Disease mechanisms revealed by transcription profiling in SOD1-G93A transgenic mouse spinal cord.код для вставкиСкачать
Disease Mechanisms Revealed by Transcription Profiling in SOD1-G93A Transgenic Mouse Spinal Cord Mary K. Olsen, BA, Steven L. Roberds, PhD, Brenda R. Ellerbrock, BA, Timothy J. Fleck, BA, Denise K. McKinley, BA, and Mark E. Gurney, PhD Mutations of copper,zinc-superoxide dismutase (cu,zn SOD) are found in patients with a familial form of amyotrophic lateral sclerosis. When expressed in transgenic mice, mutant human cu,zn SOD causes progressive loss of motor neurons with consequent paralysis and death. Expression profiling of gene expression in SOD1-G93A transgenic mouse spinal cords indicates extensive glial activation coincident with the onset of paralysis at 3 months of age. This is followed by activation of genes involved in metal ion regulation (metallothionein-I, metallothionein-III, ferritin-H, and ferritin-L) at 4 months of age just prior to end-stage disease, perhaps as an adaptive response to the mitochondrial destruction caused by the mutant protein. Induction of ferritin–H and –L gene expression may also limit iron catalyzed hydroxyl radical formation and consequent oxidative damage to lipids, proteins, and nucleic acids. Thus, glial activation and adaptive responses to metal ion dysregulation are features of disease in this transgenic model of familial amyotrophic lateral sclerosis. Ann Neurol 2001;50:730 –740 An important clue to the pathogenesis of amyotrophic lateral sclerosis (ALS) was provided by the discovery of mutations in the SOD1 gene encoding copper, zincsuperoxide dismutase (cu,zn SOD) in patients with a familial form of ALS.1 ALS causes the progressive loss of motor neurons from cortex, brainstem, and spinal cord with consequent paralysis and death. From 5 to 8% of all cases of ALS show familial inheritance, and of those roughly 20% are caused by SOD1 mutations. The disease has an autosomal dominant mode of inheritance, indicating that one copy of the mutant gene is sufficient to cause disease. The SOD1 mutations are predominantly missense mutations that substitute one amino acid for another; 92 different mutations at 67 different sites have been found to date (eg, OMIM no. 105400). These are spread throughout the length of the protein sequence. Mutant human cu,zn SOD causes motor neuron disease when expressed at high levels in transgenic mice, indicating that disease is caused by the presence of the altered protein.2 Expression of wild type human cu,zn SOD at the same level does not cause disease. Several different strains of mutant cu,zn SOD transgenic mice have been created.2–5 These differ in the SOD1 mutation analyzed, the species of the SOD1 gene into which the mutation was placed, and the level of expression of the mutant protein; all of which affects the timing and severity of motor neuron disease. Within the spinal cord, ventral horn motor neurons express the highest levels of mutant human cu,zn SOD, perhaps explaining the selective vulnerability of this neuronal population.6 The SOD1-G93A transgenic mice [TgN(SOD1G93A)Gur1] used in this study are healthy until 3 months of age, at which point they develop motor symptoms such as shaking in the limbs when suspended.7 Over the next month the mice grow weaker, their stride shortens, and eventually they become paralyzed as a result of the loss of motor neurons from the spinal cord ventral horn. Complete paralysis occurs between 4 and 5 months of age. The sequence and timing of behavioral and pathological changes in the SOD1-G93A mice are surprisingly stereotyped. Extensive pathology develops in the spinal cord well before the display of overt symptoms of motor neuron disease. As early as 1 month of age, subtle vacuolar changes are seen in motor axons where they enter the ventral roots, and high molecular weight aggregates of cu,zn SOD can be discerned by biochemical assay.7,8 By 2 months of age, vacuolar changes have extended to the motor neuron cell body. Fragmentation of the Golgi apparatus is seen, along with widespread vacuolar changes in the mitochondria caused by the swelling of the mitochondrial intermembrane space.4,9,10 Glial activation is From the Genomics Research Unit, Pharmacia Corporation, Kalamazoo, MI. Published online Nov 1, 2001; DOI 10.1002/ana.1252 Received Apr 3, 2001, and in revised form Jul 5. Accepted for publication Jul 31, 2001. 730 © 2001 Wiley-Liss, Inc. Address correspondence to Dr Gurney, deCODE Genetics, Lynghals 1, IS-110 Reykjavik, Iceland. E-mail: mark@deCODE.is Table 1. Disease Progression in SOD1-G93A Transgenic Mice Age (days) Symptoms Spinal Cord Pathology 30 60 Healthy Healthy 90 Shaking in limbs when suspended, disinhibition of spinal reflexes Weakness, shortening of stride, decrease in nightly running Slight vacuolar changes in motor axons entering the ventral root Ventral horn motor neurons show fragmentation of the golgi apparatus, extensive vacuolar changes involving both endoplasmic reticulum and mitochondria Evidence of motor neuron loss, denervation and reinervation of motor endplates, demyelination of descending spinal tracts Motor neuron loss, astrocytic and microglial cell