in the cerebral cortex and 170 in the hippocampus. Expression of Egr1 (also known as Zif268) and Arc, two genes associated with learning and memory, was significantly lower in the cortex, but not hippocampus, of LPS-treated animals. Overall, altered expression of these genes may underlie some of the inflammatory and behavioral effects of neuroinflammation.

Introduction

Inflammation plays an important role in the body’s defense against infection, but over-activation of the inflammatory cascade can have detrimental effects. In the central nervous system, inflammation is thought to be associated with normal aging¹ and to contribute to the progression of a number of pathological conditions, including Alzheimer’s disease², Parkinson’s disease³, and amyotrophic lateral sclerosis⁴.

Lipopolysaccharide (LPS) is a component of the outer membrane of Gram negative bacteria, and is commonly used to induce inflammation in the periphery and in the central nervous system. It stimulates the innate immune system through interactions with CD14 and toll-like receptor 4 (TLR4)^{5, 6}, which are expressed by microglia, the primary immunocompetent cells of the brain. Upon exposure to LPS, microglia become activated and begin producing pro-inflammatory mediators such as cytokines, chemokines, prostanoids, and reactive oxygen species⁷.

While LPS is not directly toxic to mature neurons, which do not express TLR4⁸, neurons can be indirectly damaged by LPS-activated microglia. A number of studies in vitro have shown that exposing primary neuron/glia mixed cultures to LPS leads to neuronal death, whereas no significant cell loss is observed in neuron-enriched cultures^{9, 10}. Acute administration of LPS directly to the brain recapitulates these results, and leads to neuron death in the hippocampus¹¹. Moreover, behavioral studies have shown that object recognition¹² and spatial memory¹³ are both impaired in animals treated with single doses of LPS. In line with these cognitive deficits, long-term potentiation (LTP) is largely blocked following acute LPS injections^{12, 14}.

In an effort to characterize the transcriptional changes that underlie these physiological and behavioral effects, we examined changes in gene expression in the mouse hippocampus and cerebral cortex 24 hours after a single intracerebroventricular (i.c.v.) injection of LPS using microarray technology.

Results

We began our analysis by filtering the data to include only probes with a reliable signal. A total of 9,557 annotated genes, RIKEN sequences, and expressed sequence tags passed the filtration step. Next, the raw intensity values from this dataset were normalized with a Z-Transformation¹⁵. The original data file and the filtered, normalized results are available online in the Gene Expression Omnibus (GEO Accession GSE12284). Using the calculated Z-Scores, we conducted a principal component analysis to reduce the dimensionality of the data and to enable visualization of the relationships between gene expression profiles, as shown in Figure 1a. The first principal component (PC1) separated the samples by brain region, and the second (PC2) did so on the basis of treatment. The first and second principal components accounted for 53.2% and 16.1% of total variance, respectively, indicating that brain region and treatment are responsible for the majority of the differences in gene expression between the samples. To quantify the expression changes for each probe, we compared the Z-Scores between the vehicle and LPS treatment groups using the Z-Ratio statistic¹⁵. Positive Z-Ratios indicate increased expression relative to control, and negative Z-Ratios indicate reduced expression. A plot comparing hippocampus and cortex Z-Ratios revealed a significant correlation between the changes in each brain region (r² = 0.42, P < 0.0001; Figure 1b).

Figure 1.

Global gene expression patterns. A) Principal component analysis separates the individual samples into four distinct groups using each sample’s gene expression profile. The hippocampus and cortex show unique patterns of gene expression at baseline (vehicle) and in response to LPS. The first principal component (PC1) separates samples by brain region, and the second (PC2) does so on the basis of treatment. Each point represents one independent tissue sample; eight animals (LPS, n = 4; vehicle, n = 4) are represented by two points each, one for cortex and one for hippocampus. B) LPS injection causes similar transcriptional changes in both brain regions. The changes in expression for all 9,557 transcripts that passed the filtration step are plotted as Z-Ratios for both hippocampus and cerebral cortex. There is a significant correlation between regions (r² = 0.411, P < 0.0001).

