|Home | About | Journals | Submit | Contact Us | Français|
The drug development process for CNS indications is hampered by a paucity of preclinical tests that accurately predict drug efficacy in humans. Here, we show that a wide variety of CNS-active drugs induce characteristic alterations in visual stimulus–induced and/or spontaneous eye movements in mice. Active compounds included sedatives and antipsychotic, antidepressant, and antiseizure drugs as well as drugs of abuse, such as cocaine, morphine, and phencyclidine. The use of quantitative eye-movement analysis was demonstrated by comparing it with the commonly used rotarod test of motor coordination and by using eye movements to monitor pharmacokinetics, blood-brain barrier penetration, drug-receptor interactions, heavy metal toxicity, pharmacologic treatment in a model of schizophrenia, and degenerative CNS disease. We conclude that eye-movement analysis could complement existing animal tests to improve preclinical drug development.
The development of drugs for CNS indications by the pharmaceutical industry generally follows a path from biochemical or cell-based assays to testing in animal models to human clinical trials. Attrition is high at each step. In particular, only 8% of CNS drug candidates that enter clinical trials win FDA approval, among the lowest success rates of any therapeutic area (1). The majority of these drug candidates fail in late-stage clinical trials, with attrition most commonly caused by inefficacy or undesirable side effects. This low success rate is due in part to the limited predictive power of animal testing for human CNS pharmacology (2). Additionally, drug candidates with peripheral targets can be lost from the development pipeline if they produce unwanted CNS side effects (e.g., ref. 3). Thus, there is a clear need to develop improved physiological and behavioral tests in animals that accurately predict human responses to CNS-active drugs (4, 5).
In rodents, functional tests that assess CNS drug action can be roughly divided into 4 categories based on behavioral complexity: (a) simple stimulus-response paradigms, such as the startle reflex, prepulse inhibition, paw withdrawal from a hot plate, or vocalization after isolation; (b) stereotyped motor tasks, such as balancing on a rotarod; (c) complex innate behaviors, such as circadian entrainment or open-field activity; and (d) learned responses, such as maze running (6, 7). Tests in the first 2 categories can be performed rapidly with simple equipment and can measure a variety of changes in CNS function, such as sedation, analgesia, and ataxia. By contrast, tests in the third and fourth categories require more extensive monitoring or preliminary training trials, constraints that substantially increase the time and expense of animal testing (8–10). For tests with a large volitional component, trial-to-trial variability is generally large, and data sets must be averaged over many trials.
Eye movements represent a readily monitored behavior that reflects a complex set of sensory-motor computations (11). In nonfoveate mammals, such as mice, there are 3 primary types of eye movements: (a) the optokinetic reflex (OKR), (b) the vestibular-ocular reflex (VOR), and (c) spontaneous eye movements. The OKR is an involuntary response to a moving visual stimulus that consists of a series of slow eye-tracking movements (ETMs) interrupted at regular intervals by rapid resetting movements (saccades) in the opposite direction. Foveate mammals, such as humans and other primates, have additional types of eye movements, most prominently, fixational eye movements that bring the object of regard onto the fovea and maintain it there.
Eye movements are relatively unexplored in the context of preclinical CNS drug testing, although it has long been appreciated that some CNS-active drugs — including benzodiazepines, opiates, antipsychotics, and anticonvulsants — affect ETMs in humans (11–14). In the context of clinical pharmacology, one of the best-known uses of human eye movement analysis is for monitoring toxicity in patients taking phenytoin, an antiseizure medication with a relatively narrow therapeutic window (15, 16). In rodents, eye movements can be recorded from an animal that is head restrained using an infrared camera to monitor the position of the pupil relative to a stationary corneal reflection (17).
As a behavioral measure of CNS activity, eye movements have several favorable attributes: they are rapid, they do not require training, and they show little or no adaptation. An OKR data set, with hundreds of saccades, can be recorded in a single 30-minute session (17–19). Like the electrocardiogram or the electroencephalogram, an eye-movement record lends itself to relatively simple and semiautomated analyses of peak size, shape, polarity, and spacing.
In this paper, we explore the use of eye-movement analysis for monitoring CNS drug action in mice. We find that many CNS-active drugs elicit distinctive and characteristic eye-movement responses that can be analyzed quantitatively. We demonstrate the use of eye-movement analysis in this context by comparing it with the commonly used rotarod test of motor coordination and by using eye movements to monitor pharmacokinetics, blood-brain barrier (BBB) penetration, drug-receptor interactions, pharmacologic treatment in a model of schizophrenia, and degenerative CNS disease.
To rapidly assess the extent to which a variety of CNS-active drugs might affect eye movements in mice, we examined spontaneous and visual stimulus–induced eye movements after i.p. injection of 48 drugs (Figures (Figures11 and and22 and Supplemental Tables 1 and 2; supplemental material available online with this article; doi: 10.1172/JCI45557DS1). In this initial survey, relatively high doses were chosen to increase the likelihood that a drug effect might be observed. For this and all subsequent analyses we have used a custom-built eye-tracking apparatus in which a rotating pattern of vertical black and white stripes is projected onto the inner wall of a test cylinder from a ceiling-mounted rotating liquid crystal display projector (19). Pupil position was sampled at 60 Hz with an infrared video camera.