activation, ubiquitinated inclusions visible in ventral horn 120 prominent by 3 months of age, as is formation of ubiquitinated, cytoplasmic inclusion bodies11; some or most of which contain aggregated SOD.11–14 DNA chip technology was used to compare and contrast alterations in gene expression over the course of disease in transgenic SOD1-G93A mutant mice. Gene expression monitoring was based on hybridization to high-density arrays of synthetic oligonucleotides (Affymetrix, Santa Clara, CA). Such array-based methods allow the simultaneous monitoring of large numbers of probe sequences for expression in tissues and for alteration of their expression because of disease. Monitoring of gene expression in whole spinal cord tissue allowed us to monitor changes in bulk cellular activation and overall changes in cellular composition as well as homeostatic responses to neural degeneration, myelin fragmentation, and lipid release. Extensive microglial and astrocytic activation concident with the onset of paralysis at 3 months of age is reflected in striking transcriptional alterations in genes involved in inflammation, scavenging, and resorption. As disease severity worsens, this is followed by the activation of genes involved in metal ion regulation. Materials and Methods The colonies of TgN(SOD1-G93A)1Gur and TgN(SOD1)2Gur transgenic mice, obtained from the Jackson Laboratory, were housed and bred as described previously,2,15 in accordance with Institutional Animal Care and Use Committee guidelines. For each time point and genotype, three to five mouse spinal cords were obtained by “blow out” and pooled. Genotyping of mice was by polymerase chain reaction (PCR). C57BL/6J-Apoetml UNC mice were obtained from the Jackson Laboratory. The combined silver–cholinesterase stain for motor axons at the neuromuscular junction has been described elsewhere.2 Total RNA was purified from mouse spinal cord using Tri-Reagent (Life Technologies, Carlsbad, CA), and poly(A)⫹ RNA was isolated with a Qiagen kit (Venlo, The Netherlands). Samples for hybridization were prepared by reverse transcription of cDNA followed by in vitro transcription of RNA in the presence of biotinylated nucleotides. Affymetrix murine 6500 oligonucleotide microarrays were processed as recommended by the manufacturer using a GeneChip Fluidics Station 400. Labeled samples prepared from each pool of mouse spinal cords were hybridized with a single murine 6500 GeneChip. The probed arrays were scanned with a Gene-Array Scanner (Hewlett Packard, Palo Alto, CA). The scanned images were analyzed using Affymetrix GeneChip v 3.1 software. Images were scaled globally to compensate for minor variations in fluorescence, and to bring the mean average difference intensity (ADI) for all of the genes on each array to 2,500 intensity units (IU). Each gene discussed in the text is cross-referenced by its GenBank Accession number (eg, W41883 for thymosin B4). OMIM accession numbers refer to entries in Online Inheritance in Man (www.ncbi.nlm.nih.gov/entrez/query.fcgi?db⫽OMIM). Results Gene Expression Changes in SOD1-G93A Mouse Spinal Cord The progression of disease in SOD1-G93A transgenic mice is outlined in Table 1. Figure 1 shows the timing of sample collection from SOD1-G93A transgenic mice, nontransgenic littermate controls (NonTg), and control SOD1 transgenic mice. Both transgenes are expressed at high levels in these two lines such that the SOD1-G93A and SOD1 transgenic mice are balanced for human cu,zn SOD protein and elevation of total dismutase activity toward superoxide in the mouse spinal cord.7,16 Looking at overall patterns of gene expression profiled with the Affymetrix murine 6500 chips, there were essentially no differences in gene transcript levels Fig 1. Times of tissue harvest (days) and comparisons made between SOD1-G93A, NonTg and SOD1 transgenic mice. Olsen et al: Gene Expression Profile in ALS Transgenic Mice 731 vation of scavenge pathways for proteins and lipids, and dysregulation of metal ion homeostasis. Fig 2. (A) Comparison of gene expression profiles in SOD1 transgenic versus nontransgenic mice (A; R 2 ⫽ 0.925), and in mutant SOD1-G93A at end-stage disease (120 days) versus nontransgenic mice (B; R 2 ⫽ 0.841). between the SOD1 and NonTg spinal cords sampled at 120 days (R2 ⫽ 0.925; Fig 2). Considerably more dispersion was seen when comparing transcript levels between SOD1-G93A spinal cord at 120 days and NonTg (R2 ⫽ 0.841; see Fig 2) or SOD1 mouse spinal cord (not shown). Changes in transcript profiles at 30 and 60 days of age between SOD1-G93A mice and NonTg littermate controls were negligible. Transcripts for thymosin B4 (W41883) and a 21kDa peptide under translational control (X06407) were increased more than twofold at 30 days of age, but returned to control values at older ages (Fig 3). Thymosin B4 gene transcription was developmentally regulated with a sharp increase at 120 days of age in all three genotypes. Transcripts for the KIF3-associated protein (KAP3A/B), a regulator of fast axonal transport, were previously shown to be upregulated in the SOD1-G86R transgenic model17; however, no alteration in that transcript was seen in SOD1G93A mice. Changes in gene expression in