To provide a functional interpretation of these effects, we conducted a gene set analysis with the PAGE algorithm using Gene Ontology (GO) terms and curated Molecular Signatures Database (MSigDB) gene sets^16–18. GO terms are gene lists categorized by cellular localization, functional roles, and biological processes¹⁶ that are not necessarily co-regulated. By contrast, curated MSigDB gene sets are lists of genes that have been shown to be commonly affected in human diseases or by various experimental conditions¹⁷. Our results revealed that a number of pro-inflammatory gene sets were enriched in both regions following LPS administration (Table 1). Other gene sets with increased expression included GO Terms for ribosomal proteins and actin cytoskeletal reorganization, as well as an MSigDB set previously shown to be increased in the aged mouse neocortex¹⁹. A number of GO Terms associated with nervous system function showed significantly reduced expression in animals treated with LPS. In addition, we found that LPS caused the downregulation of a set of genes that has been shown to have decreased expression in the hippocampus of patients with Alzheimer’s disease²⁰.

Table 1. Gene sets affected by LPS.

Selected gene sets affected by LPS. We used the Parametric Analysis of Gene Set Enrichment (PAGE) algorithm to find Gene Ontology (GO) terms and curated Molecular Signatures Database (MSigDB) gene sets whose expression was altered during acute neuroinflammation. Significant changes were observed for a number of gene sets associated with inflammation and nervous system function in both brain regions. The PAGE Z-Score reflects the size of a gene set and difference between the average Z-Ratio within the set and the average Z-Ratio for the entire array; Z-Scores with a magnitude greater than ± 1.96 are considered statistically significant with P < 0.05.

		PAGE Z-Score
Gene Set	Size on Array	Hippocampus	Cortex
Inflammatory Response1	77	14.88	12.28
Chemokine Activity1	22	9.96	7.00
Leukocyte Chemotaxis1	7	7.96	6.71
Cytokine Activity1	55	7.93	5.06
Ribosome1	124	8.58	4.08
Aged Mouse Neocortex, Up2	45	4.53	5.45
Neutrophil Chemotaxis1	9	6.27	3.43
Actin Cytoskeleton Reorganization1	7	4.76	3.56
Alzheimer's Disease, Down2	672	−4.29	0.31
Learning and or Memory1	19	−2.33	−2.26
Synaptic Transmission1	89	−2.83	−2.27
Neuron Migration1	31	−2.88	−3.43

¹Gene Ontology Term

²Curated MSigDB Gene Set

We next sought to identify individual genes that were affected by LPS. Using strict criteria, including a Z-Ratio cutoff of ± 1.5 and a P-value < 0.05, we detected significant changes for 170 and 224 transcripts in the hippocampus and cerebral cortex, respectively (Table 2 and Table 3). The complete lists of affected genes are available as supplemental data (Tables S1 and S2). Sixty-four were significantly changed in both brain regions, and all of these were changed in the same direction on the microarray.

Table 2. Genes affected by LPS in hippocampus.

Forty-seven of the 170 genes that were significantly affected in the hippocampus. Many of the upregulated genes are involved in either immune or stress responses, whereas the list of downregulated genes is populated by transcripts that are involved in generalized nervous system function or learning and memory. These 47 were chosen based on the relevance of their biological roles; the complete list is available as supplemental data.

