The arachidonic acid 5-lipoxygenase inhibitor nordihydroguaiaretic acid inhibits tumor necrosis factor α activation of microglia and extends survival of G93A-SOD1 transgenic mice

Authors


Address correspondence and reprint requests to Dr Kenneth Hensley, Free Radical Biology and Aging Research Program, Oklahoma Medical Research Foundation, 825 NE 13th Street, Oklahoma City, OK 73104, USA. E-mail: Kenneth-Hensley@omrf.ouhsc.edu

Abstract

Familial forms of amyotrophic lateral sclerosis (ALS) can be caused by mutations in copper, zinc-superoxide dismutase (SOD1). Mice expressing SOD1 mutants demonstrate a robust neuroinflammatory reaction characterized, in part, by up-regulation of tumor necrosis factor alpha (TNFα) and its primary receptor TNF-RI. In an effort to identify small molecule inhibitors of neuroinflammation useful in treatment of ALS, a microglial culture system was established to identify TNFα antagonists. Walker EOC-20 microglia cells were stimulated with recombinant TNFα, with or without inhibitors, and the cell response was indexed by NO2 output. Three hundred and fifty-five rationally selected compounds were included in this bioassay. The arachidonic acid 5-lipoxygenase (5LOX) and tyrosine kinase inhibitor nordihydroguaiaretic acid (NDGA), a natural dicatechol, was one of the most potent non-cytotoxic antagonists tested (IC50 8 ± 3 μm). Investigation of the G93A-SOD1 mouse model for ALS revealed increased message and protein levels of 5LOX at 120 days of age. Oral NDGA (2500 p.p.m.) significantly extended lifespan and slowed motor dysfunction in this mouse, when administration was begun relatively late in life (90 days). NDGA extended median total lifespan of G93A-SOD1 mice by 10%, and life expectancy following start of treatment was extended by 32%. Disease-associated gliosis and cleaved microtubule-associated tau protein, an indicator of axon damage, were likewise reduced by NDGA. Thus, TNFα antagonists and especially 5LOX inhibitors might offer new opportunities for treatment of ALS.

Abbreviations used
ALS

amyotrophic lateral sclerosis

C-tau

cleaved microtubule-associated tau protein

FLAP

5-lipoxygenase-associated protein

GFAP

glial fibrillary acidic protein

LOX

lipoxygenase

LTB4

leukotriene B4

LTB4-12DH

leukotriene B4 12-dehydrogenase

NDGA

nordihydroguaiaretic acid

PGE2

prostaglandin E2

ROS

reactive oxygen species

SOD

superoxide dismutase

TNFα

tumor necrosis factor α

TNF-RI

tumor necrosis factor receptor I

Levels of tumor necrosis factor α (TNFα) have been found to be increased in the CNS of transgenic mice expressing mutant copper, zinc-superoxide dismutase (SOD1) enzymes that cause familial amyotrophic lateral sclerosis (ALS). In a rapidly progressing strain of mouse that expresses a Gly93→Ala93 mutant SOD1, levels of both TNFα and its primary receptor TNF-RI are raised at late presymptomatic stages of disease (Hensley et al. 2002, 2003; Yoshihara et al. 2002). Expression of both cytokine and receptor continues to increase during the paralytic phase of disease (Hensley et al. 2002, 2003; Yoshihara et al. 2002). In this mouse, up-regulation of the TNFα system largely precedes transcriptional up-regulation of other pro-inflammatory gene products and temporally correlates with progression of the disease (Hensley et al. 2002, 2003). Similar elaboration of pro-inflammatory cytokines has been reported in slowly progressing strains of G93A-SOD1 transgenic mice and in G37R-SOD1 mice (Elliot 2001; Nguyen et al. 2001; Kriz et al. 2002). Finally, both TNFα and soluble TNF receptor levels are raised in serum of humans with ALS (Poloni et al. 2000). TNFα is a potent pro-inflammatory cytokine capable of activating microglia and causing neurotoxicity in systems in which neurons have previously been compromised, for instance by accumulation of aggregated proteins (Robertson et al. 2001). It is therefore likely that TNFα is a significant contributing factor to the pathogenesis of ALS.

Based on these findings, a microglial cell culture assay was established to screen compounds for TNFα antagonism in the hope that such agents would have value in the treatment of ALS. Walker EOC-20 murine microglia cells were treated with recombinant TNFα, with or without inhibitors, and cell response was determined by NO2 release into the culture medium. A panel of 355 rationally selected compounds was assessed for ability to inhibit NO2 production. A series of arachidonic acid 5-lipoxygenase (5LOX) inhibitors were found to be especially potent TNFα antagonists. The natural dicatechol nordihydroguaiaretic acid (NDGA), a selective 5LOX inhibitor from the creosote plant (Larrea tridentata: Zygophyllaceae), was the most potent member of this series.

The cell culture findings prompted an investigation of 5LOX expression in spinal cords of G93A-SOD1 mice. In spinal cord from 120-day-old transgenic mice, levels of 5LOX mRNA and protein were significantly increased. On the basis that 5LOX antagonists might provide therapeutic benefit in ALS, an observer-blinded study was conducted to determine the efficacy of oral NDGA against murine ALS. Dietary administration of NDGA (2500 p.p.m.) significantly extended lifespan and delayed motor deterioration in the G93A-SOD1 mice when treatment was begun at a latter stage of disease.