In Figure Figure11 and in many of the following figures, we show a representative 90-second OKR record that encompasses 3 contiguous 30-second segments. During the first and third segments, the wall of the cylindrical test chamber was illuminated with a uniform gray light (referred to as the null stimulus or rest period), and, during the second segment, the wall was illuminated with slowly rotating vertical black and white stripes, producing a strong stimulus for the OKR. For quantification of the number of ETMs per 30 seconds (referred to hereafter as ETM30), responses were typically averaged from 12 or more 30-second visual stimulus intervals per mouse.
At the doses used in Figures Figures11 and and2,2, 31 out of 48 drugs perturbed eye movements. Injection of PBS had no effect (Supplemental Figure 1). Among medicinal drugs (Figure (Figure1,1, A and C, and Figure Figure2),2), antipsychotics (chlorpromazine, clozapine, haloperidol, and perphenazine) and sedatives (chloral hydrate, phenobarbital, and the benzodiazepines, diazepam, clobazam, and zolpidem) consistently slowed or eliminated the OKR, as did the NMDA antagonist MK-801 and the antiseizure drugs gabapentin and phenytoin. The tricyclic antidepressant desipramine and the antimigraine drug methysergide reduced OKR frequency and amplitude. At 50 mg/kg, pentylenetetrazol (PTZ), a GABA(A) receptor antagonist and CNS stimulant, eliminates eye movements; at higher doses, PTZ induces seizures.
Interestingly, a number of drugs induced spontaneous eye movements, defined as movements occurring during the 30-second rest periods in Figure Figure1A.1A. This group includes amantadine and trihexyphenidyl, anticholinergic compounds used in the treatment of Parkinson disease; baclofen, a GABA-B receptor agonist used in the treatment of spasticity; and memantine, a low-affinity NMDA, 5HT3, and nicotinic acetylcholine receptor antagonist used in the treatment of Alzheimer disease. Relative to the spontaneous eye movements observed in control OKR traces, spontaneous eye movements induced by drug treatment exhibited a high frequency and a more rapid return to baseline.
For drugs of abuse and related compounds, we tested both the OKR and rotarod performance (Figure (Figure1,1, B and D–F; for quantification of rotarod performance, rotarod balancing times up to 60 seconds were typically averaged from 8 or more tests per mouse). The opiates methadone and morphine slowed the OKR and produced a similar decrement in rotarod performance. SNC-80, a selective delta opioid agonist, eliminated the OKR and substantially impaired rotarod performance. Codeine, a relatively low-potency morphine prodrug; naloxone, an opioid antagonist; and buprenorphine, a mixed opioid agonist/antagonist, had little or no effect on the OKR or on rotarod performance. Anandamide, the endogenous agonist for cannabinoid receptors CB1 and CB2, greatly slowed the OKR but had no effect on rotarod performance. By contrast, the synthetic CB2-selective agonist, GP 1a, has minimal effects on either eye movements or rotarod performance. Cocaine, phencyclidine (PCP), and ketamine, an NMDA receptor antagonist that is used as an anesthetic and a recreational drug, induced large numbers of spontaneous eye movements, but only ketamine and phencyclidine impaired rotarod performance. Interestingly, at doses with only a modest effect on the number of visual stimulus–induced ETM30s, ethanol induces a distinctive waxing and waning of OKR responsiveness to a continuous visual stimulus with a period of approximately 2–3 minutes (Figure (Figure1B).1B). Similarly, rotarod performance after ethanol administration exhibited large trial-to-trial variability, as reflected in the large standard deviation over multiple trials (Figure (Figure1F). 1F).
The survey in Figure Figure11 does not capture the full range of effects that some of the tested compounds have on spontaneous and/or visual stimulus–induced eye movements, as this experiment reveals only part of a complex time- and dose-dependent response. For example, in earlier experiments with ketamine, we observed time-dependent changes in the slope of the slow component of the OKR and the regularity of stimulus-induced eye movements as well as complete suppression of the OKR during a 5- to 10-minute period after i.p. drug administration (19). Additionally, ketamine induces spontaneous eye movements at low doses but abolishes eye movements at higher doses.
In this study, we have not explored the effects of chronic drug administration on eye movements, an important area for future research given the well-known delay in clinical response seen with many antidepressants and the delayed side effects that accompany chronic treatment with many antipsychotic drugs. We have also made no attempts to distinguish effects referable to peripheral targets — for example, direct actions on the cardiovascular system — from actions within the CNS. For most of the tested compounds, abundant evidence implicates CNS targets as their principal sites of action, although in several cases (e.g., cocaine) peripheral targets may also be relevant. Despite its limitations, this survey indicates that the action of a broad subset of psychoactive drugs can be monitored via their effects on eye movements. In the paragraphs that follow, we describe experiments based on acute drug administration to examine dosing, pharmacokinetics, tissue distribution, receptor specificity, and efficacy.
Fundamental to investigating drug action in the intact animal is quantification of the dose dependence and time course of the response. In general, the response will be a complex function of pharmacokinetics (drug absorption, distribution, metabolism, and excretion) and pharmacodynamics (the interaction between a drug and its receptor[s] and the dependence of the physiologic response on receptor occupancy). Continuous monitoring of drug action in awake, behaving animals is most readily conducted using simple physiologic measures such as heart rate and blood pressure. By contrast, continuous monitoring of more complex behaviors, such as feeding, drinking, or wheel running, is confounded by temporal variability on short time scales. As described below, the OKR is unusual in that it reflects the function of complex neural circuits, but it can be continuously monitored on time scales of tens of seconds to hours.
