Noninvasive assessment of the injured human spinal cord by means of functional magnetic resonance imaging.

STUDY DESIGN: A magnetic resonance imaging technique that enables indirect detection of neuronal activity has been developed for the spinal cord. In the present study, this method, spinal functional magnetic resonance imaging (fMRI), is applied to the first study of the injured spinal cord, with the goal of better clinical assessment of the entire cord. OBJECTIVES: The objectives of this project are: (1) to investigate the neuronal activity that can be detected in the spinal cord caudal to a chronic injury by means of spinal fMRI, and (2) to develop spinal fMRI as a clinical diagnostic tool. SETTING: Institute for Biodiagnostics, National Research Council of Canada, Winnipeg, Manitoba, Canada. METHODS: fMRI of the spinal cord was carried out in 27 volunteers with cervical or thoracic spinal cord injuries (SCIs). Of these volunteers, 18 had complete injuries, and nine had incomplete injuries. Spinal fMRI was carried out in a 1.5 T clinical MR system, using established methods. Thermal stimulation at 10 degrees C was applied to the fourth lumbar dermatome on each leg, and images were obtained of the entire lumbar spinal cord. RESULTS: Areas of neuronal activity were consistently observed in the lumbar spinal cord in response to the thermal stimulation, even when the subjects had no awareness of the sensation. The pattern of activity was notably different compared with noninjured subjects. In general, subjects with complete SCI showed absent or diminished dorsal gray matter activity, but had enhanced ventral activity, particularly contralateral to the stimulation. CONCLUSIONS: Spinal fMRI is able to provide a noninvasive assessment of the injured spinal cord that does not depend on the patient's perception of the stimulus being applied. This work was carried out on a standard clinical MRI system without modification, and so is readily applicable in most MR units. SPONSORSHIP: This work was funded by a grant from the Canadian Institutes of Health Research (CIHR).

Functional magnetic resonance imaging at 0.2 Tesla.

Functional magnetic resonance imaging of healthy human volunteers was carried out at 0.2 T, using proton-density weighted (TE = 24 ms) spin-echo imaging, in order to eliminate any contribution from the blood oxygenation-level dependent (BOLD) effect. The purpose of the study was to verify the existence of a proton-density change contribution to spin-echo functional magnetic resonance imaging (fMRI) data. Results demonstrated signal intensity changes in motor and sensory areas of the brain during performance of a motor task and cold sensory stimulation of the hand, with signal changes ranging from 1.7 to 2.3%. These values are consistent with 1.9% signal changes observed previously under similar conditions at 3 T. These findings confirm the proton-density change contribution to spin-echo fMRI data and support the theory of signal enhancement by extravascular water protons (SEEP) as a non-BOLD fMRI contrast mechanism. This study also demonstrates that fMRI based on the SEEP contrast mechanism can be carried out at low fields where the BOLD effect is expected to be negligible.

Functional magnetic resonance imaging of the human brain based on signal enhancement by extravascular protons (SEEP fMRI).

Functional magnetic resonance imaging (fMRI) studies of the human brain were carried out at 3 Tesla to investigate an fMRI contrast mechanism that does not arise from the blood oxygen-level dependent (BOLD) effect. This contrast mechanism, signal enhancement by extravascular protons (SEEP), involves only proton-density changes and was recently demonstrated to contribute to fMRI signal changes in the spinal cord. In the present study it is hypothesized that SEEP fMRI can be used to identify areas of neuronal activity in the brain with as much sensitivity and precision as can be achieved with BOLD fMRI. A detailed analysis of the areas of activity, signal intensity time courses, and the contrast-to-noise ratio (CNR), is also presented and compared with the BOLD fMRI results. Experiments were carried out with subjects performing a simple finger-touching task, or observing an alternating checkerboard pattern. Data were acquired using a conventional BOLD fMRI method (gradient-echo (GE) EPI, TE = 30 ms), a conventional method with reduced BOLD sensitivity (GE-EPI, TE = 12 ms), and SEEP fMRI (spin-echo (SE) EPI, TE = 22 ms). The results of this study demonstrate that SEEP fMRI may provide better spatial localization of areas of neuronal activity, and a higher CNR than conventional BOLD fMRI, and has the added benefit of lower sensitivity to field inhomogeneities.

Mapping of neuronal function in the healthy and injured human spinal cord with spinal fMRI.