Materials and methods

Microglial cultures and manipulations

Walker EOC-20 cells (Walker et al. 1995; Hensley et al. 2003) and L292 murine fibroblasts were obtained from the American Tissue Type Collection (Rockville, MD, USA). EOC-20 cells are a well characterized, non-virus-transformed, colony-stimulating factor-1-dependent mouse microglial cell line that expresses typical macrophage cell surface markers including IgG receptors FcγRI and II, Mac-1, Mac-2, Mac-3, CD45 and CD80 (Walker et al. 1995). EOC-20 microglia express major histocompatibility complex (MHC)-I constitutively and express MHC-II in response to γ-interferon (Walker et al. 1995). EOC-20 cells therefore closely resemble primary macrophages and microglia with respect to the presence and inducibility of cell-surface antigens. The response of EOC-20 cells to archetypal inflammatory stimuli has been described recently (Hensley et al. 2003).

EOC-20 cells were grown in 75-cm2 cell culture flasks until passaged into 24-well cell culture plates. Cells were maintained in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum and 20% L292 fibroblast-conditioned medium. L292 cells secrete colony-stimulating factor-1, which promotes microglial proliferation. Recombinant murine TNFα was obtained from Calbiochem (San Diego, CA, USA) or Sigma Chemical (St Louis, MO, USA) and reconstituted in phosphate-buffered saline plus 4% endotoxin-free bovine serum albumin. The dose of TNFα was determined for each batch of cytokine in order to stimulate a 24-h NO2 production of approximately 25 μm (typically 20 ng/mL TNFα).

EOC-20 cells were treated with test compounds dissolved in dimthylsulfoxide (DMSO), or with DMSO vehicle alone (1% final volume) for 30 min before challenge with TNFα. Each test agent was administered at serial 1 : 5 dilutions from 100 μm down to 800 nm, with four replicate wells per dose. After stimulation for 24 h with TNFα, the culture medium was removed and assayed for NO2 using the Griess assay (Marzinzig et al. 1997) with commercially available diazotization reagents (LabChem Inc., Pittsburg, PA, USA). NO2 concentrations were interpolated to calculate the concentration of drug that suppressed TNFα-induced NO2 by 50% (IC50 value). The medium was then replaced with Dulbecco's minimal essential medium lacking phenol red indicator. A commercially available MTS reagent [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethonyphenol)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt; Promega, Madison, WI, USA] was added to each well according to the manufacturer's recommendation. After incubation for 30–45 min at 37°C, aliquots of medium were removed and evaluated spectrophotometrically at 540 nm. A blank was prepared in which the MTS solution was incubated in the absence of cells; this was taken to indicate 0% viability. Viability was formally calculated as the ratio of OD540 in the drug-treated wells, relative to the same variable measured in wells treated with TNFα alone, after subtraction of the blank. An agent was considered cytotoxic if the ratio of LD50 to IC50 was less than 10. Synthetic meso-nordihydroguaiaretic acid [NDGA or 1,4-bis(3,4-dihydroxyphenyl)-2,3-dimethylbutane] was obtained from Sigma or Aldrich (Milwaukee, WI, USA) and used in all animal experiments as well as most cell culture experiments. In specific cell culture experiments natural NDGA was obtained from Sigma. ELISAs for leukotriene B4 (LTB4) and prostaglandin E2 (PGE2) were purchased from Cayman Chemical (San Diego, CA, USA).

G93A-SOD1 mice, drug and diets

Mice expressing high copy numbers of human mutant G93A-SOD1 were obtained from Jackson Laboratories (Bar Harbor, ME, USA) [strain designation B6SJL-TgN-(SOD1 G93A)-1-Gur; Gurney et al. 1994, 1996, 1998]. Transgenic mice were maintained in the hemizygous state by mating G93A males with B6SJL-TGN females. Animals were fed ad libitum standard AIN93G diets or the same diets formulated with NDGA at 2500 p.p.m. with drug administration started at 90 days. At this stage of life, G93A-SOD1 animals begin to demonstrate motor weakness and fine limb tremors (Guégan and Przedborski 2003). The feeding study used 16 mice per group. The group receiving NDGA consisted of five male and 11 female mice; the control group consisted of four male and 12 female mice. All diets and dietary drug formulations were provided by Dyets, Inc. (Bethlehem, PA, USA). All animal procedures were reviewed and approved by the Oklahoma Medical Research Foundation Institutional Animal Care and Use Committee (IACUC) in accordance with National Institutes of Health (NIH) guidelines.

Motor performance was evaluated by means of a rotarod task, by a technician blinded to the treatment groups. Mice were placed on a horizontal rod that was made to rotate at 1 r.p.m. with an acceleration rate of 1 r.p.m. every 10 s until the animal fell from the rod. Each animal was tested three times per trial. Rotarod tests were conducted at 90 days (baseline), 100 days of age and subsequent 5-day intervals. Animals were killed when no longer able to right themselves within 10 s of being placed on their sides. Statistical analyses were performed using the GraphPad PrismTM statistical analysis package (GraphPad Inc., San Diego, CA, USA).