To compare the OKR with the rotarod as assays for quantifying the duration of drug action, we monitored the response to phenytoin at 3, 24, and 48 hours after a single i.p. injection of 85 mg/kg (Figure (Figure2).2). Consistent with phenytoin’s long half-life in humans (~24 hours), the OKR was eliminated at both 3 and 24 hours after delivery and then returned by 48 hours. Interestingly, at 24 hours after injection, phenytoin induced frequent, small-amplitude, spontaneous eye movements (see insets in the top left panel of Figure Figure2),2), reminiscent of the nystagmus that characterizes phenytoin toxicity in humans (15, 16). In contrast to the large change in the OKR at 3 and 24 hours after phenytoin injection, rotarod testing showed only a small performance decrement in response to phenytoin at 3 hours after drug delivery.
To assess dose-response relationships, we quantified the dose-dependent elimination of the OKR by gabapentin and by 2 benzodiazepines, diazepam and zolpidem (Figure (Figure2).2). For all 3 drugs we observed a simple dose-dependent reduction in ETM30. The rotarod test showed no significant changes at any of the gabapentin concentrations used, and it was less sensitive than the OKR in detecting the effects of the benzodiazepines. These experiments show that, for this set of 4 drugs, the OKR provides a measure of drug action that is at least as sensitive as rotarod performance.
As described above and shown in Figure Figure1,1, A and D, several drugs induce eye movements independent of the moving visual stimulus, referred to hereafter as spontaneous eye movements. In the case of cocaine and phencyclidine, spontaneous eye movements may represent the rodent correlate of the hyper-arousal (cocaine) or delirium/hallucinations (phencyclidine) experienced by humans. To systematically explore parameters associated with spontaneous drug-induced eye movements, we have compared continuous recordings obtained in the absence of a moving visual stimulus, before and after exposure to ketamine, memantine, and phencyclidine (Figure (Figure3).3). Spontaneous eye movements in control mice typically shift the eye to a new and relatively stable position, generating a distinctive “box-car” shape (Figure (Figure3A).3A). By contrast, the spontaneous eye movements induced by the 3 drugs shown in Figure Figure33 more closely resemble a bone fide ETM in that the rapid component is followed by a slower return to baseline; this postsaccade trajectory is reminiscent of the aberrant trajectories of fixational eye movements in primates in the absence of the neural integrator (20) and is quantified for memantine in Figure Figure3B.3B. For simplicity, we will refer to these drug-induced eye movements as “OKR like” although the slow phase is substantially less linear than the slow phase of true OKR movements.
We have used a series of simple algorithms to analyze the frequency, polarity, and shape of spontaneous eye movements. In brief, the first derivative of eye position with respect to time was calculated, and, then, proceeding sequentially from the start of the record, each contiguous set of first derivative values greater than an empirically determined threshold was counted as a single saccade, and its polarity (nasal or temporal) was recorded. To insure that the rapid rebound associated with OKR-like movements was not counted as a saccade in the opposite direction, the last time point in each set of contiguous suprathreshold first-derivative points was designated as the time of the saccade, and a second saccade designation was not permitted during the following 48 milliseconds. To distinguish box-car from OKR-like movements, the average value of the first derivative during the 116 milliseconds after each saccade was calculated; values of less than 0.5–0.8 mm/s were designated as box-car movements, and values greater than 0.5–0.8 mm/s were designated as OKR-like movements, with the exact cutoff determined individually for each OKR record. Figure Figure3C3C shows the designations of box-car or OKR-like movements for a pair of recordings in the absence of a moving visual stimulus, prior to (Figure (Figure3C,3C, top) and after (Figure (Figure3C,3C, bottom) phencyclidine administration.
The number of spontaneous OKR-like eye movements was dramatically increased by ketamine, memantine, and phencyclidine treatment (Figure (Figure3D).3D). Additionally, for memantine and phencyclidine, spontaneous saccades often occurred in rapid succession, with approximately 0.5 seconds between adjacent saccades (Figure (Figure3E),3E), a pattern reminiscent of a burst of action potentials. Visual inspection further suggests that drug-induced OKR-like movements tend to occur in clusters of the same polarity (Figure (Figure3C).3C). While this phenomenon can also be seen in Figure Figure11 during the 30-second rest periods in the baclofen, cocaine, ketamine, and phencyclidine records, it is possible that, in the experiments in Figure Figure1,1, the interleaved periods of moving visual stimuli could have entrained the polarity of the spontaneous OKR-like movements. Figure Figure3F3F shows quantification of the clustering of saccade polarities during an extended recording in the absence of a moving visual stimulus (to eliminate the possibility of entrainment of the spontaneous eye movements) and compares it with the degree of clustering calculated for a randomly ordered series of saccades with a 1:1 ratio of polarities. This analysis indicates that for memantine and phencyclidine, but not for ketamine, adjacent drug-induced OKR-like movements are of the same polarity far more often than would be expected by chance. Thus, a distinct subset of psychoactive drugs can induce clusters of saccades of identical polarity and relatively uniform spacing. Presumably, these epochs reflect distinct states of brain activity.