Functional magnetic resonance imaging of the human spinal cord is carried out with a graded thermal stimulus in order to establish the relationship between signal changes and neural activity. Studies of the lumbar spinal cord in 15 healthy subjects with 10 degrees C stimulation of the skin overlying the calf demonstrate a pattern of activity that matches the neuronal anatomy of the spinal cord. This pattern shows primarily dorsal horn activity, with expected components of motor reflex activity as well. Moreover, a later response shifting to noxious cold over time is also demonstrated with a shift to more dorsal horn activity. Signal intensity changes detected at different degrees of thermal stimulation have a biphasic nature, with much larger signal changes below 15 degrees C as the stimulus becomes noxious, and agree well with electrophysiological results reported in the literature. These findings demonstrate a strong correspondence between Spinal fMRI results and neural activity in the human spinal cord. Spinal fMRI is also applied to studies of the injured spinal cord, below the site of injury. Results consistently demonstrate activity in the spinal cord even when the subjects cannot feel the stimulus being applied. Signal intensity changes demonstrate the same stimulus-response pattern as that in noninjured subjects, but the areas of activity in the spinal gray matter are notably altered. In subjects with complete injuries, activity is absent ipsilateral to the thermal stimulation, but appears to be enhanced on the contralateral side. These findings demonstrate the reliability of Spinal fMRI and its clinical potential.

Extravascular proton-density changes as a non-BOLD component of contrast in fMRI of the human spinal cord.

The fractional signal intensity change (Delta S/S) observed during activation in T(2)-weighted fMRI of the spinal cord has previously been shown to depend linearly on the echo time (TE) but to have a positive value of roughly 2.5% extrapolated to zero TE. In this study we investigated the origin of this finding by measuring the Delta S/S in spinal fMRI with very short TEs. Our results demonstrate that the Delta S/S does not approach zero, but has a value as high as 3.3% at TE = 11 ms. At TEs > 33 ms we observed the linear relationship between Delta S/S and TE as in previous studies. These data demonstrate that there is a non-BOLD contribution to signal changes observed in spinal fMRI. We hypothesize that this contribution is a local proton density increase due to increased water exudation from capillaries with increased blood flow during neuronal activation, and term this effect "signal enhancement by extravascular protons" (SEEP).

Functional magnetic resonance imaging of the human cervical spinal cord with stimulation of different sensory dermatomes.

Functional MR imaging (fMRI) of the cervical spinal cord was carried out in 13 healthy volunteers. A cold stimulus was applied, at different times, to three different sensory dermatome regions overlying the right hand and forearm: the thumb side of the palm, the little finger side of the palm, and the forearm below the elbow. Stimulation of these areas is expected to involve the 6(th), 8(th), and 5(th) cervical spinal cord segments respectively. Whereas true activations are expected to correspond to the region being stimulated, false activations such as arising from noise and motion, are not. The results demonstrate that clustering of active pixels into groups based on their intensity time courses discriminates false activations from true activations. Following clustering, the distribution of activity observed with fMRI matched the expected regions of neuronal activation with the different areas of stimulation on the hand and forearm.

Capsaicin as a source for painful stimulation in functional MRI.

Functional magnetic resonance imaging (fMRI) was used to examine the brain processing of capsaicin-induced painful stimulation in the alpha-chloralose anesthetized rat. Experiments were performed on a 9.4-T magnet (Magnex, UK) with Avance console (Bruker, Germany) using a surface coil tuned to 400.5 MHz centred over the rat forebrain. Gradient-echo images of two slices, with an echo time of 25 msec, repetition time of 70 msec, and 50 repetitions, were acquired per experiment. These images were analyzed using a fuzzy cluster analysis technique (EvIdent). Activation of areas of the brain known to be associated with the processing of pain, namely the anterior cingulate (bilateral), frontal cortex (bilateral), and sensory motor cortex (contralateral), was found in all animals (N = 6) following injection of 25 microl of capsaicin (128 microg/mL in 7.5% dimethylsulfoxide [DMSO]) into the dorsal forepaw. It is possible to reproduce the pain response in a given animal several times throughout the course of an experiment, provided that sufficient time is allowed between capsaicin injections. This acute phase of capsaicin-induced pain involving stimulation of C polymodal nociceptors was examined by functional imaging. There was a substantial initial increase in activation in regions of the brain associated with pain and there was a trend towards increasing activation with repeated stimulations. Treatment with morphine (3 mg/kg, intravenously) was found to substantially reduce, if not completely eliminate, the areas of functional activation associated with physiologic pain (anterior cingulate and frontal cortex) after C-nociceptor stimulation with capsaicin (N = 6). FMRI involving capsaicin-induced painful stimulation could prove to be an effective tool for the study of novel analgesics and the central nervous system processing of pain.