SOD1-G93A mice due to disease became prominent at 90 and 120 days of age, consistent with changes in glial activation, acti- 732 Annals of Neurology Vol 50 No 6 December 2001 Changes in Cellular Composition and Glial Activation Motor neurons are a minor neuronal population in spinal cord. In SOD1-G93A mice, they are depleted roughly 50% at end-stage disease.7 Perhaps correlated with neuron loss, a decrease in the neurofilament NF-L (M20480) and NF-H (M35131) transcripts was seen in SOD1-G93A mouse spinal cord at end-stage disease, whereas there was no change in NF-M (X05640) transcripts (see Fig 3). Changes in glial activation were the earliest expression changes seen in the SOD1-G93A mice. Astrocytic activation is a prominent histological feature of motor neuron disease in the SOD1-G93A transgenic mice.9 This was reflected by increased expression of glial fibrillary acidic protein GFAP(K01347) and vimentin (M26251) mRNA transcripts at 90 and 120 days of age. Expression of the microglial activation marker CD63 (D16432) also was increased at 90 and 120 days of age (Fig 4). CD63 is expressed on mouse microglial cells and on mouse macrophages (where activation increases expression), although its distribution in brain has not been reported.18,19 Expression of the ␣ integrin MAC-1/ CD11B (X07640) found on macrophages and microglial cells was induced at 120 days of age in SOD1G93A mice. LFA-1 (M60778) expression was also elevated slightly at 120 days of age. Astrocytic and microglial activation is normally accompanied by an increase in MCP-1, a chemotactic factor for microglia and perhaps activated astrocytes,20 but expression of MCP-1 (U50712) was not altered by disease in these mice. Of the four macrophage inflammatory protein (MIP) genes, only MIP-1a (M73061) was present on the murine 6500 chip. Its expression increased with age, but was not altered by disease. Expression of transcripts encoding tumor necrosis factor-␣ (TNF␣), interleukin-1␣ (IL-1␣), interleukin-1␤ (IL-1␤), interleukin-5 (IL-5), interleukin-10 (IL-10), and transforming growth factor-␤3 were detected in spinal cord but were not altered by disease, whereas expression of interleukin-6 (IL-6) and epidermal growth factor were not detected in mouse spinal cord with the Affymetrix murine 6500 chip. Increased expression of interleukin-1␤ convertase has been reported previously in SOD1-G93A mice,21 but the transcript was not detected by hybridization to the Affymetrix chip. Increased expression of complement C1Q␤ has been reported previously in sporadic ALS,22 and we extend that result to the SOD1-G93A transgenic mice (see Fig 4). C1Q␣ and C1Q␤ transcripts were expressed normally and at low levels at 30 and 60 days of age, but increased rapidly by 90 and 120 days of age (see Fig 4). The complement regulators clusterin (apoJ), MCP, Fig 3. The earliest significant change in gene expression in SOD1-G93A transgenic mouse spinal cord is an increase in transcripts for thymosin B4 and a 21kDa polypeptide under translation control (X06407). At end-stage disease (120 days), alterations in NF gene transcripts are observed, while glial activation as indicated by an increase in GFAP. Vimentin gene expression is detected with onset of symptoms at 90 days of age. G93A, lines with squares; NTG, lines with triangles; SOD1, arrowhead. DAF, and CD59 were not present on the mouse 6500 chip, whereas C3 and complement receptor mRNA expression were detected at low levels that did not vary. Increased complement biosynthesis by microglia is reported after transient global cerebral ischemia, in Huntington’s disease, and in Alzheimer’s disease.23–25 All three forms of nitric oxide synthase are present on the murine 6500 chip. nNOS gene expression is elevated in SOD1-G93A mice at 120 days of age (1,038 ADI) as compared with SOD1 and nonTg mice (308 ADI and 431 ADI, respectively), whereas eNOS (2,086 ADI, 2,985 ADI, and 2,948 ADI, respectively) and iNOS (888 ADI, 1,563 ADI, and 1,407 ADI, respectively) gene transcripts are unchanged. Probes for cyclooxygenase-1 (COX-1) and COX-2 are not present on the chip. COX-2 is increased significantly in symptomatic SOD1-G93A mouse spinal cord.26 Apolipoprotein E Among the most striking changes in gene expression occurring in end stage SOD1-G93A spinal cord were those associated with induction of scavenge pathway genes for lipid mobilization and protein catabolism. Apolipoprotein E (ApoE; AA036067) gene expression was normal until 3 months of age, and then was induced more than fivefold in SOD1-G93A at 120 days of age, just prior to end-stage disease (Fig 5). ApoD (L39123) gene expression similarly was increased in SOD1-G93A mice at 4 months, although to a much lesser extent; ApoA-I (X64263) and ApoA-II (U35456) Olsen et al: Gene Expression Profile in ALS Transgenic Mice 733 Fig 4. Markers of microglial activation increase as early as 90 days of age coincident with the onset of symptomatic disease in SOD1-G93A transgenic mice. Lines with squares ⫽ G93A; lines with triangles ⫽ SOD1, arrowhead. were detected at low levels but remained unchanged. ApoE and ApoD are strongly induced in peripheral nerve as a consequence of nerve damage; ApoA-I and ApoA-IV accumulate to