Semiquantitative RT–PCR

Total RNA was isolated from spinal cords of non-transgenic control and G93A+ transgenic mice by using TRI Reagent (Sigma) according to the supplier's protocol. Samples of RNA (5 μg) were reverse transcribed using oligo(dT)15 to prime the reaction in the presence of avian myeloblastosis virus (AMV) RT (Roche, Indianapolis, IN, USA) following the manufacturer's protocol. On completion, each reaction was diluted to a final volume of 50 μL with TE buffer (10 mm Tris, 1 mm EDTA, pH 8.0). PCR amplification of a 309-bp 5LOX gene product from the above-described mouse cDNAs was accomplished with Taq DNA polymerase (Roche), utilizing the buffer supplied and final concentrations of 1.5 mm MgCl2, 0.2 mm of each dNTP and 0.3 μm of each primer. Final reaction volumes were 50 μL. Mouse 5LOX primers were 5′-GGCACCGACGACTACATCTAC-3′ (forward) and 5′-CAATTTTGCACGTCCATCCC-3′ (reverse). Beta-actin primers [5′-CGGCCAGGTCATCACTATTG-3′ (forward) and ACTCCTGCTTGCTGATCCAC-3′ (reverse)] yielding a 353-bp PCR product were used as normalization controls. The number of PCR amplification cycles was determined empirically to yield detectable product bands that were approximately linear with respect to initial cDNA concentration. For 5LOX, optimal cycling conditions were: 2 min at 94°C, one cycle; 1 min at 94°C, 1 min at 56°C and 1 min at 72°C for 27 cycles; 7 min at 72°C, one cycle. Conditions for actin primers were the same except that an annealing temperature of 54°C was used and 24 cycles were performed. Samples of 25 μL from each reaction were electrophoresed in 2% agarose/TBE [tris/borate/EDTA buffer (0.09 M Tris, 0.09 M borate, 0.002 M EDTA)] gels for 1.5 h, stained with ethidium bromide and photographed with a NucleoVision (Nucleotech, Westport, CT, USA) imaging system.

Immunochemistry and histology

Polyclonal anti-SOD1 IgG was purchased from Chemicon (Temecula, CA, USA). Polyclonal anti-5LOX IgG was purchased from Cayman Chemical. Monoclonal anti-5LOX was obtained from Transduction Laboratories (Lexington KY, USA). Polyclonal antibody against glial fibrillary acidic protein (GFAP) was purchased from Research Diagnostics International (Flanders, NJ, USA). The positive control for 5LOX western blots was SL-29 fibroblast lysate (Transduction Laboratories; provided with the antibody). Electrophoresis was performed on 4–20% gradient polyacrylamide gels, and bands were visualized with chemiluminescence detection reagents (Amersham, Piscataway, NJ, USA).

Cleaved microtubule-associated tau protein (C-tau) was measured in the heat-soluble fraction of total spinal cord lysate as described previously (Zemlan 1999; Zemlan and Jauch 1999; Zemlan et al. 1999). Spinal cord samples were homogenized in in 50 mm Tris-HCl (pH 6.8), 0.3 m NaCl, 1%β-mercaptoethanol, 1 mm phenylmethylsulfonyl fluoride and 5 μm leupeptin. Homogenates were centrifuged at 30 000 g for 5 min at 4°C. Supernanants were boiled for 10 min then centrifuged for 30 min at 30 000 g and 4°C. The final supernatants were dialyzed overnight against 50 mm Tris-HCl. C-tau levels in these heat-soluble fractions were measured by sandwich ELISA as described previously, using affinity-purified monoclonal antibody 12B2 (Zemlan 1999; Zemlan and Jauch 1999; Zemlan et al. 1999).

For immunohistochemistry, mice were terminally anesthetized and perfused transcardially with phosphate-buffered saline, pH 7.4, followed by 4% paraformaldehyde in 0.1 m phosphate buffer. The spinal cord was removed and the lumbar L5 region was processed for paraffin embedding. Serial cross-sections of the L5 spinal cord region were cut at 5 μm thickness. Immunostaining was performed using serial L5 spinal cord sections with antibodies to glial GFAP (dilution 5 μg/mL; Chemicon). Negative immunohistochemical controls were treated in the same way but in the absence of primary antibody. GFAP-labeled sections were counterstained with hematoxylin.

Results

Certain 5LOX inhibitors are effective TNFα antagonists in vitro

Microglia respond to TNFα, in part, by synthesizing NO (Colton et al. 1994; Hensley et al. 2003). NO2, an autoxidation product of nitric oxide (NO), can be measured by a simple colorimetric assay of the cell medium (Colton et al. 1994; Marzinzig et al. 1997). PGE2 can also be assessed to indicate inflammatory transcription of cyclo-oxygenase (COX)-II. The NO2 output of EOC-20 microglia is dependent on the TNFα concentration(Fig. 1a). Three hundred and fifty-five individually selected, structurally distinct compounds were screened as inhibitors of TNFα-stimulated NO2 production. The test set was chosen, in part, to maximize the number of targeted enzyme systems. Specific test agents were selected to represent known or suspected anti-inflammatory pharmacophores, or inhibitors of signal transduction elements likely to be involved with an inflammatory event.

