Carboxyfullerenes as neuroprotective agents

Malonic acid derivatives of C₆₀ were synthesized as described by Lamparth and Hirsch (18). Briefly, diethyl bromomalonate was added to a solution of C₆₀ in toluene, followed by the addition of 1,8-diazobicyclo[5,4,0]undec-7-ene, which resulted in a color change from violet to dark red. After stirring for 4 days, the solvent was removed in vacuo, and the blackish residue was chromatographed on silica gel (270–230 mesh) using toluene-hexane (1:1 by volume) as eluent. The unreacted C₆₀ was obtained first, followed by a brown band that corresponds to the diester. The eluent was changed to toluene-hexane (4:1). A brown band (tetraester) was obtained after a narrow yellow band (tetraester, para addition) followed by a deep red band (tetraester). The eluent was changed again to toluene-hexane (9:1), and a red band was collected that corresponds to the D₃ isomer as the major component. Toluene (100%) then was used to elute three bands. The third band corresponded to semipure C₃ as the major component. The fractions containing semipure C₃ or D₃ were rechromatographed on silica gel (230–400 mesh) using toluene as the mobile phase to give the purified C₃ or D₃ isomers. To a solution of either C₃ or D₃ (100 mg in 100 ml toluene, ≈ 0.1 mmol) was added NaH (80%, 60 mg, 2 mmol), and the mixture was refluxed for 1 hr. After the heating source was removed, MeOH (5 ml) was added immediately to quench the reaction. Red powder soon precipitated and was collected by centrifugation. The powder was washed with toluene twice and hexane four times. The red solid was dissolved in water to which HCl (4 M) was added. A red amorphous precipitate was formed immediately, which was collected by centrifugation. The solid again was washed with HCl (4 M), and then by water twice. The solid was dissolved in MeOH, and the solvent was removed in vacuo to give the powdery pure isomer acid (red for C₃ and brownish red for D₃).

The reactivity of carboxyfullerenes with oxygen radicals was assessed by EPR spectroscopy. Aqueous samples were loaded into a quartz flat cell (60 × 10 × 0.25 mm) and analyzed on a Bruker 200 X-band spectrometer, with experimental conditions as described in the legend for Fig. 2. Signal averaging was performed on some samples.

A second set of experiments was designed to evaluate the degree of membrane interaction of the two carboxyfullerene isomers. Mouse brain lipids were extracted (19) and aliquoted into test tubes. Spin-labeled lipids, 5-doxyl or 16-doxyl ketostearic acids, were added at a ratio of 1:50, and dried under N₂. Tris saline (25 mM, pH 7.4), with or without C₃ or D₃, was added to each tube and vortexed. C₃ and D₃ were tested at concentrations of 10–300 nmol (1:30–1:1 compared with lipids). Settings for the above EPR experiments were power = 1.6 mW, modulation = 1 G, field modulation = 100 Hz, receiver gain = 3.2 × 10⁵. Low-temperature (77°K) EPR also was performed on H₂O₂-oxidized carboxyfullerenes to determine whether oxidation produces a paramagnetic (radical) species.

Cell culture reagents and supplies were obtained from standard sources. Laboratory reagents were purchased from Sigma. Media stock consisted of minimal essential media without glutamine, containing 20 mM glucose.

Mouse neocortical cultures were prepared as neuronal-astrocyte cocultures or as pure neuronal cultures (<1% astrocytes) from Swiss–Webster mice as described previously (20–21).

Stock solutions of purified C₃ or D₃ carboxyfullerenes (25 mM or 50 mM) were freshly prepared in sterile water. Brief exposure to N-methyl-d-aspartate (NMDA) was carried out as described (16). The culture media was exchanged twice with Hepes, bicarbonate-buffered balanced salt solution (9), and then NMDA (200 μM) was added alone or with each of the carboxyfullerenes (30 μM-1 mM) for 10 min. Exposure was terminated by exchanging the medium four times with media stock. The cells were returned to the 37°C (5% CO₂) incubator for 24 hr, when injury was assessed. To determine whether carboxyfullerenes could block NMDA receptor-mediated calcium entry, cells were exposed to NMDA for 10 min in the presence of tracer ⁴⁵Ca²⁺ (0.5 μCi/well; New England Nuclear), and the amount of intracellular ⁴⁵Ca²⁺ determined as previously described (22).