Spin-echo versus gradient-echo fMRI with short echo times.

Blood-oxygen level dependent signal changes in the visual cortex were investigated as a function of echo time with spin-echo and gradient-echo EPI at 1.5 T and 3 T. The linear relationship between the fractional signal change and the echo time was apparent in all cases. Relaxation rate changes determined from the slope of this linear relation agree with published values, intercept values extrapolated to an echo time of zero, however, were 0.66% to 1.0% with spin-echo EPI, and 0.11% to 0.35% with gradient-echo EPI. Spin-echo and gradient-echo EPI can therefore yield similar signal changes at sufficiently short echo times.

Characterization of contrast changes in functional MRI of the human spinal cord at 1.5 T.

Contrast changes observed in functional magnetic resonance imaging in the human spinal cord were investigated with both motor and sensory tasks over a range of echo times. Data were acquired using a single-shot fast spin-echo sequence at 1.5 Tesla. Data were analyzed with two different correlation thresholds and the effects of altering the order of repeated experiments was also investigated. Plots of the fractional signal change as a function of echo time yielded linear functions with slopes corresponding to relaxation rate changes of -0.30 sec(-1) with sensory stimulation and approximately -0.50 sec(-1) with a motor task. However, the fractional signal change extrapolated to an echo time of zero was significantly greater than zero in each case and was roughly 2.5%. This suggests that in addition to the BOLD effect there is a baseline signal change which occurs concomitant to neuronal activation in the spinal cord.

Correlation of cerebral hypoxic-ischemic T2 changes with tissue alterations in water content and protein extravasation.

BACKGROUND AND PURPOSE: Age-dependent changes in T2-weighted MR images have been reported in cerebral hypoxia-ischemia. However, the biophysical mechanisms responsible for the image changes remain poorly defined. We investigated whether cerebral hypoxia-ischemia-induced T2 changes correlate with alterations in either water content or protein extravasation. METHODS: One- and 4-week-old rats were subjected to unilateral carotid artery occlusion plus hypoxia in 8% oxygen. T2 images were acquired before, during, and 1 or 24 hours after hypoxia-ischemia. Blood-brain barrier disruption and brain edema were evaluated by immunohistological detection of IgG extravasation and measurement of water content by dry-wet weight and specific gravity methods. RESULTS: In 1-week-old rats, T2 values, areas of hyperintensity on T2-weighted images, and water content in the ipsilateral hemisphere increased during hypoxia-ischemia, recovered at 1 hour after hypoxia-ischemia, and increased again at 24 hours after hypoxia-ischemia. Extravasation of IgG occurred during hypoxia-ischemia and remained detectable 24 hours after hypoxia-ischemia. In 4-week-old rats, an increase in T2 or extravasation of IgG did not occur until 24 hours after hypoxia-ischemia despite a comparable elevation in water content during and soon after hypoxia-ischemia. CONCLUSIONS: T2 imaging appears reliable for detecting edema associated with disruption of the blood-brain barrier but not necessarily an increase in cerebral water or plasma proteins alone. The different hypoxic-ischemic changes in T2 in immature and older brain are associated with differences in alterations in water content plus extravasation of protein, consistent with age-dependent differences in hypoxic-ischemic alterations in vascular permeability.

Effect of long-term vigabatrin administration on the immature rat brain.

PURPOSE: To determine whether the neuropathologic changes produced by vigabatrin (VGB; gamma-vinyl GABA) administration in the developing rat brain are reversible. METHODS: We injected rats daily with VGB (25-40 mg/kg/day, s.c.) from age 12 days for 2 weeks followed by 2 weeks of a drug-free period. Behavioral testing, magnetic resonance (MR) imaging, biochemical assays, and histologic technique were used to assess the adverse effect of VGB in developing brain and its reversibility. RESULTS: At the end of 2 weeks' VGB administration: (a) there was a hyperactivity and a shortened latency to escape out of cool water; (b) white matter appeared hyperintense in T2 and diffusion-weighted MR images with 4-15% increases in T2; (c) microvacuolation, TUNEL-positive nuclei, and swollen axons were observed in the corpus callosum; (d) myelin staining indicated a reduction in myelination, as did the reduction in activities of myelin and oligodendrocyte-associated enzymes and the decrease in myelin basic protein on Western blots. Two weeks after stopping VGB administration: (a) MR images were normal, and microvacuolation was no longer in the white matter; (b) reduction in myelination reversed partially; (c) the T2 relaxation time remained elevated in the hypothalamus; and (d) the behavioral response remained abnormal. CONCLUSIONS: Long-term VGB administration to young rats causes brain injury, which recovers partially on its cessation. The observed cell death, disrupted myelination, and alterations in behavior indicate a need for further safety assessment in infants and children.