Figure 1.

(a) Dose-dependent NO2 output by EOC-20 microglia stimulated with TNFα. (b) Dose-dependent inhibition of NO2 production in EOC-20 cells by minocycline, curcumin and NDGA. (c) Structures of the natural dicatechols NDGA and curcumin. Values in bar graph are mean ± SD.

A range of TNFα-antagonizing activity was observed among a group of compounds that block various enzymes of arachidonic acid metabolism (Table 1). In particular, the 5LOX-inhibiting natural dicatechols curcumin and NDGA were both very effective, non-toxic antagonists of TNFα-stimulated microglial activation (Figs 1b and c). The potency of both compounds compared favorably with that of the benchmark microglia inhibitor minocycline, which suppresses microglial responses in animal models of ALS, Huntington's disease and Parkinson's disease (Chen et al. 2000; Kriz et al. 2002; Wu et al. 2002; Zhu et al. 2002). NDGA was approximately six times more potent than minocycline in vitro, with an IC50 value of 8 ± 3 μm(data are presented as mean ± SD, n = 5 independent experiments) and no toxicity at 100 μm (Fig. 1b). Significant NO2 suppression was observed at 800 nm NDGA. Similar efficacy was observed for natural and synthetic NDGA, as well as for the acetyl ester of NDGA (Table 1). Interestingly, NDGA also suppressed TNFα-stimulated PGE2 production by EOC-20 cells with an IC50 of 841 nm, probably reflecting a suppression of cytokine-induced COX-II expression.

Table 1.  Efficacy of various antagonists of arachidonic acid metabolism against TNFα-stimulated nitrite production by EOC-20 microglia
CompoundPrincipal target(s)IC50 (µM)LD50 (µM)
  • a

    Mean ± SEM (n=5 independent experiments)

  • b Values obtained by linear extrapolation from points  ≤ 100 µm. RTK, receptor tyrosine kinase.

NDGA (natural)5LOX, RTKs8 ± 3aNon-toxic
NDGA acetyl ester5LOX, RTKs5Non-toxic
Tetra-O-methyl-NDGA20125a
Curcumin5LOX, RTKs14Non-toxic
Caffeic acid phenethyl ester5LOX12Non-toxic
Zileuton5LOX105bNon-toxic
MK-886FLAP4140
EbselenLOXs18Non-toxic
SesaminFatty acyl Δ5 desaturase42Non-toxic
Aristolochic acidPLA229100
Arachidonyl trifluoromethyl ketonePLA2577
IndomethacinCOX-I, -II61124b
IbuprofenCOXI, IIInactiveNon-toxic
NS-398COX-II > COX-I108b129b
NimesulideCOX-II > COX-I20178b

A series of structurally distinct 5LOX inhibitors was tested for ability to block TNFα-induced NO2 output by EOC-20 cells; these data are summarized in Table 1. Caffeic acid phenethyl ester was approximately as effective as curcumin, whereas other selective 5LOX inhibitors were bioactive but less potent. Interestingly, tetra-O-methyl NDGA (which does not inhibit LOX; Whitman et al. 2002) displayed modest bioactivity, although it was less potent than the parent compound (Table 1). Variable activity was observed among the several archetypal non-steroidal anti-inflammatory drugs tested. Indomethacin and ibuprofen displayed weak activity, whereas the COX-II selective inhibitor NS-398 was essentially inactive (Table 1). In contrast, the COX-II selective inhibitor nimesulide, which has recently been shown to improve prognosis in ALS mice (Pompl et al. 2003), was active with an IC50 of 20 μm (Table 1). NDGA may have general antioxidant activity; however, a broad series of classical antioxidants (including monocatechols, hindered phenolic chain-breaking antioxidants, thiol reducing agents, SOD and catalase mimetics) were generally inactive against TNFα-stimulated NO2 output (data not shown). Exceptions to this included compounds such as the glutathione peroxidase-mimicking, organoselenium compound ebselen, which is known to be a potent but relatively non-selective LOX inhibitor (Table 1 and Schewe et al. 1994). The TNFα-antagonizing effects of NDGA are probably due to a combination of activities including, but not restricted to, the inhibition of 5LOX.

5LOX is increased in G93A-SOD1 mouse spinal cord

As discussed above, it is likely that the TNFα-antagonizing effects of NDGA do not map exclusively to 5LOX; nonetheless 5LOX is the primary acknowledged target for NDGA. We therefore decided to investigate whether 5LOX expression is affected by the G93A-SOD1 transgene.

5LOX expression was assessed at the message level by semiquantitative RT–PCR and at the protein level by western blot analysis. RT–PCR indicated a significant increase in 5LOX message at 120 days but not at 80 days. After 30 cycles of PCR amplification, the 5LOX PCR product was increased two-fold in samples from G93A-SOD1 mice relative to those from non-transgenic mice (Fig. 2). The increase in 5LOX message levels between 80 and 120 days is reminiscent of similar trends documented recently for pro-inflammatory cytokines including interleukin, TNFα and TNF-RI (Hensley et al. 2002, 2003).