Z-Ratio	Symbol	GenBank Accession	Biological Role
	Inflammatory Response
23.59	Saa3	NM_011315	Monocyte recruitment
22.35	Lcn2	NM_008491.1	Antimicrobial activity
13.22	Ifitm3	NM_025378.1	Induced by interferon signaling
13.21	Ifitm1	NM_026820.2	Induced by interferon signaling
13.16	C3	NM_009778.1	Complement system
9.39	Ccl12	NM_011331.1	Chemotaxis
8.38	Ly6a	NM_010738.2	Lymphocyte proliferation
8.28	B2m	NM_009735.2	MHC antigen presentation
5.49	H2-Q2	NM_010392	MHC antigen presentation
4.88	Emr1	NM_010130.1	Generation of Treg cells
4.46	Stat1	NM_009283.2	JAK/STAT signaling
3.73	Vcam1	NM_011693.2	Lymphocyte and leukocyte migration
3.22	Tnfrsf1a	NM_011609.2	TNF receptor
3.10	Psme1	NM_011189.1	Antigen processing
3.01	Stat3	NM_011486.2	JAK/STAT signaling
2.60	Clu	NM_013492.1	Tissue protection during inflammation
2.57	Slp	NM_011413	Complement system
	Stress Response
9.80	Mt1	NM_013602.2	Metal detoxification and antioxidant defense
7.45	Apod	NM_007470.1	Antioxidant defense
6.92	Mt2	NM_008630.1	Metal detoxification and antioxidant defense
6.00	S100a6	NM_011313.1	Stress response
5.06	Rbm3	NM_016809	RNA binding; stress response
3.82	Hspb1	NM_013560.1	Stress response
2.95	Gpx1	NM_008160.1	Antioxidant defense
	Cytoskeleton
4.95	Vim	NM_011701.3	Component of cytoskeleton
	Lipid Metabolism
2.92	Gpd1	NM_010271.2	Glycerophospholipid metabolism
2.24	Pnpla2	NM_025802.1	Lipid metabolism
	Ribosome
2.74	Rps18	NM_011296.1	Ribosomal protein
2.62	Rps5	NM_009095	Ribosomal protein
2.48	Rpl37a	NM_009084.2	Ribosomal protein
2.31	Rpl15	NM_025586.1	Ribosomal protein
	Learning, Memory, and Neurotransmission
2.40	Grina	NM_023168.2	Glutamate binding; NMDA receptor
2.36	Rasgrf1	NM_011245.1	Learning and memory
	General Nervous System
2.56	Ptpre	NM_011212.2	Antagonism of Kv channels
2.32	Hpca	NM_010471.2	Neuroprotective, Ca2+ binding
2.02	Ctsb	NM_007798.1	CNS maintenance
	Inflammatory Response
−4.54	Selpl	NM_009151.2	Lymphocyte tethering & rolling
	Stress Response
−2.01	Adcy5	XM_358801.1	Regulates stress resistance
	Learning, Memory, and Neurotransmission
−1.59	Grm5	XM_149971.3	Glu receptor, metabotropic
−1.84	Arc	NM_018790.1	Learning and memory
−1.84	Nrxn3	NM_172544.1	Synaptic function
−2.08	Slc6a9	NM_008135.1	Glycine neurotransmission
−2.50	Gabra1	NM_010250.2	GABA neurotransmission
−3.73	Ptn	NM_008973.1	Modulation of LTP threshold
	General Nervous System
−1.82	Neurod2	NM_010895	Neural development
−2.53	Ephb1	NM_173447.2	Neurogenesis & precursor migration
−2.77	Cntn1	NM_007727.1	Axon guidance

Table 3. Genes affected by LPS in cerebral cortex.

Forty-nine of the 224 genes that were significantly affected in the cerebral cortex. As in the hippocampus, the list of transcripts whose expression was increased by LPS includes mostly genes that are involved in either the immune or stress responses, while many of those with reduced expression are involved in nervous system function or learning and memory. These 49 genes were chosen on the basis of their biological roles; the complete list is available as supplemental data.