a lesser extent.27 Immunoblot analysis of spinal cord extracts from 4-month-old SOD1-G93A and control SOD1 and nontransgenic mice indicated that there was an increase in ApoE protein that paralleled the increase in ApoE gene expression observed by microarray analysis (see Fig 5). As disease develops in the SOD1-G93A mice, loss of motor axons from peripheral nerves stimulates sprouting and reinnervation by the motor axons remaining. Sprouting is robust until about 3 months of age, after which the rapid loss of motor neurons from spinal cord causes extensive denervation in muscle.7 ApoE is produced by macrophages infiltrating into damaged peripheral nerve and may accumulate to 2 to 5% of the total extracellular protein, suggesting that it plays a central role in the mobilization and reutilization of lipid and cholesterol contained in degenerating myelin fragments.28,29 However, peripheral nerve regeneration is apparently unimpaired.30 To determine whether the absence of ApoE might influence sprouting, gluteus muscles were harvested from SOD1-G93A and SOD1G93A;ApoE null mice at 90 days of age and stained with a combined silver–acetylcholinesterase technique to reveal the fine details of motor endplate innervation and morphology. No difference was seen in compensatory motor axon sprouting, or in the number of denervated endplates (Table 2). To examine the role of ApoE induction on disease severity, the SOD1-G93A locus was crossed into an ApoE null background. We then compared the course 734 Annals of Neurology Vol 50 No 6 December 2001 of disease in SOD1-G93A and SOD1-G93A;ApoE null mice littermates from crosses of SOD1-G93A; ApoE ⫺/⫹ mice that were heterozygous at both loci. Fewer SOD1-G93A;ApoE null mice were bred than expected, suggesting either embryonic or perinatal lethality. The distribution of other genotypes in the cross indicates that the SOD1-G93A and ApoE loci segregate independently. Interestingly, the absence of ApoE prolonged survival of SOD1-G93A transgenic mice by about 10 days (Fig 6, Table 3). The same prolongation of survival was seen in a second cohort of SOD1G93A;ApoE null mice that also were bred by crossing SOD1-G93A,ApoE⫺/⫹ mice heterozygous at both loci. Thus, an absence of ApoE appears to confer protection in the disease model; however, the cross does not rule out the less likely possibility that the prolongation of survival is caused by cosegregation of a second neuroprotective gene contributed by the C57Bl6 parent that is tightly linked to the mouse ApoE locus, and that therefore is not separated by recombination. Table 2. Denervation, Sprouting, and Reinnervation in the Gluteus Muscles of SOD1-G93A and SOD1-G93A;ApoE Null Mice Genotype SOD1-G93A SOD1-G93A; ApoE null Innervated Sprouts Denervated 77 ⫾ 6 79 ⫾ 6 4.8 ⫾ 1.3 4.9 ⫾ 2.1 19 ⫾ 6 15 ⫾ 4 Values tabulated are mean ⫾ SEM. Approximately 100 endplates were scored per muscle. Fig 6. Survival analysis for the timing to end-stage disease of SOD1-G93A (lines with triangles) and SOD1-G93A;ApoE (KO; lines with squares) null mice. The absence of ApoE causes a slight but significant prolongation of the timing to end-stage disease. Fig 5. ApoE transcripts and protein are increased in SOD1G93A transgenic mice at end-stage disease (120 days of age). The inset shows the hybridization of Affymetrix chip elements to the probes prepared from SOD1-G93A and nontransgenic NonTg mouse spinal cord. The average difference intensity is calculated as the difference in hybridization intensity of the probe to the exact match oligonucleotides (top element in each pair) as compared with the mismatch oligonucleotides (bottom element in each pair). Each gene is probed by 20 pairs of oligonucleotide elements. Scavenge Pathways The strong induction in ApoE and ApoD transcripts during late-stage disease in the SOD1-G93A transgenic mice was paralleled by induction of transcripts for a variety of other proteins involved in lipid or glycolipid mobilization and protein catabolism. Hexosaminidase is a multisubunit enzyme that degrades saccharide substrates containing a terminal, ␤-linked, N-acetylated hexosaminide such as GM2-ganglioside, particular globosides, oligosaccharides, or glycoproteins. The two subunits of hexosaminidase A (HEXA and HEXB) combine with a third, small glycolipid transport protein, the GM2 activator protein (GM2A), which acts as a substrate-specific cofactor for hexosamindase catabolism of GM2-ganglioside. A deficiency of any one of these proteins leads to storage of the ganglioside, primarily in the lysosomes of neuronal cells, and one of the three forms of GM2-gangliosidosis whose clinical phenotype includes juvenile or adult-onset loss of motor neuron viability (OMIM no. 272800, 26880). Both HEXA(U05837) and HEXB(Y00964) were strongly induced in SOD1-G93A mouse spinal cord beginning at 90 days of age, coincident with the average age of clin- ical disease onset (Fig 7). Expression of the GM2A gene remained unchanged. Expression of brain fatty acid binding protein (BFABP, U04827) declined with age but was induced strongly in SOD1-G93A during late-stage disease (see Fig 7). FABPs may act as molecular scavengers, binding oxidized fatty acids produced by oxidative