Figure 2.

Semiquantitative RT–PCR reveals raised levels of 5LOX message in G93A-SOD1 mouse spinal cord. Reverse transcription of message and PCR amplification of cDNA were performed as described in Methods. 30 cycles of PCR amplification was used for 5LOX and 24 cycles for actin. PCR products were resolved on agarose gels and stained with ethidium bromide. Each lane represents a single mouse. Values in bar graph are mean ± SD (n = 7 mice at each age). NonTg, non-transgenic; D, days. *p < 0.02 (Mann–Whitney U-test).

Western blot analysis indicated a two-fold increase in 5LOX protein in G93A-SOD1 mouse spinal cord at 120 days, corroborating the RT–PCR results (Fig. 3). Essentially no reactivity was observed when the same samples were probed with antibodies against 12LOX and 15LOX (data not shown). Thus 5LOX expression was similarly increased at both the message and protein levels of analysis in symptomatic G93A-SOD1 mice, suggesting a transcriptional up-regulation of gene for 5LOX within the transgenic mouse spinal cord.

Figure 3.

Alterations in 5LOX protein levels in spinal cords of G93A-SOD1 mice. Western blots indicate up-regulation of 5LOX in spinal cords of G93A-SOD1 mice at 120 days (D) of age. NonTg, non-transgenic.

Attempts were made to measure leukotrienes, the lipid products of 5LOX catalysis, in mouse spinal cord using commercially available ELISAs. LTB4 increased slightly but significantly in the spinal cords from 120-day-old G93A-SOD1 mice relative to levels in non-transgenic littermates (44.5 ± 4.2 vs. 36.4 ± 4.2 ng LTB4 per mg protein respectively; p < 0.033 by t-test n = 4 per group). Only marginal concentrations were observed for other leukotrienes (data not shown). However, it should be noted that conditions have not been optimized for extraction of leukotrienes from CNS tissue, and the ELISA may suffer from limitations of dynamic range and matrix effects.

Oral NDGA influences disease course in G93A-SOD1 mice

Expression of TNFα and its receptor TNF-RI begins to increase at 80 days of age in the fast-progressing strain of G93A-SOD1 mouse (Hensley et al. 2002, 2003; Yoshihara et al. 2002). However, an extreme rise in the level of these two components is not observed until the paralytic stage of disease (Hensley et al. 2002, 2003). Animals begin to experience rapid decline of motor function at 100 days, concomitant with neuron loss (Gurney et al. 1994, 1996, 1998) (Fig. 4a). Thus it should be possible to antagonize TNFα-dependent processes pharmacologically relatively late in the life of the animal. Based on this rationale, NDGA was administered orally to G93A-SOD1 animals beginning at 90 days of age. This drug treatment significantly altered the course of disease, delaying motor dysfunction and extending survival (Fig. 4). Oral NDGA extended median lifespan by 13 days, representing a 32% increase in lifespan beyond the start of treatment (Fig. 4b; p < 0.002, log rank analysis; n = 16 animals/group). This 13-day extension of survival is the same as that achieved by oral administration of riluzole, the only currently approved ALS therapeutic agent, when riluzole treatment is begun at 40–50 days of age (Gurney et al. 1996, 1998). The benefit conferred by NDGA is also similar to that reported for minocycline, which increases the lifespan of the fast-progressing G93A-SOD1 mouse by 11 days when administration is begun at 7 weeks of age (Zhu et al. 2002). It should be noted that in the G37R-SOD1 mouse minocycline extended lifespan by 21 days (Kriz et al. 2002), indicating mutation-specific drug efficacies in murine models of ALS.

Figure 4.

Oral NDGA improves prognosis of G93A-SOD1 mice. (a) Rotarod motor performance evaluation of G93A-SOD1 mice fed a standard AIN93G diet or the same diet supplemented with 2500 p.p.m. NDGA (n = 16 mice/group). Values are mean ± SEM. p < 0.05 for drug effect by repeated measures anova. *p < 0.05 versus control at individual time points (Student's t-test). (b) Survival profiles for G93A-SOD1 mice fed a standard diet or an NDGA-supplemented diet, beginning at 90 days of age (p < 0.002, log rank test; hazard ratio 0.39 for NDGA group relative to control group).

Although G93A-SOD1 mice display measurable muscle weakness between 90 and 110 days of age, obvious signs of paralysis usually become evident at around 115 days. This event can be defined by a number of indicators, including an altered leg-splaying response when the mouse is lifted by the tail (Gurney et al. 1996). NDGA significantly delayed the onset of frank paralysis, as indicated by leg-splaying criteria (Table 2). The mean duration of the paralytic phase of disease (time between onset of paralysis and death) was extended approximately 40% by oral intake of NDGA, which was marginally significant at the 95% confidence interval (p = 0.056, two-tailed t-test; Table 2).