Z-Ratio	Symbol	GenBank Accession	Biological Role
	Inflammatory Response
21.57	Saa3	NM_011315	Monocyte recruitment
19.05	Lcn2	NM_008491.1	Antimicrobial activity
14.38	Ifitm3	NM_025378.1	Induced by interferon signaling
13.73	Ifitm1	NM_026820.2	Induced by interferon signaling
13.13	Ccl12	NM_011331.1	Chemotaxis
10.04	Ly6a	NM_010738.2	Lymphocyte proliferation
9.87	B2m	NM_009735.2	MHC antigen presentation
5.89	H2-Q2	NM_010392	MHC antigen presentation
3.20	Psme1	NM_011189.1	Antigen processing
3.07	Vcam1	NM_011693.2	Lymphocyte and leukocyte migration
2.53	Ptk2b	NM_172498.1	Leukocyte migration
1.96	Lgmn	NM_011175.1	Lysosomal processing
1.89	C1qg	NM_007574.1	Complement system
	Stress Response
8.88	Mt1	NM_013602.2	Metal detoxification and antioxidant defense
6.00	Apod	NM_007470.1	Antioxidant defense
5.86	Mt2	NM_008630.1	Metal detoxification and antioxidant defense
5.75	Rbm3	NM_016809	RNA binding; stress response
4.79	S100a6	NM_011313.1	Stress response
3.47	Hspb8	NM_030704.1	Stress response
2.08	Gpx1	NM_008160.1	Antioxidant defense
	Cytoskeleton
5.47	Vim	NM_011701.3	Component of cytoskeleton
3.17	Vil2	NM_009510.1	Regulation of cytoskeleton
	Lipid Metabolism
2.53	Dbi	NM_007830.2	Lipid binding
2.14	Pnpla2	NM_025802.1	Lipid metabolism
	Ribosome
2.18	Rps27l	NM_026467.1	Ribosomal protein
1.93	Rpl39	NM_026055.1	Ribosomal protein
1.61	Rpl3	NM_013762	Ribosomal protein
1.57	Rps25	NM_024266	Ribosomal protein
1.51	Rpl7a	NM_013721	Ribosomal protein
	Inflammatory Response
−2.53	Sema4a	NM_013658.2	T cell priming & Th1/Th2 response
−4.51	Dusp1	NM_013642.1	Inhibits macrophage inflammatory response
−4.64	Selpl	NM_009151.2	Lymphocyte tethering & rolling
	Stress Response
−3.03	Hsp105	NM_013559.1	Stress response
	Lipid Metabolism
−4.74	Ppap2b	NM_080555.1	Lipid metabolism
	Learning, Memory, and Neurotransmission
−1.59	Grm5	XM_149971.3	Glu receptor, metabotropic
−1.80	Adcy8	NM_009623.1	Neural plasticity
−1.81	Prkaca	NM_008854	Learning and memory
−2.11	Ptn	NM_008973.1	Modulates LTP threshold
−2.13	Grasp	NM_019518.2	Metabotropic glu receptor trafficking
−2.30	Slc6a9	NM_008135.1	Glycine neurotransmission
−2.39	Cckbr	NM_007627.2	DA receptor sensitivity
−2.80	Ntsr2	NM_008747	Fear memory
−5.27	Egr1	NM_007913.2	Learning and memory
−5.92	Arc	NM_018790.1	Learning and memory
	General Nervous System
−2.35	Mtap1b	NM_008634.1	Neurite outgrowth
−2.45	Ephb1	NM_173447.2	Neurogenesis & precursor migration
−2.47	Olig1	NM_016968.2	Oligodendrocyte differentiation
−2.85	Cx3cl1	NM_009142.2	Neuroprotection
−3.15	Plp	NM_011123	Protein component of myelin

A number of these individually significant genes recapitulate the results of our gene set analysis; genes encoding ribosomal proteins, acute phase proteins, and stress response proteins showed significantly increased expression, whereas those associated with nervous system function, and particularly neurotransmission, learning, and memory, showed significantly reduced expression in both brain regions. Of particular interest are the genes for Activity regulated cytoskeletal-associated protein (Arc) and Early growth response 1 (Egr1, also known as Zif268), which are intimately involved in learning and memory^{21, 22}. The microarray data indicated that Arc expression was significantly reduced in both the hippocampus (Z-Ratio = −1.84) and cerebral cortex (Z-Ratio = −5.92), whereas Egr1 was affected only in the cortex (Z-Ratio = −5.27).

We then validated our microarray results with real-time PCR using primers and probes for a subset of the genes that were found to be individually significant. These genes were chosen based on their biological relevance and for their wide range of positive and negative Z-Ratios. The expression changes for 77% of the selected genes from the cortex were confirmed by PCR results, and 54% were reproducible in the hippocampus (Table 4). A plot of fold change as assessed by PCR versus microarray Z-Ratio revealed a strongly linear relationship for both brain regions (Figure 2a, b). However, one gene, Cxcl12, was shown by PCR to be increased by 34% in the hippocampus despite having a negative Z-Ratio (Z-Ratio = −3.54). PCR showed a significant reduction in Arc mRNA following LPS in the cortex, but there was no statistically significant difference in the hippocampus (Figure 3a, b). Similarly, Egr1 expression was significantly reduced in the cortex, but not in the hippocampus (Figure 3c, d).