Cultures were exposed to α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) (8 μM), with or without carboxyfullerenes, in media stock containing MK-801 (10 μM MK-801; Merck, Sharp and Dohm) to block activation of NMDA receptors by released endogenous glutamate. Combined oxygen, glucose deprivation was performed as described (23). Cultures were placed in an anoxic chamber (O₂ < 2 mm Hg), and the media was exchanged three times with a balanced salt solution (BSS₀) lacking glucose and oxygen. After 60 min, the cells were washed back into oxygen, glucose-containing medium and returned to the aerobic culture incubator for 24 hr.

To induce apoptosis, pure neuronal cultures containing less than 1% astrocytes were deprived of serum on day in vitro 7 by exchanging the medium with serum-free media stock (20). Washed controls were returned to medium with serum. Exposure to Aβ_1–42 was performed on day in vitro 10 mixed cultures by application of 20 μM Aβ_1–42 (K-Biologicals, Rancho Cucamonga, CA) for 48 hr.

Cell death was assessed by phase contrast microscopy, assay of lactate dehydrogenase in the bathing medium, and manual counting of dead cells stained with propidium iodide (10 μM for 7 min).

Cultures were loaded with 15 μM dihydrorhodamine 123 (Molecular Probes) at the onset of serum deprivation. Cells were imaged 0.5, 4, and 8 hr later on a laser scanning confocal microscope (Noran), using Exλ = 488 nm, Emλ > 515 nm as previously described (9, 24).

Mice from the G93A SOD1 G₁ strain developed by Gurney et al. (25) as a model for familial amyotrophic lateral sclerosis (FALS) were treated with carboxyfullerene beginning at 73 ± 2 days of age. Carboxyfullerene containing a mixture of isomers (>85% C₃, 10% D₃, with trace amounts of other Tris malonic acid C₆₀ adducts) was dissolved in sterile 0.9% saline, and loaded into mini-osmotic pumps (no. 2004, Alza, 28-day delivery) as per manufacturer’s instructions. Pumps were soaked in sterile normal saline for 40 hr and implanted into the peritoneum through a small midline incision. Before implantation of the second pump 4 weeks later, the depleted first pump was removed, and the volume remaining transferred to a tube and measured to verify that drug delivery had occurred at the specified rate of 6.6 μl/day, to deliver 15 mg/kg per day. Mice were maintained in the colony until moribund, determined by inability of the mouse to right itself in 30 sec after being placed on its side. This correlated well with inability to drink spontaneously or to respond to offered food or water. Determination of whether an animal was moribund was determined by a blinded observer, and the date of death recorded.

Motor performance was assessed using a scale developed by Basso et al. (26) to evaluate function after spinal cord injury. Mice were videotaped weekly starting at 89 days of age, and their gait scored by a blinded observer using this 21-point scale, which rates spontaneous ambulation, including leg position and coordination. Mice were evaluated more informally daily.

Mass spectroscopy of the two purified trisadduct regioisomer esters demonstrated a peak in each sample at 1,194/1,195 mass units, as previously described (18). Identity and purity of both the intermediate ester and the final malonic acid product of each regioisomer were confirmed by ¹³C NMR spectroscopy and mass spectrometry (18). The final malonic acids, C₃ and D₃, were soluble in water to at least 75 mM. Control experiments verified that the compounds did not interfere with the colorimetric lactate dehydrogenase assay, and solutions to at least 25 mM failed to alter the pH of the experimental solutions.