stress and perhaps induced as a response to lipid peroxidation. Perhaps the induction of B-FABP is a protective response, as lipid peroxidation is increased in SOD1G93A mouse spinal cord at late-stage disease.31 Expression of a variety of lysosomal proteases important to protein catabolism was also elevated in SOD1G93A mice. Cathepsin D and S gene expression was strongly induced beginning at 90 days of age, coincident with the average onset of clinical disease (Fig 8, see Fig 7). The cathepsins are cysteinyl lysosomal proteases that play a role in the catabolism of intracellular and extracellular proteins. Cathepsin D is present in all cell types in the brain, whereas cathepsin S is localized primarily to microglial cells.32 Alterations in Expression of Genes Regulating Metal Ion Homeostasis Metal ion homeostasis is regulated through metal uptake, intracellular storage, and utilization. Copper and zinc are transported into the cell through selective upTable 3. Timing to End-Stage Disease for SOD1-G93A and Littermate SOD1-G93A;ApoE Null Mice Genotype SOD1-G93A SOD1-G93A; ApoE null Days Number p 133 ⫾ 12 143 ⫾ 7 37 11 — 0.0113 Values tabulated are mean ⫾ SD. Olsen et al: Gene Expression Profile in ALS Transgenic Mice 735 Fig 7. Induction of scavenge pathways for lipids, glycolipids and proten catabolism occurs after the onset of symptomatic disease in SOD1-G93A transgenic mice. G93A, lines with squares; NTG, lines with triangles; SOD1, arrowhead. take systems where they are sequestered by specific chaperones or other low-molecular weight, metalbinding proteins called metallothioneins. Mutations in the Wilson’s disease copper transporting ATPase (ATP7B) and Menke’s disease copper-effluxing ATPase (ATP7A) are associated with multisystem degeneration caused by excess tissue accumulation of copper. The ATP7B protein resides in the trans-Golgi network, and transports copper into the secretory pathway where the metal is delivered to vesicle-bound ATP7B by the copper chaperone Atox1 for incorporation into ceruloplasmin.33 A parallel pathway in the cytoplasm via the copper chaperone for SOD (CCS) loads copper into cu,zn SOD. ATP7B transcripts were reduced twofold 736 Annals of Neurology Vol 50 No 6 December 2001 in SOD1-G93A mice at end-stage disease (120 days of age), suggesting downregulation of copper transport into the secretory pathway. Components of the copper regulatory pathway, including ATP7A, CCS, the plasma membrane copper transporter CTR, and ATOX1 (HAH1) were not present on the Affymetrix chip. Expression of the ZNT-1 zinc transporter (U17132) was not altered. Metallothioneins have been proposed to detoxify metals; to play a role in zinc and copper homeostasis; to regulate synthesis, assembly, or activity of zinc metalloproteins; and to protect against formation of reactive oxygen species by free metal ions. Targeted disruption of the mouse MT-I and MT-2 genes renders mice more sensitive to heavy metal toxicity and to oxidative stress.34 Three metallothionein transcripts were detected in mouse spinal cord (MT-I, MT-III, and MT– IV; Fig 9). MT-1 expression occurs widely among tissues. In brain it is expressed principally in astrocytes,35,36 whereas MT-3 expression is restricted largely to neurons.37 MT expression within cells is inducible by zinc through activation of the metallothionein transcription factor MTF-1. As previously reported,38 MT-I and MT-III gene expression was increased in SOD1-G93A mice. Both MT transcripts were induced at end-stage disease in SOD1-G93A mice but were unchanged in SOD1 mice (Fig 10). Low expression of MT-IV also was detected in mouse spinal cord, although it was expressed at similar levels in SOD1, SOD1-G93A, and nontransgenic mice. MTF-1 transcripts were not detected in the experiment. Iron levels in the brain are regulated through uptake and intracellular storage. Iron is needed as a cofactor for enzymes in the electron transport chain, for NADPH reductase activity, for myelination of axons, and for neurotransmitter synthesis. Yet when released in unbound form, Fe2⫹ has the potential to catalyze formation of the strong oxidant, hydroxyl radical, from hydrogen peroxide via the Fenton reaction. Hydroxyl radicals can damage proteins, lipids, sugars, and nucleic acids. Ferritin is the major intracellular iron storage protein,39 consisting of heavy and light subunits (FTH and FTL). The FTH and FTL transcripts accumulated to extraordinarily high levels in diseased SOD1-G93A mouse spinal cord (see Figs 9 and 10). The FTH signal reached a level of 60,000 ADI. This saturated the signal on the GeneChip, whereas the FTL signal was induced threefold to an ADI of approximately 15,500. Accumulation of FTH and FTL transcripts occurred at 120 days of age, at end-stage disease. Iron is transported into the cell bound to transferrin via uptake through the transferrin receptor. Transferrin receptor mRNA (X57349) was low or undetectable in SOD1 or nontransgenic control mouse spinal cord, but was similarly induced at 120 days of age in SOD1- Fig 8. Expression profiles for genes involved in microglial activation, antioxidant defense, electron transport, and ATP synthesis in SOD1-G93A transgenic mice and controls at