Table 4. Genes validated with real-time PCR.

Results of microarray validation with real-time PCR. We chose to validate a total of fifteen genes, eight of which were affected in both cerebral cortex and hippocampus. LPS-induced gene expression changes determined by real-time PCR are presented as fold-change versus the vehicle group. NS indicates that no significant expression change was observed on the microarray; we did not conduct real-time PCR reactions for the majority of these genes (ND).

		Cerebral Cortex		Hippocampus
Symbol	GenBank Accession	Z-Ratio	PCR FC	Z-Ratio	PCR FC
Saa3	NM_011315	21.57	178.27***	23.59	312.48***
Ccl12	NM_011331	13.13	13.05***	9.39	10.11***
Apod	NM_007470.1	6.00	2.48***	7.45	3.27***
Gpx1	NM_008160.1	2.08	1.24***	2.95	1.65***
Grm5	XM_149971.3	−1.59	1.01	−1.58	−1.10
Srf	NM_020493	−1.66	−1.05	NS	ND
Adcy8	NM_009623.1	−1.80	−1.16***	NS	ND
Prkaca	NM_008854	−1.81	1.01	NS	ND
Neurod2	NM_010895	NS	ND	−1.82	−1.06
Gabra1	NM_010250.2	NS	ND	−2.50	−1.04
Ephb1	NM_173447.2	−2.45	−1.31***	−2.53	−1.41**
Mtap1b	NM_008634.1	−2.91	−1.13**	NS	ND
Egr1	NM_007913.2	−5.27	−1.58*	NS	1.09
Arc	NM_018790.1	−5.92	−1.63**	−1.84	−1.27
Cxcl12	NM_013655.2	−6.76	−1.62*	−3.53	1.35*

^*P < 0.05

^**P < 0.01

^***P < 0.001.

Figure 2.

Quantitative real-time PCR validation of the microarray data. a–b) Gene expression changes measured by real-time PCR correlate strongly with changes found by microarray analysis in both cerebral cortex (r² = 0.949, P < 0.0001) and hippocampus (r² = 0.955, P < 0.0001). Because the Z-Ratio statistic incorporates log-normalized Z-Scores, the logarithm of the fold change as assessed by PCR must be used in calculating the linear regression. The strong correlation between PCR results and microarray data indicate that the expression changes observed using the microarray are real and reproducible with more sensitive methods.

Figure 3.

Expression changes for Arc and Egr1 as determined by Quantitative real-time PCR. a–d) Real-time PCR confirmed significant downregulation of Arc and Egr1 transcripts in the cerebral cortex of LPS-treated mice (n = 8) when compared to vehicle control (n = 7), but no significant change was found in the hippocampus. *P < 0.05, **P < 0.01.

Discussion

The goal of this study was to characterize the changes in gene expression that occur in the mouse hippocampus and cerebral cortex in response to direct activation of innate immunity by centrally injected LPS. While previous studies have described the effects of peripheral LPS administration on the whole mouse brain transcriptome^{23, 24}, and one has described the gene expression changes that occur in the hippocampus after a central injection of LPS²⁵, ours is the first to compare the gene expression profiles of these two brain regions during acute neuroinflammation. Our results reveal that while the two brain regions have unique gene expression profiles at baseline, the expression changes induced by LPS are similar in both hippocampus and cerebral cortex. This finding suggests that, on the whole, both regions respond similarly to acute neuroinflammation.

A functional interpretation of these responses was provided by surveying the effects of LPS on predefined sets of genes from the Gene Ontology (GO) Project and Molecular Signatures Database^{16, 17}. Several GO terms associated with inflammation, including Inflammatory Response, Chemokine Activity, Cytokine Activity, Neutrophil Chemotaxis, and Leukocyte Chemotaxis were highly upregulated in both hippocampus and cortex following LPS treatment. In both regions we also detected significant enrichment of a set of genes that was previously shown to have increased expression in the aged mouse neocortex¹⁹, further supporting the concept that brain inflammation increases during normal aging¹. Overall, the results of the gene set analysis in the hippocampus and cortex were strikingly similar, and are indicative of a marked inflammatory response.