In the presence of a horizontally moving visual stimulus, as used here, the amplitude of vertical eye movements is typically small. Therefore, in the preceding analyses we have been exclusively concerned with eye movements in the horizontal plane. However, an exception to this general pattern was noted after memantine treatment (Figure (Figure4).4). Both during exposure to the standard horizontally moving visual stimulus and during the interleaved rest period, OKR and OKR-like movements were tilted on an approximately 45-degree angle relative to the horizontal plane, as seen by the roughly equal magnitudes of their horizontal and vertical components. This effect was only observed with counterclockwise visual motion and with spontaneous OKR-like events that mimicked the response to counterclockwise motion. Among the drugs tested thus far, memantine is the only one that shows a substantial shift in the angle of OKR or OKR-like movements. In humans, up-beat nystagmus has been associated with lesions of the cerebellum or medulla, multiple sclerosis, and organophosphate poisoning (11).
Cocaine is of special interest, both because of its enormous societal impact and because it affects eye movements in a complex and distinctive manner. As seen in Figure Figure1A,1A, at high doses (50 mg/kg), cocaine abolishes the visual stimulus–induced OKR while simultaneously inducing stereotyped spontaneous OKR-like movements. At lower doses (e.g., 20 mg/kg), cocaine increases the frequency of spontaneous eye movements in the absence of a moving stimulus, with little or no effect on the OKR (Figure (Figure5A5A and data not shown). Compared with a protocol in which eye movements were monitored in a continuous recording session with no moving stimuli, if the 30-second periods with no moving stimuli were interleaved between 30-second periods of rotating black and white stripes (i.e., the standard OKR protocol described in Figure Figure1A),1A), then the cocaine-induced increase in eye movements was substantially greater. This observation suggests that the hyperactivity of the oculomotor system induced by low-dose cocaine is triggered or gated by recent sensory input. In light of this observation, the experiments shown in Figure Figure55 were performed with the standard OKR protocol of interleaved stimulus and rest periods.
We have used a simple algorithm to quantify spontaneous eye movements: the horizontal eye position at each time point (Xt) (sampled at 60 Hz) was subtracted from the horizontal eye position 2 seconds earlier (Xt–Xt–2). For each 30-second interval, this calculation was performed over the central 26 seconds, and the resulting values were averaged for the 15 preinjection and 15 postinjection 30-second recording intervals. The resulting histograms of Xt–Xt–2 values (averaged from 3 to 4 mice) for PBS or 5, 10, 20, or 50 mg/kg cocaine provide a measure of spontaneous activity (Figure (Figure5B).5B). To reduce this measure of spontaneous activity to a single number for each experimental condition, we calculated for each histogram the Xt–Xt–2 value that corresponds to the 50th-percentile cutoff of the distribution (Figure (Figure5C).5C). Interestingly, while Figure Figure5,5, A–C, demonstrates a clear increase in spontaneous eye movements at 20 mg/kg cocaine, spontaneous movements increased to a lesser extent with 50 mg/kg cocaine. By monitoring individual 30-second intervals, we found that at 50 mg/kg there is a rapid onset of cocaine-induced spontaneous eye movements, followed within minutes by a decline that represents a general suppression of eye movements (Figure (Figure5D).5D). This analysis establishes a rapid and quantitative method for observing cocaine-induced changes in CNS function.
One of the most important aspects of preclinical drug development is assessing BBB penetration (5). BBB function is not readily modeled in vitro, as it depends on a complex combination of diffusion barriers and active extrusion by endothelial pumps. From the data presented above, it is clear that the OKR can be used to quantify the CNS actions of a variety of drugs and thereby indirectly measure their penetration across the BBB. To directly address BBB function using the OKR, we compared the responses of WT and Abcb1a–/– littermates to a single i.p. injection of ivermectin, an antihelmintic drug that activates invertebrate glutamate-gated chloride channels and, at lower efficiency, vertebrate GABA receptors (21, 22). Abcb1a (also known as Mdrla), a broad specificity ABC transporter that is expressed in endothelial cells within the CNS vasculature, plays a major role in protecting the CNS from ivermectin toxicity (23). As shown in Figure Figure6,6, A and B, in Abcb1a–/– mice, but not in Abcb1a+/+ littermate controls, the OKR is eliminated 1 day after i.p. injection of 0.5 mg/kg ivermectin, with a slow recovery over the ensuing 1 to 2 days. The rotarod test shows a similar time course of impaired performance, although the magnitude of the performance decrement is somewhat less than it is for the OKR (Figure (Figure6C).6C). A Kaplan-Meier survival curve for this experiment is shown in Supplemental Figure 2 and confirms the sensitivity of Abcb1a–/– mice to ivermectin at this dose. Together with the responses to other CNS-active compounds (e.g., Figures Figures11 and and2),2), these data indicate that the OKR can be used to quantitatively assess BBB integrity and function.
To assess the use of eye-movement testing in the analysis of drug-receptor specificity, we compared the effect of orally administered morphine in mice with a targeted deletion of the mu opioid receptor 1 gene (Oprm1–/– mice) relative to that in Oprm1+/+ littermates. Consistent with the known specificity of morphine for mu opioid receptors, Oprm1–/– mice appeared to be completely resistant to the effects of morphine, whereas Oprm1+/+ mice exhibited pupil dilation, suppression of the OKR, and severely degraded rotarod performance (Figure (Figure7,7, B and C).