end-stage disease. G93A mice (to ⬃1,200 ADI). Transferrin probes were not present on the murine 6500 gene chip. Regulation of ferritin and transferrin receptor levels occurs at the level of mRNA stability and translation by binding of iron regulatory protein-1 (IRP-1) or IRP-2 to an mRNA structure motif called the ironresponsive element (IRE).40 In the absence of iron, the IRP bind with high affinity to IRE present in the 5⬘ UTR of ferritin mRNA and repress translation. The level of IRP-1 (X61147) mRNA expression was unchanged (⬃2,300 ADI in the three types of mice); IRP-2 gene expression was not detected with the Affymetrix probes. Antioxidant Enzymes, the Electron Transport Chain, and ATP Synthase No changes in gene expression profiles were seen for SOD1 and SOD2 gene transcripts (note that the murine 6500 chip does not detect the abundant human SOD1 gene transcripts in the SOD1-G93A and SOD1 transgenic mice; see Fig 8). Interestingly, SOD3 (AA071956) transcripts encoding extracellular cu,zn SOD were detected in brain and were down 90% in SOD1-G93A spinal cord. No changes were seen in the expression of transcripts for catalase, different forms of glutathione peroxidase, or thioredoxin-dependent peroxide reductase. No changes were seen in the expres- sion of transcripts encoding components of the electron transport pathway or ATP synthase were noted (see Fig 8), despite the increase in complex I activity reported previously.41 Discussion Debate focuses on the pathogenic mechanism in animal models of ALS with at least two views emerging.42,43 The first is that motor neuron disease in the mutant SOD1 transgenic mice is predominantly a disease of protein or lipid oxidation caused by the excess generation of reactive oxygen species by mutant cu,zn SOD. In this view, disease is caused by metal-catalyzed free radical mechanisms. The second is that disease is caused by the propensity of mutant cu,zn SOD to agFig 9. Differential induction of genes for metal ion regulation in SOD1-G93A transgenic mice at end-stage disease (120 days of age). Olsen et al: Gene Expression Profile in ALS Transgenic Mice 737 Fig 10. Time course of induction of genes important for metal ion homeostasis in SOD1-G93A transgenic mice. G93A, lines with squares; NTG, lines with triangles; SOD1, arrowhead. gregate into cytoplasmic inclusion bodies that impair neuronal function. Of course, both views may be correct because damage by oxygen radicals can cause protein misfolding and consequent aggregation. cu,zn SOD itself is more highly oxidized in SOD1-G93A transgenic mice than in controls, perhaps contributing to the formation of SOD containing cytoplasmic inclusions.44 The changes in gene expression seen in SOD1G93A transgenic mouse spinal cord are consistent with induction of homeostatic mechanisms in response to dysregulation of metal ion sequestration. The ferritin-L and -H genes are induced, as are the MT-1 and MT-3 metallothionein genes, whereas ATP7B transcripts encoding the Wilson’s disease copper transporting ATPase are reduced. In yeast, MT are important regulators of copper homeostasis, but as isolated from mammals the MT primarily are loaded with zinc and cadmium, perhaps because of the low level of free intracellular copper resulting from sequestration by copper chaperones such as CCS and ATOX1/HAH1. MT-1 gene expression is only weakly induced by copper (it is induced more strongly by cadmium and zinc), whereas MT-3 gene expression has not been shown to be inducible by metals. Thus, increased MT gene transcription in SOD1-G93A spinal cord may be an adap- 738 Annals of Neurology Vol 50 No 6 December 2001 tive response to stress rather than a response to an increase in intracellular copper. Induction of ferritin–H and –L chain gene expression in the SOD1-G93A mice is suggestive of an adaptive response to a rise in intracellular iron. What might be the source of that iron? Mitochondrial dysfunction is implied in the SOD1-G93A mice by the mitochondrial pathology that develops with age.4,10,45 Proteins comprising the electron transport chain represent the highest density of iron-binding proteins in the cell. Complex I, for example, contains two Fe2S2 clusters and four to five Fe4S4 clusters or up to 25mol Fe/mol carrier protein. In the entire electron transport chain the ratio is 42 mol Fe/mol carrier protein. Thus, mitochondrial destruction, through releasing iron into the cytoplasm, may induce ferritin–H and –L gene expression as well as induction of ferritin mRNA translation through derepression of IRP in the presence of iron. Free metals (Fe2⫹ or Cu2⫹) can catalyze the formation of the strong oxidant, hydroxyl radical, from hydrogen peroxide via the Fenton reaction, leading to oxidation of proteins, lipids, and nucleic acids. Tissue content of oxygen radicals is increased in SOD1-G93A mouse spinal cord,16,46 as is the content of oxidized proteins,44 products of lipid peroxidation,31 and protein nitration.47 Thus, the increase in ferritin gene transcripts may be a protective response intended to limit Fecatalyzed oxidative damage caused by mitochondrial destruction. The strong fingerprint of microglial activation seen in the SOD1-G93A mice suggests that