Mechanisms of beneficial effects of probucol in adriamycin cardiomyopathy.

Probucol, a lipid-lowering drug, has been shown to offer protection against adriamycin-induced cardiomyopathy. In order to define the mechanism of this protection, we examined changes in antioxidants and lipid peroxidation in hearts as well as lipids in hearts and plasma from rats treated with either adriamycin or adriamycin and probucol with appropriate controls. Any potential free radical quenching as well as growth inhibitory effects of probucol were also examined using Chinese hamster ovary (CHO) cells in culture. In animal model, adriamycin caused a significant depression in glutathione peroxidase and increased plasma and cardiac lipids as well as lipid peroxidation. Probucol treatment modulated adriamycin-induced cardiomyopathic changes and increased glutathione peroxidase and superoxide dismutase activities. In the presence of adriamycin under hypoxic conditions, formation of adriamycin semiquinone radical was detected by ESR. The cell growth in these cultures was also inhibited by adriamycin in a dose-dependent manner. Probucol had no effect on adriamycin-induced growth inhibition as well as formation of semiquinone radicals. It is proposed that probucol protection against adriamycin cardiomyopathy is mediated by increased antioxidants and lipid-lowering without any effect on free radical production.

Prevention of both T2- and diffusion-weighted increases in image intensity during cerebral hypoxia-ischemia in infant rats pretreated with dexamethasone.

The present study examines the effect of dexamethasone treatment on the intensity of changes in T2-weighted and diffusion-weighted (DW) magnetic resonance images occurring in infant rats during and after cerebral hypoxia-ischemia. The right carotid artery was occluded under isoflurane anesthesia in 7-day-old rats and images were acquired in sedated animals using a Bruker 9.4 T magnetic resonance (MR) system. Imaging changes were markedly different in rats pretreated with dexamethasone phosphate (0.1 mg/kg, i.p.) 24 h before hypoxia than in controls. In control animals, areas of hyperintensity ipsilateral to the occlusion occurred during hypoxia-ischemia in both the DW- and T2-weighted images with some recovery of the changes in early posthypoxia. In contrast, in dexamethasone-treated animals, areas of increased hyperintensity in the MR images did not occur. Thus, dexamethasone treatment prevents MR imaging changes during ischemia, suggesting that the cytotoxic edema associated with energy depletion and/or ionic disturbances during ischemia are also prevented by dexamethasone treatment.

Metabolite changes in neonatal rat brain during and after cerebral hypoxia-ischemia: a magnetic resonance spectroscopic imaging study.

Cerebral metabolite concentrations were measured in infant rats using proton magnetic resonance spectroscopic imaging. Measurements were made prior to, during and after exposure of rats (6- and 7-day-old) to unilateral cerebral hypoxia-ischemia (right carotid artery occlusion +2h 8% oxygen). Data clustered according to age and outcome-6-day-old animals with no infarct and 7-day-old animals with infarct. In 6-day-old animals, cerebral lactate concentration increased during hypoxia-ischemia, particularly ipsilateral to the occlusion, and returned to normal soon after the end of hypoxia. There were no major changes in N-acetyl-aspartate levels (NAA) in this group and no regions of hyperintensity on T2 or DW weighted images at 24 h. In the 7-day-old animals, lactate increased during hypoxia-ischemia and remained elevated in the first hour after reperfusion. Furthermore, lactate remained at 258+/-117% and 233+/-56% of pre-hypoxic levels, 24 and 48 h post-hypoxia, respectively. NAA concentrations ipsilateral to the occlusion decreased to 55+/-14% during hypoxia, recovered early post-hypoxia and again decreased to 61+/-25% and 41+/-28% at 24 and 48 h post-hypoxia-ischemia, respectively. The infarct volumes measured by diffusion weighted and T2 weighted MRI at 48 h post-hypoxia were 152+/-40 mm3 and 172+/-35 mm3, respectively. Thus, irreversible damage correlated well with measured in vivo lactate and NAA changes. Those animals in which NAA was unaltered and lactate recovered soon after hypoxia did not show long-term damage (6-day-old animals), whereas those animals in which NAA decreased and lactate remained elevated went on to infarction (7-day-old animals).