Drug-drug interactions can enhance or diminish efficacy by changing the kinetics of drug transformation, metabolism, or excretion or by perturbing intersecting physiologic pathways or different steps in the same pathway. The most direct drug-drug interactions arise from competition for the same receptor binding site(s). Drug-drug interactions can lead to therapeutically advantageous synergies or undesirable side effects and can alter optimal dosing if 2 or more drugs are used simultaneously. To explore the use of the OKR for quantifying drug-drug interactions, we have chosen a classic example of a competitive drug interaction, the blockade of morphine-induced somnolence by the opiate antagonist naloxone. Figure Figure7E7E shows the time course of the experiments: the OKR was recorded prior to drug administration, naloxone (10 mg/kg) or PBS was injected i.p., 15 minutes later morphine (200 mg/kg) was delivered orally, and then, beginning 45 minutes later, the OKR was recorded for 250 minutes. The preinjection panels in Figure Figure7E7E show the expected visual stimulus–induced OKR as well as several box-car eye movements during the rest periods. Morphine treatment paired with a PBS preinjection suppressed both types of eye movements. By contrast, naloxone preinjection completely eliminated the effect of oral morphine. These data indicate that the OKR can be used to quantify the interaction between opiate agonists and antagonists. It could presumably be used in an analogous manner to monitor other classes of drug-drug interactions.
Variations in drug responses have been extensively documented across and within species, and they represent one of the challenges in extrapolating efficacy data from animals to humans. Among inbred lines of mice, there is wide variation in behaviorally measured opiate responses, as seen, for example, in the high-morphine sensitivity of C57BL/6J mice and the low-morphine sensitivity of 129SvEv mice in locomotion and thermal analgesia tests and in the jumping response to naloxone-induced acute opiate withdrawal (24). Figure Figure7,7, A and D, extends these behavioral observations by showing that 129SvEv mice exhibited an unaltered OKR and reduced pupil dilation at a morphine dose (200 mg/kg) that completely suppresses the OKR and dilates the pupil in C57BL/6J mice.
In humans, chronic phencyclidine or amphetamine abuse can lead to schizophrenia-like symptoms, suggesting that hypoactivity of NMDA signaling (phencyclidine) and/or hyperactivity of serotonin/dopamine signaling (amphetamine) are relevant to the etiology of schizophrenia (25, 26). In rodents, pharmacologic models of schizophrenia have been developed using phencyclidine, ketamine, or amphetamine treatment (27). The large literature on eye-tracking defects in schizophrenics (e.g., ref. 28) and the observation that many antipsychotic drugs affect eye movements in both healthy volunteers and schizophrenics (14, 29) suggest that changes in eye movements could be used to quantify disease severity or treatment efficacy in animal models of schizophrenia.
As described above, phencyclidine treatment in mice induced high-frequency spontaneous OKR-like eye movements (Figure (Figure1D1D and Figure Figure3),3), an effect that lasted for at least 90 minutes after an i.p. injection of 5 mg/kg (Figure (Figure8,8, A and B). Chlorpromazine at 0.5 mg/kg had little effect on the OKR (Figure (Figure8,8, A and B), although at 5 mg/kg it substantially slowed the response (Figure (Figure1,1, A and C). Interestingly, if chlorpromazine (0.5 mg/kg) was administered 30 minutes before phencyclidine (10 mg/kg), spontaneous eye movements were substantially suppressed and the OKR was reduced from approximately 15 ETM30 with phencyclidine alone to approximately 6 ETM30 with both phencyclidine and chlorpromazine, although in the latter case the slope of the slow components was steeper than that in control ETMs (Figure (Figure8,8, B–E). This chlorpromazine-mediated suppression began approximately 45 minutes after phencyclidine administration, and it persisted for at least 1 hour. These observations suggest that the suppression of phencyclidine-induced spontaneous eye movements could be used as a screening tool for antipsychotic drug candidates.
Assessing the effects of exposure to environmental chemicals, either natural or man-made, represents a major public health endeavor. Heavy metals — in particular, lead, cadmium, mercury, and arsenic — have long been known to impair a wide variety of physiologic functions after acute or chronic exposure. Although lead exposure has declined worldwide over the past several decades, it remains a major public health risk due to its widespread industrial use, especially in the developing world (30). The neurologic sequelae of lead exposure (“lead encephalopathy”) have been extensively studied. While several studies have examined the effects of chronic lead exposure on eye movements in humans and laboratory animals (31–33), to our knowledge there are no reports on eye-movement effects in the context of acute lead toxicity.
To examine the effects of acute lead toxicity, adult mice were given a single i.p. injection of 37.5, 75, 150, or 300 mg/kg lead chloride (n = 10 mice for each group). The resulting Kaplan-Meier survival curves show a delayed and dose-dependent lethality after lead exposure, with a dose of 37.5 mg/kg showing no lethality over the ensuing 10 days and a dose of 300 mg/kg producing death within 1–3 days (Figure (Figure9D).9D). When ETMs were monitored 1 hour after injection, there was little or no effect of exposure at doses of 37.5 and 75 mg/kg, but a nearly complete loss of eye movements at doses of 150 and 300 mg/kg (Figure (Figure9,9, A and B, and data not shown). For the 150 mg/kg dose, eye movements were monitored over the ensuing 2 days and showed a full recovery to baseline (Figure (Figure9B).9B). When rotarod performance was monitored in the same mice, there was little or no effect of 150 mg/kg lead at any of these time points (Figure (Figure9C).9C). The rapid loss and subsequent recovery of the OKR after the administration of 150 mg/kg lead implies an acute toxic process that is operationally distinguishable from the progressive loss of viability that occurred over the following 2–10 days (Figure (Figure9D).9D). Chronic toxicity likely involves the direct inhibition by lead of a variety of enzymes as well as competition by lead for metal binding sites during the biosynthesis of metalloproteins (34). It is an open question whether the rapid and transient lead inhibition of the OKR can be explained by either of these mechanisms. These data support the general idea that monitoring eye movements might be a broadly useful method for assessing chemical toxicities.