inhibition of neuroinflammation might be of therapeutic benefit in this model of ALS. COX-2 is elevated in the transgenic model and in sporadic ALS spinal cord,26 whereas COX-2 inhibitors rescue motor neurons in an organotypic model of glutamate toxicity.48 It would seem worthwhile to evaluate COX-2 inhibitors in ALS, although COX-2 may not be the sole source of prostanoids in diseased or damaged spinal cord.49 Our profiling of gene expression in SOD1-G93A mouse spinal cord using high-density oligonucleotide arrays reveals changes in bulk cellular activation consistent with neuroinflammation, metal ion dysregulation, and induction of scavenge pathways for protein catabolism and lipid mobilization. Clearly, it will be of interest to profile gene expression specifically in vulnerable cells, the spinal cord motor neurons, and extend the analysis to sporadic and familial ALS tissues. We thank Carol Himes, Alla Karnofsky, and Tom Vidmar for their assistance and Joan Valentine, Tracey Roualt, and Tom O’Halloran for helpful discussions. References 1. Rosen DR, Siddique T, Patterson D, et al. Mutations in cu/zn superoxide dismutase gene are associated with familial amyotrophic lateral sclerosis. Nature 1993;362:59 – 62. 2. Gurney ME, Pu H, Chiu AY, et al. Motor neuron degeneration in mice that express a human cu,zn superoxide dismutase mutation. Science 1994;264:1772–1775. 3. Ripps ME, Huntley GW, Hof PR, et al. Transgenic mice expressing an altered murine superoxide dismutase gene provide an animal model of amyotrophic lateral sclerosis. Proc Natl Acad Sci U S A 1995;92:689 – 693. 4. Wong PC, Pardo CA, Borchelt DR, et al. An adverse property of a familial ALS-linked SOD1 mutation causes motor neuron disease characterized by vacuolar degeneration of mitochondria. Neuron 1995;14:1105–1116. 5. Bruijn LI, Becher MW, Lee MK, et al. ALS-linked SOD1 mutant G85R mediates damage to astrocytes and promotes rapidly progressive disease with SOD1-containing inclusions. Neuron 1997;18:327–338. 6. Pardo CA, Xu Z, Borchelt DR, et al. Superoxide dismutase is an abundant component in cell bodies, dendrites, and axons of motor neurons and in a subset of other neurons. Proc Natl Acad Sci U S A 1995;92:954 –958. 7. Chiu AY, Zhai P, Dal Canto MC, et al. Age-dependent penetrance of disease in a transgenic mouse model of familial amyotrophic lateral sclerosis. Mol Cell Neurosci 1995;6: 349 –362. 8. Johnston JA, Dalton MJ, Gurney ME, Kopito RR. Formation of high molecular weight complexes of mutant cu,zn-superoxide dismutase in a mouse model for familial amyotrophic lateral sclerosis Proc Natl Acad Sci U S A 2000;7:12571–12576. 9. Hall ED, Oostveen JA, Gurney ME. Relationship of microglial and astrocytic activation to disease onset and progression in a transgenic model of familial ALS. Glia 1998;23:249 –256. 10. Dal Canto MC, Gurney ME. Development of central nervous system pathology in a murine transgenic model of human amyotrophic lateral sclerosis. Am J Pathol 1994;145: 1271–1279. 11. Mourelatos Z, Gonatas NK, Stieber A, et al. The Golgi apparatus of spinal cord motor neurons in transgenic mice expressing mutant cu,zn superoxide dismutase becomes fragmented in early, preclinical stages of the disease. Proc Natl Acad Sci U S A 1996;93:5472–5477. 12. Tu PH, Raju P, Robinson KA, et al. Transgenic mice carrying a human mutant superoxide dismutase transgene develop neuronal cytoskeletal pathology resembling human amyotrophic lateral sclerosis lesions. Proc Natl Acad Sci U S A 1996;93: 3155–3160. 13. Bruijn LI, Houseweart MK, Kato S, et al. Aggregation and motor neuron toxicity of an ALS-linked SOD1 mutant independent from wild-type SOD1. Science 1998;281:1851–1854. 14. Shibata N, Hirano A, Kobayashi M, et al. Presence of Cu/Zn superoxide dismutase (SOD) immunoreactivity in neuronal hyaline inclusions in spinal cords from mice carrying a transgene for Gly93Ala mutant human cu/zn SOD. Acta Neuropathol (Berl) 1998;95:136 –142. 15. Gurney ME, Fleck TJ, Himes CS, Hall ED. Riluzole preserves motor function in a transgenic model of familial amyotrophic lateral sclerosis. Neurology 1998;50:62– 66. 16. Liu R, Althaus JS, Ellerbrock BR, et al. Enhanced oxygen radical production in a transgenic mouse model of familial amyotrophic lateral sclerosis. Ann Neurol 1998;44:763–770. 17. Dupuis L, de Tapia M, Rene F, et al. Differential screening of mutated SOD1 transgenic mice reveals early up-regulation of a fast axonal transport component in spinal cord motor neurons. Neurobiol Dis 2000;7:274 –285. 18. Briers TW, Desmaretz C, Vanmechelen E. Generation and characterization of mouse microglial cell lines. J Neuroimmunol 1994;52:153–164. 19. Miyamoto H, Homma M, Hotta H. Molecular cloning of the murine homologue of CD63/ME491 and detection of its strong expression in the kidney and activated macrophages. Biochim Biophys Acta 1994;1217:312–316. 20. Gourmala NG, Buttini M, Limonta S, et al. Differential and time-dependent expression of monocyte chemoattractant protein-1 mRNA by astrocytes and macrophages in rat brain: effects of ischemia and peripheral lipopolysaccharide administration. J Neuroimmunol 1997;74:35– 44. 21. Friedlander RM, Brown RH, Gagliardini V, et al. Inhibition of ICE slows ALS in mice. Nature 1997;388:31. 