Table 2.  Oral NDGA affects the onset of frank paralysis as well as duration of the paralytic stage of disease in G93A-SOD1 mice
 ControlNDGAp
  1. Oral NDGA affects the onset of frank paralysis as well as duration of the paralytic phase of disease in G93A-SOD1 mice. Data represent 16 animals per group; p values by t-test.

Time to onset of frank paralysis (days)
 Mean ± SD115.9 ± 7.4121 ± 7.10.029
 Median112120 
Interval between onset and death (days)
 Mean ± SD11.5 ± 5.716.5 ± 7.90.053
 Median1014 

Animal weight trends generally reflected survival and rotarod performance data. Mean animal weight began to decrease in G93A-SOD1 animals after 110 days (Table 3). By 120 days, median body mass was significantly lower in surviving control G93A-SOD1 animals than in NDGA-treated G93A-SOD1 mice (p < 0.03; Mann–Whitney U-test).

Table 3.  Body mass of G93A-SOD1 mice is a function of age and drug treatment
Age (days) Body mass (g)
Control dietNDGA diet
  1. Data were obtained from the same animals as used in the experiment illustrated in Fig. 4.

90Mean ± SD22.6 ± 5.222.7 ± 4.9
Median20.220.8
100Mean ± SD22.1 ± 4.822.3 ± 4.0
Median20.321.7
110Mean ± SD21.8 ± 4.522.6 ± 3.7
Median19.821.7
120Mean ± SD19.7 ± 3.921.6 ± 2.9
Median18.020.8*

Astrogliosis, characterized in part by the enhanced expression of GFAP, is a homotypic response of astroglia to diverse types of CNS injury (Little and O'Callagha 2001). It is a major tissue-level phenotype associated with G93A-SOD1 transgene expression (Hall et al. 1998; Drachman et al. 2002). Recent studies of COX-II inhibitors have shown that suppression of astrogliosis with non-steroidal anti-inflammatory drugs correlates with improved prognosis in the G93A-SOD1 mouse model (Drachman et al. 2002). Accordingly, astrogliosis was investigated immunochemically as a function of NDGA administration. As shown in Fig. 5, oral NDGA diminished astrogliosis in the lumbar spinal region of 120-day-old G93A-SOD1 mice relative to that in transgenic mice fed a control diet.

Figure 5.

Reduction of astrogliosis and axon loss in G93A-SOD1 mice induced by oral NDGA. Lumbar spinal cord sections from non-transgenic (a) and 120-day-old G93A-SOD1 mice fed either a basal diet (b) or diet containing NDGA (c) were labeled with anti-GFAP (magnification × 40).

Microtubule-associated tau protein is an axonally localized protein that experiences caspase-dependent proteolytic cleavage at N- and C-termini upon neuronal damage (Zemlan 1999; Zemlan and Jauch 1999; Zemlan et al. 1999). Monoclonal antibodies have been developed that exhibit a 1000-fold higher affinity for C-tau than full-length tau (Zemlan et al. 1999). Increases in C-tau cleavage accurately index experimental neuronal damage in several conditions, including stroke and traumatic brain injury (Zemlan et al. 1999). In the present study, ELISAs for C-tau were used to determine axonal damage in ALS mouse spinal cords as a function of animal age and drug treatment. As illustrated in Fig. 6, spinal cord concentrations of C-tau increased markedly between 80 and 120 days of age in G93A-SOD1 mice. The increase in C-tau approximately correlated with diminution of motor function (Fig. 4), elaboration of pro-inflammatory cytokines (Hensley et al. 2002, 2003), and increased protein oxidative damage (Hensley et al. 2002). These results indicate that C-tau is a convenient marker for neuropathy in the G93A-SOD1 mouse. Oral NDGA administration, begun at 90 days, resulted in a decrease in C-tau at 120 days that closely approached statistical significance (p = 0.062; two-tailed t-test) (Fig. 6).

Figure 6.

Increase in spinal cord C-tau, an indicator of neuronal axon damage, during disease progression in the G93A-SOD1 mouse as determined by ELISA. Values are mean ± SEM. Statistical analysis was by two-tailed Student's t-test for unpaired samples. NonTg, non-transgenic; D, days.

Oral NDGA did not significantly decrease LTB4 levels in the small group of animals assayed. However, caveats regarding leukotriene measurement in CNS tissue should be noted as described above. Taken together these data indicate that oral NDGA positively influenced disease course in the G93A-SOD1 mouse model of ALS.