Magnetic resonance imaging during cerebral hypoxia-ischemia: T2 increases in 2-week-old but not 4-week-old rats.

We investigated whether the changes detectable with magnetic resonance imaging techniques during and after an episode of cerebral hypoxia-ischemia differ in immature and older brain. Diffusion weighted (DW) and T2-weighted (T2W) images were repeatedly acquired before, during, and after an episode of cerebral hypoxia-ischemia (unilateral carotid artery occlusion plus hypoxia) in 2- and 4-wk-old rats lightly anesthetized with isoflurane. Areas of increased brightness were detected in DW images from both 2- and 4-wk-old rats by 10-20 min after the start of hypoxia. These hyperintense areas increased during hypoxia, comprising 60.8+/-4.9% and 30.5+/-2.7% of the brain image at the level of the thalamus in 2-wk-old and 4-wk-old animals, respectively (p < 0.003). Hyperintense areas (e.g. 27.0+/-8.3%) also appeared in T2W images during hypoxia-ischemia in 2-wk-old animals, but these did not occur in 4-wk-old animals (p < 0.02). This observation was reflected in T2, which increased during hypoxia-ischemia in the 2-wk-old but not the 4-wk-old group. By 60 min after the termination of hypoxia-ischemia in either age group, areas of hyperintensity resolved and then reappeared 24 h later on both DW and T2W images. Thus, irrespective of age, magnetic resonance imaging changes during transient hypoxia-ischemia generally recover with a delayed or secondary increase in DW and T2W hyperintensity hours later. Immature brain differs from older brain primarily with respect to some combination of hypoxic/ischemic cellular or biochemical changes, that are detectable as increases in T2 within 2-wk-old but not 4-wk-old animals.

A review of in vivo 1H magnetic resonance spectroscopy of cerebral ischemia in rats.

A number of metabolic alterations are initiated by cerebral ischemia including dramatic increases in lactate concentration, decreases in N-acetylaspartate, choline, and creatine concentrations, as well as changes in amino acid levels. A review of proton nuclear magnetic resonance spectroscopy studies of focal and global cerebral ischemia in rats is presented here. In particular, studies in neonatal rats have shown that a continued elevation of lactate levels without recovery after hypoxia-ischemia or a decrease in N-acetylaspartate concentration at any time are indicative of deleterious outcome. Studies of the effect of temperature on ischemic damage in a model of focal ischemia showed that outcome improved with mild hypothermia. Again, lack of recovery of lactate upon reperfusion was shown to be indicative of poor outcome. Dichloroacetic acid was used to treat rats with focal ischemic damage. Animals subjected to transient ischemia that were treated with dichloroacetic acid showed significant decreases in lactate concentration.

Semiquinone free radical formation by daunorubicin aglycone incorporated into the cellular membranes of intact Chinese hamster ovary cells.

The production of semiquinone free radicals has been measured by electron paramagnetic resonance spectroscopy (EPR) in Chinese hamster ovary cells in which 7-hydroxy daunorubicin aglycone had been incorporated. The highly lipophilic daunorubicin aglycone was incorporated into the cellular membrane by swirling a cell suspension over a thin layer of daunorubicin aglycone. Thus, the observed semiquinone free radical was likely formed directly in the lipophilic environment of the cellular membrane. The linewidth of the observed EPR signal suggested that a neutral protonated semiquinone species was formed. In the presence of the cell-impermeant paramagnetic line broadening agent chromium(III) oxalate, no detectable signal was observed. This result indicates that even though the semiquinone is embedded in the membrane, it is still partly accessible to the external chromium(III) oxalate. Analysis of chloroform extracts of the cells after EPR experiments indicated that daunorubicin aglycone was extensively metabolized. The results of a growth inhibition assay carried out on cells into which daunorubicin aglycone had been incorporated showed almost no effect on cell growth. This result indicates that in spite of significant daunorubicin aglycone-induced radical formation taking place directly in the cell membrane, little cell damage results.

Inhibition of anthracycline semiquinone formation by ICRF-187 (dexrazoxane) in cells.