The use of eye-movement testing would be substantially enhanced if it could be used not only to monitor drug responses in otherwise normal mice but also drug efficacy in the treatment of CNS disease. In humans, many degenerative CNS diseases are associated with aberrant eye movements (11). For example, patients with Huntington disease (HD) exhibit errors in saccade timing, direction, and accuracy during fixation tasks (35, 36), and presymptomatic carriers of an HD mutation produce more errors than controls on a variety of saccade tasks (37). Thus, eye-movement abnormalities represent a potential biomarker for HD.
In the mouse, overexpression of an N-terminal Huntingtin protein fragment with an expanded polyglutamine repeat from the Htt R6/2 transgene produces a progressive neurodegeneration associated with distinctive motor symptoms and transcriptional changes characteristic of HD in humans (38, 39). Htt R6/2 heterozygotes are typically asymptomatic before approximately 9–11 weeks of age and then show a rapid decline, with death occurring at approximately 10–13 weeks. To explore the use of eye-movement analysis for assessing disease progression in Htt R6/2 mice, transgenic and nontransgenic female littermates were tested, beginning at 8 weeks of age, when rotarod performance, appearance, and general health of control and experimental mice were indistinguishable. Interestingly, at this age, Htt R6/2 mice showed spontaneous eye movements that resembled those induced by phencyclidine and memantine in their propensity to occur in runs of the same polarity but resembled visually evoked ETMs in the linearity of the slow component (Figure (Figure10).10). These eye movements may represent the murine correlate of the saccade errors made by HD patients, and they could presumably be used to follow disease progression or responses to treatment in presymptomatic HD animal models.
The experiments described above establish eye-movement analysis as a broadly useful tool for rapidly and quantitatively assessing the responses of mice to a wide variety of psychoactive compounds. The eye movement effects include a slowing or complete suppression of the OKR, a shift in eye-movement direction, induction of spontaneous OKR-like movements in the absence of visual stimuli, clustering of the polarity or timing of OKR-like movements, and changes in the slope of the slow phase of the OKR-like movements. By monitoring these parameters, eye movements can be used to quantify drug dose, duration of action, BBB penetration, receptor specificity, and agonist-antagonist interactions. We further show that the frequency of OKR-like events can be used to monitor (a) the therapeutic effect of antipsychotic drug treatment in a pharmacologic model of schizophrenia and (b) the presence of disease in a mouse model of HD.
Based on the pharmacologic specificity of the drugs tested here, it appears that a wide variety of neurotransmitter systems can influence eye movements in mice, including cholinergic, glutamatergic, aminergic, GABAergic, opioid, and cannabinoid systems. At present, we do not know which drugs directly perturb eye-movement circuits and which act indirectly via changes in arousal, coordination, or higher mental states. In the paragraphs that follow, we relate our findings to the known effects of psychoactive drugs and CNS diseases on eye movements in humans, and we discuss potential applications of eye-movement analysis to drug discovery, drug abuse, and CNS disease research.
Eccentric gaze, smooth pursuit, and convergent eye movements are easily elicited in humans and are an integral part of the standard neurologic examination (40). Each of these eye movements either brings the object of regard onto the fovea or serves to maintain it there. The rodent retina lacks a specialized region devoted to high acuity vision, and, as a result, rodent eye movements are not driven by foveal mechanisms. However, rodents resemble humans and other primates in possessing reflexes for stabilizing the image on the retina. These reflexes compensate for angular and linear self-motion and are based on vestibular input (VOR) and on movement of the image across the retina (the OKR). As the midbrain and hindbrain nuclei that subserve the VOR and the OKR are highly conserved among mammals, it is reasonable to expect a corresponding evolutionary conservation of neurochemistry and neural circuitry within these systems. Current evidence supports this expectation (41).
Changes in eye movements in response to acute administration of compounds representing the major classes of psychoactive drugs have been extensively studied in humans. Smooth pursuit, convergence, and the maintenance of eccentric gaze are especially sensitive to drug administration (11). If we consider only smooth pursuit eye movements — the human response that most closely resembles the mouse OKR — we observe a high degree of similarity in the effects of a number of drugs between the 2 species (human data are summarized in refs. 11 and 14). Benzodiazepines, barbiturates, opiates, and phenytoin impair smooth pursuit in humans and diminish or eliminate the OKR in mice. In humans, acute administration of haloperidol or chlorpromazine, typical first generation antipsychotic drugs (i.e., antagonists at D2-D4 dopamine receptors), or clozapine, a second generation antipsychotic (i.e., a mixed dopamine and serotonin receptor antagonist), decreases smooth pursuit eye movements. In mice, all 3 antipsychotics produce a marked slowing of the OKR. In humans, cocaine induces opsoclonus (spontaneous large-amplitude saccades) and phencyclidine induces nystagmus (repetitive OKR-like movements). In mice, both drugs induced spontaneous eye movements. In humans, amphetamine has little effect on smooth pursuit or saccadic eye movements, and amphetamine similarly has little or no effect on the mouse OKR.