Both C₃ and D₃ isomers were unusually potent scavengers of hydroxyl radical (^⋅OH) and superoxide anion (O_2⨪) in solution. They were able to completely eliminate ^⋅OH at concentrations as low as 4–5 μM (Fig. 2), 10- to 100-fold less than required for most other free radical scavengers. Both isomers also effectively scavenged superoxide radical (O_2⨪), although 10-fold higher concentrations were required to do this (Fig. 2).

Both the C₃ and D₃ isomers produced a dose-dependent decrease in death of cortical neurons exposed to NMDA or AMPA (Fig. 3 A and B), although the C₃ compound was both more potent and more effective than D₃ (Fig. 3 A and B). The C₃ compound provided essentially complete neuroprotection against NMDA receptor-mediated neuronal death. Although the degree of protection afforded by C₃ against AMPA-induced neurotoxicity was generally somewhat less, coapplication of C₃ still reduced neuronal death by over 80%. The concentration of carboxyfullerene required to produce maximal protection against these excitotoxic insults varied somewhat with the degree of injury; i.e., insults that resulted in 100% neuronal death required slightly higher concentrations of carboxyfullerene for maximal neuroprotection. The C₃ isomer also reduced neuronal death after combined oxygen-glucose deprivation for 45–60 min (Fig. 4), an injury mediated in large part by NMDA receptors (23); the D₃ compound was not tested. To exclude the possibility that C₃ could antagonize NMDA receptors, the cellular uptake of ⁴⁵Ca²⁺ from the bathing medium induced by NMDA was assessed; coapplication of 30–300 μM C₃ did not affect Ca²⁺ uptake (data not shown).

We hypothesized that the difference in biological activity between the isomers might reflect a difference in dipole moment and resultant ability to intercalate into lipid bilayers. To test the hypothesis that C₃ can enter cell membranes better than D₃, we used EPR spectroscopy of nitroxide spin-labels incorporated into a mixture of mouse brain lipids (Fig. 5). Three parameters of carboxyfullerene-lipid interaction were determined from the EPR measurements: the order parameter (S), correlation time (t_c), and lipid/aqueous partition factor. C₃ decreased the order parameter (measured with the spin-labeled 16-doxyl lipid) more than D₃, suggesting that C₃ interacts with the interior of the bilayer to a greater degree than D₃. This is supported by the observed changes in t_c and the partition factor for C₃, which indicate substantial entry of C₃ into the bilayer. On the other hand, D₃ caused opposite changes in t_c and the partition factor, which imply interaction by D₃ with the headgroup region of the bilayer, but little entry into the membrane. The results, taken together, are consistent with greater intercalation of C₃ compared with D₃ into brain lipid membranes (Table 1).

To examine whether the carboxyfullerenes also could inhibit neuronal apoptosis, we turned to two paradigms: (i) serum deprivation, and (ii) exposure to the Alzheimer disease amyloid peptide, Aβ_1–42. Cortical neurons cultured without glia underwent apoptosis 24–48 hr after removal of serum (Fig. 6 A and B Top). As with sympathetic neurons deprived of nerve growth factor, this serum deprivation-induced neuronal apoptosis was associated with an enhancement of intracellular free radical production (Fig. 6B, Bottom), detectable by oxidation of dihydrorhodamine to fluorescent rhodamine 123. Application of C₃ at the onset of serum deprivation reduced subsequent free radical production and apoptosis (Fig. 6 A and B). In addition, application of C₃ also reduced apoptotic death of cortical neurons induced by 24–48 hr exposure to Aβ_1–42 (Fig. 7).

Mice that received C₃ carboxyfullerene via intraperitoneal mini-osmotic pumps for 2 months, beginning just after 2 months of age, showed delayed onset of symptoms, improved functional performance, and delayed death as compared with saline-treated controls. C₃-treated mice exhibited a delay in deterioration of about 10 days, and scored 4.6 ± 2.2 points (mean ± SEM, P = 0.048, by t test) better than controls on weekly testing. Death in the treated mice also was delayed by 9.0 ± 3.3 days (mean ± SEM, P = 0.023, by t test) (Fig. 8).