Discussion

Murine models of ALS exhibit a dynamic neuroinflammatory component manifest by elaboration of pro- and anti-inflammatory cytokines, with a very limited involvement of peripheral lymphocytes (Elliot et al. 2001; Nguyen et al. 2001; Hensley et al. 2002, 2003; Kriz et al. 2002; Yoshihara et al. 2002). A particular need exists to identify inflammatory components that inititate or drive disease progression in ALS. Our findings suggest that TNFα and its primary receptor TNF-RI are among the earliest responding cytokine components in the G93A-SOD1 mouse (Hensley et al. 2002). Moreover both TNFα and TNF-RI expression increase through the progression phase of disease (Hensley et al. 2002, 2003). Because TNFα is a potent activator of microglia and contributes to tissue damage in classical inflammation, it is plausible that it might contribute to ALS pathology. It should be noted that TNFα also has positive effects on neuron viability in some excitotoxicity paradigms (Gary et al. 1998). Determination of the true importance of TNFα in the pathogenesis of ALS must await animal genetic experiments wherein TNFα or its receptor is ablated in the presence of SOD1 mutant transgenes. Nonetheless, an exploration of TNFα biology might yield candidate pharmacological treatments for the neuroinflammatory component of ALS. The present study shows that naturally occurring dicatechols that inhibit 5LOX also antagonize TNFα in microglial cultures. One compound, NDGA, slows disease progression in a fast-progressing strain of G93A-SOD1 mouse when administered orally beginning at 90 days of age, when subtle symptoms become evident (Guégan and Przedborski 2003).

Our findings suggest that the 5LOX pathway may be involved with neuroinflammation in the G93A-SOD1 mouse. Considerable attention has been given to COX as an inflammatory contributor to neurological disease (McGeer and McGeer 2001), but LOX has been studied less. 5LOX selectively oxidizes arachidonate into a lipid hydroperoxide, as the first step in production of leukotriene end-products (Funk 2001; Bigby 2002; Haeggstrőm and Wetterholm 2002; Whitman et al. 2002) (Fig. 7). The 5LOX inhibitor NDGA diminishes infarct volume after stroke (Shishido et al. 2001) but NDGA has not been widely investigated in animal models of neurological disease. NDGA selectively inhibits 5LOX over 12LOX, 15LOX and COX, with reported Ki values of 200 nm, 30 µm, 30 µm and 100 µm respectively (Salari et al. 1984). The mechanism of 5LOX inhibition involves reduction of the iron center in LOX by the catechol functionality of NDGA (Kemal et al. 1987; Whitman et al. 2002). The curry spice component curcumin (turmeric), which is structurally similar to NDGA (Fig. 1), also inhibits 5LOX, although less effectively (Skrzypczak-Jankun et al. 2000).

Figure 7.

Schematic illustration of proposed relationships between TNFα, TNF-RI and the 5LOX activation pathway. *Indicates plausible sites for NDGA action. γGTP, γ-glutamyl transpeptidase; 5-HPETE, 5-hydroperoxy-eicosatetraenoic acid; LTC4S, leukotriene C4 synthase; PLA2, phospholipase A2.

The TNFα-antagonizing effects of NDGA and curcumin may not result solely from their ability to inhibit 5LOX. Both compounds probably inhibit certain tyrosine kinases independently of LOX action. For instance NDGA has been reported to antagonize autophosphorylation of the platelet-derived growth factor receptor (Domin et al. 1994) and we have observed similar effects of NDGA on epidermal growth factor receptor activation. Moreover, tetra-O-methyl-NDGA, which does not inhibit 5LOX (Whitman et al. 2002), exhibits respectable TNFα-antagonizing ability (Table 1). 5LOX is a cytosolic enzyme that translocates to the nucleus after phosphorylation via the p38 mitogen-activated protein kinase pathway (Werz et al. 2000; Bigby 2002) (Fig. 7) which, notably, is activated in SOD1 mutant mice (Kriz et al. 2002). In the nucleus, 5LOX is presented with arachidonic acid by the 5LOX-associated protein (FLAP) (Bigby 2002; Helgadottir et al. 2004). The resulting metabolite LTB4 is a natural ligand for peroxisome proliferator activated receptor (PPARα), which interacts with activator protein-1 (AP-1) to drive transcription of certain genes (Rizzo and Carlo-Stella 1996; Madamanchi et al. 1998; Funk 2001). Interestingly, NDGA has been shown to bind AP-1 directly and thereby interfere with DNA binding by the transcription complex (Kwon et al. 2001). Thus the potency of NDGA probably results from multiple actions on 5LOX, receptor protein kinases and specific transcription factors (Fig. 7).

The involvement of LOX with neuroinflammation in the G93A-SOD1 mouse is substantiated by the finding that 5LOX message and protein levels are increased in degenerating spinal cords. The neurochemical implications of this observation are unclear, because so little is known about LOX function within the CNS. Nonetheless, some possible interpretations may be formulated from existing knowledge of 5LOX function in fibroblasts and macrophages. Leukotrienes formed via LOX are released during classical inflammation, stimulating events such as bronchoconstrictrion during asthmatic episodes. Similar paracrine functions might be assumed by the leukotrienes near motor neurons. There might be another role for 5LOX metabolites in transducing signals from TNFα into the microglial gene expression machinery, via LTB4 acting through peroxisome proliferator-activated receptors and AP-1 (Rizzo and Carlo-Stella 1996; Colville-Nash et al. 1998; Madamanchi et al. 1998; Funk 2001). Thus 5LOX stimulation of transcription probably requires the independent recruitment of separate pathways, such as the C-Jun amino terminal kinase (JNK)–AP1 axis, which are known to be activated in response to TNFα (reviewed in Hallenbeck 2002) (Fig. 7). Alternatively, leukotrienes might influence the sensitivity of neurons to apoptosis-inducing insults, perhaps by stimulating production of reactive oxygen species (ROS) at the level of mitochondria or membrane NADPH oxidase (Fig. 7). Such a hypothesis is supported by the observation that LTB4 uniquely and autonomously stimulates ROS production in fibroblast cultures, and TNFα stimulation of the same ROS flux can be blocked by leukotriene receptor antagonists (Woo et al. 2000).