Other drugs show more or less divergent responses between the 2 species. For example, in humans, carbamazepine (an anticonvulsant and mood-stabilizing drug that potentiates GABA function and decreases sodium channel opening) impairs smooth pursuit eye movements, but, in the present study, it did not affect the mouse OKR. Baclofen, a GABA derivative used in the treatment of spasticity, including periodic alternating nystagmus in humans, paradoxically induces OKR-like eye movements in mice. Nicotine impairs smooth pursuit eye movements in humans but had no apparent effect on the mouse OKR. It is unclear whether these response differences reflect a fundamental divergence between humans and mice in neurochemistry and/or neural circuitry or differences in pharmacokinetics, pharmacodynamics, drug dosing, the sensitivity of eye-movement testing protocols, or some combination of these factors.
In humans, a wide variety of focal CNS lesions, metabolic disorders, and progressive degenerative diseases produce eye-movement defects as one part of a constellation of clinical signs and symptoms. Among the degenerative diseases that typically include eye-movement defects are multiple sclerosis, Parkinson disease, HD, progressive supranuclear palsy, the hereditary cerebellar ataxias, Alzheimer disease, Creutzfeldt-Jacob disease, and AIDS dementia (11). Although mouse models have been developed for many of these diseases, thus far there has been little or no discussion of eye-movement defects. Instead, the small number of studies that have examined eye movements (the OKR and/or the VOR) in genetically modified mice have focused on gene defects that lead to selective retinal, vestibular, or cerebellar dysfunction but do not model the major human neurodegenerative diseases (17, 42–46). Our observation of a progressive increase in spontaneous OKR-like movements in a mouse model of HD that begins before overt systemic motor symptoms suggests that monitoring eye movements could represent a useful screening tool for preclinical therapeutic trials in this HD model and that a careful assessment of eye movements is warranted in other mouse models of CNS disease.
Evidence for an association between eye-movement defects and neuropsychiatric disease was first reported by Diefendorf and Dodge (47). In the century since that time, a large number of studies have supported the general concept that abnormalities in smooth pursuit and voluntary saccades represent a biomarker for schizophrenia, although confounding factors, such as previous or current drug treatment and nicotine exposure from cigarette smoking, have not been entirely eliminated (28). We report here that commonly used pharmacologic models of schizophrenia — created by treatment with ketamine or phencyclidine — produce distinctive and readily quantified eye-movement abnormalities in mice. We also observed the abrogation of phencyclidine-induced OKR-like movements after chlorpromazine pretreatment, an interaction that could represent a mouse model of antipsychotic drug action. We note that the mechanism by which this abrogation occurs may be complex, as both drugs have multiple CNS targets: phencyclidine is both an NMDA receptor antagonist (like ketamine) as well as a D2 dopamine receptor agonist, whereas chlorpromazine is an antagonist at dopaminergic, adrenergic, cholinergic, and histaminergic receptors (48).
High-throughput biochemical and cell-based screens with large libraries of compounds are now routine in the pharmaceutical industry. These screens, together with lead optimization, typically generate large numbers of candidate compounds for animal testing. However, animal testing is expensive and time consuming, and, as such, represents a major bottleneck in the drug development pipeline. This limitation is especially acute in CNS drug development, because behavioral parameters are generally more challenging to quantify and interpret than are biochemical, physiological, or histological ones. One strategy to address this bottleneck uses invertebrate model organisms, such as Caenorhabditis elegans and Drosophila melanogaster, for high-throughput animal screening (49, 50). A more promising model is the zebrafish Danio rerio, which is amenable to large-scale continuous monitoring of motor activity (51, 52).
As currently practiced, preclinical drug testing focuses almost exclusively on mammals, based on the reasonable assumption that the predictive value of the preclinical test correlates with the evolutionary distance between humans and the animal model. This assumption appears to be especially reasonable in the case of psychoactive drugs, many of which interact with multiple molecular targets, as seen, for example, in the interactions of antidepressant or antipsychotic drugs with multiple classes of amine transporters or receptors, respectively (53). Eye-movement analysis represents a potentially useful addition to the existing repertoire of behaviorally based drug assays. Since it is safe and noninvasive, it could be readily extended to larger mammals, including humans.
Based on the data presented here, the OKR appears to be especially suitable for quantifying sedation, an undesirable side effect of many medications. Thus, for drug development programs aimed at peripheral targets, the OKR could be used as an initial screening tool to identify and eliminate sedating compounds. The OKR could also guide lead optimization efforts aimed at decreasing or increasing BBB penetration for drug development programs with peripheral or central targets, respectively. The distinctive eye-movement responses that mice exhibit to various drugs of abuse — including cocaine, phencyclidine, opiates, ethanol, and cannabinoids — suggest that eye-movement assays might be used to test compounds that counteract the effects of these drugs (54). In the future, it will be interesting to extend these eye-movement studies to include the full range of CNS-active compounds — both therapeutic and toxic — to which humans are exposed.
Methods for surgically inserting a head post, immobilizing the head-posted mouse, presenting a rotating visual stimulus, infrared video recording of eye position, and thresholding and smoothing of the raw eye position data are described in ref. 19. Surgical procedures as well as mouse handling and housing were approved by the Institutional Animal Care and Use Committee of the Johns Hopkins University School of Medicine. Unless otherwise noted, tests were performed with 2- to 7-month-old C57BL/6J female mice, weighing between 18 and 30 grams. The OKR stimulus consisted of alternating vertically oriented black and white stripes, the width of each subtending 4 degrees of visual angle, with the entire pattern, which covered the inside of the testing cylinder, rotating at 5 degrees per second. Head post surgery was conducted at least 48 hours prior to testing to allow adequate time for recovery. Morphine was delivered by i.p. injection (Figure (Figure1)1) or oral gavage (Figure (Figure7);7); all other drugs were delivered by i.p. injection. Unless otherwise noted, postdrug OKR recordings were performed between 30 and 90 minutes after drug administration, and the rotarod tests were performed on each mouse immediately after the OKR recording. For rotarod testing, the cylinder rotated at 7 revolutions per minute, and each mouse was given 3–15 sixty-second trials. Prior to drug injection, each mouse was given up to 3 practice runs on the rotarod. Rotarod data collection began when the mouse attained its first 60-second rotarod time.