The formation of semiquinone free radicals of doxorubicin, epirubicin, daunorubicin, and idarubicin was measured by electron paramagnetic resonance (EPR) spectroscopy in hypoxic suspensions of chinese hamster ovary (CHO) cells. The amount of semiquinone produced was in the order idarubicin >> doxorubicin > daunorubicin > epirubicin. The idarubicin semiquinone signal was both the fastest to be formed and to decay. Idarubicin, which was the most lipophilic of the anthracyclines studied, also displayed the fastest fluorescence-measured cellular uptake of drug. Thus, it was concluded that semiquinone formation was dependent upon the rate of cellular uptake. Lysed cell suspensions were also shown to be capable of producing the doxorubicin semiquinone in the presence of added NADPH. The cardioprotective agent dexrazoxane (ICRF-187) was observed to decrease the amount of doxorubicin semiquinone observed in cell suspensions. Dexrazoxane also decreased the amount of doxorubicin semiquinone observed in the NADPH-lysed cell suspension mixture. Neither bipyridine nor deferoxamine decreased NADPH-dependent doxorubicin semiquinone formation. These results suggest that dexrazoxane does not decrease doxorubicin semiquinone formation through an iron complex formed from hydrolyzed dexrazoxane. Dexrazoxane may be inhibiting an NADPH-dependent enzyme.

Hydroxyl radical production by the iron complex of the hydrolysis product of the antioxidant cardioprotective agent ICRF-187 (dexrazoxane).

Dexrazoxane (ICRF-187) is now in clinical use for the prevention of doxorubicin-induced cardiotoxicity. This cardiotoxicity is thought to be due to iron-mediated oxidative stress. Dexrazoxane may be acting through its strongly metal ion binding rings-opened hydrolysis product ADR-925 by complexing iron. Since iron-chelates are known to be able to produce hydroxyl radicals, an electron paramagnetic resonance spin trapping study was undertaken to compare the hydroxyl radical-producing ability of the ferrous-ADR-925 complex with that of the ferrous complexes of ethylenediaminetetraacetic acid (EDTA) and the tetraacid analog of ADR-925 (DAPTA). In spectrophotometric studies it was shown that the ferrous-ADR-925 complex underwent aerobic oxidation 87 and 44 times slower than the ferrous complexes of EDTA or 1,2-diaminopropane-N,N,N',N'-tetraacetic acid (DAPTA), respectively. In spite of the much slower oxidation of the ferrous-ADR-925 complex, it was, nonetheless, equally effective in producing hydrogen peroxide-dependent spin adducts. These spin adducts were produced from the reaction of the spin trap with free hydroxyl radical (HO(.)), and with a transient iron oxidant with HO(.)-like reactivity. Thus, it is concluded that ADR-925 acts by either complexing free iron or iron bound to doxorubicin, and forming a soluble iron complex that is less effective at producing site-specific oxygen radical damage.

Production of hydroxyl radical by iron(III)-anthraquinone complexes through self-reduction and through reductive activation by the xanthine oxidase/hypoxanthine system.

The iron(III) complexes of doxorubicin and epirubicin were observed to undergo a self-reduction (autoxidation) reaction in the absence of added reductants under aerobic conditions that resulted in the formation of ferrous anthracycline complexes. These self-reduction reactions resulted in significant hydrogen peroxide-mediated hydroxyl radical formation, as determined by electron paramagnetic resonance spin trapping. In contrast, the iron(III) complexes of daunorubicin, idarubicin, and mitoxantrone produced no significant amount of hydroxyl radical formation. Only the anthraquinones with an alpha-ketol side chain were observed to undergo both self-reduction and hydroxyl radical formation. Thus, the alpha-ketol side chain must be undergoing concomitant oxidation. The rate of self-reduction of the iron(III)-doxorubicin complex is consistent with a mechanism in which unbound doxorubicin binds to an iron(III)-doxorubicin complex of decreased coordination and after binding undergoes an intramolecular electron transfer. Molecular modeling was used to identify iron(III)-doxorubicin complexes that could result in electron transfer from the doxorubicin side chain hydroxyl group to the iron(III). All of the iron(III)-anthracycline complexes were able to produce hydroxyl radicals at significantly increased rates in the presence of the xanthine oxidase/hypoxanthine superoxide-generating system. In this system the iron(III)-epirubicin complex gave the greatest rate of hydroxyl radical production, with iron(III)-idarubicin giving the least.