22. Grewal RP, Morgan TE, Finch CE. C1qB and clusterin mRNA increase in association with neurodegeneration in sporadic amyotrophic lateral sclerosis. Neurosci Lett 1999;271:65– 67. 23. Schafer MK, Schwaeble WJ, Post C, et al. Complement C1q is dramatically up-regulated in brain microglia in response to transient global cerebral ischemia. J Immunol 2000;164: 5446 –5452. 24. Singhrao SK, Neal JW, Morgan BP, Gasque P. Increased complement biosynthesis by microglia and complement activation on neurons in Huntington’s disease. Exp Neurol 1999;159: 362–376. 25. Yasojima K, Schwab C, McGeer EG, McGeer PL. Up-regulated production and activation of the complement system in Alzheimer’s disease brain. Am J Pathol 1999;154:927–936. 26. Almer G, Guegan C, Teismann P, et al. Increased expression of the pro-inflammatory enzyme cyclooxygenase-2 in amyotrophic lateral sclerosis. Ann Neurol 2001;49:176 –185. Olsen et al: Gene Expression Profile in ALS Transgenic Mice 739 27. Boyles JK, Notterpek LM, Anderson LJ. Accumulation of apolipoproteins in the regenerating and remyelinating mammalian peripheral nerve. Identification of apolipoprotein D, apolipoprotein A-IV, apolipoprotein E, and apolipoprotein A-I. J Biol Chem 1990;265:17805–17815. 28. Ignatius MJ, Gebicke-Harter PJ, Skene JH, et al. Expression of apolipoprotein E during nerve degeneration and regeneration. Proc Natl Acad Sci U S A 1986;83:1125–1129. 29. Snipes GJ, McGuire CB, Norden JJ, Freeman JA. Nerve injury stimulates the secretion of apolipoprotein E by nonneuronal cells. Proc Natl Acad Sci U S A 1986;83:1130 –1134. 30. Popko B, Goodrum JF, Bouldin TW, et al. Nerve regeneration occurs in the absence of apolipoprotein E in mice. J Neurochem 1993;60:1155–1158. 31. Hall ED, Andrus PK, Oostveen JA, et al. Relationship of oxygen radical-induced lipid peroxidative damage to disease onset and progression in a transgenic model of familial ALS. J Neurosci Res 1998;53:66 –77. 32. Petanceska S, Canoll P, Devi LA. Expression of rat cathepsin S in phagocytic cells. J Biol Chem 1996;271:4403– 4409. 33. O’Halloran TV, Culotta VC. Metallochaperones, an intracellular shuttle service for metal ions. J Biol Chem 2000;275: 25057–25060. 34. Kelly EJ, Quaife CJ, Froelick GJ, Palmiter RD. Metallothionein I and II protect against zinc deficiency and zinc toxicity in mice. J Nutr 1996;126:1782–1790. 35. Blaauwgeers HG, Sillevis Smitt PA, De Jong JM, Troost D. Distribution of metallothionein in the human central nervous system. Glia 1993;8:62–70. 36. Nishimura N, Nishimura H, Ghaffar A, Tohyama C. Localization of metallothionein in the brain of rat and mouse. J Histochem Cytochem 1992;40:309 –315. 37. Palmiter RD, Findley SD, Whitmore TE, Durnam DM. MTIII, a brain-specific member of the metallothionein gene family. Proc Natl Acad Sci U S A 1992;89:6333– 6337. 38. Gong YH, Elliott JL. Metallothionein expression is altered in a transgenic murine model of familial amyotrophic lateral sclerosis. Exp Neurol 2000;162:27–36. 740 Annals of Neurology Vol 50 No 6 December 2001 39. Meyron-Holtz EG, Vaisman B, Cabantchik ZI, et al. Regulation of intracellular iron metabolism in human erythroid precursors by internalized extracellular ferritin. Blood 1999;94: 3205–3211. 40. Iwai K, Drake SK, Wehr NB, et al. Iron-dependent oxidation, ubiquitination, and degradation of iron regulatory protein 2: implications for degradation of oxidized proteins. Proc Natl Acad Sci U S A 1998;95:4924 – 4928. 41. Browne SE, Bowling AC, Baik MJ, et al. Metabolic dysfunction in familial, but not sporadic, amyotrophic lateral sclerosis. J Neurochem 1998;71:281–287. 42. Brown RH. Amyotrophic lateral sclerosis: recent insights from genetics and transgenic mice. Cell 1995;80:687– 692. 43. Cleveland DW. From Charcot to SOD1: mechanisms of selective motor neuron death in ALS. Neuron 1999;24:515–520. 44. Andrus PK, Fleck TJ, Gurney ME, Hall ED. Protein oxidative damage in a transgenic mouse model of familial amyotrophic lateral sclerosis. J Neurochem 1998;71:2041-2048. 45. Kong J, Xu Z. Massive mitochondrial degeneration in motor neurons triggers the onset of amyotrophic lateral sclerosis in mice expressing a mutant SOD1. J Neurosci 1998;18:3241–3250. 46. Liu D, Wen J, Liu J, Li L. The roles of free radicals in amyotrophic lateral sclerosis: reactive oxygen species and elevated oxidation of protein, DNA, and membrane phospholipids. FASEB J 1999;13:2318 –2328. 47. Bruijn LI, Beal MF, Becher MW, et al. Elevated free nitrotyrosine levels, but not protein-bound nitrotyrosine or hydroxyl radicals, throughout amyotrophic lateral sclerosis (ALS)-like disease implicate tyrosine nitration as an aberrant in vivo property of one familial ALS-linked superoxide dismutase 1 mutant. Proc Natl Acad Sci U S A 1997;94:7606 –7611. 48. Drachman DB, Rothstein JD. Inhibition of cyclooxygenase-2 protects motor neurons in an organotypic model of amyotrophic lateral sclerosis. Ann Neurol 2000;48:792–795. 49. Schwab JM, Brechtel K, Nguyen TD, Schluesener HJ. Persistent accumulation of cyclooxygenase-1 (COX-1) expressing microglia/macrophages and upregulation by endothelium following spinal cord injury. J Neuroimmunol 2000;111: 122–130.