Expression of mRNA encoding ribosomal proteins was also increased in both regions at the level of gene sets and individual genes, which suggests that the brain is working to increase its capacity for protein synthesis. In support of this observation, a previous study in rats reported that chronic infusion of LPS caused ultrastructural changes in neurons consistent with impaired cytosolic protein synthesis. The authors also noted deep invaginations in nuclear membranes that were filled with polyribosomes, which they interpreted as an attempt by neurons to compensate for impaired cytosolic translation²⁶. The increased transcription of ribosomal proteins that we observed would support this conclusion. However, since activated microglia are known to contain endoplasmic reticulum rich in ribosomes^{26, 27}, it is possible that the increase we found is at least in part due to transcriptional changes that occur in microglia as they become activated. Further work will be required to determine the individual contributions of neurons and glia to this effect.

The GO terms Learning and/or Memory, Synaptic Transmission, and Neuron Migration were decreased in both hippocampus and cerebral cortex, which suggests that neural dysfunction occurs in both regions as a consequence of acute neuroinflammation. Indeed, LPS-induced neuroinflammation has been shown to adversely affect neurogenesis, neurite outgrowth, and LTP^{12, 28–30}. These are important processes in the adult nervous system, and their disruption likely contributes to the impaired cognitive performance of rodents treated with LPS.

The Ephrin receptor B1 (EphB1) is involved in the differentiation and migration of neural precursors in the adult hippocampus and in the morphogenesis of neurites and dendritic spines^{31, 32}. We observed significantly reduced expression of this transcript in both brain regions (cerebral cortex: Z-Ratio = −2.45, PCR fold change = −1.31; hippocampus: Z-Ratio = −2.53, PCR fold change = −1.41), and this may contribute to the impaired neurogenesis and to the reduced number or dendritic spines observed during neuroinflammation^{28, 30, 33}.

Despite the overall similarities in transcriptional changes in hippocampus and cerebral cortex, the plasticity-associated immediate-early genes Arc and Egr1 were found to be decreased only in the cerebral cortex upon PCR validation. Arc protein has been shown to regulate actin polymerization and AMPA receptor trafficking at synapses, and Egr1 is a transcription factor that regulates the expression of a number of plasticity-associated target genes, including Arc^{21, 34–37}. Both genes are rapidly transcribed postsynaptically following stimuli that produce behavioral learning or LTP, and transgenic animals lacking either gene exhibit impaired memory and weakened LTP^{21, 22, 38–40}. Given the importance of Arc and Egr1 in memory, it is plausible that their reduced expression underlies the memory deficits observed in neuroinflammation. In support of this concept, an antisense-mediated knockdown of Arc protein translation has been shown to impair the maintenance of LTP, but not to affect its induction⁴¹. Hennigan and colleagues (2007) obtained similar results using an acute peripheral injection of LPS to induce neuroinflammation; the authors observed a significant impairment in LTP maintenance, but no change in LTP induction¹². In both studies, the deficits in LTP were evident within one hour of induction. Reduced Egr1 expression may also contribute to inflammation-induced memory impairment; both LPS-treated rats and Egr1^−/− mice perform similarly on tests of object recognition and fail to show preference for novel objects 24 hours after training^{12, 38}. These striking similarities suggest that impaired Arc and Egr1 expression may contribute to the memory problems that occur as a consequence of neuroinflammation. However, because immediate-early genes such as Arc and Egr1 can serve as markers for neural activity^{21, 40, 42}, it is also plausible that the reduction we observed is a consequence of diminished neural activity in the cerebral cortex. Indeed, a study by Vereker et al. (2000), using the same paradigm as Hennigan but with twice the dose of LPS, characterized significant deficits in both the induction and maintenance of LTP, which they attributed to lowered glutamate release¹⁴. These results point to decreased excitatory neurotransmission, which would also produce the reductions in Arc and Egr1 transcription that we observed. Since these models of neuroinflammation used peripheral injections of LPS in rats and we used centrally injected LPS in mice, it is difficult to directly compare the dosing regimens; further studies will be required to determine if glutamate release is affected in our model as well. A possible effect of the anesthetic used before euthanizing the animals should also be considered, as general anesthesia has been shown to interfere with the expression of neuronal immediate-early genes^43–46. Additional work should be conducted to address this possibility.