5LOX is expressed in neurons, astrocytes and microglia, and expression increases with age in mammalian brain (Uz et al. 1998; Manev et al. 2000). Age-related LOX up-regulation might compromise the aging nervous system in a more general fashion not limited to the pathophysiology of ALS. In addition to 5LOX, neurons also express several enzymes that modify or metabolize leukotrienes and some of these enzymes have been circumstantially associated with ALS. A very recent proteomics investigation found that LTB4 12-dehydrogenase (LTB4-12DH) expression is completely absent in human ALS neurons and in neuron-like NSC34 cells engineered to express mutant SOD1 (Allen et al. 2003). Among the 700 proteins surveyed in this study, LTB4-12DH was the most differentially down-regulated component in mutant SOD1-expressing cells (Allen et al. 2003). LTB4-12DH down-regulation would probably increase LTB4 levels, and further activate pathways dependent upon this leukotriene.

Our findings suggest new strategies for management of ALS symptoms. TNFα and 5LOX are accessible targets for pharmacological intervention, at least in the context of systemic inflammatory processes such as rheumatoid arthritis and asthma. Anti-TNFα antibodies and soluble TNF receptor are clinically efficacious for some rheumatoid conditions (Shanahan and St Clair 2002) although neither of these agents has been explored in neurodegenerative diseases. Anti-inflammatory cytokines such as interleukin-10 generally antagonize TNFα, but are not transcriptionally increased in the G93A-SOD1 mouse model (Hensley et al. 2002, 2003). Theoretically, these anti-inflammatory cytokines might suppress neuroinflammation if a way could be found to increase their concentration within the CNS.

Likewise, small molecule inhibitors of TNFα or LOX currently exist. Thalidomide in particular is well known to inhibit TNFα expression. NDGA is not currently approved for human use, although it is a relatively non-toxic substance in its pure form (Lehman et al. 1951). High-dose (> 2% dietary) NDGA causes renal cysts in rats (Evan and Gardner 1979); however, NDGA is tolerated by mice and dogs at 1% of total dietary intake for periods of 6 months to 1 year (Cranston et al. 1947; Lehman et al. 1951). Caution would be warranted in the use of NDGA, however, particularly when considering natural preparations. Crude extracts of the creosote bush have been used as an alternative cancer therapy for some years, but these preparations can cause severe hepatotoxicity (Sheikh et al. 1997). In our hands, the cytotoxic principals of chaparral are separable from NDGA (data not shown), but have not been explicitly identified. Curcumin is a major component of the curry spice turmeric and is currently available for general consumption. Unfortunately the oral bioavailabilities of both curcumin (Cheng et al. 2002) and NDGA (personal observations) are low, but might be improved by proper formulation.

Finally, some LOX inhibitors that have been, or are being developed for the treatment of asthma may be considered for use in neurological disease. For instance zileuton (Table 1) is a hydroxyurea derivative of benzo[b]thiophene approved for management of asthma (Dube et al. 1999). Further preclinical and basic scientific research is required to determine whether these agents might be exploited for clinical benefit in ALS and other neurodegenerative conditions.

The EOC-20 microglial screen that resulted in identification of NDGA as a target compound might offer other compounds for preclinical assessment. A review of the literature revealed that several compounds included in the preliminary EOC-20 screen have also been tested in the G93A-SOD1 mouse and found to be effective (summarized in Table 4), including riluzole, a surprisingly effective inhibitor of PGE2 output. Other compounds, such as creatine and co-enzyme Q10, were not effective TNFα antagonists but have been reported to improve prognosis in the ALS mouse. Thus microglial bioassays have the potential to identify novel anti-neuroinflammatory compounds, whereas other pharmacological targets such as bioenergetic agents may not be indexed by such cell culture models.

Table 4.  Summary of compounds that have reported in vivo efficacy in the fast progressing G93A-SOD1 mouse, and which were also tested in vitro against TNFα-stimulated EOC-20 microglia
CompoundIC50 (NO2)
EOC-20 cells
IC50% (PGE2)
EOC-20 cells
Life extension in G93A-SOD1
mouse (days)
Riluzole100 µm9 µm13a
Nimesulide23 µm17 nm5b
Minocycline49 µm84 µm11c
NDGA8 ± 3 µm841 nm13d
Co-enzyme Q10InactiveInactive7e
Creatineinactiveinactive13–26f
α-TocopherolInactiveInactiveNonea

Acknowledgements

This work was supported in large part by the ALS Association; the Oklahoma Center for the Advancement of Science and Technology (HR02–149RS); and the National Institutes of Health (AG20783, NS44154). Tetra-O-methyl NDGA was a generous gift from Dr George Lee of Biocure International, Edina, MN, USA.

Ancillary