The area of the pupil was calculated from the infrared video image using ISCAN software.
Water soluble compounds were dissolved in PBS or 0.1 M NaCl for i.p. injection. Hydrophobic compounds were dissolved in a small volume of DMSO or ethanol and then dissolved in Sunflower seed oil (Sigma-Aldrich) (see Supplemental Table 1). For compounds that had been dissolved in ethanol, the ethanol was removed from the ethanol/oil mixture by spinning the sample in a SpeedVac Concentrator (Savant) prior to i.p. injection.
Abcb1a–/– male mice (Taconic Laboratories; MDR1A-M-M, FVB.129P2-Abcb1atm1Bor N7) were crossed to C57BL/6J female mice, and the heterozygous progeny were intercrossed to produce littermate Abcb1a+/+ and Abcb1a–/– mice for OKR testing. The original Abcb1a–/– line from Taconic Laboratories had been backcrossed to the FVB strain, thereby introducing 2 complications for OKR analysis. First, the mice were albino, and iris pigmentation is needed for automated localization of the pupil. Second, the mice were homozygous for the retinal degeneration allele, Pde6brd, thus eliminating visually evoked eye movements. Thus, among the F2 littermates, albino and/or Pde6brd/rd mice were eliminated from testing. Oprm1–/– male mice (The Jackson Laboratory; stock no. 007559) were crossed to C57BL/6J female mice, and the heterozygous progeny were intercrossed to produce littermate Oprm1+/+ and Oprm1–/– mice for OKR testing. Htt R6/2 male mice (The Jackson Laboratory; stock no. 002810) were bred to C57BL/6J female mice, and heterozygous progeny were tested. 129SvEv mice were purchased from Taconic (stock no. 129SVE), and C57BL/6J mice were purchased from The Jackson Laboratory (stock no. 000664). PCR primers for genotyping are listed in Supplemental Table 3.
Calculations were performed in Microsoft Excel. The fast component of the saccade was identified, and its polarity was scored by thresholding the first derivative of the smoothed eye position record as described in ref. 19. In brief, the time derivative was determined for adjacent eye position data points collected at 16-millisecond intervals, and a threshold for the absolute value of the derivative was set at 1.8–3.6 mm/s and used to identify saccades, with the exact threshold determined individually for each OKR record. Box-car and OKR-like saccades were distinguished by determining the first derivative averaged over the first 116 milliseconds after the end of the fast component. Derivatives of less than or greater than 0.5–0.8 mm/s were classified as box-car or OKR-like, respectively, with the exact cutoff determined individually for each OKR record. Saccades were averaged by both normalizing the amplitude and aligning the time of the fast component.
For polarity analysis, recordings of 5- or 10-minute duration were obtained in the absence of a moving visual stimulus, saccades were identified as described above, and negative or positive saccade polarities were represented by –1 or +1, respectively. From the resulting sequence of +1’s and –1’s, the number of runs was tabulated for each possible run length, a run being defined as a series of adjacent saccades of the same polarity. From this analysis, the number of saccades present for each run length was plotted. For comparison with these experimental distributions, we have calculated the distribution of run lengths for a sequence of randomly ordered polarities. Taking the 2 polarities as equally likely, the probability of encountering a run of either polarity and of length n is 2 × 0.5n+2, the n+2 exponent reflecting the requirement for a saccade of opposite polarity immediately prior to and after the run. Therefore, the probability of a saccade being in a run of length n is proportional to n × 2 × 0.5n+2.
For the quantification of spontaneous eye movements induced by cocaine, the horizontal eye position at each time point (sampled at 60 Hz) was subtracted from the horizontal eye position 2 seconds earlier (Xt–Xt–2). For each 30-second interval, this calculation was performed over the central 26 seconds. To reduce this measure of spontaneous activity to a single number, we calculated for each distribution the Xt–Xt–2 value that corresponds to the 50th-percentile cutoff.
For the Kaplan-Meier survival curve, cohorts of 10 mice (C57BL/6 females, 3–8 months of age), weighing 17–32 grams, were injected i.p. with 37.5, 75, 150, or 300 mg/kg lead chloride in water on day 0. In addition to the usual ad libitum food and water, an aqua-gel pack and extra food were placed in the cage. These cohorts and the cohorts used in the OKR and rotarod experiments exhibited very similar dose-dependent survival curves.
Student’s t test was used for pairwise comparisons. P values less than 0.05 were considered significant. For comparisons of eye movements, a 2-tailed t test was used. For rotarod comparisons in which the control mice remained on the rotarod for the full 60 seconds, a 1-tailed t test was used.
The authors thank Jay Baraban, Alex Kolodkin, Se-Jin Lee, Randy Reed, King-Wai Yau, and David Zee for advice and/or comments on the manuscript. This work was supported by the Howard Hughes Medical Institute.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2011;121(9):3528–3541. doi:10.1172/JCI45557.