Rosi and colleagues previously linked chronic inflammation to increased Arc expression in the dorsal rat hippocampus after chronic i.c.v. infusion of low doses of LPS⁴⁷. Though the combined in situ hybridization and cell counting technique used in their study is notably different from the microarray and real-time PCR approach that we employed, the contrast between their results and ours suggest that acute and chronic neuroinflammation may have divergent effects on Arc, and possibly Egr1, transcription. Expression of these two genes has been shown to be lowered in association with β-amyloid plaques in AD patients, in animal models of AD, and in aged mice^{48, 49}. Our results indicate that the inflammatory component of Alzheimer’s and normal aging may be responsible for this effect. In line with this hypothesis, treatment with the non-steroidal anti-inflammatory drug indomethacin restores memory performance and LTP in animal models of AD⁵⁰. However, it is not known whether anti-inflammatory treatment also restores Arc and Egr1 expression to normal levels. Future studies should investigate this possibility.

The results of this study indicate that, on the whole, gene expression profiles of the hippocampus and cerebral cortex are similarly affected by acute neuroinflammation. However, despite the overall similarity between the two brain regions, we observed reduced transcription of plasticity-associated genes in the cerebral cortex, but not hippocampus. The gene expression changes that we have characterized contribute to the understanding of brain inflammation, and may aid in the development of new therapies for diseases with a neuroinflammatory component.

Materials and methods

Animals and Tissue Processing

All animal procedures were conducted under an NICHD-approved animal protocol in accordance with NIH guidelines for the care and use of laboratory animals. Twelve to 15 week-old mice bred on a C57BL/6 and SV129/Ola genetic background were anesthetized with intraperitoneal (i.p.) ketamine (100 mg/kg) and xylazine (10 mg/kg) and placed on a stereotactic table (Kopf Instruments, Tujunga, CA, USA). Using a 33 gauge needle (World Precision Instruments, Sarasota, FL, USA) attached to a microsyringe pump (Stoelting, Wood Dale, IL, USA), sterile phosphate buffered saline (PBS) vehicle (5 µL) or LPS (E. coli serotype 127:B8; 5 µg in 5 µL sterile PBS) was injected into the lateral ventricle (−2.3 mm dorsal/ventral, −1.0 mm lateral, and −0.5 mm anterior/posterior from bregma) at a rate of 1.0 µL/min, as previously described^{11, 33, 51}. Twenty-four hours after surgery, animals were euthanized with an overdose of sodium pentobarbital (100 mg/kg, i.p.) and immediately decapitated. . While other studies have shown a peak in inflammatory markers such as IL-1β or iNOS at 6 hours after systemic LPS injection ^{52, 53}, the dose of LPS used in our study, when injected intracerebroventricularly, produced a peak in most inflammatory markers examined at 24 h in a preliminary time course pilot study. This time point has been shown by us and others to produce significant levels of neuroinflammatory and oxidative stress markers when measured at 24 h after i.c.v. injection of LPS ^{51, 54–56}. The hippocampus and cerebral cortex from both hemispheres were rapidly dissected, flash frozen in 2-methylbutane at −60°C, and stored at −80°C. Total RNA was extracted from hippocampus and cortex samples using Qiagen RNeasy Lipid Tissue Mini and Midi kits, respectively (Qiagen, Valencia, CA, USA). RNA quality was assessed by measuring the A_260/280 ratio. RNA integrity was verified by visualization of the 28S and 18S ribosomal rRNA bands, with no presence of smear, using a denaturing agarose gel.

Microarray

Four samples from each brain region and treatment group were used. A 0.5 µg aliquot of total RNA from each sample was labeled using the Illumina TotalPrep RNA Amplification Kit (Ambion; Austin, TX, USA). RNA was first converted into single-stranded cDNA using reverse transcription with an oligo-dT primer containing the T7 RNA polymerase promoter, and then the single-stranded cDNA was copied to produce double-stranded cDNA molecules. The double-stranded cDNA was used in an overnight in vitro transcription reaction where single-stranded RNA (cRNA) was generated and labeled by incorporating biotin-16-UTP. A total of 0.75 µg of biotin-labeled cRNA was hybridized for 16 hours to Illumina's Sentrix MouseRef-8 Expression BeadChips (Illumina, San Diego, CA, USA). Each BeadChip has 24,000 well-annotated RefSeq transcripts per sample with approximately 30-fold redundancy. The arrays were washed and blocked. Biotinylated cRNA was detected with streptavidin-Cy3 and quantitated using Illumina's BeadStation 500X Genetic Analysis Systems scanner. The image data was extracted using BeadStudio software (Illumina).

Data Analysis

The expression data were filtered to include only probes with a consistent signal on each chip; the cutoff value was established at P < 0.02. The resulting dataset was next analyzed with DIANE 6.0, a spreadsheet based microarray analysis program. An overview of DIANE can be found online at http://www.grc.nia.nih.gov/branches/rrb/dna/diane_software.pdf. Using DIANE, the background was subtracted from the raw signal intensity of each gene, and the results were normalized with a Z-Score transformation¹⁵. Z-normalized data were then analyzed with principal component analysis (PCA). To determine the gene expression changes caused by LPS, Z-Scores for the vehicle and LPS treatment groups were compared using the Z-Ratio statistic¹⁵:

$$Z-\mathit{\text{Ratio}}=\frac{Z-{\mathit{\text{Score}}}_{\mathit{\text{LPS}}}-Z-{\mathit{\text{Score}}}_{\mathit{\text{Vehicle}}}}{\sigma [Z-{\mathit{\text{Score}}}_{\mathit{\text{LPS}}}-Z-{\mathit{\text{Score}}}_{\mathit{\text{Vehicle}}}]}.$$

Gene set analysis was performed using the Parametric Analysis of Gene-set Enrichment (PAGE) algorithm¹⁸ to detect Gene Ontology (GO) Terms¹⁶ and curated Molecular Signatures Database (MSigDB) gene sets¹⁷ affected by LPS. Only MSigDB gene sets with more than 30% representation in the filtered dataset were included in our analysis. Expression changes for individual genes were considered significant if they met four criteria: Z-Ratio above 1.5 or below −1.5; false detection rate (FDR)⁵⁷ of less than 0.30; a P-value statistic for Z-Score replicability below 0.05; and mean background-corrected signal intensity greater than zero.

Quantitative real-time Polymerase Chain Reaction

Microarray results were validated by quantitative real-time PCR (qRT-PCR), as previously described^{51, 58}. In addition to the samples used for the microarray, we added 3 and 4 mice for the vehicle (total n=7) and LPS treatment group (total n=8), respectively. The following TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA, USA) were used: Adcy8 (Mm00507722_m1), Apod (Mm00431817_m1), Arc (Mm00479619_g1), Ccl12 (Mm01617100_m1), Cxcl12 (Mm00445552_m1), Egr1 (Mm00656724_m1), Ephb1 (Mm00557961_m1), Gabra1 (Mm00439040_m1), Gpx1 (Mm00656767_g1), Grm5 (Mm00690332_m1), Mtap1b (Mm00485261_m1), Neurod2 (Mm00440465_g1), Prkaca (Mm00660092_m1), Saa3 (Mm00441203_m1), and Srf (Mm00491032_m1). Pgk1 (Mm00435617_m1) was shown not to change by comparing threshold cycle (C_T) values for both vehicle and LPS treatment groups, and was therefore used as an endogenous control. Statistical analysis was performed by calculating the difference in threshold cycle (ΔC_T) between a given target gene and Pgk1, and then by comparing the ΔC_T values between vehicle (n = 7) and LPS (n = 8) treatment groups with student’s t test. The relative expression of target genes was computed using the ΔΔC_T method⁵⁹.